Author Response:

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

Reviewer #1 (Recommendations For The Authors):

- There were no mechanistic or causation-focused investigations that could have greatly strengthened the study. The study is ultimately providing two prioritized candidate genes that may be causative, reactive, or independent of the disease.

Answer: We thank the reviewer for their positive assessment and agree that our study lacks formal causal analyses. We are aware of this limitation and have made it clear throughout the text. Through triangulation of evidence across tissues and species, we point to very interesting candidates that merit further study, which is the usual scope of such systems genetics investigations. Nevertheless, to introduce some causal inference and reinforce the human relevance of our results, we have performed Mendelian randomization (MR) analysis to investigate the potential associations between MUC4’s gene expression in human colons and the risk of IBD. EPHA6 lacks detectable eQTLs in human colon so we could not include it in this analysis. We found suggestive evidence that increased expression of MUC4 in the sigmoid, but not transverse, colon may increase the risk of IBD (nominal p = 0.033).

The description in the manuscript:

However, it is unclear through what mechanisms the genetic variants in the candidate genes affect IBD susceptibility. One possibility is that genetic variation leads to altered levels of expression of the gene, ultimately affecting disease susceptibility. To test this possibility, we examined the GTEx resource (GTEx Consortium, 2013) and found that MUC4, but not EPHA6, has cis-eQTLs in the sigmoid and transverse colon. To establish likely causal links with IBD incidence, we used these associations as instruments in a two-sample Mendelian randomization (MR) (Hemani, Tilling and Smith, 2017; Hemani et al., 2018) analysis. Using publicly available GWAS summary statistics for IBD, Crohn’s disease, and ulcerative colitis (Liu et al., 2015; Elsworth et al., 2020) as outcomes, we found suggestive evidence that increased expression of MUC4 in the sigmoid, but not transverse, colon may increase the risk of IBD (nominal P value = 0.033, Appendix 1 - Table 6). No eQTLs were reported for EPHA6 in the colon, precluding us from investigating the potential consequences of changes in its expression in these tissues.

- Figures 3 and its supplement Figure 1: Among the 39 modules, the authors have only focused on significantly overlapping up-regulated IBD-related gene modules in both CD (M28 and M32) and HFD (M9 and M28) for their follow up analyses in Figures 4 and 5 to prioritize candidate genes. However, this reviewer thinks there is great value in also focusing on significantly overlapping down-regulated IBD-related gene modules in both CD (M17) and HFD (M15 and M26) for their follow up candidate gene prioritization analyses.

Answer: Thank you for your suggestion. We had initially performed overrepresentation analyses in HFD_M15, HFD_M26 and CD_M17, but did not find enrichments related to inflammation (see Author response image 1 below). We did not include this result in the manuscript.

Author response image 1.

Dot plot showing the enrichment of IBD-related modules in hallmark genesets. Gene ratios higher than 0.1 are shown and represented by dot size. Dots are colored by -Log10(BH-adjusted P values).

We also checked the module QTL mapping for the significantly overlapping down-regulated IBD-related gene modules in both CD and HFD. We did not find any loci that are significantly associated with these modules, indicating that they are not modulated by genetic variation and hence are less likely to inform on IBD susceptibility.

The description in the manuscript:

The ModQTL analysis was also performed on the modules that are significantly enriched in IBD-downregulated genes (HFD_M15, HFD_M24, and HFD_M26), but no significant or suggestive QTLs were detected. Therefore, we focused on the QTL for IBD-induced genes in HFD_M28 and annotated its candidate genes based on three criteria (Figure 5B).

Reviewer #2 (Recommendations For The Authors):

- One small addition that would be nice would be to indicate if the two candidate genes have cis eQTL in human tissues and/or have any protein-coding variants in humans. This would provide nice additional evidence of causality for these two genes.

Answer: Thank you for your positive assessment and suggestion. MUC4 and EPHA6 both have protein-coding variants in humans that were listed in the Appendix – Table 3 and Table 4. In addition, cis-eQTLs have been found for MUC4 in both the sigmoid and transverse colon in humans (GTEx, https://gtexportal.org/home/locusBrowserPage/ENSG00000145113.21). As indicated in our response to the first comment of Reviewer #1, we have now performed mendelian randomization on human eQTL for MUC4. However, no eQTLs were reported for EPHA6 in the colon, preventing us from performing MR analysis on its expression.

- Also, it would be helpful to include the size of the modules in the text of the manuscript. Especially the two modules that were followed up on.

Answer: Thank you for your suggestion, we have indicated the size of IBD-related modules in the text of the manuscript.

The description in the manuscript:

Enrichment analyses indicated that modules HFD_M9 (484 genes), HFD_M16 (328 genes), and HFD_M28 (123 genes) were enriched with genes that are upregulated by DSS-induced colitis, while HFD_M15 (368 genes), HFD_M24 (159 genes), and HFD_M26 (135 genes) were significantly enriched with downregulated genes (Figure 3C). Of note, more than 20% of genes involved in HFD_M9 and HFD_M28 were part of the dysregulated genes of the acute phase of mouse UC (day6 and day7) (Figure 3C). Interestingly, genes perturbed during IBD pathogenesis in humans were also enriched in HFD_M9 and HFD_M28 (Figure 3C).

While IBD-related genes were predominantly found in HFD modules, we also found that two modules, CD_M28 (185 genes) and CD_M32 (142 genes), in CD-fed mouse colons were associated with IBD (Figure 3—figure supplement 1A). These two-modules significantly overlapped with the IBD-related HFD_M9 and HFD_M28 modules, respectively (BH-adjusted P value < 0.05) (Figure 3—figure supplement 1B). Moreover, the molecular signatures underlying human UC and Crohn’s disease were also clustered in these two modules (CD_M28 and CD_M32) under CD (Figure 3—figure supplement 1C). Collectively, the co-expression and enrichment analyses identify HFD_M9 and HFD_M28 as IBD-related modules on which we focus our subsequent investigation.

Author Response

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

Reviewer #1 (Public Review):

Jamge et al. sought to identify the relationships between histone variants and histone modifications in Arabidopsis by systematic genomic profiling of 13 histone variants and 12 histone modifications to define a set of "chromatin states". They find that H2A variants are key factors defining the major chromatin types (euchromatin, facultative heterochromatin, and constitutive heterochromatin) and that loss of the DDM1 chromatin remodeler leads to loss of typical constitutive heterochromatin and replacement of this state with features common to genes in euchromatin and facultative heterochromatin. This study deepens our understanding of how histone variants shape the Arabidopsis epigenome and provides a wealth of data for other researchers to explore.

Strengths:

1) The manuscript provides convincing evidence supporting the claims that: A) Arabidopsis nucleosomes are homotypic for H2A variants and heterotypic for H3 variants, B) that H3 variants are not associated with specific H2A variants, and C) H2A variants are strongly associated with specific histone post-translational modifications (PTMs) while H3 variants show no such strong associations with specific PTMs. These are important findings that contrast with previous observations in animal systems and suggest differences in plant and animal chromatin dynamics.

2) The authors also performed comprehensive epigenomic profiling of all H2A, H2B, and H3 variants and 12 histone PTMs to produce a Hidden Markov Model-based chromatin state map. These studies revealed that histone H2A variants are as important as histone PTMs in defining the various chromatin states, which is unexpected and of high significance.

3) The authors show that in ddm1 mutants, normally heterochromatic transposable element (TE) genes lose H2A.W and gain H2A.Z, along with the facultative heterochromatin and euchromatin signatures associated with H2A.Z at silent and expressed genes, respectively.

Weaknesses:

1) Following up on the finding that H2A.Z replaces H2A.W at TE genes in ddm1 mutants, the authors provide in vitro evidence that DDM1 binds to H2A.Z-H2B dimers. These results are taken together to conclude that DDM1 normally removes H2A.Z-H2B dimers from nucleosomes at TE genes and replaces them with H2A.W-H2B dimers. However, the evidence for this model is circumstantial and such a model raises a variety of other questions that are not addressed by the authors.

The Reviewer raises a series of interesting questions. We proposed that DDM1 exchanges H2A.Z to H2A.W because it is the simplest model and also because LSH - the mammalian ortholog of DDM1 exchanges H2A to macroH2A. However we do stress in the revised manuscript that this is a model and other possible models that could involve chaperones and additional remodelers are possible. Addressing why the loss of DDM1 results in a net exchange of H2A.W to H2A.Z is not the purpose of this study. Here we use the perturbation caused by ddm1 as a means to address the importance of the dynamics exchange of H2A variants in setting up the chromatin states. We do observe that perturbing this dynamic exchange causes an important perturbation of chromatin states. This further supports our main conclusion: H2A variants dynamics are one important factor that organizes chromatin states.

For example: if DDM1 does remove H2A.Z from TE genes, how does H2A.Z normally come to occupy these sites, given that they are highly DNA methylated and that H2A.Z is known to anticorrelate with DNA methylation in plants and animals?

The anticorrelation between H2A.Z and DNA methylation is observed at steady state. The exchange of H2A.Z to H2A.W that results from the action of DDM1 would indeed remove unwanted H2A.Z from regions occupied by DNA methylation as suggested by the Reviewer.

Given that H2A.Z does not accumulate in TEs in h2a.w mutants, how would H2A.X and H2A instead become enriched at these sites if DDM1 cannot bind these forms of H2A?

This is a valid question: We envisage that H2A.X and H2A are deposited by remodelers and chaperones other than DDM1 in the h2a.w mutant.

Given that there are no apparent regions with common sequence between H2A.Z and H2A.W variants that are not also shared with other H2A classes, how would DDM1 selectively bind to H2A.W-H2B and H2A.Z-H2B dimers to the exclusion of H2A(.X)-H2B dimers?

It was shown by the Muegge Lab both in vitro and in vivo that LSH - the mammalian ortholog of DDM1 binds to macroH2A and H2A, and these two H2A variants do not share similar specific region. Yet it remains to determine which region of H2A.Z and H2A.W binds to DDM1, which does not fit in the scope of this study.

Reviewer #2 (Public Review):

Jamge et al. set out to delineate the relationship between histone variants, histone modifications and chromatin states in Arabidopsis seedlings and leaves. A strength of the study is its use of multiple types of data: the authors present mass-spec, immunoblotting and ChIPseq from histone variants and histone modifications. They confirm the association between certain marks and variants, in particular for H2A, and nicely describe the loss of constitutive heterochromatin in the ddm1 mutant.

The support for some of the conclusions is weak. The title of the discussion, "histone variants drive the overall organization of chromatin states" implies a causation which wasn't investigated, and overstates the finding that some broad chromatin states can be further subdivided when one considers histone variants (adding variables to the model).

We have removed subtitles in the discussion and have taken care to avoid over simplified statements.

Adding variables to a ChromHMM model naturally increases the complexity of the models that can be built, however it is difficult to objectively define which level of complexity is optimal. The differences between states may be subtle to the point that they may be considered redundant. The authors claim that the sub-states they define are biologically important, but provide little evidence to support this claim. It is not obvious whether the 26 states model is much more useful than a 9-states model. Removing variables naturally affects the definition of states that depend on these variables, but it is also hard to define the biological significance of that change. This sensitivity analysis is thus not very developed.

We agree that adding more input tracks/ data will increase the complexity.

But we would like to mention the differences of this study and the 9-state model,

1) We have included the histone variants which have been previously missed in chromatin state definition.

2) The previous 9-state model used data from different tissue types. In this study all the data generated and analyzed is from seedlings.

3) Increasing the number of states allowed us to resolve heterochromatin states compared to 9-state model which was previously missed. (BioRXiv)

4) The biological relevance of the 26 states model is analyzed and described in depth (States BioRxiv paper).

In addition we have now updated the Figure 2F to include a more direct comparison of marks used in both models. And we have expanded the description in the methods section and our reasoning behind using 26 state model to be analyzed in depth.

There are issues with the logical sequence of arguments in Fig1 and Fig3. Fig1A shows that nucleosomes often contain both H3.1 and H3.3. Therefore pulling-down H3.1-containing nucleosomes also pulls down H3.3 and whether specific H2A variants associated with H3.1 cannot be answered in this way (Fig1B).

We thank the Reviewer for point this out. If 60% of nucleosomes are homotypic and if they would associate with a specific H2A variant this would be clearly visible on WB as a much stronger band. Also, the MS data presented in Figure1 figure supplement 1D clearly show that all H2A variants associate with both H3.1 and H3.3. We have included in the revised version more detailed explanation to clarify this point.

The same issue likely carries to the investigation of the association with H3 modifications if Fig1C and 1D, since the H3.1-HA pull-down also pulls down endogenous H3.1 (so presumably the rest of the nucleosome, with H3.3, as well).

We disagree on this point. The H3 band corresponding to the transgene copy is either H3.1 or H3.3, so all signals on upper band (T) in Figure 1C are associated with either H3.1 (H3.1 IP) or H3.3 (H3.3 IP), thus unambiguously showing that all modifications we analyzed are present on both H3.1 and H3.3. Furthermore, data shown in Figure 1D and E, where we analyzed modifications on K27 and K36 which are in the H3 region that can be distinguished between H3.1 and H3.3 by MS clearly demonstrate that these modifications are present on both H3.1 and H3.3. In order to make this clearer, we also extended the description of this part in the Results section to emphasize this.

In Fig3, the conclusion that it is the loss of H2A.Z -> H2A.W exchange in the ddm1 mutant that causes loss of constitutive heterochromatin is rushed. The fact that the h2a.w mutant does not recapitulate the loss of constitutive heterochromatin seen in ddm1 argues against this interpretation.

We agree that at first the minimal impact of the loss of H2A.W alone is surprising. However, we point to the preprint https://www.biorxiv.org/content/10.1101/2022.05.31.493688v1. There it is shown that the joint loss of H2A.W and H3K9 methylation (also observed in ddm1) affects silencing of a large range of transposons that also lose silencing in ddm1.

It's also difficult to conclude about the importance of dynamic exchanges when the ddm1 mutation has been present for generations and the chromatin landscape has fully readapted. Further work is needed to support the authors' hypothesis.

We apologize that the Reviewer could not find the information regarding the origin of ddm1 mutant material. We did not use a mutant where ddm1 mutations was kept for generations. We were in fact very careful on this point and used leaves from ddm1 first homozygous plants segregated from heterozygous ddm1 kept heterozygous.

The study also relies on a large number of custom (polyclonal) antibodies with no public validation data. Lack of specificity, a common issue with antibodies, would muddle the interpretation of the data.

We added information about validation of custom made antibodies into Methods: ”Specificities of custom made polyclonal antibodies against Arabidopsis H2A.Z.9, H2A.X, H2A.W.6, H2A.13, H2A.W.7, H2Bs, and linker histone H1 were validated in previous publications (Yelagandula et al., 2014; Lorkovic et al., 2017; Jiang et al., 2020; Osakabe et al., 2021).“ For H2A.2 and H2A.Z.11 antibodies we provide validation data as Figure 2 figure supplement 1.

Overall, this study nicely illustrates that, in Arabidopsis, histone variants (and H2A variants in particular) display specificity in modifications and genomic locations, and correlate with some chromatin sub-states. This encourages future work in epigenomics to consider histone variants with as much attention as histone modifications.

Reviewer #3 (Public Review):

How chromatin state is defined is an important question in the epigenetics field. Here, Jamge et al. proposed that the dynamics of histone variant exchange control the organization of histone modifications into chromatin states. They found 1) there is a tight association between H2A variants and histone modifications; 2) H2A variants are major factors that differentiate euchromatin, facultative heterochromatin, and constitutive heterochromatin; 3) the mutation in DDM1, a remodeler of H2A variants, causes the mis-assembly of chromatin states in TE region. The topic of this paper is of general interest and results are novel.

Overall, the paper is well-written and results are clearly presented. The biochemical analysis part is solid.

Reviewer #4 (Public Review):

This work aims at analyzing the impact of histone variants and histone modifications on chromatin states of the Arabidopsis genome. Authors claim that histone variants are as significant as histone modifications in determining chromatin states. They also study the effect of mutations in the DDM1 gene on the exchange of H2A.Z to H2A.W, which convert the silent state of transposons into a chromatin state normally found on protein coding genes.

This is an interesting and well done study on the organization of the Arabidopsis genome in different chromatin states, adding to the previous reports on this issue.

Reviewer #1 (Recommendations For The Authors):

1) The rationale for switching from using 10-day old seedlings for chromatin profiling to using mature leaves in Figure 3 and beyond is not explained and introduces additional complexity into the analyses. The reasoning should be clearly explained in the text, and there are several additional suggestions or questions related to this that should be addressed:

This was done for practical reasons. We had already obtained some profiles of marks in ddm1 mutants and extended the dataset using the same stage of development because this tied this study with our previous study. Using different stages of development provides an additional benefit. The same chromatin states are observed in 10 day old seedlings and leaves of older plants. Constitutive heterochromatin is occupied by the same chromatin states and logically euchromatin is positioned on different genes as expected by the distinct pattern of gene expression at the two stages of development.

A) In the 16-state model (Figure 3A), euchromatin states were not well defined compared to the 26-state model. Why did the authors not profile these marks also, and could this explain why ddm1 mutants did not show a significant effect on euchromatin states in this model?

We apologize for the lack of detailed explanation: In our previous study we used leaves of five weeks ld plants to show the impact of ddm1 on the profiles of H2A.W.6, H2A.X, H1, H3K9me2, H3K36me3 and H3K27me3 in leaves (Jamge, Osakabe et al., 2021). This study showed that DDM1 causes the deposition of H2A.W.6 to heterochromatin and we thus used leaves to extend this investigation to the two other marks of heterochromatin (constitutive or facultative) H3K9me1, H2A.W.7 and H2A.Z.9 and H2A.Z.11.

B) The authors state that the tissue types do not impact the definition of chromatin states. However, there is a clear difference in the portion of the genome occupied by each chromatin state between leaf and seedling (states 1, 5, 8, 13, and 14; Figure S3A).

We had missed a comment on supFig3B and have now provided more explanation: “Although the composition of the chromatin states did not vary significantly between seedlings and leaves, each state occupied a similar proportion of the genome in seedling or leaves to the exception of state 5 present primarily in leaves and state 13 only present in seedlings (Figure 3 figure supplement 3A, right column with green bars) and the euchromatin states occupied different genes (Figure 3 figure supplement 3B) as expected by the dissimilar transcriptomes of these two developmental stages.”

2) The naming of supplemental figures throughout the text is confusing as the legends refer to them as "Figure SX" but they are called out in the text as "Figure X figure supplement XA-B". The eLifeconvention is "Figure X figure supplement XA-B".

This was changed.

3) In Figure 4, Panel D is mislabeled as C in the figure, and C is lacking a label.

4) Please remove the word "the" from the title.

This was done

Reviewer #2 (Recommendations For The Authors):

Fig1D legend should also mention K37.

This was corrected.

Fig2F legend should say "no H3 modifications" rather than "no histone modifications" This was corrected.

Fig4 labels C/D do not correspond to the legend. D is missing and C should go to the ddm1 stacked barplot.

This was corrected.

H3 variants analysis: Taking the relative abundance of H3.1 and H3.3 (and transgenes) into account would be useful to interpret the results of the nucleosome composition results. If they are at equivalent amounts, the null hypothesis of independent association would give 50% heterotypic nucleosomes and 50% homotypic.

This is a valid comment. In an ideal system the last statement would be correct, but this does not take into account chromatin dynamics associated with replication, transcription, etc. Also, total amounts of H3.1 and H3.3 in tissue we used for the experiment is not known. It could possibly be inferred from RNAseq data, but if this would reflect real amounts of the protein is highly questionable. In Arabidopsis there are 5 H3.1 genes and 3 H3.3 genes. Nevertheless, we recalculated data for H3.1 and H3.3 and this has been updated in the main text (~60% of H3.1 and ~42% of H3.3 immunoprecipitated nucleosomes contained both H3 variants). Thus, from the available data these numbers are the best we can get.

p. 5 bottom paragraph. Repetition.

This was corrected

p12. The reference to LSH is dropped in without making clear how it is relevant. Expand on mechanism to suggest similar DDM1 mechanism?

This section was expanded to provide more background in the interpretation of the results.

p13. inversion between H2A.W and H2A.Z in "the loss of DDM1 prevents the replacement of H2A.W by H2A.Z".

This was corrected

p13. make it clear that the last sentence of the results is a working model, not a fully backed up conclusion.

Alternative models are mentioned in this section and in the discussion in the revised version.

p14 middle paragraph. Not clear what "in silico simulation" refers to. Simply chromatin-state classification with ChromHMM?

This refers to the Jacard index calculation in Fig. 2F that models the impact of the loss of H2A variants (or other elements of chromatin) on the definition of chromatin states by ChromHMM. This is now clarified.

p14 bottom paragraph: the H2A.Z tail repression of ubiquitin ligase but its being the favoured substrate for H2AK121Ub is apparently contradictory. Can this be explained?

This refers to H2B Ubiquitination and is now clarified

p15. Correlation between variants and modifications/chromatin states does not necessarily mean causation.

We agree and have improved the revised version in this respect.

p15 "forward feedback loop" is ambiguous (is it a feed-forward loop? A feedback loop?), just use "positive feedback loop".

This was corrected.

p23 top "$(Ingouff et al)" doesn't seem properly formatted.

This reference did not belong there and has been removed.

Data availability: GSE226469 is not public. The manuscript also mentions availability of source data for all the main figures, but I could not find it. It would be great to make the code publicly available too.

All the data and code will be public upon posting the revised version of the manuscript.

Reviewer #3 (Recommendations For The Authors):

My major concern is authors only used DDM1 as an example to show that the exchange of the histone variant contributes to definition and distribution of chromatin state on transposons (i.e., constitutive heterochromatin regions associated with H2A.W). Readers may wonder whether similar mechanisms also work at the euchromatin region. This point should be clearly discussed and mentioned in the Results (for example, cite recent work on INO80).

We discuss the impact of other remodelers in the Discussion in the revised version. We hope that the reviewer will understand that doing a study on the impact of other remodelers on chromatin states which would require dozens of new ChIP profiles and is clearly beyond the scope of revising a manuscript.

Minor:

1) Fig. 2A and 2B, what does color mean? I guess the color code is referred to chromatin states (Fig. 2F).

We have clarified on Figure 2A the attribution of a specific color to each chromatin state. This same color is used also in other panels of Figures 2 and S2.

2) Supplemental Figures: All the figure panels should be on the same page.

We rearranged supplemental figures so that each figure fits on one page. In places where this was not possible, we created additional supplemental figures.

3) "We observed that increasing state numbers from 26 to 27 gave rise to biologically redundant states.": Where are the data? Fig S2A? This figure is hard to understand.

In the updated manuscript, we have described the legend and the methods for FigS2A in more detail.

Reviewer #4 (Recommendations For The Authors):

A general concern refers to the text that frequently falls into excessive oversimplifications and/or overstatements, with the danger of being misleading for the reader. This needs to be thoroughly revised.

We added more careful statements and proposed alternative models when it was possible.

Specific comments.

1) Fig 1A. Authors found the ~40% of nucleosomes contained both H3.1 and H3.3. This is a significant finding that deserves a more detailed comment.

We now provide a more detailed description of IP and MS data presented in Figure 1. This should also help to avoid oversimplifications and/or overstatements as criticized in a general comment.

2) Fig 1C. "H3. And H3.3 bore the same sets and comparable levels of methylation and acetylation...". Too general statement, please specify. Is this also the case for H3K9me2? Others?

We did describe this part into more detail to emphasize more precisely what Figure 1 shows. We also included data on K9me into Figure 1 figure supplement 1H.

3) Fig 1D. Could you confirm the high level of H3K27me1 on H3.3?

H3K27me1 data are shown both by WB (Figure 1C) and Mass spectrometry (Figure 1D and E). We also provide a possible explanation for high levels of this mark on H3.3 by taking into account the fact that H3K27me1 is also produced by demethylation of H3K27me3 by JMJ demethylases.

4) All WB in Fig 1. They need to be quantified and normalized (plus statistical analysis) in order to provide strong support to the conclusions.

The conclusion of all WB are supported by quantified Mass spectrometry data and many WB were even repeatedly shown in Figure 1F (for example IPs for H2A variants and a large set of H3 marks used for WBs) with the same results. Also, association of H3K4me3 and H3K36me3 with H2A variants was analyzed in both ways (Figure 1F); IPs of variants and WBs of variants and marks and IPs of marks and WBs of marks and variants. For most of the data we do not have more than two repeats, so statistical analysis may not be possible.

Nevertheless, we are convinced that our major conclusions from data presented in Figure 1 and Supporting figure 1 (these are: that H3 variants form both homotypic and heterotypic nucleosomes, that H3 marks do not preferentially associate with H3 variants but some of them do so with H2A variants and that H3 modifications show very complex pattern of associations with each other) are fully valid as they were drawn from two orthogonal approaches and further supported by the chromatin states identified.

5) Fig. 2A. Authors focus on "the most parsimonious model" based on 26 chromatin states. This needs to be justified in a more explicit manner. It is surprising that this number emerges for an analysis of 27 independent variants and marks. What are the differences in the conclusions when other number of states are used? See also below (reduced number of number derived from the "concatenated model").

Why 26 states were chosen is now explained in great details in the method section. Since to the exception of H2A variants that are invariably homotypic, nucleosomes can be heterotypic for all other histone variants and histone modifications, the random combination of the 27 marks in one nucleosome representing one states is 4 H2A (without the subtypes) x 4H3 x 2H1 x 2(power16) (for each mark) which is well above the circa 26 states observed. This shows that our probabilistic model reduces the potential complexity of a theorical random association in a remarkable manner.

6) As a summary, it would be very helpful to generated a table (or similar) where is proposed chromatin state is ascribed to functional genomic elements.

This aspect of the work is presented in a preprint where the biological association with the chromatin is described in details. See Jamge et al 2002, https://www.biorxiv.org/content/10.1101/2022.06.02.494419v1

7) Fig 2F (and S2B). A comprehensive comparison a various approaches should include others and estimate the Jaccard similarity index: (1) the same of marks and variants used in the Sequeira-Mendes et al paper, and (2) the subset of marks and variants added in this study. In this way, a direct evaluation of the contributions could be more properly made.

We thank the reviewer for this suggestion and have now included a new column with the combination of marks and variants as used in Sequeira-Mendes et al., 2014 (see Figure 2F). These data clearly demonstrate that adding histone variants significantly contribute to the definition of chromatin states.

8) Fig. 3. Explain in more detail the concatenated model used here. Does the reduction in the number of chromatin states mean that the other do not add new information?

ChromHMM concatenated model allows to identify common definition of chromatin state in multiple tissue types. Here multiple cell types are concatenated leading to a shared definition of chromatin states, but specific to each cell type.

In our paper we used the concatenated model to identify common chromatin states in two different genotypes (WT and ddm1). The data for WT and ddm1 was obtained from leaves. As we had a limited number of ChIP-seq profiles in the leaves dataset The complexity of the concatenated model was also reduced compared to the extensive 26 chromatin state model. We chose to analyze 16-states in the concatenated model because this was the minimal number of states that gave rise to a similar complexity of heterochromatic states.

9) The ddm1 mutant. The text in page 14 is a bit confusing. It seems that H2A.Z is deposited on TEs and the exchanged by the H2A.W.

We have provided additional alternative models that could explain our observations.

10) Page 15: link between H2A.Z and H3K27me3. Gomez-Zambrano et al (2018, cited in the text, found that only a relatively small subset of (putative) targets are common to H2A.Z and H3K27me3. How do authors reconcile this with their statement supporting a link between both of them?

We refer to Gomez-Zambranao et al to illustrate the link between H2A.Z and H2AK121ub so we do not understand this comment. The strong link between H2A.Z and H3K27me3 is shown without ambiguity by our work and also Carter et al., 2018.

Author Response:

Reviewer #1 (Public Review):

The study investigates the nature of "trailblazer" cells in distinct tumor models, including luminal B (MMTV/PyMT) and triple negative (TNBC) tumors (C3-TAg). The authors note that the trail-blazer phenotypes in the TNBC model are more complex relative to the Luminal B model and represent distinct EMT programs associated with the expression of distinct EMT-TFs (Zeb1, Zeb2 and Fra-1). They demonstrated that of numerous EMT-TFs, Zeb1 and Fra-1 were required for increased cancer cell migration and invasion. They reveal that TGF-beta and EGF-mediated signaling are required for the diverse EMT states that are required for trailblazer cell activity and increased cell migration/invasion. TGF-beta signaling engaged Zeb 1 and Zeb2 while EGF sig-naling activated Fra-1. Indeed, inhibitors of either TGF-beta or EGF signaling could impair cell migration/invasion. While both pathways contributed to trailblazer phenotypes, EGF signaling was shown to interfere with certain TGF-beta induced transcriptional response, including the ex-pression of genes encoding extracellular matrix proteins.

One concern was the heavy reliance of the C3-TAg as the sole TNBC model in which the dis-tinct trailblazer phenotypes were described. The data in Fig. 3 of the submission reveals that the phenotypes observed in the C3-TAg model could be recapitulated in a TNBC patient-derived xenograft model (PDX). Using this PDX, the authors were able to show vimentin expression in lung metastatic TNBC cells that were intravascular, those that had extravasated and clusters of cancer cells fully within the lung parenchyma. This was an important addition to the manuscript. The additional experiments to investigate the role of Zeb1 and Zeb1 more fully, beyond the focus on Fra-1 in the initial submission was an additional strength of the new submission. Additional clarifications to the discussion also clarified the concepts articulated in the study. The study em-ploys multiple breast cancer models, utilizes numerous in vitro and in vivo assessments of the trailblazer phenotypes, and the experimental design is rigorous and the interpretation of the data is sound. The manuscript will be of general interest to the research community.

Thank you for the supportive comments. We are glad that the revisions addressed your prior concerns.

Reviewer #2 (Public Review):

This represents an important study that demonstrates a high degree of heterogeneity within trailblazer cells in clusters that participate in collective migration. Solid methods highlight this het-erogeneity and show that in TNBC cancers, trailblazer cells are defined by vimentin (and not Keratin 14) and are dependent on both TGFbeta and EGFR signaling. Additional, single cell stud-ies would further support this work.

Thank you for the suggestion. Our current data establishes that trailblazer cells are heterogene-ous using FACS, immunostaining and functional studies of fresh tumor organoids and estab-lished tumor organoid lines. In addition, our RNA-seq experiments provided deep insight into the nature of gene expression changes that corresponded with the evolution of new trailblazer states. This discovery of trailblazer cell heterogeneity was one of multiple key new discoveries in this manuscript, along with revealing a Krt14-independent invasion mechanism, the regulation of trailblazer cells by Tgfβ and Egfr signaling and a new compromise mode of signal integration. We agree that our results support further investigation of the nature and function of basal-like breast cancer heterogeneity during the progression to metastasis. However, a comprehensive implementation of scRNA-seq is mostly likely required to further unravel new aspects of hetero-geneity that substantially advance upon the conclusions supported by our current data. Such an undertaking is beyond the scope of this investigation.

We agree that scRNA-seq would be confirmatory of trailblazer cell heterogeneity that has been demonstrated with multiple approaches rather than a new discovery of heterogeneity.

Strengths:

The paper highlights that collective migration, and the nature of trailblazer cells can be highly heterogeneous. This is important as it suggests that the ability to move between states may su-persede a singular phenotype.

The paper uses animal models and organoids and in several areas attempts to correlate find-ings to human tissues.

The experiments are logically described.

Reviewer #3 (Public Review):

Cancer is a disease of many faces and in particular, the ability of cancers cells to change their phenotypes and cell behaviors - cancer cell plasticity - is a major contributor to cancer lethality and therapeutic challenge of treating this disease. In this study, Nasir, Pearson et al., investigate tumor cell plasticity through the lens of invasive heterogeneity, and in particular in models of tri-ple-negative breast cancer (TNBC), a subtype of breast cancer with particularly poor clinical prognosis and more limited treatment modalities. Using organoid models in a variety of matrix systems, microscopy, and signaling pathway inhibitors, they find that invading TNBC breast tu-mors, primarily in the C31-Tag genetically engineered mouse model of TNBC, are composed of heterogeneous invasive/"trailblazer" type tumor cells that in many cases express vimentin, a classical intermediate filament marker of epithelial-mesenchymal transition, and reduced keratin-14, another filament marker of basal epithelial cells associated with collective invasion in differ-ent breast cancer models. Supportive genetic and pharmacologic evidence is provided that gen-eration of these cells is TGF-beta signaling pathway driven, likely in vivo from the surrounding tumor microenvironment, in accord with published studies in this space. Another important as-pect of this study is the good transcriptional evidence for multiple migratory states showing dif-fering degrees of partial overlap with canonical EMT programs, dependent on TGF-beta, and suggestive but at present incomplete understanding of a parallel program involving Egfr/Fra-1 mediated effects on invasion. When taken in context with other recent studies (Grasset et al. Science Translational Medicine 2022), these data are broadly supportive of concept of targeting vimentin-dependent invasion programs in TNBC tumors.

The core conclusions of this paper are generally supported by the data, but there are some conceptual and technical considerations that should be taken into account when interpreting this study. Specific comments:

1) The contribution of the different vimentin-positive trailblazer cells to distant metastasis was not directly confirmed in vivo in this study. Given the limited proliferative potential of many fully EMT'd cells and in light of recent studies indicating that invasion can be uncoupled from meta-static potential, it seems important to directly test whether the different C31-tag isolates, varying in invasive potential in this study, produce metastases and if so do metastases abundance corre-late with the invasive potential in 3D culture. The collection of lungs at 34 days post injection de-scribed in methods is too short to evaluate metastatic frequency.

We agree that it is important to determine the contribution of trailblazer cells towards metastatic dissemination. In this manuscript, we show that Vimentin expressing cells in a triple negative breast cancer (TNBC) PDX model disseminate to the lungs (Figure 3F). We have also shown that Vimentin expressing SUM159 breast cancer (BC) trailblazer cells spontaneously metasta-size to the lungs in previous publications (Fig. 2–figure supplement 1C) and (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767 and Maine et al, Oncotarget, 2016, 10.18632/oncotarget.7408). Notably, the depletion of genes specifically expressed in trailblazer cells reduced spontaneous metastasis without significantly impinging on primary tumor growth (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767 and Maine et al, Oncotarget, 2016, 10.18632/oncotarget.7408). Our new results in Figure 5D show that Tgfβ activates genes that define the trailblazer state in the metastatic SUM159 trailblazer cell model. Thus, features of the Tgfβ regulated trailblazer program in the C3-TAg cells is active in the SUM159 trailblazer model of spontaneous metastasis. In addition, commonly employed BC cell line metastasis models, such as MDAMB231 derivatives are highly mesenchymal (Fig. 2–figure supplement 1C) and (Kang et al, Cell, 2003, 10.1016/S1535-6108(03)00132-6 and Minn et al, Nature, 2005, 10.1038/nature03799, as examples).

It is not technically feasible to establish a correlation between the relative invasion of The C3-TAg GEMM primary tumors and spontaneous metastasis. C3-TAg GEMM primary tumors de-velop rapidly and the mice must be euthanized prior to the detection of metastasis. This limitation of the model is mentioned in the Results section “Trailblazer cells are specified by Vimentin ex-pression in basal-like breast cancer patient tumors”. The aggressive primary tumor growth and limited spontaneous metastasis of the the C3-TAg model has also been previously reported by others (Green et al, Oncogene, 2000, 10.1038/sj.onc.1203280). Surgical resection of the original primary tumor is not feasible option to allow metastases to form since additional tumors develop in multiple mammary glands.

In response to reviewer requests, we initiated the growth of orthotopic primary tumors from con-trol or Tgfβ treated 1339-org cells to address the relationship between induction of the trailblazer state and primary tumor cell dissemination. We had to euthanize the mice at day 34 (d34) be-cause tumors within both cohorts had reached the maximum permitted diameter of 2 cm. This will be indicated in the Methods section with revised text. We detected CTCs from the mice bearing control and Tgfβ treated 1339-org cell tumors. However, no micrometastases were de-tected, which is indicated in the text describing Figure 4–figure supplement 3A-B. Thus, per-forming surgical resection in new experiments would not be expected to allow the later detection of metastasis, as there did not appear to be DTCs in the lungs that could initiate colonization. In addition, we would have to resect the tumors prior to d34 to successfully and humanely remove the primary tumors, further reducing the odds of metastases developing. We will continue our work to identify an experimental balance that permits sufficient primary tumor growth to initiate spontaneous metastasis. However, the time scale of resolving this technical challenge is uncer-tain and we believe that our published analysis of trailblazer cell metastasis and new findings here showing the dissemination of Vimentin expressing cells in a PDX model addresses the question of whether Vimentin expressing trailblazer cells metastasize.

We agree that certain cell states induced by EMT programs can limit the proliferative potential of tumor cells. As described in the Introduction, we previously found that the induction of a trailblaz-er state in a subset of breast cancer cell line models triggers a collateral cost in fitness that limits the ability of trailblazer cells to initiate tumor growth (Westcott et al, Cancer Res, 2020, 10.1158/0008-5472.CAN-20-0014). The traits that distinguish trailblazer cells which are capable of tumor initiation and metastasis versus trailblazer cells with reduced fitness have begun to be delineated. Our prior report suggested that cells that were dependent on p63 for growth lost their proliferative capacity when converting to a trailblazer state (Westcott et al, Cancer Res, 2020, 10.1158/0008-5472.CAN-20-0014). C3-TAg cells are not dependent on p63 for growth, which is indicated by the vast majority of the tumor cells lacking p63 expression in primary tumors and primary tumor organoids (Westcott et al, Cancer Res, 2020, 10.1158/0008-5472.CAN-20-0014), similar to the metastatic SUM159 breast cancer cell line model. We were also able to derive clonal trailblazer cell lines that lacked detectable p63 expression from a C3-TAg tumor (Figure 2—figure supplement 1B) and grow organoids even when the limited extent of p63 expression was further reduced by Tgfβ (Figure 5C). Additionally, the persistent Tgfβ treated 1339-org cells, which were enriched for trailblazer cells and had reduced p63 expression, were capable of initiating primary tumor growth (Figure 4F). Together, these results indicate that C3-TAg trail-blazer cells are capable of initiating metastatic colonization. However, given the heterogeneity in trailblazer states that we discovered, it is possible that a subset of trailblazer cell states have re-duced proliferative capacity. Our analysis approach in this manuscript would not necessarily de-tect these low fitness trailblazer cells if they were a relatively small fraction of the total trailblazer population. We will clarify this point in the Discussion section in the revised manuscript. Our re-sults have begun to reveal mechanisms for the transcriptional regulation of trailblazer cell heter-ogeneity. We plan to continue delineating the regulatory programs conferring specific transcrip-tion state, defining approaches for the prospective isolation of distinct trailblazer subpopulations and determining trailblazer subpopulation specific biomarkers to understand the specific contri-bution of distinct trailblazer subpopulations towards metastasis. Given the scope of this analysis, it is not feasible to incorporate these future studies into this manuscript.

2) The invasion of cancer cells is dependent on 3D matrix composition. In other studies, collec-tive cancer invasion is performed in exclusively collagen type 1 gels or in other instances entirely in 3D reconstituted basement membrane gel, e.g. lung cancer invasion studies. In this study, the authors use a mixture composed of both matrices. Given the invasion suppressive effects of matrigel, particularly for epithelial type cells, further studies would be important to determine whether the invasion phenotypes seen in this study are generalizable across matrix environ-ments.

The invasion of C3-TAg and PyMT organoids embedded in a 100% pure reconstituted base-ment is shown in Fig. 1–figure supplement 1G. We will emphasize that trailblazer invasion was evaluated in multiple ECM compositions with revised text and figure graphic. We also provide images for the reviewer showing that C3-TAg organoids collectively invade in a pure Collagen I ECM. Importantly, these findings are consistent with our results showing that Vimentin express-ing cells are associated with basal-like mammary tumor cell invasion in the complex ECM of C3-TAg GEMM primary tumors (Figure 2G) and patient primary tumors (Figure 3D). Moreover, Vimentin expressing cells disseminated to the lungs in the TNBC PDX that we evaluated (Figure 3F).

The ECM composition selected for experiments is dictated by the experimental question(s) being addressed. It is unlikely that mammary tumor cells would only ever collectively invade through an ECM that is either pure Collagen I or pure reconstituted basement membrane (BM). Indeed, it has been proposed that mixtures of Collagen I and BM proteins best reconstitute the complexity of primary tumor ECM (Hooper et al, Methods Enzymol, 2006, 10.1016/S0076-6879(06)06049-6). In line this observation, mixtures of Collagen I and BM proteins have been routinely used for the past 20 years to define mechanisms of 3D invasion; Xiang and Muthuswamy, Methods En-zymol, 2006, 10.1016/S0076-6879(06)06054-X; Calvo et al, Nat Cell Biol, 2013 10.1038/ncb2756; and Kato et al, eLife, 2023, 10.7554/eLife.76520, as examples).

Consistent with the known complexity of the ECM in the tumor microenvironment (TME), we detect Collagen I and Collagen IV (a key component of experimental BM) in the TME of primary breast cancer tumor models (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767). Important-ly, we have found that a mixture of collagen I and experimentally derived BM proteins reliably reveals breast cancer trailblazer cell invasion mechanisms that promote the malignant progres-sion and metastasis of primary tumors and whose expression correlates with poor patient out-come (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767 and Westcott et al, Cancer Res, 2020, 10.1158/0008-5472.CAN-20-0014, as examples). Notably, the relative differences in trail-blazer and opportunist cell invasive phenotypes are not dictated by the ECM composition used in our 3D assays. We have previously tested the invasion of trailblazer and opportunist subpopula-tions in different ECM compositions using both spheroid vertical invasion assays (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767). Increasing collagen I concentration enhanced the rela-tive rate of trailblazer cell invasion, with trailblazer cells always showing a significantly enhanced invasion relative to opportunist cells.

The relationship between trailblazer and opportunist cells that we have detected in primary tu-mors is recapitulated when using mixtures of Collagen I and BM proteins in our past publications and in this manuscript. The clonal opportunist cell lines derived from a C3-TAg tumor expressed high levels of the transcription factor p63 (Figure 2–figure supplement 1A-B). We previously showed that p63 restricts induction of a trailblazer state in human breast cancer trailblazer cell lines (Westcott et al, Cancer Res, 2020, 10.1158/0008-5472.CAN-20-0014). Notably, we showed that p63 expressing C3-TAg cells were not able to initiate collective invasion in the same ECM composition used in our current manuscript. Moreover, p63 cells in primary C3-TAg tumors were noninvasive opportunist cells that were limited to trailing p63-low trailblazer cells when collective-ly invading in primary tumors and in organoids (Westcott et al, Cancer Res, 2020). We now show that p63 expressing opportunist cell lines are limited to invading behind primary C3-TAg trailblazer cells and trailblazer cell lines in our 3D invasion assays (Figure 1B and Figure 1–figure supplement 1D-E). Together, these results indicate that the ECM employed in our 3D assays reveals the mechanistic underpinnings of both trailblazer and opportunist cell invasion in primary tumors.

With respect to lung cancer invasion, leader cells that we would classify as trailblazer cells have been isolated from 2 non-small cell lung cancer cell line spheroid models grown in pure reconsti-tuted BM extract (Konen et al, Nat Comm, 2017, 10.1038/ncomms15078). However, it unclear whether these cell line derived NSCLC trailblazer cells are more intrinsically invasive than non-trailblazer siblings in primary NCSCLC tumors or if the traits associated cell line NSCLC trail-blazer cells are required for metastasis. These tests have never been reported to the best of our knowledge. Similarly, it is not clear whether these NSCLC cell line derived trailblazer cells reflect features of primary NSLC primary tumor cells, as we are unaware of any such comparisons be-ing reported. Thus, there is no reason to consider pure reconstituted BM to be an equivalent or preferred experimental option to define trailblazer cell features. Nevertheless, as we mentioned before, our discovery approach identifies trailblazer cells that are intrinsically more invasive than opportunist siblings across multiple ECM conditions, including pure reconstituted BM and, im-portantly, in primary tumors.

3) TGF-beta is well known to induce EMT. Although this study identifies potential transcriptional mediators of the invasion/trailblazer program, is this program reversible?

We have previously shown the breast cancer trailblazer cells can convert to an opportunist state, demonstrating that trailblazer states are reversible (Westcott et al, J Clin Invest, 2015, 10.1172/JCI77767). In this manuscript. we show that C3-TAg organoid lines derived in the Tgfbr1 inhibitor A83-01 have few if any cells with a trailblazer phenotype relative to C3-TAg pri-mary tumors, suggesting a reversion of the trailblazer state (Fig. 4C and Figure 4–figure sup-plement 2A-C). However, our results do not entirely rule out the possibility that only non-trailblazer cells grew to establish the organoid lines. Indeed, the problem of tracing phenotypic conversions when evaluating heterogeneous populations is a systemic challenge that extends beyond our analysis of trailblazer cells. Clearly defining the conversion rates for trailblazer cells will require multiple genetic markers to distinguish the different trailblazer states we have now identified, in addition to phenotypic and molecular analysis over multiple days, or possibly weeks. Thus, further definition of the rate of reversion of different trailblazer cells is worthy line of future investigation rather than a feasible objective of this study.

Author Response:

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

Reviewer #1 (Public Review):

This paper describes the discovery, functional analysis and structure of TcaP, a protein encoded by the Vibrio phage satellite PLE that forms a size-determining scaffold around PLE procapsids made from helper phage ICP1 structural proteins. The system displays a fascinating similarity to the P2/P4 system, which had previously been unique in its use of a size-determining external scaffolding protein, Sid. The work is interesting, comprehensive and of high quality. The presentation could be improved as listed in the suggestions below.

An interesting observation is that PLE appears to be dependent on small capsids for efficient transduction. This is not completely surprising if the element uses a cos site type mechanism for packaging, since this requires an integer number of genomes to be packaged when the capsid is full, and this might be more difficult to accomplish when the helper capsid is much larger than the satellite, as is the case with ICP1. The authors mention in a few places that this is the first known satellite to have this requirement. However, this is not quite correct: a similar defect was seen in phi12/SaPIbov5, where the large phi12 capsid was not quite the right size for either two or three copies of the wildtype ("unevolved") SaPIbov5 (Carpena et al. 2016).

We thank the reviewer for bringing up this point. First, we agree that for cos type packaging systems, this would not be surprising. However, ICP1 is a pac type phage and we have evidence that PLE is also a pac rather than a cos type packaging satellite; therefore, PLE is the first headful satellite to show such a defect. For cos packaging elements, both SaPIbov5 and P4, non-integer genome lengths have been shown to pack less efficiently into capsids as pointed out above and shown in Carpena et al 2016 and Shore 1978. However, in both of these cases, the genomes were manipulated to change their size, suggesting that naturally occurring cos satellites maintain their genome sizes to be proportional to their capsid sizes or in integer proportion to their helper capsids. We will include a short summary of these previous findings in the main text to provide context for the rare decreases in transduction efficiency reported in the cos satellites.

The authors present several micrographs showing capsids formed in the presence or absence of wildtype or mutant TcaP and CP (Fig. 1, Fig 2., Fig 3). However, each micrograph shows only a handful of particles of the "correct" size, in addition to a few shells that are aberrant or of a different size. I miss a more statistically rigorous enumeration of shells of different size (PLE or ICP1 sized, or different), empty vs. full, aberrant shells etc. This could be presented as a size distribution graph, a histogram or in table form.

We thank the reviewer for this recommendation and agree that it would add to the manuscript. We will quantify these particles and present the data in the main text.

In the abstract, the term "divergent satellite P4" is vague and unclear. Divergent from what? Probably they mean distinct from or unrelated to PLE. Please clarify.

Yes, we did mean unrelated to PLE, and we will clarify in the text.

How do they know that gp123 is a decoration protein? Was this previously determined, does it have (sequence) similarity to other known decoration proteins, or is it simply the most likely designation based on its position in the genome?

Gp123 was annotated based on its position. While there is sequence similarity to other annotated Vibrio phages’ decoration proteins, we will clarify in the text that Gp123 is a putative decoration protein.

Although the reconstruction and modeling statistics are good, it is difficult to assess the quality of the map and the model from the presented figures. Details of the density and FSC curves (half-map and model-to-map) should be shown. It is also difficult to see the TcaP structure and how it compares to Sid from the figures presented.

We will address this concern in the revised manuscript.

Introduction, Paragraph 3: "...which is the number of coat proteins divided by 60" is not strictly speaking the definition of T number. The T number corresponds to the number of subtriangles that one triangular face of the icosahedron is divided into. It corresponds to the number of coat proteins divided by 60 in the canonical case, but in tailed phages, 5 copies are removed to make way for the portal protein. (Other viruses could be described as having architecture corresponding to a specific T number, but with divergent numbers of subunits, e.g. adenoviruses or polyomaviruses.)

We agree that our simplified explanation of the T number is not entirely accurate and will modify the sentence appropriately.

Reviewer #2 (Public Review):

Phage satellites are fascinating elements that have evolved to hijack phages for induction, packaging, and transfer, promoting their widespread dissemination in nature. It is remarkable how different satellites use conserved strategies of parasitism, utilising unrelated proteins that perform similar roles in their cognate elements. In the current manuscript, Dr. Seed and coworkers elucidated the mechanism used by one family of satellites, the PLEs, to produce small capsids, a process that inhibits phage reproduction while increasing PLE transmission. The work is presented beautifully, and the results are astonishing. The authors identified the gene responsible for generating the small capsids, characterised its role in the PLE transfer and phage inhibition, and determined the structure of the PLE-sized small capsids. It is a truly impressive piece of work.

We thank the reviewer for their positive evaluation of our work.

Reviewer #3 (Public Review):

The manuscript by Boyd and co-authors "A Vibrio cholerae viral satellite maximizes its spread and inhibits phage by remodelling hijacked phage coat proteins into small capsids" reports important results related to self-defending mechanisms that bacteria are used against phages that infect them. It has been shown previously that bacteria produce phage-inducible chromosomal island-like elements (PLE) that encode proteins that are integrated into bacterial genome. These proteins are used by bacteria to amend the phage capsids and to create phage-like particles (satellites) that move between cells and transfer the genetic material of PLE to another bacteria. That study highlights the interactions between a PLE-encoded protein, TcaP, and capsid proteins of the phage ICP1.

The manuscript is well written, provides a lot of new information and the results are supported by biochemical analysis.

We thank the reviewer for their supportive evaluation of our work.

Author Response:

We would like to thank the reviewers for their time in evaluating our manuscript. The reviewers provided constructive comments and suggested changes to improve our manuscript. The main comment was about the framing. We agree with the reviewers and will rewrite the manuscript to focus more on migration patterns than conservation. We will add and expand the paper's theoretical framework and include the studies and descriptions of migration patterns of individual species suggested by the reviewers. At the same time, some of the reviewers' comments (especially on the terms and suggestions for changing the title of the paper) are mutually exclusive. We will pay particular attention to this issue and improve the paper's theoretical basis.

Author Response

Joint Public Review

Strengths

Overall, the idea that the PAG interacts with the BLA via the midline thalamus during a predator vs. foraging test is new and quite interesting. The authors have used appropriate tools to address their questions. The major impact in the field would be to add evidence to claims that the BLA can be downstream of the dPAG to evoke defensive behaviors. The study also adds to a body of evidence that the PAG mediates primal fear responses.

Weaknesses

(Anatomical concerns)

1) The authors claim that the recordings were performed in the dorsal PAG (dPAG), but the histological images in Fig. 1B and Supplementary S2 for example show the tip of the electrode in a different subregion of PAG (ventral/lateral). They should perform a more careful histological analysis of the recording sites and explain the histological inclusion and exclusion criteria. Diagrams showing the sites of all PAG and BLA recordings, as well as all fiber optics, would be helpful.

The PAG is composed of dorsomedial (dm), dorsolateral (dl), lateral (l), and ventrolateral (vl) columns that extend along the rostro-caudal axis of the aqueduct. The term “dorsal PAG” (dPAG) generally encompasses dmPAG, dlPAG, and lPAG, as substantiated by track-tracing, neurochemical, and immunohistochemical techniques (e.g., Bandler et al., 1991; Bandler & Keay, 1996; Carrive, 1993). As Bandler and Shipley (1994) summarized, “These findings suggest that what has been traditionally called the 'dorsal PAG' (a collective term for regions dorsal and lateral to the aqueduct), consists of three anatomically distinct longitudinal columns: dorsomedial and lateral columns…and a dorsolateral column…" Similarly, Schenberg et al. (2005) clarified in their review that, “According to this parcellation...the defensive behaviors (freezing, flight or fight) and aversion-related responses (switchoff behavior) were ascribed to the DMPAG, DLPAG, and LPAG (usually named the ‘dorsal’ PAG).” In our study, all recordings were conducted within the dPAG. Also, Figures 1B and S2 in our manuscript correspond to the -6.04 mm template from Paxinos & Watson’s atlas (1998), which is shown in the left panel in Author response image 1 and is considerably anterior to the location where the vlPAG emerges, as shown in the right panel. In our revised manuscript, we will provide a detailed definition of the dPAG, inclusive of dmPAG, dlPAG, and lPAG, and support this with the referenced literature.

Author response image 1.

2) Prior studies investigating the role of BLA neurons during a foraging vs. robot test similar to the one used in this study should be also cited and discussed (e.g., Amir et al 2019; Amir et al 2015). These two studies demonstrated that most neurons in the basal portion of the BLA exhibit inhibitory activity during foraging behavior and only a small fraction of neurons (~4%) display excitatory activity in response to the robot (in contrast to the 25% reported in the present study). A very accurate histological analysis of BLA recording sites should be performed to clarify whether distinct subregions of the BLA encode foraging and predator-related information, as previously shown in the two described studies.

In the revised manuscript, we will discuss papers by Amir et al. (2015) and Amir et al. (2019) that utilized a similar 'approach food-avoid predator' paradigm. These studies found a correlation between the neuronal activities in the basolateral amygdala (BL) and the velocity of animal movement during foraging, regardless of the presence or absence of predators. Specifically, the majority of BL neurons were inhibited in both conditions, with only 4.5% being responsive to predators. Consequently, Amir et al. posited that amygdala activity predominantly aligns with behavioral output such as foraging, rather than with responses to threats.

In contrast, our body of work (Kim et al., 2018; Kong et al., 2021; the present study) reveals that the majority of neurons in the BA/BLA displayed distinct responses in pre-robot and robot sessions. Kong et al. (2021) discussed in depth several factors that may account for this discrepancy, given that both Amir et al. and our research used similar behavioral paradigms. Differences in apparatus features, experimental procedures, and data analysis methodologies (refer to Amir et al., 2019) could be contributing to the conflicting results and interpretations concerning the significance of amygdalar neuronal activities.

Additionally, our studies uniquely monitored the same set of amygdalar neurons during pre-robot and robot sessions, affording us the opportunity for a direct comparison of neuronal activities under different threat conditions.

Another salient difference lines in the foraging success rates, which were markedly higher in Amir et al (~80%) compared to our studies (<3-4%). We hypothesize that there may be an inverse relationship between the pellet procurement rate and the intensity of fear. The high foraging success rate in Amir et al., which correlates with subdued amygdalar activity, stands in contrast to our findings of heightened amygdalar activity associated with a lower foraging success rate. Supporting this notion, optogeneticallyinduced amygdalar activity led naïve rats to abandon foraging and escape to the nest (Kong et al., 2021, the present study).

3) An important claim of this study that the PAG sends predator-related signals to BLA via the PVT (Fig. 4). The authors stated that PVT neurons labeled by intra-BLA injection of the retrograde tracer CTB were activated by the predator, but a proper immunohistochemical quantification with a control group was not provided to support this claim. To provide better support for their claim, the authors should quantify the doublelabeled PVT neurons (cFos plus CTB positive neurons) during the robot test.

As recommended, we will include a revised Fig. 4 in the manuscript to present the quantification of neurons that are double-labeled with c-Fos and CTB in the PVT. This updated figure will provide a more rigorous analysis and visual representation of the data.

4) The AVV anterograde tracer deposit spread to a large part of the PAG, including dorsolateral and lateral PAG, and supraoculomotor regions (Fig. 4B). Is the projection to the PVT from the dPAG or other regions of the PAG?

As previously addressed in response to Comment #1, the dPAG comprises the dmPAG, dlPAG, and lPAG. In the revised manuscript, we will acknowledge the diffusion of the AAV to the adjacent deep gray layer of the superior colliculus. Additionally, we are considering conducting more restricted AAV injections into the dPAG to verify terminal expressions in the PVT.

(Concerns about the strength of the evidence supporting a role for the PVT)

5) The authors conclude in the discussion section that the dPAG-amygdala pathway is involved in generating antipredatory defensive behavior. However, the current results are entirely based on correlational analyses of neural firing rate and there is no direct demonstration that the PAG provides information about the robot to the BLA. Therefore, the authors should tone down their interpretation or provide more evidence to support it by performing experiments applying inhibitory tools in the dPAG > PVT > BLA pathway and examining the impact on behavior and downstream neural firing.

As suggested, we will moderate the assertions about the functional implications of the PVT, based on the data from anterograde and retrograde tracers, to present a more measured interpretation in the manuscript.

(Other concerns)

6) One of the main findings of this study is the observation that BLA neurons that are responsive to PAG photostimulation are preferentially recruited during the foraging vs. robot test (Fig. 3). However, the experimental design used to address this question is problematic because the laser photostimulation of PAG neurons preceded the foraging vs. robot test. Prior photoactivation of PAG may have caused indirect shortterm synaptic plasticity in BLA cells, which would favor the response of these cells to the robot. Please see Oishi et al, 2019 PMID: 30621738, which demonstrated that 10 trains of 20Hz photoactivation (300 pulses each) was sufficient to induce LTP in brain slices.

After approximately eight photostimulation trials of the dPAG, with 40 pulses each, the animals entered a post-photostimulation testing phase (referred to as "Post"; Fig. 3C), lasting 10-15 minutes over an average of eight trials before robot testing. Although the PAG does not directly project to the BLA, the remote possibility of trans-synaptic plasticity in the BLA cannot be completely excluded and will be acknowledged. Additionally, it is noteworthy that Oishi et al's (2019) study applied a total of 3,000 pulses (i.e., 10 15-s trains of 20-Hz pulses) and investigated CA3-CA3 synaptic plasticity, as opposed to a total of 320 pulses (i.e., 8 2-s trains of 20-Hz pulses) in our study.

7) The authors should perform a longitudinal analysis of the behavioral responses of the rats across the trials to clarify whether the animals habituate to the robot or not. In Figure 1E, it appears that PAG neurons fire less across the trials, which could be associated with behavioral habituation to the predator robot. If that is the case, the activity of many other PAG and BLA neurons will also most likely vary according to the trial number, which would impact the current interpretation of the results.

In Figure 1E, the y-axis represents the Z scores of individual dPAG neurons, instead of representing repeated tests of the same neuron across multiple trials. The raster plot in Figure 1F clearly depicts that the same dPAG neurons consistently display heightened neural activity in response to the approaching robot across successive trials.

8) In Figure 1, it is unclear why the authors compared the activity of neurons that respond to the robot activation against the activity of the neurons during the retrieval of the food pellets in the pre-robot and postrobot sessions. The best comparison would be aligning the cells that were responsive to the activation of the robot with the moment in which the animals run back to the nest after consuming the pellets during the prerobot or post-robot sessions. This would enable the authors to demonstrate that the PAG responses are directly associated with the expression of escaping behavior in the presence of the robot rather than associated with the onset of goal-directed movement in direction to the next during the pre- and post-robot sessions. A graphic showing the correlation between PAG firing rate and escape response would be also informative.

Figure 1E compares the dPAG neural activity when animals enter a designated pellet zone (time-stamped by camera tracking) during both pre-robot and post-robot trials to the dPAG neural activity when entering the robot trigger zone (time-stamped by robot activation). We wish to clarify that rats carry the large (0.5 g) pellet back to the nest for consumption rather than consume it in the open arena before returning to the nest.

In our study, we aimed to investigate the direct response of dPAG neurons to the looming predator and explore the communication between dPAG and BLA in relation to antipredatory defensive responses. To build upon our previous research that suggests a potential role of dPAG in conveying such responses to the BLA (Kim et al., 2013) and the immediate firing of BLA neurons in response to predatory threats (Kim et al., 2018; Kong et al., 2021), we chose to narrow our testing window to a short latency period (< 500 ms) following robot activations. This specific time window allowed us to focus on the initial stages of the threat stimulus processing and minimize potential confounding factors such as the presence of residual firing activity triggered by the robot during the animals’ escape or any activity changes induced by the animals' behavior.

Furthermore, Figure S1C clearly demonstrates that (i) increased activity of dPAG robot cells preceded the animals’ actual turning and fleeing behavior toward the nest, as indicated by the peak values of movement speed (dark yellow), and (ii) the presence of pellets did not affect activity changes of the robot cells during pre- and post-robot sessions. These observations suggest that the heightened activity of dPAG robot cells was not due to movement changes or pellet motivation.

Lastly, as stated in the original manuscript, the vast majority of robot cells (90.9%) did not show significant correlations between movement speed and firing rates, lending further support to the interpretation that the dPAG activity observed was not merely a reflection of movement changes.

References

Bandler, R., Carrive, P., & Depaulis, A. (1991). Emerging principles of organization of the midbrain periaqueductal gray matter. The midbrain periaqueductal gray matter: functional, anatomical, and neurochemical organization, 1-8.

Bandler, R. & Keay, K. A. (1996). Columnar organization in the midbrain periaqueductal gray and the integration of emotional expression. Progress in brain research, 107, 285-300.

Bandler, R. & Shipley, M. T. (1994) Columnar organization in the midbrain periaqueductal gray: modules for emotional expression? Trends in Neurosciences, 17(9), 379-89.

Carrive, P. (1993). The periaqueductal gray and defensive behavior: functional representation and neuronal organization. Behavioural brain research, 58(1-2), 27-47.

Oishi, N., Nomoto, M., Ohkawa, N., Saitoh, Y., Sano, Y., Tsujimura, S., ... & Inokuchi, K. (2019). Artificial association of memory events by optogenetic stimulation of hippocampal CA3 cell ensembles. Molecular brain, 12, 1-10.

Paxinos, G. & Watson, C. (1998). The Rat Brain in Stereotaxic Coordinates. Academic Press, San Diego. Schenberg, L. C., Póvoa, R. M. F., Costa, A. L. P., Caldellas, A. V., Tufik, S., & Bittencourt, A. S. (2005). Functional specializations within the tectum defense systems of the rat. Neuroscience & Biobehavioral Reviews, 29(8), 1279-1298.

Author Response

We are grateful for the constructive feedback and the possibility of further improving our manuscript in terms of quality and clarity. Below, we have prepared a brief answer to the points raised in the reviewers’ feedback. We plan to address all these issues fully in the revised version of the manuscript.

We agree that some of our claims were overly enthusiastic. We will rewrite parts of the manuscript to tame our statements. Additionally, we are thankful for the comments on the use of language, which we will certainly apply while editing the manuscript. Below, we focus on the main comments.

Both reviewers: We appreciate advice on possible confounding factors. We should note here that there is substantial evidence on the effects of alpha rhythm amplitude on the excitability of a neuronal network and, as a consequence, on the amplitude of evoked responses (Baumgarten et al., 2016 Cerebral Cortex; Iemi et al., 2017 eLife; Stephani et al., 2021 eLife). This effect is due to changing the gain for evoked responses, and it is quite different compared to the baseline-shift mechanism (BSM). In BSM, the changes in the amplitude of evoked responses occur due to the generation of an additional evoked response component, which we tried to reveal in our current work. Still, we agree with suggestions to test additional factors, such as earlier evoked responses, baseline window, and head size, and we will test those.

Reviewer #2 Comment 2: Certainly, for low-density recordings, some method of data transformation is required. Here we would like to show our reasoning for why we did not use current-source density (CSD) but rather utilised other approaches. First, the CSD transform performs well for spatially localised activities since it is a spatial high-pass filter. In our case, P300 and alpha amplitude dynamics are fairly widespread with low spatial frequency, and we believe we would not benefit from applying CSD. Second, CSD has been shown to be more sensitive to surface sources in the crowns of gyri. For activity in the P300 window, we have no reason to believe that this is the case. Third, as we completely agree that low density montage is a limitation, we used source reconstruction with eLoreta (Fig. 5) to refine the spatial localisation of potential sources of P300 and alpha amplitude change.

Reviewer #1 Comment 4: Our study is indeed based on a sample of older participants. However, in our previous work (Studenova et al., 2022), we compared young and elderly participants using resting-state data. There, we measured the baseline-shift index (BSI). We found that BSIs for elderly participants were lower in comparison to those for young participants. Therefore, despite these limitations, in the current study, we were still able to detect a correspondence between BSIs and evoked responses in elderly participants. Therefore, we believe that for a sample of young participants, the results should not be different.

Reviewer #2 Comment 4: We agree that mediation analysis will provide additional insights, and we will add it to the revised version of the manuscript.

Overall, we found the reviewer's comments very helpful. We will update the manuscript accordingly.

Author Response:

We would like to thank the reviewers for their comments on the manuscript. The primary concern that they raised is that the imaging data are largely qualitative. This is a fair assessment, and we agree that a careful quantitative characterization of TF clustering with and without IDRs using high resolution imaging would provide valuable insight that would extend our findings. Our goal for this study was to conduct a high level survey of IDR localization, for which we believe a qualitative overview was sufficient. We hope that this work can serve as a useful foundation for future studies of the complex roles that IDRs play in TF function.

Author Response

Reviewer #1 (Public Review):

1) Only one PITAR siRNA was tested in majority of the experiments, which compromises the validity of the results. Some results are inconsistent. For example, Fig 2G indicates that PITAR siRNA caused G1 arrest. However, PITAR overexpression in the same cell line did not show any effect on cell cycle progression in Fig 5I.

We thank the reviewer for this comment. Indeed, we have used two siRNAs in experiments related to Fig. 2C, 2D, and 2E. Keeping the reviewer’s comment, we plan to reproduce the results of Fig. 2F, 2G, 2H, 2I, 5A, 5B, 5E, and supplementary Fig. 5A using additional siRNA targeting PITAR.

The reason for the fact that “PITAR silencing showed a robust G1 arrest, but PITAR overexpression failed to show any effect on the cell cycle profile” is as follows: since glioma cells overexpress PITAR (which keeps the p53 suppressed), silencing PITAR (which will elevate p53 levels) in glioma cells will show a robust phenotype in cell cycle profile (in the form of increase G1 arrest). In contrast, the overexpression of PITAR in glioma cells (which already has high levels of PITAR and hence drastically reduced p53 levels) is unlikely to show any significant change in the cell cycle profile. But, a phenotype for PITAR overexpression on cell cycle profile can be shown in DNA-damaged (which induces p53 levels) glioma cells. Indeed, we have done this experiment in Fig. 5L, which shows G2/M arrest (42.34%) induced by DNA damage is reduced significantly (19%) in PITAR overexpressed condition (34.42%). However, keeping reviewers' comments in the right spirit, we plan to repeat this experiment with appropriate modifications to arrive at a more robust phenotype for PITAR overexpression.

2) The conclusion that PITAR inactivates p53 through regulating TRIM28, which is highlighted in the title of the manuscript, is not supported by convincing results. Although the authors showed that a PITAR siRNA increased while PITAR overexpression decreased p53 level, the siRNA only marginally increased the stability of p53 (Fig 5E). The p53 ubiquitination level was barely affected by PITAR overexpression in Fig 5F. To convincingly demonstrate that PITAR regulates p53 through TRIM28, the authors need to show that this regulation is impaired/compromised in TRIM28-knockout conditions. The authors only showed that TRIM28 overexpression suppressed PITAR siRNAinduced increase of p53, which is not sufficient. Note that only one cell line was investigated in Fig 5.

To address this issue, we will overexpress PITAR in TRIM28 silenced cells to show the requirement of TRIM28 for PITAR to inhibit p53. In addition, we also plan to carry out PITAR silencing and overexpression experiments in another glioma cell line as recommended by the reviewer.

3) Another major weakness of this manuscript is that the authors did not provide any evidence indicating that the glioblastoma-promoting activities of PITAR were mediated by its regulation of p53 or TRIM28 (Fig 6 and Fig 7). Thus, the regulation of glioblastoma growth and the regulation of TRIM28/p53 appear to be disconnected.

We would like to respectfully disagree with the reviewer on this particular point. We have indeed provided the following evidence in the current version of the manuscript glioblastomapromoting activities of PITAR were mediated by its regulation of p53 or TRIM28.

A) In Fig. 6, we demonstrate that PITAR silencing-induced reduction in the neurosphere growth is accompanied by a reduction in TRIM28 RNA and an increase in the CDKN1A RNA without a change in p53 RNA levels. We also demonstrate that PITAR overexpression-induced neurosphere growth is accompanied by an increase in the TRIM28 RNA, and a decrease in CDKN1A RNA without a change in p53 RNA levels.

B) To add strength to the above results, we plan to do western blot experiments under similar conditions to demonstrate the appropriate changes in TRIM28, p53, and CDKN1A levels. Also, we will do a TRIM28 rescue experiment in RG5 neurosphere cells.

C) In supplementary Fig. 6 (related to Fig. 6), we show that PITAR silencing failed to decrease neurosphere growth in mutant p53 containing GSC line (MGG8).

D) In supplementary Fig. 7 (related to Fig. 6), we show that PITAR silencing failed to inhibit colony growth of p53-silenced U87 glioma cells (U87/shp53#1). We also show that while PITAR silencing decreased TRIM28 RNA levels in U87/shNT and U87/shp53#1 glioma cells, it failed to increase CDKN1A and MDM2 (p53 targets) at the RNA level.

E) In Fig. 7, we show that the TRIM28 protein level is drastically reduced in small tumors formed by U87/siPITAR cells.

F) In supplementary Fig. 8 (related to Fig. 7), we show that glioma tumor formed by U87/PITAR OE express high levels of TRIM28 protein but reduced levels of p21 protein.

G) We also plan to do additional experiments, as described below, to demonstrate that glioblastoma-promoting activities of PITAR are indeed mediated by its regulation of p53 or TRIM28. We will demonstrate the inability of PITAR overexpression to induce the growth of glioma-tumor initiated by TRIM28 silenced U87 cells.

4) It is not clear what kind of message the authors tried to deliver in Fig 7F/G. Based on the authors' hypothesis, DNA-damaging agents like TMZ would induce PITAR to inactivate p53, which would compromise TMZ's anti-cancer activity. However, the data show that TMZ was very effective in the inhibition of U87 growth. The authors may need to test whether PITAR downregulation, which would increase p53 activity, have any effects on TMZ-insensitive tumors. Such results are more therapeutically relevant.

Reviewer #1 rightly pointed out that TMZ induces PITAR expression, which should compromise TMZ's anti-cancer activity. In addition, overexpression of PITAR also promotes glioma-tumor growth. Figure 7F&G demonstrates the following two facts:1. PITAR overexpression increases the glioma-tumor growth (Figure 7G, compare red line with the blue line), 2. PITAR overexpressing glioma-tumor are resistant to TMZ chemotherapy (Figure 7G, compare the pink line with the green line).

In addition, in Figure 2I, we indeed show that PITAR-silenced cells are more sensitive to TMZ and Adriamycin chemotherapy.

However, considering reviewers’ comments, we plan to repeat Figure 7A, combining TMZ chemotherapy and PITAR silencing to demonstrate that TMZ chemotherapy-induced PITAR indeed promotes chemo-resistance.

5) Lastly, the model presented in Fig 7H is confusing. It is not clear what the exact role of PITAR in the DNA damage response based on this model. If DNA damage would induce PITAR expression, this would lead to inactivation of p53 as revealed by this manuscript. However, DNA damage is known to activate p53. Do the authors want to imply that PITAR induction by DNA damage would help to bring down the p53 level at the end of DNA damage response? The presented data do not support this role unfortunately.

We appreciate reviewer #1 comments. Based on our model in 7H, we believe DNA damageinduced PITAR attenuates DNA damage response by increasing TRIM28 protein levels. TRIM28 ubiquitinates p53 in an MDM2-dependent manner ( Wang et al., 2005). Based on this, we hypothesised that PITAR-induced TRIM28 also contributes to MDM2 mediated ending of DNA damage response.

Considering the reviewers' comments, we plan to do the following experiment.

The kinetics of p53, TRIM28, p21, MDM2 protein levels, and PITAR RNA levels after DNA damage will be monitored in PITAR-silenced conditions. It is known that reduction in the DNA damage-induced p53 levels coincides with high levels of MDM2 accumulation. We believe that in PITAR-silenced cells, p53 levels will remain high for a longer time compared to control cells because of the lack of PITAR-induced TRIM28-mediated degradation of p53.

Author Response:

Reviewer #1 (Public Review):

[…] The major strength of the study is the elegant and well-powered data set. Longitudinal data on this scale is very difficult to collect, especially with patient cohorts, so this approach represents an exciting breakthrough. Analysis is straightforward and clearly presented. However, no multiple comparison correction is applied despite many different tests. While in general I am not convinced of the argument in the citation provided to justify this, I think in this case the key results are not borderline (p<0.001) and many of the key effects are replications, so there are not so many novel/exploratory hypothesis and in my opinion the results are convincing and robust as they are. The supplemental material is a comprehensive description of the data set, which is a useful resource.

The authors achieved their aims, and the results clearly support the conclusion that the AD and mean confidence in a perceptual task covary longitudinally. I think this study provides an important impact to the project of computational psychiatry.Sspecifically, it shows that the relationship between transdiagnostic symptom dimensions and behaviour is meaningful within as well as across individuals.

Response: We thank the reviewer for their appraisal of our paper and positive feedback on the main manuscript and supplementary information. We agree with the reviewer that the lack of multiple comparison corrections can also justified by key findings being replications and not borderline significance. We have added this additional justification to the manuscript (Methods, Statistical Analyses, page 15, line 568: “Adjustments for multiple comparisons were not conducted for analyses of replicated effects”)

Reviewer #2 (Public Review):

[…] The major strength and contribution of this study is the use of a longitudinal intervention design, allowing the investigation of how the well-established link between underconfidence and anxious-depressive symptoms changes after treatment. Furthermore, the large sample size of the iCBT group is commendable. The authors employed well-established measures of metacognition and clinical symptoms, used appropriate analyses, and thoroughly examined the specificity of the observed effects.

However, due to the small effect sizes, the antidepressant and control groups were underpowered, reducing comparability between interventions and the generalizability of the results. The lack of interaction effect with treatment makes it harder to interpret the observed differences in confidence, and practice effects could conceivably account for part of the difference. Finally, it was not completely clear to me why, in the exploratory analyses, the authors looked at the interaction of time and symptom change (and group), since time is already included in the symptom change index.

Response: We thank the reviewer for their succinct summary of the main results and strengths of our study. We apologise for the confusion in how we described that analysis. We examine state-dependence., i.e. the relationship between symptom change and metacognition change, in two ways in the paper – perhaps somewhat redundantly. (1) By correlating change indices for both measures (e.g. as plotted in Figure 3D) and (2) by doing a very similar regression-based repeated-measures analysis, i.e. mean confidence ~ time*anxious-depression score change. Where mean confidence is entered with two datapoints – one for pre- and one for post-treatment (i.e. within-person) and anxious-depression change is a single value per person (between-person change score). This allowed us to test if those with the biggest change in depression had a larger effect of time on confidence. This has been added to the paper for clarification (Methods, Statistical Analysis, page 14, line 553-559: “To determine the association between change in confidence and change in anxious-depression, we used (1) Pearson correlation analysis to correlate change indices for both measures and, (2) regression-based repeated-measures analysis: mean confidence ~ time*anxious-depression score change, where mean confidence is entered with two datapoints (one for pre- and one for post-treatment i.e., within-person) and anxious-depression change is a single value per person (between-person change score)”).

The analyses have also been reported as regression in the Results for consistency (Treatment Findings: iCBT, page 5, line 197-204: ‘To test if changes in confidence from baseline to follow-up scaled with changes in anxious-depression, we ran a repeated measure regression analyses with per-person changes in anxious-depression as an additional independent variable. We found this was the case, evidenced by a significant interaction effect of time and change in anxious-depression on confidence (b=-0.12, SE=0.04, p=0.002)… This was similarly evident in a simple correlation between change in confidence and change in anxious-depression (r(647)=-0.12, p=0.002)”).

This longitudinal study informs the field of metacognition in mental health about the changeability of biases in confidence. It advances our understanding of the link between anxiety-depression and underconfidence consistently found in cross-sectional studies. The small effects, however, call the clinical relevance of the findings into question. I would have found it useful to read more in the discussion about the implications of the findings (e.g., why is it important to know that the confidence bias is state-dependent; given the effect size of the association between changes in confidence and symptoms, is the state-trait dichotomy the right framework for interpreting these results; suggestions for follow-up studies to better understand the association).

Response: Thank you for this comment. We have elaborated on the implications of our findings in the Discussion, including the relevance of the state-trait dichotomy to future research and how more intensive, repeated testing may inform our understanding of the state-like nature of metacognition (Discussion, Limitations and Future Directions, page 10, line 378-380: “More intensive, repeating testing in future studies may also reveal the temporal window at which metacognition has the propensity to change, which could be more momentary in nature.”).

Reviewer #3 (Public Review):

[…] I think these findings are exciting because they directly relate to one of the big assumptions when relating cognition to mental health - are we measuring something that changes with treatment (is malleable), so might be mechanistically relevant, or even useful as a biomarker?

This work is also useful in that it replicates a finding of heightened confidence in those with compulsivity, and lowered confidence in those with elevated anxious-depression.

One caveat to the interest of this work is that it doesn't allow any causal conclusions to be drawn, and only measures two timepoints, so it's hard to tell if changes in confidence might drive treatment effects (but this would be another study). The authors do mention this in the limitations section of the paper.

Another caveat is the small sample in the antidepressant group.

Some thoughts I had whilst reading this paper: to what extent should we be confident that the changes are not purely due to practice? I appreciate there is a relationship between improvement in symptoms and confidence in the iCBT group, but this doesn't completely rule out a practice effect (for instance, you can imagine a scenario in which those whose symptoms have improved are more likely to benefit from previously having practiced the task).

Response: We thank the reviewer for commenting on the implications of our findings and we agree with the caveats listed. We thank the reviewer for raising this point about practice effects. A key thing to note is that this task does not have a learning element with respect to the core perceptual judgement (i.e., accuracy), which is the target of the confidence judgment itself. While there is a possibility of increased familiarity with the task instructions and procedures with repeated testing, the task is designed to adjust the difficulty to account of any improvements, so accuracy is stable. We see that we may not have made this clear in some of our language around accuracy vs. perceptual difficulty and have edited the Results to make this distinction clearer (Treatment Findings: iCBT, pages 4-5, lines 184-189: “Although overall accuracy remained stable due to the staircasing procedure, participants’ ability to detect differences between the visual stimuli improved. This was reflected as the overall increase in task difficulty to maintain the accuracy rates from baseline (dot difference: M=41.82, SD=11.61) to follow-up (dot difference: M=39.80, SD=12.62), (b=-2.02, SE=0.44, p<0.001, r2\=0.01)”.)

However, it is true that there can be a ‘practice’ effect in the sense that one may feel more confident (despite the same accuracy level) due to familiarity with a task. One reason we do not subscribe to the proposed explanation for the link between anxious-depression change and confidence change is that the other major aspect of behaviour that improved with practice did so in a manner unrelated to clinical change. As noted above in the quoted text, participants’ discrimination improved from baseline to follow-up, reflected in the need for higher difficulty level to maintain accuracy around 70%. Crucially, this was not associated with symptom change. This speaks against a general mechanism where symptom improvement leads to increased practice effects in general. Only changes in confidence specifically are associated with improved symptoms. We have provided more detail on this in the Discussion (page 9, lines 324-326: “This association with clinical improvements was specific to metacognitive changes, and not changes in task performance, suggesting that changes in confidence do not merely reflect greater task familiarity at follow-up.”).

Relatedly, to what extent is there a role for general task engagement in these findings? The paper might be strengthened by some kind of control analysis, perhaps using (as a proxy for engagement) the data collected about those who missed catch questions in the questionnaires.

Response: Thank you for your comment. We included the details of data quality checks in the Supplement. Given the small number of participants that failed more than one attention checks (1% of the iCBT arm) and that all those participants passed the task exclusion criteria, we made the decision to retain these individuals for analyses. We have since examined if excluding these small number of individuals impacts our findings. Excluding those that failed more than one catch item did not affect the significance of results, which has now been added to the Supplementary Information (Data Quality Checks: Task and Clinical Scales, page 5, lines 181-185: “Additionally, excluding those that failed more than one catch item in the iCBT arm did not affect the significance of results, including the change in confidence (b=0.16, SE=0.02, p<0.001), change in anxious-depression (b=-0.32, SE=0.03, p<0.001), and the association between change in confidence and change in anxious-depression (r(638)=-0.10, p=0.011)”).

I was also unclear what the findings about task difficulty might mean. Are confidence changes purely secondary to improvements in task performance generally - so confidence might not actually be 'interesting' as a construct in itself? The authors could have commented more on this issue in the discussion.

Response: Thank you for this comment and sorry it was not clear in the original paper. As we discussed in a prior reply, accuracy – i.e. proportion of correct selections (the target of confidence judgements) are different from the difficulty of the dot discrimination task that each person receives on a given trial. We had provided more details on task difficulty in the Supplement. Accuracy was tightly controlled in this task using a ‘two-down one-up’ staircase procedure, in which equally sized changes in dot difference occurred after each incorrect response and after two consecutive correct responses. The task is more difficult when the dot difference between stimuli is lower, and less difficult when the dot difference between stimuli is greater. Therefore, task difficulty refers to the average dot difference between stimuli across trials. Crucially, task accuracy did not change from baseline to follow-up, only task difficulty. Moreover, changes in task difficulty were not associated with changes in anxious-depression, while changes in confidence were, indicating confidence is the clinically relevance construct for change in symptoms.

We appreciate that this may not have been clear from the description in the main manuscript, and have added more detail on task difficulty to the Methods (Metacognition Task, page 14, lines 540-542: “Task difficulty was measured as the mean dot difference across trials, where more difficult trials had a lower dot difference between stimuli.”) and Results (Treatment Findings: iCBT, pages 4-5, lines 184-186: “Although overall accuracy remained stable due to the staircasing procedure, participants’ ability to detect differences between the visual stimuli improved.”). We have also elaborated more on how improvements in symptoms are associated with change in confidence, not task performance in the Discussion (page 9, lines 324-326: “This association with clinical improvements was specific to metacognitive changes, and not changes in task performance, suggesting that changes in confidence do not merely reflect greater task familiarity at follow-up”).

To make code more reproducible, the authors could have produced an R notebook that could be opened in the browser without someone downloading the data, so they could get a sense of the analyses without fully reproducing them.

Response: Thank you for your comment. We appreciate that an R notebook would be even better than how we currently share the data and code. While we will consider using Notebooks in future, we checked and converting our existing R script library into R Notebooks would require a considerable amount of reconfiguration that we cannot devote the time to right now. We hope that nonetheless the commitment to open science is clear in the extensive code base, commenting and data access we are making available to readers.

Rather than reporting full study details in another publication I would have found it useful if all relevant information was included in a supplement (though it seems much of it is). This avoids situations where the other publication is inaccessible (due to different access regimes) and minimises barriers for people to fully understand the reported data.

Response: We agree this is good practice – the Precision in Psychiatry study is very large, with many irrelevant components with respect to the present study (Lee et al., BMC Psychiatry, 2023). For this reason, we tried to provide all that was necessary and only refer to the Precision in Psychiatry study methods for fine-grained detail. Upon review, the only thing we think we omitted that is relevant is information on ethical approval in the manuscript, which we have now added (Methods, Participants, page 11, lines 412-417: “Further details of the PIP study procedures that are not specific to this study can be found in a prior publication (21). Ethical approval for the PIP study was obtained from the Research Ethics Committee of School of Psychology, Trinity College Dublin and the Northwest-Greater Manchester West Research Ethics Committee of the National Health Service, Health Research Authority and Health and Care Research Wales”). If any further information is lacking, we are happy to include it here also.

Author Response

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

Reviewer #1 (Public Review):

She et al studied the evolution of gene expression reaction norms when individuals colonise a new environment that exposes them to physiologically challenging conditions. Their objective was to test the "plasticity first" hypothesis, which suggest that traits that are already plastic (their value changes when facing a new environment compared to the original environment) facilitates the colonisation of novel environments, which, if true, would be predicted to result in the evolution of gene expression values that are similar in the population that colonised the new environment and evolved under these particular selection pressures. To test this prediction, they studied gene expression in cardiac and muscle tissues in individuals originating from three conditions: lowland individuals in their natural environment (ancestral state), lowland individuals exposed to hypoxia (the plastic response state), and a highland population facing hypoxia for several generations (the coloniser state). They classified gene expression patterns as maladaptive or adaptive in lowland individuals responding to short term hypoxia by classifying gene expression patterns using genes that differed between the ancestral state (lowland) and colonised state (highland). Genes expressed in the same direction in lowland individuals facing hypoxia (the plastic state) as what is found in the colonised state are defined as adaptative, while genes with the opposite expression pattern were labelled as maladaptive, using the assumption that the colonised state must represent the result of natural selection. Furthermore, genes could be classified as representing reversion plasticity when the expression pattern differed between the plasticity and colonised states and as reinforcement when they were in the same direction (for example more expressed in the plastic state and the colonised state than in the ancestral state). They found that more genes had a plastic expression pattern that was labelled as maladaptive than adaptive. Therefore, some of the genes have an expression pattern in accordance with what would be predicted based on the plasticity-first hypothesis, while others do not.

Thank you for a precise summary of our work. We appreciate the very encouraging comments recognizing the value of our work. We have addressed concerns from the reviewer in greater detail below.

Q1. As pointed out by the authors themselves, the fact that temperature was not included as a variable, which would make the experimental design much more complex, misses the opportunity to more accurately reflect the environmental conditions that the colonizer individuals face at high altitude. Also pointed out by the authors, the acclimation experiment in hypoxia lasted 4 weeks. It is possible that longer term effects would be identifiable in gene expression in the lowland individuals facing hypoxia on a longer time scale. Furthermore, a sample size of 3 or 4 individuals per group depending on the tissue for wild individuals may miss some of the natural variation present in these populations. Stating that they have a n=7 for the plastic stage and n= 14 for the ancestral and colonized stages refers to the total number of tissue samples and not the number of individuals, according to supplementary table 1.

We shared the same concerns as the reviewer. This is partly because it is quite challenging to bring wild birds into captivity to conduct the hypoxia acclimation experiments. We had to work hard to perform acclimation experiments by taking lowland sparrows in a hypoxic condition for a month. We indeed have recognized the similar set of limitations as the review pointed out and have discussed the limitations in the study, i.e., considering hypoxic condition alone, short time acclimation period, etc. Regarding sample sizes, we have collected cardiac muscle from nine individuals (three individuals for each stage) and flight muscle from 12 individuals (four individuals for each stage). We have clarified this in Supplementary Table 1.

Q2. Finally, I could not find a statement indicating that the lowland individuals placed in hypoxia (plastic stage) were from the same population as the lowland individuals for which transcriptomic data was already available, used as the "ancestral state" group (which themselves seem to come from 3 populations Qinghuangdao, Beijing, and Tianjin, according to supplementary table 2) nor if they were sampled in the same time of year (pre reproduction, during breeding, after, or if they were juveniles, proportion of males or females, etc). These two aspects could affect both gene expression (through neutral or adaptive genetic variation among lowland populations that can affect gene expression, or environmental effects other than hypoxia that differ in these populations' environments or because of their sexes or age). This could potentially also affect the FST analysis done by the authors, which they use to claim that strong selective pressure acted on the expression level of some of the genes in the colonised group.

The reviewer asked how individual tree sparrows used in the transcriptomic analyses were collected. The individuals used for the hypoxia acclimation experiment and represented the ancestral lowland population were collected from the same locality (Beijing) and at the same season (i.e., pre-breeding) of the year. They are all adults and weight approximately 18g. We have clarified this in the Supplementary Table S1 and Methods. We did not distinguish males from females (both sexes look similar) under the assumption that both sexes respond similarly to hypoxia acclimation in their cardiac and flight muscle gene expression.

The Supplementary Table 2 lists the individuals that were used for sequence analyses. These individuals were only used for sequence comparisons but not for the transcriptomic analyses. The population genetic structure analyzed in a previously published study showed that there is no clear genetic divergence within the lowland population (i.e., individuals collected from Beijing, Tianjing and Qinhuangdao) or the highland population (i.e., Gangcha and Qinghai Lake). In addition, there was no clear genetic divergence between the highland and lowland populations (Qu et al. 2020).

Q4. Impact of the work

There has been work showing that populations adapted to high altitude environments show changes in their hypoxia response that differs from the short-term acclimation response of lowland population of the same species. For example, in humans, see Erzurum et al. 2007 and Peng et al. 2017, where they show that the hypoxia response cascade, which starts with the gene HIF (Hypoxia-Inducible Factor) and includes the EPO gene, which codes for erythropoietin, which in turns activates the production of red blood cell, is LESS activated in high altitude individuals compared to the activation level in lowland individuals (which gives it its name). The present work adds to this body of knowledge showing that the short-term response to hypoxia and the long term one can affect different pathways and that acclimation/plasticity does not always predict what physiological traits will evolve in populations that colonize these environments over many generations and additional selection pressure (UV exposure, temperature, nutrient availability). Altogether, this work provides new information on the evolution of reaction norms of genes associated with the physiological response to one of the main environmental variables that affects almost all animals, oxygen availability. It also provides an interesting model system to study this type of question further in a natural population of homeotherms.

Erzurum, S. C., S. Ghosh, A. J. Janocha, W. Xu, S. Bauer, N. S. Bryan, J. Tejero et al. "Higher blood flow and circulating NO products offset high-altitude hypoxia among Tibetans." Proceedings of the National Academy of Sciences 104, no. 45 (2007): 17593-17598.

Peng, Y., C. Cui, Y. He, Ouzhuluobu, H. Zhang, D. Yang, Q. Zhang, Bianbazhuoma, L. Yang, Y. He, et al. 2017. Down-regulation of EPAS1 transcription and genetic adaptation of Tibetans to high-altitude hypoxia. Molecular biology and evolution 34:818-830.

Thank you for highlighting the potential novelty of our work in light of the big field. We found it very interesting to discuss our results (from a bird species) together with similar findings from humans. In the revised version of manuscript, we have discussed short-term acclimation response and long-term adaptive evolution to a high-elevation environment, as well as how our work provides understanding of the relative roles of short-term plasticity and long-term adaptation. We appreciate the two important work pointed out by the reviewer and we have also cited them in the revised version of manuscript.

Reviewer #2 (Public Review):

This is a well-written paper using gene expression in tree sparrow as model traits to distinguish between genetic effects that either reinforce or reverse initial plastic response to environmental changes. Tree sparrow tissues (cardiac and flight muscle) collected in lowland populations subject to hypoxia treatment were profiled for gene expression and compared with previously collected data in 1) highland birds; 2) lowland birds under normal condition to test for differences in directions of changes between initial plastic response and subsequent colonized response. The question is an important and interesting one but I have several major concerns on experimental design and interpretations.

Thank you for a precise summary of our work and constructive comments to improve this study. We have addressed your concerns in greater detail below.

Q1. The datasets consist of two sources of data. The hypoxia treated birds collected from the current study and highland and lowland birds in their respective native environment from a previous study. This creates a complete confounding between the hypoxia treatment and experimental batches that it is impossible to draw any conclusions. The sample size is relatively small. Basically correlation among tens of thousands of genes was computed based on merely 12 or 9 samples.

We appreciate the critical comments from the reviewer. The reviewer raised the concerns about the batch effect from birds collected from the previous study and this study. There is an important detail we didn’t describe in the previous version. All tissues from hypoxia acclimated birds and highland and lowland birds have been collected at the same time (i.e., Qu et al. 2020). RNA library construction and sequencing of these samples were also conducted at the same time, although only the transcriptomic data of lowland and highland tree sparrows were included in Qu et al. (2020). The data from acclimated birds have not been published before.

In the revised version of manuscript, we also compared log-transformed transcript per million (TPM) across all genes and determined the most conserved genes (i.e., coefficient of variance ≤  0.3 and average TPM ≥ 1 for each sample) for the flight and cardiac muscles, respectively (Hao et al. 2023). We compared the median expression levels of these conserved genes and found no difference among the lowland, hypoxia-exposed lowland, and highland tree sparrows (Wilcoxon signed-rank test, P<0.05). As these results suggested little batch effect on the transcriptomic data, we used TPM values to calculate gene expression level and intensity. This methodological detail has been further clarified in the Methods and we also provided a new supplementary Figure (Figure S5) to show the comparative results.

The reviewer also raised the issue of sample size. We certainly would have liked to have more individuals in the study, but this was not possible due to the logistical problem of keeping wild bird in a common garden experiment for a long time. We have acknowledged this in the manuscript. In order to mitigate this we have tested the hypothesis of plasticity following by genetic change using two different tissues (cardiac and flight muscles) and two different datasets (co-expressed gene-set and muscle-associated gene-set). As all these analyses show similar results, they indicate that the main conclusion drawn from this study is robust.

Q2. Genes are classified into two classes (reversion and reinforcement) based on arbitrarily chosen thresholds. More "reversion" genes are found and this was taken as evidence reversal is more prominent. However, a trivial explanation is that genes must be expressed within a certain range and those plastic changes simply have more space to reverse direction rather than having any biological reason to do so.

Thank you for the critical comments. There are two questions raised we should like to address them separately. The first concern centered on the issue of arbitrarily chosen thresholds. In our manuscript, we used a range of thresholds, i.e., 50%, 100%, 150% and 200% of change in the gene expression levels of the ancestral lowland tree sparrow to detect genes with reinforcement and reversion plasticity. By this design we wanted to explore the magnitudes of gene expression plasticity (i.e., Ho & Zhang 2018), and whether strength of selection (i.e., genetic variation) changes with the magnitude of gene expression plasticity (i.e., Campbell-Staton et al. 2021).

As the reviewer pointed out, we have now realized that this threshold selection is arbitrarily. We have thus implemented two other categorization schemes to test the robustness of the observation of unequal proportions of genes with reinforcement and reversion plasticity. Specifically, we used a parametric bootstrap procedure as described in Ho & Zhang (2019), which aimed to identify genes resulting from genuine differences rather than random sampling errors. Bootstrap results suggested that genes exhibiting reversing plasticity significantly outnumber those exhibiting reversing plasticity, suggesting that our inference of an excess of genes with reversion plasticity is robust to random sampling errors. We have added these analyses to the revised version of manuscript, and provided results in the Figure 2d and Figure 3d.

In addition, we adapted a bin scheme (i.e., 20%, 40% and 60% bin settings along the spectrum of the reinforcement/reversion plasticity). These analyses based on different categorization schemes revealed similar results, and suggested that our inference of an excess of genes with reversion plasticity is robust. We have provided these results in the Supplementary Figure S2 and S4.

The second issue that the reviewer raised is that the plastic changes simply have more space to reverse direction rather than having any biological reason to do so. While a causal reason why there are more genes with expression levels being reversed than those with expression levels being reinforced at the late stages is still contentious, increasingly many studies show that genes expression plasticity at the early stage may be functionally maladapted to novel environment that the species have recently colonized (i.e., lizard, Campbell-Staton et al. 2021; Escherichia coli, yeast, guppies, chickens and babblers, Ho and Zhang 2018; Ho et al. 2020; Kuo et al. 2023). Our comparisons based on the two genesets that are associated with muscle phenotypes corroborated with these previous studies and showed that initial gene expression plasticity may be nonadaptive to the novel environments (i.e., Ghalambor et al. 2015; Ho & Zhang 2018; Ho et al. 2020; Kuo et al. 2023; Campbell-Staton et al. 2021).

Q3. The correlation between plastic change and evolved divergence is an artifact due to the definitions of adaptive versus maladaptive changes. For example, the definition of adaptive changes requires that plastic change and evolved divergence are in the same direction (Figure 3a), so the positive correlation was a result of this selection (Figure 3d).

The reviewer raised an issue that the correlation between plastic change and evolved divergence is an artifact because of the definition of adaptive versus maladaptive changes, for example, Figure 3d. We agree with the reviewer that the correlation analysis is circular because the definition of adaptive and maladaptive plasticity depends on the direction of plastic change matched or opposed that of the colonized tree sparrows. We have thus removed previous Figure 3d-e and related texts from the revised version of manuscript. Meanwhile, we have changed Figure 3a to further clarify the schematic framework.

Reviewer #1 (Recommendations For The Authors):

Q1. Here are private recommendations that I think could help improve the manuscript. West-Eberhard was a pioneer back in 2003 in explicating the hypothesis of "plasticity first". I think it is important to cite their main work in the first paragraph of introduction and to use the term "plasticity-first", which is widely known among evolutionary biologists studying phenotypic plasticity, instead of "plasticity followed by genetic change", since the three papers cited in paragraph 1 call it « plasticity first ».

West-Eberhard, M.J. (2003) Developmental Plasticity and Evolution, Oxford University Press.

Thank you for suggesting West-Eberhard (2003) and we have cited this important work. We have also changed “plasticity followed by genetic change” to “plasticity first”.

Q2. Introduction. Line 5, Change for « On the one hand, if plasticity changes ... »

We have modified as suggested.

Q3. Line 52, Change for « ...same direction as adaptive evolution does ...»

We have modified as suggested.

Q4. Line 66,When presenting papers that address the plasticity and evolution of gene expression in response to environmental variables, paper by Morris et al is another example that could be useful to include (but this is only a suggestion in case the authors missed it).

Thank you for suggesting this nice work. We have cited Morris et al. (2014).

Q5. Line 94, Change for "We acclimated"

We have modified as suggested.

Q6. In Figure 3, the figure in panel A and B is labelled "normaxia", but I think that "normoxia" is usually the term used.

Thank you for spot the typo. We have modified Figure 3a and we no longer used the term “normaxia”.

Material and methods

It would be important to merge supplementary table 1 and 2 and only present the individuals that were used with their respective cardiac and muscle libraries (if they come from the same individual?). Also, the origin of the individuals used in the hypoxia experiment should be explained at the beginning of the methods section and explicated in the supplementary table. Information on sex or stage of development (juvenile? Adult? Male? female?) and time of year (in breeding stage? Pre-migration (if any), etc) would allow the reader to see that individuals from lowland differed only in their exposure to hypoxia or not, or if other variables may affect gene expression patterns. Similarly, if all individuals form the highland are males and the lowland hypoxia exposed individuals are females (or juveniles versus breeders, or different time of year, etc) this should be stated in the methods. Gene expression is labile so the reader should know if other variables influence the results presented or not.

Thank you for suggestion. We have added detailed information (i.e., age, collecting time and season) to the supplementary Table 1. We have also added this information to the Methods. Because the birds used in transcriptomic analysis (Supplementary Table 1) were different individuals from those used in the sequence analyses (Supplementary Table 2), these two tables cannot be merged.

References:

Campbell-Staton SC, Velotta JP, Winchell KM. 2021. Selection on adaptive and maladaptive genes expression plasticity during thermal adaptation to urban heat islands. Nat. Commun. 12: 6195.

Ghalambor CK, Hoke KL, Ruell EW, Fischer EK, Reznick DN, Hughes KA. 2015. Non-adaptive plasticity potentiates rapid adaptive evolution of gene expression in nature. Nature 525:372–375.

Hao et al. 2023. Divergent contributions of coding and noncoding sequences to initial high-altitude adaptation in passerine birds endemic to the Qinghai–Tibet Plateau. Mol. Ecol. Doi: 10.1111/mec.16942.

Ho WC, Zhang J. 2018. Evolutionary adaptations to new environments generally reverse plastic phenotypic changes. Nat. Commun. 9: 350.

Ho WC, Zhang J. 2019. Genetic gene expression changes during environmental adaptations tend to reverse plastic changes even after correction for statistical nonindependence. Mol. Biol. Evol. 36: 604–612.

Ho WC, Li D, Zhu Q, Zhang J. 2020. Phenotypic plasticity as a long-term memory easing readaptations to ancestral environments. Sci. Adv. 6: eaba3388.

Kuo KC, Yao CT, Liao BY, Weng MP, Dong F, Hsu YC, Hung CM. 2023. Weak gene-gene interaction facilitates the evolution of gene expression plasticity. BMC Biol. 21: 57.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

I would recommend the authors check the results section, it seems to me that the first two paragraphs are not results, but methods.

We would like to express our appreciation to both reviewers for bringing this to our attention. Indeed, we discussed this in detail, but decided that because the methods come after the results section. We believe that providing the basic methodological approach to readers before the results is essential for better comprehension. Once again, we sincerely thank the reviewers for their valuable feedback, however, we would prefer to leave this part as it is.

In Figure 3B, why there is not male and female shown in different lines, as in the rest of figures? I recommend following the same pattern everywhere.

Has been changed accordingly, and the respective sex-specific lines were also added to Figure 4.

I recommend checking carefully all the articles included in Table 2. Maybe some of the included information here is not precise.

We thank the reviewer for highlighting this. We carefully checked the articles again, and made some small adjustments.

In Material and methods: just note that when ages are estimated, usually there is a variable accounting for the amount of estimated years, that should be included in the model, and see that it has no effect on the dependent variable. I recommend including this variable.

We sincerely appreciate the helpful comment from the reviewer, which we have carefully considered and implemented in our manuscript. However, we would like to highlight that addressing age estimation error is complex, as it involves measurement error. Thus, simply adding it as an independent variable may not fully capture its potential impact, as the effect may be positive or negative depending on the individual. Hence, the potential effect would be better accounted for by the implementation of individual random intercepts and smooths to adjust the confidence intervals, which is part of our model structures. Furthermore, we would like to emphasize that we have also conducted analyses on a reduced dataset that only included zoo-born individuals with precisely known birthdates, and the results remained consistent. So instead of changing our analyses, we now emphasize how our approach also addresses this aspect.

Creatinine: Is there any other reference, more recent and in English, to complement the original one cited?

We have now supplemented the original citation with an additional English citation: Anestis et al. 2009.

Reviewer #2 (Recommendations For The Authors):

Minor corrections

Please, in Study population, the citation of table 2 is in fact Table 3. For table 3 (in Methodology), please provide the units Body weight having a mean of 32.4, has it a median of 9 ?

Please, provide results separately for males and females

We changed the table as requested, though the table only reports sample sizes and thus only numbers without units. The values for body weight are accurate.

In Results

The two first paragraphs have to be included in methods and structured with those already present.

We would like to express our appreciation to both reviewers for bringing this to our attention. Indeed, we discussed this in detail, but decided that because the methods come after the results section, we believe that providing the basic methodological approach to readers before the results is essential for better comprehension. Once again, we sincerely thank the reviewers for their valuable feedback, however, we would prefer to leave this part as it is.

In Table 1, indicate what 'Est' means.

Has been changed accordingly

Author Response

Reviewer #1 (Public Review):

The cerebral cortex, or surface of the brain, is where humans do most of their conscious thinking. In humans, the grooves (sulci) and bumps (convolutions) have a particular pattern in a region of the frontal lobe called Broca's area, which is important for language. Specialists study features imprinted on the internal surfaces of braincases in early hominins by casting their interiors, which produces so-called endocasts. A major question about hominin brain evolution concerns when, where, and in which fossils a humanlike Broca's area first emerged, the answer to which may have implications for the emergence of language. The researchers used advanced imaging technology to study the endocast of a hominin (KNM-ER 3732) that lived about 1.9 million years ago (Ma) in Kenya to test a recently published hypothesis that Broca's remained primitive (apelike) prior to around 1.5 Ma. The results are consistent with the hypothesis and raise new questions about whether endocasts can be used to identify the genus and/or species of fossils.

We would like to thank Rev. 1 for their comments on our paper.

Reviewer #2 (Public Review):

The authors tried to support the hypothesis that early Homo still had a primitive condition of Broca's cap (the region in fossil endocasts corresponding to Broca's area in the brain), being more similar to the condition in chimpanzees than in humans. The evidence from the described individual points to this direction but there are some flaws in the argumentation.

We are grateful to Rev. 2 for their comments, although we partially agree with some of them.

First, we would like to rectify the statement of Rev. 2 that we “tried to support the hypothesis that early Homo still had a primitive condition of Broca's cap”, indeed, our aim was to test this hypothesis and not to try to validate it.

First, only one human and one chimpanzee were used for comparison, although we know that patterns of brain convolutions (and in addition how they leave imprints in the endocranial bones) are very variable.

We understand the point raised by Rev. 2 about the variation of brain convolutions in humans and chimpanzees. We used atlases published by Connolly (1950), Falk et al. (2018) and de Jager et al. (2019, 2022) to analyse the endocast of KNM-ER 3732 and compare it to the extant human and chimpanzee cerebral conditions. However, in Figure 2, for the sake of clarity only two Homo and Pan specimens were used to illustrate the comparison (as it has been done in other published papers, e.g., Carlson et al., 2011; Science, Gunz et al., 2020 Sci Adv). In the revised version, we modified the manuscript to explain further our approach (line 156) “We used brain and endocast atlases published in Connolly (1950), Falk et al. (2018) and de Jager et al. (2019, 2022; see also www.endomap.org) for comparing the pattern identified in KNM-ER 3732 to those described in extant humans and chimpanzees. To the best of our knowledge, these atlases are the most extensive atlases of extant human and chimpanzee brains/endocasts available to date and are widely used in the literature to explore variability in sulcal patterns. In Figure 2, the extant human and chimpanzee conditions are illustrated by one extant human (adult female) and one extant chimpanzee (adult female) specimens from the Pretoria Bone Collection at the University of Pretoria (South Africa) and in the Royal Museum for Central Africa in Tervuren (Belgium), respectively (Beaudet et al., 2018).”.

Second, the evidence from this fossil specimen adds to the evidence of previously describe individuals but still not yet fully prove the hypothesis.

We tempered our discussion by concluding that (line 116) “Overall, the present study not only demonstrates that Ponce de León et al.’s (2021) hypothesis of a primitive brain of early Homo cannot be rejected, but also adds information […]”.

Third, there is a vicious circle in using primitive and derived features to define a fossil species and then using (the same or different) features to argue that one feature is primitive or derived in a given species. In this case, we expect members of early Homo to be derived compared to their predecessors of the genus Australopithecus and that's why it seems intriguing and/or surprising to argue that early Homo has primitive features. However, we should expect that there is some kind of continuum or mosaic in a time in which a genus "evolves into" another genus. This discussion requires far more discussions about the concepts we use, maybe less discussion about what is different between the two groups but more discussion about the evolutionary processes behind them.

We fully agree with Rev. 2 on this aspect. We believe that identifying these differences/similarities between fossil and extant hominids constitute the first step of a better understanding of the evolutionary mechanisms. Our work suggests indeed a certain continuity between genera and raises questions on the genus concept and how to interpret the specimens currently attributed to early Homo. In the revised version of the manuscript we included a reference to this possible scenario (line 134): “[…] or to the absence of a definite threshold between the two genera based on the morphoarchitecture of their endocasts (Wood and Collard, 1999).”.

Fourth, the data of convolutional imprints presented are rather subjective when identifying which impressions represent which brain convolutions. Not seeing an impression does not necessarily mean that the corresponding brain feature did not exist. Interestingly, the manuscript does not mention and discuss at all the frontoorbital sulcus. This is a sulcus that usually runs from the orbital surface of the frontal lobe up to divide the inferior frontal gyrus in chimpanzees, a condition totally different than in humans who do not have a frontoorbital sulcus. Could such a sulcus be identified, this would provide a far more convincing argument for a primitive condition in this specimen. In Australopithecus sediba, e.g., the condition in this region seems to be a mosaic in which some aspects of the morphology seem to be more modern while one of the sulcual impressions can well be interpreted as a short frontoorbital sulcus. For this specimen, by the way, I would come back to my third point above: some experts in the field might argue that this specimen could belong to Homo rather than Australopithecus...

We agree that the presence of a fronto-orbital sulcus would be more conclusive. However, this sulcus has not been identified in KNM-ER3732 and the region in which we would expect to find it is not preserved. As demonstrated by Ponce de León et al. (2021), because of the topographic relationships between sulci (and cranial structures), it is possible to interpret imprints on endocasts and the evolutionary polarity of some traits even in the absence of landmarks such as the fronto-orbital sulcus. In Australopithecus sediba the main derived feature of the endocast corresponds to the ventrolateral bulge in the left inferior frontal gyrus, and not to the sulcal pattern itself (Carlson et al., 2011 Science). However, the discussion around the taxonomic status of this taxon confirms the urgent need for reconsidering specimens from that time period and clarifying the mosaic-like or concerted evolution of the derived Homo-like traits within our lineage. Regarding the subjective nature of this approach, we invite readers to examine the specimen on MorphoSource (https://www.morphosource.org/concern/media/000497752?locale=en) and to request access to the National Museums of Kenya to the physical or virtual specimen to falsify our hypothesis.

According to my arguments above, I think that this manuscript might revive interesting discussions about this topic but it is not likely to settle them because the data presented are not strong enough to fully support the hypothesis.

We would be more than happy to consider new/other specimens with similar chronological and geographical contexts and investigate further this hypothesis in the future.

Reviewer #3 (Public Review):

The authors provide a detailed analysis of the sulcal and sutural imprints preserved on the natural endocast and associated cranial vault fragments of the KNM-ER3732 early Homo specimen. The analyses indicate a primitive ape-like organization of this specimen's frontal cortex. Given the geological age of around 1.9 million years, this is the earliest well-documented evidence of a primitive brain organization in African Homo.

In the discussion, the authors re-assess one of the central questions regarding the evolution of early Homo: was there species diversity, and if yes, how can we ascertain it? The specimen KNM-ER1470 has assumed a central role in this debate because it purportedly shows a more advanced organization of the frontal cortex compared to other largely coeval specimens (Falk, 1983). However, as outlined in Ponce de León et al. 2021 (Supplementary Materials), the imprints on the ER1470 endocranium are unlikely to represent sulcal structures and are more likely to reflect taphonomic fracturing and distortion. Dean Falk, the author of the 1983 study, basically shares this view (personal communication). Overall, I agree with the authors that the hypothesis to be tested is the following: did early Homo populations with primitive versus derived frontal lobe organizations coexist in Africa, and did they represent distinct species?

I greatly appreciate that the authors make available the 3D surface data of this interesting endocast.

We are grateful to Rev. 3 for their comments and for contextualizing our finding. We would also like to point out that, although the 3D surface can be viewed on MorphoSource, permission from the National Museums of Kenya has to be requested for studying the specimen and getting access to the physical specimen and/or the 3D model.

Author Response

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

We thank the reviewers for their positive and constructive evaluations. Based upon the reviewers’ helpful comments, we have performed complementary experiments. In particular, we additionally show that:

• a complete analysis of CXCR1/2 binding chemokines in the secretions of tissular CD8+ T cells reinforces the key role of CXCL8 in CD8+ T cell-induced fibrocyte chemotaxis (new panel D in Figure 2)

• a direct contact between fibrocytes and CD8+ T cells triggers CD8+ T cell cytotoxicity against primary basal bronchial epithelial cells (new Figure 6)

• the interaction between CD8+ T cells and fibrocytes is bidirectional, with CD8+ T cells triggering the development of fibrocyte immune properties (new Figure 7)

• the characteristic time to reach a stationary state reminiscent of a resolution of the COPD condition was estimated to be about 2.5 years using the simulations. Interfering with chemotaxis and adhesion processes by inhibiting CXCR1/2 and CD54, respectively was not sufficient to reverse the COPD condition, as predicted by the mathematical model (new Figure 9)

• the massive proliferation effect induced by fibrocytes is specific to CD8+ T cells and not CD4+ T cells (new Figure 3-figure supplement 2), and that fibrocytes moderately promote the death of unactivated CD8+ T cells in direct co-culture (new Figure 3-figure supplement 3)

We have graphically summarized our findings (new Figure 10) suggesting the existence of a positive feedback loop playing a role in the vicious cycle that promotes COPD. A new table describing patient characteristics for basal bronchial epithelial cell purification has also been added (new Supplementary File 9), the Supplementary Files 7 and S8 have been up-dated to take into account the new experiments.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifier PXD041402.  

Reviewer #1 (Recommendations For The Authors):

The experimental approaches are all rationally designed and the data clearly presented, with appropriate analyses and sample sizes. I could find no technical or interpretative concerns. The interrelationship between the observational data (histology) with the quantitative live cell imaging and the follow-on functional investigations is especially laudable. The data nicely unifies several years of accumulated data regarding the (separate) participation of CD8 T cells and fibrocytes in COPD.

We thank the reviewer for his/her comments.

I have only minor comments:

1) Line 79: The observation that T cells may influence fibrocyte differentiation/function was initially made some years earlier by Abe et al (J Immunol 2001; 7556), and should be cited in addition to the follow-on work of Niedermeyer.

This reference has been added to acknowledge this seminal work.

2) Line 632: Corticosteroids originate from the cortex of the adrenal gland. Budenoside and fluticasone are glucocorticoids, not corticosteroids.

This mistake has been corrected in the discussion of the revised manuscript (see line 802 in the revised manuscript).

3) Given the state of T cell immunotherapies, cytokine/chemokine antagonists, and emerging fibrocyte-targeted drugs, can the authors possibly speculate as to desired pathways to target therapeutically?

Chemokine-receptor based therapies could be used to inhibit fibrocyte recruitment into the lungs, such as CXCR4 blockade. We have very recently shown that using the CXCR4 antagonist, plerixafor, alleviates bronchial obstruction and reduces peri-bronchial fibrocytes density (Dupin et al., 2023). Because CXCR4 expression in human fibrocytes is dependent on mTOR signaling and is inhibited by rapamycin in vitro (Mehrad et al., 2009), alternative strategies consisting of targeting fibrocytes via mTOR have been proposed. This target has proven effective in bronchiolitis obliterans, idiopathic pulmonary fibrosis, and thyroid-associated ophthalmopathy, using rapamycin (Gillen et al., 2013; Mehrad et al., 2009), sirolimus (Manjarres et al., 2023) or an insulin-like growth factor-1 (IGF-I) receptor blocking antibody (Douglas et al., 2020; Smith et al., 2017). Inhibiting mTOR is also expected to have effects on CD8+ T cells, ranging from an immunostimulatory effect by activation of memory CD8+ T-cell formation, to an immunosuppressive effect by inhibition of T cell proliferation (Araki et al., 2010). Last, chemokine-receptor base therapies could also include strategies to inhibit the CD8+-induced fibrocyte chemotaxis, such as dual CXCR1-CXCR2 blockade. We were able to test this latter strategy in our mathematical model, see response to point 6 of reviewer 2.

Immunotherapies directly targeting the interaction between fibrocytes and CD8+ T cells could also be considered, such as CD86 or CD54 blockade. The use of abatacept and belatacept, that interfere with T cell co-stimulation, is effective in patients with rheumatoid arthritis (Pombo-Suarez & Gomez-Reino, 2019) and in kidney-transplant recipients (Vincenti et al., 2016), respectively. Targeting the IGF-I receptor by teprotumumab in the context of thyroid-associated ophthalmopathy also improved disease outcomes, possibly by altering fibrocyte-T cell interactions (Bucala, 2022; Fernando et al., 2021).

We also tested this CD86 and CD54 blocking strategy for COPD treatment by simulations, see response to point 6 of reviewer 2.

However, such therapies should be used with caution as they may favour adverse events such as infections, particularly in the COPD population (Rozelle & Genovese, 2007). Additionally, the fibrocytes-lymphocytes interaction has recently been shown to promote anti-tumoral immunity via the PD1-PDL1 immunological synapse (Afroj et al., 2021; Mitsuhashi et al., 2023). Therefore, care should be taken in the selection of patients to be treated and/or timing of treatment administration with regards to the increased risk of lung cancer in COPD patients.

The discussion section has been altered accordingly.

4) The authors may want to consider mentioning (and citing) recent insight into the immune-mediated fibrosis in thyroid-associated ophthalmopathy

These important publications are now cited in a dedicated paragraph about the possible therapeutical interventions (see answer to point 3, and discussion in the revised manuscript).

Reviewer #2 (Recommendations For The Authors):

Specific comments

1) The rationale for the selection of chemokines overexpressed by CD8+ T cells in COPD is based on literature data of n=2 patients per group. This is limited and risky. I am less concerned about false positives given the selection of chemokines and the available literature but am worried about the possibility that many chemokines may not have been selected based on insufficient power to do meaningful stats on this comparison. For example, many other CXCR1/2 binding CXCL chemokines exist and these could contribute to the migration effect in Fig 2C as well. Given the currently available single-cell resources it should be possible to extend these observations and to investigate CXCL chemokine expression in COPD CD8 T cells to the benefit of Fig 2A in full detail.

We agree with the reviewer that the rationale for the selection of chemokines of interest could be reinforced by the analysis of supplementary single-cell resources. We used data from the COPD cell atlas (Gene Expression Omnibus GSE136831 (Sauler et al., 2022)) to perform such an analysis of chemokine expression by CD8+ CD103+ and CD8+ CD103- T cells. However, the expression level of all chemokines was globally very low, and was not different between control and COPD patients (see Author response image 1).

Author response image 1.

Expression of CXC chemokines in lung CD8+ CD103+ and CD8+ CD103- T cells from patients with COPD (n=18 independent samples) in comparison with healthy control subjects (n=29 independent samples) under resting conditions by Single-Cell RNA sequencing analysis (GEO accession GSE136831). The heatmaps show the normalized expression of genes (horizontal axes) encoding CXC chemokines. PF4=CXCL4, PPBP= CXCL7.

The latter results are in discrepancy with those resulting from transcriptomic analysis of microarray data obtained on purified lung CD8+ CD103+ and CD8+ CD103- T cells, showing a significant level of chemokines expression (Hombrink et al., 2016), and a differential expression of CCL2, CCL26, CXCL2, CXCL8 and CCL3L1 between CD8+ T lymphocytes of control and COPD patients (Figure 2A in the revised manuscript). The reason for these differences is unclear, and could be attributed to biological differences (samples obtained from different patients) or, more likely, to differences in sample processing (cell sorting by flow cytometry for microarray analysis, that could activate minimally CD8+ cells) and/or methodological differences (differences of sensitivity between microarray and scRNA seq).

Nevertheless, microarray data regarding CXCL8 expression are in good agreement with our in vitro experiments, showing an enhanced CXCL8 expression by CD8+ T cells purified from COPD lungs, in comparison with that of control subjects. In addition, the CXCL8 blocking antibody fully abrogates the increase of migration induced by secretion of COPD CD8+ T cells, to the same extent as the blocking of CXCR1/2 by reparixin. This suggests that this supplementary chemotaxis is mainly due to CXCL8 and not other CXCR1/2 binding CXCL chemokines, and correlates CXCL8 measurements to functional experiments. This precision has been now added in the results section of the revised version.

2) Equally, it would strengthen the work if multiplex ELISA assays could be provided on the supernatants used in Fig 2D to provide a more comprehensive view of CXCR1/2 binding chemokines.

In order to have a complete view of CXCR1/2 binding chemokines, we have now performed supplementary ELISA assays to measure the concentrations of CXCL1, 3, 5, 6 and 7, in addition of the measurements of CXCL2 and CXCL8 already presented in the previous version of the manuscript (Figure 2D). Results of these new assays are now presented in the revised version of Figure 2. Concentrations of CXCL1, 3, 5, 6 and 7 were unchanged between the control and COPD conditions.

3) In the functional analyses, I missed information on the activation of the fibrocytes. Equally, the focus on CD8 T cells was mainly on proliferation in the functional work. RNAseq analyses on the cells, comparing CD8 T cells and fibrocytes, alone and in co-culture to each other would help to identify interaction patterns in comprehensive detail. Such an experiment would bolster the significance of the studies by providing impact analysis not only on the T cells beyond proliferation but by expanding on the effect of the interaction on the fibrocyte as well.

Regarding the activation state of fibrocytes, we apologize if this was not clear: in our in vitro co-culture experiments, we chose not to activate the fibrocytes. This setting is in agreement with previous findings, demonstrating an antigen-independent T cell proliferation effect driven by fibrocytes (Nemzek et al., 2013), and it is now explicitly written in the results of the revised manuscript.

Regarding the focus of the functional analyses:

First, we have pushed forward the analysis of the consequences of the interaction beyond CD8+ T cells proliferation. In particular, having shown that fibrocytes promote CD8+ T cells expression of cytotoxic molecules such as granzyme B, we decided to investigate the cytotoxic capacity of CD8+ T cells against primary basal bronchial epithelial cells (see new Supplementary File 9 in the revised manuscript for patient characteristics).

Direct co-culture with fibrocytes increased total and membrane expression of the cytotoxic degranulation marker CD107a, which was only significant in non-activated CD8+ T cells (see new Figure 6A-E in the revised manuscript). A parallel increase of cytotoxicity against primary epithelial cells was observed in the same condition (see new Figure 6F-H in the revised manuscript). This demonstrates that following direct interaction with fibrocytes, CD8+ T cells have the ability to kill target cells such as bronchial epithelial cells. This is now included in the results section of the revised manuscript.

Second, we have now performed proteomic analyses on fibrocytes, alone or in co-culture during 6 days with CD8+ T cells either non-activated or activated (see new Figure 7A in the revised manuscript). Of the top ten pathways that were most significantly activated in co-cultured vs mono-cultured fibrocytes, largest upregulated genes were those of the dendritic cell maturation box, the multiple sclerosis signaling pathway, the neuroinflammation signaling pathway and the macrophage classical signaling pathway, irrespective of the activation state of CD8+ T cells (see new Figure 7B in the revised manuscript). The changes were globally identical in the two conditions of CD8+ T cell activation, with some upregulation more pronounced in the activated condition. They were mostly driven by up-regulation of a core set of Major Histocompatibility Complex class I (HLA-B, C, F) and II (HLA-DMB, DPA1, DPB1, DRA, DRB1, DRB3) molecules, co-simulatory and adhesion molecules (CD40, CD86 and CD54). Another notable proteomic signature was that of increased expression of IFN signaling-mediators IKBE and STAT1, and the IFN-responsive genes GBP2, GBP4 and RNF213. We also observed a strong downregulation of CD14, suggesting fibrocyte differentiation, and an upregulation of the matrix metalloproteinase-9 (MMP9) in the non-activated condition only. Altogether, these changes suggest that the interaction between CD8+ T cells and fibrocytes promotes the development of fibrocyte immune properties, which could subsequently impact the activation of CD4+ T cells activation.

Up-regulated pathways identified in proteomic profile of fibrocytes co-cultured with CD8+ T cells are very consistent with a shift towards a proinflammatory phenotype rather than towards a reparative role. The activation of IFN-γ signaling could be triggered by CD8+ T cell secretion of IFN upon fibrocyte interaction, suggesting the existence of a positive feedback loop (see new Figure 10). Additionally, the priming of fibrocytes by CD8+ T cells could also induce CD4+ T cell activation.

4) I suggest rewording the abstract to capture the main storyline and wording more. The abstract is good, but I see so many novelties in the paper that are not well sold in the abstract, particularly the modelling aspects.

As suggested by the reviewer, we revised the abstract, as shown below and in the revised manuscript. The changes are indicated in red:

Revised abstract:

Bronchi of chronic obstructive pulmonary disease (COPD) are the site of extensive cell infiltration, allowing persistent contacts between resident cells and immune cells. Tissue fibrocytes interaction with CD8+ T cells and its consequences were investigated using a combination of in situ, in vitro experiments and mathematical modeling. We show that fibrocytes and CD8+ T cells are found in vicinity in distal airways and that potential interactions are more frequent in tissues from COPD patients compared to those of control subjects. Increased proximity and clusterization between CD8+ T cells and fibrocytes are associated with altered lung function. Tissular CD8+ T cells from COPD patients promote fibrocyte chemotaxis via the CXCL8-CXCR1/2 axis. Live imaging shows that CD8+ T cells establish short-term interactions with fibrocytes, that trigger CD8+ T cell proliferation in a CD54- and CD86-dependent manner, pro-inflammatory cytokines production, CD8+ T cell cytotoxic activity against bronchial epithelial cells and fibrocyte immunomodulatory properties. We defined a computational model describing these intercellular interactions and calibrated the parameters based on our experimental measurements. We show the model’s ability to reproduce histological ex vivo characteristics, and observe an important contribution of fibrocyte-mediated CD8+ T cell proliferation in COPD development. Using the model to test therapeutic scenarios, we predict a recovery time of several years, and the failure of targeting chemotaxis or interacting processes. Altogether, our study reveals that local interactions between fibrocytes and CD8+ T cells could jeopardize the balance between protective immunity and chronic inflammation in bronchi of COPD patients.

5) The probabilistic model appears to suggest that reduced CD8 T cell death may also explain the increase in the pathology in COPD. Did the authors find that fibrocytes reduce cell death of the CD8 T cells?

Taking advantage of the staining of CD8+ T cells with the death marker Zombie NIR™, we have quantified CD8+ T cell death in our co-culture assay. The presence of fibrocytes in the indirect co-culture assay did not affect CD8+ T cell death (see new Figure 3-figure supplement 3A-B in the revised manuscript). In direct co-culture, the death of CD8+ T cells was significantly increased in the non-activated condition but not in the activated condition (see new Figure 3-figure supplement 3C-D in the revised manuscript). Of note, these results are in agreement with a recent study showing the existence of CD8+ T cell-population-intrinsic mechanisms regulating cellular behavior, with induction of apoptosis to avoid an excessive increase in T cell population (Zenke et al., 2020). This is taken into account in our mathematical model by an increased probability p_(dC+) of dying when a CD8+ T cell is surrounded by many other T cells in its neighborhood. It also suggests that the reduced CD8+ T cell death evidenced in tissues from patients with COPD (Siena et al., 2011) might not be due to the specific interplay between fibrocyte and CD8+ T cells, but rather to a global pro-survival environment in COPD lungs.

These new data have been described in the results section.

6) Following the modeling in Figure 6, curiosity came to mind, which is how long it would take for the pathology to disappear if a drug would be applied to the patient. How much should the interactions be reduced and how long would it take to reach clinical benefit? Could such predictions be made? I understand that this may be outside the main message of the manuscript but perhaps this could be included in the discussion.

This is a very interesting question, that we have addressed by performing additional simulations to investigate the outcomes of possible therapeutic interventions. First, we applied a COPD dynamics during 20 years, to generate the COPD state, that provide the basis for treatment implementation. Then, we applied a COPD dynamic during 7 years, that mimics the placebo condition (see new Figure 9A in the revised manuscript, and below), that we compared to a control dynamics (“Total inhibition”), that mimics an ideal treatment able to restore all cellular processes. As expected the populations of fibrocytes and CD8+ T cells, as well as the density of mixed clusters, decreased. These numbers reached levels similar of healthy subjects after approximately 2.5 years, and this time point can therefore be considered as the steady state (Figure 9B-E).

Monitoring of the different processes revealed that these effects were mainly due to a reduction in fibrocyte-induced CD8+ T duplication, and a transient or more prolonged increase in basal fibrocyte and CD8+ T death (Figure 9C-D).

Then, three possible realistic treatments were considered (Figure 9A). We tested the effect of directly inhibiting the interaction between fibrocytes and CD8+ T cells by blocking CD54. This was implemented in the model by altering the increased probability of a CD8+ T cell to divide when a fibrocyte is in its neighbourhood, as shown by the co-culture results (Figure 4). We also chose to reflect the effect of a dual CXCR1/2 inhibition by setting the displacement function of fibrocyte similar to that of control dynamics, in agreement with the in vitro experiments (Figure 2E). Blocking CD54 only slightly reduced the density of CD8+ T cells compared to the placebo condition, and had no effect on fibrocyte and mixed cluster densities (Figure 9B). CXCR1/2 inhibition was a little bit more potent on the reduction of CD8+ T cells than CD54 inhibition, and it also significantly decreased the density of mixed clusters (Figure 9B). As expected, this occurred through a reduction of fibrocyte-induced duplication, which was affected more strongly by CXCR1/2 blockage than by CD54 blockage (Figure 9C-E). Combining both therapies (CD54 and CXCR1/2 inhibition) did not strongly major the effects (Figure 9B-E). In all the conditions tested, the size of the fibrocyte population remained unchanged, suggesting that other processes such as fibrocyte death or infiltration should be targeted to expect broader effects.

The results section has been altered accordingly.

Using the simulations, we were also able to estimate the characteristic time to reach a stationary state reminiscent of a resolution of the COPD condition. This time of approximately 2.5 years was totally unpredictable by in vitro experiments, and indicates that a treatment aiming at restoring these cellular processes should be continued during several years to obtain significant changes.

We have also investigated the outcomes of more realistic treatments, modifying specifically processes such as chemotaxis or targeting directly the intercellular interactions. The modification of parameters controlling these processes only slightly affected the final state, suggesting that such treatments may be more effective when used in combination with other drugs e.g. those affecting fibrocyte infiltration and/or death.

The discussion section has been altered accordingly.

Reviewer #3 (Recommendations For The Authors):

1) Broader assessment of cell types in the lung: Staining for other cell types such as dendritic cells, CD4 cells, and interstitial macrophages, and comparing their proximity to fibrocytes with that of CD8 cells would better justify the CD8 focus.

We agree with the reviewer that multiple stainings would have better justified the focus on CD8+ T cells. However, it is difficult to distinguish fibrocytes, dendritic cells and interstitial macrophages on the basis of immunohistochemistry, as we and others previously showed (Dupin et al., 2019; Mitsuhashi et al., 2015; Pilling et al., 2009). On the other hand, the study of Afroj et al. indicated the possible interaction between fibrocytes and CD8+ T cells in cancer context, with the induction of CD8+ T cell proliferation (Afroj et al., 2021). This T cell-costimulatory function of fibrocytes and CD8+ T cells was further confirmed in a very recent study, together with the antitumor effects of PD-L1 and VEGF blockade (Mitsuhashi et al., 2023). These data, along with the specific implication on CD8+ T cells in COPD, relying mainly on their abundance in COPD bronchi (O’Shaughnessy et al., 1997), their overactivation state (Roos-Engstrand et al., 2009), their cytotoxic phenotype (Freeman et al., 2010; Wang et al., 2020) and the protection against lung inflammation and emphysema induced by their depletion (Maeno et al., 2007) justified the CD8 focus.

To further justify this focus, we have now performed co-culture between fibrocytes and CD4+ T cells, indicating that the massive fibrocyte-mediated proliferation was specific to CD8+ T cells (see answer to comment 3 below). This is in agreement with the results obtained with the simulations, showing that considering fibrocytes and CD8+ T cells only was sufficient to reproduce the spatial patterns in the bronchi of healthy and COPD patients. Altogether, we think that focusing on the CD8+ T cell-fibrocyte interplay was pertinent in the context of COPD. It does obviously not exclude the possibility of other interactions, that could be the focus of other studies.

2) Transcriptomic analysis: Using n=2 and only showing the chemokines as well as selected adhesion receptor data narrows the focus but does not provide broader insights into the interactions. Using a more robust sample size and performing a comprehensive pathway analysis would represent an unbiased analysis to determine the most dysregulated pathways. Importantly, the authors could use a single-cell RNA-seq dataset to broadly assess the transcriptomes of several cell types in the lung (such as the data from (Sauler et al, Characterization of the COPD alveolar niche using single-cell RNA sequencing).

This very pertinent suggestion has also been raised by reviewer 2, see our answer to comment 1 of reviewer 2, and below:

We agree with the reviewer that the rationale for the selection of chemokines of interest could be reinforced by the analysis of supplementary single-cell resources. We used data from the COPD cell atlas (Gene Expression Omnibus GSE136831 (Sauler et al., 2022)) to perform such an analysis of chemokine expression by CD8+ CD103+ and CD8+ CD103- T cells. However, the expression level of all chemokines was globally very low, and was not different between control and COPD patients (see Figure scRNAseq, in the answer to comment 1 of reviewer 2).

These latter results are in discrepancy with those resulting from transcriptomic analysis of microarray data obtained on purified lung CD8+ CD103+ and CD8+ CD103- T cells, showing a significant level of chemokines expression (Hombrink et al., 2016), and a differential expression of CCL2, CCL26, CXCL2, CXCL8 and CCL3L1 between CD8+ T lymphocytes of control and COPD patients (Figure 2A in the revised manuscript). The reason for these differences is unclear, and could be attributed to biological differences (samples obtained from different patients) or, more likely, to differences in sample processing (cell sorting by flow cytometry for microarray analysis, that could activate minimally CD8+ cells) and/or methodological differences (differences of sensitivity between microarray and scRNA seq).

Nevertheless, microarray data regarding CXCL8 expression are in good agreement with our in vitro experiments, showing an enhanced CXCL8 expression by CD8+ T cells purified from COPD lungs, in comparison with that of control subjects. In addition, the CXCL8 blocking antibody fully abrogates the increase of migration induced by secretion of COPD CD8+ T cells, to the same extent as the blocking of CXCR1/2 by reparixin. This suggests that this supplementary chemotaxis is mainly due to CXCL8 and not other CXCR1/2 binding CXCL chemokines, and correlates CXCL8 measurements to functional experiments. This precision has been now added in the text of the revised version.

3) Inclusion of control/comparison cell types in co-culture studies would help establish that CD8 cells are more relevant for interactions with fibrocytes than for example CD4 cells.

We have now performed co-cultures between fibrocytes and CD4+ T cells, with the same settings than for CD8+ T cells. The results from these experiments show that fibrocytes did not have any significant effect of CD4+ T cells death, regardless of their activation state (see new Figure 3-figure supplement 2A-C in the revised manuscript, and below). Fibrocytes were able to promote CD4+ T cells proliferation in the activated condition but not in the non-activated condition (see new Figure 3-figure supplement 2A-D in the revised manuscript). Altogether this indicates that although fibrocyte-mediated effect on proliferation is not specific to CD8+ T cells, the amplitude of the effect is much larger on CD8+ T cells than on CD4+ T cells.

These new data have been added in the results section.

4) In vitro analysis of cells from non-COPD patients would also help assess whether the circulating cells from COPD patients have a level of baseline activation which promotes the vicious cycle but may not exist in healthy cells.

Regarding circulating cells, the present study relies on the COBRA cohort (COhort of BRonchial obstruction and Asthma), which includes only asthma and COPD patients, and therefore does not grant access to healthy subjects’ blood samples (Pretolani et al., 2017). Unfortunately, we have no other ongoing study with healthy subjects that would allow us to retrieve blood for research, and fibrocytes can only be grown from freshly drawn blood samples. We agree with the reviewer that it is a limitation of our study, which is now acknowledged at the end of the discussion section.  

References

Afroj, T., Mitsuhashi, A., Ogino, H., Saijo, A., Otsuka, K., Yoneda, H., Tobiume, M., Nguyen, N. T., Goto, H., Koyama, K., Sugimoto, M., Kondoh, O., Nokihara, H., & Nishioka, Y. (2021). Blockade of PD-1/PD-L1 Pathway Enhances the Antigen-Presenting Capacity of Fibrocytes. The Journal of Immunology, 206(6), 1204‑1214. https://doi.org/10.4049/jimmunol.2000909

Araki, K., Youngblood, B., & Ahmed, R. (2010). The role of mTOR in memory CD8+ T-cell differentiation. Immunological reviews, 235(1), 234‑243. https://doi.org/10.1111/j.0105-2896.2010.00898.x

Bucala, R. J. (2022). Targeting fibrocytes in autoimmunity. Proceedings of the National Academy of Sciences, 119(5), e2121739119. https://doi.org/10.1073/pnas.2121739119

Douglas, R. S., Kahaly, G. J., Patel, A., Sile, S., Thompson, E. H. Z., Perdok, R., Fleming, J. C., Fowler, B. T., Marcocci, C., Marinò, M., Antonelli, A., Dailey, R., Harris, G. J., Eckstein, A., Schiffman, J., Tang, R., Nelson, C., Salvi, M., Wester, S., … Smith, T. J. (2020). Teprotumumab for the Treatment of Active Thyroid Eye Disease. The New England Journal of Medicine, 382(4), 341‑352. https://doi.org/10.1056/NEJMoa1910434

Dupin, I., Henrot, P., Maurat, E., Abohalaka, R., Chaigne, S., Hamrani, D. E., Eyraud, E., Prevel, R., Esteves, P., Campagnac, M., Dubreuil, M., Cardouat, G., Bouchet, C., Ousova, O., Dupuy, J.-W., Trian, T., Thumerel, M., Begueret, H., Girodet, P.-O., … Berger, P. (2023). CXCR4 blockade alleviates pulmonary and cardiac outcomes in early COPD (p. 2023.03.10.529743). bioRxiv. https://doi.org/10.1101/2023.03.10.529743

Dupin, I., Thumerel, M., Maurat, E., Coste, F., Eyraud, E., Begueret, H., Trian, T., Montaudon, M., Marthan, R., Girodet, P.-O., & Berger, P. (2019). Fibrocyte accumulation in the airway walls of COPD patients. The European Respiratory Journal, 54(3), Article 3. https://doi.org/10.1183/13993003.02173-2018

Fernando, R., Caldera, O., & Smith, T. J. (2021). Therapeutic IGF-I receptor inhibition alters fibrocyte immune phenotype in thyroid-associated ophthalmopathy. Proceedings of the National Academy of Sciences, 118(52), e2114244118. https://doi.org/10.1073/pnas.2114244118

Freeman, C. M., Han, M. K., Martinez, F. J., Murray, S., Liu, L. X., Chensue, S. W., Polak, T. J., Sonstein, J., Todt, J. C., Ames, T. M., Arenberg, D. A., Meldrum, C. A., Getty, C., McCloskey, L., & Curtis, J. L. (2010). Cytotoxic potential of lung CD8+ T cells increases with COPD severity and with in vitro stimulation by IL-18 or IL-15. Journal of immunology (Baltimore, Md. : 1950), 184(11), 6504‑6513. https://doi.org/10.4049/jimmunol.1000006

Gillen, J. R., Zhao, Y., Harris, D. A., LaPar, D. J., Stone, M. L., Fernandez, L. G., Kron, I. L., & Lau, C. L. (2013). Rapamycin Blocks Fibrocyte Migration and Attenuates Bronchiolitis Obliterans in a Murine Model. The Annals of thoracic surgery, 95(5), 1768‑1775. https://doi.org/10.1016/j.athoracsur.2013.02.021

Hombrink, P., Helbig, C., Backer, R. A., Piet, B., Oja, A. E., Stark, R., Brasser, G., Jongejan, A., Jonkers, R. E., Nota, B., Basak, O., Clevers, H. C., Moerland, P. D., Amsen, D., & van Lier, R. A. W. (2016). Programs for the persistence, vigilance and control of human CD8+ lung-resident memory T cells. Nature Immunology, 17(12), Article 12. https://doi.org/10.1038/ni.3589

Maeno, T., Houghton, A. M., Quintero, P. A., Grumelli, S., Owen, C. A., & Shapiro, S. D. (2007). CD8+ T Cells are required for inflammation and destruction in cigarette smoke-induced emphysema in mice. Journal of Immunology (Baltimore, Md.: 1950), 178(12), 8090‑8096. https://doi.org/10.4049/jimmunol.178.12.8090

Manjarres, D. C. G., Axell-House, D. B., Patel, D. C., Odackal, J., Yu, V., Burdick, M. D., & Mehrad, B. (2023). Sirolimus suppresses circulating fibrocytes in idiopathic pulmonary fibrosis in a randomized controlled crossover trial. JCI Insight. https://doi.org/10.1172/jci.insight.166901

Mehrad, B., Burdick, M. D., & Strieter, R. M. (2009). Fibrocyte CXCR4 regulation as a therapeutic target in pulmonary fibrosis. The International Journal of Biochemistry & Cell Biology, 41(8‑9), 1708‑1718. https://doi.org/10.1016/j.biocel.2009.02.020

Mitsuhashi, A., Goto, H., Saijo, A., Trung, V. T., Aono, Y., Ogino, H., Kuramoto, T., Tabata, S., Uehara, H., Izumi, K., Yoshida, M., Kobayashi, H., Takahashi, H., Gotoh, M., Kakiuchi, S., Hanibuchi, M., Yano, S., Yokomise, H., Sakiyama, S., & Nishioka, Y. (2015). Fibrocyte-like cells mediate acquired resistance to anti-angiogenic therapy with bevacizumab. Nature Communications, 6(1), Article 1. https://doi.org/10.1038/ncomms9792

Mitsuhashi, A., Koyama, K., Ogino, H., Afroj, T., Nguyen, N. T., Yoneda, H., Otsuka, K., Sugimoto, M., Kondoh, O., Nokihara, H., Hanibuchi, M., Takizawa, H., Shinohara, T., & Nishioka, Y. (2023). Identification of fibrocyte cluster in tumors reveals the role in antitumor immunity by PD-L1 blockade. Cell Reports, 112162. https://doi.org/10.1016/j.celrep.2023.112162

Nemzek, J. A., Fry, C., & Moore, B. B. (2013). Adoptive transfer of fibrocytes enhances splenic T-cell numbers and survival in septic peritonitis. Shock (Augusta, Ga.), 40(2), 106‑114. https://doi.org/10.1097/SHK.0b013e31829c3c68

O’Shaughnessy, T. C., Ansari, T. W., Barnes, N. C., & Jeffery, P. K. (1997). Inflammation in bronchial biopsies of subjects with chronic bronchitis : Inverse relationship of CD8+ T lymphocytes with FEV1. American Journal of Respiratory and Critical Care Medicine, 155(3), 852‑857. https://doi.org/10.1164/ajrccm.155.3.9117016

Pilling, D., Fan, T., Huang, D., Kaul, B., & Gomer, R. H. (2009). Identification of markers that distinguish monocyte-derived fibrocytes from monocytes, macrophages, and fibroblasts. PloS One, 4(10), e7475. https://doi.org/10.1371/journal.pone.0007475

Pombo-Suarez, M., & Gomez-Reino, J. J. (2019). Abatacept for the treatment of rheumatoid arthritis. Expert Review of Clinical Immunology, 15(4), 319‑326. https://doi.org/10.1080/1744666X.2019.1579642

Pretolani, M., Soussan, D., Poirier, I., Thabut, G., Aubier, M., COBRA Study Group, & COBRA cohort Study Group. (2017). Clinical and biological characteristics of the French COBRA cohort of adult subjects with asthma. The European Respiratory Journal, 50(2), 1700019. https://doi.org/10.1183/13993003.00019-2017

Roos-Engstrand, E., Ekstrand-Hammarström, B., Pourazar, J., Behndig, A. F., Bucht, A., & Blomberg, A. (2009). Influence of smoking cessation on airway T lymphocyte subsets in COPD. COPD, 6(2), 112‑120. https://doi.org/10.1080/15412550902755358

Rozelle, A. L., & Genovese, M. C. (2007). Efficacy results from pivotal clinical trials with abatacept. Clinical and Experimental Rheumatology, 25(5 Suppl 46), S30-34.

Sauler, M., McDonough, J. E., Adams, T. S., Kothapalli, N., Barnthaler, T., Werder, R. B., Schupp, J. C., Nouws, J., Robertson, M. J., Coarfa, C., Yang, T., Chioccioli, M., Omote, N., Cosme, C., Poli, S., Ayaub, E. A., Chu, S. G., Jensen, K. H., Gomez, J. L., … Rosas, I. O. (2022). Characterization of the COPD alveolar niche using single-cell RNA sequencing. Nature Communications, 13(1), Article 1. https://doi.org/10.1038/s41467-022-28062-9

Siena, L., Gjomarkaj, M., Elliot, J., Pace, E., Bruno, A., Baraldo, S., Saetta, M., Bonsignore, M. R., & James, A. (2011). Reduced apoptosis of CD8+ T-lymphocytes in the airways of smokers with mild/moderate COPD. Respiratory Medicine, 105(10), 1491‑1500. https://doi.org/10.1016/j.rmed.2011.04.014

Smith, T. J., Kahaly, G. J., Ezra, D. G., Fleming, J. C., Dailey, R. A., Tang, R. A., Harris, G. J., Antonelli, A., Salvi, M., Goldberg, R. A., Gigantelli, J. W., Couch, S. M., Shriver, E. M., Hayek, B. R., Hink, E. M., Woodward, R. M., Gabriel, K., Magni, G., & Douglas, R. S. (2017). Teprotumumab for Thyroid-Associated Ophthalmopathy. The New England Journal of Medicine, 376(18), 1748‑1761. https://doi.org/10.1056/NEJMoa1614949

Vincenti, F., Rostaing, L., Grinyo, J., Rice, K., Steinberg, S., Gaite, L., Moal, M.-C., Mondragon-Ramirez, G. A., Kothari, J., Polinsky, M. S., Meier-Kriesche, H.-U., Munier, S., & Larsen, C. P. (2016). Belatacept and Long-Term Outcomes in Kidney Transplantation. The New England Journal of Medicine, 374(4), 333‑343. https://doi.org/10.1056/NEJMoa1506027

Wang, X., Zhang, D., Higham, A., Wolosianka, S., Gai, X., Zhou, L., Petersen, H., Pinto-Plata, V., Divo, M., Silverman, E. K., Celli, B., Singh, D., Sun, Y., & Owen, C. A. (2020). ADAM15 expression is increased in lung CD8+ T cells, macrophages, and bronchial epithelial cells in patients with COPD and is inversely related to airflow obstruction. Respiratory Research, 21(1), 188. https://doi.org/10.1186/s12931-020-01446-5

Zenke, S., Palm, M. M., Braun, J., Gavrilov, A., Meiser, P., Böttcher, J. P., Beyersdorf, N., Ehl, S., Gerard, A., Lämmermann, T., Schumacher, T. N., Beltman, J. B., & Rohr, J. C. (2020). Quorum Regulation via Nested Antagonistic Feedback Circuits Mediated by the Receptors CD28 and CTLA-4 Confers Robustness to T Cell Population Dynamics. Immunity, 52(2), 313-327.e7. https://doi.org/10.1016/j.immuni.2020.01.018

Author Response:

We are grateful to the reviewers for their insightful comments, suggestions, and criticism. In the updated version of the manuscript, all these will be properly reflected. Here we briefly address the main points raised:

Reviewer #1:

1.1. Patient selection and tumor area selection are crucial for this study but not very carefully defined. Why are some core and others not? Figure referral is an issue here (sup figure 6 where all core and non-core samples are supposed to be according to the legend of Fig 4 is likely sup fig 7 but this is then a complete copy paste of Figure 4). In the methods it is stated that the core samples are based on limited contamination of additional morphotypes (<20%) but Fig 4 suggests that all tumours listed have multiple morphotypes.

The tissue samples were obtained from a hospital cohort of patients with stage II-IV colorectal cancer (at diagnostic time), with no particular selection criteria imposed, as this was an exploratory study.

Tumor regions were marked for macro-dissection by an experienced pathologist following the standard practice for whole-tumor transcriptomics studies. The subregions (morphological regions) were marked by the same experienced pathologist for macro-dissection (in an adjacent section) and reassessed later with respect to their “morphological purity”. It is impossible to macro-dissect regions containing a single morphological pattern. Hence, those regions which contained significant amount (>=20%) of other morphologies were considered “non-core”, while the rest were called “core” regions. This distinction applies to morphological regions solely and not to whole-tumor samples.

Indeed, the reference in caption to Figure 4, should refer to Supp. Fig. 7 (which needs to be updated).

1.2. CMS subtype should be performed with single sample predictor rather than CMScaller.

We agree that a single-sample predictor for CMS is needed, however CMScaller is the de facto classifier for CMS (>130 citations) so we used it to illustrate the practical implications.

1.3. A couple of surprising observations need specification. MUC2 is a strong CMS3 reporter gene yet Mucinous tumours appear to end up in CMS4 rather than 3. Can the authors show that indeed stroma cells are very evident in these samples?

We do not have a direct estimation of the amount of stromal cells, but the high scores of the various fibroblast-related signatures in mucinous regions (Fig2 B, D) indicate that, indeed, there is an enrichment in stroma. In the follow-up study we plan to perform specific staining as well as spatial transcriptomics of these regions to further investigate our findings.

1.4. The SE PP and CT are assigned to CMS2, but in Figure 4 this appears a lot more variable than the authors would make the reader believe. The full data are not completely clear (see point 1).

In the paper, we transparently state that PP, SE, and CT were assigned to CMS2 in 62.5%, 41.7% and 41.9% of cases, respectively. These proportions referred to all samples for which CMSCaller made a prediction. In Fig.4, we also show the proportion of cases in which CMSCaller did not predict any subtype.

1.5. The tumor response rates are rather weird as this is likely dependent on the complete tumour and not so much the subareas. It is not very well described what we see in this analysis.

We did not compute any response rates but simple prognostic scores as (weighted, if weights were provided) means of genes in the specific signatures (see Methods). The question addressed was whether these scores were comparable between whole tumor and corresponding tumor regions (within same tumor). Given the observed (relative) variability, the more important follow-up question - which we cannot answer with our limited survival data – is whether a higher score in a region in comparison with whole-tumor is indeed indicative of a higher risk of relapse.

1.6. Serrated adenomas have previously been aligned with CMS4. Is this different from serrated areas in cancers?

We do not have data from adenomas to compare with the serrated carcinoma regions. But a comparison of (regions of) both traditional serrated and sessile serrated adenomas to serrated carcinoma would be interesting.

1.7. The fact that iCMS2 and iCMS3 align rather well with the current analysis of the distinct regions suggests that the analysis that was reported last year is the proper way to view tumor intrinsic signatures. The authors now propose a rather similar outcome to this issue which does take away a lot of the novelty of the findings of this study.

Our goal was not to propose another stratification paradigm for colorectal cancer, but rather to study the associations between morphology and transcriptome and its implications in practice. As such, our analyses are not limited to molecular subtypes and the respective observations are but a small part of our findings. Indeed, the intrinsic subtypes (iCMS 2/3) are stable and robust, as they are based on the genes expressed in epithelial cells, and they may well prove to be of clinical importance too. However, they do not cover all aspects (e.g. fibroblasts subtypes) and, as stated in Joanito et al. Nat Gen 54, pages 963–975 (2022), “iCMS, MSI status and CMS jointly inform the molecular classification of CRC”. Last, in our opinion, the molecular classification of CRC, while a useful point of view in tumour classification, is not covering all the necessary perspectives on tumour heterogeneity.

Reviewer #2:

2.1. Overall, the manuscript provides an interesting histological/morphological framework through which we can consider heterogeneity in colorectal carcinoma and an approach by which we might improve the performance of gene expression-based classifiers in predicting clinical behaviour and/or responses to therapy. Exploration of CRC morphotypes and their differences was quite interesting. However, more work is needed to support the claims made by the authors. While I appreciate that the authors themselves identify limitations of their study within the manuscript, I believe awareness of these limitations is not reflected in some of the claims made in the abstract and at points in the main text when discussing the use of expression-based classifiers.

We will improve the manuscript to stress the exploratory nature of our analyses and their limitations.

Author Response

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

This important work reports the identification of a list of proteins that may participate in the clearance of paternal mitochondria during fertilization, which is known as essential for normal fertilization and embryonic and fetal development. While the main method used is state of the art and the supporting data are solid, the vigor of the biochemical assays and function validation is inadequate. This work will be of interest to developmental and reproductive biologists working on fertilization. Key revisions (for the authors) include 1) Use a mitochondria-enriched fraction instead of whole sperm for the assays, and add more control samples to monitor what got lost during sperm and oocyte treatments before the coincubation step. 2) Functional validation of the key proteins identified.

We thank Editors of eLife, as well as Special Issue Guest-Editors and Reviewers for a favorable assessment and helpful recommendations for key revisions. Provisional revisions included in our revised article are detailed below. We agree with Editors’ comment about the use of mitochondrion enriched fractions and additional functional validation of key proteins. In fact, we are developing experimental protocols for oocyte extract coincubation with isolated sperm heads and tails, and eventually with purified mitochondrial sheaths, to separate the ooplasmic sperm nucleus remodeling factors from the mitophagic ones. Such experiments, as well as functional validations using porcine zygotes are contingent upon anticipated post-pandemic rebound in the availability of porcine oocytes, obtained from ovaries harvested on slaughterhouse floors, requiring currently unavailable workforce which has hampered our access to this necessary resource.

Reviewer #1 (Peer Review):

Could the authors make clear how much the presented pictures reflect the described localisation? There is no information on the number of spermatozoa and embryos observed nor the fraction of these embryos showing the presented pattern of localisation. This must be included.

Two hundred spermatozoa were counted per replicate of the cell-free system co-incubation and 20 zygotes per replicate, with 3 replicates of immunolabelling for each phase/picture which were examined to establish the typical localization patterns that were observed. The displayed patterns were observed in 65 to 88% of examined spermatozoa/zygotes; varying dependent on protein, replicate, and phase of immunolabelling. In all cases, the signal displayed is the typical pattern that was displayed in most cells. This information has been added to the Materials and Methods section for clarification.

It is not clear if the authors also examined the localization of other proteins and obtained a different pattern than anticipated from the proteomic approach or if they only tested these 6 proteins and got a 100% of correlation.

These are the 6 proteins which were selected based on extensive literature review into known functions of all identified proteins, as well as extensive research into available and reliable antibodies to detect such proteins within our porcine systems. Even so, no particular localization patterns were anticipated; instead, we presented the patterns actually observed and even some patterns which defied our expectations (i.e., the localization of BAG5 in the sperm acrosome).

The authors use "MS" in the text to indicate "mitochondrial Sheath" and "Mass spectrometry". this is confusing.

The authors agree and the usage of MS as an acronym for either has been removed entirely to avoid confusion.

In the introduction the author refers to Ankel-Simons and Cummins, 1996 as a reference for the number of sperm mitochondria in mammalian species, this is incorrect since the quoted paper is about the number of mtDNA molecules and mentioned an earlier publication.

This has been revised and the appropriate citation has been used.

Reviewer #2 (Peer Review):

Major:

1) It has been proved from the earlier studies from this group that the porcine cell-free system is useful to observe spermatozoa interacting with ooplasmic proteins in a single trial and could recapitulate fertilization sperm mitophagy events that take place in a zygote without affecting later cell-division process. However, the post-fertilization sperm mitophagy process is a complex time-associated event that many processes that occur sequentially and interactively, which means ooplasmic proteins might be involved in this process but may not directly interact with sperm or may associate with sperm-ooplasmic protein complex at different time points. It is certainly a great advance already in knowledge to identify "the candidate players" from the list of 185 proteins; however, with the time-resolution (4 and 24hr) in the current study and without functional validation experiments at this stage, it is still difficult to postulate the importance of these identified proteins. The functional validation experimental designs, in my opinion, is critically important for better interpretation of the data.

The authors agree with this reviewer’s sentiments and do plan to conduct further functional analysis. This project was able to generate a list of candidate, sperm-mitophagy promoting proteins and we were further able to show that many of these proteins were detectable both via mass spectrometry and via immunocytochemistry in spermatozoa exposed to our cell-free system. Furthermore, similar localization patterns were found in spermatozoa that were detected within newly fertilized zygotes. These results boost our confidence in our cell-free system and show that our list of candidate proteins is truly a useful list for future localization and functional analyses. We are certainly aware that we have not captured every protein that may play a role in post-fertilization sperm mitophagy and that the proteins captured are just candidates until proven otherwise. Likewise, we have almost certainly captured multiple proteins that are currently candidates that will likely not be shown to play a role in postfertilization sperm mitophagy, while it is plausible that at least some of these candidate proteins do play a role in mitophagy and some of them likely participate (perhaps have yet to be described roles) in other fertilization events, in which we would be extremely interested in as well.

2) As shown in Figure 1, whole sperm was used in the co-incubation and the later MS analysis; thus, proteins identified in the current study might be relevant in fertilization processes other than postfertilization sperm mitophagy, as proteins identified in the current study may be associated with other parts of the sperm (e.g. sticky sperm head, e.g. PSMG2 associated with sperm midpieces, tail at 4hr coincubation, but then only associate with sperm head at 24hr co-incubation) rather than sperm midpiece, despite the fact that authors applied immunohistochemistry to show the localization of this protein, but the evidence is indirect, so how authors functionally differentiate these 6 identified proteins from sperm mitophagy process with other processes and to confirm (or to associate) the relevance of these proteins with sperm mitophagy process?

The authors agree that the 6 proteins which were further studied by using immunocytochemistry may be playing roles in other processes such as pronuclear formation. We discussed some potential roles including and beyond post-fertilization mitophagy, in the Supplemental Discussion. After reviewer comments, we moved the Supplemental Discussion back in the main Discussion section. Thus, this section now considers additional putative pathways in which the said 6 proteins cold participate, though we concede that thorough functional studies must still be performed.

3) Class 3 proteins were present in both the gametes or only the primed control spermatozoa, but are decreased in the spermatozoa after co-incubation, which authors interpreted as sperm-borne mitophagy determinants and/or sperm-borne proteolytic substrates of the oocyte autophagic system, this data categorization may need to be revised as sperm-borne proteolytic substrates of the oocyte autophagic system only, not for sperm borne mitophagy determinants. The argument for this disagreement is due to the fact that if the protein is a sperm-borne mitophagy determinant, after coincubation, to execute the mitophagy process, this protein should still be associated with the sperm at least at the early stage (of 4hr) (constant under MS detection when comparing control with 4hr treated) rather than being released from the sperm. Or alternatively, they could result in class 3 proteins (but not all those 6 were in class 3). Nevertheless, if these proteins serve as substrates, they can be used (consumed) and show decreased under MS detection.

This argument for redefining the Class 3 proteins more accurately is understood and we agree. The definition is revised in the paper.

4) Of particular interest among the 6 proteins that were further investigated. Unlike other proteins, MVP was highly significant (p<0.001) after 4hr incubation, but the significance became less after 24hr (p=0.19). Interpretation of this dynamic change in the relevance of the mitophagy process would facilitate the readers to understand the relevance and the role of MVP.

The differences in significance are likely influenced by the abundance of MVP detectable by mass spectrometry. As the time of cell-free system incubation increases, the variability between replicates also seemed to increase, likely due to the sustained proteolytic activity taking place in our system. This work was based on three replicates of mass spectrometry for each time point; additional replicates likely would have reduced the p-value for the 24hr cell-free data set, for MVP and potentially other proteins also. At both time points, MVP was only detectable in spermatozoa after they had been exposed to the cell-free system treatment which is the criteria that truly interested us more than the actual differences in content between the timepoints and is why it was added to our list of candidate proteins.

5) In figure 3, the association of ooplasmic MVP to sperm midpiece is not convincing enough as sperm midpiece and tail often show some levels of non-specific signals under fluorescent microscopy. And the dynamic association of ooplasmic MVP to sperm midpiece in Fig. 3F-G is difficult to reach a conclusion solely based on data presented in the manuscript. Additional negative control of sperm MVP staining from the primed and treated sperm would be helpful. Additionally, a quantitative comparison (15 vs 25hr) of sperm-associated MVP signals from the fertilized embryo or a stack image from different angles would clarify the doubts raised here.

For all images and all replicates, serum controls were also generated. These controls were then viewed under fluorescent microscope, and light intensities and exposures thresholds for each fluorescent light channel were set based on the background intensity that came from these nonimmune serum-treated control samples. We set our light intensity/acquisition time below a threshold where the non-specific signal began to appear. All the presented patterns are based on setting this peak intensity threshold and as such the signal we see should be the true signal. Furthermore, 200 spermatozoa were counted per treatment per replicate of the cell-free system co-incubation and 20 zygotes per replicate, with 3 replicates of immunolabelling for each protein and data point, which was used to represent the typical localization patterns that were observed. The displayed patterns were observed between in 65- 88% of examined spermatozoa/zygotes. Invariably, the signal displayed in the manuscript is the typical pattern that was seen in a majority of cells. This information has now been added to the Materials & Methods section for clarification.

6) Same concerns for the other 5 proteins (PSMG2, PSMA3, FUNDC2, SAMM50, BAG5) as indicated above.

See response to Question 5.

7) The patterns of these 6 proteins under the immunofluorescent study are confusing as the pattern varies after co-incubation (treated), and mostly, the signal of these proteins observed from the fertilized embryos is not really associated with sperm midpieces. Therefore, the evidence of these proteins involving in post-fertilization sperm mitophagy is, at this moment, weak based on the data presented. But the relevance of these proteins in events post-fertilization or early embryo development is certainly (evidence did not strong enough to support "sperm mitophagy," in my opinion).

The authors agree that some of these proteins seem to be playing roles beyond postfertilization sperm mitophagy and that there is a need for true functional studies before the authors can state with certainty that these proteins play a role in any of the discussed fertilization events. We state this in the discussion: “Considering the dynamic proteomic remodeling of both the oocyte and spermatozoa which takes place during early fertilization, these 185 proteins which have been identified likely play roles in processes beyond sperm mitophagy.” It should be noted that the authors went into greater detail about potential alternative protein functions based on the present data and literature review in the Supplemental Discussion. Based on this comment and other reviewer comments we have now included the Supplemental Discussion as part of the main Discussion section, and this will hopefully help clarify some of the authors’ thoughts about the 6 candidate proteins which were further analyzed during this study.

Minor:

1) To my understanding, statistical significance (relevance) is normally set at a p-value of either <0.1 or 0.05. The reason for loosening the p-value of 0.2 in the current study needs to be justified as this was not a common statistical criterium, and the interpretation of those candidates from this loosened criterium should also be careful.

The loosening of statistical relevance in this study to 0.2, only applied to our Class 1 proteins. This is because for a protein to fall into the Class 1 proteins it was a protein that was only present in samples after they were exposed to the cell-free system. In the case of these Class 1 proteins, this happened for all 3 replicates at each stated timepoint. We found this pattern of detection to be important whether the p-value fell under 0.1 or 0.2. As such, we loosened our statistical threshold for our Class 1 proteins. Any proteins added to our candidate list will be subject to further investigation before definitive conclusions can be drawn, and as such we think that capturing more proteins was more important for the goals of this study than limiting the number of proteins captured, especially for those Class 1 proteins. An explanation of this has been added to the Materials & Methods section Mass Spectrometry Data Statistical Analysis.

2) First cell cleavage of porcine embryo normally occurs within 48hr post-insemination or activation; therefore, the 4 and the 24hr time points used in the current study require justification included in the discussion or methods and material section.

First cleavage of porcine embryos normally occurs around 24 - 28 hours post-insemination. Thus, for both the cell-free system and the embryo studies we were capturing an advanced 1 cell stage zygote/zygote like system with our 24 hour and 25-hour time points.

3) In figure 2, colors used in different time points and in two different classes represent (sometimes) different protein categories, would be easier for the readers for quick comparisons if the same color could be used to represent the same protein category throughout the graph. (E.g, proteins for early zygote development are shown in red in "A", but blue in "B")

This has been corrected and the color scheme for Figure 2 has been revised for easier comparisons.

Reviewer #3 (Peer Review):

I am not used to seeing a supplementary discussion in a manuscript. I also believe it should be incorporated into normal discussion.

The Supplemental Discussion has been incorporated into the main Discussion now.

It would be very helpful to make an additional figure in which the proposed interactome of identified factors with the sperm mitochondria before and after incubation are drawn schematically and also which factors are not IDed in both cases (when comparing to somatic mito- or autophagy). This eases to get through the discussion and will beautifully summarize and illustrate the importance and progress that the authors have made with this assay.

We made a diagram that depicts the changes in protein localization patterns overtime within our cell-free system. This diagram has been added to the manuscript as Figure 9.

Reviewer #1 (Public Review):

In this manuscript, the authors used an unbiased method to identify proteins from porcine oocyte extracts associated with permeabilised boar spermatozoa in vitro. The identification of the proteins is done by mass spectrometry. A previous publication of this lab validated the cell-free extract purification methods as recapitulating early events after sperm entry in the oocyte. This novel method with mammalian gametes has the advantage that it can be done with many spermatozoa at the time and allows the identification of proteins associated with many permeabilised boar spermatozoa at the time. This allowed the authors to establish a list of proteins either enriched or depleted after incubation with the oocytes extract or even only associated with spermatozoa after incubation for 4h or 24h. The total number of proteins identified in their test is around 2 hundred and with very few present in the sample only when spermatozoa were incubated with the extracts. The list of proteins identified using this approach and these criteria provide a list of proteins likely associated with spermatozoa remnants after their entry and either removed or recruited for the transformation of spermatozoa-derived structures. Using WB and histochemistry labelling of spermatozoa and early embryos using specific antibodies the authors confirmed the association/dissociation of 6 proteins suspected to be involved in autophagy.

While this unique approach provides a list of potential proteins involved in sperm mitochondria clearance it's (only) a starting point for many future studies and does not provide the demonstration that any of these proteins has indeed a role in the processes leading to sperm mitochondria clearance since the protein identified may also be involved in other processes going-on in the oocyte at this time of early development.

We thank reviewer 1 for positive comments. We added a sentence in Discussion addressing the obvious shortcoming of present study, as further functional validations of candidate mitophagy factors are planned.

Concerning the localisation of the 6 proteins further analysed, the authors must add how much the presented picture represents the observed patterns. They must include the details on the fraction of spermatozoa and embryos displaying the presented pattern.

We now specify that the patterns depicted in manuscript are typical and representative of data from at least three replicates of immunolabeling in spermatozoa and zygotes. For each of these replicates, 200 spermatozoa were examined per replicate of the cell-free system co-incubation or 20 zygotes per replicate. The displayed patterns were observed between 65-88% in examined spermatozoa/zygotes. Invariably, the signal displayed in manuscript is the typical pattern that was seen in a majority of cells. This information has now been added to the Materials & Methods section for clarification.

Reviewer #2 (Public Review):

Mitochondria are essential cellular organelles that generate ATPs as the energy source for maintaining regular cellular functions. However, the degradation of sperm-borne mitochondria after fertilization is a conserved event known as mitophagy to ensure the exclusively maternal inheritance of the mitochondrial DNA genome. Defects on post-fertilization sperm mitophagy will lead to fatal consequences in patients. Therefore, understanding the cellular and molecular regulation of the postfertilization sperm mitophagy process is critically important. In this study, Zuidema et. al applied mass spectrometry in conjunction with a porcine cell-free system to identify potential autophagic cofactors involved in post-fertilization sperm mitophagy. They identified a list of 185 proteins that might be candidates for mitophagy determinants (or their co-factors). Despite the fact that 6 (out of 185) proteins were further studied, based on their known functions, using a porcine cell-free system in conjunction with immunocytochemistry and Western blotting, to characterize the localization and modification changes these proteins, no further functional validation experiments were performed. Nevertheless, the data presented in the current study is of great interest and could be important for future studies in this field.

We thank reviewer 2 for positive comments. As we explain in our response to Editors and Reviewer 1, further validation studies will be resumed once the availability of slaughterhouse ovaries for such studies improves. Examples of such functional validation of pro-mitophagic proteins SQSTM1 and VCP are included in our previous studies (DOI: 10.1073/pnas.1605844113 and DOI: 10.3390/cells10092450) that led to the development of cell-free system reported here, and are cited in present study.

Reviewer #3 (Public Review):

In this manuscript, a cytosolic extract of porcine oocytes is prepared. To this end, the authors have aspirated follicles from ovaries obtained from by first maturing oocytes to meiose 2 metaphase stage (one polar body) from the slaughterhouse. Cumulus cells (hyaluronidase treatment) and the zona pellucida (pronase treatment) were removed and the resulting naked mature oocytes (1000 per portion) were extracted in a buffer containing divalent cation chelator, beta-mercaptoethanol, protease inhibitors, and a creatine kinase phosphocreatine cocktail for energy regeneration which was subsequently triple frozen/thawed in liquid nitrogen and crushed by 16 kG centrifugation. The supernatant (1.5 mL) was harvested and 10 microliters of it (used for interaction with 10,000 permeabilized boar sperm per 10 microliter extract (which thus represents the cytosol fraction of 6.67 oocytes). The sperm were in this assay treated with DTT and lysoPC to prime the sperm's mitochondrial sheath. After incubation and washing these preps were used for Western blot (see point 2) for Fluorescence microscopy and for proteomic identification of proteins.

Points for consideration:

1) The treatment of sperm cells with DTT and lysoPC will permeabilize sperm cells but will also cause the liberation of soluble proteins as well as proteins that may interact with sperm structures via oxidized cysteine groups (disulfide bridges between proteins that will be reduced by DTT).

This is certainly a possibility, the lysoPC and DTT permeabilization steps were designed to mimic natural processing (plasma membrane removal and sperm protein disulfide bond reduction), which the spermatozoa would undergo during fertilization. However, we do realize that this is a chemically induced processing and thus is not a perfect recapitulation of fertilization processes. However, in this study and in previous studies with this system, we were able to show alignment between proteomic interactions taking place in the cell-free system and within the zygotes.

2) Figure 3: Did the authors really make Western blots with the amount of sperm cells and oocyte extracts as the description in the figures is not clear? This point relates to point 1. The proteins should also be detected in the following preparations (1) for the oocyte extract only (done) (2) for unextracted nude oocytes to see what is lost by the extraction procedure in proteins that may be relevant (not done) (3) for the permeabilized (LPC and DTT treated and washed) sperm only (not done) (4) For sperm that were intact (done) (5) After the assay was 10,000 permeabilized sperm and the equivalent of 6.67 oocyte extracts were incubated and were washed 3 times (or higher amounts after this incubation; not done). Note that the amount of sperm from one assay (10,000) likely will give insufficient protein for proper Western blotting and or Coomassie staining. In the materials and methods, I cannot find how after incubation material was subjected to western blotting the permeabilized sperm. I only see how 50 oocyte extracts and 100 million sperm were processed separately for Western blot.

The authors did make Western blots with the number of spermatozoa and oocytes stated in the materials and methods, a total protein equivalent of 10 to 20 million spermatozoa (equivalent to ~20-40 µg of total protein load) and 100 MII oocytes (equivalent to ~20 µg of total protein load). These numbers have been corrected in the Materials & Methods. Also, we did find in the Materials & Methods section that the Co-Incubation of Permeabilized Mammalian Spermatozoa with Porcine Oocyte Extracts section refers to using cell-free exposed spermatozoa for electrophoresis; however, for none of the presented Western blot work was this true. Rather, all of the presented Western blots as per their descriptions are utilizing ejaculated or capacitated sperm or oocytes. This line has been removed from the Materials & Methods to reduce confusion.

Regarding preparation (2), we have previously assessed the difference between oocyte extract and intact oocytes in this manner internally and we are certainly losing proteins due to the oocyte extraction process. We make caveats in this vein throughout the article such as: “Furthermore, this cell-free system while useful does not perfectly capture all the events which take place during in vivo fertilization. The cell-free system is intended to mimic early fertilization events but is presumably not the exact same as in vitro fertilization.”

3) Figures 4, 5, 6, 7, and 8 see point 2. I do miss beyond these conditions also condition 1 despite the fact that the imaged ooplasm does show positive staining.

For all the presented Western blots, the tissue type is stated in the image description and the protocol which was used to prepare these samples is stated in the Materials & Methods.

4) These points 1-3 are all required for understanding what is lost in the sperm and oocyte treatments prior to the incubation step as well as the putative origin of proteins that were shown to interact with the mitochondrial sheath of the oocyte extract incubated permeabilized sperm cells after triple washing. Is the origin from sperm only (Figs 5-8) or also from the oocyte? Is the sperm treatment prior to incubation losing factors of interest (denaturation by DTT or dissolving of interacting proteins preincubation Figs 3-8)?

The authors understand that there are proteins and interactions lost on both sides of the cellfree system equation and we have added a sentence to the Discussion to caveat this limitation in the system.

5) Mass spectrometry of the permeabilized sperm incubated with oocyte extracts and subsequent washing has been chosen to identify proteins involved in the autophagy (or cofactors thereof). The interaction of a number of such factors with the mitochondrial sheath of sperm has been shown in some cases from sperm and others for an oocyte origin. Therefore, it is surprising that the authors have not sub-fractionated the sperm after this incubation to work with a mitochondrial-enriched subfraction. I am very positive about the porcine cell-free assay approach and the results presented here. However, I feel that the shortcomings of the assay are not well discussed (see points 1-5) and some of these points could easily be experimentally implemented in a revised version of this manuscript while others should at least be discussed.

We agree that the use of a mitochondrial-enriched subfraction for further analysis would be interesting and useful. We are actively developing experimental protocols for oocyte extract coincubation with isolated sperm heads and tails, and eventually with purified mitochondrial sheaths. However, such experiments are contingent upon our access to porcine oocytes, which has continued to be a struggle since the COVID-19 pandemic compromised our ability to attain oocytes in large, cheap, and reliable quantities. This was a continuous problem with preparing materials for this very paper and has continued to be an issue for our laboratory as well as many others at our university and across the country. We continue to maximize oocytes every time we can get access to them, but the unfortunate reality is that this access has become sparce and unreliable over the past three years.

Author Response

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

Reviewer #1 (Public Review):

The expression and localization of Foxc2 strongly suggest that its role is mainly confined to As undifferentiated spermatogonia (uSPGs). Lineage tracing demonstrated that all germ cells were derived from the FOXC2+ uSPGs. Specific ablation of the FOXC2+ uSPGs led to the depletion of all uSPG populations. Full spermatogenesis can be achieved through the transplantation of Foxc2+ uSPGs. Male germ cell-specific ablation of Foxc2 caused Sertoli-only testes in mice. CUT&Tag sequencing revealed that FOXC2 regulates the factors that inhibit the mitotic cell cycle, consistent with its potential role in maintaining a quiescent state in As spermatogonia. These data made the authors conclude that the FOXC2+ uSPG may be the true SSCs, essential for maintaining spermatogenesis. The conclusion is largely supported by the data presented, but two concerns should be addressed: 1) terminology used is confusing: primitive SSCs, primitive uSPGs, transit amplifying SSCs... 2) the GFP+ cells used for germ cell transplantation should be better controlled using THY1+ cells.

Thanks for your good comments. According to your suggestions, we have addressed your two concerns as follows:

1> Overall our work suggest that FOXC2+ SSCs are a subpopulation of SSCs in a quiescent state, thus we have replaced the term ‘primitive’ with ‘quiescent’ in the revised manuscript. In general, ‘transient amplifying SSCs’ is considered to be ‘progenitors’, thus we have replaced ‘transient amplifying SSCs’ with ‘progenitors’ in the revised manuscript.

2> The transplantation experiment was conducted using MACS-sorted THY1+, FACS sorted THY1+, and FACS-sorted GFP+ (FOXC2+) uSPGs simultaneously. To be consistent with the single-cell RNA-seq using the MACS-sorted THY1+ uSPGs, we only presented the results from MACS-sorted THY1+ and FACS-sorted GFP+ (FOXC2+) uSPGs in the previous manuscript. Following the reviewer’s suggestion, we have included the results derived from FACS sorted THY1+ uSPGs as the control. The overall conclusion is still fully supported by the more comprehensive dataset, i.e. FOXC2+ cells generated significant higher numbers of colonies than THY1+ cells after transplantation (Figure 2D, E).

Reviewer #2 (Public Review):

The authors found FOXC2 is mainly expressed in As of mouse undifferentiated spermatogonia (uSPG). About 60% of As uSPG were FOXC2+ MKI67-, indicating that FOXC2 uSPG were quiescent. Similar spermatogonia (ZBTB16+ FOXC2+ MKI67-) were also found in human testis.

The lineage tracing experiment using Foxc2iCreERT2/+;Rosa26LSL-T/G/LSL-T/G mice demonstrated that all germ cells were derived from the FOXC2+ uSPG. Furthermore, specific ablation of the FOXC2+ uSPGs using Foxc2iCreERT2/+;Rosa26LSL-DTA/+ mice resulted in the depletion of all uSPG population. In the regenerative condition created by busulfan injection, all FOXC2+ uSPG survived and began to proliferate at around 30 days after busulfan injection. The survived FOXC2+ uSPGs generated all germ cells eventually. To examine the role of FOXC2 in the adult testis, spermatogenesis of Foxc2f/-;Ddx4Cre/+ mice was analyzed. From a 2-month-old, the degenerative seminiferous tubules were increased and became Sertoli cell-only seminiferous tubules, indicating FOXC2 is required to maintain normal spermatogenesis in adult testes. To get insight into the role of FOXC2 in the uSPG, CUT&Tag sequencing was performed in sorted FOXC2+ uSPG from Foxc2iCreERT2/+;Rosa26LSL-T/G/LSL-T/G mice 3 days after TAM diet feeding. The results showed some unique biological processes, including negative regulation of the mitotic cell cycle, were enriched, suggesting the FOXC2 maintains a quiescent state in spermatogonia.

Lineage tracing experiments using transgenic mice of the TAM-inducing system was well-designed and demonstrated interesting results. Based on all data presented, the authors concluded that the FOXC2+ uSPG are primitive SSCs, an indispensable subpopulation to maintain adult spermatogenesis.

The conclusion of the mouse study is mostly supported by the data presented, but to accept some of the authors' claims needs additional information and explanation. Several terminologies define cell populations used in the paper may mislead readers.

1) "primitive spermatogonial stem cell (SSC)" is confusing. SSCs are considered the most immature subpopulation of uSPG. Thus, primitive uSPGs are likely SSCs. The naming, primitive SSCs, and transit-amplifying SSCs (Figure 7K) are weird. In general, the transit-amplifying cell is progenitor, not stem cell. In human and even mouse, there are several models for the classification of uSPG and SSCs, such as reserved stem cells and active stem cells. The area is highly controversial. The authors' definition of stem cells and progenitor cells should be clarified rigorously and should compare to existing models.

Thanks for your good comments. Considering that our results showed that FOXC2+ SSCs are in a quiescent state and that Mechanistically FOXC2 maintained the quiescent state of SSCs by promoting the expression of negative regulators of cell cycle, we have replaced ‘primitive SSCs’ with ‘quiescent SSCs’ in the revised manuscript. We agree with the reviewer that ‘transient amplifying SSCs’ is considered to be ‘progenitors’, thus we have replaced ‘transient amplifying SSCs’ with ‘progenitors’ in the revised manuscript. Further，from our point of view, the FOXC2+Ki67+ SSCs could be regarded as active stem cells, and the FOXC2+Ki67- SSCs could be regarded as reserved stem cells, although further research evidence is still needed to confirm this.

2) scRNA seq data analysis and an image of FOXC2+ ZBTB16+ MKI67- cells by fluorescent immunohistochemistry are not sufficient to conclude that they are human primitive SSCs as described in the Abstract. The identity of human SSCs is controversial. Although Adark spermatogonia are a candidate population of human SSCs, the molecular profile of the Adark spermatogonia seems to be heterogeneous. None of the molecular profiles was defined by a specific cell cycle phase. Thus, more rigorous analysis is required to demonstrate the identity of FOXC2+ ZBTB16+ MKI67- cells and Adark spermatogonia.

We agree with the reviewer that the identity of human SSCs remain elusive even though Adark population demonstrates certain characteristics of SSCs. To acknowledge this notion, we have revised our conclusion as such that only suggests FOXC2+ZBTB16+MKI67- represents a quiescent state of human SSCs.

3) FACS-sorted GFP+ cells and MACS-THY1 cells were used for functional transplantation assay to evaluate SSC activity. In general, the purity of MACS is significantly lower than that of FACS. Therefore, FACS-sorted THY1 cells must be used for the comparative analysis. As uSPGs in adult testes express THY1, the percentage of GFP+ cells in THY1+ cells determined by flow cytometry is important information to support the transplantation data.

Thanks for your good comments. According to your suggestions, we have addressed your concerns as follows:

1> The transplantation experiment was conducted using MACS-sorted THY1+, FACS sorted THY1+, and FACS-sorted GFP+ (FOXC2+) uSPGs simultaneously. To be consistent with the single-cell RNA-seq using the MACS-sorted THY1+ uSPGs, we only presented the results from MACS-sorted THY1+ and FACS-sorted GFP+ (FOXC2+) uSPGs in the previous manuscript. Following the reviewer’s suggestion, we have included the results derived from FACS sorted THY1+ uSPGs as the control. The overall conclusion is still fully supported by the more comprehensive dataset, i.e. FOXC2+ cells generated significant higher numbers of colonies than THY1+ cells after transplantation (Figure 2D, E).

2> We performed FACS analysis to determine the proportion of GFP+ cells in FACS-sorted THY1+ cells from Rosa26LSL-T/G/LSL-T/G or Foxc2iCreERT2/+;Rosa26LSL-T/G/LSL-T/G mice at day 3 post TAM induction, and the result showed that GFP+ cells account for approximately 20.9±0.21% of THY1+ cells, See Author response image 1.

Author response image 1.

4) The lineage tracing experiments of FOXC2+-SSCs in Foxc2iCreERT2/+;Rosa26LSL-T/G/LSL-T/G showed ~95% of spermatogenic cells and 100% progeny were derived from the FOXC2+ (GFP+) spermatogonia (Figure 2I, J) at month 4 post-TAM induction, although FOXC2+ uSPG were quiescent and a very small subpopulation (~ 60% of As, ~0.03% in all cells). This means that 40% of As spermatogonia and most of Apr/Aal spermatogonia, which were FOXC2 negative, did not contribute to spermatogenesis at all eventually. This is a striking result. There is a possibility that FOXC2CRE expresses more widely in the uSPG population although immunohistochemistry could not detect them.

Thanks for your good comments. From our lineage tracing results, over 95% of the spermatogenic cells are derived from the FOXC2+ SSCs in the testes of 4-month-old mice, which means that FOXC2+ SSCs maintain a long-term stable spermatogenesis. In addition, previous studies have shown that only a portion of As spermatogonia belong to SSCs with complete self-renewal ability (PMID: 28087628, PMID: 25133429), which is consistent with our findings. Therefore, we speculate that 40% of As spermatogonia and most of Apr/Aal spermatogonia, which were FOXC2 negative, did contribute to spermatogenesis but cannot maintain a long-term spermatogenesis due to limited self-renewal ability.

5) The CUT&Tag_FOXC2 analysis on the FACS-sorted FOXC2+ showed functional enrichment in biological processes such as DNA repair and mitotic cell cycle regulation (Figure 7D). The cells sorted were induced Cre recombinase expression by TAM diet and cut the tdTomato cassette out. DNA repair process and negative regulation of the mitotic cell cycle could be induced by the Cre/lox recombination process. The cells analyzed were not FOXC2+ uSPG in a normal physiological state.

We do appreciate the reviewer’s concern on the possibility of the functions enriched in the analysis as referred might be derived from Cre/lox recombination. However, we think it is unlikely that the Cre/lox recombination process, supposed to be rather local and specific, can trigger such a systemic and robust response by the DNA damage and cell cycle regulatory pathways. The reasons are as follows: First, as far as we are aware, there has been sufficient data to support this suggested scenario. Second, we did not observe any alteration in either the SSC behaviors or spermatogenesis in general upon the TAM-induced genomic changes, suggesting the impact from the Cre/lox recombination on DNA damage or cell cycle was not significant. Third, no factors associated with the DNA repair process were revealed in the differential analysis of single-cell transcriptomes of FOXC2-WT and FOXC2-KO.

6) Wei et al (Stem Cells Dev 27, 624-636) have published that FOXC2 is expressed predominately in As and Apr spermatogonia and requires self-renewal of mouse SSCs; however, the authors did not mention this study in Introduction, but referred shortly this at the end of Discussion. Their finding should be referred to and evaluated in advance in the Introduction.

Thanks for your good comments. According to your suggestion, we have revised the introduction to refer this latest parallel work on FOXC2. We are happy to see that our discoveries are converged to the important role of FOXC2 in regulating SSCs in adult mammals.  

Reviewer #3 (Public Review):

By popular single-cell RNA-seq, the authors identified FOXC2 as an undifferentiated spermatogonia-specific expressed gene. The FOXC2+-SSCs can sufficiently initiate and sustain spermatogenesis, the ablation of this subgroup results in the depletion of the uSPG pool. The authors provide further evidence to show that this gene is essential for SSCs maintenance by negatively regulating the cell cycle in adult mice, thus well-established FOXC2 as a key regulator of SSCs quiescent state.

The experiments are well-designed and conducted, the overall conclusions are convincing. This work will be of interest to stem cell and reproductive biologists.

Thanks for the positive feedback.  

Reviewer #1 (Recommendations for the Authors):

The authors should address the following concerns:

1) The most primitive uSPGs should be the true SSCs. The term "primitive SSCs" is very confusing.

2) In addition to FACS-sorted GFP+ cells, FACS-sorted THY1+ cells should also be used for transplantation.

Thanks for your good comments. According to your suggestions, we have addressed your two concerns as follows:

1) Overall our work suggest that FOXC2+ SSCs are a subpopulation of SSCs in a quiescent state, thus we have replaced the term ‘primitive’ with ‘quiescent’ in the revised manuscript.

2) The transplantation experiment was conducted using MACS-sorted THY1+, FACS sorted THY1+, and FACS-sorted GFP+ (FOXC2+) uSPGs simultaneously. To be consistent with the single-cell RNA-seq using the MACS-sorted THY1+ uSPGs, we only presented the results from MACS-sorted THY1+ and FACS-sorted GFP+ (FOXC2+) uSPGs in the previous manuscript. Following the reviewer’s suggestion, we have included the results derived from FACS sorted THY1+ uSPGs as the control. The overall conclusion is still fully supported by the more comprehensive dataset, i.e. FOXC2+ cells generated significant higher numbers of colonies than THY1+ cells after transplantation (Figure 2D, E).

Reviewer #3 (Recommendations for the Authors):

The experiments are well-designed and conducted, the overall conclusions are convincing. The only concerns are the writing, especially the introduction which was not well-rationalized. Sounds the three subtypes and three models for SSCs' self-renew are irrelevant to the major points of this manuscript. I don't think you need to talk too much about the markers of SSCs. Instead, I suggest you provide more background about the quiescent or activation states of the SSCs. In addition to that, as a nuclear-localized protein, it cannot be used to flow cytometric sorting, I don't think it should be emphasized as a marker. You identified a key transcription factor for maintaining the quiescent state of the primitive SSCs, that's quite important!

Appreciate the positive feedback and constructive suggestions on the writing. We have substantially revised our manuscript to include the relevant advances and understanding from the field as well as highlight the importance of FOXC2 in regulating the quiescent state of SSCs.

Author Response

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

Reviewer 1 (Recommendations For The Authors):

1) The strikingly different conclusion from the previous Bourane study seems to stem from the experimental approaches. Rather than using genetic crosses that target all neurons from the hindbrain and spinal cord that express Npy at any point in development, Boyle et al target their manipulations specifically to the lumbar region of the superficial dorsal horn in adult mice using direct viral injections. Thus, Boyle is almost certainly manipulating much fewer neurons that the original study. How then is their behavioral effects so much greater? At the minimum, the authors need to discuss this discrepancy head on. Better would be a direct molecular/anatomical comparison of the neurons targeted by each approach. This could be done using Nyp-Cre mice crossed to a Rosa-LSL-reporter strain and quantifying the overlap with the same markers used here. Perhaps, the intersectional approach with Lbx1 resulted in labeling of a different population of neurons than the adult AAV injections? Although likely outside the scope, given this work directly questions the main conclusion of the Bourane paper, it will be important to see a replication of the original finding of selectivity to mechanical itch.

We agree that our approach should be manipulating a smaller population of neurons, and that it is therefore suprising that we see greater behavioural effects. Please see our response to "Weakness 1" of Reviewer 2 for consideration of this point. We have already provided a direct molecular comparison as requested by the reviewer, and this appears in Figure 1 supplement 1. Here we used tissue from NPY::Cre that had been crossed with Ai9 mice (i.e. a Rosa-LSL-reporter) and had received intraspinal injections of AAV.flex.GFP. We then characterised the neurochemistry of tdTomato+ cells that were GFP+ or GFP-negative.

2) The authors state that, "91.6% ± 0.3% of cells classed as Cre-positive cells were also Npy-positive, and these accounted for 62.1% ± 0.6% of Npy-positive cells" If I am reading this correctly, does that mean that 40% of the Npy+ cells are Cre negative? If so, how is this possible?

This interpretation is correct. For quantification of RNAscope data we used a cut-off level of 4 transcripts, and cells with fewer than 4 transcripts were classed as negative. It is likely that some of the NPY cells classified as negative for Cre would have had some Cre mRNA (sufficient to cause recombination), but at a level below this threshold. It is also possible that some NPY+ cells would fail to express Cre, since this is a BAC transgenic mouse, rather than a knock-in.

3) Similarly, the authors state that "great majority of FP-expressing neurons in laminae I-III were immunoreactive (IR) for NPY (78.5% ± 3.6%), and these accounted for 74.6% ± 109 1.9% of the NPY-IR neurons in this area". So does this mean 20% of the recombination is non-specific/in other cell types that could be involved in pain/itch sensation?

Our finding that 91.6% of cells with Cre mRNA were also positive for Npy mRNA (see above) indicates that Cre expression was largely restricted to NPY cells. The failure to detect NPY peptide in some of these cells probably results from the relatively low level of peptide seen in the cell bodies of peptidergic neurons, which results from the rapid transport of peptides into their axons.

4) Comparing Fig 3B and Fig4B it seems the control baseline von Frey responses are different. In fact, baseline response in Fig4b is quite like the CNO effect in Fig 3B. Unless I'm misunderstanding something, this seems quite odd?

We agree that there is a difference between the baseline responses. We are not aware of any particular reason for this, and we think that it reflects a degree of variability that is seen with the von Frey test. Interestingly, the baseline values for the SNI cohort (Fig 4E) lies between the values in Fig 3B and Fig 4B.

5) In Fig 4E, the behavior of the CNO treated mice is quite variable. Can the authors comment as to how this might be happening? Does the effect correlate with viral transduction?

We did not see any obvious correlation between the extent of viral transduction and the behaviour of individual mice.

6) Fig6, the PDyn-Cre experiment, is a bit of a non sequitur?

Please see our response to "Weakness 2" of Reviewer 2 for consideration of this point.

7) The conclusion is unusually long. I recommend trimming it to make it more concise.

We presume that this refers to the Discussion. However, this was ~1550 words, and we do not feel that that is unusually long.

Reviewer 2 (Public Review):

Weaknesses

1) There is inadequate discussion about previous studies of NPY interneurons. Specifically, the authors should address why a more restricted subset of these neurons (this study) have broader effects than seen previously.

We have expanded the discussion on the discrepancies between our findings and those reported previously. We state at the outset that we are targeting a more restricted population (lines 509-10), and we now go into more detail concerning both similarities and differences between our findings and the reasons that we think may underlie any discrepancies (various changes between lines 522-575).

2) I cannot see the reason for including results from manipulation of Dyn+ interneurons in this paper. First, the title does not reflect roles of spinal Dyn+ population. In addition, without further experiments characterizing relationships between NPY and Dyn interneurons in modulating itch and/or nociception, Dyn datasets seem to deviate from the main theme.

We had previously shown that activating Dyn-INs suppressed pruritogen-evoked itch (Huang et al 2018), but it was important to test whether silencing these cells would have the opposite effect. Our finding of overlap in function (i.e. both NPY-INs and Dyn-INs suppress itch, and that both innervate GRPR cells) provides strong evidence against the idea that neurochemically-defined interneuron populations have highly specific functions, and we now state this in the Discussion. The anatomical experiments (which follow on from the functional studies) provide important new information concerning synaptic circuitry of the dorsal horn, by showing that NPY-INs preferentially innervate GRPR cells, and provide around twice as many synapses on these cells, compared to the Dyn-INs. Interestingly, this correlates with the relatively large optogenetically-evoked IPSCs that we saw when NPY-INs were activated, compared to those reported by Liu et al (2019) when galanin-expressing (which largely correspond to Dyn-INs) were activated. By including these findings in the paper, we are able to make comparisons between these two populations.

3) While the authors provided convincing evidence that GRPR+ neurons serve as a downstream effector of NPY+ neuron evoked itch, the relationship between GRPR and NPY neurons in modulating pain is not examined. Therefore, Fig. 7B is pure speculation and should be removed.

We feel that our recent findings that GRPR neurons correspond to vertical cells, that they respond to noxious stimuli, and that activating them results in pain-related behaviours, makes it reasonable to speculate that the NPY/GRPR circuit may also be involved in the anti-nociceptive action of NPY cells. The legend for Fig 7B already refers to this as a "potential circuit", and we have toned down the corresponding part of the discussion to say that our findings "raise the possibility" that this is the case (lines 605-7). We feel that this part of the figure is important, as otherwise our summary diagram ignores some of the main findings of the paper, and we hope that this is now acceptable.

Recommendations For The Authors

1) Fig. 1G: the "misexpression" of tdTomato neurons was much more prominent in deep dorsal horn laminae but not in the superficial ones. Was this representative? Can the authors perform a laminae specific characterization?

We did test for this possibility in 2 NPY::Cre;Ai9 mice that had received intraspinal injections of AAV.flex.GFP, and found that there was a modest difference - 62% of tdTomato+ cells in laminae I-II, but only 39% of those in lamina III, were GFP+. This suggests that "misexpression" may have differed slightly between these regions. However, since the difference was quite modest, and we were only able to analyse tissue from two mice in this way, we did not include these findings in the paper.

2) I have a lot of problems interpreting the c-Fos data in Fig. 2 E and F. For the mCherry- population, how was the quantification performed? From the image, it does not look like 2030% of cells express c-Fos; at a minimum a clear stain of neurons would be needed. Similarly, the identification of NPY cells is not particularly convincing (e.g., middle arrowhead lower 2 panels in C).

We have provided further details on how the analysis was performed (changes made to lines 1016-29). NeuN staining was used to reveal all neurons, and a modified optical disector method was performed from somatotopically appropriate regions of the dorsal horn. As noted by the Reviewer, NeuN staining was required to allow identification of mCherrynegative cells. However, we have not included the NeuN immunoreactivity in the image, as this would add considerably to the complexity. These images are from single optical sections, and therefore the overall numbers of cells are low (in comparison to what would be seen in a projected image). The intensity of mCherry staining varied between cells. However, for all mCherry-positive cells (including the example referred to by the Reviewer), there was clear staining in the membrane, which could be followed in serial sections.

3) Please add individual data points for all quantifications.

These have been added.

Reviewer 3 Recommendations For The Authors:

1) It is somewhat surprising that there is no effect on CPP after activating spinal NPY neurons in neuropathic mice, given the almost complete rescue of hypersensitivity to baseline values in the nociceptive tests. Based on the methods, it appears that conditioning was carried out already 5 min after CNO injection. Yet, suppression of c-fos activity in excitatory spinal dh neurons was observed 30min after CNO injection. Also, it is not clear to me when CNO was injected prior to the nociceptive or CQ testing?

Have the authors considered that conditioning from 5-35 min after CNO injection might be too short after CNO injection to achieve a profound analgetic effect?

In a previous study (Polgár et al 2023), we had observed the timecourse of CNO-evoked itch and pain behaviours in mice in which GRPR cells expressed hM3Dq. We found that these started within 5 minutes of i.p. CNO injection (e.g. Fig S2 in that paper). In addition, the timecourse of action of gabapentin and CNO (both given i.p.) are likely to be similar, and there was a preference for the chamber paired with gabapentin. We are therefore confident that the conditioning period with CNO was adequate. We now explain this in the Methods section (lines 846-52). The timing of CNO injections for the nociceptive and CQ tests is now described (lines 749-55).

2) The authors claim that tonic pain was not affected based on the conditioned place preference test. Efficacy in withdrawal response tests and in the CPP differ by more than duration of the stimulus. I'd suggest using more cautious wording here.

We agree that caution is needed in interpreting the results of the CPP experiments. We have therefore replaced "does" with "may" in the Results section (line 336) and "did" with "may" in the Discussion (line 620).

3) On page 9 the authors state "...suggesting that they suppress the transmission of pain- and itch-related information in the dorsal horn." However, pain is not affected in the loss of function experiments suggesting some qualitative differences in the role of the NPY neurons in itch and pain. This should also be reflected more clearly in this statement and in the discussion e.g. "suppress itch" and "can suppress pain".

We accept the point made by the Reviewer. We have slightly altered the wording in lines 249-51 and 610 to reflect this.

Author Response

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

Reviewer #1 (Public Review):

[...] Weaknesses

Showing that A-2 and especially A-3 are outliers in the PCA analysis is useful, but it may be hiding other interesting signals in the data. The other strains are remarkably colinear on these plots, hinting that if the outliers were removed, one main component would emerge along which they are situated. It also seems possible that this additional analysis step would allow the second dimension to better differentiate them in a way that is interesting with respect to their mutator status or mutations in key metabolic or regulatory genes.

We thank the reviewer for their positive comments and their constructive feedback on the manuscript. Following reviewer’s recommendation, we performed the PCA analysis on metabolism data after removing A-2 and A-3 data. We have detailed those results below. Consistent with a similar analysis performed on RNA-seq datasets in our previous publication, we find that removing these outliers has only a modest effect on separating mutators from non-mutators. We find that, while the new PC2 separates most mutators from the non-mutators, the separation is rather weak. Moreover, we do not see a similar distinction when looking at metabolic data in the Stationary phase. In the interest of improving the readability of the manuscript, we recommend not including these analysis in the final manuscript. We have presented the data for the reviewer’s benefit in Author response image 1, 2 and 3.

Author response image 1.

Author response image 2.

Author response image 3.

There is a missed opportunity to connect some key results to what is known about LTEE mutations that reduce the activity of pykF (pyruvate kinase I). This gene is mutated in all 12 LTEE populations, and often these mutations are frameshifts or transposon insertions that should completely knock out its activity. At first glance, inactivating an enzyme for a step in glycolysis does not make sense when the nutrient source in the growth medium is glucose, even though PykF is only one of two isozymes E. coli encodes for this reaction. There has been speculation that inactivating pykF increases the concentration of phosphoenolpyruvate (PEP) in cells and that this can lead to increased rates of glucose import because PEP is used by the phosphotransferase system of E. coli to import glucose (see https://doi.org/10.1002/bies.20629). The current study has confirmed the higher PEP levels, which is consistent with this model.

We thank the reviewer for pointing out this missed opportunity. We have expanded the discussion around the role of pykF mutations and the elevated concentrations of PEP observed in our data in section 3.4.

In the introduction, the papers cited to show the importance of changes in metabolism for adaptation do not seem to fit the focus of this study very well. They stress production of toxins and secondary metabolites, which do not seem to be mechanisms that are at work in the LTEE. I can think of two areas of background that would be more relevant: (1) studies of how bacterial metabolism evolves in adaptive laboratory evolution (ALE) experiments to optimize metabolic fluxes toward biomass production (for example, https://doi.org/10.1038/nature01149), and (2) discussions of how cross-feeding, metabolic niche specialization, and metabolic interdependence evolve in microbial communities, including in other evolution experiments (for example, https://doi.org/10.1073/pnas.0708504105 and https://doi.org/10.1128/mBio.00036-12).

We thank the reviewer for pointing out missed citations in our introduction. We agree that these papers are relevant to the topic and have added their citations. Additionally, following the suggestion of another reviewer, we have reorganized the introduction so that the concept of the role of metabolism in evolution is presented first and the LTEE second.

Reviewer #2 (Public Review):

[...] Overall, this is a significant and well-executed research study. It offers new insights into the complex relationship between genetic changes and observable traits in evolving populations and utilizes metabolomics in the LTEE, a novel approach in combination with RNA-seq and mutation datasets.

However, the paper's overall clarity is lacking. It is spread too thin and covers many topics without a clear focus. I strongly recommend a substantial rewrite of the manuscript, emphasizing structure and readability. The science is well executed, but the current writing does not do it justice.

We thank the reviewer for their positive comments and their constructive feedback on the lack of clarity in writing. Following the reviewer’s suggestions, we have rewritten parts of the manuscript and reorganizd a few sections to improve readability. We hope the revised manuscript is significantly improved.

Recommendations for the authors

Reviewer #1 (Recommendations For The Authors):

1) Title and Abstract: Add the study organism to the abstract, and probably also the title. Currently, E. coli is not mentioned in either! I'm also not sure that the LTEE is a sufficiently well-known acronym to abbreviate this in the title.

We have revised the title of the manuscript and now spell out LTEE and included E. coli in the title and the abstract.

2) Abstract: I would switch the usage of metabolome to metabolism in a few more places. For example, "changes in its metabolism", "networked and convoluted nature of metabolism". The metabolome, the concentrations of all metabolites, is what is being measured, but I think of this as a phenotypic readout of how metabolism evolving.

We have changed “metabolome” to “metabolism” in cases where we refer to what is evolving and use “metabolome” when we refer to what is being measured.

3) Line 16: Technically, the 12 LTEE populations were not initially identical. The Ara- differed from the Ara+ ancestors by one intentional mutation and one unintentional mutation that was not discovered until whole genomes were sequenced. I would rephrase this to "where 12 replicate populations of E. coli are propagated" or something similar so that it can be correct without needing to describe this unnecessary detail.

The line has been rephrased as suggested.

4) General Note: The text refers to populations as Ara-3 but the figures use A-3. I'd suggest going with A-3 and similar throughout for consistency.

Instances of Ara have been changed to A+/-, and a sentence specifying as such has been added to the intro to make mention of this.

5) Lines 43-44, 97-98. My understanding is that both S and L ecotypes in A-2 can use both glucose and acetate, but that the differentiation is related to their specialization that leads to each one being better on one or the other nutrient. The descriptions make it sound like each grows at a different time. Also, by definition, cells are not growing during "stationary phase". The change from glucose utilization (and acetate secretion) to acetate utilization during one cycle of growth is better described as a diauxic shift.

We have reworded this part to remove mention of “growth” during stationary phase and changed the wording such that it no longer sounds like they grow at different times.

6) Line 54: The statement "provide the ability to test hypotheses from previous data" is vague. Either provide an example or delete.

We have removed this sentence as suggested.

7) Lines 71-72: The terms "interphase" and "intraphase" sound too much like parts of the cell cycle. I'd suggest describing the comparisons as between and within growth phases.

The use of intra and interphase have been changed as suggested.

8) Line 79: The citrate is presumably still a chelating agent, so change phrasing to "Citrate is present in the medium because it was originally added as a chelating agent" or something similar.

This sentence has been rewritten as suggested.

9) Line 83: Write out "mutation accumulations" so it is easier to understand as "the number of mutations that have accumulated".

The phrase has been changed as suggested.

10) Line 116: It's unclear whether the abundances of metabolites are "strategies of survival" in stationary phase. An equally valid explanation is that there is less selection on the metabolome to have a specific composition during stationary phase to have high fitness.

We have added a line about the possibility for alternative hypotheses.

11) Figure 1: There seems to be some information missing from the legend. What are R06 and R07 in Panels A and B? Is panel D exponential phase and panel E stationary phase?

This information was inadvertently missing from the caption and has been added.

12) Figures 2 and 3: Gene names should be in italics. To me, the gray for deleted genes is hard to tell apart from the blue/red. Perhaps you could put a little X in these boxes instead? I think that having a little triangle pointing from each gene or metabolite name its corresponding abundance panel would help the reader track which information goes with which features. In Fig. 3 the placement of L-aspartate is a bit awkward. I'd suggest moving it down so the dashed line does not have to go through the abundance panel.

These figures have been edited to include small triangles that link a gene or metabolite and its heatmap. Additionally, an X has been added where genes have suffered inactivating mutations and the placement of some elements has been moved to improve overall clarity.

13) Lines 183-185: It would be easier to see and judge the consistency of these argR related relationships if a correlation graph of some kind was shown, probably as a supplemental figure. This plot could, for example, have genes/metabolites across the x-axis and fold-change on the y-axis with lines connecting points corresponding to each of the twelve populations across these categories (like Fig S8 but with lines added). Alternatively, it could be a heat map with the populations across one axis and the genes/metabolites across the other axis (like Fig S3).

We have added a supplementary figure consisting of heatmaps showing the consistency of these changes within an evolved line. It is now figure S9.

14) Line 195: I think adding a sentence elaborating on what exactly mutation accumulation means in this context would be helpful to readers.

We have attempted to clarify the meaning of this by specifically stating that it is due to the accumulation of deleterious mutations.

15) Line 293: Is standard LTEE medium DM25? These omics experiments with the LTEE sometimes use similar media with different glucose concentrations, and this is a very important detail to precisely specify.

We reference “standard” LTEE medium in the methods section and have additionally specified the amount of sugar to make it clear that we are not supplementing the media with additional sugar.

16) Figure S8B. Is "cystine" used instead of "cysteine" on purpose here since the compound is oxidized in the metabolomics treatment?

The use of cystine is intentional, we detect the oxidized compound.

Reviewer #2 (Recommendations For The Authors):

Title:

The abbreviation "LTEE" should not be in the title. Most readers will not recognize what it means. Instead, either the full name of the experiment, "Long-Term Evolution Experiment with E. coli," should be used, or the title should be rephrased to "Linking genotypic and phenotypic changes during a long-term evolution experiment using metabolomics."

We have spelled out LTEE and included E. coli in the title.

Abstract:

Sentence 1: Consider softening the statement: "Do changes in an organism's environment, genome, or gene expression patterns often lead to changes in its metabolome?"

We have rephrased this sentence to “Changes in an organism's environment, genome, or gene expression patterns can lead to changes in its metabolism”.

Sentence 4: Use a hyphen for "Long-Term."

This addition has been made.

Sentence 4: Replace "transduce" with a more appropriate term: "...how the effects of mutations can be distributed through a cellular network to eventually affect metabolism and fitness."

We have rewritten this sentence as “to understand how mutations can eventually affect metabolism and perhaps fitness”.

Sentence 5: Clarify the use of "both" to refer to the ancestor of the LTEE and its descendant populations as two classes.

We have reworded this sentence so it’s clear that the ancestors and evolved lines are two separate classes “We used mass-spectrometry to broadly survey the metabolomes of the ancestral strains and all 12 evolved lines…”.

Sentence 6: Reverse the order for better emphasis: "Our work provides a better understanding of how mutations might affect fitness through the metabolome in the LTEE, and thus provides a major step in developing a complete genotype-phenotype map for this experimental system."

We have rearranged this sentence per the reviewers suggestion.

Introduction:

Revise the introduction for clarity, readability, and logical narrative progression. Start with the second paragraph to set up the basic scientific principles being studied and then transition to describing the LTEE as a model system to examine those principles.

The introduction has been rearranged and reworded in parts to increase clarity.

Sentence 1: Revise for clarity: "The Long-Term Evolution Experiment (LTEE) has studied 12 initially identical populations of Escherichia coli as they have evolved in a carbon-limited, minimal glucose medium under a daily serial transfer regime."

Sentence 2: Suggestion: "Begun in 1988, the LTEE populations have evolved for more than 75,000 generations, making it the longest-running experiment of its kind."

Paragraph 2, sentence 2: Italicize "Drosophila."

Paragraph 3, sentence 2: Make an important distinction: "Ara-3 is unique in that it evolved the ability to grow aerobically on citrate."

Paragraph 3, sentence 4: Introduce the IS-mediated loss of the rbs operon in the LTEE as if it has not been described elsewhere.

These suggestions have been incorporated into the manuscript.

Results:

Section 3.1: The use of samples from hours 2 and 24 to represent exponential and stationary phase may present some issues. For instance, capturing Ara-3 during its exponential growth on glucose, but not citrate, at hour 2. Furthermore, except for Ara-3, the LTEE populations reach stationary phase after approximately 4 hours, and there could be significant differences between early, mid, and late stationary phase. This possibility should be acknowledged, and future follow-up work should consider exploring these differences.

We have added sentences in the first paragraph of the results section to include these details. We have also added a short paragraph to the conclusions suggesting additional studies of stationary phase, citing work on evolution of E. coli during long term stationary phase.

Paragraph 3: While Turner et al. 2017 is an essential reference regarding resource use differences between Ara-3 and other LTEE populations, it would be more suitable to reference Blount et al. 2012 for the mutations that enabled access to citrate. Also, it is important to note that the difference lies in the ability to grow aerobically on citrate, rather than the ability to metabolize it.

This citation has been added.

Paragraph 4: As mentioned elsewhere, most LTEE populations exhibit balanced polymorphisms. Therefore, it is more appropriate to state that Ara-2 is the best-understood example of long-term diversity. It is likely that there are important metabolic differences between co-existing lineages in other LTEE populations.

We now refer to Ara-2 as being the best-understood example of long term diversity..

Paragraph 5: The first sentence of this paragraph should likely end with "levels."

The word “levels” was added to the end of this sentence.

Figure 3: It is preferable to refer to the "Superpathway of arginine and polyamine biosynthesis," citing EcoCyc as a reference, rather than a descriptor.

This has been changed to a reference.

Section 3.3, Paragraph 3: While higher intracellular amino acid abundances may facilitate higher translation rates and faster growth, the higher abundances themselves do not evaluate the hypothesis. To evaluate the hypothesis, it is necessary to demonstrate that higher abundances are associated with higher translation or growth rates. Therefore, the final sentence of this paragraph is not meaningful.

We have reworded this sentence to say that it’s not possible to tell what the additional amino acids are being used for given only this data and that additional experiments are needed to confirm this hypothesis.

Section 3.4: The first paragraph of this section misstates how evolution works. The low level of glucose in the LTEE does not drive innovation; instead, innovation occurs at random through the introduction of variation by mutation. Although the existence of the citrate resource acts as a reward that selects for variation that provides access to it, it is essential to remember that evolution is blind to such a reward. Moreover, regarding the evolution of the Cit+ trait, it is incorrect to assert that low glucose contributed to its evolution. As shown by Quandt et al. (2015), it seems probable that Cit+ evolution was potentiated by adaptation to specialization on acetate, which is produced by overflow metabolism resulting from rapid growth on glucose. This rapid growth only occurs when glucose is relatively abundant. The level of glucose seems low to us because it is low relative to traditional levels in bacteriological media, but not to the bacteria.

We agree that this is a semantical, but important distinction. We have reworded this part as to not suggest that evolution has any forward thinking properties and is indeed blind to any rewards that might occur as the result of adaptation.

In general, all instances of "utilize" and its cognates should be replaced with "use" and its cognates.

Instances of “utilize” have been changed to use and its cognates.

There is some uncertainty about the expectation of ramping up the TCA cycle in the LTEE. Overflow metabolism and acetate production appear to be prevalent in the LTEE, suggesting that many lineages only partially oxidize carbon derived from glucose, thereby bypassing the TCA cycle. While it is possible that this interpretation is incorrect, it would be helpful to see it addressed in the manuscript.

We agree that this is a plausible hypothesis, we have added a paragraph at the end of this section that discusses the implications of overflow metabolism as an alternative hypothesis.

Author Response

Reviewer #1 (Public Review):

In this study, the authors study the effect of dynactin disruption on kinetochore fiber (k-fiber) length in spindles of dividing cultured mammalian cells. Dynactin disruption is known to interfere with dynein function and hence spindle pole formation. The main findings are that poles are not required for correct average k-fiber length and that severed k-fibers can regrow to their correct length both in the presence and absence of poles by modulating their dynamic properties at both k-fiber ends. In the presence of poles, regrowth is faster and the variation between k-fiber lengths is smaller. This is a very interesting study with high-quality quantitative imaging data that provides important new insight into potential mechanisms of spindle scaling, extending in an original manner previous work on this topic in cultured cells and in Xenopus egg extract. The Discussion is interesting to read as several possible mechanisms for k-fiber length control are discussed. The technical quality of the study is very high, the experiments are very original, and most conclusions are well supported by the data. Especially, the experiments observing the regrowth of k-fibers after severing and the study of the dynamic properties of these k-fibers provide very novel insight. Addressing the following concerns could potentially improve the manuscript:

We thank the reviewer for their fair, rigorous, and conceptually engaging remarks.

(1) The phenotype generated here by disrupting dynactin via overexpressing p50 appears to be different from that caused by knocking down NuMA or dynein - as previously reported by the Dumont lab (Hueschen et al., 2019). In this study here, unfocused spindles are observed whereas earlier turbulent spindles were observed. This raises the question of whether dynein activity that contributes to pole focusing is really completely inhibited here. These discrepancies in phenotypes seem to deserve an explanation. Is k-fiber length in cultured mammalian cells only maintained in the case of this specific type of inhibition?

We thank the reviewer for the important point about the different phenotypes observed in different dynein inhibition conditions and we refer them to our response to Essential Revision #1. In summary, we believe that different dynein inhibition phenotypes are similar. Unfocused spindles appear turbulent on longer timescales and appear to reach a steady-state on shorter timescales. The amount of pole-unfocusing also seems to correspond to the severity of dynein inhibition (Figure 1—figure supplement 1). We have chosen to study inhibited spindles that were steady-state and unfocused. We have added this discussion in line 129 as well as better characterized our system of dynein inhibition by adding two new figures (Figure 1—figure supplement 1, Figure 1—figure supplement 3).

Furthermore, we address the question of whether dynein might still be responsible for length regulation despite poles being unfocused in line 433 of the Discussion: “recent work has revealed that mammalian spindles can achieve similar architecture whether or not dynein (or its recruiter NuMA) is knocked out (Neahring et al., 2021). This suggests that the severe defects in spindle coordination (Figure 1, Figure 5) and maintenance (Figure 2) observed in p50-unfocused spindles are more likely due to the loss of spindle poles than due to the loss of dynein activity per se.”

We have additionally overexpressed p50 in human RPE1 cells and observed qualitatively similarly unfocused yet generally bi-oriented spindles as in rat kangaroo PtK2 cells, showing that the formation of unfocused spindles in PtK2 is not an artifact unique to that cell line (see newly added Figure 1—figure supplement 3). However, these unfocused RPE1 spindles did not have clear, resolvable k-fibers as in PtK2, so length was not quantified. The only method we are aware of that robustly unfocuses poles in PtK2 spindles is p50 overexpression.

(2) p50 addition and also p150-cc1 addition was often used in Xenopus egg extract in order to inhibit dynein function. Considerably larger concentrations of p50 than p150-cc1 needed to be used. Can the authors estimate the level of overexpression of p50 in the cells they study? It seems that could be possible given that a mCherry fusion protein can be overexpressed. Was it necessary to select cells with a particular level of mCherry-p50 overexpression to observe the reported phenotypes?

We thank the reviewers for the suggestion to quantify p50 expression and have added Figure 1—figure supplement 1. Due to gradual red laser power loss over months, data from a single day were plotted for proper comparison, but trends were always consistent within any given day. As discussed above, we observed that higher levels of mean p50 intensity corresponded to unfocused spindles. We have clarified that we chose to study these highly overexpressing unfocused spindles in the text and methods, and we speculate that level of p50 overexpression correlates with amount of dynein inhibition and subsequent pole-unfocusing. This is also consistent with the higher concentrations of p50 needed to inhibit dynein in Xenopus.

(3) Some comparison to previous experiments using p50 and p150-cc1 addition to Xenopus egg extract spindles could put this study better into the context of the available literature. It seems from previous publications that the p50 addition produced short, unfocused, barrel-shaped spindles, indicating that spindle length is maintained without poles, whereas the p150-cc1 addition produced elongating spindles (e.g. Gaetz & Kapoor, 2004).

We appreciate the reviewer’s discussion of dynein inhibition in the Xenopus context.

While Xenopus has been used to study spindle size regulation, it has not been as useful to study k-fiber length regulation, which we focus on. Xenopus spindles have a different architecture, with k-fibers that are not discrete and continuous like in mammalian spindles. Indeed, while p50 and p150-CC1 overexpression alter spindle length in Xenopus, they do not have the same effect in mammalian spindles. Additionally, p150-CC1 does not robustly unfocus poles in mammalian spindles as it does in Xenopus; instead, it leads to an inconsistent variety of spindle disorganization phenotypes with frequently focused poles in PtK2 (data not shown). We speculate this variety of spindle phenotypes arise from a different mechanism of dynein inhibition that does not fully target pole-focusing.

However, we agree that referencing prior Xenopus work establishes important context and precedent. In line 95 of the Introduction, we state “…inhibiting dynein unfocuses poles but spindles still form albeit with altered lengths in Drosophila (Goshima et al., 2005) and Xenopus (Gaetz and Kapoor, 2004; Heald et al., 1996; Merdes et al., 1996), and without a clear effect on mammalian spindle length (Guild et al., 2017; Howell et al., 2001),” addressing the different effects of dynein inhibition in Xenopus compared to mammalian spindles. We have also added direct mentions of p50 in Xenopus in line 129 (see Essential Revision #1 response).

Finally, we have added a figure showing overexpression of p50 in a human RPE1 cells to show reproducibility of pole unfocusing across other mammalian cell types (see newly added Figure 1—figure supplement 3).

(4) In this context, it seems that some more explanation is required for the observations presented in Fig. 1D and 1E. It appears that spindle length and k-fiber length have been measured quite differently. Not much information is provided for how spindle length was defined and measured (please expand this part of the Methods). Could the two different methods of measurement be the reason for the mean k-fiber length remaining unaltered in dynactin-disrupted spindles, whereas the spindle length increases in these cells? If not, do non-k-fiber microtubules contribute to unfocused spindles being longer or are chromosomes not aligned in the metaphase plate causing the increase in spindle length by misalignment of k-fiber sister pairs?

We thank the reviewers for pointing out the lack of clarity in Figures 1D and 1E. We have expanded and clarified the Methods section describing how spindle axes were measured and how k-fiber lengths were measured, as well as included examples and cartoons to illustrate them (see newly added Figure1—figure supplement 4).

To clarify, we did not intend to directly measure spindle length, but we did approximate the size of each spindle’s “footprint” in Figure 1D as well as measure individual k-fiber length in Figure 1E. It is now clarified in the Methods line 898 as “Spindle minor and major axes lengths were determined by cropping, rotating, then thresholding spindle images with the Otsu filter using SciKit. Ellipses were fitted to thresholded spindles to approximate the length of their major and minor axes using SciKit’s region properties measurement (Figure1—figure supplement 4A). In control spindles, the major axis corresponded to spindle length along the pole-to-pole axis, and the minor axis corresponded to spindle width along the metaphase plate axis. However, unfocused spindles were disorganized along both axes to the extent where the minor axis did not always correspond to the metaphase plate axis. Thus, Figure 1D reports ”spindle minor axis length” and “spindle major axis length” rather than “spindle width” and “spindle length”. Furthermore, it is worth noting that in unfocused spindles, spindle length is decoupled from k-fiber length because of k-fiber disorganization along both axes. Thus, spindle length was not measured in unfocused spindles...”

We additionally removed the potentially confusing terminology of “wider” and “longer” in the Results section to make clear that we are approximating spindle size, not spindle length and width, and we now state in line 168,“ k-fibers were more spread out in the cell, with spindles covering a larger area compared to control along both its major and minor axes (Figure 1D).”

We believe our clarification and expansion of the Methods section, as well as inclusion of a new supplementary figure and cartoon address the reviewer’s points, and we thank them for pointing out the lack of clarity.

(5) It seems that in the Discussion it is implied that k-fibers can respond to severing in both focused and unfocused spindles by modulating their dynamics at both ends of the k-fibers, but in the Results section the wording is more cautious because of the difference in 'flux' in severed and unsevered unfocused spindles is not significant (Fig. 4D, blue data). It appears indeed that there is also a difference in flux between severed and unsevered unfocused spindles, but the number of data points is too small. Depending on how difficult these experiments are, it could be worth increasing the size of the data set to come to a clear conclusion, given that the data shown in Figs. 3 and 4 are quite remarkable and form the core of the study.

We appreciate the reviewer’s close reading and pertinent suggestions.

As detailed in our response to Essential Revision #3, we did not increase the sample size for unfocused spindles since it would not be reasonably feasible to show significant differences in flux. However, we performed more ablations and photomarking in control spindles as detailed in our response to this reviewer’s point 6 below, a different but related point.

(6) Can the authors exclude that the stopping of 'flux' at minus ends after severing is due to some sort of permanent damage induced by ablation? In other words, do severed spindles begin to flux again once they have regrown to their original length?

We thank the reviewer for their important points.

We have addressed this question in the newly added Figure 4—figure supplement 1 as described in our response to Essential Revision #3 to show that flux resumes after length recovery. In summary, we observed no adverse effects of ablation on k-fiber minus-ends. Severed k-fibers have restored lengths, and minus-end dynamics several minutes after ablation.

(7) To this reader, the conceptualization of distinguishing between 'global' and 'local' effects/behavior was a little confusing, both in the title and also later in the text. The concept of 'local' regulation of k-fiber length appears to contradict the observation that k-fiber length can be regained after severing by changes in the dynamics at both ends (so at two very different locations) which is a rather remarkable finding. Maybe distinguishing between 'individual' and 'collective' k-fiber behavior could be clearer.

We appreciate the reviewer’s consideration of terminology. We have addressed this by clearly defining our use of ‘local’ to refer to individual k-fibers as a unit where appropriate in the text (lines 271, 449). We chose these terms since they can help describe individual versus collective properties, while simultaneously emphasizing the aspects of global architecture and spatial organization in the spindle.

(8) Can the authors exclude that some of the differences between unfocused and focused spindles could be due to altered dynein activity at kinetochores? Or due to the dynein-dependent accumulation of certain spindle proteins along microtubules towards the minus ends of k-fibers or other spindle microtubules, instead of being due to only the presence versus absence of poles? Could this be tested by ablating both poles? If this is too challenging, a discussion of these possibilities could be justified.

We appreciate the reviewer’s consideration of kinetochore activity as well as other methods of removing poles. However, p50 overexpression is currently the only method to robustly unfocus spindles in PtK2 cells – ablating poles or removing pole-associated structures such as centrosomes does not abolish pole-focusing in this system (Khodjakov et al., 2000). Furthermore, we now discuss the possibility that altered dynein activity (such as activity at kinetochores) may give rise to the phenotypes we describe in our work in line 433: “…recent work has revealed that mammalian spindles can achieve similar architecture whether or not dynein (or its recruiter NuMA) is knocked out (Neahring et al., 2021). This suggests that the severe defects in spindle coordination (Figure 1, Figure 5) and maintenance (Figure 2) observed in p50-unfocused spindles are more likely due to the loss of spindle poles than due to the loss of dynein activity per se. Though we cannot exclude it, this also suggests that the findings we make in unfocused spindles are not due changes in activity of the dynein population at kinetochores.”

Reviewer #2 (Public Review):

The mitotic spindle of eukaryotic cells is a microtubule-based assembly responsible for chromosome segregation during cell division. For a given cell type, the steady-state size and shape of this structure are remarkably consistent. How this morphologic consistency is achieved, particularly when one considers the complex interplay between dynamic microtubules, spatial and temporal regulation of microtubule nucleation, and the activities of several microtubule-based motor proteins, remains a fundamental unanswered question in cell biology. In this work by Richter et al., the authors use biochemical and biophysical perturbations to explore the feedback between mitotic spindle shape and the dynamics of one of its main structural elements, kinetochore fibers (k-fibers) - bundles of microtubules that extend from kinetochores to spindle poles. Overexpression of the p50 dynactin subunit in mammalian tissue culture cells (Ptk2) was used to inhibit the microtubule motor cytoplasmic dynein resulting in misshapen spindles with unfocused poles. Measurements of k-fiber lengths in control and unfocused conditions showed that although mean k-fiber length was not statistically different, the variation of length was significantly higher in unfocused spindles, suggesting that k-fiber length is set locally, occurring in the absence of focused poles. With a clever combination of live-cell imaging with photoablation and/or photobleaching of fluorescently-labeled k-fibers, the authors went on to explore the mechanistic bases of this length regulation. K-fiber regrowth following ablation occurred in both conditions, albeit more slowly in unfocused spindles. Paired ablation and localized photobleaching on the same k-fiber revealed that microtubule dynamics, specifically those at the plus-end, can be tuned at the level of individual k-fiber. Lastly, the authors show that chromosome segregation is severely impaired when cells with unfocused spindles are forced to enter mitosis. The work's biggest strength is the application of an innovative experimental approach to address thoughtful and well-articulated hypotheses and predictions. Conclusions stemming from the experiments are generally well-supported, though the experiments addressing the "tuning" of k-fiber dynamics could be bolstered by additional data points and perhaps better presented. The manuscript would also benefit from the inclusion of some investigation of spatial differences in the observed effects as well as the molecular and biophysical basis of the observed feedback between k-fiber length and focused poles.

We appreciate the reviewer providing pertinent, rigorous, and intellectually astute suggestions.

Comments/Concerns/Questions:

1) In the discussion, the authors acknowledge that the changes in spindle morphology resulting from p50 overexpression are likely also causing changes in the well-characterized RanGTP/SAF gradients that radiate from chromosome surfaces. Why did the authors did not include an analysis of k-fiber length as a function of positioning within the spindle? The inclusion of this data would not require more experimentation and could be added as a plot showing K-fiber length versus distance from the geometric center of the spindle (defined by the intersection of the major and minor axes perhaps?).

We thank the reviewer for this pertinent suggestion and refer them to our response to Essential Revision #2. Briefly, we have added the recommended analyses to Figure 1—figure supplement 6 by correlating k-fiber length to position along the spindle’s longitudinal and latitudinal axes.

2) The authors also acknowledge the established relationship between MT length and MT end dynamics, yet in their ablation studies, the average initial k-fiber length at ablation in control spindles was higher than that for k-fibers in unfocused spindles. It seems that this difference makes the interpretation of the data, particularly the conclusion that fiber growth rates differ due to the absence of focused poles, a bit tenuous. To address this, the authors should consider including plots of grow-back rates versus k-fiber length (again, this should not require additional experiments, just more analysis).

We thank the reviewer for their critical thinking about experiments. We would like to clarify to the reviewer that initial k-fiber lengths within unfocused spindles preceding ablation were not actually longer on average compared to the average length of control k-fibers from Figure 1E (Figure 2—figure supplement 1). We apologize that this unexpected artifact was not clear in the text and have now reworded line 232 to be more straightforward: “Mean k-fiber lengths in unfocused spindles before ablation appeared to be shorter (Figure 2D); however, this was due to not capturing the full length of k-fibers in a single z-plane while imaging ablated k-fibers. Indeed, length analysis of full z-stacks from unfocused spindles before ablation yielded an indistinguishable mean k-fiber length compared to control k-fibers in Figure 1E (Figure 2—figure supplement 1). Thus, ablated k-fibers were compared to their unablated neighbors as internal controls.”

We believe that this language clearly calls out the perceived inconsistency, and that our use of internal controls overcomes this confounding factor to make meaningful conclusions. We address the relationship of k-fiber length and growth rate in our response to Essential Revision #2. We are not including it in the manuscript based on our inability to make any meaningful conclusion to either support or exclude the possibility of length-dependent growth rates.

3) As presented, the data shown in Figure 4 is confusing and does not seem very compelling. The relationship between the kymographs and time series is unclear as is the relationship between the dashed lines in the kymographs and the triangles and the plots in the 4B time series and 4C, respectively. Furthermore, it's not always clear what the triangles are pointing to (e.g. in the unfocused condition time series). The authors might want to consider reworking this figure and providing more measurements of flux following ablation in both the control and unfocused conditions. Lastly, the authors should clarify what negative displacement means.

We apologize for the unclear figure annotations and thank reviewers for their suggestions. As discussed in our response to Essential Revision #3, we believe we have improved the clarity and presentation of figures and kymographs. More measurements of flux after ablation in unfocused spindles was not feasible as discussed; however, we have performed these measurements in control spindles and added Figure 4—figure supplement 1 to strengthen conclusions about turning flux off/on after ablation.

We have additionally clarified axis titles by replacing “negative displacement” with the more intuitive descriptor “photomark position relative to minus-end” and clearly defining it in the figure legends in line 565 as follows: “Figure 3 […] (D) Minus-end dynamics, where photomark position over time describes how the mark approaches the k-fiber’s minus-end over time in control and unfocused k-fibers.”

We thank reviewers for their suggestions to improve clarity and bolster our conclusions.

Author Response

We thank the Editor for his assessment. We agree that the data we present in this manuscript can be a starting point for more in-depth analysis. We are currently developing a mathematical model of HIV transmission dynamics; we plan to use the data that we present in this paper as parameter values.

Reviewer #1 (Public Review):

One aim of this paper was to study historical migration from Botswana during the time of the development of the HIV epidemic. The second aim was to test whether the migration networks impacted the development of the epidemic. The first aim was achieved: this paper used historical census data in a clear way, to describe the qualities of characteristics of migration in the country at four points in time, from 1981 to 2011. Very detailed data are presented in clear ways, using network chord diagrams, sharing age- and sex-specific migration rates, and urban-rural classifications. However, data was not presented to achieve the second aim. The authors reviewed some important literature about migration and HIV. They suggested that the migration patterns, such as from specific mining towns and mostly between districts, could have been important in supporting the generalized spread of HIV. But without evidence linking HIV prevalence over time in the linked districts in Botswana, this aim was not supported.

We have now made it clear that we are not testing whether the migration networks impacted the development of Botswana’s HIV epidemic: this is what the Reviewer describes as the second aim of our paper. We have only one aim: to test the hypothesis that, during the development of Botswana’s HIV epidemic, the population was extremely mobile and highly connected through migratory flows and counter-flows. This is based on the fact that these conditions are necessary for the development of a generalized HIV epidemic. However – previous to our analysis – these conditions have not been shown to occur during the development of a generalized HIV epidemic. Given that our results support our mobility hypothesis (i.e., that the population was very mobile and essentially all the districts were connected throughout the country), in the discussion (lines 338-362) we describe how the migration networks that we have identified may have impacted the development of the generalized hyperendemic HIV epidemic in Botswana. We have also clarified that our study has only one hypothesis that we are testing by referring to this single hypothesis as the mobility hypothesis (Abstract: lines 25-29).

One other limitation of the paper was that very little context, outside of migration rates, was provided. Is there any additional information about economic growth, or political event for example, that could clarify or add context to these migration flows? As it stands now, these analyses are quite basic and don't take into account underlying demographic, economic, or political trends.

In response to this concern we have expanded the text in the introduction to provide more context regarding political, demographic and economic factors (Introduction: lines 66-75). We have also expanded our discussion of the implications of our results (and of additional results that we have included: lines 263-283) for understanding the role of internal migration on urbanization in Botswana (Discussion: lines 379-420); urbanization occurred simultaneously to the development of Botswana’s generalized hyperendemic HIV epidemic.

The data presented in this paper has potential impact. As the paper stands now, it could be quite useful for future work when linked to additional data sources on HIV prevalence over time (or other questions that could have been influenced by migration patterns).

We thank this Reviewer for their helpful comments.

Reviewer #2 (Public Review):

To provide context into the HIV epidemic in Botswana over the latter half of the 20th century and the beginning of the 21st, the authors have analyzed micro census data to examine patterns of migration. They use this dataset to show how patterns between urban and rural areas have changed over several decades, and the demographic characteristics of migrants. The dataset used for this study is a very reliable source, and the insights in terms of migration patterns are interesting. The primary weakness of the analyses regards the link to HIV transmission: micro-census data only examine mobility that leads to individuals changing residence for longer periods of time, without accounting for shorter-term trips that may also lead to HIV transmission, such as seasonal migration or short trips. This is likely less of an issue with HIV than other diseases, however, due to its transmission often involving new sexual partners, which will generally be less likely to occur during short trips. Broadly, however, this is an interesting report on the migration patterns during a critical period for HIV transmission nationwide.

We thank the Reviewer for their comments.

In our current manuscript, we have discussed the potential impact of mobility on Botswana’s HIV epidemic, and focused on migration, i.e., one directional movement in terms of a permanent re-location of residency. This type of migration, by changing an individual’s sexual network and social environment, has been shown to increase the risk of acquiring HIV for both women and men. Short-term mobility (e.g., short-term circular migration, where the trip can range in duration from overnight to an entire season) can also affect HIV transmission dynamics. Circular migrants have been shown to both have an increased risk of acquiring HIV, and of transmitting HIV. The greater the number of trips and/or the duration of the trip, the greater the risk. We note that both migration and short-term mobility are important, and their relative importance to each other is likely to evolve over time as a generalized HIV epidemic diffuses through the population. Their relative importance is also likely to vary amongst countries in sub-Saharan Africa.

We have added all of the previous paragraph, with citations, to the text (Discussion: lines 364-377).

Author Response

Reviewer #1 (Public Review):

1) Although I found the introduction well written, I think it lacks some information or needs to develop more on some ideas (e.g., differences between the cerebellum and cerebral cortex, and folding patterns of both structures). For example, after stating that "Many aspects of the organization of the cerebellum and cerebrum are, however, very different" (1st paragraph), I think the authors need to develop more on what these differences are. Perhaps just rearranging some of the text/paragraphs will help make it better for a broad audience (e.g., authors could move the next paragraph up, i.e., "While the cx is unique to mammals (...)").

We have added additional context to the introduction and developed the differences between cerebral and cerebellar cortex, also re-arranging the text as suggested.

2) Given that the authors compare the folding patterns between the cerebrum and cerebellum, another point that could be mentioned in the introduction is the fact that the cerebellum is convoluted in every mammalian species (and non-mammalian spp as well) while the cerebrum tends to be convoluted in species with larger brains. Why is that so? Do we know about it (check Van Essen et al., 2018)? I think this is an important point to raise in the introduction and to bring it back into the discussion with the results.

We now mention in the introduction the fact that the cerebellum is folded in mammals, birds and some fishes, and provide references to the relevant literature. We have also expanded our discussion about the reasons for cortical folding in the discussion, which now contains a subsection addressing the subject (this includes references to the work of Van Essen).

3) In the results, first paragraph, what do the authors mean by the volume of the medial cerebellum? This needs clarification.

We have modified the relevant section in the results, and made the definition of the medial cerebellum more clear indicating that we refer to the vermal region of the cerebellum.

4) In the results: When the authors mention 'frequency of cerebellar folding', do they mean the degree of folding in the cerebellum? At least in non-mammalian species, many studies have tried to compare the 'degree or frequency of folding' in the cerebellum by different proxies/measurements (see Iwaniuk et al., 2006; Yopak et al., 2007; Lisney et al., 2007; Yopak et al., 2016; Cunha et al., 2022). Perhaps change the phrase in the second paragraph of the result to: "There are no comparative analyses of the frequency of cerebellar folding in mammals, to our knowledge".

We have modified the subsection in the methods referring to the measurement of folial width and folial perimeter to make the difference more clear. The folding indices that have been used previously (which we cite) are based on Zilles’s gyrification index. This index provides only a global idea of degree of folding, but it’s unable to distinguish a cortex with profuse shallow folds from one with a few deep ones. An example of this is now illustrated in Fig. 3d, where we also show how that problem is solved by the use of our two measurements (folial width and perimeter). The problem is also discussed in the section about the measurement of folding in the discussion section:

“Previous studies of cerebellar folding have relied either on a qualitative visual score (Yopak et al. 2007, Lisney et al. 2008) or a “gyrification index” based on the method introduced by Zilles et al. (1988, 1989) for the study of cerebral folding (Iwaniuk et al. 2006, Cunha et al. 2020, 2021). Zilles’s gyrification index is the ratio between the length of the outer contour of the cortex and the length of an idealised envelope meant to reflect the length of the cortex if it were not folded. For instance, a completely lissencephalic cortex would have a gyrification index close to 1, while a human cerebral cortex typically has a gyrification index of ~2.5 (Zilles et al. 1988). This method has certain limitations, as highlighted by various researchers (Germanaud et al. 2012, 2014, Rabiei et al. 2018, Schaer et al. 2008, Toro et al. 2008, Heuer et al. 2019). One important drawback is that the gyrification index produces the same value for contours with wide variations in folding frequency and amplitude, as illustrated in Fig. 3d. In reality, folding frequency (inverse of folding wavelength) and folding amplitude represent two distinct dimensions of folding that cannot be adequately captured by a single number confusing both dimensions. To address this issue we introduced 2 measurements of folding: folial width and folial perimeter. These measurements can be directly linked to folding frequency and amplitude, and are comparable to the folding depth and folding wavelength we introduced previously for cerebral 3D meshes (Heuer et al. 2019). By using these measurements, we can differentiate folding patterns that could be confused when using a single value such as the gyrification index (Fig. 3d). Additionally, these two dimensions of folding are important, because they can be related to the predictions made by biomechanical models of cortical folding, as we will discuss now.”

5) Sultan and Braitenberg (1993) measured cerebella that were sagittally sectioned (instead of coronal), right? Do you think this difference in the plane of the section could be one of the reasons explaining different results on folial width between studies? Why does the foliation index calculated by Sultan and Braitenberg (1993) not provide information about folding frequency?

The measurement of foliation should be similar as far as enough folds are sectioned perpendicular to their main axis. This will be the case for folds in the medial cerebellum (vermis) sectioned sagittally, and for folds in the lateral cerebellum sectioned coronally. The foliation index of Sultan and Braitenberg does not provide a similar account of folding frequency as we do because they only measure groups of folia (what some called lamellae), whereas we measure individual folia. It is not easy to understand exactly how Sultan and Braitenberg proceeded from their paper. We contacted Prof. Fahad Sultan (we acknowledge his help in our manuscript). Author response image 1 provides a more clear description of their procedure:

Author response image 1.

As Author response image 1 shows, each of the structures that they call a fold is composed of several folia, and so their measurements are not comparable with ours which measure individual folia (a). The flattened representation (b) is made by stacking the lengths of the fold axes (dashed lines), separating them by the total length of each fold (the solid lines), which each may contain several folia.

6) Another point that needs to be clarified is the log transformation of the data. Did the authors use log-transformed data for all types of analyses done in the study? Write this information in the material and methods.

Yes, we used the log10 transformation for all our measurements. This is now mentioned in the methods section, and again in the section concerning allometry. We are including a link to all our code to facilitate exact replication of our entire method, including this transformation.

7) The discussion needs to be expanded. The focus of the paper is on the folding pattern of the cerebellum (among different mammalian species) and its relationship with the anatomy of the cerebrum. Therefore, the discussion on this topic needs to be better developed, in my opinion (especially given the interesting results of this paper). For example, with the findings of this study, what can we say about how the folding of the cerebellum is determined across mammals? The authors found that the folial width, folial perimeter, and thickness of the molecular layer increase at a relatively slow rate across the species studied. Does this mean that these parameters have little influence on the cerebellar folding pattern? What mostly defines the folding patterns of the cerebellum given the results? Is it the interaction between section length and area? Can the authors explain why size does not seem to be a "limiting factor" for the folding of the cerebellum (for example, even relatively small cerebella are folded)? Is that because the 'white matter' core of the cerebellum is relatively small (thus more stress on it)?

We have expanded the discussion as suggested, with subsections detailing the measuring of folding, the modelling of folding for the cerebrum and the cerebellum, and the role that cerebellar folding may play in its function. We refer to the literature on cortical folding modelling, and we discuss our results in terms of the factors that this research has highlighted as critical for folding. From the discussion subsection on models of cortical folding:

“The folding of the cerebral cortex has been the focus of intense research, both from the perspective of neurobiology (Borrell 2018, Fernández and Borrell 2023) and physics (Toro and Burnod 2005, Tallinen et al. 2014, Kroenke and Bayly 2018). Current biomechanical models suggest that cortical folding should result from a buckling instability triggered by the growth of the cortical grey matter on top of the white matter core. In such systems, the growing layer should first expand without folding, increasing the stress in the core. But this configuration is unstable, and if growth continues stress is released through cortical folding. The wavelength of folding depends on cortical thickness, and folding models such as the one by Tallinen et al. (2014) predict a neocortical folding wavelength which corresponds well with the one observed in real cortices. Tallinen et al. (2014) provided a prediction for the relationship between folding wavelength λ and the mean thickness (𝑡) of the cortical layer: λ = 2π𝑡(µ/(3µ𝑠))1/3. (...)”

From this biomechanical framework, our answers to the questions of the Reviewer would be:

• How is the folding of the cerebellum determined across mammals? By the expansion of a layer of reduced thickness on top of an elastic layer (the white matter)

• Folial width, folial perimeter, and thickness of the molecular layer increase at a relatively slow rate across the species studied. Does this mean that these parameters have little influence on the cerebellar folding pattern? On the contrary, that indicates that the shape of individual folia is stable, providing the smallest level of granularity of a folding pattern. In the extreme case where all folia had exactly the same size, a small cerebellum would have enough space to accommodate only a few folia, whereas a large cerebellum would accommodate many more.

• What mostly defines the folding patterns of the cerebellum given the results? Is it the interaction between section length and area? It’s the mostly 2D expansion of the cerebellar cortical layer and its thickness.

• Can the authors explain why size does not seem to be a "limiting factor" for the folding of the cerebellum? Because even a cerebellum of very small volume would fold if its cortex were thin enough and expanded sufficiently. That’s why the cerebellum folds even while being smaller than the cerebrum: because its cortex is much thinner.

8) One caveat or point to be raised is the fact that the authors use the median of the variables measured for the whole cerebellum (e.g., median width and median perimeter across all folia). Although the cerebellum is highly uniform in its gross internal morphology and circuitry's organization across most vertebrates, there is evidence showing that the cerebellum may be organized in different functional modules. In that way, different regions or folia of the cerebellum would have different olivo-cortico-nuclear circuitries, forming, each one, a single cerebellar zone. Although it is not completely clear how these modules/zones are organized within the cerebellum, I think the authors could acknowledge this at the end of their discussion, and raise potential ideas for future studies (e.g., analyse folding of the cerebellum within the brain structure - vermis vs lateral cerebellum, for example). I think this would be a good way to emphasize the importance of the results of this study and what are the main questions remaining to be answered. For example, the expansion of the lateral cerebellum in mammals is suggested to be linked with the evolution of vocal learning in different clades (see Smaers et al., 2018). An interesting question would be to understand how foliation within the lateral cerebellum varies across mammalian clades and whether this has something to do with the cellular composition or any other aspect of the microanatomy as well as the evolution of different cognitive skills in mammals.

We now address this point in a subsection of the discussion which details the implications of our methodological decisions and the limitations of our approach. It is true that the cerebellum is regionally variable. Our measurements of folial width, folial perimeter and molecular layer thickness are local, and we should be able to use them in the future to study regional variation. However, this comes with a number of difficulties. First, it would require sampling all the cerebellum (and the cerebrum) and not just one section. But even if that were possible that would increase the number of phenotypes, beyond the current scope of this study. Our central question about brain folding in the cerebellum compared to the cerebrum is addressed by providing data for a substantial number of mammalian species. As indicated by Reviewer #3, adding more variables makes phylogenetic comparative analyses very difficult because the models to fit become too large.

Reviewer #2 (Public Review):

1) The methods section does not address all the numerical methods used to make sense of the different brain metrics.

We now provide more detailed descriptions of our measurements of foliation, phylogenetic models, analysis of partial correlations, phylogenetic principal components, and allometry. We have added illustrations (to Figs. 3 and 5), examples and references to the relevant literature.

2) In the results section, it sometimes makes it difficult for the reader to understand the reason for a sub-analysis and the interpretation of the numerical findings.

The revised version of our manuscript includes motivations for the different types of analyses, and we have also added a paragraph providing a guide to the structure of our results.

3) The originality of the article is not sufficiently brought forward:

a) the novel method to detect the depth of the molecular layer is not contextualized in order to understand the shortcomings of previously-established methods. This prevents the reader from understanding its added value and hinders its potential re-use in further studies.

The revised version of the manuscript provides additional context which highlights the novelty of our approach, in particular concerning the measurement of folding and the use of phylogenetic comparative models. The limitations of the previous approaches are stated more clearly, and illustrated in Figs. 3 and 5.

b) The numerous results reported are not sufficiently addressed in the discussion for the reader to get a full grasp of their implications, hindering the clarity of the overall conclusion of the article.

Following the Reviewer’s advice, we have thoroughly restructured our results and discussion section.

Reviewer #3 (Public Review):

1) The first problem relates to their use of the Ornstein-Uhlenbeck (OU) model: they try fitting three evolutionary models, and conclude that the Ornstein-Uhlenbeck model provides the best fit. However, it has been known for a while that OU models are prone to bias and that the apparent superiority of OU models over Brownian Motion is often an artefact, a problem that increases with smaller sample sizes. (Cooper et al (2016) Biological Journal of the Linnean Society, 2016, 118, 64-77).

Cooper et al.’s (2016) article “A Cautionary Note on the Use of Ornstein Uhlenbeck Models in Macroevolutionary Studies” suggests that comparing evolutionary models using the model’s likelihood leads often to incorrectly selecting OU over BM even for data generated from a BM process. However, Grabowski et al (2023) in their article ‘A Cautionary Note on “A Cautionary Note on the Use of Ornstein Uhlenbeck Models in Macroevolutionary Studies”’ suggest that Cooper et al.’s (2016) claim may be misleading. The work of Clavel et al. (2019) and Clavel and Morlon (2017) shows that the penalised framework implemented in mvMORPH can successfully recover the parameters of a multivariate OU process. To address more directly the concern of the Reviewer, we used simulations to evaluate the chances that we would decide for an OU model when the correct model was BM – a similar procedure to the one used by Cooper et al.’s (2016). However, instead of using the likelihood of the fitted models directly as Cooper et al. (2016) – which does not control for the number of parameters in the model – we used the Akaike Information Criterion, corrected for small sample sizes: AICc. The standard Akaike Information Criterion takes the number of parameters of the model into account, but this is not sufficient when the sample size is small. AICc provides a score which takes both aspects into account: model complexity and sample size. This information has been added to the manuscript:

“We selected the best fitting model using the Akaike Information Criterion (AIC), corrected for 𝐴𝐼𝐶 = − 2 𝑙𝑜𝑔(𝑙𝑖𝑘𝑒𝑙𝑖ℎ𝑜𝑜𝑑) + 2 𝑝. This approximation is insufficient when the𝑝 sample size small sample sizes (AICc). AIC takes into account the number of parameters in the model: is small, in which case an additional correction is required, leading to the corrected AIC: 𝐴𝐼𝐶𝑐 = 𝐴𝐼𝐶 + (2𝑝2 + 2𝑝)/(𝑛 − 𝑝 − 1), where 𝑛 is the sample size.”

In 1000 simulations of 9 correlated multivariate traits for 56 species (i.e., 56*9 data points) using our phylogenetic tree, only 0.7% of the times we would decide for OU when the real model was BM.

2) Second, for the partial correlations (e.g. fig 7) and Principal Components (fig 8) there is a concern about over-fitting: there are 9 variables and only 56 data points (violating the minimal rule of thumb that there should be >10 observations per parameter). Added to this, the inclusion of variables lacks a clear theoretical rationale. The high correlations between most variables will be in part because they are to some extent measuring the same things, e.g. the five different measures of cerebellar anatomy which include two measures of folial size. This makes it difficult to separate their effects. I get that the authors are trying to tease apart different aspects of size, but in practice, I think these results (e.g. the presence of negative coefficients in Fig 7) are really hard or impossible to interpret. The partial correlation network looks like a "correlational salad" rather than a theoretically motivated hypothesis test. It isn't clear to me that the PC analyses solve this problem, but it partly depends on the aims of these analyses, which are not made very clear.

PCA is simply a rigid rotation of the data, distances among multivariate data points are all conserved. Neither our PCA nor our partial correlation analysis involve model fitting, the concept of overfitting does not apply. PCA and partial correlations are also not used here for hypothesis testing, but as exploratory methods which provide a transformation of the data aiming at capturing the main trends of multivariate change. The aim of our analysis of correlation structure is precisely to avoid the “correlational salad” that the Reviewer mentions. The Reviewer is correct: all our variables are correlated to a varying degree (note that there are 56 data points per variable = 56*9 data points, not just 56 data points). Partial correlations and PCA aim at providing a principled way in which correlated measurements can be explored. In the revised version of the manuscript we include a more detailed description of partial correlations and PCA (phylogenetic). Whenever variables measure the same thing, they will be combined into the same principal component (these are the combinations shown in Fig. 8 b and d). Additionally, two variables may be correlated because of their correlation with a third variable (or more). Partial correlations address this possibility by looking at the correlations between the residuals of each pair of variables after all other variables have been covaried out. We provide a simple example which should make this clear, providing in particular an intuition for the meaning of negative correlations:

“All our phenotypes were strongly correlated. We used partial correlations to better understand pairwise relationships. The partial correlation between 2 vectors of measurements a and b is the correlation between their residuals after the influence of all other measurements has been covaried out. Even if the correlation between a and b is strong and positive, their partial correlation could be 0 or even negative. Consider, for example, 3 vectors of measurements a, b, c, which result from the combination of uncorrelated random vectors x, y, z. Suppose that a = 0.5 x + 0.2 y + 0.1 z, b = 0.5 x - 0.2 y + 0.1 z, and c = x. The measurements a and b will be positively correlated because of the effect of x and z. However, if we compute the residuals of a and b after covarying the effect of c (i.e., x), their partial correlation will be negative because of the opposite effect of y on a and b. The statistical significance of each partial correlation being different than 0 was estimated using the edge exclusion test introduced by Whittaker (1990).”

The rationale for our analyses has been made more clear in the revised version of the manuscript, aided by the more detailed description of our methods. In particular, we describe better the reason for our 2 measurements of folial shape – width and perimeter – which measure independent dimensions of folding (this is illustrated in Fig. 3d).

3) The claim of concerted evolution between cortical and cerebellar values (P 11-12) seems to be based on analyses that exclude body size and brain size. It, therefore, seems possible - or even likely - that all these analyses reveal overall size effects that similarly influence the cortex and cerebellum. When the authors state that they performed a second PC analysis with body and brain size removed "to better understand the patterns of neuroanatomical evolution" it isn't clear to me that is what this achieves. A test would be a model something like [cerebellar measure ~ cortical measure + rest of the brain measure], and this would deal with the problem of 'correlation salad' noted below.

The answer to this question is in the partial correlation diagram in Fig. 7c. This analysis does not exclude body weight nor brain weight. It shows that the strong correlation between cerebellar area and length is supported by a strong positive partial correlation, as is the link between cerebral area and length. There is a significant positive partial correlation between cerebellar section area and cerebral section length. That is, even after covarying everything else, there is still a correlation between cerebellar section area and cerebral section length (this partial correlation is equivalent to the suggestion of the Reviewer). Additionally, there is a positive partial correlation between body weight and cerebellar section area, but not significant partial correlation between body weight and cerebral section area or length. Our approach aims at obtaining a general view of all the relationships in the data. Testing an individual model would certainly decrease the number of correlations, however, it would provide only a partial view of the problem.

4) It is not quite clear from fig 6a that the result does indeed support isometry between the data sets (predicted 2/3 slope), and no coefficient confidence intervals are provided.

We have now added the numerical values of the CIs to all our plots in addition to the graphical representations (grey regions) in the previous version of the manuscript. The isometry slope (0.67) is either within the CIs (both for the linear and orthogonal regressions) or at the margin, indicating that if the relationships are not isometric, they are very close to it.

Referencing/discussion/attribution of previous findings

5) With respect to the discussion of the relationship between cerebellar architecture and function, and given the emphasis here on correlated evolution with cortex, Ramnani's excellent review paper goes into the issues in considerable detail, which may also help the authors develop their own discussion: Ramnani (2006) The primate cortico-cerebellar system: anatomy and function. Nature Reviews Neuroscience 7, 511-522 (2006)

We have added references to the work of Ramnani.

6) The result that humans are outliers with a more folded cerebellum than expected is interesting and adds to recent findings highlighting evolutionary changes in the hominin human cerebellum, cerebellar genes, and epigenetics. Whilst Sereno et al (2020) are cited, it would be good to explain that they found that the human cerebellum has 80% of the surface area of the cortex.

We have added this information to the introduction:

“In humans, the cerebellum has ~80% of the surface area of the cerebral cortex (Sereno et al. 2020), and contains ~80% of all brain neurons, although it represents only ~10% of the brain mass (Azevedo et al. 2009)”

7) It would surely also be relevant to highlight some of the molecular work here, such as Harrison & Montgomery (2017). Genetics of Cerebellar and Neocortical Expansion in Anthropoid Primates: A Comparative Approach. Brain Behav Evol. 2017;89(4):274-285. doi: 10.1159/000477432. Epub 2017 (especially since this paper looks at both cerebellar and cortical genes); also Guevara et al (2021) Comparative analysis reveals distinctive epigenetic features of the human cerebellum. PLoS Genet 17(5): e1009506. https://doi.org/10.1371/journal. pgen.1009506. Also relevant here is the complex folding anatomy of the dentate nucleus, which is the largest structure linking cerebellum to cortex: see Sultan et al (2010) The human dentate nucleus: a complex shape untangled. Neuroscience. 2010 Jun 2;167(4):965-8. doi: 10.1016/j.neuroscience.2010.03.007.

The information is certainly important, and could have provided a wider perspective on cerebellar evolution, but we would prefer to keep a focus on cerebellar anatomy and address genetics only indirectly through phylogeny.

8) The authors state that results confirm previous findings of a strong relationship between cerebellum and cortex (P 3 and p 16): the earliest reference given is Herculano-Houzel (2010), but this pattern was discovered ten years earlier (Barton & Harvey 2000 Nature 405, 1055-1058. https://doi.org/10.1038/35016580; Fig 1 in Barton 2002 Nature 415, 134-135 (2002). https://doi.org/10.1038/415134a) and elaborated by Whiting & Barton (2003) whose study explored in more detail the relationship between anatomical connections and correlated evolution within the cortico-cerebellar system (this paper is cited later, but only with reference to suggestions about the importance of functions of the cerebellum in the context of conservative structure, which is not its main point). In fact, Herculano-Houzel's analysis, whilst being the first to examine the question in terms of numbers of neurons, was inconclusive on that issue as it did not control for overall size or rest of the brain (A subsequent analysis using her data did, and confirmed the partially correlated evolution - Barton 2012, Philos Trans R Soc Lond B Biol Sci. 367:2097-107. doi: 10.1098/rstb.2012.0112.)

We apologise for this oversight, these references are now included.

Author Response

Reviewer #2 (Public Review):

Root growth is driven by cell elongation, and its local control allows roots to navigate the complex soil environment. Cell growth is driven by the relaxation of the cell wall, a process requiring a drop in pH. Auxin is a key regulator of root development that inhibits root growth. Auxin effects on proton dynamics are complex, it can promote both acidification and alkalinization of the extracellular space through different signaling modules, some only recently uncovered. Serre et al. report on using a new dye to monitor extracellular pH in the region surrounding the Arabidopsis thaliana root. Their manuscript aims to clarify the relationships between pH around the root, proton flux, auxin, cell elongation, and root growth with this tool. They show a typical zonation of pH values along the root: a more acidic domain corresponding to the transit-amplifying compartment, followed by a more alkaline one at the transition and early elongation zones and a more acidic one in the late elongation/root hair zone. This zonation is in agreement with previous reports obtained by other methods. A particularly puzzling aspect is the origin of the more alkaline domain. Serre et al. present evidence supporting the involvement of the AUX1-AFB1-CNGC14 module for the emergence of this more alkaline domain and how it can contribute to the ability of the root to navigate its environment.

Serre et al. show that the more alkaline domain in the transition zone is not directly determined by the activity or localization of the AHA proton pumps but rather by the auxin influx carrier AUX1. They show that the components of the rapid auxin response pathway, in particular, the auxin co-receptor AFB1 and the calcium channel CNGC14, contribute to the emergence of this more alkaline domain. Finally, they show that mutants in these two genes, impaired in the rapid auxin response pathway, show less efficient navigation of the root tip.

The manuscript is clear and well-written. The logic is sound, and the conclusions are supported by the data.

The new dye appears as a promising tool for monitoring the pH in the rhizosphere with advantages over the previous ones. Yet, as pointed out by the authors in the discussion, it reports on pH at the organ scale in the region around the root, not in the apoplast or the cell wall, which can eventually complexify the elaboration of a mechanistic model joining auxin, proton efflux, cell wall properties, cell elongation, and root growth. Although several of the findings confirm previous reports, the manuscript brings novelty by demonstrating the involvement of the rapid auxin response. I am overall supportive of the manuscript. Yet, several points should be addressed:

• The presentation of the more acidic and alkaline domains could be easier to visualize.

• The authors refer to acidic and alkaline domains but do not report on absolute pH values; they monitor the emission ratio of the dye. They justify why to use relative pH value in the discussion and refer there to internal controls that are not clearly defined. In my opinion, the wording should be more consistent across the text and figures and refer to more acidic and more alkaline domains rather than acidic (pH<7) and alkaline (pH>7) domains.

• The data related to the unaltered distribution of AHA using antibody staining should be backed up.

• The way the pH profile and the statistical analyses should be improved.

• The authors should test the effect of extracellular auxin perception (tmk, abp) mutants on pH zonation.

• Conclusion could be strengthened by moving several pieces of data currently in supplemental material to the main text.

We agree with the comment to the definition of ‘acidic’ and ‘alkaline’ domains; we altered the text and explained that we observe ‘relatively alkaline’ and ‘relatively acidic’ domains in comparison to the medium pH in the first part of results.

We defined the ‘internal controls’ in the text – by this we mean mock treated or wild type plants imaged together with the treated or mutant plants.

To address the role of the apoplastic auxin pathway in the root surface pH, we analyzed the tmk1, tmk4 and abp1 mutants. Surprisingly, all three mutants appear undistinguishable from the controls, showing the crucial importance of the cytoplasmic AFB1 auxin perception pathway. We have included the data as Fig.S4-1.

Author Response

Reviewer #1 (Public Review):

This paper studies color vision in anemonefish. The central conclusion of the paper is that anemonefish use signals from their UV cones to discriminate colors that would not otherwise be distinguishable; this differs from other fish in which UV cones extend the range of wavelengths of sensitivity but do not add a dimension to color vision. The work fits into a rich history of studies investigating how color vision fits into an animal's ecological niche. My primary concerns regard the microspectrophotometry data from single cones and some aspects of the presentation of the behavioral data.

Microspectrophotometry

The spectral properties of the cone types are a key issue for interpreting the results. These were measured using MSP, and fits are shown in Figure 2. The raw data shown in Fig. S1 appears more complicated than indicated in the main text. The templates miss the measurements across broad wavelength bands in each cone type. Particularly concerning is the high UV absorbance across cone types and the long-wavelength absorbance in the UV cone. It is not clear how this picture supports the relatively simple description of cone types and spectral sensitivities given in the main text and which forms the basis of the modeling.

Microspectrophotometry is an inherently noise-prone measurement technique, particularly for very small photoreceptor outer segments such as that of single cones, which are also difficult to detect as intact, isolated (nonoverlapping) cells. As such, the absorbance curve fitting and derived lambda max (λmax) values should be treated as estimates. The accuracy of these estimates is adequate for this type of study, and visual modelling results have been shown to be robust against small errors (±10 nm λmax) in photoreceptor sensitivity for multiple species [see Lind, O. & Kelber, A. (2009). Vis Res. 49(15), 1939-1947; and Bitton, PP. et al. (2017). PLOS ONE, 12: e0169810]. We consider it highly unlikely that small shifts in cone λmax from measurement error would make a meaningful difference to the colour discrimination thresholds.

It should be noted that the raw data shown in the original Supplementary Figure 1, included all scans overlain with an average absorbance curve for presentation purposes; however, the actual lambda max values for different cone types were measured and then averaged among individual scans fitted with photopigment absorbance curve templates. For clarity and transparency, we have now provided three multipaned plots (see Figure 1 – figure supplements 1-3) showing the individual pre- and post-bleach scans of absorbance spectra, fitted absorbance curve templates, and R2 values from the best visual pigment template fit.

It is worth noting that most of the cone absorbance spectra found in our study closely resemble those in λmax and quality to those measured in another anemonefish species (Amphiprion akindynos) [see Supplementary Figure 1 in Stieb S. et al. (2019). Sci Rep. 9, 16459]. These cone λmax values can also be reconciled with previous estimates on opsin λmax based on amino acid sequences and cone opsin expression in the A. ocellaris retina characterised in Mitchell LJ et al. (2021). GBE, 13: evab184.

Evidence that the unusual long-wavelength absorbance detected in a couple of the single cone (pre-bleach) measurements were not of visual pigment in origin comes from post-bleach scans, which showed their persistence (i.e., did not show a photobleaching response) and were likely instead contaminants (e.g., blood, RPE pigment). UV absorbance in some of the double cone measurements (above that expected of the prebleached beta peak from chromophore spectral absorption) can be attributed to either noise from scans as is quite typical of MSP and/or partial (accidental) bleaching from stray light sources. Although utmost care was taken to minimise contamination and unintended bleaching sometimes it is unavoidable.

We refer the Reviewer to multiple published studies for further examples of typical MSP measurements that share similar levels of noise to ours e.g., see Figure 1 in Knott B. et al. (2013). JEB, 216:4454-4461; Figure 3 in Schott, RK et al. (2015). PNAS, 113(2): 356-361; Figure 2 in Dalton BE et al. (2014). Proc R Soc B. 281; Figure 5 in Tosetto, JE et al. (2021). Brain Behav Evol. 96: 103-123.

Presentation

The results are not presented in a straightforward way - at least for this reviewer. What is missing for me is a clear link between the psychometric curves in Figure 3A and the discrimination thresholds indicated in Figure 3B and Figure 4. Figure 3A is only discussed in the text on line 289 - after Figure 4 has been introduced and discussed. It would have been very helpful for me if the psychometric curves were first introduced and described, then the relation to Figure 3B was clearly indicated (perhaps with a single psychometric curve as an example). Similarly for Figure 4 the relationship between specific psychometric curves and the threshold plotted would be quite helpful. Currently it takes a careful reading to understand why being below the dashed line in Figure 4 is important.

We have made the following changes, including the introduction of the psychometric curves earlier in the results (lines 236-249) and moved the psychometric function comparison before the mention of Figure 4. Additionally, to make the association between the plotted colour loci and psychometric curves clearer, we have added a smaller psychometric curve plot adjacent to the colour space (in Figure 3B) using red as an example which has an averaged psychometric curve overlying the individual fish curves. The figure caption (lines 250-274) explains that the plotted colour loci and given thresholds are mean values calculated from the individual fish behavioural data.

We have also added a brief reminder that the theoretical limit of colour discrimination is predicted by the RNL model as 1∆S, where in our task fish should be just able to distinguish targets from grey distractors (see lines 222-224). To clarify, the plotted values in Figure 4B are both the individual fish thresholds (points) and average threshold (black bar) per colour set. The individual threshold values are taken at a correct choice probability of 50% from fitted psychometric curves of fish behavioural performance (shown in Figure 3A).

RNL model

The data is fit and interpreted in the context of the receptor noise limited model. The paragraph in the discussion about complementary color pairs suggests that this model is incorrect (text around line 332). Consideration of how the results depend on the RNL model is important, especially given the interpretation here.

The inability of the RNL model to account for the observed asymmetry between color discrimination thresholds implies that they cannot be solely attributed to photoreceptor noise. We can therefore infer from the asymmetry that thresholds are set by a higher-level process, whether that involves post-receptor processes within the inner retina or in the brain remains to be investigated. As explained in lines 396-397 one possibility is that activation of the UV receptor suppresses noise in the visual pathway or enhances the saliency of colors for anemonefish. The high sensitivity to violet-green, which was found in all six of the fish tested, is consistent with the heightened saliency of this color (lines 397-399).

Figure 3B

This is the key figure in the paper. But several issues make seeing the data in this figure difficult. First, the important part of the figure is buried near the origin and hard to see. Can you show a surface that connects the thresholds in the different chromatic directions, or otherwise highlight the regions of discriminable and not discriminable colors?

See previous comment. In short, we have taken the advice of the Reviewer and added highlighted areas around the regions of discriminable colors in Figure 3B to help visually separate them from the non-discriminable regions of colors (from grey). Additionally, we have added an inset showing an enlarged image of the area surrounding the centre of colour space.

Reviewer #2 (Public Review):

Mitchell and colleagues examined the contribution of a UV-sensitive cone photoreceptor to chromatic detection in Amphiprion ocellaris, a type of anemonefish. First, they used biophysical measurements to characterize the response properties of the retinal receptors, which come in four spectrally-distinct subtypes: UV, M1, M2, and L. They then used these spectral sensitivities to construct a 4-dimensional (tetrahedral) color space in which stimuli with known spectral power distributions can be represented according to the responses they elicit in the four cone types. A novel five-LED display was used to test the fish's ability to detect "chromatic" modulations in this color space against a background of random-intensity, "achromatic" distractors that produce roughly equal relative responses in the four cone types. A subset of stimuli, defined by their high positive UV contrast, were more readily detected than other colors that contained less UV information. A well-established model was used to link calculated receptor responses to behavioral thresholds. This framework also enabled statistical comparisons between models with varying number of cone types contributing to discrimination performance, allowing inferences to be drawn about the dimensionality of color vision in anemonefish.

The authors make a compelling case for how UV light in the anemonefish habitat is likely an important ecological source of information for guiding their behavior. The authors are to be commended for developing an elegant behavioral paradigm to assess visual performance and for incorporating a novel display device especially suited to addressing hypotheses about the role of UV light in color perception. While the data are suggestive of behavioral tetrachromacy in anemonefish, there are some aspects of the study that warrant additional consideration:

1) One challenge faced by many biological imaging systems is longitudinal chromatic aberration (LCA) - that is, the focal power of the system depends on wavelength. In general, focal power increases with decreasing wavelength, such that shorter wavelengths tend to focus in front of longer wavelengths. In the human eye, at least, this focal power changes nonlinearly with wavelength, with the steepest changes occurring in the shorter part of the visible spectrum (Atchison & Smith, 2005). In the fish eye, where the visible spectrum extends to even shorter wavelengths, it seems plausible that a considerable amount of LCA may exist, which could in turn cause UV-enriched stimuli to be more salient (relative to the distractor pixels) due to differences in perceived focus rather than due solely to differences in their respective spectral compositions. Such a mechanism has been proposed by Stubbs & Stubbs (2016) as a means for supporting "color vision" in monochromatic cephalopods (but see Gagnon et al. 2016). It would be worth discussing what is known about the dispersive properties of the crystalline lens in A. ocellaris (or similar species), and whether optical factors could produce sufficient cues in the retinal image that might explain aspects of the behavioral data presented in the current study.

This is an interesting point, and we appreciate the reviewer’s thoughtful comment regarding this topic especially as LCA increases exponentially in the UV. Although we certainly cannot disprove such a mechanism in the present study, we are highly sceptical that LCA could be used by reef fish and is involved in the heightened saliency of UV stimuli. Previous work has found that LCA is mostly corrected for in the teleost retina of both marine and freshwater species by graded, multifocal lenses that focus different wavelengths at the same depth as their maximally sensitive cone photoreceptors [e.g., for evidence in African cichlids see Kröger, R. H. H. et al. (1999). J Comp Physiol. A, 184, 361-369; Malkki, P. E. & Kröger, R. H. H. (2005). J Opt. A, 7, 691-700; and for various reef fishes see Karpestam, B. et al. (2007). J Exp Biol., 210, 16: 2923-2931]. In essence, LCA is corrected in the eyes of many teleosts by accurately tuning longitudinal spherical aberration through having a graded density lens. We draw particular attention to the latter reference which comparatively examined the optical properties of reef fish lenses, including diurnal, planktivorous damselfishes (from the same family as anemonefishes, Pomacentridae). They found that not only were the lenses of these species highly UV-transmissive (as we show in anemonefish), but all were multifocal and capable of focusing both visible (non-UV) and UV wavelengths. Considering the coastal cephalopod species examined thus far, all of them contain only one type of visual pigment which is packed in their long photoreceptor (150-450µm long outer segment) across an entire retina (Chung and Marshall 2016, Proceeding B). Theoretically, given these long photoreceptors, the LCA and the resulting differentials of focal length onto different patches of photoreceptors or different depth of the outer segment might provide cues for colour discrimination even though no behavioural evidence exists to prove this hypothesis yet. Unlike the cephalopod case, the four specific spectral cones arranged in a mosaic pattern along with their very short outer segments (5-10µm) in the anemonefish retina likely makes the LCA less effective in this retinal design.

We have added a short paragraph (Lines 400-412) discussing the possibility of an optical mechanism contributing to heightened UV saliency with a particular focus on LCA and our thoughts on why we consider it an unlikely mechanism in anemonefish.

2) The authors provide a quantitative description of anemonefish visual performance within the context of a well-developed receptor-based framework. However, it was less clear to me what inferences (if any) can be drawn from these data about the post-receptoral mechanisms that support tetrachromatic color vision in these organisms. Would specific cone-opponent processes account for instances where behavioral data diverged from predictions generated with the "receptor noise limited" model described in the text? The general reader may benefit from more discussion centered on what is known (or unknown) about the organization of cone-opponent processing in anemonefish and related species.

In short, we do not know the specific opponent interactions of anemonefish cones. The RNL model assumes all possible opponent interactions in its calculations. From our results, very little can be said about the post-receptor mechanisms involved in their putative tetrachromatic vision. We would like to avoid overreaching beyond what our data can show. A future directions section has now been added to the discussion (lines 467-497), which briefly mentions the known UV opponency in larval zebrafish and that future investigation in anemonefish should attempt to disentangle the specific opponent (chromatic) and non-opponent (achromatic) circuits in the anemonefish retina.

Reviewer #3 (Public Review):

The comments below focus mainly on ways that the data and analysis as currently present do not to this reviewer compel the conclusions the authors wish to draw. It is possible that further analysis and/or clarification in the presentation would more persuasively bolster the authors' position. It also seems possible that a presentation with more limited conclusions but clarity on exactly what has been demonstrated and where additional future work is needed would make a strong contribution to the literature.

• Fig 3A. It might be worth emphasizing a bit more explicitly that the x-axis (delta S) is the result of a model fit to the data being shown, since this then means that if RNL model fit the data perfectly, all of the thresholds would fall at deltaS = 1. They don't, so I would like to see some evaluation from the authors' experience with this model as to whether they think the deviations (looks like the delta S range is ~0.4 to ~1.6 in Figure 4B) represent important deviations of the data from the model, the non-significant ANOVA notwithstanding. For example, Figure 4B suggests that the sign of the fit deviations is driven by the sign of the UV contrast and that this is systematic, something that would not be picked up by the ANOVA. Quite a bit is made of the deviations below, but that the model doesn't fully account for the data should be brought out here I think. As the authors note elsewhere, deviations of the data from the RNL model indicate that factors other than receptor noise are at play, and reminding the reader of this here at the first point it becomes clear would be helpful.

We have now stated more explicitly in the figure caption for Figure 3A, that the delta S values presented were calculated by fitting fish behavioral data to the RNL model. To test the overall effect that the sign of the UV contrast had on the discrimination threshold, we have now included ‘contrast’ (positive or negative) as another fixed effect in the linear mixed effects model. We have now included details of this test in the results which shows the systematic effect (lines 338-340). Additionally, as suggested we now briefly introduce in the results the idea that factors other than receptor noise are causing the observed deviations in data from the RNL model.

• Line 217 ff, Figure 4, Supplemental Figure 4). If I'm understanding what the ANOVA is telling us, it is that the deviations of the data across color directions and fish (I think these are the two factors based on line 649) is that the predictions deviate significantly from the data, relative to the inter-fish variability), for the trichromatic models but not the tetrachromatic model. If that's not correct, please interpret this comment to mean that more explanation of the logic of the test would be helpful.

The interpretation of the ANOVA by the Reviewer is mostly correct. We had the variables color set and Fish ID, with threshold delta S as the dependent variable. This showed that deviations from the predicted threshold were significant relative to the inter-fish variability for the trichromatic models. Missing details describing the ANOVA have now been added to the methods (lines 789-798).

Assuming that the above is right about the nature of the test, then I don't think the fact that the tetrachromatic model has an additional parameter (noise level for the added receptor type) is being taken into account in the model comparison. That is, the trichromatic models are all subsets of the tetrachromatic model, and must necessarily fit the data worse. What we want to know is whether the tetrachromatic model is fitting better because its extra parameter is allowing it to account for measurement noise (overfitting), or whether it is really doing a better job accounting for systematic features of the data. This comparison requires some method of taking the different number of parameters into account, and I don't think the ANOVA is doing that work. If the models being compared were nested linear models, than an F-ratio test could be deployed, but even this doesn't seem like what is being done. And the RNL model is not linear in its parameters, so I don't think that would be the right model comparison test in any case.

Typical model comparison approaches would include a likelihood ratio test, AIC/BIC sorts of comparisons, or a cross-validation approach.

If the authors feel their current method does persuasively handle the model comparison, how it does so needs to be brought out more carefully in the manuscript, since one of the central conclusions of the work hinges at least in part on the appropriateness of such a statistical comparison.

Our visual model comparisons were aimed at assessing whether a trichromatic or tetrachromatic model best fit the colour discrimination data. The trichromatic and tetrachromatic models assume two and three opponency pathways, respectively. If the fish were not tetrachromatic, and instead trichromatic, then we would expect that the RNL model should better fit the data with two opponency mechanisms (rather than three). Our reason for making this assessment, is because of the possibility that not all the cones could be contributing to colour vision and could be used exclusively for achromatic tasks (e.g., luminance vision or motion detection). However, according to our finding that the data best fit the tetrachromatic model (i.e., how the behavioural discrimination thresholds more closely fitted the theoretical prediction of 1∆S), it is likely that anemonefish used all four cones for colour vision.

We have also now repeated our analysis using unweighed delta S values which are calculated using general n-dimensional models of colour vision (using the PAVO2 package). These models essentially follow the same initial steps followed by the RNL model (and many others) but omit the receptor noise correction stage. After comparing (using ANOVA, see lines 303-311) the predicted thresholds with the data in this non-RNL space, it was found that again the tetrachromatic model predictions did not deviate significantly from the data relative to individual fish performance; however, we also found that the trichromatic model without M2 cone input no longer differed from the predicted values. In this case, it seems that the extra noise parameter did contribute to the difference in fit. Whether this is a biologically meaningful comparison (as all photoreceptors contain noise) is an open question. We have added a short statement explicitly framing our interpretation of anemonefish having a 3-D colour space to being in accordance with the closeness of RNL model predictions (lines 370-371, 506-508).

• Also on the general point on conclusions drawn from the model fits, it seems important to note that rejecting a trichromatic version of the RNL model is not the same as rejecting all trichromatic models. For example, a trichromatic model that postulates limiting noise added after a set of opponent transformations will make predictions that are not nested within those of RNL trichromatic models. This point seems particularly important given the systematic failures of even the tetrachromatic version of the RNL model.

This is a good point. We have limited our conclusions to specifically address trichromatic models generated within the framework of the RNL model by adding in the conclusion section that fish psychophysical thresholds were best explained by the RNL model when all four cone types contributed to colour vision (see lines 370-371, 506-508). In this same sentence, we have also added in parentheses that “suggesting (but not proving) tetrachromacy” (line 508). We have also edited the abstract to state that our results were “…best described by a tetrachromatic model using all four cone types…”, rather than stating we have shown tetrachromacy (lines 36-37).

• More generally, attempts to decide whether some human observers exhibit tetrachromacy have taught us how hard this is to do. Two issues, beyond the above, are the following. 1) If the properties of a trichromatic visual system vary across the retina, then by imaging stimuli on different parts of the visual field an observer can in principle make tetrachromatic discriminations even though visual system is locally trichromatic at each retinal location. 2) When trying to show that there is no direction in a tetrachromatic receptor space to which the observer is blind, a lot of color directions need to be sampled. Here, 9 directions are studied. Is that enough? How would we know? The following paper may be of interest in this regard: Horiguchi, Hiroshi, Jonathan Winawer, Robert F. Dougherty, and Brian A. Wandell. "Human trichromacy revisited." Proceedings of the National Academy of Sciences 110, no. 3 (2013): E260-E269. Although I'm not suggesting that the authors conduct additional experiments to try to address these points, I do think they need to be discussed. We agree with the reviewer, that colour discriminability achieved by tetrachromatic vision could in theory be achieved by the combined effect of localised, distinct forms of trichromacy. Evidence in other fishes suggests that such multiple forms of trichromacy across the retina likely exist in many species. However, the behavioural effects of this retinal setup remain to be studied likely due to its extremely difficult nature. We have added a new section titled “future directions” (Lines 474-489), in which we discuss the possibility that distinct forms of trichromacy in the anemonefish retina could in theory achieve colour discrimination on par with tetrachromatic vision. We also give suggestions on how this could be investigated.

Although we tried to include as many colour directions as practically possible in our experiment, we have certainly not provided an exhaustive range that completely encompasses anemonefish colour space. Whether 9 colour directions are adequate to assess the dimensionality of their color vision is difficult to say. As addressed in the previous comment, we now acknowledge this limitation by refining our conclusion, saying that our results do not prove tetrachromacy.

• Line 277 ff. After reading through the paper several times, I remain unsure about what the authors regard as their compelling evidence that the UV cone has a higher sensitivity or makes an omnibus higher contribution to sensitivity than other cones (as stated in various forms in the title, Lines 37-41, 56-57, 125, 313, 352 and perhaps elsewhere).

At first, I thought they key point was that the receptor noise inferred via the RNL model as slightly lower (0.11) for the UV cone than for the double cones (0.14). And this is the argument made explicitly at line 326 of the discussion. But if this is the argument, what needs to be shown is that the data reject a tetrachromatic version of the RNL model where the noise value of all the cones is locked to be the same (or something similar), with the analysis taking into account the fewer parametric degrees of freedom where the noise parameters are so constrained. That is, a careful model comparison analysis would be needed. Such an analysis is not presented that I see, and I need more convincing that the difference between 0.11 and 0.14 is a real effect driven by the data. Also, I am not sanguine that the parameters of a model that in some systematic ways fails to fit the data should be taken as characterizing properties of the receptors themselves (as sometimes seems to be stated as the conclusion we should draw).

We have performed various modelling scenarios where receptor noise was adjusted for each channel; however, the UV channel was consistently found to be more sensitive than the other channels. In (the original) Supplementary Figure 6 (now Figure 4 – figure supplements 1 and 2), we show predicted dS values calculated using receptor noise levels in the exact manner that the Reviewer suggests by ranging from 0.05 to 0.15, and most importantly, included scenarios where receptor noise was held equal across cone types and others where it was varied between single cones and double cones. None of the models adjusted the data so that sensitivity was equal across all four channels, which means that by an unknown mechanism, the UV channel is more sensitive, but this is unrelated to noise levels. Our best-fit receptor noise values of 0.11 (for single cones) and 0.14 (for double cones) are estimate values and should be treated as such till actual receptor noise measurements are made.

Then, I thought maybe the argument is not that the noise levels differ, but rather that the failures of the model are in the direction of thresholds being under predicted for discriminations that involve UV cone signals. That's what seems to be being argued here at lines 277 ff, and then again at lines 328 ff of the discussion. But then the argument as I read it more detail in both places switches from being about the UV cones per se to being about postive versus negative UV contrast. That's fine, but it's distinct from an argument that favors omnibus enhanced UV sensitivity, since both the UV increments and decrements are conveyed by the UV cone; it's an argument for differential sensitivity for increments versus decrements in UV mediated discriminations. The authors get to this on lines 334 of the discussion, but if the point is an increment/decrement asymmetry the title and many of the terser earlier assertions should be reworked to be consistent with what is shown.

To clarify our argument, we found that the colour discrimination thresholds were systematically lower than predicted by the RNL model for colours which elicited higher UV cone stimulation relative to other cone types. These colours we refer to as UV positive based on the sign direction of their contrast against grey distractors produced by higher UV/V LED channel (i.e., in a positive direction). Whereas colours with UV negative chromatic contrast had lower UV cone stimulation relative to the other cone types. Therefore, our interpretation of the importance of UV cone signals for colour discrimination are congruent with the results. In the discussion, we suggest a possibility that activation of the UV receptor suppresses noise downstream in the visual pathway or enhances the saliency of colours (see lines 397-398). This activation of the UV receptor would, of course, be at its highest for colours with positive UV chromatic contrast.

Note that we have added to the discussion the possibility that colour preferences or a difference in attentiveness might have contributed to differences in discrimination thresholds (see discussion lines 412-413, 427-428, 433-435, 456-466, and 469-473). However, we consider it a less likely explanation due to a couple of reasons, including 1) a lack of difference in responsiveness across colour sets in their timing to peck the target, and 2) any non-learnt bias would have likely been overridden or at least weakened by training prior to the experiment where colours were rewarded equally (see lines 462-466).

We have edited the results (lines 334-352) to make our point clearer and by changing the subtitle to be more explicit: “Lower discrimination thresholds induced by positive UV contrast”. The subsection begins by explaining the different types of UV chromatic contrast by elevation angle and, finally, how this division among colour sets was a major determinant of colour discrimination thresholds.

Perhaps the argument with respect to model deviations and UV contrast independent of sign could be elaborated to show more systematically that the way the covariation with the contrasts of the other cone stimulations in the stimulus set goes, the data do favor deviations from the RNL in the direction of enhanced sensitivity to UV cone signals, but if this is the intent I think the authors need to think more about how to present the data in a manner that makes it more compelling than currently, and walk the reader carefully through the argument.

We have added to the results the linear mixed-effects model output with ‘contrast’ (positive/negative) added as a fixed effect. This analysis shows that the sign direction of UV contrast was a strong predictor of threshold (see address to previous comments and lines 399-401, 790-799).

• On this point, if the authors decide to stick with the enhanced UV sensitivity argument in the revision, a bit more care about what is meant by "the UV cone has a comparatively high sensitivity (line 313 and throughout)" needs more unpacking. If it is that these cones have lower inferred noise (in the context of a model that doesn't account for at least some aspects of the data), is this because of properties of the UV cones, or the way that post-receptoral processing handles the signals from these cones mimicking a cone effect in the model. And if it is thought that it is because of properties of the cones, some discussion of what those properties might be would be helpful. As I understand the RNL model, relative numbers of cones of each type are taken into account, so it isn't that. But could it be something as simple as higher photopigment density or larger entrance aperture (thus more quantum catches and higher SNR)?

It is unknown what aspect of the cone morphology or physiology sets the activation or inactivation threshold. Electrophysiological data collected from the UV cones of other fish species e.g., in goldfish and zebrafish [see Hawryshyn & Beauchamp (1985). 25, Vis Res.; and Yoshimatsu et al. (2020). 107, Neuron.] show that they have exceptionally high sensitivity. What has not been shown is that having a UV cone can improve colour discrimination.

Previous quantitative cone opsin gene expression analysis showed that the single cone opsins (SWS1 and SWS2B) are expressed at lower levels than all double cone opsin genes. This difference in expression combined with the smaller size of single cone outer segments than the double cones make it unlikely that a larger photoreceptor size, higher volume or packing density of visual pigment is responsible. Contrary to our findings, these aspects of the different cone types (if they had an effect) would instead predict that double cones have a higher SNR, and non-UV colours would be more discriminable. We have now added these details to the discussion (see lines 391-397).

• Line 288 ff. The fact that the slopes of the psychometric functions differed across color directions is, I think, a failure of the RNL model to describe this aspect of the data, and tells us that a simple summary of what happens for thresholds at delta S = 1 does not generalize across color directions for other performance levels. Since one of the directions where the slope is shallower is the UV direction, this fact would seem to place serious limits on the claim that discrimination in the UV direction is enhanced relative to other directions, but it goes by here without comment along those lines. Some comment here, both about implications for fit of RNL model and about implications for generalizations about efficacy of UV receptor mediated discrimination and UV increment/decrement asymmetries, seems important.

The variation in the psychometric functions is difficult to interpret and cannot be explained by the RNL model. What the RNL model predicts is delta S based on low level factors (namely receptor noise). In the discussion, we completely agree with the notion that the asymmetry in thresholds from predicted values, and the variation in psychometric slopes cannot be explained by the RNL model, e.g., this is heavily implied by “colour discrimination thresholds cannot be directly attributed to noise in the early stages of the visual pathway…” (lines 388-390). To clarify the inability of the RNL model to account for this aspect of the data, we have included a statement (see line 390).

It is a good point that this could be an indication of heterogeneity in colour space. Heterogeneity in discrimination thresholds across animal colour space (both surrounding the threshold area and for more saturated regions) has been explored in detail using trichromatic triggerfish by Green N. F. et al. (2022). JEB, 7(225):jeb243533. We have added this idea to the discussion (see lines 490-498). For UV, it seems that two of the five fish (#34 and 20) had noticeably shallower curves than the others tested for UV (fish #19, 33, 36). Both also varied more in their ability to distinguish targets, as shown by their wider confidence intervals. One of these two fish (#34) was retested for UV at the end of the experiment, and in the secondary assessment had a steeper psychometric curve more in line with the other fish in the experiment (see Figure 3 – figure supplement 1 and added lines 247-250). Based on this discrepancy in performance between assessments, it is also possible that individual learning effects had a role in impacting the shape of the psychometric curve. Note, this had minimal effect on colour discrimination thresholds and any differences were in the direction of change observed across colour sets in the experiment (i.e., lower dS for UV positive directions).

• Line 357 ff. Up until this point, all of the discussion of differences in threshold across stimulus sets has been in terms of sensitivity. Here the authors (correctly) raise the possibility that a difference in "preference" across stimulus sets could drive the difference in thresholds as measured. Although the discussion is interesting and germaine, it does to some extent further undercut the security of conclusions about differential sensitivity across color directions relative to the RNL model predictions, and that should be brought out for the reader here. The authors might also discuss about how a future experiment might differentiate between a preference explanation and a sensitivity explanation of threshold differences.

We have now added a paragraph (see lines 469-473) discussing that future work should test for color preferences and suggest how this could be done using a similar foraging task. We also include our thoughts immediately prior on why it is unlikely that a colour preference was a major contribution towards the results. In short, we consider it unlikely as fish showed no evidence of reduced latency for pecking at targets across the colour sets and because the training regime prior to the experiment equally rewarded fish for all colours and would likely have overridden a strong preference (at least in this specific foraging context).

• RNL model. The paper cites a lot of earlier work that used the RNL model, but I think many readers will not be familiar with it. A bit more descriptive prose would be helpful, and particularly noting that in the full dimensional receptor space, if the limiting noise at the photoreceptors is Gaussian, then the isothreshold contour will be a hyper-ellipsoid with its axes aligned with the receptor directions.

There is now added explanation of the RNL model (see lines 141-151), particularly on its assumptions that it only receives chromatic input and that discrimination is limited by noise arising in the photoreceptors and not by any specific opponent mechanisms. We also added the mention of the expected hyper-ellipsoid shape of isothreshold contours if receptor noise is Gaussian. Note, while we appreciate the importance of the reader to understand the basic functionality of the model, we wanted to avoid overloading the introduction with details on the RNL model which is not the focus of the paper. The RNL model is well-established in the field of visual ecology and animal vision research for well over a decade and has been thoroughly dissected by previous methodological reviews. We refer to one of these more recent reviews by Olsson et al. (2018) Behav Ecol. 29(2):273-282, and direct the reader to the methods section for further details on the RNL model.

• Use of cone isolating stimuli? For showing that all four cone classes contribute to what the authors call color discrimination, a more direct approach would seem to be to use stimuli that target stimulation of only one class of cone at a time. This might require a modified design in which the distractors and target were shown against a uniform background and approximately matched in their estimated effect on a putative achromatic mechanism. Did the authors consider this approach, and more generally could they discuss what they see as its advantages and disadvantages for future work.

The Reviewer is correct in that a targeted approach of isolated cone stimulation would be the optimal approach to demonstrating tetrachromatic colour vision. However, the extreme spectral overlap in the absorption curves of anemonefish cones, particularly in the mid-wavelength region makes this problematic in using the current LED display. We added to the discussion ways that this could be studied in the future (see lines 474-489). This might be possible (but still challenging) using a monochromator, but such technology severely limits the diversity of stimuli which can be created and usually restricts experiments to a simple paired choice design (or grey card experiment). The traditional paired choice experiment requires animals to be trained to distinguish a specific colour, while the Ishihara-like task trains animals to distinguish targets using an odd-one-out approach. This latter approach is highly efficient, as it does not require retraining when testing a new colour (i.e., fish learnt the task not a specific colour). Here, we wanted to assess colour discrimination in multiple directions to compare performance, and the flexible LED display combined with a generalisable task was important.

The above assumes that anemonefish do not use multiple trichromatic systems. In which case, the use of standard experimental stimuli (e.g., a monochromator, an LED display) would be unsuitable as they illuminate the whole retina. To definitively test the range of opponent interactions, it would be necessary to make electrophysiological measurements targeting the transmitting neurons using a retinal multielectrode array (MEA) approach or by in-vivo calcium imaging (lines 484-486).

We understand that our results are not a direct test of the dimensionality of anemonefish colour vision and should not be interpreted as such, as we do not have direct evidence of tetrachromacy. To recognize this limitation of our data, we have drawn back some of our conclusive statements that claimed to have demonstrated tetrachromacy.

Author Response

Reviewer #1 (Public Review):

Precise regulation of gamete fusion ensures that offspring will have the same ploidy as the parents. However, breaking this regulation can be useful for plant breeding. Haploid induction followed by chemical-induced genome doubling can be used to fix desirable genotypes, while triparental hybrids where two sperm cells with two different genotypes fertilize an egg cell can be advantageous for bypassing hybridization barriers to create interspecies hybrids with increased fitness. This manuscript follows up on a previous study from the same research group that used a clever high throughput polyspermy detection assay (HIPOD) to show that wild-type Arabidopsis naturally forms triparental hybrids at very low frequencies (less than 0.05% of progeny) and that these triparental hybrids can bypass dosage barriers in the endosperm (Nakel, et al., 2017). Mao and co-authors hypothesized that mutants that conferred polytubey, the attraction of multiple pollen tubes by mutant female gametophytes, would also increase the rate of triparental hybrids. They used a double mutant in the endopeptidase genes ECS1 and ECS2 which had previously been reported to induce supernumerary pollen tube attraction to test this hypothesis with their two-component HIPOD system in which one pollen donor constitutively expresses the mGAL4-VP16 transcription factor while the second pollen donor carries an herbicide resistance gene regulated by the GAL4-responsive UAS promoter. Triparental hybrids are detected as herbicide-resistant progeny from wild-type Arabidopsis flowers that have been pollinated by the two paternal genotypes. The authors convincingly show that the ecs1 ecs2-1 double mutant more than doubled the frequency of triparental, triploid hybrids in HIPOD crosses. They next tested the hypothesis that this increase in triparental hybrids was due to a gametophytic effect by using an ecs1-/- ecs2-1/ECS2 maternal parent in the HIPOD assay and testing whether the ecs2-1 mutant allele was preferentially inherited in triparental hybrids. The mutant allele was inherited at a much higher rate than expected, confirming their hypothesis.

The triparental hybrid results with the ecs1 ecs2 mutant were not that surprising since the presence of extra sperm cells gives more opportunities for triparental hybrids to form, especially if gamete fusion is misregulated. However, an unexpected result came when the authors used aniline blue staining to analyze the ecs1 ecs2 polytubey phenotype. They confirmed that the double mutant had increased levels of polytubey compared to wild-type ovules, but they also noticed that 13% of seeds were not developing normally. This phenotype was confirmed with a second ecs2 allele and was complemented with both ECS1 and ECS2 transgenes under their native promoters. Microscopic analysis revealed normal gametophyte morphology before fertilization, but 8% of pollinated ovules failed to develop an embryo and 7% failed to develop endosperm, suggesting single fertilization events. In a logical set of experiments, they followed up on this result by crossing ecs1 ecs2 with pollen carrying a fluorescent reporter that would be expressed in developing embryos and endosperm. In this experiment, they were again surprised. Some of the wild-type-looking seeds lacked a paternal contribution (i.e. no fluorescent signal from the paternal reporter construct) in the embryo. This prompted them to look more closely at the progeny, upon which they detected small plants that were haploid. They confirmed the haploid nature by chromosome spreads. Finally, they used interaccession crosses between ecs1 ecs2 (Col-0) and Landsberg to verify that haploid progeny only carried maternal alleles of markers on all five chromosomes, indicating that the ecs1 ecs2 genotype can induce maternal haploids.

This interesting study highlights the importance of following up on unexpected results. The conclusions are well-supported by the data and quite exciting. Paternal haploid inducers have been discovered in several species, but this is one of only two examples of maternal haploid induction. While the percentage of maternal haploids is very low, this phenomenon could be useful for plant breeding.

Weaknesses

The data in the manuscript is intriguing, but the question of how the same mutant combination promotes the formation of both triploid and haploid progeny remains unanswered and is not thoroughly discussed, nor is any model suggested for how the ECS1/2 peptidases could play a role in regulating gamete fusion and/or repressing parthenogenesis. A second unanswered question is whether the maternal haploids are a result of failed plasmogamy or karyogamy between the egg and sperm leading to parthenogenesis or a result of paternal genome elimination after plasmogamy. In figure 3B, the authors attempted to test whether plasmogamy occurs between the male and female gametes in ecs1 ecs2 ovules by crosses with pollen that expresses a mitochondrial marker under control of the pRPS5a promoter which is active in sperm cells as well as embryos and endosperm of fertilized ovules. This experiment allowed them to detect sperm cells that had not fused with the egg and central cell at 2 days after pollination. They also counted the percentage of seeds that expressed the mitochondrial marker in both embryo and endosperm at 2 DAP and found that ecs1 ecs2 mutants had a 20% reduction of visible mitochondria in embryo sacs compared to wildtype. They conclude that the result indicates a potential plasmogamy defect. However, the dependability of this marker is questionable since only ~55% of wild-type seeds had detectable signal in the embryo and endosperm. The authors imply that this experiment could be used to test plasmogamy, but it is not clear how any conclusions related to the abnormal seed phenotype could be drawn from examining the rate of signal in both the embryo and endosperm. Since the mitochondrial marker was not expressed from a sperm-specific promoter, the fluorescent signal at 2DAP is likely due to new gene expression from pRPS5a in the fertilized embryo and endosperm, not an indication of the presence of sperm-derived mitochondria. Perhaps an earlier timepoint could be used as well as a spermspecific promoter instead of pRPS5a to answer the question of whether plasmogamy is happening in the ecs1 ecs2 ovules.

Thanks for the suggestion. We here provide two additional new data sets to provide evidence that ecs1 ecs2 mutant plants indeed exhibit single fertilization that lead to fertilization recovery.

We determined the fertilization failure by checking the decondensation HTR10-RFP labelled sperm nuclei 8-10 HAP (Figure 3B) and the frequency of heterofertilization through dual pollination experiment (Figure 3C-E) (see above).

Reviewer #2 (Public Review):

The manuscript reports the triploid and haploid productions using an ecs1ecs2 mutant as the maternal donor, in addition to the evaluation of the sexual process observed in the mutant. The indicated data show exquisite quality. To improve the content, I recommend carefully reconsidering the descriptions because some of the insights would cause a stir in the controversy regarding ECS1&2 functions in plant reproduction.

Strengths

Triploid production by a combination of ecs1ecs2 mutant and HIPOD system has potential as a future plant breeding tool. Moreover, it's intriguing that both triploid and haploid productions were achieved using the same mutant as a maternal donor. I think authors can claim the value of their results more by adding descriptions about the usefulness of the aneuploid plants in plant breeding history.

The evidence of the persistent synergid nucleus (Figure 3A) is critical insight reported by this study. As Maruyama et al. (2013) reported by live cell imaging, synergid-endosperm fusion had occurred at the two endosperm nuclei stage. It would be valuable to claim the observed fact by citing Maruyama's previous observation.

Weakness

As the authors suggested, the higher triploid frequency observed in ecs1ecs2 than WT was likely caused by the increased polyspermy. However, it also could be that reduction of normal seed number in ecs1ecs2 (whichever is due to failure of fertilization or embryo development arrest) accounts for the increased frequency of the triploid compared to WT.

The results in Figure 3C-E suggested the single fertilization for both egg and central cells at similar frequencies. This is an exciting result, but it is still possible that the fertilized egg or central cell degenerated after fertilization resulting in the disappearance of paternally inherited fluorescence. Evaluation of fertilization patterns at 7-10HAP in ecs1ecs2 mutant may provide more confident insight, although unfused sperm cell was evaluated at 1DAP (Figure 3-figure supplement 1B). The fertilization states can be distinguished depending on the HTR10RFP sperm nuclei morphology and their positions, as reported by Takahashi et al (2018).

Thank you for your suggestion. We added the requested experiment see Figure 3B in the revised manuscript. In addition, we conducted a dual pollination experiment, that provides evidence for the activation of the fertilization recovery machinery (Figure 3C-E) (see above).

Several recent studies have reported exciting insights on ECS1&2 functions; however, various results from different laboratories have raised controversy. Though, the commonly found feature is the repression of polytubey. For readers, it would be helpful to organize the explanation about which insights are concordant or different.

Thank you for your suggestion. We now indicate using terms like in line with or in contrast to, where our data confirms /or contradicts with previous data.

In addition, a drawing that explains the time course in the process from pollination to seed development (up to 6DAP) based on WT would help to understand which point is evaluated in each data.

Thank you for your suggestion. We added a model figure (Figure 4E) at the end of the manuscript that brings the concepts together and facilitates the understandings.

Reviewer #3 (Public Review):

In this manuscript, Mao et al. reported that the two proteases ECS1 and ECS2 participate in both polyspermy block and gamete fusion in Arabidopsis thaliana. The authors could observe polytubey phenotype which has been reported previously and obtain both triparental plants and haploids in ecs1 ecs2 mutants. Therefore, they proposed that the triparental plants resulted from the polytubey block defect, whereas the haploids were caused by the gamete fusion defect. Together with two other previous reports, I think it is very interesting to see these two proteases participating in so many different but connected processes. Although they did not provide the molecular mechanism of how ECS participated in polyspermy block and gamete fusion, their findings provide more options for and thus promote plant breeding. The work may have a wide application in the future and will be of broad interest to cell biologists working on gamete fusion and plant breeders.

We thank the reviewer for their positive comments.

Although most of the conclusions in this paper are well supported by the data, it could be improved with a minor revision including providing clearer data analysis and descriptions, images with higher resolution, and more discussions.

Author Response

Reviewer #2 (Public Review):

In the discussion, the authors suggest that the binding of CHAPS could be an inspiration to develop compounds, targeting, for instance, mammalian receptors, that would bind to both the orthosteric site and a potential groove underneath loop C (where the sterol moiety of CHAPS binds in Alpo4). A figure (SI4) shows a few homologues in surface representation, giving an idea of whether this groove is generally present in the family.

Seeing this figure, I wondered if it would be relevant to compare several conformations of one or a few chosen homologues. Given that gating always impacts the quaternary assembly, is this groove more pronounced in say the inhibited state of a given homologue than in its agonist-bound state?

The width of the groove in 7 does change as the channel transition from apo to open state. This is now demonstrated with an additional Figure 3 – figure supplement 1b and the discussion was adjusted accordingly p 18, line 379:

“The sterol group connected by a linker binds in between subunits and induces conformational changes which also change the width of the groove in Alpo4 (Figure 3f, g), therefore it likely plays an active role in the observed quaternary twist. The changes in the groove shape are not specific to Alpo4 but are also observed for example in nicotinic 7 receptor (Figure 3 – supplement 1b) suggesting that the groove can be targeted for allosteric modulation of the channel. ”

A related thought was that some of the protein binders affecting pLGIC function (toxins, VHH) contact two subunits and wrap around/below loop C. Do these have binding sites that overlap with the groove?

We inspected the structures of pLGICs homologs with bound -bungarotoxin (6UWZ, 4HQP, 7Z14, and 7KOO) and 2 with bound VHHs (6SSI and 6HJY). The toxins were bound in similar conformations but not the VHHs. The examples of the complexes are now shown as Supplementary Figure 13a (see above). In the case of ELIC, the nanobody Nb72 was bound on top of the sterol-binding cavity, but it did not interact with the interior of the cavity. This is now explained on p 17 from line 374:

“When binding sites of larger know binders, including VHH47,48 and -bungarotoxin10,49 were examined (Figure 3 – supplement 1a) a nanobody bound to ELIC in the site covering the sterol-binding groove was identified, however, its interactions with ELIC did not overlap significantly with the interior of the sterol-binding groove. This suggests that the latter is a novel target location for binders.”

Very interestingly, the binding of CHAPS stabilizes a conformation that differs from the apo one. It includes a twist of the ECDs but does not lead to a significant opening of the M2 bundle. The authors note that the direction of the twist is reversed to that often associated with the binding of ligands in homologues. This reversion is quite a feature, which deserves to be shown in a supplementary movie (e.g overlay of the Alpo apo>CHAPs transition with the nico>apo transition of a7).

We have re-examined the rotation and compared it to the conformational changes in nACh 7 and 5-HT3 receptors. Upon closer examination, it became clear that relative rotation of the ECD and the TMD provides a very simplistic view of the quaternary conformational changes which are more complex 3D quaternary changes than a simple relative domain rotation. Careful alignment of the structures to the extracellular side of the trans-membrane pore showed that in both channels resting-> open state transition is associated with clockwise rotation, but resting-> desensitized state transition in 5-HT3 involves a counterclockwise rotation. Thus, 1) the direction of rotation is not a ‘universal’ feature of pLGICs and 2) the clockwise rotation is the direction of channel activation for α7 nACh receptor and 5-HT3 and shares similarities with rearrangements observed in Alpo4. However, the relative movement of the ECDs is different between Alpo upon CHAPS binding and α7 nACh and 5-HT3 receptor upon activation. To demonstrate this, we added Video 2 which shows quaternary changes for all 3 channels and the text has been modified as follows on page 11 line 208:

“Quaternary changes in Alpo4 induced upon CHAPS binding and those associated with the activation of related α7 nACh and 5-HT3 receptors induced rotation of ECD relative to TMD in the same direction, however, the shifts of principal relative to complementary subunits were different (Video 2). In Alpo4, the complementary subunit slides upward whereas in the two other channels it consistently shifts towards the principal subunit and tilts relative to the TMD. The tilt is less pronounced in Alpo4 which is probably why it does not lead to the pore dilation.”

We are grateful to the reviewer for drawing our attention to this point, which permitted us to correct initially inaccurate statements.

Author Response

Reviewer #2 (Public Review):

Here, a simple model of cerebellar computation is used to study the dependence of task performance on input type: it is demonstrated that task performance and optimal representations are highly dependent on task and stimulus type. This challenges many standard models which use simple random stimuli and concludes that the granular layer is required to provide a sparse representation. This is a useful contribution to our understanding of cerebellar circuits, though, in common with many models of this type, the neural dynamics and circuit architecture are not very specific to the cerebellum, the model includes the feedforward structure and the high dimension of the granule layer, but little else. This paper has the virtue of including tasks that are more realistic, but by the paper’s own admission, the same model can be applied to the electrosensory lateral line lobe and it could, though it is not mentioned in the paper, be applied to the dentate gyrus and large pyramidal cells of CA3. The discussion does not include specific elements related to, for example, the dynamics of the Purkinje cells or the role of Golgi cells, and, in a way, the demonstration that the model can encompass different tasks and stimuli types is an indication of how abstract the model is. Nonetheless, it is useful and interesting to see a generalization of what has become a standard paradigm for discussing cerebellar function.

We appreciate the Reviewer’s positive comments. Regarding the simplifications of our model, we agree that we have taken a modeling approach that abstracts away certain details to permit comparisons across systems. We now include an in-depth discussion of our simplifying assumptions (Assumptions & Extensions section in the Discussion) and have further noted the possibility that other biophysical mechanisms we have not accounted for may also underlie differences across systems.

Our results predict that qualitative differences in the coding levels of cerebellum-like systems, across brain regions or across species, reflect an optimization to distinct tasks (Figure 7). However, it is also possible that differences in coding level arise from other physiological differences between systems.

Reviewer #3 (Public Review):

1) The paper by Xie et al is a modelling study of the mossy fiber-to-granule cell-to-Purkinje cell network, reporting that the optimal type of representations in the cerebellar granule cell layer depends on the type task. The paper stresses that the findings indicate a higher overall bias towards dense representations than stated in the literature, but it appears the authors have missed parts of the literature that already reported on this. While the modelling and analysis appear mathematically solid, the model is lacking many known constraints of the cerebellar circuitry, which makes the applicability of the findings to the biological counterpart somewhat limited.

We thank the Reviewer for suggesting additional references to include in our manuscript, and for encouraging us to extend our model toward greater biological plausibility and more critically discuss simplifying assumptions we have made. We respond to both the comment about previous literature and about applicability to cerebellar circuitry in detail below.

2) I have some concerns with the novelty of the main conclusion, here from the abstract: ’Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classic theories.’ Stated like this, this has in principle already been shown, i.e. for example: Spanne and Jo¨rntell (2013) Processing of multi-dimensional sensorimotor information in the spinal and cerebellar neuronal circuitry: a new hypothesis. PLoS Comput Biol. 9(3):e1002979. Indeed, even the 2 DoF arm movement control that is used in the present paper as an application, was used in this previous paper, with similar conclusions with respect to the advantage of continuous input-output transformations and dense coding. Thus, already from the beginning of this paper, the novelty aspect of this paper is questionable. Even the conclusion in the last paragraph of the Introduction: ‘We show that, when learning input-output mappings for motor control tasks, the optimal granule cell representation is much denser than predicted by previous analyses.’ was in principle already shown by this previous paper.

We thank the Reviewer for drawing our attention to Spanne and Jo¨rntell (2013). Our study shares certain similarities with this work, including the consideration of tasks with smooth input-output mappings, such as learning the dynamics of a two-joint arm. However, our study differs substantially, most notably the fact that we focus our study on parametrically varying the degree of sparsity in the granule cell layer to determine the circumstances under which dense versus sparse coding is optimal. To the best of our ability, we can find no result in Spanne and J¨orntell (2013) that indicates the performance of a network as a function of average coding level. Instead, Spanne and Jo¨rntell (2013) propose that inhibition from Golgi cells produces heterogeneity in coding level which can improve performance, which is an interesting but complementary finding to ours. We therefore do not believe that the quantitative computations of optimal coding level that we present are redundant with the results of this previous study. We also note that a key contribution of our study is mathemetical analysis of the inductive bias of networks with different coding levels which supports our conclusions.

We have included a discussion of Spanne and Jo¨rntell (2013) and (2015) in the revised version of our manuscript:

"Other studies have considered tasks with smooth input-output mappings and low-dimensional inputs, finding that heterogeneous Golgi cell inhibition can improve performance by diversifying individual granule cell thresholds (Spanne and J¨orntell, 2013). Extending our model to include heterogeneous thresholds is an interesting direction for future work. Another proposal states that dense coding may improve generalization (Spanne and Jo¨rntell, 2015). Our theory reveals that whether or not dense coding is beneficial depends on the task."

3) However, the present paper does add several more specific investigations/characterizations that were not previously explored. Many of the main figures report interesting new model results. However, the model is implemented in a highly generic fashion. Consequently, the model relates better to general neural network theory than to specific interpretations of the function of the cerebellar neuronal circuitry. One good example is the findings reported in Figure 2. These represent an interesting extension to the main conclusion, but they are also partly based on arbitrariness as the type of mossy fiber input described in the random categorization task has not been observed in the mammalian cerebellum under behavior in vivo, whereas in contrast, the type of input for the motor control task does resemble mossy fiber input recorded under behavior (van Kan et al 1993).

We agree that the tasks we consider in Figure 2 are simplified compared to those that we consider elsewhere in the paper. The choice of random mossy fiber input was made to provide a comparison to previous modeling studies that also use random input as a benchmark (Marr 1969, Albus 1971, Brunel 2004, Babadi and Sompolinsky 2014, Billings 2014, LitwinKumar et al., 2017). This baseline permits us to specifically evaluate the effects of lowdimensional inputs (Figure 2) and richer input-output mappings (Figure 2, Figure 7). We agree with the Reviewer that the random and uncorrelated mossy fiber activity that has been extensively used in previous studies is almost certainly an unrealistic idealization of in vivo neural activity—this is a motivating factor for our study, which relaxes this assumption and examines the consequences. To provide additional context, we have updated the following paragraph in the main text Results section:

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a lowdimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks."

4) The overall conclusion states: ‘Our results....suggest that optimal cerebellar representations are task-dependent.’ This is not a particularly strong or specific conclusion. One could interpret this statement as simply saying: ‘if I construct an arbitrary neural network, with arbitrary intrinsic properties in neurons and synapses, I can get outputs that depend on the intensity of the input that I provide to that network.’ Further, the last sentence of the Introduction states: ‘More broadly, we show that the sparsity of a neural code has a task-dependent influence on learning...’ This is very general and unspecific, and would likely not come as a surprise to anyone interested in the analysis of neural networks. It doesn’t pinpoint any specific biological problem but just says that if I change the density of the input to a [generic] network, then the learning will be impacted in one way or another.

We agree with the Reviewer that our conclusions are quite general, and we have removed the final sentence as we agree it was unspecific. However, we disagree with the Reviewer’s paraphrasing of our results.

First, we do not select arbitrary intrinsic properties of neurons and synapses. Rather, we construct a simplified model with a key quantity, the neuronal threshold, that we vary parametrically in order to assess the effect of the resulting changes in the representation on performance. Second, we do not vary the intensity/density of inputs provided to the network – this is fixed throughout our study for all key comparisons we perform. Instead, we vary the density (coding level) of the expansion layer representation and quantify its effect on inductive bias and generalization. Finally, our study’s key contribution is an explanation of the heterogeneity in average coding level observed across behaviors and cerebellum-like systems. We go beyond the empirical statement that there is a dependence of performance on the parameter that we vary by developing an analytical theory. Our theory describes the performance of the class of networks that we study and the properties of learning tasks that determine the optimal expansion layer representation.

To clarify our main contributions, we have updated the final paragraph of the Introduction. We have also removed the sentence that the Reviewer objects to, as it was less specific than the other points we make here.

"We propose that these differences can be explained by the capacity of representations with different levels of sparsity to support learning of different tasks. We show that the optimal level of sparsity depends on the structure of the input-output relationship of a task. When learning input-output mappings for motor control tasks, the optimal granule cell representation is much denser than predicted by previous analyses. To explain this result, we develop an analytic theory that predicts the performance of cerebellum-like circuits for arbitrary learning tasks. The theory describes how properties of cerebellar architecture and activity control these networks’ inductive bias: the tendency of a network toward learning particular types of input-output mappings (Sollich, 1998; Jacot et al., 2018; Bordelon et al., 2020; Canatar et al., 2021; Simon et al., 2021). The theory shows that inductive bias, rather than the dimension of the representation alone, is necessary to explain learning performance across tasks. It also suggests that cerebellar regions specialized for different functions may adjust the sparsity of their granule cell representations depending on the task."

5) The interpretation of the distribution of the mossy fiber inputs to the granule cells, which would have a crucial impact on the results of a study like this, is likely incorrect. First, unlike the papers that the authors cite, there are many studies indicating that there is a topographic organization in the mossy fiber termination, such that mossy fibers from the same inputs, representing similar types of information, are regionally co-localized in the granule cell layer. Hence, there is no support for the model assumption that there is a predominantly random termination of mossy fibers of different origins. This risks invalidating the comparisons that the authors are making, i.e. such as in Figure 3. This is a list of example papers, there are more: van Kan, Gibson and Houk (1993) Movement-related inputs to intermediate cerebellum of the monkey. Journal of Neurophysiology. Garwicz et al (1998) Cutaneous receptive fields and topography of mossy fibres and climbing fibres projecting to cat cerebellar C3 zone. The Journal of Physiology. Brown and Bower (2001) Congruence of mossy fiber and climbing fiber tactile projections in the lateral hemispheres of the rat cerebellum. The Journal of Comparative Neurology. Na, Sugihara, Shinoda (2019) The entire trajectories of single pontocerebellar axons and their lobular and longitudinal terminal distribution patterns in multiple aldolase C-positive compartments of the rat cerebellar cortex. The Journal of Comparative Neurology.

6) The nature of the mossy fiber-granule cell recording is also reviewed here: Gilbert and Miall (2022) How and Why the Cerebellum Recodes Input Signals: An Alternative to Machine Learning. The Neuroscientist. Further, considering the re-coding idea, the following paper shows that detailed information, as it is provided by mossy fibers, is transmitted through the granule cells without any evidence of re-coding: Jo¨rntell and Ekerot (2006) Journal of Neuroscience; and this paper shows that these granule inputs are powerfully transmitted to the molecular layer even in a decerebrated animal (i.e. where only the ascending sensory pathways remains) Jo¨rntell and Ekerot 2002, Neuron.

We agree that there is strong evidence for a topographic organization in mossy fiber to granule cell connectivity at the microzonal level. We thank the Reviewer for pointing us to specific examples. We acknowledge that our simplified model does not capture the structure of connectivity observed in these studies.

However, the focus of our model is on cerebellar neurons presynaptic to a single Purkinje cell. Random or disordered distribution of inputs at this local scale is compatible with topographic organization at the microzonal scale. Furthermore, while there is evidence of structured connections at the local scale, models with random connectivity are able to reproduce the dimensionality of granule cell activity within a small margin of error (Nguyen et al., 2022). Finally, our finding that dense codes are optimal for learning slowly varying tasks is consistent with evidence for the lack of re-coding – for such tasks, re-coding may absent because it is not required.

We have dedicated a section on this issue in the Assumptions and Extensions portion of our Discussion:

"Another key assumption concerning the granule cells is that they sample mossy fiber inputs randomly, as is typically assumed in Marr-Albus models (Marr, 1969; Albus, 1971; LitwinKumar et al., 2017; Cayco-Gajic et al., 2017). Other studies instead argue that granule cells sample from mossy fibers with highly similar receptive fields (Garwicz et al., 1998; Brown and Bower, 2001; J¨orntell and Ekerot, 2006) defined by the tuning of mossy fiber and climbing fiber inputs to cerebellar microzones (Apps et al., 2018). This has led to an alternative hypothesis that granule cells serve to relay similarly tuned mossy fiber inputs and enhance their signal-to-noise ratio (Jo¨rntell and Ekerot, 2006; Gilbert and Chris Miall, 2022) rather than to re-encode inputs. Another hypothesis is that granule cells enable Purkinje cells to learn piece-wise linear approximations of nonlinear functions (Spanne and J¨orntell, 2013). However, several recent studies support the existence of heterogeneous connectivity and selectivity of granule cells to multiple distinct inputs at the local scale (Huang et al., 2013; Ishikawa et al., 2015). Furthermore, the deviation of the predicted dimension in models constrained by electron-microscopy data as compared to randomly wired models is modest (Nguyen et al., 2022). Thus, topographically organized connectivity at the macroscopic scale may coexist with disordered connectivity at the local scale, allowing granule cells presynaptic to an individual Purkinje cell to sample heterogeneous combinations of the subset of sensorimotor signals relevant to the tasks that Purkinje cell participates in. Finally, we note that the optimality of dense codes for learning slowly varying tasks in our theory suggests that observations of a lack of mixing (J¨orntell and Ekerot, 2002) for such tasks are compatible with Marr-Albus models, as in this case nonlinear mixing is not required."

7) I could not find any description of the neuron model used in this paper, so I assume that the neurons are just modelled as linear summators with a threshold (in fact, Figure 5 mentions inhibition, but this appears to be just one big lump inhibition, which basically is an incorrect implementation). In reality, granule cells of course do have specific properties that can impact the input-output transformation, PARTICULARLY with respect to the comparison of sparse versus dense coding, because the low-pass filtering of input that occurs in granule cells (and other neurons) as well as their spike firing stochasticity (Saarinen et al (2008). Stochastic differential equation model for cerebellar granule cell excitability. PLoS Comput. Biol. 4:e1000004) will profoundly complicate these comparisons and make them less straight forward than what is portrayed in this paper. There are also several other factors that would be present in the biological setting but are lacking here, which makes it doubtful how much information in relation to the biological performance that this modelling study provides: What are the types of activity patterns of the inputs? What are the learning rules? What is the topography? What is the impact of Purkinje cell outputs downstream, as the Purkinje cell output does not have any direct action, it acts on the deep cerebellar nuclear neurons, which in turn act on a complex sensorimotor circuitry to exert their effect, hence predictive coding could only become interpretable after the PC output has been added to the activity in those circuits. Where is the differentiated Golgi cell inhibition?

Thank you for these critiques. We have made numerous edits to improve the presentation of the details of our model in the main text of the manuscript. Indeed, granule cells in the main text are modeled as linear sums of mossy fiber inputs with a threshold-linear activation function. A more detailed description of the model for granule cells can now be found in Equation 1 in the Results section:

"The activity of neurons in the expansion layer is given by: h = φ(Jeffx − θ), (1) where φ is a rectified linear activation function φ(u) = max(u,0) applied element-wise. Our results also hold for other threshold-polynomial activation functions. The scalar threshold θ is shared across neurons and controls the coding level, which we denote by f, defined as the average fraction of neurons in the expansion layer that are active."

Most of our analyses use the firing rate model we describe above, but several Supplemental Figures show extensions to this model. As we mention in the Discussion, our results do not depend on the specific choice of nonlinearity (Figure 2-figure supplement 2). We have also considered the possibility that the stochastic nature of granule cell spikes could impact our measures of coding level. In Figure 7-figure supplement 1 we test the robustness of our main conclusion using a spiking model where we model granule cell spikes with Poisson statistics. When measuring coding level in a population of spiking neurons, a key question is at what time window the Purkinje cell integrates spikes. For several choices of integration time windows, we show that dense coding remains optimal for learning smooth tasks. However, we agree with the Reviewer that there are other biological details our model does not address. For example, our spiking model does not capture some of the properties the Saarinen et al. (2008) model captures, including random sub-threshold oscillations and clusters of spikes. Modeling biophysical phenomena at this scale is beyond the scope of our study. We have added this reference to the relevant section of the Discussion:

"We also note that coding level is most easily defined when neurons are modeled as rate, rather than spiking units. To investigate the consistency of our results under a spiking code, we implemented a model in which granule cell spiking exhibits Poisson variability and quantify coding level as the fraction of neurons that have nonzero spike counts (Figure 7-figure supplement 1; Figure 7C). In general, increased spike count leads to improved performance as noise associated with spiking variability is reduced. Granule cells have been shown to exhibit reliable burst responses to mossy fiber stimulation (Chadderton et al., 2004), motivating models using deterministic responses or sub-Poisson spiking variability. However, further work is needed to quantitatively compare variability in model and experiment and to account for more complex biophysical properties of granule cells (Saarinen et al., 2008)."

A second concern the Reviewer raises is our implementation of Golgi cell inhibition as a homogeneous rather than heterogeneous input onto granule cells. In simplified models, adding heterogeneous inhibition does not dramatically change the qualitative properties of the expansion layer representation, in particular the dimensionality of the representation (Billings et al., 2014, Cayco-Gajic et al., 2017, Litwin-Kumar et al., 2017). We have added a section about inhibition to our Discussion:

"We also have not explicitly modeled inhibitory input provided by Golgi cells, instead assuming such input can be modeled as a change in effective threshold, as in previous studies (Billings et al., 2014; Cayco-Gajic et al., 2017; Litwin-Kumar et al., 2017). This is appropriate when considering the dimension of the granule cell representation (Litwin-Kumar et al., 2017), but more work is needed to extend our model to the case of heterogeneous inhibition."

Regarding the mossy fiber inputs, as we state in response to paragraph 3, we agree with the Reviewer that the random and uncorrelated mossy fiber activity that has been used in previous studies is an unrealistic idealization of in vivo neural activity. One of the motivations for our model was to relax this assumption and examine the consequences: we introduce correlations in the mossy fiber activity by projecting low-dimensional patterns into the mossy fiber layer (Figure 1B):

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a low-dimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks.

We therefore assume that the inputs to our model lie on a D-dimensional subspace embedded in the N-dimensional input space, where D is typically much smaller than N (Figure 1B). We refer to this subspace as the “task subspace” (Figure 1C)."

The Reviewer also mentions the learning rule at granule cell to Purkinje cell synapses. We agree that considering online, climbing-fiber-dependent learning is an important generalization. We therefore added a new supplemental figure investigating whether we would still see a difference in optimal coding levels across tasks if online learning were used instead of the least squares solution (Figure 7-figure supplement 2). Indeed, we observed a similar task dependence as we saw in Figure 2F. We have added a new paragraph in the Discussion under Assumptions and Extensions describing our rationale and approach in detail:

"For the Purkinje cells, our model assumes that their responses to granule cell input can be modeled as an optimal linear readout. Our model therefore provides an upper bound to linear readout performance, a standard benchmark for the quality of a neural representation that does not require assumptions on the nature of climbing fiber-mediated plasticity, which is still debated. Electrophysiological studies have argued in favor of a linear approximation (Brunel et al., 2004). To improve the biological applicability of our model, we implemented an online climbing fiber-mediated learning rule and found that optimal coding levels are still task-dependent (Figure 7-figure supplement 2). We also note that although we model several timing-dependent tasks (Figure 7), our learning rule does not exploit temporal information, and we assume that temporal dynamics of granule cell responses are largely inherited from mossy fibers. Integrating temporal information into our model is an interesting direction for future investigation."

Finally, regarding the function of the Purkinje cell, our model defines a learning task as a mapping from inputs to target activity in the Purkinje cell and is thus agnostic to the cell’s downstream effects. We clarify this point when introducing the definition of a learning task:

"In our model, a learning task is defined by a mapping from task variables x to an output f(x), representing a target change in activity of a readout neuron, for example a Purkinje cell. The limited scope of this definition implies our results should not strongly depend on the influence of the readout neuron on downstream circuits."

8) The problem of these, in my impression, generic, arbitrary settings of the neurons and the network in the model becomes obvious here: ‘In contrast to the dense activity in cerebellar granule cells, odor responses in Kenyon cells, the analogs of granule cells in the Drosophila mushroom body, are sparse...’ How can this system be interpreted as an analogy to granule cells in the mammalian cerebellum when the model does not address the specifics lined up above? I.e. the ‘inductive bias’ that the authors speak of, defined as ‘the tendency of a network toward learning particular types of input-output mappings’, would be highly dependent on the specifics of the network model.

We agree with the Reviewer that our model makes several simplifying assumptions for mathematical tractability. However, we note that our study is not the first to draw analogies between cerebellum-like systems, including the mushroom body (Bell et al., 2008; Farris, 2011). All the systems we study feature a sparsely connected, expanded granule-like layer that sends parallel fiber axons onto densely connected downstream neurons known to exhibit powerful synaptic plasticity, thus motivating the key architectural assumptions of our model. We have constrained anatomical parameters of the model using data as available (Table 1). However, we agree with the Reviewer that when making comparisons across species there is always a possibility that differences are due to physiological mechanisms we have not fully understood or captured with a model. As such, we can only present a hypothesis for these differences. We have modified our Discussion section on this topic to clearly state this.

"Our results predict that qualitative differences in the coding levels of cerebellum-like systems, across brain regions or across species, reflect an optimization to distinct tasks (Figure 7). However, it is also possible that differences in coding level arise from other physiological differences between systems."

9) More detailed comments: Abstract: ‘In these models [Marr-Albus], granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli.’ Yes, I would agree with the first part, but I contest the second part of this statement. I think what is true for sparse coding is that the learning of random stimuli will be faster, as in a perceptron, but not necessarily better. As the sparsification essentially removes information, it could be argued that the quality of the learning is poorer. So from that perspective, it is not optimal. The authors need to specify from what perspective they consider sparse representations optimal for learning.

This is an important point that we would like to clarify. It is not the case that sparse coding simply speeds up learning. In our study and many related works (Barak et al. 2013; Babadi and Sompolinsky 2014; Litwin-Kumar et al. 2017), learning performance is measured based on the generalization ability of the network – the ability to predict correct labels for previously unseen inputs. As our study and previous studies show, sparse codes are optimal in the sense that they minimize generalization error, independent of any effect on learning speed. To communicate this more effectively, we have added the following sentence to the first paragraph of the Introduction:

"Sparsity affects both learning speed (Cayco-Gajic et al., 2017), and generalization, the ability to predict correct labels for previously unseen inputs (Barak et al., 2013; Babadi and Sompolinsky, 2014; Litwin-Kumar et al., 2017)."

10) Introduction: ‘Indeed, several recent studies have reported dense activity in cerebellar granule cells in response to sensory stimulation or during motor control tasks (Knogler et al., 2017; Wagner et al., 2017; Giovannucci et al., 2017; Badura and De Zeeuw, 2017; Wagner et al., 2019), at odds with classic theories (Marr, 1969; Albus, 1971).’ In fact, this was precisely the issue that was addressed already by Jo¨rntell and Ekerot (2006) Journal of Neuroscience. The conclusion was that these actual recordings of granule cells in vivo provided essentially no support for the assumptions in the Marr-Albus theories.

In our reading, the main finding of J¨orntell and Ekerot (2006) is that individual granule cells are activated by mossy fibers with overlapping receptive fields driven by a single type of somatosensory input. However, there is also evidence of nonlinear mixed selectivity in granule cells in support of the re-coding hypothesis (Huang et al., 2013; Ishikawa et al., 2015). Jo¨rntell and Ekerot (2006) also suggest that the granule cell layer shares similar topographic organization as mossy fibers, organized into microzones. The existence of topographic organization does not invalidate Marr-Albus theories. As we have suggested earlier, a local combinatorial expansion can coexist with a global topographic organization.

We have described these considerations in the Assumptions and Extensions portion of the Discussion:

"Another key assumption concerning the granule cells is that they sample mossy fiber inputs randomly, as is typically assumed in Marr-Albus models (Marr, 1969; Albus, 1971; LitwinKumar et al., 2017; Cayco-Gajic et al., 2017). Other studies instead argue that granule cells sample from mossy fibers with highly similar receptive fields (Garwicz et al., 1998; Brown and Bower, 2001; J¨orntell and Ekerot, 2006) defined by the tuning of mossy fiber and climbing fiber inputs to cerebellar microzones (Apps et al., 2018). This has led to an alternative hypothesis that granule cells serve to relay similarly tuned mossy fiber inputs and enhance their signal-to-noise ratio (Jo¨rntell and Ekerot, 2006; Gilbert and Chris Miall, 2022) rather than to re-encode inputs. Another hypothesis is that granule cells enable Purkinje cells to learn piece-wise linear approximations of nonlinear functions (Spanne and J¨orntell, 2013). However, several recent studies support the existence of heterogeneous connectivity and selectivity of granule cells to multiple distinct inputs at the local scale (Huang et al., 2013; Ishikawa et al., 2015). Furthermore, the deviation of the predicted dimension in models constrained by electron-microscopy data as compared to randomly wired models is modest (Nguyen et al., 2022). Thus, topographically organized connectivity at the macroscopic scale may coexist with disordered connectivity at the local scale, allowing granule cells presynaptic to an individual Purkinje cell to sample heterogeneous combinations of the subset of sensorimotor signals relevant to the tasks that Purkinje cell participates in. Finally, we note that the optimality of dense codes for learning slowly varying tasks in our theory suggests that observations of a lack of mixing (J¨orntell and Ekerot, 2002) for such tasks are compatible with Marr-Albus models, as in this case nonlinear mixing is not required."

We have also included the Jo¨rntell and Ekerot (2006) study as a citation in the Introduction:

"Indeed, several recent studies have reported dense activity in cerebellar granule cells in response to sensory stimulation or during motor control tasks (Jo¨rntell and Ekerot, 2006; Knogler et al., 2017; Wagner et al., 2017; Giovannucci et al., 2017; Badura and De Zeeuw, 2017; Wagner et al., 2019), at odds with classic theories (Marr, 1969; Albus, 1971)."

11) Results: 1st para: There is no information about how the granule cells are modelled.

We agree that this should information should have been more readily available. We now more completely describe the model in the main text. Our model for granule cells can be found in Equation 1 in the Results section and also the Methods (Network Model):

"The activity of neurons in the expansion layer is given by: h = φ(Jeffx − θ), (2)

where φ is a rectified linear activation function φ(u) = max(u,0) applied element-wise. Our results also hold for other threshold-polynomial activation functions. The scalar threshold θ is shared across neurons and controls the coding level, which we denote by f, defined as the average fraction of neurons in the expansion layer that are active."

12) 2nd para: ‘A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space.’ Yes, I agree, and this is in fact in conflict with the known topographical organization in the cerebellar cortex (see broader comment above). Mossy fiber inputs coding for closely related inputs are co-localized in the cerebellar cortex. I think for this model to be of interest from the point of view of the mammalian cerebellar cortex, it would need to pay more attention to this organizational feature.

As we discuss in our response to paragraphs 5 and 6, we see the random distribution assumption at the local scale (inputs presynaptic to a single Purkinje cell) as being compatible with topographic organization occurring at the microzone scale. Furthermore, as discussed earlier, we specifically model low-dimensional input as opposed to the random and high-dimensional inputs typically studied in prior models.

"A typical assumption in computational theories of the cerebellar cortex is that inputs are randomly distributed in a high-dimensional space (Marr, 1969; Albus, 1971; Brunel et al., 2004; Babadi and Sompolinsky, 2014; Billings et al., 2014; Litwin-Kumar et al., 2017). While this may be a reasonable simplification in some cases, many tasks, including cerebellumdependent tasks, are likely best-described as being encoded by a low-dimensional set of variables. For example, the cerebellum is often hypothesized to learn a forward model for motor control (Wolpert et al., 1998), which uses sensory input and motor efference to predict an effector’s future state. Mossy fiber activity recorded in monkeys correlates with position and velocity during natural movement (van Kan et al., 1993). Sources of motor efference copies include motor cortex, whose population activity lies on a low-dimensional manifold (Wagner et al., 2019; Huang et al., 2013; Churchland et al., 2010; Yu et al., 2009). We begin by modeling the low dimensionality of inputs and later consider more specific tasks. We therefore assume that the inputs to our model lie on a D-dimensional subspace embedded in the N-dimensional input space, where D is typically much smaller than N (Figure 1B). We refer to this subspace as the “task subspace” (Figure 1C)."

References

Albus, J.S. (1971). A theory of cerebellar function. Mathematical Biosciences 10, 25–61.

Apps, R., et al. (2018). Cerebellar Modules and Their Role as Operational Cerebellar Processing Units. Cerebellum 17, 654–682.

Babadi, B. and Sompolinsky, H. (2014). Sparseness and expansion in sensory representations. Neuron 83, 1213–1226.

Badura, A. and De Zeeuw, C.I. (2017). Cerebellar granule cells: dense, rich and evolving representations. Current Biology 27, R415–R418.

Barak, O., Rigotti, M., and Fusi, S. (2013). The sparseness of mixed selectivity neurons controls the generalization–discrimination trade-off. Journal of Neuroscience 33, 3844– 3856.

Bell, C.C., Han, V., and Sawtell, N.B. (2008). Cerebellum-like structures and their implications for cerebellar function. Annual Review of Neuroscience 31, 1–24.

Billings, G., Piasini, E., Lo˝rincz, A., Nusser, Z., and Silver, R.A. (2014). Network structure within the cerebellar input layer enables lossless sparse encoding. Neuron 83, 960–974.

Bordelon, B., Canatar, A., and Pehlevan, C. (2020). Spectrum dependent learning curves in kernel regression and wide neural networks. International Conference on Machine Learning 1024–1034.

Brown, I.E. and Bower, J.M. (2001). Congruence of mossy fiber and climbing fiber tactile projections in the lateral hemispheres of the rat cerebellum. Journal of Comparative Neurology 429, 59–70.

Brunel, N., Hakim, V., Isope, P., Nadal, J.P., and Barbour, B. (2004). Optimal information storage and the distribution of synaptic weights: perceptron versus Purkinje cell. Neuron 43, 745–757.

Canatar, A., Bordelon, B., and Pehlevan, C. (2021). Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks. Nature Communications 12, 1–12.

Cayco-Gajic, N.A., Clopath, C., and Silver, R.A. (2017). Sparse synaptic connectivity is required for decorrelation and pattern separation in feedforward networks. Nature Communications 8, 1–11.

Chadderton, P., Margrie, T.W., and Ha¨usser, M. (2004). Integration of quanta in cerebellar granule cells during sensory processing. Nature 428, 856–860.

Churchland, M.M., et al. (2010). Stimulus onset quenches neural variability: a widespread cortical phenomenon. Nature Neuroscience 13, 369–378.

Farris, S.M. (2011). Are mushroom bodies cerebellum-like structures? Arthropod structure & development 40, 368–379.

Garwicz, M., Jorntell, H., and Ekerot, C.F. (1998). Cutaneous receptive fields and topography of mossy fibres and climbing fibres projecting to cat cerebellar C3 zone. The Journal of Physiology 512 ( Pt 1), 277–293.

Gilbert, M. and Chris Miall, R. (2022). How and Why the Cerebellum Recodes Input Signals: An Alternative to Machine Learning. The Neuroscientist 28, 206–221.

Giovannucci, A., et al. (2017). Cerebellar granule cells acquire a widespread predictive feedback signal during motor learning. Nature Neuroscience 20, 727–734.

Huang, C.C., et al. (2013). Convergence of pontine and proprioceptive streams onto multimodal cerebellar granule cells. eLife 2, e00400.

Ishikawa, T., Shimuta, M., and Ha¨usser, M. (2015). Multimodal sensory integration in single cerebellar granule cells in vivo. eLife 4, e12916.

Jacot, A., Gabriel, F., and Hongler, C. (2018). Neural tangent kernel: Convergence and generalization in neural networks. Advances in Neural Information Processing Systems 31.

Jo¨rntell, H. and Ekerot, C.F. (2002). Reciprocal Bidirectional Plasticity of Parallel Fiber Receptive Fields in Cerebellar Purkinje Cells and Their Afferent Interneurons. Neuron 34, 797–806.

Jorntell, H. and Ekerot, C.F. (2006). Properties of Somatosensory Synaptic Integration in Cerebellar Granule Cells In Vivo. Journal of Neuroscience 26, 11786–11797.

Knogler, L.D., Markov, D.A., Dragomir, E.I., Stih, V., and Portugues, R. (2017). Senso-ˇ rimotor representations in cerebellar granule cells in larval zebrafish are dense, spatially organized, and non-temporally patterned. Current Biology 27, 1288–1302.

Litwin-Kumar, A., Harris, K.D., Axel, R., Sompolinsky, H., and Abbott, L.F. (2017). Optimal degrees of synaptic connectivity. Neuron 93, 1153–1164. Marr, D. (1969). A theory of cerebellar cortex. Journal of Physiology 202, 437–470.

Nguyen, T.M., et al. (2022). Structured cerebellar connectivity supports resilient pattern separation. Nature 1–7.

Saarinen, A., Linne, M.L., and Yli-Harja, O. (2008). Stochastic Differential Equation Model for Cerebellar Granule Cell Excitability. PLOS Computational Biology 4, e1000004.

Simon, J.B., Dickens, M., and DeWeese, M.R. (2021). A theory of the inductive bias and generalization of kernel regression and wide neural networks. arXiv: 2110.03922.

Sollich, P. (1998). Learning curves for Gaussian processes. Advances in Neural Information Processing Systems 11.

Spanne, A. and Jo¨rntell, H. (2013). Processing of Multi-dimensional Sensorimotor Information in the Spinal and Cerebellar Neuronal Circuitry: A New Hypothesis. PLOS Computational Biology 9, e1002979.

Spanne, A. and Jo¨rntell, H. (2015). Questioning the role of sparse coding in the brain. Trends in Neurosciences 38, 417–427.

van Kan, P.L., Gibson, A.R., and Houk, J.C. (1993). Movement-related inputs to intermediate cerebellum of the monkey. Journal of Neurophysiology 69, 74–94.

Wagner, M.J., Kim, T.H., Savall, J., Schnitzer, M.J., and Luo, L. (2017). Cerebellar granule cells encode the expectation of reward. Nature 544, 96–100.

Wagner, M.J., et al. (2019). Shared cortex-cerebellum dynamics in the execution and learning of a motor task. Cell 177, 669–682.e24.

Wolpert, D.M., Miall, R.C., and Kawato, M. (1998). Internal models in the cerebellum. Trends in Cognitive Sciences 2, 338–347.

Yu, B.M., et al. (2009). Gaussian-process factor analysis for low-dimensional single-trial analysis of neural population activity. Journal of Neurophysiology 102, 614–635.

Author Response

Reviewer #1 (Public Review):

This study combines in vitro somatic and dendritic recordings and computational modeling to study how cholinergic agonists modulate the response of CA1 pyramidal neurons to triangular current injections. The authors have previously used a similar approach (Upchurch, 2022, JNeuroscience) to show that CA1 neurons exhibit asymmetric AP firing (more firing on the upward ramp) in response to such current injections and that this effect is due to Na channel inactivation. The present work builds on these results by showing that cholinergic modulation changes this response, i.e., there is more firing on the downward part of the ramp. This change appears to require an intracellular Ca2+ concentration increase (mediated via IP3 and voltage-gated Ca2+ channels), which activates TRPM4 channels. In this scheme, cholinergic activity increases IP3, and the depolarizing current injection opens voltage-gated Ca2+ channels. This study will be of some interest to cellular neurophysiology experts working on the hippocampus.

1) This study claims that the triangular current injections recapitulate hippocampal place cell activity. However, it has been shown recently that the asymmetric firing of CA1 place cells is due to synaptic weight changes resulting from synaptic plasticity (e.g., Bittner et al., 2017). This suggests that the asymmetric firing of place cells is primarily the result of asymmetric synaptic input. Therefore, the authors should test whether carbachol similarly affects a synaptically driven membrane potential ramp. If this is not the case, the strong claim that this work has implications for place cell firing is not justified, in my opinion.

We have added the results showing the effects of cholinergic modulation on a synaptically-driven membrane potential ramp, obtained by electrically stimulating the Schaffer collaterals with a stimulation frequency that was adjusted according to a linear, symmetric ramp (see also Hsu et al, Neuron 99,147-162, 2018). These results have been added to the manuscript in the Results section for new Figure 2 (lines 169-197) and in the Methods section (lines 716-726).

2) Along the same lines, it has been shown before that the precision of spike timing depends on the stimulation pattern in vitro (Mainen and Sejnowski, 1995). Constant stimuli led to imprecise AP firing trains, whereas current injections that included fluctuations resembling synaptic input generated spike trains that were more reliable and reproducible in terms of timing. This study concluded that a low intrinsic noise level in spike generation was essential in generating informative spike sequences. Following this pivotal work, the authors could add noise to their current stimulus and observe the effect on the AP firing patterns. If this is not possible, the authors should at least report the sweep-to-sweep variability for the data shown, e.g., in panels 1A2, 1B2, 1D2, and 1E2.

We thank the reviewer for this suggestion to acknowledge the variability in the data across trials and we have added the Mainen and Sejnowski, 1995 citation to the manuscript (see Results lines 128-134). We addressed sweep-to-sweep variability among the various trials.

3) In most of the data presented in this manuscript, Carbachol appears to induce a 3 mV hyperpolarization and increase input resistance. As a result, the amount of current injected during Carbachol is drastically lower than during the controls. This should be emphasized more, and the input resistance should be quantified for each experimental condition. It should also be discussed whether this change in input resistance can account for the changes in the firing pattern observed. Finally, it should be clearly stated how the amount of the current injected was chosen for each cell, and data from a range of injected current ramps should be shown for each cell.

We thank the reviewers for this comment, which made us realize that our initial presentation was not clear, in particular with regard to the traces that were chosen as examples in the initial submission of the paper. We now clarify on page 5 (lines 113-125) of the manuscript as follows:

“In some trials, under control conditions, we applied a baseline depolarization prior to the ramp, in order to capture the variability observed in vivo (Harvey et al Nature 461:941–946, 2009; Epsztein et al. Neuron 70:109–120, 2011). Application of the cholinergic agonist carbachol (CCh, 2 µM) caused a depolarization of 2-6 mV. We compensated for this depolarization by injecting tonic hyperpolarizing current to reestablish the original membrane potential (see also Losonczy, et al., Nature 452, 436-442, 2008), as indicated by an offset from the 0 pA current level in the traces of the injected current ramps. The amplitude of background fluctuations in the resting membrane potential increased from a few tenths of a mV in control to 2-4 mV in CCh. Moreover, the threshold for action potential generation became more hyperpolarized. For all these reasons, we were not able to consistently vary the membrane potential using baseline depolarizations in the presence of CCh, because baseline depolarization alone frequently evoked spiking.”

For this reason, many of the carbachol example traces in the initial submission had more hyperpolarized Vm than their control counterparts. Acetylcholine also caused a depolarization in a dose-dependent manner, that was compensated for in the same way. In this new version of the manuscript, we systematically report the effects of cholinergic agonists on membrane potential and neuronal excitability. Further, we show example traces with resting membrane potentials within 1 mV for each pharmacological comparison, therefore removing this variable and hopefully making results clearer. We also now state how the amount of injected current was chosen for each condition, and that the amount of injected current was generally lower in the presence of cholinergic agonists. Both the tonic hyperpolarizing current and the amplitude of the injected ramp for each example can now be appreciated in each figure.

Finally, the reviewers’ comment also made us realize that, in principle, the center of mass of firing could be systematically skewed by the initial membrane potential, the amplitude of the current ramp injection and/or the input resistance. For this reason, we added a supplementary figure (1-2) where the adaptation index was plotted as a function of each these variables. In all cases, it is apparent that the main factor determining whether the center of mass of firing is shifted earlier or later in the ramp is the presence or absence of carbachol rather than initial membrane potential, current injection amplitude, or input resistance.

4) It remains unclear how the current result that TRPM4 channels can mediate the firing pattern change relates to the previous finding that the current injection evoked CA1 neuronal firing pattern is due to long-term Na channel inactivation.

We thank the reviewers for this suggestion, which helps to clarify our initial results. New Figure 8 addresses the connection between long-term inactivation of Na+ channels and the activation of TRPM4 channels, as characterized by the model (see Results lines 375-391). Furthermore, the model was instrumental in assessing how the Ca2+ and voltage-dependence of TRPM4 channels synergize to contribute to the shift in the center of mass of firing (Figure 9). Figure 9 illustrates the positive feedback loop between Ca2+ entry and the additional depolarization produced by Ca2+ activation of TRPM4 channels that can potentially accelerate firing (see Results lines 392-427).

5) Figure 8: Panel C is supposed to confirm the prediction from the model that the carbachol-mediated change of firing activity is related to intracellular Ca2+ domains. However, the example cell shown is depolarized to -52 mV, and there is no hyperpolarization following Carbachol. Is this an effect of the high concentration of BAPTA? Again, what was the current injected under this experimental condition?

Again, we thank the reviewer for pointing out the lack of clarity in the presentation of our results. We have now rewritten the results section for former Figure 8 (now Figure 10) to more clearly present these findings. The reviewer is correct that with the combination of 30 mM BAPTA + 10 nM free Ca2+ added to the intracellular solution (panel C of current Figure 10) the addition of carbachol did not change the membrane potential, as there were no changes in the holding current. Also, the amplitude of the ramp is comparable in control conditions and in the presence of carbachol under these conditions.

We have now added all these details in the Results section for figure 10C.

Reviewer #2 (Public Review):

The manuscript focuses on the cholinergic modulation of TRPM4 channels in the CA1 pyramidal neurons. The authors presented solid convincing evidence that TRPM4 but not TRPC channels are the Ca2+-activated nonselective cation channel in CA1 pyramidal neurons being modulated by activation of muscarinic receptors. Using bi-directional ramp protocol, the authors revealed that ACh modulation could lead to forward shifts in place field center of mass, whereas decreased ACh modulation could contribute to backward shifts. This represents a significant molecular/cellular finding that links neuromodulation of intrinsic properties to place field shifts, a phenomenon seen in vivo. The authors used a computational approach to model this CA1 neuron spiking to further reveal the mechanism.

To further improve the manuscript, I have the following suggestions/questions:

1) The triangular ramp stimulation (introduced by the same group; Upchurch et al., 2022) makes it possible to emulate the hill-shaped depolarization during place field firing. However, one concern is the time scale/duration of the ramp (2 sec) compared to the physiological pattern (100ms~200ms in the in vivo recording in freely moving rat, Epsztein et al., 2011). Using a longer ramp to generate more spikes for calculating the adaptation index is understandable. However, considering the Ca entry/accumulation during prolonged depolarization, repeating one set of experiments with a shorter ramp is crucial to verify the major findings.

When determining the duration of the current injections for our ramps, we relied on the data recorded in vivo in freely moving rats (Epsztein et al. Neuron 70:109–120, 2011) or in head-fixed mice running on spherical a treadmill immersed in virtual reality (Harvey et al Nature 461:941–946, 2009). In those papers, the voltage deflections are shown as a function of time, and gray bars or boxes represent the time the animals spend traversing the place field. We interpret those figures as showing that the hill-shaped depolarizations have variable durations, on the order of 1-20 s; we therefore think that our experiments with 2 and 10 second-long ramps cover a fair range of these durations. The place fields in Epsztein et al., 2011 were 4 cm long, and the authors give an example in Figure 3, in which the 2 meter track is traversed 1.5 times in 3 minutes. At that rate, the rat spent on average 2.4 seconds in each place field. We interpret the numerous shorter epochs of firing on the order of 100-200 ms shown Figure 2 in Epsztein et al. as the result of ongoing theta modulation within one overall depolarization during a single place field traversal. The following quote from that paper supports our interpretation “Some (Figure 2E, trace 1), but not all (trace 2), passes revealed spiking associated with a series of large (to ~-25 mV), long-lasting (~100 ms) depolarizations (Kandel and Spencer, 1961; Wong and Prince, 1978; Traub and Llinás, 1979; Takahashi and Magee, 2009) occurring rhythmically at ~4–5 Hz (theta frequency).” We thank the reviewer for pointing out these traces; our results are more directly applicable to the traces without theta modulation. Adding theta modulation is beyond the scope of this study but will be considered in future studies. Our average results in Figure 1 show that carbachol similarly affects 2 s and 10 s ramps, therefore we decided to present only the data on 2 second ramps for all the subsequent figures (see Results lines 156-157).

2) Strictly speaking, the term "Ca2+-induced Ca2+ release (CICR)" is only used in ER Ca2+ release via ryanodine receptors (RyR) rather than IP3Rs. The author should be careful since it is used in the abstract (Line 36). In addition, pharmacology inhibition experiments should be incorporated to further dissect the role of RyR-induced CICR.

We thank the reviewer for pointing out the possible confusion regarding the use of the term Ca2+-induced Ca2+ release (CICR) and we removed it from the text. Further, for this resubmission, we have pharmacologically dissected the role of IP3 vs ryanodine receptors in the cholinergic shift in the center of mass of firing due to the activation of TRPM4 channels, as suggested by the reviewer (see new Figure 6). To our surprise, neither the IP3R antagonist, Xestospongin C (1-2 µM), nor the RyR antagonist ryanodine (40 µM) were effective in preventing the cholinergic shift of the center of mass of firing when added to the intracellular solution (see Results lines 310-340).

3) Applying strong buffering BAPTA not only removed the IP3R-TRPM nanodomain but also hindered Ca entry via VGCC. To validate the role of ER Ca2+ release in regulating TRPM, depletion of ER Ca2+ pool with SERCA inhibitor (e.g. thapsigargin) would be a more direct way to test the model (also make sure to add TRPC inhibitor to avoid the store-operated Ca2+ entry).

We agree with the reviewer that 30 mM BAPTA also disrupts intracellular Ca2+ elevation via voltage-dependent Ca2+ channels on the neuronal membrane. Given that our experiments excluded a role of Ca2+ release from the intracellular stores (see below), our new model includes a nanodomain where, during cholinergic activation, the Ca2+ entry through VGCC is amplified to reach micromolar concentrations, through a currently unknown mechanism. As pointed out by the reviewer, the experimental results with 30 mM BAPTA support the existence of a nanodomain for the activation of TRPM4 channels, regardless of the nature of the calcium source.

We have also addressed the role of ER Ca2+ release in our experiments.

4) How does the TRPM current overcome the long-term inactivation of Nav? A channel state model should be added to the manuscript to make it easier to understand.

Figure 11C now shows the Markov model of the NaV channel and new Figure 8 is devoted to explaining the mechanism by which current through the TRPM4 channels overcomes the long-term inactivation of the NaV channel.

Reviewer #3 (Public Review):

Combining slice physiology and simulation, Combe and colleagues discovered that TRPM4 channels activated by Ca2+ in nanodomains mediate ICAN currents in CA1 pyramidal neurons that drive the cholinergic modulation of firing rate. The finding is novel and interesting.

Strengths:

1) Identification of TRPM4 channels as the carrier of ICAN currents with independent pharmacological inhibitors and other supporting evidence.

2) Physiological and simulational verification of physically closely located Ca2+ source and TRPM4 channels required for ICAN activation.

Weaknesses:

1) The conclusion of the cholinergic role in down-ramp or backward firing shifts is not convincing.

We agree with the reviewer that our interpretation is somewhat speculative, and we have now included disclaimers throughout the manuscript as well as placed most of these interpretations in a portion of the discussion titled “Ideas and speculations: Implications of our results for place fields in intact rodents”. In addition, we added the word “potential” in the title.

Author Response

Reviewer #1 (Public Review):

The manuscript by Masschelin et al. describes how Vitamin B2 deficiency affects body composition, energy expenditure, and glucose metabolism. B2 deficient mice have lower O2 consumption, and locomotor activity, with no difference in food intake. These mice also have lower liver FAD levels, which is expected given that B2 is a necessary cofactor for this coenzyme. Additionally, these mice have lower blood glucose levels following pyruvate injection, implying a lower capacity for gluconeogenesis. Using PPAR KO mice, they show that this effect on pyruvate tolerance is due to PPARα activation, though there is still a minor difference between wildtype and KO mice. Importantly, they show that fenofibrate PPAR agonism can improve glucose output following pyruvate injection in the absence of B2. The authors also perform robust metabolomics in each experimental condition and phenotype of the mouse well.

Thank you for the positive input.

1) The authors have yet to explore other explanations of differences in glucose metabolism under B2D +/Fenofibrate. The canonical targets of PPARα are involved in fatty acid oxidation, ketogenesis, and VLDL/HDL metabolism, in addition to gluconeogenesis (Bougarne et al. 2018). Gluconeogenesis is more of a fasting response due to CREB, FOXO1/PGC1a activation rather than PPAR. In response to B2D, the PPARα KO mice have increased plasma TGs, which may suggest a difference in VLDL TG secretion (Suppl. S3). Perhaps lipid metabolism is more directly affected, and changes in glucose metabolism are secondary to that of triglyceride metabolism. Regarding ketogenesis, the fenofibrate+ B2D fed mice have decreased plasma betahydroxybutyrate, suggesting decreased ketogenesis, which is a more canonical PPARα pathway (Suppl. S3). Testing each of these processes would help control that this mechanism is specific to gluconeogenesis and not secondary to something else.

We value this reviewer’s comment. To address this point, we considered other mechanisms in our revised Discussion. In future studies, we plan to further explore these metabolic effects and to use ATAC-Seq to understand the transcription factors responsive to B2D. We anticipate these studies will take additional years to complete. Nonetheless, the present studies set the foundation for future work to investigate how FAD influences transcriptional regulation of metabolism.

2) Is the effect on ISR dependent on PPARα? Is the mechanism of Fenofibrate on the liver, or on another cell type? In Figure 1, the authors state that Riboflavin deficiency alters body composition and energy expenditure, and then focuses on the liver. However, FAD levels are also increased in the heart and kidneys in addition to the liver. These tissues also respond to PPARα agonism, in addition to the muscle which plays a role in regulating glucose metabolism (B2D mice also have a higher lean mass (Fig 1e)). Additionally, the authors haven't shown specifically if the effects of Fenofibrate on electron transport and the ISR are dependent on the presence of PPARα (Figure 5, 6).

We agree that knowing whether the effects of Fenofibrate on the ISR require liver PPARA is a critical issue, which will require dedicated studies for a thorough and meaningful conclusion. In new experiments, we knocked down Ppara in the liver using AAV8-Cre administration to Pparaflox/flox mice. Our data show liver-specific Ppara knockdown recapitulates whole-body B2D effects on pyruvate tolerance and hepatic steatosis (Figure 3I). These results agree with findings in whole-body Ppara knockout mice (Supplemental Figure 4), reinforcing the idea that the direct impact of B2D mainly occurs via PPARA activity in the liver. We acknowledge in the discussion ATF4 and ISR activation may contribute to PPARA-independent responses to B2D (Biochem J 443:165–71, 2012; Gut 65:1202-1214, 2016).

An assessment of genetic requirements will require a large, rigorous set of experiments to identify the ratelimiting responses for fenofibrate activities during B2D, which we plan to do in the future. For this report, we decided to focus exclusively on tissue-specific knockout of Ppara. We will establish evidence for ISR responses to B2D in a separate study based on the feedback received here.

Reviewer #2 (Public Review):

The objective of this work by Masschelin et al. is to investigate the physiological relevance of flavin adenine dinucleotide (FAD). In particular, FAD supports the activity of flavoproteins involved in the production of cellular energy. Mutations in genes encoding flavoproteins often are associated with inborn errors of metabolism (IEMs), thus the clinical interest in investigating in more depth the physiological role of FAD. In this study, the authors first subjected male mice to a vitamin B12 deficient diet (B2D), demonstrating that loss of B12 replicates the phenotypes often observed with IEMs, including loss of body weight, hypoglycemia, and fatty liver. Using a combination of metabolomic phenotyping, transcriptomic analyses, and pharmacology (treatment with Fenofibrate, a PPARa agonist), the authors then reach the general conclusion that activation of the nuclear receptor PPARa can rescue the B2D phenotypes, thus revealing that PPARa directly controls the metabolic responses to FAD availability. Although the phenotypic analysis of the mice subjected to B2D increases our knowledge of the physiological impact of depleting the FAD pools on global energy metabolism, not all conclusions and statements made by the authors are totally supported by the data. In particular, the study is overall too descriptive and lacks mechanistic insights. While PPARa is likely an important player in the metabolic response to FAD availability, the molecular details on how FAD controls the activity of PPARa either directly or indirectly are entirely missing. Therefore, the authors are encouraged to directly assess whether B2D directly influences PPARa activity on the genes identified in the study, perform rescue experiments in the liver of PPARa KO mice and explore the possibility that other factors (including nuclear receptors) also participate in the response to B2 deficiency and diminished FAD pools.

We appreciate the input from Reviewer 2. The direct and indirect effects of B2D on PPARA activity are likely not trivial. However, we performed experiments to determine how FAD depletion affects PPARA transcriptional activity using the riboflavin analog and competitive inhibitor lumiflavin (Figure 3L). We found lumiflavin reduced PPRE-luciferase activity in the presence of PPARA agonist. Although the assay is a synthetic reporter expressed in vitro, the experiment provides evidence of how B2D influences PPARA transcriptional activity. And, yes, we agree that our manuscript does not completely reconcile the factor(s) explaining the effects of B2D on gene expression, and expanded the discussion to comment on this point. In future studies, we intend to identify which transcription factor(s) regulate the liver responses to B2D, and further elucidation of the molecular mechanisms will be a central objective of future work.

Author Response

Reviewer #1 (Public Review):

In this manuscript, Scagliotti and colleagues investigate the role of Dlk1 in regulating pituitary size in multiple mouse models with different Dlk1 gene dosages in order to understand the mechanisms of organ size control. They find that overexpression of Dlk1 leads to pituitary overgrowth and loss of Dlk1 causes undergrowth. Authors find two compartments of Dlk1 expression in the pituitary, in the marginal zone stem cell compartment and the parenchymal differentiated cell compartment, and by combing genetic mouse models show that a specific interaction of Dlk1 expression in both regions is necessary to affect pituitary organ size. They present to suggest that Dlk1 may repress Wnt signaling during development to control a shift from progenitor proliferation to differentiation. The data are meticulous, high quality, and clear.

I have some questions about the interpretation of their data regarding the mechanism of Dlk1 regulation of pituitary organ size, as I believe there could be potential alternative explanations for their observations:

I was wondering about the cause of the enlargement of the pituitary gland in Fig 1E, and whether it is caused by an increased number of cells (hyperplasia), an increased cell size (hypertrophy), or both. Line 104 states it is hyperplasia, and that cell size was not affected in WT-TG ('not shown', line 121). However, line 444 says the TG is hypertrophic. It would be good if the authors could elaborate on this and show or state how cell size was determined. Figs 5/6 show that WT-Tg proliferation is generally similar to WT, which suggests the increased size is not hyperplasia. It would be good to know whether this is correct. Some previous studies have shown that in pregnancy, lactotroph hypertrophy can be responsible for pituitary enlargement without hyperplasia (Castrique 2010, Hodson 2012).

We have now clarified this point throughout the manuscript. We had previously counted cells per field in the analysis shown in Figure 1D as a proxy for cell number (these did not significantly differ by genotype). We have now performed a more robust examination. Cell number was determined using a well-established stereological technique: For each animal the maximal cross-sectional area (CSA) was determined from the volumetric analysis. At this level 3 independent sections were used to measure anterior pituitary CSA and count haematoxilin-stained nuclei, giving a mean cells/CSA measurement per individual. This number was multiplied by the AP volume to give an estimate of cell number.

This analysis was performed on mice from the new cohort of animals containing litter matched adults of all 4 genotypes, and shown in Figure 4E. WT-TG animals had a significant increase in cell number compared to WT littermates (p = 0.0443), therefore pituitary expansion occurs by hyperplasia.

Related to the organ size question above, I had a question about the cell number and proportions in Fig 1D/E/F, which shows the maintenance of endocrine cell proportions and an increase in the volume of ~30% in WT-Tg. For the cell proportions to be maintained, I thought the increase in volume per cell type (Fig 1G) would therefore have to also increase proportionally in every cell type, while 1G appears to show an increase in GH (sig) and PRL/TSH cells (ns). It would be good if the authors could discuss this briefly.

We agree and indeed we see this trend across all cell types. When the data in Figure 1G is compared by 2-Way ANOVA we see a significant effect by cell type (p< 0.0001) and by genotype (p = 0.0009). However, for other hormone producing cells the effect size is does not overcome the variation in a smaller cell population so the difference between genotypes does not pass multiple significance testing with the relatively small sample size used. We have modified the legend to Figure 1G to make the ANOVA result clearer.

This study is impactful and will be of interest to several research communities, including those interested in pituitary development and function, organ size control, and gene imprinting mechanisms.

Reviewer #2 (Public Review):

Scagliotti et al address how organ size is regulated by imprinted genes. Using a series of mouse models to modulate the dosage of the paternally expressed gene, Dlk1, the authors demonstrate that DLK1 is important for the maintenance of the stem cell compartment leading to the growth of the pituitary gland and the expansion of growth hormone-producing cells. The authors show that overexpression of Dlk1 leads to pituitary hyperplasia while deletion of the paternal allele leads to reduced pituitary size. Reduced pituitary size is accompanied by reduced cell proliferation in the cleft at e13.5 and an increase in the number of POU1F1+ cells, suggesting that loss of Dlk1 alters the balance between the number of cells remaining in the replicating stem cell pool and those differentiating into the POU1F1 lineage. An elegant caveat of this paper is the rescue of Dlk1 expression in the population of cells expressing Pou1f1 but not in SOX2+ stem cells. Expression of Dlk1 only in POU1F1+ cells is not sufficient to rescue pituitary size. The authors suggest that this is because DLK1 must be present in stem cells which then activate paracrine WNT signaling to promote cell proliferation in POU1F1+ cells.

Strengths:

This is an important study that provides a mechanistic understanding of how the imprinted gene, Dlk1, regulates organ size. The study employs an elegant experimental design to address the dosage requirement for Dlk1 in regulating pituitary gland size. Rescuing Dlk1 in the POU1F1+ cells, but not the marginal zone SOX2+ cells provides intriguing results about a possible role for DLK1 in paracrine signaling between these different pituitary cell types. The study uses publicly available scRNAseq and ChIPseq data to further support their findings and identify Dlk1 as a likely target of POU1F1.

Weaknesses:

The study only analyzes females for the adult time point. For embryonic and postnatal time points sexes are pooled. Gender differences in pituitary gene expression embryonically or postnatally could potentially affect experimental outcomes.

We have now added adult data for both sexes.

The authors employ a mouse model that rescues Dlk1 expression starting at e15.5 in POU1F1+ parenchymal cells but not in marginal zone stem cells. Rescuing Dlk1 expression in a specific population of cells is one of the strengths of this study. Based on this information and the fact that overexpression of Dlk1 leads to increased pituitary size, the authors suggest that DLK1+ marginal zone stem cells and DLK+ parenchymal cells may interact to promote postnatal proliferation. However, the ability to more carefully parse out the complex spatial and temporal contributions of DLK1 to pituitary size would be enhanced by the addition of a mouse model that rescues Dlk1 expression only in SOX2+ cells and a model that rescues expression in both stem cells and POU1F1+ cells.

We agree that the addition of a model where Dlk1 is only expressed in SOX2+ cells would add significant mechanistic insight. To our knowledge an inducible gain-of-function Dlk1 model does not yet exist. Moreover, use of a SOX2-Cre driver would also increase Dlk1 expression in the hypothalamus as well as Rathke’s pouch, further complicating the analysis.

Author Response

Reviewer #1 (Public Review):

In this manuscript, Huang et al., assess cognitive flexibility in rats trained on an animal model of anorexia nervosa known as activity-based anorexia (ABA). For the first time, they do this in a way that is fully automated and free from experimenter interference, as apparently experimenter interference can affect both the development of ABA as well as the effect on behaviour. They show that animals that are more cognitively flexible (i.e. animals that had received reversal training) were better able to resist weight loss upon exposure to ABA, whereas animals exposed to ABA first show poorer cognitive flexibility (reversal performance).

Strengths:

• The development of a fully-automated, experimenter-free behavioural assessment paradigm that is capable of identifying individual rats and therefore tracking their performance.

• The bidirectional nature of the study - i.e. the fact that animals were tested for cognitive flexibility both before and after exposure to ABA, so that direction of causality could be established.

• The analyses are rigorous and the sample sizes sufficient.

• The use of touchscreens increases the translational potential of the findings.

Weaknesses

• Some descriptions of methods and results are confusing or insufficiently detailed.

We have been through all methods and results to include additional details as requested by this reviewer below.

It seems to me that performance on the pairwise discrimination task cannot be directly (statistically) compared to performance on reversal (as in Figure 4E), as these are tapping into fundamentally different cognitive processes (discrimination versus reversal learning). I think comparing groups on each assessment is valid, however.

We agree that discrimination and reversal are different cognitive processes, and statistical comparisons between these two components of the task were only made when examining the speed of learning in the validation of the novel testing system. Moreover, our inclusion of the pink and purple bars on graphs such as Figure 4C & 4E represent “main effects of ABA exposure”, regardless of learning phase (PD or reversal) rather than, as you describe, comparing PD to R1. Perhaps this comparison wasn’t clear, so we have amended the text to say ‘main effect of ABA exposure p=.0017’ rather than just “exposure”.

Not necessarily a 'weakness' but I would have loved to see some assessment of the alterations in neural mechanisms underlying these effects, and/or some different behavioural assessments in addition to those used here. In particular, the authors mention in the discussion that this manipulation can affect cholinergic functioning in the dorsal striatum We (Bradfield et al., Neuron, 2013) and a number of others have now demonstrated that cholinergic dysfunction in the dorsomedial striatum impairs a different kind of reversal learning that based on alterations in outcome identity and thus relies on a different cognitive process (i.e. 'state' rather than 'reward' prediction error). It would be interesting perhaps in the future to see if the ABA manipulation also alters performance on this alternative 'cognitive flexibility' task.

This is an excellent suggestion and we have already begun exploring this in other ongoing work in the laboratory. Due to ‘compulsive’ wheel running being a hallmark of ABA, we are interested in determining if this also translates to a goal-directed action impairment using the well-established outcome-specific devaluation task. Perhaps with ABA it may be more relevant to investigate outcome-reversals rather than stimulus-reversals, and if this is the case, it would further support the use of the ABA model for investigating cognitive dysfunction relevant to AN. We have included an additional section in the discussion text relating to our hypotheses regarding outcome-specific reversal learning in the ABA model.

Nevertheless, I certainly think the manuscript provides a solid appraisal of cognitive flexibility using more traditional tasks, and that the authors have achieved their aims. I think the work here will be of importance, certainly to other researchers using the ABA model, but perhaps also of translational importance in the future, as the causal relationship between ABA and cognitive inflexibility is near impossible to establish using human studies, but here evidence points strongly towards this being the case.

Reviewer #2 (Public Review):

Huang and colleagues present data from experiments assessing the role of cognitive inflexibility in the vulnerability to weight loss in the activity-based anorexia paradigm in rats. The experiments employ a novel in-home cage touchscreen system. The home cage touch screen system allows reduced testing time and increased throughput compared with the more widely used systems resulting in the ability to assess ABA following testing cognitive flexibility in relatively young female rats. The data demonstrate that, contrary to expectations, cognitive inflexibility does not predispose to greater ABA weight loss, but instead, rats that performed better in the reversal learning task lost more weight in the ABA paradigm. Prior ABA exposure resulted in poorer learning of the task and reversal. An additional experiment demonstrated that rats that had been trained in reversal learning resisted weight loss in the ABA paradigm. The findings are important and are clearly presented. They have implications for anorexia nervosa both in terms of potentially identifying those at risk also in understanding the high rates of relapse.

Thanks for a great summary of the manuscript.

Reviewer #3 (Public Review):

Activity-based anorexia (ABA), which combines access to a running wheel and restricted access to food, is a most common paradigm used to study anorexic behavior in rodents. And yet, the field has been plagued by persistent questions about its validity as a model of anorexia nervosa (AN) in humans. This group's previous studies supported the idea that the ABA paradigm captures cognitive inflexibility seen in AN. Here they describe a fully automated touchscreen cognitive testing system for rats that makes it possible to ask whether cognitive inflexibility predisposes individuals to severe weight loss in the ABA paradigm. They observed that cognitive inflexibility was predictive of resistance to weight loss in the ABA, the opposite of what was predicted. They also reported reciprocal effects of ABA and cognitive testing on subsequent performance in the other paradigm. Prior exposure to the ABA decreased subsequent cognitive performance, while prior exposure to the cognitive task promoted resistance to the ABA. Based on these findings, the authors argue that the ABA model can be used to identify novel therapeutic targets for AN.

The strength of this manuscript is primarily as a methods paper describing a novel automated cognitive behavioral testing system that obviates the need for experimentalist handling and single housing, which can interfere with behavioral testing, and accelerate learning on the task. Together, these features make it feasible to perform longitudinal studies to ask whether cognitive performance is predictive of behavior in a second paradigm during adolescence, a peak period of vulnerability for many psychiatric disorders. The authors also used machine learning tools to identify specific behaviors during the cognitive task that predicted later susceptibility to the ABA paradigm. While the benefits of this system are clear, the rigor and reproducibility of experiments using this paradigm would be enhanced if the authors provided clear guidelines about which parameters and analyses are most useful. In their absence, the large amount of data generated can promote p-hacking.

The authors use their automated behavioral testing paradigm to ask whether cognitive inflexibility is a cause or consequence of susceptibility to ABA, an issue that cannot be addressed in AN. They provide compelling evidence that there are reciprocal effects of the two behavioral paradigms, but do not perform the controls needed to evaluate the significance of these observations. For example, the learning task involves sucrose consumption and food restriction, conditions that can independently affect susceptibility to the ABA. Similarly, the ABA paradigm involves exercise and restricted access to food, which can both affect learning.

In the Discussion, the authors hypothesize that the ABA paradigm produces cognitive inflexibility and argue that uncovering the underlying mechanism can be used to identify new therapeutic targets for AN. The rationale for their claim of translational relevance is undermined by the fact that the biggest effect of the ABA paradigm is seen in the pair discrimination task, and not reversal learning. This pattern does not fit clinical observations in AN.

In summary, the significance of this manuscript lies in the development of a new system to test cognitive function in rats that can be combined with other paradigms to explore questions of causality. While the authors clearly demonstrate that cognitive flexibility does not promote susceptibility to ABA, the experiments presented do not provide a compelling case that their model captures important features of the pathophysiology of AN.

We thank the reviewer for this detailed review and note that we have now both explicitly defined the most useful parameters for analyses from the novel touchscreen system as well as removed some comparisons that could be considered superfluous. We argue that the additional information provided by the machine learning analyses are, at this stage, exploratory, and rather than reveal independent descriptions of behavioural change in ABA exposed versus naïve rats this information will aid in the generation of hypotheses to be tested in future studies. Therefore, the figures pertaining to these analyses have now been provided as supplements to Figures 3 & 4 (Figure 3-figure supplement 3; Figure 4-figure supplements 3&4). We have also clarified our intention to explore possible behavioural differences using this technique in the methods and discussion.

We have also completed the essential control experiment, defined in the “essential revisions” section of this review, whereby we show only moderate impairments in reversal learning following a matched period of food restriction without rapid weight loss, suggesting that the substantial impairment seen following ABA exposure was not due to food restriction alone (see updated Figure 4 and supplements).

However, we do not agree with this reviewer “that the biggest effect of the ABA paradigm is seen in the pair discrimination task” and point to the outcomes of both reciprocal experiments.

In the first experiment, rats that went onto be susceptible or resistant to ABA did not differ on pairwise discrimination learning but specifically on performance at the reversal of reward contingencies (Figure 3B & E). Although this result was not in the hypothesised direction, this suggests that reversal learning specifically and not pairwise discrimination can differentiate those rats that go on to be susceptible to weight loss. We have included additional discussion in the text related to this finding (see line 490-497).

In the second experiment, it is clear by the number of ABA exposed rats that were unable to learn the reversal component even after being able to learn pairwise discrimination, that flexible learning is more impaired by ABA. While it is true that ABA exposed rats that were successful in learning the reversal task were slower to learn the pairwise discrimination component than naïve rats (Figure 4E), this was not related to their ability to learn the reversal task overall – with equivalent learning rates in pairwise discrimination to ABA exposed rats that failed to learn the reversal component (Figure 4G-I). The absence of significant differences between ABA exposed and naïve animals in Figure 4F relates to the fact that the large proportion of ABA exposed animals never reached performance criterion in the reversal phase of the task and therefore data from these animals could not be included in the figure. This is where the trials completed within each session becomes important for interpretation (i.e. Figure 4-figure supplement 1M-O), whereby ABA exposure caused impaired responding specifically within the reversal phase of the task. The results text has been updated to better reflect this critical point.

Overall, this suggests that the impairment in cognitive flexibility caused by ABA exposure was related both to an associative learning impairment (slower to learn PD than naïve animals) and an impairment in the integration of new and existing learning (failure to learn R1 in a large proportion of animals).

Author Response

Reviewer #1 (Public Review):

Weaknesses

1) I was curious as to how novel this setup is. Although I do not do head-fixed research myself, I thought there were already some open-source, relatively cheap systems available. I'm not sure how the current setup differs from those already available. Personally, even if this system involves only the wheel turning, as this is a truly operant response, that is novel enough for my liking.

The novelty of the system stems from the synergistic combination of functionality, the low-cost open source nature of the design, and the breadth of behavioral procedures the system is able to support. The use of a wheel as an operant response was adapted from the International Brain Laboratory rig which has been used extensively for visual discrimination tasks. We adapted this wheel design to make the response closer to lever pressing through the use of the wheel brake, which ensures that subjects have to rotate the wheel in discrete rotational bouts rather than continuously spinning the wheel and potentially disengaging and allowing the wheel to rotate independently. There are no examples of systems capable of delivering 5+ solutions within a behavioral session or conducting valence testing with a modification of real-time place preference without the cost and complexity associated with virtual reality. We believe that the combination of factors, the flexibility and scalability of the system makes OHRBETS a novel and useful system for diverse motivation and consumption behaviors in head-fixed mice.

2) It would be useful to have a bit more detail in the manuscript (not just on the GitHub link - in supplemental material perhaps?) on how to build such a system, just to get a sense of how difficult building such a system might be and how many components it has.

With this submission we have included detailed assembly instructions as a supplement to the main manuscript and added reference to the file within the methods section. We have also added details, including time estimates, to the methods section.

3) I wasn't sure how to feel about the comparisons across experimental set-ups in Figures 2 and 3. Usually, these sorts of comparisons are not considered statistically valid due to the many variables that differ between set-ups. However, I do see that the intent here is a bit different - i.e. is to show that despite all these alterations in variables the behavioural outputs are still highly correlated. However, without commenting on this intent, I did find these comparisons a little jarring to read.

Thank you for highlighting this. We have added in a justification for why we measured the consistency in behavior measured with each head-fixed system.

4) The only dataset I was not wholly convinced by was that in Figure 3 (real-time place preference and aversion). I think the authors have done the best job that they can of replicating such a procedure in a head-fixed mouse, but the head-fixed version is going to necessarily differ from the freely moving version in a fundamental way when the contextual cues and spatial navigation form part of the RTPT task. Giving a discrete cue, such as a tone, just is not a sufficient substitute for contextual cues, and I think the two types of task would engage fundamentally different brain cells and circuits (e.g. only the free-moving version is likely to engage place cells in the hippocampus).

To avoid confusion regarding the place component of the real-time place preference assay name, we have renamed the head-fixed assay for assessing valence to Wheel-Time Preference (WTP). We have also added a full paragraph to the discussion where we outline the differences in the task requirements and relevant neuronal circuits between the freely-moving RTPP and head-fixed WTP. We understand that the head-fixed task is not a perfect analog of the RTPP task, however based on the similarity in the resulting time spent in the stimulation chamber/zone we believe that the WTP is able to replicate the valence assessment that many in the field uses RTPP to measure. We believe that the WTP with OHRBETS opens up new possibilities for assessing preference in head-fixed mice and this justifies keeping the figure within the main manuscript.

To thoroughly address the potential confound of spatial information during the multi-spout experiment, we have added an additional supplemental figure (Figure 4- figure supplement 5) that depicts the proportion of trials with licking and added a paragraph to the discussion centered on the potential confound associated with learning the solution identity.

5) Personally, I found having the statistics in a separate file confusing.

Thank you for raising this concern. With our initial submission, we were concerned that including all of the statistics within the main text would make the paper difficult to read due to the extensive amount of statistics. With this submission, in addition to the statistics table, we have included statistics within the figure legends and main text where applicable.

6) Line 589-594. Suggesting the medial/lateral shell recording results mean that the medial shell 'tracks value, and the range of values during the multi-spout consumption of gradients of NaCl is greater than the range of values during multi-spout consumption of gradients of sucrose" seems to engage in circular logic to me. That is, the authors should use behavioural data to infer what the animal is experiencing and whether it is a change in value, and/or a greater change in value during NaCl vs. sucrose consumption, and only then should they make an inference about what the larger medial shell response means.

Thank you for identifying this potential site of confusion. To address this concern we have modified the language to better communicate our interpretation of the data.

“If we assume that the range of values is greater during multi-spout consumption of gradients of NaCl compared to gradients of sucrose, as indicated by a greater range in licking behavior (Figure 8- Figure Supplement 4), then the greater range of dopamine release in the NacShM could imply that dopamine release in this structure tracks value.”

Author Response

Reviewer #1 (Public Review):

Wang, Y. et al. investigated the role of TPL2 signaling in acute and chronic neuroinflammatory conditions using small molecule inhibitors and a TPL2 kinase-dead mutant mouse line. They find that TPL2 is upregulated by various brain-resident cells, including microglia, astrocytes, and endothelial cells, during neurodegenerative disease progression and following peripheral LPS injection. They show that upon pharmacological and genetic inhibition during acute LPS stimulation, pro-inflammatory cytokine concentration, microgliosis, and neuronal loss can be reversed. In chronic neuroinflammation, as seen in a tauopathy mouse model, the loss of TPL2 rescues reactive gliosis, immune cell infiltration, neurodegeneration, and cognitive health. Interestingly, TPL2 loss of function was not significantly beneficial in models of nerve injury and stroke. By analyzing their multiple sequencing datasets and those of other research teams, the authors find that TPL2 aids to upregulate transcripts for the DAM signature, immediate early genes, and astrocyte reactivity. These data build together to further emphasize the intricacy and importance of the immune component in neurodegeneration and other neuroinflammatory conditions.

The conclusions of this paper are mostly well supported by their data, but further confirmation of sequencing results and microglia intrinsic mechanisms need to be expanded.

1) In the discussion section, it will be important to highlight that TPL2 could also be directly contributing to tauopathy disease progression through its actions in brain-resident endothelial cells. They spend a lot of time characterizing the effects of TPL2 on in vitro microglial responses and do not adequately discuss the potential that their disease phenotypes in the tauopathy model have more to do with TPL2's ability to regulate BBB permeability or facets of endothelial biology. It will be important to highlight that there are various discrete cellular mechanisms (e.g. functions for TPL2 in microglia, endothelial cells, astrocytes, peripheral immune cells, etc.) that could be underlying the disease readouts seen in their global TPL2 kinase-dead mice. They should discuss this in the context of previous literature demonstrating roles for TPL2 in other non-microglial cell types (e.g. Nanou et al PMID: 34038728).

Thank you for this comment. We agree that while TPL2 is most highly expressed in microglia in the brain, TPL2 expression in endothelial cells and other cell types could potentially contribute to the disease. We have added discussion of this to the manuscript including discussion of the Nanou et al paper which raises the possibility that the TPL2-dependent infiltration of peripheral immune cells in TauP301S mice could be due to regulation of the BBB by TPL2 activity in endothelial cells. We also discuss potential roles for TPL2 in the various other cell types. In addition, we have now added characterization of cell-autonomous TPL2-dependent phenotypes in cultured astrocytes and have provided additional analysis of TPL2-dependent changes in endothelial cells in the scRNAseq experiment in TauP301S mice.

2) Hippocampal single-cell RNA sequencing led the authors to report that TLP2KD in the PS19 model of tauopathy reduced the number of T-cell and dendritic cell (DC) infiltrates into the brain. The authors should corroborate this finding with immunohistochemistry or flow cytometry to confirm the presence of changing CD4+, CD8+, and DC populations. Most notably, it is critical for them to enumerate the cell numbers in an effort to validate that there are indeed empirical, and not just proportional, reductions in these cell populations.

Thank you for the suggestion. We have performed immunohistochemistry to examine T cells in fixed brain tissue sections. We have included the data for T cell staining in Figure 5-figure supplement 2. We focused the IHC analysis on staining for CD8+ T cells based on the substantially greater abundance of CD8+ T cells compared to CD4+ T cells or DC in the single cell data (Figure 5C, Figure 5-figure supplement 5) and the availability of an antibody that worked well in our hands. These results corroborate the single cell data by empirically showing significantly increased numbers of T cells in TauP301S mice and significantly reduced numbers in the TauP301S x TPL2KD mice (Figure 5-figure supplement 2).

3) The authors concluded from Figure 3 that TPL2 plays a key role in in vivo microglia and astrocyte activation. Adding in an in vitro study, like those done in Figures 1, 2, and S4, that looks at a cell-autonomous role for TPL2 in astrocyte reactivity would strengthen this claim and rule out a microglial-independent pathway of TPL2 inflammation.

Thank you for the suggestion. To investigate the potential cell-autonomous role of TPL2 in astrocytes, we cultured primary mouse astrocyte and stimulated astrocytes with either LPS or cytokines, in the absence or presence of TPL2 inhibitor and measured stimulation induced changes in cytokine release and gene expression. Data are included in Figure 3-figure supplement 1 and the results are discussed in the manuscript. In contrast to the broader TPL2-dependence of cytokine release by cultured microglia only a much more restricted set of cytokines exhibited TPL2-dependence in cultured astrocytes. Furthermore, RT-qPCR analysis of TPL2-dependent activated astrocyte genes identified in the LPS in vivo study found much less TPL2-dependent activation in cultured astrocytes. We discuss that these results suggest that the TPL2-dependent astrocyte activation observed in vivo was probably largely contributed to indirectly by the function of TPL2 in microglia, but there was also potentially some contribution of cell-autonomous function of TPL2 in astrocytes.

4) Although the TPL2KD mouse line is a valuable tool to impair TPL2's function while retaining its expression, the researchers failed to comment on the potential effects a global mutation in TPL2 could have in their model systems. Peripheral immunological challenges, like their IP injections of LPS, could behave differently and affect the nervous system in a microglia-independent pathway if monocyte/macrophage signaling is also impaired.

We agree that during peripheral immunological challenges TPL2 could affect the nervous system in a microglia-independent manner. We have added this point to the discussion.

5) Oligodendrocytes and OPCs have comparable numbers of DEGs to astrocytes (Figure S11a). What is changing within their transcriptional profile?

In this manuscript we focused on TPL2-dependent DEGs in the Tauopathy model, which were all in microglia. We agree the TPL2-independent changes in the TauP301S mice in other cell types are also interesting. This data set has been uploaded to public data repository (GSE180041) and analysis of the changes in oligodendrocytes has been performed from this data set, as well as other disease models, in a recent publication: “Disease-associated oligodendrocyte responses across neurodegenerative diseases” (PMID: 36001972).

Author Response

Reviewer #1 (Public Review):

Strengths

This paper is well situated theoretically within the habit learning/OCD literature. Daily training in a motor-learning task, delivered via smartphone, was innovative, ecologically valid and more likely to assay habitual behaviors specifically. Daily training is also more similar to studies with non-humans, making a better link with that literature. The use of a sequential-learning task (cf. tasks that require a single response) is also more ecologically valid. The in-laboratory tests (after the 1 month of training) allowed the researchers to test if the OCD group preferred familiar, but more difficult, sequences over newer, simpler sequences.

The authors achieved their aims in that two groups of participants (patients with OCD and controls) engaged with the task over the course of 30 days. The repeated nature of the task meant that 'overtraining' was almost certainly established, and automaticity was demonstrated. This allowed the authors to test their hypotheses about habit learning. The results are supportive of the authors' conclusions.

We truly appreciate the positive assessment of referee 1, particularly the consideration that our study is theoretically strong and that ‘the results are supportive of the authors' conclusions’. This is an important external endorsement of our conclusions, contrasting somewhat with the views of referee 2.

Weaknesses

The sample size was relatively small. Some potentially interesting individual differences within the OCD group could have been examined more thoroughly with a bigger sample (e.g., preference for familiar sequences). A larger sample may have allowed the statistical testing of any effects due to medication status.

The authors were not able to test one criterion of habits, namely resistance to devaluation, due to the nature of the task

We agree with the reviewer that the proof of principle established in our study opens new avenues for research into the psychological and behavioral determinants of the heterogeneity of this clinical population. However, considering the study timeline and the pandemic constraints, a bigger sample was not possible. Our sample can indeed be considered small if one compares it with current online studies, which do not require in-person/laboratory testing, thus being much easier to recruit and conduct. However, given the nature of our protocol (with 2 demanding test phases, 1-month engagement per participant and the inclusion of OCD patients without comorbidities only) and the fact that this study also involved laboratory testing, we consider our sample size reasonable and comparable to other laboratory studies (typically comprising on average between 30-50 participants in each group).

This article is likely to be impactful -- the delivery of a task across 30 days to a patient group is innovative and represents a new approach for the study of habit learning that is superior to an inlaboratory approach.

An interesting aspect of this manuscript is that it prompts a comparison with previous studies of goal-directed/habitual responding in OCD that used devaluation protocols, and which may have had their effects due to deficits in goal-directed behavior and not enhanced habit learning per se.

Thank you for acknowledging the impact of our study, in particular the unique ability of our task to interrogate the habit system.

Reviewer #2 (Public Review):

In this study, the researchers employed a recently developed smartphone application to provide 30 days of training on action sequences to both OCD patients and healthy volunteers. The study tested learning and automaticity-related measures and investigated the effects of several factors on these measures. Upon training completion, the researchers conducted two preference tests comparing a learned and unlearned action sequences under different conditions. While the study provides some interesting findings, I have a few substantial concerns:

1) Throughout the entire paper, the authors' interpretations and claims revolve around the domain of habits and goal-directed behavior, despite the methods and evidence clearly focusing on motor sequence learning/procedural learning/skill learning. There is no evidence to support this framing and interpretation and thus I find them overreaching and hyperbolic, and I think they should be avoided. Although skills and habits share many characteristics, they are meaningfully distinguishable and should not be conflated or mixed up. Furthermore, if anything, the evidence in this study suggests that participants attained procedural learning, but these actions did not become habitual, as they remained deliberate actions that were not chosen to be performed when they were not in line with participants' current goals.

We acknowledge that the research on habit learning is a topic of current controversy, especially when it comes to how to induce and measure habits in humans. Therefore, within this context referee’s 2 criticism could be expected. Across disQnct fields of research, different methodologies have been used to measure habits, which represent relaQvely stereotyped and autonomous behavioral sequences enacted in response to a specific sQmulus without consideraQon, at the Qme of iniQaQon of the sequence, of the value of the outcome or any representaQon of the relaQonship that exists between the response and the outcome. Hence these are sQmulus-bound responses which may or may not require the implementaQon of a skill during subsequent performance. Behavioral neuroscienQsts define habits similarly, as sQmulus-response associaQons which are independent of reward or outcome, and use devaluaQon or conQngency degradaQon strategies to probe habits (Dickinson and Weiskrantz, 1985; Tricomi et al., 2009). Others conceptualize habits as a form of procedural memory, along with skills, and use motor sequence learning paradigms to invesQgate and dissect different components of habit learning such as acQon selecQon, execuQon and consolidaQon (Abrahamse et al., 2013; Doyon et al., 2003; Squire et al., 1993). It is also generally agreed that the autonomous nature of habits and the fluid proficiency of skills are both usually achieved with many hours of training or pracQce, respecQvely (Haith and Krakauer, 2018).

We consider that Balleine and Dezfouli (2019) made an excellent attempt to bring all these different criteria within a single framework, which we have followed. We also consider that our discussion in fact followed a rather cautious approach to interpretation solely in terms of goaldirected versus habitual control.

Referee 2 does not actually specify criteria by which they define habits and skills, except for asserting that skilled behavior is goal-directed, without mentioning what the actual goal of the implantation of such skill is in the present study: the fulfillment of a habit? We assume that their definition of habit hinges on the effects of devaluation, as a single criterion of habit, but which according to Balleine and Dezfouli (2019) is only 1 of their 4 listed criteria. We carefully addressed this specific criterion in our manuscript: “We were not, however, able to test the fourth criterion, of resistance to devaluation. Therefore, we are unable to firmly conclude that the action sequences are habits rather than, for example, goal-directed skills. Regardless of whether the trained action sequences can be defined as habits or goal-directed motor skills, it has to be considered…”. Therefore, we took due care in our conclusions concerning habits and thus found the referee’s comment misleading and unfair.

We note that our trained motor sequences did in fact fulfil the other 3 criteria listed by Balleine and Dezfouli (2019), unlike many studies employing only devaluation (e.g. Tricomi et al 2009; Gillan et al 2011). Moreover, we cited a recent study using very similar methodology where the devaluation test was applied and shown to support the habit hypothesis (Gera et al., 2022).

Whether the initiation of the trained motor sequences in experiment 3 (arbitration) are underpinned by an action-outcome association (or not) has no bearing on whether those sequences were under stimulus-response control after training (experiment 1). Transitions between habitual and goal-directed control over behavior are quite well established in the experimental literature, especially when choice opportunities become available (Bouton et al (2021), Frölich et al (2023), or a new goal-directed schemata is recruited to fulfill a habit (Fouyssac et al, 2022). This switching between habits and goal-directed responding may reflect the coordination of these systems in producing effective behavior in the real world.

• Fouyssac M, Peña-Oliver Y, Puaud M, Lim NTY, Giuliano C, Everitt BJ, Belin D. (2021).Negative Urgency Exacerbates Relapse to Cocaine Seeking After Abstinence. Biological Psychiatry. doi: 10.1016/j.biopsych.2021.10.009

• Frölich S, Esmeyer M, Endrass T, Smolka MN and Kiebel SJ (2023) Interaction between habits as action sequences and goal-directed behavior under time pressure. Front. Neurosci. 16:996957. doi: 10.3389/fnins.2022.996957

• Bouton ME. 2021. Context, attention, and the switch between habit and goal-direction in behavior. Learn Behav 49:349– 362. doi:10.3758/s13420-021-00488-z

2) Some methodological aspects need more detail and clarification.

3) There are concerns regarding some of the analyses, which require addressing.

We thank referee 2 for their detailed review of the methods and analyses of our study and for the helpful feedback, which clearly helps improve our manuscript. We will clarify the methodological aspects in detail and conduct the suggested analysis. Please see below our answers to the specific points raised.

Introduction:

4) It is stated that "extensive training of sequential actions would more rapidly engage the 'habit system' as compared to single-action instrumental learning". In an attempt to describe the rationale for this statement the authors describe the concept of action chunking, its benefits and relevance to habits but there is no explanation for why sequential actions would engage the habit system more rapidly than a single-action. Clarifying this would be helpful.

We agree that there is no evidence that action sequences become habitual more readily than single actions, although action sequences clearly allow ‘chunking’ and thus likely engage neural networks including the putamen which are implicated in habit learning as well as skill. In our revised manuscript we will instead state: “we have recently postulated that extensive training of sequential actions could be a means for rapidly engaging the ‘habit system’ (Robbins et al., 2019)]”

5) In the Hypothesis section the authors state: “we expected that OCD patients... show enhanced habit attainment through a greater preference for performing familiar app sequences when given the choice to select any other, easier sequence”. I find it particularly difficult to interpret preference for familiar sequences as enhanced habit attainment.

We agree that choice of the familiar response sequence should not be a necessary criterion for habitual control although choice for a familiar sequence is, in fact, not inconsistent with this hypothesis. In a recent study, Zmigrod et al (2022) found that 'aversion to novelty' was a relevant factor in the subjective measurement of habitual tendencies. It should also be noted that this preference was present in patients with OCD. If one assumes instead, like the referee, that the familiar sequence is goal-directed, then it contravenes the well-known 'egodystonia' of OCD which suggests that such tendencies are not goal-directed.

To clarify our hypothesis, we will amend the sentence to the following: “Finally, we expected that OCD patients would generally report greater habits, as well as attribute higher intrinsic value to the familiar app sequences manifested by a greater preference for performing them when given the choice to select any other, easier sequence”.

A few notes on the task description and other task components:

6) It would be useful to give more details on the task. This includes more details on the time/condition of the gradual removal of visual and auditory stimuli and also on the within practice dynamic structure (i.e., different levels appear in the video).

These details will be included in the revised manuscript. Thank you for pointing out the need for further clarification of the task design.

7) Some more information on engagement-related exclusion criteria would be useful (what happened if participants did not use the app for more than one day, how many times were allowed to skip a day etc.).

This additional information will be added to the revised manuscript. If participants omitted to train for more than 2 days, the researcher would send a reminder to the participant to request to catch up. If the participant would not react accordingly and a third day would be skipped, then the researcher would call to understand the reasons for the lack of engagement and gauge motivation. The participant would be excluded if more than 5 sequential days of training were missed. Only 2 participants were excluded given their lack of engagement.

8) According to the (very useful) video demonstrating the task and the paper describing the task in detail (Banca et al., 2020), the task seems to include other relevant components that were not mentioned in this paper. I refer to the daily speed test, the daily random switch test, and daily ratings of each sequence's enjoyment and confidence of knowledge.

If these components were not included in this procedure, then the deviations from the procedure described in the video and Banca al. (2020) should be explicitly mentioned. If these components were included, at least some of them may be relevant, at least in part, to automaticity, habitual action control, formulation of participants' enjoyment from the app etc. I think these components should be mentioned and analyzed (or at least provide an explanation for why it has been decided not to analyze them).

This is also true for the reward removal (extinction) from the 21st day onwards which is potentially of particular relevance for the research questions.

The task procedure was indeed the same as detailed in Banca et al., 2020. We did not include these extra components in this current manuscript for reasons of succinctness and because the manuscript was already rather longer than a common research article, given that we present three different, though highly inter-dependent, experiments in order to answer key interrelated questions in an optimal manner. However, since referee 2 considers this additional analysis to be important, we will be happy to include it in the supplementary material of the revised manuscript.

Training engagement analysis:

9)I find referring to the number of trials including successful and unsuccessful trials as representing participants "commitment to training" (e.g. in Figure legend 2b) potentially inadequate. Given that participants need at least 20 successful trials to complete each practice, more errors would lead to more trials. Therefore, I think this measure may mostly represent weaker performance (of the OCD patients as shown in Figure 2b). Therefore, I find the number of performed practice runs, as used in Figure 2a (which should be perfectly aligned with the number of successful trials), a "clean" and proper measure of engagement/commitment to training.

We acknowledge referee’s concern on this matter and agree to replace the y-axis variable of Figure 2b to the number of performed practices (thus aligning with Figure 2a). This amendment will remove any potential effect of weaker performance on the engagement measurement and will provide clearer results.

10) Also, to provide stronger support for the claim about different diurnal training patterns (as presented in Figure 2c and the text) between patients and healthy individuals, it would be beneficial to conduct a statistical test comparing the two distributions. If the results of this test are not significant, I suggest emphasizing that this is a descriptive finding.

We will conduct the statistical test and report accordingly.

Learning results:

11) When describing the Learning results (p10) I think it would be useful to provide the descriptive stats for the MT0 parameter (as done above for the other two parameters).

Thank you for pointing this out. The descriptive stats for MT0 will be added to the revised version of the manuscript.

12) Sensitivity of sequence duration and IKI consistency (C) to reward:

I think it is important to add details on how incorrect trials were handled when calculating ∆MT (or C) and ∆R, specifically in cases where the trial preceding a successful trial was unsuccessful. If incorrect trials were simply ignored, this may not adequately represent trial-by-trial changes, particularly when testing the effect of a trial's outcome on performance change in the next trial.

This is an important question. Our analysis protocol was designed to ensure that incorrect trials do not contaminate or confound the results. To estimate the trial-to-trial difference in ∆MT (or C) and ∆R, we exclusively included pairs of contiguous trials where participants achieved correct performance and received feedback scores for both trials. For example, if a participant made a performance error on trial 23, we did not include ∆R or ∆MT estimates for the pairs of trials 23-22 and 24-23. Instead of excluding incorrect trials from our analyses, we retained them in our time series but assigned them a NaN (not a number) value in Matlab. As a result, ∆R and ∆MT was not defined for those two pairs of trials. Similarly for C. This approach ensured that our analyses are not confounded by incremental or decremental feedback scores between noncontiguous trials. In the past, when assessing the timing of correct actions during skilled sequence performance, we also considered events that were preceded and followed by correct actions. This excluded effects such as post-error slowing from contaminating our results (Herrojo Ruiz et al., 2009, 2019). Therefore, we do not believe that any further reanalysis is required.

• Ruiz MH, Jabusch HC, Altenmüller E. Detecting wrong notes in advance: neuronal correlates of error monitoring in pianists. Cerebral cortex. 2009 Nov 1;19(11):2625-39.

• Bury G, García-Huéscar M, Bhattacharya J, Ruiz MH. Cardiac afferent activity modulates early neural signature of error detection during skilled performance. NeuroImage. 2019 Oct 1;199:704-17.

13) I have a serious concern with respect to how the sensitivity of sequence duration to reward is framed and analyzed. Since reward is proportional to performance, a reduction in reward essentially indicates a trial with poor performance, and thus even regression to the mean (along with a floor effect in performance [asymptote]) could explain the observed effects. It is possible that even occasional poor performance could lead to a participant demonstrating this effect, potentially regardless of the reward. Accordingly, the reduced improvement in performance following a reward decrease as a function of training length described in Figure 5b legend may reflect training-induced increased performance that leaves less room for improvement after poor trials, which are no longer as poor as before. To address this concern, controlling for performance (e.g., by taking into consideration the baseline MT for the previous trial) may be helpful. If the authors can conduct such an analysis and still show the observed effect, it would establish the validity of their findings."

Thank you for raising this point. Figure 5b illustrates two distinct effects of reward changes on behavioral adaptation, which are expected based on previous research.

I. Practice effects: Firstly, we observe that as participants progress across bins of practice, the degree of improvement in behavior (reflected by faster movement time, MT) following a decrease in reward (∆R−) diminishes, consistent with our expectations based on previous work. Conversely, we found that ∆MT does not change across bins of practices following an increase in reward (∆R+). We appreciate the reviewer's suggestion regarding controlling for the reference movement time (MT) in the previous trial when examining the practice effect in the p(∆T|∆R−) and p(∆T|∆R+) distributions. In the revised manuscript, we will conduct the proposed control analysis to better understand whether the sensitivity of MT to score decrements changes across practice when normalising MT to the reference level on each trial. But see below for a preliminary control analysis.

II. Asymmetry of the effect of ∆R− and ∆R+ on performance: Figure 5b also depicts the distinct impact of score increments and decrements on behavioural changes. When aggregating data across practice bins, we consistently observed that the centre of the p(∆T|∆R−) distribution was smaller (more negative) than that of p(∆T|∆R+). This suggests that participants exhibited a greater acceleration following a drop in scores compared to a relative score increase, and this effect persisted throughout the practice sessions. Importantly, this enhanced sensitivity to losses or negative feedback (or relative drops in scores) aligns with previous research findings (Galea et al., 2015; Pekny et al., 2014; van Mastrigt et al., 2020).

We have conducted a preliminary control analysis to exclude the potential impact that reference movement time (MT) values could have on our analysis. We have assessed the asymmetry between behavioural responses to ∆R− and ∆R+ using the following analysis: We estimated the proportion of trials in which participants exhibited speed-up (∆T < 0) or slow-down (∆T > 0) behaviour following ∆R− and ∆R+ across different practice bins (bins 1 to 4). By discretising the series of behavioural changes (∆T) into binary values (+1 for slowing down, -1 for speeding up), we can assess the type of changes (speed-up, slow-down) without the absolute ∆T or T values contributing to our results. We obtained several key findings:

• Consistent with expectations (sanity check), participants exhibited more instances of speeding up than slowing down across all reward conditions.

• Participants demonstrated a higher frequency of speeding up following ∆R− compared to ∆R+, and this asymmetry persisted throughout the practice sessions (greater proportion of -1 events than +1 events). 53% events were speed-up events in the in the p(∆T|∆R+) distribution for the first bin of practices, and 55% for the last bin. Regarding p(∆T|∆R-), there were 63% speed-up events throughout each bin of practices, with this proportion exhibiting no change over time.

• Accordingly, the asymmetry of reward changes on behavioural adaptations, as revealed by this analysis, remained consistent across the practice bins.

Thus, these preliminary findings provide an initial response to referee 2 and offer valuable insights into the asymmetrical effects of positive/negative reward changes on behavioural adaptations. We plan to include these results in the revised manuscript, as well as the full control analysis suggested by the referee. We will further expand upon their interpretation and implications.

14) Another way to support the claim of reward change directionality effects on performance (rather than performance on performance), at least to some extent, would be to analyze the data from the last 10 days of the training, during which no rewards were given (pretending for analysis purposes that the reward was calculated and presented to participants). If the effect persists, it is less unlikely that the effect in question can be attributed to the reward dynamics.

The reviewer’s concern is addressed in the previous quesQon. Also, this analysis would not be possible because our Gaussian fit analyses use the Qme series of conQnuous reward scores, in which ∆R− or ∆R+ are embedded. These events cannot be analyzed once reward feedback is removed because we do not have behavioral events following ∆R− or ∆R+ anymore.

15) This concern is also relevant and should be considered with respect to the sensitivity of IKI consistency (C) to reward. While the relationship between previous reward/performance and future performance in terms of C is of a different structure, the similar potential confounding effects could still be present.

We will conduct this analysis for the revised manuscript, similarly to the control analysis suggested by referee 2 on MT. Our preliminary control analysis, as explained above, suggests that the fundamental asymmetry in the effect of ∆R+ and ∆R+ on behavioral changes persists when excluding the impact of reference performance values in our Gaussian fit analysis.

16) Another related question (which is also of general interest) is whether the preferred app sequence (as indicated by the participants for Phase B) was consistently the one that yielded more reward? Was the continuous sequence the preferred one? This might tell something about the effectiveness of the reward in the task.

We have now conducted this analysis. There is in fact no evidence to conclude that the continuously rewarded sequence was the preferred one. The result shows that 54.5% of HV and 29% of the OCD sample considered the continuous sequence to be their preferred one. Of note, this preference may not necessarily be linked to the trial-by-trial reward sensitive analysis. The latter assesses how learning may be affected by reward. The overall preference may be influenced by many other factors, such as, for example, the aesthetic appeal of particular combinations of finger movements.

Regarding both experiments 2 and 3:

17) The change in context in experiment 2 and 3 is substantial and include many different components. These changes should be mentioned in more detail in the Results section before describing the results of experiments 2 and 3.

Following referee’s advice, we will move these details (currently written in the Methods section) to the Results section, when we introduce Phase B and before describing the results of experiments 2 and 3.

Experiment 2:

18) In Experiment 2, the authors sometimes refer to the "explicit preference task" as testing for habitual and goal-seeking sequences. However, I do not think there is any justification for interpreting it as such. The other framings used by the authors - testing whether trained action sequences gain intrinsic/rewarding properties or value, and preference for familiar versus novel action sequences - are more suitable and justified. In support of the point I raised here, assigning intrinsic rewarding properties to the learned sequences and thereby preferring these sequences can be conceptually aligned with goal-directed behavior just as much as it could be with habit.

We clearly defined the theoretical framing of experiment 2 as a test of whether trained action sequences gain intrinsic value and we are pleased to hear that the referee agrees with this framing. If the referee is referring to the paragraph below (in the Discussion), we actually do acknowledge within this paragraph that a preference for the trained sequences can either be conceptually aligned with a habit OR a goal-directed behavior.

“On the other hand, we are describing here two potential sources of evidence in favor of enhanced habit formation in OCD. First, OCD patients show a bias towards the previously trained, apparently disadvantageous, action sequences. In terms of the discussion above, this could possibly be reinterpreted as a narrowing of goals in OCD (Robbins et al., 2019) underlying compulsive behavior, in favor of its intrinsic outcomes”

This narrowing of goals model of OCD refers to a hypothetically transiQonal stage of compulsion development driven by behavior having an abnormally strong, goal-directed nature, typically linked to specific values and concerns.

If the referee is referring to the penulQmate sentence of hypothesis secQon, this has been amended in response to Q5. We cannot find any other possible instances in this manuscript stating that experiment 2 is a test of habitual or goal-directed behavior.

Experiment 3:

19) Similar to Experiment 2, I find the framing of arbitration between goal-directed/habitual behavior in Experiment 3 inadequate and unjustified. The results of the experiment suggest that participants were primarily goal-directed and there is no evidence to support the idea that this reevaluation led participants to switch from habitual to goal-directed behavior.

Also, given the explicit choice of the sequence to perform participants had to make prior to performing it, it is reasonable to assume that this experiment mainly tested bias towards familiar sequence/stimulus and/or towards intrinsic reward associated with the sequence in value-based decision making.

This comment is aligned with (and follows) the referee’s criticism of experiment 1 not achieving automatic and habitual actions. We have addressed this matter above, in response 1 to Referee 2.

Mobile-app performance effect on symptomatology: exploratory analyses:

20) Maybe it would be worth testing if the patients with improved symptomatology (that contribute some of their symptom improvement to the app) also chose to play more during the training stage.

We have conducted analysis to address this relevant question. There is no correlation between the YBOCS score change and the number of total practices, meaning that the patients who improved symptomatology post training did not necessarily chose to play the app more during the training stage (rs = 0.25, p = 0.15). Additionally, we have statistically compared the improvers (patients with reduced YBOCS scores post-training) and the non-improvers (patients with unchanged or increased YBOCS scores post-training) in their number of app completed practices during the training phase and no differences were observed (U = 169, p = 0.19).

Discussion:

21) Based on my earlier comments highlighting the inadequacy and mis-framing of the work in terms of habit and goal-directed behavior, I suggest that the discussion section be substantially revised to reflect these concerns.

We do not agree that the work is either "inadequate or mis-framed" and will not therefore be substantially revising the Discussion. We will however clarify further the interpretation we have made and make explicit the alternative viewpoint of the referee. For example, we will retitle experiment 3 as “Re-evaluation of the learned action sequence: possible test of goal/habit arbitration” to acknowledge the referee’s viewpoint as well as our own interpretation.

22) In the sentence "Nevertheless, OCD patients disadvantageously preferred the previously trained/familiar action sequence under certain conditions" the term "disadvantageously" is not necessarily accurate. While there was potentially more effort required, considering the possible presence of intrinsic reward and chunking, this preference may not necessarily be disadvantageous. Therefore, a more cautious and accurate phrasing that better reflects the associated results would be useful.

We recognize that the term "disadvantageously" may be semantically ambiguous for some readers and therefore we will remove it.

Materials and Methods:

23) The authors mention: "The novel sequence (in condition 3) was a 6-move sequence of similar complexity and difficulty as the app sequences, but only learned on the day, before starting this task (therefore, not overtrained)." - for the sake of completeness, more details on the pre-training done on that day would be useful.

Details of the learning procedure of the novel sequence (in condition 3, experiment 3) will be provided in the methods of the revised version of the manuscript.

Minor comments:

24) In the section discussing the sensitivity of sequence duration to reward, the authors state that they only analyzed continuous reward trials because "a larger number of trials in each subsample were available to fit the Gaussian distributions, due to feedback being provided on all trials." However, feedback was also provided on all trials in the variable reward condition, even though the reward was not necessarily aligned with participants' performance. Therefore, it may be beneficial to rephrase this statement for clarity.

We will follow this referee’s advice and will rephrase the sentence for clarity.

25) With regard to experiment 2 (Preference for familiar versus novel action sequences) in the following statement "A positive correlation between COHS and the app sequence choice (Pearson r = 0.36, p = 0.005) further showed that those participants with greater habitual tendencies had a greater propensity to prefer the trained app sequence under this condition." I find the use of the word "further" here potentially misleading.

The word "further" will be removed.

Author Response:

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

Thank you for considering our manuscript “An Unexpected Role of Neutrophils in Clearing Apoptotic Hepatocytes In Vivo". We also thank the referees for their review. We have addressed their comments in detail and added new data to buttress our conclusions.

Reviewer #1 (Public Review):

This study by Cao et al. demonstrates role of Neutrophil in clearing apoptotic hepatocytes by directly burrowing into the apoptotic hepatocytes and ingesting the effete cells from inside without causing inflammation. The authors applied intravital microscopy, Immunostaining and electron microscopy to visualize perforocytosis of neutrophil in hepatocytes. They also found that neutrophil depletion impairs the clearance of apoptotic hepatocytes causing impaired liver function and generation of autoantibodies, implying a role of defective neutrophil- mediated clearance of apoptotic cells in Autoimmune Liver disease. The experiments were well designed and conducted, the results were reasonably interpreted, and the manuscript was clearly written with logical inputs.

Thank you for your comments.

One weak point is that the signals/mechanisms that determine why neutrophil specifically target apoptotic hepatocytes in liver and no other organs or cells is not clearly understood.

We are still studying why neutrophils selectively phagocytose hepatocytes but not HUVEC or 293 cells. We have some intriguing preliminary data so far showing that apoptotic 293 cells have no significant increase of IL-1β production as compared with their nonapoptotic controls; both apoptotic 293 cells and HUVECs do not have increased surface selectin proteins (new Fig. S3C).

Reviewer #2 (Public Review):

[…] By examination of HE-stained, noncancerous liver tissue sections from patients with hepatocellular carcinoma and hepatic hemangioma, the authors observed that cells with neutrophil nuclear morphology were inside apoptotic hepatocytes. The authors also further characterized this observation by staining the sections with neutrophil and apoptosis markers. In addition, the authors observed the same phenomena in mouse livers using intravital microscopy, which also recorded the time course of the disappearance of a neutrophil-associated apoptotic cell. The author went on further characterization of neutrophil-mediated efferocytosis of cultured hepatic cells in vitro and demonstrated the process was specific for apoptotic hepatic cells, but not HEK293 or endothelial cells. The in vitro system was then used to characterize the molecular bases for neutrophil-mediated efferocytosis of apoptotic hepatic cells. The evidence was provided to suggest that IL1b and IL-8 released from and selectins upregulated in apoptotic hepatic cells were important. Importantly, the authors used two methods to deplete the neutrophils and showed that the neutrophil depletion increased apoptotic cells in livers. Finally, the authors showed that neutrophil depletion caused defects in liver function parameters. At the end, the authors presented evidence to suggest that AIL disease may be due to defective neutrophils that fail to perform "perforocytosis."

Thank you for your comments.

Point #1. Although the evidence in its totality indicates that neutrophils burrow into apoptotic hepatocytes, the significance of this "perforocytosis" phenomenon and the circumstances under which it may occur remain to be better defined. In both neutrophil depletion models, the TNUEL-positive cells were not definitively identified rather than assuming they were hepatocytes.

Anatomically, the apoptotic hepatocytes are randomly distributed in the hepatic plate from the central vein to the portal region (please refer to the image below: hematoxylin staining of liver tissues, black arrowhead indicates perforocytosis sites).

Author response image 1.

Histologically, the structure of liver/hepatic lobe are well defined, and the cell types in the livers are easy to histologically identify based on their location, morphology and the relationship to hepatic plate and sinusoid. In addition, the hepatocytes are well known for its rich cytoplasmic components, cellular connection and prominent large round nucleus. Thus, hepatocytes are very easy to identify even without using specific molecular markers such as E-cadherin or albumin. Based on these characteristics, the TUNEL positive cells that we displayed in Fig. 5A are apoptotic hepatocytes.

Point #2. In addition, there are discrepancies in the number of neutrophils and apoptotic cells in mouse liver studies; Fig. 2a WT (many neutrophils; locations unclear) vs Fig. 5A Ctr (a few neutrophils that appear in or near a vessel), and Fig. 2a DTR (a few apoptotic cells) vs Fig. 5A Depletion (many apoptotic cells).

In response, Fig. 2A demonstrates a larger area of the mouse liver (bar, 100 µm), while Fig. 5A exhibits a relatively small area of the liver sample (bars, 20 µm for Ctrl and 15 µm for DTR). Similarly, apoptotic cells in Fig. 2A DTR need to zoom in to quantify. We apologize for the confusion, and we did quantify the apoptotic cells in Fig.2A WT vs DTR (see the bar graph next to the images in Fig. 2A).

Point #3. Importantly, Fig 5a Ctrl, which is presumably a section from a mouse without any surgical treatment or without inflammation, the sole TUNNEL signal does not appear to be associated with neutrophils. Does this mean that "perforocytosis" primarily occurs in inflamed livers (Of note, human liver samples in Fig 1 are from patient with tumors. There should be inflammation in the livers of these patients).

In Fig 5A Ctrl, the TUNEL signal indicates apoptotic hepatocytes. The neutrophils (stained with anti-NE antibody, red) are associated with the apoptotic hepatocyte (Fig. 5A). We observed that perforocytosis primarily occurs in normal noninflamed livers.

Human liver samples in Fig 1 are from patient with tumors, hence it is possible that neutrophil burrowing is somehow associated with cancerous/inflammatory livers as the reviewer pointed out. This possibility was ruled out based on our method of sample preparation and experimental results themselves.

1) Both noncancerous and cancerous liver samples were sliced based on the anatomical appearance of normal and cancer tissues (differences were rather easy to identify, and these samples were prepared by highly experienced pathologists from the Liver Cancer Center of Zhongshan Hospital, Shanghai). Furthermore, the results were confirmed by determining whether the surrounding tissue contained microlesions characteristic of metastatic tumors. We only counted apoptotic hepatocytes in noncancerous regions having normal liver lobes and morphologically normal hepatocytes, plates, sinusoid and Kupffer cells. We also excluded hepatoma, chronic inflammatory regions, and necrotic regions.

2) We did not observe recruitment of neutrophils into apoptotic HCC cells, indicating that the clearance of apoptotic cancer cells was not mediated by neutrophils (unpublished observations).

3) It is hard for us to obtain normal human liver samples; however, we did study samples from patients with liver hemangioma characterized by aberrant vasculature in livers but with normal liver functions and the structure of hemangioma livers that we analyzed are nearly identical to a healthy liver in histology (these liver samples contained no cancerous regions and there was no apparent cirrhosis or inflammation). And here we obtained similar results (these are shown in Fig. 1B; a total of 40 apoptotic hepatocytes were examined).

4) Our data from normal mouse livers, isolated primary cells (hepatocytes and neutrophils) and cell lines (NCTC and HL60) all confirmed the central findings in this paper (Fig. 2, 3).

Point #4. The data on human AIL patient neutrophils raises more questions: how many AIL patients have been examined? Do these AIL neutrophils lack IL1, IL8 receptors, and/or selectin ligands? Are there increases in apoptotic hepatocytes in AIL patients?

In response, we have analyzed 16 AIL patient samples (see table below).

Author response table 1.

We performed microarray assay to screen the differential gene expression of neutrophils from normal and liver autoimmune patients. We have identified that IL-1β receptor, IL1R1 and selectin binding protein, P- selectin glycoprotein ligand 1 (PSGL-1) were all decreased in neutrophils from the AIL patients (new Fig 7D). These findings are consistent with our observations using cells and mouse models.

Point #5. Additionally, the overall numbers of apoptotic cells even in the absence of neutrophils are rare; thus, it is questionable that such rarity of apoptotic cells can cause significant AIL phenotypes.

We quantified apoptotic liver cells in percentages instead of overall numbers (Fig. 5, we were not able to precisely calculate the overall numbers, which could be large since billions of cells undergoing apoptosis daily). Depletion of neutrophils increased the percentage of apoptotic cells about 5-6-fold in livers, and we observed the generation of autoantibodies (Fig. 6).

Reviewer #1 (Recommendations For The Authors):

This study by Cao et al. was well designed and conducted, the results were reasonably interpreted, and the manuscript was clearly written with logical inputs.

It would further gain the significance of this study if authors could address the following questions:

1.  What are the mechanisms/ signals that prevents AIL Liver neutrophils from burrowing into hepatocytes?

We have identified that IL-1β receptor, IL1R1 and selectin binding protein, P-selectin glycoprotein ligand 1 (PSGL-1) were all decreased in neutrophils from the AIL patients (new Fig 7D).

2.  Have authors looked if autoantigens expressed on hepatocytes, which are often found in autoimmune liver disease trigger signaling events that activate neutrophils to burrow?

Thank you for the comment, we have not examined autoantigens expressed in hepatocytes and plan to carry out this research as suggested.

3.  Is perforocytosis observed in apoptotic hepatocytes induced by different agents like LPS, TNF-a , rapamycin, alcohol etc?

We did not observe perforocytosis in LPS or TNF-a treated hepatocytes. One possible reason is that LPS or TNF-a we used induced massive necrosis instead of apoptosis. Howere, we did observe neutrophil perforocytosis in FasL-induced apoptotic hepatocytes (unpublished observations).

Reviewer #2 (Recommendations For The Authors):

In addition to the questions raised in the "Public review" section, the authors are also recommended to address the following issues:

1) Why is CD11b+ not associated with the apoptotic sites as neutrophils express CD11b

We have co-immunostained human liver samples with CD11b antibody (from Abcam: ab133357) and MPO antibody (from R&D: AF3667) and observed that tissue infiltrating neutrophils in livers have low to undetectable levels of CD11b expression (please refer the image below; white arrowheads point to neutrophils). Few CD11b+ cells in liver tissues express MPO (the CD11b+ cells are mostly macrophages, unpublished observations).

Based on these data, we conclude that CD11b is hardly expressed in neutrophils inside livers.

Author response image 2.

2) Can TUNEL signals in Fig. S1C be from apoptotic neutrophils?

In response, the fragmentation of nucleus is a hallmark of apoptosis hence TUNEL staining will uniformly label all fragmented parts of apoptotic nucleus. The nucleus of NE+ neutrophils are not labelled by TUNEL staining in Fig. S1C. The TUNEL+ nuclear fragments seen inside neutrophils are nuclear debris of apoptotic hepatocytes phagocytosed by neutrophils (Fig. S1C).

3) The Fig 2B experiment may be done with induced apoptosis so that neutrophil burrowing steps may be recorded from the very beginning and a better time course for the entire process can be assessed.

Thank you for the suggestions, we had tried many times with various conditions, yet still had no success to capture the very beginning of perforocytosis in vivo. We are continuing to work on this.

4) In "we found thatU937 cells exhibited much lower phagocytosis of apoptotic NCTC cells than did HL60 cells (Fig. S2B, C)," the citation should be only S2C

Thank you for pointing this out, we have corrected this in the manuscript.

5) Both neutrophil depletion models cause neutrophil death, which may complicate the interpretation of the liver function and AIL disease phenotypes. A neutropenic model such as G-CSFR−/− or Cebpe-/- mice may be used to avoid the caveat of antibody/DTR-dependent depletion models.

Thank you for this thoughtful suggestion. We have also induced AIL phenotypes in mice by using α- Galcer. α-Galcer did not cause neutrophil death but impaired neutrophil perforocytosis and futher generated AIL phenotypes in mice (unpublished observations). We plan to perform the simiarl experiments in G-CSFR−/− or Cebpe−/− mice as the reviewer suggested.

6) RNAi silencing experiments need additional controls for off-target effects

These RNAi silencing constructs were purchased from Santa Cruz Biotechnology and the off-target effects have been tested by the company. No significant off-target effects have been detected according to the manufacture report.

Author Response

Joint Public Review

The molecular composition of synaptic vesicles (SVs) has been defined in substantial detail, but the function of many SV-resident proteins are still unknown. The present study focused on one such protein, the 'orphan' SV-resident transporter SLC6A17. By utilizing sophisticated and extensive mouse genetics and behavioral experiments, the authors provide convincing support for the notion that certain SLC6A17 variants cause intellectual disability (ID) in humans carrying such genetic variations. This is an important and novel finding. Furthermore, the authors propose, based on LCMS analyses of isolated SVs, that SLC6A17 is responsible for glutamine (Gln) transport into SVs, leading to the provocative idea that Gln functions as a neurotransmitter and that deficits in Gln transport into SVs by SLC6A17 represents a key pathogenetic mechanism in human ID patients carrying variants of the SLC6A17 gene.

This latter aspect of the present paper is not adequately supported by the experimental evidence so that the main conceptual claims of the study appear insufficiently justified at this juncture. Key weaknesses are as follows:

A) Detection of Gln, along with classical neurotransmitters such as glutamate, GABA, or ACh, in isolated SV fractions does not prove that Gln is transported into SVs by active transport. Gln is quite abundant in extracellular compartments. Its appearance in SV samples can therefore also be explained by trapping in SVs during endocytosis, presence in other - contaminating - organelles, binding to membrane surfaces, and other processes. Direct assays of Gln uptake into SVs, which have the potential to stringently test key postulates of the authors, are lacking.

We have conducted multiple control experiments to exclude the possibility of contamination.

1). Western blot analysis of SLC6A17-HA immunoisolation (Figure 4D and Figure 4—figure supplement 1) has shown that this faction contained little other organelles and membranes. These results are strong argument that contaminations in our isolated fraction were in very low level.

2). We then examined the proportion of SLC6A17 localized SVs through quantifying the co-localization of Syp and SLC6A17 by anti-Syp immunoisolation in Slc6a17-2A-HA-iCre mice. We found that SLC6A17 is predominately localized on SVs (with 98.7% compared with classical SV marker, Author response image 1A). This further showed that immunoisolated SLC6A17 fraction was mainly composed of SVs.

3). We also analyzed other SV marker proteins such as Syt1 and Syb2 for IP-LC-MS, all results supported Gln enrichment (Author response image 1B).

4). Importantly, immunoisolation of the SLC6A17P633R-HA protein, which caused SLC6A17 mislocalization away from the SVs (Figure 3B and Figure 3—figure supplement 1C, D), showed no Gln enrichment (Author response image 1C).

5). Moreover, immunoisolation of AAV-PHP.eb overexpressed cytoplasmic membrane Gln transporter SLC38A1-HA did not show Gln enrichment (Author response image 1D).

6). We also tested whether trafficking organelles such as the lysosome could enrich Gln. As is shown in Author response image 1E, immunoisolation of AAV-PHP.eb overexpressed TMEM192-HA did not show Gln enrichment. For active transport, we tested the effects of proton dissipator FCCP, v-ATPase inhibitor NEM and ΔpH dissipator nigercin. As is shown in Author response image 1F, 1G, Gln level was reduced by these inhibitors, supporting active transport of Gln.

Author response image 1.

Control experiments to test for contamination. A. Anti-Syp immunoisolation in Slc6a17-2A-HA-iCre mice. B. Quantification of Gln level in anti-Syt1 and anti-Syb2 immunoisolated fraction. C. Anti-HA immunoisolation in SLC6A7-2A-HA and anti-Slc6a17P633R mice. D. Anti-HA immunoisolation in AAV-PHP.eb-hSyn-SLC38A1-HA overexperssion mice. E. Anti-HA immunoisolation in AAV-PHP.eb-hSyn-TMEM192-HA overexperssion mice. F. Anti-HA immunoisolation in SLC6A7-2A-HA mice under FCCP (50 μM) and NEM (200 μM). G. Anti-Syp immunoisolation in wild type mice under FCCP (50 μM) and Nigercin (20 μM).

B) The authors generated multiple potentially very useful genetic tools and models. However, the validation of these models is incomplete. Most importantly, it remains unclear whether the different mutations affect SLC6A17 expression levels, subcellular localization, or the expression and trafficking of other SV and synapse components.

The verification of transgenic mouse line is described in the Material and Methods section of our manuscript. There are numerous literatures published for CRISPR mediated gene editing in animals and the off-target effect of CRISPR-Cas9 system is widely studied with optimized design tools developed by many groups (Platt et al., 2014; Chu et al., 2015, 2016; Liu et al., 2017; Gemberling et al., 2021; Singh et al., 2022). The gRNAs used for animal generation were chosen carefully based on publically available tools. Apart from basic genomic PCR sequencing of target regions of all gene edited mouse models, Southern blots were performed by Biocytogen company for Slc6a17-HA-2A-iCre and Slc6a17P633R mice to rule out random insertions. Expression levels in Slc6a17-KO and Slc6a17P633R mice were not affected, as shown in Figure R2. HA-tagged protein in Slc6a17-HA-2A-iCre and Slc6a17P633R mice were detected by immunoisolation, immunofluorescence, and fractionation (Figure 3, 4, Figure 3—figure supplement 1, Figure 4—figure supplement 1). Both showed localizations expected from previous reports ().

C) Apart from the caveats mentioned above regarding Gln uptake into SVs, the data interpretation provided by the authors lacks stringency with respect to the biophysics of plasma membrane and SV transporters.

The biophysics of SLC6A17 was carefully studied (Para et al 2008; Zaia and Reimer, 2009). Our work focused on in vivo biochemical results, not biophysics.

Author response image 2.

Verification of genetic mouse models. A. q-PCR verification of Slc6a17-KO mice; B. q-PCR verification of Slc6a17P633R mice; C. Example of genomic primer design for Slc6a17-HA-2A-iCre mice founder mice screen; D. Example of genomic PCR for Slc6a17-HA-2A-iCre mice founder mice screen; E. Southern blot performed for Slc6a17-HA-2A-iCre mice.

Reference

Chu, Van Trung et al. “Increasing the efficiency of homology-directed repair for CRISPR-Cas9-induced precise gene editing in mammalian cells.” Nature biotechnology vol. 33,5 (2015): 543-8. doi:10.1038/nbt.3198

Chu, Van Trung, et al. "Efficient generation of Rosa26 knock-in mice using CRISPR/Cas9 in C57BL/6 zygotes." BMC biotechnology 16.1 (2016): 1-15.

Gemberling, Matthew P et al. “Transgenic mice for in vivo epigenome editing with CRISPR-based systems.” Nature methods vol. 18,8 (2021): 965-974. doi:10.1038/s41592-021-01207-2

Liu, Edison T., et al. "Of mice and CRISPR: The post‐CRISPR future of the mouse as a model system for the human condition." EMBO reports 18.2 (2017): 187-193.

Madisen, Linda, et al. "A robust and high-throughput Cre reporting and characterization system for the whole mouse brain." Nature neuroscience 13.1 (2010): 133-140.

Parra, Leonardo A., et al. "The orphan transporter Rxt1/NTT4 (SLC6A17) functions as a synaptic vesicle amino acid transporter selective for proline, glycine, leucine, and alanine." Molecular pharmacology 74.6 (2008): 15211532.

Platt, R.J., Chen, S., Zhou, Y., Yim, M.J., Swiech, L., Kempton, H.R., Dahlman, J.E., Parnas, O., Eisenhaure, T.M., Jovanovic, M., et al. (2014). CRISPR-Cas9 knockin mice for genome editing and cancer mode Yang, Hui, Haoyi Wang, and Rudolf Jaenisch. "Generating genetically modified mice using CRISPR/Cas-mediated genome engineering." Nature protocols 9.8 (2014): 1956-1968.ling. Cell 159, 440-455.

Singh, Surender et al. “Opportunities and challenges with CRISPR-Cas mediated homologous recombination based precise editing in plants and animals.” Plant molecular biology, 10.1007/s11103-022-01321-5. 31 Oct. 2022, doi:10.1007/s11103-022-01321-5

Zaia, K.A., and Reimer, R.J. (2009). Synaptic vesicle protein NTT4/XT1 (SLC6A17) catalyzes Na+-coupled neutral amino acid transport. J Biol Chem 284, 8439-8448.

Author Response

eLife assessment

This study assesses homeostatic plasticity mechanisms driven by inhibitory GABAergic synapses in cultured cortical neurons. The authors report that up- or down-regulation of GABAergic synaptic strength, rather than excitatory glutamatergic synaptic strength, is critical for homeostatic regulation of neuronal firing rates. The reviewers noted that the findings are potentially important, but they also raised questions. In particular, the evidence supporting the findings is currently incomplete and demonstration of independent regulation of mEPSCs and mIPSCs is a necessary experiment to support the major claims of the study.

We appreciate the detailed, thoughtful assessment of our paper by the reviewers and editors and will submit a revised version in the future that addresses the reviewers’ comments as detailed below in response to each concern. We will include a more open discussion of alternative possibilities. Further, we will repeat the optogenetic experiments assessing AMPAergic scaling in our mouse cortical cultures in order to demonstrate independent regulation of mEPSCs and mIPSCs as suggested.

Reviewer #1 (Public Review):

In the manuscript titled "GABAergic synaptic scaling is triggered by changes in spiking activity rather than transmitter receptor activation," the authors present an investigation of the role of GABAergic synaptic scaling in the maintenance of spike rates in networks of cultured neurons. Their main findings suggest that GABAergic scaling exhibits features consistent with a key homeostatic mechanism that contributes to the stability of neuronal firing rates. Their data demonstrate that GABAergic scaling is multiplicative and emerges when postsynaptic spike rates are altered. Finally, their data suggest that, in contrast to their prior data on glutamatergic scaling, GABAergic scaling is driven by spike rates. The authors set the paper up as an argument that GABAergic scaling, rather than glutamatergic scaling, serves as the critical homeostatic mechanism for spike rate regulation.

While the paper is ambitious in its rhetorical scope and certainly presents intriguing findings, there are several serious concerns that need to be addressed to substantiate the interpretations of the data. For example, the CTZ data do not support the interpretations and conclusions drawn by the authors. Summarily, the authors argue that GABAergic scaling is measuring spiking (at the time scale of the homeostatic response, which they suggest is a key feature of a homeostat) yet their data in figure 5B show more convincingly that CTZ does not influence spiking levels - only one out of four time points is marginally significant (also, I suspect that the bootstrapping method mentioned in line 454-459 was conducted as a pairwise comparison of distributions. There is no mention of multiple comparisons corrections, and I have to assume that the significance at 3h would disappear with correction).

We certainly understand the criticism here (similar to reviewer 2’s third point). In our resubmission we will do a better job discussing these complications, which we now summarize. First, we are presenting our entire dataset to be as transparent as possible. Unlike most synaptic scaling studies (including our own) that apply drugs to alter activity and assess mPSC amplitude at the final time point, here we are actually showing CTZ’s effect on spiking activity within the culture over time. This is critical because it has informed us of the drug’s true effect on spiking, the variability that is associated with these perturbations, and the ability and timing of the cultured network to homeostatically recover initial levels. This was important because it revealed that the drugs do not always influence activity in the way we assume, and this provides greater context to our results. Second, we are showing all of our data, and presenting it using estimation statistics which go beyond the dichotomy of a simple p value yes or no (Ho J, Tumkaya T, Aryal S, Choi H, Claridge-Chang A. 2019. Moving beyond P values: data analysis with estimation graphics. Nat Methods 16: 565-66). Estimation statistics have become a more standard statistical approach in the last 15 years and is the preferred method for the Society for Neuroscience’s eNeuro Journal. This method shows the effect size and the confidence interval of the distribution. For the 3 hr time point in Fig. 5B the CTZ/ethanol vs. ethanol data points exhibit very little overlap and the effect size demonstrates a near doubling of spike frequency, and the confidence interval shows a clear separation from 0. This was a pairwise comparison as we compared values at each time point after the addition of ethanol or ethanol/CTZ. Third, the plots illustrate an upward trend in spike frequency at 1 and 6 hrs, but that there is also clear variability. It is important to note that while these recordings help us to understand effects on spiking across the cultured network, they cannot directly speak to spiking activity in the principal neurons that we target. This complication along with the variability inherent in these cultures could make simple comparisons difficult to interpret. Regardless, we do see some increase in spiking with CTZ and we clearly see increases in mIPSC amplitude, thus providing some support for the idea that spiking could be a critical player in terms of GABAergic scaling, particularly when put in the context of our other findings. However, it is important to recognize that something other than total spike rate may contribute to GABAergic scaling, such as the pattern of spiking that produces a particular calcium transient, and this will be discussed in the resubmission.

Then, the fact that TTX applied on top of CTZ drives a increase in mIPSC amplitude is interpreted as a conclusive demonstration that GABAergic scaling is sensing spiking. It is inevitable, however, that TTX will also severely reduce AMAP-R activation - a very plausible alternative explanation is that the augmentation of AMPAR activation caused by CTZ is not sufficient to overcome the dramatic impact of TTX. All together, these data do not provide substantial evidence for the conclusion drawn by the authors.

We understand this point when considering the CTZ/TTX experiments by themselves. However, spiking appears to be a more straightforward trigger when the CTZ/TTX results are coupled with the prevention of GABAergic downscaling by optogenetic restoration of spiking in the presence of AMPAR antagonists. Further, an important point here is that our results with TTX vs. TTX + CTZ are different for GABAergic scaling (no difference) and AMPAergic scaling (CTZ diminished upward scaling) suggesting different triggers for the two forms of scaling. We will make this more clear in our resubmission.

Specific points:

• The logic of the basis for the argument is somewhat flawed: A homeostat does not require a multiplicative mechanism, nor does it even need to be synaptic. Membrane excitability is a locus of homeostatic regulation of firing, for example. In addition, synapse-specific modulation can also be homeostatic. The only requirement of the homeostat is that its deployment subserves the stabilization of a biological parameter (e.g., firing rate).

We agree with the reviewer and should not have suggested that this was a necessary requirement for a spike rate hemostat. What we should have said was that historically this definition has been attributed to AMPAergic scaling, which is thought to be a spike rate homeostat. We will correct this in the resubmission.

• Line 63 parenthetically references an important, but contradictory study as a brief "however". Given the tone of the writing, it would be more balanced to give this study at least a full sentence of exposition.

Agreed, we will do this.

• The authors state (line 11) that expression of a hyperpolarizing conductance did not trigger scaling. More recent work ('Homeostatic synaptic scaling establishes the specificity of an associative memory') does this via expression of DREADDs and finds robust scaling.

The purpose of citing this study was to argue that the spike rate homeostat hypothesis doesn’t make sense for AMPAergic scaling based on a study that hyperpolarized an individual cell while leaving the rest of the network unaltered and therefore leaving network activity and neurotransmission largely normal. In this case scaling was not triggered, suggesting reduced spike rate within an individual cell was insufficient to trigger scaling. The study that the reviewer refers to hyperpolarizes a majority of cells in the network and therefore will also alter neurotransmission throughout the network, which does not separate the importance of spiking and receptor activation as in the above-mentioned study. We will make this point more clearly in the resubmission.

• Supplemental figure 1 looks largely linear to me? Out of curiosity, wouldn't you expect the left end to be aberrant because scaling up should theoretically increase the strength of some synapses that would have been previously below threshold for detection?

We agree that the scaling ratio plot is largely linear. To be clear, the linearity of the ratio plot was interesting but our main point here was that this line had a positive slope meaning ratios (CNQX mPSC amplitudes/control mPSC amplitudes) got bigger for the larger CNQX-treated mPSCs. Alternatively, a multiplicative relationship where mPSCs are all increased by a single factor (e.g. 2X) would be a flat line with 0 slope at the multiplicative value (e.g. 2). In terms of the left side of the plot, we do see values that rise abruptly from 1 - this is partially obstructed by the Y axis in this figure and we will adjust this. This left part of the plot is likely due the CNQX-induced increases in mPSC amplitudes of mini’s that were below our detection threshold of 5pA. Therefore, mini’s that were 4pAs could now be 5pAs after CNQX treatment and these are then divided by the smallest control mPSCs which are 5 pAs (ratio of 1). We will try to do a better job describing this in the resubmission.

Given that figure 2B also shows warping at the tail ends of similar distributions, how is this to be interpreted?

The left side of the ratio plot shows evidence consistent with the idea that mIPSCs are dropping into the noise after CNQX treatment (similar to above argument), while most of the distribution suggests mIPSCs are reduced to 50% by CNQX treatment. On the right side of the ratio plot the values appear to mostly increase. We are not sure why this is happening, but it looks like some mIPSCs are not purely multiplicative at 0.5, particularly in TTX. It is also important to point out that this is a relatively small percent of the total population and the biggest mPSCs can vary to a great degree from one cell to the next. We will discuss this in the resubmission.

• The readability of the figures is poor. Some of them have inconsistent boundary boxes, bizarre axes, text that appears skewed as if the figures were quickly thrown together and stretched to fit.

We will address these issues in the resubmission.

• I'm concerned about the optogenetic restoration of activity experiment. Cortical pyramidal neuron mean firing rates are log normally distributed and span multiple orders of magnitude. The stimulation experiments can only address the total firing at a network-level - given than a network level "mean" is meaningless in a lognormal distribution, how are we to think about the effect of this manipulation when it comes to individual neurons homeostatically stabilizing their own activities? In essence, the argument is made at the single-neuron level, but the experiment is conducted with a network-level resolution.

As described above, we do not have the capacity to know what the actual firing rate of a particular neuron was before and after introducing a drug and so we cannot absolutely say that we have restored the original firing rates of neurons. However, there is reason to believe that this is achieved to some extent. Our optogenetic stimulation is only 50-100 ms long activating a subset of neurons. This is sufficient to provide a synaptic barrage that then triggers a full blown network burst where the majority of spikes occur, but this is after the light is off. In other words, the optogenetic light pulse only initiates what becomes a normal network burst that fortunately allows the individual cells to express their relatively normal (pre-drug) activity pattern. In our previous study we show that this is the case for individual units - the spiking of an individual unit during a burst is similar before and after CNQX/optostim (see Figure 4b and Suppl. Fig 4 in Fong et al. 2015 Nat. Comm.). We are not claiming that we have restored spiking to exactly the pre-drug state, but bring it back toward those levels and we see this is associated with a return of the mIPSC amplitude to near control levels. We will include a description of this in the resubmission.

• Line 198-99: multiplicativity is not a requirement of a homeostatic mechanism.

• Line 264-265 - again, neither multiplicativity and synaptic mechanisms are fundamentally any more necessary for a homeostatic locus than anything else that can modulate firing rate in via negative feedback.

Agreed, see above discussion of homeostat requirement. Will adjust these statements in our resubmission.

• 277: do you mean AMPAR?

We were not clear enough here. We actually do mean GABAR. The idea is that CTZ increases network activity and thus increases both AMPAergic and GABAergic transmission. We will clarify this in the resubmission.

• Example: Figure 1A is frustratingly unreadable. The axes on the raster insets are microscopic, the arrows are strangely large, and it seems unnecessary to fill so much realestate with 4 rasters. Only one is necessary to show the concept of a network burst. The effect of time+CNQX on the frequency of burst is shown in B and C.

• Example: Figure 2 appears warped and hastily assembled. Statistical indications are shown within and outside of bounding boxes. Axes are not aligned. Labels are not aligned. Font sizes are not equal on equivalent axes.

We will adjust these issues in the resubmission.

• The discussion should include mention of the limitations and/or constraints of drawing general conclusions from cell culture.

We agree and will adjust the discussion. Also, this is why we cited studies that argue GABAergic neurons have a particularly important role in homeostatic regulation of firing following sensory deprivations in vivo.

• The discussion should include mention of the role of developmental age in the expression of specific mechanisms. It is highly likely that what is studied at ~P14 is specific to early postnatal development.

We will discuss caveats of cortical cultures at DIV 14-20.

It is essential to ensure that the data presented in the paper adequately supports the conclusions drawn. A more cautious approach in interpreting the results may lead to a stronger argument and a more robust understanding of the underlying mechanisms at play.

Agreed.

Reviewer #2 (Public Review):

Synaptic scaling has long been proposed as a homeostatic mechanism for the regulation for the activity of individual neurons and networks. The question of whether homeostasis is controlled by neuronal spiking or by the activation of specific receptor populations in individual synapses has remained open. In a previous work, the Wenner group had shown that upscaling of glutamatergic transmission is triggered by direct blockade of glutamate receptors rather than by the concomitant reduction in firing rate (Nat Comm 2015). In this manuscript they investigate the mechanisms regulating scaling of GABA-mediated responses in cortical cell cultures using whole-cell recordings to detect GABAergic currents and multielectrode arrays to monitor global firing activity, and find that spiking plays a fundamental role in scaling.

Initially, the authors show that chronic blockade (24 h) of glutamatergic transmission by CNQX first reduces spontaneous spiking (at 2 h), but later (24 h) firing grows back towards higher frequencies, suggesting a compensatory mechanism. Then it is shown that either chronic CNQX treatment or TTX cause a reduction in the amplitude of GABAergic mIPSCs. Effects of CNQX on IPSCs are then reverted by replacing spontaneous network firing by chronic optogenetic stimulation of the entire culture, also indicating that GABAergic transmission is homeostatically regulated by global firing. Enhancing glutamatergic transmission with CTZ increases mIPSC amplitude, while addition of TTX in the presence of CTZ causes the opposite effect. Finally, increasing spiking activity using bicuculline also increases mIPSC amplitude, and the authors conclude that spiking activity rather than neurotransmission control homeostatic GABA scaling. The manuscript shows interesting properties in the regulation of global GABAergic transmission and highlight the important role of spiking activity in triggering GABA scaling. However, it is strongly recommended to address some caveats in order to better support the conclusions presented in the manuscript.

Major points:

1) The reason why CNQX does not completely eliminate spiking is unclear (Fig. 1). What is the circuit mechanism by which spiking continues, although at lower frequency, in the absence of AMPA-mediated transmission and what the mechanism by which spiking frequency grows back after 24h (still in the absence of AMPA transmission)?

Is it possible that NMDA-mediated transmission takes over and triggers a different type of network plasticity?

The bursting in AMPAR blockade is due to the remaining NMDA receptor mediated transmission. We showed this in our previous study in Suppl. Figure 2 and 6 of Fong et al., 2015 Nat. Comm.. Our ability to optically induce normal looking bursts of spikes was also dependent NMDAR activation. Further, in Dr Fong’s PhD dissertation it was shown that the bursting activity was abolished when AMPA and NMDA receptors were both blocked. There are likely many factors that contribute to the recovery of activity, and certainly one of them is likely to be the weakening of inhibitory GABAergic currents. These points will be discussed in the resubmission.

2) A possible activation of NMDARs should be considered. One would think that experiments involving chronic glutamatergic blockade could have been conducted in the presence of NMDAR blockers. Why this was not the case?

Unfortunately, it was not possible to optogenetically restore normal bursting in the presence of NMDAR blockade (even when AMPAergic transmission was intact), as NMDARs appeared to be critical for the optical restoration of the normal duration of the burst (see Suppl. Figure 6 Fong et al., 2015 Nat. Comm). The reviewer raises an excellent point about a possible NMDAR contribution to altered synaptic strength, however. It is likely that NMDAR signaling is reduced in the presence of CNQX since burst frequency was reduced along with AMPAR-mediated depolarizations. We cannot rule out the possibility that NMDAR signaling could contribute to the alterations in GABAergic mIPSCs and will discuss this in the resubmission. However, previous work suggests that 24/48 hour block NMDARs (APV) did not trigger AMPAergic scaling in cortical or hippocampal cultures (see Figure 1 Turrigiano et al., 1998 Nature and Suppl. Figure 4 Sutton et al., 2006 Cell), moreover, our previous study showed that restoring NMDAergic transmission optogentically, at least to some point, had no influence on AMPAergic scaling (Fong et al., 2015, Nat. Comm.). Regardless, we cannot rule out a role for NMDAergic transmission in GABAergic scaling and this discussion will be included in the resubmission.

Also, experiments with global ChR2 stimulation with coincident pre and postsynaptic firing might also activate NMDARs and result in additional effects that should be taken into consideration for the global scaling mechanism.

To be clear, our optical stimulation was turned off before the vast majority of spiking that occurred in the bursts, which played out in a relatively natural manner (see lower panel of Figure 3B optogenetic stimulation – short duration only at onset of burst – we will make this clearer in resubmission). Therefore, we were unlikely to trigger significant synchronous activation that does not normally occur in network bursts.

3) Cultures exposed to CTZ to enhance AMPA receptors generated variable results (Fig. 5), somewhat increasing spiking activity in a non-significant manner but, at the same time, strengthening mIPSC amplitude. This result seems to suggest that spiking might be involved in GABAergic scaling, but it does not seem to prove it.Then, addition of TTX that blocked spiking reduced mIPSC amplitude. It was concluded here that the ability of CTZ to enhance GABAergic currents was primarily due to spiking, rather than the increase in AMPA-mediated currents. However, in addition to blocking action potentials, TTX would also prevent activation of AMPARs in the presence of CTZ due to the lack of glutamatergic release. Therefore, under these conditions, an effect of glutamatergic activation on GABAergic scaling cannot be ruled out.

These concerns were very similar to reviewer 1’s first comments. We will address these issues in the resubmission, but to briefly repeat our responses: We are going a step beyond most scaling studies by assessing MEA-wide firing rate, but this still provides an incomplete picture of the particular cells that we target for patch recordings in terms of their firing before and after a drug. Further, we see considerable variability in effect on firing rate from culture to culture, which we will better recognize in the resubmission. Finally, While the CTZ results are not conclusive, taken together with the optogenetic results we think our results are most consistent with idea that GABAergic scaling is a strong candidate as a spike rate homeostat.

4) The sample size is not mentioned in any figure. How many cells/culture dishes were used in each condition?

The individual dots represent either individual cells for mIPSC amplitude or individual cultures in MEA experiments. Number of cultures for figures were: Figure 2 – con = 10, TTX = 3, CNQX = 6, Figure 4 – CNQX = 4, con = 10, CNQX/photostim = 6, Figure 5 – ethanol = 3, CTZ = 3, CTZ + TTX =3, Figure 6 – con = 10, bicuculline = 4. We will include the number of cultures for mIPSC amplitude experiments in the figure legends upon resubmission.

5) Cortical cultures may typically contain about 5-10% GABAergic interneurons and 90-95 % pyramidal cells. One would think that scaling mechanisms occurring in pyramidal cells and interneurons could be distinct, with different impact on the network. Although for whole-cell recordings the authors selected pyramidal looking cells, which might bias recordings towards excitatory neurons, naked eye selection of recording cells is quite difficult in primary cultures. Some of the variability in mIPSC amplitude values (Fig. 2A for example) might be attributed to the cell type? One could use cultures where interneurons are fluorescently labeled to obtain an accurate representation. The issue of the possible differential effects of scaling in pyramidal cells vs. interneurons and the consequences in the network should be discussed.

We will include this discussion in the resubmission. Briefly, we chose large cells, which will be predominantly glutamatergic neurons as suggested by the reviewer. Ultimately, even among glutamatergic principal cells there may be variability in the response to drug application. All of these issues could contribute to variability and we will expand our description of the variability in our results, including that based on cellular heterogeneity.

Reviewer #3 (Public Review):

This paper concerns whether scaling (or homeostatic synaptic plasticity; HSP) occurs similarly at GABA and Glu synapses and comes to the surprising conclusion that these are regulated separately. This is surprising because these were thought to be co-regulated during HSP and in fact, the major mechanisms thought to underlie downscaling (TTX or CNQX driven), retinoic acid and TNF, have been shown to regulate both GABARs and AMPARs directly. (As a side note, it is unclear that the manipulations used in Josesph and Turrigiano represent HSP, and so might not be relevant). Thus the main result, that GABA HSP is dissociable from Glu HSP, is novel and exciting. This suggests either different mechanisms underlie the two processes, or that under certain conditions, another mechanism is engaged that scales one type of synapse and not the other.

However, strong claims require strong evidence, and the results presented here only address GABA HSP, relying on previous work from this lab on Glu HSP (Fong, et al., 2015). But the previous experiments were done in rat cultures, while these experiments are done in mice and at somewhat different ages (DIV). Even identical culture systems can drift over time (possibly due to changes in the components of B27 or other media and supplements). Therefore it is necessary to demonstrate in the same system the dissociation. To be convincing, they need to show the mEPSCs for Fig 4, clearly showing the dissociation. Doing the same for Fig 5 would be great, but I think Fig 4 is the key.

We understand the concern of the reviewer as we do see significant variability within our cultures and they were plated in different places, by different people, in different species (rat vs mouse). Therefore, in the resubmission to strengthen the conclusions we will repeat our optogenetic studies restoring activity in the presence of AMPAergic blockade in our mouse cortical cultures and measuring AMPA mEPSCs to assess scaling.

The paper also suggests that only receptor function or spiking could control HSP, and therefore if it is not receptor function then it must be spiking. This seems like a false dichotomy; there are of course other options. Details in the data may suggest that spiking is not the (or the only) homeostat, as TTX and CNQX causes identical changes in mIPSC amplitude but have different effects on spiking. Further, in Fig 5, CTZ had a minimal effect on spiking but a large effect on mIPSCs. Similar issues appear in Fig 6, where the induction of increased spiking is highly variable, with many cells showing control levels or lower spiking rates. Yet the synaptic changes are robust, across all cells. Overall, this is not persuasive that spiking is necessarily the homeostat for GABA synapses.

Together our results argue against AMPAR or GABAR activation as a trigger for GABAergic scaling and that this is different than our results for AMPAergic scaling. These points alone are important to recognize. While changes in spiking do not perfectly follow the changes in GABAergic scaling they do always trend in the right direction. As mentioned above, total spiking activity is only one measure of spiking. It is possible that these drugs alter the pattern of spiking that translates into an altered calcium transient that is important for triggering the plasticity. Again, it is important to note that we are going a step beyond most homeostatic plasticity studies that add a drug and simply assume it is having an effect on spiking (e.g. CNQX was initially thought to completely abolish spiking, but clearly does not). Based on the variability that we observe and the nature of our MEA recordings we cannot precisely determine how the total activity or pattern of activity changes with drug application in the specific cells that we target for whole cell recordings. However, we believe our results are more consistent with our proposal that GABAergic scaling is a strong candidate as a spike rate homeostat. Regardless, in the resubmission we will include a broader discussion about these possibilities, and the reality that there could be multiple homeostatic mechanisms that act to recover spiking activity.

The paper also suggests that the timing of the GABA changes coincides with the spiking changes, but while they have the time course of the spiking changes and recovery, they only have the 24h time point for synaptic changes. It is impossible to conclude how the time courses align without more data.

We can only say that by the 24 hour CNQX time point, when overall spiking is recovered, that GABAergic scaling has already occurred. We will state this more clearly in the resubmission.

Author Response:

We are grateful to the editors for getting our study reviewed, and are pleased that the reviewers found value in our findings. We plan to submit a revision that we believe can resolve much of the remaining doubt about the major conclusions.

Our current understanding is that much of the uncertainty stems from extensive diversity among synapses. The FM-dye de-staining technique does have single synapse resolution, so it should be possible to develop new kinds of analysis that can make each of our points at the level of individual synapses. For a preview, see Figure 2D (explained in lines 126-141), and Figure 2-Figure supplement 5 of the current version.

Author Response

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

We thank the reviewers for their time in evaluating the strengths and weaknesses of our manuscript.

We are pleased to see that all reviewers recognized the high significance of our work, noting that the manuscript addresses “longstanding question of which cell types are infected during congenital or perinatal rubella virus infection”. As noted by reviewer 1, “This study reveals a new cellular target that will have important implications for basic studies on rubella virus-host interactions and for the potential development of therapies or improved vaccines targeting this virus. As the rubella virus is a pathogen of high concern during human pregnancy, this study also has important implications in the field of neonatal infectious diseases”.

Below, we provide responses (in blue) to specific critiques:

Reviewer #1 (Public Review):

A weakness is that the current data do not provide information on the full replicative potential of the rubella virus in microglia, or whether the virus persists in this system.

See our response below. Briefly, we include new experimental evidence from primary tissue, microglia-transplanted organoids, and Vero cells to further characterize the dynamics of viral infection.

Reviewer #1 (Recommendations for the authors):

Most of the viral assays in the brain slices and organoids examine viral protein synthesis, which is a surrogate for genome replication. However, basic virological characterization is lacking and would improve the robustness of the model and its potential utility to understand better rubella virus-microglia interactions. Questions the authors should consider with new experiments include:

Are new virions produced? Can viruses be detected in the media?

Or, are the infections abortive, with viral protein synthesis occurring, but no virus production?

We performed RV titering experiments in dissociated microglia co-cultured with other cell types, as well as Vero cells as a control. While we can detect a robust increase in viral titer from Vero cells, it fell below detection levels in microglia co-cultures. See Author response image 1. We now include these data in Supplementary Figure 2D.

Author response image 1.

Rubella virus titering experiment performed in Vero cells (positive control) or dissociated microglia co-cultures. In primary microglia co- cultures, viral titer falls below detection levels after several days of infection.

While we could not detect an increase in the viral particles from microglia mixed cultures, we confirmed the presence of GFP from the RV-GFP reporter construct, and we believe it serves as a proof that the virus can infect microglia cells and lead to production of functional viral protein (Author response image 2, Figure 1E-F of the current manuscript):

Author response image 2.

We also observed an increase in RV RNA over time in tissue slice infections, using qPCR (Author response image 3, not included in the manuscript).

Author response image 3.

Modest increase in RV RNA over time in brain slice infections. Rubella virus RNA measured by qPCR relative to GAPDH gene, in n=3 samples (2 technical replicates each condition). Brain slices were exposed to RV, then collected at end of inoculation (4 hours post infection), or at 3 or 5 days post infection, and processed for RNA extraction and RT-qPCR.

How long do the infections persist in the model? What is the fate of infected microglia over time? Time courses to monitor infection and cell health would be useful.

We performed a longer infection with RV in organoids transplanted with microglia, and after two weeks of infection, we can detect multiple microglia cells positive for the RV capsid. These data are now included in Figure 4 of the current manuscript.

Author response image 4.

After 2 weeks post infection, microglia remain positive for RV capsid.

Reviewer #2 (Public Review):

Weaknesses

The set of data is rather descriptive. It suggests that microglia are the predominant brain target of RV in vivo, without identifying the targeting mechanism that provides cell type specificity. Moreover, what are the diffusible cues released from the brain environment that increase microglia infection and RV replication?

We agree with the reviewer that identifying molecular mechanisms that underlie this phenotype will be very interesting to explore in future research, and we acknowledge the limitation of the study in the Discussion.

It is unclear why brain organoids not supplemented by microglia are susceptible to RV inoculation.

We could not detect RV capsid in organoids without microglia after 72 hours of inoculation. We attribute any changes seen at the level of single cell transcriptomics in the absence of microglia transplantation to exposure to virus-associated particles, including but not limited to viral RNA species, viral proteins, or even other components of the viral stocks made in Vero cells. These factors may induce transcriptomic differences even in the absence of RV infection. In the text, we take care to refer to these condition as “Rubella virus-exposed” rather than “Rubella virus- infected”. We now include the following panel from Author response image 5 in Figure 4B of the current manuscript.

Author response image 5.

Organoids without microglia do not show positive RV immunofluorescence.

Reviewer #2 (Recommendations for the authors):

Several points could be further addressed to improve the data set and shed more light on some aspects of this manuscript:

• Figure 1. Additional microglia markers should be used to reinforce the evidence that microglia cells are the principal RV targets. Since Iba1 is a marker of activated microglia, does RV have a selective tropism to all microglia or only to activated ones in human fetal brain slices?

The reviewer brings up an interesting point that, in our mind, can be separated into two independent questions:

1. Are Iba1-positive cells bona fide microglia, or are there other cell populations of macrophage/monocyte origin that are labeled with Iba1? Therefore, additional markers should be used for immunolabeling;

2. Is RV infection selective for microglia “activation” status, when only 5mmune-primed cells can be infected?

For the first point, we have previously shown that in the developing human brain, virtually all Iba1-positive cells are also P2RY12-positive (unpublished; Author response image 6). Therefore, in primary human brain slices, there is a negligible amount of non-microglia macrophages. However, in culture microglia quickly lose their “homeostatic” identity, including P2RY12 expression, as quickly as six hours after ex vivo extraction (Gosselin et al., 2017; DOI: 10.1126/science.aal3222).

Author response image 6.

P2RY12 co-localizes with Iba1 in primary brain tissue from gestational week 17.5, including cells with more ameboid morphology (arrows)

However, in organoids at 2 weeks post-RV exposure, we found microglia with both ameboid and more ramified morphology (Author response image 7). It would be challenging and beyond the scope of this manuscript to use morphology or Iba1 intensity levels to determine cause and effect as microglia activation state relates to RV infectivity (i.e. do activated microglia preferentially get infected with the virus, or do infected microglia become activated and upregulate Iba1 levels and change morphology).

Author response image 7.

Examples of microglia with round (top) and ramified (bottom) morphology that co-localize with RV capsid staining.

Regarding RV tropism in the 2D culture of microglia, some Iba- cells are infected by RV as they show capsid staining. What are these cells? Are neurons and/or glia also susceptible to RV in vitro infection? Are non-microglial cells getting RV infected in the absence of microglia?

In the absence of microglia cells, a small proportion of non-microglia cells get infected with RV. There is no statistically significant difference in the number of cells that get infected with RV in the presence or absence of microglia across different cell types. We add these data as Supplement Figure 3.

Author response image 8.

Rubella infection in non-microglia cells. A. Representative images of different cell types depleted of microglia. Cell cultures were stained RV capsid (green) and DAPI. B. Quantification of total cells that are positive for RV capsid across conditions. C. Quantification of RV+ cells that are not microglia across different cell populations. No statistically significant difference was detected in RV infectivity in cells c-cultured with or without microglia.

• Figure 3. The low rate of Rubella virus infection in homogenous CD11b+ cell culture raises the question of whether the Rubella virus can infect microglia at a specific activation stage. It is also surprising that there is no infection of such cell population (also CD11b+) alone while cultured in 2D, as reported in figure 2. Why such a difference?

It is well established that culture of microglial cells isolated from brain tissue alters their molecular properties, which likely alters the cell surface protein composition. In the revised discussion, we include activation as a possible mechanism that will require further investigation.

• Fig 4A-B, it is unclear whether organoids that are not engrafted with microglia get infected upon RV (with active viral replication) inoculation. If non-microglia-supplemented organoids are indeed infected and allow RV replication, this suggests that organoids might not be the ideal system to model human fetal brain RV infection at GW18-23.

We could not detect RV capsid in organoids without microglia after 72 hours of inoculation. We include the following panel from Author respone image 9 in Figure 4 now.

Author response image 9.

Organoids without microglia do not show positive RV immunofluorescence.

• Figure 4E, why are cells derived from microglia-free organoids so much enriched in the UMAP plots as compared to the other organoid condition? Is RV impacting cell fitness, proliferation, or neurodifferentiation?

This perceived difference is due to data presentation. Based on cell proportions, cells from organoids that were treated with microglia are more represented in the scRNAseq data, and this difference most likely comes from user-introduced imbalance in cell loading and possible cell losses during demultiplexing (Author response image 10, panel A). Cell number composition across different conditions and cell types, including RV and MG treatment, are shown in Supplement Figure 4 of the current manuscript (Author response image 10, panel B).

Contribution of each condition can be visualized via UCSC single cell data browser: https://cells.ucsc.edu/?ds=rubella-organoids

Author response image 10.

Data composition depending on condition. A. Cell number contribution from organoids with and without microglia. B. Contribution of each condition to each cluster composition.

• Figure 4F-H. If microglia is the predominant target for RV in the brain, why are microglia-free organoids susceptible to RV and who are the other cellular targets, whose infection leads to activation of interleukin pathway genes and dysregulation of brain developmental markers in selected subpopulations (RGCs, ENs..).

Thank you for bringing this point. We did not detect any appreciable RV genomic RNA in our published single cell data, nor did we identify RV capsid in the RV-exposed organoids without microglia. Our experiments on dissociated cell cultures show that a small population (~1-4%) of other cell types was positive for the RV capsid, including neuron-enriched and glial-enriched fractions (Author response image 11; Supplementary Figure 3C in current manuscript). We expect a similar proportion of non-microglia cells to be infected in the brain organoids. One possible explanation for the robust interferon response even in the absence of productive infection in other cell types is exposure to virions and virus-associated particles, including but not limited to viral RNA species, viral proteins, or even other components of the viral stocks made in Vero cells (which is a cell line that should not produce interferons, but may produce other unmeasured cytokines as a virally infected cell culture).

Author response image 11.

Quantification of RV+ cells that are not microglia across different cell populations. No statistically significant difference was detected in RV infectivity in cells cultured with or without microglia.

• QRT-PCR validations of some of these key brain targets should be performed.

We agree with the reviewer that further validation of the predicted molecular changes downstream of Rubella exposure would be valuable. We have opted to validate IFITM3 and NOVA1 expression differences using immunostaining, and the results are consistent with our predictions from scRNAseq, and the data is presented in revised Figure 5 and 6 of the current manuscript.

Reviewer #3 (Public Review):

Weaknesses of the paper: Overall, additional control experiments are needed to support the stated conclusions. Affinity chromatography is used to purify microglia and other cell types, but the overall cell enrichment is not quantified.

We appreciate the reviewer concern. However, affinity based enrichments rarely guarantee purity of the enrichment, and we do not believe accurate estimation of the purification purity would alter the biological interpretation of the data.

In cell mixing experiments, the authors do not rule out the possibility that the added non- microglia cells also become infected, releasing additional infectious viruses. The finding that a diffusible factor is required for RV infection would be unusual if not unprecedented; therefore, additional data are required to support this claim and rule out other interpretations.

We provide quantification of non-microglia cells that are positive for RV capsid in the presence and absence of microglia. Small (~1-4%) of non-microglia cells get infected with the virus and can potentially release more of the virus (see Author response image 12), but we do not know how this newly produced virus would be different from the one that was applied to the cells directly. To follow up our co-culture experiments, we wanted to exclude a possibility of microglia engulfing RV- infected cells in co-cultures, therefore we separated the two cell fractions by a liquid-permeable membrane (Figure 3 of the current manuscript). It is possible that factors secreted by other cell populations in the transwell assay experiments act on microglia cells to upregulate a yet unidentified receptor on microglia surface or other infection-dependent molecule rendering them infectable by the virus.

We re-phrase the text by de-emphasizing “soluble factors” and focusing on excluding phagocytosis of infected cells as a possible mechanism of RV capsid immunoreactivity in microglia cells.

Author response image 12.

Rubella infection in non-microglia cells. A. Representative images of different cell types depleted of microglia. Cell cultures were stained RV capsid (green) and DAPI. B. Quantification of total cells that are positive for RV capsid across conditions. C. Quantification of RV+ cells that are not microglia across different cell populations. No statistically significant difference was detected in RV infectivity in cells c-cultured with or without microglia.

The methods section would be improved by including details about the iPSC line that was used.

We include the following section in Materials and Methods:

iPSC lines.

All work related to human iPS cells has been approved by the UCSF Committee on Human Research and the UCSF GESCR (Gamete, Embryo, and Stem Cell Research) Committee. Human iPS cell line “WTC-10” derived from healthy 30-year-old Japanese male fibroblasts was from the Conklin Lab, UCSF (Bershteyn et al., 2017; Kreitzer et al., 2013). Human iPSC line “13325” was derived from 9-year-old female fibroblasts originally obtained from Coriell cell repository. Human iPSC line “1323-4” derived from healthy 48-year-old Caucasian female fibroblasts (gift from the Conklin Lab, UCSF) was used for immunofluorescence validation analysis as we found that this line generates more reproducible brain organoid differentiations.

and by a more thorough description of virus-specific details, including the numbers of infectious particles added per volume of incubation media.

We now include the following data in the Materials and Methods section:

Rubella virus infection

Cells cultured in 2D were inoculated by adding RV stock virus to culture media in 1:1 dilution (250 ul of media to the equal volume of viral stock, 1.75x105 total ffu/well) to achieve a multiplicity of infection (MOI) of 2. After four hours, media was exchanged with fresh cell culture media. Cortical brain slices were treated with 500 ul of RV viral stock (3.5x105 total ffu/slice) applied over the slice culture filter for four hours, and then the viral culture media was removed and replaced with fresh slice culture media. Organoids were treated in 6-well plates with 2ml of 1:1 dilution of viral stock:organoid maintenance media (7x105 total ffu) for four hours, and then viral media was exchanged for fresh media. For all experimental conditions, cells were fixed and processed for downstream analysis at 72 hours post infection. Supernatant from non-infected Vero cells (mock) or heat-inactivated RV (650C, 30 mins) was used as control.

In addition to immunofluorescence, adding additional data to demonstrate and quantify virus infection (PCR and plaque assays. or immunofluorescence using an anti-double-stranded RNA antibody such as J2) from the infected brain slices and organoids would provide greater assurance that the virus is indeed replicating under the experimental conditions.

We performed RV titering experiment in dissociated microglia co-cultured with other cell types, as well as Vero cells control. While we can detect a robust increase in viral titer from Vero cells, it fell below detection levels in microglia co-cultures. We now include these data in Supplementary Figure 2D.

Author response image 13.

Rubella virus titering experiment performed in Vero cells (positive control) or dissociated microglia co-cultures. In primary microglia co- cultures, viral titer falls below detection levels after several days of infection.

Unfortunately, we did not find J2 staining informative because we could detect signal in both wild type RV infection conditions and in heat-inactivated RV, presumably due to native dsRNA species present in cells. We did not detect any increase or difference in the pattern of staining between RV and heat-inactivated virus-exposed conditions (Author response image 14; not included in the manuscript).

Author response image 14.

J2 antibody labels dsRNA in both RV-exposed and control heat- inactivated virus conditions, presumably due to native dsRNA that is not unique to the viral replication.

Organoid imaging with immunofluorescence would be very informative in demonstrating the presence of microglia and also in showing which cells are virus-infected in the context of organoid structures.

We provide images from 72hrs and 2 week RV infection, providing a zoomed-out view of organoids with microglia and RV capsid staining. We also provide images of 72hrs post- infection in organoids without microglia Author response image 15, Figure 4C in current manuscript).

Author response image 15.

Microglia in organoids co-localize with RV capsid staining.

GenBank accession numbers are listed for the recombinant RV and GFP-RV reporter, but a search using those numbers did not locate the deposits--perhaps the deposits were very recent?

Both viral construct information is now available on GenBank:

M33 RV strain can be found here: https://www.ncbi.nlm.nih.gov/nuccore/OM816674

RV-GFP can be found here: https://www.ncbi.nlm.nih.gov/nuccore/OM816675

The authors incorrectly refer to the GFP virus as a new strain; it is not a viral strain and should be referred to as a reporter virus.

Thank you, we changed the description to

“To confirm functional transcription and translation of the viral genome, a new reporter construct of RV designed to express GFP within the non-structural P150 gene was generated (RV-GFP, GenBank Accession OM816675)”

Given that the authors show that Vero cell cultures are infected by the Rubella virus in the absence of other cells, additional evidence is needed to demonstrate that a diffusible factor from other cells enables microglia to be infected by the Rubella virus.

We have revised the manuscript to indicate that our data is consistent with the possibility that a diffusible factor is involved. Our experiment utilizing transwell assay argues against phagocytosis and physical interactions as primary drivers, but future studies will be needed to determine if soluble factors are involved.

The authors did not detect Rubella virus transcripts in the single-cell RNA sequencing experiment, nor was a microglia cluster found.

Indeed, microglia recovery using scRNAseq is very inefficient. We note this limitation in the discussion.

Innate immune responses can be activated in the presence of viral particles but without virus replication, as in inactivated viral vaccines; therefore changes in interferon responses do not necessarily prove virus replication.

We agree with the reviewer on this point, it is difficult, if at all possible, to entirely eliminate the possibility that some of the transcriptomic changes, particularly the interferon responses, are not induced by the exposure to viral particles. We have revised the manuscript to more rigorously described the conditions as “RV-exposed”.

Figure 4: it would be helpful to define the abbreviations used in the figure legend (e.g. IPC, RG, EN). In the volcano plots, the gene names are blocked by the dots, and the figure becomes very pixelated when enlarged to read the text.

We have added abbreviations and replaced the figure files with higher resolution images (Figure 6 in current manuscript).

The value of including Supplemental Figure 2 (MOG) is not clear because it receives little mention in the text and also seems to be previously published data that could be cited.

We have removed the figure and replaced it with a citation and a link to the Cell Browser.

Supplemental Figure 4: In panel G, the legend shows "YH10" and "13325". These terms are not described in the Figure legend, nor did a search of the manuscript identify these terms. In its current form Supp. Fig. 4G is not interpretable. In addition, would be more clear to use the term "RV-infected" instead of "treated" to describe the addition of the virus.

We have expanded the Methods section to include the description of different organoid lines and added a revised legend for Supplementary Figure 4. We do not provide evidence of RV infecting organoids without microglia, therefore we have revised the claims that organoid cells become infected with the virus and replaced it with “RV-exposed” to better reflect the conditions studied.

Reviewer #3 (Recommendations for the authors):

1) Demonstrate and quantify virus replication to provide data to complement the imaging. In order of data quality, plaque assays would be most convincing in demonstrating infection and release of infectious virus, while a time course of PCR on RV transcripts would support a conclusion of replicating virus. Further, staining with an anti-double-stranded RNA antibody (J2) would represent evidence of virus replication.

In response to the reviewer’s comment, we performed an RV titering experiment in dissociated microglia co-cultured with other cell types, as well as Vero cells control. While we can detect a robust increase in viral titer from Vero cells, it fell below detection levels in microglia co-cultures. We now include these data in Supplementary Figure 2D.

Author response image 16.

Rubella virus titering experiment performed in Vero cells (positive control) or dissociated microglia co-cultures. In primary microglia co- cultures, viral titer falls below detection levels after several days of infection.

We detected a very modest increase in RV RNA in infected brain slices over time using RT- qPCR (see Author response image 17, not included in current manuscript)

Author response image 17.

Modest increase in RV RNA over time in brain slice infections. Rubella virus RNA measured by qPCR relative to GAPDH gene, in n=3 samples (2 technical replicates each condition). Brain slices were exposed to RV, then collected at end of inoculation (4 hours post infection), or at 3 or 5 days post infection, and processed for RNA extraction and RT-qPCR.

Unfortunately, we did not find J2 staining informative because we could detect signal in both wild type RV infection conditions and in heat-inactivated RV, presumably due to native dsRNA species present in cells. We did not detect any increase of difference in the pattern of staining between RV and heat-inactivated virus-exposed conditions (Author response image 18; not included in the manuscript).

Author response image 18.

J2 antibody labels dsRNA in both RV-exposed and control heat- inactivated virus conditions, presumably due to native dsRNA that is not unique to the viral replication.

We utilized FISH to detect negative-stranded (non-genomic) RV RNA as an alternative to J2 to indicate RNA replication. However, it proved to be not very sensitive, as a small quantity of negative-strand RV RNA could be detected in highly infected Vero cells, but negative-strand RV RNA was not detected in more modestly infected microglia (based on positive-strand RV RNA quantification), as in Author response image 19, not included in current manuscript.

Author response image 19.

FISH probes to positive strand (genomic) and negative strand (replication template) RV RNA in Vero cells and microglia co-cultures. A: representative images of Vero cells infected with RV (top row) or Zika virus as control (bottom row). At 72hpi, cells were fixed and processed for immunofluorescence with anti-RV capsid antibody (RVcap) or Zika virus antibody (Zika4G2), and then FISH was performed using probes to positive strand (+) or negative strand (-) RV RNA. Negative strand RV RNA difficult to visualize at low-power magnification, and required quantification within cell borders defined by wheat germ agglutinin staining with results in panel B. B: In Vero cells, negative strand RV RNA is detected in strongly infected cells. Infection strength determined by intensity of RV capsid immunofluorescence staining and positive strand RV RNA (RVcap/(+) 2/3 indicates robust infection, RVcap/(+) 1 indicates weak infection). ZIKVinf = Zika virus infected control. C: In microglia co-cultures, positive strand RV RNA detected in cells with RV capsid immunopositivity (RVcap_pos). RVinf = RV infected. RVHI = heat-inactivated RV. D: In microglia co-cultures, negative strand RV RNA quantification not significantly different between mock, heat-inactivated RV (RVHI), or RV- infected conditions (RVinf), including cells with weak positive-strand RV RNA (RVinf, (+)<8) or cells with stronger positive-strand RV RNA ((RVinf, (+)>=8). Two biological replicates (bHR60 and bHR61), n indicates number of cells counted.

While we could not detect an increase in the viral particles from microglia mixed cultures, we confirmed the presence of GFP from the RV-GFP reporter construct, and we believe it serves as a proof that the virus can infect microglia cells and lead to production of functional viral protein (see Author response image 20, Figure 1E-F of the current manuscript)

Author response image 20.

Thus, overall we detect replication of viral RNA and protein (qPCR, RV-GFP), but not an appreciable increase in released newly-made virions. The discussion now reflects this more clearly in the current manuscript.

2) The claim of requiring a diffusible factor to enable RV infection requires additional data. A suggestion would be to include further characterization of affinity-purified cells to define the levels of cell enrichment and to determine which other cell types are present, It is also important to test the RV infection of the fractionated cell types alone before adding to the microglia, in order to demonstrate whether RV is replicating in cell types other than microglia.

We performed quantifications of RV capsid-positive cells in each of the affinity-purified cell populations: neuron-enriched (purified with PSA-NCAM beads), glia-enriched (PSA-NCAM depleted cell fraction), or non-microglia fraction (“Flow through”, depleted of CD11b+ cells). We show that across each condition, we have low infectivity (ranging from ~1 to 4% of total cell population) after 72 hours post-infection. We include these data in Supplementary Figure 3.

Author response image 21.

Rubella infection in non-microglia cells. A. Representative images of different cell types depleted of microglia. Cell cultures were stained RV capsid (green) and DAPI. B. Quantification of total cells that are positive for RV capsid across conditions. C. Quantification of RV+ cells that are not microglia across different cell populations. No statistically significant difference was detected in RV infectivity in cells c-cultured with or without microglia.

Another approach to limit cell heterogeneity would be to use iPSC-derived cells, which are highly enriched as a single cell type as a specific cell type, to test the requirement for additional cell types to achieve RV infection of microglia.

In our prior publication (Popova et al. 2021) we have identified a number of molecular differences between primary and iPSC derived microglia. iPSC derived microglia like cells could show differences in infection tropism from primary microglia, and those results may be difficult to interpret biologically. We agree with the reviewer that iPSC derived cells would be an interesting model, there are now several distinct protocols for deriving microglia like cells from pluripotent stem cells and we feel that embarking on a protocol comparison project would fall outside the scope of the current manuscript.

3) Consider a longer organoid infection. The authors did not identify viral RNA transcripts in their organoid scRNAseq data after a 72-hour infection. Although the 72-hour time point seems right for cells in 2D culture, it’s possible that the infection in the organoids is slower because the virus has to spread inwardly. It would be worth trying a time course out to 2 weeks, collecting organoids every few days and then imaging and doing pcr or plaque assays. Zoomed-out views that show immunofluorescence of the entire organoid would also be beneficial in assessing organoid quality and immunofluorescent staining to identify cell types,

We performed longer RV infection for two weeks and now present data on RV capsid in microglia in 72 hrs and 2 weeks post-infection (Author response image 22, Figure 4C of the current manuscript). We have also validated one of the scRNAseq-generated gene candidates in combination with different cell type markers and present data on whole organoids immunostained with NeuN for neurons and EOMES for intermediate progenitor cells that demonstrate the overall structure of the organoids (Author response image 23; Figure 6 of the current manuscript).

Author response image 22.

Microglia in organoids co-localize with RV capsid staining. Organoid with microglia were exposed to RV for 72 hrs or two weeks.

Author response image 23.

Organoids labeled with splice regulator NOVA1 (magenta), neuronal marker NeuN (green) and intermediate progenitor cell marker EOMES (cyan).

Author Response

Reviewer #1 (Public Review):

While the CTD human brain organoids show a decrease in Cr (in absence of Cr in the culture medium) as compared to control organoids (4 times less), they are not devoid of Cr. Do these organoids express the two enzymes allowing Cr synthesis (AGAT and GAMT), and in which brain cell types? If yes, how to explain the decrease in Cr in the CTD organoids?

There is a lack of functional CRT in the CTD human brain organoids. The basal level of creatine in CTD human brain organoid is significantly lower than in healthy human brain organoids. The intracerebral creatine synthesis is due to different expression of the AGAT and GAMT enzymes and relies on functional CRT for the transport of the GAA intermediate Litterature pointed out that both enzymes are rarely co-expressed (Braissant et al., 2001, PMID: 11165387) meaning that GAA intermediate needs to be transported by CRT to neurones for complete creatine synthesis. Even if we evidenced a slight mRNA expression of AGAT and GAMT enzymes, the creatine synthesis is not effective since the GAA intermediate could not be transporterd in cell expressing GAMT due to the non functional creatine transporter in the CTD human brain organoids.

The rescue experiment, re-establishing a functional Cr transporter (CRT or SLC6A8) in the CTD human brain organoids, is very interesting, as this may help the design and development of new treatments for CTD. However, authors claim that the functional CRT expressed in the rescued CTD organoids was expressed in each cell. This may be a difficulty in the development of new CTD treatments, as CRT should be expressed in neurons and oligodendrocytes, but not in astrocytes. Authors may want to comment on this point.

As shown in Figure S2C, the whole brain organoid in the resue experiment shows the expression of the GFP protein, thus also the co-expressed wild-type CRT. In these experiments we did not make a detailed cellular characterization of the rescued organoids, and this may be a task in our next experiments for an exact characterization of the cell-specific CRT expresion and function in the rescued brain organoids. According to this, we will correct in the revision version of manuscript the statement on page 6: “SLC6A8 expressing brain organoids showed GFP fluorescence in the whole area of the organoid (Fig S2C).”

Author Response

Reviewer #2 (Public Review):

The current work was basically a follow-up of a previous study in juvenile mice, and the results were also very similar to the juvenile results (Sommeijer et al., 2017). One possible interpretation of the results is that the lack of OD plasticity in adult V1 and dLGN was caused by an early blockade of the development of the inhibitory circuit in dLGN, which retains the dLGN in an immature stage till adulthood. The authors indeed claimed in the discussion that the 2-day OD shift is intact in juvenile dLGN and V1 in KO mice, and provided evidence in supplementary figure that GABAergic and cholinergic synapse amount are similar between WT and KO mice. However, the 7-day OD shift is indeed defected in juvenile V1 and dLGN in KO mice (Sommeijer et al., 2017), and it is possible that this early functional deficit didn't induce a structural remodeling in adulthood. To better support the author's claim that the lack of adult V1 OD plasticity is specifically due to reduced dLGN synaptic inhibition, the author should generate conditional KO mice that dLGN synaptic inhibition was only interfered in adulthood.

In order to address this point it is important to discuss the plasticity deficits in dLGN and V1 separately.

Concerning V1 plasticity: We have previously shown that brief MD during the standard critical period induces an OD shift in V1 of mice lacking thalamic synaptic inhibition in dLGN (Sommeijer et al, 2017). OD plasticity induced by brief MD is a hallmark of critical period plasticity in V1, and it thus seems unlikely that critical period onset in V1 is defective or that development of V1 is halted in an immature state that does not support OD plasticity in thalamus-specific GABRA1 deficient mice.

The observed plasticity deficit during the critical period was limited to the second stage of the OD shift in V1, which requires long-term monocular deprivation. The straightforward explanation for this result and our current findings is that both during the critical period and in adulthood, the second stage of OD plasticity in V1 induced by long-term monocular deprivation requires thalamic plasticity or inhibition. The proposed alternative, that lack of thalamic synaptic inhibition during development results in a possible lack of structural change in V1 that would cause a lifelong deficiency selectively affecting OD plasticity induced by long-term monocular deprivation, is not impossible but requires many more assumptions.

Concerning dLGN plasticity: The simplest explanation for the observed lack of OD plasticity in dLGN is that it is a direct consequence of the absence of synaptic inhibition in the KO mice. However, an alternative explanation could indeed be that dLGN is kept in an immature (pre-critical period-like) state due to the developmental absence of synaptic inhibition. This situation would be analogous to that in V1 of GAD65 deficient mice (which have reduced GABA release), in which OD plasticity cannot be induced by brief monocular deprivation during the critical period or in adulthood (Fagiolini and Hensch, 2000). Because this deficit can be reversed by treating the mice with benzodiazepines (positive allosteric modulators of GABA receptors) at any age, it is thought that development of V1 in GAD65 mice is halted in a pre-critical period-like state until inhibition is strengthened. We cannot exclude that something similar occurs in dLGN of mice lacking thalamic synaptic inhibition, although we did not observe any changes in hallmarks of dLGN maturity, such as reduced receptive field size (Fig. 1C), and increased cholinergic and inhibitory bouton densities (Suppl. Fig. 1).

However, if the analogy with the developmental deficit in V1 of GAD65 deficient mice is valid, the reduced plasticity is still likely to be a direct consequence of reduced inhibition. In GAD65 deficient mice, long-term monocular deprivation during the critical period causes a full OD shift, showing that no additional deficits (besides reduced inhibition) limit OD plasticity in V1 of these mice (Fagiolini and Hensch, 2000). And, as already mentioned, increasing inhibition rescues OD plasticity in GAD65 KO mice. Thus, the immature state of V1 in these mice is probably a situation in which inhibition tone is too low to support efficient OD plasticity. In dLGN, knocking out GABRA1 at a later age could therefore also create a situation in which inhibition is too low to support thalamic OD plasticity, which is not different from the situation in which the gene is inactivated at birth. Only if lack of synaptic inhibition in thalamus affects another, unknown developmental process that is of importance later in life to support OD plasticity in dLGN, the proposed experiment would result in a different outcome. We are not convinced that this scenario is likely enough to justify repeating most of this study, but now using mice in which GABRA1 is inactivated in dLGN through bilateral AAV-cre injections.

Independently of the exact cause of the plasticity deficit in dLGN, our results make clear that a cortical plasticity deficit in adulthood can have a thalamic origin, which we believe is an important insight that is highly relevant.

2) The authors found that in juveniles, dLGN OD shift is dependent on V1 feedback, but not in adults. However, a recent work showed that the effects of V1 silencing on dLGN OD plasticity could differ with various starting points and duration of the V1 silencing and MD (Li et al., 2023). Could the authors provide more details of the MD and V1 silencing for an in-depth discussion?

We would be happy to include some discussion about this interesting new paper in a revised manuscript. Some of the results may appear to contradict our findings. Most strikingly, the study by Li et al does not find evidence for OD plasticity in dLGN of 60-day old mice after 7 days of monocular deprivation. This seems to be at odds with the current work and with that of (Jaepel et al 2017) and (Huh et al. 2020). However, in the (Li et al, 2022) study, only the binocular neurons which responded to both contralateral and ipsilateral stimulus were included to measure the OD. This has important consequences for assessing OD and its plasticity. To illustrate: if dLGN neurons are monocularly responsive to the contralateral eye and become binocular after deprivation of the contralateral eye, they are excluded from analysis before deprivation but included after. This would cause an underestimation of the size of this OD shift. In our experiments, all dLGN neurons with receptive fields that were within 30o degrees away from the center of the visual field were included in the analysis, potentially explaining the different outcome of the studies.

Also, Li et al observed that an OD shift in dLGN was still present after silencing V1 at p24. This observation is not necessarily at odds with our observation that the OD shift reduces at p30 upon silencing V1, as we find that the ODI does not return to normal but remains slightly lower (though not significantly so). Moreover, the age and the duration of deprivation were different and as mentioned before, analysis was performed differently.

Interestingly, an excitotoxic lesion of V1 was found to alter OD in dLGN during development and affect OD plasticity in dLGN at various ages in the work of Li et al. This suggests that continuous crosstalk between thalamus and cortex during development guides plasticity, possibly optimizing thalamocortical and corticothalamic connections. The continued absence of corticothalamic feedback is likely to have a much larger impact on dLGN plasticity than the acute silencing we performed.

Fagiolini M, Hensch TK. Inhibitory threshold for critical-period activation in primary visual cortex. Nature. 2000 Mar 9;404(6774):183-6.

Huh CYL, Abdelaal K, Salinas KJ, Gu D, Zeitoun J, Figueroa Velez DX, Peach JP, Fowlkes CC, Gandhi SP. Long-term Monocular Deprivation during Juvenile Critical Period Disrupts Binocular Integration in Mouse Visual Thalamus. J Neurosci. 2020 Jan 15;40(3):585-604. doi: 10.1523/JNEUROSCI.1626-19.2019

Jaepel J, Hübener M, Bonhoeffer T, Rose T. Lateral geniculate neurons projecting to primary visual cortex show ocular dominance plasticity in adult mice. Nat Neurosci. 2017 Dec;20(12):1708-1714

Li N, Liu Q, Zhang Y, Yang Z, Shi X, Gu Y. Cortical feedback modulates distinct critical period development in mouse visual thalamus.. iScience. 2022 Dec 7;26(1):105752.

Sommeijer JP, Ahmadlou M, Saiepour MH, Seignette K, Min R, Heimel JA, Levelt CN. Thalamic inhibition regulates critical-period plasticity in visual cortex and thalamus. Nat Neurosci. 2017 Dec;20(12):1715-1721.

Author Response

We sincerely appreciate the reviewers for investing their valuable time in assessing our manuscript. We understand the considerable effort involved in the review process, and we will make use of these suggestions in order to make the revised manuscript more complete in terms of explanation, discussion, additional simulations, experiments and analyses.

-Specifically, we will experimentally and computationally investigate how activation via anti-CD3 antibodies relates to our mechanism.

-We will also utilize a weaker pMHC binder in the pMHC-mediated T cell activation experiments.

-We will improve the description of the function of the FG loop and the role of the connecting peptide (CP).

-Furthermore, we will improve our description of and justification for the simulation methodology. We want to emphasize that all potentials have been described, and we will draw attention to these methodological descriptions where needed.

The reviewers also suggested a number of additional simulations that are probably beyond our current capability. These include:

-simulations of TCR in complex with a weaker agonist -simulations of the proline and alanine TCR mutants in complex with a pMHC.

While we agree that such simulations would provide new insights into the mechanism of TCR triggering, they simply are not feasible at this time. We will give a more detailed explanation for these arguments in the revised manuscript.

Below, please find our point-by-point planned action items:

Reviewer #1 (Public Review):

The manuscript entitled: "TCR-pMHC complex formation triggers CD3 dynamics" by Van Eerden et al. mainly uses coarse-grained molecular dynamics to probe the dynamic changes, in terms of CDε spatial arrangements around 226 TCRs, before and after the engagements of MCC/I-Ek. The broader distributions of CDε iso-occupancies after pMHC binding correlate with the decreases of TCR-CD3 contacts and extensions of TCR conformations. Given the observed release of motion restrictions upon antigen recognition, the authors proposed a "drawbridge" model to describe the initial triggering processes from pMHC association to TCR straightening, FG-loop getaway, and CD3 enhanced mobility. In addition, the authors briefly investigated the functional effects of the rigidified connecting peptide (CP) in T-cell activation using in silico and in vitro mutagenesis. The manuscript raises an important and exciting hypothesis about the allostery of TCR-CD3 during TCR triggering; however, due to current not-yet-convincing evidence, both computationally and experimentally, in supporting their conclusions.

I would like to see additional work before supporting the publication of this manuscript in Life. See details below:

1) As mentioned by the authors, the TCR triggering and T cell activation have been illustrated by a number of models, such as mechanosensing and kinetic proofreading, "in which TCRs discriminate agonistic from antagonistic pMHCs." However, the critical feature of antigen discrimination is lacking in the drawbridge model. So far, the CDε movements qualitatively distinguish on and off states. The simulation of the antagonist or weaker binder would strengthen the manuscript by demonstrating the relevance of CDε mobility in the triggering mechanism. 226 TCR associated with K99E/I-Ek has been resolved in Ref (DOI: 10.4049/jimmunol.1100197), which can potentially serve as the "intermediate" system to formulate the gradual increase of CDε dynamics.

Planned actions:

-Explain why the current study can not easily address pMHC discrimination

-Explain why simulation of antagonist or weaker binding pMHC is technically difficult

2) The linkage between conserved motifs in CP and CDε mobility is less apparent to this reviewer. The notion of the rigidified hinge (PP) requires further clarification. Computationally, the details of fine-grained simulations are required to justify the origin of the apparent mobility increase in PP. The direct comparison between Fig. 2 and Fig. 7 can help assess the relevance of CP through the alignment by FG-loop at a fixed direction in polar coordinates. Experimentally, anti-CD3 positive experiments and, ideally, another antagonist on 3A9 TCRs can strengthen the current functional assay. The baseline level of TCR expression (after positive selection) and 0h activation (Fig. S8) is missing.

Planned actions:

-Provide additional analysis of the role of CP as a hinge

-Better clarify the FG simulation methodology

-Align the CG and the FG polar plots

-Perform experiments with anti-CD3 antibody 2C11

-Perform additional experiment using weaker agonist (HEL peptide mutant)

-Measure baseline-level TCR expression

-Perform T cell activation experiments at t=0 h

3) Regarding the section "The TCRβ FG loop acts as a gatekeeper," besides contact analysis, additional motion analysis, such as RMSF or PCA, can further establish the importance of FG loops.

Planned actions:

-Perform additional analyses of FG loop dynamics

4) The discussion on anti-CD3 antibody effects and their potential contribution to CD3 mobility is highly recommended.

Planned actions:

-We will add the discussion of anti-CD3 antibody effects

Reviewer #2 (Public Review):

In this research article a new allosteric mechanism for T cell receptor (TCR) triggering upon peptide-MHC complex binding is presented in which conformational change in the TCR facilitates activation by controlling CD3 dynamics around the TCR. The authors find that the Cb FG loop acts as a gatekeeper and Cb connecting peptide acts as a hinge to control TCR flexibility.

As an initial result, the authors set up two sets of simulations - TCR-CD3 and pMHC-TCR-CD3 in POPC bilayers and identified that the CD3e chains exhibit a wider range of mobility in the pMHC-TCR-CD3 system as compared to the TCR-CD3 system. Next, they examined the contacts between all subunits during the course of both simulations and established that CD3g and CD3eg made far fewer contacts with TCRb in the pMHC-TCR-CD3 simulations. Next, they identified that the TCR is extended in the pMHC-TCR-CD3 simulations with larger tilt angle of 150º and FG loop acts as gatekeeper that allows CD3 movements upon pMHC binding. Finally, Mutations in Cb connecting peptide regions indicated rigidified TCR leading to hypersensitive TCR, proved both by simulations and in vitro experiments.

The following major concerns must be addressed.

Major concerns:

1) The simulations were performed with intracellular regions unfolded and free from the membrane. A more complete system should have the intracellular regions embedded in the membrane. An NMR structure of CD3e is available (Xu et al., Cell, 2008) and could have been modeled into the TCR-CD3 system before the simulation. Prakaash et al., (PLoS, Comput Biol, 2021) studied cytoplasmic domain motions during in their simulation experiments.

Planned actions:

-Explain why we can not perform adequate additional simulations of ITAMs

2) Comparing Fig. 2C and Fig.7C, the movement of CD3eg is more restricted in Fig.7C. Is this because the simulation time is lower in the mutation experiments?

Planned actions:

-Explain the differences between the CG and FG polar plots

3) Only TCR-CD3 simulation were performed for PP and AA mutants. A simulation with pMHC (pMHC-TCRmutants-CD3) should be performed to show increased CD3 mobility.

Planned actions:

-Explain why TCR-CD3-pMHC simulations of the mutants are not feasible at this time

4) Using CD3e antibody, OKT3, for activation instead of pMHC as a separate experiment would add more value to this study. They can look at CD3 mobility and TCR elongation in the system with OKT3 antibody and compare it to the CD3 mobility and TCR elongation with the pMHC system. They can also use OKT3 with AA and PP mutants and perform both simulation and in vitro activation experiments.

Planned actions:

-Perform anti-CD3 (2C11) experiments

-Perform CG simulation of TCR with CD3 Fab fragment

-Explain why we cannot perform FG simulations of TCR mutants with CD3

5) The activation experimental data is rather underwhelming. The difference between WT and PP in 2hr experiment at 0.016 ug/mL looks exceedingly low. A stronger TCR-pMHC system should be considered for the in vitro activation experiments.

Planned actions:

-Explain that this is a dilution curve, which is why at lower concentrations the effect is smaller, but at higher concentrations the effect is clear

6) There is some concern that the scientific work lacks solid experimental functional data and lack of novelty due to earlier TCR-CD3 simulation studies (Pandey et al., 2021, eLife) that already reported flexibility and elongation of the complex.

Planned actions:

-Explain the similarities and difference between this and Pandey’s work; clarify how our study contributes novel findings

Reviewer #3 (Public Review):

The authors first explore structural differences of unbound TCR-CD3 complexes and pMHC-bound TCR-CD3 complexes with coarse-grained simulations. In the simulations with pMHC-bound complexes, the transmembrane (TM) domains of the TCR-CD3 complex and of pMHC are embedded in two opposing membrane patches. In the pMHC membrane patch, a pore is created and stabilised in the simulation setup with the aim to allow water transport in and out of the compartment between the membranes. The authors report a more upright conformation of the TCR extracellular (EC) domain in the simulations in which this EC domain is bound to pMHC, compared to simulations with unbound TCR, and postulate an allosteric signalling model based on these apparent conformational changes and associated changes in TCR-CD3 quaternary arrangements. Subsequently, the authors identify a GxxG motif in the TCRbeta connecting peptide between EC domain and TM domain as putative hinge in allosteric signalling, and explore the effect of double proline and double alanine substitutions in atomistic simulations and experiments.

While these simulation and experimental setups and the addressed questions are of interest in the field, the following weaknesses prevail in my overall assessment of the work:

(1) I am not convinced that the reported coarse-grained simulation results are sound or allow to draw the conclusions stated in the work. In the simulations with a pMHC-bound TCR-CD3 complex, the intermembrane distance in the setup shown in Figure S1 appears excessively large and likely leads to a rather strong force in the membrane-vertical direction and to the reported upright conformation of the TCR EC domain. This upright confirmation thus appears to be a consequence of force from the simulation setup, rather than a consequence of pMHC binding alone as suggested by the authors. While the membrane pore in principle allows water exchange, the relaxation of the intermembrane distance resulting from this water exchange in the 10 microsecond long simulation trajectories is not (but needs to be) addressed. This relaxation eventually would lead to an equilibrated membrane separation, in which essentially no force is exerted on the TCR-pMHC EC complex. However, I suspect that this computationally demanding equilibration is not achieved in the simulations, with the consequence that forces on the TCR-pMHC EC complex in the membrane-vertical direction remain.

In addition, I am not convinced that the Martini force field of the coarse-grained simulations allows a reliable assessment of the quaternary interactions between the TCR and CD3 EC domains. Getting protein structures and interactions right in coarse-grained simulations is notoriously difficult. In simulations with the coarse-grained Martini force field, secondary protein structures are constrained as a standard procedure, and the authors also use a recommended Go-potential procedure, likely to stabilise tertiary protein structures. The quaternary interactions between the TCR EC domain and the pMHC EC domain are modelled by rather strong harmonic constraints to prevent dissociation. While the treatment of the quaternary interactions between the TCR EC domain and the CD3 EC domains in the simulations is not (but needs to be) addressed in detail, I suspect that there are no additional, or only weak constraints to stabilise these interactions. In any case, I think that a faithful representation of these quaternary interactions is beyond the reach of the Martini force field, as is the reported diffusion of the CD3 EC domains around the TCR EC domain, which plays a central role in the allosteric mechanism proposed by the authors (see Fig 2 and 5).

Planned actions:

-We will provide further description and justification for the CG simulations

(2) The allosteric model suggested by the authors is motivated in an introduction that appears to omit central controversial aspects in the field, as well as experimental evidence that is not compatible with allosteric conformational changes in the TCR. These aspects are:

• The mechanosensor model is controversial. In original versions of this model, a transversal force has been postulated to be required for T cell activation. However, more recent single-molecule force-sensor experiments reported in J Goehring et al., Nat Commun 12, 1 (2021) provide no evidence for a scenario in which transversal forces beyond 2 pN are associated with T cell activation.

• The role of catch bonds is controversial. Evidence for TCR catch bonds has been mainly obtained in experimental setups using the biomembrane force probe, in which force is applied to TCRs on the surface of T cells, but is not reproduced in experimental setups using isolated TCRs, see e.g. L Limozin et al., PNAS 116, 16943 (2019)

• Ref. 1 of the manuscript prominently discusses the kinetic segregation model of T cell activation, which is not (but needs to be) addressed in the introduction. In this model, exclusion of CD45 from close-contact zones around pMHC-bound TCRs triggers T cell activation. The model is supported by evidence from diverse experiments, see for example M Aramesh et al., PNAS 118, e2107535118 (2021) and Ref. 1. At least part of this evidence is not compatible with mechanosensing or allosteric models of T cell activation.

Planned actions:

-We will add the requested literature references and include a better description of the kinetic segregation model

Author Response:

The major criticism from the reviewers is that factors other than high-impact rare variants – such as environmental factors or epistasis – could have produced the complex tail architecture that we test for and detect. While we did explain this point in the Discussion, we agree with the reviewers that this should have been emphasized more and earlier in the manuscript.

Regarding suggestions for more complex simulations and methods, we absolutely agree that much more work is needed here to produce optimised inference of all the causes of complex tail architecture. We are performing multiple projects at various stages of completion that we hope will contribute to this, but we felt that this was a good stopping-point in this project to publish what we had completed so far, in order to: (1) introduce the idea of inferring complex genetic architecture from siblings without requiring genetic data, (2) outline an initial theoretical framework for inferring complex tail architecture from sibling data, (3) provide simple tests powered to identify enrichments of de novo or ‘Mendelian’ variants in the tails (albeit tests that make several strong simplifying assumptions), (4) enable others interested in the topic to build upon this work now. However, we plan to expand our simulations and analyses in a revised manuscript based on reviewer feedback.

We thank the reviewers for their comments about the value of our work, its mathematical robustness and the promise of our method.

Author Response:

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

Reviewer #1 (Public Review):

[…] Overall, the authors build a convincing case for TEs being an important source of regulatory information. I don't have any issues with the analysis, but I am concerned about the sweeping claims made in the title. Once you get rid of eQTLs that could be altered by either SNPs or TIPs and include only those insertions that show strong evidence of selection, the number of genes is reduced to only 30. And even in those cases, the observed linkage is just that, not definitive evidence for the involvement of TEs. Although clearly beyond the scope of this analysis, transgenic constructs with the TEs present or removed, or even segregating families, would have been far more convincing. 

We notice that the referee thinks that we "built a convincing case for TEs being an important source of regulatory information". This is what we wanted to convey in the title, were we were cautious to not claiming that TEs are the most important contributor to gene expression variability in rice populations. However, we agree with the referee that the title may be improved to better describe the results presented. We have therefore changed the title to "Transposons are an important contributor to gene expression variability under selection in rice populations".

With respect to demonstrating causality by removing or introducing the TEs, this is indeed a work we plant to do but that, as stated by the referee, is beyond the scope of this analysis.

The fact that many of the eQTL-TIPs were relatively old is interesting because it suggests that selection in domesticated rice was on pre-existing variation rather than new insertions. This may strengthen the argument because those older insertions are less likely to be purged due to negative effects on gene expression. Given that the sequence of these TEs is likely to have diverged from others in the same family, it would have been interesting to see if selection in favor of a regulatory function had caused these particular insertions to move away from more typical examples of the family. 

The TIP-eQTL are from different classes, superfamilies and families and the number of TIP-eQTLs of the same family is too small to deduce sequence communalities (4.6 TIP-eQTLs/family in indica and 3.6 TIP-eQTLs/family in japonica). On the other hand the effect of TIPs on expression can be positive or negative (we show actually that it is often negative). In the later case, a plausible scenario would be of the insertion inactivating a promoter element, and in this case it would be the insertion itself, and not the actual sequence of the TE what would be selected.

Also, previous work done in our lab has shown that TEs can amplify and mobilize transcription factor binding sites that are bound by the TF even when they are not close to a gene and therefore probably not directly affecting gene expression (Hénaff et al.,2014. The Plant Journal). In that case, the sequence of the eQTL TEs and those that are far away from genes will not necessarily differ. 

Reviewer #2 (Public Review):

In this manuscript, Castanera et al. investigated how transposable elements (TEs) altered gene expression in rice and how these changes were selected during the domestication of rice. Using GWAS, the authors found many TE polymorphisms in the proximity of genes to be correlated to distinct gene expression patterns between O. sativa ssp. japonica and O. sativa ssp. indica and between two different growing conditions (wet and drought). Thereby, the authors found some evidence of positive selection on some TE polymorphisms that could have contributed to the evolution of the different rice subspecies. These findings are underlined by some examples, which illustrate how changes in the expression of some specific genes could have been advantageous under different conditions. In this work, the authors manage to show that TEs should not be ignored when investigating the domestication of rise as they could have played an important role in contributing to the genetic diversity that was selected. However, this study stops short of identifying causations as the used method, GWAS, can only identify promising correlations. Nevertheless, this study contributes interesting insights into the role TEs played during the evolution of rice and will be of interest to a broader audience interested in the role TEs played during the evolution of plants in general. 

We agree with the referee that the results presented do not allow concluding on causality, and we have been careful not to pretend they would in the manuscript. We plan to perform analysis of adding or removing TEs by CRIPR/Cas 9 approaches to address this, but, in line with referee's 1 comment, we think this is beyond the scope of this analysis.

---------- 

Reviewer #1 (Recommendations For The Authors): 

Everything that I need to say is provided in the public portion of my review. 

Reviewer #2 (Recommendations For The Authors): 

Major concerns:

1. The authors compare the proportion of the variance explained by the most significant TIP and SNP on the observed eQLTs associated with TIPs and SNPs. Thereby the authors conclude that TIPs explain more variance than SNPs. If I am not mistaken the GWAS was run separately for TIPs and SNPs, however, I am wondering if running the GWAS on the combined TIP and SNP dataset might be the better way to compare the variance explained by TIPs and SNPs on gene expression differences. It would be nice to see if these results also hold true if a TIP and SNP combined dataset is used as the most significant marker in a GWAS might not be the causal mutation but might just be linked to the causal mutation. Further in the TIP dataset, the number of markers is only 45k and in the SNP dataset, it is 1 000k, which could bias the GWAS toward finding markers that explain more of the variation in the dataset with fewer markers. 

We addressed the reviewer concern by using two complementary approaches, whose results are described in the text (lines 119-121) and in the new Figure 1-figure supplement 1.

First, we addressed the concern regarding the independent GWAS for TIPs and SNPs vs a combined strategy. For this, we built new japonica/indica genotype matrices containing all TIP and SNP matrix together and ran eQTL mapping again. Using the same strategy (association + FDR adjust), we found 100% of the previous TIP-eQTLs and 99% of the previous SNP-eQTLs. We repeated the same analysis (proportion of expression variance), and the results were mostly the same (Figure 1-figure supplement 1A).

Second, we addressed the two concerns (combined genotypes and different amount of TIP and SNP markers) using a single approach. SNP matrices were LD pruned using a r2 = 0.9 and later subsampled to the exact number of TIPs (Indica = 30,396, Japonica = 25,168). We verified that these SNPs covered well the 12 rice chromosomes. SNP and TIP genotypes were later merged into a single matrix, and eQTL mapping was repeated for each of the subspecies and conditions using the same parameters as in the previous version of the manuscript. 100 % of the previously reported TIP-eQTL associations were found using this new approach. Nevertheless, we found a very important drop of sensitivity in the SNP-eQTLs (only 15-20% of the previous associations were detected), possibly due to the strong reduction in the number of SNPs (> 95 %), which results in much lower number of markers at < 5Kb from genes). We repeated the analysis of Figure 1D, and observed very similar results (Figure 1-figure supplement 1D). There is a very important number of TIP-eQTL associations that do not coincide with SNP-eQTLs, (74% in indica, 83% in japonica) indicating that TIP-eQTL mapping is complementary to SNP-eQTL mapping as it uncovers additional associations (note that in this case the overlap between TIP-eQTLs and SNP-eQTLs is lower than in the previous analysis due to the lower sensitivity of SNP-eQTL mapping using less markers). In the cases were both a TIP and a SNP coincide as eQTL, TIPs explained slightly more variance than SNPs in both indica and japonica (in 54% of the cases TIP variance > SNP variance).

2. Line 146 to 152: in this section, the authors describe overlaps between TIP-eQTLs in two different growth conditions, however, in the text it is not mentioned if the TIPs have the same effect on gene expression in the two conditions or if the gene expression is up-regulated in one condition but down-regulated in the other. This information would be interesting to have here, especially as the authors go on to say that only a small number of TIP-eQTLs are stress-specific. The same comment also goes for the eQTL overlap described on lines 167 to 170. 

We checked the effect type (positive or negative) of TIP-eQTLs in both scenarios (associations shared between wet/dry conditions, and associations shared between subspecies). In both cases, 100 % of the shared TIP-eQTLs have the same effect type in the two conditions or subspecies. We have updated the text accordingly (Lines 55-157 and Lines 179-181)

3. Lines 192 to 196: the authors mention that the frequency of non-eQTL-TIPs was at the same frequency in indica and japonica, which is in contrast to eQTL-TIPs. However, on line 132 it is mentioned that eQTL-TIPs were overrepresented in 1 kb regions upstream of genes. Hence, is the pattern of the frequency of non-eQTL-TIPs being at the same frequency in indica and japonica also observed in the 1 kb regions upstream of genes and/or if the distribution of non-eQTL-TIPs is matched to one of the eQTL-TIPs? Or is this pattern driven by non-eQTL-TIPs far away from genes?

We checked the frequencies of TIPs at 1Kb upstream genes and found that the general pattern is maintained, with the frequencies of TIP no-eQTLs being more correlated than that of TIP-eQTLs. We have included this information (lines 204-206) an added a new supplementary file (Figure 2-figure supplement 2)

4. In the discussion, the authors could briefly discuss how linked selection affecting TIPs could contribute to the observed results. After reading the second example in the result section where one of the example TIPs (TIP_50059) is found on the Hap B which contains "some additional structural differences" (line 290), I was left wondering how much of the increase in TIP frequency can be attributed to genetic hitchhiking? And how much of the results could be caused by linked selection, especially when considering that structural variations are not included in the GWAS analyses. 

We agree with the referee in that some of the TIP eQTLs here described might be not the actual cause of expression variability (ej, TIP linked with the causal mutation), although we cannot know the exact fraction. This is stated in several places of the results and discussion sections. However, the fact that TIPs tend to explain more variance than SNPs and that TIP eQTL, but not SNP eQTL, tend to concentrate in the upstream proximal region of genes where most transcription regulatory sequences are located (Figure 1), suggest that TIP eQTLs could be more frequently the causal than SNP eQTLs. We revised the text to ensure that we convey this message appropriately.

Minor comments: 

• Lines 80 to 83: the description of the rice phylogeny should be moved to the introduction. 

Done (Lines 68-72)

• Line 177 to 186: It was unclear to me if the authors checked in the ancestral rice population laced the TIPs described in this section as recently inserted in the indica and japonica ssp. It would be nice to add this information to this section. 

Thanks to the referee comment we noted an imprecision in the text. The approximate 1/3 of subspecies specific TIP-eQTLs refers to the TIPs at 3% MAF (ie, some of these insertions could be present at > 3% in indica, but at < 3% MAF in japonica). We now indicate only the TIPs that are truly specific to any of the two subspecies (frequency is zero in one of the two) and looked for their presence in rufipogon:

59 insertions are indica-specific. Of those, 33 are present in rufipogon.

21 insertions are japonica-specific. Of those, 5 are present in rufipogon.

We have incorporated this information in the manuscript (Lines 185-189). The species-specific TIPs are also available in the Supplementary File 3.

• Line 353: "have two of more TIPs" should be "two or more" 

Done (Line 369)

• Figure 1D: Using a square layout instead of a rectangle layout for the plot will make it easier to interpret. 

Done.

Author Response:

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

Reviewer #1 (Public Review):

[…] This novel system could serve as a powerful tool for loss-of-function experiments that are often used to validate a drug target. Not only this tool can be applied in exogenous systems (like EGFRdel19 and KRASG12R in this paper), the authors successfully demonstrated that ARTi can also be used in endogenous systems by CRISPR knocking in the ARTi target sites to the 3'UTR of the gene of interest (like STAG2 in this paper).

We thank the referee for highlighting the novelty and potential of the ARTi system.

ARTi enables specific, efficient, and inducible suppression of these genes of interest, and can potentially improve therapeutic target validations. However, the system cannot be easily generalized as there are some limitations in this system:

• The authors claimed in the introduction sections that CRISPR/Cas9-based methods are associated with off-target effects, however, the author's system requires the use CRISPR/Cas9 to knock out a given endogenous genes or to knock-in ARTi target sites to the 3' UTR of the gene of interest. Though the authors used a transient CRISPR/Cas9 system to minimize the potential off-target effects, the advantages of ARTi over CRISPR are likely less than claimed.

We thank the reviewer for raising these very valid concerns about potential off-target effects related to the CRISPR/Cas9-based gene knockout or engineering of endogenous ARTi target sites. In contrast to conventional RNAi- and CRISPR-based approaches, such off-target effects can be investigated prior to loss-of-function experiments through comparison between parental and engineered cells, which in the absence of CRISPR-induced off-target events are expected to be identical. Subsequent ARTi experiments provide full control over RNAi-induced off-target activities through comparison of target-site engineered and parental cells. However, we agree that undetected CRISPR/Cas9-induced off-target events cannot be ruled out in a definitive way, which we have pointed out in our revised manuscript.

• Instead of generating gene-specific loss-of-function triggers for every new candidate gene, the authors identified a universal and potent ARTi to ensure standardized and controllable knockdown efficiency. It seems this would save time and effort in validating each lost-of-function siRNAs/sgRNAs for each gene. However, users will still have to design and validate the best sgRNA to knock out endogenous genes or to knock in ARTi target sites by CRISPR/Cas9. The latter is by no-means trivial. Users will need to design and clone an expression construct for their cDNA replacement construct of interest, which will still be challenging for big proteins.

We fully agree that the required design of gene-specific sgRNAs and subsequent CRISPR-engineering steps are by no means trivial. However, we believe that decisive advantages of the method, in particular the robustness of LOF perturbations and additional means for controlling off-target activities, can make ARTi an investment that pays off. In our experience, much time can be lost in the search for effective LOF reagents, and even when these are found, questions about off-target activity remain. While ARTi overcomes many of these challenges by providing a standardized experimental workflow, we do not propose to replace all other LOF approaches by this method. Instead, we would position ARTi as a unique orthogonal approach for the stringent validation and in-depth characterization of candidate target genes, as we have highlighted in our revised discussion.

• The approach of knocking-out an endogenous gene followed by replacement of a regulatable gene can also be achieved using regulated degrons, and by tet-regulated promoters included in the gene replacement cassette. The authors should include a discussion of the merits of these approaches compared with ARTi.

We thank the reviewer for pointing out these alternative LOF methods. We had already included a brief discussion of chemical-genetic LOF methods based on degron tags. While we certainly share the current excitement about degron technologies, they inevitably require changes to the coding sequence of target proteins, which can alter their regulation and function in ways that are hard to control for. In our revised discussion, we have added a brief comparison to conventional tet-regulatable expression systems, which unlike ARTi require the use of ectopic tet-responsive promoters. Overall, we would position ARTi as an orthogonal tool that enables inducible and reversible LOF perturbations without changing the coding sequence and the endogenous transcriptional control of candidate target genes.

Reviewer #2 (Public Review):

[…] The ARTi system is based on expression of a transgene with an artificial RNAi target site in the 3'-UTR as well as a TET-inducible miR-E-based shRNAi. Using this system, the authors convincingly show that they can target strong oncogenes such as EGFRdel19 or KRasG12 as well as synthetic lethal interactions (STAG1/2) in various human cancer cell lines in vivo and in vitro.

The system is very innovative, likely easy to be established and used by the scientific community and thus very meaningful.

We thank the reviewer for her/his enthusiasm about ARTi.

Reviewer #1 (Recommendations For The Authors):

• The authors claimed that ARTi enables specific, efficient, inducible, and reversible suppression of any gene of interest. However, there are no experiments supporting the reversible suppression of their gene of interest. Data are required to support this statement.

We thank the reviewer for pointing this out. The statement about the reversibility ARTi-mediated effects was based on a rich body of literature that has demonstrated the reversibility of Tet-shRNAmir-induced LOF perturbations and associated phenotypes. As ARTi employs the same Tet-shRNAmir expression vectors, we have no reason to believe that this feature would be lost. However, since we have not demonstrated this in our study, we have removed this statement in our revised manuscript.

• In Figure 1E, the authors did make the point by including trametinib treated samples as positive controls. However, the trametinib treated samples also made the transcriptome changes in the ARTi groups hard to read. I wonder what the PCA analysis will look like if the authors exclude the trametinib treated groups.

In Figure 1E, we used PCA as a common and easy-to-digest visualization tool to showcase the neutrality of ARTi shRNAmirs. Given the complete absence of significantly deregulated genes for all three ARTi shRNAmirs (Figure 1F), we believe that a PCA analysis of just these samples would merely represent experimental noise and not yield additional insights.

• This universal and potent ARTi should ensure standardized and controllable knockdown efficiency, however, the knockdown efficiency for KRASG12R is not as potent as that for EGFRdel19. The authors should discuss the differences.

We thank the reviewer for pointing this out. The exact level of knockdown on the protein level is hard to determine due to detection limits of the used method. The differences in the extent to mRNA knockdown could be attributable to cleavage efficiencies due to potential secondary structures in the respective mRNAs. We suspect that the KRASG12R transgene expresses at higher levels, compared to EGFRdel19. We might therefore still be looking at the same overall magnitude of knockdown. As we did not perform a detailed analysis of the respective knockdown levels, we do not feel comfortable in stating differences in knockdown levels and therefore do not think that addressing potential differences are justified.

• It is interesting to see that, unlike other cancer types, tumor burdens did not decrease after inducing knockdown of STAG1 in STAG2 knockout HCT116 lines in Figure 2L. Have the authors examined senescence markers in this set of mice?

We have not investigated these markers and thank the reviewer for this suggestion. More detailed analyses of the phenotype are planned.

• Have the authors carefully examined the transcriptome changes induced or if not across all targets at least in the case of ARTi knock into the 3'UTR of STAG1?

We thank the reviewer for this suggestion. This would indeed be interesting to conduct for STAG1/2, especially for genes with an integration of the ARTi into the 3’UTR. The reason why we did not perform this analysis with our cell lines is that we used a construct that also adds an AID tag to STAG1 (STAG1_AID_V5_P2A_Blasti_STOP_ARTi), as outlined in the methods section. After the engineering, STAG1 carries the ARTi sequence in the 3’UTR but is also fused to AID::V5. In addition a P2A::Blasticidin_resistance Protein is made from the same transcript. We chose to use this complex strategy with the aim of comparing AID mediated degradation with ARTi-mediated knockdown. Unfortunately, the AID-based approach did not work, and we were not able to observe a reduction in protein levels. We however observed lower expression of STAG1 in the engineered versus the parental cells, likely caused by the tag, and consequently did not conduct gene expression analyses, as we would not be able to assess if transcriptome changes could be solely ascribed to the changes in the 3’UTR. The knockdown levels are hence only analyzed on the protein level.

Reviewer #2 (Recommendations For The Authors):

This is a fantastic paper, easy to read and provides a very meaningful new and innovative approach for drug target validation. I think the manuscript could be further improved by adding a section to the discussion outlining other approaches that could be used to solve the same problem. For example, Bill Kaelin came up with a strategy of expressing shRNA- or sgRNA-resistant and rtTA- or tTA-regulated cDNAs of essential gene-of-interest followed by sh/sgRNA-mediated ablation of the endogenous gene (e.g.PMID: 28082722), which is conceptually quite similar to the ARTi approach. Similarly, people have used conditional degron tags such as AID tags, dTags, HALOTags, IHZF3 degrons or SMASh either knocked into the endogenous locus or as rescue transgene. Comparing and contrasting the pros and cons of these methods to the ARTi-based approach would be certainly beneficial to the readers.

We thank the referee for pointing out these alternative LOF methods. We certainly share the current excitement about various degron tags and are applying them in our own research. In our view, a major downside of these strategies is that they inevitably require changes to the coding sequence of target proteins, which can alter their regulation and function in ways that are hard to predict and control for. We had briefly mentioned this distinguishing feature in our discussion. The strategy proposed by Bill Kaelin, i.e. rescue of the the endogenous gene through Tet-regulated expression of sh/sgRNA-resistant cDNAs, indeed shares many features of the ARTi system, but requires expression of the candidate target from an ectopic promoter element. In contrast, ARTi enables similar perturbations of candidate genes without altering their endogenous transcriptional regulations – a feature that we have highlighted in our revised discussion.

All my other comments outlined below should be considered minor and are not essential.

1, Suppl Fig.1 C: Please explain what the red star means. How can the knock-out be more than 100%. Please specify what the controls are. Why does shRNA660 exhibit no knockdown at all?

The red star indicates ARTi-shRNAmirs that were selected for further characterization. Depicted GFP knockdown levels are normalized to the performance of shRen.713, a well-characterized potent control shRNA targeting Renilla Luciferase. Values of more than 100% mean that the respective shRNA exceeded effects of shRNA.713. shRNA.660 served as a neutral control – its target site was not included in the reporter construct. We thank the reviewer for bringing up these points, which we have clarified in the legend.

2, x-axis label in Suppl Fig. 1D is missing

We thank the referee for spotting this and have added this information to the figure and its legend.

3, I would argue that ARTi6634 also has a slight effect in MV4-11 similar to its effect to RN2. Maybe add that to the text.

We thank the reviewer and have added this observation to our revised text.

4, Suppl. Figure Legend 1F - specify which cell line was used (HT-1080 presumably)

We apologize for this mistake and now have indicated the cell line in the legend.

5, Fig. 2A and E, it might be nice to add the dsRED fusion to the schematics so that the reader sees the difference between the endogenous and the endogenous. One could then also change the color to red instead of blue.

We thank the reviewer for this suggestion and adapted the figure accordingly.

6, Fig. 2B - In the third lane, there appears to be a residual band of the endogenous EGFR despite the fact that it should be KO. Is this a EGFR wt lysate with EGFR::dsRED::ARTi overexpression and as such a type in the legend or is this a non-complete KO? It might be beneficial to label the legend with EGFR::dsRED::ARTi instead of EGFR::ARTi have one arrow depicting EGFR and one additional arrow showing the EGFR::dsRED fusion (as in Fig. 1F).

We thank the reviewer for this insightful comment. We interpret the WB signal in lane three as potential cleavage/degradation products of the transgene as all signal disappears upon ARTi-mediated knockdown. Due to space reasons, we would prefer to keep the label as it is. The exact nature of the transgene is stated in the text and in the methods section.

7, Suppl Fig. 2d: It is interesting that there is such a huge upregulation of DUSP6 in cells that express EGFR::ARTi compared to parental? The figure legend states: expression levels of DUSP6 in parental and engineered PC-9 cells. I assume the first box (EGFR::ARTi -/ dox -) is the parental line? Is there really a 5x upregulation of DUSP6 upon overexpression of EGFR::ARTi compared to parental (despite the fact that the endogenous EGFR::ARTi is expressed to similar levels compared to the endogenous EGFR)? Please clarify a little better which of the cells are parental and which are EGFR KO and which are transduced with EGFR::ARTi. Might suffice to just explain in the supplmental figure legend that expression of the exogenous EGFR::ARTi in EGFR KO cells leads to increased expression of ERK targets such as DUSP6 and EPHA2 etc.

We thank the reviewer for this comment. We ascribe the increased expression of DUSP6 to the forced expression of the oncogenic variant of EGFR (EGFRdel19) while only a subset of EGFR genes in PC-9 cells is mutated and the rest is wild-type. Therefore, the net-output of EGFR signaling would be higher, even if the EGFR protein levels were exactly the same, as the EGFR gene is only present in the oncogenic form in the engineered cells but a mixture of mutant and wild-type proteins would make up the EGFR pool in the parental cells. The figure legend was changed accordingly, highlighting that DUSP6 is a MAPK downstream gene.

8, Suppl Fig. 2e: Similar to my comment #7. Expression of endogenous EGFR is lost upon KO of EGFR, but cylcinD1 expression as well as expression of other ERK target genes increases upon loss of the endogenous EGFR gene with concomitant expression of EGFR::ARTi . It is nice to see that most of those genes are down-regulated upon DOX treatment. However, CyclinD1 is strongly up-regulated - any idea why? Might be nice to comment on this in the supplemental material to make it easy for the reader to interpret the data.

We agree with the reviewer that the direct MAPK target genes follow the expected pattern of strong downregulation. We have not studied the expression of CCND1 in detail and therefore cannot comment on the mechanistic basis of this observation.

9, Fig. 2F - might be nice to provide some densitometry data to quantify the effect of ARTi-mediated KRasG12R knock-down.

We thank the reviewer for this suggestion and apologize that this data is not available for this study. We will include densitometry data in upcoming studies involving ARTi. As the observed knockdown was almost complete and hence readily observable by eye, we did not measure the effects using densitometry. In addition, we would like to mention that the sensor assay contains a quantitative analysis of the knockdown levels.

10, Fig. 2I, it might be nice to add the V5 tag to the schematic and mention the V5 tag in the text: ... and homozygously inserted ARTi target sites into the 3'-UTR as well as a V5 tag to the endogenous STAG1 alleles (Fig. 2i)

We thank the reviewer for the suggestion and explained the exact makeup of the construct better in the main text. We would however like to keep the figure as simple as possible and put the focus on the endogenous engineering here.

11, Fig. 2J - might be nice to provide some densitometry data to quantify the effect of ARTi-mediated STAT1::V5 knock-down.

We thank the reviewer for this suggestion and apologize that this data is not available for this study. We will include densitometry data in upcoming studies involving ARTi. As the observed knockdown was almost complete and hence readily observable by eye, we did not measure the effects using densitometry. In addition, we would like to mention that the sensor assay contains a quantitative analysis of the knockdown levels.

12, Suppl. Fig 4B: the authors write: 'Western blotting confirmed ... the homozygous insertion of the targeting cassette into the STAG1 locus, ...' . I think the WB nicely shows insertion of the V5 tag into the STAG1 locus, but it I think WB cannot show homozygous insertion. The fact that in Suppl Fig 1B STAG1 expression is (almost) completely ablated, is a good indication, but in Fig. 2J, there is still about 50% expression. As such, proofing homozygous insertion by PCR/Sanger sequencing or densitometry over several experiments or just rephrasing the text a little might be beneficial.

We agree with the reviewer and have adapted the respective passage in the main text.

Competing interests statement: A patent application related to the design and use of the ARTi system entitled ‘Methods and molecules for RNA interference (RNAi)’ has been submitted by T.H., M.H., J.Z. and R.N. to the European Patent Office (application EP21217407.2).

Author Response:

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

Reply to Public Reviews:

Reply to Reviewer #1:

This is a carefully performed and well-documented study to indicate that the FUS protein interacts with the GGGGCC repeat sequence in Drosophila fly models, and the mechanism appears to include modulating the repeat structure and mitigating RAN translation. They suggest FUS, as well as a number of other G-quadruplex binding RNA proteins, are RNA chaperones, meaning they can alter the structure of the expanded repeat sequence to modulate its biological activities.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript. We are very happy to see the reviewer for highly appreciating our manuscript.

1. Overall this is a nicely done study with nice quantitation. It remains somewhat unclear from the data and discussions in exactly what way the authors mean that FUS is an RNA chaperone: is FUS changing the structure of the repeat or does FUS binding prevent it from folding into alternative in vivo structure?

Response: We appreciate the reviewer’s constructive comments. Indeed, we showed that FUS changes the higher-order structures of GGGGCC [G4C2] repeat RNA in vitro, and that FUS suppresses G4C2 RNA foci formation in vivo. According to the established definition of RNA chaperone, RNA chaperones are proteins changing the structures of misfolded RNAs without ATP use, resulting in the maintenance of proper RNAs folding (Rajkowitsich et al., 2007). Thus, we consider that FUS is classified into RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Reply to Reviewer #2:

Fuijino et al. provide interesting data describing the RNA-binding protein, FUS, for its ability to bind the RNA produced from the hexanucleotide repeat expansion of GGGGCC (G4C2). This binding correlates with reductions in the production of toxic dipeptides and reductions in toxic phenotypes seen in (G4C2)30+ expressing Drosophila. Both FUS and G4C2 repeats of >25 are associated with ALS/FTD spectrum disorders. Thus, these data are important for increasing our understanding of potential interactions between multiple disease genes. However, further validation of some aspects of the provided data is needed, especially the expression data.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript and also for her/his important comments that helped to strengthen our manuscript.

Some points to consider when reading the work:

1. The broadly expressed GMR-GAL4 driver leads to variable tissue loss in different genotypes, potentially confounding downstream analyses dependent on viable tissue/mRNA levels.

Response: We thank the reviewer for this constructive comment. In the RT-qPCR experiments (Figures 1E, 3C, 4G, 6D and Figure 1—figure supplement 1C), the amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts expressed in the same tissue, to avoid potential confounding derived from the difference in tissue viability between genotypes, as the reviewer pointed out. To clarify this process, we have made the following change to the revised manuscript.

(1) On page 30, line 548-550, the sentence “The amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts in the same sample” was changed to “The amounts of G4C2 repeat transcripts were normalized to those of gal4 transcripts expressed in the same tissue to avoid potential confounding derived from the difference in tissue viability between genotypes”.

2. The relationship between FUS and foci formation is unclear and should be interpreted carefully.

Response: We appreciate the reviewer’s important comment. We apologize for the lack of clarity. We showed the relationship between FUS and RNA foci formation in our C9-ALS/FTD fly, that is, FUS suppresses RNA foci formation (Figures 3A and 3B), and knockdown of endogenous caz, a Drosophila homologue of FUS, enhanced it conversely (Figures 4E and 4F). We consider that FUS suppresses RNA foci formation through altering RNA structures and preventing aggregation of misfolded G4C2 repeat RNA as an RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Reply to Reviewer #3:

In this manuscript Fujino and colleagues used C9-ALS/FTD fly models to demonstrate that FUS modulates the structure of (G4C2) repeat RNA as an RNA chaperone, and regulates RAN translation, resulting in the suppression of neurodegeneration in C9-ALS/FTD. They also confirmed that FUS preferentially binds to and modulates the G-quadruplex structure of (G4C2) repeat RNA, followed by the suppression of RAN translation. The potential significance of these findings is high since C9ORF72 repeat expansion is the most common genetic cause of ALS/FTD, especially in Caucasian populations and the DPR proteins have been considered the major cause of the neurodegenerations.

Response: We would like to thank the reviewer for her/his time for evaluating our manuscript. We are grateful to the reviewer for the insightful comments, which were very helpful for us to improve the manuscript.

1. While the effect of RBP as an RNA chaperone on (G4C2) repeat expansion is supposed to be dose-dependent according to (G4C2)n RNA expression, the first experiment of the screening for RBPs in C9-ALS/FTD flies lacks this concept. It is uncertain if the RBPs of the groups "suppression (weak)" and "no effect" were less or no ability of RNA chaperone or if the expression of the RBP was not sufficient, and if the RBPs of the group "enhancement" exacerbated the toxicity derived from (G4C2)89 RNA or the expression of the RBP was excessive. The optimal dose of any RBPs that bind to (G4C2) repeats may be able to neutralize the toxicity without the reduction of (G4C2)n RNA.

Response: We appreciate the reviewer’s constructive comments. We employed the site-directed transgenesis for the establishment of RBP fly lines, to ensure the equivalent expression levels of the inserted transgenes. We also evaluated the toxic effects of overexpressed RBPs themselves by crossbreeding with control EGFP flies, showing in Figure 1A. To clarify them, we have made the following changes to the revised manuscript.

(1) On page 8, line 166-168, the sentence “The variation in the effects of these G4C2 repeat-binding RBPs on G4C2 repeat-induced toxicity may be due to their different binding affinities to G4C2 repeat RNA, and their different roles in RNA metabolism.” was changed to “The variation in the effects of these G4C2 repeat-binding RBPs on G4C2 repeat-induced toxicity may be due to their different binding affinities to G4C2 repeat RNA, and the different toxicity of overexpressed RBPs themselves.”.

(2) On page 29, line 519-522, the sentence “By employing site-specific transgenesis using the pUASTattB vector, each transgene was inserted into the same locus of the genome, and was expected to be expressed at the equivalent levels.” was added.

2. In relation to issue 1, the rescue effect of FUS on the fly expressing (G4C2)89 (FUS-4) in Figure 4-figure supplement 1 seems weaker than the other flies expressing both FUS and (G4C2)89 in Figure 1 and Figure 1-figure supplement 2. The expression level of both FUS protein and (G4C2)89 RNA in each line is important from the viewpoint of therapeutic strategy for C9-ALS/FTD.

Response: We appreciate the reviewer’s important comment. The FUS-4 transgene is expected to be expressed at the equivalent level to the FUS-3 transgene, since they are inserted into the same locus of the genome by the site-directed transgenesis. Thus, we suppose that the weaker suppressive effect of FUS-4 coexpression on G4C2 repeat-induced eye degeneration can be attributed to the C-terminal FLAG tag that is fused to FUS protein expressed in FUS-4 fly line. Since the caz fly expresses caz protein also fused to FLAG tag at the C-terminus, we used this FUS-4 fly line to directly compare the effect of caz on G4C2 repeat-induced toxicity to that of FUS.

3. While hallmarks of C9ORF72 are the presence of DPRs and the repeat-containing RNA foci, the loss of function of C9ORF72 is also considered to somehow contribute to neurodegeneration. It is unclear if FUS reduces not only the DPRs but also the protein expression of C9ORF72 itself.

Response: We thank the reviewer for this comment. We agree that not only DPRs, but also toxic repeat RNA and the loss-of-function of C9ORF72 jointly contribute to the pathomechanisms of C9-ALS/FTD. Since Drosophila has no homolog corresponding to the human C9orf72 gene, the effect of FUS on C9orf72 expression cannot be assessed. Our fly models are useful for evaluating gain-of-toxic pathomechanisms such as RNA foci formation and RAN translation, and the association between FUS and loss-of function of C9ORF72 is beyond the scope of this study.

4. In Figure 5E-F, it cannot be distinguished whether FUS binds to GGGGCC repeats or the 5' flanking region. The same experiment should be done by using FUS-RRMmut to elucidate whether FUS binding is the major mechanism for this translational control. Authors should show that FUS binding to long GGGGCC repeats is important for RAN translation.

Response: We would like to thank the reviewer for these insightful comments. Following the reviewer’s suggestion, we perform in vitro translation assay again using FUS-RRMmut, which loses the binding ability to G4C2 repeat RNA as evident by the filter binding assay (Figure 5A), instead of BSA. The results are shown in the figures of Western blot analysis below. The addition of FUS to the translation system suppressed the expression levels of GA-Myc efficiently, whereas that of FUS-RRMmut did not. FUS decreased the expression level of GA-Myc at as low as 10nM, and nearly eliminated RAN translation activity at 100nM. At 400nM, FUS-RRMmut weakly suppressed the GA-Myc expression levels probably because of the residual RNA-binding activity. These results suggest that FUS suppresses RAN translation in vitro through direct interactions with G4C2 repeat RNA.

Unfortunately, RAN translation from short G4C2 repeat RNA was not investigated in our translation system, although the previous study reported the low efficacy of RAN translation from short G4C2 repeat RNA (Green et al., 2017).

Author response image 1.

(A) Western blot analysis of the GA-Myc protein in the samples from in vitro translation.

(B) Quantification of the GA-Myc protein levels.

We have made the following changes to the revised manuscript.

(1) Figure 5F was replaced to new Figures 5F and 5G.

(2) On page 14-15, line 326-330, the sentence “Notably, the addition of FUS to this system decreased the expression level of GA-Myc in a dose-dependent manner, whereas the addition of the control bovine serum albumin (BSA) did not (Figure 5F).” was changed to “Notably, upon the addition to this translation system, FUS suppressed RAN translation efficiently, whereas FUS-RRMmut did not. FUS decreased the expression levels of GA-Myc at as low as 10nM, and nearly eliminated RAN translation activity at 100nM. At 400nM, FUS-RRMmut weakly suppressed the GA-Myc expression levels probably because of the residual RNA-binding activity (Figure 5F and 5G).”.

(3) On page 15, line 330-332, the sentence “Taken together, these results indicate that FUS suppresses RAN translation from G4C2 repeat RNA in vitro as an RNA chaperone.” was changed to “Taken together, these results indicate that FUS suppresses RAN translation in vitro through direct interactions with G4C2 repeat RNA as an RNA chaperone.”.

(4) On page 37, line 720-723, the sentence “For preparation of the FUS protein, the human FUS (WT) gene flanked at the 5¢ end with an Nde_I recognition site and at the 3¢ end with a _Xho_I recognition site was amplified by PCR from pUAST-_FUS.” was changed to “For preparation of the FUS proteins, the human FUS (WT) and FUS-RRMmut genes flanked at the 5¢ end with an Nde_I recognition site and at the 3¢ end with a _Xho_I recognition site was amplified by PCR from pUAST-_FUS and pUAST- FUS-RRMmut, respectively.”.

(5) On page 41, line 816-819, the sentence “FUS or BSA at each concentration (10, 100, and 1,000 nM) was added for translation in the lysate.” was changed to “FUS or FUS-RRMmut at each concentration (10, 100, 200, 400, and 1,000 nM) was preincubated with mRNA for 10 min to facilitate the interaction between FUS protein and G4C2 repeat RNA, and added for translation in the lysate.”.

5. It is not possible to conclude, as the authors have, that G-quadruplex-targeting RBPs are generally important for RAN translation (Figure 6), without showing whether RBPs that do not affect (G4C2)89 RNA levels lead to decreased DPR protein level or RNA foci.

Response: We appreciate the reviewer’s critical comment. Following the suggestion by the reviewer, we evaluate the effect of these G-quadruplex-targeting RBPs on RAN translation. We additionally performed immunohistochemistry of the eye imaginal discs of fly larvae expressing (G4C2)89 and these G-quadruplex-targeting RBPs. As shown in the figures of immunohistochemistry below, we found that coexpression of EWSR1, DDX3X, DDX5, and DDX17 significantly decreased the number of poly(GA) aggregates. The results suggest that these G-quadruplex-targeting RBPs regulate RAN translation as well as FUS.

Author response image 2.

(A) Immunohistochemistry of poly(GA) in the eye imaginal discs of fly larvae expressing (G4C2)89 and the indicated G-quadruplex-targeting RBPs.

(B) Quantification of the number of poly(GA) aggregates.

We have made the following changes to the revised manuscript.

(1) Figures 6E and 6F were added.

(2) On page 6-7, line 135-137, the sentence “In addition, other G-quadruplex-targeting RBPs also suppressed G4C2 repeat-induced toxicity in our C9-ALS/FTD flies.” was changed to “In addition, other G-quadruplex-targeting RBPs also suppressed RAN translation and G4C2 repeat-induced toxicity in our C9-ALS/FTD flies.”.

(3) On page 15, line 344-346, the sentence “As expected, these RBPs also decreased the number of poly(GA) aggregates in the eye imaginal discs (Figures 6E and 6F).” was added.

(4) On page 15, line 346-347, the sentence “Their effects on G4C2 repeat-induced toxicity and repeat RNA expression were consistent with those of FUS.” was changed to “Their effects on G4C2 repeat-induced toxicity, repeat RNA expression, and RAN translation were consistent with those of FUS.”

(5) On page 16, line 355-357, the sentence “Thus, some G-quadruplex-targeting RBPs regulate G4C2 repeat-induced toxicity by binding to and possibly by modulating the G-quadruplex structure of G4C2 repeat RNA.” was changed to “Thus, some G-quadruplex-targeting RBPs regulate RAN translation and G4C2 repeat-induced toxicity by binding to and possibly by modulating the G-quadruplex structure of G4C2 repeat RNA.”

(6) On page 19, line 417-421, the sentence “We further found that G-quadruplex-targeting RNA helicases, including DDX3X, DDX5, and DDX17, which are known to bind to G4C2 repeat RNA (Cooper-Knock et al., 2014; Haeusler et al., 2014; Mori et al., 2013a; Xu et al., 2013), also alleviate G4C2 repeat-induced toxicity without altering the expression levels of G4C2 repeat RNA in our Drosophila models.” was changed to “We further found that G-quadruplex-targeting RNA helicases, … ,also suppress RAN translation and G4C2 repeat-induced toxicity without altering the expression levels of G4C2 repeat RNA in our Drosophila models.”.

Reply to Recommendations For The Authors:

1) It is not clear from the start that the flies they generated with the repeat have an artificial vs human intronic sequence ahead of the repeat. It would be nice if they presented somewhere the entire sequence of the insert. The reason being that it seems they also tested flies with the human intronic sequence, and the effect may not be as strong (line 234). In any case, in the future, with a new understanding of RAN translation, it would be nice to compare different transgenes, and so as much transparency as possible would be helpful regarding sequences. Can they include these data?

Response: We thank the editors and reviewers for this comment. We apologize for the lack of clarity. We used artificially synthesized G4C2 repeat sequences when generating constructs for (G4C2)n transgenic flies, so these constructs do not contain human intronic sequence ahead of the G4C2 repeat in the C9orf72 gene, as explained in the Materials and Methods section. To clarify the difference between our C9-ALS/FTD fly models and LDS-(G4C2)44GR-GFP fly model (Goodman et al., 2019), we have made the following change to the revised manuscript.

(1) Schema of the LDS-(G4C2)44GR-GFP construct was presented in Figure 3—figure supplement 1.

Furthermore, to maintain transparency of the study, we have provided the entire sequence of the insert as the following source file.

(2) The artificial sequences inserted in the pUAST vector for generation of the (G4C2)n flies were presented in Figure 1—figure supplement 1—source data 1.

2) It is really nice how they quantitated everything and showed individual data points.

Response: We thank the editors and reviewers for appreciating our data analysis method. All individual data points and statistical analyses are summarized in source data files.

3) So when they call FUS an RNA chaperone, are they simply meaning it is changing the structure of the repeat, or could it just be interacting with the repeat to coat the repeat and prevent it from folding into whatever in vivo structures? Can they speculate on why some RNA chaperones lead to presumed decay of the repeat and others do not? Can they discuss these points in the discussion? Detailed mechanistic understanding of RNA chaperones that ultimately promote decay of the repeat might be of highly significant therapeutic benefit.

Response: We appreciate these critical comments. Indeed, we showed that FUS changes the higher-order structures of G4C2 repeat RNA in vitro, and that FUS suppresses G4C2 RNA foci formation. According to the established definition of RNA chaperone, RNA chaperones are proteins changing the structures of misfolded RNAs without ATP use, resulting in the maintenance of proper RNAs folding (Rajkowitsich et al., 2007). Thus, we consider that FUS is classified into RNA chaperone. To clarify these interpretations, we revised the manuscript as follows.

(1) On page 10, line 215-219, the sentence “These results were in good agreement with our previous study on SCA31 showing the suppressive effects of FUS and other RBPs on RNA foci formation of UGGAA repeat RNA as RNA chaperones …” was changed to “These results were in good agreement with … RNA foci formation of UGGAA repeat RNA through altering RNA structures and preventing aggregation of misfolded repeat RNA as RNA chaperones …”.

(2) On page 17, line 363-366, the sentence “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure, as evident by CD and NMR analyses (Figure 5), suggesting its functional role as an RNA chaperone.” was changed to “FUS directly binds to G4C2 repeat RNA and modulates its G-quadruplex structure as evident by CD and NMR analyses (Figure 5, Figure 5—figure supplement 2), and suppresses RNA foci formation in vivo (Figures 3A and 3B), suggesting its functional role as an RNA chaperone.”

Besides these RNA chaperones, we observed the expression of IGF2BP1, hnRNPA2B1, DHX9, and DHX36 decreased G4C2 repeat RNA expression levels. In addition, we recently reported that hnRNPA3 reduces G4C2 repeat RNA expression levels, leading to the suppression of neurodegeneration in C9-ALS/FTD fly models (Taminato et al., 2023). We speculate these RBPs could be involved in RNA decay pathways as components of the P-body or interactors with the RNA deadenylation machinery (Tran et al., 2004; Katahira et al., 2008; Geissler et al., 2016; Hubstenberger et al., 2017), possibly contributing to the reduced expression levels of G4C2 repeat RNA. To clarify these interpretations, we revised the manuscript as follows.

(3) On page 18, line 392-398, the sentences “Similarly, we recently reported that hnRNPA3 reduces G4C2 repeat RNA expression levels, leading to the suppression of neurodegeneration in C9-ALS/FTD fly models (Taminato et al., 2023). Interestingly, these RBPs have been reported to be involved in RNA decay pathways as components of the P-body or interactors with the RNA deadenylation machinery (Tran et al., 2004; Katahira et al., 2008; Geissler et al., 2016; Hubstenberger et al., 2017), possibly contributing to the reduced expression levels of G4C2 repeat RNA.” was added.

4) What is the level of the G4C2 repeat when they knock down caz? Is it possible that knockdown impacts the expression level of the repeat? Can they show this (or did they and I miss it)?

Response: We thank the editors and reviewers for this comment. The expression levels of G4C2 repeat RNA in (G4C2)89 flies were not altered by the knockdown of caz, as shown in Figure 4G.

5) A puzzling point is that FUS is supposed to be nuclear, so where is FUS in the brain in their lines? They suggest it modulates RAN translation, and presumably, that is in the cytoplasm. Is FUS when overexpressed now in part in the cytoplasm? Is the repeat dragging it into the cytoplasm? Can they address this in the discussion? If FUS is never found in vivo in the cytoplasm, then it raises the point that the impact they find of FUS on RAN translation might not reflect an in vivo situation with normal levels of FUS.

Response: We appreciate these important comments. We agree with the editors and reviewers that FUS is mainly localized in the nucleus. However, FUS is known as a nucleocytoplasmic shuttling RBP that can transport RNA into the cytoplasm. Indeed, FUS is reported to facilitate transport of actin-stabilizing protein mRNAs to function in the cytoplasm (Fujii et al., 2005). Thus, we consider that FUS binds to G4C2 repeat RNA in the cytoplasm and suppresses RAN translation in this study.

6) When they are using 2 copies of the driver and repeat, are they also using 2 copies of FUS? These are quite high levels of transgenes.

Response: We thank the editors and reviewers for this comment. We used only 1 copy of FUS when using 2 copies of GMR-Gal4 driver. Full genotypes of the fly lines used in all experiments are described in Supplementary file 1.

7) In Figure5-S1, FUS colocalizing with (G4C2)RNA is not clear. High-magnification images are recommended.

Response: We appreciate this constructive comment on the figure. Following the suggestion, high-magnification images are added in Figure 5—figure supplement 1.

8) I also suggest that the last sentence of the Discussion be revised as follows: Thus, our findings contribute not only to the elucidation of C9-ALS/FTD, but also to the elucidation of the repeat-associated pathogenic mechanisms underlying a broader range of neurodegenerative and neuropsychiatric disorders than previously thought, and it will advance the development of potential therapies for these diseases.

Response: We appreciate this recommendation. We have made the following change based on the suggested sentence.

(1) On page 20-21, line 455-459, “Thus, our findings contribute not only towards the elucidation of repeat-associated pathogenic mechanisms underlying a wider range of neuropsychiatric diseases than previously thought, but also towards the development of potential therapies for these diseases.” was changed to “Thus, our findings contribute to the elucidation of the repeat-associated pathogenic mechanisms underlying not only C9-ALS/FTD, but also a broader range of neuromuscular and neuropsychiatric diseases than previously thought, and will advance the development of potential therapies for these diseases.”.

Authors’ comment on previous eLife assessment:

We thank the editors and reviewers for appreciating our study. We mainly evaluated the function of human FUS protein on RAN translation and G4C2 repeat-induced toxicity using Drosophila expressing human FUS in vivo, and the recombinant human FUS protein in vitro. To validate that FUS functions as an endogenous regulator of RAN translation, we additionally evaluated the function of Drosophila caz protein as well. We are afraid that the first sentence of the eLife assessment, that is, “This important study demonstrates that the Drosophila FUS protein, the human homolog of which is implicated in amyotrophic lateral sclerosis (ALS) and related conditions, …” is somewhat misleading. We would be happy if you modify this sentence like “This important study demonstrates that the human FUS protein, which is implicated in amyotrophic lateral sclerosis (ALS) and related conditions, …”.

Author Response:

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

Reviewer #1 (Recommendations For The Authors):

This is a list of suggestions the authors could use to improve the details of the manuscript:<br /> - it is not immediately clear what is meant by "modular" on line 38 and the corresponding paragraph. This aspect is not mentioned or developed in the Results.<br /> - the discussion of remapping vectors on lines 119-137 is particularly illuminating. It could have been interesting to generate surrogate manifolds separated by arbitrary remapping vectors and see how much the alignment metric (Procrustes shape) is sensitive to the dimensionality or amplitude of remapping vectors.<br /> - A visual comparison between Fig 1 D and H suggests a difference between the manifold geometry in experiments and in the model. It seems that the embedding dimensionality of ring manifolds is higher in the data than in the model. Is that the case? It could have been interesting to explore how much embedding dimensionality influences the alignment metric.<br /> - I could not find information about the initialization of the connectivity weights. An important possibility is that the degree of alignment (and the organization of remapping vectors) depends on the strength of initial random connectivity.<br /> - It might have been interesting to comment on the relationship between the top three PCS in Fig1 and the three readout vectors. To which extent are they aligned?<br /> - I found panels C and G in Fig 1 somewhat difficult to read. In panel C, the remapping seems to be aligned to the same position across all trials. This is not the case in panel G. I am not certain what the comparison is meant to convey, but it would help to have a similar alignment in C and G. Similarly, I was not sure what to conclude from the matrix in the right part of panel C, perhaps the legend should be expanded.<br /> - the comparison with remapping models of Misha Tsodyks could be expanded. The current discussion implies that the model of Romani & Tsodyks leads to less alignment than found in trained networks, but no direct evidence is given for that statement as far as I can tell.

Reviewer #2 (Recommendations For The Authors):

Minor points:

All mentions of 'modularity' should be replaced with 'compositionality'.

I found Supplementary Figure 2 highly confusing. I thought it was meant to help understand the analysis in Figure 1K and related figures. In the end, I never really understood what was happening in these figures. Do authors make perturbations along these different coding dimensions and compare the resulting maps? I wasn't sure what exactly the authors were calculating cosine similarity for. Maybe more exposition on this in the methods would help other readers as well.

Was there any behavioral difference when the maps were not aligned?

Why did the authors only go up to 10 contexts? Was this dependent on size of the network? Sorry if I missed this.

Are remapping event aligned to unit axes? Would this change with different nonlinearities? This could be interesting in the context of (Driscoll et all 2022) and (Wittington et al 2022).

Reviewer #3 (Recommendations For The Authors):

Cueva, Ardalan, et al. 2021 arXiv:2111.01275 showed that RNNs trained to remember two circular variables develop a toroidal geometry to store this information, so consider citing this in your section on the toroidal manifolds.

We thank the reviewers for their thoughtful comments. We appreciate that all three reviewers affirmed the importance of our work and the rigor of our approach. We believe that no major weaknesses were identified by the reviews. In our view, the comparisons between recurrent neural network models and experimental data are one of the most important contributions of our work, and all reviewers agreed that this was a core strength of the manuscript.

The reviewers highlighted several future modeling directions that are raised by our results and that we did not explore in the manuscript. For example, Reviewer 2 suggests that we train networks on a navigation task alone, freeze the weights, and then train on a context discrimination task. We agree that this kind of contextual learning paradigm is of interest and could provide insight into biological remapping, such as that observed by Low et al. (2021). We also agree with Reviewer 3’s broader point that “There are many choices that must be made when simulating RNNs and there is a growing awareness that these choices can influence the kinds of solutions RNNs develop.” It is notable that we were able to reproduce the qualitative features of the experimental data without finely tuning hyperparameters (we used default settings in PyTorch layers), using a very basic training protocol (gradient descent with gradient clipping), and without adding any hand crafted regularization (though we agree that regularization could make the RNN solution look even more like the data).

We believe that readers will benefit from reading the reviewers' suggestions, which are insightful and well-motivated. Having weighed the reviewer comments carefully, we feel that our manuscript stands as a complete scientific story. We hope that the public reviewer comments will inspire future investigations to fully explore these possibilities and unpack their outcomes at a level of detail that would not be possible in the context of our manuscript.

Thus, we have chosen to implement the following minor changes suggested by the reviewers, which we hope will improve the clarity of the text and figures (summarized below). These changes do not alter the fundamental content of the manuscript.

Text:

• We corrected a few minor typos.

• We updated the citations to follow the eLife citation style.

• To address comments from Reviewers 1 and 2: we reworded the final paragraph of the Introduction (p. 3) to remove the term “modularity” and clarify our main finding. Those sentences now read, “The RNN geometry and algorithmic principles readily generalized from a simple task to more complex settings. Furthermore, we performed a new analysis of experimental data published in Low et al.26 and found a similar geometric structure in neural activity from a subset of sessions with more than two stable spatial maps.”

• To address comments from Reviewer 1: in the first paragraph of the Results section A recurrent neural network model of 1D navigation and context inference remaps between aligned ring manifolds (p. 3), we added the sentence, “Remapping was not aligned to particular track positions, rewards, or landmarks.” to clarify that experimental result from Low et al. (2021).

• To address comments from Reviewer 3: in the final paragraph of the Results section Aligned toroidal manifolds emerge in a 2D generalization of the task (p. 11) we clarified that models were trained “to estimate position on a 2D circular track.” We also added a citation to Cueva, Ardalan et al. (2021) with the following sentence, “Notably, each toroidal manifold alone is reminiscent of networks trained to store two circular variables without remapping.”

• To address a question from Reviewer 2: in the final paragraph of the Results section Manifold alignment generalizes to three or more maps (p. 13), we added the following clarification: “In Supplemental Figure 3, we show that RNNs are capable of solving this task with larger numbers of latent states (more than three; for simplicity, we consider up to 10 states).”

• To address a comment from Reviewer 1: in the fourth paragraph of the Discussion (p. 17), we removed the sentence, “Notably, our model captured aspects of the data that these previous forward-engineered models did not explore—namely, that the ring manifolds corresponding to the correlated spatial maps were much more aligned than expected by chance and than strictly required by the task.” to focus on the key point in the following sentence that, “forward-engineered models provide insights into how neural circuits may remap, but do not answer why they do so.”

• To address comments from Reviewers 1 and 2: we reworded the penultimate paragraph of the Discussion (p. 17–18) to clarify our findings and remove the term “modularity” (except when referencing papers that themselves use that term (Driscoll et al., 2022; Yang et al., 2019)). Those sentences now read:

“When RNN architecture is explicitly designed to include dedicated neural subpopulations, these subpopulations can improve model performance on particular types of tasks (Beiran et al., 2021; Dubreuil et al., 2022). Thus, there is an emerging conclusion that RNNs use simple dynamical motifs as building blocks for more general and complex computations, which our results support. In particular, aligned ring attractors are a recurring, dynamical motif in our results, appearing first in a simple task setting (2 maps of a 1D environment) and subsequently as a component of RNN dynamics in more complex settings (e.g., as sub-manifolds of toroidal attractors in a 2D environment, see Figure 4). We can therefore conceptualize a pair of aligned ring manifolds as a dynamical “building block” that RNNs utilize to solve higher-dimensional generalizations of the task. Intriguingly, our novel analysis of neural data from Low et al. (2021) revealed that similar principles may hold in biological circuits—when three or more spatial maps were present in a recording, the pairs of ring manifolds tended to be aligned.”

• To address questions from Reviewers 2 and 3: in the first paragraph of the Methods section RNN Model and Training Procedure (p. 21), we added the sentence: “The connection weights were randomly initialized from the uniform distribution 𝑈(−√1/N, √1/N), which is the default initialization scheme in PyTorch.”

• To address a question from Reviewer 2: we added a third paragraph to the Methods section Manifold Geometry Analysis (p. 23), as follows:

“In Figure 1K, 4G, 5G, and Supplementary Figure 2B, we calculate the angles between the input and output weights and the position subspace or remapping dimension. To find this angle, we calculated the cosine similarity between each weight vector and each subspace. Cosine similarity of 0 indicates that the weights were orthogonal to the subspace, while a similarity of 1 indicates that the weight vector was contained within the subspace.”

• To address a question from Reviewer 1: we added the following sentence to the second paragraph of the Methods section Experimental Data (p. 24), “We performed the same analysis of trial-by-trial spatial stability to obtain the similarity matrices in Figure 1C and G.”

Figures and legends:

• To address a question from Reviewer 1: in Figure 1C and G, we added x-axis labels to the similarity matrices to clarify that these are trial-by-trial correlations.

• To address a question from Reviewer 1: we expanded the Figure 1C legend to clarify the experimental results as follows:

Old legend:

(C, left) An example medial entorhinal cortex neuron switches between two maps of the same track (top, raster; bottom, average firing rate by position; red, map 1; black, map 2). (C, right/top) Network-wide trial-by-trial correlations for the spatial firing pattern of all co-recorded neurons in the same example session (colorbar indicates correlation). (C, right/bottom) k-means map assignment.

New legend:

(C, left) An example medial entorhinal cortex neuron switches between two maps of the same track (top, spikes by trial and track position; bottom, average firing rate by position across trials from each map; red, map 1; black, map 2). (C, right/top) Correlation between the spatial firing patterns of all co-recorded neurons for each pair of trials in the same example session (dark gray, high correlation; light gray, low correlation). The population-wide activity is alternating between two stable maps across blocks of trials. (C, right/bottom) K-means clustering of spatial firing patterns results in a map assignment for each trial.

• To address comments from Reviewer 3: in the legend of Figure 4C, we added the sentence “Note that the true tori are not linearly embeddable in 3 dimensions, so this projection is an approximation of the true torus structure.”

• To address a question from Reviewer 2: we expanded the legend for Supplementary Figure 2 to clarify the purpose of the figure schematics as follows:

Old legend:

(A)  Schematic showing the orthogonalization of the position and context input and output weights.

(B)  Reproduced from Figure 1K.

(C-D) Schematic: How a single velocity input (blue arrows) updates the position estimate (yellow to red points) from the starting position (blue points).

(C)  Velocity input lies in the position tuning subspace (gray plane)(hypothetical). Note that the same velocity input results in different final positions.

(D)  Velocity input is orthogonal to the position tuning subspace (observed).

(E)  Schematic of possible flow fields in each of the three planes (numbers correspond to planes in C and D), which would result in the correct positional estimate given orthogonal velocity inputs at different positions (D).

New legend:

(A)  Schematic showing the relative orientation of the position output weights and the context input and output weights to the position and state tuning subspaces.

(B)  Reproduced from Figure 1K.

(C-D) Schematic to interpret why the position input weights are orthogonal to the position tuning subspace. These schematics illustrate how a single velocity input (blue arrows) updates the position estimate (yellow to red points) from a given starting position (blue points).

(C, not observed) Velocity input lies in the position tuning subspace (gray plane). Note that the same velocity input pushes the network clockwise or counterclockwise along the ring depending on the circular position

(D, observed) Velocity input is orthogonal to the position tuning subspace and pushes neural activity out of the subspace.

(E) Schematic of possible flow fields in each of three planes (numbers correspond to planes in C and D). We conjecture that these dynamics would enable a given orthogonal velocity input to nonlinearly update the position estimate, resulting in the correct translation around the ring regardless of starting position (as in D).

Author Response:

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

We thank both reviewers for their comments, which have suggested changes that have improved the manuscript.

Reviewer #1 (Public Review): 

[…] A weakness in the methodology is the link to tissue tension and conclusions about tissue mechanics. Methods that directly affect tissue tension and a more thorough and systematic application of laser ablation experiments would be needed to profoundly investigate mechanosensation and consequential effects on tissue tension by the various genetic perturbations.

Response: In revision, we have added some additional experiments that examine altered tension.

While the in-silico analysis of competing for F-actin binding sites for βH-Spec and myosin appears logical and supports the authors' claims, no point mutation or truncations were used to test these results in vivo.

In its current structure the manuscript's strength, the genetic perturbations, is compromised by missing clear assessments of knockdown efficiencies early in the manuscript and other controls such as the actual effect on myosin by ROCK overactivation. 

Response: In revision, we reorganized the manuscript and figures to document the knockdown efficiency earlier in the manuscript, and have added additional figure panels illustrating the effects of altered tension on myosin levels.

Reviewer #2 (Public Review):

[…] The authors suggest that Ajuba is required for the effect of beta-heavy spectrin. However, it is still formally possible that this could be a parallel pathway that is being masked by the strong phenotype of Ajuba RNAi flies. 

Response: While it is formally true that the genetic requirement for Jub could reflect a role in parallel to, rather than downstream of, spectrins, our conclusion that spectrins act through Jub is based not only on the genetic requirement for Jub, but also on the influence of spectrins on junctional tension and Jub localization, which indicate that spectrins influence Jub activity in a manner consistent with their affecting the Hippo pathway through Jub.

One of the major points of the manuscript is the observation that alpha- and beta-heavy-spectrin are potentially working independently and not as part of a spectrin tetramer. This is mostly dependent on the observation that alpha- and beta-heavy-spectrin appear to have non-overlapping localizations at the membrane and the fact that alpha- and beta-heavy-spectrin localize at the membrane seemingly independently. It is not entirely obvious that a potential lack of colocalization and the fact that protein localization at the membrane is not affected when the other partner is absent is sufficient to argue that alpha- and beta-heavy-spectrin do not form a complex. Moreover, it is possible that the spectrin complexes are only formed in specific conditions (e.g. by modulating tissue tension). 

Response: Our results argue that alpha- and beta-heavy-spectrin do not form a detectable complex in the wing disc under the conditions examined, and thus that they act independently is this context. However, we agree that it is possible that they could function together contexts, eg in other tissues or under different conditions, and we have revised the text in the Discussion to note this.

If indeed spectrins function independently, would it not be expected to see additive effects when both spectrins are depleted? 

Response: Not necessarily, since both alpha- and beta-heavy-spectrin act through Jub, and there may be a limit as to how much Yki activity can be increased by Jub (eg the increases in wing size induced by spectrin RNAi are similar to the increases in wing size observed with constitutive recruitment of Jub through alpha-catenin mutation (Alegot et al 2019).

Related to the two previous points, the fact that the authors suggest that both alpha- and beta-heavy-spectrin regulate Hippo signaling via Ajuba would be consistent with the necessity of an alpha- and beta-heavy-spectrin complex being formed. How would the authors explain that both spectrins require Ajuba function but work independently? 

Response: The different spectrins both affect Jub because they both affect cytoskeletal tension, but our results suggest that they act in different ways to affect tension. We have made some revisions to the Discussion section to try to make this clearer.

Another major point of the manuscript is the potential competition between beta-heavy-spectrin and myosin for F-actin binding. The authors suggest that there is a mutual antagonism between the two proteins regarding apical F-actin. However, this has not been formally assessed. Moreover, despite the arguments put forward in the discussion, it seems hard to justify a competition for F-actin when beta-heavy-spectrin seems to be unable to compete with myosin. Myosin can displace beta-heavy-spectrin from F-actin but the reciprocal effect seems unlikely given the in vitro data. 

Response: We show in vivo, in vitro, and in silico data that are all consistent with the inference that beta-heavy-spectrin and myosin compete for binding to F-actin. As the reviewer notes, and as we discuss, the in vitro competition experiments were limited because, for technical reason, we were unable to increase the protein concentrations higher. We also note that our in vitro experiments used an active form of myosin, which binds F-actin much more strongly than inactive myosin.

Reviewer #1 (Recommendations For The Authors): <br /> While the flow of experiments is logical in general, I see major problems regarding the structure of the manuscript and essential controls: 

• It is very confusing to have samples (kst-CRISPRa) in figures 1-3 that were not introduced in the text until the second-last paragraph of the results. I would suggest introducing this elegant overexpression experiment early in the manuscript as it fits well in the scope of these experiments or alternatively (if the authors prefer) make a new figure containing all the data regarding the overexpression in the end. 

Response: We have now moved these results to a new figure (new Fig 7) that is described later in the text.

• At the beginning of the manuscript, essential controls regarding the knockdown efficiency are missing in the main figure. Many of the key experiments are based on KD and as a reader, I want to assess their efficiency. Only in Figure 4, at the end of the manuscript, KST and α-Spec KD efficiency is revealed - this should be shown earlier and quantified properly. While reading the manuscript in its current form, the doubt remains that differences e.g. in α-Spec and KST KD can be explained by varying knockdown efficiencies as their levels can't be assessed. 

Response: We have now moved these results to a new supplemental figure (Fig 1-supplement 1) that is cited earlier in the text.

• On a similar line, in Figure 5 where myosin activity is perturbed, induction or repression of myosin activity is only suggested but not formally shown. The authors have to demonstrate that this is indeed the case by showing the myosin signal, ideally accompanied by measurement of tissue tension. 

Response: This was not included because we and others have assessed these manipulations in earlier publications. However, as requested we have now added a supplemental figure (Fig 6 supplement 1) showing myosin levels in these genotypes.

• On p. 7, the authors claim that "The epistasis of jub to kst suggests that βH-Spec regulates wing size through its tension-dependent regulation of Jub." While the authors show that KST KD increases myosin and junctional Jub, and that the wing overgrowth phenotype of KST KD depends on Jub, the tension-dependency was not demonstrated. To make that claim, the tension profile should be perturbed e.g. by overexpression of rok, myosin mutants (as the authors do in Fig 5) and the effect on Jub should be analyzed. Induction of tension in these conditions should be measured by laser ablation or a suitable alternative method. It might well be that the induction of Jub in KST KD is not via tension but an alternative mechanism such as the release of steric hindrance, interaction competition, etc. Also: Does KD of Jub affect spectrin localization? 

Response: The effect of tension on Jub, and the effects of the myosin activity changes we employed on tension, have been analyzed in prior publications (eg Rauskolb et al 2014). To further address the issue raised by the reviewer here as to whether Kst affects Jub and wing growth via tension, we have also now added an additional experiment (Fig 3 supplement 1) in which we decreased tension in a βH-Spec RNAi wing disc by simultaneously expressing RNAi targeting Rok. The results show that the wing growth and Jub accumulation associated with βH-Spec RNAi are suppressed by Rok RNAi, consistent with our conclusion that these effects are mediated via cytoskeletal tension.

As KD of Jub alters the pattern of myosin accumulation in wing discs (Rauskolb et al 2019) it could be expected to have a complementary influence on βH-Spec localization, but we have not examined this.

• The authors make a very strong point in saying "The influence of βH-Spec on junctional tension is thus a direct consequence of its competition with myosin for overlapping binding sites on F-actin." While the authors provide some in vitro and in silico evidence, it was for example not possible to outcompete myosin by increasing levels of KST CH1-CH2 domains in vitro (for possible reasons the authors discuss). More importantly, the hypothesis that competition for actin binding is the definite cause of the antagonizing effect was not tested in vivo. Overexpression of a mutant version of KST that is unable to bind F-actin, or that has an increased affinity (etc) for actin was not tested. Such an experiment would be very valuable to enrich this manuscript but at least, claims like that have to be less bold and need to be written in a more speculative language. 

Response: We consider creating and analyzing mutant forms of Kst in vivo to be beyond the scope of this manuscript, but as suggested we have now modified the text highlighted by the Reviewer to be more cautious.

Further points: 

• Why does the thickness of the wing disc epithelium change due to KST and α Spec KD, the authors should introduce this experiment better and draw a proper conclusion. Is there any relocalization of myosin along the apical-basal axis? Can the authors speculate about the differences between KST and α Spec KD? 

Response: The epithelium thickness changes with α-Spec KD, but does not change with Kst KD. We think the explanation is provided by work from the Pan lab (done mainly in pupal eyes), which reported decreased cortical tension and increased apical area when α-Spec is lost. The interpretation in essence is that with the loss of attachment of F-actin to membranes along the lateral sides of the cells, the sides of the cells are "softer" and the cells expand laterally and thus also (by conservation of volume) shorten apical-basally. This is somewhat speculative, and it's not a focus of our study, but we have added some text to try to explain this better. Myosin along apical-basal axis was not visibly altered, but it is harder to analyze as it is very weak compared to junctional myosin.

• Given the authors' observation of differences in the relative localization of KST and α Spec (Figure 4), proper quantification of KST, α Spec and myosin levels along the apical-basal cell axis would be important. This would also ease data interpretation. 

Response: We have now added a higher resolution image and also a line scan of Kst, α-Spec  and Myo in a new supplemental figure (Fig 6 supplement 1)

• KD of α Spec seems to induce myosin activity more, causes a bigger reduction of wing thickness, a stronger induction of Jub, and a similar effect on wing size. What lead the authors to focus on KST rather than α Spec regarding the detailed analysis of myosin competition? 

Response: Our observations identify a competition between Kst and myosin, but we have no indication that α-Spec competes with myosin. (It's conceivable that β-Spec might also compete with myosin in some contexts, but wing discs would not be a good place to examine this because the localization profiles of β-Spec and Myosin are so different).

• A big criticism regarding the figures is the bad color choice which makes it difficult to decipher the fluorescent signals. Likewise, the labels are difficult to read with the present coloring. They should really be changed. 

Response: We have now changed the single color images to gray scale (for multi-color images we retain RGB coloring).

A minor point: 

• To make the manuscript more accessible for researchers outside the Drosophila field, I'd suggest adding explanatory labels for Drosophila-specific terms such as hyperactive myosin for sqhEE, a scheme to show where UAS-dcr2 is active, explain the purpose of Rfp expression as a control for tissue specificity, etc. 

Response: We have added some explanations to the text to try to make this clearer.

Reviewer #2 (Recommendations For The Authors): <br /> Major points: 

In lines 99-101, the authors mention that Deng et al., 2015 report that the depletion of spectrins leads to an increase in pMLC, with no associated changes in the colocalization of myosin and F-actin. It is more accurate to mention that Deng et al. suggest that the levels of a GFP-tagged rescue construct of MLC (Sqh) are unchanged in alpha-spectrin mutants, although this was not formally quantified. Moreover, there was not a formal assessment of colocalization between MLC and F-actin, but rather a suggestion that F-actin levels are unaffected by the alpha-spectrin mutation. Finally, Deng et al. mostly analyzed alpha-spectrin so it remains possible that the new results shown by the authors are compatible with the initial observations from Deng and colleagues. 

Response: As suggested, we revised the text to note that Deng et al., 2015 specifically examined Sqh:GFP. While we agree that our focus is more on Kst and Deng et al focused on α-Spec, we also examined α-Spec, and as described our results examining Myosin and Jub differ from what was reported by Deng et al 2015.

As mentioned above, it is still possible that spectrins and Ajuba are working in parallel and Ajuba is not necessarily downstream of spectrins. The strong phenotype of Ajuba RNAi flies in adult wings could mask the effect of spectrins. Are the results similar in other settings, such as in the absence of Dicer2? Also, can Ajuba RNAi phenotypes be modified by overexpression of spectrins? This would provide further evidence of a link to Ajuba function. 

Response: While formally it is true that the genetic requirement for Jub could reflect a role in parallel to, rather than downstream of, spectrins, our conclusion that spectrins act through Jub is based not only on the genetic requirement for Jub, but also on the influence of spectrins on junctional tension and Jub localization, which indicate that spectrins influence Jub activity in a manner consistent with their affecting the Hippo pathway through Jub.

We would not expect over-expression of spectrins in a jub RNAi background to further reduce Hippo signaling, and as the jub RNAi phenotype is much stronger than the Kst over-expression phenotype even if there were an effect it would likely be difficult to detect.

Regarding the potential independent functions of spectrins, it would be interesting to determine if alpha- and beta-heavy-spectrin can still interact at the level of the AJ despite the fact that their distributions appear to be partly non-overlapping. Would it be possible to assess this using PLA? If an interaction is not detected via PLA, it would be more convincing that spectrins are functioning independently. 

Response: We have now performed this experiment, and no significant signal was detected by PLA. As a control, we used identical antibodies (GFP and α-Spec) to conduct PLA on α-Spec and β-Spec, and we did detect signal by PLA. These results (included in a revised Figure 4) further support the conclusion that α-Spec and βH-Spec are not physically associated in wing discs.

Related to this point, if the spectrins work independently, it is reasonable to assume that they could display additive effects. Is this the case? If alpha- and beta-heavy-spectrin are simultaneously depleted are the phenotypes more severe than either depletion alone? 

Response: We disagree here. Since both alpha- and beta-heavy-spectrin act through tension and Jub, and there is likely a limit as to how much Yki activity can be increased by this pathway. For example, the increases in wing size induced by spectrin RNAi are similar to the increases in wing size observed with constitutive recruitment of Jub through alpha-catenin mutation (Alegot et al 2019), which may thus represent the maximum increase that can be induced through this pathway (as there are multiple, independent factors that regulate Hippo signaling).

Authors should modulate membrane tension and assess if this affects the localization of alpha- and beta-heavy-spectrin and, specifically, their colocalization, as their interaction could be regulated. 

Response: As reported, we do see effects of tension on βH-Spec localization. We would not expect significant effects of membrane tension on α-Spec localization, but we consider analysis of this outside the scope of this manuscript.

In lines 185-187, the authors mention that beta-spectrin depletion does not affect beta-heavy-spectrin localization. Interestingly, Figure 4E appears to show that the levels of Kst-YFP appear to be lower in the beta-spectrin-depleted tissue. The localization of beta-heavy-spectrin is not necessarily affected but the overall levels could be. 

Response: Indeed the levels appear slightly lower, but elucidating the reason for this will require further experiments that are beyond the scope of this manuscript (we suspect it is because cytoskeletal tension increases in β-Spec-depleted tissue as it does in α-Spec depleted tissue, which based on our observations should decrease levels of Kst at near junctions). The key point of these experiments was to show that α-Spec localization does not require βH-Spec, but does require β-Spec, which supports our conclusion that in wing discs α-Spec forms a complex with β-Spec but not with βH-Spec.

In lines 200-203, the authors state that beta-heavy-spectrin and myosin colocalize extensively at the apical region. However, this colocalization is not as clear as stated. Do the authors have alternative data that suggests that the two proteins are indeed colocalizing? Would it be possible to perform PLA to detect a potential colocalization? 

Response: Unfortunately we do not have antibodies against both proteins that work well enough for PLA. However, we quantified the co-localization by analysis of Pearson's correlation coefficient, as reported in the manuscript. We also added an additional higher magnification image, and a line scan, in a supplemental figure (Fig. 6 supplement 1).

Authors should try to assess and quantify colocalization with F-actin for both beta-heavy-spectrin and myosin in wild-type conditions and when the levels (and/or activity) for each of them are modulated. 

Response: We have added quantification of the co-localization of βH-Spec with F-actin and of myosin with F-actin to the revised manuscript.

Minor points: 

In lines 122-124, the authors should clarify the relevance of the observation that alpha-spectrin knockdown affects the thickness of the wing disc epithelium. 

Response: We have added some text to try to elaborate on this.

In the intro, it is perhaps necessary to mention that there are conflicting reports regarding the role of spectrins in the regulation of cell proliferation, at least in the follicular epithelium. For instance, Ng et al., 2016 argued that spectrins do not regulate cell proliferation in FECs. 

Response: Rather than wading into a detailed discussion of issues that are peripheral to this study, we modified the text in the Introduction to avoid implying that spectrins control cell proliferation in the ovary.

In Figures 1, 2, 3, and 4 (and respective supplements), it is encouraged that, wherever appropriate, the authors mark the different compartments or the relevant boundary using dashed lines, to more clearly indicate the regions to compare. 

Response: We have now done this.

In Figure 2, supplement 1 panels C and D should have an indication of the genotype for clarity. 

Response: We have now added this.

In lines 362-367, the authors suggest that other actin-binding proteins are likely to influence the role of beta-heavy-spectrin. Have the authors tested the role of spectrin interactors such as Ankyrin and Adducin?

Response: No, we have not examined this.

Author Response:

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

We were pleased with the overall enthusiastic comments of the reviewers:

• Reviewer #1: “This manuscript by Mahlandt, et al. presents a significant advance in the manipulation of endothelial barriers with spatiotemporal precision”

• Reviewer #2: “The immediate and repeatable responses of barrier integrity changes upon light-on and light-off switches are fascinating and impressive.”

• Reviewer #3: “, these molecular tools will be of broad interest to cell biologists interested in this family of GTPases.”

We thank the reviewers for their fair and constructive comments that helped us to improve the manuscript.

Reviewer #1 (Recommendations For The Authors):

1) This paper is likely to attract a diverse audience. However, the order of data presented in this manuscript can be confusing or challenging to follow for the naive reader. This is because the tool characterization is split into two parts: before the barrier strength assay (selection of optogenetic platform and tool expression) and after (characterization of cell morphology with global and local optogenetic stimulation). Reorganizing the results such that the barrier strength results follows from an understanding of individual cell responses to stimulation may improve the ability of this readership to understand the factors at play in the changes in barrier strength observed when opto-RhoGEFs are activated.

We appreciate this idea, and we initially structured the paper in the proposed order and then decided, that we wanted to put more focus on the barrier strength results by already presenting them in the second figure. Therefore, we prefer to keep this order of figures.

2) While the description of the selection of iLID as the study's optogenetic platform is clear, a better job could be done motivating the need for engineering new optogenetic tools for the control of GEF recruitment. Given that iLID-based tools for GEFs of RhoA, Rac1, and Cdc42 already exist, some of which are cited in the introduction, more information on why these tools were not used would be helpful-were these tools tested in endothelial cells and found lacking.

The original system has the domain structure DHPH-tagRFP-SspB. But we wanted to work with a SspB-FP-GEF construct, which would allow easy exchange of the FP and the DHPH domain. This modular approach allowed us to generate and compare the mCherry, iRFP647 and HaloTag version. We don’t want to claim that we engineered an entirely new optogenetic tool but rather optimized an existing one with different tags. To make this more clear we added : ‘The membrane tag of the original iLID was changed to an optimized anchor. In addition, we modified the sequence of the domains to SspB, tag, GEF to simplify the exchange of GEF and genetically encoded tag. A set of plasmids with different fluorescent tags was created for more flexibility in co-imaging.’

3) Comment on the reason behind using DHPH vs. DH domains for each GEF is needed.

We have previously found (and this is supported by biochemical analysis of GEF activity) that the selected domains provide the best activity. We will add reference and the following to the text: ‘Their catalytic active DHPH domains were used for ITSN1 and TIAM1 (Reinhard et al., 2019).  In case of p63 the DH domain only was used, because the PH domain of p63 inhibits the GEF activity (Van Unen et al., 2015) (Fig. 1E).’

4) Since multiple Rho GTPases (e.g., RhoA, RhoB, RhoC) exist and Rho is used as the name of the GTPase family, please use RhoA where applicable for clarity.

Since the RhoGEFp63 will activate RhoA/B/C we would rather not refer to RhoA only. We will clarify this in the text: ‘Three GEFs were selected, ITSN1, TIAM1 and RhoGEFp63, which are known to specifically activate respectively Cdc42, Rac and Rho and their isoforms.’

5) A brief comment on the use of HeLa cells for protein engineering and characterization (versus the endothelial cells motivated in the introduction) may be helpful.

We added the following to the text: ‘HeLa cells were used for the tool optimization because of easier handling and  higher transfection rate in comparison to endothelial cells.’

Minor suggestions:

In figure 1C, line sections showing intensity profiles before and after protein dimerization might further emphasize the change in biosensor localization.

We are not a fan of intensity profiles as the profile depends strongly on the position of the line and it basically turns a 2D image in 1D data, for a single image. So, we prefer to stick to the quantification as shown in panel 1B (which shows data from multiple cells).

Reviewer #2 (Recommendations For The Authors):

1)The study has analyzed the effects of light-induced activation of the three optogenetic constructs in endothelial cells on their barrier function (electrical resistance) at high cell density and correlated the findings with the cellular overlap-producing effects on endothelial cells cultured at sparse cell density. It should be tried to show these effects at a cell density where these light-induced effects increase electrical resistance. Lifeact with different chromophores in adjacent cells might be useful.

We had attempted to measure the overlap in a monolayer by taking advantage of the Halotag and the variety of dyes available by staining one pool of cells red with JF 552 nm and the other far red with the JF 635 nm dye. However, the cells need at least 24 h to form a monolayer and by then they had exchanged the dye and red and far red pool could not be distinguished any longer.

Therefore, we used the Lck-mTq2-iLID construct, which already marks the plasma membrane of the cells. We created a mosaic monolayer of cells expressing mScarlet-CaaX and cells expressing Lck-mTq2-iLID + SspB-HaloTag-TIAM(DHPH). We observed and increase in the overlap between cells under this condition. The results have been added to figure 4 - figure supplement 2I&J. To the text we added:

'Additionally, cell-cell membrane overlap increased about 20 %, up on photo-activation of OptoTIAM, in a mosaic expression monolayer (figure 4 - figure supplement 2I,J, Animation 22)‘

2) The authors correctly state that some reports have shown that S1P can increase endothelial barrier function in VE-cadherin independent ways and these are related to Rac and Cdc42. This was also shown for Tie-2 in vitro and even in vitro in the absence of VE-cadherin and should also be mentioned.

We added the following to the text: ‘Not only S1P promotes endothelial barrier independent from VE-cadherin, also Tie2 can increase barrier resistance in the absence of VE-cadherin (Frye et al. 2015).’

Since a blocking antibody against VE-cadherin was used, a negative control antibody should be tested which also binds to endothelial cells.

To visualize the cell-cell junctions in the experiment shown in Supplemental Fig 3.1, we added a non-blocking VE-cadherin antibody that is directly labeled with ALEXA 647 and shows normal junction morphology. These experiments already give an indication that the live labeling antibody of VE-cadherin does not disturb the junction morphology. However, when we added the blocking antibody against VE-cadherin, known to interfere with the trans-interactions of VE-cadherin, a rapid disruption of the junctions is observed.

Additionally, previous work has shown, that VE-cadherin labeling antibody does not interfere with junction dynamics and function (see Figure 2.A, Kroon et al. 2014 ‘Real-time imaging of endothelial cell-cell junctions during neutrophil transmigration under physiological flow’, jove.). We have added the figures below, showing that addition of the control IgG and VE-cadherin 55-7H1 Abs at the timepoint where the dotted line is, did not interfere with the resistance whereas the blocking Ab drastically reduced resistance. We have added this reference to the results. ‘Previous work has shown the specific blocking effect of this antibody in comparison to the VE-cadherin (55-7H1) labeling antibody (Kroon et al., 2014).’

Author response image 1.

Reviewer #3 (Recommendations For The Authors):

Additional comments for the authors:

1) The introduction is very long and would benefit from a more concise emphasis on the information required to put the work and results in context and understand their importance.

Comment: we appreciate the comment of the reviewer. However, we wish to introduce the topic and the tools thoroughly and therefore we chose to keep the introduction as it is.

2) The N-terminal membrane-binding domain does not homogeneously translocate to the plasma membrane, since lck is a raft-associated kinase. Please comment on this.

In our hands, the Lck is among the most selective and efficient tags for plasma membrane localization (https://doi.org/10.1101/160374). We do observe homogeneous translocation, but our resolution is limited to ~200 nm and so we cannot exclude that the Lck concentrates in structures smaller than 200 nm. Given the robust performance of the lck-based iLID anchor in the optogenetics experiments, we think that the Lck anchor is a good choice.

3) Figure 1D is not very clear. What does 25 or 36% change mean? If iLID tg is conjugated to these sequences, its cytosolic localization should be reduced versus iLID alone. Is this what the graph wants to express? If so, please, label properly the ordinate axis in the graph (% of non-tagged iLID values?)

The graph is representing the recruitment efficiency of SspB to the plasma membrane for the two different membrane tags, targeting iLID to the plasma membrane. The recruitment efficiency was measured by the depletion of SspB-mScarlet intensity in the cytosol, up on light activation, and represented as a change in percentage.

We added the following to the title of the graph_: SspB recruitment efficiency for Plasma Membrane tagged iLID._

4) Supplemental figures in the main text. Fig S1D in the text refers to data in Fig S1E and Fig S1E is supposed to be Fig S1F? (page 11).

That is correct. The mistakes have been corrected (and this is now renamed to figure 1 - figure supplement 1E and 1F).

5) Figure 3. Contribution of VE-cadherin. Other junctional complexes, such as tight junctions may also intervene. However, these results would also suggest that cell-substrate adhesion rather than cell-cell junctions may modulate the barrier properties, as it has been previously demonstrated for example by imatinib-mediated activation of Rac1 (Aman et al. Circulation 2012). The ECIS system used to measure TEER in the quantitative barrier function assays can modulate these measurements and discriminate between paracellular permeability (Rb) and cell-substrate adhesion (alpha). Please, provide whether the optogenetic modulation of these GTPases does indeed regulate Rb or alpha.

The measured impedance is made up of two components: capacitance and resistance. At relatively high AC frequencies (> 32,000 Hz) more current capacitively couples directly through the plasma membranes. At relatively low frequencies (≤ 4000 Hz), the current flows in the solution channels under and between adjacent endothelial cells’ (https://www.biophysics.com/whatIsECIS.php).

Therefore, the high frequency impedance is representing cell-substrate adhesion whereas the low frequency responds more strongly to changes in cell-cell junction connections.

We only measured at 4000 Hz, representing the paracellular permeability. We chose a single frequency to maximize time resolution.

We have added this extra comment to the legend of the figure: ‘(B) Resistance of a monolayer of BOECs stably expressing Lck-mTurquoise2-iLID, solely as a control (grey), and either SspB-HaloTag-TIAM1(DHPH)(purple)/ ITSN1(DHPH) (blue) or p63RhoGEF(DH) (green) measured with ECIS at 4000 Hz, representing paracellular permeability, every 10 s.’

Author Response

eLife assessment

In this work, the authors provide important mechanistic insights into how the intracellular effector protein Calcineurin B homologous protein 3 (CHP3) can be regulated in a calcium-independent manner to expose its lipid binding site. Compelling evidence demonstrates a binding partner protein (NHE1) triggers a conformation change and exposure of the myristoyl group in CHP3 resulting in membrane association. This provides mechanistic insight into the signalling mechanisms achieved by CHP3 in a target-dependent manner, which will be of broad scientific interest.

Thank you for providing an accompanying eLife assessment. As we slightly modified the name of the novel mechanism to meet the suggestion of reviewer 2, and to emphasize the binding to a lipid membrane, we suggest the following update:

“In this work, the authors provide important mechanistic insights into how the intracellular effector protein Calcineurin B homologous protein 3 (CHP3) can be regulated in a calcium-independent manner to expose its lipid membrane binding site. Compelling evidence demonstrates a binding partner protein (NHE1) triggers a conformation change and exposure of the myristoyl group in CHP3 resulting in membrane association. This provides mechanistic insight into the signalling mechanisms achieved by CHP3 in a target-binding dependent manner, which will be of broad scientific interest.

Reviewer #1 (Public Review):

This study examines the effects of Ca2+ and NHE1 peptide binding on the conformation of CHP3, one of three related calcineurin-homologous proteins. One question that is addressed is whether Ca2+ binding triggers membrane association of the myristoyl group, a so-called "Ca2+-myristoyl switch". This is convincingly demonstrated to not be the case by the experiment in Figure 6B: unlike myristoylated recoverin, mCHP3 does not show enhanced association with liposomes. In the presence of a target peptide, however, myristoylation enhances membrane association. Curiously, this interaction is not Ca2+ dependent, but the membrane association of the non-myristoylated CHP3 is Ca2+-dependent.

My concerns with this study relate to physiological relevance. First, it is unclear if Ca2+ binding has a regulatory function in any of the CHP proteins. The authors state that CHP1 and CHP2 have Ca2+ binding affinities <100 nM, so these proteins are likely saturated with Ca2+ under all physiological conditions. On the other hand, CHP3 binds Ca2+ with a Kd of 8 micromolar (in the presence of physiological concentrations of Mg2+) so it will be largely unbound under most normal cellular concentrations of Ca2+ which are in the submicromolar range. Free Ca2+ rarely reaches 1 micromolar under non-pathological concentrations, and if it does, the fraction of CHP3 bound to Ca2+ should be estimated for context. Given these caveats, I am not convinced that experiments done with millimolar concentrations of Ca2+ (e.g., Figures 2, 3, 6) are physiologically informative.

Precise knowledge on the distinct and isoform-specific molecular basis of the important physiological roles of calcineurin homologous proteins is only emerging. Here, we ruled out the suggested Ca2+-myristoyl switch and showed that instead, target-binding (NHE1-peptide) induces membrane association of myristoylated CHP3. In respect to Ca2+ response, we showed in this work and previous studies that all CHPs undergo Ca2+-induced conformational changes, a feature that is required for EFCaBPs to act as Ca2+ sensor. Millimolar Ca2+ concentrations are commonly used in this type of in vitro characterization to ensure uniform conformational states of the protein, thus we followed this approach. We agree that in future studies, the distinct molecular responses to Ca2+ signals have to be studied in cellular context. So far, one can state that for CHP1 and CHP2, affinities for Ca2+ were reported with Kd values of ~90 nM determined in vitro in the absence of Mg2+. This is close to the cellular Ca2+ concentration in the resting cell, but would not lead to saturation of all CHP1 or CHP2 molecules in the cell with Ca2+. The presence of Mg2+ in the cell may further attenuate the affinity of CHPs for Ca2+. One cannot exclude, that CHP1 and CHP2 could respond to Ca2+ signals in the cell. For target-free CHP3, a Kd of 3.5 µM for Ca2+ in the presence of Mg2+ was reported, so it is unlikely to respond to Ca2+-signals. However, target binding (at least for NHE1) does not require the presence of Ca2+ (as shown in the present study), and target binding can increase Ca2+-binding affinity of EFCaBPs up to 100 fold (reported 45-fold for CHP1 and 42-fold for CHP2). Target-bound CHP3 might have an affinity for Ca2+ that enables a response to Ca2+-signals.

Reviewer #2 (Public Review):

The manuscript by Becker and coworkers describes a target-binding myristoyl switch in the calcium-binding EF hand protein CHP3 using one of its targets, the NHE1. The work uses a suite of biophysical methods including SEC, nanoDSF, fluorescence, and native MS, to address conformations, ligand binding (Ca2+, Mg2+, NHE1), and liposome association, pinpointing a conformation switch which they term a target-dependent myristoyl switch. The strength of the manuscript is a convincing mapping of the different conformations and the conclusion that target binding, and not Ca2+ binding is necessary to expel the lipid from the protein, and that this jointly enhances membrane binding. It would have been even stronger if additional structural data had been included to address the properties of the different states and hence support if there indeed are changes in dynamics and flexibility.

We are thankful to Reviewer #2 for a number of valuable comments on our manuscript which we addressed systematically to enhance description and discussion of our results. Specifically, we clarified the use of conformation, flexibility, state, dynamics and now consistently refer to distinct states of the protein (Ca2+-, Mg2+- and apo-state) as well as defined conformations (open, closed and target-bound). We agree that structural characterization is important, yet, it is beyond the focus of the present biochemical and biophysical characterization and needs to be addressed in future studies.

Reviewer #3 (Public Review):

This work provides new insights into the regulation of the intracellular effector protein Calcineurin B homologous protein 3 (CHP3). The authors precisely delineate how intracellular calcium signals and myristoylation affect the binding of CHP3 to lipid membranes and the sodium/proton exchanger NHE1. Different mechanisms are known to trigger the exposure of the myristoyl-moiety in the calcium-binding protein family and CHP3 was proposed to use a "calcium-myristoyl switch", which leads to exposure of the myristoyl group due to conformational changes in the protein triggered by calcium-binding. Becker and Fuchs et al. now demonstrate that CHP3 uses a novel mechanism, in which not calcium-binding but binding to the target protein NHE1 triggers exposure of its myristoyl-group. This paper represents a detailed functional characterization of CHP3 and the maximum level of mechanistic interpretation that can be achieved without high-resolution structural information.

The conclusions of this paper are fully supported by the data.

Strengths

The protein biochemistry is of an exceptionally high level, both with respect to the quality of the material and the stringency with which the authors assess and assure the protein quality. The authors purify CHP3 without any affinity tags, and thus in its most representative relevant state. Their validations indicate that complete myristoylation of CHP3 is achieved and that all protein is functional with respect to calcium binding.

The authors go to extensive lengths to convince themselves of the quality of their data and their interpretation. They use an extensive amount of replicates, including both biological and technical replicates. Assays and experimental procedures are verified using model proteins, such as Recoverin. In addition, the authors employ an extensive set of complementary approaches to assure their observations are universal.

We highly appreciate the positive feedback of Reviewer #3 on our experimental design and quality of biochemical data.

Weaknesses

A small weakness is the fact that the interpretation in terms of mechanistic insights contributed by some of the assays employed is rather limited, resulting in comparably unprecise descriptions of the state of the protein such as "affects the conformation and/or flexibility of CHP3" or the "open" and "closed" conformations. As indicated by the authors, structural studies are required to precisely detail the conformational states and delineate their mechanism of action.

We updated the manuscript for a stringent use of the descriptions “conformation”, “state” and “flexibility” to match terminology commonly used for EFCaBPs.

The authors imply that the major form of CHP3 is the myristoylated state. However, it remains unclear whether the source of the biological material, which appears to be membrane-only, already implies a significant experimental bias that only allows (or highly favors) the identification of myristoylated CHP3. Without a calcium-signal, unmyristoylated CHP may not associate with membranes, or be less strong, resulting in its depletion upon isolation of the vesicles.

We agree that our data are based on membrane fractions, so referring to the “major form of CHP3” was misleading. We updated two sentences as follows: “Finally, we investigated the N-terminal myristoylation status of membrane associated CHP3 in vivo using liquid-chromatography coupled mass spectrometry (LC-MS/MS). ………Together, this suggests that myristoylated CHP3 is both NHE1-associated and membrane-anchored in agreement with a target-induced exposure and membrane integration of the N-terminal myristoyl moiety.”

Author Response

Reviewer #1 (Public Review):

The Introduction starts by setting up a straw-man argument, claiming that the assumption is that gene expression is set up as stable expression domains that undergo little or no subsequent change. I don't think that any current developmental biologist thinks this is true. The references used to support this claim are from the 1990s up to the early 2000s. There are numerous examples since then that show that developmental gene expression is dynamic as a rule.

Our argument might seem like a strawman for certain sector of developmental biologists who work in the field of pattern formation, or aware of the latest advances in the field. However, a look at current publications on developmental enhancers reveals that the dominant model with which enhancer biologists interpret their data is still the French Flag model (specifically, the eve-stripe-2 model of enhancer function). We meant to address this audience, and attempted to clarify this from the very beginning by stating that “Much of our models of how enhancers work during development relies on the assumption that …”. Please, note here that we are talking about “models of how enhancers work”, not models of pattern formation in general.

The Introduction then continues as a rather detailed review of enhancers, Tribolium methodology, tools for identifying enhancers, and more. The Introduction cites 99 references, which seems excessive for what is essentially an experimental paper. Significant parts of the Introduction can be trimmed or removed. There is no need to mention all the tools available for Tribolium if they are not used in the described experiments. A thorough analysis of the advantages and disadvantages of different modes of ATAC-seq is also beyond the scope of the Introduction. The authors should explain why they chose the tools they chose without excessive background.

In the revised manuscript, we shortened the discussion of Tribolium methodologies and imaging techniques. However, we think that the paragraph discussing ATAC-seq strategies are important to justify our choices as why we took the effort to cut the embryos to perform tissue-specific ATAC-seq analysis, instead of performing whole-embryo ATAC-seq.

Having said that, the Introduction actually overlooks a lot of significant work that is relevant to the subject of the paper. Specifically, the authors completely ignore all of the work on development in hemimetabolous insects such as Oncopeltus and Gryllus - the omission is glaring. There has been a lot of relevant work on dynamic gene expression patterns coming out of these species.

You are right indeed. We apologize for that. We added now citations to relevant works from those to insect to the manuscript.

The experimental setup involves cutting embryos into three sections at two time points. The results then discuss differences in "space" and "time" but there is no discussion of the embryological meaning of these terms. What is happening at the two time points from a developmental perspective? What is the difference between the three sections? There is a lot of relevant development going on at these stages and important regional differences, which have been well-studied in Tribolium and in other insects but are not even mentioned.

A good point. Correlating chromatin landscape changes with embryological events is an interesting point that needs further analysis and the application of ATAC-seq to further timepoints. We chose leaving this to future work (possibly using single cell ATAC-seq). In this work, we restricted our analysis to the benefits of applying time- and tissue-specific ATAC-seq in predicting active enhancers. We added a note on this point in the discussion.

In the preliminary results of the ATAC-seq analysis, it is clear that there are significant differences between the sections, which should come as no surprise, but fairly minor differences between the same section at the two time points. This could be because the two time points are pretty close together at a stage when there is a lot of repetitive patterning going on. A possible interpretation, which the authors don't mention because it goes against their main thesis, is that maybe most of the processes that are taking place at this stage are not dynamic enough to show up at the temporal resolution they have applied. This is worth at least a mention.

We agree with this observation. We would like to draw the reviewer’s attention to our statement “Together, our findings indicate that changes in chromatin accessibility in Tribolium at this developmental stage are primarily associated with space rather than time…””. Detailed analysis of the chromatin dynamics across time would need taking more datapoints, which is something we plan to do in future work.

The authors link each accessible site to the nearest gene when looking at putative enhancer function. This is a risky assumption since there are many examples of enhancer sites that are far upstream or downstream of the target gene and often closer to an unrelated gene than to the target gene. The authors should at least acknowledge this problem with their functional annotation.

The reviewer is correct in that, in particular for large eukaryotic genomes, enhancers are often located far away from their target genes. We have no comprehensive enhancer-target data that would enable us to perform a more accurate analysis. Furthermore, the assumption that at least for some of the enhancers the nearest genes will also be their targets, and hence, provide insight into the function of the enhancers themselves seems reasonable given the relatively compact organization of the Tribolium genome. In any case, the analysis was just presented as one of several sanity checks for our ATAC-seq data; for the sake of streamlining the manuscript we no longer include this analysis in the current version of the manuscript.

In the Discussion, the authors claim that contrary to how it may seem, the question they are addressing is not a "fringe problem". Once again, I think this is a straw man. No active researcher thinks that the question of dynamic regulation of gene expression during development is a fringe problem. On the contrary, most researchers will accept that this is one of the most interesting and important questions in current developmental biology.

This whole argument was removed from the Discussion in the revised manuscript.

Perhaps the most significant problem with the manuscript is that it is all built around the premise of enhancer switching between dynamic enhancers and static enhancers. The authors find one site that is consistent with their prediction for a dynamic enhancer and one site - regulating a different gene - that is consistent with their prediction for a static enhancer and claim that they have provided support for their model. I think this claim is grossly exaggerated. They present data that can be seen as consistent with their model but are a long way from providing evidence for it.

We actually thought we were cautious enough about this. Nowhere in our text did we mention that our data “support” the enhancer switching model. We stated quite early (in the abstract, actually) that:

“We found our data consistent with a model in which the timing of gene expression during embryonic pattern formation is mediated by a balancing act between enhancers that induce rapid changes in gene expressions (that we call ‘dynamic enhancers’) and enhancers that stabilizes gene expressions (that we call ‘static enhancers’).”

To make this message clearer, we added the following sentence to the abstract of the revised manuscript: “However, more data is needed for a strong support for this or any other alternative models.” And again at the end of the Introductions: “While these data are in line with our Enhancer Switching model, more data is needed as a strong support for the model.” Also, at the end of the Results section examining runB enhancer dynamics, we stated: “However, this merely shows that runB activity dynamics are consistent with our model, but is still far from strongly supporting the model (more on that in the Discussion).” Also for the Results section on enhancer hbA dynamics: “Again, this merely shows that hbA activity dynamics are consistent with our model, but is still far from strongly supporting it.”.

Moreover, in the opening paragraph of the Discussion, we explicitly and quite openly addressed this point, and suggested what kind of observations and experiments needed in the future to qualify as a “strong support” for the model. We even ran simulations for what kind of observation should one expect in enhancer deletion experiments if the model is correct (Figure 7).

But it seems like discussing the enhancer switching model in detail gives the impression of its central importance to the paper. In our view, our experimental system is quite general and does not depend on that model, but the point of mentioning it is that it is an example of how could an alternative model of enhancer regulation be of relevance to the problem of dynamic gene expression. This wouldn’t be obvious without this or a similar model that is showing this, even if it is hypothetical. But since our presentation is obviously giving the impression that our claims are stronger that they really are, we altered our phrasing in the introduction of the revised manuscript to make our point clearer:

“Despite its potential inaccuracies, the Enhancer Switching model exemplifies the type of alternative frameworks we need to explore in order to elucidate the mechanisms driving the generation of gene expression waves during development. Consequently, an appropriate model system is required, allowing us to test not only the Enhancer Switching model but also any other prospective model that provides a satisfactory explanation for the initiation of gene expression waves at the enhancer level.”

We hope that this addresses the reviewer’s quite legitimate concerns.

Like the Introduction, the Discussion includes long paragraphs (lines 450-480) that are more suitable for a review/hypothesis paper. The data presented in this manuscript has little relevance to the question of kinematic vs. trigger waves, and therefore there is no real reason for the question to be discussed here.

We have now significantly shortened the discussion.

Reviewer #2 (Public Review):

Open questions:

What happens with the runB enhancer at later stages of embryogenesis? With what kind of dynamics do the anterior-most stripes fade and does that agree with the model? Do they show the same dynamics throughout segmentation? I think later stages need to be shown because the prediction from the model would be that the dynamics are repeated with each wave. I am not so sure about the prediction for ageing stripes – yet it would have been interesting to see the model prediction and the activity of the static enhancer.

Yes, the dynamics repeats in the germband. This is shown in Supplementary Figure 8. The dynamics in germband were shown by visualizing yellow mRNA and intronic probes. MS2 imaging was not possible to be used because the embryo dive into the yolk for a while, and then it becomes difficult to capture the germband in the right orientation for imaging. We are currently working to use light sheet microscopy for imaging germband stages.

I understand that the mRNA of the reporter gene yellow is more stable than the runt mRNA. This might interfere with the possibility to test your prediction for static enhancers: The criterion is that the stripes should increase in strength as the wave migrates towards the anterior. You show this for runB – but given that yellow has a more stable transcript – could this lead to a “false positive” increase in intensity with the slower migration and accumulation of transcripts? I would feel more comfortable with the statement that this is a static enhancer if you could exclude that the signal is blurred by an artifact based on different mRNA stability. What about re-running the simulation (with the p–rameters that have shown to well reflect endogenous –unt mRNA levels) but i“creasing the parameter for the stability of the mRNA? Are static and dynamic enhancers still distinguishable? The claim of having found a static enhancer rests on this increase in signal, hence, other explanations need to be excluded carefully.

Good questions. Note that runB reporter dynamics were examined not only by visualizing yellow mRNAs (which indeed seem to be more stable than endogenous run mRNA; see Supplementary Figure 10), but also using MS2 (with virtually zero mRNA stability; although stability was simulated in the shown movies to show virtual mRNA dynamics), and intronic yellow mRNA (showing de novo transcription; Supplementary Figure 10; you will need to zoom in to see intronic de novo transcripts). The expected dynamics of a static enhancer reporter is quite unique: it progressively increases initially as it propagates from posterior to anterior, then it progressively decreases as it slows down and stabilizes at the anterior. Then they eventually fade. These full range of dynamics is obvious in germband embryos stained for intronic yellow to show de novo transcription of runB enhancer reporter (Supplementary Figure 10; you will need to zoom in to see intronic de novo transcripts).

Running the simulation for the model using different degradation rates for the enhancer reporter made the static enhancer’s expression either less or more persistent, but gave the same overall result: the static enhancer expression has diminished expression at the very posterior, but high expression as its expression wave exiting the growth-zone/SAZ. This is consistent with not only yellow mRNA expressions of runB, but with its intronic expression as well (Supplementary Figure 10; you will need to zoom in to see intronic de novo transcripts).

What about the head domain of the runB enhancer (e.g. Fig. 6A lowest row): This seems to be different from endogenous expression in your work and in Choe et al. Is that aspect different from endogenous expression and can this be reconciled with your model?

Yes, indeed this aspect cannot be explained by our model. We believe that head patterning in insects is regulated by a different regulatory network. This network might be (de)-activated by missing repressors in the selected DNA segment for runB enhancer. We mentioned this issue in the revised manuscript.

The claim of similar dynamics of expression visualized by in situ and MS2 in vivo relies on comparing Fig. 6C with 6A. To compare these two panels, I would need to know to what stage in A the embryo in C should be compared. Actually, the stripe in 6C appears more crisp than the stripes in 6A.

Were the enhancer dynamics tested in vivo at later stages as well? I would appreciate a clear statement on what stages can be visualized and where the technical boundaries are because this will influence any considerations by others using this system.

One really cannot be that super-precise about the timing of a very dynamic process in space and time like this one we are studying. We believe that Figure 6D shows clearly that runB activity dynamics are similar to endogenous run expression.

How do the reported accessibility dynamics of runA enhancer correlate with the activity of the reporter: E.g. is the enhancer open in the middle body region but closed at the posterior part of the embryo? Or is it closed at the anterior – and if so: why is there a signal of the reporter in the head?

You show that chromatin accessibility dynamics help in identifying active enhancers. Is this idea new or is it based on previous experience with Drosophila (e.g. PMID: 29539636 or works cited in https://doi.org/10.1002/bies.201900188)? Or in what respect is this novel?

Our manuscript contributes to the growing body of evidence confirming that accessibility per se does not imply activity. Of course, this is not a new idea, but given the widely use of accessibility as a proxy for enhancer activity in the genomics community, we do feel it is important to reiterate the message. As the reviewer correctly indicates, several published findings point to a correlation between accessibility dynamics and enhancer activity. However, to our knowledge, this is the first example in Tribolium. It is important to point out that what “dynamic” means strongly depends on the experimental design. Even in Drosophila, not enough studies have been conducted to fully understand the relationship (e.g., ideally, this should be done on a continuous time scale and at single cell level). We acknowledge in the manuscript that this relationship has been observed before in other species (and have added the references suggested by the reviewer, for which we are very grateful), but still believe that our observations are highly significant to the Tribolium community.

Reviewer #3 (Public Review):

I have two major concerns: First, the claim about differential accessibility being related to enhancer activity is not really established from the presented data, in my view. This needs to be clarified. (I do believe in the claim to some extent, but not based on presented evidence.)

We agree with the reviewer that more data – and, more importantly, independent replication – are necessary to confirm this finding. Please, refer to our response to your comment regarding the statistical significance of the findings.

Second, the evidence in support of the Enhancer Switching model for runt should be accompanied by identification of and spatiotemporal profiling of the “speed regulator”, if this is not established yet.

Experiments supporting the role of Cad as a speed regulator for both pair-rule and gap genes have been published in El-Sherif et al PLOS Genetics 2014 and Zhu et al PNAS 2017. We added a comment stressing this fact.

In addition to these two concerns, the simulations of the Enhancer Switching model need to be described, at least in the outline, in the Methods section.

Done

Author Response

Reviewer #1 (Public Review):

Specifically, the authors define "efficacy" (eta) of a ligand as the fractional change in binding free energy between the open and the closed states of the channel.

We assume that the word in quotes is a typo; ղ is efficiency, not efficacy (now given the symbol λ). We now emphasize the distinction immediately after Eq. 2.

1) One concern regards the clustering of the data sets in Fig. 5 into exactly 5 eta-classes. First, two clusters contain only two data points each. Second, the proposed "catch&hold LFER model" (Fig. 2) does not predict the existence of a discrete number of such eta-classes. How strong is the evidence that there are exactly 5 classes as opposed to a continuum of possible eta values.

Statistical (x means cluster) analysis indicates that the 23 agonists segregate into 5 ղ classes. Groups with only 2 members (plus the intercept) are less well defined (Fig 4) but are supported by the 5 mutational ղ classes (Fig. 7). (see above)

2) The authors do not discuss the uniqueness of the proposed model.

see above. Ln 405 Induced fits are common.

In fact, it seems to me that the existence of eta-classes might be explained just as well by an alternative model which assumes a single gating mechanism for the receptor,

We are not sure what a “single gating mechanism” means. Does non-single refer to i) the2 stage induced fits (catch-hold LFER)? … ղ classes makes this conclusion unavoidable. ii) our conjecture that are there are 5 different C versus O binding site structural pairs…? Energy derives from structure, so we the 5 energy ratios indicate 5 structural pairs. iii) multiple steps inside gating (ϕ)? …So far there have not been any alternative explanations for the organized map of ϕ. iv) catch itself?... Evidence for this induced fit is given in Fig 2 and 7 SI, and on Ln 528-547 we discuss the implications of kon to C versus O. Ln 405 Local ‘Induced fit’ rearrangements in enzymes are common. We think the evidence is strong for the bottom scheme in Fig 2A.

but distinct patterns of ligand-protein interactions for the different agonists.

ղ classes derive from distinct interactions for different agonists, but what these are and whether the ‘contact number’ idea is useful are uncertain (see above).

The pore opening-associated increase in agonist affinity is typically caused by a tightening of the substrate binding site (often called clamshell closure) …

Ln 379-386 In the Discussion we now relate catch-hold to induced fit

Ln 455, 461-463, 471-474 Fig 2SI and the induced fit to clamshell closure

Reviewer #2 (Public Review):

This is an interesting manuscript with a worthwhile approach to receptor mechanisms. The paper contains an impressive amount of new data. These single molecule concentration response curves have been compiled with care and the authors deserve great credit for obtaining these data.

Ln 233 ղ can be estimated from a CRC built from whole-cell currents…

Ln 150 …or indeed any method that estimates KdC and KdO (for example binding assays, or perhaps in silico simulations of AC and AO structures)

I judge the main result to be that there are different values of the recently-proposed agonist-related quantity "efficiency".

Ln 21, 26-27, 535-547 OK, but to us the most interesting insight is that in AChRs binding IS gating.

These values are clustered into 5 quite closely spaced groups. The authors propose that these groups are the same whether considering mutations in the binding site or different agonists.

see above

It was unclear to me in several places, what new data and what old data are included in each figure. Therefore readers may have difficulty judging the claimed advance. This difficulty is not helped by the discussion, which includes some previous findings as "results".

see above.

A further weakness is that it is unclear how general or how specific these concepts are. The authors assert that they are, by definition, completely universal. However, we do not have reference to previous work or current data on any other receptor than the muscle nicotinic. I could not square the concept that "every receptor works like this" with the evident lack of desire to demonstrate this for any other receptor.

Ln 132-136 There are reasons to think that receptors in general work according to Figure 1A. A thermalized ligand (for instance TriMA, MW 60) has the momentum of only ~3 water molecules. A momentum sensor would have terrible signal/noise.

Reviewer #3 (Public Review):

This work attempts to introduce a new attribute of the receptor- efficiency, a fraction of an agonist binding energy consumed by conformational transition of the receptor from resting to active (open) states. Furthermore, the authors use an impressive set of experimental data (single channel recordings with 23 agonists and 53 mutations) to measure the efficiency for each agonist and mutant receptor. All the estimated efficiencies fall into a few groups and inside each of the efficiency groups there is a strong correlation between agonist affinity and receptor opening efficacy.

The main finding in this study is that estimated efficiencies fall into 5 groups.

see above.

There is no clear description of the method how the efficiencies were allocated into different groups. Most importantly, it is not clear if the method used takes into account the uncertainty of the efficiency estimate. The study does not show any statistical metrics of the efficiency estimates as well as any other calculated variable such as dissociation equilibrium constants to resting or open states. Surely, the uncertainty of the efficiency should matter especially considering how near the efficiency group values are (eg. difference about 10% between 0.51 and 0.56 or 0.41 and 0.45).

see above

All the tested agonists fell into groups according to the efficiency value attributed to them. It is difficult to see why some of the agonists belong to the same group. For example, it is not obvious at all why such agonists as epibatidine, decamethonium and TMP are in the same group. The question, I guess, arises if this grouping based on efficiency has any predictability value. Furthermore, if a series of mutations with the same agonist fall into different groups, the prediction power of this approach is very limited if one attempts to design a new agonist or look for a new mutation.

see above and Ln 548-561 (last para of text). Efficiency is a relatively new idea. This report is one of only a few on the subject. More experiments with different receptors by more labs using other approaches are needed to ascertain whether ղ is general.

Author Response

Reviewer #1 (Public Review):

This manuscript will interest cognitive scientists, neuroimaging researchers, and neuroscientists interested in the systems-level organization of brain activity. The authors describe four brain states that are present across a wide range of cognitive tasks and determine that the relative distribution of the brain states shows both commonalities and differences across task conditions.

The authors characterized the low-dimensional latent space that has been shown to capture the major features of intrinsic brain activity using four states obtained with a Hidden Markov Model. They related the four states to previously-described functional gradients in the brain and examined the relative contribution of each state under different cognitive conditions. They showed that states related to the measured behavior for each condition differed, but that a common state appears to reflect disengagement across conditions. The authors bring together a state-of-the-art analysis of systemslevel brain dynamics and cognitive neuroscience, bridging a gap that has long needed to be bridged.

The strongest aspect of the study is its rigor. The authors use appropriate null models and examine multiple datasets (not used in the original analysis) to demonstrate that their findings replicate. Their thorough analysis convincingly supports their assertion that common states are present across a variety of conditions, but that different states may predict behavioural measures for different conditions. However, the authors could have better situated their work within the existing literature. It is not that a more exhaustive literature review is needed-it is that some of their results are unsurprising given the work reported in other manuscripts; some of their work reinforces or is reinforced by prior studies; and some of their work is not compared to similar findings obtained with other analysis approaches. While space is not unlimited, some of these gaps are important enough that they are worth addressing:

We appreciate the reviewer’s thorough read of our manuscript and positive comments on its rigor and implications. We agree that the original version of the manuscript insufficiently situated this work in the existing literature. We have made extensive revisions to better place our findings in the context of prior work. These changes are described in detail below.

1) The authors' own prior work on functional connectivity signatures of attention is not discussed in comparison to the latest work. Neither is work from other groups showing signatures of arousal that change over time, particularly in resting state scans. Attention and arousal are not the same things, but they are intertwined, and both have been linked to large-scale changes in brain activity that should be captured in the HMM latent states. The authors should discuss how the current work fits with existing studies.

Thank you for raising this point. We agree that the relationship between low-dimensional latent states and predefined activity and functional connectivity signatures is an important and interesting question in both attention research and more general contexts. Here, we did not empirically relate the brain states examined in this study and functional connectivity signatures previously investigated in our lab (e.g., Rosenberg et al., 2016; Song et al., 2021a) because the research question and methodological complexities deserved separate attention that go beyond the scope of this paper. Therefore, we conceptually addressed the reviewer’s question on how functional connectivity signatures of attention are related to the brain states that were observed here. Next, we asked how arousal relates to the brain states by indirectly predicting arousal levels of each brain state based on its activity patterns’ spatial resemblance to the predefined arousal network template (Goodale et al., 2021).

Latent states and dynamic functional connectivity

Previous work suggested that, on medium time scales (~20-60 seconds), changes in functional connectivity signatures of sustained attention (Rosenberg et al., 2020) and narrative engagement (Song et al., 2021a) predicted changes in attentional states. How do these attention-related functional connectivity dynamics relate to latent state dynamics, measured on a shorter time scale (1 second)?

Theoretically, there are reasons to think that these measures are related but not redundant. Both HMM and dynamic functional connectivity provide summary measures of the whole-brain functional interactions that evolve over time. Whereas HMM identifies recurring low-dimensional brain states, dynamic functional connectivity used in our and others’ prior studies captures high-dimensional dynamical patterns. Furthermore, while the mixture Gaussian function utilized to infer emission probability in our HMM infers the states from both the BOLD activity patterns and their interactions, functional connectivity considers only pairwise interactions between regions of interests. Thus, with a theoretical ground that the brain states can be characterized at multiple scales and different methods (Greene et al., 2023), we can hypothesize that the both measures could (and perhaps, should be able to) capture brain-wide latent state changes. For example, if we were to apply kmeans clustering methods on the sliding window-based dynamic functional connectivity as in Allen et al. (2014), the resulting clusters could arguably be similar to the latent states derived from the HMM.

However, there are practical reasons why the correspondence between our prior dynamic functional connectivity models and current HMM states is difficult to test directly. A time point-bytime point matching of the HMM state sequence and dynamic functional connectivity is not feasible because, in our prior work, dynamic functional connectivity was measured in a sliding time window (~20-60 seconds), whereas the HMM state identification is conducted at every TR (1 second). An alternative would be to concatenate all time points that were categorized as each HMM state to compute representative functional connectivity of that state. This “splicing and concatenating” method, however, disrupts continuous BOLD-signal time series and has not previously been validated for use with our dynamic connectome-based predictive models. In addition, the difference in time series lengths across states would make comparisons of the four states’ functional connectomes unfair.

One main focus of our manuscript was to relate brain dynamics (HMM state dynamics) to static manifold (functional connectivity gradients). We agree that a direct link between two measures of brain dynamics, HMM and dynamic functional connectivity, is an important research question. However, due to some intricacies that needed to be addressed to answer this question, we felt that it was beyond the scope of our paper. We are eager, however, to explore these comparisons in future work which can more thoroughly address the caveats associated with comparing models of sustained attention, narrative engagement, and arousal defined using different input features and methods.

Arousal, attention, and latent neural state dynamics

Next, the reviewer posed an important question about the relationship between arousal, attention, and latent states. The current study was designed to assess the relationship between attention and latent state dynamics. However, previous neuroimaging work showed that low-dimensional brain dynamics reflect fluctuations in arousal (Raut et al., 2021; Shine et al., 2016; Zhang et al., 2023). Behavioral studies showed that attention and arousal hold a non-linear relationship, for example, mind-wandering states are associated with lower arousal and externally distracted states are associated with higher arousal, when both these states indicate low attention (Esterman and Rothlein, 2019; Unsworth and Robison, 2018, 2016).

To address the reviewer’s suggestion, we wanted to test if our brain states reflected changes in arousal, but we did not collect relevant behavioral or physiological measures. Therefore, to indirectly test for relationships, we predicted levels of arousal in brain states by applying the “arousal network template” defined by Dr. Catie Chang’s group (Chang et al., 2016; Falahpour et al., 2018; Goodale et al., 2021). The arousal network template was created from resting-state fMRI data to predict arousal levels indicated by eye monitoring and electrophysiological signals. In the original study, the arousal level at each time point was predicted by the correlation between the BOLD activity patterns of each TR to the arousal template. The more similar the whole-brain activation pattern was to the arousal network template, the higher the participant was predicted to be aroused at that moment. This activity pattern-based model was generalized to fMRI data during tasks (Goodale et al., 2021).

We correlated the arousal template to the activity patterns of the four brain states that were inferred by the HMM. The DMN state was positively correlated with the arousal template (r=0.264) and the SM state was negatively correlated with the arousal template (r=-0.303) (Author response image 1). These values were not tested for significance because they were single observations. While speculative, this may suggest that participants are in a high arousal state during the DMN state and a low arousal state during the SM state. Together with our results relating brain states to attention, it is possible that the SM state is a common state indicating low arousal and low attention. On the other hand, the DMN state, a signature of a highly aroused state, may benefit gradCPT task performance but not necessarily in engaging with a sitcom episode. However, because this was a single observation and we did not collect a physiological measure of arousal to validate this indirect prediction result, we did not include the result in the manuscript. We hope to more directly test this question in future work with behavioral and physiological measures of arousal.

Author response image 1.

Changes made to the manuscript

Importantly, we agree with the reviewer that a theoretical discussion about the relationships between functional connectivity, latent states, gradients, as well as attention and arousal was a critical omission from the original Discussion. We edited the Discussion to highlight past literature on these topics and encourage future work to investigate these relationships.

[Manuscript, page 11] “Previous studies showed that large-scale neural dynamics that evolve over tens of seconds capture meaningful variance in arousal (Raut et al., 2021; Zhang et al., 2023) and attentional states (Rosenberg et al., 2020; Yamashita et al., 2021). We asked whether latent neural state dynamics reflect ongoing changes in attention in both task and naturalistic contexts.”

[Manuscript, page 17] “Previous work showed that time-resolved whole-brain functional connectivity (i.e., paired interactions of more than a hundred parcels) predicts changes in attention during task performance (Rosenberg et al., 2020) as well as movie-watching and story-listening (Song et al., 2021a). Future work could investigate whether functional connectivity and the HMM capture the same underlying “brain states” to bridge the results from the two literatures. Furthermore, though the current study provided evidence of neural state dynamics reflecting attention, the same neural states may, in part, reflect fluctuations in arousal (Chang et al., 2016; Zhang et al., 2023). Complementing behavioral studies that demonstrated a nonlinear relationship between attention and arousal (Esterman and Rothlein, 2019; Unsworth and Robison, 2018, 2016), future studies collecting behavioral and physiological measures of arousal can assess the extent to which attention explains neural state dynamics beyond what can be explained by arousal fluctuations.”

2) The 'base state' has been described in a number of prior papers (for one early example, see https://pubmed.ncbi.nlm.nih.gov/27008543). The idea that it might serve as a hub or intermediary for other states has been raised in other studies, and discussion of the similarity or differences between those studies and this one would provide better context for the interpretation of the current work. One of the intriguing findings of the current study is that the incidence of this base state increases during sitcom watching, the strongest evidence to date is that it has a cognitive role and is not merely a configuration of activity that the brain must pass through when making a transition.

We greatly appreciate the reviewer’s suggestion of prior papers. We were not aware of previous findings of the base state at the time of writing the manuscript, so it was reassuring to see consistent findings. In the Discussion, we highlighted the findings of Chen et al. (2016) and Saggar et al. (2022). Both studies highlighted the role of the base state as a “hub”-like transition state. However, as the reviewer noted, these studies did not address the functional relevance of this state to cognitive states because both were based on resting-state fMRI.

In our revised Discussion, we write that our work replicates previous findings of the base state that consistently acted as a transitional hub state in macroscopic brain dynamics. We also note that our study expands this line of work by characterizing what functional roles the base state plays in multiple contexts: The base state indicated high attentional engagement and exhibited the highest occurrence proportion as well as longest dwell times during naturalistic movie watching. The base state’s functional involvement was comparatively minor during controlled tasks.

[Manuscript, page 17-18] “Past resting-state fMRI studies have reported the existence of the base state. Chen et al. (2016) used the HMM to detect a state that had “less apparent activation or deactivation patterns in known networks compared with other states”. This state had the highest occurrence probability among the inferred latent states, was consistently detected by the model, and was most likely to transition to and from other states, all of which mirror our findings here. The authors interpret this state as an “intermediate transient state that appears when the brain is switching between other more reproducible brain states”. The observation of the base state was not confined to studies using HMMs. Saggar et al. (2022) used topological data analysis to represent a low-dimensional manifold of resting-state whole-brain dynamics as a graph, where each node corresponds to brain activity patterns of a cluster of time points. Topologically focal “hub” nodes were represented uniformly by all functional networks, meaning that no characteristic activation above or below the mean was detected, similar to what we observe with the base state. The transition probability from other states to the hub state was the highest, demonstrating its role as a putative transition state.

However, the functional relevance of the base state to human cognition had not been explored previously. We propose that the base state, a transitional hub (Figure 2B) positioned at the center of the gradient subspace (Figure 1D), functions as a state of natural equilibrium. Transitioning to the DMN, DAN, or SM states reflects incursion away from natural equilibrium (Deco et al., 2017; Gu et al., 2015), as the brain enters a functionally modular state. Notably, the base state indicated high attentional engagement (Figure 5E and F) and exhibited the highest occurrence proportion (Figure 3B) as well as the longest dwell times (Figure 3—figure supplement 1) during naturalistic movie watching, whereas its functional involvement was comparatively minor during controlled tasks. This significant relevance to behavior verifies that the base state cannot simply be a byproduct of the model. We speculate that susceptibility to both external and internal information is maximized in the base state—allowing for roughly equal weighting of both sides so that they can be integrated to form a coherent representation of the world—at the expense of the stability of a certain functional network (Cocchi et al., 2017; Fagerholm et al., 2015). When processing rich narratives, particularly when a person is fully immersed without having to exert cognitive effort, a less modular state with high degrees of freedom to reach other states may be more likely to be involved. The role of the base state should be further investigated in future studies.”

3) The link between latent states and functional connectivity gradients should be considered in the context of prior work showing that the spatiotemporal patterns of intrinsic activity that account for most of the structure in resting state fMRI also sweep across functional connectivity gradients (https://pubmed.ncbi.nlm.nih.gov/33549755/). In fact, the spatiotemporal dynamics may give rise to the functional connectivity gradients (https://pubmed.ncbi.nlm.nih.gov/35902649/). HMM states bear a marked resemblance to the high-activity phases of these patterns and are likely to be closely linked to them. The spatiotemporal patterns are typically obtained during rest, but they have been reported during task performance (https://pubmed.ncbi.nlm.nih.gov/30753928/) which further suggests a link to the current work. Similar patterns have been observed in anesthetized animals, which also reinforces the conclusion of the current work that the states are fundamental aspects of the brain's functional organization.

We appreciate the comments that relate spatiotemporal patterns, functional connectivity gradients, and the latent states derived from the HMM. Our work was also inspired by the papers that the reviewer suggested, especially Bolt et al.’s (2022), which compared the results of numerous dimensionality and clustering algorithms and suggested three spatiotemporal patterns that seemed to be commonly supported across algorithms. We originally cited these studies throughout the manuscript, but did not discuss them comprehensively. We have revised the Discussion to situate our findings on past work that used resting-state fMRI to study low-dimensional latent brain states.

[Manuscript, page 15-16] “This perspective is supported by previous work that has used different methods to capture recurring low-dimensional states from spontaneous fMRI activity during rest. For example, to extract time-averaged latent states, early resting-state analyses identified task-positive and tasknegative networks using seed-based correlation (Fox et al., 2005). Dimensionality reduction algorithms such as independent component analysis (Smith et al., 2009) extracted latent components that explain the largest variance in fMRI time series. Other lines of work used timeresolved analyses to capture latent state dynamics. For example, variants of clustering algorithms, such as co-activation patterns (Liu et al., 2018; Liu and Duyn, 2013), k-means clustering (Allen et al., 2014), and HMM (Baker et al., 2014; Chen et al., 2016; Vidaurre et al., 2018, 2017), characterized fMRI time series as recurrences of and transitions between a small number of states. Time-lag analysis was used to identify quasiperiodic spatiotemporal patterns of propagating brain activity (Abbas et al., 2019; Yousefi and Keilholz, 2021). A recent study extensively compared these different algorithms and showed that they all report qualitatively similar latent states or components when applied to fMRI data (Bolt et al., 2022). While these studies used different algorithms to probe data-specific brain states, this work and ours report common latent axes that follow a long-standing theory of large-scale human functional systems (Mesulam, 1998). Neural dynamics span principal axes that dissociate unimodal to transmodal and sensory to motor information processing systems.”

Reviewer #2 (Public Review):

In this study, Song and colleagues applied a Hidden Markov Model to whole-brain fMRI data from the unique SONG dataset and a grad-CPT task, and in doing so observed robust transitions between lowdimensional states that they then attributed to specific psychological features extracted from the different tasks.

The methods used appeared to be sound and robust to parameter choices. Whenever choices were made regarding specific parameters, the authors demonstrated that their approach was robust to different values, and also replicated their main findings on a separate dataset.

I was mildly concerned that similarities in some of the algorithms used may have rendered some of the inter-measure results as somewhat inevitable (a hypothesis that could be tested using appropriate null models).

This work is quite integrative, linking together a number of previous studies into a framework that allows for interesting follow-up questions.

Overall, I found the work to be robust, interesting, and integrative, with a wide-ranging citation list and exciting implications for future work.

We appreciate the reviewer’s comments on the study’s robustness and future implications. Our work was highly motivated by the reviewer’s prior work.

Reviewer #3 (Public Review):

My general assessment of the paper is that the analyses done after they find the model are exemplary and show some interesting results. However, the method they use to find the number of states (Calinski-Harabasz score instead of log-likelihood), the model they use generally (HMM), and the fact that they don't show how they find the number of states on HCP, with the Schaeffer atlas, and do not report their R^2 on a test set is a little concerning. I don't think this perse impedes their results, but it is something that they can improve. They argue that the states they find align with long-standing ideas about the functional organization of the brain and align with other research, but they can improve their selection for their model.

We appreciate the reviewer’s thorough read of the paper, evaluation of our analyses linking brain states to behavior as “exemplary”, and important questions about the modeling approach. We have included detailed responses below and updated the manuscript accordingly.

Strengths:

• Use multiple datasets, multiple ROIs, and multiple analyses to validate their results

• Figures are convincing in the sense that patterns clearly synchronize between participants

• Authors select the number of states using the optimal model fit (although this turns out to be a little more questionable due to what they quantify as 'optimal model fit')

We address this concern on page 30-31 of this response letter.

• Replication with Schaeffer atlas makes results more convincing

• The analyses around the fact that the base state acts as a flexible hub are well done and well explained

• Their comparison of synchrony is well-done and comparing it to resting-state, which does not have any significant synchrony among participants is obvious, but still good to compare against.

• Their results with respect to similar narrative engagement being correlated with similar neural state dynamics are well done and interesting.

• Their results on event boundaries are compelling and well done. However, I do not find their Chang et al. results convincing (Figure 4B), it could just be because it is a different medium that explains differences in DMN response, but to me, it seems like these are just altogether different patterns that can not 100% be explained by their method/results.

We entirely agree with the reviewer that the Chang et al. (2021) data are different in many ways from our own SONG dataset. Whereas data from Chang et al. (2021) were collected while participants listened to an audio-only narrative, participants in the SONG sample watched and listened to audiovisual stimuli. They were scanned at different universities in different countries with different protocols by different research groups for different purposes. That is, there are numerous reasons why we would expect the model should not generalize. Thus, we found it compelling and surprising that, despite all of these differences between the datasets, the model trained on the SONG dataset generalized to the data from Chang et al. (2021). The results highlighted a robust increase in the DMN state occurrence and a decrease in the base state occurrence after the narrative event boundaries, irrespective of whether the stimulus was an audiovisual sitcom episode or a narrated story. This external model validation was a way that we tested the robustness of our own model and the relationship between neural state dynamics and cognitive dynamics.

• Their results that when there is no event, transition into the DMN state comes from the base state is 50% is interesting and a strong result. However, it is unclear if this is just for the sitcom or also for Chang et al.'s data.

We apologize for the lack of clarity. We show the statistical results of the two sitcom episodes as well as Chang et al.’s (2021) data in Figure 4—figure supplement 2 in our original manuscript. Here, we provide the exact values of the base-to-DMN state transition probability, and how they differ across moments after event boundaries compared to non-event boundaries.

For sitcom episode 1, the probability of base-to-DMN state transition was 44.6 ± 18.8 % at event boundaries whereas 62.0 ± 10.4 % at non-event boundaries (FDR-p = 0.0013). For sitcom episode 2, the probability of base-to-DMN state transition was 44.1 ± 18.0 % at event boundaries whereas 62.2 ± 7.6 % at non-event boundaries (FDR-p = 0.0006). For the Chang et al. (2021) dataset, the probability of base-to-DMN state transition was 33.3 ± 15.9 % at event boundaries whereas 58.1 ± 6.4 % at non-event boundaries (FDR-p < 0.0001). Thus, our result, “At non-event boundaries, the DMN state was most likely to transition from the base state, accounting for more than 50% of the transitions to the DMN state” (pg 11, line 24-25), holds true for both the internal and external datasets.

• The involvement of the base state as being highly engaged during the comedy sitcom and the movie are interesting results that warrant further study into the base state theory they pose in this work.

• It is good that they make sure SM states are not just because of head motion (P 12).

• Their comparison between functional gradient and neural states is good, and their results are generally well-supported, intuitive, and interesting enough to warrant further research into them. Their findings on the context-specificity of their DMN and DAN state are interesting and relate well to the antagonistic relationship in resting-state data.

Weaknesses:

• Authors should train the model on part of the data and validate on another

Thank you for raising this issue. To the best of our knowledge, past work that applied the HMM to the fMRI data has conducted training and inference on the same data, including initial work that implemented HMM on the resting-state fMRI (Baker et al., 2014; Chen et al., 2016; Vidaurre et al., 2018, 2017) as well as more recent work that applied HMMs to the task or movie-watching fMRI (Cornblath et al., 2020; Taghia et al., 2018; van der Meer et al., 2020; Yamashita et al., 2021). That is, the parameters—emission probability, transition probability, and initial probability—were estimated from the entire dataset and the latent state sequence was inferred using the Viterbi algorithm on the same dataset.

However, we were also aware of the potential problem this may have. Therefore, in our recent work asking a different research question in another fMRI dataset (Song et al., 2021b), we trained an HMM on a subset of the dataset (moments when participants were watching movie clips in the original temporal order) and inferred latent state sequence of the fMRI time series in another subset of the dataset (moments when participants were watching movie clips in a scrambled temporal order). To the best of our knowledge, this was the first paper that used different segments of the data to fit and infer states from the HMM.

In the current study, we wanted to capture brain states that underlie brain activity across contexts. Thus, we presented the same-dataset training and inference procedure as our primary result. However, for every main result, we also showed results where we separated the data used for model fitting and state inference. That is, we fit the HMM on the SONG dataset, primarily report the inference results on the SONG dataset, but also report inference on the external datasets that were not included in model fitting. The datasets used were the Human Connectome Project dataset (Van Essen et al., 2013), Chang et al. (2021) audio-listening dataset, Rosenberg et al. (2016) gradCPT dataset, and Chen et al. (2017) Sherlock dataset.

However, to further address the concern of the reviewer whether the HMM fit is reliable when applied to held-out data, we computed the reliability of the HMM inference by conducting crossvalidations and split-half reliability analysis.

(1) Cross-validation

To separate the dataset used for HMM training and inference, we conducted cross-validation on the SONG dataset (N=27) by training the model with the data from 26 participants and inferring the latent state sequence of the held-out participant.

First, we compared the robustness of the model training by comparing the mean activity patterns of the four latent states fitted at the group level (N=27) with the mean activity patterns of the four states fitted across cross-validation folds. Pearson’s correlations between the group-level vs. cross-validated latent states’ mean activity patterns were r = 0.991 ± 0.010, with a range from 0.963 to 0.999.

Second, we compared the robustness of model inference by comparing the latent state sequences that were inferred at the group level vs. from held-out participants in a cross-validation scheme. All fMRI conditions had mean similarity higher than 90%; Rest 1: 92.74 ± 5.02 %, Rest2: 92.74 ± 4.83 %, GradCPT face: 92.97 ± 6.41 %, GradCPT scene: 93.27 ± 5.76 %, Sitcom ep1: 93.31 ± 3.92 %, Sitcom ep2: 93.13 ± 4.36 %, Documentary: 92.42 ± 4.72 %.

Third, with the latent state sequences inferred from cross-validation, we replicated the analysis of Figure 3 to test for synchrony of the latent state sequences across participants. The crossvalidated results were highly similar to manuscript Figure 3, which was generated from the grouplevel analysis. Mean synchrony of latent state sequences are as follows: Rest 1: 25.90 ± 3.81%, Rest 2: 25.75 ± 4.19 %, GradCPT face: 27.17 ± 3.86 %, GradCPT scene: 28.11 ± 3.89 %, Sitcom ep1: 40.69 ± 3.86%, Sitcom ep2: 40.53 ± 3.13%, Documentary: 30.13 ± 3.41%.

Author response image 2.

(2) Split-half reliability

To test for the internal robustness of the model, we randomly assigned SONG dataset participants into two groups and conducted HMM separately in each. Similarity (Pearson’s correlation) between the two groups’ activation patterns were DMN: 0.791, DAN: 0.838, SM: 0.944, base: 0.837. The similarity of the covariance patterns were DMN: 0.995, DAN: 0.996, SM: 0.994, base: 0.996.

Author response image 3.

We further validated the split-half reliability of the model using the HCP dataset, which contains data of a larger sample (N=119). Similarity (Pearson’s correlation) between the two groups’ activation patterns were DMN: 0.998, DAN: 0.997, SM: 0.993, base: 0.923. The similarity of the covariance patterns were DMN: 0.995, DAN: 0.996, SM: 0.994, base: 0.996.

Together the cross-validation and split-half reliability results demonstrate that the HMM results reported in the manuscript are reliable and robust to the way we conducted the analysis. The result of the split-half reliability analysis is added in the Results.

[Manuscript, page 3-4] “Neural state inference was robust to the choice of 𝐾 (Figure 1—figure supplement 1) and the fMRI preprocessing pipeline (Figure 1—figure supplement 5) and consistent when conducted on two groups of randomly split-half participants (Pearson’s correlations between the two groups’ latent state activation patterns: DMN: 0.791, DAN: 0.838, SM: 0.944, base: 0.837).”

• Comparison with just PCA/functional gradients is weak in establishing whether HMMs are good models of the timeseries. Especially given that the HMM does not explain a lot of variance in the signal (~0.5 R^2 for only 27 brain regions) for PCA. I think they don't report their own R^2 of the timeseries

We agree with the reviewer that the PCA that we conducted to compare with the explained variance of the functional gradients was not directly comparable because PCA and gradients utilize different algorithms to reduce dimensionality. To make more meaningful comparisons, we removed the data-specific PCA results and replaced them with data-specific functional gradients (derived from the SONG dataset). This allows us to directly compare SONG-specific functional gradients with predefined gradients (derived from the resting-state HCP dataset from Margulies et al. [2016]). We found that the degrees to which the first two predefined gradients explained whole-brain fMRI time series (SONG: 𝑟! = 0.097, HCP: 0.084) were comparable to the amount of variance explained by the first two data-specific gradients (SONG: 𝑟! = 0.100, HCP: 0.086). Thus, the predefined gradients explain as much variance in the SONG data time series as SONG-specific gradients do. This supports our argument that the low-dimensional manifold is largely shared across contexts, and that the common HMM latent states may tile the predefined gradients.

These analyses and results were added to the Results, Methods, and Figure 1—figure supplement 8. Here, we only attach changes to the Results section for simplicity, but please see the revised manuscript for further changes.

[Manuscript, page 5-6] “We hypothesized that the spatial gradients reported by Margulies et al. (2016) act as a lowdimensional manifold over which large-scale dynamics operate (Bolt et al., 2022; Brown et al., 2021; Karapanagiotidis et al., 2020; Turnbull et al., 2020), such that traversals within this manifold explain large variance in neural dynamics and, consequently, cognition and behavior (Figure 1C). To test this idea, we situated the mean activity values of the four latent states along the gradients defined by Margulies et al. (2016) (see Methods). The brain states tiled the two-dimensional gradient space with the base state at the center (Figure 1D; Figure1—figure supplement 7). The Euclidean distances between these four states were maximized in the two-dimensional gradient space, compared to a chance where the four states were inferred from circular-shifted time series (p < 0.001). For the SONG dataset, the DMN and SM states fell at more extreme positions of the primary gradient than expected by chance (both FDR-p values = 0.004; DAN and SM states, FDRp values = 0.171). For the HCP dataset, the DMN and DAN states fell at more extreme positions on the primary gradient (both FDR-p values = 0.004; SM and base states, FDR-p values = 0.076). No state was consistently found at the extremes of the secondary gradient (all FDR-p values > 0.021).

We asked whether the predefined gradients explain as much variance in neural dynamics as latent subspace optimized for the SONG dataset. To do so, we applied the same nonlinear dimensionality reduction algorithm to the SONG dataset’s ROI time series. Of note, the SONG dataset includes 18.95% rest, 15.07% task, and 65.98% movie-watching data whereas the data used by Margulies et al. (2016) was 100% rest. Despite these differences, the SONG-specific gradients closely resembled the predefined gradients, with significant Pearson’s correlations observed for the first (r = 0.876) and second (r = 0.877) gradient embeddings (Figure 1—figure supplement 8). Gradients identified with the HCP data also recapitulated Margulies et al.’s (2016) first (r = 0.880) and second (r = 0.871) gradients. We restricted our analysis to the first two gradients because the two gradients together explained roughly 50% of the entire variance of functional brain connectome (SONG: 46.94%, HCP: 52.08%), and the explained variance dropped drastically from the third gradients (more than 1/3 drop compared to second gradients). The degrees to which the first two predefined gradients explained whole-brain fMRI time series (SONG: 𝑟! = 0.097, HCP: 0.084) were comparable to the amount of variance explained by the first two data-specific gradients (SONG: 𝑟! = 0.100, HCP: 0.086; Figure 1—figure supplement 8). Thus, the low-dimensional manifold captured by Margulies et al. (2016) gradients is highly replicable, explaining brain activity dynamics as well as data-specific gradients, and is largely shared across contexts and datasets. This suggests that the state space of whole-brain dynamics closely recapitulates low-dimensional gradients of the static functional brain connectome.”

The reviewer also pointed out that the PCA-gradient comparison was weak in establishing whether HMMs are good models of the time series. However, we would like to point out that the purpose of the comparison was not to validate the performance of the HMM. Instead, we wanted to test whether the gradients introduced by Margulies et al. (2016) could act as a generalizable lowdimensional manifold of brain state dynamics. To argue that the predefined gradients are a shared manifold, these gradients should explain SONG data fMRI time series as much as the principal components derived directly from the SONG data. Our results showed comparable 𝑟!, both in predefined gradient vs. data-specific PC comparisons and predefined gradient vs. data-specific gradient comparisons, which supported our argument that the predefined gradients could be the shared embedding space across contexts and datasets.

The reviewer pointed out that the 𝑟2 of ~0.5 is not explaining enough variance in the fMRI signal. However, we respectfully disagree with this point because there is no established criterion for what constitutes a high or low 𝑟2 for this type of analysis. Of note, previous literature that also applied PCA to fMRI time series (Author response image 4A and 4B) (Lynn et al., 2021; Shine et al., 2019) also found that the cumulative explained variance of top 5 principal components is around 50%. Author response image 4C shows cumulative variances to which gradients explain the functional connectome of the resting-state fMRI data (Margulies et al., 2016).

Author response image 4.

Finally, the reviewer pointed out that the 𝑟! of the HMM-derived latent sequence to the fMRI time series should be reported. However, there is no standardized way of measuring the explained variance of the HMM inference. There is no report of explained variance in the traditional HMMfMRI papers (Baker et al., 2014; Chen et al., 2016; Vidaurre et al., 2018, 2017). Rather than 𝑟!, the HMM computes the log likelihood of the model fit. However, because log likelihood values are dependent on the number of data points, studies do not report log likelihood values nor do they use these metrics to interpret the goodness of model fit.

To ask whether the goodness of the HMM fit was significant above chance, we compared the log likelihood of the HMM to the log likelihood distribution of the null HMM fits. First, we extracted the log likelihood of the HMM fit with the real fMRI time series. We iterated this 1,000 times when calculating null HMMs using the circular-shifted fMRI time series. The log likelihood of the real model was significantly higher than the chance distribution, with a z-value of 2182.5 (p < 0.001). This indicates that the HMM explained a large variance in our fMRI time series data, significantly above chance.

• Authors do not specify whether they also did cross-validation for the HCP dataset to find 4 clusters

We apologize for the lack of clarity. When we computed the Calinski-Harabasz score with the HCP dataset, three was chosen as the most optimal number of states (Author response image 5A). When we set K as 3, the HMM inferred the DMN, DAN, and SM states (Author response image 5C). The base state was included when K was set to 4 (Author response image 5B). The activation pattern similarities of the DMN, DAN, and SM states were r = 0.981, 0.984, 0.911 respectively.

Author response image 5.

We did not use K = 3 for the HCP data replication because we were not trying to test whether these four states would be the optimal set of states in every dataset. Although the CalinskiHarabasz score chose K = 3 because it showed the best clustering performance, this does not mean that the base state is not meaningful to this dataset. Likewise, the latent states that are inferred when we increase/decrease the number of states are also meaningful states. For example, in Figure 1—figure supplement 1, we show an example of the SONG dataset’s latent states when we set K to 7. The seven latent states included the DAN, SM, and base states, the DMN state was subdivided into DMN-A and DMN-B states, and the FPN state and DMN+VIS state were included. Setting a higher number of states like K = 7 would mean that we are capturing brain state dynamics in a higher dimension than when using K = 4. Because we are utilizing a higher number of states, a model set to K = 7 would inevitably capture a larger variance of fMRI time series than a model set to K = 4.

The purpose of latent state replication with the HCP dataset was to validate the generalizability of the DMN, DAN, SM, and base states. Before characterizing these latent states’ relevance to cognition, we needed to verify that these latent states were not simply overfit to the SONG dataset. The fact that the HMM revealed a similar set of latent states when applied to the HCP dataset suggested that the states were not merely specific to SONG data.

To make our points clearer in the manuscript, we emphasized that we are not arguing for the four states to be the exclusive states. We made edits to Discussion as follows.

[Manuscript, page 16] “Our study adopted the assumption of low dimensionality of large-scale neural systems, which led us to intentionally identify only a small number of states underlying whole-brain dynamics. Importantly, however, we do not claim that the four states will be the optimal set of states in every dataset and participant population. Instead, latent states and patterns of state occurrence may vary as a function of individuals and tasks (Figure 1—figure supplement 2). Likewise, while the lowest dimensions of the manifold (i.e., the first two gradients) were largely shared across datasets tested here, we do not argue that it will always be identical. If individuals and tasks deviate significantly from what was tested here, the manifold may also differ along with changes in latent states (Samara et al., 2023). Brain systems operate at different dimensionalities and spatiotemporal scales (Greene et al., 2023), which may have different consequences for cognition. Asking how brain states and manifolds—probed at different dimensionalities and scales—flexibly reconfigure (or not) with changes in contexts and mental states is an important research question for understanding complex human cognition.”

• One of their main contributions is the base state but the correlation between the base state in their Song dataset and the HCP dataset is only 0.399

This is a good point. However, there is precedent for lower spatial pattern correlation of the base state compared to other states in the literature.

Compared to the DMN, DAN, and SM states, the base state did not show characteristic activation or deactivation of functional networks. Most of the functional networks showed activity levels close to the mean (z = 0). With this flattened activation pattern, relatively low activation pattern similarity was observed between the SONG base state and the HCP base state.

In Figure 1—figure supplement 6, we write, “The DMN, DAN, and SM states showed similar mean activity patterns. We refrained from making interpretations about the base state’s activity patterns because the mean activity of most of the parcels was close to z = 0”.

A similar finding has been reported in a previous work by Chen et al. (2016) that discovered the base state with HMM. State 9 (S9) of their results is comparable to our base state. They report that even though the spatial correlation coefficient of the brain state from the split-half reliability analysis was the lowest for S9 due to its low degrees of activation or deactivation, S9 was stably inferred by the HMM. The following is a direct quote from their paper:

“To the best of our knowledge, a state similar to S9 has not been presented in previous literature. We hypothesize that S9 is the “ground” state of the brain, in which brain activity (or deactivity) is similar for the entire cortex (no apparent activation or deactivation as shown in Fig. 4). Note that different groups of subjects have different spatial patterns for state S9 (Fig. 3A). Therefore, S9 has the lowest reproducible spatial pattern (Fig. 3B). However, its temporal characteristics allowed us to distinguish it consistently from other states.” (Chen et al., 2016)

Thus, we believe our data and prior results support the existence of the “base state”.

• Figure 1B: Parcellation is quite big but there seems to be a gradient within regions

This is a function of the visualization software. Mean activity (z) is the same for all voxels within a parcel. To visualize the 3D contours of the brain, we chose an option in the nilearn python function that smooths the mean activity values based on the surface reconstructed anatomy.

In the original manuscript, our Methods write, “The brain surfaces were visualized with nilearn.plotting.plot_surf_stat_map. The parcel boundaries in Figure 1B are smoothed from the volume-to-surface reconstruction.”

• Figure 1D: Why are the DMNs further apart between SONG and HCP than the other states

To address this question, we first tested whether the position of the DMN states in the gradient space is significantly different for the SONG and HCP datasets. We generated surrogate HMM states from the circular-shifted fMRI time series and positioned the four latent states and the null DMN states in the 2-dimensional gradient space (Author response image 6).

Author response image 6.

We next tested whether the Euclidean distance between the SONG dataset’s DMN state and the HCP dataset’s DMN state is larger than would be expected by chance (Author response image 7). To do so, we took the difference between the DMN state positions and compared it to the 1,000 differences generated from the surrogate latent states. The DMN states of the SONG and HCP datasets did not significantly differ in the Gradient 1 dimension (two-tailed test, p = 0.794). However, as the reviewer noted, the positions differed significantly in the Gradient 2 dimension (p = 0.047). The DMN state leaned more towards the Visual gradient in the SONG dataset, whereas it leaned more towards the Somatosensory-Motor gradient in the HCP dataset.

Author response image 7.

Though we cannot claim an exact reason for this across-dataset difference, we note a distinctive difference between the SONG and HCP datasets. Both datasets largely included resting-state, controlled tasks, and movie watching. The SONG dataset included 18.95% of rest, 15.07% of task, and 65.98% of movie watching. The task only contained the gradCPT, i.e., sustained attention task. On the other hand, the HCP dataset included 52.71% of rest, 24.35% of task, and 22.94% of movie watching. There were 7 different tasks included in the HCP dataset. It is possible that different proportions of rest, task, and movie watching, and different cognitive demands involved with each dataset may have created data-specific latent states.

• Page 5 paragraph starting at L25: Their hypothesis that functional gradients explain large variance in neural dynamics needs to be explained more, is non-trivial especially because their R^2 scores are so low (Fig 1. Supplement 8) for PCA

We address this concern on page 21-23 of this response letter.

• Generally, I do not find the PCA analysis convincing and believe they should also compare to something like ICA or a different model of dynamics. They do not explain their reasoning behind assuming an HMM, which is an extremely simplified idea of brain dynamics meaning they only change based on the previous state.

We appreciate this perspective. We replaced the Margulies et al.’s (2016) gradient vs. SONGspecific PCA comparison with a more direct Margulies et al.’s (2016) gradient vs. SONG-specific gradient comparison as described on page 21-23 of this response letter.

More broadly, we elected to use HMM because of recent work showing correspondence between low-dimensional HMM states and behavior (Cornblath et al., 2020; Taghia et al., 2018; van der Meer et al., 2020; Yamashita et al., 2021). We also found the model’s assumption—a mixture Gaussian emission probability and first-order Markovian transition probability—to be the most suited to analyzing the fMRI time series data. We do not intend to claim that other data-reduction techniques would not also capture low-dimensional, behaviorally relevant changes in brain activity. Instead, our primary focus was identifying a set of latent states that generalize (i.e., recur) across multiple contexts and understanding how those states reflect cognitive and attentional states.

Although a comparison of possible data-reduction algorithms is out of the scope of the current work, an exhaustive comparison of different models can be found in Bolt et al. (2022). The authors compared dozens of latent brain state algorithms spanning zero-lag analysis (e.g., principal component analysis, principal component analysis with Varimax rotation, Laplacian eigenmaps, spatial independent component analysis, temporal independent component analysis, hidden Markov model, seed-based correlation analysis, and co-activation patterns) to time-lag analysis (e.g., quasi-periodic pattern and lag projections). Bolt et al. (2022) writes “a range of empirical phenomena, including functional connectivity gradients, the task-positive/task-negative anticorrelation pattern, the global signal, time-lag propagation patterns, the quasiperiodic pattern and the functional connectome network structure, are manifestations of the three spatiotemporal patterns.” That is, many previous findings that used different methods essentially describe the same recurring latent states. A similar argument was made in previous papers (Brown et al., 2021; Karapanagiotidis et al., 2020; Turnbull et al., 2020).

We agree that the HMM is a simplified idea of brain dynamics. We do not argue that the four number of states can fully explain the complexity and flexibility of cognition. Instead, we hoped to show that there are different dimensionalities to which the brain systems can operate, and they may have different consequences to cognition. We “simplified” neural dynamics to a discrete sequence of a small number of states. However, what is fascinating is that these overly “simplified” brain state dynamics can explain certain cognitive and attentional dynamics, such as event segmentation and sustained attention fluctuations. We highlight this point in the Discussion.

[Manuscript, page 16] “Our study adopted the assumption of low dimensionality of large-scale neural systems, which led us to intentionally identify only a small number of states underlying whole-brain dynamics. Importantly, however, we do not claim that the four states will be the optimal set of states in every dataset and participant population. Instead, latent states and patterns of state occurrence may vary as a function of individuals and tasks (Figure 1—figure supplement 2). Likewise, while the lowest dimensions of the manifold (i.e., the first two gradients) were largely shared across datasets tested here, we do not argue that it will always be identical. If individuals and tasks deviate significantly from what was tested here, the manifold may also differ along with changes in latent states (Samara et al., 2023). Brain systems operate at different dimensionalities and spatiotemporal scales (Greene et al., 2023), which may have different consequences for cognition. Asking how brain states and manifolds—probed at different dimensionalities and scales—flexibly reconfigure (or not) with changes in contexts and mental states is an important research question for understanding complex human cognition.”

• For the 25- ROI replication it seems like they again do not try multiple K values for the number of states to validate that 4 states are in fact the correct number.

In the manuscript, we do not argue that the four will be the optimal number of states in any dataset. (We actually predict that this may differ depending on the amount of data, participant population, tasks, etc.) Instead, we claim that the four identified in the SONG dataset are not specific (i.e., overfit) to that sample, but rather recur in independent datasets as well. More broadly we argue that the complexity and flexibility of human cognition stem from the fact that computation occurs at multiple dimensions and that the low-dimensional states observed here are robustly related to cognitive and attentional states. To prevent misunderstanding of our results, we emphasized in the Discussion that we are not arguing for a fixed number of states. A paragraph included in our response to the previous comment (page 16 in the manuscript) illustrates this point.

• Fig 2B: Colorbar goes from -0.05 to 0.05 but values are up to 0.87

We apologize for the confusion. The current version of the figure is correct. The figure legend states, “The values indicate transition probabilities, such that values in each row sums to 1. The colors indicate differences from the mean of the null distribution where the HMMs were conducted on the circular-shifted time series.”

We recognize that this complicates the interpretation of the figure. However, after much consideration, we decided that it was valuable to show both the actual transition probabilities (values) and their difference from the mean of null HMMs (colors). The values demonstrate the Markovian property of latent state dynamics, with a high probability of remaining in the same state at consecutive moments and a low probability of transitioning to a different state. The colors indicate that the base state is a transitional hub state by illustrating that the DMN, DAN, and SM states are more likely to transition to the base state than would be expected by chance.

• P 16 L4 near-critical, authors need to be more specific in their terminology here especially since they talk about dynamic systems, where near-criticality has a specific definition. It is unclear which definition they are looking for here.

We agree that our explanation was vague. Because we do not have evidence for this speculative proposal, we removed the mention of near-criticality. Instead, we focus on our observation as the base state being the transitional hub state within a metastable system.

[Manuscript, page 17-18] “However, the functional relevance of the base state to human cognition had not been explored previously. We propose that the base state, a transitional hub (Figure 2B) positioned at the center of the gradient subspace (Figure 1D), functions as a state of natural equilibrium. Transitioning to the DMN, DAN, or SM states reflects incursion away from natural equilibrium (Deco et al., 2017; Gu et al., 2015), as the brain enters a functionally modular state. Notably, the base state indicated high attentional engagement (Figure 5E and F) and exhibited the highest occurrence proportion (Figure 3B) as well as the longest dwell times (Figure 3—figure supplement 1) during naturalistic movie watching, whereas its functional involvement was comparatively minor during controlled tasks. This significant relevance to behavior verifies that the base state cannot simply be a byproduct of the model. We speculate that susceptibility to both external and internal information is maximized in the base state—allowing for roughly equal weighting of both sides so that they can be integrated to form a coherent representation of the world—at the expense of the stability of a certain functional network (Cocchi et al., 2017; Fagerholm et al., 2015). When processing rich narratives, particularly when a person is fully immersed without having to exert cognitive effort, a less modular state with high degrees of freedom to reach other states may be more likely to be involved. The role of the base state should be further investigated in future studies.”

• P16 L13-L17 unnecessary

We prefer to have the last paragraph as a summary of the implications of this paper. However, if the length of this paper becomes a problem as we work towards publication with the editors, we are happy to remove these lines.

• I think this paper is solid, but my main issue is with using an HMM, never explaining why, not showing inference results on test data, not reporting an R^2 score for it, and not comparing it to other models. Secondly, they use the Calinski-Harabasz score to determine the number of states, but not the log-likelihood of the fit. This clearly creates a bias in what types of states you will find, namely states that are far away from each other, which likely also leads to the functional gradient and PCA results they have. Where they specifically talk about how their states are far away from each other in the functional gradient space and correlated to (orthogonal) components. It is completely unclear to me why they used this measure because it also seems to be one of many scores you could use with respect to clustering (with potentially different results), and even odd in the presence of a loglikelihood fit to the data and with the model they use (which does not perform clustering).

(1) Showing inference results on test data

We address this concern on page 19-21 of this response letter.

(2) Not reporting 𝑹𝟐 score

We address this concern on page 21-23 of this response letter.

(3) Not comparing the HMM model to other models

We address this concern on page 27-28 of this response letter.

(4) The use of the Calinski-Harabasz score to determine the number of states rather than the log-likelihood of the model fit

To our knowledge, the log-likelihood of the model fit is not used in the HMM literature. It is because the log-likelihood tends to increase monotonically as the number of states increases. Baker et al. (2014) illustrates this problem, writing:

“In theory, it should be possible to pick the optimal number of states by selecting the model with the greatest (negative) free energy. In practice however, we observe that the free energy increases monotonically up to K = 15 states, suggesting that the Bayes-optimal model may require an even higher number of states.”

Similarly, the following figure is the log-likelihood estimated from the SONG dataset. Similar to the findings of Baker et al. (2014), the log-likelihood monotonically increased as the number of states increased (Author response image 8, right). The measures like AIC or BIC, which account for the number of parameters, also have the same issue of monotonic increase.

Author response image 8.

Because there is “no straightforward data-driven approach to model order selection” (Baker et al., 2014), past work has used different approaches to decide on the number of states. For example, Vidaurre et al. (2018) iterated over a range of the number of states to repeat the same HMM training and inference procedures 5 times using the same hyperparameters. They selected the number of states that showed the highest consistency across iterations. Gao et al. (2021) tested the clustering performance of the model output using the Calinski-Harabasz score. The number of states that showed the highest within-cluster cohesion compared to the across-cluster separation was selected as the number of states. Chang et al. (2021) applied HMM to voxels of the ventromedial prefrontal cortex using a similar clustering algorithm, writing: “To determine the number of states for the HMM estimation procedure, we identified the number of states that maximized the average within-state spatial similarity relative to the average between-state similarity”. In our previous paper (Song et al., 2021b), we reported both the reliability and clustering performance measures to decide on the number of states.

In the current manuscript, the model consistency criterion from Vidaurre et al. (2018) was ineffective because the HMM inference was extremely robust (i.e., always inferring the exact same sequence) due to a large number of data points. Thus, we used the Calinski-Harabasz score as our criterion for the number of states selected.

We agree with the reviewer that the selection of the number of states is critical to any study that implements HMM. However, the field lacks a consensus on how to decide on the number of states in the HMM, and the Calinski-Harabasz score has been validated in previous studies. Most importantly, the latent states’ relationships with behavioral and cognitive measures give strong evidence that the latent states are indeed meaningful states. Again, we are not arguing that the optimal set of states in any dataset will be four nor are we arguing that these four states will always be the optimal states. Instead, the manuscript proposes that a small number of latent states explains meaningful variance in cognitive dynamics.

• Grammatical error: P24 L29 rendering seems to have gone wrong

Our intention was correct here. To avoid confusion, we changed “(number of participantsC2 iterations)” to “(#𝐶!iterations, where N=number of participants)” (page 26 in the manuscript).

Questions:

• Comment on subject differences, it seems like they potentially found group dynamics based on stimuli, but interesting to see individual differences in large-scale dynamics, and do they believe the states they find mostly explain global linear dynamics?

We agree with the reviewer that whether low-dimensional latent state dynamics explain individual differences—above and beyond what could be explained by the high-dimensional, temporally static neural signatures of individuals (e.g., Finn et al., 2015)—is an important research question. However, because the SONG dataset was collected in a single lab, with a focus on covering diverse contexts (rest, task, and movie watching) over 2 sessions, we were only able to collect 27 participants. Due to this small sample size, we focused on investigating group-level, shared temporal dynamics and across-condition differences, rather than on investigating individual differences.

Past work has studied individual differences (e.g., behavioral traits like well-being, intelligence, and personality) using the HMM (Vidaurre et al., 2017). In the lab, we are working on a project that investigates latent state dynamics in relation to individual differences in clinical symptoms using the Healthy Brain Network dataset (Ji et al., 2022, presented at SfN; Alexander et al., 2017).

Finally, the reviewer raises an interesting question about whether the latent state sequence that was derived here mostly explains global linear dynamics as opposed to nonlinear dynamics. We have two responses: one methodological and one theoretical. First, methodologically, we defined the emission probabilities as a linear mixture of Gaussian distributions for each input dimension with the state-specific mean (mean fMRI activity patterns of the networks) and variance (functional covariance across networks). Therefore, states are modeled with an assumption of linearity of feature combinations. Theoretically, recent work supports in favor of nonlinearity of large-scale neural dynamics, especially as tasks get richer and more complex (Cunningham and Yu, 2014; Gao et al., 2021). However, whether low-dimensional latent states should be modeled nonlinearly—that is, whether linear algorithms are insufficient at capturing latent states compared to nonlinear algorithms—is still unknown. We agree with the reviewer that the assumption of linearity is an interesting topic in systems neuroscience. However, together with prior work which showed how numerous algorithms—either linear or nonlinear—recapitulated a common set of latent states, we argue that the HMM provides a strong low-dimensional model of large-scale neural activity and interaction.

• P19 L40 why did the authors interpolate incorrect or no-responses for the gradCPT runs? It seems more logical to correct their results for these responses or to throw them out since interpolation can induce huge biases in these cases because the data is likely not missing at completely random.

Interpolating the RTs of the trials without responses (omission errors and incorrect trials) is a standardized protocol for analyzing gradCPT data (Esterman et al., 2013; Fortenbaugh et al., 2018, 2015; Jayakumar et al., 2023; Rosenberg et al., 2013; Terashima et al., 2021; Yamashita et al., 2021). The choice of this analysis is due to an assumption that sustained attention is a continuous attentional state; the RT, a proxy for the attentional state in the gradCPT literature, is a noisy measure of a smoothed, continuous attentional state. Thus, the RTs of the trials without responses are interpolated and the RT time courses are smoothed by convolving with a gaussian kernel.

References

Abbas A, Belloy M, Kashyap A, Billings J, Nezafati M, Schumacher EH, Keilholz S. 2019. Quasiperiodic patterns contribute to functional connectivity in the brain. Neuroimage 191:193–204.

Alexander LM, Escalera J, Ai L, Andreotti C, Febre K, Mangone A, Vega-Potler N, Langer N, Alexander A, Kovacs M, Litke S, O’Hagan B, Andersen J, Bronstein B, Bui A, Bushey M, Butler H, Castagna V, Camacho N, Chan E, Citera D, Clucas J, Cohen S, Dufek S, Eaves M, Fradera B, Gardner J, Grant-Villegas N, Green G, Gregory C, Hart E, Harris S, Horton M, Kahn D, Kabotyanski K, Karmel B, Kelly SP, Kleinman K, Koo B, Kramer E, Lennon E, Lord C, Mantello G, Margolis A, Merikangas KR, Milham J, Minniti G, Neuhaus R, Levine A, Osman Y, Parra LC, Pugh KR, Racanello A, Restrepo A, Saltzman T, Septimus B, Tobe R, Waltz R, Williams A, Yeo A, Castellanos FX, Klein A, Paus T, Leventhal BL, Craddock RC, Koplewicz HS, Milham MP. 2017. Data Descriptor: An open resource for transdiagnostic research in pediatric mental health and learning disorders. Sci Data 4:1–26.

Allen EA, Damaraju E, Plis SM, Erhardt EB, Eichele T, Calhoun VD. 2014. Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 24:663–676.

Baker AP, Brookes MJ, Rezek IA, Smith SM, Behrens T, Probert Smith PJ, Woolrich M. 2014. Fast transient networks in spontaneous human brain activity. Elife 3:e01867.

Bolt T, Nomi JS, Bzdok D, Salas JA, Chang C, Yeo BTT, Uddin LQ, Keilholz SD. 2022. A Parsimonious Description of Global Functional Brain Organization in Three Spatiotemporal Patterns. Nat Neurosci 25:1093–1103.

Brown JA, Lee AJ, Pasquini L, Seeley WW. 2021. A dynamic gradient architecture generates brain activity states. Neuroimage 261:119526.

Chang C, Leopold DA, Schölvinck ML, Mandelkow H, Picchioni D, Liu X, Ye FQ, Turchi JN, Duyn JH. 2016. Tracking brain arousal fluctuations with fMRI. Proc Natl Acad Sci U S A 113:4518–4523.

Chang CHC, Lazaridi C, Yeshurun Y, Norman KA, Hasson U. 2021. Relating the past with the present: Information integration and segregation during ongoing narrative processing. J Cogn Neurosci 33:1–23.

Chang LJ, Jolly E, Cheong JH, Rapuano K, Greenstein N, Chen P-HA, Manning JR. 2021. Endogenous variation in ventromedial prefrontal cortex state dynamics during naturalistic viewing reflects affective experience. Sci Adv 7:eabf7129.

Chen J, Leong YC, Honey CJ, Yong CH, Norman KA, Hasson U. 2017. Shared memories reveal shared structure in neural activity across individuals. Nat Neurosci 20:115–125.

Chen S, Langley J, Chen X, Hu X. 2016. Spatiotemporal Modeling of Brain Dynamics Using RestingState Functional Magnetic Resonance Imaging with Gaussian Hidden Markov Model. Brain Connect 6:326–334.

Cocchi L, Gollo LL, Zalesky A, Breakspear M. 2017. Criticality in the brain: A synthesis of neurobiology, models and cognition. Prog Neurobiol 158:132–152.

Cornblath EJ, Ashourvan A, Kim JZ, Betzel RF, Ciric R, Adebimpe A, Baum GL, He X, Ruparel K, Moore TM, Gur RC, Gur RE, Shinohara RT, Roalf DR, Satterthwaite TD, Bassett DS. 2020. Temporal sequences of brain activity at rest are constrained by white matter structure and modulated by cognitive demands. Commun Biol 3:261.

Cunningham JP, Yu BM. 2014. Dimensionality reduction for large-scale neural recordings. Nat Neurosci 17:1500–1509.

Deco G, Kringelbach ML, Jirsa VK, Ritter P. 2017. The dynamics of resting fluctuations in the brain: Metastability and its dynamical cortical core. Sci Rep 7:3095.

Esterman M, Noonan SK, Rosenberg M, Degutis J. 2013. In the zone or zoning out? Tracking behavioral and neural fluctuations during sustained attention. Cereb Cortex 23:2712–2723.

Esterman M, Rothlein D. 2019. Models of sustained attention. Curr Opin Psychol 29:174–180.

Fagerholm ED, Lorenz R, Scott G, Dinov M, Hellyer PJ, Mirzaei N, Leeson C, Carmichael DW, Sharp DJ, Shew WL, Leech R. 2015. Cascades and cognitive state: Focused attention incurs subcritical dynamics. J Neurosci 35:4626–4634.

Falahpour M, Chang C, Wong CW, Liu TT. 2018. Template-based prediction of vigilance fluctuations in resting-state fMRI. Neuroimage 174:317–327.

Finn ES, Shen X, Scheinost D, Rosenberg MD, Huang J, Chun MM, Papademetris X, Constable RT. 2015. Functional connectome fingerprinting: Identifying individuals using patterns of brain connectivity. Nat Neurosci 18:1664–1671.

Fortenbaugh FC, Degutis J, Germine L, Wilmer JB, Grosso M, Russo K, Esterman M. 2015. Sustained attention across the life span in a sample of 10,000: Dissociating ability and strategy. Psychol Sci 26:1497–1510.

Fortenbaugh FC, Rothlein D, McGlinchey R, DeGutis J, Esterman M. 2018. Tracking behavioral and neural fluctuations during sustained attention: A robust replication and extension. Neuroimage 171:148–164.

Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC, Raichle ME. 2005. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A 102:9673–9678.

Gao S, Mishne G, Scheinost D. 2021. Nonlinear manifold learning in functional magnetic resonance imaging uncovers a low-dimensional space of brain dynamics. Hum Brain Mapp 42:4510–4524.

Goodale SE, Ahmed N, Zhao C, de Zwart JA, Özbay PS, Picchioni D, Duyn J, Englot DJ, Morgan VL, Chang C. 2021. Fmri-based detection of alertness predicts behavioral response variability. Elife 10:1–20.

Greene AS, Horien C, Barson D, Scheinost D, Constable RT. 2023. Why is everyone talking about brain state? Trends Neurosci.

Greene DJ, Marek S, Gordon EM, Siegel JS, Gratton C, Laumann TO, Gilmore AW, Berg JJ, Nguyen AL, Dierker D, Van AN, Ortega M, Newbold DJ, Hampton JM, Nielsen AN, McDermott KB, Roland JL, Norris SA, Nelson SM, Snyder AZ, Schlaggar BL, Petersen SE, Dosenbach NUF. 2020. Integrative and Network-Specific Connectivity of the Basal Ganglia and Thalamus Defined in Individuals. Neuron 105:742-758.e6.

Gu S, Pasqualetti F, Cieslak M, Telesford QK, Yu AB, Kahn AE, Medaglia JD, Vettel JM, Miller MB, Grafton ST, Bassett DS. 2015. Controllability of structural brain networks. Nat Commun 6:8414.

Jayakumar M, Balusu C, Aly M. 2023. Attentional fluctuations and the temporal organization of memory. Cognition 235:105408.

Ji E, Lee JE, Hong SJ, Shim W (2022). Idiosyncrasy of latent neural state dynamic in ASD during movie watching. Poster presented at the Society for Neuroscience 2022 Annual Meeting.

Karapanagiotidis T, Vidaurre D, Quinn AJ, Vatansever D, Poerio GL, Turnbull A, Ho NSP, Leech R, Bernhardt BC, Jefferies E, Margulies DS, Nichols TE, Woolrich MW, Smallwood J. 2020. The psychological correlates of distinct neural states occurring during wakeful rest. Sci Rep 10:1–11.

Liu X, Duyn JH. 2013. Time-varying functional network information extracted from brief instances of spontaneous brain activity. Proc Natl Acad Sci U S A 110:4392–4397.

Liu X, Zhang N, Chang C, Duyn JH. 2018. Co-activation patterns in resting-state fMRI signals. Neuroimage 180:485–494.

Lynn CW, Cornblath EJ, Papadopoulos L, Bertolero MA, Bassett DS. 2021. Broken detailed balance and entropy production in the human brain. Proc Natl Acad Sci 118:e2109889118.

Margulies DS, Ghosh SS, Goulas A, Falkiewicz M, Huntenburg JM, Langs G, Bezgin G, Eickhoff SB, Castellanos FX, Petrides M, Jefferies E, Smallwood J. 2016. Situating the default-mode network along a principal gradient of macroscale cortical organization. Proc Natl Acad Sci U S A 113:12574–12579.

Mesulam MM. 1998. From sensation to cognition. Brain 121:1013–1052.

Munn BR, Müller EJ, Wainstein G, Shine JM. 2021. The ascending arousal system shapes neural dynamics to mediate awareness of cognitive states. Nat Commun 12:1–9.

Raut R V., Snyder AZ, Mitra A, Yellin D, Fujii N, Malach R, Raichle ME. 2021. Global waves synchronize the brain’s functional systems with fluctuating arousal. Sci Adv 7.

Rosenberg M, Noonan S, DeGutis J, Esterman M. 2013. Sustaining visual attention in the face of distraction: A novel gradual-onset continuous performance task. Attention, Perception, Psychophys 75:426–439.

Rosenberg MD, Finn ES, Scheinost D, Papademetris X, Shen X, Constable RT, Chun MM. 2016. A neuromarker of sustained attention from whole-brain functional connectivity. Nat Neurosci 19:165–171.

Rosenberg MD, Scheinost D, Greene AS, Avery EW, Kwon YH, Finn ES, Ramani R, Qiu M, Todd Constable R, Chun MM. 2020. Functional connectivity predicts changes in attention observed across minutes, days, and months. Proc Natl Acad Sci U S A 117:3797–3807.

Saggar M, Shine JM, Liégeois R, Dosenbach NUF, Fair D. 2022. Precision dynamical mapping using topological data analysis reveals a hub-like transition state at rest. Nat Commun 13.

Schaefer A, Kong R, Gordon EM, Laumann TO, Zuo X-N, Holmes AJ, Eickhoff SB, Yeo BTT. 2018. Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cereb Cortex 28:3095–3114.

Shine JM. 2019. Neuromodulatory Influences on Integration and Segregation in the Brain. Trends Cogn Sci 23:572–583.

Shine JM, Bissett PG, Bell PT, Koyejo O, Balsters JH, Gorgolewski KJ, Moodie CA, Poldrack RA. 2016. The Dynamics of Functional Brain Networks: Integrated Network States during Cognitive Task Performance. Neuron 92:544–554.

Shine JM, Breakspear M, Bell PT, Ehgoetz Martens K, Shine R, Koyejo O, Sporns O, Poldrack RA. 2019. Human cognition involves the dynamic integration of neural activity and neuromodulatory systems. Nat Neurosci 22:289–296.

Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, Filippini N, Watkins KE, Toro R, Laird AR, Beckmann CF. 2009. Correspondence of the brain’s functional architecture during activation and rest. Proc Natl Acad Sci 106:13040–13045.

Song H, Emily FS, Rosenberg MD. 2021a. Neural signatures of attentional engagement during narratives and its consequences for event memory. Proc Natl Acad Sci 118:e2021905118.

Song H, Park B-Y, Park H, Shim WM. 2021b. Cognitive and Neural State Dynamics of Narrative Comprehension. J Neurosci 41:8972–8990.

Taghia J, Cai W, Ryali S, Kochalka J, Nicholas J, Chen T, Menon V. 2018. Uncovering hidden brain state dynamics that regulate performance and decision-making during cognition. Nat Commun 9:2505.

Terashima H, Kihara K, Kawahara JI, Kondo HM. 2021. Common principles underlie the fluctuation of auditory and visual sustained attention. Q J Exp Psychol 74:705–715.

Tian Y, Margulies DS, Breakspear M, Zalesky A. 2020. Topographic organization of the human subcortex unveiled with functional connectivity gradients. Nat Neurosci 23:1421–1432.

Turnbull A, Karapanagiotidis T, Wang HT, Bernhardt BC, Leech R, Margulies D, Schooler J, Jefferies E, Smallwood J. 2020. Reductions in task positive neural systems occur with the passage of time and are associated with changes in ongoing thought. Sci Rep 10:1–10.

Unsworth N, Robison MK. 2018. Tracking arousal state and mind wandering with pupillometry. Cogn Affect Behav Neurosci 18:638–664.

Unsworth N, Robison MK. 2016. Pupillary correlates of lapses of sustained attention. Cogn Affect Behav Neurosci 16:601–615.

van der Meer JN, Breakspear M, Chang LJ, Sonkusare S, Cocchi L. 2020. Movie viewing elicits rich and reliable brain state dynamics. Nat Commun 11:1–14.

Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, Ugurbil K. 2013. The WU-Minn Human Connectome Project: An overview. Neuroimage 80:62–79.

Vidaurre D, Abeysuriya R, Becker R, Quinn AJ, Alfaro-Almagro F, Smith SM, Woolrich MW. 2018. Discovering dynamic brain networks from big data in rest and task. Neuroimage, Brain Connectivity Dynamics 180:646–656.

Vidaurre D, Smith SM, Woolrich MW. 2017. Brain network dynamics are hierarchically organized in time. Proc Natl Acad Sci U S A 114:12827–12832.

Yamashita A, Rothlein D, Kucyi A, Valera EM, Esterman M. 2021. Brain state-based detection of attentional fluctuations and their modulation. Neuroimage 236:118072.

Yeo BTT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, Roffman JL, Smoller JW, Zöllei L, Polimeni JR, Fisch B, Liu H, Buckner RL. 2011. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 106:1125–1165.

Yousefi B, Keilholz S. 2021. Propagating patterns of intrinsic activity along macroscale gradients coordinate functional connections across the whole brain. Neuroimage 231:117827.

Zhang S, Goodale SE, Gold BP, Morgan VL, Englot DJ, Chang C. 2023. Vigilance associates with the low-dimensional structure of fMRI data. Neuroimage 267.

Author Response

Reviewer #1 (Public Review):

Here the authors set out to disentangle neural responses to acoustic and linguistic aspects of speech. Participants heard a short story, which could be in a language they understood or did not (French vs. Dutch stories, presented to Dutch listeners). Additional predictors included a combination of acoustic and linguistic factors: Acoustic, Phoneme Onsets, Phoneme Surprisal, Phoneme Entropy and, Word Frequency. Accuracy of reconstruction of the acoustic amplitude envelope was used as an outcome measure.

The use of continuous speech and the use of comprehended vs. uncomprehended speech are both significant strengths of the approach. Overall, the analyses are largely appropriate to answer the questions posed.

1) The reconstruction accuracies (e.g., R^2 values Figure 1) seem lower perhaps than might be expected - some direct comparisons with prior literature would be welcome here. Specifically, the accuracies in Figure 1A are around .002-.003 whereas the range seen in some other papers is about an order of magnitude or more larger (e.g. Broderick et al. 2019 J Neurosci; Ding and Simon 2013 J Neurosci).

We thank the reviewer for their constructive comments and careful review of our paper. The important point the reviewer makes stems from whether the reconstruction accuracies presented are from the whole brain/sensor space (as in our submission) or from selected channels (Broderick) or selected sources (Ding & Simon). Moreover, we used R2 score for reconstruction accuracy which is generally of a different order of magnitude than correlation coefficients (as used in Ding and Simon 2013). Crucially when we now selected the “auditory cortex,” we can also report reconstruction accuracies around the language network on the same scale as in the previous studies. In Figure 2 A and B (Figure 1 in the first version of the manuscript), we took the average of model accuracies of each source point over whole brain, without selecting any region of interest, to investigate if each speech feature is incrementally increasing the averaged model accuracy which was a more conservative method than selecting the sources with a stronger response to the stimuli (e.g., the average R2 value over all participants of acoustic model in auditory cortex for French stories is 0.01187 and it is 0.01315 for Dutch stories, which is similar in magnitude to e.g. Broderick et al. 2019 J Neuroscience). TRF accuracies on the brain regions outside of the language network are quite small, so the average accuracy on Figure 2 A and B is almost an order of magnitude lower than previous studies. (Ding and Simon 2013 J Neurosci : “To reduce computational complexity, the MEG sensors in each hemisphere were compressed into 3 components using denoising source separation”, averaged accuracy over all subjects is around 0.2 because they used both correlation as a measure of accuracy (not R2) and backward modeling (decoding) instead of forward modeling. Reconstruction accuracy of decoding models are usually higher than forward models; Broderick et al. 2019 J Neurosci: Averaged across frontocentral channels, averaged R2 over all subjects is 0.0171) Figure 2 C shows sources where accuracies of base acoustic model were significantly different than 0. Reconstruction accuracies around the language network is in the similar scale with the previous studies. Figure 2 D shows the sources where each feature significantly improved the reconstruction accuracy compared to the previous model. Accuracy values are smaller than the accuracies of base acoustic model because they are the values that shows how much each speech feature incrementally increased the accuracy. (E.g Phoneme onset accuracy = (Accuracy of the model Acoustic features + Phoneme Onset) – (Accuracy of the model Acoustic Features). Figure captions are updated on the manuscript.

Figure 2. A) Accuracy improvement (averaged over the sources in whole brain) by each feature for Dutch Stories B) Accuracy improvement (averaged over the sources in whole brain) by each feature for French Stories (Braces in Figure A and B shows the significance values of the contrasts (difference between consecutive models, ** <0.0001, *** <0.001, <0.01, * < 0.05) in linear mixed effect models (Table 2 and 3) C) Source points where accuracies of base acoustic model were significantly different than 0 D) Source points where reconstruction accuracies of the model were significantly different than previous model. Accuracy values shows how much each linguistic feature increased the reconstruction accuracy compared to the previous model.

2) One theoretical point relevant to this and similar studies concerns the use of acoustic envelope reconstruction accuracy as the dependent measure. On the one hand, reconstruction accuracy provides an objective measure of "success", and a satisfying link between stimulus and brain activity. On the other hand, as the authors point out, envelope reconstruction is probably not the primary goal of listeners in a conversation: comprehension is. Some discussion of the implications of envelope reconstruction accuracy might be useful in guiding interpretation of the current work, and importantly, helping the field as a whole grapple with this issue.

Overall, the results support the authors' conclusions that acoustic edges and phoneme features are treated differently depending on whether a listener comprehends the language being spoken. In particular, phoneme features contribute to a greater degree when language is comprehended, whereas acoustic edges contribute similarly regardless of comprehension. These findings are important in part because of prior work suggesting that acoustic edges are critically important for "chunking" continuous speech into linguistic units; the current results re-center language units (phonemes) as critical to comprehension.

Reviewer #2 (Public Review):

In this study, the authors used an audiobook listening paradigm and encoding analysis of MEG to examine the independent contributions to MEG responses of putative acoustic and phoneme-level linguistic features in speech and their modulation by higher-level sentence/discourse constraints and language proficiency. The results indicate that:

1) Acoustic and phoneme features do indeed make independent contributions to MEG responses in frontotemporal language regions (with a left-hemisphere bias for phoneme features).

2) Brain responses to acoustic and phoneme features are enhanced when sentence/discourse constraints are low (i.e. when word entropy is high).

3) While brain responses to phoneme features are enhanced when the language is comprehended (or word entropy is high), the opposite is observed for acoustic features.

These results are taken to support widely held views on the nature of information flow during language processing. On the one hand, processing is hierarchical, consistent with finding 1 above. On the other hand, information flow between lower and high-levels of language processing is also flexible and interactive (finding 2) and modulated by behavioural goals (finding 3).

This is a methodologically sophisticated study with useful findings that I think will be of interest to the burgeoning community investigating 'neural speech tracking' and also to the wider community interested in language processing and predictive coding. Moreover, the evidence appears convincing.

I thought the impact was somewhat limited by the results presentation, which I think missed some key details and made the study somewhat hard to follow (but this issue can be addressed).

Perhaps more major, I do wonder about the novelty of the study as each of the main findings has precedent in the literature. Finding 1 (e.g. Brodbeck, Simon et al.), Finding 2 (e.g. Broderick, Lalor et al.; Molinaro et al.), Finding 3 (e.g. Brodbeck, Simon et al. although here the manipulation of behavioural goals was through a cocktail party listening manipulation and there were was no opposing modulation of acoustic vs phoneme level representations). Thus, while the study appears well executed, overall I am unsure how significant the advance is. Related to this point, the study's findings and theoretical interpretations (e.g. the brain as a hierarchical 'filter') are consistent with widely held views of language processing (at least within cognitive neuroscience) and so again I question the potential advance of the study.

We are thanking the reviewer for bringing this up. While we started our work with the aim to replicate these patterns seem in the literature – which is especially important in the burgeoning area of neural tracking of speech and language - our key extension of these findings is that we can show that phonemic features are encoded more strongly both in a comprehended language compared to an uncomprehended language, and as a function of word-level statistical information, and that there is a tradeoff between acoustic and linguistic features encoding. As the Reviewer mentions, there is a patchwork of consistent findings from very different experimental circumstances, but in order to have strong evidence for the “tradeoff” of hierarchical feature encoding, it is even more crucial to have a design where features can directly compared as we do, and where acoustic differences are carefully controlled in contrast to the presence of linguistic features and language comprehension.

While our results are consistent with Molinaro et al. (2021). – as we also provide support for a cost minimization perspective rather than the perception facilitation perspective discussed in Molinaro et al. - it is important to note that Molinaro et al. only examined the tracking of acoustic features, specifically the speech envelope, using the Phase Locking Value, and did not examine the contribution of lower-level linguistic features. Secondly, Molinaro et al. use a condition-based experimental design in contrast to our naturalistic stimulus approach. In our study, our aim was to investigate the dynamics of encoding both acoustic and linguistic features, and we utilized a multivariate linear regression method on low and high constraining words which ‘naturally’ occurred in our audiobook stimulus across languages. Our results revealed a trade-off between the encoding of acoustic and linguistic features that was dependent on the level of comprehension. Specifically, in the comprehended language, the predictability of the following word had a greater influence on the tracking of phoneme features as opposed to acoustic features, while in the uncomprehended language, this trend was reversed. To best of our knowledge, Brodbeck et al. (2020) showed an effect of attention on the tracking of acoustic features only in cocktail party problem but didn’t investigate the encoding of linguistic features. Brodbeck et al. (2018) showed that linguistic features are represented only in the attended speech but they didn’t explicitly compare the acoustics features as in the previous study. Both studies used a mixed speech and investigated the effect of attention rather than comprehension. In our study, we investigated the effect of comprehension where both stimuli were attended. We found that even in the uncomprehended language, linguistic features are represented as opposed to unattended speech in Brodbeck et al. (2018) study, however it was less strong than the comprehended language. Additionally, one of the goals in this study was to investigate the effect of context on the representations of acoustic and phoneme level features. Opposing modulation of acoustic and phonemic features in our study was driven by the contextual information. However, as we also mentioned in the discussion, we don’t expect the effect of context on the uncomprehended language so the modulation of acoustic features could be related to statistical chunking of acoustic signal for frequent words, essentially reflecting recognition of those single function words such as le, la, un, une.

We have now revised the Discussion (we revised manuscript as highlighted in red in this text) to clarify the advance of this study and how this study adds more on previous studies.

Author Response

eLife assessment

Mizukami et al. propose a scenario for the evolutionary origin of the coronary artery in amniotes by comparing the morphologies of the vasculatures across several species and developmental timepoints. They show that the coronary arteries of non-amniotes most closely resemble embryonic amniote aortic subepicardial vessels (ASVs), which are replaced by the true coronary arteries during amniote development. While the identification of common vascular structures in diverse taxa is a valuable contribution, additional developmental evidence is needed to confirm that such vessels are truly homologous.

We have extensively revised our paper by including additional animal data and references. While we were unable to obtain useful data on lungfish or coelacanth, we have obtained new data related to the physiology of coronary artery, which has been added to Fig. 7. We have also attempted to compare blood vessels at the molecular level, but found that gene expression patterns in blood vessels throughout the body were not always conserved between lineages, making it difficult to make comparisons between amphibians and amniotes. However, based on comparative morphological analysis using newly added three-dimensional data, it is reasonable to consider the amniotes' ASVs and amphibians' ASV-like vessels to be homologous.

Reviewer #3 (Public Review):

Mizukami et al. compare the structure of the coronary arteries in multiple species of amniotes, amphibians, and fish. By selecting species from each of these taxa, the authors were able to evaluate modifications to the coronary arteries during key evolutionary transitions. In mice and quail, they show two populations of vessels that are visible on the developing heart-true coronary arteries on the ventricle and a second population of vessels on the outflow tract known as the ASV., They found that in amphibians, outflow tract vessels were present but ventricular coronary arteries were completely absent. In zebrafish (a more ancestral species) an arterial branch off the rostral section of the hypobranchial artery was shown to have similar anatomical features to outflow tract vessels found in higher organisms. These zebrafish outflow tract arteries also appeared conserved in several chondriichthyes specimens. The authors conclude that rearrangement of the outflow tract vasculature or hypobranchial arteries in fish during evolution, could be homologous to the ASV population of coronary arteries in amphibians and amniotes. These data give new insight into the evolutionary origins of the coronary vasculature.　

Major Points

1) The manuscript presents important data on the coronary vascular structure of several different species. However, these data alone do not conclusively demonstrate whether the developmental origins of ASV like vessels are homologous. Therefore, care should be taken when concluding that the outflow tract vessels found in all different species are conserved features. While this is a reasonable hypothesis and should be presented, the manuscript could be improved by also discussing alternate explanations. For example, ASVs in mice originate during embryonic development, while in fish and amphibians outflow tract vessels are formed only in mature animals.

We have added data on mice and amphibians (e.g., Fig. 2) and substantially revised the overall development and discussion of the paper. Morphological homology is evident for ASVs and amphibian ASV-like vessels, but the homologous relationship with the hypobranchial artery only suggests a similarity in the embryonic region.

Comparisons of developmental timings of the various structures among diferent lineages of vertebrates reveal that heterochronical shifts are not uncommon. For example, ossification of the head skeleton and vertebrae occurs during the fetal stage in amniotes, but after hatching in larval amphibians and teleosts. A similar trend is observed in the development of the limb bud (paired fins). Overall, the larval stages of amphibians and teleosts are comparable to the fetal stages of amniotes for many structures. We did not suppose this to be particularly unusual, and we did not include it in the text.

2) Figure 3 A-D: The authors state that "the ASV ran through the outflow tract, then entered the aortic root before reaching the ventricle to form a secondary orifice". Do the authors have serial sections to conclude that the vessel branching off the carotid runs the length of the aorta and is continuous with an orifice at the aortic root? The endothelial projection off the aorta in panel C could reasonably be an independent projection. For example, Chen et al., described similar looking projections in the base of the aorta that were not attached to external vessels. A whole mount approach would be the most convincing to show the attachments of the ASV vessel.

We added the data of the whole-mount immunohistochemistry. Please refer Figs. 2 and S2.

3) Figure 3E: Similar as above, how is it concluded that the orifice is continuous with the ASV and that this projection is not the coronary artery stem?

As for quail, we could not achieve as a clear whole-mount staining as in mice. It was also difficult to trace the route in sections because in quail, ASVs are not restricted to a few lines as in mice, but are the plexus of small vessels. Thus, we added the detailed data from mice (Fig. 2, S2) and we emphasized that the position of orifice in quail is exactly same as that in mice.

4) The discussion section could be improved by making some statements more consistent, using more precise or appropriate terminology accepted in the field, and being more cognizant of how the authors' findings fit within the history of the field. For example, when referring to coronary arteries, please clarify whether this refers to ASV/ outflow tract coronary arteries, or true ventricular coronary arteries. In addition, the first sentence of the discussion makes it seem like the origins of coronary arteries were unknown prior to this study, however, their origins have been described in multiple papers previously. The authors could revise their statement to acknowledge these previous findings.

We rewrote the entire text to clarify what each "coronary artery" refers to. We also changed the first section of the discussion as suggested by the reviewer.

Author Response:

The following is the authors' response to the current reviews.

We appreciate the thoughtful critiques of the reviewers. While we agree that performing additional experiments and analyses probing the sensitivity of the technique would be useful for future studies, we are unable to perform additional experiments as our lab has closed. We share this technique as a starting point for further investigation, but it may need to be modified for success in other contexts. We have provided details of the scenarios (life stage, feeding, day, number of ticks) where we successfully sequenced B. burgdorferi from ticks, as well as one where we did not (unfed nymphs) as a starting point. We will clarify in proofing that our qPCR experiments show that we capture the vast majority of B. burgdorferi flaB mRNA from our input samples, suggesting that we are likely capturing the majority of the B. burgdorferi.

In this work, we were most interested in using RNA-seq to perform differential expression analysis between annotated mRNAs across our timepoints. We have provided the number of genes detected in each sample (92% of annotated transcripts on average) as well as the median number of reads covering each gene (604 on average) in the supplemental file containing sequencing statistics. This coverage is highly reproducible across replicates, with an average Pearson correlation of 0.99 between gene expression levels (as Transcripts Per Million) between any two replicates. These data and the fact that many of the gene expression changes we observed align with previous observations of others give us confidence in our differential expression analysis. For those interested in tRNAs or sRNAs, we think that it would be best to modify the protocol to focus specifically on capturing those sequences in the library preparation. We encourage others interested in other aspects of our data to download it and explore it.

We will correct remaining wording issues in proofing.

—————

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

Dear Reviewing Editor,

We thank you and the reviewers for the thoughtful comments on our manuscript, and we are excited to submit a revised version of our manuscript “Longitudinal map of transcriptome changes in the Lyme pathogen Borrelia burgdorferi during tick-borne transmission.” In response to the reviews, we have made the following changes to our manuscript:

1. We updated the text for increased clarity around experimental details, including statistical analyses.

2. We added additional details about the mapping of non-Bb reads as well as more information about Bb read coverage.

3. We compared our differentially expressed genes to 4 previous studies of global transcriptional changes in different tick feeding contexts.

4. We updated the discussion to address these comparisons as well as caveats of our study more directly.

Please see our responses to individual comments below.

Reviewer #1 (Public Review):

In this study, Sapiro et al sought to develop technology for a transcriptomic analysis of B. burgdorferi directly from infected ticks. The methodology has exciting implications to better understand pathogen RNA profiles during specific infection timepoints, even beyond the Lyme spirochete. The authors demonstrate successful sequencing of the B. burgdorferi transcriptome from ticks and perform mass spectrometry to identify possible tick proteins that interact with B. burgdorferi. This technology and first dataset will be useful for the field. The study is limited in that no transcripts/proteins are followed-up by additional experiments and no biological interactions/infectious-processes are investigated.

Critiques and Questions:

We thank the reviewer for these thoughtful critiques and helping us improve our manuscript.

This study largely develops a method and is a resource article. This should be more directly stated in the abstract/introduction.

We edited the abstract and introduction to more directly state that we are sharing a new method and a resource for future investigations. (Lines 29-32; 101-103)

Details of the infection experiment are currently unclear and more information in the results section is warranted. State the species of tick and life-stage (larval vs nymphal ticks) used for experiments. For RNA-seq, are mice are infected and ticks are naïve or are ticks infected and transmitting Borrelia to uninfected mice?

We updated the results section to more clearly state the tick species and life stage and to make it more clear that infected ticks are transmitting Bb to naïve mice. (Lines 113-115)

What is the limit of detection for this protocol? Experimental data should be provided about the number of B. burgdorferi required to perform this approach.

We performed this protocol on pools of 6 (for later feeding stages) to 14 (for early stages) infected nymphs. Published studies (PMID: 7485694, PMID: 11682544) suggest that one day after attachment, there may be a few thousand Bb per tick, suggesting what we’ve measured here may come from on the order of 104 Bb. We were not able to capture consistent data from Bb from unfed ticks, which may be due to lower numbers or to an altered transcriptional state caused by lack of nutrients in the unfed tick. We updated the discussion to reflect some of these limitations and uncertainties. (Lines 461-465)

More information regarding RNA-seq coverage is required. Line 147-148 "read coverage was sufficient"; what defines sufficient? Browser images of RNA-seq data across different genes would be useful to visualize the read coverage per gene. What is the distribution of reads among tRNAs, mRNAs, UTRs, and sRNAs?

As we were interested in differential expression analysis, we defined sufficient as the number of reads needed per gene to determine statistically significant expression changes across days, which with DESeq2 is typically 10 reads. We reworded this section for clarity and added additional information about the median number of reads per gene which is also useful in thinking about differential expression analysis. (Lines 163-170) As we chose to focus on differential expression analysis here, we believe these are most relevant metrics to cover.

My lab group was excited about the data generated from this paper. Therefore, we downloaded the raw RNA-seq data from GEO and ran it through our RNA-seq computational pipeline. Our QC analysis revealed that day 4 samples have a different GC% pattern and that a high percentage of E. coli sequences were detected. This should be further investigated and addressed in the paper: Are other bacteria being enriched by this method? Why would this be unique to day 4 samples? Does this affect data interpretation?

We appreciate the interest in our data and pointing out this anomaly. We found that the day 4 samples do have a high percentage of reads that mapped to a bacterial species, Pseudomonas fulva, rather than ticks as we expected. (The reads that map to E. coli also map to P. fulva.) We have updated the results to include this information (Lines 156-165). We believe this is likely due to contamination from collecting ticks after they have fallen off mice in cages on day 4, rather than pulling ticks off the mice as in days 1-3. Unfortunately, as our lab has shut down, we cannot investigate the source further. We do think the high percentage of P. fulva reads suggests that other bacteria can be enriched with the anti-Bb antibody we used. We’ve updated the discussion to highlight this caveat. (Lines 459-460)

While the presence of these bacterial reads did lower our overall Bb mapping rate and necessitate deeper sequencing for the day 4 samples, the Bb sequencing coverage of these samples is on par with samples from the other days in terms of percentage of genes with at least 10 reads and median number of reads per gene. Fewer than 0.0002% of the reads that map to Bb genes in any day 4 sample also map to P. fulva. We found that this small fraction of reads is dispersed across 334 genes in which an average of 0.05% (maximally 2.3%) of day 4 reads also map to P. fulva. Therefore, these bacterial reads do not change our interpretation of the results comparing gene expression across days, including day 4.

Comprehensive data comparisons of this study and others are warranted. While the authors note examples of known differentially expressed genes (like lines 235-241), how does this global study compare to other global approaches? Are new expression patterns emerging with this RNA-seq approach compared to other methods? What differences emerged from day 1 to day 4 ticks compared to differences observed in unfed to fed ticks or fed ticks to DMC experiments? Directly compare to the following studies (PMID: 11830671; PMID: 25425211; PMID: 36649080.

We added comparisons of our list of DE genes to those noted to change between “unfed tick” and “fed tick” culture conditions (PMID: 11830671 and 12654782), as well as fed nymph to DMC (PMID: 25425211 and 36649080) (Lines 231-252, Figure S4). These comparisons pointed us to two main findings: that global changes to Bb in different culture conditions generally agreed with the most dramatic changes we saw in our data, and that the timing of expression increases during feeding may relate to whether genes are more highly expressed in fed ticks or in mammalian conditions. Overall, the majority of our DE genes have been identified in at least one of these studies or in the other studies we compared to outlining RpoS, Rrp1, and RelBbu regulons. As many of these studies were asking slightly different questions and using different conditions and vastly different technology, we would expect some differences to arise from different contexts and some to be purely technical. The genes that were not seen in these previous studies tended to follow the same functional patterns we saw overall, heavily skewing towards genes of unknown function, outer surface proteins, and a handful of genes related to other functions. With the current state of the functional annotation of the genome, it is difficult to assess whether these amount to new expression patterns in and of themselves, so we focused on the overall trends in our data rather than those that were different from other studies.

Details about the categorization of gene functions should be further described. The authors use functional analysis from Drechtrah et al., 2015, but that study also lacks details of how that annotation file was generated. Here, the authors have seemed to supplement the Drechtrah et al., 2015 list with bacteriophage and lipoprotein predictions - which are the same categories they focus their findings. Have they introduced a bias to these functional groups? While it can be noted that many lipoproteins are upregulated (or comment on specific genes classes), there are even more "unknown" proteins upregulated. I argue that not much can be inferred from functional analysis given the current annotation of the B. burgdorferi genome.

We strongly agree that the current annotation of the Bb genome makes it difficult to perform meaningful global functional analysis, but we feel it is useful to get a general overview of gene functions. We described our methods for classifying genes into functional categories in the methods, in which we relied on previously published papers to make our best estimate of gene category (noted for each gene in the Table S4). Due to the lack of annotations for many genes, we focused on the relatively well-defined category of lipoproteins, as these are overrepresented as a group in our upregulated genes, as well as phage genes, which are not necessarily overrepresented, but are still interesting to us. We hope that others will look at the data (particular in Table S4, but also Table S3, or download the raw data and do their own analysis) with their own interests and biases and dig more into genes that we did not highlight specifically. We provide this data as a resource with the hope that some of the genes of unknown function that we see change here will be the subject of future functional studies so that this is less of problem in the future.

Reviewer #1 (Recommendations For The Authors):

In general, the paper is well written and digestible for a broad audience. However, some of the figure graphics are unnecessary and take away from the data. Please label tick species and tick life-stage in Figure 1 drawings. The legend of Figure 1 requires citations. The Figure 4B graphic is unnecessary and the colors are confusing as they are too similar to the color palette of Figure 4A, where the colors have meaning. The Figure 5A graphic is unnecessary and takes away from the data embedded within it.

We more clearly labeled the species in Figure 1 and added citations to the legend. We have simplified Figures 4A and 5A for clarity.

Clarify lines 220-259 and Figure 3. What days are being compared? Downregulated genes should also be commented on.

We considered our set of differentially expressed genes as those that changed two-fold (multiple hypothesis adjusted p-value < 0.05) in any of the three comparisons shown in Figure 2 (day1 to day2, day1 to day3, day1 to day4). We clarified this at multiple points in the results (i.e Line 273). We commented on downregulated genes throughout, although as there were fewer genes and the magnitude of change was smaller, we focused more on upregulated genes.

Line 327-329, state numbers not percentages. How many Bb proteins were actually detected?

We updated this section to include numbers (Lines 371-374). In concordance with our sequencing data, we found (and were looking for) mainly tick proteins in this experiment.

Data availability: B. burgdorferi and tick oligo sequences used for DASH should be provided in a supplemental table.

We added a supplemental table of these sequences (Table S9). Please note they have been previously published in Dynerman et al. 2020 and Ring et al. 2022.

Reviewer #2 (Recommendations For The Authors):

The manuscript is overall well written and easy to follow. The data are compelling and support the conclusions. The discussion of this work is however highly insufficient and needs to be thoroughly edited:

- Statistical analysis: The authors mention that DESeq2 was used. Please provide information on the type and the stringency of the tests used for differential gene expression analysis, including any additional potential correction for p-values (Bonferroni). The authors mention that genes with fold changes >2 were used for analysis, yet there is no information on the p-value cut off or if the genes with fold changes >2 were statistically significant. Please provide detail and rationale for the analysis.

We clarified in the results and methods (Lines 200, 642-644) that we required a adjusted p-value < 0.05 from DESeq2’s Wald test with Benjamini-Hochberg correction along with a two-fold change when determining our genes of interest. As small fold changes showed statistically significant differences, we chose to set a fold change cutoff in most of our analysis to help us focus on the most highly expressed genes, like other studies we compared our data to. We included all of the DESeq2 results in Table S3 so that others may explore the data with different cutoffs if desired.

- The field has been generating data on gene expression in ticks for decades. Yet, many of these studies are not referenced here. There is no discussion of how the data described here compares to what is known in the literature. For example, Venn diagrams or tables could be included for comparison with the data described lines 208-216. Extensive description and comparison of the data to the literature should be added in the discussion, and similarities/discrepancies should be discussed appropriately.

We added additional comparisons to four different papers looking at global gene expression in Bb in the fed tick or tick-like culture conditions (Lines 231-252, Figure S4). This information as well as comparisons to transcriptional regulons (Figure S3) is available in Table S4. In addition to discussing some examples in the results, we added more information in the discussion regarding these comparisons (Lines 420-425). The majority of the genes that we see change over feeding have been previously noted to change expression during the enzootic cycle or be regulated by transcriptional programs active during this timeframe, and we have more clearly stated that. We focused on similarities here as these papers all ask slightly different questions in different contexts and use different technology which could all account for the many differences in individual genes between all of them and our work.

- There is no discussion of the caveats of the study: for example, the authors are using an anti-OspA antibody, which could induce bias. The authors provide in-vitro pull down data supporting that this should not be an issue, but the pull down is performed from BSK-grown bacteria. This caveat should be discussed.

We’ve added a paragraph to the discussion including this caveat and others (Lines 453-463).

- Timing of RNA extraction: There is over 1h of delay between initial tick collection and RNA fixation. The effects of time on gene expression should be discussed.

Although we were able to show that this timeframe did not affect cultured Bb gene expression, we added this to the discussion.

- Gene expression is compared to Day 1. This introduces analyses bias as it does not allow identification of transcripts that first change upon initial feeding. This caveat should also be discussed

We added this caveat – that we may miss gene expression changes in the first 24 hours of feeding – to the discussion.

- This study is performed with 1 strain of B. burgdorferi on one tick species. Please provide perspective on the impact of these findings on Lyme disease causing spirochetes and their vectors broadly.

We believe this method could be easily adaptable to study gene expression in other spirochete/vector pairs to determine similarities and differences and we added a comment to the discussion.

- The discussion should also include insights on how to build on this work and include additional areas of method development to increase the recovery of B. burgforferi from ticks or other organisms and facilitate future transcriptomic studies.

We added a few ideas to the discussion noting that this protocol could be modified for use in other timeframes, with other antibodies, or in other organisms. We also highlight the recent advent of TBDCapSeq by Grassmann et al. that may be used in conjunction with this type of protocol.

Minor comments:

- Consider re-wording the description of the methods and findings to the third person for coherence.

The majority of the methods are now written in third person.

- Over 90% of the reads did not map to B. burgdorferi: please provide additional information on what these reads mapped to (tick or mouse), and if the data reflects what is known in the literature

We have updated the results and discussion with information about the reads that do not map to Bb (Lines 156-166). The majority of reads mapped the tick genome, which is what we expected. While a large number of reads in our day 4 samples unexpectedly mapped to Pseudomonas fulva, we do not believe this affects the interpretation of our data as we were still able to get broad genome coverage of Bb in these samples.

- Please be more clear in the result section on the life stage of the ticks used for these studies.

We have updated the results to clarify throughout.

- Indicate how many total reads were generated for each sample

This information is present in Table S1.

- Provide statistical analyses for Figures 1C and D.

We added t tests to determine statistical differences for these panels.

Reviewing Editor (Recommendations for The Authors):

1. It is important to mention in the abstract (line 27) that 'upregulated genes' is in comparison to day 1. This is also true in the introduction (lines 92-93).

We updated in the results and introduction to more clearly include that day 1 is our baseline measurement.

2. It is also important to discuss in the manuscript that because your 'controls' are day 1 samples, initial transcriptome changes in response to the tick environment might be missed.

This has been added in the discussion as a caveat (Lines 460-463).

3. As someone who does not work with Bb, I would like to have seen a clearer description of what the feeding event looks like. Although there is some text in the introduction that touches on that ('prolonged nature of I. scapularis feeding'), I would like to see something even clearer. Maybe stating that feeding may take from x-y days would clarify that for the non-specialist.

We updated the results to more clearly state that the tick falls off of the mice by around 4 days after feeding, our last time point (Lines 113-115). Additional details of tick feeding are also in the Figure 1 legend.

4. In Fig. 3 linear DNA molecules seem to be drawn to scale. Is that also the case for plasmids? This could be clarified in the legend.

The genome is drawn approximately to scale. We noted this and updated the legend with more information about how linear and circular plasmid names denote their size.

5. Figure 5C: Colors are a bit confusing here. The legend indicates that they refer to fold changes, but the scale in the panel shows expression levels, not fold changes. Please clarify. Also, is this really TPM or RPKM? If comparisons of relative levels between different genes are made, number of reads should be normalized by gene length.

The heatmap in Figure 4C does show expression levels, and we updated the legend to more clearly state this. The highlighted gene names are meant to show which genes change two-fold during this time (those present in panel A). The data are presented as TPM (transcripts per million), which, like RPKM, is normalized by gene length (PMID: 20022975).

Author Response:

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

We have now incorporated the changes recommended by the reviewers to improve the interpretations and clarity of the manuscript. We are grateful for their thoughtful comments and suggestions, which have significantly strengthened the manuscript.

Reviewer #1 (Public Review):

Park et al demonstrate that cells on either side of a BM-BM linkage strengthen their adhesion to that matrix using a positive feedback mechanism involving a discoidin domain receptor (DDR-2) and integrin (INA-1 + PAT-3). In response to its extracellular ligand (Collagen IV/EMB-9), DDR-2 is endocytosed and initiates signaling that in turn stabilizes integrin at the membrane. DDR-2 signaling operates via Ras/LET-60. This work's strength lies in its excellent in vivo imaging, especially of endogenously tagged proteins. For example, tagged DDR-2:mNG could be seen relocating from seam cell membranes to endosomes. I also think a second strength of this system is the ability to chart the development of BM-BM linkage over time based on the stages of worm larval development. This allows the authors to show DDR signaling is needed to establish linkage, rather than maintain it. It likely is relevant to many types of cells that use integrin to adhere to BM and left me pondering a number of interesting questions.

We thank the reviewer for highlighting the strengths and impact of our work in expanding our understanding of tissue linkages and how DDR and integrins might work in other contexts.

For example: (1) Does DDR-2 activation require integrin? Perhaps integrin gets the process started and DDR-2 positively reinforces that (conversely is DDR-2 at the top of a linear pathway)?

DDR activation by receptor clustering upon exposure to its ligand collagen is well documented (Juskaite et al., 2017 eLife PMID: 285ti0245). Clustered DDR is rapidly internalized into endocytic vesicles, where full activation of tyrosine kinase activity is thought to occur (Fu et al., 2013 J Biol Chem PMID: 23335507). Supporting this model, we found that concentrated type IV collagen is required for vesicular DDR-2 localization in the utse and seam cells at the utse-seam connection. Whether DDR-2 activation requires integrin has not been fully established. However, one study using mouse and human cell lines showed that DDR1 activation occurs independent of integrin (Vogel et al., 2000 J Biol Chem PMID: 10681566), consistent with the latter possibility raised by the reviewer that DDR-2 is upstream of integrin.

To test these hypotheses, we require an experimental condition where loss or near complete loss of INA- 1 integrin is achieved by the mid-to-late L4 larval stage, when DDR-2 is activated by collagen and taken into endocytic vesicles. Currently, we can only partially deplete INA-1 by RNAi (Figure 5—figure supplement 2E), and strong loss of function mutations in ina-1 result in early larval arrest and lethality (Baum and Garriga, 1titi7 Neuron PMID: ti247263). To overcome these obstacles, we are adapting the new FLP-ON::TIR1 system developed for precise spatiotemporal protein degradation in worms (Xiao et al., 2023 Genetics PMID: 36722258). We hope to achieve a near complete knockdown of ina-1 with this timed depletion strategy. In the future, we will use this system to block DDR-2 and integrin function specifically in the utse or seam cells, to complement our current dominant negative mis-expression approach.

(2) In ddr-2(qy64) mutants, projections seem to form from the central portion of the utse cell. Does this reveal a second function for DDR-2, regulating perhaps the cytoskeleton?

We thank the reviewer for their observation and agree with their interpretation. We think it is important to comment on this and have stated in the results text, lines 208-212: “In addition, membrane projections emanating from the central body of the utse were detected in ddr-2(qy64) animals. These projections were first observed at the mid L4 stage and persisted to young adulthood (Figure 2C). These observations suggest that DDR-2 functions around the mid L4 to late L4 stages to promote utse-seam attachment, and that DDR-2 may also regulate utse morphology.”

And (3) can you use the forward genetic tools available in C. elegans to find new genes connecting DDR-2 and integrin?

This is an excellent suggestion. We found that loss of ddr-2 strongly enhanced the uterine prolapse (Rup) defect caused by RNAi mediated depletion of integrin. To find new genes connecting DDR-2 and integrin, a targeted screen for the Rup phenotype could be performed in an integrin reduction of function condition. As we cannot work with null or strong loss-of-function ina-1 alleles (described above), the screen could be conducted with either timed depletion of INA-1 with candidate RNAi treatments, or combinatorial ina-1 RNAi with candidate RNAi treatments.

I do see two areas where the manuscript could be improved. First, the authors rely on imprecise genetic methods to reach their conclusions (i.e. systemic RNAi, or expression of dominant negative constructs.) I think their conclusion would be stronger if they used tissue specific degradation to block ddr-2 function specifically in the utse or seam cells. Methods to do this are now regularly used in C. elegans and the authors have already developed the necessary tissue-specific promoters.

We agree with the reviewer that tissue specific degradation of DDR-2 in the utse and seam cells will complement and strengthen our evidence for the site of action of DDR-2. As described earlier, we are currently adapting the FLP-ON::TIR1 tissue degradation system to perform these experiments and will provide our findings in a follow-up manuscript.

Second, the manuscript is presented in the introduction as a study on formation and function of BM-BM linkage. The authors start the discussion in a similar manner. But their results are about adhesion between cells and BM. In fact they show the BM-BM linkage forms normally in ddr-2 mutants. Thus it seems like what they have really uncovered is an adhesion mechanism that works in parallel to the BM-BM linkage. Since ddr-2 appears to function equally in both utse + seam cells (based on their dominant negative data), there are likely three layers of adhesion (utse-BM, BM-BM, BM-seam) and if any of those break down, you get a partially penetrant rupture phenotype.

The reviewer raises an important and interesting point, and we agree that we did not articulate the organization of the utse-seam tissue connection clearly. The utse-seam connection is comprised of the utse and seam BMs each ~50nm thick, and a connecting matrix bridging the two BMs, which is ~100nm thick (Vogel and Hedgecock, 2001 Development PMID: 11222143). Type IV collagen builds up to high levels within the connecting matrix and links the utse and seam BMs, and its concentration is required for DDR-2 vesiculation. An important point we did not highlight is that type IV collagen is approximately 400 nm long (Timpl et al. 1ti81, Eur J Biochem PMID: 6274634). Thus, collagen molecules within the connecting matrix could span the entire length of the utse-seam connection and project into the utse and seam BMs to interact with cell surface receptors. Consistent with this possibility, we found that buildup of type IV collagen that spans the utse-seam BM-BM linkage correlated with the timing of DDR-2 activation/vesiculation within utse and seam cells. In addition, super-resolution imaging of the mouse kidney glomerular basement membrane (GBM), a tissue connection between endothelial BM and epithelial (podocyte) BM, showed type IV collagen, which spans the BMs, projects into the endothelial and podocyte BMs (Suleiman et al., 2013 eLife PMID: 24137544 ). We carefully considered these points to generate the schematics in Figure 1A and Figure 8, but failed to articulate this point in the manuscript. We are grateful for the reviewer for bringing up our error and have now stated these details in the text to address the reviewer’s concern as outlined below.

In the introduction (lines ti3-ti6): “A BM-BM tissue connection between the large, multinucleated uterine utse cell and epidermal seam cells stabilizes the uterus during egg laying. The utse-seam connection is formed by BMs of the utse and the seam cells, each ~50 nm thick, which are bridged by an ~100 nm connecting matrix (Vogel and Hedgecock 2001, Morrissey, Keeley et al. 2014, Gianakas, Keeley et al. 2023).”

In the discussion (lines 507-520): “We also found that internalization of DDR-2 at the utse-seam connection correlated with the assembly of type IV collagen at the BM-BM linkage and was dependent on type IV collagen deposition. Type IV collagen is ~400 nm in length and the utse-seam connecting matrix spans ~100 nm, while the utse and seam BMs are each ~50 nm thick (Timpl, Wiedemann et al. 1ti81, Vogel and Hedgecock 2001). Thus, collagen molecules in the connecting matrix could project into the utse and seam BMs to interact with DDR-2 on cell surfaces. Consistent with this possibility, super- resolution imaging of the mouse kidney glomerular basement membrane (tiBM), a tissue connection between podocytes and endothelial cells, showed type IV collagen within the tiBM projecting into the podocyte and endothelial BMs (Suleiman, Zhang et al. 2013). As DDR-2 is activated by ligand-induced clustering of the receptor (Juskaite, Corcoran et al. 2017, Corcoran, Juskaite et al. 201ti), it suggests that the BM-BM linking type IV collagen network, which is specifically assembled at high levels, clusters and activates DDR-2 in the utse and seam cells to coordinate cell-matrix adhesion at the tissue linkage site.”

These concerns do not undercut the significance of this work, which identifies an interesting mechanism cells use to strengthen adhesion during BM linkage formation. In fact, I am excited to read future papers detailing the connection between DDR-2 and integrin. But before undertaking those experiments the authors should be certain which cells require DDR-2 activity, and that should not be determined based solely on mis expression of a dominant negative.

We thank the reviewer for recognizing the significance of our work and reiterate that we will use tissue-specific degradation for site of action experiments in future studies on the biology of the utse- seam tissue linkage.

Reviewer #2 (Public Review):

This paper explores the mechanisms by which cells in tissues use the extracellular matrix (ECM) to reinforce and establish connections. This is a mechanistic and quantitative paper that uses imaging and genetics to establish that the Type IV collagen, DDR-2/collagen receptor discoidin domain receptor 2, signaling through Ras to strengthen an adhesion between two cell types in C. elegans. This connection needs to be strong and robust to withstand the pressure of the numerous eggs that pass through the uterus. The major strengths of this paper are in crisply designed and clear genetic experiments, beautiful imaging, and well supported conclusions. I find very few weaknesses, although, perhaps the evidence that DDR-2 promotes utse-seam linkage through regulation of MMPs could be stronger. This work is impactful because it shows how cells in vivo make and strengthen a connection between tissues through ECM interactions involving collaboration between discoidin and integrin.

We appreciate the reviewer’s assessment of the impact of our work in detailing a mechanism for how cells increase their adhesion to the ECM to establish connections between adjacent tissues. We have softened the interpretation of our MMP localization data to address the reviewer’s concern (detailed below).

Reviewer #1 (Recommendations For The Authors):

Regarding Figure 1D, is it possible to show when the BM forms on the cartoons more clearly (something like the 3rd section of Fig 3A)? I can see it in the timeline but it's hard to follow in the diagrams.

We agree with the reviewer that we could show when the BM-BM connecting matrix forms more clearly in Figure 1D. Hemicentin and fibulin, the earliest components of the connecting matrix, are detected at very low levels at the utse-seam connection during the mid-L4 stage and are more prominently localized by the mid-to-late L4 stage (Gianakas et al., 2023 J Cell Biol PMID: 36282214). For this reason, we only show the connecting matrix in yellow from the mid-to-late L4 stages onward. We have now made the BM-BM connection more prominent in the figure 1D cartoons with boxed outlines (similar to Figure 3A as the reviewer suggested). We also added a label for the time window when the BM-BM connection forms.

Regarding the RNAi induced prolapse phenotype, looking at 2B, it appears that between 5% and 10% of animals have uterine prolapse when fed control RNAi. Is this correct, it seems very high? This prolapse in control animals was not observed other RNAi experiments such as Figure 5C.

We thank the reviewer for pointing this out. For Figure 2B, the control used was wild-type N2 animals fed with OP50 E. coli bacteria, rather than HT115 bacteria carrying the L4440 empty vector (control RNAi). This is because the main comparisons were to five ddr-1 and ddr-2 mutant strains. We did notice a slightly higher baseline uterine prolapse frequency (5% on average, detailed in Figure 2—Source data 1) in wild-type animals fed OP50 bacteria, compared to HT115 bacteria fed animals (approximately 1-2% on average). It is possible this could be linked to the nutritional differences in the two bacterial strains. However, we are confident of our data in Figure 2B as we carried out 3 independent trials, and the uterine prolapse frequencies in ddr-1 mutant animals matched the baseline in wild-type animals, while the frequencies for ddr-2 mutants were all increased over the baseline in all trials (as detailed in Figure 2—Source data 1).

Relating to the point above, in reading the methods to try to understand how they did the RNAi, I noticed that they measure prolapse continually over five days. I didn't realize it takes a long time to occur. I think they should explain this in the text and in the figures. Reading the manuscript I thought prolapse occurred as soon as mutant animals began laying eggs. In the text they should explain this when they first assay the phenotype (page 7), and for figures the Y axis on the graphs could say "% uterine prolapse after 5 days."

We thank the reviewer for their suggestions. We did not articulate clearly that the utse-seam connection is able to withstand some mechanical stress, even when key components are lost. It’s only over time and repeated use that the connection breaks down. This is likely because a number of components contribute to the connection and as we have shown previously, there is feedback, such that when one components is reduced, such as collagen, hemicentin is increased in levels at the BM-BM connection. Since ruptures arising from utse-seam detachments typically occur sometime after the onset

of egg-laying, we screened the entire egg-laying period (days two to five post-L1) as described in Gianakas et al. 2023. We have now incorporated these points in the text and figures as follows:

In the introduction, we clarified that utse-seam BM-BM connection breaksdown over time, by adding (lines titi-105): “Hemicentin promotes the recruitment of type IV collagen, which accumulates at high levels at the BM-BM tissue connection and strengthens the adhesion, allowing it to resist the strong mechanical forces of egg-laying. The utse-seam connection is robust, with each component of the tissue- spanning matrix contributing to the BM-BM connection (Gianakas, Keeley et al. 2023). This likely accounts for the ability of the utse-seam connection to initially resist mechanical forces after loss of any one of these components, delaying the uterine prolapse phenotype until sometime after the initiation of egg-laying.”

We expanded the results text when we first describe the Rup phenotype (lines 183-184): “We first screened for the Rup phenotype caused by uterine prolapse, observing animals every day during the egg-laying period, from its onset (48 h post-L1) to end (120 h) (Methods)”.

We provided more detail in the Methods section (lines 784-7ti0): “Uterine prolapse frequency was assessed as described previously (Gianakas et al 2023). Briefly, synchronized L1 larvae were plated (~20 animals per plate) and after 24 h, the exact number of worms on each plate was recorded. Plates were then visually screened for ruptured worms (uterine prolapse) every 24 h during egg-laying (between 48 h to 120 h post-L1). We chose to examine the entire egg-laying period as ruptures arising from utse-seam detachments do not usually occur at the onset of egg-laying, but after cycles of egg-laying that place repeated mechanical stress on the utse-seam connection (Gianakas et al 2023).”

Finally, we modified the Y-axes of graphs in Figure 2B and 5C and the respective figure legends as suggested by the reviewer.

Then I went back and compared to the previous publication (Gianakas, 2023). I would be interested to see a time course of how many animals prolapse after 1 day, 2 days, etc.? Is this consistent with their data on hemicentrin?

We agree with the reviewer that a time course of uterine prolapse would be interesting as we saw ruptures occur throughout the egg-laying period. However, for the hemicentin knockdown experiments in Gianakas et al. 2023 as well as the experiments in this study, we recorded only the pooled number of animals with ruptures at the end of the experimental window. In future studies we will also record the uterine prolapse frequencies on each day to generate time courses that will provide more insight into the function of proteins at the utse-seam connection.

Lines 183-184: I'm not sure what it means to say "trended towards displaying a significant Rup phenotype?" Since the difference was not statistically significant, it would be better to say something like "increased but not statistically significant."

We agree with the reviewer and have now modified this sentence (lines 190-193): “Animals carrying the ddr-2(ok574) allele, which deletes a portion of the intracellular kinase domain (Unsoeld, Park et al. 2013),also showed an increased frequency of the Rup phenotype compared to wild-type animals, although this difference was not statistically significant (Figure 2A and B)”.

Line 186: 'penetrant' needs a qualifier to indicate the magnitude of the proportion of individuals with the phenotype.

As we provide the Rup frequency numbers in Figure 2—Source data 1, we modified the sentence as follows (lines 1ti3-1ti5): “We further generated a full-length ddr-2 deletion allele, ddr-2(qy64), and confirmed that complete loss of ddr-2 led to a significant uterine prolapse defect (Figure 2A and B).”

Lines 206-208; could the mounting/imaging procedure (which I assume requires squeezing the worm between agarose pad and coverslip) alter the occurrence of prolapse? I would think prolapse would occur more frequently under these conditions as compared to worms laying eggs on a plate.

The reviewer brings up an important concern. The mounting and imaging procedure does require placing the worm between an agarose pad and a coverslip. However, this did not alter the occurrence of uterine prolapse in this experiment. We were careful to perform the same procedure on both wild-type and ddr- 2(qy64) animals to control for this. As detailed in the manuscript, none of the eight wild-type animals we mounted underwent uterine prolapse after recovery off the coverslip, and among the ddr-2(qy64) mutants we mounted, only the ones that exhibited utse-seam detachments went on to rupture later.

We articulated these points more clearly by modifying lines 214-216 as follows: “Wild-type and ddr- 2(qy64) animals were mounted and imaged at the L4 larval stage for utse-seam attachment defects, recovered, and tracked to the 72-hour adult stage, where they were examined for the Rup phenotype.”

In seam cells you can see that DDR-2:mNG is present at membranes from early to mid L4, which makes sense. But I cannot see it on the membrane at any time point in the utse. Perhaps it is obscured by the yellow dotted line. Should it be visible on utse membranes before it is endocytosed?

The reviewer raises an interesting question. We think it is likely that DDR-2 is initially on the membrane of the utse like it is on the seam cells. However, we have not observed this, possibly due to the complex shape and thin membrane extensions of the utse. We are unable even to detect clear membrane enrichment of membrane markers in the utse (for example, compare the utse and seam membrane markers in Figure 3B). Thus, we refrained from speculating on DDR-2 utse membrane localization in the manuscript, and instead focused on the pattern of vesicular DDR-2 peaking at the late L4 stage, which was clearly visible in both the utse and seam cells.

Sup Fig 3A - please show quantification of seam cells not contacting utse at the same Y-axis scale as for regions that do contact utse.

We have modified the Y-axis scale for the quantification of the seam region not contacting the utse.

Figure 4A - I don't see a difference between WT and ok574 - what am I missing?

In the representative ok574 animal shown, a portion of the utse arm on the top right is detached from the seam. To make this phenotype clearer, we have recropped the image panels, readjusted the brightness and contrast of the utse and the seam, and redrawn the outline of the detachment to make this clearer.

Figure 4C+D, and lines 296-298: I'd bet that both are needed to recruit DDR-2 to membranes. But him-4 has a more severe phenotype because the RNAi knockdown is much more effective (perhaps b/c they are using the newer t444t vector).

We agree with the reviewer that the him-4 knockdown phenotype is likely more severe than emb-9 knockdown. Type IV collagen at the utse-seam connection is very stable compared to hemicentin (Gianakas et al 2023, J Cell Biol PMID: 36282214, see Fig. 5C), which could explain the lower knockdown efficiency.

We modified our interpretation of the data in the text as follows (lines 308-312): “In addition, we did not detect DDR-2 at the cell surface, suggesting that hemicentin has a role in recruiting DDR-2 to the site of utse-seam attachment. It is possible that collagen could also function in DDR-2 recruitment, but we could not assess this definitively due to the lower knockdown efficiency of emb-9 RNAi (Figure 4—figure supplement 1A).”

Reviewer #2 (Recommendations For The Authors):

Line 218 DDR-2 (typo)

We have corrected this typo.

Evidence (line 344-348) may not be strong enough to say whether or not DDR-2 promotes utse- seam linkage through regulation of MMPs.

We agree with the reviewer and have softened our conclusions as follows (lines 356-363): “The C. elegans genome harbors six MMP genes, named zinc metalloproteinase 1-6 (zmp-1-6) (Altincicek, Fischer et al. 2010). We examined four available reporters of ZMP localization (ZMP-1::tiFP, ZMP-2::tiFP, ZMP-3::tiFP, and ZMP-4::tiFP) (Kelley, Chi et al. 201ti).Only ZMP-4 was detected at the utse-seam connection and its localization was not altered by knockdown of ddr-2 (Figure 5—figure supplement 1F). These observations suggest that DDR-2 does not promote utse-seam linkage through regulation of MMPs, although we cannot rule out roles for DDR-2 in promoting the expression or localization of ZMP-5 or ZMP-6.”

The authors show the critical period is in late L4, however, is the signaling needed later too? For example, is the linkage strengthening moderated by DDR-2 important as more eggs accumulate?

The reviewer raises an interesting question. We observed that the vesicular localization of DDR-2 sharply declined before the onset of egg-laying. By young adulthood, very few punctate structures of DDR-2 were observed in the seam cells, and none in the utse (Figure 3B). Furthermore, the frequency of utse- seam detachments in ddr-2 mutant animals peaked by the late L4 stage and did not increase after this time, suggesting DDR function is no longer required after the late L4 stage (Figure 2D). Thus, we believe that DDR-2 signaling strengthens tissue linkage only during the early formation of the utse-seam connection between the mid and late L4 stage.

We incorporated these points in the discussion (lines 477-485): “Through analysis of genetic mutations in the C. elegans receptor tyrosine kinase (RTK) DDR-2, an ortholog to the two vertebrate DDR receptors (DDR1 & DDR2) (Unsoeld, Park et al. 2013), we discovered that loss of ddr-2 results in utse-seam detachment beginning at the mid L4 stage. The frequency of detachments in ddr-2 mutant animals peaked around the late L4 stage and did not increase after this time. This correlated with the levels of DDR-2::mNG at the utse-seam connection, which peaked at the late L4 stage and then sharply declined by adulthood. Together, these findings suggest that DDR-2 promotes utse-seam attachment in the early formation of the tissue connection between the mid and late L4 stage.”

Fig. 3B is the fluorescence quantification normalized to the area?

Yes, it is. We used mean fluorescence intensity for all fluorescence quantifications to normalize for the area where the signal was measured. We added a line in Methods to emphasize this (lines 73ti-740): “We measured mean fluorescence intensity for all quantifications in order to account for linescan area.”

Fig. 4B a statistical assessment of the degree of co-localization of DDR-2::mNG and the endosomal markers might be a nice addition.

We believe the reviewer is referring to Figure 3—figure supplement 1B. We have now added the statistical assessment of the degree of co-localization of DDR-2::mNG and the endosomal markers.

We want to sincerely thank the two reviewers for their thoughtful comments and suggestions. The changes we have made in response to these comments have substantially improved the manuscript.

Author Response:

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

We appreciate the in depth review of our manuscript, and the excellent suggestions from the two reviewers. We have addressed all concerns as described in the point by point response below. We have also added all of these changes to a revised version posted to biorxiv on May 23rd 2023 (BIORXIV/2023/536585).

Reviewer #1 (Recommendations For The Authors):

It is sometimes difficult to connect the rationalizations behind the transitions between NB7 binding interaction, the compare/contrast of p84 and p101 effectors, and the synergy with phosphorylation. More explanation of the rationalizations behind these transitions in the Results would be helpful.

We agree that the manuscript would benefit from better transitions between the sections. We have added a new paragraph in the final section describing the nanobody structure before the helical domain phosphorylation that fully describes the rationale for how both inform on the critical role of helical domain dynamics in kinase activity. This paragraph is shown below.

‘The interface of NB7 with p110_g _is distant from both the putative membrane binding surface, as well as the catalytic machinery of the kinase domain. To further understand how this nanobody could so potently inhibit PI3K activity we examined any other potential modulators of PI3K activity localised in this region. There are two regulatory phosphorylation sites in the helical (Walser et al., 2013) and kinase domain (Perino et al., 2011) localised at the NB7 interface. This is intriguing as helical domain phosphorylation is activating, and kinase domain phosphorylation is inhibitory. This suggested a critical role in the regulation of p110_g _is the dynamics of this kinase-helical interface. To fully define the role of NB7 in altering the dynamics of the helical domain we needed to study other modulators of helical domain dynamics.’

The Methods section would benefit from careful copy editing for clarity and consistency.

We have gone through the methods section and edited for clarity and consistency throughout.

There's a minor ambiguity throughout when referring to the phosphorylation of S594/S595. Although close inspection makes it clear that this refers to the monophosphorylation of either site S594 or S595, there are several references to "S594/S595" that could be interpreted as phosphorylation of both residues.

We agree that this was ambiguous in the original text. We have added an explicit statement describing this as a single phosphorylation event.

‘The modification at this site results in a single phosphorylation event , but due to CID MS/MS fragmentation we cannot determine which site is modified, and will be described as S594/S595 throughout the manuscript.’

In Figure 2B, the authors show the cryo-EM density map and the structural model based on this map. It would be helpful to also include an image of the structural model fit in the density map to allow readers to evaluate the quality of the map and model. The 2F panel provides an important view of this fit, but CD3 models are difficult to discern.

We agree that this would help interpret model quality. We have added a new supplemental figure showing the fit of both p110 and NB7 into this Cryo EM density (see new Fig S2).

Paragraph starting at line 258: The shift to monitoring ATPase activity is confusing here. ATPase activity indicates production of ADP + phosphate (rather than ADP + PIP3). However, an explanation is provided that states that measuring ADP production serves as a surrogate for measuring PIP3 production. The apparent absence of membrane PIP2 substrate in Figure 4E (left) suggests that there is a true ATPase background activity in this kinase. If so, does the increase in ADP production in Fig. 4E reflect the inclusion of PIP2 substrate, increased background ATPase activity, or both?

We agree that this was worded confusingly in the original version. We have now clarified exactly what we are observing in these ATPase assays. The new paragraph is appended below

‘To further explore the potential role of phosphorylation in mediating p110g activity, we examined the kinase activity of p110g under two conditions: basal ATP turnover, and with PIP2 containing lipid membranes. The experiments in the absence of PIP2 measure turnover of ATP into ADP and phosphate, and is a readout of basal catalytic competency.  Experiments with PIP2, measured ATP consumed in the generation of PIP3, as well as in non-productive ATP turnover. The p110g enzyme in the absence of stimulators is very weakly active towards PIP2 substrate with only ~2 fold increased ATP turnover compared to in the absence of membranes. This is consistent with very weak membrane recruitment of p110g complexes in the absence of lipid activators (Rathinaswamy et al., 2023). PKCb-mediated phosphorylation enhanced the ATPase activity of p110g ~2-fold in both the absence and presence of membranes (Fig. 4E). This suggests that the effect of phosphorylation is to change the intrinsic catalytic efficiency of phosphorylated p110g, with limited effect on membrane binding.’

In the section "Nanobody blocks p110gamma phosphorylation," it's not entirely convincing that "the presence of NB7 showed even lower phosphorylation than p110gamma-p101." This does not seem to be subject to a significance test in Figure 5A/B. The follow-up point about "complete abrogation of phosphorylation," however, is readily apparent.

We agree that we could have been more precise with our language, as this is not a complete block of phosphorylation, it is merely a significant decrease in phosphorylation. We have removed the comparison to p110-p101, and also removed the statement about complete abrogation of phosphorylation. This is now reworded to

‘The presence of NB7 showed a significant decrease in p110g phosphorylation at both sites (Fig. 5A-B).’

Figure 1: Legend needs to include more detail to define the data. (1A) Representations of variance need to be clarified (e.g., replicates, error bar meaning). Consider "Normalized lipid kinase activity" as a y-axis label and expand on the activity measurement and normalization in the legend. (1B) How was error calculated? (1C) Mislabeled as 1B? Also, consider clarifying the first title highlighting the comparison to class IA PI3Ks. (1D) Typo: "Y647-p84/p110gamma." Also, would it not be more accurate to say "effect of nanobody NB7 on PI3K displacement..." for this experiment?

We apologise for these oversights. See details on what has been changed in Fig 1A, 1B, 1C, and 1D.

For Fig 1A we now show the data where each replicate is indicated in the graph in the absence of error bars, and have also more clearly expanded on this activity measurement in the figure legend and also stated the replicate number.

For Fig 1B we now clearly state how the error was generated.

For Fig 1C we have fixed the typo

For Fig 1D, we have fixed this typo and also changed the sub-heading as suggested.

New figure legend is below as well

Figure 1. The inhibitory nanobody NB7 binds tightly to all p110γ complexes and inhibits kinase activity, but does not prevent membrane binding

A. Cartoon schematic depicting nanobody inhibition of activation by lipidated Gβγ (1.5 µM final concentration). Lipid kinase assays show a potent inhibition of lipid kinase activity with increasing concentrations of NB7 (3-3000 nM) for the different complexes. Experiments are carried out in triplicate (n=3) with each replicate shown. The y-axis shows lipid kinase activity normalised for each complex activated by Gβγ in the absence of nanobody. Concentrations of each protein were selected to give a lipid kinase value in the detectable range of the ATPase transcreener assay. The protein concentration of p110γ (300 nM), p110γ-p84 (330 nM) and p110γ-p101 (12 nM) was different due to intrinsic differences of each complex to be activated by lipidated Gβγ, and is likely mainly dependent for the difference seen in NB7 response.

B. Association and dissociation curves for the dose response of His-NB7 binding to p110γ, p110γ-p84 and p110γ-p101 (50 – 1.9 nM) is shown. A cartoon schematic of BLI analysis of the binding of immobilized His-NB7 to p110γ is shown on the left. Dissociation constants (KD) were calculated based on a global fit to a 1:1 model for the top three concentrations and averaged with error shown. Error was calculated from the association and dissociation value (n=3) with standard deviation shown. Full details are present in the source data.

C. Association and dissociation curves for His-NB7 binding to p110γ, p110a-p85a, p110b-p85b, and p110d-p85b. Experiments were performed in duplicate with a final concentration of 50 nM of each class I PI3K complex.

D. Effect of NB7 on PI3K recruitment to supported lipid bilayers containing H-Ras(GTP) and farnesyl-Gbg as measured by Total Internal Reflection Fluorescence Microscopy (TIRF-M). DY647-p84/p110g displays rapid equilibration kinetics and is insensitive to the addition of 500 nM nanobody (black arrow, 250 sec) on supported lipid bilayers containing H-Ras(GTP) and farnesyl-Gbg.

E. Kinetics of 50 nM DY647-p84/p110g membrane recruitment appears indistinguishable in the absence and presence of nanobody. Prior to sample injection, DY647-p84/p110g was incubated for 10 minutes with 500 nM nanobody.

F. Representative TIRF-M images showing the localization of 50 nM DY647-p84/p110g visualized in the absence or presence of 500 nM nanobody (+NB7). Membrane composition for panels C-E: 93% DOPC, 5% DOPS, 2% MCC-PE, Ras(GTP) covalently attached to MCC-PE, and 200 nM farnesyl-Gbg.

Figure 2: (1A) For consistency with the rest of the paper, p110g can be updated with the Greek character. (1B) This may have been intentional to attract attention to subdomains interacting with NB7, but "colored according to the schematic" omits the purple RBD. (2F) the figure legend should specify whether p110gamma surfaces depicted are the cryo-EM density or a surface rendition of the structural model.

We agree and have fixed the p110 typo, and have also colored the schematic the same as shown in the cartoon model.

The data shown in Fig 1B is indeed the Cryo EM density and this is now clearly indicated in the legend.

Figure 3: (3B) Specifying the [M+H] as [M+2H]2+ and [M+4H]4+ would help the reader understand the delta mass for monophosphorylation here. Given the broad readership of this journal, it would be useful to define 't' and 'e' as 'theoretical' and 'experimental' in the legend. It may also help to be explicit about the meaning of the red spectra and residues in the legend. (3C-E) autocorrect typo for "(C)" and an opportunity to update "b" for Greek character beta.

We agree that clearly defining the charge state of each spectra will make it more obvious that we are dealing with a mono-phosphorylation and have made this change as suggested in the figure. We have also clearly define m/z t and m/z e in the figure legend, as well as the black and red lines, and characters. Finally we have added PKCb for all descriptions of PKC treatment in the figure, and fixed the incorrect PKC’b’ in the legend.

Figure 4: (4C) Given the common use of "ND" for other terms, it would be useful to spell out "no deuterium" or "undeuterated." (4E) the parenthetical "(concentration, 12nM to 1000nM)" could be clarified. How are the (presumably p100gamma) concentration ranges reflected in the three plotted data points per treatment? See also Figure 5E.

We agree and have redefined ND as undeuterated. We apologise for the typo in the figure legend, as the concentrations of p110 gamma were the same for both phosphorylated and non-phosphorylated, with this being a typo (all concentrations of enzyme were 1000 nM). This has been changed here and in Fig 5E.

Figure 5: (5A/B) Some clarification that we're looking at extracted ion chromatograms would be very useful in this figure legend. On a related note, the experimental details on the LC-MS methodology for this data appear to be split between two sections of the Methods: the "Phosphorylation analysis" paragraph (line 526) and the HDX-MS section. Some more explicit cross-referencing would clarify this experiment. (5E) Clarify inclusion of PIP2 membranes here.

We have clearly described that we are looking at extracted ion chromatograms in both panel A and B. We also have normalized the experimental methods in the LC-MS as these used exactly the same procedure. Finally, we now clearly describe the assays shown in Fig 5E were performed in the absence of PIP2 membranes.

Miscellaneous typos:<br /> Line 205: reference omitted for "Previous study.."

We have added this reference

Line 196: "unambiguous"

Fixed to unambiguously

Reviewer #2 (Recommendations For The Authors):

The only mistake I spotted was that on line 729 there is a reference to Fig 3C that should actually be Fig. 4C

We have changed this to the correct Fig 4C.

Author Response

eLife assessment

In this valuable study, the authors investigate the mechanism of amyloid nucleation in a cellular system using their novel ratiometric measurements and uncover interesting insights regarding the role of polyglutamine length and the sequence features of glutamine-rich regions on amyloid formation. Overall, the problem is significant and being able to assess nucleation in cells is of considerable relevance. The data, as presented and analyzed, are currently still incomplete. The specific claims would be stronger if based on in vitro measurements that avoid the intricacies of specific cellular systems and that are more suitable for assessing sequence-intrinsic properties.

We are pleased that the editors find our study valuable. We find that the reviewers’ criticisms largely arise from misunderstandings inherent to the conceptually challenging nature of the topic, rather than fundamental flaws, as we will elaborate here. We are grateful for the opportunity afforded by eLife to engage reviewers in a constructive public dialogue.

Reviewer #1 (Public Review):

The authors take on the challenge of defining the core nucleus for amyloid formation by polyglutamine tracts. This rests on the assertion that polyQ forms amyloid structures to the exclusion of all other forms of solids. Using their unique assay, deployed in yeast, the authors attempt to infer the size of the nucleus that templates amyloid formation by polyQ. Further, through a series of sequence titrations, all studied using a single type of assay, the authors converge on an assertion stating that a single polyQ molecule is the nucleus for amyloid formation, that 12-residues make up the core of the nucleus, that it takes ca. 60 Qs in a row to unmask this nucleation potential, and that polyQ amyloid formation belongs to the same universality class as self-poisoned crystallization, which is the hallmark of crystallization from polymer melts formed by large, high molecular weight synthetic polymers. Unfortunately, the authors have decided to lean in hard on their assertions without a critical assessment of whether their findings stand up to scrutiny. If their findings are truly an intrinsic property of polyQ molecules, then their findings should be reconstituted in vitro. Unfortunately, careful and rigorous experiments in vitro show that there is a threshold concentration for forming fibrillar solids. This threshold concentration depends on the flanking sequence context on temperature and on solution conditions. The existence of a threshold concentration defies the expectation of a monomer nucleus. The findings disagree with in vitro data presented by Crick et al., and ignored by the authors. Please see: https://doi.org/10.1073/pnas.1320626110. These reports present data from very different assays, the importance of which was underscored first by Regina Murphy and colleagues. The work of Crick et al., provides a detailed thermodynamic framework - see the SI Appendix. This framework dove tails with theory and simulations of Zhang and Muthukumar, which explains exactly how a system like polyQ might work (https://doi.org/10.1063/1.3050295). The picture one paints is radically different from what the authors converge upon. One is inclined to lean toward data that are gleaned using multiple methods in vitro because the test tube does not have all the confounding effects of a cellular milieu, especially when it comes to focusing on sequence-intrinsic conformational transitions of a protein. In addition to concerns about the limitations of the DAmFRET method, which based on the work of the authors in their collaborative paper by Posey et al., are being stretched to the limit, there is the real possibility that the cellular milieu, unique to the system being studied, is enabling transitions that are not necessarily intrinsic to the sequence alone. A nod in this direction is the work of Marc Diamond, which showed that having stabilized the amyloid form of Tau through coacervation, there is a large barrier that limits the loss of amyloid-like structure for Tau. There may well be something similar going on with the polyQ system. If the authors could show that their data are achievable in vitro without anything but physiological buffers one would have more confidence in a model that appears to contradict basic physical principles of how homopolymers self-assemble. Absent such additional evidence, numerous statements seem to be too strong. There are also several claims that are difficult to understand or appreciate.

Rebuttal to the perceived necessity of in vitro experiments

The overarching concern of this reviewer and reviewing editor is whether in-cell assays can inform on sequence-intrinsic properties. We understand this concern. We believe however that the relative merit of in-cell assays is largely a matter of perspective. The truly sequence-intrinsic behavior of polyQ, i.e. in a vacuum, is less informative than the “sequence-intrinsic” behaviors of interest that emerge in the presence of extraneous molecules from the appropriate biological context. In vitro experiments typically include a tiny number of these -- water, ions, and sometimes a crowding agent meant to approximate everything else. Obviously missing are the myriad quinary interactions with other proteins that collectively round out the physiological solvent. The question is what experimental context best approximates that of a living human neuron under which the pathological sequence-dependent properties of polyQ manifest. We submit that a living yeast cell comes closer to that ideal than does buffer in a test tube.

The reviewer’s statements that our findings must be validated in vitro ignores the fact -- stressed in our introduction -- that decades of in vitro work have not yet generated definitive evidence for or against any specific nucleus model. In addition to the above, one major problem concerns the large sizes of in vitro systems that obscure the effects of primary nucleation. For example, a typical in vitro experimental volume of e.g. 1.5 ml is over one billion-fold larger than the femtoliter volume of a cell. This means that any nucleation-limited kinetics of relevant amyloid formation are lost, and any alternative amyloid polymorphs that have a kinetic growth advantage -- even if they nucleate at only a fraction the rate of relevant amyloid -- will tend to dominate the system (Buell, 2017). Novel approaches are clearly needed to address these problems. We present such an approach, stretch it to the limit (as the reviewer notes) across multiple complementary experiments, and arrive at a novel finding that is fully and uniquely consistent with all of our own data as well as the collective prior literature.

That the preceding considerations are collectively essential to understand relevant amyloid behavior is evident from recent cryoEM studies showing that in vitro-generated amyloid structures generally differ from those in patients (Arseni et al., 2022; Bansal et al., 2021; Radamaker et al., 2021; Schmidt et al., 2019; Schweighauser et al., 2020; Yang et al., 2022). This is highly relevant to the present discourse because each amyloid structure is thought to emanate from a different nucleating structure. This means that in vitro experiments have broadly missed the mark in terms of the relevant thermodynamic parameters that govern disease onset and progression. Note that the rules laid out via our studies are not only consistent with structural features of polyQ amyloid in cells, but also (as described in the discussion) explain why the endogenous structure of a physiologically relevant Q zipper amyloid differs from that of polyQ.

A recent collaboration between the Morimoto and Knowles groups (Sinnige et al.) investigated the kinetics of aggregation by Q40-YFP expressed in C. elegans body wall muscle cells, using quantitative approaches that have been well established for in vitro amyloid-forming systems of the type favored by the reviewer. They calculate a reaction order of just 1.6, slightly higher than what would be expected for a monomeric nucleus but nevertheless fully consistent with our own conclusions when one accounts for the following two aspects of their approach. First, the polyQ tract in their construct is flanked by short poly-Histidine tracts on both sides. These charges very likely disfavor monomeric nucleation because all possible configurations of a four-stranded bundle position the beginning and end of the Q tract in close proximity, and Q40 is only just long enough to achieve monomeric nucleation in the absence of such destabilization. Second, the protein is fused to YFP, a weak homodimer (Landgraf et al., 2012; Snapp et al., 2003). With these two considerations, our model -- which was generated from polyQ tracts lacking flanking charges or an oligomeric fusion -- predicts that amyloid nucleation by their construct will occur more frequently as a dimer than a monomer. Indeed, their observed reaction order of 1.6 supports a predominantly dimeric nucleus. Like us and others, Sinnige et al. did not observe phase separation prior to amyloid formation. This is important because it not only argues against nucleation occurring in a condensate, it also suggests that the reaction order they calculated has not been limited by the concentration-buffering effect of phase separation.

While we agree that our conclusions rest heavily on DAmFRET data (for good reason), we do provide supporting evidence from molecular dynamics simulations, SDD-AGE, and microscopy.

To summarize, given the extreme limitations of in vitro experiments in this field, the breadth of our current study, and supporting findings from another lab using rigorous quantitative approaches, we feel that our claims are justified without in vitro data.

Rebuttal to the perceived incompatibility of monomeric nucleation with the existence of a critical concentration for amyloid

We appreciate that the concept of a monomeric nucleus can superficially appear inconsistent with the fact that crystalline solids such as polyQ amyloid have a saturating concentration, but this is only true if one neglects that polyQ amyloids are polymer crystals with intramolecular ordering. The perceived discrepancy is perhaps most easily dispelled by protein crystallography. Folded proteins form crystals. These crystals have critical concentrations, and the protein subunits within them each have intramolecular crystalline order (in the form of secondary structure). To extrapolate these familiar examples to our present finding with polyQ, one need only appreciate the now well-established phenomenon of secondary nucleation, whereby transient interactions of soluble species with the ordered species leads to their own ordering (Törnquist et al., 2018). Transience is important here because it implies that intramolecular ordering can in principle propagate even in solutions that are subsaturated with respect to bulk crystallization. This is possible in the present case because the pairing of sufficiently short beta strands (equivalent to “stems” in the polymer crystal literature) will be more stable intramolecularly than intermolecularly, due to the reduced entropic penalty of the former. Our elucidation that Q zipper ordering can occur with shorter strands intramolecularly than intermolecularly (Fig. S4C-D) demonstrates this fact. It is also evident from published descriptions of single molecule “crystals” formed in sufficiently dilute solutions of sufficiently long polymers (Hong et al., 2015; Keller, 1957; Lauritzen and Hoffman, 1960).

In suggesting that a saturating concentration for amyloid rules out monomeric nucleation, the reviewer assumes that the Q zipper-containing monomer must be stable relative to the disordered ensemble. This is not inherent to our claim and in fact opposes the definition of a nucleus. The monomeric nucleating structure need not be more stable than the disordered state, and monomers may very well be disordered at equilibrium at low concentrations. To be clear, our claim requires that the Q zipper-containing monomer is both on pathway to amyloid and less stable than all subsequent species that are on pathway to amyloid. The former requirement is supported by our extensive mutational analysis. The latter requirement is supported by our atomistic simulations showing the Q zipper-containing monomer is stabilized by dimerization (see our 2021 preprint). Hence, requisite ordering in the nucleating monomer is stabilized by intermolecular interactions. We provide in Author response image 1 an illustration to clarify what we believe to be the discrepancy between our claim and the reviewer’s interpretation.

Author response image 1.

That the rate-limiting fluctuation for a crystalline phase can occur in a monomer can also be understood as a consequence of Ostwald’s rule of stages, which describes the general tendency of supersaturated solutes, including amyloid forming proteins (Chakraborty et al., 2023), to populate metastable phases en route to more stable phases (De Yoreo, 2022; Schmelzer and Abyzov, 2017). Our findings with polyQ are consistent with a general mechanism for Ostwald’s rule wherein the relative stabilities of competing polymorphs differ with the number of subunits (De Yoreo, 2022; Navrotsky, 2004). As illustrated in Fig. 6 of Navrotsky, a polymorph that is relatively stable at small particle sizes tends to give way to a polymorph that -- while initially unstable -- becomes more stable as the particles grow. The former is analogous to our early stage Q zipper composed of two short sheets with an intramolecular interface, while the latter is analogous to the later stage Q zipper composed of longer sheets with an intermolecular interface. Subunit addition stabilizes the latter more than the former, hence the initial Q zipper that is stabilized more by intra- than intermolecular interactions will mature with growth to one that is stabilized more by intermolecular interactions.

We apologize to the Pappu group for neglecting to cite Crick et al. 2013 in the current preprint. Contrary to the reviewer’s assessment, however, we find that the conclusions of this valuable study do more to support than to refute our findings. Briefly, Crick et al. investigated the aggregation of synthetic Q30 and Q40 peptides in vitro, wherein fibrils assembled from high concentrations of peptide were demonstrated to have saturating concentrations in the low micromolar range. As explained above, this finding of a saturating concentration does not refute our results. More relevant to the present work are their findings that “oligomers” accumulated over an hours-long timespan in solutions that are subsaturated with respect to fibrils, and these oligomers themselves have (nanomolar) critical concentrations. The authors postulated that the oligomers result from liquid–liquid demixing of intrinsically disordered polyglutamine. However, phase separation by a peptide is expected to fix its concentration in both the solute and condensed phases, and, because disordered phase separation is inherently faster than amyloid formation, the postulated explanation removes the driving force for any amyloid phase with a critical solubility greater than that of the oligomers. In place of this interpretation that truly does appear to -- in the reviewer’s words -- “contradict basic physical principles of how homopolymers self-assemble”, we interpret these oligomers as evidence of our Q zipper-containing self-poisoned multimers, rounded as an inherent consequence of self-poisoning (Ungar et al., 2005), and likely akin to semicrystalline spherulites that have been observed in other polymer crystal and amyloid-forming systems (Crist and Schultz, 2016; Vetri and Foderà, 2015). That Crick et al. also observed the formation of a relatively labile amyloid phase when the reactions were started with 50 uM peptide is unsurprising in light of the aforementioned kinetic advantage that large reaction volumes can confer to labile polymorphs, and that high concentrations (in this case, orders of magnitude higher than the likely physiological concentration of polyQ (Wild et al., 2015)) can favor the formation of labile amyloid polymorphs (Ohhashi et al., 2010). Indeed, a contemporaneous study by the Wetzel group using very similar peptide constructs and polyQ lengths -- but beginning with lower concentrations -- found that the relevant saturating concentrations for amyloid lie below their limit of detection of 100 nM (Sahoo et al., 2014).

Rebuttals to other critiques

The reviewer states that we found nucleation potential to require 60 Qs in a row. Our data are collectively consistent with nucleation occurring at and above approximately 36 Qs, a point repeated in the paper. The reviewer may be referring to our statement, ”Sixty residues proved to be the optimum length to observe both the pre- and post-nucleated states of polyQ in single experiments”. The purpose of this statement is simply to describe the practical consideration that led us to use 60 Qs for the bulk of our assays. We do appreciate that the fraction of AmFRET-positive cells is very low for lengths just above the threshold, especially Q40. They are nevertheless highly significant (p = 0.004 in [PIN+] cells, one-tailed T-test), and we will modify the figure and text to clarify this.

The reviewer characterizes self-poisoning as the hallmark of crystallization from polymer melts, which would be problematic for our conclusions if self-poisoning were limited to this non-physiological context. In fact the term was first used to describe crystallization from solution (Organ et al., 1989), wherein the phenomenon is more pronounced (Ungar et al., 2005).

Reviewer #2 (Public Review):

Numerous neurodegenerative diseases are thought to be driven by the aggregation of proteins into insoluble filaments known as "amyloids". Despite decades of research, the mechanism by which proteins convert from the soluble to insoluble state is poorly understood. In particular, the initial nucleation step is has proven especially elusive to both experiments and simulation. This is because the critical nucleus is thermodynamically unstable, and therefore, occurs too infrequently to directly observe. Furthermore, after nucleation much faster processes like growth and secondary nucleation dominate the kinetics, which makes it difficult to isolate the effects of the initial nucleation event. In this work Kandola et al. attempt to surmount these obstacles using individual yeast cells as microscopic reaction vessels. The large number of cells, and their small size, provides the statistics to separate the cells into pre- and post-nucleation populations, allowing them to obtain nucleation rates under physiological conditions. By systematically introducing mutations into the amyloid-forming polyglutamine core of huntingtin protein, they deduce the probable structure of the amyloid nucleus. This work shows that, despite the complexity of the cellular environment, the seemingly random effects of mutations can be understood with a relatively simple physical model. Furthermore, their model shows how amyloid nucleation and growth differ in significant ways, which provides testable hypotheses for probing how different steps in the aggregation pathway may lead to neurotoxicity.

In this study Kandola et al. probe the nucleation barrier by observing a bimodal distribution of cells that contain aggregates; the cells containing aggregates have had a stochastic fluctuation allowing the proteins to surmount the barrier, while those without aggregates have yet to have a fluctuation of suitable size. The authors confirm this interpretation with the selective manipulation of the PIN gene, which provides an amyloid template that allows the system to skip the nucleation event.

In simple systems lacking internal degrees of freedom (i.e., colloids or rigid molecules) the nucleation barrier comes from a significant entropic cost that comes from bringing molecules together. In large aggregates this entropic cost is balanced by attractive interactions between the particles, but small clusters are unable to form the extensive network of stabilizing contacts present in the larger aggregates. Therefore, the initial steps in nucleation incur an entropic cost without compensating attractive interactions (this imbalance can be described as a surface tension). When internal degrees of freedom are present, such as the conformational states of a polypeptide chain, there is an additional contribution to the barrier coming from the loss of conformational entropy required to the adopt aggregation-prone state(s). In such systems the clustering and conformational processes do not necessarily coincide, and a major challenge studying nucleation is to separate out these two contributions to the free energy barrier. Surprisingly, Kandola et al. find that the critical nucleus occurs within a single molecule. This means that the largest contribution to the barrier comes from the conformational entropy cost of adopting the beta-sheet state. Once this state is attained, additional molecules can be recruited with a much lower free energy barrier.

There are several caveats that come with this result. First, the height of the nucleation barrier(s) comes from the relative strength of the entropic costs compared to the binding affinities. This balance determines how large a nascent nucleus must grow before it can form interactions comparable to a mature aggregate. In amyloid nuclei the first three beta strands form immature contacts consisting of either side chain or backbone contacts, whereas the fourth strand is the first that is able to form both kinds of contacts (as in a mature fibril). This study used relatively long polypeptides of 60 amino acids. This is greater than the 20-40 amino acids found in amyloid-forming molecules like ABeta or IAPP. As a result, Kandola et al.'s molecules are able to fold enough times to create four beta strands and generate mature contacts intramolecularly. The authors make the plausible claim that these intramolecular folds explain the well-known length threshold (L~35) observed in polyQ diseases. The intramolecular folds reduce the importance of clustering multiple molecules together and increase the importance of the conformational states. Similarly, manipulating the sequence or molecular concentrations will be expected to manipulate the relative magnitude of the binding affinities and the clustering entropy, which will shift the relative heights of the entropic barriers.

The reviewer correctly notes that the majority of our manipulations were conducted with 60-residue long tracts (which corresponds to disease onset in early adulthood), and this length facilitates intramolecular nucleation. However, we also analyzed a length series of polyQ spanning the pathological threshold, as well as a synthetic sequence designed explicitly to test the model nucleus structure with a tract shorter than the pathological threshold, and both experiments corroborate our findings.

The authors make an important point that the structure of the nucleus does not necessarily resemble that of the mature fibril. They find that the critical nucleus has a serpentine structure that is required by the need to form four beta strands to get the first mature contacts. However, this structure comes at a cost because residues in the hairpins cannot form strong backbone or zipper interactions. Mature fibrils offer a beta sheet template that allows incoming molecules to form mature contacts immediately. Thus, it is expected that the role of the serpentine nucleus is to template a more extended beta sheet structure that is found in mature fibrils.

A second caveat of this work is the striking homogeneity of the nucleus structure they describe. This homogeneity is likely to be somewhat illusory. Homopolymers, like polyglutamine, have a discrete translational symmetry, which implies that the hairpins needed to form multiple beta sheets can occur at many places along the sequence. The asparagine residues introduced by the authors place limitations on where the hairpins can occur, and should be expected to increase structural homogeneity. Furthermore, the authors demonstrate that polyglutamine chains close to the minimum length of ~35 will have strict limitations on where the folds must occur in order to attain the required four beta strands.

We are unsure how to interpret the above statements as a caveat. We agree that increasing sequence complexity will tend to increase homogeneity, but this is exactly the motivation of our approach. We explicitly set out to determine the minimal complexity sequence sufficient to specify the nucleating conformation, which we ultimately identified in terms of secondary and tertiary structure. We do not specify which parts of a long polyQ tract correspond to which parts of the structure, because, as the reviewer points out, they can occur at many places. Hence, depending on the length of the polyQ tract, the nucleus we describe may have any length of sequence connecting the strand elements. We do not think that the effects of N-residue placement can be interpreted as a confounding influence on hairpin position because the striking even-odd pattern we observe implicates the sides of beta strands rather than the lengths. Moreover, we observe this pattern regardless of the residue used (Gly, Ser, Ala, and His in addition to Asn).

A novel result of this work is the observation of multiple concentration regimes in the nucleation rate. Specifically, they report a plateau-like regime at intermediate regimes in which the nucleation rate is insensitive to protein concentration. The authors attribute this effect to the "self-poisoning" phenomenon observed in growth of some crystals. This is a valid comparison because the homogeneity observed in NMR and crystallography structures of mature fibrils resemble a one-dimensional crystal. Furthermore, the typical elongation rate of amyloid fibrils (on the order of one molecule per second) is many orders of magnitude slower than the molecular collision rate (by factors of 10^6 or more), implying that the search for the beta-sheet state is very slow. This slow conformational search implies the presence of deep kinetic traps that would be prone to poisoning phenomena. However, the observation of poisoning in nucleation during nucleation is striking, particularly in consideration of the expected disorder and concentration sensitivity of the nucleus. Kandola et al.'s structural model of an ordered, intramolecular nucleus explains why the internal states responsible for poisoning are relevant in nucleation.

We thank the reviewer for noting the novelty and plausibility of the self-poisoning connection. We would like to elaborate on our finding that self-poisoning inhibits nucleation (in addition to elongation), as this could prove confusing to some readers. While self-poisoning is claimed to inhibit primary nucleation in the polymer crystal literature (Ungar et al., 2005; Zhang et al., 2018), the semantics of “nucleation” in this context warrants clarification. Technically, the same structure can be considered a nucleus in one context but not in another. The Q zipper monomer, even if it is rate-limiting for amyloid formation at low concentrations (and is therefore the “nucleus”), is not necessarily rate-limiting when self-poisoned at high concentrations. Whether it comprises the nucleus in this case depends on the rates of Q zipper formation relative to subunit addition to the poisoned state. If the latter happens slower than Q zipper formation de novo, it can be said that self-poisoning inhibits nucleation, regardless of whether the Q zipper formed. We suspect this to be the mechanism by which preemptive oligomerization blocks nucleation in the case of polyQ, though other mechanisms may be possible.

To achieve these results the authors used a novel approach involving a systematic series of simple sequences. This is significant because, while individual experiments showed seemingly random behavior, the randomness resolved into clear trends with the systematic approach. These trends provided clues to build a model and guide further experiments.

Reviewer #3 (Public Review):

Kandola et al. explore the important and difficult question regarding the initiating event that triggers (nucleates) amyloid fibril growth in glutamine-rich domains. The researchers use a fluorescence technique that they developed, dAMFRET, in a yeast system where they can manipulate the expression level over several orders of magnitude, and they can control the length of the polyglutamine domain as well as the insertion of interfering non-glutamine residues. Using flow cytometry, they can interrogate each of these yeast 'reactors' to test for self-assembly, as detected by FRET.

In the introduction, the authors provide a fairly thorough yet succinct review of the relevant literature into the mechanisms of polyglutamine-mediated aggregation over the last two decades. The presentation as well as the illustrations in Figure 1A and 1B are difficult to understand, and unfortunately, there is no clear description of the experimental technique that would allow the reader to connect the hypothetical illustrations to the measurement outcomes. The authors do not explain what the FRET signal specifically indicates or what its intensity is correlated to. FRET measures distance between donor and acceptor, but can it be reliably taken as an indicator of a specific beta-sheet conformation and of amyloid? Does the signal increase with both nucleation and with elongation, and is the signal intensity the same if, e.g., there were 5 aggregates of 10 monomers each versus 50 monomeric nuclei? Is there a reason why the AmFRET signal intensity decreases at longer Q even though the number of cells with positive signal increases? Does the number of positive cells increase with time? The authors state later that 'non-amyloid containing cells lacked AmFRET altogether', but this seems to be a tautology - isn't the lack of AmFRET taken as a proof of lack of amyloid? Overall, a clearer description of the experimental method and what is actually measured (and validation of the quantitative interpretation of the FRET signal) would greatly assist the reader in understanding and interpreting the data.

We believe the difficulty in understanding the illustrations in Figure 1A and 1B is inherent to the subject. We agree that elaborating how DAmFRET works would help the reader, and will add a few sentences to this end. Beyond this, we refer the reviewer and readers to our cited prior work describing the theory and interpretation of DAmFRET. Note that the y-axes of DAmFRET plots are not raw FRET but rather “AmFRET”, a ratio of FRET to total expression level. As explained thoroughly in our cited prior work, the discontinuity of AmFRET with expression level indicates that the high AmFRET-population formed via a disorder-to-order transition. When the query protein is predicted to be intrinsically disordered, the discontinuous transition to high AmFRET invariably (among hundreds of proteins tested in prior published and unpublished work) signifies amyloid formation as corroborated by SDD-AGE and tinctorial assays.

When performed using standard flow cytometry as in the present study, every AmFRET measurement corresponds to a cell-wide average, and hence does not directly inform on the distribution of the protein between different stoichiometric species. As there is only one fluorophore per protein molecule, monomeric nuclei have no signal. DAmFRET can distinguish cells expressing monomers from stable dimers from higher order oligomers (see e.g. Venkatesan et al. 2019), and we are therefore quite confident that AmFRET values of zero correspond to cells in which a vast majority of the respective protein is not in homo-oligomeric species (i.e. is monomeric or in hetero-complexes with endogenous proteins). The exact value of AmFRET, even for species with the same stoichiometry, will depend both on the effect of their respective geometries on the proximity of mEos3.1 fluorophores, and on the fraction of protein molecules in the species. Hence, we only attempt to interpret the plateau values of AmFRET (where the fraction of protein in an assembled state approaches unity) as directly informing on structure, as we did in Fig. S3A.

We believe that AmFRET decreases with longer polyQ because the mass fraction of fluorophore decreases in the aggregate, simply because the extra polypeptide takes up volume in the aggregate.

Yes, the fraction of positive cells in a discontinuous DAmFRET plot does increase with time. However, given the more laborious data collection and derivation of nucleation kinetics in a system with ongoing translation, especially across hundreds of experiments with other variables, ours is a snapshot measurement to approximately derive the relative contributions of intra- and intermolecular fluctuations to the nucleation barrier, rather than the barrier’s magnitude.

We will revise the tautological statement by removing “non-amyloid containing”.

The authors demonstrate that their assay shows that the fraction of cells with AmFRET signal increases strongly with an increase in polyQ length, with a 'threshold around 50-60 glutamines. This roughly correlates with the Q-length dependence of disease. The experiments in which asparagine or other amino acids are inserted at variable positions in the glutamine repeat are creative and thorough, and the data along with the simulations provide compelling support for the proposed Q zipper model. The experiments shown in Figure 5 are strongly supportive of a model where formation of the beta-sheet nucleus is within a monomer. This is a potentially important result, as there are conflicting data in the literature as to whether the nucleus in polyQ is monomer.

We thank the reviewer for these comments. We wish to clarify one important point, however, concerning the correlation of our data with the pathological length threshold. As we state in the first results section, “Our data recapitulated the pathologic threshold -- Q lengths 35 and shorter lacked AmFRET, indicating a failure to aggregate or even appreciably oligomerize, while Q lengths 40 and longer did acquire AmFRET in a length and concentration-dependent manner”. Hence, most of our experiments were conducted with 60Q not because it resembles the pathological threshold, but rather because it was most convenient for DAmFRET experiments.

I did not find the argument, that their data shows the Q zipper grows in two dimensions, compelling; there are more direct experimental methods to answer this question. I was also confused by the section that Q zippers poison themselves. It would be easier for the reader to follow if the authors first presented their results without interpretation. The data seem more consistent with an argument that, at high concentrations, non-structured polyQ oligomers form which interfere with elongation into structured amyloid assemblies - but such oligomers would not be zippers.

Self-poisoning is a widely observed and heavily studied phenomenon in polymer crystal physics, though it seems not yet to have entered the lexicon of amyloid biologists. We were new to this concept before it emerged as an extremely parsimonious explanation for our results. As described in the text, two pieces of evidence exclude the alternative mechanism suggested by the reviewer -- that non-structured oligomers form and subsequently engage and inhibit the template. Specifically, 1) inhibition occurs without any detectable FRET, even at high total protein concentration, indicating the species do not form in a concentration-dependent manner that would be expected of disordered oligomers; and 2) inhibition itself has strict sequence requirements that match those of Q zippers. Hence our data collectively suggest that inhibition is a consequence of the deposition of partially ordered molecules onto the templating surface.

Although some speculation or hypothesizing is perfectly appropriate in the discussion, overall the authors stretch this beyond what can be supported by the results. A couple of examples: The conclusion that toxicity arises from 'self-poisoned polymer crystals' is not warranted, as there is no relevant data presented in this manuscript. The authors refer to findings 'that kinetically arrested aggregates emerge from the same nucleating event responsible for amyloid formation', but I cannot recall any evidence for this statement in the results section.

We restricted any mention of toxicity to the introduction and a section in the discussion that is not worded as conclusive. Nevertheless, we will soften the subheading and text of the relevant section in the discussion to more clearly indicate the speculative nature of the statements.

We stand by our statement 'that kinetically arrested aggregates emerge from the same nucleating event responsible for amyloid formation', as this follows directly from self-poisoning.

Bibliography

Arseni D, Hasegawa M, Murzin AG, Kametani F, Arai M, Yoshida M, Ryskeldi-Falcon B. 2022. Structure of pathological TDP-43 filaments from ALS with FTLD. Nature 601:139–143. doi:10.1038/s41586-021-04199-3

Bansal A, Schmidt M, Rennegarbe M, Haupt C, Liberta F, Stecher S, Puscalau-Girtu I, Biedermann A, Fändrich M. 2021. AA amyloid fibrils from diseased tissue are structurally different from in vitro formed SAA fibrils. Nat Commun 12:1013. doi:10.1038/s41467-021-21129-z

Buell AK. 2017. The Nucleation of Protein Aggregates - From Crystals to Amyloid Fibrils. Int Rev Cell Mol Biol 329:187–226. doi:10.1016/bs.ircmb.2016.08.014

Chakraborty D, Straub JE, Thirumalai D. 2023. Energy landscapes of Aβ monomers are sculpted in accordance with Ostwald’s rule of stages. Sci Adv 9:eadd6921. doi:10.1126/sciadv.add6921 Crist B, Schultz JM. 2016. Polymer spherulites: A critical review. Prog Polym Sci 56:1–63. doi:10.1016/j.progpolymsci.2015.11.006

De Yoreo JJ. 2022. Casting a bright light on Ostwald’s rule of stages. Proc Natl Acad Sci USA 119. doi:10.1073/pnas.2121661119

Hong Y, Yuan S, Li Z, Ke Y, Nozaki K, Miyoshi T. 2015. Three-Dimensional Conformation of Folded Polymers in Single Crystals. Phys Rev Lett 115:168301. doi:10.1103/PhysRevLett.115.168301

Keller A. 1957. A note on single crystals in polymers: Evidence for a folded chain configuration. Philosophical Magazine 2:1171–1175. doi:10.1080/14786435708242746

Landgraf D, Okumus B, Chien P, Baker TA, Paulsson J. 2012. Segregation of molecules at cell division reveals native protein localization. Nat Methods 9:480–482. doi:10.1038/nmeth.1955

Lauritzen JI, Hoffman JD. 1960. Theory of Formation of Polymer Crystals with Folded Chains in Dilute Solution. J Res Natl Bur Stand A Phys Chem 64A:73–102. doi:10.6028/jres.064A.007

Navrotsky A. 2004. Energetic clues to pathways to biomineralization: precursors, clusters, and nanoparticles. Proc Natl Acad Sci USA 101:12096–12101. doi:10.1073/pnas.0404778101

Ohhashi Y, Ito K, Toyama BH, Weissman JS, Tanaka M. 2010. Differences in prion strain conformations result from non-native interactions in a nucleus. Nat Chem Biol 6:225–230. doi:10.1038/nchembio.306

Organ SJ, Ungar G, Keller A. 1989. Rate minimum in solution crystallization of long paraffins. Macromolecules 22:1995–2000. doi:10.1021/ma00194a078

Radamaker L, Baur J, Huhn S, Haupt C, Hegenbart U, Schönland S, Bansal A, Schmidt M, Fändrich M. 2021. Cryo-EM reveals structural breaks in a patient-derived amyloid fibril from systemic AL amyloidosis. Nat Commun 12:875. doi:10.1038/s41467-021-21126-2

Sahoo B, Singer D, Kodali R, Zuchner T, Wetzel R. 2014. Aggregation behavior of chemically synthesized, full-length huntingtin exon1. Biochemistry 53:3897–3907. doi:10.1021/bi500300c

Schmelzer JWP, Abyzov AS. 2017. How do crystals nucleate and grow: ostwald’s rule of stages and beyond In: Šesták J, Hubík P, Mareš JJ, editors. Thermal Physics and Thermal Analysis, Hot Topics in Thermal Analysis and Calorimetry. Cham: Springer International Publishing. pp. 195–211. doi:10.1007/978-3-319-45899-1_9

Schmidt M, Wiese S, Adak V, Engler J, Agarwal S, Fritz G, Westermark P, Zacharias M, Fändrich M. 2019. Cryo-EM structure of a transthyretin-derived amyloid fibril from a patient with hereditary ATTR amyloidosis. Nat Commun 10:5008. doi:10.1038/s41467-019-13038-z

Schweighauser M, Shi Y, Tarutani A, Kametani F, Murzin AG, Ghetti B, Matsubara T, Tomita T, Ando T, Hasegawa K, Murayama S, Yoshida M, Hasegawa M, Scheres SHW, Goedert M. 2020. Structures of α-synuclein filaments from multiple system atrophy. Nature 585:464–469. doi:10.1038/s41586-020-2317-6

Snapp EL, Hegde RS, Francolini M, Lombardo F, Colombo S, Pedrazzini E, Borgese N, Lippincott-Schwartz J. 2003. Formation of stacked ER cisternae by low affinity protein interactions. J Cell Biol 163:257–269. doi:10.1083/jcb.200306020

Törnquist M, Michaels TCT, Sanagavarapu K, Yang X, Meisl G, Cohen SIA, Knowles TPJ, Linse S. 2018. Secondary nucleation in amyloid formation. Chem Commun 54:8667–8684. doi:10.1039/c8cc02204f

Ungar G, Putra EGR, de Silva DSM, Shcherbina MA, Waddon AJ. 2005. The Effect of Self-Poisoning on Crystal Morphology and Growth Rates In: Allegra G, editor. Interphases and Mesophases in Polymer Crystallization I, Advances in Polymer Science. Berlin, Heidelberg: Springer Berlin Heidelberg. pp. 45–87. doi:10.1007/b107232

Vetri V, Foderà V. 2015. The route to protein aggregate superstructures: Particulates and amyloid-like spherulites. FEBS Lett 589:2448–2463. doi:10.1016/j.febslet.2015.07.006

Wild EJ, Boggio R, Langbehn D, Robertson N, Haider S, Miller JRC, Zetterberg H, Leavitt BR, Kuhn R, Tabrizi SJ, Macdonald D, Weiss A. 2015. Quantification of mutant huntingtin protein in cerebrospinal fluid from Huntington’s disease patients. The Journal of Clinical Investigation.

Yang Y, Arseni D, Zhang W, Huang M, Lövestam S, Schweighauser M, Kotecha A, Murzin AG, Peak-Chew SY, Macdonald J, Lavenir I, Garringer HJ, Gelpi E, Newell KL, Kovacs GG, Vidal R, Ghetti B, Ryskeldi-Falcon B, Scheres SHW, Goedert M. 2022. Cryo-EM structures of amyloid-β 42 filaments from human brains. Science 375:167–172. doi:10.1126/science.abm7285

Zhang X, Zhang W, Wagener KB, Boz E, Alamo RG. 2018. Effect of Self-Poisoning on Crystallization Kinetics of Dimorphic Precision Polyethylenes with Bromine. Macromolecules 51:1386–1397. doi:10.1021/acs.macromol.7b02745