Preprint Review

This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to version 1 of the manuscript.

Summary

This paper analyzes the evolution of the KRAB-containing zinc finger protein (KZFP) family of proteins. While the reviewers were all interested in the topic, several major concerns came up during review. These include technical limitations of the methods chosen to analyze this challenging protein family (e.g., determination of orthology, selection analysis, and so on), and that new ideas, including claims about non-coding evolution and positive selection, are not convincingly supported by the analysis presented.

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

Learn more at Review Commons

Reply to the reviewers

R-We would like to thank the reviewers for their constructive feedback. We respond to all the reviewers points below. We highlighted major changes introduced to the manuscript in response to both reviewers’ comments in the attached revised version of the manuscript.

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

The work described in this manuscript by Tan and Marques aims to address if the splicing of enhancer-associated long noncoding RNAs (elncRNAs) has a direct impact on enhancer activity or just reflects their cognate's enhancer's high activity.

For this purpose, the authors started by integrating RNA-seq data for human lymphoblastoid cell lines with ENCODE enhancer annotations and ChIP-seq data for enhancer function-associated chromatin modifications to show that multi-exonic elncRNAs are more transcriptionally active than single-exonic elncRNAs and eRNAs. They then show that regions flanking elncRNA splice sites are enriched in splicing-associated sequence elements and that these are under stronger purifying selection (suggesting some functional relevance), both when compared to promoter-associated lncRNAs. They also show the concomitance of cis-disrupted splicing in elncRNAs and drops in expression in their target genes. Finally, they use causal inference analysis of joint seQTLs and joint scQTLs to investigate the causal relationship between splicing of elncRNAs and expression of putative gene targets and chromatin states at elncRNA cognate enhancers, respectively. They conclude that, in both cases, most associations are causally mediated by splicing of elncRNAs and therefore that this contributes to their enhancer activity.

This manuscript is generally well written, targets an original question and potentially sets the seeds for a new exciting line of research on transcriptional regulation, by providing some evidence for the functional relevance of the splicing of elncRNAs. However, the overlooking of some important aspects of the regulation of RNA splicing led to biases in the design of data analyses and in the interpretation of the biological implications of some results that need to be dealt with before the described work can be considered sound enough for publication.

R:We would like to thank the reviewer for taking the time to assess our manuscript and for the constructive comments.

**Essential revisions:**

1.Results, page 10, lines 21-24: The statement that "the impact of SS variants on gene splicing efficiency depends on the total number of alternative transcripts and exons" is not properly substantiated. The four examples given in Figure S3 do not illustrate any dependence or trend. If such dependence is "expected" the underlying concept must be explained, i.e. why the impact of SS variants on splicing efficiency should depend on the number of alternative transcripts and exons. The reader is unrealistically expected to be used to the chosen splicing efficiency metric to intuit its dependence on the number of exons. Moreover, it is not obvious where the dependence on the number of alternative transcripts comes from, particularly given that alternative splicing (e.g. the skipping of a neighbouring exon, if internal) is not profiled.

R-In the previous version of the manuscript, we estimated gene level changes in splicing, which includes all alternative splicing events within a gene. Therefore, the more exons an elncRNA has, the more “diluted” we expect the overall impact of SS variant on elncRNA splicing efficiency to be. After considering the reviewer’s comment, we realized that only splicing events directly impacted by the SS variant should be considered in this analysis.

In the revised version of the manuscript, we considered only alternative splicing events that include the splice donor acceptor site changed by the SS variant, and are therefore a direct consequence, of the SS variant. As suggested by the reviewer, for these splicing events, we report the fold difference in Percentage-Spliced-In (PSI) (estimated by Leafcutter (Li et al. 2018)) between samples that carry reference and alternative alleles at these SS variants. To further illustrate these changes, we now include a diagram, for each SS variant, with differential splicing information and the fold difference in PSI for each affected splicing events (Figure 3B,C, Supplementary Figure S3). In addition, the overall change across all affected splicing events is also plotted in Figure 3D and Supplementary Figure S4.

We have modified this section to account for this and the next comment from the reviewer.

“To estimate the impact of SS variant on splicing efficiency, we calculated the Percentage-Spliced-In (PSI) (Li et al. 2018) per individual and for each elncRNA splicing event involving the splice donor or acceptor site disrupted by the SS variants (Figure 3B,C, Supplementary Figure S3). PSI measures exon inclusion and considers spliced reads spanning exon junctions (Li et al. 2018). We compared the average difference in PSI, as a proxy for change in splicing efficiency, of all affected splicing events between individuals that carry the reference and alternative canonical splice donor/acceptor sites (GT-AG). Alongside decreased exon inclusion, SS variants can also promote exon skipping events (Figure 3B,C, Figure S3). Despite some increase in exon skipping, SS variants are associated with an overall decrease in splicing efficiency (Figure 3D and Supplementary Figure S4).” (Page 10).

Along these same lines, and more importantly, why haven't the authors looked at the possibility that a variant disrupting a splice site would lead to skipping of the neighbouring exon (if internal)? Given how the spliceosome operates (in terms of intron and exon recognition), wouldn't this be the most likely scenario? When calculating the splicing index, are reads spanning junctions between non-consecutive exons considered? Otherwise, not profiling alternative isoforms generated by exon skipping will necessarily bias splicing efficiency quantifications by overlooking fully efficient splicing associated with such isoforms. Similarly, how did the authors make sure that splicing changes did not bias elncRNA expression estimates? How was the effective transcript length determined for the calculation of RPKMs? The authors need to make these methodological clarifications, as well as why exon skipping was not considered as a splicing disruption with potential functional implications. Calculating the percent spliced-in (PSI) for all internal exons would be much informative.

R-Regarding the methodology, what we refer to as splicing efficiency is Percentage-Spliced-In (PSI). We calculated PSI for all, including alternative, splicing events. We now make this clearer throughout the manuscript and in the figure axis/legends.

As detailed in the methods section, to minimize the impact of alternative splicing on gene expression estimates, we quantified expression at the gene, and not at the transcript, level using HTSeq across all annotated exons. This approach allows us to assess elncRNA and target gene expression while masking differences in alternative transcript abundance, which are not relevant in the context of this analysis.

As suggested by the reviewer, instead of considering PSI of all possible splicing events of the gene, in the revised version of the manuscript, we considered only splicing events that are directly impacted by the SS variant. This change does not impact our conclusions, but certainly provides a better understanding of how SS variants impact splicing and we would like to thank the reviewer for raising this point. As predicted by the reviewer and as expected given how the spliceosome operates, exon skipping is a frequent outcome of SS variants. However, the increase in exon skipping is not sufficient to compensate for the decrease in the inclusion of these exons, which is directly impacted by the SS variants. This is demonstrated by the lower overall splicing efficiency for each elncRNAs in individuals that carry SS variants that disrupt canonical splice/donor acceptor sites (Figure 3D and Supplementary Figure S4).

3.All results in panels 3B-F are presented as fold differences. It is actually not clear what those differences refer to. For instance, the grey boxes are the distributions of the fold differences in splicing index / expression between individuals carrying reference alleles and what?

R-The boxplots represent the distribution of the fold difference in PSI or expression for each individual relative to the median PSI or expression in individuals with the reference genotype. As expected, the distribution of log fold difference in either PSI or expression for individuals carrying the reference allele is centered at 0.

We have clarified this in the methods section and figure legends.

4.It is expectable that most joint seQTLs result from variants directly impacting splicing in cis. As the quantification of splicing is noisier than that of expression, a stronger effect is required for the detection of an sQTL than an eQTL. In other words, joint seQTLs are essentially sQTLs. This illustrated by the example in Figure 4A, with the SNP in an intronic region of the elncRNA being associated with strong differences in splicing and tiny (R-We agree with the reviewer that the quantification of splicing is noisier than that of expression. However, and in contrast with the reviewer’s hypothesis, higher “noise” in splicing quantification compared to expression led to weaker associations between splicing and seQTLs, as illustrated in the figure below. This is in line with splicing being measured with higher error rate, which would ultimately lead to smaller detectable sQTL effect than what they would be with perfect measurements. This also demonstrates that since a priori, eQTLs association are stronger, if a bias exists in the causal inference analysis, it should favour detection of non-causal associations. Therefore, our approach is not biased in detecting causal seQTLs.

We agree with the reviewer that this potential bias may be a concern to readers and should be addressed. We have added this analysis to the text (Supplementary Figure S7E) and explained why the causal inference testing approach is not biased in detecting causal seQTLs.

“To assess whether this approach was biased towards the detection of causal seQTLs we compared the slope and adjusted p-value of the associations between the variant and splicing or expression for all causal seQTLs. As illustrated in Supplementary Figure S7E this analysis revealed there is no evidence that stronger sQTLs would favour causal model predictions.” (Page 16).

5.It is not totally clear what message the authors intend to convey with the result of panel 4D. Are they talking about the relative position of the variant to the elncRNA transcript or the target transcript? If the former, shouldn't the known synergy between transcription and 5´ end splicing reflect on elncRNA expression? If the latter, it is not obvious how the result connects to the mentioned synergy.

R:In Figure 4D, we show the relative position, within a transcript, of the exonic splicing junction which is associated with causal seQTLs. The enrichment in associations with splicing junctions located at 5’ end of elncRNAs is consistent with the synergy between 5’ end splicing and transcription. We clarify this in the text:

“Importantly, 90% of seQTL associations that support elncRNA splicing as a mediator of target expression are associated with splicing junctions located at the 5´ end of the transcript, which is consistent with the known synergy between transcription and 5´ end splicing (Furger et al. 2002; Damgaard et al. 2008)(Figure 4D).” (Page 17).

**Proofreading edits:**

R:We would like to thank the reviewer for identifying all the typos listed below. We have corrected them in the revised version of the manuscript.

6.Introduction, page 3, line 13: double "in".

7.Figure S2A, leftmost panel X-axis label: "intrno" instead of "intron".

8.Results, page 10, line 30: remove "of".

9.It is 5´ and 3´(prime) not 5' and 3' (apostrophe).

**Other suggestions:**

10.Violin plots (with included boxplots) would more comprehensively convey the differences in distributions than the chosen notched boxplots.

R-We thank the reviewer for this suggestion. Although we appreciate the added information a violin plot can provide, this also renders, in our opinion, their interpretation less intuitive. Because boxplots are simpler and easier to interpret, after consideration, we decided to continue using these to represent the distribution of the data.

Reviewer #1 (Significance (Required)):

It is hard for me to assess the significance of this work (beyond some evidence for the potential functional relevance of the splicing of elncRNAs) until the aforementioned concerns are addressed but it is of potential interest to the broad RNA research community.

I am a computational biologist with experience in the analysis of high-throughput transcriptomic data and a focus on transcriptional and alternative splicing regulation.

========================================================================

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

This manuscript addresses an interesting question - whether the splicing of transcriptional enhancer-associated RNAs influences their transcriptional enhancement activity. The analyses appear carefully done, using appropriate datasets and statistical methods. The authors find, for example, that marks of active chromatin are enriched near spliced elncRNAs, that splicing-related motifs of elncRNAs are under selective constraint, and that splicing of elncRNAs is associated with higher elncRNA expression and to very slightly higher expression of target genes.

R:We thank the reviewer for the constructive feedback on our manuscript. We have extended our analysis to address the reviewers concerns that we detail in our response to the comments below.

However, I did not find the main results convincing of the main conclusion for the following reasons:

1.The most direct evidence is shown in Figure 3, where SNPs that occur in 3' splice sites of elncRNA introns are explored, and it is shown that variants predicted to disrupt splicing of elncRNA introns are associated with reduced expression of target but not non-target genes. But the fold difference in expression of target genes is extremely small - a few percent - and is actually less than the fold difference in expression of the elncRNAs themselves (which appears closer to 10%), raising the question of whether elncRNA expression rather than splicing may be more important for activity. Furthermore, the entire analysis has an anecdotal quality, being based on only 4 splice-disrupting elncRNA variants. I did not find the figure at all convincing of the conclusion the authors draw from it.

R:We agree with the reviewer that our analysis is limited by the available genotyping data that is restricted to common genetic variants. Our evolutionary constraint analysis (Figure 2) indicates that variants that disrupt elncRNA splicing are depleted by natural selection and so we expected to identify a relatively small number of elncRNAs (n=4) suitable for this analysis. Despite the anticipated challenges in identifying elncRNA splice site mutations, we nevertheless believe this unbiased natural mutational analysis is analogous to experimentally disrupting splice sites of these 4 elncRNA candidates.

Regarding the strength of impact of splicing on target expression: in the absence of a comparable experiment, we could not anticipate the magnitude of the effect. We acknowledge that previous studies, which sought to completely remove splicing by either deleting all elncRNA introns (Yin et al. 2015) or terminating transcription after its 1st exon (Engreitz et al. 2016), were both associated with significantly stronger impact on elincRNA splicing and target expression than what we report here. The analysis we present here involves single nucleotide polymorphisms and so it is not surprising to have resulted in more moderate impact on overall splicing. Furthermore, whether the differences in the impact on target expression between this and previous analysis is the result of stronger effect of complete removal of splicing or a consequence of the genetic changes introduced remains unclear. The small yet consistent decrease in target expression we observed, even with minimal changes in splicing of an unbiased set of 4 candidates, is in our opinion strong evidence that modulation in elncRNA splicing is sufficient to impact, albeit moderately, target expression.

Importantly, we replicated the impact of decreased splicing on target expression of the 4 elncRNA candidates using 89 samples of Yoruba (YRI) population from the Geuvadis dataset (Supplementary Figure S5). The robustness of the mutational study consistently supports the physiologically relevant effect of elncRNA splicing on cognate enhancer function.

As pointed out by the reviewer, elncRNA SS variants led to stronger impact on the expression of the elncRNAs compared to that of their targets (Figure 3F,H and Supplementary Figure S4), which suggests that target expression regulation is likely a consequence of changes in elncRNA expression as a result of changes in its splicing. This is described as our working model in the discussion section of the manuscript.

2.Figure 4 uses a causal inference approach and involves larger datasets. While causal inference can be a useful tool to identify candidate causal relationships, it does not prove causality, which still requires some sort of experimental perturbation. Thus, I found these results suggestive but still not satisfying to justify, e.g., the title of the paper or claims made in the abstract. As in Figure 3, the specific example shown in Fig. 4A again shows a relatively tiny effect on target gene expression, which again appears to be a few percent at most.

R:For the reasons explained above, we had no expectation that the effect size of the association between elncRNA splicing and target expression would be high. It is nevertheless key that these associations are robust, which would provide reliable support for our hypothesis. To assess this, we used 2 independent datasets to replicate elncRNA target associations with sQTL variants associated with elncRNA splicing: 1) 147 LCL samples from GTEx and 2) 31,684 blood samples from eQTLgen. Using these datasets, we replicated the association between sQTL and target expression for targets of up to 77% of elncRNAs. As expected, replicated associations have significantly higher effect size in both datasets (Supplementary Figure S9) and 1.2 times more associations can be replicated in the eQTLgen blood samples with a larger cohort size. LCL-specific effect of elncRNA splicing likely explains why not all associations are replicated in these blood samples. We report these analyses in the manuscript (Supplementary Figure S9).

We agree with the reviewer that the causal inference analysis is only suggestive per se. However, we would argue that conclusions of the present manuscript do not rely on this analysis alone, but instead on the combined evidence of several experiments, including the natural mutational analysis that is analogous to the experiment the reviewer proposes.

Considering the reviewers concern, we realized that previous version of Figure 4A did not reflect the average strength of the association between seQTL variant and target expression (median=0.319, ranging from 0.16 to 0.81, Rebuttal Figure 1). For this reason, we replaced the previous illustration by a more representative example (Figure 4A).

The text illustrating reproducibility of our results in GTEx and eQTLgen have been added to Page 17 of the manuscript.

“We used two independent datasets to assess the robustness of elncRNA target association with sQTL variants we predict to be associated with the splicing of these elncRNAs in LCLs. Using a smaller cohort of LCLs (n=147 (GTEx Consortium 2013)), we found a significant association in the same direction between sQTL and target expression for targets of 70% of elncRNAs (45% of variants). A larger fraction of associations (77% of elncRNAs and 52% of variants) could be replicated in a larger cohort of blood samples (n=31,684 (Võsa et al. 2018)). The difference in size between these two cohorts is likely to explain the difference in replication rate. The association between elncRNA splicing variants and target expression that were replicated have significantly higher effect size relative to non-replicated associations (Supplementary Figure S9). Furthermore, LCL-specific effect also likely explains why not all associations can be replicated in the large blood cohort.” (Page 17).

3.Figure 2 shows that splicing-related signals are under selective constraint in spliced elncrRNAs, which is convincing. But this does not prove that splicing of elncRNAs is directly related to enhancer activity. It is equally plausible that elncRNA expression directly impacts enhancer activity and that elncRNA splicing is conserved because it boost elncRNA expression, for example.

R:The reviewer is right and the sentence “If splicing of elncRNAs is important for enhancer function, …” does not faithfully describe the conclusions that can be drawn from the analysis reported in Figure 2. This portion of the text now reads: “If splicing of elncRNAs is functionally relevant, one would expect selection to have prevented the accumulation of deleterious mutations in their splicing-associated motifs during evolution” (Page 8). We would like to thank the reviewer for pointing this out.

Other points:

4.Are the ChiP profiles in Figures 1A-E significantly different from each other in a statistical sense? Probably yes, but a specific test should be done.

R:We now added boxplots representing the distribution of read density centered at transcript promoters. Statistical difference in the distribution is also tested. We show this in the revised Figure 1A-E and Supplementary Figure S1B-C.

5.This sentence (p. 10) was hard to follow and should be clarified: "As expected, the impact of SS variants on gene splicing efficiency depends on the total number of alternative transcripts and exons and ranges from 11% to 24% for elncRNA with 6 to 2 number of exons, respectively (Supplementary Figure S3)."

R:We had previously estimated the average amount of change in splicing for all alternative splicing events at each elncRNA candidate. To calculate this, we considered the difference in Percentage-Spliced-In (PSI) for all splicing events and divided this by the total number of considered events. Given that only a subset of events is affected by a splice site variant, the more exons an elncRNA has, the more alternative splicing events are likely to occur and the lower the average impact of a SS variant on overall gene splicing efficiency is expected to be. Following a comment from reviewer 1 (comment 1), we now only consider splicing events directly disrupted by the SS variant. We agree this sentence was not clear and have removed it from the manuscript.

6.Related to point 5 above, Supplementary Figure 3 is somewhat confusing because two splicing change and three expression change plots are shown for each locus, without labels of what each one is, or explanation of what the red and green colors mean.

R:We apologize for the confusion and thank the reviewer for pointing this out. In the figure, we plot the fold difference in elncRNA splicing, target gene splicing, target gene expression, non-target gene expression, and elncRNA expression. elncRNA features are plotted in red and target gene features are plotted in green. We have added labels to clarify the relevant plots (Figure 3D-H, Supplementary Figure S4,5).

**Minor points:**

1.Top of p. 16: "90% of those that support elncRNA splicing as a mediator 3 of target expression are located at the 5' end of the transcript, which is 4 consistent with the known synergy between transcription and 5' end splicing" - a reference is needed

R:We thank the reviewer for pointing this out and we have now added the appropriate reference.

“Importantly, 90% of seQTL associations that support elncRNA splicing as a mediator of target expression are associated with splicing junctions located at the 5´ end of the transcript, which is consistent with the known synergy between transcription and 5´ end splicing (Furger et al. 2002; Damgaard et al. 2008)(Figure 4D).” (Page 17)

2.Figure 5B,C y-axes indicate "fold difference", but scales include negative numbers, which is confusing. Probably should redo the analysis showing log of fold difference.

R:We thank the reviewer for the suggestion. Since the fold difference in Percentage-Spliced-In (PSI) used to estimate the amount of splicing at each exon junction can be of both positive and negative values, we now plot the log modulus transformation (John and Draper, 1980) of the data, which is equivalent to a log transformation while preserving the sign of the data. The analysis has been redone for Figure 3D-H, 5B,C, and Supplementary Figure S4, S5. This change does not impact the conclusions and makes the interpretation of the results more intuitive.

3.p. 20 Describes U1 snRNP as "a protein essential for the 4 recognition of nascent RNA 5' splice site and assembly of the spliceosome". U1 is a large RNA-protein complex, not a protein.

R:We thank the reviewer for pointing this out and this has now been corrected.

“Chromatin-bound lncRNAs have been recently shown to be enriched in U1 small nuclear ribonucleoprotein (snRNP) RNA-protein complex, a protein essential for the recognition of nascent RNA 5´ splice site and assembly of the spliceosome (Yin et al. 2020).” (Page 22)

4.Typo: p. 11, l. 2 missing word (genes): "expression levels of other nearby was unaffected"

R:This has been corrected.

Reviewer #2 (Significance (Required)):

The question addressed is very interesting, given recent work the significance of transcription from enhancers, and work addressing functional relationships between splicing and expression. The work is suggestive of effects of enhancer splicing on expression but I did not find it fully convincing as the effects observed are extremely small, and other explanations are not ruled out, as discussed above.

Prior literature has shown that many active enhancers are transcribed, that enhancer transcription can preced and is positively correlated with target gene expression, and work from both Ulitsky and from the authors indicates that splicing of enhancer-associated lncRNAs is positively correlated with enhancer activity. A variety of studies have also shown that splicing of protein-coding genes generally has a strong positive effect on gene expression. Here, the authors attempt to go further and show that splicing of enhancers causes increased transcriptional enhancement of target genes. A variety of public genotype, expression, chromatin and other types of data are analyzed to address this question. The statistical genetics crowd may find the work of interest, but molecular biologists will not be convinced of the conclusions. My expertise is in computational biology, genomics and RNA biology.

Engreitz JM, Haines JE, Perez EM, Munson G, Chen J, Kane M, McDonel PE, Guttman M, Lander ES. 2016. Local regulation of gene expression by lncRNA promoters, transcription and splicing. Nature 539: 452-455.

John, J., & Draper, N. (1980). An Alternative Family of Transformations. Journal of the Royal Statistical Society. Series C (Applied Statistics), 29(2), 190-197. doi:10.2307/2986305

Li YI, Knowles DA, Humphrey J, Barbeira AN, Dickinson SP, Im HK, Pritchard JK. 2018. Annotation-free quantification of RNA splicing using LeafCutter. Nature genetics 50: 151-158.

Yin Y, Yan P, Lu J, Song G, Zhu Y, Li Z, Zhao Y, Shen B, Huang X, Zhu H et al. 2015. Opposing Roles for the lncRNA Haunt and Its Genomic Locus in Regulating HOXA Gene Activation during Embryonic Stem Cell Differentiation. Cell stem cell 16: 504-516.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

This manuscript addresses an interesting question - whether the splicing of transcriptional enhancer-associated RNAs influences their transcriptional enhancement activity. The analyses appear carefully done, using appropriate datasets and statistical methods. The authors find, for example, that marks of active chromatin are enriched near spliced elncRNAs, that splicing-related motifs of elncRNAs are under selective constraint, and that splicing of elncRNAs is associated with higher elncRNA expression and to very slightly higher expression of target genes.

However, I did not find the main results convincing of the main conclusion for the following reasons:

1.The most direct evidence is shown in Figure 3, where SNPs that occur in 3' splice sites of elncRNA introns are explored, and it is shown that variants predicted to disrupt splicing of elncRNA introns are associated with reduced expression of target but not non-target genes. But the fold difference in expression of target genes is extremely small - a few percent - and is actually less than the fold difference in expression of the elncRNAs themselves (which appears closer to 10%), raising the question of whether elncRNA expression rather than splicing may be more important for activity. Furthermore, the entire analysis has an anecdotal quality, being based on only 4 splice-disrupting elncRNA variants. I did not find the figure at all convincing of the conclusion the authors draw from it.

2.Figure 4 uses a causal inference approach and involves larger datasets. While causal inference can be a useful tool to identify candidate causal relationships, it does not prove causality, which still requires some sort of experimental perturbation. Thus, I found these results suggestive but still not satisfying to justify, e.g., the title of the paper or claims made in the abstract. As in Figure 3, the specific example shown in Fig. 4A again shows a relatively tiny effect on target gene expression, which again appears to be a few percent at most.

3.Figure 2 shows that splicing-related signals are under selective constraint in spliced elncrRNAs, which is convincing. But this does not prove that splicing of elncRNAs is directly related to enhancer activity. It is equally plausible that elncRNA expression directly impacts enhancer activity and that elncRNA splicing is conserved because it boost elncRNA expression, for example.

Other points:

4.Are the ChiP profiles in Figures 1A-E significantly different from each other in a statistical sense? Probably yes, but a specific test should be done.

5.This sentence (p. 10) was hard to follow and should be clarified: "As expected, the impact of SS variants on gene splicing efficiency depends on the total number of alternative transcripts and exons and ranges from 11% to 24% for elncRNA with 6 to 2 number of exons, respectively (Supplementary Figure S3)."

6.Related to point 5 above, Supplementary Figure 3 is somewhat confusing because two splicing change and three expression change plots are shown for each locus, without labels of what each one is, or explanation of what the red and green colors mean.

Minor points:

1.Top of p. 16: "90% of those that support elncRNA splicing as a mediator 3 of target expression are located at the 5' end of the transcript, which is 4 consistent with the known synergy between transcription and 5' end splicing" - a reference is needed

2.Figure 5B,C y-axes indicate "fold difference", but scales include negative numbers, which is confusing. Probably should redo the analysis showing log of fold difference.

3.p. 20 Describes U1 snRNP as "a protein essential for the 4 recognition of nascent RNA 5' splice site and assembly of the spliceosome". U1 is a large RNA-protein complex, not a protein.

4.Typo: p. 11, l. 2 missing word (genes): "expression levels of other nearby was unaffected"

Significance

The question addressed is very interesting, given recent work the significance of transcription from enhancers, and work addressing functional relationships between splicing and expression. The work is suggestive of effects of enhancer splicing on expression but I did not find it fully convincing as the effects observed are extremely small, and other explanations are not ruled out, as discussed above.

Prior literature has shown that many active enhancers are transcribed, that enhancer transcription can preced and is positively correlated with target gene expression, and work from both Ulitsky and from the authors indicates that splicing of enhancer-associated lncRNAs is positively correlated with enhancer activity. A variety of studies have also shown that splicing of protein-coding genes generally has a strong positive effect on gene expression. Here, the authors attempt to go further and show that splicing of enhancers causes increased transcriptional enhancement of target genes. A variety of public genotype, expression, chromatin and other types of data are analyzed to address this question. The statistical genetics crowd may find the work of interest, but molecular biologists will not be convinced of the conclusions. My expertise is in computational biology, genomics and RNA biology.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

The work described in this manuscript by Tan and Marques aims to address if the splicing of enhancer-associated long noncoding RNAs (elncRNAs) has a direct impact on enhancer activity or just reflects their cognate's enhancer's high activity.

For this purpose, the authors started by integrating RNA-seq data for human lymphoblastoid cell lines with ENCODE enhancer annotations and ChIP-seq data for enhancer function-associated chromatin modifications to show that multi-exonic elncRNAs are more transcriptionally active than single-exonic elncRNAs and eRNAs. They then show that regions flanking elncRNA splice sites are enriched in splicing-associated sequence elements and that these are under stronger purifying selection (suggesting some functional relevance), both when compared to promoter-associated lncRNAs. They also show the concomitance of cis-disrupted splicing in elncRNAs and drops in expression in their target genes. Finally, they use causal inference analysis of joint seQTLs and joint scQTLs to investigate the causal relationship between splicing of elncRNAs and expression of putative gene targets and chromatin states at elncRNA cognate enhancers, respectively. They conclude that, in both cases, most associations are causally mediated by splicing of elncRNAs and therefore that this contributes to their enhancer activity.

This manuscript is generally well written, targets an original question and potentially sets the seeds for a new exciting line of research on transcriptional regulation, by providing some evidence for the functional relevance of the splicing of elncRNAs. However, the overlooking of some important aspects of the regulation of RNA splicing led to biases in the design of data analyses and in the interpretation of the biological implications of some results that need to be dealt with before the described work can be considered sound enough for publication.

Essential revisions:

1.Results, page 10, lines 21-24: The statement that "the impact of SS variants on gene splicing efficiency depends on the total number of alternative transcripts and exons" is not properly substantiated. The four examples given in Figure S3 do not illustrate any dependence or trend. If such dependence is "expected" the underlying concept must be explained, i.e. why the impact of SS variants on splicing efficiency should depend on the number of alternative transcripts and exons. The reader is unrealistically expected to be used to the chosen splicing efficiency metric to intuit its dependence on the number of exons. Moreover, it is not obvious where the dependence on the number of alternative transcripts comes from, particularly given that alternative splicing (e.g. the skipping of a neighbouring exon, if internal) is not profiled.

2.Along these same lines, and more importantly, why haven't the authors looked at the possibility that a variant disrupting a splice site would lead to skipping of the neighbouring exon (if internal)? Given how the spliceosome operates (in terms of intron and exon recognition), wouldn't this be the most likely scenario? When calculating the splicing index, are reads spanning junctions between non-consecutive exons considered? Otherwise, not profiling alternative isoforms generated by exon skipping will necessarily bias splicing efficiency quantifications by overlooking fully efficient splicing associated with such isoforms. Similarly, how did the authors make sure that splicing changes did not bias elncRNA expression estimates? How was the effective transcript length determined for the calculation of RPKMs? The authors need to make these methodological clarifications, as well as why exon skipping was not considered as a splicing disruption with potential functional implications. Calculating the percent spliced-in (PSI) for all internal exons would be much informative.

3.All results in panels 3B-F are presented as fold differences. It is actually not clear what those differences refer to. For instance, the grey boxes are the distributions of the fold differences in splicing index / expression between individuals carrying reference alleles and what?

4.It is expectable that most joint seQTLs result from variants directly impacting splicing in cis. As the quantification of splicing is noisier than that of expression, a stronger effect is required for the detection of an sQTL than an eQTL. In other words, joint seQTLs are essentially sQTLs. This illustrated by the example in Figure 4A, with the SNP in an intronic region of the elncRNA being associated with strong differences in splicing and tiny (<1%) and barely significant differences in expression. Moreover, current knowledge and reported evidence strongly suggests that cis regulation of splicing is essentially "local", i.e. directly involves the processed sequences and not the interference of neighbouring RNAs. Similarly, to my knowledge there is no evidence suggesting a trend for genes encoding splicing factors being associated to the same eQTL variants as those of their target RNAs. I would therefore predict that most joint seQTLs result from variants within the elncRNA loci directly impacting their splicing. If this is the case, causal inference analysis will naturally be biased towards more strongly linking the variants with elncRNA splicing and thereby suggesting its causal role. The same rationale applies to scQTLs. The authors need to control for that potential bias in their analyses or explain why there is no bias.

5.It is not totally clear what message the authors intend to convey with the result of panel 4D. Are they talking about the relative position of the variant to the elncRNA transcript or the target transcript? If the former, shouldn't the known synergy between transcription and 5´ end splicing reflect on elncRNA expression? If the latter, it is not obvious how the result connects to the mentioned synergy.

Proofreading edits:

6.Introduction, page 3, line 13: double "in".

7.Figure S2A, leftmost panel X-axis label: "intrno" instead of "intron".

8.Results, page 10, line 30: remove "of".

9.It is 5´ and 3´(prime) not 5' and 3' (apostrophe).

Other suggestions:

10.Violin plots (with included boxplots) would more comprehensively convey the differences in distributions than the chosen notched boxplots.

Significance

It is hard for me to assess the significance of this work (beyond some evidence for the potential functional relevance of the splicing of elncRNAs) until the aforementioned concerns are addressed but it is of potential interest to the broad RNA research community.

I am a computational biologist with experience in the analysis of high-throughput transcriptomic data and a focus on transcriptional and alternative splicing regulation.

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

Learn more at Review Commons

Reply to the reviewers

Response to reviewer comment for manuscript RC-2020-00207

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

**Major Comments:**

The authors of the paper start the paper with just one protein narrowed down ie. HRG. The rest of the paper uses affinity based proteomics, antibody validation, GWAS and survival analysis to validate this target and support their claim that HRG is an age associate protein linked to mortality and certain clinical outcomes. How did the authors conclude that HRG was the only target to explore further in this paper? What methods or analysis was done for this? What were the other proteins if any that showed up in these studies?

We appreciate this comment which reveals unclear explanation how the protein was chosen for further analysis. The protein profile obtained using HPA045005 was the top and single hit out of 7258 protein profiles using a threshold of adjusted P-value below 0.01. In other words, only the profile of HRG was statistically significantly associated with age in the screening sample set (N = 156). The results of all protein profiles were attached as Supporting Table 1. Phrases about the alpha level were added to the text to make the threshold clear. Because antibody validation of these exploratory studies requires enormous efforts and time, we could not choose a more liberal and inclusive threshold.

For mortality outcome, it is not clear which class of disease is most strongly associated with increased risk of mortality from elevated HRG levels. If cause-specific mortality exists among the cohorts, could authors provide a more exact breakdown of the type of associated mortality by a disease class?

We thank the reviewer for the question and have now added cause-specific data in the manuscript. Using cause of death data, mortality risk by diseases in circulatory system were compared with the risk by neoplasm and others. ElevatedHPA045005-HRG profiles were found to associate with mortality risk by diseases of the circulatory system (HR = 1.46 per SD, P = 2.80 × 10‑4, ICD-10 code I00-I99). It was larger than the risk by malignant neoplasms (HR = 1.28 per SD, P = 1.73 × 10‑2, ICD-10 code C00-C97). We chose big categories as ICD-10 codes "I" and "C" because the number of events was too small to get enough power in the survival analysis.

Page 4 Section 3 (Results)-

The authors say "We found consistent age-associated trends with HPA045005 across all eight replication sets (Supporting Figure 3)". On examining the supporting figure we noticed that the slope for the set with the largest number of subjects (Set 3 with ~3000 people) is visually negligibly positive (showing weakest age associated trends with HPA045005). Some comments from the authors on why they think the largest data set showed the weakest association.

The plot for each cohort (in Supporting Figure 3) had different ranges in the y-axes. To make those plots comparable, the ranges in the y-axes of the different panels in the figure were modified to be the same for all cohorts. In the new version of the plot, it is easier to notice that there in fact is an increasing trend of the profiles in set 3. As we briefly discussed in Discussion, weaker age-association of the sample set may be due to the set was near to a random sample of population in the age range. Set 1, however, had over-representation of older people by selecting equal number of people in every age-intervals.

From Figure 2 C in the main manuscript one concludes that for HPA045005, binding for CC individuals is ~ 2 times higher than TT individuals. Is it possible the age association showing up for HPA045005 is primarily a function of changing/increase in allele frequency as a function of age?

The authors could consider adding a clarifying plot of Age vs Allele frequency or adding an interaction term of Age and Allele Frequency in the regression and survival analysis to address this question.

As suggested, we now added a test of age association, and average age was compared by genotype. The result was added in Supporting Table 3. The heterozygote (CT) group has slightly higher average age without statistical significance (ANOVA P = 0.096).

It is interesting that the signals were significant with the HPA045005 antibody but not with the BSI037 antibody. This is in spite of the fact that the GWAS for BSI0137 signals had an even stronger hit to the same locus. Can the authors please comment on why the signals from HPA045005 and BSI0137 were not highly correlated with one another and why the better antibody could not replicate the survival analysis results?

We thank the reviewer for the comments. We believe that our text about our findings were not clear enough, though it is a primary finding. We modified the main text to easily distinguish the HPA045005-derived profiles that were influenced by the 204th amino-acid of HRG protein, from the BSI0137-derived profiles influenced by the 493the amino-acid. The signals from those two antibodies were likely obtained by capturing different parts of HRG, which are schematically illustrated in Figure 2D. What we found is that only one binder's profiles, not the other's, had predictive power for mortality risk within about 8.5 years. That suggests some age-dependent changes around the 204th residue of HRG reflected biological aging rather than whole protein level. To make our finding clearer, the two binders were compared in Table 2.

**Minor Comments:**

Figure 1: The authors description of the figure could use more clarification. "For each sample set, the estimated effect from the linear regression model.." estimated effect of what on what? On reading the main text one concludes it is the effect of age on HPA045005. This needs to be clarified in the label.

We agree with the reviewer and have added these words.

Figure 3: The X axis for the Kaplan Meir survival curve is labelled as Age. Survival is usually time to event and time is usually the follow up time. Further clarification for the choice of this label might be helpful.

We clarified the choice of the time scale in the figure legend with a reference, where it was further discussed (Thiébaut & Bénichou, 2004). We chose age as the time scale, seeing age is the strongest risk factor for all-cause mortality, as the suggestion in the reference. We attempted to use follow-up time as the time scale with age adjustment before, which gave us almost the same results but violated the proportionality assumption of COX models.

Figure 3: it would be good to include a table with the number of individuals at risk at the bottom of the plot at defined time intervals. The figure currently compares the bottom and top quartiles of HRP for visual assessment of mortality risk, it would also be informative to include middle quantiles.

The figure was updated accordingly. The risk table was included and the results of the middle group were presented.

Supporting Table 5: The note at the bottom of this table states "standardized HRG values by linear regression and scaling." What does standardization by linear regression mean?

A sentence that explains the standardization was added in the footnote of the table.

Supporting Table 5: It would be useful to understand that HRG carries additional risk beyond known Age and known clinical biomarkers listed in Table 2 (APOA1, APOB, TC, TG, Glucose, LDL). Could authors include a multivariate CoxPH regression with just Age? and with Age + clinical covariates?

The impact of those clinical variables on survival models was examined and the results were added to Supporting Table 6 (which was Table S5). It turned out that the addition of those variables barely changed the results of the model for the HRG profile affected by 202th amino-acid.

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

**Summary**

The manuscript by Hong et al. describes the identification and validation of histidine-rich glycoprotein (HRG) as a marker of chronological age and all-cause mortality. HRG was determined using proteomics of serum and plasma samples in 9 different cohorts (total sample size ~4,100). The association with mortality was tested in the largest available cohort (TwinGene), comprising ~3,000 samples. The association with mortality seems to be stronger in women in comparison to men and could not be explained by CRP or diabetes-related traits. The HRG levels determined using an alternative antibody, BSI0137, did not show any association with mortality, indicating that the effect on mortality is likely isoform-dependent. The performed analyses seem to be statistically solid. However, the association with mortality still needs to be replicated in independent studies and the HRG measurement does not yet seem to be ready for standardized high-throughput measurement, which is necessary to make it usable as biomarker.

**Major comments**

• Although the authors have convincingly identified HRG to be associated with chronological age and mortality, it will require quite some additional work (including replication of the observed association with mortality in independent cohorts, testing the predictive ability, and making the measurement standardized and high-throughput) to prove its use as potential biomarker. At the moment, this is not at all discussed in the manuscript. Moreover, there have been some recent large-scale studies that identified biomarkers at the metabolic level that are not at all mentioned by the authors. The authors only refer once to the recent proteomic study by Lehallier in the Introduction, but do not at all discuss their findings in relation to this paper. Last but not least, HRG has already been associated with mortality in a previous study (https://www.ncbi.nlm.nih.gov/pubmed/29303798), but there is no mention of this anywhere in the manuscript. Hence, I think it would be good if the authors perform a thorough literature search to place their findings into context and rewrite their Discussion accordingly.

We appreciate the reviewer's comments on the limitation of our paper. We are aware of the requirement of further investigation on HPA045005-HRG profiles as a biomarker to confirm it with independent cohorts. Instead, we supported our findings with a set of confirmatory analyses; we validated and annotated age-associated profile applying GWAS, sandwich assays, peptide arrays and mass spectrometry. Comparing two antibody profiles, we narrowed down to age-associated region within the protein HRG. The approach and finding, we believe, is novel.

We added some discussion about recent large-scale proteomic studies such as Tanaka et al, 2018 and Lehallier et al, 2019. Unexpectedly, HRG was found not measured in those studies despite of the protein is one of the abundant proteins in blood (Poon et al, 2011). It may reflect challenges in assay development and missing piece in those large studies. The papers lack further investigation for molecular targets, which is common in proteomic papers, and makes it difficult to compare between studies and technologies. In that sense, our approach is different from other proteomic studies, because we invested time and efforts to investigate the molecular target.

We are though thankful for the introduction of the suggested HRG publication, which we did not know about. We concluded that there are substantial differences in the subjects and suggested functions for the protein. Kuroda et al. found HRG as a biomarker for sepsis of ICU patients, while our study was done on the general population. They were measuring HRG protein level, whereas we found one particular region in HRG as a biomarker for all-cause mortality. Hence, we briefly discussed the reference in the paragraph about general information about HRG.

• The authors need to add a Supplementary Table showing the association of all their 7,258 HPA antibodies with chronological age. Although I trust the authors, I can currently not tell if it is indeed correct that only one antibody was significantly associated with age in set 1.

We agree with the reviewer. The table of association test results of all 7258 antibody profiles was attached to the paper as Supporting Table 1. We were also surprised that only one passed a conventional P-value threshold 0.01 after Bonferroni correction. It might be due to the low number of samples in the sample set 1 (N=156), compared to the number of antibodies or tests.

• According to description in the Supporting Information, several samples in set 3-5 were overlapping with set 1 (45 in total). These samples should be removed from datasets 3-5 to make sure that there are no overlapping samples in the meta-analysis. However, I am not sure if the authors have actually done this. For the GWAS the overlapping samples from set 3 could still be included, given that set 1 is not involved in that. The authors could actually use these 45 overlapping samples to provide additional details about the reproducibility of HPA045005 between different measurements, for example by showing a correlation plot.

We agree with the reviewer. Those 45 overlapping samples were excluded in the meta-analysis. As the reviewer's comment, only the data of sample set 3 was used for the GWAS.

We also appreciate the comment regarding reproducibility and acknowledge that there are limitations to the technical performance of our exploratory SBA method. The procedure is tailored to handle large number of antibodies and profile 384 sample in the analysis plates. This setup allowed us to process relatively large number of samples per batch but it might be affected by batch effects. In our study set 3, there were 2999 samples randomized and analyzed in 8 different 384-well plates. The 44 overlapping samples between sets 1 and 3 were added to one of these 8 plates. This resulted in 1-11 samples to be analyzed on the same plate, hence, comparing these 44 with previous assays might be influenced if not dominated by plate effects. We went back to the initial data set generated during 2011/2012 and compared the first data with replicated assays using the same freeze-thawed samples. For HPA045005 we found the data to correlate by r=0.45. The next analyses of these 44 samples were conducted during 2015 using different sample aliquots and preparations as well as different SBAs. The correlation to previous assays was r

• When looking at the effect of the rs9898-stratified analysis (Table S2) it seems that there only is an effect in the presence of the C-allele. Have the authors considered the presence of a potential recessive effect of this variant when looking at mortality?

Average age of the individuals of each genotype of the SNP was compared and added into Supporting Table 3 (which was Table S2). No significant difference between the genotypes was found. As the reviewer noted, the mortality association of the HRG profiles affected by 204th amino-acid in the TT genotype group of rs9898 was milder and did not reach statistical significance. We believe that it is due to substantially smaller sample size and number of deaths in the genetic group. To clarify the difference in numbers, those numbers were added into the Supporting Table 3 (which was Table S2).

• The authors need to discuss in more detail the implications of the difference between the two HRG antibodies in their association with mortality, for example in light of the use of HRG levels as a potential biomarker (i.e. how should one deal with the fact the way the levels are measured influences the outcome).

We appreciated this valuable comment, which clearly reveals that our claim was not explained sufficiently. We modified the main text to distinguish those two antibody profiles more clearly. We also added Figure 2D and changed the structure of Table 2 to highlight the difference between the two antibody profiles.

• Why did the authors put part of their Discussion in the Supplement? This is not common practice. They should either move it to the manuscript or remove it completely.

We moved the discussion in the supplement to main text as the reviewer's suggestion.

Reviewer #2 (Significance (Required)):

The manuscript is clearly written and the analyses seem to be solid. However, although the findings described in the manuscript are interesting for the ageing field, they only provide a small step in the process of the usability of HRG as biomarker, i.e. many validation and follow-up studies will be necessary to prove its value. There have been some recent biomarker studies that have been much more advanced in this respect, which limits the novelty of this manuscript. I therefore feel that this manuscript may be best suitable for a medium-impact ageing-specific journal. My fields of expertise are ageing, genetics, and molecular epidemiology. Given my limited expertise when it comes to proteomics, I was not able to provide detailed comments on the methodology concerning this part.

We thank the reviewer for the honest and constructive assessment of our work and agree with the suggestion to transfer this work to a medium-impact journal covering aspects of ageing research.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary

The manuscript by Hong et al. describes the identification and validation of histidine-rich glycoprotein (HRG) as a marker of chronological age and all-cause mortality. HRG was determined using proteomics of serum and plasma samples in 9 different cohorts (total sample size ~4,100). The association with mortality was tested in the largest available cohort (TwinGene), comprising ~3,000 samples. The association with mortality seems to be stronger in women in comparison to men and could not be explained by CRP or diabetes-related traits. The HRG levels determined using an alternative antibody, BSI0137, did not show any association with mortality, indicating that the effect on mortality is likely isoform-dependent. The performed analyses seem to be statistically solid. However, the association with mortality still needs to be replicated in independent studies and the HRG measurement does not yet seem to be ready for standardized high-throughput measurement, which is necessary to make it usable as biomarker.

Major comments

• Although the authors have convincingly identified HRG to be associated with chronological age and mortality, it will require quite some additional work (including replication of the observed association with mortality in independent cohorts, testing the predictive ability, and making the measurement standardized and high-throughput) to prove its use as potential biomarker. At the moment, this is not at all discussed in the manuscript. Moreover, there have been some recent large-scale studies that identified biomarkers at the metabolic level that are not at all mentioned by the authors. The authors only refer once to the recent proteomic study by Lehallier in the Introduction, but do not at all discuss their findings in relation to this paper. Last but not least, HRG has already been associated with mortality in a previous study (https://www.ncbi.nlm.nih.gov/pubmed/29303798), but there is no mention of this anywhere in the manuscript. Hence, I think it would be good if the authors perform a thorough literature search to place their findings into context and rewrite their Discussion accordingly.

• The authors need to add a Supplementary Table showing the association of all their 7,258 HPA antibodies with chronological age. Although I trust the authors, I can currently not tell if it is indeed correct that only one antibody was significantly associated with age in set 1.

• According to description in the Supporting Information, several samples in set 3-5 were overlapping with set 1 (45 in total). These samples should be removed from datasets 3-5 to make sure that there are no overlapping samples in the meta-analysis. However, I am not sure if the authors have actually done this. For the GWAS the overlapping samples from set 3 could still be included, given that set 1 is not involved in that. The authors could actually use these 45 overlapping samples to provide additional details about the reproducibility of HPA045005 between different measurements, for example by showing a correlation plot.

Minor comments

• When looking at the effect of the rs9898-stratified analysis (Table S2) it seems that there only is an effect in the presence of the C-allele. Have the authors considered the presence of a potential recessive effect of this variant when looking at mortality?

• The authors need to discuss in more detail the implications of the difference between the two HRG antibodies in their association with mortality, for example in light of the use of HRG levels as a potential biomarker (i.e. how should one deal with the fact the way the levels are measured influences the outcome).

• Why did the authors put part of their Discussion in the Supplement? This is not common practice. They should either move it to the manuscript or remove it completely.

Significance

The manuscript is clearly written and the analyses seem to be solid. However, although the findings described in the manuscript are interesting for the ageing field, they only provide a small step in the process of the usability of HRG as biomarker, i.e. many validation and follow-up studies will be necessary to prove its value. There have been some recent biomarker studies that have been much more advanced in this respect, which limits the novelty of this manuscript. I therefore feel that this manuscript may be best suitable for a medium-impact ageing-specific journal.

My fields of expertise are ageing, genetics, and molecular epidemiology. Given my limited expertise when it comes to proteomics, I was not able to provide detailed comments on the methodology concerning this part.

Joris Deelen

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

The paper applied affinity based proteomics and antibody validation to choose and validate histidine-rich glycoprotein (HRG) as a protein/target of interest. Survival analysis techniques were used to show associations between this protein and certain biomarkers, age and all cause mortality.<br> These results and findings were used to conclude that HRG may serve as a molecular indicator of age and mortality risk.

Major Comments:

The authors of the paper start the paper with just one protein narrowed down ie. HRG. The rest of the paper uses affinity based proteomics, antibody validation, GWAS and survival analysis to validate this target and support their claim that HRG is an age associate protein linked to mortality and certain clinical outcomes. How did the authors conclude that HRG was the only target to explore further in this paper? What methods or analysis was done for this? What were the other proteins if any that showed up in these studies?

For mortality outcome, it is not clear which class of disease is most strongly associated with increased risk of mortality from elevated HRG levels. If cause-specific mortality exists among the cohorts, could authors provide a more exact breakdown of the type of associated mortality by a disease class?

Page 4 Section 3 (Results)-

The authors say "We found consistent age-associated trends with HPA045005 across all eight replication sets (Supporting Figure 3)". On examining the supporting figure we noticed that the slope for the set with the largest number of subjects (Set 3 with ~3000 people) is visually negligibly positive (showing weakest age associated trends with HPA045005). Some comments from the authors on why they think the largest data set showed the weakest association.

From Figure 2 C in the main manuscript one concludes that for HPA045005, binding for CC individuals is ~ 2 times higher than TT individuals. Is it possible the age association showing up for HPA045005 is primarily a function of changing/increase in allele frequency as a function of age? The authors could consider adding a clarifying plot of Age vs Allele frequency or adding an interaction term of Age and Allele Frequency in the regression and survival analysis to address this question.

It is interesting that the signals were significant with the HPA045005 antibody but not with the BSI037 antibody. This is in spite of the fact that the GWAS for BSI0137 signals had an even stronger hit to the same locus. Can the authors please comment on why the signals from HPA045005 and BSI0137 were not highly correlated with one another and why the better antibody could not replicate the survival analysis results?

Minor Comments:

Figure 1: The authors description of the figure could use more clarification. "For each sample set, the estimated effect from the linear regression model.." estimated effect of what on what? On reading the main text one concludes it is the effect of age on HPA045005. This needs to be clarified in the label.

Figure 3: The X axis for the Kaplan Meir survival curve is labelled as Age. Survival is usually time to event and time is usually the follow up time. Further clarification for the choice of this label might be helpful.

Figure 3: it would be good to include a table with the number of individuals at risk at the bottom of the plot at defined time intervals. The figure currently compares the bottom and top quartiles of HRP for visual assessment of mortality risk, it would also be informative to include middle quantiles.

Supporting Table 5: The note at the bottom of this table states "standardized HRG values by linear regression and scaling." What does standardization by linear regression mean?

Supporting Table 5: It would be useful to understand that HRG carries additional risk beyond known Age and known clinical biomarkers listed in Table 2 (APOA1, APOB, TC, TG, Glucose, LDL). Could authors include a multivariate CoxPH regression with just Age? and with Age + clinical covariates?

Significance

The authors have identified a new biomarker for aging and mortality. Understanding the mechanism and pathways involved in HRG homeostasis and how aging causes dysregulation of this HRG could be a topic for further research. Overall, this pathway provides an opportunity of a new molecular target for aging-based drugs and research.

This article should be of interest to researchers interested in the biology of aging and for researchers developing drugs to slow down the process of aging. In addition, it should be of interest to researchers studying the HRG as a biomarker (for example, in sepsis (https://ccforum.biomedcentral.com/articles/10.1186/s13054-018-2127-5, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3437790).

This paper was reviewed by 3 co-reviewers, a senior principal investigator with extensive bioinformatics, metabolomics/proteomics, epidemiological experience, a highly experienced computational biologist with a record of developing and applying methods in bioinformatics and computational biophysics and lastly an computational biologist with a background in applied mathematics and statistical analysis. All three scientists are interested in aging research and understanding how human physiology and biomarkers in specific, change as a function of age.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary

The manuscript describes an extensive and detailed investigation into the structure and function(s) of MORN domains. It has to be acknowledged that, despite the considerable amount of work reported, the conclusions are rather limited. From a technical viewpoint, the experiments have been appropriately executed and, generally, I concur with the conclusions drawn. However, the manuscript is over-long: in general, I would recommend concentrating on positive conclusions which can be drawn from the data and avoid excessive speculation or inference (some examples given below).

Major Comments

There are three general- perhaps rather obvious- points to make. First, there is no particular reason to think that conservation of structure necessarily indicates conservation of a particular function. There seems to be an implicit assumption that MORN domains are associated with a specific, well-defined biological function. Given their diversity, are there particular reasons to think that this is the case? Second, a strategy which examines the properties of just the recombinant MORN domains in vitro, removed from the context of the whole protein (eg junctophilin) or- importantly- its interacting partners in vivo, has obvious limitations. Frequently a reductionist approach is successful; however, in this case, MORN domains appear to be less tractable to that kind of approach. For all the in vitro binding and structural experiments presented, there is always a concern that the absence of other parts of the relevant MORN-containing protein or its partners could explain failure or inconsistency of in vitro biological activity measurements. Third, the possibility that MORN domains might mediate interactions with other proteins seems to be given little consideration, in spite of the Li et al (2019) paper. An experimental strategy which looked for binding partners (eg by pulldown assay) might have provided more insight.

Minor Comments

1. In the abstract and elsewhere the authors refer to a possible function of MORN domains as 'dimerisation and oligomerisation devices' (line 53). What is the evidence that dimer formation is important for function in vivo?
2. Did the authors attempt to co-crystallize TbMORN1(7-15) with PI(4,5)P2?
3. Fig 2C: did the authors also estimate binding stoichiometry as well as the equilibrium binding constants for these data? This should be determined by fitting a single binding site model to the data. Other methods (eg ITC) can probably determine this with more accuracy. The value of stoichiometry is sometimes forgotten in such binding measurements- is one ligand bound per monomer or dimer, for example?
4. Lines 674-678 I found it hard to work out whether these constructs harbour the natural C-terminal sequence without truncation or addition of an affinity tag. I think the answer is 'yes' but it was difficult working this out from the details in M&M.
5. Lines 688-694 The PISA interface analysis is useful here in distinguishing crystal contacts from those which persist in solution. The discussion of the results is unclear, however, on this critical point: were the dimer interfaces the only contacts which were significant in the various crystal forms?
6. Lines 754-763 This paragraph seems rather speculative and is a good example where the text could be cut down.
7. Line 765-788 This section is also rather overdone: such observations are only useful if they are subsequently tested by recording dimer conformation for a representative selection of MORN dimers from different species.
8. Lines 800-801 I don't think this statement is strictly correct. The SAXS data show that PfMORN1(7-15) adopts an extended conformation, with no evidence of the 'V' shaped structure. Related to that point, from what I could glean from the SAXS Methods section, all solution conditions for these experiments were conducted without Zn2+? If some dimer interfaces require Zn2+, should it not be included?

Significance

There is certainly value in establishing that MORN domains do not, in vitro, appear to bind to lipid vesicles, and to define their lipid binding capability (although it is rather complex). The crystal structures and SAXS data extend the rather limited structural data on MORN domains. Despite the effort involved, conclusions about likely functions of MORN domains in vivo are rather limited. My expertise lies in X-ray crystallography and protein biochemistry.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

MORN (membrane occupation and recognition nexus) repeat proteins are found in prokaryotes and eukaryotes. They feature characteristic repeats in their primary sequence, have been assumed to play a role in lipid binding, but remain poorly characterized on the functional and structural level. This manuscript tries to address both these questions and is organized in major parts. In the first part the authors characterize a putative role of MORN repeat proteins in lipid binding and membrane association. In the second part, the authors use X-ray crystallography to establish the structure of MORN repeat proteins and to investigate the dimerization.

As a cleverly chosen point of departure, they focus their study particularly on MORN1 from Trypanosoma brucei (TbMORN1), which is composed solely on MORN repeats. The structures of MORN repeats (from several species) in part two provide interesting insights into their mode of homotypic interactions and their role as dimerization or oligomerization devices. The lipid binding and membrane association of MORN proteins in the first part remains somewhat confusing and unclear, despite the use of a whole battery of techniques. It is questionably, why the authors invest so many figures and words to inform the reader on negative results. The authors suggest that MORN proteins can bind to lipids via their hydrophobic acyl chains- which is 'very hard to imagine under physiological conditions unless TbMORN1 is a lipid carrier and not a membrane-binding proteins.' Unfortunately, a role as lipid carrier has not been rigorously tested. In this sense the first part remains somewhat immature and incoherent. Furthermore, they suggest based on the lack-of-evidence that MORN proteins do not bind membranes in vivo and in vitro.

The main issue of this manuscript is, in my view, the way the data were presented.The manuscript is generally well-written, but much too long. The structural work is important and concise. The first part, however, reports in five separate figures on a lack of membrane binding by a MORN protein and its ability to bind individual lipids. The physiologically relevance of this lipid binding is questionable as acknowledged by the authors. Even though I find it important that the membrane/lipid binding ability of MORN proteins is rigorously tested, I would highly recommend to separate the current manuscript in two independent stories. Alternatively, I would recommend to reduce the first part into a single figure and to remove the most artifactual assays. In the current form, the first part and the second part of the manuscript remain somewhat detached from each other. The characterization of the lipid binding/membrane binding properties has a number of substantial weaknesses (e.g. use of quite different, non-physiological buffers for membrane binding assays; use of deletion mutants for the binding assays, which do not show the full potential of oligomerization). This which makes it hard to read and confuses the reader. Even though I have no reason to doubt the conclusions by the authors, I do not think that all necessary caution has been invested to rule out other possibilities.

In summary, even though the technical quality of the individual performed assays is high, there are some conceptual issues that make it hard to make a strong case based on a collection of individual, clear datasets. Even though I find the structures of the MORN proteins important, timely, and interesting, I would not recommend this study for publication in its current form. The manuscript would be more fun to read if both of the parts would be shortened substantially and more focused. While I agree that most evidence provided on lipid/membrane binding of TbMORN1 argue against a direct role of MORN proteins in membrane binding, I feel that the experimental approach is not coherent enough. See a few major points of criticism below.

Major Points:

1) The authors decide to characterize the membrane binding of a MORN repeat protein using a deletion variant that lacks the N-terminal repeat. However, in Figure 1B they show that the N-terminal repeat is important for the formation of higher-order oligomers. While I fully understand that the presence of the most N-terminal repeat does hamper the structural work, I find it problematic to remove it for the lipid/membrane-binding assays. The formation of higher oligomeric species beyond the dimer, may be important for membrane binding/recruitment (avidity effects).

2) (Related to point 1) I do not understand the choice of the buffers used for some of the assays. The use of pH 8.5 and NaCl concentrations of 200 mM are non-physiological. For CD spectroscopy, a high ionic strength was obtained by the use of 200 mM NaF. If a high ionic strength is required to prevent the formation of higher oligomers of MORN, it raises the question if the formation of higher oligomers (under physiological conditions) may also contribute to their function. It is unclear, in which buffer the fluorescence anisotropy measurements were performed. The sucrose-loaded vesicles were hydrated in a 20 mM HEPES pH 7.4, 0.3 M Sucrose. The composition of the buffer after the addition of MORN proteins is not clear. Despite the use of an impressive array of techniques, this first part of the manuscript remains somewhat immature and incoherent. Due to the use of constructs that have not the full ability to oligomerize (point 1) and due to the inconsistent use of experimental conditions, it is hard to draw firm conclusions from this first part. For example: In Figure 2E TbMORN(2-15) does show some concentration-dependent binding, which -however- is interpreted as background binding. What are the results using this assay (or better: a liposome floatation assay) when using full-length TbMORN(1-15) in a more physiological buffer? The statement that MORN proteins bind to lipids, but not to liposomes/membranes is -in my view- not sufficiently addressed to make a strong case.

3) The physiological relevance of lipid binding to MORN proteins remains obscure (as also acknowledged by the authors). Does the binding of PE lipids to the MORN protein have a physiological role? Does the binding of fluorescent PI(4,5)P2 point to a physiological role of MORN proteins?

4) In light of recent data from the Chris Stefan lab (PMID: 31402097) a co-incidence detection of PI(4,5)P2, PS, and cholesterol seems possible. Can the authors address this possibility? Furthermore, the role of Ca2+ signaling / Ca2+ ions has not been addressed. In light of the important role of Ca2+ for the recognition of PI(4,5)P2 (PMID: 28177616), this point should be addressed.

5) For characterizing the binding of lipids to MORN proteins, the authors use non-physiological fluorescent and short-chain lipid analogues at concentrations, which are unlikely to occur for endogenous PIPs in the cytosol of cells. Why choosing such an artificial system? Why introducing this system at length, if other -less artifact-prone- assays are available? I would recommend to not feature this assay as prominently as it was in the current study.

6) How would PE find its way to the lipid binding region in MORN? Would it diffuse to the MORN protein via the aqueous phase or would the MORN protein pickup PE form membranes up collision? The authors should address this point, by separating the lipid-depleted MORN protein from donor-vesicles containing PE by a dialysis membrane. If PE would not find its way to the lipid binding site of MORN, this would imply that MORN protein can extract lipids only upon colliding with the membrane. What is the stoichiometry of PE to MORN?

Despite my critique raised above, I agree with the authors that the data point in the direction that MORN proteins (or at least TbMORN1) does not directly bind to membranes. Their data, however, would still be consistent with a role as lipid transfer protein and a recruitment of MORN proteins to the membrane by other proteins. Have the authors performed any additional experiments in this direction? Also, the potential role of palmitoylation is only mentioned in the discussion (page 22), while palmitoylation would provide a simple means for membrane recruitment.

Minor Points:

Figure 1B: The authors should provide information on the void volume of the column.

Page 17, line 696-701: The authors point out that the C2 crystal form is stabilized by two disulfide bridges. The authors should comment on the physiological relevance of these disulfide bridges.

Page 18, line 734-740: The authors should provide data on the potential role of Zn2+ on MORN function in a physiological context. The section describing that the dimer is stabilized by Zn2+ ions (pages 18 and 19) lacks a discussion if Zn2+ are functionally relevant. There is only a beautiful sequence analysis and a discussion of the conservation of the Zn2+ coordinating residues. Can the authors perform Zn2+ titrations and SEC-MALS experiments (or alternatives such as SAXS) to show that Zn2+ indeed affects the oligomeric state of only the PfMORN, but not the other MORN proteins that form alternative dimers?

Significance

A putative role of MORN proteins in membrane and lipid binding is addressed. The view the MORN proteins bind directly to membranes is challenged. Structures of dimeric MORN proteins provide important insight into the modes of dimerization.

There is a recent structure of MORN proteins (which is referenced by the authors), but I feel that additional structural work is important and justified. The work on membrane vs. lipid binding is important, but not sufficiently addressed in the current manuscript.

I would recommend to separate the study in two parts. The audience is likely to confused (or bored) by the lengthy discussion on whether or not MORN proteins bind lipids and or membrane or not.

I am not an expert in structural biology, but have a fair understanding of structural biology. I have worked on lipid binding proteins and have a very good understanding of lipid/membrane-binding assays.

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Reviewer #3

Vuorio and colleagues combine atomic resolution molecular dynamics simulations and NMR experiments to probe how glycosylation can bias binding of hyaluronan to one of several binding sites/modes on the CD44 hyaluronan binding domain. The results are of interest specifically to the field of CD44 biophysics and more generally to the broad field of glycosylation-dependent protein-ligand binding. The manuscript is clearly written, and the combination of data from computational and experimental methodologies is convincing. I especially commend the authors on the thorough molecular dynamics work, wherein they ran multiple simulations at microsecond timescale and tried different force fields to minimize the likelihood of their findings being an artifact of a particular force field.

Reviewer #2

This manuscript is focused on understanding how N-linked glycosylation regulates the binding of the (very large) polysaccharide hyaluronan (HA) to its major cell surface receptor CD44, a question relevant, for example to the role of CD44 in mediating leukocyte migration in inflammation. The paper concludes that multiple binding sites for HA exist and that their occupancy is determined by the nature of the glycosylation, a suggestion first made by Teriete et al. (2004). The work is based on atomistic simulations with different glycan compositions and NMR spectroscopy on a non-glycosylated CD44 HA-binding domain (HABD) expressed in E. coli. While the question being researched is interesting and of biological relevance, there are flaws in the work.

The paper describes how the well-established HA-binding site on CD44 (determined by a co-crystal structure; Banerji et al., 2007) is blocked by N-linked glycosylation (principally at N25 with a contribution from glycans at N100 and N110) and how certain glycans favour binding at a completely distinct binding site that lies perpendicular to the canonical 'crystallographic' binding site. This alternative 'upright' binding site, which has been proposed previously by the authors (Vuorio et al., 2017), needs further supporting experimental data.

Firstly, unlike the 'crystallographic' binding site that forms an open-ended shallow groove on the surface of the protein allowing polymeric HA to bind (and multivalent interactions to take place), the 'upright' binding site is closed at one end and can thus only accommodate the reducing end of the polysaccharide (as apparent from Appendix 1 Figure 1). Its configuration means that it would be impossible for this mode of binding to allow multivalent interactions with polymeric HA. This is a major problem since biologically relevant CD44-HA interactions are multivalent where a single HA polymer interacts with a large number of CD44 molecules (e.g. see Wolny et al., 2010 J. Biol. Chem. 285, 30170-30180). So even if this binding site existed, an interaction between a single CD44 molecule on the cell surface with the reducing terminus of an HA polymer would be exceptionally weak.

Secondly the NMR experiments performed in this study, purporting to provide evidence for multiple modes of binding, are problematic. Why weren't differentially glycosylated proteins used, i.e. where individual sites were mutated (e.g. +/- N25); this would have allowed comparisons of the glycosylation patterns hypothesised (based on the computer simulations) to favour the 'crystallographic' versus 'upright' modes. Furthermore, previous NMR studies have shown that the binding of HA to CD44 causes a considerable number of chemical shift changes due to the induction of a large conformational change in the protein (Teriete et al., 2004; Banerji et al., 2007), making it very difficult to identify amino acids directly involved in HA binding based on the NMR data. Moreover, this conformational change has been fully characterised for mouse CD44 with structures available in the absence and presence of HA (Banerji et al., 2007); this information should have been used to inform the interpretation of the shift mapping. In fact, the way in which the shift mapping data are interpreted is simplistic and doesn't fully take account of the reasons that NMR spectra can exhibit different exchange regimes.

Reviewer #1

The authors use MD simulations and NMR to study the cell surface adhesion receptor CD44 with the purpose of understanding the binding of carbohydrate polymer, hyaluronan (HA). In particular, this study focuses on the effects of N-glycosylation of the CD44 glycoprotein on potential HA binding. The authors previously proposed two lower affinity HA binding modes as alternatives to the primary mode seen in the crystal structure of the HA binding domain of CD44, driven by different arginine interactions, but overlapping with glycosylation sites that will affect HA binding. This study suggests that, because the canonical site appears blocked by glycans attached to the surface, HA would instead likely bind to an alternate parallel site with lower affinity, thus changing receptor affinity. The authors do not study HA binding to the glycosylated form directly, but undertake simulations of bound glycans to draw their conclusion. They do, however, place HA near the non-glycosylated CD44 in simulations, although it is not clear that MD sampling has been designed to provide unbiased observations of HA binding, or how the simulations help explain the NMR experiments.

The data rely on libraries of MD simulation, which are substantial, with several replicas of a microsecond each. But what have these simulations really proved with reliability? Figure 2a shows that, while glycans stay roughly where they started, they are dynamic and cover much of the canonical HA binding site, which may be the case. From this the authors imply that the crystallographic site is significantly obstructed, the lower-affinity upright mode remains most accessible, and that the level of occlusion of the main site depends on the degree of glycosylation and size of the oligosaccharides. However, a full simulation of HA binding to this glycosylated surface was not attempted. It would have been good to see the glycans actually block unbiased simulation of canonical binding to the crystallographic site on long timescales (not being dislodged), but allow alternative binding to the parallel site, without initial placement there.

HA was, however, added to the non-glycosylated CD44-HABD surface in simulations, but no clear data is shown to illustrate the extent of sampling, convergence and reproducibility, beyond some statistical analysis of contacts. It seems a total of 30 microseconds of the non-glycosylated protein with 2 or 3 nearby HA placed was run, leading to contacts. But how well did these 30 simulations sample HA movement and relative binding to sites, if at all? Figure 4 suggests that the HA stay where they have been put. As the MD is the dominant source of data for the paper, the extent of sampling and how the outcomes depend on the initial placement of molecules requires proof. Was any sampling of HA movement, such as between canonical and alternative parallel conformations seen in MD?

The NMR is suggested to show that a short HA hexamer can bind to non-glycosylated CD44-HABD simultaneously in several modes at distinct binding sites, and that MD "correlates" with this. But is this MD biased by initial choices of where and how many HAs are placed, given HA movement is likely not well sampled?

No MD seems to have been used to examine the blocking or lack thereof by antibody MEM-85 in glycosylated or non-glycosylated CD44.

Preprint Review

This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to Version 1 of the preprint.

Summary

This manuscript examines how N-linked glycosylation regulates the binding of polysaccharide hyaluronan (HA) to cell surface receptor CD44, to conclude that multiple sites exist but are controlled by the nature of the glycosylation. The reviewers appreciated many aspects of the work, but they have raised serious concerns about the experimental and simulation design. The reviewers suggested that the proposed alternative binding site may not be biologically relevant, as the relevant CD44-HA interactions are multivalent and cannot be supported by that site. They also suggested that the findings are not well supported by the NMR experiments, which could have been extended to allow comparisons of the glycosylation patterns hypothesised. Moreover, the MD simulations, despite being considerable in size, were limited in sampling different possibilities without bias from the initial HA placement, and there is not enough data to convince the readers of thorough sampling and reproducibility.

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

Learn more at Review Commons

Reply to the reviewers

OVERALL RESPONSE

We showed that chronic adaptation to mPOS, a new mechanism of cell stress initially discovered in yeast, induces muscle atrophy in a mouse model. We are pleased to see the overall enthusiasm from the reviewers about our work. The reviewers are unanimous in (1) that the work represents “a huge amount of work” that “has been well conducted regarding the characterization of the Ant1(TG/+) murine model that exhibits a muscle loss phenotype”, and (2) that our study linking mitochondria-induced proteostatic stress to muscle atrophy opens “a new field of investigation” and “is of interest to the scientific communities studying skeletal muscle pathophysiology and mitochondrial homeostasis”.

Mitochondrial alterations are important hallmarks in skeletal muscles under many pathological conditions. Given that bioenergetic deficiency alone is not sufficient to explain muscle wasting as shown by other groups, the reviewers commented that our work “is original and opens new perspectives in the field of mitochondrial dysfunction related to myopathies”.

Additional strengths noted by the reviewers include the demonstration of mPOS in an animal model, and the potential implication of our work for FSHD that is one of the most common muscle disease in humans.

We are also pleased to learn that the reviewers reached “cross-referee” recommendations to improve the paper. We are excited to see these recommendations. We are motivated and have the capacity to implement all the four series of experiments recommended by the reviewers as outlined below.

REVIEWERS’ COMMENTS AND OUR RESPONSE

1. It would also be useful to perform cell fractionation and measure the accumulation of Ant1 and other unimported mitochondrial precursor proteins in the cytosol (reviewer 1). Characterizing the aggregates observed in muscles, to see whether they contain ANT1, ubiquitin, p62 and unimported mitochondrial proteins (reviewer 2 & 3). It would be informative to measure the formation of soluble and insoluble protein aggregates (reviewer 1&2).

Response

        We propose to perform subcellular fractionation of muscle lysates using sucrose gradient centrifugation, coupled with western-blot. This will enable us to learn whether Ant1-induced stress increases the retention of unimported mitochondrial (pre)proteins (e.g., Ant1, Tom20, MDH2, TFAM, SDHA and Aco2) in the cytosol or extramitochondrial aggregates. We routinely practice this technique and we have all these antibodies validated in the lab.

        Our previous work showed that the giant aggresomes induced by ANT1 overexpression contain Ant1 and mitochondrial proteins in HEK293T cells (Liu et al., 2019, MBoC 30:1272-1284). However, the aggresomes we observed in the ANT1-transgenic muscles have sizes often comparable to mitochondria. Protein import stress may also lead to the accumulation and misfolding of precursors on the mitochondrial surface that are subject to ubiquitination and autophagic removal. It would be difficult to distinguish between aggresomes and mitochondria by IHC using antibodies against Ant1, ubiquitin, p62 and unimported mitochondrial proteins. To overcome this, we will take advantage of the subcellular fractionation technique described above.  This should enable us to clarify whether the cytosolic small aggregates co-fractionate with p62 and ubiquitin.

        As suggested by reviewer 1 & 2, we will determine whether NP-40 insoluble but SDS-soluble aggregates can be detected in the Ant1-transgenic muscles, using the dot blot technique as we previously published (Liu et al., 2019, MBoC 30:1272-1284). Antibodies against mitochondrial proteins, p62 and ubiquitin will be used to determine whether the aggregates are enriched in mitochondrial protein, p62 and ubiquitin.

        Collectively, the experiments proposed about will provide biochemical support for the retention, and possibly ubiquitination and p62-mediated aggregation of unimported mitochondrial proteins in the cytosol of Ant1-transgenic muscles.

2. Finalizing the characterization of the EM analysis (reviewer 2 & 3) – The reviewers suggested that we should try to quantify the different aggresomal/autophagic/vacuolar structures in the transgenic and control muscles in the TEM experiments.

Response

  Yes, we will perform the quantitation with the grids we prepared as suggested by the reviewers.

3. Evidence reduced altered protein synthesis rate (Reviewer 1 & 3) – Reviewer 1 suggested that it may be helpful to provide biochemical evidence for potential changes to protein synthesis rate, in order the validate the RNA-Seq data. This is also echoed by reviewer 3. The non-radioactive SUnSET technique (FASEB J. 2011 Mar;25(3):1028-39. doi: 10.1096/fj.10-168799) was recommended for measuring protein synthesis rate in vivo.

Response

  We appreciate reviewers’ suggestions and will be happy to set up this experiment. Briefly, we will inject the mice (n=3 for the transgenic and control mice) with puromycin to bind neosynthesized peptides. Muscle tissues will be collected and analyzed by western blot using an anti-puromycin antibody (#MABE343, Millipore Sigma). Quantitation of the western blot signals will inform whether relative protein synthesis rate is decreased in transgenic muscles compared with wild-type controls.

4. Quantifying different lysosomal markers (reviewer 2) – Reviewer 2 suggested that we should quantify the levels of LC3I/II, Lmap2 and other lysosomal markers (e.g, Beclin1).

Response

  We agree with this and the western blot experiments will be performed accordingly, using frozen muscle samples that we collected. We will also extend to the analysis using antibodies against CTSL (Abcam, #ab103574) and V-ATPase subunits such as ATP6V1H (Abcam, #ab187706) and ATP6V1G2 (Sigma, # WH0000534M2) that are upregulated in the transgenic muscles as revealed by RNA-Seq.

ADDITIONAL RECOMMENDATIONS.

The reviewers made additional minor recommendations that we found very constructive and helpful for improving the manuscript. These include providing for details of animal number and muscle types used in the experiments, statistical analysis, image analysis, and method sections for protein extraction and western blot. The reviewers also made suggestions for reorganization of some of the Figures. We have implemented some of these suggestions in the revised version of the manuscript.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

In this study the authors generated a transgenic mouse model of impaired mitochondrial import/loading by overexpressing ANT1, a mitochondrial carrier protein. The moderate overexpression of ANT1 was sufficient to cause progressive muscle wasting, characterized by a reduction of myofiber size. By transmission electron microscopy, the authors observed that aggresome-like structures in the cytosol of ANT1 Tg muscle cells, suggesting that occurrence of mitochondrial Precursor Over-accumulation Stress (mPOS). The authors characterized the transcriptional profile of the muscle of ANT1 Tg and WT mice by RNA-sequencing. The data supports that mPOS leads to changes in gene expression to promote: 1) mitochondrial protein import and proteostasis; 2) inhibition of protein synthesis; and 3) protein degradation. The authors speculate that this is an adaptative response to cope with the cytosolic proteostatic stress induced by ANT1 overloading but that, if chronically sustained, leads to a reduction of protein content and consequent muscle wasting.

Major comments:

The key conclusions of this work are that ANT1 overexpression induces accumulation of proteins in aggresomes in the cytosol, supporting the existence of overaccumulation of unimported mitochondrial precursor proteins (or mPOS) in an in vivo animal model. As an adaptive response, the cell's transcription profile changes to inhibit protein synthesis and promote protein degradation. This adaptation creates a protein imbalance that leads to muscle wasting.

The data and methods are clearly presented. However, this work would benefit from more evidence to strengthen the main conclusions. I have the following comments:

1) By transmission electron microscopy, the authors detected aggresome-like structures in the cytosol of ANT1Tg/+ but not in that of WT muscles. Is it possible to quantify the frequency of each type of structure in WT and ANT1Tg/+? Does this frequency increase with age or correlate with the degree of myofiber phenotype (reduction in size)?

2) mPOS is characterised by the accumulation of unimported mitochondrial precursor proteins. Is there any evidence either that the aggresome-like structures contain unimported mitochondrial proteins or that mitochondrial precursor proteins, still containing their mitochondrial targeting sequence, accumulate in the cytosol? Does unimported Ant1 precursor protein accumulate in ANT1Tg/+ mice?

3) Previous studies have shown that expression of mutant Ant1 protein causes mitochondrial morphology defects and mtDNA deletions. Do ANT1Tg/+ mice have altered mitochondrial morphology or deletions/loss of mtDNA in muscle?

4) The transcriptomic data and the amplification of the lysosomal compartment suggest an activation of multiple protein degradation processes, that could contribute to the reduced protein content in ANT1Tg/+ muscles. The increase in P-4E-BP and eIF2alpha expression in ANT1Tg/+ muscles suggest protein synthesis may be decreased. These data are consistent with unbalanced protein synthesis versus degradation. Figure 7A demonstrates a reduction in steady state protein levels in ANT1Tg/+ muscles. However, Figure 7A is insufficient to confirm a mechanistic explanation of the muscle wasting phenotype as the authors state at the end of the Results. This would require direct evidence of altered protein synthesis rates and protein degradation, which, although challenging in vivo, have not been directly demonstrated. The authors should therefore modify the final sentence in the Results.

Minor comments:

Suggestions to improve the presentation of data and conclusions: On page 7, it reads "No myofiber type grouping was observed in muscle samples stained for mitochondrial activities, suggesting the lack of chronic neuropathy." This sentence is lacking a reference to a figure (maybe Fig 2, D?).

On page 12, the authors say "These genes are known to be activated as an important regulatory circuit in the Integrated Stress Response (ISR), an elaborating signaling network that is stimulated by divers cellular stresses to decrease global protein synthesis and to activate selected genes in the benefit of cellular recovery (42)." The word diverse is misspelled.

On page 13, the authors wrote: "First, we found that the transcription of genes encoding proteasomal subunits, NFE2L1 and NFE2L2 are upregulated (Fig. 6A & 6B). NFE2L1 and NFE2L2 activate the transcription of proteasomal genes." This section could be rephrased for clarity. The first sentence suggests that NFE2L1 and NFE2L2 are proteasomal subunits. But then the authors say they activate transcription. I believe what the authors meant was that the genes encoding proteasomal subunits were upregulated (Fig. 6A), as well as NFE2L1 and NFE2L2 (Fig. 6B), that activate the transcription of proteasomal genes.

On page 13 (last paragraph), the authors mention that "numerous genes involved in autophagy, cytoskeletal organization and intracellular trafficking are upregulated in ANT1Tg/+ muscles" but they only explain what STBD1, ARHGAP33 and ARHGEF2 do. The other genes are not mentioned. If the authors want to speculate that the differences in gene expression are relevant, they should explain the role of the different genes/proteins.

On page 14, where it reads "we found that Lamp2-possitive lysosomes and/or lysosome-derived structures are amplified in the ANT1Tg/+ muscles", the word positive is misspelled.

Regarding Fig 1: in C there are two Tg bars, do they represent the two independent transgenic mice referred in the text? Or are these males and females, like in the following graphs? The authors should clarify this. What do the values and error bars represent and which statistical tests were used?

Regarding Fig 2: How was the coefficient of variability calculated? A sentence explaining this might be helpful in the methods' section. There is no scale bar in D. Which statistical test was used in I-J? The authors mention " P values were calculated by unpaired Student's t test." for E-H but no for I-J.

Fig 3. is missing a scale bar.

In Fig 5, B-C: the labelling of males and females is missing from the graphs. Authors should indicate which statistical test was performed for each graph.

In Fig S1: Why split A from B? Why only showing 1 of the Tg for each timepoint and not both? If the images are representative of both transgenics, then it should be mentioned. Suggestion: adding the timepoint info (3, 6, 17 months) above each panel makes it easier to understand the figure without reading the legend. E-H: the graphs show grouped variables, wouldn't a Two-way ANOVA be more appropriate to test for statistical significance?

Significance

mPOS has been observed in yeast (Wang and Chen 2015) and human cells (Liu et al. 2019). This works shows, for the first time, evidence for mPOS is an in vivo animal model. This work suggests that a moderate overexpression of ANT1 is sufficient to impair mitochondrial import and loading, causing mPOS and an adaptative response to cytosolic proteostatic stress. These findings are relevant to understand pathologies where ANT1 is affected, such as facioscapulohumeral muscular dystrophy (FSHD) (Laoudj-Chenivesse et al 2005). mPOS represents a novel mechanism through which mitochondrial dysfunction impacts on muscle wasting. Moreover, since impairment of the mitochondrial import and loading machinery is observed in aging and disease (MacKenzie and Payne 2007), these findings are relevant to better understand the impact of mitochondrial dysfunction in different cell types.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary:

The authors investigate the role of mitochondria - other than bioenergetic or oxidative stress - in the loss of muscle mass. They hypothesize that the accumulation of immature mitochondrial proteins in the cytosol is responsible for muscle atrophy, independently to mitochondria metabolism. For this purpose, they generated a murine model that over-expresses ANT1, a mitochondrial protein. The two-fold increase in Ant1 protein level leads to an overload in mitochondria import machinery, thus an accumulation of ANT1 in the cytosol, and thus to mitochondrial Precursor Over-accumulation Stress (mPOS). Consequently, the protein degradation pathway is stimulated, leading to an imbalance between protein synthesis/degradation, and in long term in muscle wasting.

Major comments:

The authors have executed a huge amount of work. Before fully reaching the conclusion that mPOS induce an imbalance between protein degradation/protein synthesis, by mainly increasing the lysosomal pathway, the authors should test/validate few more things:

1-Quantify the increase in LAMP2 muscle (WB and/or immunolabeling quantifications

2-Quantify other lysosomal markers

3-Characterize better the aggresomes observed in ANT1Tg/+ muscles

4-Add details for the methods (details are missing and make it difficult to judge this section, see comments below) These experiments should be easily done, 3 months of work: regular WB analysis, immunohistology and analyses EM images the authors already have, cost for supply <$2000.

Results:

Figure 1, movie 1-3 and paragraph "Moderate Ant1 overexpression causes progressive muscle wasting." The authors generated two independent hemizygous transgenic mice (ANT1Tg/+) and characterized them. The authors show a greater level of ANT1 in the transgenic mice. Could they show the localization of ANT1 in ANT1Tg/+ muscles: cytosol? Near the mitochondria? Sub-sarcolemmal mitochondria or else? Does ANT1 form aggregates? If yes, do the aggregates co-localised with ubiquitin? Proteasome? Lysosomal markers?

Figure 3 and paragraph "Cytosolic aggresome formation supports mPOS." The authors show EM images of muscle section of ANT1Tg/+ muscles at 1 and 2 years old. The authors wrote that there is an increase of aggresomes: they show in figure 3Q and M structures that look like mature lysosomes, or in 3F and 3R early mitophagy ... The authors should try to classify the different structures they observed and quantify these structures (eg number of autophagic vacuole per sarcomere). They should then perform some immunostaining on muscle sections at same age to confirm an increase in lyosomal markers for example. They still should do the same analysis (quantification and immunostaining) in WT muscle tissue same age. Figure 3 B and C suggest lipid vacuoles. Can the authors check using Oil red-O staining for example (or another staining)? The accumulation of lipid drops in transgenic muscle would suggest an impact on the metabolism, and more specifically on the lipid metabolism. All these structures are classic and should be observable in WT muscle, but probably at a lower frequency. Attempting to quantify these parameters and confirm by histochemistry would help to characterise better the murine model.

Figure 4, Supplemental Figure 4, Supplemental Table1 and paragraph "Ant1 overloading activates genes involved in mitochondrial protein import and proteostasis, and those encoding small heat shock family B chaperones consistent with mPOS." The authors generated Supplemental table 2 but never mentioned it in the text. Figure 4 and supplemental Figure 4, can the authors add the stat. The authors conclude from these figures (Figure 4 and Supplemental Figure 4) that ANT1 overexpression causes a protein import stress on the mitochondria. This is based on transcriptomic analysis and RTqPCR. They should validate at the protein level: eg level of HSPBs, NACA and HsP90 by WB and localisation in muscle section by immunostaining (counterstaining with mitochondria marker)

Figure 6 and 7 and paragraph "Activation of multiple protein degradation processes and reduced protein content in ANT1Tg/+ muscles." Figure 6H: the authors should quantify LAMP2 level. Other markers of the lysosomal should be assessed at the protein level (LC3I/II, Beclin1 etc) The proteasome pathway does not seem strongly stimulated as no increase in ubiquitinylation nor in P62 are observed by Western blot. However, the authors should check whether The aggresomes observed, do they colocalise with ubiquitin and/or P62 proteins in muscle section (if yes try to find a way to quantify this if there is some colocalization). Are the aggresomes soluble or non-soluble proteins? The latter could interfere on the absence of detection n of increases in protein ubiquitinylation.

Material and methods

Paragraph describing the statistical analysis is missing. Number of mice, sample used for each experiment should be added in the Mat and Methods as well.Which muscle was used for which experiments (for histology, EM and RNAseq in Mat & Methods)? The procedure for image analysing is missing: objective used, number of images analysis per sample, how many muscle were studied? Protein extraction and Western blot procedures is missing

Minor comments:

-Figure 1I: typo in the x axe legend: quadriceps instead of quadrucep

-Figure 2 and paragraph "Mitochondrial respiration is moderately decreased in ANT1Tg/+ muscles." Figure 2A-B-C: the authors should move the supplemental figure 2A and B in the main figure, and place figure 2Band C in sup data. To confirm that there is a difference in fiber size distribution, the authors should perform a Kolmogorov Smirnov test. Can the author clarify if whether they are using minimum diameter of fibres throughout the results (figure 2c and supplemental figure 2A and B), and if this is what is meant by the term "lesser diameter"? Figure 2I-J: It would be interesting to compare the different respiratory state at different age using an ANOVA2 factor and post-hoc test.

-Page 13: full stop missing after the reference (49): "ligases respectively, are frequently upregulated (49),"

-Supplemental Figure 4: reorganise the plot: put the reference ANT1 in first position, then organise per pathway involvement (eg: put together SLC7A1, SLC7A5 and ASNS for the acid transport, MTHFD2 and PSAT1 together for the one-carbon metabolism etc).

-The authors describe figure 6E and F before A,B,C,D... they thus may need to switch them around.

Significance

Muscle loss associated with cachexia, sarcopenia, or neuromuscular disorders, if of current interest to the field, with much work ongoing to study the role of inflammation, denervation, REDOX homeostasis and proteostasis. The current paper suggests a new mechanism that could be involved in muscle atrophy: mitochondrial protein load and import. The authors generated a new murine model that would be useful to the muscle community to investigate pathways involved in muscle wasting, in different physiological and pathological context. Working on different neuromuscular disorders and muscle ageing, the existence of such a model would be an interesting tool to investigate the role of mitochondrial dysfunctions (dysfunctions other than mitochondrial metabolism) in muscle wasting.

REFEREES CROSS COMMENTING

Reading the comments from other reviewers, it seems that there is general agreement that this paper has been well conducted regarding the characterization of Ant1TG/+ murine model, and muscle loss.

Similarly, all the reviewers seem to agree that: Finalizing the characterization of the EM analysis, characterizing the aggregates observed in muscles (containing ANT1?, ub?, p62?, soluble or non-soluble aggregates), as well as quantifying the protein synthesis and different lysosomal markers would improve the paper.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary In this manuscript, Wang and colleagues show that chronic adaptation to mitochondria-dependent proteostatic stress in the cytosol induces muscle atrophy. Although mitochondrial dysfunction is known to cause muscle wasting, the underlying mechanism is unclear. The authors generated a transgenic mouse (ANT1Tg/+) in which the expression of the nuclear-encoded mitochondrial carrier protein Ant1 is increased by two-fold. These mice are characterized by the progressive loss of body weight and muscle mass. As revealed by muscle histology and immunocytochemistry analysis, ANT1Tg/+ mice have decreased myofiber size and increased myofiber size variability. Consistent with muscle wasting, ANT1Tg/+ mice are characterized by decreased home cage activities and exercise tolerance. Mechanistically, the authors found that ANT1 overexpression has a relatively mild effect on mitochondrial respiration. However, ANT1 overexpression induces cytosolic proteostasis stress (mPOS), the formation of aggresome-like structures and the activation of small heat shock proteins in the cytosol accompanied by the upregulation of the stress-activated transcriptions factors. The drastically remodeled transcriptome of ANT1Tg/+ mice muscles is indicated by the authors as an adaptive response to counteract mPOS.

Major comments

The phenotypic characterization of the ANT1Tg/+ skeletal muscle is well conducted and detailed. However, the key conclusions are mainly based on the interpretation of RNA-seq data with little experimental evidence of the underlying mechanism. For this reason, the authors should qualify some of their claims as preliminary or speculative. For instance, they find upregulation of SENS2 gene, which has been demonstrated to take part in mitophagy. In addition, they corroborate this data with electron microscopy images where mitophagy structures are present. However, these two data are not enough to state that mitophagy is involved. It would be advisable to focus on fewer genes but with a stronger validation process.

I recommend performing the following experiments:

To better characterize the effects of Ant1 overexpression on mitochondrial function, the author should address whether ANT1Tg/+ mitochondria are more prone to depolarize or not.

It would be informative to measure the formation of soluble and insoluble protein aggregates.

It would also be useful to perform cell fractionation and measure the accumulation of Ant1 and other mitochondrial proteins in the cytosol.

Finally, I would suggest to measure protein synthesis by the non-radioactive SUNSET technique (FASEB J. 2011 Mar;25(3):1028-39. doi: 10.1096/fj.10-168799).

Minor comments: The author should mention how many muscles were used for the EM studies. I would encourage discussion of the atrophic (and not dystrophic) phenotype of the ANT1Tg/+ mice related to the connection between Ant1 and FSHD.

Significance

The authors investigate a still poorly explored mechanism underlying muscle wasting based on mitochondrial import machinery dysfunction. The work is original and opens new perspectives in the field of mitochondrial dysfunction related to myopathies.

Mitochondria alterations play a key role in the context of muscle decline in many diseases and in aging. It is well known that the alteration of mitochondrial respiration and the oxidative stress increase are hallmarks of mitochondria dysfunction in skeletal muscles under pathological conditions. However, there is evidence that these features are not sufficient to explain the severe phenotype of muscle wasting. This work opens the way to the possibility that non-bioenergetics factors could take part in the pathological scenario. In detail, the involvement of the mitochondrial import mechanism, which causes a cytosolic proteostatic stress, is a new field of investigation. A few years ago, the same authors demonstrated that the mitochondrial precursor over-accumulation stress (mPOS) triggers a cytosolic proteostatic stress in yeast, however until now there was no evidence whether this phenomenon could occur in animals and which tissues would be involved. Thanks to this work, the authors demonstrated that mPOS occurs in skeletal muscle.

This study is of interest to the scientific communities studying skeletal muscle pathophysiology and mitochondrial homeostasis. My main research field is the role of mitochondrial homeostasis in skeletal muscle function in health and disease.

Reviewer #3

This study addresses the role of the miR29 micro RNAs in the regulation of melanoma development. Expression of miR29 that is generated from pre-miRs from two clusters is regulated by oncogenic BRAF in human melanocytes. Levels of the mature miR29 are down-regulated in melanoma compared to untransformed melanocytes or nevi and inhibition of miR29 function increases melanoma growth in a murine in vivo model. From RNA-seq data and computational analyses, the authors identify the small MAF protein MAFG as a novel target of miR29 that is involved in melanoma growth. . This study is focused on the function of miR29 in melanoma. The necessity of having the first two figures relevant only for the role of oncogenic BRAF and NRAS in regulating miR29 expression in MEFs is not obvious. Perhaps only one of the two should be shown as a main figure while the other could be moved to the supplemental figures.

The authors state that TPA regulates the MAPK pathway, but this is misleading as the primary target for TPA is PKC. This should be corrected.

The comparison of miR29 expression in the collection of melanocyte and melanoma lines uses a poor logic. BRAFV600E expression in primary melanocytes leads to senescence, the HERMES lines are already immortalized by exogenous expression of various genes (CDK4 etc), but this is not mentioned. What would be the effect of BRAFV600E expression in primary melanocytes on Mir29 and MAFG expression? The comparison between these melanocyte lines and the melanoma lines is also misleading as while they all share the BRAFV600E mutation, the melanoma lines have very different transcriptional signatures some being of melanocytic phenotype and other de-differentiated phenotypes. This is not mentioned and how the differences in transcriptional phenotype and P53 status affect miR29 and MAFG expression is not mentioned (see also comment below).

The description and characterization of the mouse melanoma models is not acceptable as presented. There are no images of tumours, no measure of number and size of tumours or tumour progression only Kaplan-Meier plots of viability. It is impossible for the reader to assess the conclusions from the figure, the additional data should be added. Also, can the authors show that the mouse tumours (or the cells established in vitro) express Mafg and that its levels are altered in the different genetic backgrounds. If not then another mechanism is maybe operative in the mouse tumours.

The RNA-seq data following expression of the miR29 mimics is not fully described, how many genes were changed, what is fold change of the genes that were subsequently selected for further study, in particular MAFG?

The changes in MAFG protein expression in Figure 6A are minor. What is the evidence that such small changes can really impact cell growth (see below)? At face value, basal MAFG expression in H1B melanocytes appears higher than in WM164 cells and its levels in H1B cells can only be mildly affected by modulating miR29. Can the authors comment. More importantly, in Figure 6F some highly tumorigenic lines like 1205Lu or SK-Mel-28 have MAFG levels comparable to the HERMES lines. This does not support the authors’ hypothesis that MAFG levels are major regulators of tumorigenic capacity. There is no obvious correlation between the MAFG mRNA and protein levels comparing panels E and F. Also, what are the relative levels of miR29 in these different cell types, do they correlate with MAFG protein levels or are differences in MAFG levels explained by other regulatory mechanisms? Is there any correlation between MAFG protein levels and cell growth rates and clonogenic capacity amongst the different analyzed melanoma lines? Resolving these issues would strengthen the conclusions.

To fully demonstrate that the effects of miR29 in regulating tumour growth are principally mediated via MAFG, the authors must show they can rescue cell growth defects upon miR29 expression, by expressing MAFG from a cDNA that is insensitive to miR29 regulation. This experiment will help to exclude the implication of other potential miR29 targets in regulation of melanoma cell growth.

Reviewer #2

The manuscript by Vera and colleagues dissects the mechanism of miR-29 family expression in melanoma and provides a possible target to support its tumour-suppressive functions. Towards this the expression of miR-29 family upon MAPK and P53 signalling is carefully followed in transgenic mice and humans and classical target analysis is performed. However a few points remain to be addressed:

Subsection “The MAPK pathway regulates miR-29 expression in human melanocytes and melanoma cells”: "Our results indicate that BrafV600E-induced expression of miR-29 may form a tumour suppressive barrier that restricts the full transformation of melanocytes." This is an overstatement. While the authors clearly show a tumour suppressor role for miR-29 and clearly show that it is induced by MAPK signalling, they never prove that inhibition of miR-29 supports melanocyte transformation.

Discussion section: "Thus, our work has uncovered that miR-29 prevents melanoma progression downstream of MAPK signalling by repressing MAFG." Again overstatement. Although the authors prove that MAFG is important in melanoma and it is a target of miR-29, they never prove that the activity of miR-29 is mediated by MAFG. A rescue experiment is missing here. The sentence needs a rewording.

Additionally, the authors could add (expression) correlation analysis between miR-29 and MAFG in human melanoma samples from publicly available databases.

Reviewer #1

The goal of this manuscript is appealing. The authors wish to evaluate the importance of mir-29 and MAFG in melanoma progression that would be linked to the activation of the MAPK pathway. This article presents a huge amount of biochemical experiments; it has a potential. However, a significant number of issues must be clarified.

The MAPK pathway is induced in the very large majority of melanoma, the role of mir29 and MAFG should then be observed in the vast majority of melanomas. Is it the case? If not, what is(are) the main cause(s)? The authors used BRAF V600E, which is perfectly understandable in the case of melanoma, and they also used KRAS G12D. This last mutation is very rare in melanoma. Why not address a similar question with NRAS Q61K/R?

Choice of the cellular models:

1) The authors focus on mouse embryonic fibroblasts (MEF) in the two first figures. What is the significance of MEF for human melanoma? Why not using primary melanocytes from human (NHEM) and/or established mouse melanocyte cell lines?

2) In this study, as models the authors use mouse (MEF and transgenic) and human (melanoma and melanocyte) species. A crucial question is: are the targets for mir29 the same in humans and mice? The conservation of miR and targets is very poor between species. This needs to be addressed.

3) The authors have to better explain their conclusion of Figure 1. All the presented experiments were performed in MEF. What would happen if the authors used cells from the intestine to evaluate the consequence of BRAFV600E on miR-29? The choice of intestine is not random. What is the link with melanoma? The reason for using mouse embryonic fibroblasts is fine to study molecular issues for this type of cells. However, it is fully accepted that the responses of melanoma to various agents are highly variable.

Quality of the presented results and reproducibility:

I will not go through all of the experiments. I will make some remarks.

1) Figure 1A: The abundance of pERK is moderately induced after expression of KRAS G12D. The authors have to show quantifications on several independent experiments to be convincing.

2) Figure 3E: The culture media are different in melanocyte and melanoma cell lines. It is therefore difficult to compare the level of miRs. For nevi and melanoma, there is also a pitfall. What is the level of these miR in the stroma? What is the percentage of stromal cells in these biopsies?

3) Figure 4: There is a clear action of miR-29 sponge in melanoma initiation in mice. What are the targets in mice? According to their models, are they the same in humans? According to the claim of the authors on progression, we expect that the mice have more metastasis? Is it the case? The authors present an overall survival curve. Knowing the ethical rules associated with mouse studies, the authors do not show the survival since they have to sacrifice the mice. The authors have to show the associated raw data.

4) Figure 7: To further test the importance of MAFG as an oncogene, the authors have to evaluate the growth proliferation in a medium lacking major supplement allowing melanocytes to grow in culture, to reduce the amount of serum, and to test the ability of these cells to grow in 3D or/and in mice.

Terminologies are vague and/or not defined:

1) What do the authors refer to "melanoma progression"? In vivo, the authors address the question of melanoma initiation. There is no information on invasion or metastasis. This is crucial according to their title.

2) AOf course according to the title we wonder if this function is attributed to miR-29a? miR-29b? miR-29c? All? Proper introduction of these three miRs must be done including the known targets of these Mirs. Of course, it has to include the knowledge associated with the different species. In particular, they have to make the point for mouse and humans.

3) The authors refer to physiological conditions in vitro on plastic in the presence of calf serum. The authors must reformulate the text accordingly and tone down their conclusions.

4) The authors refer to "full transformation of melanocytes". What do they refer to? It is too vague. Molecular? Cellular?

Bypass of senescence in melanomagenesis:

Bypass of senescence is mainly due to the RB/INK4A during melanomagenesis. P53 may be involved but appears to occur later. The authors must address this issue, especially when they use Hermes cells.

TPA induces mainly PKC and not the MAPK pathway as the authors mention. The authors should clearly show that the MAPK pathway is indeed induced, not only using pERK. Here, in this context, the WB analyses are not sufficient. Moreover, what would be the action of dbcAMP and aMSH?

Additional comment:

The authors could present a clear and comprehensive scheme for humans (and mice?) representing the associated pathways.

Preprint Review

This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to Version 2 of the preprint.

Summary

This study addresses the role of the miR29 microRNAs in the regulation of melanoma development. Expression of miR29 that is generated from pre-miRs from two clusters is regulated by oncogenic BRAF in human melanocytes. Levels of the mature miR29 are down-regulated in melanoma compared to untransformed melanocytes or nevi and inhibition of miR29 function increases melanoma growth in a murine in vivo model. From RNA-seq data and computational analyses, the authors identify the small MAF protein MAFG as a novel target of miR29 that is involved in melanoma growth.

We found this study interesting, but we are of the opinion that the central hypothesis that miR29 regulates MAFG levels to influence melanoma is not yet fully substantiated by the data. Critical experiments could be added, for example, the rescue of growth defects upon miR29 mimic expression with a miR-insensitive form of MAFG, or evidence that Mir29 regulation of Mafg is involved in the mouse melanoma. Furthermore, we do not feel that the immunoblots support the idea that MAFG promotes tumour growth as the 1205LU cells that are highly tumorigenic in nude mice have MAFG levels comparable to the melanocytes lines.

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

Learn more at Review Commons

Reply to the reviewers

Reviewer #1:

In this manuscript the authors explore the requirements for centromere transcription using single-molecule FISH. Previous studies have found that centromeres are transcriptionally active in a wide variety of organisms. Centromere transcription has been proposed to facilitate Cenp-A deposition through chromatin remodeling and to directly contribute to centromere/kinetochore function by producing a functional ncRNA. However, we currently know almost nothing about how transcription is initiated at the centromere or how levels of centromere transcripts are controlled. This manuscript makes several major findings that are potentially of importance to groups studying centromere transcription. 1.) Centromere RNAs are produced by RNA Polymerase II and are localized in the nucleus of a wide-range of cell types. 2.) Centromere RNAs do not localize to the centromere, which is in contrast to several recent studies. 3.) Centromere proteins are not required for transcription of alpha-satellite sequences. 4.) Localization of centromeres to the nucleolus represses centromere transcription. Overall, this is a solid manuscript and has the potential to make a significant impact in the field. Below I suggest a couple of experiments and modification to the data presentation that could improve the manuscript.

We thank this reviewer for their interest in this paper and agree with their clear articulation of the key points.

1. • All of the experiments in this manuscript rely on detection of centromere RNAs using single molecule FISH probes. These probes are validated by showing the RNase treatment removes the FISH signal. A strength of this approach is that the authors use multiple different probe sets and achieve comparable results. However, there is no orthogonal validation that the probes detect alpha satellite RNA. All of the experiments in this manuscript would be significantly improved by showing that the results presented here can be confirmed by a different approach. I suggest that the authors use Q-RT-PCR to validate the smFISH results. * The smFISH probes provide a powerful and unique strategy to detect alpha-satellite transcripts. To ensure that these experiments are carefully controlled, we analyzed multiple distinct probe sequences that recognize alpha-satellite transcripts derived from different chromosomes, as this reviewer highlights. We also conducted an in-depth computational analysis to ensure that these probes do not match genomic sequences outside of alpha-satellite regions. However, we recognize and agree that a complementary method to detect these transcripts would be a useful addition to this paper. We are currently highly constrained in our ability to conduct these experiments due to COVID-19-related laboratory closures, but if feasible our goal for a revised manuscript would be to conduct qPCR experiments for a subset of the conditions that are the most central to the key results in this paper (focusing particularly on HeLa and Rpe1 control cell lines, CENP-C iKO, Ki67 KO, and RNA Polymerase I and RNA Pol II inhibitors).

• Several results in this manuscript directly contradict results in published studies, but these discrepancies are not discussed. I believe the authors need to discuss the following discrepancies between their results and those in the literature: *

• McNulty et al. Dev. Cell. 2017. Show that alpha-satellite RNA is transcribed from all centromeres and remains localized to the site of transcription. The different results and possible explanations for the differences should be discussed. *

1. • Additionally, Rosic et al. JCB 2014, Blower Cell Reports 2016 and Bobkov et al. JCB 2018 all show that centromere RNAs localize to centromere regions. The differences between these studies and the authors results should be discussed. *

• The authors show that satellite RNA cannot be detected on mitotic chromosomes. However, Johnson et al. Elife 2017, Bobkov et al. JCB. 2018, and Perea-Resa et al. Mol. Cell. 2020 show that EU-labeled RNA can be detected at the centromere during mitosis. The authors should discuss the discrepancy between their results and these studies. Is it possible that their smFISH probes do not detect nascent, chromatin-bound transcripts? *

We believe that a strength of our paper is that it assesses alpha-satellite transcripts in individual intact cells using fixation conditions that preserve the native behaviors without disruptive and harsh extraction. As our results differ from those of other laboratories in some cases, we agree that it would be helpful to comment more directly on these differences with prior work. Points a, b, and c above all relate to the presence of alpha-satellite transcripts at centromeres. For the revised paper, we will include a discussion of these prior observations and some possible reasons for the differing results. In particular, we think that these discrepancies reflect two key differences:

1. Other strategies with harsh extraction conditions likely eliminate soluble alpha-satellite transcripts that are not tightly associated with centromeres, whereas our work preserves these.
2. It is possible that we are unable to detect nascent transcripts by smFISH as these are embedded within the RNA polymerase. Extraction conditions: An advantage of the smFISH probes used in our paper is that these require mild fixation conditions without prior extraction to better preserve cellular structures allowing us to analyzed intact cells, rather than chromosome spreads. Thus, our approach maintains the diverse alpha-satellite transcripts that are not bound to centromeres, and which may have been washed away in other studies. In contrast, some prior studies used stringent extraction conditions and primarily conducted experiments in chromosome spreads (not intact cells). Although it is not feasible to precisely determine the basis for differences without repeating this work the precise approaches and conditions from each paper and working closely with each group, we believe that these substantial technical differences explain our differing observations that reveal that the majority of alpha-satellite transcripts do not remain at centromeres..

Nascent transcripts: As suggested by this reviewer, we agree that our differing conditions may mean that we are unable to detect nascent transcripts that are closely associated with the RNA polymerase, inaccessible due to their chromatin proximity, or that are not sufficiently elongated such that they are present to hybridize to multiple copies of the smFISH probes to be detectable. The alpha-satellite transcripts must be derived from centromeric and pericentromeric regions and so must exist there at some point (as also attested to the EU signals that this reviewer mentions in the work from our collaborative the Blower lab; we have also detected EU signal at centromeres). However, our work suggests that alpha-satellite transcripts do not persist at centromeres indefinitely once generated, with mature transcripts in the nucleoplasm and liberated from chromosomes during mitosis. We believe that the combination of the relative inability of our smFISH probes to detect nascent transcripts, but stringent conditions disrupting non-centromere bound transcripts for prior work likely explain these distinctions.

• The authors show nicely that deletion of Ki-67 reduces centromere localization to the nucleolus and increases centromere transcription. However, this has no effect on centromere function. Studies from the Earnshaw lab (e.g. Nakano et al. Dev Cell 2008 and Bergmann et al. EMBO J. 2011) show that increasing or decreasing centromere transcription results in loss of kinetochore function on a human artificial chromosome. The authors should discuss the differences between their results and these studies. Is it possible that the small size of the HAC exaggerates the importance of the correct levels of centromere transcription? *

We are big fans of the Earnshaw lab work. In this case, there are a couple of possibilities to explain the strong effect that the Earnshaw lab observed on kinetochore function by perturbing centromere transcription. First, the degree of the change in centromere transcription may make a big difference. The Ki-67 results in an approximately 2-fold increase in alpha-satellite smFISH foci, which may still be within a permissive range for normal kinetochore function. Second, the experiments from the Earnshaw lab rely on targeting activating or silencing proteins to the centromere region, and it is possible that changes in centromere chromatin downstream of these factors contribute to the observed phenotypes in addition to altering the amount of centromere transcription. We will include a brief discussion of the Earnshaw work in a revised paper.

• The authors treat cells with transcriptional inhibitors for 24 hours. I am concerned that this may result in massive cell death. It would be helpful to include cell viability data from these experiments. *

We appreciate this point and agree that cell lethality is an important consideration given the essential role of the RNA polymerases. For the inhibitors, we first treated the cells for a variety of different time points to evaluate these behaviors. For example, we found that we could treat cells with RNA Polymerase II inhibitors for as much 48-72 hours without detecting noticeable cell death. Thus, at the 24 hour time point, the cells remain viable and intact, as is also visible in the images showing DNA staining for these treatments in Figure 3. We also note that this timing is consistent with prior studies that block transcription or translation. However, we did additionally conduct these experiments at earlier time points (5 hours and 12 hours post-drug addition) and obtained similar results. For example, for the Cdk7 inhibitor using the ASAT probe, we observed the following smFISH foci/cell: Control (3.4 foci/cell), 5 h (1.5 foci/cell), 12 h (1.2 foci/cell), 24 h (0.9 foci/cell). There is a clear effect even at 5 hours of treatment and a continued downward trend. Both for simplicity and because the replicates and number of cells that were quantified were lower for these conditions, we chose not to include these in the paper. We will include a statement regarding these earlier time points in the revised version.

• In Figure 3C the authors examine the effects of centromere protein knock outs on centromere transcription. To me this is the most important experiment in the manuscript and is a major step forward for the field. The authors use inducible CRISPR knock out cell lines that are not 100% penetrant. It would be helpful if the authors could describe how they ensured that cells included in the image quantification were knock out cells. *

Based on this comment and the other questions from the other reviewers, we recognize that we need to provide a much better description of the CRISPR knockout strategy, the prior validation of these cell lines, and the strategies that allow us to use these cell lines in a robust manner to ensure that we are effectively eliminating the target genes. We have systematically tested this strategy in multiple cases and find that this strategy is superior to RNAi for its efficacy and the potency of the phenotype, particularly for this type of cell biological assay.

The Cas9-based strategy is a highly effective way to conditionally eliminate essential genes. In this case, the efficiency of the Cas9 nuclease ensures that the genomic locus is cleaved in essentially 100% of cases. As this is repaired in an error prone manner and typically using non-homologous end joining, 66% of individual events result in frame shifts mutations that disrupt the coding sequence of a target gene, with ~50% of cells resulting in frame shifts in both copies of a gene. In addition, if a sgRNA targets a region of a gene that cannot tolerate mis-sense mutations, this will result in an even greater fraction of mutant cells. Thus, these inducible knockout cell lines result in robust and irreversible gene knockout, with a large fraction of cells (50% or more) displaying a clear phenotype. However, it is also true that there are a subset of cells within the population that will repair the DNA damage following Cas9 cleavage in a way that preserves protein function such that they behave similarly to control cells. Importantly, this means that there will be two classes of cells within a population – those that are unaffected, and those that are strongly affected. As we are analyzing each cell individually instead of creating a population average, this will capture this phenotypic diversity to reveal two populations of behaviors in cases where eliminating a gene results in a substantial change in smFISH foci. For example, the smFISH foci/cell data for the CENP-C inducible knockout (Fig. 3C and 3E) indicates that many cells have smFISH foci numbers that are comparable to control cells, but others that display substantial differences and highly increased numbers. An ideal control in these experiments would be to additionally analyze the levels of the target protein together with the smFISH analysis. Unfortunately, many of the antibodies are not compatible with the conditions needed for the smFISH. For CENP-C, the antibody that we have is not compatible with the conditions that we are using for the smFISH, so it is not feasible to co-stain these cells as suggested. Instead, for our analysis of the centromere-nucleoli localization (for example), we used the presence of a clear CENP-C interphase phenotype (“bag of grapes” resulting from chromosome mis-segregation) as an indication that the cells had been knocked out for CENP-C.

The majority of the Cas9-based inducible knockouts that we used for this paper were generated previously in the lab (McKinley et al. 2015; McKinley et al. 2017). For the centromere protein knockouts (McKinley et al. 2015), these were analyzed previously with respect to phenotype and monitored for the depletion of each gene target over time. For the larger collection of cell cycle and cell division inducible knockouts, for our prior work we systematically validated each of these with respect to their phenotype (see http://cellcycleknockouts.wi.mit.edu). Thus, we are confident that each of these cell lines is functional and effective for eliminating the target gene.

For conducting the experiments using the inducible Cas9 cell lines in this paper, we used the presence of these previously-defined phenotypes within the population as a validation that the strategy is working. Again, in general we find these knockouts are both penetrant and severe in their phenotypes. Importantly, for this diverse set of genes, we note that our goal was to broadly survey diverse factors to identify changes in alpha-satellite transcript levels. We intended this analysis as a “screen” where we would identify factors that resulted in a substantial change in the number of smFISH foci. As with any larger analysis, it is possible that there are false negatives where we did not detect a strong effect on transcript levels (such that they may contribute to centromere transcription). We have tried to use caution not to indicate that this data excludes any possible role for these factors in transcript levels, although in general the majority of the tested factors did not show a substantial change in smFISH foci. For the revised paper, we will make an explicit statement to this effect.

• On p8. The authors cite Quenet and Dalal. eLife 2014 for the idea that transcription during G1 is important for new Cenp-A loading. They should also cite Chen et al. Dev. Cell 2015 and Bobkov et al. JCB. 2018. *

Thank you for these helpful suggestions. We will update the text to incorporate these references.

Reviewer #2:

The study by Bury et al. investigates the formation of two different types of alpha-satellite transcripts (ASAT, SF1 and 3) in different human cell lines. Using smFISH they find that during the cell cycle these centromeric transcripts don’t stay at the centromere and are found in the cytoplasm after mitosis. Using specific inhibitors, they find that transcription is dependent on RNAPII, but not on various centromere and kinetochore proteins taking advantage of an inducible CRISPR-depletion system that the lab had previously developed. Interestingly, they find that CENP-C, a major component of the centromere and previously characterised as an RNA-binding protein, negatively regulates alpha-satellite transcript levels. Another regulator for transcript levels appears to be centromere-nucleolus interactions (as also indicated in the title) acting to suppress expression of these non-coding RNAs.

This is overall a really interesting study and indeed, transcription at the centromere is little understood at this point. Given the importance of the centromere the findings in this manuscript will be of high interest to both researchers in the field and a general audience. There are novel and interesting insights into centromeric transcripts but the study still requires some controls.

We appreciate this reviewer’s kind words and their clear description of our work.

1) The authors state that the majority of smFISH foci do not colocalise with centromeres in a combined IF/FISH experiment (some quantification and a % of that subpopulation should be given somewhere). This is a bit concerning but of course could also be true. It either means that alpha-satellite transcripts leave the centromere as suggested by the authors (although some should be visible at the centromeres during the act of transcription). Alternatively, a trivial explanation would be that there is a lot of unspecific staining, which can occur in FISH-experiments to varying degrees. The RNase treatment to control for the absence of potential DNA hybridization is convincing, but the FISH probe could also interact with non-centromeric cellular RNA. With the centromere localisation as a reference point gone, some control is needed to validate that the RNA-FISH signals are indeed recognising alpha-satellite RNA that emerged from centromeres. The authors could try competition experiments titrating unlabelled specific or unspecific DNA probes alongside their labelled specific FISH probe into their FISH experiment to see if they lose or maintain the signal and the number of foci. The specific RNA FISH probes could also be used in DNA FISH, to demonstrate they are working and recognising specific centromeres.

For understanding this behavior, we believe that an important feature of alpha-satellite transcripts is that they are relatively stable (protected from nucleases within the nucleus), but that their overall number is low, consistent with transcription of other non-coding regions across the genome. Thus, if a transcript were produced at centromeres, but subsequently diffuses away, only a small subset would be detectable at centromeres. In addition to our validation these probes using RNAse, we would like to highlight that we have analyzed multiple distinct sequences that recognize different subsets of alpha-satellite repeats. In each case, the observed behaviors are very similar. In addition, the nature of the oligo FISH method requires multiple individual probes to anneal to the same transcript such that a signal is only detected if a sufficient number of oligos bind to the same transcript. This makes nonspecific binding unlikely to contribute to a false signal. Finally, a subset of the perturbations that we tested that are relevant to centromere function (including the CENP-C inducible knockout) clearly affect the levels of these transcripts, supporting a centromere origin. The additional control experiments suggested by the reviewer could be useful, but are technically complex with their own caveats in interpretation and we do not feel that they would add substantially to the existing paper. Instead, as discussed in response to Reviewer #1, point #1, we plan to validate key results described in the paper using qRT-PCR (if possible based on current experimental constraints in the lab associated with COVD-19).

As described above in response to Reviewer #1, point #2, we also believe that some differences with prior work suggesting that alpha-satellite transcripts localize to centromeres may be due to stringent extraction conditions that eliminated non-centromere bound transcripts, while at the same time reflecting our inability to detect nascent transcripts. Quantifying “colocalization” within the nucleus is limited by the resolution in light microscopy, and we would prefer to use caution in defining which transcripts in our smFISH analysis overlap with centromeres. However, we believe that our work clearly highlights the fact that a general feature of mature alpha-satellite transcripts is that they localize throughout the nucleoplasm and are not strongly associated with mitotic chromosomes.

2) Apart from Figure 4, there is no analysis shown for statistical significance. This should be done for most if not all quantifications. Are indeed ASAT and antisense RNA Foci number not significantly different? The authors say that the levels of alpha-sat RNA in Rpe1 cells are not substantially different from other cell lines, but is it also not significant (Fig 1F)? In Figure 2D it is concluded that transcripts foci number are increased in S/G2 (from G1) and remain stable in mitosis, but it looks like there is an increase in mitosis. Again, it looks like the higher number of smFISH foci/Cell is significantly higher for both ASAT and SF1, so some statistical analysis would be required here.

For this paper, we quantified hundreds of cells for each condition, measuring the number of foci/cell in each case. Because of these large n’s, even relatively small differences between samples become statistically significant when tested using standard statistical comparisons (unpaired T test and one-way ANOVA test amongst others). For our experiments, every sample condition included an analysis of control cells, allowing us to compare the control condition to any perturbations on the same day. However, there is some variability between these different replicates, with the average number of ASAT smFISH foci/cell in HeLa cells ranging from 3.4 to 5.6. When compared relative to each other, a subset of these control samples will appear to be statistically different from each other despite the fact that this is not a substantial difference between replicates. Similarly, the majority of the tested inducible knockout cell lines are statistically different from control cells, even when the differences are relatively minor. Therefore, we have tried to use caution when applying the double-edged sword of statistics to these analyses. Instead, we have tried to consider differences with a “substantial magnitude” instead of “statistically significant” differences that may make modest, but statistically significant differences seem artificially more important. We believe that the graphs in which every data point is represented, together with listing the average number of foci/cell in each condition allow the reader to evaluate this data for themselves. Many of the trends that this reviewer highlights are indeed interesting comparisons to consider for future work.

3) Starting with the description of Figure 1E in the main text the paper equates foci count of smFISH per cell with RNA transcript levels. I'm not convinced that these are necessarily the same. You could have many weak foci or few very bright with the same amount of overall transcripts in both. The authors start out introducing smFISH as highly sensitive "for accurate characterisation of number ...of RNA transcripts". This suggests that foci intensity could be used as a read-out for transcript levels. It should be possible to measure individual intensity of the foci for a subset of images. Do foci intensity correlate or anti-correlate with foci numbers? Is the sum of the intensities of all the foci less variable than the foci number for an individual cell type?

Due to the repetitive nature of alpha-satellite sequences, an increased intensity of a smFISH foci could reflect either the close proximity of multiple separable transcripts, or a longer transcript with multiple binding sites for the smFISH probes. Because of this, throughout the paper, we have referred to these as “foci” instead of stating a specific transcript number. As part of the automated computational analysis of the smFISH images, we additional analyzed foci intensity. In general, these values were similar across a cell population and between various perturbations with the key results and findings consistent whether we measured foci number or overall foci intensity per cell. However, foci intensity can vary slightly across a coverslip (technical constraints, not biological differences), and thus we have focused on foci number as a more consistent metric that correlates with the production of alpha-satellite transcripts.

4) I really like the use of the inducible CRISPR system to remove various centromere factors. However, some validation would be required to show that the system is effective in removing the proteins of interest in these experiments. For instance it would be helpful to show in Figure 3D an additional panel with CENP-C staining. Also for a subset of factors, some antibody staining co-staining with the smFISH could be provided in the supplemental material.

We appreciate this point. However, we feel that the existing experiments appropriately consider the nature of the knockout. First, we primarily used Cas9-based inducible knockouts that were generated previously in the lab (McKinley et al. 2015 and McKinley et al. 2017). As these knockouts have been described previously and extensively validated with respect to phenotype (in every case; see http://cellcycleknockouts.wi.mit.edu for example) and antibody staining (in selected cases), we have not repeated this here for the diverse cell cycle knockouts used. In general, we find these knockouts are both penetrant and severe in their phenotypes. Given the broad number of knockouts that we tested, this is not feasible in every case. We also intended this analysis as a type of “screen” where we could validate any “hits” that were observed, and will use caution in our wording not to imply that a negative result is decisive.

The important exceptions to this are CENP-C (which we analyzed more closely) and Ki67 (for which both the inducible and stable knockouts were generated for this paper). For Ki67, the antibody staining is shown and we believe that this is clear. For CENP-C, the antibody that we have is unfortunately not compatible with the conditions that we are using for the smFISH, so it is not feasible to co-stain these cells as suggested. For the smFISH analysis in the inducible CENP-C knockout, we analyzed every single cell, including some cells that are likely to have intact CENP-C levels. Thus, if anything, the potent increase in smFISH foci underrepresents the dramatic effect of CENPC depletion. Based on our prior work (McKinley et al. 2015) we found that the CENP-C knockout results in a pervasive “bag of grapes” phenotype in which chromosomes mis-segregate during mitosis and are packaged into separable interphase nuclei. For the analysis of the nucleoli, we selected cells that displayed this clear phenotype (as shown in the figures).

5) Since none of the CRISPR iKO has a particular inhibiting phenotype it would be useful to include some positive control in the CRISPR experiment. Would it be possible to use a CRISPR iKO target that affect some factor of the transcription machinery (RNA Pol II or similar) to reduce transcript levels?

Generating additional Cas9 iKO cell lines is feasible, but would be time consuming. In this case, we are not convinced of the value of generating and validating these additional cell lines (particularly with the additional current constraints due to COVID-19). For evaluating the role of the RNA polymerases, we believe that the effect of the drug treatment is clear. For creating a positive control to assess whether the CRISPR iKO strategy is a feasible way to conduct these experiments, we would like to highlight the CENP-C iKO cell line, which has a potent effect in this assay.

6) The authors find a negative correlation between the nucleolus-centromere association and the number of alpha sat foci. This is really interesting and they suggest that the nucleolus association could negatively regulate centromere transcription. However, this correlation is rather indirect in the sense that cells with a higher-degree of nucleolus-centromere localisation have fewer smFISH foci and the inverse, disruption of the nucleolus increases smFISH foci number as a whole. A model based on physical association would suggest that a nucleolus associated centromere produces less or no transcripts. Given that this is not a population-based assay, it should be possible to address this directly by analysing the location of individual centromeres and corresponding transcripts to strengthen the hypothesis. This could be done by either analysing the smaller subset of centromere-associated foci that colocalise with the smFISH signal and test whether the majority of these signals are proximal or distal to the nucleolus (this would not work or be less meaningful if the subpopulation is very small). Or doing a combined DNA/RNA FISH experiment. The expectation would be that DNA FISH signals of centromeres close to the nucleolus would not produce an RNA FISH signal somewhere else, and vice versa.

We predict that centromere-nucleolar associations are dynamic. Thus, we anticipate that centromeres would be associated transiently with the nucleolus (perhaps for a few hours), and that a given centromere would not be associated with the nucleolus in every cell at a specific time point. Thus, we believe that analyzing these behaviors across a diverse range of cells, as we did for this paper, is appropriate. In addition, technical considerations make these suggested experiments prohibitive. Defining the relationship between a centromere RNA and its originating centromere would require combined DNA and RNA FISH. The repetitive nature of alpha-satellite repeats and the strong similarity of these sequences between chromosomes makes it highly complex to visualize an individual centromere. Even if we were able to do this, the conditions required to simultaneously detect nucleoli (immunofluorescence), RNA (smFISH), and DNA (requires denaturation and hybridization) make this such that it would be complex to correlate the localization of an individual centromere with the levels of the corresponding alpha-satellite transcripts. In addition, these RNAs are likely to persist for an extended duration (possibly throughout the course of an entire cell cycle), such that they would not necessarily correlate with the current localization behavior of the centromere from which they are derived. For future work (beyond the scope of this paper), we plan to create cell lines expressing both centromere (CENP-A) and nucleolar markers (for example, Ki67) to conduct time lapse imaging to assess the dynamic associations between these structures.

7) At the end of the abstract, the authors conclude that the control of centromere transcription might be regulated by the centromere-nucleolar contacts to modulate chromatin dynamics. What does that really mean? One possibility they give in the discussion is rejuvenating centromeric chromatin. It would be nice if they could show some effect along those lines at the centromere in one of the manipulations they did (either through inhibiting or increase transcription). At least as discussed in the paper (Supp. Fig 3 D) it appears that overall levels of CENP-A are not affected. Is this different for newly loaded CENP-A? Or some other aspect of chromatin dynamics that is modulated? I realise that this might have been difficult to detect and therefore missing in the current study.

In a separate study from our lab as part of our recent work (Swartz et al. 2018), we found that CENP-A is gradually incorporated at centromeres in non-dividing quiescent cells, including non-transformed human Rpe1 cells and starfish oocytes. In the case of oocytes, which contain a substantial pool of mRNAs such that they do not require ongoing transcription for viability, we found that inhibiting RNA Polymerase II and preventing ongoing transcription blocked the incorporation of newly synthesized histones, including both canonical histone H3 and CENP-A. We realize that our description of this prior work was not sufficient to understand our integrated model, which relies on information from both papers. For the revised paper, we will update our discussion to better describe this data and present our model.

• Page 8: The authors state that as cells entered mitosis, dissociation of smFISH foci from chromatin was observed. While the absence of co-localisation of DAPI and smFISH signals is obvious in mitotic cells, what evidence is there that smFISH foci are chromatin associated in interphase nuclei? Rephrasing this bit might avoid confusion here. *

We appreciate this point. We did not mean to imply that the smFISH foci are bound to (or associate with) chromatin in the interphase nucleus. We will reword this as suggested.

Reviewer #3:

The manuscript of Bury et al. addresses how alpha-satellite transcription around centromeres is regulated. Using smFISH to detect alpha-satellite RNA transcripts, the authors find that alpha satellites are transcribed by RNA pol II, but their transcription is independent of centromeric proteins. In addition, they present evidence that nucleolar association represses alpha-satellite transcription. The data is convincing, solid and generally supports the conclusions. The manuscript includes appropriate control experiments, such as test for the validity of the RNA FISH probes. The manuscript is well-written and easy to follow, also for someone who is not directly an expert in the field.

The authors use a single-cell technique (smFISH) to look at the localization and transcription of alpha-satellite transcription from centromeres. The technical advance of this paper is limited, as smFISH is a well-established technique by now. Nevertheless, applying this single-cell approach to these repetitive regions has resulted in new insights regarding the regulation of alpha-satellite transcription, especially their localization of centromeres to nucleoli. Regarding the significance of these insights in the context of centromere biology/regulation and its literature is hard to evaluate for me, because this is not my field of expertise (my background is in single-cell transcription regulation). As a researcher from a related research field, I think the findings of this manuscript are mostly relevant for the direct research community of centromere and alpha-satellite biology, but not for researchers outside the field.

We appreciate these comments regarding the carefully controlled nature of our paper and the value of the advances for understanding alpha-satellite transcription. We also agree that smFISH is an established technique, although it has not been applied to these repetitive alpha-satellite sequences in prior work, allowing us to make important new observations usng the studies in this paper.

1. • The description of the inducible knock out cell lines is very limited. My main concern is how is checked that the gene is actually knocked out. I went back to the referenced paper, but it is still is not clear to me whether the new knockouts are sufficiently checked. It would be more convincing if the authors could show western blots or other evidence that their knockouts are working. In any case, the description of the knockout generation should be more elaborate. * This important point was also noted by the other reviewers. Please see our responses to Reviewer #1 point 4 and Reviewer #2 point 4. As described above, for a revised paper, we will provide an improved description of these knockout cell lines, our validation of these tools, and how we conducted the experiments in this paper.

• The authors nicely show that there is an inverse correlation between nucleolar association of the centromere and alpha-satellite transcription. The data supports this claim, but given the many knockouts and cell lines that were tested, with many intermediate phenotypes (such as CENP-B), I find the correlation based on 4 points a bit sparse. I would recommend filling up figure 4C with a few more mutants, to show that the inverse correlation holds for all mutants. These experiments would be straightforward for the authors, as the knockout/cell lines and techniques are already available. *

We see a compelling general correlation between the fraction of nucleolar-localized centromeres and alpha-satellite transcript levels. Our goal for Figure 4C was to highlight this correlation for a selected subset of conditions. However, we do not believe that there will be a precise linear correlation between transcript levels and nucleolar centromeres under every condition. Indeed, it is quite possible that some perturbations would affect transcript levels without altering nucleolar associations. This is particularly true for perturbations that cause subtle phenotypes. Systematically analyzing centromere-nucleolar co-localization for each of the knockouts represents a substantial undertaking that we do not feel would contribute substantially to this existing paper.

• The nucleolar repression is also supported by the Fibrillarin and Ki67 knockout. These are nice experiments which support their findings. What I am missing is whether these data quantitatively agree with the inverse correlation. Are these mutants completely lacking nucleoli, and if so, would you not expect both mutants to show the same upregulation? Similar to my point above, where do these mutants fall in the graph of figure 4C? *

For the perturbations described in this paper, we believe that inhibiting RNA Polymerase I most closely approximates the condition where nucleolar function is eliminated. Although Ki67 is a nucleolar protein in interphase, loss of Ki67 does not cause lethality indicating that nucleolar function is largely intact. We agree that it would be a good experiment to assess nucleolar-centromere associations in the Ki67 knockout. In fact, we have tried these experiments several times. However, due to the absence of Ki67 (for which we have the best localization tools), we instead needed to use Fibrillarin to monitor nucleoli. We have found this antibody to be much more finicky and not as readily compatible with the fixation conditions needed to detect centromeres. Thus far, we have not been able to generate clear data for this behavior.

• Related to this, since their imaging techniques have single-cell resolution, I wonder if cells that contain many centromeres in the nucleolus have less alpha satellite transcripts than cell with few centromeres. *

The correlation between centromere-nucleolar associations and alpha-satellite transcript numbers is strongly supported by our data across a population. However, analyzing this in individual cells is additionally complicated by the fact that we found that transcript levels vary over the cell cycle (low in G1, higher in S/G2). In addition, monitoring each of these markers in individual cells is technically complicated. Thus, while we appreciate this suggestion, we believe that our data stands on its own.

• One claim that is a bit speculative is the suggestion that transcription itself and not the RNA may be required for the function of the alpha-satellites. This is indeed supported by the fact that most transcripts are not localized at the centromeres. However, this contrasts to the findings of the papers that increasing alpha-satellite transcription in different mutants does not appear to result in any phenotype on centromere function. For a non-expert, the function of these transcripts/transcription itself is not clear from the current manuscript, so I would recommend discussing the nuances of its functions in more detail in the discussion. *

We agree that our model is speculative, but have chosen to include this to provide our perspective on the possible roles for centromere transcription based on this paper and our other recent work (Swartz et al. 2018). We believe that our data provide a context and set of constraints for potential roles of centromere transcription, but also agree that future work is needed to resolve these. Based on this comment and those from the other reviewers, we will also provide a better description of the data in the Swartz et al. paper, which analyzed different features of centromere transcription.

• To quantify the smFISH data, the authors count the number of foci. From the images, it looks like the different foci have very different intensities. This may occur if the transcripts are different length when transcribed from different genomic regions. However, this may also occur if several RNA co-localize to the same spot, i.e. if one spot contains several RNAs. Can the authors verify that the distribution of spot intensities matches the expected intensities based on the different transcribed alpha-satellite regions? *

Please see our response to Reviewer #2, point #3.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

The manuscript of Bury et al. addresses how alpha-satellite transcription around centromeres is regulated. Using smFISH to detect alpha-satellite RNA transcripts, the authors find that alpha satellites are transcribed by RNA pol II, but their transcription is independent of centromeric proteins. In addition, they present evidence that nucleolar association represses alpha-satellite transcription.

The data is convincing, solid and generally supports the conclusions. The manuscript includes appropriate control experiments, such as test for the validity of the RNA FISH probes. The manuscript is well-written and easy to follow, also for someone who is not directly an expert in the field.

Major comments:

The description of the inducible knock out cell lines is very limited. My main concern is how is checked that the gene is actually knocked out. I went back to the referenced paper, but it is still is not clear to me whether the new knockouts are sufficiently checked. It would be more convincing if the authors could show western blots or other evidence that their knockouts are working. In any case, the description of the knockout generation should be more elaborate.

The authors nicely show that there is an inverse correlation between nucleolar association of the centromere and alpha-satellite transcription. The data supports this claim, but given the many knockouts and cell lines that were tested, with many intermediate phenotypes (such as CENP-B), I find the correlation based on 4 points a bit sparse. I would recommend filling up figure 4C with a few more mutants, to show that the inverse correlation holds for all mutants. These experiments would be straightforward for the authors, as the knockout/cell lines and techniques are already available.

The nucleolar repression is also supported by the Fibrillarin and Ki67 knockout. These are nice experiments which support their findings. What I am missing is whether these data quantitatively agree with the inverse correlation. Are these mutants completely lacking nucleoli, and if so, would you not expect both mutants to show the same upregulation? Similar to my point above, where do these mutants fall in the graph of figure 4C?

Related to this, since their imaging techniques have single-cell resolution, I wonder if cells that contain many centromeres in the nucleolus have less alpha satellite transcripts than cell with few centromeres.

Minor comments

One claim that is a bit speculative is the suggestion that transcription itself and not the RNA may be required for the function of the alpha-satellites. This is indeed supported by the fact that most transcripts are not localized at the centromeres. However, this contrasts to the findings of the papers that increasing alpha-satellite transcription in different mutants does not appear to result in any phenotype on centromere function. For a non-expert, the function of these transcripts/transcription itself is not clear from the current manuscript, so I would recommend discussing the nuances of its functions in more detail in the discussion.

To quantify the smFISH data, the authors count the number of foci. From the images, it looks like the different foci have very different intensities. This may occur if the transcripts are different length when transcribed from different genomic regions. However, this may also occur if several RNA co-localize to the same spot, i.e. if one spot contains several RNAs. Can the authors verify that the distribution of spot intensities matches the expected intensities based on the different transcribed alpha-satellite regions?

Significance

The authors use a single-cell technique (smFISH) to look at the localization and transcription of alpha-satellite transcription from centromeres. The technical advance of this paper is limited, as smFISH is a well-established technique by now. Nevertheless, applying this single-cell approach to these repetitive regions has resulted in new insights regarding the regulation of alpha-satellite transcription, especially their localization of centromeres to nucleoli. Regarding the significance of these insights in the context of centromere biology/regulation and its literature is hard to evaluate for me, because this is not my field of expertise (my background is in single-cell transcription regulation). As a researcher from a related research field, I think the findings of this manuscript are mostly relevant for the direct research community of centromere and alpha-satellite biology, but not for researchers outside the field.

REFEREES CROSS COMMENTING

I agree with all the points raised. There is indeed a lot of overlap. The experiments should not take long in normal circumstances, but given the current situation, some extra time may indeed be required.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

The study by Bury et al. investigates the formation of two different types of alpha-satellite transcripts (ASAT, SF1 and 3) in different human cell lines. Using smFISH they find that during the cell cycle these centromeric transcripts don't stay at the centromere and are found in the cytoplasm after mitosis. Using specific inhibitors, they find that transcription is dependent on RNAPII, but not on various centromere and kinetochore proteins taking advantage of an inducible CRISPR-depletion system that the lab had previously developed. Interestingly, they find that CENP-C, a major component of the centromere and previously characterised as an RNA-binding protein, negatively regulates alpha-satellite transcript levels. Another regulator for transcript levels appears to be centromere-nucleolus interactions (as also indicated in the title) acting to suppress expression of these non-coding RNAs.

Major points

1) The authors state that the majority of smFISH foci do not colocalise with centromeres in a combined IF/FISH experiment (some quantification and a % of that subpopulation should be given somewhere). This is a bit concerning but of course could also be true. It either means that alpha-satellite transcripts leave the centromere as suggested by the authors (although some should be visible at the centromeres during the act of transcription). Alternatively, a trivial explanation would be that there is a lot of unspecific staining, which can occur in FISH-experiments to varying degrees. The RNase treatment to control for the absence of potential DNA hybridization is convincing, but the FISH probe could also interact with non-centromeric cellular RNA. With the centromere localisation as a reference point gone, some control is needed to validate that the RNA-FISH signals are indeed recognising alpha-satellite RNA that emerged from centromeres.¬ The authors could try competition experiments titrating unlabelled specific or unspecific DNA probes alongside their labelled specific FISH probe into their FISH experiment to see if they lose or maintain the signal and the number of foci. The specific RNA FISH probes could also be used in DNA FISH, to demonstrate they are working and recognising specific centromeres.

2) Apart from Figure 4, there is no analysis shown for statistical significance. This should be done for most if not all quantifications. Are indeed ASAT and antisense RNA Foci number not significantly different? The authors say that the levels of alpha-sat RNA in Rpe1 cells are not substantially different from other cell lines, but is it also not significant (Fig 1F)? In Figure 2D it is concluded that transcripts foci number are increased in S/G2 (from G1) and remain stable in mitosis, but it looks like there is an increase in mitosis. Again, it looks like the higher number of smFISH foci/Cell is significantly higher for both ASAT and SF1, so some statistical analysis would be required here.

3) Starting with the description of Figure 1E in the main text the paper equates foci count of smFISH per cell with RNA transcript levels. I'm not convinced that these are necessarily the same. You could have many weak foci or few very bright with the same amount of overall transcripts in both. The authors start out introducing smFISH as highly sensitive "for accurate characterisation of number ...of RNA transcripts". This suggests that foci intensity could be used as a read-out for transcript levels. It should be possible to measure individual intensity of the foci for a subset of images. Do foci intensity correlate or anti-correlate with foci numbers? Is the sum of the intensities of all the foci less variable than the foci number for an individual cell type?

4) I really like the use of the inducible CRISPR system to remove various centromere factors. However, some validation would be required to show that the system is effective in removing the proteins of interest in these experiments. For instance it would be helpful to show in Figure 3D an additional panel with CENP-C staining. Also for a subset of factors, some antibody staining co-staining with the smFISH could be provided in the supplemental material.

5) Since none of the CRISPR iKO has a particular inhibiting phenotype it would be useful to include some positive control in the CRISPR experiment. Would it be possible to use a CRISPR iKO target that affect some factor of the transcription machinery (RNA Pol II or similar) to reduce transcript levels?

6) The authors find a negative correlation between the nucleolus-centromere association and the number of alpha sat foci. This is really interesting and they suggest that the nucleolus association could negatively regulate centromere transcription. However, this correlation is rather indirect in the sense that cells with a higher-degree of nucleolus-centromere localisation have fewer smFISH foci and the inverse, disruption of the nucleolus increases smFISH foci number as a whole. A model based on physical association would suggest that a nucleolus associated centromere produces less or no transcripts. Given that this is not a population-based assay, it should be possible to address this directly by analysing the location of individual centromeres and corresponding transcripts to strengthen the hypothesis. This could be done by either analysing the smaller subset of centromere-associated foci that colocalise with the smFISH signal and test whether the majority of these signals are proximal or distal to the nucleolus (this would not work or be less meaningful if the subpopulation is very small). Or doing a combined DNA/RNA FISH experiment. The expectation would be that DNA FISH signals of centromeres close to the nucleolus would not produce an RNA FISH signal somewhere else, and vice versa.

7) At the end of the abstract, the authors conclude that the control of centromere transcription might be regulated by the centromere-nucleolar contacts to modulate chromatin dynamics. What does that really mean? One possibility they give in the discussion is rejuvenating centromeric chromatin. It would be nice if they could show some effect along those lines at the centromere in one of the manipulations they did (either through inhibiting or increase transcription). At least as discussed in the paper (Supp. Fig 3 D) it appears that overall levels of CENP-A are not affected. Is this different for newly loaded CENP-A? Or some other aspect of chromatin dynamics that is modulated? I realise that this might have been difficult to detect and therefore missing in the current study.

Minor points

Page 8: The authors state that as cells entered mitosis, dissociation of smFISH foci from chromatin was observed. While the absence of co-localisation of DAPI and smFISH signals is obvious in mitotic cells, what evidence is there that smFISH foci are chromatin associated in interphase nuclei? Rephrasing this bit might avoid confusion here.

Significance

This is overall a really interesting study and indeed, transcription at the centromere is little understood at this point. Given the importance of the centromere the findings in this manuscript will be of high interest to both researchers in the field and a general audience. There are novel and interesting insights into centromeric transcripts but the study still requires some controls.

REFEREES CROSS COMMENTING

I agree with the comments of the other reviewers. I appears that there is a lot of overlap between the referees regarding the exciting parts and those that raise concerns. In particular I share the view of Reviewer 1 on the imporantance of validating the FISH probes and the knockouts (also raised by R3) and the concern that a 24h transcription inhibition is prone to secondary effects. I would agree with both that less time might be required to complete revisions (may be 1-3 months) but was factoring in some extra time for wet experiments which likely take longer under the current conditions.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary

In this manuscript the authors explore the requirements for centromere transcription using single-molecule FISH. Previous studies have found that centromeres are transcriptionally active in a wide variety of organisms. Centromere transcription has been proposed to facilitate Cenp-A deposition through chromatin remodeling and to directly contribute to centromere/kinetochore function by producing a functional ncRNA. However, we currently know almost nothing about how transcription is initiated at the centromere or how levels of centromere transcripts are controlled. This manuscript makes several major findings that are potentially of importance to groups studying centromere transcription. 1.) Centromere RNAs are produced by RNA Polymerase II and are localized in the nucleus of a wide-range of cell types. 2.) Centromere RNAs do not localize to the centromere, which is in contrast to several recent studies. 3.) Centromere proteins are not required for transcription of alpha-satellite sequences. 4.) Localization of centromeres to the nucleolus represses centromere transcription. Overall, this is a solid manuscript and has the potential to make a significant impact in the field. Below I suggest a couple of experiments and modification to the data presentation that could improve the manuscript.

Major points:

1.All of the experiments in this manuscript rely on detection of centromere RNAs using single molecule FISH probes. These probes are validated by showing the RNase treatment removes the FISH signal. A strength of this approach is that the authors use multiple different probe sets and achieve comparable results. However, there is no orthogonal validation that the probes detect alpha satellite RNA. All of the experiments in this manuscript would be significantly improved by showing that the results presented here can be confirmed by a different approach. I suggest that the authors use Q-RT-PCR to validate the smFISH results.

2.Several results in this manuscript directly contradict results in published studies, but these discrepancies are not discussed. I believe the authors need to discuss the following discrepancies between their results and those in the literature:

a. NcNulty et al. Dev. Cell. 2017. Show that alpha-satellite RNA is transcribed from all centromeres and remains localized to the site of transcription. The different results and possible explanations for the differences should be discussed.

b. Additionally, Rosic et al. JCB 2014, Blower Cell Reports 2016 and Bobkov et al. JCB 2018 all show that centromere RNAs localize to centromere regions. The differences between these studies and the authors results should be discussed.

c. The authors show that satellite RNA cannot be detected on mitotic chromosomes. However, Johnson et al. Elife 2017, Bobkov et al. JCB. 2018, and Perea-Resa et al. Mol. Cell. 2020 show that EU-labeled RNA can be detected at the centromere during mitosis. The authors should discuss the discrepancy between their results and these studies. Is it possible that their smFISH probes do not detect nascent, chromatin-bound transcripts?

d. The authors show nicely that deletion of Ki-67 reduces centromere localization to the nucleolus and increases centromere transcription. However, this has no effect on centromere function. Studies from the Earnshaw lab (e.g. Nakano et al. Dev Cell 2008 and Bergmann et al. EMBO J. 2011) show that increasing or decreasing centromere transcription results in loss of kinetochore function on a human artificial chromosome. The authors should discuss the differences between their results and these studies. Is it possible that the small size of the HAC exaggerates the importance of the correct levels of centromere transcription?

Minor Points

1.The authors treat cells with transcriptional inhibitors for 24 hours. I am concerned that this may result in massive cell death. It would be helpful to include cell viability data from these experiments.

2.In Figure 3C the authors examine the effects of centromere protein knock outs on centromere transcription. To me this is the most important experiment in the manuscript and is a major step forward for the field. The authors use inducible CRISPR knock out cell lines that are not 100% penetrant. It would be helpful if the authors could describe how they ensured that cells included in the image quantification were knock out cells.

3.On p8. The authors cite Quenet and Dalal. eLife 2014 for the idea that transcription during G1 is important for new Cenp-A loading. They should also cite Chen et al. Dev. Cell 2015 and Bobkov et al. JCB. 2018.

Significance

Describes the requirements of transcription of centromere RNAs. Identifies factors that regulate centromere transcription.

Audience: centromere biologists.

REFEREES CROSS COMMENTING

I agree with all the concerns raised by the other reviewers. I think that all three reviews taken together are a fair and constructive review of this manuscript.

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

Learn more at Review Commons

Reply to the reviewers

Reply to the Reviewers

We thank the reviewers for their thoughtful comments and suggestions how to improve our manuscript. Most of the remarks are now addressed in a new version of the manuscript with modifications marked in blue. We found especially interesting the idea to explore the changes in dynamics of microtubules that make up bridging fibers, which we will do in revision. In addition, we will perform Western blot analysis of PRC1 and acquire better images of cells with SiR-DNA for Fig. 2 A.

** Major issues: **

Reviewer #1:

The use of blue light necessary to relocate opto-PRC1 from the spindle to the membrane is a concern, specially given the strongest phenotype associated with acute vs. constitutive inactivation of PRC1. While these differences may indeed reflect distinct cellular adaptation responses to each procedure, the authors must rule out that phototoxicity caused by blue light (e.g. see Douthwright S, Sluder G. Live Cell Imaging: Assessing the Phototoxicity of 488 and 546 nm Light and Methods to Alleviate it. J Cell Physiol. 2017 Sep;232(9):2461-2468. doi: 10.1002/jcp.25588. PubMed PMID:27608139) is not responsible for the observed stronger phenotypes. A control of U2OS cells expressing the centromere marker (without opto-PRC1) in metaphase after exposure to the same blue light regimen (i.e. 200 ms every 10 sec for 20 min and same laser power) should be provided.

Response: We thank the reviewer for raising this important point. We added the suggested experiments on U2OS cells without opto-PRC1 filmed with same blue light regimen, and updated Fig. S2 A-C, E to contain also the new measurements of inter-kinetochore distance (dKC), distance from equatorial plane (dEQ), corresponding time-lapse images, and angle between sister kinetochore axis and spindle long axis (αKC), respectively. We added the following text in Results: “The effects of PRC1 removal were found neither in control experiments without iLID, nor in a different set of control experiments where cells without opto-PRC1 and without iLID were exposed to the same laser illumination protocol (Fig. 2 B,E,F,H; Fig. S2, A-C, E), suggesting that the observed effects were not a consequence of laser photodamage (Douthwright and Sluder, 2017).”

Reviewer #1:

I could not find in the manuscript whether opto-PRC1 is RNAi resistant. I would assume so, as the authors are targeting the 3'-UTR of endogenous PRC1, but at least a western blot should be provided: 1) to properly ascertain depletion efficiency of the endogenous protein; and 2) the levels of opto-PRC1 after depletion.

Response: We added a note in Methods that opto-PRC1 is RNAi-resistant. We have assessed the depletion efficiency of the endogenous PRC1 and the levels of opto-PRC1 after depletion by using immunofluorescence of PRC1 on the spindle (Fig. S1 A and B). Additionally, we will perform Western blot analysis to show depletion efficiency of endogenous PRC1 and the levels of transfected opto-PRC1 after depletion of endogenous PRC1. However, as we observed that the efficiency of opto-PRC1 plasmid transfection is low, Western blot analysis may provide biased levels of PRC1 in the complete population, not specific to the analyzed opto cells.

Reviewer #1:

One aspect related with data interpretation and the proposed model: if PRC1 selectively bundles anti-parallel microtubules, how could it mechanically couple sister k-fibers that are made of parallel MTs? This should be explained in detail, ideally supported by data.

Response: This is an important issue, which we now explain in detail in Discussion: “As midzone-crossing microtubules associate with k-fibers on either side of the metaphase plate (O'Toole et al., 2020), PRC1 and probably also other microtubule-associated proteins crosslink antiparallel overlaps between k-fiber microtubules extending from one pole and bridging microtubules extending from the opposite spindle half, as well as antiparallel overlaps within the bridging fiber.”

Reviewer #1:

The author should find a way to unequivocally demonstrate that opto-PRC1 is fully functional and can rescue depletion of endogenous PRC1. The fact that recovery of PRC1 on spindles never fully rescue spindle architecture and chromosome properties might indicate that opto-PRC1 is not fully functional. For example, can it rescue anaphase or cytokinesis roles of PRC1?

Response: To demonstrate functionality of opto-PRC1, we added images of cell's progression to cytokinesis in both control and opto cells in new Fig. S1D and added the following text to Results: “Importantly, after exposure to the blue light, opto cells were able to progress to cytokinesis (Fig. S1 D)”. Furthermore, as PRC1's major binding partners, Kif4A and MKLP1 (Fig. S4A, and new Fig. S4J, respectively), which depend on its localization in the spindle midzone in anaphase, are found to co-localize with opto-PRC1 in anaphase, opto-PRC1 is fully functional and rescues depletion of endogenous PRC1. We added the following text to Results: “In anaphase, MKLP1 also co-localized with opto-PRC1 in the spindle midzone (Fig. S4 J) (Gruneberg et al., 2006; Kurasawa et al., 2004)”.

** Minor issues: **

Reviewer #1:

Abstract: the authors introduce the problem by stating that chromosome position at the spindle equator is mainly regulated by forces by kMTs. We do not know this, actually there is evidence in the literature that kif4a on chromosome arms is required to maintain chromosomes aligned by exerting forces on ipMTs (e.g. Wandke et al., JCB, 2012). Along the same line, there is evidence from the Dumont lab that sister k-fibers are not mechanically coupled. These alternative views should be discussed and taken into account when formulating the problem under investigation in the present study.

Response: We changed the sentence in Abstract to include polar ejection forces: “During metaphase, chromosome position at the spindle equator is regulated by the forces exerted by kinetochore microtubules and polar ejection forces”. When formulating the problem in Introduction, we discuss polar ejection forces and cite Wandke et al., 2012, and several other papers. We also discuss the findings about PRC1-mediated coupling of sister k-fibers from the Dumont lab in relation to our local effect of PRC1 removal on a fraction of sister kinetochore pairs: “This local effect is in line with weak mechanical coupling between neighboring k-fibers, yet strong coupling between sister k-fibers (Elting et al., 2017; Suresh et al., 2020).” In addition, we mention the Dumont lab results when we suggest that the persistent misorientation of kinetochores after PRC1 return to the spindle is due to perturbed overlap geometry during the absence of PRC1: “This is in agreement with a recent finding that PRC1 restricts pivoting of k-fibers near kinetochores by promoting tight coupling between sister k-fibers (Suresh et al., 2020).”

Reviewer #1:

The authors refer to kinetochore alignment or lagging kinetochores throughout the text. Although this is unquestionable, it might be more appropriate to refer to chromosome alignment or lagging chromosomes instead, as this is the object to me moved.

Response: We agree with the reviewer and changed this at several places throughout the text.

Reviewer #1:

page 2: "...PRC1 regulates forces acting on kinetochores". The authors should mention that this would be indirect, as PRC1 is not at kinetochores itself.

Response: This is true and therefore we added the word indirectly in this sentence.

Reviewer #1:

page 6: "PRC1 removal did not activate the spindle assembly checkpoint". Although this might be considered semantics, given that the SAC is constitutively active and needs to be satisfied, the authors might adopt a more accurate description such as "PRC1 removal did not prevent spindle assembly checkpoint satisfaction".

Response: We changed the sentence into the suggested one.

Reviewer #1:

page 13: the authors mention about the localization of Kif18a on bridging fibers. Was this known? From the images it is unclear if we are looking at bridging fibers or k-fibers. Co-localization with PRC1 would help clarifying this issue. If indeed associated with bridging fibers, this would raise an alternative interpretation of how Kif18a contributes to maintain chromosome alignment.

Response: We thank the reviewer for raising this important point. Kif18A localization in the bridge is a new observation, and to make it clearer we introduced merged images where both Kif18A and PRC1 are shown during optogenetic experiment (Fig. S4E) and four examples of enlarged regions around kinetochores with Kif18A-GFP to show its localization in the bridging fiber in mock treated cells and its lack of localization in the bridging fiber after PRC1 siRNA (Fig. S4F). We also added a discussion of a new potential role of Kif18A (and Kif4A and MKLP1) in chromosome alignment: “Interestingly, we found that Kif4A, MKLP1, and Kif18A localize in the bridging fibers in metaphase and this localization was lost after optogenetic or siRNA-mediated PRC1 removal. During anaphase, the PRC1-dependent Kif4A and MKLP1 in the bridging fibers are involved in sliding of antiparallel microtubules to elongate the spindle (Vukušić et al., 2019). Kif4A and MKLP1 may have a similar role in metaphase, and thus Kif4A removal from the bridging fibers induced by PRC1 removal may affect chromosome alignment by affecting microtubule sliding in the bridging fiber. This possibility is in agreement with previous work showing that Kif4A depletion reduces microtubule flux (Wandke et al., 2012). Similarly, Kif18A in the bridging fiber may have microtubule-sliding and crosslinking activities similar to those of the yeast kinesin-8 (Su et al., 2013), which may promote chromosome alignment. The roles of these and other motors within bridging fibers in chromosome alignment will be an intriguing topic for future studies.”

** Major concerns: **

Reviewer #2:

Regarding lagging chromosomes: Page 6 and Fig. 2F: "Kinetochore remains displaced even after opto-PRC1 return": Why is this? The reasoning in the discussion is not clear/convincing. Is it possible that these irreversible changes reflect light-induced deactivation of protein? Or, could these irreversible changes arise from a perturbation in the structure of microtubules at the end of the 'light' period? Discussion or additional supportive evidence to address this will be helpful.

Response: We thank the reviewer for raising this question. As we have not observed these effects in control cells with opto-PRC1 and without iLID that relocates opto-PRC1 to the membrane, we do not find light-induced deactivation of opto-PRC1 likely. We find the latter possibility more realistic, thus we added the following text to Discussion: “Kinetochore positions and orientations did not revert to the initial values within 10 min of PRC1 return. We speculate that upon PRC1 removal the geometry of the overlap structures is perturbed due to a change in the force balance in the spindle. When PRC1 returns to the perturbed overlaps, it likely confines the chromosomes in new positions and orientations. This is in agreement with a recent finding that PRC1 restricts pivoting of k-fibers near kinetochores by promoting tight coupling between sister k-fibers (Suresh et al., 2020).”

Reviewer #2:

The correlation between misaligned kinetochores and lagging chromosomes is not clear. Are lagging chromosomes more frequently attached to kinetochores that show high deq values (Fig. S2G) in metaphase?

Response: This is an interesting point. To clarify our results, we rewrote the text: “Opto cells that showed lagging kinetochores in anaphase had a slightly smaller inter-kinetochore distance before anaphase than opto cells without lagging kinetochores, but we did not find correlation between lagging kinetochores in anaphase and kinetochore misalignment in metaphase (Fig. S2 H).” Fig. S2 H is the old Fig. S2 G. Please note that we were not able to backtrack individual lagging kinetochores to metaphase to see if they had a higher value of d_eq. Instead, we measured mean d_eq of all kinetochores in opto cells that had a lagging chromosome and in those that did not.

Reviewer #2:

Regarding the contribution of other motors: Looking at the contributions of various other microtubule-associated proteins in accounting for effects of PRC1 removal is a good addition to the paper (Fig. 4). However, the consequences of the depletion of Kif18A and MKLP1 from the bridging fiber are not elaborated upon. Is the presence of these motors at the bridging fiber functionally important? It would be good to incorporate their known activity in the final model for how PRC1-crosslinked fibers align chromosomes. In particular, a recent biorxiv submission from this group has a thorough examination of the consequences of motor removal in anaphase, and perhaps some of their findings and other literature can be used to draw some insights into if and how the presence of these motors on PRC1-crosslinked fibers contribute to chromosome alignment.

Response: We thank the reviewer for this important idea, which we now elaborate in Discussion: “Interestingly, we found that Kif4A, MKLP1, and Kif18A localize in the bridging fibers in metaphase and this localization was lost after optogenetic or siRNA-mediated PRC1 removal. During anaphase, the PRC1-dependent Kif4A and MKLP1 in the bridging fibers are involved in sliding of antiparallel microtubules to elongate the spindle (Vukušić et al., 2019). Kif4A and MKLP1 may have a similar role in metaphase, and thus Kif4A removal from the bridging fibers induced by PRC1 removal may affect chromosome alignment by affecting microtubule sliding in the bridging fiber. This possibility is in agreement with previous work showing that Kif4A depletion reduces microtubule flux (Wandke et al., 2012). Similarly, Kif18A in the bridging fiber may have microtubule-sliding and crosslinking activities similar to those of the yeast kinesin-8 (Su et al., 2013), which may promote chromosome alignment. The roles of these and other motors within bridging fibers in chromosome alignment will be an intriguing topic for future studies.”

Reviewer #2:

Page 13 and Fig. 4A & 4B: "The localization of Kif18A in the bridge was perturbed by both acute and long-term PRC1 removal." However, this is not apparent from Figures 4A and 4B. It would be helpful to clarify how this interpretation was made from the data in the figure.

Response: We thank the reviewer for pointing this out. To clarify this issue, we added merged-channel images of the cell from Fig. 4 A to Fig. S4 E to show colocalization of Kif18A and PRC1. Moreover, we added enlargements from spindles without and with PRC1 depletion in Fig. S4 F to show presence and absence of Kif18A in the bridging fibers, respectively. To clarify how the images were evaluated, we added the following text in Methods: “Localization test of GFP-Kif4A or Kif4A-GFP, MKLP1-GFP, Kif18A-GFP, EGFP-CLASP1, and CENP-E-GFP in the bridging fibers of either opto cells or cells treated with mock siRNA or PRC1 siRNA was performed by visually inspecting the GFP signal through the z-stack, in the region where PRC1-labeled fibers were found, i.e., in the region that spans between sister kinetochores and continues ~2 µm laterally from sister kinetochores.”

Additional suggested experiment and analysis:

Reviewer #2:

One factor that could potentially contribute to the changes in chromosome alignment and increase in lagging chromosomes upon PRC1 removal, is changes in dynamics of microtubules that make up bridging fibers. This may also provide insights on the role of associated proteins. One possible experiment is to look at tubulin turnover in the bundles (example by FRAP). Another alternative possibility is to examine EB3 comets in the presence and absence of PRC1 (note: these are just some potential suggestions; the authors may have other ways of addressing the question). Examining the dynamics would help in addressing if bridging fibers is dynamically remodeled through metaphase and early anaphase and whether the loss of PRC1 causes a change in the dynamics of these microtubules.

Response: We thank the reviewer for this exciting suggestion. We will perform experiments to examine EB3 comets (their numbers and velocities) in the bridging fibers in the presence and absence of PRC1.

Reviewer #2:

Do kinetochores oscillate / fluctuate about the metaphase plate over time? Does the absence of PRC1 affect these fluctuations? Since the authors already have the data (Fig. 2), they can track the trajectories of sister kinetochore displacement from the equatorial plane as a function of time from prometaphase on, both in the presence and absence of PRC1. This analysis will be informative in understanding how kinetochores and bridging fibers act together to maintain force balance in the spindle and how misalignments are corrected.

Response: This is an interesting point. We added the following results: “Kinetochore displacement was not a result of higher oscillation amplitude because kinetochores fluctuated to a similar extent in the presence and absence of opto-PRC1, but in its absence the displaced kinetochores fluctuated within a region that was offset from the equatorial plane (Fig. S2 D).”

** Minor points: ** Reviewer #2:

Is the number of microtubules that make up the bridging fiber the same for outermost and inner kinetochores?

Response: This is an interesting question. Even though we could not measure this here, we suggest that the outermost bridges may have more microtubules: “The misaligned kinetochores were found in the inner part of the spindle, where PRC1 signal disappeared faster than on the outer part, which indicates that the inner bridging fibers were more severely affected by PRC1 removal and/or that they are made up of fewer microtubules than the outer bridges.”

Reviewer #2:

The quantity dax that is plotted in fig. S2F has not been defined in the text.

Response: We now define dAX in the caption of Fig. S2 F: “Graphs show aKC versus corresponding dEQ and the distance from the midpoint between sister kinetochores to the long spindle axis, dAX (left), ...”

Reviewer #2:

Discussion of these findings in the context of recent work from the lab of Sophie Dumont will be interesting (Suresh et al. eLife 2020;9:e53807).

Response: We thank the reviewer for reminding us to discuss this highly relevant recent paper by the Dumont lab. We included a discussion of the findings about PRC1-mediated coupling of sister k-fibers in relation to our local effect of PRC1 removal on a fraction of sister kinetochore pairs: “This local effect is in line with weak mechanical coupling between neighboring k-fibers, yet strong coupling between sister k-fibers (Elting et al., 2017; Suresh et al., 2020).” In addition, we mention these results when we suggest that the persistent misorientation of kinetochores after PRC1 return to the spindle is due to perturbed overlap geometry during the absence of PRC1: “This is in agreement with a recent finding that PRC1 restricts pivoting of k-fibers near kinetochores by promoting tight coupling between sister k-fibers (Suresh et al., 2020).”

Reviewer #3:

Some of the images are sub-optimal. For example Fig 2A, there doesn't seem to be much/any PRC1 on the spindle in the "Dark 0 min" condition, although some is visible after the reversal. Do the authors have a better example to show here? In Figures 1 and 2 we can see the removal clearly yet in later images, the spindle is zoomed such that the relocation cannot be observed.

Response: We agree that PRC1 is not properly visible in Fig. 2 A and we will do new experiments to obtain better images. Regarding the later images in Figs. 3 and 4 that are zoomed, they are displayed in this manner to show the localization of proteins in the bridging fiber and/or at the ends of kinetochore fibers. In these experiments, the removal of PRC1 was the same as in earlier images, which is visible in examples shown in Figs. S3 and S4.

Reviewer #3:

Have the authors looked at whether the cells progress normally after removal and reversal of PRC1? In the paper the authors describe how the knockdown and re-expression of opto-PRC1 does not interfere with mitotic progression, but we wondered whether cells recover after the optogenetic operation, compared to a control with similar illumination.

Response: We found that cells were able to progress to cytokinesis and added an example to Fig. S1 D and the following text to Results: “Importantly, after exposure to the blue light, opto cells were able to progress to cytokinesis (Fig. S1 D).”

Reviewer #3:

Is there a reason why no Dark-state images are shown in Fig 2C and I?

Response: We swapped those images with Dark-state images.

Reviewer #3:

For some of the plots, the y-axis is not shown scaled from 0. This is misleading because it exaggerates differences. Examples are 2E,F,G,H, 3G,J, S3E,G,H, S5C,D,E,F,G.

Response: We agree and we now show graphs with the y-axis starting at 0 for Fig. 2 F,H, Fig. S2 B,E, Fig. 3 G,J, Fig. S3 G, and Fig. S5 C,F,G. However, we did not change the graphs for d_kc, theta, spindle length and width, and widths of bridging and k-fibers, because these values span a rather narrow range, which is far from zero. Because the differences are statistically significant, we chose a scale at which they can be easily visualized.

Reviewer #3:

In the legend, formulae should be written in correct notation.

Response: Corrected: Formulae y=A*exp(-τ*x) and y=A*exp(-τ*x)+c were used for opto-PRC1 removal and return, respectively.

Reviewer #3:

In Fig 2 legend it says that a 0.5-pixel-radius Gaussian blur is applied. Doesn't the kernel for transformation result in an identity matrix?

Response: To clarify this, we replaced “0.5-pixel-radius Gaussian blur” with “0.5-pixel-sigma Gaussian blur” in figure captions and added the following to Methods: “To remove high frequency noise in displayed images a Gaussian blur filter with a 0.5-pixel sigma (radius) was applied where stated”.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

This manuscript by Jagrić et al. shows a role for PRC1 at metaphase by using an optogenetic method to rapidly remove PRC1 from bridging fibres in the mitotic spindle. They show that this methodology (which uses light to relocate PRC1 temporarily on the plasma membrane) is superior to long-term depletion by siRNA and, because it is reversible, has advantages over chemically-induced protein translocation. They put the method to use to examine PRC1's role in bridging fibres the results are consistent with siRNA approaches but cleaner due to the acute nature of the method. Overall the paper is convincing and is likely to be of interest to cell biologists working on mitosis.

We have only minor comments that can be easily addressed during the current crisis. Note that we covered this paper in our lab journal club when it went up on bioRxiv and our comments in that pre-pandemic time were the same as now.

1.Some of the images are sub-optimal. For example Fig 2A, there doesn't seem to be much/any PRC1 on the spindle in the "Dark 0 min" condition, although some is visible after the reversal. Do the authors have a better example to show here? In Figures 1 and 2 we can see the removal clearly yet in later images, the spindle is zoomed such that the relocation cannot be observed.

2.Have the authors looked at whether the cells progress normally after removal and reversal of PRC1? In the paper the authors describe how the knockdown and re-expression of opto-PRC1 does not interfere with mitotic progression, but we wondered whether cells recover after the optogenetic operation, compared to a control with similar illumination.

3 Is there a reason why no Dark-state images are shown in Fig 2C and I?

4.For some of the plots, the y-axis is not shown scaled from 0. This is misleading because it exaggerates differences. Examples are 2E,F,G,H, 3G,J, S3E,G,H, S5C,D,E,F,G

5.In the legend, formulae should be written in correct notation.

6.In Fig 2 legend it says that a 0.5-pixel-radius Gaussian blur is applied. Doesn't the kernel for transformation result in an identity matrix?

Significance

We thought the paper is likely to be of significant interest to cell biologists working on mitosis.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Jagric and colleagues adapt an optogenetic method for acute and reversible removal of spindle associated proteins to the cell membrane. They apply this technique to deplete the microtubule crosslinking protein PRC1 from the metaphase spindle with high temporal accuracy. They establish that the spindle localization of PRC1 can be perturbed in a fast and reversible manner, on a timescale of minutes, using this method.

Next, they use this system to show that acute depletion of PRC1, which has previously been shown to localize to the bridging fibers that link kinetochore pairs. They find that PRC1 depletion modestly disrupts chromosome alignment on the metaphase plate and results in an increased frequency of lagging kinetochores during anaphase. The advantage of the optogenetic system is that they can look at the reversibility and compare the effects of acute depletion to long-time course methods such as RNAi. This comparison is well done and well presented in the paper. The authors further probe the mechanism underlying the defects associated with PRC1 depletion and find a decrease in the number of microtubules that make up a bridging fiber. The localization of other proteins to the kinetochores are not affected by the removal of PRC1, but the localization of Kif18a and MKLP1 to bridging fibers is disrupted. Together, the authors propose a model where the movement of bi-oriented chromosomes is restricted to the region containing PRC1-crosslinked bridging fibers, and this buffering is important in maintaining chromosome alignment.

Overall, the paper is well written, and the schematics, figures and descriptions of experiments are easy to follow. The microscopy experiments and data analysis are carefully performed and thorough, and the representation of data in figures and tables is very clear. The comparison between RNAi and opto-depletion has been well executed and a great addition. The advance in this paper is the establishment of an optogenetics system to selectively and reversibly perturb PRC1. While the method is not novel (Guntas et al., 2015), its development and application to a spindle protein will be of interest to researchers in the field, and I expect this work to be a major resource in that regard. I am less enthusiastic about the biological findings as the effects of PRC1-removal from the bridging fiber are modest. In addition, some effects, such as kinetochore misalignment and decrease in the number of microtubules in the bridging fiber, are not reversible, which raises some concerns about whether these effects are directly mediated by specific protein depletion. I have outlined my specific comments below:

Major concerns:

1 . Regarding lagging chromosomes:

•Page 6 and Fig. 2F: "Kinetochore remains displaced even after opto-PRC1 return": Why is this? The reasoning in the discussion is not clear/convincing. Is it possible that these irreversible changes reflect light-induced deactivation of protein? Or, could these irreversible changes arise from a perturbation in the structure of microtubules at the end of the 'light' period? Discussion or additional supportive evidence to address this will be helpful.

•The correlation between misaligned kinetochores and lagging chromosomes is not clear. Are lagging chromosomes more frequently attached to kinetochores that show high deq values (Fig. S2G) in metaphase?

2 . Regarding the contribution of other motors

•Looking at the contributions of various other microtubule-associated proteins in accounting for effects of PRC1 removal is a good addition to the paper (Fig. 4). However, the consequences of the depletion of Kif18A and MKLP1 from the bridging fiber are not elaborated upon. Is the presence of these motors at the bridging fiber functionally important? It would be good to incorporate their known activity in the final model for how PRC1-crosslinked fibers align chromosomes. In particular, a recent biorxiv submission from this group has a thorough examination of the consequences of motor removal in anaphase, and perhaps some of their findings and other literature can be used to draw some insights into if and how the presence of these motors on PRC1-crosslinked fibers contribute to chromosome alignment.

•Page 13 and Fig. 4A & 4B: "The localization of Kif18A in the bridge was perturbed by both acute and long-term PRC1 removal." However, this is not apparent from Figures 4A and 4B. It would be helpful to clarify how this interpretation was made from the data in the figure.

Additional suggested experiment and analysis:

1 . One factor that could potentially contribute to the changes in chromosome alignment and increase in lagging chromosomes upon PRC1 removal, is changes in dynamics of microtubules that make up bridging fibers. This may also provide insights on the role of associated proteins. One possible experiment is to look at tubulin turnover in the bundles (example by FRAP). Another alternative possibility is to examine EB3 comets in the presence and absence of PRC1 (note: these are just some potential suggestions; the authors may have other ways of addressing the question). Examining the dynamics would help in addressing if bridging fibers is dynamically remodeled through metaphase and early anaphase and whether the loss of PRC1 causes a change in the dynamics of these microtubules.

2 . Do kinetochores oscillate / fluctuate about the metaphase plate over time? Does the absence of PRC1 affect these fluctuations? Since the authors already have the data (Fig. 2), they can track the trajectories of sister kinetochore displacement from the equatorial plane as a function of time from prometaphase on, both in the presence and absence of PRC1. This analysis will be informative in understanding how kinetochores and bridging fibers act together to maintain force balance in the spindle and how misalignments are corrected.

Minor points:

1 . Is the number of microtubules that make up the bridging fiber the same for outermost and inner kinetochores?

2 . The quantity dax that is plotted in fig. S2F has not been defined in the text.

3 . Discussion of these findings in the context of recent work from the lab of Sophie Dumont will be interesting (Suresh et al. eLife 2020;9:e53807).

Significance

The advance in this paper is the establishment of an optogenetics system to selectively and reversibly perturb PRC1. While the method is not novel (Guntas et al., 2015), its development and application to a spindle protein will be of interest to researchers in the field, and I expect this work to be a major resource in that regard. I am less enthusiastic about the biological findings as the effects of PRC1-removal from the bridging fiber are modest.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

The manuscript by Jagric et al. investigates the role of PRC1 in the maintenance of chromosome alignment at the spindle equator using acute inactivation by optogenetic control. This is an elegant system based on the iLID system that recruits a protein of interest to the cell membrane in a reversible way. Accordingly, acute removal of PRC1 resulted in reduction of bridging fibers and decreased inter-kinetochore distances, while widening the metaphase plate and increasing the frequency of lagging chromosomes in anaphase. The authors investigate whether acute PRC1 removal from bridging fibers compromise other proteins and conclude that PRC1 acts essentially by coupling bridging and kinetochore fibers. They propose that PRC1 uses this role to buffer kinetochore movements to promote chromosome alignment. Overall, this is a very high quality study that adds to our knowledge about the roles of PRC1 and bridging fibers in spindle mechanics and will be of interest to a specialized readership of mitosis researchers. Nevertheless, there are still few remaining issues, mostly concerning additional controls and interpretations that should be addressed prior to publication.

Major issues:

1- The use of blue light necessary to relocate opto-PRC1 from the spindle to the membrane is a concern, specially given the strongest phenotype associated with acute vs. constitutive inactivation of PRC1. While these differences may indeed reflect distinct cellular adaptation responses to each procedure, the authors must rule out that phototoxicity caused by blue light (e.g. see Douthwright S, Sluder G. Live Cell Imaging: Assessing the Phototoxicity of 488 and 546 nm Light and Methods to Alleviate it. J Cell Physiol. 2017 Sep;232(9):2461-2468. doi: 10.1002/jcp.25588. PubMed PMID:27608139) is not responsible for the observed stronger phenotypes. A control of U2OS cells expressing the centromere marker (without opto-PRC1) in metaphase after exposure to the same blue light regimen (i.e. 200 ms every 10 sec for 20 min and same laser power) should be provided.

2- I could not find in the manuscript whether opto-PRC1 is RNAi resistant. I would assume so, as the authors are targeting the 3'-UTR of endogenous PRC1, but at least a western blot should be provided: 1) to properly ascertain depletion efficiency of the endogenous protein; and 2) the levels of opto-PRC1 after depletion.

3- One aspect related with data interpretation and the proposed model: if PRC1 selectively bundles anti-parallel microtubules, how could it mechanically couple sister k-fibers that are made of parallel MTs? This should be explained in detail, ideally supported by data.

4- The author should find a way to unequivocally demonstrate that opto-PRC1 is fully functional and can rescue depletion of endogenous PRC1. The fact that recovery of PRC1 on spindles never fully rescue spindle architecture and chromosome properties might indicate that opto-PRC1 is not fully functional. For example, can it rescue anaphase or cytokinesis roles of PRC1?

Minor issues:

1- Abstract: the authors introduce the problem by stating that chromosome position at the spindle equator is mainly regulated by forces by kMTs. We do not know this, actually there is evidence in the literature that kif4a on chromosome arms is required to maintain chromosomes aligned by exerting forces on ipMTs (e.g. Wandke et al., JCB, 2012). Along the same line, there is evidence from the Dumont lab that sister k-fibers are not mechanically coupled. These alternative views should be discussed and taken into account when formulating the problem under investigation in the present study.

2- The authors refer to kinetochore alignment or lagging kinetochores throughout the text. Although this is unquestionable, it might be more appropriate to refer to chromosome alignment or lagging chromosomes instead, as this is the object to me moved.

3- page 2: "...PRC1 regulates forces acting on kinetochores". The authors should mention that this would be indirect, as PRC1 is not at kinetochores itself.

4- page 6: "PRC1 removal did not activate the spindle assembly checkpoint". Although this might be considered semantics, given that the SAC is constitutively active and needs to be satisfied, the authors might adopt a more accurate description such as "PRC1 removal did not prevent spindle assembly checkpoint satisfaction".

5- page 13: the authors mention about the localization of Kif18a on bridging fibers. Was this known? From the images it is unclear if we are looking at bridging fibers or k-fibers. Co-localization with PRC1 would help clarifying this issue. If indeed associated with bridging fibers, this would raise an alternative interpretation of how Kif18a contributes to maintain chromosome alignment.

Significance

If additional controls are provided, this manuscript represents a significant technical advance in the study of PRC1 function. The results however are just incremental relative to previous state-of-the-art and will be of interest to more specialized researchers working on mitosis and spindle architecture. The concept of buffer for kinetochore movement is interesting, but how exactly PRC1 contributes to this is not addressed in the present work. Maybe some modeling would help test some ideas.

Preprint Review

This preprint was reviewed using eLife’s Preprint Review service, which provides public peer reviews of manuscripts posted on bioRxiv for the benefit of the authors, readers, potential readers, and others interested in our assessment of the work. This review applies only to Version 4 of the preprint.

Summary

While all three reviewers found this study to be conceptually of considerable interest, a number of major concerns were highlighted. Most notably, the reviewers do not feel that the central claim of the paper that phospho-eIF4E and S6K1 "interaction is sufficient to overcome rapamycin sensitivity and mTORC1 dependence of S6K1" is sufficiently supported by the evidence presented.

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

Learn more at Review Commons

Reply to the reviewers

We are grateful for the reviewers’ appreciation and comments. We have tried to address all concerns, and believe that those changes have greatly ameliorated the precision and presentation of our findings. All of our responses are in green in this document, and so are the changes in the manuscript and figure legends.

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

In "An asymmetry in the frequency and position of mitosis in the epiblast precedes gastrulation and suggests a role for mitotic rounding in cell delamination during primitive streak epithelial-mesenchymal transition", Mathiah, Despin-Guitard and colleagues study divisions during mouse gastrulation. They perform ex vivo culture, live imaging and immunostaining to observe the frequency and position of mitosis within the embryo as well as the destiny of daughter cells after their divisions. The find that divisions on the posterior side of the embryo tend to be more basally located and could contribute to cell delamination into the mesodermal layer. Authors also affect antero-posterior signaling by genetically preventing the migration of the anterior visceral endoderm, which leads to mitosis away from the apical side of the epithelium on all lateral parts of the embryo.

This study tackles a key developmental process which is poorly understood in mammals due to its concomitance with the implantation phase. Therefore, any carefully-made description of this process has the capacity to be eye-opening. This is potentially the case for this report, which provides nice images that most likely required skills and important efforts to obtain. The authors have written a clear manuscript with an interesting narrative. However, the quantifications are very poorly described, which makes it impossible for anyone to reproduce these results. I describe below a number of suggestions to clarify the quantifications, which is in my opinion a prerequisite to consider the conclusions from the authors.

Fig1: Authors describe differences in the formation of rosettes between the anterior and posterior sides of the embryo. The microscopy images and movies provided are overlaid with drawings from the authors but without this visual help, I, and I assume other readers, see more rosettes than highlighted and fail to see some of the rosettes that are marked. To avoid this subjectivity, a clear methodology is required. In the methods, the authors state: "For quantification, rosettes were manually annotated and counted on Z sections located 5 μm from the basal side of the epiblast." And that is all. What defines a rosette? How many cells need to share a vertex to be considered as part of a rosette? How long do they need to persist not to be considered as occurring by chance? What about cells part of multiple rosettes? Does the rosette organization need to be apical, basal or all the way? Having those clearly defined criteria would be essential for anyone else to reproduce this quantification and would also offer a much more comprehensive description of the phenomenon and allow for more powerful conclusions.

A rosette is defined as a multicellular transient structure composed of at least 5 cells converging to a central vertex. Practically, a region of interest where cell contours are in focus is determined on the Z-section located 5 mm from the epiblast basal side, which is easily identified as the epithelial architecture changes radically when one enters the visceral endoderm. Only rosettes that are visible throughout the epiblast layer, from the basal to the apical side, are counted. To ensure this, manual segmentation of all cells (for all Z plan acquired, from the basal plane to the apical side) contributing to a rosette was performed for lightsheet imaging. This is illustrated in Video 2. For confocal imaging, segmentation was annotated only at the basal plane, but visual verification that the rosette structure is persistent throughout the layer was performed. One cell could be part of several rosettes, and rosette events were counted even when visible only on one timeframe, but this was consistent for all embryo sides. Due to the time resolution of confocal imaging, rosettes could not be followed overtime. However, the time resolution of lightsheet imaging allowed observing rosettes lifespan and resolution. The protocol for image analysis has been better detailed in the results (lines 162-163) and the methods section of the revised version of the manuscript (lines 452-467, copied hereunder).

"Rosettes: For lightsheet imaging, embryos were dissected at E5.75. Images were acquired for 10 to 12 hours. Quantification focused on the first 20 to 30 frames (around 3 hours) to capture pregastrulation events and reduce the risk of bias from imaging. The rest of the frames showed that the embryo continued growing for several hours. Z-stacks from 4 sides were fused using Zeiss plugin for lightsheet Imaging. Images were then processed using Arivis Vision4D v2.12.3 (Arivis, Germany). Embryo contours were segmented manually on each Z-slice and time point, in order to adjust for embryo rotation manually if necessary. For each side of the embryo, Z stack was cropped to an average of 30 Z slices, from the basal side (5 microns from VE layer, which can be morphologically distinguished due to cell shape and membrane Tomato distribution) to the cavity, marking the apical side. Rosettes were identified and counted on Z sections located 5 µm from the basal side of the epiblast. Practically, vertices were systematically scanned to find those in which 5 cells or more met. Cells contributing to a rosette were then manually segmented on each Z-slice and time point by highlighting cellular membranes using Wacom’s Cintiq 13HD, to create a 3D reconstruction. For confocal imaging, rosettes were identified using the same method, and counted on Z sections located 5 µm from the basal side of the epiblast after visual verification that it was present throughout the Z-stack. For both techniques, presence of associated apical rounding was assessed for each vertex. Cells could contribute to several rosettes."

In addition, the data are given as "rosette/frame" and as "rosette/mm2". What is the point of giving both data, which are essentially the same? The frame is irrelevant. It would be more interesting to know how many cells there are in this area, as cell packing could be a determinant of rosette formation. "Rosette/mm2/min" is very confusing. It should state "rosette.mm-2.min-1" or "rosette/mm2.min".

Following this comment, we indeed chose to get rid of the data expressed as “rosette/frame”. Cells were counted in the area of the epiblast in focus to present data as number of rosettes normalized by the number of cells in the region of interest for both lightsheet and confocal microscopy data (described in results section lines 140 and 164). These measurements led to a similar conclusion, confirming that rosettes are more frequent in posterior epiblast. Difference in cell packing was indeed essential to rule out. We estimated cell packing as the ratio of cell number to surface area, and found it to be similar in posterior, anterior, and lateral sides of embryos at a given stage, which indicates that cell packing is not a determinant for difference in rosette frequency in this context. We discussed packing in the Results section (lines 169-173, copied hereunder).

"The cell number per surface area was similar on all sides, which indicates that the higher number of rosettes was not due to increased cell packing. Rosettes have also been identified in the chick PS (Wagstaff, Bellett, Mogensen, & Münsterberg, 2008), where they were proposed to facilitate ingression during gastrulation."

We modified the legend to use "rosette/mm2.min”.

On a conclusive note, I fail to understand how relevant the formation of rosettes would be. The authors should clarify this point.

Epithelial rosettes have been observed as common intermediates in numerous morphogenesis events. In particular cases, such as Drosophila germ band extension, or zebrafish lateral line development, the mechanisms of formation (planar cell polarity (PCP) and apical constriction, respectively) and resolution have been very well described. In the mouse embryo, anterior visceral endoderm (AVE) migration has been linked to PCP signaling-dependent rosette formation (Trichas 2012). In primitive streak (PS) formation, rosettes with actin-rich centers were described in the chick PS and found to be Nodal dependent (Wagstaff 2008 and Yaganawa 2011). Their mode of formation or resolution is currently unknown. Our observations confirm the findings in chick and highlight the presence of rosettes at an earlier stage, before PS can be identified. Interestingly, rosettes are enriched on the posterior side at the same time when Nodal signaling becomes asymmetric, leading to posterior restriction of basal membrane perforations (Kyprianou 2020). To progress towards understanding rosettes’ significance in the mouse gastrulation context, it would be interesting to study whether the distribution of rosettes is homogenous before anterior-posterior axis specification. Additionally, it would be important to assess whether random epiblast cells delaminate before PS formation, as observed in chick (Voiculescu 2014). We could not attempt those experiments so far, as we perform most experiments by two-photon microscopy, by which only one embryo side can be recorded at a time, and have no way to distinguish embryo orientation before AVE migration. A better understanding of rosette mode of formation and resolution, including the role of Nodal, would also be necessary to assess the importance of our observations. The technical evolution in mouse embryo imaging will probably permit solving those questions in the near future, through prolonged imaging with tracking of every cell fate (McDole 2018). We have tried to improve the discussion (lines 314-326, copied hereunder), and acknowledge the limitations of our findings to a description of a phenomenon without proven significance at this stage.

"However, since we observed a marked imbalance in rosette frequency as soon as the anterior-posterior axis was specified, it is possible that rosettes reflect increased epithelium fluidity in posterior epiblast, which is exposed to a distinct mechanical context, at the very beginning of PS morphogenesis. Indeed, a posterior shift in the distribution of basement membrane perforations was identified just after AVE migration, due to an asymmetry in Nodal signaling dependent metalloproteinase activity (Kyprianou et al., 2020). To progress towards understanding rosette formation significance in this context, it would be interesting to study whether the distribution of rosettes is homogenous before anterior-posterior axis specification, and to assess whether random epiblast cells delaminate before PS formation, as observed in chick (Voiculescu, Bodenstein, Lau, & Stern, 2014). As Nodal plays a major role in PS initiation, the presence and distribution of rosettes should be studied in models in which Nodal signaling can be tuned (Kumar, Lualdi, Lewandoski, & Kuehn, 2008)."

Fig2: I have essentially the same issue for bottle cells and delamination counting as for rosettes. In this case, there is nothing in the methods section.

We have added a paragraph to describe the mosaic analysis in the Methods section (lines 472-488):

"Mosaic: Embryos were recorded in a lateral position. As the proportion of GFP positive cells varied between mosaic embryos, normalisation was performed by dividing by the number of green cells in a given embryo. Anterior and posterior halves were defined by drawing a line perpendicular to the embryonic/extraembryonic boundary and passing through the distal tip. Bottle-shaped cells were identified as having a thin attachment on the apical surface (less than a third of the larger section), and the majority of the cell body located in the basal side. Quantification was performed both on the 3D rendering, and through navigating through the Z-stack. The same criteria where used on all sides of the embryo, and quantification was verified by two independent investigators. Delamination was defined as retraction of the apical process, and displacement of the cell body in the mesoderm layer, which could be identified because of the ubiquitous membrane Tomato labelling. Cell division was characterized by cell rounding followed by the appearance of daughter cells. Cell dispersion after mitosis was defined as absence of basolateral contact between daughter cells, which implies presence of at least one epiblast cell (more often 2 or 3) between daughter cells. Mitosis was considered “non-apical” when happening at least 10 µm away from the apical pole, hence not in the first pseudo-layer of nuclei lining the apical pole."

What defines a cell as bottle shape and not bottle shape (apical vs basal width for example)?

Bottle-shaped cells were visually identified as having a thin attachment on the apical surface (less than a third of the larger section), and the majority of the cell body located in the basal side. Quantification was performed both on the 3D rendering, and by navigating through the Z-stack. Due to the large variation in shape, no systematic measurement was performed. However, the same criteria were used on all sides of the embryo, and quantification was verified by two independent investigators. As proposed by Reviewer 2, those criteria would include scutoids with smaller apical surface, which explains why we observe bottle-shaped cells both on the anterior and posterior sides. In addition to Methods, we included a better description of the methodology in the Results (lines 196-200).

"The quantification of bottle-shaped cells was performed in 3D and through Z-stack navigation and included all cells with an apical section smaller than a third of the basal section. Some cells had a round basal cell body and a thin apical extension while others resembled the recently described scutoids performing apico-basal transitions (Gómez-Gálvez et al., 2018)."

Where does a cell need to be to be counted as delaminated (a distance needs to be stated, absolute (better) or relative)?

Delamination is defined as retraction of the apical process, and displacement of the cell body in the mesoderm layer. Using the ubiquitous membrane tomato marker we could easily distinguish the epiblast, mesoderm and visceral endoderm layers, notably through cell packing, morphology and arrangement. This was described in Results (lines 200-204).

"Asymmetrical cells were present on both sides, but more frequent on the posterior side, and cell delamination (retraction of the apical process and cell body shift in the mesoderm layer) only took place on the posterior side. Cells maintained an apical attachment until their basally located cell body had begun crossing the PS/mesoderm border, and only fully detached after delamination."

What defines sister cells as dispersing after division? How far apart do they have to be? After how much time? From the movies provided, the acquisition time seems to short to assess cell dispersal.

Cell dispersion after mitosis was defined as absence of basolateral contact between daughter cells as they extend towards the basal side, which implies intercalation of at least one epiblast cell (more often 2 or 3) between daughter cells. After cytokinesis was completed, extension and separation of daughter cells was visible in the next time point (after 25 min). The time resolution was thus sufficient to note that daughter cells were not adjacent, which is consistent with other studies (Abe 2018).

We have modified the Methods (copied above) and the Results section of the revised version of the manuscript (lines 213-217).

"Upon elongation of daughter cells to reach the basal pole of the epiblast, the majority displayed no basolateral connection between each other and were instead separated by intercalating epiblast cells, which would be expected to result in daughter cells dispersion over time, as described in (Abe, Kutsuna, Kiyonari, Furuta, & Fujimori, 2018)."

Fig3: Mitotic index calculation is described in the figure legend but not in the methods section. It should also be in the methods section and made explicit that the number of mitotic cells is normalized to green cells only, not the entire cell population. The mitotic index seems higher in this population than in the entire embryo as seen in Fig4.

The mitotic index (MI) was indeed calculated differently so numbers cannot be directly compared. MI identified for anterior and posterior epiblast is not statistically different from the ones found in Figure 4 for E7 embryos. In mosaic embryos, we do not have a way to delimitate the PS. In Figure 4, measurements of MI in the PS (delimitated by the area where the basement membrane is degraded) include cells that are destined to delaminate as wells as those that won't. In the mosaic embryos, MI is measured in cells that delaminate only, and is indeed higher. This represents a small population, which likely explains why it does not reach statistical significance and manifests as a trend.

We have fixed the Methods (see above) and Results (lines 222-228) sections.

"For systematic quantification, epiblast regions were defined as anterior or posterior by tracing a line passing by the distal pole and perpendicular to the embryonic/extraembryonic border, and GFP positive cells undergoing rounding were followed overtime (Fig. 3a-c). Although the frequency of cell division (normalized to the total number of GFP positive cells) was similar in anterior and posterior epiblast, there was a trend towards a higher division rate specifically in cells undergoing delamination to become mesoderm (Fig. 3d)."

What defines an exiting cell**?

An exiting cell is characterized by morphological remodelling, apical retraction, as well as the position of the cell body across the mesoderm/epiblast border visualized by the precise membrane Tomato labelling. It is now described in Methods and in Results (lines 201-204: " cell delamination (retraction of the apical process and cell body shift in the mesoderm layer) only took place on the posterior side. Cells maintained an apical attachment until their basally located cell body had begun crossing the PS/mesoderm border, and only fully detached after delamination".

Regarding the non-apical rounding, why not calling it basal rounding? How far from the apical side does a cell need to be counted as non-apical?

The reason for that denomination is that these so-called “non-apical mitoses” are not strictly basal either. Indeed, mitosis is considered “non-apical” when happening at least 10µm away from the apical pole, meaning that these mitoses do not occur within the first pseudo-layer of nuclei lining the apical pole. This is described in Methods.

In the panel h, with the posterior division outcome, is that for all divisions or only for non-apical divisions?

The panel (Fig. 3g, there was an error in figure labelling in the previous version) has been modified to better precise cell outcomes. It represents all posterior divisions, and quantifies the outcome according to the position of mitosis along the apical-basal axis of the cell. See Results, line 230-232: "Non-apical mitosis in the posterior epiblast was preferentially associated with EMT, as it resulted in formation of one or two mesoderm cells (Fig. 3g)."

Do basal divisions give rise to more epi?

No, non-apical divisions mainly give rise to mesoderm cells. Indeed, approximatively 66% of basal divisions give rise to two mesoderm cells, and 33% to an epiblast and a mesenchymal cell (Figure 3g). We never observed a non-apical division resulting in two epiblast cells.

Is epi or meso fate only determined by location in a different layer or are fate markers used?

Epiblast or mesenchymal fate was determined by both morphological and localization criteria. Epiblast cells have an apical and a basal pole. Mesoderm cells have no apical process, and display initiation of front-rear polarity often defined by the presence of nascent migration appendix. As stated before, membrane Tomato labelling allows exact distinction of germ layers.

What happens to the non-apical mitosis on the anterior side?

On the anterior side, the very few anterior non-apical mitoses only give epiblast cells (not shown).

Fig4: Methods state "For Phospho-histone H3 quantifications, sections were chosen at least 10 μm apart to ensure that each cell was only counted once, and counting was performed using the Icy software" and legend states "The PS region is defined by the area where the basal membrane (yellow) is degraded, and the posterior region quantification excludes counts from the PS region".

What about cells at the boundary between PS and non-PS regions? This needs to be extended and brought together in the methods section. **Also, the tissue architecture in the PS is not as well defined as in the rest of the tissue.

A cell was counted as being part of the PS region if at least 50% of its cell body (visual measuring) was within the area where the basal membrane is non-ambiguously degraded, and if the cell retained its attachment to the apical pole (cell contours were determined by F-actin detection using Phalloidin). The Methods section has been completed in the revised version of the manuscript (lines 490-501).

"Phospho-histone H3: For Phh3 quantifications, sections were chosen at least 10 mm apart to ensure that each cell was only counted once, and counting was performed using the Icy software (http://icy.bioimageanalysis.org). For sagittal sections, anterior and posterior regions were defined by drawing a line perpendicular to the embryonic/extraembryonic boundary and passing through the distal tip. For transverse sections, anterior-posterior boundary was placed at mid-distance between the anterior and posterior poles. The PS region was defined by the area where the basement membrane was degraded, and the posterior region quantification excluded counts from the PS region. A cell was counted as being part of the PS region if at least 50% of its cell body was within the area where the basement membrane was non-ambiguously degraded, and if the cell retained its attachment to the apical pole (cell contours were defined by F-actin detection using Phalloidin)."

Is the epithelial polarity clear enough to be determined without AB marker in the PS?

We considered that a cell retained its AB polarity if the cell extended to both apical and basal pole. Even if the pseudostratified epithelium architecture is complex, most cell contours could be delimited when navigating through the Z-stack.

Finally, the number of cells counted is missing. This has been fixed in the Figure legend.

Supp Fig5: based on available images of the Rac1KO embryo, I am not sure that epithelial architecture is established well enough to assess the location of mitosis along the apico-basal axis.

Indeed, the architecture of the Rac1 KO mutants is vastly altered. As a consequence, only a small number of Rac1 mutants in which we could delimitate the germ layers were analysed, and only the cells we could unambiguously locate were considered. The Rac1 VE-deleted phenotype, on the other hand, was not severe enough as there is only a partial AVE migration defect in most mutants (Migeotte et al., 2010). This is why we confirmed the data on AVE migration defective embryos by using the RhoA VE-deleted mutant, which has a strong AVE migration defect but retains good tissue architecture. We tried to increase figure clarity by annotating the embryo cavity as well as the embryonic/extraembryonic boundary. We also submit a less compressed version of the figures, which we hope will facilitate image analysis.

Reviewer #1 (Significance (Required)):

Although I am not as familiar with mouse gastrulation as I would like to be, I am familiar with gastrulation, live imaging and analysis. At this point, I find it difficult to discuss the conclusions of the study since the methodology is so unclear. Nevertheless, any carefully-made description of mammalian gastrulation has the capacity to be eye-opening. This is potentially the case for this report, which provides nice images that most likely required skills and important efforts to obtain.

We hope the changes we made help better understanding the methodology, and thank Reviewer 1 for positive comments and the help in identifying the points we had failed to properly describe.

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

This manuscript, from Mathiah and colleagues, describes an in-depth analysis of differences in cell organization and division within the epiblast of the very early mouse embryo, and in particular, with the onset of gastrulation. Their data indicate a difference in the organization of cells between the anterior/lateral and posterior regions of the epiblast even before gastrulation has commenced, as well as differences in the location of mitoses relative to the apical and basal ends of the cells. The data provide new insight into the early regionalization of the epiblast. However, the authors should include reference to, and discussion of, the paper by Michael Snow on growth and regionalization of the epiblast (Snow MHL (1977) Gastrulation in the mouse: Growth and regionalization of the epiblast. J. Emb. Exp. Morph. 42: 293-303), where he did a much more fine-grained analysis of mitotic index across the entire epiblast, defining a proliferative zone in the anterior part of the primitive streak where the mitotic index was higher between E6.5 and E7.5. He also describes non-apical mitoses specifically in the primitive streak region as compared to all other regions of the epiblast. The results of the present study dovetail nicely with the results presented by Snow.

This was indeed a major oversight, and we apologize for it. The work of Snow identifies very nicely a proliferative zone in the anterior part of the PS. We did not comment on that as our study focuses on posterior PS. We included the reference in the revised version of the manuscript, and pointed the fact that he first described non-apical mitosis in the PS (lines 233-236).

"Remarkably, this concurs with the observation by Snow (Snow, 1977) that in the PS of E6.5 and E7 embryos, mitosis could be found at all levels of the tissue, including adjacent to the endoderm, while it was located at the apical surface of the pseudostratified tissue everywhere else."

Overall, this is a very nice study, but some revisions would help with clarity at certain points. The data on rosette formation are interesting, but it is not clear what an increase in rosettes in the posterior region means. The authors contend (lines 169-170) that this represents a dynamic epithelium primed for EMT, but it is not clear how rosettes facilitate or promote EMT, and especially why that would be seen at E5.75 before EMT has begun. An alternative interpretation might be that the shape of the epithelium may be changing and the packing of the epithelial cells has to change to accommodate this. We do know that the overall shape of the embryo changes from elongate medial-lateral to elongate anterior-posterior just as EMT is initiated (Perea-Gomez et al., (2004) Current Biology 14: 197-207) and it may be that changes in cell packing are required to accommodate this. The authors may want to consider whether the rosettes that they observed represent scutoids (Gomez-Galvez et al., (2018) Scutoids are a geometrical solution to three-dimensional packing of epithelia. Nat. Commun. 9:2960). An analysis of the 3-dimensional organization of the cells within the rosettes (i.e. at all Z levels) may shed some light on this.

Following on the comments by Reviewer 1 and 2, we quantified cell packing, and found it to be identical on all sides at a given stage. We have added a better description of rosette quantification (lines 169-172 and lines 452-470), a video showing 3D reconstruction of cells in a rosette (Video 2), and an extended discussion (lines 314-326) in the revised version of the manuscript. Some cells within the epiblast are indeed likely to be shaped as scutoids, some with an apical-basal asymmetry (lines 196-202). The reference was added to the manuscript (line 200).

Figure 2b,c and Figure 3a, a', b, c - Addition of dotted lines to indicate the apical and basal ends of the epiblast would be helpful in orienting the reader**.

We have added lines to indicate apical and basal ends of the epiblast.

Figure 2c' - what these graphs represent exactly is somewhat vague, and the figure legend is also very vague. In particular, the third graph on cell dispersion is not clear. Does this mean that the daughter cells are separated from one another following division? Or that they are in different compartments (epiblast/mesoderm) after division? A better description should be included in the figure legend.

Following on the comments by Reviewer 1 and 2, we have added a better description of cell dispersion in the Results (lines 213-217), Methods (lines 483-486) and figure legend.

Figure 3g would appear to show the proportion of the total number of posterior divisions that give rise to particular combinations of daughter cells (epi/epi, epi/meso, meso/meso). However, the discussion of this graph in the text (lines 215-219) suggests that it demonstrates that non-apical mitoses always result in meso/meso and epi/meso daughter cells, which it does not. That analysis would be very interesting to add to Figure 3, with the daughter cell types broken down into those coming from apical mitoses and those coming from basal mitoses.

The analysis was broken down as suggested, and has been added to the revised version of the manuscript (line 352, Figure 3g).

In Figure 4, it is not clear how anterior and posterior are defined, and what criteria were used to distinguish posterior from primitive streak. This is nicely demonstrated in Supplementary Figure 3 - maybe panels A and B could be included in Figure 4 to improve the clarity of the analysi**s.

We have better described the quantification methodology in the Methods section (lines 490-501), moved panel a from Supplementary Figure 3 to Figure 4a as suggested, and added an explanatory drawing (Figure 4b) to the revised version of the manuscript.

The data on mitotic index in Figures 2, 4, and 5 do not appear to be consistent. The mitotic index for E7.25 in Figure 2e is similar between anterior and posterior, even though the posterior includes the primitive streak, while the mitotic index presented for the three stages in Figure 4b would imply that the mitotic index for the entire posterior region should be higher than the anterior at all three stages. Similarly, in Figure 5a' and b', the mitotic index in anterior and posterior regions of E5.75 and E6.25 embryos are not significantly different despite the primitive streak being included in the posterior count, while the data presented in Figure 4 would imply that the entire posterior region including primitive streak should be much higher than the anterior. The authors should clarify this in the Results.

In Figure 2 and 3, the mitotic index (MI) is calculated as number of cell division among GFP+ cells divided by the total number of GFP+ cells, while in Figure 4 and 5 it is quantified as Phospho-histone H3+ cells per total number of cells (DAPI). We have clarified this in the revised graphs and legends of the novel version of the manuscript. Those numbers cannot be directly compared. Nonetheless, we found no statistical difference between the MI shown in Figure 3d, and the MI shown in Figure 4c third row (E7). In sagittal view, the PS area cannot be delimited, so we compared anterior and posterior regions, with the PS included in the posterior region, and saw no difference in MI. In transverse section, there was no MI difference when comparing anterior and posterior embryo halves. However, when we refined the analysis by defining the PS as the area where the basal membrane was degraded, a higher MI emerged specifically in the PS compared to anterior and posterior (not including PS) regions. This difference was thus lost by dilution when the PS area was included in the posterior region. We have also stated this distinction more clearly in the revised version of the manuscript (lines 269-271).

The data on non-apical mitoses in the RhoA-VE deleted (Figure 6) and Rac1ko embryos (Supplemental figure5) are not particularly compelling. It is hard to see the basal mitoses in the new AVE-opposed regions in the mutant embryos in the images presented. Perhaps the graphs in these two figures could have the AVE-opposed data broken down into two groups - the region that is posterior and the region that is anterior but not adjacent to AVE. Better images would improve the clarity of these data as well.

As explained in response to Reviewer 1, we have attempted to clarify the anatomy through annotation, and provide less compressed images. We agree that the embryos are altered. Nonetheless, especially in RhoA-VE deleted, the germ layers could be distinguished and non-apical mitosis identified through combining 3D analysis and navigation through the Z-stack. We honestly admit those are the best images we could get, and we believe that they allow to make the point that non-apical mitosis are only found in the area further away from the AVE.

Reviewer #2 (Significance (Required)):

The data on differential proliferation and apical vs. basal mitoses are complementary to data already published, but the present study updates the existing data by the addition of live imaging and 3-dimensional reconstruction of cell shapes, providing a more complete insight into the process. The observation that rosettes are detectable at the basal ends of the epiblast, and more so in the posterior, is novel, but the significance for embryonic development is not well rationalized.

These data are of interest to those investigating the mechanisms of early morphogenesis, as well as those interested in the cellular correlates of molecular regionalization that results from the well-described signaling pathways regulating axis specification.

My background is in early mouse embryo morphogenesis, therefore I feel that I have sufficient expertise to evaluate the data presented.

We thank Reviewer 2 for positive comments, and are grateful for the constructive criticism and important references.

\*Referees Cross Commenting***

I agree with Reviewer 1 on the lack of detail about the methods - my comments stemmed from the same confusion about how measurements were made, but Reviewer 1 more articulately addressed the key points. I agree with Reviewer 3 on the quality of the videos. It is very difficult to see how they could follow cells with a 20 minute interval.

I would like to address the comment by Reviewer 3 on the use of agarose in the imaging experiments. The methods section states that agarose was used to make the culture "chambers" used for light-sheet imaging, which was not the major approach used for imaging in this study. Only the data in Figure 1A came from those experiments, and it was validated by confocal data in 1B,C where the embryos were cultured in Ibidi chambers with culture medium and no agarose present. So I don't think agarose effects on embryo development are a major worry. Also, this same approach was used by Ryan Udan in Mary Dickinson's lab to visualize yolk sac vasculogenesis, and it did not appear to have a deleterious effect on development in that case, although the embryos imaged here were much earlier and are definitely differentially sensitive to culture conditions from those cultured at E8.

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

This is an essentially descriptive study, looking at primitive streak formation and cell ingression from the epiblast in the mouse from about E5.75 to E7.5 or so using time-lapse microscopy (light sheet and confocal) of cultured embryos. The study also takes advantage of some genetically encoded reporters, some of them inducible by tamoxifen, which allow following cells, closer examination of their shapes, and in some cases unambiguous orientation of the embryos based on expression of the reporter. Overall the study is well designed.

I have two very major concerns about this paper - first, the culture system used in most experiments uses agarose, which has been found in several labs to affect normal cell movements and other cell behaviours. It is essential to determine that embryos cultured in these conditions develop normally for much longer than the period of imaging to ensure that the findings are relevant to normal development rather than an artefact. This is particularly important because mouse embryos develop rather poorly at peri-implantation stages with any culture method, and this one could make matters even worse.

Embryos were mounted into an agarose cylinder in which a tunnel had been created with a 150 microns wide copper wire. Embryos were mounted vertically, with the cone oriented on the bottom, to avoid restriction of growth at distal tip of the embryo. As embryos had a smaller diameter than the tunnel, they could comfortably grow without being restricted (Methods, lines 413-414). Although embryos could not been recovered after the long imaging period (12h), embryos similarly mounted in the agarose cylinder but not imaged were kept in culture, and showed normal growth compared to a free-floating embryo (Methods, lines 431-433). In addition, we focused on the first hours of imaging to reduce the risk of phototoxicity-induced anomalies (Methods, lines 453-455). Moreover, although we identified the asymmetry in rosette abundance through lightsheet imaging, we confirmed the finding through confocal imaging of free-floating embryos, and found similar results (Results lines 153-167, Figure 1b and c, Video 3).

While it has been reported that agarose can affect the development of chick embryos in culture, agarose has been a widely used culture matrix for live imaging particularly for lightsheet imaging in other organisms including drosophila, zebrafish, and mouse. We thank Reviewer 2 (in cross-comments) for highlighting that in Udan et al., (2014), a report from Mary Dickinson’s lab, embryos are cultured in agarose “chambers” for lightsheet. Although some of the experiments in Udan et al., (2014) are performed at E8, this paper also focuses on pre-gastrulation mouse embryos as they culture E6.5 embryos for 24 hours, image from 5 view angles, analyze 572 z-slices representing half of the embryo (Fig5 and Fig6 Udan et al., 2014) and show no adverse effects.

The second concern is that for a paper that is almost entirely about time-lapse microscopy observations of live embryos, the movies are very poor. Although the images are generally good and the 3-d sequences/images from the light-sheet microscope sequences are quite impressive (and have good spatial resolution), the time resolution is extremely poor and the movies very short. It is largely impossible to follow cell behaviours or movements in these sequences.

Indeed, the time lapse between time points as well as the total duration of the acquisition is limited, especially when embryos are imaged by confocal microscopy. These measures were taken mainly to preserve the integrity of the embryo and thus ensure that growth conditions were the closest to optimal in vivo conditions. For rosette analysis, the 20 minutes interval was too long to follow rosette resolution, as stated in the manuscript. For mosaically labelled embryos, we quantified only the cells for which the fate and/or progeny could be identified without ambiguity, which was made easier as we chose a 4OH-tamoxifen posology that resulted in a low proportion of labelling. As both cell delamination and mitosis are relatively slow processes, this time resolution proved sufficient. Time resolution for lightsheet was 7 minutes, which is similar compared to other works on mouse gastrulation (such as Williams et al., 2012), and actually higher than most two-photon or confocal studies, including that of our previous reports (Migeotte et al., 2010, Saykali et al., 2019, Trichas et al., 2012) in which cell tracking could be efficiently performed. This high time resolution allowed following individual rosettes overtime (Sup. Fig.2c).

Reviewer #3 (Significance (Required))

The study focuses on cell shape changes and various processes that accompany ingression and reports that ingression may occur through a variety of different mechanisms that occur at the same time, including rosette formation, individual ingression of bottle-shaped cells, and larger population ingression events. This is very similar to what has been described in chick embryos (eg. Voiculescu et al. eLife 2014 - surprisingly this is not cited), although in rodents primitive streak formation occurs in the absence of large-scale movements of cell sheets. Basically there are no surprises in the findings either for mouse or in comparison with other species, but the study is OK in terms of contributing useful information about streak formation and function in mouse (if the above problems are fixed).

We thank Reviewer 3 for helpful comments and references. We respectfully disagree concerning the risk of bias due to agarose cylinder culture, as exposed above. Concerning the videos, we have provided less compressed videos to retain as much image quality as possible. Although it would evidently be better to have a higher time resolution and longer movies, we believe it is not a limitation for the events we study and describe as they can be reliably followed with the time resolution and observation length we provide. The reference to Voiculescu et al., 2014 is indeed important, we have added it to the revised version of the manuscript (line 324) and apologize for the oversight.

\*Referees Cross Commenting***

In response to reviewer 2: One issue with this is that one does not know whether there is a "deleterious" effect of the agarose on movements until one is sure that (a) one understands what the movements would look like without agarose and that there are no differences, and (b) (a serious shortcoming here) that embryos need to be shown to develop completely normally in those culture conditions WAY beyond the period of imaging. There are lots of observations by several labs (some unpublished of course, but some are published) suggesting that agar and agarose do interfere with cell movements. In chick for example the Chapman and Schoenwolf method where embryos are placed on agarose, there are always head defects due to impaired movements and the agarose interfering with tissue tensile forces.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

This is an essentially descriptive study, looking at primitive streak formation and cell ingression from the epiblast in the mouse from about E5.75 to E7.5 or so using time-lapse microscopy (light sheet and confocal) of cultured embryos. The study also takes advantage of some genetically encoded reporters, some of them inducible by tamoxifen, which allow following cells, closer examination of their shapes, and in some cases unambiguous orientation of the embryos based on expression of the reporter. Overall the study is well designed.

I have two very major concerns about this paper - first, the culture system used in most experiments uses agarose, which has been found in several labs to affect normal cell movements and other cell behaviours. It is essential to determine that embryos cultured in these conditions develop normally for much longer than the period of imaging to ensure that the findings are relevant to normal development rather than an artefact. This is particularly important because mouse embryos develop rather poorly at peri-implantation stages with any culture method, and this one could make matters even worse.

The second concern is that for a paper that is almost entirely about time-lapse microscopy observations of live embryos, the movies are very poor. Although the images are generally good and the 3-d sequences/images from the light-sheet microscope sequences are quite impressive (and have good spatial resolution), the time resolution is extremely poor and the movies very short. It is largely impossible to follow cell behaviours or movements in these sequences.

Significance

The study focuses on cell shape changes and various processes that accompany ingression and reports that ingression may occur through a variety of different mechanisms that occur at the same time, including rosette formation, individual ingression of bottle-shaped cells, and larger population ingression events. This is very similar to what has been described in chick embryos (eg. Voiculescu et al. eLife 2014 - surprisingly this is not cited), although in rodents primitive streak formation occurs in the absence of large-scale movements of cell sheets. Basically there are no surprises in the findings either for mouse or in comparison with other species, but the study is OK in terms of contributing useful information about streak formation and function in mouse (if the above problems are fixed).

Referees Cross Commenting

In response to reviewer 2

One issue with this is that one does not know whether there is a "deleterious" effect of the agarose on movements until one is sure that (a) one understands what the movements would look like without agarose and that there are no differences, and (b) (a serious shortcoming here) that embryos need to be shown to develop completely normally in those culture conditions WAY beyond the period of imaging.

There are lots of observations by several labs (some unpublished of course, but some are published) suggesting that agar and agarose do interfere with cell movements. In chick for example the Chapman and Schoenwolf method where embryos are placed on agarose, there are always head defects due to impaired movements and the agarose interfering with tissue tensile forces.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

This manuscript, from Mathiah and colleagues, describes an in-depth analysis of differences in cell organization and division within the epiblast of the very early mouse embryo, and in particular, with the onset of gastrulation. Their data indicate a difference in the organization of cells between the anterior/lateral and posterior regions of the epiblast even before gastrulation has commenced, as well as differences in the location of mitoses relative to the apical and basal ends of the cells. The data provide new insight into the early regionalization of the epiblast. However, the authors should include reference to, and discussion of, the paper by Michael Snow on growth and regionalization of the epiblast (Snow MHL (1977) Gastrulation in the mouse: Growth and regionalization of the epiblast. J. Emb. Exp. Morph. 42: 293-303), where he did a much more fine-grained analysis of mitotic index across the entire epiblast, defining a proliferative zone in the anterior part of the primitive streak where the mitotic index was higher between E6.5 and E7.5. He also describes non-apical mitoses specifically in the primitive streak region as compared to all other regions of the epiblast. The results of the present study dovetail nicely with the results presented by Snow.

Overall, this is a very nice study, but some revisions would help with clarity at certain points. The data on rosette formation are interesting, but it is not clear what an increase in rosettes in the posterior region means. The authors contend (lines 169-170) that this represents a dynamic epithelium primed for EMT, but it is not clear how rosettes facilitate or promote EMT, and especially why that would be seen at E5.75 before EMT has begun. An alternative interpretation might be that the shape of the epithelium may be changing and the packing of the epithelial cells has to change to accommodate this. We do know that the overall shape of the embryo changes from elongate medial-lateral to elongate anterior-posterior just as EMT is initiated (Perea-Gomez et al., (2004) Current Biology 14: 197-207) and it may be that changes in cell packing are required to accommodate this. The authors may want to consider whether the rosettes that they observed represent scutoids (Gomez-Galvez et al., (2018) Scutoids are a geometrical solution to three-dimensional packing of epithelia. Nat. Commun. 9:2960). An analysis of the 3-dimensional organization of the cells within the rosettes (i.e. at all Z levels) may shed some light on this.

Figure 2b,c and Figure 3a, a', b,c - Addition of dotted lines to indicate the apical and basal ends of the epiblast would be helpful in orienting the reader.

Figure 2c' - what these graphs represent exactly is somewhat vague, and the figure legend is also very vague. In particular, the third graph on cell dispersion is not clear. Does this mean that the daughter cells are separated from one another following division? Or that they are in different compartments (epiblast/mesoderm) after division? A better description should be included in the figure legend.

Figure 3g would appear to show the proportion of the total number of posterior divisions that give rise to particular combinations of daughter cells (epi/epi, epi/meso, meso/meso). However, the discussion of this graph in the text (lines 215-219) suggests that it demonstrates that non-apical mitoses always result in meso/meso and epi/meso daughter cells, which it does not. That analysis would be very interesting to add to Figure 3, with the daughter cell types broken down into those coming from apical mitoses and those coming from basal mitoses.

In Figure 4, it is not clear how anterior and posterior are defined, and what criteria were used to distinguish posterior from primitive streak. This is nicely demonstrated in Supplementary Figure 3 - maybe panels A and B could be included in Figure 4 to improve the clarity of the analysis.

The data on mitotic index in Figures 2, 4, and 5 do not appear to be consistent. The mitotic index for E7.25 in Figure 2e is similar between anterior and posterior, even though the posterior includes the primitive streak, while the mitotic index presented for the three stages in Figure 4b would imply that the mitotic index for the entire posterior region should be higher than the anterior at all three stages. Similarly, in Figure 5a' and b', the mitotic index in anterior and posterior regions of E5.75 and E6.25 embryos are not significantly different despite the primitive streak being included in the posterior count, while the data presented in Figure 4 would imply that the entire posterior region including primitive streak should be much higher than the anterior. The authors should clarify this in the Results.

The data on non-apical mitoses in the RhoA-VEdeleted (Figure 6) and Rac1ko embryos (Supplemental figure5) are not particularly compelling. It is hard to see the basal mitoses in the new AVE-opposed regions in the mutant embryos in the images presented. Perhaps the graphs in these two figures could have the AVE-opposed data broken down into two groups - the region that is posterior and the region that is anterior but not adjacent to AVE. Better images would improve the clarity of these data as well.

Significance

The data on differential proliferation and apical vs. basal mitoses are complementary to data already published, but the present study updates the existing data by the addition of live imaging and 3-dimensional reconstruction of cell shapes, providing a more complete insight into the process. The observation that rosettes are detectable at the basal ends of the epiblast, and moreso in the posterior, is novel, but the significance for embryonic development is not well rationalized.

These data are of interest to those investigating the mechanisms of early morphogenesis, as well as those interested in the cellular correlates of molecular regionalization that results from the well-described signaling pathways regulating axis specification.

My background is in early mouse embryo morphogenesis, therefore I feel that I have sufficient expertise to evaluate the data presented.

Referees Cross Commenting

I agree with Reviewer 1 on the lack of detail about the methods - my comments stemmed from the same confusion about how measurements were made, but Reviewer 1 more articulately addressed the key points.

I agree with Reviewer 3 on the quality of the videos. It is very difficult to see how they could follow cells with a 20 minute interval.

I would like to address the comment by Reviewer 3 on the use of agarose in the imaging experiments. The methods section states that agarose was used to make the culture "chambers" used for light-sheet imaging, which was not the major approach used for imaging in this study. Only the data in Figure 1A came from those experiments, and it was validated by confocal data in 1B,C where the embryos were cultured in Ibidi chambers with culture medium and no agarose present. So I don't think agarose effects on embryo development are a major worry. Also, this same approach was used by Ryan Udan in Mary Dickinson's lab to visualize yolk sac vasculogenesis, and it did not appear to have a deleterious effect on development in that case, although the embryos imaged here were much earlier and are definitely differentially sensitive to culture conditions from those cultured at E8.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

In "An asymmetry in the frequency and position of mitosis in the epiblast precedes gastrulation and suggests a role for mitotic rounding in cell delamination during primitive streak epithelial-mesenchymal transition", Mathiah, Despin-Guitard and colleagues study divisions during mouse gastrulation. They perform ex vivo culture, live imaging and immunostaining to observe the frequency and position of mitosis within the embryo as well as the destiny of daughter cells after their divisions. The find that divisions on the posterior side of the embryo tend to be more basally located and could contribute to cell delamination into the mesodermal layer. Authors also affect antero-posterior signaling by genetically preventing the migration of the anterior visceral endoderm, which leads to mitosis away from the apical side of the epithelium on all lateral parts of the embryo.

This study tackles a key developmental process which is poorly understood in mammals due to its concomitance with the implantation phase. Therefore, any carefully-made description of this process has the capacity to be eye-opening. This is potentially the case for this report, which provides nice images that most likely required skills and important efforts to obtain. The authors have written a clear manuscript with an interesting narrative. However, the quantifications are very poorly described, which makes it impossible for anyone to reproduce these results. I describe below a number of suggestions to clarify the quantifications, which is in my opinion a prerequisite to consider the conclusions from the authors.

Fig1: Authors describe differences in the formation of rosettes between the anterior and posterior sides of the embryo. The microscopy images and movies provided are overlaid with drawings from the authors but without this visual help, I, and I assume other readers, see more rosettes than highlighted and fail to see some of the rosettes that are marked. To avoid this subjectivity, a clear methodology is required. In the methods, the authors state: "For quantification, rosettes were manually annotated and counted on Z sections located 5 μm from the basal side of the epiblast." And that is all. What defines a rosette? How many cells need to share a vertex to be considered as part of a rosette? How long do they need to persist not to be considered as occurring by chance? What about cells part of multiple rosettes? Does the rosette organization need to be apical, basal or all the way? Having those clearly defined criteria would be essential for anyone else to reproduce this quantification and would also offer a much more comprehensive description of the phenomenon and allow for more powerful conclusions. In addition, the data are given as "rosette/frame" and as "rosette/mm2". What is the point of giving both data, which are essentially the same? The frame is irrelevant. It would be more interesting to know how many cells there are in this area, as cell packing could be a determinant of rosette formation. "Rosette/mm2/min" is very confusing. It should state "rosette.mm-2.min-1" or "rosette/mm2.min". On a conclusive note, I fail to understand how relevant the formation of rosettes would be. The authors should clarify this point.

Fig2: I have essentially the same issue for bottle cells and delamination counting as for rosettes. In this case, there is nothing in the methods section. What defines a cell as bottle shape and not bottle shape (apical vs basal width for example)? Where does a cell need to be to be counted as delaminated (a distance needs to be stated, absolute (better) or relative)? What defines sister cells as dispersing after division? How far apart do they have to be? After how much time? From the movies provided, the acquisition time seems to short to assess cell dispersal.

Fig3: Mitotic index calculation is described in the figure legend but not in the methods section. It should also be in the methods section and made explicit that the number of mitotic cells is normalized to green cells only, not the entire cell population. The mitotic index seems higher in this population than in the entire embryo as seen in Fig4. What defines an exiting cell? Regarding the non-apical rounding, why not calling it basal rounding? How far from the apical side does a cell need to be to be counted as non-apical? In the panel h, with the posterior division outcome, is that for all divisions or only for non-apical divisions? Do basal divisions give rise to more epi? Is epi or meso fate only determined by location in a different layer or are fate markers used? What happens to the non-apical mitosis on the anterior side?

Fig4: Methods state "For Phospho-histone H3 quantifications, sections were chosen at least 10 μm apart to ensure that each cell was only counted once, and counting was performed using the Icy software" and legend states "The PS region is defined by the area where the basal membrane (yellow) is degraded, and the posterior region quantification excludes counts from the PS region". This needs to be extended (what about cells at the boundary between PS and non-PS regions?) and brought together in the methods section. Also the tissue architecture in the PS is not as well defined as in the rest of the tissue. Is the epithelial polarity clear enough to be determined without AB marker in the PS? Finally, the number of cells counted is missing.

Supp Fig5: based on available images of the Rac1KO embryo, I am not sure that epithelial architecture is established well enough to assess the location of mitosis along the apico-basal axis.

Significance

Although I am not as familiar with mouse gastrulation as I would like to be, I am familiar with gastrulation, live imaging and analysis. At this point I find it difficult to discuss the conclusions of the study since the methodology is so unclear. Nevertheless, any carefully-made description of mammalian gastrulation has the capacity to be eye-opening. This is potentially the case for this report, which provides nice images that most likely required skills and important efforts to obtain.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

The work reports finding a molecular genetic basis for individual differences in behavior in different strains of outbred mice, even including individual behavioral differences between mice of the same inbred genetically isogenic strain. The authors were able to measure copy numbers for the tandemly repeated intronic snoRNA clusters SNORD115 and SNORD116 and found correlation with measures of anxiety in open-field test and elevated plus maze. Expression data for previously proposed targets of these snoRNAs are also provided.

Major comments:

1.The techniques to measure copy numbers are challenging and the authors' conclusion that ddPCR was their method of choice is convincing. They were able to obtain limited optical mapping (Bionano zephyr) data, only for SNORD116 and only in mouse, but these data are useful to corroborate those obtained with ddPCR.

2.Figure 3 reports single copy numbers for individuals that are presumably heterozygous. Do we have to assume that the numbers reported represent the larger alleles since the ddPCR method does not allow to distinguish two different size alleles, as was shown for optical mapping?

3.The analyses reported do not take into account the specific parental origin of the alleles used in the regression analyses. Since PWSCR-specific SNORDs are only expressed from the paternal chromosomes, this generates some uncertainty about the whole dataset.

4.Lines 353-365: The ankrd11 exon-specific RNAseq data are confusing and too preliminary. More work needs to be done to resolve the splice variants in this region and their relationship to SNORD116 copy numbers. Alternatively lines 356-361 should be deleted.

5.In all tested rodents, higher SNORD copy number was correlated with higher relative anxiety score. In the human samples, however, higher anxiety scores were associated with lower copy numbers. These apparently contradictory results are not mentioned in the abstract, and are not satisfactory explained in the text.

6.Extension to other species would be desirable but was limited by availability of genomic data: Results are presented for wood mouse only for SNORD115 and for the guinea pig for SNORD116.

Minor comments:

1.The authors present skull shape data related to SNORD116 copy numbers, but fail to consider how these data are relevant to the craniofacial abnormalities reported in an ankrd11 mutation. Barbaric et al (2008) reported a dominant ENU- induced mutation caused shortened snouts, wider skull, deformed nasal bones, reduced BMD, reduced osteoblast activity and reduced leptin levels. This phenotype was traced to a heterozygous missense mutation (conserved glutamate to lysine change) in an HDAC binding site. They postulated that the mutation fails to recruit HDACs to a transcription complex and to inhibit hormone-receptor activated gene transcription. What is the postulated link between this mechanism and the here reported skull shape data correlated with SNORD copy number variation?

2.The observed co-variation of copy numbers between the two SNORD clusters could indicate a duplication involving the entire region, This could be tested by determining the dosage of IPW, UBE3a and Snrpn genes.

3.Line 129 "the RNA coding region" and Line 148 "snoRNA coding parts" (and elsewhere) does seems correct, as by definition, this is non-coding RNA. The region they are referring to could be called the "processed C/D box snoRNA". The mechanism that generates these C/D box snoRNAs is well established: the "genes" are located in introns of host genes - and after transcription - the spliced out introns are exonucleolytically trimmed to the functional sizes. Both SNORD115 and 116 clusters are within a large transcript that originates from the SNRPN promoter of the paternal allele.

4.Figure 2 does not show data on skull shape as claimed in the legend.

5.S1 Figure: Snprn should be Snrpn

Significance

This provocative work proposes the regulation of behavioral variance by dosage changes of a regulatory RNA. The dosage changes are apparently caused by dynamic and frequent alteration in copy number. This is a novel concept and worthy of publicizing. Extensive data documentation is provided for others to analyze and possibly replicate. The data potentially throw light on the function of the tandemly repeated imprinted snoRNA clusters in the PWS critical region.

Novel aspects of this work include the discovery of copy number variation of these snoRNAs; and validation of a target of SNORD116: Ankrd11 is one of many potential targets of SNORD116 that was previously computationally predicted, this paper reports experimental evidence for this interaction.

The work would be of interest to researchers in behavioral evolution, non-coding RNA function, epigenetics and overall genome evolution.

Define your field of expertise with a few keyword: Molecular genetic disorders, Prader-Willi syndrome, mouse models

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

Learn more at Review Commons

Reply to the reviewers

Dear dr. Monaco,

We thank you and the reviewers for the positive and encouraging reviews on our manuscript entitled “Protective anti-prion antibodies in human immunoglobulin repertoires” and are glad to address the reviewer’s suggestions

In the following you will find a point-by-point response to the referees' critique.

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

Comment 1: Abtract: Although it is clear and direct, the last sentence where it refers to "a link to the low incidence of spontaneous prion diseases in human populations", is not easy to understand without a detailed explanation as given in the Discussion. I suggest a re-wording.

Response 1: We have reworded the sentence in the abstract and given more explanation in the discussion (see also Reviewer 2, Comment 2).

Comment 2: Results: It is clear how these Fabs act in preventing prion-induced neurotoxicity as shown in the COCS model. In addition to this effect, they also inhibit prion spreading, although this appears to be a lesser effect than inhibition of neurotoxicity. Thus, it would be interesting to discuss the possible effect of a Fab therapy, which provide a fully inhibition of the neurotoxicity but only partially inhibition of the prion propagation.

• Response 2: As suggested by the reviewer, we have added appropriate text to the discussion to comment on the option of a potential Fab therapy with a fully inhibition of neurotoxicity and partially inhibition of prion propagation. Comment 3: The therapeutic effect of the Fabs in the cell model was performed by adding the Fabs to the medium 1 h after infection and during splitting. Is there any study that evaluates the effect of Fabs added to the medium before inoculation or at later times?

Response 3: The goal of these experiments was to investigate whether the antibodies in question would counteract prion infections in principle, rather than performing a precise range-finding of the optimal therapeutic window. We have opted to not add the Fabs before inoculation, because past experience (and many papers) show that the “prophylactic” treatment rarely correlated with post-exposure efficacy. We also have not treated the cells after prion infection at later time points, because the data at later time points may be less pronounced and more variable. As for the treatment of cells with anti-PrP antibodies prior to exposure to prions, a study has been conducted in N2a cells (Pankievicz J et al., 2006, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1779824/). There, preincubation of N2a cells with mouse monoclonal anti-PrP antibodies (Mabs) before prion infection (22L) and preincubation of the inoculum with Mabs before infection of the cells led to a significant reduction in PrPSc levels as assessed by proteinase-K Western blot. This paper is now discussed in our manuscript.

Comment 4: Discussion: The authors repeatedly refer to the toxicity that antibodies against GD might have. Related to this, there is currently a therapy (experimental medicine) in humans using an antibody against this region. Perhaps it would be interesting to make a comment on this.

Response 4: Our findings (Sonati et al, Nature 2013, and several following papers) are fundamentally incompatible with those of the London lab on the toxicity of anti-GD antibodies, and elsewhere I have warned loudly against the use of such antibodies in humans. However, this discussion is peripheral to the findings presented here. We have added some text to the discussion but we would rather not expand on this specific issue.

Comment 5: Page 15. I have found the speculative comment: "Accordingly, clinically silent prion generation may occasionally occur in healthy individuals. PrPSc aggregates arising de novo may result in exposure of neoepitopes and/or epitopes occluded in cell-borne PrPC." interesting. However, some of the auto-antibodies found in healthy humans are against a region believed to be structurally unaltered in PrPSc, which it doesn't fit with the theory of exposure to neo-epitopes.

• Response 5: I still believe that my hypothesis is viable, but of course I concede that – thus far – I have no supporting data. We have therefore modified the text to alleviate this comment.

Reviewer #2:

Comment 1: This is a technically advanced and carefully executed study that clearly demonstrate the presence of natural autoantibodies to PrP, some of which show protective properties, in an unselected human population. Although this finding is interesting on its own right, its impact on issues such as incidence of sporadic prion diseases is unclear given that apparently only 0.06% of the nearly 38,000 subjects examined carried these antibodies "in high titer".

Response 1: We agree with the reviewer and have modified the statement as follows: “The frequency of high-titer anti-PrP antibody carriers (0.06%) is much lower than the occurrence of Fab71-like HCDR3 sequences in published human repertoires. This discrepancy could mean that most anti-PrP specificities exist in a dormant state, or are expressed as B-cell receptors, but do not produce circulating antibodies. It will be interesting to discover the triggers that may ignite antibody production and, possibly, afford protection against prions”. The discrepancy between the frequency of anti-PrP antibodies found in the plasma screen and by analysis of the antibody repertoires in the NGS datasets could stem from the fact that most anti-PrP specificities exist in a dormant state, or are expressed as B-cell receptors, but do not produce circulating antibodies (Joseena Iype et al., J Immunol 2019; now also included in the manuscript).

Comment 2: Furthermore, this reviewer could not locate the base of the pivotal statement made in the Abstract that these autoantibodies lack in carriers of disease-associated PRNP mutations. These two points need to be clarified.

Response 2: The statement refers to the study by Frontzek et al. (citation #48: Frontzek, K. et al. Autoantibodies against the prion protein in individuals with PRNP mutations Neurologyhttps://n.neurology.org/content/early/2020/02/25/WNL.0000000000009183?rss=1). Although listed in the references, the citation got lost in the discussion. We have inserted the reference again.

Comment 3: The manuscript suffers for the excessive amount of data that are crammed in the five figures. Combined these figures display a total of 33 panels some of which are quite complicated. The authors should be more selective and roll over some of the nonessential information i.e. that related to methodology, to the Supplement.

• Response 3: We agree with the reviewer and have moved several panels to the Supplement.

Comment 4: The use of acronyms is excessive and should be reduced (see for example COCS).

• Response 4: We have attempted to reduce the number of acronyms. We have however introduced the term COCS in Falsig et al., Nature Neuroscience 2007, and have used it regularly in more than a dozen follow-up papers. Comment 5: The legends need to be carefully checked for clarity, especially figure 4

• Response 5: We have revised the legends to improve their clairity.

Reviewer #3:

Comment 1: On page 10, the authors state that Fab71 (Figure 3e) and Fab100 (Extended data Figure 7) substantially lowered PrPSc levels in prion-infected cells. However, in both cases, only about half of the cultures tested showed less PrPSc than either the control samples or samples treated with other Fabs. This variability undercuts the conclusion that what they are observing is a substantial, reproducible effect. The authors should consider moderating their conclusion somewhat to better fit the data.

Response 1: We agree with the reviewer. The effect of Fab71 and Fab100 in reducing PrPSc levels in cells as compared to control samples and samples treated with the other Fabs is only partially present and variable among the replicates, but still statistically significant (One-way ANOVA; p

Comment 2: In figure 2, the legend to panel a does not match the figure. Fab3 and Fab71 are represented by the blue lines, not the red lines as stated in the legend.

• Response 2: We thank the referee for pointing this out. We have now corrected it.

Comment 3: In the legend to Extended data figure S4, please give the epitopes to Fab10 and Fab53.

Response 3: We have included the epitopes of these two Fabs (OR51-91 for Fab10 and CC2-HC92-120 for Fab53) in the Figure legend.

Comment 4: In Figure 3c, the lines indicating the significant groups are not well-aligned. In the left side of the panel, the lines should connect the dark gray control group squares with the Fab25 pink diamonds. Likewise, in the right side of the panel, the lower set of lines should connect the dark gray control group squares with the Fab83 dark blue triangles.

• Response 4: We have corrected this issue.

Comment 5: I agree with the comments of both reviewers. The suggestion of reviewer #2 to move methodology-related panels in the main figures to supplemental data would make it much easier for the reader to focus on the critical experimental data.

• Response 5: See response to comment 2 of reviewer 2. With all issues addressed, we hope that our revised manuscript will now be found suitable for proceeding to the next steps.

Best regards,

Adriano Aguzzi

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

The manuscript by Senatore et al. is large scale study looking for natural human antibodies directed against prion protein (PrP). Using a synthetic human Fab phage display library, they found and characterized multiple human anti-PrP Fabs most of which recognized epitopes to a region of PrP from amino acid residues 92-120. Based on this information, they searched for and found low affinity, long-lived anti-PrP antibodies in both a repertoire of human antibodies and in 27 of almost 38,000 human clinical samples. They speculate that anti-PrP antibodies may help to protect against sporadic forms of prion disease and conclude that they may represent a source of potential immunotherapeutics against human prion infection.

Minor comments:

1) On page 10, the authors state that Fab71 (Figure 3e) and Fab100 (Extended data Figure 7) substantially lowered PrPSc levels in prion-infected cells. However, in both cases, only about half of the cultures tested showed less PrPSc than either the control samples or samples treated with other Fabs. This variability undercuts the conclusion that what they are observing is a substantial, reproducible effect. The authors should consider moderating their conclusion somewhat to better fit the data.

2) In figure 2, the legend to panel a does not match the figure. Fab3 and Fab71 are represented by the blue lines, not the red lines as stated in the legend.

3) In the legend to Extended data figure S4, please give the epitopes to Fab10 and Fab53.

4) In Figure 3c, the lines indicating the significant groups are not well-aligned. In the left side of the panel, the lines should connect the dark gray control group squares with the Fab25 pink diamonds. Likewise, in the right side of the panel, the lower set of lines should connect the dark gray control group squares with the Fab83 dark blue triangles.

Significance

This is an extensive, well-written study which provides significant data suggesting that humans can make anti-PrP antibodies. This is a novel finding that raises important questions about how the body may respond to spontaneous formation of infectious prions. Technically, the study is sound with appropriately interpreted data. Overall the study and the antibodies it characterizes, some of which are novel, will be of interest to prion researchers.

Referees cross commenting

I agree with the comments of both reviewers. The suggestion of reviewer #2 to move methodology-related panels in the main figures to supplemental data would make it much easier for the reader to focus on the critical experimental data.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary:

The authors extensively and rigorously characterized a subset of antibodies to PrP identified in a human Fab phage display library. These selected antibodies were compared and found to be similar to repertoires of naturally occurring human antibodies present in circulating B cells. Profiling of antibodies harvested from an unbiased 38,000 patient population uncovered the presence of high titer anti-PrP autoantibodies in 21 individuals sharing no specific pathologies. This finding demonstrates the presence of apparently innocuous immunity to prion in an unselected population. Based also on "the reported lack of such antibodies in carriers of disease-associated PRNP mutations" the authors propose that the low incidence of "spontaneous" prion diseases may be linked to the presence of these protective antibodies in the general population.

Major comments:

This is a technically advanced and carefully executed study that clearly demonstrate the presence of natural autoantibodies to PrP, some of which show protective properties, in an unselected human population. Although this finding is interesting on its own right, its impact on issues such as incidence of sporadic prion diseases is unclear given that apparently only 0.06% of the nearly 38,000 subjects examined carried these antibodies "in high titer". Furthermore, this reviewer could not locate the base of the pivotal statement made in the Abstract that these autoantibodies lack in carriers of disease-associated PRNP mutations. These two points need to be clarified. The manuscript suffers for the excessive amount of data that are crammed in the five figures. Combined these figures display a total of 33 panels some of which are quite complicated. The authors should be more selective and roll over some of the nonessential information i.e. that related to methodology, to the Supplement.

Minor comments:

The use of acronyms is excessive and should be reduced (see for example COCS). The legends need to be carefully checked for clarity, especially figure 4

Significance

Significance

See above

Referees Cross Commenting

I agree with most of the comments by Reviewers 1 and 3. However, my queries remain.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This is a very interesting article with important implications in the prion field. It is extremely well detailed and exquisitely well written. The objective of the article is very clear and the results obtained are not only interesting but also have very important implications for understanding prion diseases.

I have some comments and a few minor concerns.

Abtract:

Although it is clear and direct, the last sentence where it refers to "a link to the low incidence of spontaneous prion diseases in human populations", is not easy to understand without a detailed explanation as given in the Discussion. I suggest a re-wording.

Results:

It is clear how these Fabs act in preventing prion-induced neurotoxicity as shown in the COCS model. In addition to this effect, they also inhibit prion spreading, although this appears to be a lesser effect than inhibition of neurotoxicity. Thus, it would be interesting to discuss the possible effect of a Fab therapy, which provide a fully inhibition of the neurotoxicity but only partially inhibition of the prion propagation.

The therapeutic effect of the Fabs in the cell model was performed by adding the Fabs to the medium 1 h after infection and during splitting. Is there any study that evaluates the effect of Fabs added to the medium before inoculation or at later times?

Discussion:

The authors repeatedly refer to the toxicity that antibodies against GD might have. Related to this, there is currently a therapy (experimental medicine) in humans using an antibody against this region. Perhaps it would be interesting to make a comment on this.

Page 15. I have found the speculative comment: "Accordingly, clinically silent prion generation may occasionally occur in healthy individuals. PrPSc aggregates arising de novo may result in exposure of neoepitopes and/or epitopes occluded in cell-borne PrPC." interesting. However, some of the auto-antibodies found in healthy humans are against a region believed to be structurally unaltered in PrPSc, which it doesn't fit with the theory of exposure to neo-epitopes.

Significance

The advance is highly significance for two reasons: 1) the tools that the authors have generated are really useful for the community and 2) The fact the healthy humans can generate anti-PrP antibodies is completely new and open new ways to understand the prion diseases mechanisms.

The audience is principally for those working on prion and prion-like diseases.

My expertise is in prion and prion-like diseases.

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

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)): \*Summary:** Reproducibility of genetic interactions across studies is low. The authors identify reproducible genetic interactions and ask the question of what are properties of robust genetic interactions. They find that 1. oncogene addiction tends to be more robust than synthetic lethality and 2. genetic interactions among physically interacting proteins tend to be more robust. They then use protein-protein interactions (PPIs) to guide the detection of genetic interactions involving passenger gene alterations. **Major comments:** The claims of the manuscript are clear and well supported by computational analyses. My only concern is the influence of (study) bias on the observed enrichment of physical protein interactions among genetic interactions. 1. Due to higher statistical power the here described approach favors genetic interactions involving frequently altered cancer genes (as acknowledged by the authors). 2. Also some of the libraries in the genetic screens might be biased towards better characterized screens. 3. PPI networks are highly biased towards well studied proteins (in which well studied proteins - in particular cancer-related proteins - are more likely to interact). The following tests would help to clarify if and to which extend these biases contribute to the described observations:*

Our response: We thank the reviewer for the positive assessment of our manuscript and have addressed the issue of study bias in response to the specific queries below.

* 1 . The authors should demonstrate that the PPI enrichment in reproducible vs non-reproducible genetic interactions is not solely due to the biased nature of PPI networks. One simple way of doing so would be to do the same analysis with a PPI network derived from a single screen (eg PMID: 25416956). I assume that due to the much lower coverage the effect will be largely reduced but it would be reconfirming to see a similar trend in addition to the networks on which the authors are already testing. Another way would be to use a randomized network (with the same degree distribution as the networks the authors are using and then picking degree matched random nodes) in which the observed effect should vanish.

*

Our response: We appreciate the reviewer’s point and have now assessed both of the suggested approaches.

The overlap with unbiased yeast two-hybrid (y2h) screens, even the recent HuRI dataset (Luck et al, Nature 2020), was too small in scale to draw any conclusions. Among the ~140,000 interactions tested for genetic interactions, only 51 overlap with y2h interactions. Two of the discovered genetic interactions were supported by a y2h interaction, while one of the robust genetic interactions was supported by a y2h interaction. While this is actually more than would be expected based on the overlap of interactions in the test space the numbers are not especially convincing.

We therefore focused on two alternative assessments. We first compared our results with the network derived from the systematic AP-MS mapping of protein interactions in HEK293 cells (BioPlex 3.0, Huttlin et al, Biorxiv 2020). We restricted our analysis of genetic interactions to gene pairs that could conceivably be observed in the BioPlex dataset (i.e. between baits screened and preys expressed in HEK293T). We found that although the numbers were small, the same pattern of enrichment was observed:

This analysis has now been added to the revised manuscript as Supplementary Table S4 and Figure S3E (shown below):

We next compared the results we observed with the real STRING protein-protein interaction network to 100 degree-matched randomisations of this network. We observed that the number of discovered and validated genetic interactions observed using the real STRING interaction network was greater than that observed using the randomised networks. With this in mind, we have now revised the manuscript to state:

‘Previous work has demonstrated that the protein-protein interaction networks aggregated in databases are subject to significant ascertainment bias – some genes are more widely studied than others and this can result in them having more reported protein-protein interaction partners than other genes(Rolland et al., 2014). As cancer driver genes are studied more widely than most genes, they may be especially subject to this bias. To ensure the observed enrichment of protein-protein interactions among genetically interacting pairs was not simply due to this ascertainment bias, we compared the results observed for the real STRING protein-protein interaction network with 100 degree-matched randomised networks and again found that there was a higher than expected overlap between protein-protein interactions and both discovered and validated genetic interactions (Supplemental Fig. S4).’

Supplemental Figure S4. Genetic interactions are more enriched in real protein-protein interaction networks than randomised networks. Histograms showing the overlap between 100 degree matched randomisations of the STRING medium confidence protein-protein interaction and discovered (a and b) and validated (c and d) genetic interactions. The observed overlap with the real STRING protein interaction are highlighted with the orange lines.

* 2 . What's the expected number of robust genetic interactions involving passenger gene alterations? Is it surprising to identify 11 interactions? This question could be addressed with some sort of randomization test: When selecting (multiple times) 47,781 non-interacting random pairs between the 2,972 passenger genes and 2,149 selectively lethal genes, how many of those pairs form robust genetic interactions?

*

Our response: We have now addressed this as follows:

“At an FDR of 20% we found 11 robust genetic interactions involving passenger gene alterations (Supplemental Table S6). To assess whether this is more than would be expected by chance we randomly sampled 47,781 gene pairs from the same search space 100 times. The median number of robust genetic interactions identified amongst these randomly sampled gene pairs was 1 (mean 1.27, min 0, max 6) suggesting that the 11 robust genetic interactions observed among protein-protein interacting pairs was more than would be expected by chance.”

\*Minor comments:**

Two additional analyses would add in my opinion value to the manuscript:

-The authors state that reasons for irreproducibility of genetic interactions are of technical or biological nature. Is it possible to disentangle the contribution of the two factors given the available data? Eg how many genetic interactions are reproducible in two different screening platforms using the same cell line vs how similar are results of screens from two different cell lines in the same study?

*

Our response: We are also very interested in this question, but with the available data, we are not confident that we could draw solid conclusions.

-The authors state that "some of the robust genetic dependencies could be readily interpreted using known pathway structures" and argue that they recover for example MAPK or Rb pathway relationships. Is this a general trend? Do genes forming a robust genetic interactions have a higher tendency to be in the same pathway as opposed to different pathways?

Our response: We have now systematically tested the robust genetic interactions for each driver gene for enrichment in specific pathways. Relevant text is as follows:

‘To test if this enrichment of pathway members among the robust dependencies associated with specific driver genes was a common phenomenon, for each driver gene with at least three dependencies we asked if these dependencies were enriched in specific signalling pathways (see Methods). Of the twelve driver genes tested, we found that five of these were enriched in specific pathways and in all five cases found that the driver gene itself was also annotated as a member of the most enriched pathway (Table SX). As expected RB1 (most enriched pathway ‘G1 Phase’) and BRAF (most enriched pathway ‘Negative feedback regulation of MAPK pathway’) were among the five driver genes, alongside PTEN (‘PI3K/AKT activation’), CDKN2A (‘Cell cycle’), and NRAS (‘Ras signaling pathway’).’

Details in the methods are as follows:

‘Pathway enrichment was assessed using gProfiler (Raudvere et al., 2019) with KEGG (Kanehisa et al., 2017) and Reactome (Jassal et al., 2020) as annotation databases and the selectively lethal genes as the background list.’

*I think the pathway topic could be in general better exploited: eg does pathway (relative) position play a role?**

*

Our response: We agree that pathway position, especially distance from driver gene in an ordered pathway, would be very interesting to tease out but we don’t think that current pathway annotations are reliable enough nor the set of robust genetic interactions large enough to analyse this properly.

*Reviewer #1 (Significance (Required)):**

Personalized cancer medicine aims at the identification of patient-specific vulnerabilites which allow to target cancer cells in the context of a specific genotype. Many oncogenic mutations cannot be targeted with drugs directly. The identification of genetic interactions is therefore of crucial importance. Unfortunately, genetic interactions show little reproducibility accross studies. The authors make an important contribution to understanding which factors contribute to this reproducibility and thereby providing means to also identify more reliable genetic interactions with high potential for clinical exploitation or involving passenger gene alterations (which are otherwise harder to detect for statistical reasons).

REFEREES CROSS COMMENTING

Reviewer 2 raises a few valid points, which if addressed would certainly increase the clarity of the paper. In particular addressing the first point (the self interactions of tumor suppressors) seems important to me. From what I can see all of reviewer 2's comments can be addressed easily.

*

End of Reviewer 1 comments

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

*In this manuscript, Lord et al. describe the analysis of loss-of-function (LOF) screens in cancer cell lines to identify robust (i.e., technically reproducible and shared across cell lines) genetic dependencies. The authors integrate data from 4 large-scale LOF studies (DRIVE, AVANA, DEPMAP and SCORE) to estimate the reproducibility of their individual findings and examine their agreement with other types of functional information, such as physical binding. The main conclusions from the analyses are that: a) oncogene-driven cancer cell lines are more sensitive to the inhibition of the oncogene itself than any other gene in the genome; b) robust genetic interactions (i.e., those observed in multiple datasets and cell lines driven by the same oncogene/tumour suppressor) are enriched for gene pairs encoding physically interacting proteins.

**Main comments:**

I think this study is well designed, rigorously conducted and clearly explained. The conclusions are consistent with the results and I don't have any major suggestions for improving their support. I do, however, have a few suggestions for clarifying the message.

*

Our response: We thank the reviewer for this positive assessment of our manuscript and have addressed the requests for clarity below.

-Could the authors provide some intuitive explanation (or speculation) about the 2 observed cases of tumour suppressor "addiction" (TP53 and CDKN2A)? While the oncogene addiction cases are relatively easy to interpret, the same effects on tumour suppressors are less clear. Is it basically an epistatic effect, which looks like a relative disadvantage? For example, if we measure fitness: TP53-wt = 1, TP53-wt + CRISPR-TP53 = 1.5, TP53-mut = 1.5, TP53-mut + CRISPR-TP53 = 1.5. That is, inhibiting TP53 in TP53 mutant cells appears to be disadvantageous (relative to WT) only because inhibiting TP53 in wild-type cells is advantageous?

Our response: The reviewer is correct – the TP53 / TP53 dependency is similar to an epistatic effect. In a TP53 mutant background targeting TP53 with shRNA or CRISPR has a neutral effect, while in a TP53 wild type background targeting TP53 with shRNA or CRISPR often causes an increase in cell growth. We have clarified this in the text below (new text in bold)

‘We also identified two (2/9) examples of ‘self vs. self’ dependencies involving tumour suppressors -TP53 (aka p53) and CDKN2A (aka p16/p14arf) (Supplemental Fig. S2c). This type of relationship has previously been reported for TP53: TP53 inhibition appears to offer a growth advantage to TP53 wild type cells but not to TP53 mutant cells(Giacomelli et al., 2018). Inhibiting TP53 in TP53 mutant cells has a largely neutral effect, while on average inhibiting TP53 in TP53 wild type cells actually increases fitness growth. Consequently, we observed an association between TP53 status and sensitivity to TP53 inhibition. A similar effect was observed for CDKN2A, although the growth increase resulting from inhibiting CDKN2A in wild-type cells is much lower than that seen for TP53 (Supplemental Fig. S2c).;

*-In the analysis of overlap between genetic and physical interactions, the result should be presented more precisely. Currently, the text reads "when considering the set of all gene pairs tested, gene pairs whose protein products physically interact were more likely to be identified as significant genetic interactors". However, the referenced figure (Fig. 5a) shows an orthogonal perspective: relative to all gene pairs tested, those that have a significant genetic interaction are more likely to have a physical interaction as well. In other words, in the text, we are comparing the relative abundance of genetic interactions in 2 sets: tested and physically interacting. However, in the figure, we are comparing the relative abundance of protein interactions in 2 sets -- tested and genetically interacting. The odds ratio and the p-values stay the same but the result would be more clear if the figure matched the description in the text.

*

Our response: Due to the fact that genetic interactions are rare (~1% of all gene pairs tested have a discovered genetic interaction, ~0.1% have a validated genetic interaction) it’s hard to convey the enrichment effectively. This is demonstrated in the below figure – it’s clear that there are more discovered / validated genetic interaction pairs among the protein-protein interaction pairs but the scale is hard to appreciate:

Focusing only on the discovered/validated genetic interactions makes the picture a little clearer but does not effectively show that the discovered pairs themselves are enriched among protein-protein interaction pairs

As we feel the original figures convey the main message most effectively, we have altered the text rather than the images as follows:

“We found that, when considering the set of all gene pairs tested, gene pairs identified as significant genetic interactors in at least one dataset are more likely to encode proteins that physically interact (Fig. 5a)”

\*Minor comments:**

There're a few places where the more explicit explanation would improve the readability of the manuscript.

-Page 5: The multiple regression model used to identify genetic interactions is briefly mentioned in the text (and described more extensively in the methods). I think it would be better to explicitly describe the dependent and independent variables of the model in the text, so that the reader can intuitively understand what is being estimated*.

Our response: We have added additional information to the main text as follows:

‘This model included tissue type, microsatellite instability and driver gene status as independent variables and gene sensitivity score as the dependent variable (Methods). Microsatellite instability was included as a covariate as it has previously been shown to be associated with non-driver gene specific dependencies (Behan et al., 2019), while tissue type was included to avoid confounding by tissue type.’*

-Page 5: "Using this approach, we tested 142,477 potential genetic dependencies..." -- could the authors provide a better explanation of where that number is coming from? E.g., 142,477 = ... driver genes x 2470 selectively lethal genes?*

Our response: Because not every selectively lethal gene is tested in every dataset (e.g. DRIVE only screened ~8,000 genes instead of the whole genome) the 142,477 number does not correspond to a simple multiplication of number of driver genes times number of selectively lethal gene. However, we have added additional information in bold as follows:

‘Using this approach, we tested 142,477 potential genetic dependencies between 61 driver genes and 2,421 selectively lethal genes. We identified 1,530 dependencies that were significant in at least one discovery screen (Fig. 2a, Supplemental Fig. S1). All 61 driver genes had at least one dependency that was significant in at least one discovery screen while less than half of the selectively lethal genes (1,141 / 2,421) had a significant association with a driver gene. Of the 1,530 dependencies that were significant in at least one discovery screen, only 229 could be validated in a second screen (Supplemental Table S3, Fig. 2a). For example, in the AVANA dataset TP53 mutation was associated with resistance to inhibition of both MDM4 and CENPF, but only the association with MDM4 could be validated in a second dataset (Fig. 2b, 2c). Similarly, in the DEPMAP dataset NRAS mutation was associated with increased sensitivity to the inhibition of both NRAS itself and ERP44, but only the sensitivity to inhibition of NRAS could be validated in a second dataset (Fig. 2b, 2c).

The 229 reproducible dependencies involved 31 driver genes and 204 selectively lethal genes.’

-Page 5: Repeating the number of findings of each type would help understanding the landscape of the genetic dependencies (suggested numbers in brackets): "Of the (229?) reproducible genetic dependencies nine were 'self vs self' associations". "The majority (7/9?) of these ... were oncogene addiction effects". "We also identified 2 (2/9?) examples of 'self vs self' dependencies involving tumour suppressors".

Our response: We have taken the reviewer’s advice and added these figures to the main text for clarity

* -Page 12: "Three of these interactions involve genes frequently deleted with the tumour suppressor CDKN2A (CDKN2B and MTAP) and mirror known associations with CDKN2A". It is not clear what "mirror" means -- do they recapitulate known interactions?

*

Our response: Yes, we meant to indicate that they recapitulate known CDKN2A interactions and have now replaced ‘mirror’ with ‘recapitulate’.

-Page 15: "Although we have not tested them here, other features predictive of between-species conservation may also be predictive of robustness to genetic heterogeneity" -- could the authors explicitly list the features?

Our response: We have now explicitly listed these features as follows:

“Previous work has also shown that genetic interactions between gene pairs involved in the same biological process, as indicated by annotation to the same gene ontology term, are more highly conserved across species (Ryan et al., 2012; Srivas et al., 2016). Similarly, genetic interactions that are stable across experimental conditions (e.g. that can be observed in the presence and absence of different DNA damaging agents) are more likely to be conserved across species (Srivas et al., 2016). Although we have not tested them here, these additional features predictive of between-species conservation may also be predictive of robustness to genetic heterogeneity.”

*Reviewer #2 (Significance (Required)):

The identification of a significant overlap between genetic and physical interactions in cancer cell lines is an interesting and promising observation that will help understanding known genetic dependencies and predicting new ones. However, similar observations have been made in other organisms and biological systems. These past studies should be referenced to provide a historical perspective and help define further analyses in the cancer context. In particular, studies in yeast S. cerevisiae have shown that, not only there is a general overlap between genetic interactions (both positive and negative) and physical interactions, but at least 2 additional features are informative about the relationship: a) the relative strength of genetic interactions and b) the relative density of physical interactions (i.e., isolated interaction vs protein complexes). Here's a sample of relevant studies: 1) von Mering et al., Nature, 2002; 2) Kelley & Ideker, Nat Biotechnol, 2005; 3) Bandyopadhyay et al., PLOS Comput Biol, 2008; 4) Ulitsky et al., Mol Syst Biol, 2008; 5) Baryshnikova et al., Nat Methods, 2010; 6) Costanzo et al., Science, 2010; 7) Costanzo et al., Science, 2016.

Similar observations have also been made in mammalian systems: e.g., in mouse fibroblasts (Roguev et al., Nat Methods, 2013) and K562 leukemia cells (Han et al., Nat Biotech, 2017). I don't think that past observations negate the novelty of this manuscript. The analysis presented here is more focused and more comprehensive as it is based on a large integrated dataset and is driven by a series of specific hypotheses. However, a reference to previous publications should be made.

As a frame of reference: my expertise is in high-throughput genetics of model organisms, including mapping and analyzing genetic interactions.

*

Our response: We thank the reviewer for highlighting this point.

We have attempted to provide better context for our work in the discussion as follows:

‘In budding and fission yeast, multiple studies have shown that genetic interactions are enriched among protein-protein interaction pairs and vice-versa (Costanzo et al., 2010; Kelley and Ideker, 2005; Michaut et al., 2011; Roguev et al., 2008). Pairwise genetic interaction screens in individual mammalian cell lines have also revealed an enrichment of genetic interactions among protein-protein interaction pairs (Han et al., 2017; Roguev et al., 2013). Our observation that discovered genetic interactions are enriched in protein-protein interaction pairs is consistent with these studies. However, these studies have not revealed what factors influence the conservation of genetic interactions across distinct genetic backgrounds, i.e. what predicts the robustness of a genetic interaction. In yeast, the genetic interaction mapping approach relies on mating gene deletion mutants and consequently the vast majority of reported genetic interactions are observed in a single genetic background (Tong et al., 2001). In mammalian cells, pairwise genetic interaction screens across multiple cell lines have revealed differences across cell lines but not identified what factors influence the conservation of genetic interactions across cell lines(Shen et al., 2017). While variation of genetic interactions across different strains or different genetic backgrounds has been poorly studied, previous work has analysed the conservation of genetic interactions across species and shown that genetic interactions between gene pairs whose protein products physically interact are more highly conserved (Roguev et al., 2008; Ryan et al., 2012; Srivas et al., 2016). Our analysis here suggests that the same principles may be used to identify genetic interactions conserved across genetically heterogeneous tumour cell lines.’

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Lord et al. describe the analysis of loss-of-function (LOF) screens in cancer cell lines to identify robust (i.e., technically reproducible and shared across cell lines) genetic dependencies. The authors integrate data from 4 large-scale LOF studies (DRIVE, AVANA, DEPMAP and SCORE) to estimate the reproducibility of their individual findings and examine their agreement with other types of functional information, such as physical binding. The main conclusions from the analyses are that: a) oncogene-driven cancer cell lines are more sensitive to the inhibition of the oncogene itself than any other gene in the genome; b) robust genetic interactions (i.e., those observed in multiple datasets and cell lines driven by the same oncogene/tumour suppressor) are enriched for gene pairs encoding physically interacting proteins.

Main comments:

I think this study is well designed, rigorously conducted and clearly explained. The conclusions are consistent with the results and I don't have any major suggestions for improving their support. I do, however, have a few suggestions for clarifying the message.

-Could the authors provide some intuitive explanation (or speculation) about the 2 observed cases of tumour suppressor "addiction" (TP53 and CDKN2A)? While the oncogene addiction cases are relatively easy to interpret, the same effects on tumour suppressors are less clear. Is it basically an epistatic effect, which looks like a relative disadvantage? For example, if we measure fitness: TP53-wt = 1, TP53-wt + CRISPR-TP53 = 1.5, TP53-mut = 1.5, TP53-mut + CRISPR-TP53 = 1.5. That is, inhibiting TP53 in TP53 mutant cells appears to be disadvantageous (relative to WT) only because inhibiting TP53 in wild-type cells is advantageous?

-In the analysis of overlap between genetic and physical interactions, the result should be presented more precisely. Currently, the text reads "when considering the set of all gene pairs tested, gene pairs whose protein products physically interact were more likely to be identified as significant genetic interactors". However, the referenced figure (Fig. 5a) shows an orthogonal perspective: relative to all gene pairs tested, those that have a significant genetic interaction are more likely to have a physical interaction as well. In other words, in the text, we are comparing the relative abundance of genetic interactions in 2 sets: tested and physically interacting. However, in the figure, we are comparing the relative abundance of protein interactions in 2 sets -- tested and genetically interacting. The odds ratio and the p-values stay the same but the result would be more clear if the figure matched the description in the text.

Minor comments:

There're a few places where the more explicit explanation would improve the readability of the manuscript.

-Page 5: The multiple regression model used to identify genetic interactions is briefly mentioned in the text (and described more extensively in the methods). I think it would be better to explicitly describe the dependent and independent variables of the model in the text, so that the reader can intuitively understand what is being estimated.

-Page 5: "Using this approach, we tested 142,477 potential genetic dependencies..." -- could the authors provide a better explanation of where that number is coming from? E.g., 142,477 = ... driver genes x 2470 selectively lethal genes?

-Page 5: Repeating the number of findings of each type would help understanding the landscape of the genetic dependencies (suggested numbers in brackets): "Of the (229?) reproducible genetic dependencies nine were 'self vs self' associations". "The majority (7/9?) of these ... were oncogene addiction effects". "We also identified 2 (2/9?) examples of 'self vs self' dependencies involving tumour suppressors".

-Page 12: "Three of these interactions involve genes frequently deleted with the tumour suppressor CDKN2A (CDKN2B and MTAP) and mirror known associations with CDKN2A". It is not clear what "mirror" means -- do they recapitulate known interactions?

-Page 15: "Although we have not tested them here, other features predictive of between-species conservation may also be predictive of robustness to genetic heterogeneity" -- could the authors explicitly list the features?

Significance

The identification of a significant overlap between genetic and physical interactions in cancer cell lines is an interesting and promising observation that will help understanding known genetic dependencies and predicting new ones. However, similar observations have been made in other organisms and biological systems. These past studies should be referenced to provide a historical perspective and help define further analyses in the cancer context. In particular, studies in yeast S. cerevisiae have shown that, not only there is a general overlap between genetic interactions (both positive and negative) and physical interactions, but at least 2 additional features are informative about the relationship: a) the relative strength of genetic interactions and b) the relative density of physical interactions (i.e., isolated interaction vs protein complexes). Here's a sample of relevant studies: 1) von Mering et al., Nature, 2002; 2) Kelley & Ideker, Nat Biotechnol, 2005; 3) Bandyopadhyay et al., PLOS Comput Biol, 2008; 4) Ulitsky et al., Mol Syst Biol, 2008; 5) Baryshnikova et al., Nat Methods, 2010; 6) Costanzo et al., Science, 2010; 7) Costanzo et al., Science, 2016.

Similar observations have also been made in mammalian systems: e.g., in mouse fibroblasts (Roguev et al., Nat Methods, 2013) and K562 leukemia cells (Han et al., Nat Biotech, 2017). I don't think that past observations negate the novelty of this manuscript. The analysis presented here is more focused and more comprehensive as it is based on a large integrated dataset and is driven by a series of specific hypotheses. However, a reference to previous publications should be made.

As a frame of reference: my expertise is in high-throughput genetics of model organisms, including mapping and analyzing genetic interactions.

REFEREES CROSS COMMENTING

I agree with the questions raised by reviewer #1. And I think the authors should be able to address them (either through analyses or reasoning) within 1-3 months.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

Reproducibility of genetic interactions across studies is low. The authors identify reproducible genetic interactions and ask the question of what are properties of robust genetic interactions. They find that 1. oncogene addiction tends to be more robust than synthetic lethality and 2. genetic interactions among physically interacting proteins tend to be more robust. They then use protein-protein interactions (PPIs) to guide the detection of genetic interactions involving passenger gene alterations.

Major comments:

The claims of the manuscript are clear and well supported by computational analyses. My only concern is the influence of (study) bias on the observed enrichment of physical protein interactions among genetic interactions. 1. Due to higher statistical power the here described approach favors genetic interactions involving frequently altered cancer genes (as acknowledged by the authors). 2. Also some of the libraries in the genetic screens might be biased towards better characterized screens. 3. PPI networks are highly biased towards well studied proteins (in which well studied proteins - in particular cancer-related proteins - are more likely to interact). The following tests would help to clarify if and to which extend these biases contribute to the described observations:<br> 1 . The authors should demonstrate that the PPI enrichment in reproducible vs non-reproducible genetic interactions is not solely due to the biased nature of PPI networks. One simple way of doing so would be to do the same analysis with a PPI network derived from a single screen (eg PMID: 25416956). I assume that due to the much lower coverage the effect will be largely reduced but it would be reconfirming to see a similar trend in addition to the networks on which the authors are already testing. Another way would be to use a randomized network (with the same degree distribution as the networks the authors are using and then picking degree matched random nodes) in which the observed effect should vanish.

2 . What's the expected number of robust genetic interactions involving passenger gene alterations? Is it surprising to identify 11 interactions? This question could be addressed with some sort of randomization test: When selecting (multiple times) 47,781 non-interacting random pairs between the 2,972 passenger genes and 2,149 selectively lethal genes, how many of those pairs form robust genetic interactions?

Minor comments:

Two additional analyses would add in my opinion value to the manuscript:

-The authors state that reasons for irreproducibility of genetic interactions are of technical or biological nature. Is it possible to disentangle the contribution of the two factors given the available data? Eg how many genetic interactions are reproducible in two different screening platforms using the same cell line vs how similar are results of screens from two different cell lines in the same study?

-The authors state that "some of the robust genetic dependencies could be readily interpreted using known pathway structures" and argue that they recover for example MAPK or Rb pathway relationships. Is this a general trend? Do genes forming a robust genetic interactions have a higher tendency to be in the same pathway as opposed to different pathways? I think the pathway topic could be in general better exploited: eg does pathway (relative) position play a role?

Significance

Personalized cancer medicine aims at the identification of patient-specific vulnerabilites which allow to target cancer cells in the context of a specific genotype. Many oncogenic mutations cannot be targeted with drugs directly. The identification of genetic interactions is therefore of crucial importance. Unfortunately, genetic interactions show little reproducibility accross studies. The authors make an important contribution to understanding which factors contribute to this reproducibility and thereby providing means to also identify more reliable genetic interactions with high potential for clinical exploitation or involving passenger gene alterations (which are otherwise harder to detect for statistical reasons).

REFEREES CROSS COMMENTING

Reviewer 2 raises a few valid points, which if addressed would certainly increase the clarity of the paper. In particular addressing the first point (the self interactions of tumor suppressors) seems important to me. From what I can see all of reviewer 2's comments can be addressed easily.

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

Learn more at Review Commons

Reply to the reviewers

Reviewer #1

1.The stimulatory effect of LPHC diet on browning of some WATs has been previously reported (Nutrition, 42, 37-45 Oct 2017). Also, the activation of AMPK was observed in this study. However, the detailed mechanism responsible for AMPK activation by LPHC diet remains elusive in the present study, which lowers its scientific importance.

Response: OK. We will adequately mention previous study in which the AMPK activation was observed upon LPHC diet and more deeply decipher the molecular mechanisms that lead to AMPK activation by analyzing AMP- and ROS/Ca2+-dependent pathways according to the Reviewer’s suggestions (for more details, see point 2-4).

2.Different with WAT, LPHC diet increases glucose uptake and FA synthesis in BAT (Nutrition. 30 (4), 473-80 Apr 2014). __Is it possible that AMPK activation in WAT due to the lowered glucose uptake, which might increase AMP/ATP ratio? It is recommended to determine the uptake of glucose in WAT.__

Response: In the article cited by the Reviewer, authors measured glucose uptake only in BAT and found that it was significantly increased. On the contrary, no data are reported regarding glucose metabolism in WAT. Our in vitro data clearly indicate that AMPK activation occurs upon amino acid restriction (AAR) and in the presence of glucose in the culture medium (see Fig. 6C, Suppl. Fig. 5I). Moreover, glucose uptake is increased upon this condition (see Fig. 5G). Hence a decreased glucose uptake by WAT and the activation of AMPK via a decrease of AMP/ATP ratio has to be likely excluded. However, we will test AMP and ATP levels both in vivo and in vitro and this, together with experiments aimed at deciphering the contribution of mitochondrial ROS (mtROS) and CaMKK (see points 3, 4), we will hopefully clarify the mechanisms of AMPK activation upon LPHC diet.

The present study indicates that the promotional effect of LPHC diet on WAT browning is dependent on mitochondrial ROS generation. However, it is still unknown why the production of ROS increased and why ROS could activate AMPK. The authors should clarify these critical steps.

Response: Redox unbalance is widely reported to directly or indirectly stimulate AMPK activation (Shao et al., 2015, Cell Metab; Hinchi et al., 2018, J Biol Chem). Moreover, it has been demonstrated that activation of AMPK could depend on mtROS and be independent of an increase in AMP/ATP ratio (Emerlin et al., 2009, Free Radic Biol Med). Based on this evidence and our results, we believe that, upon AAR or LPHC diet, the recorded increase of mtROS concentration could not derive from an enhanced production but rather to a decrease of intracellular availability of the sulfur amino acid cysteine that represents an efficient ROS scavenger. Actually, by replenishing cysteine through N-acetyl cysteine (NAC) treatment we were able to buffer mtROS increase (see Fig. 6A), as assayed by cytofluorimetric analyses through mitoSox staining, and avoid AMPK phosphorylation (see Fig. 6C and 6J) as well as the downstream upregulation of brown fat and muscular genes (see Fig 6B and 6I). In line with this result, treatment with erastin, a cysteine depleting agent, was able to mimic the effects of AAR and LPHC diet by up-regulating the expression of brown-like and muscular genes (see Fig. 6G). Therefore, to more deeply decipher the mechanisms involved in AMPK activation and to further involve cysteine depletion in mtROS increase and AMPK activation, we could assay mtROS and AMPK levels also following erastin treatment. Of course, to involve cysteine decrease in AMPK activation, measuring intracellular cysteine levels upon AAR and LPHC diet is mandatory and will be carried out. Importantly, we have preliminary data, not included in the present manuscript, indicating that cysteine is decreased both upon AAR and LPHC diet; hence, after increasing the sample size, we will include this result in the revised version.

4.The relationship between cytosolic calcium and AMPK was not clear. In addition to the fact that AMPK regulates SERCA to increase cytosolic Ca depicted in the present study, AMPK could also be activated by increased Ca via CaMKK. A recent study indicates that the activation of AMPK requires TRPV4-mediated Ca release from ER (Cell Metabolism Volume 30, Issue 3, 3 September 2019, Pages 508-524.e12). This issue should also be clarified.

Response: Regarding the possible involvement of TRPV4-mediated Ca release from ER, through RNAseq we found that TRPV4 mRNA is slightly expressed in subcutaneous white adipose tissue and changes in its expression were not found upon LPHC diet. Moreover, TRPV4 protein was not detected in our samples by proteomic analysis. Notably, by integrating transcriptomic and proteomic data, it emerged that cell membrane intracellular calcium transporters (i.e. CACNG1, CACNA2D1), which are interconnected to the network of sarcoplasmic reticulum calcium cycle, are upregulated upon LPHC diet (see Fig. 5I). Therefore, we will evaluate the effects of a calcium channel blocker (e.g. Verapamil) and/or extracellular calcium chelator (e.g. BAPTA) on AMPK activation and its downstream gene expression cascade. In parallel, to possibly involve CAMKK in the activation of AMPK, treatment with a CAMKK inhibitor (e.g. Sto-609) will be carried out. Importantly, mtROS are upstream inducers of intracellular calcium raise (Mungai et al., 2011, Mol Cell Biol) and therefore an involvement of mtROS-Ca2+ axis could not be ruled out. In line with this hypothesis, by buffering mtROS through NAC treatment, we were able to abrogate intracellular calcium raise elicited by AAR (see Fig. 6F). Therefore, by performing the above described experiments and by evaluating CAMKK following NAC treatment, we will be hopefully able to establish whether AMPK activation is AMP-(in)dependent and/or relies on mtROS/Ca2+/CAMKK pathway.

Reviewer #2

o Interesting paper but see comments below.

Response: OK, thanks

o The relevance of the described effects for whole-body energy balance regulation is not shown. Indirect calorimetry could be interesting. The only whole-body effect (slightly improved glucose clearance in oGTT) was very small.

Response: OK. As suggested by this Reviewer we can include indirect calorimetry to give a more comprehensive view of the effects of LPHC diet on the whole-body energy balance (see also the following point).

o …1) Indirect calorimetry could be very helpful to show effects on energy metabolism. 2) Can the authors discuss why they didn't conduct the experiment also under thermoneutral conditions?

Response: OK. As stated above, we will add indirect calorimetry experiments and, as suggested by this Reviewer, we will discuss this issue in the revised version. Importantly, we already have indirect calorimetry data that were not included in the present version of the manuscript and that we will add in the revised version.

o Maybe an additional collaborator is necessary.

Response: Yes, collaborators who performed indirect calorimetry will be included as co-authors in the revised version.

o Article numbers of all diets must be added and information if the all diets were purified diets. This could have effects on the gut microbiome.

Response: OK. We will add the article numbers as well as more detailed information about all the diets.

o Sample sizes are very low. The authors should explain why only males were used in the experiments. oGTT analysis should also include calculation of area under the curve. No explicit statement if correction for multiple testing is required or other measures to reduce false positive results.

Response: We have used only male mice to avoid sex bias. We will edit the OGTT analysis graph to include calculation area under the curve. Regarding the sample size, a mistake occurred when the figure legends have been written. Actually, in materials and methods section, we clearly indicated the number of animals used (n=8 mice for WD and n=6 mice for LPHC diet and not n=3). Information regarding the statistical analyses was included in Bioinformatics and Statistical Analysis section. In this section, we described how the correction for multiple testing was carried out (i.e. one-way ANOVA followed by Dunnetts correction). In the revised version, we will dedicate a separate section for statistical analysis to avoid misreading.

**Minor comments:**

o Are prior studies referenced appropriately?* Relevant reference: Desjardins, E.M., Steinberg, G.R. Emerging Role of AMPK in Brown and Beige Adipose Tissue (BAT): Implications for Obesity, Insulin Resistance, and Type 2 Diabetes. Curr Diab Rep 18, 80 (2018). __https://doi.org/10.1007/s11892-018-1049-6____ __

Response: We thank the Reviewer for this suggestion and we will include and appropriately discuss this paper.

o *Are the text and figures clear and accurate?* YES

Response: Ok, thanks for this positive evaluation.

Reviewer #3

My main critique, coming from the perspective of a dietitian that works in human trials in the US, is that the diet called a "Western" diet is not similar to the diet that humans with metabolic problems typically eat…

Response: OK. The aim of this work was to study at molecular level the responses of white adipose tissue to changes in protein to carbohydrate ratio. We completely agree with the Reviewer that “Western” diet is not an appropriate term to describe the diet that we have used; hence, we will change “Western diet” in “Control diet” throughout the manuscript. Actually, according to the general guidelines for nutrition studies on mice, when experimental animals are fed a special diet (i.e. LPHC in our study), the control animals should be fed a diet matched in every way to the special diet, except of course for the dietary variable (i.e. P/C ratio in our study) that the researcher is studying (Pellizzon and Ricci, The common use of improper control diets in diet-induced metabolic disease research confounds data interpretation: the fiber factor (2018). Nutrition & Metabolism 15:3).

**Major comments:**

-The authors provide strong support their key findings.

Response: We thank this reviewer for this positive evaluation.

-The mice were on the LPHC diet for a short period of time (2 weeks). Ongoing amino acid deficiency has potential to promote frailty and other deleterious outcomes. No long-term diet outcomes can be inferred from this study.

Response: OK. We will discuss this issue, highlighting that this dietary regimen should be recommended on human only for a short period and that further study is needed for understanding the long-term effects of LPHC diet.

-The authors have provided no evidence that a LPHC diet improves human health, so I think they need to scale back those assertions, particularly as it relates to people shifting to a LPHC from what they currently eat, since people don't typically eat what the authors refer to as a "Western" diet as it's defined in this paper.

Response: OK. We will reference studies in which LPHC diet has been suggested to improve human health.

-As far as I can tell, no additional experiments are needed to support their claims identifying how the LPHC affects AMPK activated pathways in mice.

Response: We thank this reviewer for this positive evaluation.

-The methods are rigorous and sufficiently described to be reproducible.

Response: We thank this reviewer for this positive evaluation.

**Minor comments:**

-__Minor grammatical issues through e.g. "It is worth to notice" in last paragraph on page 12; there are font differences in the methods section __

Response: OK. We will correct these minor grammatical/font issues.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

The authors use a mouse model to compare molecular responses to a 23% protein/57 carbohydrate/20 fat diet to a 7% protein/73% carbohydrate/20% fat diet. The authors show that the low protein diet enhanced activation of biological pathways related to fatty acid catabolism including FAO, TCA cycle and electron transport chain in sWAT but not BA, similar to cold exposure. The authors use redundant assays and experiments in cell models to validate the genes and molecular pathways involved in the sWAT response to a low protein diet in mice. The authors show that AMPK activation promotes the induction of typical brown fat and muscular genes in sWAT. The authors identify novel non-canonical pathways (Serca1 and Serca2a,) that are upregulated in sWAT browning.

My main critique, coming from the perspective of a dietitian that works in human trials in the US, is that the diet called a "Western" diet is not similar to the diet that humans with metabolic problems typically eat. The typical US diet is closer to approximately a 17% protein/50% carbohydrate/33% split (https://doi.org/10.1016/j.nut.2015.02.007, https://doi.org/10.1038/s41430-017-0031-8). This level of protein utilized for the experimental "Western" diet here is comparable to levels used for "high protein" diets in some human studies (https://doi.org/10.1111/nure.12111).

Since the experimental diet differs substantially from what metabolically sick people typically eat, the ability to speculate how the findings from this study may apply to humans with metabolic diseases is very limited. This paper is really well-done, but I think the authors should call the experimental diet a high-protein, moderate carbohydrate diet (HPMC), not a "Western" diet. There are many who argue that such a HPMC is metabolically advantageous and promotes weight loss/improved body composition, so this study lays the groundwork for refuting that guidance. It would be exciting to see a head to head comparison of the two diets in humans in the future!

Major comments:

-The authors provide strong support their key findings

-The mice were on the LPHC diet for a short period of time (2 weeks). Ongoing amino acid deficiency has potential to promote frailty and other deleterious outcomes. No long-term diet outcomes can be inferred from this study.

-The authors have provided no evidence that a LPHC diet improves human health, so I think they need to scale back those assertions, particularly as it relates to people shifting to a LPHC from what they currently eat, since people don't typically eat what the authors refer to as a "Western" diet as it's defined in this paper.

-As far as I can tell, no additional experiments are needed to support their claims identifying how the LPHC affects AMPK activated pathways in mice.

-The methods are rigorous and sufficiently described to be reproducible

Minor comments:

-Minor grammatical issues through e.g. "It is worth to notice" in last paragraph on page 12; there are font differences in the methods section

Significance

The work is significant as it describes the metabolic effects of a LPHC at the molecular level for the first time. This paper demonstrates how a low protein diet may promote longevity and improve glucose metabolism, which has been shown to some extent in humans, but hasn't had a mechanistic explanation until now.

If similar findings were supported in longer term animal and human trials, it could lay the groundwork for modifying dietary recommendations to promote metabolic health and longevity.

This paper is of interest to basic scientists studying diet and energy metabolism. The potential health implications are interesting to people in healthcare and scientists studying human metabolism.

I am a dietitian who has conducted weight loss trials in humans, emphasizing varying macronutrient ratios. I have also done whole body metabolism work in humans using metabolic chambers. I have experience in urinary proteomics, but I lack sufficient expertise to scrutinize much of the methodology of the basic work you present here.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Aquilano et al. submitted a manuscript investigating the effects of a low-protein/high-carbohydrate diet on AMPK-dependent thermogenic activity in subcutaneous adipose tissue in mice presumably resulting in stimulated energy dissipation. Based on the observation that LPHC diets may promote metabolic benefits the authors aimed to study the underlying molecular functions. They focused mainly on a comparison of molecular markers for thermogenesis and the related metabolic pathways in brown and subcutaneous white adipose tissue in response to feeding mice a LPHC diet for two weeks. Using a proteomics approach first, they identified 75 proteins differentially present in sWAT compared to BAT. These could be linked both to canonical as well as non-canonical (muscular) thermogenic functions as the authors state. Overall, they conclude that feeding a LPHC diet induces a white-to-brown conversion in sWAT. Deep RNA-sequencing identified 416 up and 52 down-regulated gene transcripts in sWAT. GO terms analysis showed enrichment for biological processed related to mitochondrial fatty acid catabolism, response to cold, and muscle contraction genes. Following up this rational, they conducted several experimental approaches to identify regulators in this system. For example, they tried to rule out that changes in gut microbiome composition could mediate metabolic benefits in response to LPHC diet. Finally, they hypothesized that nutrient shortage in particular amino acid lowering is responsible for sWAT browning. Here, AMPK seems to play a central role in the browning of sWAT in response to LPHC diet.

Major comments:

o Are the key conclusions convincing? YES interesting paper but see comments below.

o Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? YES - The relevance of the described effects for whole-body energy balance regulation is not shown. Indirect calorimetry could be interesting. The only whole-body effect (slightly improved glucose clearance in oGTT) was very small.

o Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary to evaluate the paper as it is, and do not ask authors to open new lines of experimentation. YES - 1) Indirect calorimetry could be very helpful to show effects on energy metabolism. 2) Can the authors discuss why they didn't conduct the experiment also under thermoneutral conditions?

o Are the suggested experiments realistic for the authors? It would help if you could add an estimated cost and time investment for substantial experiments. Maybe an additional collaborator is necessary.

o Are the data and the methods presented in such a way that they can be reproduced? YES mostly - but article numbers of all diets must be added and information if the all diets were purified diets. This could have effects on the gut microbiome.

o Are the experiments adequately replicated and statistical analysis adequate? Sample sizes are very low. The authors should explain why only males were used in the experiments. oGTT analysis should also include calculation of area under the curve. No explicit statement if correction for multiple testing is required or other measures to reduce false positive results.

Minor comments:

o Are prior studies referenced appropriately? Relevant reference: Desjardins, E.M., Steinberg, G.R. Emerging Role of AMPK in Brown and Beige Adipose Tissue (BAT): Implications for Obesity, Insulin Resistance, and Type 2 Diabetes. Curr Diab Rep 18, 80 (2018). https://doi.org/10.1007/s11892-018-1049-6

o Are the text and figures clear and accurate? YES

Significance

My expertise: Energy metabolism in rodent models for metabolic disease, body temperature regulation, body mass regulation

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This work found that a LPHC meal activates browning of sWAT by ROS/AMPK pathway, and tried to clarify the detailed mechanism of the beneficial effect of LPHC diet. Although the paper contains scientific novelty and is well-written, most of the results are descriptive and deeper mechanistic study seems lacking. Here listed some comments and questions.

1.The stimulatory effect of LPHC diet on browning of some WATs has been previously reported (Nutrition, 42, 37-45 Oct 2017). Also, the activation of AMPK was observed in this study. However, the detailed mechanism responsible for AMPK activation by LPHC diet remains elusive in the present study, which lowers its scientific importance.

2.Different with WAT, LPHC diet increases glucose uptake and FA synthesis in BAT (Nutrition. 30 (4), 473-80 Apr 2014). Is it possible that AMPK activation in WAT due to the lowered glucose uptake, which might increase AMP/ATP ratio? It is recommended to determine the uptake of glucose in WAT.

1. The present study indicates that the promotional effect of LPHC diet on WAT browning is dependent on mitochondrial ROS generation. However, it is still unknown why the production of ROS increased and why ROS could activate AMPK. The authors should clarify these critical steps.

4.The relationship between cytosolic calcium and AMPK was not clear. In addition to the fact that AMPK regulates SERCA to increase cytosolic Ca depicted in the present study, AMPK could also be activated by increased Ca via CaMKK. A recent study indicates that the activation of AMPK requires TRPV4-mediated Ca release from ER (Cell Metabolism Volume 30, Issue 3, 3 September 2019, Pages 508-524.e12). This issue should also be clarified.

Significance

This work indicates that LPHC diet promotes browing of WAT through activation of AMPK by elevating mitochondrial ROS production. Compared to previous studies, this work firstly found the critical importance of mitochondrial ROS in activation of AMPK through a series of works on omics data. However, they failed to clearly explain the detailed mechanism responsable for either enhanced mitochondrial ROS production by LPHC diet or activation of AMPK by mitochondrial ROS. Therefore, due to most of the conclusions have been presented in some previous published papers, the main novelty of the present work should be greatly improved by further mechanistic stidies.

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

Learn more at Review Commons

Reply to the reviewers

I thank the referees for their enthusiasm and time providing critical feedbacks to our manuscript. The novelty of our work is the identification of the importance of Mfn2 in regulating the Rac signaling and neutrophil migration& adhesion, which is significantly relevant to the mitochondrial field and cell biology in general. Below please find our point-to-point response to the comments.

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

Introduction:

"Although mitochondria-derived ATP possibly regulates neutrophil chemotaxis in vitro (Bao et al., 2015), removal of extracellular ATP improves neutrophil chemotaxis in vivo (Li et al., 2016). These conflicting reports prompted us to search for mechanisms delineating the role of mitochondria in neutrophil migration outside the realm of ATP or cellular energy (Bi et al., 2014; Schuler et al., 2017; Zanotelli et al., 2018)." This sentence is superficial and misleading: extracellular ATP may interfere with chemotaxis through various energy-independent mechanisms (see for example Zumerle et al. Cell Reports 2019) and this is not conflicting with the role of intracellular ATP in migration.*

We were not clear in the writing that Bao et al suggest that neutrophils secret ATP at the leading edge and mitochondria at the leading edge is the source of the extracellular ATP. Both studies focused on extracellular ATP. We agree that the reports are not necessarily conflicting since exogenous ATP can induce additional signaling. We rewrote this sentence emphasizing that we are looking for mechanisms in addition to ATP, which is distinct from previous studies.

Figure 1: The authors didn't show evidence of the genome edition (PCR, RFLP or Sequencing over the sgRNA target) or at least RT-PCR or WB for MFN2. In Fig 1b, 1c the scale bar is missing. "Neutrophils were sorted from both lines and their respective loci targeted by the 4 sgRNAs were deep sequenced." There are no data about sorting strategies for zebrafish neutrophils in the figure. Moreover, only 2 sgRNAs are shown and there are no sequencing data.

To show evidence of the genome edition, we have deep sequenced this loci of mfn2 and opa1 and the mutation frequencies were stated in the original text. The sorting strategies were described in Methods-Mutational Efficiency Quantification. Each mfn2 KO has 2 individual sgRNAs, and two KO (mfn2 KO and mfn2 KO#2) were shown in Fig 1b, so there are 4 sgRNAs targeting mfn2. Since each embryos have approximately 150 neutrophils, WB is not feasible. Sequencing is the standard method (Ablain et al., 2015; Zhou et al., 2018). We only stated the mutation efficiency in the manuscript because amplifying the genomic DNA from the sorted cells introduces PCR bias and the numbers are not a quantitative reflection of the degree of gene disruption. We will include the sequencing result of the sgRNA target sites in a supplemental Figure.

We used one scale bar for all the panels in Fig 1b,c. All panels are at the some magnification.

Figure 2:** In the WB, reconstitution is not obvious. In general, all WBs are not quantified (and they should be quantified). The in vivo experiment does not have proper controls. For example, can the authors exclude that in these mice there is reduced inflammation because neutrophils have defective activation? What about NETs? And cytokines/chemokines? And exocytosis? In the absence of these controls, the experiment cannot be properly interpreted.

We have quantified all WBs in our study. The results were sometimes stated in the text only. We will add the quantifications to each blot.

The mice model we chose is used to evaluate in vivo neutrophil migration. We used a neutrophil specific promoter to delete mfn2 in mice and collected data at a very early time point when the tissue inflammatory environment is determined by tissue resident sentinel cells, such as macrophages. Although our results support that mfn2 is required for neutrophil migration in mammals, we agree that we can not fully rule out that other neutrophil functions are also regulated by mfn2.

To address whether other neutrophil functions are affected by MFN2, we will performed assays to evaluate NETosis and degranulation in MFN2 KD HL-60 cells to evaluate the other neutrophil functions.

Figure 3: The conclusion of the authors "In summary, Mfn2 modulates the actin cytoskeleton and cell migration in MEFs" should be supported by experiments to distinguish between the specific role of Mfn2 and the role of mitochondrial dynamics (Opa1, Drp1, Mfn1). It is also not clear why the authors decided to use MEFs instead of other cells (more similar to neutrophils which are not adherent cells). The results obtained in MEFs may be irrelevant for neutrophils.

We agree that MEFs are very different from neutrophils. We chose MEFs since the function of Mfn2 in MEF is well characterized (Chen et al., 2003; de Brito and Scorrano, 2008; Naon et al., 2016). Both Mfn1 and Mfn2 MEF have fragmented mitochondria. Mfn1, which is very similar to Mfn2, serves as the best control. We will confirm the mitochondria structure in the KO cells.

For specificity, in addition to mfn2, we looked at Mfn1 and opa1 in different systems. We did not select Drp1 since the mitochondrial network in neutrophils is highly fused (Fig 4 and 5)(Maianski et al., 2002; Zhou et al., 2018).

We have also knocked down Opa 1 in HL-60s. We observed massive cell death in this line and cell migration is affected, possibly due to a depletion of cellular ATP as reported (Amini et al., 2018). We will include the data showing cell death, qRT to show knockdown efficiency and chemotaxis. In zebrafish neutrophils, knocking out Opa1 also reduced cell migration (Fig 1S).

Figure 4-5: Fig 5a: in ctrl and sh1 the ER seems to be larger than the phalloidin (=cytoskeleton=cell border approximately) in a few regions. Only the sh1+T seems to fit correctly.

We use the F-actin staining as an indicator of cell front. F-Actin is predominant at cell front, but much less in the cell body and uropod. Here we set the confocal laser power at a certain level to give us a good resolution of brighter signals which may not be strong enough to detect signals in the cell body. That’s why the fluorescence is very dim or even absent in the cell body. However, the majority of ER do fit in the cell border if look closer.

The TEM image (only 1 in supplementary) is not sufficient to convince that the tethering is lost. Quantification of number of contacts and distance between ER and mitochondria should be included.

Using EM method, Mfn2 ablation decreases the ER-contacting mitochondrial surface by ∼20–35% (Naon et al., 2017). Using the same cells, different groups reached different conclusions using TEM(Filadi et al., 2017). We reason that ER-mitochondria contact sites are rare events in TEM since the samples are sliced. We will try to take more TEM images to quantify the distance. However, we are not sure that we can come up with a definitive conclusion by TEM. Nevertheless, we observed significant mitochondrial structural changes using IF and observed the changes in cytosolic calcium levels, which is consistent with the known function of Mfn2 as a ER-mitochondrial tether (Naon et al., 2016).

The title of figure 5 is wrong. However, in these figures, it is clear that cells are beautifully polarized, with mitochondria accumulating at the uropod (and even more in the absence of Mfn2). When comparing these images with those published by Campello et al (JEM 2006), there are 2 observations that can be made: first of all, these data confirm that mitochondrial fission promotes cell polarity; second, they suggest that the defect is not at the level of cell polarity/chemotaxis.

We have fixed the title of figure 5.

We agree that mfn2 defective neutrophils does not have a defect in cell polarization. The defects in migration is possibly due to other reasons such as poor adhesion or regulation in the actin cytoskeleton dynamics. However, our data is not sufficient to support that mitochondrial fission promotes cell polarization and chemotaxis.

Figure 6: Calcium data are, in general, very weak. First of all, controls with ionomycin are missing. Statistical analyses of the curves should be included. As for the use of the MCU inhibitor Ru360, is there any evidence that it is cell-permeant in this context? Is it blocking MCU? Since the authors can show mitochondrial calcium upon FMLP, they should also demonstrate that Ru360 is indeed working and inhibiting mitochondrial calcium uptake. The sentence "The MCU inhibitor Ru360did not cause further reduction of chemotaxis in MFN2 knockdown dHL-60 cells (Supplementary Fig. 6c, d and Supplementary Movie. 12), indicating that MCU and MFN2 lies in the same pathway in terms of regulating chemotaxis in dHL-60 cells" is speculative. In general, there is no solid demonstration that the effect is calcium-mediated.

We will include the control of ionomycin and include statistics of the results.

Ru360 is a widely used MCU inhibitor. The fact that Ru360 itself inhibited neutrophil migration supported that the chemical enters cells. We agree that stating “indicating that MCU and MFN2 lies in the same pathway in terms of regulating chemotaxis in dHL-60 cells" is speculative. In addition, we tried to reduce cytosolic calcium levels in mfn2 KD cells either using Ca2+ chelator (BAPTA, in Fig S6) or an IP3 receptor inhibitor. In both cases we observed reduced migration blocking calcium signal alone. The mfn2 KD phenotype was not rescued. This could due to that multiple molecules/pathways are calcium dependent in cell migration. We will include all the negative data. We thus far are still unable to establish a functional link of the calcium with mfn2 regulated signaling.

We have moved the calcium data to Fig 4. The elevated calcium signal is an indirect evidence to support the loss of ER-mitochondria tether. We have modified our conclusion to leave out calcium as a relevant signal regulated by mfn2 for neutrophil migration.

As for Rac, it is surprising to see that Rac inhibition has no effect on cell migration. Rac is known to promotes migration in fibroblasts and other cell types and Rac deficiency inhibits migration (see for example Steffen et al, JCS 2013). Two sets of experiments are absolutely required: 1) verify this in fibroblasts since it has been elegantly shown that Rac is essential in these cells for migration; 2) analyse the effect of Rac inhibitors in pPak kinetics.

Rac is required for cell migration and the growth of branched actin network. The Rac inhibitors we selected here are specific to two rac GEFs, vav and Tiam. Steffen et al, JCS 2013 used Rac1 KO MEF, which is different from ours. Thus the works are not contradictory. MEFs are very different from neutrophils. We chose MEFs since there are knockout cells available and well characterized. The MFN2 KO cells display prominent lamellipodia, which is also consistent with the observation in Steffen et al, JCS 2013. We have used these two inhibitors in MEF wound closure and did not observe a strong phenotype.

We will analyze the effect of Rac inhibitors in pPak kinetics in the control and Mfn2 deficient dHL-60 cells.

*Reviewer #1 (Significance (Required)):**

As presented here, the manuscript has a modest significance. The audience would be specialised: cell migration, cell signalling. My expertise is immunology, cell activation, cell migration, cell signalling.

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

**Major comments:**

Although the results could be very interesting, and could be significantly relevant to the mitochondrial field and the cell biology one in general, major points need to be addressed to fully support conclusions of the authors. Different controls and quantification are missing, Actin dynamics analysis should be improved, effects of the artificial tether is weakly characterized and the demonstration of the specific role of mito-ER contacts via mfn2 in migration should be reinforced.

-In figure 1, quantification of circulating neutrophils is required in Mfn2 KO embryos. The authors should also show these quantified results for OPA1KO, which are just mentioned in the text. In addition, in figure 1b and d, the neutrophils from the Mfn2KO embryos seem bigger compared to control. Can the authors comment on neutrophils size and potential contribution to the phenotype? Finally, the authors propose a defect in neutrophil migration in Mfn2-KO, however neutrophils are found in the circulation. The authors should explain these results.*

Since the cells are all in circulation, we can only estimate the percentage. Overall, the phenotype is drastic, shown in movie S1. We will state how many fish embryos we have imaged and how often we observe this phenotype (only 1 or 2 in the tissue (mfn2 KO) or in circulation (control)). The bigger spots are resulted from cells outside the focal plane-zebrafish embryos are thick tissues. We agree that since neutrophils in the KO fish are all circulation, we cannot make a conclusion whether they can migrate in tissue in zebrafish. We conclude that “mfn2 regulates neutrophil tissue retention and extravasation in zebrafish”, but did not comment on chemotaxis.

-The authors need to reinforce the Mfn2 specificity for their phenotype. In particular in Fig S1, they show that loss of OPA1 significantly decreases neutrophil migration in vivo. However, they then only study the effect of Mfn1 silencing in neutrophil and MFN1 KO MEFs (Sup Fig s3). The authors should perform the same experiments in neutrophil and MEF upon loss of OPA1 (similar to Fig S3). Does loss of OPA1 and Mfn1 decrease neutrophil arrest to activated endothelial cells?

We knocked down OPA1 in HL-60 cells. The cells appear unhealthy and display a migration defect, consistent with the data in zebrafish. We are not comfortable making conclusions here since secondary affects in dying cells may cause any phenotype not directly attributed to the loss of OPA1. Nevertheless, we will include the data.

We have decided not to include Opa1 KO MEF since the cell morphology as documented in ATCC is similar to that of WT MEF. Only the MEF2 KO MEF is more circular. MFN1 KO MEF is a better specificity control which we have characterized in depth.

Since Mfn1 KD HL-60 cells migrate well on surface, they are not expected to have adhesion defects. Nevertheless, we will determine whether loss of Mfn1 decrease neutrophil arrest to activated endothelial cells and include the data.

-Using their images, the authors should also document on the directionality of the cell during cell migration. Do Mfn2 depleted cells do not migrate because they are arrested or because they are lacking directionality? Environment/chemokine sensing defects?

We will quantified the directionality of the cells. As pointed out by reviewer 1, mfn2 deficient cells can polarize and not defective in chemokine sensing. We do not expect a significant change in directionality defect.

-Actin dynamics analysis should be improved. Loss of Mfn1 and Mfn2 lead to cell shape changes. The authors should quantify this phenotype by analysing cell circularity (as well as for Opa1 loss). Stress fibres number or Phalloidin intensity quantification in cell body should also be performed.

We will quantified the circularity, stress fiber numbers and phalloidin intensity in Mfn1 and Mfn2 KO MEFs.

-Can the migration defects could be attributed to Focal adhesion protein dynamics defects? The authors shown an hyperactivation of Rac1 and an hyperphosphorylation of PAK, which can control FAP (focal adhesion proteins) dynamics. In addition, immunofluorescence analysis shows a decreased signal and cellular misdistribution of paxillin. The authors should characterize these phenotypes. FAP levels (Paxillin/Phospho-Paxillin and Vinculin) should be analysed by immunoblot, the number of FAP/cell, distribution and size should also be quantified. Their dynamics should also be analysed by live cell imaging. Finally, Paxillin level and distribution seems to be also impacted in Mfn1KO cells. Can the authors comment on that? The different quantifications would help to better understand the effect of different mitofusins in cytoskeleton dynamic.

We thank the reviewer for the great advices for our follow up work. So far our results supports Rac over activation as a relevant pathway how mfn2 regulates neutrophil migration. Although Rac can regulate focal adhesion dynamics in other cells (Rooney et al., 2010), how Rac activation regulates focal adhesion dynamics in neutrophils is not clear. Mfn2 regulated membrane tether could affect lipid trafficking, cellular metabolism and other signaling molecules. It will take substantial amount of work to make a conclusion and it is more suitable a separate report. This is one of the directions we will pursuit in our future studies.

-Please perform rescue experiments for cell migration in MFN2KO and MFN1KO MEFs. Immunoblots showing protein levels of these proteins would be appreciated. To really discriminate how Mfn2 regulates cell migration, the authors should also perform rescue experiments using a fusogenic mutant Mfn2 ((K109A). It will help to demonstrate the relevance of mito-ER contacts and not mitochondrial fusion in the phenotype.

For the reason mentioned above, we do not plan to do additional experiments in MEF cells since this work is focused on neutrophils. It is documented that Mfn2 K109A cannot restore mitochondrial fusion. However, it is not clear whether this construct can restore ER-tether. Result using this construct will be hard to interpret.

-Figure 4, the authors stipulate that Mfn2 regulates ER-mitochondria tethering. However, the authors present no evidence for this conclusion. The authors should perform manders coefficient in MFN2 KO cells and compared it to control. Also, loss of Mfn2 induces mitochondrial fragmentation, which can lead to problem for mito-er contacts quantification by light microscopy. The authors should use their TEM pictures to quantify mito-ER contacts (Number, length and % of mito perimeter), not only mitochondrial morphology. Mfn1 should be used as negative control. it would be interesting also to determine the status of the mito-ER contact in the different conditions used in the manuscript to stimulate cell migration like fMLP treatment.

We have performed manders coefficient in the mfn2 KD cells and observed no difference compared with the control. It is possibly due to the prevalent ER structure in the cells-despite the structural change, mitochondria are still mostly on top of ER when examined using IF. Using EM method, Mfn2 ablation decreases the ER-contacting mitochondrial surface by ∼20–35% (Naon et al., 2017). Using the same cells, different groups reached different conclusions using TEM(Filadi et al., 2017). We reason that ER-mitochondria contact sites are rare events in TEM since the samples are sliced. We will try to take more TEM images to quantify the distance. However, we are not sure that we will come up with a definitive conclusion by TEM. Nevertheless, we observed significant mitochondrial structural changes using IF and observed the changes in cytosolic calcium levels, which is consistent with the known function of Mfn2 as a ER-mitochondrial tether (Naon et al., 2016).

-The authors use an artificial tether to manipulate mito-ER contacts in cellulo. However, no information from its origin, or its design are documented in the manuscript. In addition, the authors should show that this tether efficiently works by analyzing mito-ER contacts upon expression by EM and mitochondrial calcium uptake. Does this tether rescue mito-ER contacts defects induced by loss of Mfn2? How the authors explain that the tether rescues mitochondrial morphology defects in MFN2KO? In these conditions, mitochondria should not be able to fuse anymore as Mfn2 is lost? This is really intriguing results. Does the tether rescue the other parameters? Mitochondrial distribution (with quantification)? Cell shape? Paxillin defects? ROS and membrane potential? These rescue experiments analyses are important to determine which parameters are really involved in cell migration defects due to the decreased tethering. Finally, it would be of great interest to analyse the effect of the tether alone on cell migration, Rac1 activity, cell shape? Gain of function? These results may reinforce the idea that contact sites regulate cell migration.

The tether is a GFP protein carrying both ER and mitochondrial localization sequences at the ends (Kornmann et al., 2009). The details are now added to the manuscript.

In HL-60 mfn2 KD cells, tether expression partially rescues mitochondrial distribution (quantified in Fig 5c), cell migration and Rac over activation. Although ROS and membrane potential are slightly affected by Mfn2 deletion in HL-60 cells, it is not clear whether they play any roles in mfn2 regulated cell adhesion or migration. We will attempt to use TEM to determine the mitochondrial structure upon tether rescue.

Despite multiple attempts, we could not obtain a line over-expressing the tether in wt HL-60 cells. We suspect that further increase in the tether is toxic to the cells.

-It is well established that a decrease of membrane potential leads to a decrease of mitochondrial calcium uptake. Calcium results obtained by the authors without any information on the roles of the tether could not lead to any conclusion. Does the tether rescues membrane potential and calcium uptake by the mitochondria? So far, the decrease of mitochondrial calcium upon stimulation in Mfn2KO cab be attributed to both mito-ER contact or membrane potential defects. It has been shown that MEFs MFN2 KO can lead to a decrease of MCU provel level leading to a decrease of mitochondrial calcium uptake (PMID: 25870285). The authors should also check MCU protein level.

We observed that mfn2 deficiency resulted in a minor reduction in membrane potential. Although Mfn2 KO MEF has reduced level of Mcu, Mfn2 silence in MEF does not affect Mcu levels (Filadi et al., 2015). Another group also concluded that Mfn2 deletion does not necessarily affect Mcu levels (Naon et al., 2016). Nevertheless, we will measure the MCU protein level in the Mfn2 knockdown HL-60 cells.

-Hyperactivation of Rac1 is only based on phosphorylation of PAK, which is quite weak. The authors should better describe the hyperactivation of RAC1 or other RhoGTPases in their Mfn2 KO MEFs. What are the levels of RAC1 and other RhoGTPases? Subcellular distribution in the cell? Kits are also available to determine RhoGTPase activity by pull down assay (Cell biolabs).

In Mfn2 KO MEFs, Rac overactivation is suggested by the increased lamellipodia formation, classical Rac readouts. Since the current manuscript focuses on neutrophils, we will performed the Rac GFP pull down experiments in HL-60 cells. We will also stain Rac GTP in HL-60 cells.

*-The references are up-to-date. The text and the figures are clear and accurate.**

**Minor comments:**

-The authors should show the efficiency of the KO generated for Mfn2 and Opa1 in zebrafish embryos. Sequencing results to highlight the position of the mutations and their consequences on the coding protein should be shown, as well as immunoblot analysis should be performed to analyse Mfn1, Mfn2 and OPA1 protein levels. The generation of a MFN1-KO transgenic line would have been appreciated to finely compare the roles of the 3 GTPases involved in mitochondrial fusion during neutrophil infiltration and migration in vivo.*

To show evidence of the genome edition, we have deep sequenced this loci of mfn2 and opa1 and the mutation frequencies were stated in the original text. Since each embryos have approximately 150 neutrophils, WB is not feasible. Sequencing is the standard method (Ablain et al., 2015; Zhou et al., 2018). We only stated the mutation efficiency in the manuscript because amplifying the genomic DNA from the sorted cells introduces PCR bias and the numbers are not a quantitative reflection of the degree of gene disruption. We will include the sequencing result of the sgRNA target sites in a supplemental Figure.

The mfn1 gene in zebrafish is duplicated. We are not sure whether we can obtain efficient disruption at both loci. We hope the results using Mfn1 KO MEF and MFN1 KD HL-60 cells are enough to show a specific role of Mfn2 in cell migration.

-MFN1, MFN2, AND OPA1 protein levels should be analysed by immunoblot in the Mfn1 and Mfn2 KO MEFs.

It is unlikely that mfn1/2 KO will affect OPA1 levels (Saita et al., 2016). Both MFN1 and MFN2 MEF display fragmented mitochondrial network which can be rescued by overexpression of MFN1 or MFN2 (Chen et al., 2003). The level of OPA1 in the cells are not relevant. We will stain mitochondria in the mfn1/2 KO MEFs to make sure that the cells have fragmented mitochondria as expected.

-In cell spreading assay, it would be great to identify cells during the process, by an asterix for example. "wt MEFs extended transient filopodia and lamellipodia and eventually elongated, whereas Mfn2-null MEFs generated extensive membrane ruffles and retained the circular shape". It would be interesting to quantify these different parameters.

We will add Asterixes to the cells. We will quantify the percentage of cells that can rearrange their cell shape in the WT and Mfn2 KO MEFs.

-For all their immunoblot analysis, the authors should use a mitochondrial marker as loading control (VDAC1, TOM20, HSP60...). In figure 5, Vinculin should not be used a loading control, with its role in focal adhesion dynamics.

Vinculin is stable in HL-60 cells under multiple conditions and selected as a control. The signal intensity correlates well with the amount of protein loaded. Using mitochondrial proteins as loading controls is not common and may be risky as the amount of mitochondria in cells can be variable.

-Legends for figures 5 and 6 are inverted.

Thanks, we have changed the heading of figure 5. The legends were correct.

-Please document in the material and methods section, how confocal images have been acquired: number of z-stacks, reconstitution, 3D analysis...

We will update in the method the parameters of imaging acquisition.

-The authors should show their results of blood cell composition quantification in ctrl vs MFn2 depletion.

We will include the results of blood cell composition in a supplemental figure.

-The authors should describe all the acronyms used throughout all the manuscript. For example, LTB4, fMLP...

We have describe all the acronyms in the updated manuscript.

*Reviewer #2 (Significance (Required)):

Beyond their role in energy production, mitochondria are involved in numerous cellular functions including cell migration. Mitochondria form a network balanced by fission and fusion events, where membrane contact sites with the endoplasmic reticulum are crucial. These contact sites are also involved in mitochondrial and cellular functions via their capacity to exchange lipids, metabolites and calcium. The role of mitochondria in cell migration has started to emerge where mitochondrial fragmentation and/or mitochondrial calcium homeostasis are acknowledged to drive cancer cell migration and to regulate actin dynamics. In this manuscript, Zhou W and colleagues proposed for the first time the role of mitochondria-ER contacts in cell migration. Mechanistically, this can be associated to the capacity of these contacts to control mitochondrial functions or mitochondrial calcium homeostasis. These findings are physiologically relevant and of particular interest to the mitochondrial and cell migration field but also to general cell biology. It represents a novel function associated to these membrane contact sites and point-out these contacts as signalling platform creating microdomains of metabolites exchanges involved in cell migration.

Keywords: Mitochondria - Membraned dynamics - calcium homeostasis - Membrane contact sites*

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

Mitofusin 2 (Mfn2) is a mitochondrial outer membrane protein that is important for mitochondrial fusion and the establishment of mitochondrial ER contacts. It has been published before that these contact sites are important for calcium signaling. Zhou et al. examined the role of Mfn2 in neutrophils. They propose a model in which mitochondrial ER contacts established via Mfn2 are crucial for regulation of Rac, which is a small GTPase driving cell migration by promoting actin polymerization. Loss of Mfn2 results in elevated cytosolic calcium, over-activation of Rac, and defects of chemotactic movements. These defects can be partially rescued by restoration of mitochondrial ER contacts through expression of an artificial tether protein.

**Major points**

1.The authors claim on p. 6 that decreased neutrophil retention is not simply due to defects in mitochondrial fusion. However, the experimental setup they used for mfn2 (Fig. 1) is different from that for opa1 (Fig. S1), and therefore the results are not directly comparable. Unfortunately, the authors don't show fragmentation of mitochondria, neither in mfn2 nor in opa1 depleted cells. To support their statement they must show that lack of Mfn2 and Opa1 causes mitochondrial fragmentation to the same extent and then examine neutrophil retention in the same assay. Also, it would make sense to include Mfn1 in this analysis, as the authors later claim that the effects they observed are specific for Mfn2.*

Since Mfn2 KO neutrophils are not in tissue, the experiment in Fig 1S to look at cell speed in tissue is not feasible. Since the cells are all in circulation, we can only estimate the percentage. Overall, the phenotype is very drastic, see movie S1. We will state how many fish embryos we have imaged and how often we observe this phenotype (only 1 or 2 in the tissue (mfn2 KO) or in circulation (control and opa1 KO)). Opa1 KO neutrophils are not in circulation.

To show evidence of the genome edition, we have deep sequenced this loci of mfn2 and opa1 and the mutation frequencies were stated in the original text. Since each embryos have approximately 150 neutrophils, WB and other biochemical assays are not feasible. Sequencing is the standard method (Ablain et al., 2015; Zhou et al., 2018). We only stated the mutation efficiency in the manuscript because amplifying the genomic DNA from the sorted cells introduces PCR bias and the numbers are not a quantitative reflection of the degree of gene disruption. We will include the sequencing result of the sgRNA target sites in a supplemental Figure.

Since Mfn2 KO neutrophils are all in circulating, we cannot observe their mitochondrial morphology. This is the reason why we used HL-60 cells for the mechanistic study. The mfn1 gene in zebrafish is duplicated. We have generated an mfn1b KO line and did not observe any phenotype. However we are not sure whether we can obtain efficient disruption at both loci. We hope the results using Mfn1 KO MEF and MFN1 KD HL-60 cells are enough to show a specific role of Mfn2 in cell migration.

We will have stained the mitochondria structure in the MEF1/2 MEF cells and the in Mfn1/2 KD dHL-60 cells. Opa1 KD HL-60 cells display extensive cell death and we are not confident interpreting any results from this line.

2.The authors should examine mitochondrial morphology in MFN2 shRNA treated cells (Fig. 2) and in mfn2-null MEFs (Fig. 3).

Mitochondrial morphology is examined in MFN2 shRNA treated cells (Fig 4c and 5a). The mitochondrial morphology in mfn2-null MEFs are published (Chen et al., 2003). We will further confirmed their results by staining mitochondrial structure in the KO MEFs.

3.The authors claim that chemotaxis defects of neutrophils are specific for MFN2 knock down, but not for MFN1. They show a Western blot of mfn1 knock down cells in Figure S3s. There is a band clearly visible, which appears to be much stronger than the MFN2 band in sh1 cells in Fig. 2a. Therefore, this conclusion is not valid.

The band intensities are dependent on the antibody quality and imaging acquisition and display. We don’t feel comfortable comparing the amount of two different proteins from two separate blots.

4.The colocalization of MFN2 with mitochondria and ER, shown in Fig. 4a, should be improved. Both mitochondria and ER appear abnormally clumped. The authors should stain mitochondria, ER and Mfn2 in the same cells. Images should be displayed much larger. The same is true for Fig. 5a. The authors claim that an artificial tether restored mitochondrial morphology in mfn2 knock down cells. They should state in the text which tether was used. Furthermore, they should explain their criteria for judgement of mitochondrial morphology. At least in my exes, mitochondria appear highly clumped in the image shown for sh1+T cells. In Fig. 5c it is not indicated how many cells were scored.

We will replace Fig 4a with a more representative image.

Neutrophils are blood cells and do not spread as well as adherent cells. We have also overexposed the images to show the smaller mitochondria, which cannot be visualized without saturating the signals. We tried to stain the cells with Mfn2 and Calnexin. However we cannot retain the mitotracker signal in fixed cells and could not do a triple label in dHL-60 cells. For this reason we have done double staining of mitochondria-ER, MFN2-mitochondria and MFN2-ER.

We have included the citation and the description of the tether. The tether is composed of a GFP protein carrying both ER and mitochondrial localization sequences at the ends, which functions independently of MFN2.

The criteria for the judgement of the mitochondrial morphology is now included in the methods, clustering analysis.

\*Minor points**

5.The Western blot shown in Fig. 5d suggests that expression of the tether construct reduced the amount of MFN2. How can this be explained?*

This Mfn2 amount is not significantly altered by the tether expression when quantified. We will add the quantification of all WB to the figures.

6.The paper is sometimes hard to digest for a reader who is not familiar with the authors' experimental systems. The description of the experiments in the main text is highly condensed.

We will elaborate on the experimental system in the results section.

7.Page 11: "5 m post stimulation" should read "5 min post stimulation".

Thank you. We have made this correction.

8.Some references are incomplete (page numbers are lacking).

We will reformat our references and checked for page numbers.

*Reviewer #3 (Significance (Required)):

Apparently, the manuscript is written for an audience with a special interest in chemotactic movements of neutrophils. I guess that the results reported in this manuscript will be of interest for this field. My background is mitochondrial biology and dynamics and I don't have the expertise to evaluate the aspects specific for neutrophils.*

It is well established that mfn2 mediates mitochondrial fusion and ER contact. Our novelty is the discovery that mfn2 suppresses Rac activation, which is essential for neutrophil adhesion and migration.

References:

Ablain, J., E.M. Durand, S. Yang, Y. Zhou, and L.I. Zon. 2015. A CRISPR/Cas9 vector system for tissue-specific gene disruption in zebrafish. Developmental cell. 32:756-764.

Amini, P., D. Stojkov, A. Felser, C.B. Jackson, C. Courage, A. Schaller, L. Gelman, M.E. Soriano, J.M. Nuoffer, L. Scorrano, C. Benarafa, S. Yousefi, and H.U. Simon. 2018. Neutrophil extracellular trap formation requires OPA1-dependent glycolytic ATP production. Nature communications. 9:2958.

Chen, H., S.A. Detmer, A.J. Ewald, E.E. Griffin, S.E. Fraser, and D.C. Chan. 2003. Mitofusins Mfn1 and Mfn2 coordinately regulate mitochondrial fusion and are essential for embryonic development. The Journal of cell biology. 160:189-200.

de Brito, O.M., and L. Scorrano. 2008. Mitofusin 2 tethers endoplasmic reticulum to mitochondria. Nature. 456:605-610.

Filadi, R., E. Greotti, G. Turacchio, A. Luini, T. Pozzan, and P. Pizzo. 2015. Mitofusin 2 ablation increases endoplasmic reticulum-mitochondria coupling. Proceedings of the National Academy of Sciences of the United States of America. 112:E2174-2181.

Filadi, R., E. Greotti, G. Turacchio, A. Luini, T. Pozzan, and P. Pizzo. 2017. On the role of Mitofusin 2 in endoplasmic reticulum-mitochondria tethering. Proceedings of the National Academy of Sciences of the United States of America. 114:E2266-E2267.

Kornmann, B., E. Currie, S.R. Collins, M. Schuldiner, J. Nunnari, J.S. Weissman, and P. Walter. 2009. An ER-mitochondria tethering complex revealed by a synthetic biology screen. Science. 325:477-481.

Maianski, N.A., F.P. Mul, J.D. van Buul, D. Roos, and T.W. Kuijpers. 2002. Granulocyte colony-stimulating factor inhibits the mitochondria-dependent activation of caspase-3 in neutrophils. Blood. 99:672-679.

Naon, D., M. Zaninello, M. Giacomello, T. Varanita, F. Grespi, S. Lakshminaranayan, A. Serafini, M. Semenzato, S. Herkenne, M.I. Hernandez-Alvarez, A. Zorzano, D. De Stefani, G.W. Dorn, 2nd, and L. Scorrano. 2016. Critical reappraisal confirms that Mitofusin 2 is an endoplasmic reticulum-mitochondria tether. Proceedings of the National Academy of Sciences of the United States of America. 113:11249-11254.

Naon, D., M. Zaninello, M. Giacomello, T. Varanita, F. Grespi, S. Lakshminaranayan, A. Serafini, M. Semenzato, S. Herkenne, M.I. Hernandez-Alvarez, A. Zorzano, D. De Stefani, G.W. Dorn, 2nd, and L. Scorrano. 2017. Reply to Filadi et al.: Does Mitofusin 2 tether or separate endoplasmic reticulum and mitochondria? Proceedings of the National Academy of Sciences of the United States of America. 114:E2268-E2269.

Rooney, C., G. White, A. Nazgiewicz, S.A. Woodcock, K.I. Anderson, C. Ballestrem, and A. Malliri. 2010. The Rac activator STEF (Tiam2) regulates cell migration by microtubule-mediated focal adhesion disassembly. EMBO reports. 11:292-298.

Saita, S., T. Ishihara, M. Maeda, S. Iemura, T. Natsume, K. Mihara, and N. Ishihara. 2016. Distinct types of protease systems are involved in homeostasis regulation of mitochondrial morphology via balanced fusion and fission. Genes to cells : devoted to molecular & cellular mechanisms. 21:408-424.

Zhou, W., L. Cao, J. Jeffries, X. Zhu, C.J. Staiger, and Q. Deng. 2018. Neutrophil-specific knockout demonstrates a role for mitochondria in regulating neutrophil motility in zebrafish. Disease models & mechanisms. 11.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Mitofusin 2 (Mfn2) is a mitochondrial outer membrane protein that is important for mitochondrial fusion and the establishment of mitochondrial ER contacts. It has been published before that these contact sites are important for calcium signaling. Zhou et al. examined the role of Mfn2 in neutrophils. They propose a model in which mitochondrial ER contacts established via Mfn2 are crucial for regulation of Rac, which is a small GTPase driving cell migration by promoting actin polymerization. Loss of Mfn2 results in elevated cytosolic calcium, over-activation of Rac, and defects of chemotactic movements. These defects can be partially rescued by restoration of mitochondrial ER contacts through expression of an artificial tether protein.

Major points

1.The authors claim on p. 6 that decreased neutrophil retention is not simply due to defects in mitochondrial fusion. However, the experimental setup they used for mfn2 (Fig. 1) is different from that for opa1 (Fig. S1), and therefore the results are not directly comparable. Unfortunately, the authors don't show fragmentation of mitochondria, neither in mfn2 nor in opa1 depleted cells. To support their statement they must show that lack of Mfn2 and Opa1 causes mitochondrial fragmentation to the same extent and then examine neutrophil retention in the same assay. Also, it would make sense to include Mfn1 in this analysis, as the authors later claim that the effects they observed are specific for Mfn2.

2.The authors should examine mitochondrial morphology in MFN2 shRNA treated cells (Fig. 2) and in mfn2-null MEFs (Fig. 3).

3.The authors claim that chemotaxis defects of neutrophils are specific for MFN2 knock down, but not for MFN1. They show a Western blot of mfn1 knock down cells in Figure S3s. There is a band clearly visible, which appears to be much stronger than the MFN2 band in sh1 cells in Fig. 2a. Therefore, this conclusion is not valid.

4.The colocalization of MFN2 with mitochondria and ER, shown in Fig. 4a, should be improved. Both mitochondria and ER appear abnormally clumped. The authors should stain mitochondria, ER and Mfn2 in the same cells. Images should be displayed much larger. The same is true for Fig. 5a. The authors claim that an artificial tether restored mitochondrial morphology in mfn2 knock down cells. They should state in the text which tether was used. Furthermore, they should explain their criteria for judgement of mitochondrial morphology. At least in my exes, mitochondria appear highly clumped in the image shown for sh1+T cells. In Fig. 5c it is not indicated how many cells were scored.

Minor points

5.The Western blot shown in Fig. 5d suggests that expression of the tether construct reduced the amount of MFN2. How can this be explained?

6.The paper is sometimes hard to digest for a reader who is not familiar with the authors' experimental systems. The description of the experiments in the main text is highly condensed.

7.Page 11: "5 m post stimulation" should read "5 min post stimulation".

8.Some references are incomplete (page numbers are lacking).

Significance

Apparently, the manuscript is written for an audience with a special interest in chemotactic movements of neutrophils. I guess that the results reported in this manuscript will be of interest for this field. My background is mitochondrial biology and dynamics and I don't have the expertise to evaluate the aspects specific for neutrophils.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary:

The manuscript by Zhou W and colleagues entitled "Mitofusin 2 regulated neurophil adhesive migration and actin cytoskeleton" proposed that mitochondrial outer membrane GTPase Mitofusin 2 controls cell migration via its capacity to regulate mitochondria-endoplasmic reticulum (ER) contacts, independently of its fusogenic activity. Using transgenic Mfn2 zebrafish, they first show that Mfn2 mutant embryos exhibit circulating neutrophils and defects in neutrophil recruitment to generated wound, compared to control. Then, using a combination of in cellulo and in vivo mouse models, they show that loss of Mfn2 decreases neutrophil migration, their adhesion under sheer stress and their infiltration to the peritoneal cavity in vivo. Third, they confirm these results using Mfn2 KO MEFs, where they show migration and actin skeleton defects, in contrast to Mfn1 KO MEFs. Mechanistically, they propose that migration defects induced by Mfn2 loss are associated to a decrease of membrane contact sites between the ER and the mitochondria. Using different in cellulo cell migration assays, they show that migration defects in Mfn2-null was rescued upon an artificial mito-ER tether. Finally, they propose that the loss of Mfn2 leads to cytosolic calcium accumulation, inducing hyper-activation of the RhoGTPase, Rac1, a key regulator of actin dynamics and cell migration. Together, the authors proposed a new function of Mfn2 in regulating cell migration via mito-ER contacts tethering.

Major comments:

Although the results could be very interesting, and could be significantly relevant to the mitochondrial field and the cell biology one in general, major points need to be addressed to fully support conclusions of the authors. Different controls and quantification are missing, Actin dynamics analysis should be improved, effects of the artificial tether is weakly characterized and the demonstration of the specific role of mito-ER contacts via mfn2 in migration should be reinforced.

-In figure 1, quantification of circulating neutrophils is required in Mfn2 KO embryos. The authors should also show these quantified results for OPA1KO, which are just mentioned in the text. In addition, in figure 1b and d, the neutrophils from the Mfn2KO embryos seem bigger compared to control. Can the authors comment on neutrophils size and potential contribution to the phenotype? Finally, the authors propose a defect in neutrophil migration in Mfn2-KO, however neutrophils are found in the circulation. The authors should explain these results.

-The authors need to reinforce the Mfn2 specificity for their phenotype. In particular in Fig S1, they show that loss of OPA1 significantly decreases neutrophil migration in vivo. However, they then only study the effect of Mfn1 silencing in neutrophil and MFN1 KO MEFs (Sup Fig s3). The authors should perform the same experiments in neutrophil and MEF upon loss of OPA1 (similar to Fig S3). Does loss of OPA1 and Mfn1 decrease neutrophil arrest to activated endothelial cells?

-Using their images, the authors should also document on the directionality of the cell during cell migration. Do Mfn2 depleted cells do not migrate because they are arrested or because they are lacking directionality? Environment/chemokine sensing defects?

-Actin dynamics analysis should be improved. Loss of Mfn1 and Mfn2 lead to cell shape changes. The authors should quantify this phenotype by analysing cell circularity (as well as for Opa1 loss). Stress fibres number or Phalloidin intensity quantification in cell body should also be performed.

-Can the migration defects could be attributed to Focal adhesion protein dynamics defects? The authors shown an hyperactivation of Rac1 and an hyperphosphorylation of PAK, which can control FAP (focal adhesion proteins) dynamics. In addition, immunofluorescence analysis shows a decreased signal and cellular misdistribution of paxillin. The authors should characterize these phenotypes. FAP levels (Paxillin/Phospho-Paxillin and Vinculin) should be analysed by immunoblot, the number of FAP/cell, distribution and size should also be quantified. Their dynamics should also be analysed by live cell imaging. Finally, Paxillin level and distribution seems to be also impacted in Mfn1KO cells. Can the authors comment on that? The different quantifications would help to better understand the effect of different mitofusins in cytoskeleton dynamic.

-Please perform rescue experiments for cell migration in MFN2KO and MFN1KO MEFs. Immunoblots showing protein levels of these proteins would be appreciated. To really discriminate how Mfn2 regulates cell migration, the authors should also perform rescue experiments using a fusogenic mutant Mfn2 ((K109A). It will help to demonstrate the relevance of mito-ER contacts and not mitochondrial fusion in the phenotype.

-Figure 4, the authors stipulate that Mfn2 regulates ER-mitochondria tethering. However, the authors present no evidence for this conclusion. The authors should perform manders coefficient in MFN2 KO cells and compared it to control. Also, loss of Mfn2 induces mitochondrial fragmentation, which can lead to problem for mito-er contacts quantification by light microscopy. The authors should use their TEM pictures to quantify mito-ER contacts (Number, length and % of mito perimeter), not only mitochondrial morphology. Mfn1 should be used as negative control. it would be interesting also to determine the status of the mito-ER contact in the different conditions used in the manuscript to stimulate cell migration like fMLP treatment.

-The authors use an artificial tether to manipulate mito-ER contacts in cellulo. However, no information from its origin, or its design are documented in the manuscript. In addition, the authors should show that this tether efficiently works by analysing mito-ER contacts upon expression by EM and mitochondrial calcium uptake. Does this tether rescue mito-ER contacts defects induced by loss of Mfn2? How the authors explain that the tether rescues mitochondrial morphology defects in MFN2KO? In these conditions, mitochondria should not be able to fuse anymore as Mfn2 is lost? This is really intriguing results. Does the tether rescue the other parameters? Mitochondrial distribution (with quantification)? Cell shape? Paxillin defects? ROS and membrane potential? These rescue experiments analyses are important to determine which parameters are really involved in cell migration defects due to the decreased tethering. Finally, it would be of great interest to analyse the effect of the tether alone on cell migration, Rac1 activity, cell shape? Gain of function? These results may reinforce the idea that contact sites regulate cell migration.

-It is well established that a decrease of membrane potential leads to a decrease of mitochondrial calcium uptake. Calcium results obtained by the authors without any information on the roles of the tether could not lead to any conclusion. Does the tether rescues membrane potential and calcium uptake by the mitochondria? So far, the decrease of mitochondrial calcium upon stimulation in Mfn2KO cab be attributed to both mito-ER contact or membrane potential defects. It has been shown that MEFs MFN2 KO can lead to a decrease of MCU provel level leading to a decrease of mitochondrial calcium uptake (PMID: 25870285). The authors should also check MCU protein level.

-Hyperactivation of Rac1 is only based on phosphorylation of PAK, which is quite weak. The authors should better describe the hyperactivation of RAC1 or other RhoGTPases in their Mfn2 KO MEFs. What are the levels of RAC1 and other RhoGTPases? Subcellular distribution in the cell? Kits are also available to determine RhoGTPase activity by pull down assay (Cell biolabs).

-The references are up-to-date. The text and the figures are clear and accurate.

Minor comments:

-The authors should show the efficiency of the KO generated for Mfn2 and Opa1 in zebrafish embryos. Sequencing results to highlight the position of the mutations and their consequences on the coding protein should be shown, as well as immunoblot analysis should be performed to analyse Mfn1, Mfn2 and OPA1 protein levels. The generation of a MFN1-KO transgenic line would have been appreciated to finely compare the roles of the 3 GTPases involved in mitochondrial fusion during neutrophil infiltration and migration in vivo.

-MFN1, MFN2, AND OPA1 protein levels should be analysed by immunoblot in the Mfn1 and Mfn2 KO MEFs.

-In cell spreading assay, it would be great to identify cells during the process, by an asterix for example. "wt MEFs extended transient filopodia and lamellipodia and eventually elongated, whereas Mfn2-null MEFs generated extensive membrane ruffles and retained the circular shape". It would be interesting to quantify these different parameters.

-For all their immunoblot analysis, the authors should use a mitochondrial marker as loading control (VDAC1, TOM20, HSP60...). In figure 5, Vinculin should not be used a loading control, with its role in focal adhesion dynamics.

-Legends for figures 5 and 6 are inverted.

-Please document in the material and methods section, how confocal images have been acquired: number of z-stacks, reconstitution, 3D analysis...

-The authors should show their results of blood cell composition quantification in ctrl vs MFn2 depletion.

-The authors should describe all the acronyms used throughout all the manuscript. For example, LTB4, fMLP...

Significance

Beyond their role in energy production, mitochondria are involved in numerous cellular functions including cell migration. Mitochondria form a network balanced by fission and fusion events, where membrane contact sites with the endoplasmic reticulum are crucial. These contact sites are also involved in mitochondrial and cellular functions via their capacity to exchange lipids, metabolites and calcium. The role of mitochondria in cell migration has started to emerge where mitochondrial fragmentation and/or mitochondrial calcium homeostasis are acknowledged to drive cancer cell migration and to regulate actin dynamics. In this manuscript, Zhou W and colleagues proposed for the first time the role of mitochondria-ER contacts in cell migration. Mechanistically, this can be associated to the capacity of these contacts to control mitochondrial functions or mitochondrial calcium homeostasis. These findings are physiologically relevant and of particular interest to the mitochondrial and cell migration field but also to general cell biology. It represents a novel function associated to these membrane contact sites and point-out these contacts as signalling platform creating microdomains of metabolites exchanges involved in cell migration.

Keywords: Mitochondria - Membraned dynamics - calcium homeostasis - Membrane contact sites

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

The manuscript by Zhou et al. describes the role of mitofusin-2 in neutrophil adhesive migration. The authors suggest that MFN-2 is required to sustain neutrophil migration and link this observation to the role of MFN-2 in maintaining mitochondria-ER contacts and suppressing Rac activation. Although some of the experiments are convincing, the authors come to conclusions that are not entirely supported by their data and a few statements appear the result of inductive reasoning. A major problem is the distinction between adhesion and migration: in several parts of the manuscript, there is confusion between these two events and the experiments are not designed (and not discussed) in order to clarify this point. For example, the fact that in zebrafish embryos lacking Opa1 there is no defect in neutrophil retention but reduced neutrophil migration should suggest that MFN-2 controls adhesion rather than migration. But this is not properly elaborated. The same problem comes with the role of Rac, which has been elegantly shown to be required for cell migration but not for cell spreading or focal adhesion formation (Steffen et al, JCS 2013). Again, it is necessary to distinguish between migration and other functions requiring the actin cytoskeleton.

Specific comments:

Introduction:

"Although mitochondria-derived ATP possibly regulates neutrophil chemotaxis in vitro (Bao et al., 2015), removal of extracellular ATP improves neutrophil chemotaxis in vivo (Li et al., 2016). These conflicting reports prompted us to search for mechanisms delineating the role of mitochondria in neutrophil migration outside the realm of ATP or cellular energy (Bi et al., 2014; Schuler et al., 2017; Zanotelli et al., 2018)." This sentence is superficial and misleading: extracellular ATP may interfere with chemotaxis through various energy-independent mechanisms (see for example Zumerle et al. Cell Reports 2019) and this is not conflicting with the role of intracellular ATP in migration.

Figure 1: The authors didn't show evidence of the genome edition (PCR, RFLP or Sequencing over the sgRNA target) or at least RT-PCR or WB for MFN2. In Fig 1b, 1c the scale bar is missing. "Neutrophils were sorted from both lines and their respective loci targeted by the 4 sgRNAs were deep sequenced." There are no data about sorting strategies for zebrafish neutrophils in the figure. Moreover, only 2 sgRNAs are shown and there are no sequencing data.

Figure 2: In the WB, reconstitution is not obvious. In general, all WBs are not quantified (and they should be quantified). The in vivo experiment does not have proper controls. For example, can the authors exclude that in these mice there is reduced inflammation because neutrophils have defective activation? What about NETs? And cytokines/chemokines? And exocytosis? In the absence of these controls, the experiment cannot be properly interpreted.

Figure 3: The conclusion of the authors "In summary, Mfn2 modulates the actin cytoskeleton and cell migration in MEFs" should be supported by experiments to distinguish between the specific role of Mfn2 and the role of mitochondrial dynamics (Opa1, Drp1, Mfn1). It is also not clear why the authors decided to use MEFs instead of other cells (more similar to neutrophils which are not adherent cells). The results obtained in MEFs may be irrelevant for neutrophils.

Figure 4-5: Fig 5a: in ctrl and sh1 the ER seems to be larger than the phalloidin (=cytoskeleton=cell border approximately) in a few regions. Only the sh1+T seems to fit correctly. The TEM image (only 1 in supplementary) is not sufficient to convince that the tethering is lost. Quantification of number of contacts and distance between ER and mitochondria should be included. The title of figure 5 is wrong. However, in these figures, it is clear that cells are beautifully polarized, with mitochondria accumulating at the uropod (and even more in the absence of Mfn2). When comparing these images with those published by Campello et al (JEM 2006), there are 2 observations that can be made: first of all, these data confirm that mitochondrial fission promotes cell polarity; second, they suggest that the defect is not at the level of cell polarity/chemotaxis.

Figure 6: Calcium data are, in general, very weak. First of all, controls with ionomycin are missing. Statistical analyses of the curves should be included. As for the use of the MCU inhibitor Ru360, is there any evidence that it is cell-permeant in this context? Is it blocking MCU? Since the authors can show mitochondrial calcium upon FMLP, they should also demonstrate that Ru360 is indeed working and inhibiting mitochondrial calcium uptake. The sentence "The MCU inhibitor Ru360did not cause further reduction of chemotaxis in MFN2 knockdown dHL-60 cells (Supplementary Fig. 6c, d and Supplementary Movie. 12), indicating that MCU and MFN2 lies in the same pathway in terms of regulating chemotaxis in dHL-60 cells" is speculative. In general, there is no solid demonstration that the effect is calcium-mediated. As for Rac, it is surprising to see that Rac inhibition has no effect on cell migration. Rac is known to promotes migration in fibroblasts and other cell types and Rac deficiency inhibits migration (see for example Steffen et al, JCS 2013). Two sets of experiments are absolutely required: 1) verify this in fibroblasts since it has been elegantly shown that Rac is essential in these cells for migration; 2) analyse the effect of Rac inhibitors in pPak kinetics.

Significance

As presented here, the manuscript has a modest significance. The audience would be specialised: cell migration, cell signalling. My expertise is immunology, cell activation, cell migration, cell signalling.

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

Learn more at Review Commons

Reply to the reviewers

We thank the reviewers and the editor for the insightful and thorough assessment of our manuscript. In this response to review letter, we have listed the original review (black text) and responded to each critique after it.

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

Yang et al. submitted a manuscript describing the detection of pseudogenes ("retrocopies") of APOBEC3 (A3) genes in primates. The evolutionary history and relationship to specific A3s was analyzed and speculated that the maintained A3 retrocopies had a functional role at least early in the evolution and some may have still now. Functional data on some of the expressed retrocopies are presented on L1 and HIV.

The authors claim that "retrocopying expands the functional repertoire of A3 antiviral proteins in primates". While almost of the genetic findings were published recently (Ito et al. 2020), the authors should more clearly describe how their data differ or confirm the data of Ito et. al.

We thank the reviewer for their helpful comments which have guided revisions to our manuscript. We have taken steps to clarify the dramatic differences between our work and the recent publication from Ito, Gifford, and Sato.

Foremost, we respectfully disagree with the reviewer that the genetic findings in our work were contained within the Ito, et al manuscript. Using a computational screen of assembled mammalian genome, the Sato group cataloged the gain and loss of APOBEC3 genes during the evolution of mammals. They found a fascinating correlation between the dynamics of A3s and ERVs that formed the precis of the paper. From their genome-wide search for A3s, Ito et al describe several retrocopies of A3s in two New World Monkey species, one of which retains a full-length open reading frame, leading to the statement that this gene may be functional.

We note that the retrocopies found in the Ito et al paper span only two of the more than 20 species in which we identify A3 retrocopies. Further, as a result of the breadth of our search for A3s, we find additional retrocopies in the same two New World Monkey species that were examined in the Ito et al paper. Finally, our study also examined functional capabilities of these additional A3s. These differences are highlighted by reviewer 3 who writes that relative to Ito et al, our manuscript studies the phenomenon of A3 retrocopies “more deeply both by in silico analyses and cell culture experiments.” Reviewer 3 also summarizes the most important difference in our studies – our work presents a “conceptual advance that the antiviral gene expansion has achieved not only via tandem gene duplication but also via gene retrocopying”.

Lastly, we want to point out that the findings of our manuscript and Ito et al. 2020 were made concurrently. Indeed, throughout the preparation process of this manuscript, we were both aware of each other’s findings and shared preprints with each other. Most of the participating journals in Review Commons have “scoop protection” mechanisms that typically extend 6 months after the publication of the first article (Ito et al was published Jan 2020), and our article was first submitted to Reviewer Commons on February 14, 2020. Therefore, we feel confident that the ‘no scoop’ policy applies to the minimal overlap between our paper and that of Ito et al.

Nevertheless, we have modified the text to more clearly acknowledge the parallel finding of some New World monkey retrogenes in the Ito, et al. paper.

The functional data (Fig. 6) are interesting, but in the current form not complete. The authors have to show protein expression in the transfected cells (A3, L1, HIV) and level of encapsidation into viral particles. In addition, please analyze if the retrocopies express cytidine deaminase active enzymes.

We thank the reviewer for this comment, and we have added a Western blot of the six long-ORF-containing retrocopies as Figure S5. In this blot (from early in the project), we detected protein production in 293T cells for 3/6 retrocopies. In later optimizations of subsets of this blot, we were able to detect expression of the marmoset A3G and the other two marmoset retrocopies (marmoset-2 and marmoset-4). Despite optimization attempts, we were unable to detect protein for one of the retrocopies that restricts HIV-1ΔVif (capuchin-C1). Unfortunately, at this time the included blot is the only one we have in which all 6 constructs are included on a single blot. Optimally, all 6 constructs would be side-by-side in a single blot with optimized conditions, and we are happy to complete this experiment as soon as we are able to return to our lab after the SARS-2 quarantine is lifted. However, we think the added blot shows that some of the retrocopies produce protein and the absence of detectable protein from capuchin-C1 could suggest that this retrocopy is especially potent in its restriction function or an idiosyncratic problem with detecting this protein using Western blot analyses.

We have not previously tested our lentiviral particles for levels of encapsidation of protein from each retrocopy. The value we see in this experiment is in explaining why some of the retrocopies that are expressed in producer cells may not restrict in target cells. While we note that precedent in the literature suggests that A3 proteins which restrict HIV-1ΔVif are invariably encapsidated, we would be happy to carry out this experiment when our lab reopens.

In response to the reviewer’s request to test deaminase activity for each retrocopy, we note that Figure 4 shows the intactness of the deaminase motif in each retrocopy. However, we feel that a description of the mechanisms of restriction of these retrocopies is not a major point of this paper and is beyond the scope of the current investigation.

Reviewer #1 (Significance (Required)):

Minor advance compared to Ito et al. 2020.

We respectfully and rigorously disagree with this assessment. Please refer back to the reviewer’s first comment. We defer, again, to Reviewer 3’s assessment that our work presents a “conceptual advance that the antiviral gene expansion has achieved not only via tandem gene duplication but also via gene retrocopying”. Moreover, we must point out that the Ito et al 2020 paper was entirely computational; indeed, several retrogenes that could computationally be predicted to be ‘dead’ were confirmed by us as having antiviral activity.

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

**Summary:**

Yang et al. study the expansion of APOBEC3 (A3) cytidine deaminases genes in primates. Authors find A3 retrocopies in several lineages in primates using Blast searches. Some are old and some are species specific. Some have disablements and some have intact ORFs. Authors study their mode of evolution, expression and functionality. Authors have performed detailed analyses including functional analyses. Some A3 retrocopies are broadly expressed and some have retained ability to restrain retroelements. I agree with the authors that their data supports that retrocopying has contributed the turnover in the repertoire of host retroelement restriction factors. Authors show that some retrocopies have remained active for long periods of time and they still show that they can restrict retroelements/retrovirus. This work provides an interesting example of immune system diversification. This study of the A3 family of proteins that are part of the vertebrate innate immune system and the data supporting turnover of these kind of immune system genes is strong. The work underscores that this is a way immunity genes evolve and it has parallels in the evolution of the TRIM gene family of immune genes. I just have a few comments. I think the work can gain from analyzing some aspects of the data in more detail and presenting the big picture in a summary table, even if it is just supplementary.

**Major comments:**

A3I is in many species. Does this mean it was preserved (i.e., functional for a while)? For how long have disabling mutations been accumulating? Can we get a sense of that? Even for other retrocopies, do we have a sense of how recent has the pseudogenization been? If it is very recent that means that the gene was active until not long ago.

Our analyses suggest that A3I was born in the common ancestor of simian primates and pseudogenized before the Catarrhini/Platyrrhini split. It is possible that A3I was functional within this extended period (~12-15 million years), but the presence of a shared truncating stop codon amongst all simian A3Is suggests the gene was no longer full-length at the time of diversification of the simians. Instead, the simian LCA likely encoded an A3I with a predicted ORF of 261 codons; if this truncated ORF were functional, it was then further truncated/pseudogenized with additional frame-breaking mutations which follow the phylogeny of primates.

We estimated the timeline of pseudogenization of each retrocopy using the species distribution of each syntenic retrocopy. We also note that we find full-length ORFs in three retrocopies which have been retained for a period of time at least as long as the age of the last common ancestor of the four New World monkeys. These old but intact retrocopies motivated our simulations of ORF retention rates (Figure 5).

In the PAML analyses test could be performed to test if the rate of evolution that are higher or lower than 1 for particular genes are actually significantly higher or lower than 1 for the particular gene comparing the likelihoods of the modes with the given rate with the one with the rate fixed to 1. Is there enough power to do this?

We thank the reviewer for pointing out this omission in our analysis. We did perform these tests and find a significant p-value for two of the nodes p=0.058 and p=0.025 respectively). We have updated the legend for figure S4 to incorporate these p-values

Page 9. It seems to me that the synteny data Figure S2 reveals they are derived from independent retroposition events and not duplications of segments because those would include flanking genes. Is this correct? Authors could comment on that.

Yes, we think that each retrocopy we show in Figure S2 is likely created via an independent retrotransposition event. We have clarified in the text that Figure S2 shows the genes used to establish synteny to support orthology of the retrocopies shared amongst multiple species and that each of these ortholog groups presumably originated via distinct retrotransposition events.

In figure S4, I am not sure why orthologous genes are not grouped together in the phylogeny and why p is smaller than 0.05. How should that figure and the probability be interpreted?

We thank the reviewer for their comments on this figure. First, the reviewer identified an error in the tree in which the branch labels for ‘night monkey-C2’ and ‘night monkey-SS1’ were inadvertently switched. The corrected tree now follows the pattern expected by the reviewer. Second, we employed RELAX to “determine whether selective strength was relaxed or intensified in one of these subsets relative to the other” (Wertheim, et al. MBE 2014). In this case, the p-value corresponds to the finding that the retrocopies (test branches) show intensification of selection relative to the intron-containing A3Gs (reference branches).

We have modified Figure S4 and the associated text to more clearly explain the specific hypothesis test we report.

It would be good to have a summary table that summarizes what genes have support for past or current functionality (preservation for long time or recent pseudogenization, expression, purifying or positive selection, ability to restrict retroelements) and in what lineages.

We agree with this reviewer suggestion. We have added the additional information including the number of frame disrupting mutations as a measure of age, intactness, and ability to restrict retroelements to Table S1. Thanks to this suggestion, Table S1 now serves as the master table to summarize the analyses of each retrocopy.

**Minor comments:**

1. Page 3. Authors say "...the exons and UTRs..." but UTRs are part of exons. Authors could talk about exons only that include protein-coding regions and UTRs.

Changed the text to "exons".

Page 7. I would say disabled instead of "... becoming degraded by mutation."

Fixed according to the reviewer's suggestion.

I would say neutral evolution not neutral selection.

Fixed according to the reviewer's suggestion.

Reviewer #2 (Significance (Required)):

This work provides an interesting example of immune system diversification. Authors study the APOBEC3 family of proteins that is part of the vertebrate innate immune system and the data supporting turnover of these kind of immune system genes. The work underscores that this is a way immunity genes evolve and it has parallels in the evolution of the TRIM gene family of immune genes revealing patterns in the mode of evolution of immunity genes. The audience of this work will be people interested in evolution of immunity, arms races and gene diversification and all evolutionary biologists interested in adaptation. I work in the field of comparative genomics and molecular evolution.

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

**Summary:**

This manuscript by Yang et al. is an well-written, intriguing paper highlighting the evolutionary significance of the gene creation via "retrocopying". The authors investigated the expansion of antiviral A3 genes via retrocopy in Primates, and found that A3G-like retrocopies have been generated repeatedly during primate evolution. A part of A3 retrocopies found in New World monkeys retained full length open reading flames and anti-lentiviral capacities. Interestingly, the spectrum of anti-retroelement activity of A3 retrocopies was different from the original (i.e., intron-containing) A3G gene in these species, suggesting the occurrence of the functional differentiation followed by gene amplification. However, one of the main findings that many A3 retrocopies are present in New World monkey is in-line to a previous report (i.e., Ito et al., 2020, PNAS), and the experimental validations were based on the human (not New World monkey's) retroelements. Nevertheless, this study deeply investigated the possible importance of A3 retrocopies for the host defense system evolution both by in silico analyses and cell culture experiments. This study provides the findings that can potentially expand our knowledge on the evolutionary arms races between retroelements and the hosts.

**Major:**

To strengthen the impact of this work, it would be better to increase the numbers of retroviruses in which the anti-retroviral capacities are investigated. I understand that it is difficult to examine retroviruses or L1s that are colonized naturally with New World monkeys, but I suppose it is not so difficult to investigate a variety of representative retroviruses such as murine leukemia virus (MLV) or the reconstructed human endogenous retrovirus K (HERV-Kcon). This additional experiment would be helpful to highlight that the spectrum of anti-retroviral activity of A3 retrocopies is divergent from the original A3G gene in these species and strengthen the concept to be proposed by this study.

The reviewer raises a fascinating question about whether retrocopies might have different restriction abilities relative to the other A3s in a given species. First, we feel that showing activity against one pathogen is sufficient for our claim that some of the A3 retrocopies have antiviral potential. Second, we discuss in the paper the idea that HIV-1 is not the actual target of these (or any) innate immune genes in New world primates. We argue that any other targets we might test would also be surrogates for the ‘true’ target of these genes.

**Specific:**

1, Since the authors found the expansion of "functional" repertoire of A3 retrocopies specifically in New World monkey, it would be better to rephrase the title as "Retrocopying expands the functional repertoire of APOBEC3 antiviral proteins in New World monkeys".

We thank the reviewer for this comment but point out that a large portion of our manuscript presents our work on primates outside the New World monkeys. The reviewer is correct to note that our finding of restriction activity is limited to New World monkey retrocopies, but we feel that the current title will attract a broader audience and reflects the broader relevance of this work.

2, It might be better to add a figure summarizing which A3 retrocopies in which species retain nearly full length ORFs. For example, how about making a figure like Fig. 2A for all the four representative New World monkey species?

We agree. We have added the length of the longest ORF for each retrocopy to Table S1.

3, Fig. 3

It would be helpful to clarify that which cell of the heatmap corresponds to the intact A3 retrocopies.

We have added labels to indicate the intact A3 retrocopies and adjusted the legend accordingly.

4, Page 4, line 5

It would be better to replace the word "protected" with "escaped" because this retrocopy subset should include the ones that are intact but not functional.

Changed as suggested.

5, Page 4, line 25

It would be better to rephrase "the common ancestor of mammals" as "the common ancestor of placental mammals" because A3 gene is absent in Marsupial.

Changed as suggested.

6, Page 5, line 5

Please rephrase "ongoing" as "recently-occurred".

Changed as suggested.

7, Page 6, line 19

I checked the multiple sequence alignment in File S1 and suspect that the codon (alignment) position of the shared premature stop codon is 261 (not 264).

We thank the reviewer for pointing out this discrepancy. We have revised the text to reflect the correct position of the shared stop.

8, Page 6, line 23

I could not understand the meaning of the sentence "Intriguingly, one lineage-specific mutation...".

Please specify the position of mutation which the authors mentioned (in File S1 or Fig. 1B).

This portion of the text refers to a reversion of a stop codon in the orangutan A3I; specifically, the stop codon shared in all simians acquired a second mutation that created a longer ORF in only this species. We have removed this sentence from the text for the sake of clarity.

9, Page 12, line 8

Please refer Fig. S4 here.

Changed as suggested.

10, Page 12, line 8

Please say "Significant relaxed selection was not detected" rather than "Our analysis detected no relaxation...".

Changed as suggested.

11, Page 12, line 8

Fig. S4 indicates "p=0.015", but the authors regard it as "not significant"?

We thank the reviewer for pointing out this confusing wording. We employ RELAX to “determine whether selective strength was relaxed or intensified in one of these subsets relative to the other” (Wertheim, et al. MBE 2014). In this case, the p-value corresponds to the finding that the retrocopies show intensification of selection.

We have modified Figure S4 to more clearly explain the specific hypothesis test for this p-value. We have also modified the text to clarify this point.

12, Page 12, line 9

Please here refer the data showing the claim "Instead, these A3G retrocopies have evolved more rapidly than...".

Changed as suggested; see previous point.

13, Page 12, line 11

Did the authors perform the statistical test on the dN/dS ratio analysis? If so, please mention the result of the test.

Yes we did. Please refer to Reviewer 2’s ‘Major Point 3’.

14, Page 12, line 15

It would be better to modify the phrase "show evidence of recurrent selection for functional innovation"

Changed as suggested.

Reviewer #3 (Significance (Required)):

This study provides a conceptual advance that the antiviral gene expansion has achieved not only via tandem gene duplication but also via gene retrocopying.

Compare to existing published knowledge.

Although one of the main findings that many A3 retrocopies are present in New World monkey is in-line to a previous report (i.e., Ito et al., 2020, PNAS), this study investigated the above finding more deeply both by in silico analyses and cell culture experiments.

Audience.

Evolutionary biologists and researchers in the field of viruses (particularly retroviruses including HIV-1) and transposable elements would be interested in this work.

Your expertise.

Bioinformatics, genome biology, viruses, and transposable elements

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

This manuscript by Yang et al. is an well-written, intriguing paper highlighting the evolutionary significance of the gene creation via "retrocopying". The authors investigated the expansion of antiviral A3 genes via retrocopy in Primates, and found that A3G-like retrocopies have been generated repeatedly during primate evolution. A part of A3 retrocopies found in New World monkeys retained full length open reading flames and anti-lentiviral capacities. Interestingly, the spectrum of anti-retroelement activity of A3 retrocopies was different from the original (i.e., intron-containing) A3G gene in these species, suggesting the occurrence of the functional differentiation followed by gene amplification. However, one of the main findings that many A3 retrocopies are present in New World monkey is in-line to a previous report (i.e., Ito et al., 2020, PNAS), and the experimental validations were based on the human (not New World monkey's) retroelements. Nevertheless, this study deeply investigated the possible importance of A3 retrocopies for the host defense system evolution both by in silico analyses and cell culture experiments. This study provides the findings that can potentially expand our knowledge on the evolutionary arms races between retroelements and the hosts.

Major:

To strengthen the impact of this work, it would be better to increase the numbers of retroviruses in which the anti-retroviral capacities are investigated. I understand that it is difficult to examine retroviruses or L1s that are colonized naturally with New World monkeys, but I suppose it is not so difficult to investigate a variety of representative retroviruses such as murine leukemia virus (MLV) or the reconstructed human endogenous retrovirus K (HERV-Kcon). This additional experiment would be helpful to highlight that the spectrum of anti-retroviral activity of A3 retrocopies is divergent from the original A3G gene in these species and strengthen the concept to be proposed by this study.

Specific:

1, Since the authors found the expansion of "functional" repertoire of A3 retrocopies specifically in New World monkey, it would be better to rephrase the title as "Retrocopying expands the functional repertoire of APOBEC3 antiviral proteins in New World monkeys".

2, It might be better to add a figure summarizing which A3 retrocopies in which species retain nearly full length ORFs. For example, how about making a figure like Fig. 2A for all the four representative New World monkey species?

3, Fig. 3 It would be helpful to clarify that which cell of the heatmap corresponds to the intact A3 retrocopies.

4, Page 4, line 5 It would be better to replace the word "protected" with "escaped" because this retrocopy subset should include the ones that are intact but not functional.

5, Page 4, line 25 It would be better to rephrase "the common ancestor of mammals" as "the common ancestor of placental mammals" because A3 gene is absent in Marsupial.

6, Page 5, line 5 Please rephrase "ongoing" as "recently-occurred".

7, Page 6, line 19 I checked the multiple sequence alignment in File S1 and suspect that the codon (alignment) position of the shared premature stop codon is 261 (not 264).

8, Page 6, line 23 I could not understand the meaning of the sentence "Intriguingly, one lineage-specific mutation...". Please specify the position of mutation which the authors mentioned (in File S1 or Fig. 1B).

9, Page 12, line 8 Please refer Fig. S4 here.

10, Page 12, line 8 Please say "Significant relaxed selection was not detected" rather than "Our analysis detected no relaxation...".

11, Page 12, line 8 Fig. S4 indicates "p=0.015", but the authors regard it as "not significant"?

12, Page 12, line 9 Please here refer the data showing the claim "Instead, these A3G retrocopies have evolved more rapidly than...".

13, Page 12, line 11 Did the authors perform the statistical test on the dN/dS ratio analysis? If so, please mention the result of the test.

14, Page 12, line 15 It would be better to modify the phrase "show evidence of recurrent selection for functional innovation"

Significance

This study provides a conceptual advance that the antiviral gene expansion has achieved not only via tandem gene duplication but also via gene retrocopying.

Compare to existing published knowledge.

Although one of the main findings that many A3 retrocopies are present in New World monkey is in-line to a previous report (i.e., Ito et al., 2020, PNAS), this study investigated the above finding more deeply both by in silico analyses and cell culture experiments.

Audience.

Evolutionary biologists and researchers in the field of viruses (particularly retroviruses including HIV-1) and transposable elements would be interested in this work.

Your expertise.

Bioinformatics, genome biology, viruses, and transposable elements

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary: Yang et al. study the expansion of APOBEC3 (A3) cytidine deaminases genes in primates. Authors find A3 retrocopies in several lineages in primates using Blast searches. Some are old and some are species specific. Some have disablements and some have intact ORFs. Authors study their mode of evolution, expression and functionality. Authors have performed detailed analyses including functional analyses. Some A3 retrocopies are broadly expressed and some have retained ability to restrain retroelements. I agree with the authors that their data supports that retrocopying has contributed the turnover in the repertoire of host retroelement restriction factors. Authors show that some retrocopies have remained active for long periods of time and they still show that they can restrict retroelements/retrovirus. This work provides an interesting example of immune system diversification. This study of the A3 family of proteins that are part of the vertebrate innate immune system and the data supporting turnover of these kind of immune system genes is strong. The work underscores that this is a way immunity genes evolve and it has parallels in the evolution of the TRIM gene family of immune genes. I just have a few comments. I think the work can gain from analyzing some aspects of the data in more detail and presenting the big picture in a summary table, even if it is just supplementary.

Major comments:

1. A3I is in many species. Does this mean it was preserved (i.e., functional for a while)? For how long have disabling mutations been accumulating? Can we get a sense of that? Even for other retrocopies, do we have a sense of how recent has the pseudogenization been? If it is very recent that means that the gene was active until not long ago.
2. In the PAML analyses test could be performed to test if the rate of evolution that are higher or lower than 1 for particular genes are actually significantly higher or lower than 1 for the particular gene comparing the likelihoods of the modes with the given rate with the one with the rate fixed to 1. Is there enough power to do this?
3. Page 9. It seems to me that the synteny data Figure S2 reveals they are derived from independent retroposition events and not duplications of segments because those would include flanking genes. Is this correct? Authors could comment on that.
4. In figure S4, I am not sure why orthologous genes are not grouped together in the phylogeny and why p is smaller than 0.05. How should that figure and the probability be interpreted?
5. It would be good to have a summary table that summarizes what genes have support for past or current functionality (preservation for long time or recent pseudogenization, expression, purifying or positive selection, ability to restrict retroelements) and in what lineages.

Minor comments:

1. Page 3. Authors say "...the exons and UTRs..." but UTRs are part of exons. Authors could talk about exons only that include protein-coding regions and UTRs.
2. Page 7. I would say disabled instead of "... becoming degraded by mutation."
3. I would say neutral evolution not neutral selection.

Significance

This work provides an interesting example of immune system diversification. Authors study the APOBEC3 family of proteins that is part of the vertebrate innate immune system and the data supporting turnover of these kind of immune system genes. The work underscores that this is a way immunity genes evolve and it has parallels in the evolution of the TRIM gene family of immune genes revealing patterns in the mode of evolution of immunity genes. The audience of this work will be people interested in evolution of immunity, arms races and gene diversification and all evolutionary biologists interested in adaptation. I work in the field of comparative genomics and molecular evolution.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Yang et al. submitted a manuscript describing the detection of pseudogenes ("retrocopies") of APOBEC3 (A3) genes in primates. The evolutionary history and relationship to specific A3s was analyzed and speculated that the maintained A3 retrocopies had a functional role at least early in the evolution and some may have still now. Functional data on some of the expressed retrocopies are presented on L1 and HIV.

The authors claim that "retrocopying expands the functional repertoire of A3 antiviral proteins in primates". While almost of the genetic findings were published recently (Ito et al. 2020), the authors should more clearly describe how their data differ or confirm the data of Ito et. al.

The functional data (Fig. 6) are interesting, but in the current form not complete. The authors have to show protein expression in the transfected cells (A3, L1, HIV) and level of encapsidation into viral particles. In addition, please analyze if the retrocopies express cytidine deaminase active enzymes.

Significance

Minor advance compared to Ito et al. 2020.

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

Learn more at Review Commons

Reply to the reviewers

We thank the reviewers for their close reading and constructive comments on our manuscript. We believe that their insight has substantially strengthened our manuscript. Please find our response/revision plan for each comment below (in bold).

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

This is a very interesting study that addresses an important topic. In brief, the authors build on their previous data showing that LSD1 seems to be neuroprotective. Here they follow the hypothesis that Tau-dependent sequestration of LSD1 to the cytoplasm leads to loss of nuclear LSD1 function. Crossing Tau mutant mice (PS19) to heterozyous LSD1 knock out mice exacerbates phenotypes in PS19 mice, while viral overexpression of LSD1 rescues part of these phenotypes.

As said the data is interesting but lack mechanistic explanation that would allow in my view publication in a very high profile journal. Moreover, there are some data such as the RNA-seq that would not be acceptable in the present for by any journal. However, all of these issues could be addressed by the authors in case reviewers would refer to them.

The sequestration of LSD1 in the cytoplasm by tau, along with the co-localization of LSD1 with tau in human cases (in our previous Nature Communications paper-Christopher et al. 2017) provide a mechanistic explanation (sequestration) for why we are able to exacerbate and rescue tau-mediated neurodegeneration by modulating LSD1. As the reviewer pointed out, we believe that we can address all of the critiques brought up (see responses below). By addressing these critiques we believe that we can further substantiate the mechanism underlying our ability to functionally modulate tau-mediated neurodegeneration in vivo.

**Here are some specific issues.**

1 . Especially the proposed link of Tau-mediated sequestration of LSD1 to the cytoplasm is not fully supported by the data. A key finding shows that LSD1 is seen more in cytoplasm in PS19 mice. However, the biological relevance of this observation cannot be fully appreciated at present, since the magnitude of this phenotype is unclear. Approaches to perform a quantitative analysis in addition to the representative IHC images would be helpful.

The change in localization of LSD1 from nuclear to cytoplasmic that we observe in Tau PS19 mice is dramatic. We tried to convey this magnitude of sequestration in different brain regions by showing a range of representative images. Consistent with this, Reviewer 2 commented that “These data are very strong, the effect is impressive.” Nevertheless, we can attempt to further quantify the change in localization. To accomplish this, we can try two different methods. (1) We can add a nuclear marker and attempt to quantify the level of nuclear versus cytoplasmic LSD1 from the immunofluorescence images. (2) We can also attempt to generate nuclear versus cytoplasmic fractions and quantify LSD1 levels by western blot.

2 . Point 1 might be of specific importance since the subsequent experiments built on the idea that mice with already recued LSD1 levels should have a more severe phenotype in case of Tau pathology. However, they do not really address the role of Tau-mediated sequestration of LSD1 anymore. The authors employ mice that constitutively lack one allele of LSD1 which generally leads to a more severe phenotype in PS19 mice. This is very interesting, but I wonder if reduced LSD1 levels might generally put the network in a more vulnerable state and that other detrimental stimuli that do not cause intracellular protein aggregation might have a similar effect. The authors realize this and address this question by comparing via RNA-seq the gene-expression changes observed in Lsd1+/+, Lsd1Δ/+, PS19 Tau, and PS19;Lsd1Δ/+ littermates. Comparatively few changes are observed. However, the major issue with this experiment is that an n=2/group is simply no acceptable anymore to be published in any serious journal. Thus, this data is not interpretable as it stands.

To further address the role of tau-mediated sequestration of LSD1, we can attempt to quantify (see above) nuclear versus cytoplasmic LSD1 in PS19 Tau mice with heterozygous Lsd1, and compare it to the level of sequestration observed in PS19 Tau mice alone.

To strengthen the RNAseq data, we can perform two additional replicates.__However, because (1) the RNAseq results were only used for genome-wide comparisons, (2) the replicates were very tight, and (3) the results were clear, it is very unlikely that additional replicates are going to alter the result. Thus, alternatively we might be able to alter the language of the manuscript to qualify the result somewhat. In this regard, it should be noted that reviewer 2 commented that “The data are very convincing, and provide a strong molecular base showing a tight overlap in the effected molecular pathways associated with both pathological tau and Lsd1 heterozygosity.” Reviewer 3 also commented that the transcriptomic dataset “__strengthen some of the conclusions.”

Reviewer #1 (Significance (Required)):

The data will be interesting to the field and help to further understand the role of LSD1 in neuroegenerative dieases linked to tauopathy.

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

The manuscript "The inhibition of LSD1 via sequestration contributes to tau-mediated neurodegeneration" by Amanda K. Engstrom, Alicia C. Walker, Rohitha A. Moudgal, Dexter A. Myrick, Stephanie M. Kyle and David J. Katz, is an excellent study that beautifully describes the implication of the epigenetic enzyme LSD1, as downstream mediator of tau pathology in neurodegenerative disease.

The same authors in a previous paper showed that i) loss of LSD1 in the adult mouse brain, leads to significant neuronal cell death and ii) loss of LSD1 induces genome-wide expression changes that significantly overlap with those observed in the brains of postmortem human AD. In this work, are presented initial evidences that in AD brain, LSD1 nuclear function could be disrupted by mislocalization to pathological tau.

In the present work, using the PS19 mouse model, the authors provide the first cytological evidence that pathological tau can prevent LSD1 from properly localizing to the nucleus in hippocampal and cortical neurons. These data are very strong, the effect is impressive. Crossing the PS19 mouse model of taupathology with a mouse model of LSD1 brain heterozygosity LSD1Δ /+, the authors provide functional data that the inhibition of LSD1 function contributes to tau induced neurodegeneration. Indeed, several pathological parameters measured in the PS19 mouse model, are exacerbated in a reduced genetic LSD1 background. Survival rate, motor activity measured with a rotarod test. The behavioral analysis is nicely paralleled by the analysis of spinal cord motor neurons, showing abnormal morphology in the double mutant mice, compared to the PS19. Overall morphology of the hippocampus shows decreased brain size and brain weight. The analysis is accompanied by MRI analysis, showing again very impressive results, with the double mutant being the most affected and the LSD1Δ /+ very similar to WT.

The second part of the work is aimed at demonstrating specificity of the functional interaction between tau pathology and LSD1. The authors provide a very well planned transcriptional profiling of the different mouse models, choosing the most relevant time point (prior the onset of neuronal cell death), very clearly justifying the rational of their choice. The data are very convincing, and provide a strong molecular base showing a tight overlap in the effected molecular pathways associated with both pathological tau and Lsd1 heterozygosity. As final approach, the authors rescue neurodegeneration in the hippocampus of PS19 Tau mice overexpressing LSD1 using a neuron-specific virus. Overall, these data establish LSD1 as a major downstream effector of tau-mediated neurodegeneration indicating that the LSD1 pathway is a potential late stage target for intervention in tauopathies, such as AD.

**Minor point:**

In material and Methods is missing a section dedicated to a detailed description of statistical analysis.

We have added a section to the materials and methods dedicated to a detailed description of statistical analysis (lines 509-518).

Reviewer #2 (Significance (Required)):

I believe that this work will be of great interest for the neurodegenerative together with the neuro-epigenetic field (my personal area of expertise). The identification of a clear new pathway implicated in AD and neurodegeneration together with the suggestion of a possible new therapeutic target (disruption of tau-LSD1 interaction) is of high potential impact for future studies.

We really appreciate this very positive review, which acknowledges the thoroughness of our results, the mechanistic insight that we provided and the “high potential impact” of our work.

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

The work on LSD1 in this manuscript is based on earlier studies that deleting the histone demethylase Lsd1 in adult mice leads to neuronal cell death and that the neurofibrillary tangles in Alzheimer's disease brains can be stained for LSD1.

The manuscript first shows that LSD1 is sequestered in PS19 tau transgenic mice, that a reduction of Lsd1 exacerbates the pathology and that overexpression rescues, complemented by a separate transcriptomic dataset used to strengthen some of the conclusions. As it currently stands, in my view, this work is very preliminary, and I am not sharing all conclusions made by the authors.

**My specific points are following the headers of the Results section:**

(1) Tau pathology depletes LSD1 from the nucleus in the PS19 Tau mouse model: What is clear from the images is that in the PS19 Tau tg mice LSD1 is depleted from the nucleus. What is not correct is that in WT it is only localized to the nucleus. What should be done, is to quantify the relative localization to the two compartments. In addition, a subcellular fractionation could be performed (see point further below).

LSD1 is strictly a nuclear protein with a well-defined nuclear localization signal that interacts with the importin __a__complex (Jin et al, J. Biochem 2014). Reference to this has been added to the text in the introduction (lines 53-54) and in the results (line 87-89). Nevertheless, we can also attempt subcellular fractionation and localization of LSD1 in the nuclear versus cytoplasmic fraction (see response to reviewer 1 above).

(2) Reduction of LSD1 increases the mouse tauopathy phenotype: 2.1. The PS19 Tau Tg mice have been crossed with an Lsd1 heterozygous mutant (LSD1 delta/-). I tried to find the reference as to how this mutant has actually been made (refs 32-34). Ref 32 describes a conditional KO (The position of the gene trap insertion (STOP), downstream of exon 3, truncates the LSD1 open reading frame within the SWIRM domain prior to the amine oxidase domain, which is essential for the catalytic activity of LSD1), which leaves a 210 amino acid truncated protein which is an obvious confound which should be mentioned and discussed. Besides from that, it is not clear to me how the Lsd1 gene was deleted for the current study, i.e. which promoter has been used.

The Lsd1 allele used in this study was generated in the Rosenfeld lab (Wang et al., Nature 2007). This was stated in the acknowledgements, but has now been added to both the main text (lines 105-109) and the methods (lines 386-391) for clarity. ThisLsd1 allele is a null allele that has also been used previously by both our group, as well as by additional groups (For example: Christopher et al., Nature Communications 2017 and Lyons et al., Cell 2013). In this current study, Lsd1 was deleted with the Vasa**-Cre transgenic line. Once the deletion allele passes through the germline, Lsd1 is heterozygous throughout the mouse. We deeply regret this oversight.

2.2. Fig S2 shows that LSD1 is reduced in the heterozygote, but increased in PS19 by 20% and then again decreased in the PS19 x LSD1 delta/-. Clearly, a subcellular fractionation or histological quantification is needed to understand what the levels are in the cytoplasm as compared to the nucleus.

The data referred to in Figure S2 is from bulk brain homogenate showing that there is a reduction of LSD1 in mice carry the deletion allele both in a wild-type and PS19 Tau background. Nevertheless, we can attempt subcellular fractionation and quantification of LSD1 localization in the nucleus versus cytoplasm (see response to reviewer 1 above) to further clarify this result.

2.3. The rescue in Fig 2A is really modest. Certainly, with tau in PS19 potentially trapping LSD1 in the cytoplasm there should be less of LSD1 in the nucleus when there is only one functioning allele. What is needed is a quantification of nuclear and cytoplasmic LSD1 in the genotypes.

We can attempt subcellular fractionation and quantification of LSD1 localization in the nucleus versus cytoplasm (see response to reviewer 1 above) to further clarify this result.

2.4. I don't agree with the statement: 'started only slightly earlier than PS19 Tau mice, but after the appearance of pathological tau in neurons (p. 6)' as tau pathology develops gradually and is present before the age of 6 months in this strain.

This statement refers specifically to the AT8 positive pathology that was quantified in this manuscript (Figure S6). This quantification shows that AT8 positive pathology is present in the hippocampus and cortex, when PS19 Tau mice with reduced LSD1 begin to decline. The text has been amended to clarify this (line 124-125).

(3) Tau pathology is not affected by change in LSD1 levels: This is to be expected as Tau is upstream of LSD1 in a pathocascade.

The quantification of tau pathology was included as a negative control. As we expected, tau pathology is not affected by the change in LSD1 levels. As the reviewer correctly points out, this data is consistent with our model, that tau pathology is upstream of LSD1.

(4) The functional interaction between tau pathology and LSD1 inhibition is specific: The specificity of the interaction needs to be tested by co-immunoprecipitations or proximity ligation assays and by mapping which domains of LSD1 and Tau have a role in trapping, using the appropriate positive and negative controls, as is being routinely done for these kinds of studies.

We too are very interested in whether LSD1 interacts directly or indirectly with tau pathology, and what domains of LSD1 are required for LSD1 to co-localize with tau pathology. However, to address these questions, we will need to perform multiple biochemical experiments (such as the ones suggested by the reviewer) on mice of different ages, as well as human cases. We believe that this is significantly beyond the scope of the current study, which is focused on the functional interaction between tau pathology and LSD1 in mice.

(5) Overexpression of LSD1 rescues neurodegeneration in the hippocampus of PS19 Tau mice: The data in Figure 5 are not convincing.

It is not clear why the reviewer is not convinced by the rescue data in Figure 5. Reviewer 1 acknowledged that “viral overexpression of LSD1 rescues part of these phenotypes” and reviewer 2 agreed that “the authors rescue neurodegeneration in the hippocampus of PS19 Tau mice overexpressing LSD1 using a neuron-specific virus.”

**Minor points:**

Abstract: The statement 'However, the mechanism through which tau contributes to neurodegeneration remains unknown.' is not correct and should be removed. There is a wealth of information on tau-based pathomechanisms available and several studies have identified proteins which become, as seems to be the case for LSD1, trapped by tau in the cytosol.

This statement in the abstract has been modified (lines 13-15).

Reviewer #3 (Significance (Required)):

This form asks me about my expertise. I am working on tau pathomechanisms since more than two decades and the revision experiments I am asking for is what we are doing in our own studies. I find the data on LSD1 interesting, but definitely more work needs to be done to substantiate the claims.

We thank the reviewer for their careful reading of the manuscript and appreciate that they found that the data on LSD1 are interesting.

Overall, we feel that the reviews of our manuscript are very positive. We hope that our response/revision plan will be suitable for publication.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

The work on LSD1 in this manuscript is based on earlier studies that deleting the histone demethylase Lsd1 in adult mice leads to neuronal cell death and that the neurofibrillary tangles in Alzheimer's disease brains can be stained for LSD1.

The manuscript first shows that LSD1 is sequestered in PS19 tau transgenic mice, that a reduction of Lsd1 exacerbates the pathology and that overexpression rescues, complemented by a separate transcriptomic dataset used to strengthen some of the conclusions. As it currently stands, in my view, this work is very preliminary, and I am not sharing all conclusions made by the authors.

My specific points are following the headers of the Results section:

(1) Tau pathology depletes LSD1 from the nucleus in the PS19 Tau mouse model: What is clear from the images is that in the PS19 Tau tg mice LSD1 is depleted from the nucleus. What is not correct is that in WT it is only localized to the nucleus. What should be done, is to quantify the relative localization to the two compartments. In addition, a subcellular fractionation could be performed (see point further below).

(2) Reduction of LSD1 increases the mouse tauopathy phenotype: 2.1. The PS19 Tau Tg mice have been crossed with an Lsd1 heterozygous mutant (LSD1 delta/-). I tried to find the reference as to how this mutant has actually been made (refs 32-34). Ref 32 describes a conditional KO (The position of the gene trap insertion (STOP), downstream of exon 3, truncates the LSD1 open reading frame within the SWIRM domain prior to the amine oxidase domain, which is essential for the catalytic activity of LSD1), which leaves a 210 amino acid truncated protein which is an obvious confound which should be mentioned and discussed. Besides from that, it is not clear to me how the Lsd1 gene was deleted for the current study, i.e. which promoter has been used.

2.2. Fig S2 shows that LSD1 is reduced in the heterozygote, but increased in PS19 by 20% and then again decreased in the PS19 x LSD1 delta/-. Clearly, a subcellular fractionation or histological quantification is needed to understand what the levels are in the cytoplasm as compared to the nucleus.

2.3. The rescue in Fig 2A is really modest. Certainly, with tau in PS19 potentially trapping LSD1 in the cytoplasm there should be less of LSD1 in the nucleus when there is only one functioning allele. What is needed is a quantification of nuclear and cytoplasmic LSD1 in the genotypes.

2.4. I don't agree with the statement: 'started only slightly earlier than PS19 Tau mice, but after the appearance of pathological tau in neurons (p. 6)' as tau pathology develops gradually and is present before the age of 6 months in this strain.

(3) Tau pathology is not affected by change in LSD1 levels: This is to be expected as Tau is upstream of LSD1 in a pathocascade.

(4) The functional interaction between tau pathology and LSD1 inhibition is specific: The specificity of the interaction needs to be tested by co-immunoprecipitations or proximity ligation assays and by mapping which domains of LSD1 and Tau have a role in trapping, using the appropriate positive and negative controls, as is being routinely done for these kinds of studies.

(5) Overexpression of LSD1 rescues neurodegeneration in the hippocampus of PS19 Tau mice: The data in Figure 5 are not convincing.

Minor points:

Abstract: The statement 'However, the mechanism through which tau contributes to neurodegeneration remains unknown.' is not correct and should be removed. There is a wealth of information on tau-based pathomechanisms available and several studies have identified proteins which become, as seems to be the case for LSD1, trapped by tau in the cytosol.

Significance

This form asks me about my expertise. I am working on tau pathomechanisms since more than two decades and the revision experiments I am asking for is what we are doing in our own studies. I find the data on LSD1 interesting, but definitely more work needs to be done to substantiate the claims.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

The manuscript "The inhibition of LSD1 via sequestration contributes to tau-mediated neurodegeneration" by Amanda K. Engstrom, Alicia C. Walker, Rohitha A. Moudgal, Dexter A. Myrick, Stephanie M. Kyle and David J. Katz, is an excellent study that beautifully describes the implication of the epigenetic enzyme LSD1, as downstream mediator of tau pathology in neurodegenerative disease.

The same authors in a previous paper showed that i) loss of LSD1 in the adult mouse brain, leads to significant neuronal cell death and ii) loss of LSD1 induces genome-wide expression changes that significantly overlap with those observed in the brains of postmortem human AD. In this work, are presented initial evidences that in AD brain, LSD1 nuclear function could be disrupted by mislocalization to pathological tau.

In the present work, using the PS19 mouse model, the authors provide the first cytological evidence that pathological tau can prevent LSD1 from properly localizing to the nucleus in hippocampal and cortical neurons. These data are very strong, the effect is impressive. Crossing the PS19 mouse model of taupathology with a mouse model of LSD1 brain heterozygosity LSD1Δ /+, the authors provide functional data that the inhibition of LSD1 function contributes to tau induced neurodegeneration. Indeed, several pathological parameters measured in the PS19 mouse model, are exacerbated in a reduced genetic LSD1 background. Survival rate, motor activity measured with a rotarod test. The behavioral analysis is nicely paralleled by the analysis of spinal cord motor neurons, showing abnormal morphology in the double mutant mice, compared to the PS19. Overall morphology of the hippocampus shows decreased brain size and brain weight. The analysis is accompanied by MRI analysis, showing again very impressive results, with the double mutant being the most affected and the LSD1Δ /+ very similar to WT.

The second part of the work is aimed at demonstrating specificity of the functional interaction between tau pathology and LSD1. The authors provide a very well planned transcriptional profiling of the different mouse models, choosing the most relevant time point (prior the onset of neuronal cell death), very clearly justifying the rational of their choice. The data are very convincing, and provide a strong molecular base showing a tight overlap in the effected molecular pathways associated with both pathological tau and Lsd1 heterozygosity. As final approach, the authors rescue neurodegeneration in the hippocampus of PS19 Tau mice overexpressing LSD1 using a neuron-specific virus. Overall, these data establish LSD1 as a major downstream effector of tau-mediated neurodegeneration indicating that the LSD1 pathway is a potential late stage target for intervention in tauopathies, such as AD.

Minor point:

In material and Methods is missing a section dedicated to a detailed description of statistical analysis.

Significance

I believe that this work will be of great interest for the neurodegenerative together with the neuro-epigenetic field (my personal area of expertise). The identification of a clear new pathway implicated in AD and neurodegeneration together with the suggestion of a possible new therapeutic target (disruption of tau-LSD1 interaction) is of high potential impact for future studies.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This is a very interesting study that addresses an important topic. In brief, the authors build on their previous data showing that LSD1 seems to be neuroprotective. Here they follow the hypothesis that Tau-dependent sequestration of LSD1 to the cytoplasm leads to loss of nuclear LSD1 function. Crossing Tau mutant mice (PS19) to heterozyous LSD1 knock out mice exacerbates phenotypes in PS19 mice, while viral overexpression of LSD1 rescues part of these phenotypes.

As said the data is interesting but lack mechanistic explanation that would allow in my view publication in a very high profile journal. Moreover, there are some data such as the RNA-seq that would not be acceptable in the present for by any journal. However, all of these issues could be addressed by the authors in case reviewers would refer to them.

Here are some specific issues.

1 . Especially the proposed link of Tau-mediated sequestration of LSD1 to the cytoplasm is not fully supported by the data. A key finding shows that LSD1 is seen more in cytoplasm in PS19 mice. However, the biological relevance of this observation cannot be fully appreciated at present, since the magnitude of this phenotype is unclear. Approaches to perform a quantitative analysis in addition to the representative IHC images would be helpful.

2 . Point 1 might be of specific importance since the subsequent experiments built on the idea that mice with already recued LSD1 levels should have a more severe phenotype in case of Tau pathology. However, they do not really address the role of Tau-mediated sequestration of LSD1 anymore. The authors employ mice that constitutively lack one allele of LSD1 which generally leads to a more severe phenotype in PS19 mice. This is very interesting, but I wonder if reduced LSD1 levels might generally put the network in a more vulnerable state and that other detrimental stimuli that do not cause intracellular protein aggregation might have a similar effect. The authors realize this and address this question by comparing via RNA-seq the gene-expression changes observed in Lsd1+/+, Lsd1Δ/+, PS19 Tau, and PS19;Lsd1Δ/+ littermates. Comparatively few changes are observed. However, the major issue with this experiment is that an n=2/group is simply no acceptable anymore to be published in any serious journal. Thus, this data is not interpretable as it stands.

Significance

The data will be interesting to the field and help to further understand the role of LSD1 in neuroegenerative dieases linked to tauopathy.

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

Learn more at Review Commons

Reply to the reviewers

We thank the Reviewers for the positive assessment of our work and their insightful remarks. Please find below a point-by-point response to each comment.

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

*The authors present a well written article describing distinct transcriptomic profiles generated by RNA sequencing analysis of hippocampus, a distinct anatomical area, at well spaced and defined time points of clinical progression following prion inoculation in an established mouse model. The authors contribute significantly in the detailed transcriptomic definition of changes during disease progression, especially during the early and almost asymptomatic stages.

The brain region chosen to perform their analysis is logical as the hippocampus shows clear signs of neuronal degeneration in prion disease progression and furthermore provides a well defined area for analysis that is easily accessible experimentally; Although, more information would be needed to strengthen this choice in relation to the hippocampus playing a key role in the initiation stages of the disease. It remains an anatomical subset of the whole brain and the study would benefit if extended to include other affected areas. *

The hippocampus is one of the most affected and therefore most studied regions during prion disease (Moreno et al., 2012, Nature). We have clarified this in the text (page 3). While the analysis of transcriptional changes in additional brain regions would be of interest, the main conclusions derived by the present analyses on the hippocampus already opens new perspectives to our understanding of this complex disease (e.g. premature pathological changes at 8 weeks occur long before the development of neuropathological and clinical signs). In light of the new findings observed from the first cohort of experimental animals, we designed the rest of the study to prioritize more analyses (e-g- splicing and RNA editing) and validations (e.g. second cohort, aging cohort, plasma administration cohort etc.) in order to provide comprehensive and robust dataset and corroborate our findings. We are currently working on a follow-up study thoroughly describing transcriptional changes during prion disease development in other brain regions. We believe that the inclusion of these data would not be instrumental to support the main conclusions of the present study and may unduly add complexity to the current manuscript.

The article presents comprehensive bioinformatics analysis of the gene expression profiles, during disease progression and continues focusing on two early stages whose profiles clearly cluster together. The authors elegantly query the transcriptomic data extrapolating clusters representative of different cell types and conclude that at preclinical stages microglial-related DEGs are enriched. Importantly, data trends are replicated in an independent animal cohort supporting the experimental design, reproducibility and bioinformatics analysis. Enriched microglial populations from challenged animals compared to controls, would have added more value to the approach.

We agree with the reviewer that certain cell types, including microglia should be investigated in more detail. We are currently working on a study investigating prion-induced changes in a cell-type specific manner using ribosomal profiling. While space reasons prevent us from adding these studies to the present manuscript, we are planning to publish a comprehensive searchable database that will include both the transcriptomics and translatomics data.

The authors proceed to conclude that these transcriptomic enrichment of microglial related DEGs are suggestive of driver events in the initiation of prion disease. Although the statement is gaining a lot of interest in the current literature, it is yet immature to conclude from only RNA sequencing data that microglial neuroinflammation is the causative driver event and not the result of the infection and subsequent neurodegeneration. Taking also into consideration the route of infection (ic) which is expected to initiate an acute immune response in the brain.

Towards that comment, the immunohistochemistry data should show increased immune reaction from the early time points pi.

While the simultaneous occurrence of microglia-related changes and motor decline suggests that microglia may be the ultimate drivers of prion disease progression, we agree that correlation does not prove causation, and have toned down our conclusions to this respect.

Clearly, microglia activation does not play a major role during the early stages of prion disease: we do not see any increased immune reaction at the early time points as the reviewer pointed out, nor do we see any RNA expression changes in microglia-enriched genes at the early time points.

We also don’t believe that the infection is the source of microglia activation for the following reason: if the inoculation itself would induce microglia activation we would expect a strong microglia response directly after the injection that should progressively decrease. Instead, we see no expression change in microglia-enriched genes until 16 wpi. We have clarified the corresponding sections in the text.

To address the reviewer’s point that the route of infection may contribute to the observed changes we have added the following datasets as new Supplemental Fig. 4:

We have analyzed prion induced changes 8wpi and the terminal stage from intraperitoneally inoculated mice (new Supplemental Fig. 4). The prion induced changes between the different routes of administration correlate at the respective timepoint, indicating that the induced changes are independent of the route of prion inoculation. To strengthen the point that the 8 wpi changes are indeed prion-dependent (and thus require in vivo prion replication by incorporation of cellular prion protein PrP), we have additionally included 8 wpi samples from PrP knock-out mice. The knockout mice don’t show any prion-induced changes at 8 wpi (new Supplemental Fig. 4), suggesting that the 8 wpi changes are not the result of the infection and more importantly, are in fact prion-dependent.

Also, the paper would gain significantly, if there were random as well as targeted (eg microglial specific) molecular targets selected, for independent validation by Real-time QC PCR and immunohistochemistry. This would be especially interesting if it was combined with the targets that showed selective splicing like Ctsa, a microglial related gene.

We respectfully disagree with the reviewer on this issue. In the early days of RNAseq, most scientists would validate their results with qPCR of select genes. However, by now RNAseq is widely accepted as the state-of-the-art technique to profile whole transcriptomes and is considered to be more reliable, accurate and sensitive compared to orthogonal methods such as RT-qPCR. Also, RNAseq and RT-qPCR data are highly correlated (typically ~85% and well above 90% when genes with a low expression are neglected; PMID: 28484260). The inclusion of an orthogonal technology is thus only needed when a) no biological replicates are available (potentially detrimental intra-group variability); b)definitive conclusions depend on genes with extremely low expression levels (potentially detrimental high dispersion); c) the main findings of an experiment revolve around one or a handful of genes (potentially detrimental false positives). None of the above applies to this study. Moreover, in terms of overall validation, we already include data from a second, independent cohort of mice with the same experimental settings (Supplementary Fig. 3), as well as from aged mice and from mice with plasma/saline treatment. We therefore maintain that qPCR verification is unnecessary in this instance and may potentially even produce confounders.

*RNA binding deaminase proteins show a similar pattern to a recent report, strengthening the finding that protein levels do not change and/or compensate with other RNA binding and editing enzymes, even though edited targets and editing frequency shift significantly.

The authors continue with RNA editing analysis concluding that they did not find any (apart from two targets being edited) differential RNA editing sites contradictory to a recent study. We believe that this contradiction is a premature conclusion since, the analysis was based on an older protocol that was published by the same group based on GTAK version 3.4.0 from 2011. The predicted RNA edited sites were only based in previously catalogued samples from hippocampus of young mice by Stilling et al 2014. They did not take into account C-U editing in all genomic locations in the whole brain regardless of aging or region. Also, the depth of sequencing was not taken into account which would increase the novel identification of editing sites instead of being limited to previously identified non-validated RNA editing. The study would significantly benefit from Sanger sequencing validation of random and non random edited targets. How do the identified targets validate? *

As suggested by the Reviewer, we have reanalyzed RNA editing using the same editing pipeline as Kanata et al. (PMID: 31492812), neither restricting the analysis to a pre-existing list of candidate sites, nor limiting the analysis to A-to-I editing events. Following this approach, a number of editing sites comparable to those reported by Kanata et al. were identified. However, we did not observe a statistical difference between control and PrD samples at the locus level.

As discussed in the manuscript Kanata et al, analyzed a different brain region using a different infection model. Furthermore, the fact that we assessed triplicates, and the application of strict filters in the selection of putative editing sites might have contributed to us not detecting differentially edited sites. While we used the same parameters linked to quality and depth of coverage, we only considered the intersection of both REDItools and VarScan2, and required that at least 2 out of 3 samples were edited. Regarding the validation of the editing events through Sanger sequencing, we believe it is outside of the scope of the present study because our main goal is not that of exactly pinpointing specific editing sites and hypothesizing on their potential effects. We rather view the editing analysis as an auxiliary layer to the main conclusions of the manuscript, and through the updated analysis and results we believe we have reached such a goal.

Finally, the study concludes with the administration of young plasma at 8 weeks (early stage of the disease) and the authors support that this intervention improves the phenotype of the affected animals without lifespan changes. In our view, this part of the study should either be omitted, or full transcriptomic and clinicopathological improvement should be demonstrated with clear emphasis on microglial-related molecular targets.

While plasma administration does not prolong lifespan and terminal prion-induced changes are very similar in plasma vs saline-treated animals (Fig. 6d-e), we did in fact observe a full transcriptomic improvement upon plasma administration at 8 wpi (Fig. 6b). We currently don’t know if prion-induced 8 wpi changes and the plasma-induced improved health span are linked to microglia-related changes (see also response above). We have therefore not put any additional emphasis on microglia-related targets. We therefore feel that the plasma experiments do add to the present paper, but we would be prepared to discuss with the editors whether it may be appropriate to omit this part and publish it separately. \*Minor comment:**

Other behavioral tests such as T-maze, Morris water-maze, novel object recognition, wouldn't it be better suited for memory assessment? *

Although we agree with the reviewer that these tests are better suited for memory assessment, the purpose of the rotarod evaluation (together with histological and biochemical tests) was to obtain an objective monitoring of clinical disease development. Rotarod assessment has been instrumental to objectivate the genetic or pharmacological modulation of prion disease development (e.g PMID: 29176838; PMID: 26246168; PMID:25502554). A more sophisticated behavioral assessment would go beyond the scope of this study and would require access to specific infrastructures which are not available in our veterinary bio contained research facility allowing the handling of prion-infected mice.

*Reviewer #1 (Significance (Required)):

The authors present a very detailed and informative transcriptomic profiling of a well structured in vivo experiment with a satisfactory number of time points that has provided significant transcriptomic and splicing information at the preclinical stage of the disease. The field would definitely benefit from such a profile oriented approach however the above should be sufficiently addressed. *

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

In the present paper by Sorce et al. the authors studied mRNA changes, splicing and editing alterations during the progression of prion disease in an experimental animal model (inocculated mice). The main findings are that changes in RNA processing and abundance occured very early at around 8 wpi. Interestingly, changes in microglia-enriched genes appeared early and coincided with the onset of clinical symptoms while neuronal genes were unchanged and played a more prominent role at later stages of the disease suggesting that glial cells might be the driving force and pivotal for the early stages and disease progression. Young plasma restored mRNA alterations and was beneficial for delaying neurological symptoms. This conclusion seems to be supported by the data and overall the study was well performed. The findings are clearly presented, the discussion is insightful and balanced and the figures are in general of high quality but there are some concerns that need to be clarified.

The authors could tackle the following comments with a straightforward revision:

1) On Page 3 it is mentioned that RNA sequencing was performed in n=3 samples per time point and for the 20 wpi time point in controls only n=2 samples have been used. Overall this is a very low sample size that needs to be increased. More samples need to be analyzed in order to provide biological relevance. *

We agree with the reviewer that, like in any other biological study entailing a certain degree of experimental variability, increased sample sizes always increase the statistical power and may allow for the detection of changes that might otherwise go undetected. However, there are opportunity costs that go along with enlarging the study. Also, the Swiss Animal Protection Law requires us to adhere to the 3R principles (Replacement, Reduction and Refinement of animal experimentations). We have aimed at using the minimum number of animals allowing us to identify a subset statistically significant and robust changes. Animal welfare considerations and an attempt to prevent an escalation of cost resulted in the majority of experiments being performed with 3 samples per time point.

It is accepted in the field that three replicates are sufficient to identify the vast majority of biologically relevant changes in mRNA abundance. Unless major claims are made about individual genes at the lower end of the expression’s dynamic range (which is not the case in this study), three replicates ensure that about 85 % of the relevant changes are accurately captured (PMID: 29767357; PMID: 26813401). This is particularly true when the variability across replicates is low and appropriate analysis tools, such as edgeR, are employed (PMID: 30726870).

We used age and gender matched inbred C57BL/6J mice in a microbiologically tightly-controlled environment (see Methods) to minimize interindividual variability. This allowed us to identify thousands of statistically significant prion-dependent changes, despite the low sample number.In few exceptions we sequenced two instead of three replicates (eg because a sample got lost, the RNA was degraded, or the sample did not pass quality control after sequencing). In these instances, we ensured that both replicates showed a high correlation and could thus still yield reliable results. Furthermore, we have validated the RNA expression changes in an independent cohort of mice with the same experimental settings (Supplementary Fig. 3), as well as from aged mice and from mice with plasma/saline treatment, indicating that the observed changes are robust.

2) On page 4 it is written that 'While clusters 2 and 3 consist predominantly of microglia and neuronal genes, cluster 1 and 4 contain genes enriched in multiple cell types'. A few sentences later, the authors write, that 'Neuronal genes almost exclusively belonged to clusters 3 and 4................, whereas microglia genes were essentially contained in clusters 1 and 2. These two statements are contradictory. Please explain and clarify.

Compared to clusters 2 and 3, the enriched genes in clusters 1 and 4 don’t predominantly fall into one category (eg cluster 4 contains ~30% neuronal-enriched genes, ~25% oligodendrocyte enriched genes, 20% endothelial-enriched genes – see Fig. 1d). However, of 203 neuronal-enriched DEGs, 143 are cluster 3 genes (~70%), 50 are cluster 4 genes (~25%), while only 10 are cluster 1 genes (~5%) and 0 are cluster 2 genes. To better illustrate this point, we have included these numbers as an additional Table in Supplementary Fig. 2.

3) On page 5 the authors claim that 'astro- and microgliosis became evident at 16 wpi...........' This statement is based solely on histological images and needs to be confirmed by quantification. However in Supplementary figure 5c astrocytes and microglia (GFAP and Iba1 staining) are almost not visible and the overview images too superficial. I recommend high resolution images and additional inserts and a solid quantification.

We have added high resolution pictures, additional inserts and a quantification of the stainings (new Supplementary Fig. 6).

4) On Page 6 the authors write 'We observed progressive decline in motor performance starting 18 wpi'. However, the graph in figure 3a clearly shows only a significant difference at 19 wpi'. This needs to be corrected.

• The decline in motor performance shows a visible trend at 18 wpi but only becomes statistically significant at 19 wpi. We have clarified this in the text.*

5) Figure 6c: it would make sense to combine both graphs (saline and plasma) for a direct comparison of prion infected mice that received saline or plasma so that potential differences would be easier to recognize ......although they seem to be pretty modest.

We have combined both graphs from Fig. 6c into one but believe that it becomes very difficult to extract any information from the figure. We shall defer to the reviewer’s judgment but we would prefer to keep the original figure.

Reviewer #2 (Significance (Required)):

The findings are of interest to a wide readership and the paper thus seems suited to be published, but there are some concerns that need to be clarified (see specific comments above).

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

In the present paper by Sorce et al. the authors studied mRNA changes, splicing and editing alterations during the progression of prion disease in an experimental animal model (inocculated mice). The main findings are that changes in RNA processing and abundance occured very early at around 8 wpi. Interestingly, changes in microglia-enriched genes appeared early and coincided with the onset of clinical symptoms while neuronal genes were unchanged and played a more prominent role at later stages of the disease suggesting that glial cells might be the driving force and pivotal for the early stages and disease progression. Young plasma restored mRNA alterations and was beneficial for delaying neurological symptoms. This conclusion seems to be supported by the data and overall the study was well performed. The findings are clearly presented, the discussion is insightful and balanced and the figures are in general of high quality but there are some concerns that need to be clarified.

The authors could tackle the following comments with a straightforward revision:

1) On Page 3 it is mentioned that RNA sequencing was performed in n=3 samples per time point and for the 20 wpi time point in controls only n=2 samples have been used. Overall this is a very low sample size that needs to be increased. More samples need to be analyzed in order to provide biological relevance.

2) On page 4 it is written that 'While clusters 2 and 3 consist predominantly of microglia and neuronal genes, cluster 1 and 4 contain genes enriched in multiple cell types'. A few sentences later, the authors write, that 'Neuronal genes almost exclusively belonged to clusters 3 and 4................, whereas microglia genes were essentially contained in clusters 1 and 2. These two statements are contradictory. Please explain and clarify.

3) On page 5 the authors claim that 'astro- and microgliosis became evident at 16 wpi...........' This statement is based solely on histological images and needs to be confirmed by quantification. However in Supplementary figure 5c astrocytes and microglia (GFAP and Iba1 staining) are almost not visible and the overview images too superficial. I recommend high resolution images and additional inserts and a solid quantification.

4) On Page 6 the authors write 'We observed progressive decline in motor performance starting 18 wpi'. However, the graph in figure 3a clearly shows only a significant difference at 19 wpi'. This needs to be corrected.

5) Figure 6c: it would make sense to combine both graphs (saline and plasma) for a direct comparison of prion infected mice that received saline or plasma so that potential differences would be easier to recognize ......although they seem to be pretty modest.

Significance

The findings are of interest to a wide readership and the paper thus seems suited to be published, but there are some concerns that need to be clarified (see specific comments above).

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

The authors present a well written article describing distinct transcriptomic profiles generated by RNA sequencing analysis of hippocampus, a distinct anatomical area, at well spaced and defined time points of clinical progression following prion inoculation in an established mouse model. The authors contribute significantly in the detailed transcriptomic definition of changes during disease progression, especially during the early and almost asymptomatic stages.

The brain region chosen to perform their analysis is logical as the hippocampus shows clear signs of neuronal degeneration in prion disease progression and furthermore provides a well defined area for analysis that is easily accessible experimentally; Although, more information would be needed to strengthen this choice in relation to the hippocampus playing a key role in the initiation stages of the disease. It remains an anatomical subset of the whole brain and the study would benefit if extended to include other affected areas.

The article presents comprehensive bioinformatics analysis of the gene expression profiles, during disease progression and continues focusing on two early stages whose profiles clearly cluster together. The authors elegantly query the transcriptomic data extrapolating clusters representative of different cell types and conclude that at preclinical stages microglial-related DEGs are enriched. Importantly, data trends are replicated in an independent animal cohort supporting the experimental design, reproducibility and bioinformatics analysis. Enriched microglial populations from challenged animals compared to controls, would have added more value to the approach.

The authors proceed to conclude that these transcriptomic enrichment of microglial related DEGs are suggestive of driver events in the initiation of prion disease. Although the statement is gaining a lot of interest in the current literature, it is yet immature to conclude from only RNA sequencing data that microglial neuroinflammation is the causative driver event and not the result of the infection and subsequent neurodegeneration. Taking also into consideration the route of infection (ic) which is expected to initiate an acute immune response in the brain.

Towards that comment, the immunohistochemistry data should show increased immune reaction from the early time points pi. Also, the paper would gain significantly, if there were random as well as targeted (eg microglial specific) molecular targets selected, for independent validation by Real-time QC PCR and immunohistochemistry. This would be especially interesting if it was combined with the targets that showed selective splicing like Ctsa, a microglial related gene.

RNA binding deaminase proteins show a similar pattern to a recent report, strengthening the finding that protein levels do not change and/or compensate with other RNA binding and editing enzymes, even though edited targets and editing frequency shift significantly.

The authors continue with RNA editing analysis concluding that they did not find any (apart from two targets being edited) differential RNA editing sites contradictory to a recent study. We believe that this contradiction is a premature conclusion since, the analysis was based on an older protocol that was published by the same group based on GTAK version 3.4.0 from 2011. The predicted RNA edited sites were only based in previously catalogued samples from hippocampus of young mice by Stilling et al 2014. They did not take into account C-U editing in all genomic locations in the whole brain regardless of aging or region. Also, the depth of sequencing was not taken into account which would increase the novel identification of editing sites instead of being limited to previously identified non-validated RNA editing. The study would significantly benefit from Sanger sequencing validation of random and non random edited targets. How do the identified targets validate?

Finally, the study concludes with the administration of young plasma at 8 weeks (early stage of the disease) and the authors support that this intervention improves the phenotype of the affected animals without lifespan changes. In our view, this part of the study should either be omitted, or full transcriptomic and clinicopathological improvement should be demonstrated with clear emphasis on microglial-related molecular targets.

Minor comment:

Other behavioral tests such as T-maze, Morris water-maze, novel object recognition, wouldn't it be better suited for memory assessment?

Significance

The authors present a very detailed and informative transcriptomic profiling of a well structured in vivo experiment with a satisfactory number of time points that has provided significant transcriptomic and splicing information at the preclinical stage of the disease. The field would definitely benefit from such a profile oriented approach however the above should be sufficiently addressed.

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

Learn more at Review Commons

Reply to the reviewers

Reviewer #1:

**Summary**

Jang et al., address the important question of spatially localized or compartmentalized metabolic enzymes with a focus on the glycolytic enzyme PFK1. Using a good strategy of inserting a fluorescent tag at the endogenous PFK1 locus with tissue-specific inducible expression in C. elegans, combined with strong quantitative longitudinal imaging and innovative bioengineered microfluidic-hydrogels to control oxygen availability as well as optogenetic approaches, they show PFK1 condensates, which are not stress granules and not seen in normoxia, assemble with hypoxia. PFK1 condensates are dynamic, reversible, localized at the synapse in neurons, and recruit aldolase, another glycolytic enzyme. Although glycolytic proteins were previously shown to compartmentalize near the plasma membrane, and PFK1 was previously shown to assemble into filaments in vitro and be punctate at the plasma membrane in mammalian cells, evidence for cellular localized PFK1 condensates in animals is highly significant. The work includes strong biophysical characterization of PFK1 phase-separated condensates, but no clear indication of the composition of condensates. More significantly, the findings lack functional significance related to PFK1 activity or glycolytic flux with hypoxia vs normoxia. Despite previous work by this group showing that disrupting subcellular localization of glycolytic enzymes impairs neuronal activity in response with hypoxia, the reader is left with questions on the importance of localized and PFK1 condensates and their make-up .

**Major comments:**

Key conclusions are convincing, and most experimental approaches, biophysical characterization including thermodynamic principles, and data analysis are exemplary and well described. However, as indicated above, the work is limited to a descriptive analysis of cellular localization of PFK1 condensates and their biophysical properties without insights on functional significance relative to enzyme activity - or at least glycolytic flux or metabolic reprogramming with hypoxia. At best, only correlations can be drawn from hypoxia-induced localized PFK1 condensates and the authors' previous report (Jang et al., 2016) on hypoxia-regulated neuronal activity. Some insight or at least prediction in the discussion on the differences in spatially localized PFK1 in muscle vs neurons with regard to metabolic or energy distinctions should be included.

We have added additional discussions on the differences of the spatially localized PFK-1.1 in muscles versus neurons, explaining that in both tissues the cellular enrichment appears to be at sites predicted to have high ATP consumption (lines 128-133; 482-484).

Despite the strong biophysical analysis of condensates, several important features are not determined. First is at best a rudimentary analysis of the composition of condensates and also how PFK1 is assembled into these structures. For the former, is the core of the condensate predominantly PFK1 with perhaps aldolase only recruited to the periphery or is aldolase an integral component of the structure. Hence, is it a PFK1 condensate or a glycolytic condensate? For the latter question, is there a particular orientation for PFK1 in condensates, i.e a collection of filaments as previously reported, which might provide insight on assembly? Finally, and less critical but also important is the criterion for spherical, which is not well defined, and at least some idea or speculation on determinants for a spherical morphology - compared with filaments that have been reported for other non-glycolytic metabolic enzymes.

We have now co-expressed PFK-1.1 and ALDO-1 and examined their dynamic formation during hypoxic conditions. We observe PFK-1.1 and ALDO-1 form condensates simultaneously, with gradual enrichment of both molecules. We now include this new data in Figure 7E and Video 8; lines 422-441, 964-989). We also include genetic data demonstrating the ALDO-1 requires pfk-1.1 to form condensates, and that PFK-1.1 requires aldo-1 as well. Therefore, the enzymes are interdependent on each other to form condensates (Figures 7G, 7H, S7B, and S7C).

The spheroid geometry reflects liquid-like properties, which arises from surface tension of molecules loosely held together via multi-valent interactions. Filamentous arrangements reflect crystalline-like structures resulting from more stable interactions between molecules into solid-like states. While we did not perform high resolution studies, like Cryo-EM, to resolve this question, the spheroid geometry of PFK-1.1 condensates, along with its fluid-like properties, suggest the condensates are liquid-like compartment distinct to filamentous structures. We now add this discussion in lines 467-470.

The work is an important advance in our understanding on the self-assembly of metabolic enzymes by showing hypoxia-induced PFK1 condensates in vivo, their spatially-restricted subcellular localization in muscle cells and neurons, and their biophysical properties, the latter being distinct from those of stress granules. Taken together, these findings are more extensive than many previous reports on the assembly of metabolic enzymes into filaments or condensates, but fall short for new insights on functional significance.

We focus this study on the biophysical characterization of the condensates, and how that results in compartmentalized enrichment of glycolytic proteins. Examination of the functional significance of the phase separation to the enzymatic reactions in vivo is not currently possible because we lack probes we can use in vivo to measure the metabolites resulting from the reaction. We have now added discussion acknowledging this and framing its significance in the context of what has been published in the field (lines 484-492). For example, a recent manuscript in ChemRxiv demonstrated, in vitro, that the enzymatic activity of glycolytic proteins, hexokinase and glucose-6 phosphate dehydrogenase, promote these enzymes condensing into liquid droplets. The authors further found that the condensation accelerated the glycolytic reactions (Ura et al., 2020). This raises the question whether glycolytic proteins compartmentalize, and form condensates, in vivo, which we address in this manuscript. We capture this point in (lines 444-464) where we explain that, while it has long been hypothesized that glycolytic proteins like PFK-1 could be compartmentalized, this remained controversial due to lack of dynamic in vivo imaging. In our study, and through a systematic examination of endogenous PFK-1.1 via the use of a hybrid microfluidic-hydrogel device, we conclusively determine that PFK-1.1 indeed displays distinct patterns of subcellular localization in specific tissues in vivo.

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

This paper reports on the condensation of the glycolytic enzyme PFK-1 in response to hypoxic conditions in neurons of C. elegans. The authors employ a microfluidic-hydrogel device to dynamically monitor the relocalisation of PFK-1 from a mostly diffuse state to clusters in response to hypoxia and show that PFK-1 can undergo multiple rounds of PFK-1 clustering and dissolution. The authors work through the key features of a liquid-like compartment (sphericity, fusion, fast internal rearrangements) and give evidence that PFK-1 may have all three. Finally, the authors tag PFK-1 with the light-inducible multimerization domain Cry2 and find that even without light PFK-1 will constitutively form clusters that sequestrate endogenous PFK-1 as well as other glycolytic proteins. The strength of this work is that it is characterizing what appears to likely be phase separation in the context of a whole animal experiencing a stress that it could encounter in the natural world. A limitation of the work is that it is unclear what the functional implications are of condensates of PFK-1 at the molecular or cell scale.

**Major comments:**

-All experiments were performed using fluorescently tagged PFK-1 expressed from endogenous promoter or from the native genetic locus which is important for excluding overexpression artifacts. However, there is still risk that the GFP tag is driving the assembly process. In order to exclude tag-specific effects that may cause aggregation of the tetrameric PFK-1, ideally a control would be done in which PFK-1 is visualized through immunofluorescence experiments of WT cells. Alternatively, a short tag (e.g HA, His) as epitope for is an alternative .

We used fluorescent tags to observe the dynamic relocalization in vivo. While in the study we have not performed immunofluorescence, we established the validity of the labeling method by: 1) using monomeric versions of GFP; 2) using different fluorophores to show the same condensation phenomenon; 3) performing CRISPR for single copy insertions; 4) Demonstrating that different glycolytic proteins form condensates; 5) demonstrating the GFP-tagged versions of the protein are capable of rescuing the loss-of-function alleles and 6) Now adding new data demonstrating the observed localization specifically depend on the presence of other glycolytic proteins. This last result supports that GFP tag is not driving the assembly process of glycolytic condensate and that the glycolytic condensate formation requires the presence of specific molecules in the pathway. I add that we routinely use fluorophore markers to over a dozen distinct proteins that label subcellular compartments, and we have never observed the dynamic relocalization reported here, with the exception of other glycolytic proteins that interact with PFK, suggesting this is a property specific to glycolytic proteins, and, based on the genetic studies, dependent on the glycolytic reaction. We add and discuss these findings in Figures 7G, 7H, S7B, and S7C; lines 422-441, 964-989.

-For the Cry2-section, the complementation of the pfk-1 mutant supports functionality of the synaptic clustering phenotype. Are there other features of function that can be evaluated or could you look at how Cry-2 vs wt worms recover from different durations of stress or frequencies. Could you see if the Cry-2-fusion will rescue function to a partial-loss-of-function allele or a tetramerization deficient allele? A detailed analysis of the effects of constitutive presence of PFK-1-Cry2 clusters would be necessary to bolster claims that this is fully functional construct. Can enzyme activity be somehow monitored?

We did not observe any difference between wild-type worms and CRY2-expressing worms with regards to their development, survival, locomotive behavior or synaptic phenotype. While we can not discard the possibility that this is not a full rescue, with available tools, we can not distinguish the recue with PFK-1-Cry2 from that of just PFK-1.

-The analysis of the sphericity of clusters (4A) is limited due to the diffraction limit of light which limits an analysis of a compartment of this size. While this is a limitation of the live organism, this should be more clearly acknowledged.

We have included in the Methods section our criteria for quantifying condensates and avoiding diffraction limit artifacts. Briefly, “Considering the resolution limit of a spinning disc confocal (approximately 300nm), any structure with a diameter less than 500nm and an area smaller than 0.2 µm2 was excluded from the analyses”. To better clarify this point, we also now add a description of the criteria used in the main text (lines 242-243).

In addition, we observed that PFK-1.1 condensates are not perfect spheres, but constrained spheroids (which can not be explained by diffraction-limited point spread functions). We can explain the observed spheroid shapes based on liquid-like properties of the condensates, and the constrains of the diameter of the neurite. To better highlight this finding, we have now moved Figure S4E into the main figure (Figure 4B’).

-Fusion experiments (4C) do not fully exclude that clusters overlap instead of merging. It would be beneficial to show the foci for several subsequent frames. One would expect that upon fusion, the condensate size would increase, but video 3 suggests the opposite. It would be useful to quantify condensate size before and after fusion for several separate fusion events. -an alternative possible experiment would be the tagging of PFK-1 with a photoconvertible fluorophore (e.g. Dendra2) and subsequent analysis of fusion events

To better show the fusion events in Figure 4C, we now include all xy, yz, and zx plane views of before and after fusion events of Figure 4C (Figure S5B). We also added a quantification of four independent fusion events in which we compare the sum of the areas of the two puncta before fusion and the size of the area of the single punctum after fusion (Figure S5C). These data support that we are observing fusions events.

-4D). It is unclear if foci are indeed undergoing fission or if two clusters next to each other are moving apart.

For Figure 4D, in all the frames we had recorded, a single structure maintains a continuous signal until fission occurs and splits into two structures. To better present this event, we now include an unabridged version of figure of 4D in the supplement that shows all the frames captured (Figure S5D).

-The analysis of side-by-side growth and dissolution kinetics are interesting and a novel view into the non-equilibrium aspects of phase separation in cells.

-Purification of PFK-1 and in vitro reconstitution of condensates would be supportive of liquid-like characteristics although I don't think it is necessary however it would add a lot to the relevance to show enzyme activity is different +/- condensate state but I am not sure if an easy enzymatic assay exists in vitro.

We agree. But the significance of this particular paper, specifically in the context of the in vitro enzymatic work on glycolytic proteins, is to examine the dynamic in vivo localization and the biophysical characteristics of the condensates. To better underscore this in the context of the field, we add discussion of a recent in vitro manuscript demonstrating that liquid droplet formation of glycolytic proteins affect their enzymatic activity (Ura et al., 2020) (lines 444-464; 484-492). While we see the value of future studies reconstituting the glycolytic particles, we believe that is beyond the scope of this particular in vivo study.

**Minor comments:**

-Stress granules in other organisms (yeast paper) have different composition depending on stress type. To make the claim that the PFK-1 compartments are independent of SGs one would ideally test multiple different SG markers.

We selected the stress granule protein TIAR-1 because it is one of the most studied stress granule markers in C. elegans and it is reportedly one of the core proteins and universal components of stress granules irrespective of a stress type (Buchan et al., 2011; Gilks et al., 2004; Huelgas-Morales et al., 2016; Kedersha et al., 1999). Although we did not include images in the manuscript, we had tested a total of three stress granule markers: TIAR-1, TDP-43, and G3BP1 with similar results. We now added that as data not shown (lines 193-194).

-it should be stated in the main text that the microfluidic-hydrogel device was fabricated following previously published protocols

We have added the reference in the main text (line 170) to supplement what we had written in the Methods section: “A reusable microfluidic PDMS device was fabricated to deliver gases through a channel adjacent to immobilized animals, following protocols as previously described (Lagoy and Albrecht, 2015)”.

-Figure 4b: Y-axis should be changed from probability to fraction of occurrence

We have corrected this in both the figure and the figure legends (Figure 4B).

-The discussion should be less speculative concerning any effects seen in PFK1-Cry2 expressing C. elegans

We have modified the discussion as suggested.

-it is perplexing that a protein known to tetramerize with no disordered or RNA-binding domains forms condensates like this. Is there anything known from other systems of additional interacting proteins that may have features that promote liquidity and serve to fluidize these assemblies?

Condensates can form via multivalent interactions, which include, but is not exclusive, to disordered or RNA-binding domains. Because glycolytic proteins have dihedral symmetries that can facilitate multivalent interactions, we believe these structural properties, in combination with regulated conformational changes, promote multivalent interactions leading to their condensation. We had a statement in the discussion (lines 494-519) now add this more clearly in the results (lines 395-398).

Reviewer #2 (Significance (Required)):

Stimulus-induced phase separation has been observed for dozens of metabolic enzymes from various different pathways (reviewed in Prouteau, 2018). Several studies have published the formation of condensates through PFK-1 in diverse organisms (C. elegans, Yeast, human cancer cells) in response to hypoxia or in some cancer lines also without hypoxia (Jin, 2017, Jang, 2016, Kohnhorst 2017, etc.). A yeast study showed that PFK-1 condensates contain various other glycolytic enzymes and that condensate formation enhances glycolytic rates (Jin, 2017).

This study gives the advance of analyzing the dynamics of PFK-1 condensate formation in vivo in the context of a live animal using a microfluidic-hydrogel device and showing that PFK-1 relocalizes to reversible condensates within minutes of hypoxia. If further appropriate experiments (as mentioned above) are performed, this study would strongly suggest that the underlying process of PFK-1 condensate formation is liquid-liquid phase separation. Ideally, if at all feasible, it would be strengthened if there was some insight into the functional consequences of the condensed assemblies formed in hypoxia. These findings may be interesting to researchers working on glycolysis and metabolism in different cells but particularly in neurons.

Field of expertise

-Phase separation, microscopy, in vitro reconstitution

-no experience with C. elegans biology and do not have a practical handle on ease or difficulties of genetic manipulation of C. elegans or metabolic assays for PFK-1

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

**Summary:**

In this manuscript, the authors focus on the subcellular localization of the key glycolytic enzyme PFK-1.1 in C. elegans, initially in whole animals through GFP tagging of the endogenous locus and subsequently in single cells/tissues using a clever genome editing strategy that permitted tissue-specific expression of GFP-tagged PFK-1.1 from its endogenous locus. They observe that PFK-1.1 localization differs from cell-type to cell-type and can be dynamically reorganized in response to exogenous cues. Focusing on hypoxia, they observe that PFK-1.1 forms foci near synapses in neurons under this stress condition. These foci are not stress granules and they are dissolved upon re-oxygenation. These condensates have properties of liquid droplets and can mature (harden) over time. PFK-1.1 fused to the CRY domain can trigger condensate formation under normoxic conditions, which can co-recruit WT PFK-1.1 as well as aldolase.

**Major comments:**

The conclusions are convincing but the impact could be increased if the authors were able to demonstrate the physiological role that the observed phase separation plays in this stress response. Would it be possible to assess glycolytic flux under hypoxia vs normoxia?

It is currently not possible to assess glycolytic flux in vivo in our system, as we lack metabolic sensors (an active area of work we are trying to address, but will take several years to perform correctly). We have added discussion of new in vitro studies examining the consequences of metabolic flux due to glycolytic compartmentation into liquid droplets (Ura et al., 2020), and the significance of those findings in the context of our in vivo studies (lines 444-464; 484-492).

The authors should comment on viability during the hypoxia time course.

C. elegans can survive anoxic condition for a day (Powell-Coffman, 2010). Our hypoxic conditions last minutes, and we can rescue live C. elegans upon completion of the assays. We now include a description of this in the Methods (lines 1216-1218).

The co-clustering of ALDO-1 and PFK-1.1::mCh::CRY2 in Figure 7 should be properly quantified/statistically analyzed

We quantified the fraction of animals that displays ALDO-1 clustering in PFK-1.1::mCh::CRY2 co-expressing animals, as suggested (Figure S7C).

A control of mCh::CRY2 + ALDO-1::EGFP is missing from the experiments shown in Figure 7. Is the presence of mCh::CRY2 sufficient to drive ALDO-1::EGFP clustering?

As a control for the CRY2 tag promoting the formation of glycolytic condensates, we had co-expressed mCh::CRY2 with PFK-1.1::EGFP, which is insufficient to cause the formation of the condensate (Figure 7C). We have now added a new data where we show that in pfk-1.1 deletion mutants, ALDO-1 condensate formation is suppressed, which further demonstrates the dependency between PFK-1.1 and ALDO-1 (Figures 7H and S7C).

Does hypoxia trigger co-clustering of ALDO-1 and PFK-1.1?

To answer this question, we examined the dynamic formation of ALDO-1 and PFK-1.1 condensates by co-expressing the two proteins together and observed that hypoxia triggers their co-clustering. We now include this in Figure 7E and Video 8.

The authors speculate that hypoxia acts via diminished energy (altered ATP AMP ratios). Can this be measured? To support this hypothesis, the authors may wish to test if similar phase separation is triggered by mitochondrial poisons.

We currently lack sensors that can reliably measure, in vivo, the subcellular changes in energy or metabolic flux in C. elegans neurons. However, we previously did test mitochondrial mutants and observed that in those mutants we observe glycolytic condensates (Jang et al., 2016), supporting the idea that defects in energy production promotes the formation of glycolytic condensates.

**Minor comments:** Is 21% O2 not hyperoxic for worms?

While C. elegans are known to prefer lower percentage of oxygen than those in air, in the lab animals are reared in normal air. We therefore used 21% oxygen present in air as our normoxic conditions.

Can the authors speculate more on how do these condensates exhibit "memory" (how they're able to cluster in the same place repeatedly)? Is there any role for the cytoskeleton in mediating nucleation and/or condensation of PFK and glycolytic enzymes?

When we were testing the reversibility of PFK-1.1 condensates, we were not expecting the reappearance of PFK-1.1 condensates in the same place repeatedly. Our current speculation is that, because many glycolytic enzymes, such as PFK-1.1, are allosterically regulated by nucleotides, AMP/ATP ratio may play a role on where glycolytic condensates appear. In other words, the specific synaptic areas, where PFK-1.1 condensate repeatedly reappeared, may have different AMP/ATP ratio that may trigger the condensation of the glycolytic proteins in those locationsupon conformational changes in PFK-1. We can’t exclude, currently, the presence of nucleating factors at synapses that facilitate PFK-1 clustering, but we have not yet identified them. We now include a discussion of this (lines 494-519).

Do the authors think that these clusters are effectively G-bodies from yeast?

G-bodies from yeast also shows glycolytic proteins changing from its diffuse localization to punctate localization in response to hypoxia (Jin et al., 2017). G-bodies, like C. elegans glycolytic condensates, are forms of subcellular glycolytic organization within eukaryotic cells. Yet, G-bodies take 24 hours to form, while we observe the glycolytic clusters in C. elegans within minutes of hypoxic conditions. We will need to understand the composition and function of both to determine if these forms of glycolytic subcellular organization represent the same structure. We note that glycolytic clusters have also been observed in some human cancer cell lines (Kohnhorst et al., 2017). Observation of glycolytic compartments in multiple different species and cell types suggest that, although the regulation, composition and formation kinetics of the glycolytic condensates may differ, compartmentalization of glycolytic enzymes may be a conserved feature. We now add a sentence discussing this (line 535-537).

Reviewer #3 (Significance (Required)):

It is much appreciated that this study tackles the cell biology of signaling and metabolism, which is a fascinating but difficult to study aspect of molecular biology. This work conclusively documents the dynamic reorganization of metabolic enzymes in vivo in response to physiological stimuli. Such reorganization had been proposed previously but was controversial and difficult to study in a controlled way. This work not only confirms previous observations but further demonstrates that the dynamic reorganization is mediated by a liquid-liquid phase separation. What is lacking is a demonstration that this phase separation is physiologically important. Such observations would generate interest from a much broader audience; the present audience presently targeting people specifically interested in non-membrane organelles per se. The reviewer has expertise in cell signalling and its regulation by phase separation.

As we explain for Reviewer 1, we focus this study on the biophysical characterization of the condensates, and how that results in compartmentalized enrichment of glycolytic proteins. Examination of the functional significance of the phase separation to the enzymatic reactions in vivo is not currently possible because we lack probes we can use in vivo to measure the metabolites resulting from the reaction. We have now added discussion acknowledging this and framing its significance in the context of what has been published in the field (lines 484-492). For example, a recent manuscript in ChemRxiv demonstrated, in vitro, that the enzymatic activity of glycolytic proteins, hexokinase and glucose-6 phosphate dehydrogenase, promote these enzymes condensing into liquid droplets. The authors further found that the condensation accelerated the glycolytic reactions (Ura et al., 2020). This raises the question whether glycolytic proteins compartmentalize, and form condensates, in vivo, which we address in this manuscript. We capture this point in (lines 444-464) where we explain that, while it has long been hypothesized that glycolytic proteins like PFK-1 could be compartmentalized, this remained controversial due to lack of dynamic in vivo imaging. In our study, and through a systematic examination of endogenous PFK-1.1 via the use of a hybrid microfluidic-hydrogel device, we conclusively determine that PFK-1.1 indeed displays distinct patterns of subcellular localization in specific tissues in vivo.

**REFEREES CROSS-COMMENTING** Globally it seems that all reviewers feel that impact would be increased if the physiological consequence of PFK-1.1 condensates was examined. Other, specific comments seem fair.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

In this manuscript, the authors focus on the subcellular localization of the key glycolytic enzyme PFK-1.1 in C. elegans, initially in whole animals through GFP tagging of the endogenous locus and subsequently in single cells/tissues using a clever genome editing strategy that permitted tissue-specific expression of GFP-tagged PFK-1.1 from its endogenous locus. They observe that PFK-1.1 localization differs from cell-type to cell-type and can be dynamically reorganized in response to exogenous cues. Focusing on hypoxia, they observe that PFK-1.1 forms foci near synapses in neurons under this stress condition. These foci are not stress granules and they are dissolved upon re-oxygenation. These condensates have properties of liquid droplets and can mature (harden) over time. PFK-1.1 fused to the CRY domain can trigger condensate formation under normoxic conditions, which can co-recruit WT PFK-1.1 as well as aldolase.

Major comments:

The conclusions are convincing but the impact could be increased if the authors were able to demonstrate the physiological role that the observed phase separation plays in this stress response. Would it be possible to assess glycolytic flux under hypoxia vs normoxia?

The authors should comment on viability during the hypoxia time course.

The co-clustering of ALDO-1 and PFK-1.1::mCh::CRY2 in Figure 7 should be properly quantified/statistically analyzed

A control of mCh::CRY2 + ALDO-1::EGFP is missing from the experiments shown in Figure 7. Is the presence of mCh::CRY2 sufficient to drive ALDO-1::EGFP clustering?

Does hypoxia trigger co-clustering of ALDO-1 and PFK-1.1?

The authors speculate that hypoxia acts via diminished energy (altered ATP AMP ratios). Can this be measured? To support this hypothesis, the authors may wish to test if similar phase separation is triggered by mitochondrial poisons.

Minor comments: Is 21% O2 not hyperoxic for worms? Can the authors speculate more on how do these condensates exhibit "memory" (how they're able to cluster in the same place repeatedly)? Is there any role for the cytoskeleton in mediating nucleation and/or condensation of PFK and glycolytic enzymes? Do the authors think that these clusters are effectively G-bodies from yeast?

Significance

It is much appreciated that this study tackles the cell biology of signalling and metabolism, which is a fascinating but difficult to study aspect of molecular biology. This work conclusively documents the dynamic reorganization of metabolic enzymes in vivo in response to physiological stimuli. Such reorganization had been proposed previously but was controversial and difficult to study in a controlled way. This work not only confirms previous observations but further demonstrates that the dynamic reorganization is mediated by a liquid-liquid phase separation. What is lacking is a demonstration that this phase separation is physiologically important. Such observations would generate interest from a much broader audience; the present audience presently targeting people specifically interested in non-membrane organelles per se. The reviewer has expertise in cell signalling and its regulation by phase separation.

REFEREES CROSS-COMMENTING Globally it seems that all reviewers feel that impact would be increased if the physiological consequence of PFK-1.1 condensates was examined. Other, specific comments seem fair.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

This paper reports on the condensation of the glycolytic enzyme PFK-1 in response to hypoxic conditions in neurons of C. elegans. The authors employ a microfluidic-hydrogel device to dynamically monitor the relocalisation of PFK-1 from a mostly diffuse state to clusters in response to hypoxia and show that PFK-1 can undergo multiple rounds of PFK-1 clustering and dissolution. The authors work through the key features of a liquid-like compartment (sphericity, fusion, fast internal rearrangements) and give evidence that PFK-1 may have all three. Finally, the authors tag PFK-1 with the light-inducible multimerization domain Cry2 and find that even without light PFK-1 will constitutively form clusters that sequestrate endogenous PFK-1 as well as other glycolytic proteins. The strength of this work is that it is characterizing what appears to likely be phase separation in the context of a whole animal experiencing a stress that it could encounter in the natural world. A limitation of the work is that it is unclear what the functional implications are of condensates of PFK-1 at the molecular or cell scale.

Major comments:

-All experiments were performed using fluorescently tagged PFK-1 expressed from endogenous promoter or from the native genetic locus which is important for excluding overexpression artifacts. However, there is still risk that the GFP tag is driving the assembly process. In order to exclude tag-specific effects that may cause aggregation of the tetrameric PFK-1, ideally a control would be done in which PFK-1 is visualized through immunofluorescence experiments of WT cells. Alternatively, a short tag (e.g HA, His) as epitope for is an alternative .

-For the Cry2-section, the complementation of the pfk-1 mutant supports functionality of the synaptic clustering phenotype. Are there other features of function that can be evaluated or could you look at how Cry-2 vs wt worms recover from different durations of stress or frequencies. Could you see if the Cry-2-fusion will rescue function to a partial-loss-of-function allele or a tetramerization deficient allele? A detailed analysis of the effects of constitutive presence of PFK-1-Cry2 clusters would be necessary to bolster claims that this is fully functional construct. Can enzyme activity be somehow monitored?

-The analysis of the sphericity of clusters (4A) is limited due to the diffraction limit of light which limits an analysis of a compartment of this size. While this is a limitation of the live organism, this should be more clearly acknowledged.

-Fusion experiments (4C) do not fully exclude that clusters overlap instead of merging. It would be beneficial to show the foci for several subsequent frames. One would expect that upon fusion, the condensate size would increase, but video 3 suggests the opposite. It would be useful to quantify condensate size before and after fusion for several separate fusion events.

-an alternative possible experiment would be the tagging of PFK-1 with a photoconvertible fluorophore (e.g. Dendra2) and subsequent analysis of fusion events

-4D). It is unclear if foci are indeed undergoing fission or if two clusters next to each other are moving apart.

-The analysis of side-by-side growth and dissolution kinetics are interesting and a novel view into the non-equilibrium aspects of phase separation in cells.

-Purification of PFK-1 and in vitro reconstitution of condensates would be supportive of liquid-like characteristics although I don't think it is necessary however it would add a lot to the relevance to show enzyme activity is different +/- condensate state but I am not sure if an easy enzymatic assay exists in vitro.

Minor comments:

-Stress granules in other organisms (yeast paper) have different composition depending on stress type. To make the claim that the FPK-1 compartments are independent of SGs one would ideally test multiple different SG markers.

-it should be stated in the main text that the microfluidic-hydrogel device was fabricated following previously published protocols

-Figure 4b: Y-axis should be changed from probability to fraction of occurrence

-The discussion should be less speculative concerning any effects seen in PFK1-Cry2 expressing C. elegans

-it is perplexing that a protein known to tetramerize with no disordered or RNA-binding domains foms condensates like this. Is there anything known from other systems of additional interacting proteins that may have features that promote liquidity and serve to fluidize these assemblies?

Significance

Stimulus-induced phase separation has been observed for dozens of metabolic enzymes from various different pathways (reviewed in Prouteau, 2018). Several studies have published the formation of condensates through PFK-1 in diverse organisms (C. elegans, Yeast, human cancer cells) in response to hypoxia or in some cancer lines also without hypoxia (Jin, 2017, Jang, 2016, Kohnhorst 2017, etc.). A yeast study showed that PFK-1 condensates contain various other glycolytic enzymes and that condensate formation enhances glycolytic rates (Jin, 2017).

This study gives the advance of analyzing the dynamics of PFK-1 condensate formation in vivo in the context of a live animal using a microfluidic-hydrogel device and showing that PFK-1 relocalizes to reversible condensates within minutes of hypoxia. If further appropriate experiments (as mentioned above) are performed, this study would strongly suggest that the underlying process of PFK-1 condensate formation is liquid-liquid phase separation. Ideally, if at all feasible, it would be strengthened if there was some insight into the functional consequences of the condensed assemblies formed in hypoxia. These findings may be interesting to researchers working on glycolysis and metabolism in different cells but particularly in neurons.

Field of expertise

-Phase separation, microscopy, in vitro reconstitution

-no experience with C. elegans biology and do not have a practical handle on ease or difficulties of genetic manipulation of C. elegans or metabolic assays for PFK-1

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary

Jang et al., address the important question of spatially localized or compartmentalized metabolic enzymes with a focus on the glycolytic enzyme PFK1. Using a good strategy of inserting a fluorescent tag at the endogenous PFK1 locus with tissue-specific inducible expression in C. elegans, combined with strong quantitative longitudinal imaging and innovative bioengineered microfluidic-hydrogels to control oxygen availability as well as optogenetic approaches, they show PFK1 condensates, which are not stress granules and not seen in normoxia, assemble with hypoxia. PFK1 condensates are dynamic, reversible, localized at the synapse in neurons, and recruit aldolase, another glycolytic enzyme. Although glycolytic proteins were previously shown to compartmentalize near the plasma membrane, and PFK1 was previously shown to assemble into filaments in vitro and be punctate at the plasma membrane in mammalian cells, evidence for cellular localized PFK1 condensates in animals is highly significant. The work includes strong biophysical characterization of PFK1 phase-separated condensates, but no clear indication of the composition of condensates. More significantly, the findings lack functional significance related to PFK1 activity or glycolytic flux with hypoxia vs normoxia. Despite previous work by this group showing that disrupting subcellular localization of glycolytic enzymes impairs neuronal activity in response with hypoxia, the reader is left with questions on the importance of localized and PFK1 condensates and their make-up .

Major comments:

Key conclusions are convincing, and most experimental approaches, biophysical characterization including thermodynamic principles, and data analysis are exemplary and well described. However, as indicated above, the work is limited to a descriptive analysis of cellular localization of PFK1 condensates and their biophysical properties without insights on functional significance relative to enzyme activity - or at least glycolytic flux or metabolic reprogramming with hypoxia. At best, only correlations can be drawn from hypoxia-induced localized PFK1 condensates and the authors' previous report (Jang et al., 2016) on hypoxia-regulated neuronal activity. Some insight or at least prediction in the discussion on the differences in spatially localized PFK1 in muscle vs neurons with regard to metabolic or energy distinctions should be included.

Despite the strong biophysical analysis of condensates, several important features are not determined. First is at best a rudimentary analysis of the composition of condensates and also how PFK1 is assembled into these structures. For the former, is the core of the condensate predominantly PFK1 with perhaps aldolase only recruited to the periphery or is aldolase an integral component of the structure. Hence, is it a PFK1 condensate or a glycolytic condensate? For the latter question, is there a particular orientation for PFK1 in condensates, i.e a collection of filaments as previously reported, which might provide insight on assembly? Finally, and less critical but also important is the criterion for spherical, which is not well defined, and at least some idea or speculation on determinants for a spherical morphology - compared with filaments that have been reported for other non-glycolytic metabolic enzymes.

Significance

The work is an important advance in our understanding on the self-assembly of metabolic enzymes by showing hypoxia-induced PFK1 condensates in vivo, their spatially-restricted subcellular localization in muscle cells and neurons, and their biophysical properties, the latter being distinct from those of stress granules. Taken together, these findings are more extensive than many previous reports on the assembly of metabolic enzymes into filaments or condensates, but fall short for new insights on functional significance.

Expertise is published on topic

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

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary:

In this work the authors investigate the dynamics of cell morphogenesis in a convenient in vivo system. They use the terminal cells of the embryonic tracheal system and comprehensively address how cell shape change (elongation in this case) takes place and how membrane is remodeled during the process. By combining different high-resolution techniques, i.e. in vivo imaging of terminal cells expressing different membrane markers and serial-section electron tomography, they describe the organelle organization (ER, Golgi, different types of vesicles) in the terminal cells during their elongation. They identify the presence of membrane structures/vesicles particularly abundant at the tip of the cell ahead of the growing tube. When they block endocytosis they find increased tube membrane and lack of basal membrane growth. In addition, in tomograms, they observe clear membrane defects, like invaginations that could even connect the tube and basal membranes. This correlates with the absence of vesicles at the tip observed in normal conditions.

The analysis of the nature of the vesicles observed indicated the accumulation of late endosomes and MVB, particularly at the tip of the growing terminal cells. Interfering with the formation of these MVB led to defects in the growth of both the tube and the basal membrane.

Altogether the authors propose a model in which the newly formed membrane (and transmembrane proteins) passes through the ER and Golgi and reaches the apical membrane. The incorporated membrane is then rapidly endocytosed and follows a maturation pathway through MVB, from where different cargoes and membrane would be sorted and recycled back to the apical (tube) membrane, or to the basal membrane through a transcytosis mechanism

Major comments:

Are the key conclusions convincing?

Most of the conclusions presented are convincing and supported by the results observed.

However, to my understanding, one of the key conclusions of this work (that membrane is transcytosed from the apical to the basal domain) is not fully convincing. A critical result to support the author's conclusion of apical-basal transcytosis is to find clear evidence of basal accumulation of a transcytosed marker.

1. The authors show accumulation of FGFR-GFP and Myospheroid as evidence. However, I find the results presented not very convincing. The accumulation of FGFR-GFP at the basal membrane in the control is not very clear in the images and movie presented. In addition, in the shibire mutant, some basal accumulation of FGFR-GFP seems to be detected (particularly in the movie). In the figures the authors show an increase of FGFR-GFP intensity when endocytosis is blocked, but this is not explained in the text.

If we understand the referee correctly, there are two parts to her/his concern:

1. Are the proteins we use in fact present basally in normal tracheal cells, i.e. are they good candidates for transcytosed cargo?

2.Do they change their localisation when endocytosis is disrupted? And this point can be divided into two aspects: a. do they change at the basal membrane? b. do they change at the apical membrane (this latter point is not questioned the referee)?

1. The FGFR and beta-integrin are the only known basal markers in tracheal cells. A major reason for being confident of their presence in the basal membrane, even though they are difficult to visualize, is that the biological function of both is at the basal membrane, with the FGFR receiving growth or chemotactic signals from the surrounding tissue, and integrins anchoring the branches on the underlying tissue. However, it is indeed the case that their expression levels are very low, and it is difficult to visualize them, whether by expression of GFP-labelled constructs or by immunofluorescence. We have pushed to the limit a number of methods to improve the detection, but we seem to be constrained by the biology of these molecules.

In addition to the low but detectable signal at the outer boundary cell, some signal is always visible within the cell, which we had in the past always interpreted as an artefact or background, but for which our findings here might provide alternative interpretations.

1. a. We agree with the referee’s assessment that FGFR::GFP is still detectable in the basal membrane after blocking endocytosis. This is, in our view, no contradiction to our model. The most parsimonious interpretation is that this is the FGFR that had already been delivered basally before we interfered with endocytosis, and which remains there after endocytosis is blocked.

b. In addition to this basal pool of FGFR, cells with blocked endocytosis accumulate abnormally high levels of FGFR at the apical membrane, in fact at much higher intensities than at the basal membrane. This is the more dramatic aspect of the phenotype, and our conclusions therefore rely not so much on a possible reduction of basal signal after blocking endocytosis (which would not be possible to demonstrate reliably), but rather on the abnormally enriched presence in the apical membrane.

A technical point: The increase in FGFR::GFP on endocytosis blockage that we show in Figure S4 corresponds to the cytoplasmic + apical pool of the FGFR. We used the Dof signal in the cell to create a mask of the total cell volume to 3D-segment the FGFR signal. Therefore, this analysis does not take into account the FGFR that is present in the basal membrane. We had explained this only in the methods section but will now describe it more explicitly in the Results section.

On the other hand, and very importantly, how does FGFR localization relates to its activity? The authors show that when endocytosis is prevented dpERK (i.e. a reporter of FGFR activation) is not decreased, indicating that FGFR is normally active. Wouldn't this suggest that FGFR is still localized basally to receive Bnl signal?

Indeed, and this is also what we see. This is not in conflict with any of the results or known functions of the receptor. If endocytosis is blocked, FGFR cannot be internalized and removed from the basal membrane, where it is needed to receive the FGF from the surrounding cells.

Our concern, and the reason for doing the experiment, had been that endocytosis might be required for FGF signaling, and that this might account for the failure of the cell to grow. But it turns out that our results show that this is not the case, at least not for the terminal cell in this time frame.

Actually, as the authors indicate, in larval tracheal cells the intracellular accumulation of the FGFR leads to a reduced FGF signal transduction (Chanut-Delalande et al., 2010), suggesting that reduced FGF signaling activity in these cells is due to less FGFR reaching the basal membrane.

That is true, and again, not inconsistent with our own results. In the cited study, trafficking was blocked at a step downstream of endocytosis. In this experimental situation, the internalisation of the FGFR would therefore occur as normal in these cells, but due to the impaired function of the ESCRT complex, intracellular processing, and therefore potential re-delivery to the basal membrane would be impaired. Furthermore, if (as we now propose) FGFR is also delivered via late endosomes in this context, blocking the ESCRT pathway should also impair initial FGFR delivery. In either way, initial delivery or re-delivery of the receptor being blocked, it is reasonable to assume that reduced signal transduction is the result of reduced basal FGFR.

Thus in our study we see no reduction in basal FGFR and no reduction (and even an increase) in signaling, while Chanut et al see reduced basal FGFR and reduced signaling, and the reason for this is that they interfere with a different step of the membrane trafficking pathway.

1. The results with Myospheroid are not very convincing either, as the authors just show a single confocal section of control and shibire mutants. In summary, I consider that this very important point needs to be better documented before concluding that apical membrane material containing basal cargoes is transcytosed to the basal membrane.

We observed this phenotype in several instances. We will quantify it for the resubmission.

1. Another conclusion that, to my opinion, should be better explained and documented, is the coordination of tube and basal membrane growth. Following the movements of the vesicles the authors conclude that there is a net displacement of these vesicles to the tip of the cell. This correlates with the presence of mature endosomes there. So the results postulate that transcytosis occurs at the tip, and therefore the growth of the basal membrane would occur preferentially at the tip. It has been demonstrated, as the authors indicate, that the tube membrane grows all along the length of the tube. How is then coordinated the tube and basal membrane growth? If, as the authors propose in their model, the tube membrane also grows after a process of endocytosis and recycling, wouldn't it be expected to have preferential tip growth? How do the authors reconcile all these observations with previously published results (Gervais and Casanova, 2010)?

The elegant study by Gervais and Casanova (2010) used a very clever method, which was however entirely non- quantitative (and did not make any claims to the contrary, either). The conclusion that material is delivered to the tube throughout its length was based on looking at the displacement relative to the base and the tip of the cell of short and transient secondary branches (seen for example in Fig. 1H in our paper). If the tube would primarily receive material at the tip, these secondary branches would not change their position with respect to the base of the cell. Instead, these branches tend to be displaced towards the tip, which shows that material is also added between the branch and the base of the cell. These branches are seen only in a fraction of wild type terminal cells and quantification is therefore difficult. Thus, the experiment shows convincingly that material is also delivered behind the transient branch and excludes a model by which all growth occurs only at the tip. But it does not discuss what proportion of the total is delivered along the length vs the tip of the branch.

Our model also does not contest the idea of ubiquitous membrane delivery over the length of the tube, either in the initial delivery step, or during redistribution. On the contrary, the presence along the length of the cell of vesicles carrying FYVE::GFP and Rab7, and of smaller MVB-like bodies in the EM sections, suggest that the pathway can also be deployed at a distance from the tip.

1. The serial-section electron tomography analysis is very interesting and identifies different sorts of vesicles. However, it is very unclear what the different vesicles referred in the models correspond to (in the in vivo imaging for instance). For instance, the small granular or the dense-core vesicles correspond to endocytic vesicles at different stages of maturation?

Based on distribution alone, it is almost impossible to determine which of these vesicles correspond to endosomes or to secretory vesicles, even for the extremely experienced EM experts in the team. We would require high resolution CLEM, and a wide range of fluorescent markers to be able to determine which population of vesicles found on EM correspond to each marker. Due to the broad distribution of these vesicles within the cell and their small sizes right now it would be extremely technically challenging to pursue this question (even though we too would love to know)

1. If there is a constant endocytosis from the apical membrane to generate the basal and build the definitive apical membrane, wouldn't it be expected to find many more vesicles around the tube? Wouldn't it be expected to find coated vesicles around or budding from the tube, as the coated vesicles observed budding from the basal membrane in Fig 2D? Or is the endocytosis observed mediated by non-clathrin coated vesicles?

We agree, and we had observed this to be the case, but had not quantified this. We have now analysed the distribution of coated pits and their density in the apical or the basal membrane of the cell. Overall, we found a higher density of endocytic events in the apical membrane than in the basal membrane. As the reviewer noticed, we also found that the majority of endocytic events in the basal membrane occur towards the tip of the cell. We will add these data to Figure S3.

Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

The results of Serp accumulation upon MVB interference can lead to confusion (from line 404). The authors seem to suggest that Serp protein is exclusively produced in the FB and transported by transcytosis to reach the tracheal lumen. However, Serp is also produced in the tracheal cells themselves. In fact, serp expression in the FB seems to be detected by late embryogenesis, while expression in tracheal cells is detected much earlier (Dong et al. 2014; Luschnig et al. 2006; Wang et al. 2006). It was also shown that Serp undergoes a recycling mechanism from the lumen to the lumen, through the endosomes-TGN retrograde trafficking, that may also require Shrub (Dong et al, 2014; Dong et al 2013). Thus, it is unclear (and even unlikely) that the Serp found in the vesicles in Shrub-GFP mutants is derived exclusively from the transcytosed component from the FB. I suggest to better explore this issue or to remove this part.

We agree that the notion that the Serp we see is exclusively derived from the fat body is not correct. We observed Serp accumulation around Shrb::GFP sites in embryos at early and late stages, so it is likely that what we see is the result both of apical-to-apical redelivery of Serp (as reported in Dong et al., 2014a), and transcytosis of Serp from the basal membrane to the apical. We will therefore rewrite this. But regardless of whether we are looking at transcytosis, or apical-to-apical cycling, this experiment still reinforces the idea of our work that late endosomes serve as stations that collect material from and re-deliver it towards various compartments in the cell.

Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

As indicated before, more conclusive results for transcytosis should be provided.

I suggest that the authors determine the presence of apical and basal cargoes (FGFR) in the late endosomes found when Shrub activity is impaired. According to his model, both types of markers should accumulate there. A high accumulation of these markers in those endosomes would reinforce the hypothesis proposed.

We agree with the reviewer that this would be a great experiment. Since we only have an FGFR tagged with GFP we cannot do this experiment using the Shrb::GFP line, so instead we would have to use shrb mutants. This in itself is not a problem (see also below – we will be adding some data), but the experiment would require multi-generation crosses which could only be started once the current Covid-19 restrictions are lifted and labs opened again. Instead, we propose to cite the following supporting evidence. Chanut-Delalande et al., (2010) showed that mutants for components of the ESCRT pathway hrs and stam show intracellular accumulation of overexpressed FGFR::GFP in tracheal cells of the air sac primordium, and Dong et al., (2014a) show that shrb mutants accumulate Crb in late endocytic compartments in tracheal cells of the dorsal trunks. We would suggest it is very likely that terminal cells therefore also accumulate Crb and FGFR in late endosomes in the absence of Shrb. An experiment would be nicer, but we fear this is the best we can do at the moment.

While it has been reported that shrub-GFP act as a dominant negative in different contexts (Dong et al, 2014; Sweeney et al 2006) it is unclear why. So it would be desirable to confirm the results with a loss of function condition (either mutant or RNAi line).

We will now add data on shrb mutants where we find a phenotype that is similar as in Shrb::GFP overexpression.

Are the data and the methods presented in such a way that they can be reproduced?

The materials and methods section would benefit from more detailed explanations.

We agree.

Are the experiments adequately replicated and statistical analysis adequate?

Many of the experiments presented in this work are technically very challenging, like the in vivo analyses and particularly the serial-section electron tomography. This prevents having high numbers of replicates on occasions.

Minor comments:

1. In the abstract the authors state: "We show that apical endocytosis and late endosome-mediated trafficking determine the membrane allocation to the apical and basal membrane domains". I think that the authors show that "that apical endocytosis and late endosome-mediated trafficking is required for correct membrane growth", but I am not that sure that they show that it determines the membrane allocation

We agree and will change this.

1. References for the PH-GFP localization in cells should be provided. Which is the evidence that it only localizes to plasma membrane?

Apical localization of PH-GFP is preferential rather than exclusive. The construct has nevertheless been widely used as an ‘apical’ marker, based on the fact that this PH domain binds to PIP2, which is enriched in the apical domain of epithelial cells (e.g. Pilot, et al., 2006; Román-Fernández, et al., 2018).

1. It would be more adequate to always use the same terms to facilitate the reading. For instance, in several figures the membranes are referred as basal plasma membrane and tube membrane, but in others outer and inner membrane

We will go through the entire manuscript and use a consistent nomenclature.

1. Figure 3C,D and corresponding text are difficult to understand. The increase of fluorescence of the inner membrane seems to be very high, even higher than the corresponding to the outer membrane. Can the authors explain better this point and also describe better the method applied in the materials and methods section?

Rather than absolute amounts the graphs show fold increase over the amount of membrane at the beginning of the recording. We had used this representation for two reasons. First, the overall signal intensity can vary from one imaging set to the next, so comparing and representing absolute amounts from different datasets is not easily possible. However, we understand now why the representation in our plots was confusing. We will now show in Figure 3D how the total amount of plasma membrane in shibirets cells increases in the same way as in controls. For Figure 3C, we have found a better way of representing what percentage of the total membrane is present in each compartment as the cell grows. We will rewrite this part to make it clearer, and we will describe more thoroughly in the methods section how the analysis was done.

Significance:

This work represents an important advance for the field for several reasons. First of all it represents a technical advance because the authors are able to combine the traditional genetic analysis with two powerful techniques (in vivo imaging and serial-section electron tomography) to analyze single cell behavior at high resolution (temporal and spatial). In addition it represents a conceptual advance as it proposes a mechanism through which membrane growth is coordinated to regulate cell morphogenesis. The mechanism presented (endocytosis and transcytosis) is not new but they find evidence in an in vivo system.

It was previously known that tracheal terminal cells undergo a process of intracellular tube formation and cell elongation at the same time, but the mechanisms coordinating these two cell events were not known. The proposed mechanism may not only be relevant for the morphogenesis of tracheal terminal cells, but could represent a general mechanism of cell morphogenesis. Therefore, the paper should be relevant for research in the morphogenesis area but also in the cell biology field, as it shows how regulated membrane trafficking can control tissue morphogenesis

REFEREES CROSS-COMMENTING:

I agree with reviewer #4 on her/his comments and suggestions about analyzing the involvement of the recycling endosome in the process.

Reviewer #2 (Evidence, reproducibility and clarity):

In their manuscript, Mathew et al present a model in which transcytosis is utilized to deliver endocytosed apical membranes to supply basal membrane growth. The authors examined the developing terminal cell of the fruitfly tracheal system, which is a well established tubulogenesis model, as these cells form subcellular tubes by apical plasma membrane invagination. The authors show that basal membrane growth stops when endocytosis or endosomal transport is blocked, while the apical membrane grows excessively or membrane material accumulates in the cytosol, respectively.

Significance:

The authors used high-end microscopy (including CLEM and electron tomography) to support their model and in my opinion, the quality and the quantity of the presented data are indeed adequate for this. The text is well written, the figures are of superb quality, and several cartoons help to understand the presented data/experiments. Therefore I highly recommend submitting the manuscript to a cell biology journal in its present form.

Reviewer #3 (Evidence, reproducibility and clarity):

This paper investigates the role of membrane trafficking in growth of the polarized tracheal cell that forms a cellular projection containing a subcellular tube. The authors show that apical endocytosis and late endosome-mediated trafficking determine the membrane allocation to the apical and basal membrane domains. Basal plasma membrane growth stops if endocytosis is blocked, whereas the apical membrane grows excessively. Plasma membrane is initially delivered apically, and then appears to be continuously endocytosed, together with apical and basal cargo. The sorting and recycling of apical and basolateral membrane appears to occur in a novel organelle carrying markers of late endosomes and multivesicular bodies (MVBs). Inhibiting endocytosis eliminates this compartment.

The work in this paper is generally of high quality, and with one exception, quite comprehensive. The writing and figures are clear.

Major concerns:

1. A central focus of the paper is that balance between apical and basal membrane and the role of transcytosis in moving membrane from the apical compartment to the basolateral compartment. The current view is that transcytosis in mammalian cells usually goes through the recycling endosomes which are marked by rab11, although there is evidence for some trafficking through MVBs as well. In Drosophila, Rab11 positive recycling endosomes are frequently examined as part of endocytic system analyses. However, rab11 is not used as a marker in this paper and indeed there is no mention at all of recycling endosomes, even though recycling is at the core of the work. Since the authors do not examine rab11 or other possible markers of recycling endosomes, it is unclear whether the organelle they identify as carrying markers of late endosome and MVBs is some MVB/recycling endosome hybrid, or whether the organelle is completely distinct from the recycling endosome. Consequently, it is not possible to assess whether the observed trafficking either uses or does not use involve the recycling endosome. This ambiguity make is difficult to relate the observed trafficking to other systems. Minimally the authors should stain for rab-11 in WT and in some of the conditions where trafficking is perturbed and determine if the MVB-like compartment they are observing is rab11 positive, and whether the recycling endosome are affected by the perturbations. Further experiments may be needed to resolve whether any trafficking is going via the recycling endosome or this new MVB-type structure, but without even preliminary data on the relationship between the MVB compartment and the recycling endosome, its hard to say what might be appropriate or exactly how long addressing this will take. But just staining for rab11 in WT and a few mutant conditions to get a handle on what is up with the recycling endosomes in these cells should take less than a month.

We had done a number of experiments on Rab11 but did not include them because we felt they did not add any crucial insights on the mechanism we describe here. However, as the reviewer rightfully points out, Rab11 is a classical marker for recycling and transcytosis and we agree that the reader should know our results. We found that overexpressed Rab11::GFP as well as endogenously tagged Rab11:YFP are both highly enriched around the tube. Unlike Rab5 which is seen in widely spaced discrete vesicles, Rab11 forms a cloud of small puncta. We found very low overlap of Rab11 with CD4::mIFP-positive vesicles at the tip of the cell. This suggests that Rab11 is unlikely to be directly involved in the transcytosis pathway we describe here.

Loss of Rab11 was harder to analyse; Rab11-RNAi did not show any obvious phenotype, which could be due to high maternal contribution or to low knockdown efficiency, none of which we analysed in detail. Expression of a dominant negative Rab11 resulted in very early defects (reported by Le Droguen et al., 2015) which prevented us from analysing the role of Rab11 in tube formation. But because Rab11 does not localize to the compartment at the tip of the cell, we believe that this structure does not rely on Rab11 to transfer material from the apical to the basal membrane of the cell.

We will add the data on Rab11 distribution in Figure S6.

1. In addition to the above, I would recommend more discussion of how the authors' results relate to membrane trafficking and transcytosis in other systems. The recycling endosome should be considered, and it may be appropriate to draw comparisons to membrane trafficking in neurons that goes through MVBs (e.g. reviewed in VON BARTHELD and ALTICK Prog Neurobiol. 2011 Mar; 93(3): 313-340.). Although neurons are not hollow, they have definite morphological resemblance to tracheal terminal branches.

We thank the reviewer for the observation and will expand the discussion on MVB-mediated transcytosis in other systems.

1. line 204-208 "To test whether raised levels of Crb were responsible for the excessive apical membrane, as reported in other contexts (Pellikka et al., 2002; Schottenfeld-Roames et al., 2014; Wodarz et al., 1995), we knocked down Crb (Fig. S4G-H)." According to the legend for Fig s4, the authors express an RNAi construct against crbs. However, there does not appear to be any quantification of the amount knockdown of crb that was achieved. This is a concern for two reasons: 1) RNAi in the embryonic trachea works poorly for most genes (for unknown reasons) 2) This does not appear to be a clonal experiment but rather a pan tracheal driving of Crb RNAi. Loss of crbs would be expected to have very negative effects on tracheal morphogenesis (although this hasn't been rigorously tested to this reviewer's knowledge), but there doesn't appear to be any adverse effects of pan-tracheal crbs RNAi, suggesting that little if any knockdown of crb was actually achieved.

The authors either need to document the reduction of crbs or remove this paragraph. Preferably, they would be able to document the reduction of crb because they are trying to address an important point and if they can show the apical expansion is crb-independent, that would be an nice result.

Loss of Crumbs definitely is detrimental to embryonic epithelia, to different degrees (Tepass and Knust, 1990). We therefore could not use mutants but expressed an RNAi, which does not abolish Crb completely. We have now determined the degree of crb knockdown in our experiments. We stained Crb in embryos that expressed crb-IR in the entire tracheal system (but leaving the epidermal expression intact) and quantified the amount of Crb in the tracheal dorsal trunks, normalized to the signal in the epidermis. We found that in crb-IR embryos, Crb levels were reduced by around 50% compared to control siblings. We will add these results to Fig. S4 in the new version of the manuscript.

Minor concerns:

1. Lines 250. "By this interpretation, unscissioned membrane invaginations protruding from the subcellular tube would occasionally have touched the basal plasma membrane or its protrusions and fused with it, as transcytosing vesicles would have done in the normal situation."

I am not convinced by the argument that the bridging invaginations are fusing analogously to transcytosing vesicles because a protrusion/nascent vesicle coming from the apical surface should have rabs and V-SNAREs that should dock the protrusion/nascent vesicle with an endosomal compartment, not the basolateral surface. A transcytotic vesicle would have the rabs and V-SNAREs for the basolateral membrane. So it would seem that a fusion of the apical surface directly to the basolateral surface would have to be an ectopic event outside of the normal situation.

We agree that the fusion events could also be unrelated to the normal physiology of the animal. In other contexts (e.g. the embryonic epidermis, the synaptic bouton) blocking dynamin results in long membrane invaginations as a result of failure in membrane scission. In terminal cells, the apical and basal plasma membranes are very close to each other, and we believe this increases the chance of membrane invaginations meeting and fusion to take place. In addition, the long membrane invaginations we see seem to have been stripped of their clathrin coat suggesting that at least some aspects of ‘vesicle’ maturation proceed even though scission had failed. We also find evidence of small vesicles that resemble the contents of MVBs being deposited within the aberrant membrane invaginations. This suggests that MVBs are able to fuse with these unscissioned tubes and sheets, again indicating that the appropriate molecular markers are present, and the machinery in charge of generating these vesicles is active at the invaginated pits directly. In either case, we will rephrase our interpretations of these data and present it as speculation in the discussion section.

1. Significant figures. This is not a big deal, but the authors are over reporting their significant figures. E.g. "a 7.05-fold increase (+/-2.98 SD)" . With an SD that is 50% the value of the measurement, reporting to hundreds is definitely beyond the accuracy of measurement. Rounding to tenths would be more appropriate.

We agree with the reviewer. We will rephrase this section and use more appropriate metrics. As prompted by Reviewer #1, we modified the analysis that corresponds to this sentence, which also modified the way data is normalized also reducing the spread. This happened because in the new analysis we compare apical and basal signal for each timepoint, which allows better comparison between different cells.

Significance:

As there have not been that many studies on the dynamics of membrane trafficking during morphogenesis, the results should be of broad interest to those studying the endocytic system and the role of membrane trafficking in morphogenesis. However, the paper would be greatly strengthened if the authors considered the recycling endosome in their analysis and write up. As a well-known compartment for trafficking cargo and membrane to both the apical and basolateral surface, it is hard to know how to interpret the observed trafficking without knowing the involvement, or lack thereof, of recycling endosomes in this system.

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

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

This paper investigates the role of membrane trafficking in growth of the polarized tracheal cell that forms a cellular projection containing a subcellular tube. The authors show that apical endocytosis and late endosome-mediated trafficking determine the membrane allocation to the apical and basal membrane domains. Basal plasma membrane growth stops if endocytosis is blocked, whereas the apical membrane grows excessively. Plasma membrane is initially delivered apically, and then appears to be continuously endocytosed, together with apical and basal cargo. The sorting and recycling of apical and basolateral membrane appears to occur in a novel organelle carrying markers of late endosomes and multivesicular bodies (MVBs). Inhibiting endocytosis eliminates this compartment. The work in this paper is generally of high quality, and with one exception, quite comprehensive. The writing and figures are clear.

Major concerns:

-A central focus of the paper is that balance between apical and basal membrane and the role of transcytosis in moving membrane from the apical compartment to the basolateral compartment. The current view is that transcytosis in mammalian cells usually goes through the recycling endosomes which are marked by rab11, although there is evidence for some trafficking through MVBs as well. In Drosophila, Rab11 positive recycling endosomes are frequently examined as part of endocytic system analyses. However, rab11 is not used as a marker in this paper and indeed there is no mention at all of recycling endosomes, even though recycling is at the core of the work. Since the authors do not examine rab11 or other possible markers of recycling endosomes, it is unclear whether the organelle they identify as carrying markers of late endosome and MVBs is some MVB/recycling endosome hybrid, or whether the organelle is completely distinct from the recycling endosome. Consequently, it is not possible to assess whether the observed trafficking either uses or does not use involve the recycling endosome. This ambiguity make is difficult to relate the observed trafficking to other systems. Minimally the authors should stain for rab-11 in WT and in some of the conditions where trafficking is perturbed and determine if the MVB-like compartment they are observing is rab11 positive, and whether the recycling endosome are affected by the perturbations. Further experiments may be needed to resolve whether any trafficking is going via the recycling endosome or this new MVB-type structure, but without even preliminary data on the relationship between the MVB compartment and the recycling endosome, its hard to say what might be appropriate or exactly how long addressing this will take. But just staining for rab11 in WT and a few mutant conditions to get a handle on what is up with the recycling endosomes in these cells should take less than a month.

-In addition to the above, I would recommend more discussion of how the authors' results relate to membrane trafficking and transcytosis in other systems. The recycling endosome should be considered, and it may be appropriate to draw comparisons to membrane trafficking in neurons that goes through MVBs (e.g. reviewed in VON BARTHELD and ALTICK Prog Neurobiol. 2011 Mar; 93(3): 313-340.). Although neurons are not hollow, they have definite morphological resemblance to tracheal terminal branches.

-line 204-208 "To test whether raised levels of Crb were responsible for the excessive apical membrane, as reported in other contexts (Pellikka et al., 2002; Schottenfeld-Roames et al., 2014; Wodarz et al., 1995), we knocked down Crb (Fig. S4G-H)." According to the legend for Fig s4, the authors express an RNAi construct against crbs. However, there does not appear to be any quantification of the amount knockdown of crb that was achieved. This is a concern for two reasons: 1) RNAi in the embryonic trachea works poorly for most genes (for unknown reasons) 2) This does not appear to be a clonal experiment but rather a pan tracheal driving of Crb RNAi. Loss of crbs would be expected to have very negative effects on tracheal morphogenesis (although this hasn't been rigorously tested to this reviewer's knowledge), but there doesn't appear to be any adverse effects of pan-tracheal crbs RNAi, suggesting that little if any knockdown of crb was actually achieved. The authors either need to document the reduction of crbs or remove this paragraph. Preferably, they would be able to document the reduction of crb because they are trying to address an important point and if they can show the apical expansion is crb-independent, that would be an nice result.

Minor concerns:

-Lines 250. "By this interpretation, unscissioned membrane invaginations protruding from the subcellular tube would occasionally have touched the basal plasma membrane or its protrusions and fused with it, as transcytosing vesicles would have done in the normal situation." I am not convinced by the argument that the bridging invaginations are fusing analogously to transcytosing vesicles because a protrusion/nascent vesicle coming from the apical surface should have rabs and V-SNAREs that should dock the protrusion/nascent vesicle with an endosomal compartment, not the basolateral surface. A transcytotic vesicle would have the rabs and V-SNAREs for the basolateral membrane. So it would seem that a fusion of the apical surface directly to the basolateral surface would have to be an ectopic event outside of the normal situation.

-Significant figures. This is not a big deal, but the authors are over reporting their significant figures. E.g. "a 7.05-fold increase (+/-2.98 SD)" . With an SD that is 50% the value of the measurement, reporting to hundreds is definitely beyond the accuracy of measurement. Rounding to tenths would be more appropriate.

Significance

As there have not been that many studies on the dynamics of membrane trafficking during morphogenesis, the results should be of broad interest to those studying the endocytic system and the role of membrane trafficking in morphogenesis. However, the paper would be greatly strengthened if the authors considered the recycling endosome in their analysis and write up. As a well-known compartment for trafficking cargo and membrane to both the apical and basolateral surface, it is hard to know how to interpret the observed trafficking without knowing the involvement, or lack thereof, of recycling endosomes in this system.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

In their manuscript, Mathew et al present a model in which transcytosis is utilized to deliver endocytosed apical membranes to supply basal membrane growth. The authors examined the developing terminal cell of the fruitfly tracheal system, which is a well established tubulogenesis model, as these cells form subcellular tubes by apical plasma membrane invagination. The authors show that basal membrane growth stops when endocytosis or endosomal transport is blocked, while the apical membrane grows excessively or membrane material accumulates in the cytosol, respectively.

Significance

The authors used high-end microscopy (including CLEM and electron tomography) to support their model and in my opinion, the quality and the quantity of the presented data are indeed adequate for this. The text is well written, the figures are of superb quality, and several cartoons help to understand the presented data/experiments. Therefore I highly recommend submitting the manuscript to a cell biology journal in its present form.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

In this work the authors investigate the dynamics of cell morphogenesis in a convenient in vivo system. They use the terminal cells of the embryonic tracheal system and comprehensively address how cell shape change (elongation in this case) takes place and how membrane is remodeled during the process. By combining different high-resolution techniques, i.e. in vivo imaging of terminal cells expressing different membrane markers and serial-section electron tomography, they describe the organelle organization (ER, Golgi, different types of vesicles) in the terminal cells during their elongation. They identify the presence of membrane structures/vesicles particularly abundant at the tip of the cell ahead of the growing tube. When they block endocytosis they find increased tube membrane and lack of basal membrane growth. In addition, in tomograms, they observe clear membrane defects, like invaginations that could even connect the tube and basal membranes. This correlates with the absence of vesicles at the tip observed in normal conditions. The analysis of the nature of the vesicles observed indicated the accumulation of late endosomes and MVB, particularly at the tip of the growing terminal cells. Interfering with the formation of these MVB led to defects in the growth of both the tube and the basal membrane. Altogether the authors propose a model in which the newly formed membrane (and transmembrane proteins) passes through the ER and Golgi and reaches the apical membrane. The incorporated membrane is then rapidly endocytosed and follows a maturation pathway through MVB, from where different cargoes and membrane would be sorted and recycled back to the apical (tube) membrane, or to the basal membrane through a transcytosis mechanism

Major comments:

- Are the key conclusions convincing?

Most of the conclusions presented are convincing and supported by the results observed. However, to my understanding, one of the key conclusions of this work (that membrane is transcytosed from the apical to the basal domain) is not fully convincing. A critical result to support the author's conclusion of apical-basal transcytosis is to find clear evidence of basal accumulation of a transcytosed marker. The authors show accumulation of FGFR-GFP and Myospheroid as evidence. However, I find the results presented not very convincing. The accumulation of FGFR-GFP at the basal membrane in the control is not very clear in the images and movie presented. In addition, in the shibire mutant, some basal accumulation of FGFR-GFP seems to be detected (particularly in the movie). In the figures the authors show an increase of FGFR-GFP intensity when endocytosis is blocked, but this is not explained in the text. On the other hand, and very importantly, how does FGFR localization relates to its activity? The authors show that when endocytosis is prevented dpERK (i.e. a reporter of FGFR activation) is not decreased, indicating that FGFR is normally active. Wouldn't this suggest that FGFR is still localized basally to receive Bnl signal? Actually, as the authors indicate, in larval tracheal cells the intracellular accumulation of the FGFR leads to a reduced FGF signal transduction (Chanut-Delalande et al., 2010), suggesting that reduced FGF signaling activity in these cells is due to less FGFR reaching the basal membrane. The results with Myospheroid are not very convincing either, as the authors just show a single confocal section of control and shibire mutants. In summary, I consider that this very important point needs to be better documented before concluding that apical membrane material containing basal cargoes is transcytosed to the basal membrane.

Another conclusion that, to my opinion, should be better explained and documented, is the coordination of tube and basal membrane growth. Following the movements of the vesicles the authors conclude that there is a net displacement of these vesicles to the tip of the cell. This correlates with the presence of mature endosomes there. So the results postulate that transcytosis occurs at the tip, and therefore the growth of the basal membrane would occur preferentially at the tip. It has been demonstrated, as the authors indicate, that the tube membrane grows all along the length of the tube. How is then coordinated the tube and basal membrane growth? If, as the authors propose in their model, the tube membrane also grows after a process of endocytosis and recycling, wouldn't it be expected to have preferential tip growth? How do the authors reconcile all these observations with previously published results (Gervais and Casanova, 2010)?

The serial-section electron tomography analysis is very interesting and identifies different sorts of vesicles. However, it is very unclear what the different vesicles referred in the models correspond to (in the in vivo imaging for instance). For instance, the small granular or the dense-core vesicles correspond to endocytic vesicles at different stages of maturation?. If there is a constant endocytosis from the apical membrane to generate the basal and build the definitive apical membrane, wouldn't it be expected to find many more vesicles around the tube? Wouldn't it be expected to find coated vesicles around or budding from the tube, as the coated vesicles observed budding from the basal membrane in Fig 2D? Or is the endocytosis observed mediated by non-clathrin coated vesicles?.

- Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

The results of Serp accumulation upon MVB interference can lead to confusion (from line 404). The authors seem to suggest that Serp protein is exclusively produced in the FB and transported by transcytosis to reach the tracheal lumen. However, Serp is also produced in the tracheal cells themselves. In fact, serp expression in the FB seems to be detected by late embryogenesis, while expression in tracheal cells is detected much earlier (Dong et al. 2014; Luschnig et al. 2006; Wang et al. 2006). It was also shown that Serp undergoes a recycling mechanism from the lumen to the lumen, through the endosomes-TGN retrograde trafficking, that may also require Shrub (Dong et al, 2014; Dong et al 2013). Thus, it is unclear (and even unlikely) that the Serp found in the vesicles in Shrub-GFP mutants is derived exclusively from the transcytosed component from the FB. I suggest to better explore this issue or to remove this part.

- Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

As indicated before, more conclusive results for transcytosis should be provided.

I suggest that the authors determine the presence of apical and basal cargoes (FGFR) in the late endosomes found when Shrub activity is impaired. According to his model, both types of markers should accumulate there. A high accumulation of these markers in those endosomes would reinforce the hypothesis proposed.

While it has been reported that shrub-GFP act as a dominant negative in different contexts (Dong et al, 2014; Sweeney et al 2006) it is unclear why. So it would be desirable to confirm the results with a loss of function condition (either mutant or RNAi line)

- Are the data and the methods presented in such a way that they can be reproduced?

The materials and methods section would benefit from more detailed explanations.

- Are the experiments adequately replicated and statistical analysis adequate?

Many of the experiments presented in this work are technically very challenging, like the in vivo analyses and particularly the serial-section electron tomography. This prevents having high numbers of replicates on occasions.

Minor comments:

-In the abstract the authors state: "We show that apical endocytosis and late endosome-mediated trafficking determine the membrane allocation to the apical and basal membrane domains". I think that the authors show that "that apical endocytosis and late endosome-mediated trafficking is required for correct membrane growth", but I am not that sure that they show that it determines the membrane allocation

-References for the PH-GFP localization in cells should be provided. Which is the evidence that it only localizes to plasma membrane?

-It would be more adequate to always use the same terms to facilitate the reading. For instance, in several figures the membranes are referred as basal plasma membrane and tube membrane, but in others outer and inner membrane

-Figure 3C,D and corresponding text are difficult to understand. The increase of fluorescence of the inner membrane seems to be very high, even higher than the corresponding to the outer membrane. Can the authors explain better this point and also describe better the method applied in the materials and methods section?

Significance

This work represents an important advance for the field for several reasons. First of all it represents a technical advance because the authors are able to combine the traditional genetic analysis with two powerful techniques (in vivo imaging and serial-section electron tomography) to analyze single cell behavior at high resolution (temporal and spatial). In addition it represents a conceptual advance as it proposes a mechanism through which membrane growth is coordinated to regulate cell morphogenesis. The mechanism presented (endocytosis and transcytosis) is not new but they find evidence in an in vivo system. It was previously known that tracheal terminal cells undergo a process of intracellular tube formation and cell elongation at the same time, but the mechanisms coordinating these two cell events were not known. The proposed mechanism may not only be relevant for the morphogenesis of tracheal terminal cells, but could represent a general mechanism of cell morphogenesis. Therefore, the paper should be relevant for research in the morphogenesis area but also in the cell biology field, as it shows how regulated membrane trafficking can control tissue morphogenesis

REFEREES CROSS-COMMENTING: I agree with reviewer #4 on her/his comments and suggestions about analyzing the involvement of the recycling endosome in the process.