1,938 Matching Annotations
  1. Last 7 days
    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      We are grateful for the constructive and highly supportive reviews provided by our Reviewers. We especially appreciate the efforts they have made to provide suggestions on how to make our revised manuscript even more robust. We have incorporated many of these suggestions into the revised manuscript that will post to Biorxiv and will be submitted to an affiliate journal. We have provided point-by-point responses to each Reviewer below each item (starting with Response: …), along with any changes made in response to that comment/suggestion (starting with In our revised manuscript, …).

      Finally, we agree with all Reviewers that this work should be of broad interest to the molecular biology, cell biology, and parasitology communities. Our discovery that Plasmodium and two related genera have taken the unorthodox approach of duplicating their NOT1 protein, and that Plasmodium has dedicated it for its unique transmission strategy, is a fascinating adaptation of the use of this core eukaryotic complex. We believe that those that focus on diverse aspects of RNA biology, including RNA preservation/decay, the maternal to zygotic transition, translational repression, and beyond will find this work to be of interest and relevant to their own research questions.

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

      The manuscript „The Plasmodium NOT1-G paralogue acts as an essential nexus for sexual stage maturation and parasite transmission" investigates the two forms of NOT1 in rodent malaria parasites. The authors found out that the original NOT1 is crucial for gametocyte induction as well as transmission to the mosquito, they therefore renamed it NOT1-G. The paralogous proteins, on the other hand, appears to be crucial for intraerythrocytic growth, since it cannot be knocked out. The authors then investigated NOT1-G in more detail, using standard phenotyping assays. They found a slightly increased gametocytemia and a minor effect on transmission to the mosquito.

      Response: In our submitted manuscript, we do focus on PyNOT1-G because of the exciting role it has for both sexes of gametocytes, which results in a complete defect in transmission to mosquitoes. Our investigations of what domains of PyNOT1-G focused on the most likely suspect: the putative tristetraprolin-binding domain (TTPbd). It was through deletion of this domain that we observed only a minor defect in the prevalence of infection of mosquitoes, indicating that the portion of PyNOT1-G that is required for transmission lies elsewhere (in part or in total). It is also important to correct Reviewer 1’s statement regarding the other (perhaps canonical) PyNOT1. To our surprise, PyNOT1 could be deleted, but resulted in a parasite that has an extreme fitness cost and a very slow growth phenotype. This is in stark contrast to other eukaryotes, where NOT1 is essential.

      Reviewer #1 (Significance (Required)):

      If the authors are able to provide convincing data that NOT1-G is indeed important for gametocyte induction and transmission to the mosquito, then the report would be of high significance for the malaria and molecular cell biology fields.

      Response**: We have in fact shown this and more in the originally submitted manuscript, and thus we are grateful that Reviewer 1 considers this work to be of high significance in a broad readership (molecular and cell biology, parasitology). In our revised manuscript, we have added text throughout to make these results even more apparent and clear for the reader.

      My expertise: molecular cell biology of gametocytes, translational regulation, parasite transmission

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

      **Summary**

      The manuscript by Hart et al. builds upon a fascinating finding presented in a previous manuscript by the same authors, in which they show that CCR4 seems to be able to associate with two members of the NOT1 family. In this work, the authors first re-annotate the two NOT1 paralogs in Plasmodium yoelii and then perform an in depth characterization of the role of NOT1-G during gametocytogenesis and early mosquito development. Using gene knockout and different genetic crosses, the authors show that NOT1-G is essential for male gametocyte development and leads to an arrest of development in zygotes arising from female gametocytes. Using RNA-seq the authors show that NOT1-G leads to lower transcript abundances, leading to the hypothesis that NOT1-G might be involved in preserving mRNAs in a larger RNA-binding complex. Lastly, the authors characterize a NOT1-G defining TPP domain and find that it is not essential for either male/female phenotype observed for the whole gene KO.

      Response**: We appreciate the concise and accurate summary of these findings.

      **Major comments:**

      • Are the key conclusions convincing?

        The phenotypic characterization of NOT1-G during gametocytogenesis / early mosquito development is nicely presented and the experiments are well performed. Because a duplication of NOT1 with possibly opposing roles of the paralogs is a very unique feature with broad implication on RNA metabolism, it would have been great to see two select experiments on the molecular level adding evidence that 1) NOT1/NOT1-G are mutually exclusive in a complex with CCR4/CAF1 and 2) NOT1-G acts post-transcriptionally in an antagonistic way to NOT1 (i.e. as a mRNA 'stabilizer' as proposed by the authors).

      Response**: We agree that inclusion of those two aspects would make for a more complete story about these two NOT1 paralogues.

      First, we also think that it is highly likely that NOT1 and NOT1-G are mutually exclusive, as in other eukaryotes NOT1 acts as a scaffold protein upon which effector proteins bind and bridging interactions are made. In our original manuscript, we did not include a mention of our previous attempts to address this question through colocalization and proteomic approaches, as they were largely unsuccessful. Specifically, we generated rabbit polyclonal antisera to PyNOT1-G’s tristetraprolin-binding domain but it did not pass our rigorous quality control (e.g. too much staining persisted in pynot1-g- parasites). Using both asexual and sexual blood stage parasites, we also attempted immunoprecipitation (with and without chemical crosslinking) and proximal labeling approaches via BioID and TurboID but all approaches did not produce rigorous results and thus we did not report them in our original manuscript. However, this question of whether the two NOT1 paralogues were mutually exclusive in complexes was also taken up by the Bozdech Laboratory in their 2020 preprint (Liu et al.) where they were able to capture the P. falciparum NOT1-G and NOT1 proteins (called Not1.1 and Not1.2 in that work). While their proteomic evidence showed that they could capture these bait proteins and that the NOT1 paralogues were not in the same complex, these results should be taken with a grain of salt: all mass spectrometry-based proteomic approaches are limited in that an absence of evidence does not mean that the protein is not present/interacting. Moreover, these efforts only identified a few other proteins that were already known to interact with the CAF1/CCR4/NOT complex, but even so, they did not use statistically rigorous methods in an attempt to quantify these results. In our revised our manuscript, we have included additional text to describe our unsuccessful efforts to do these capture proteomics experiments, and we have expanded our discussion of the Liu et al findings that provide some evidence in support of a mutually exclusive complex.

      Second, we also hypothesize that PyNOT1-G acts post-transcriptionally to affect mRNA abundance and translation. However, it is important to emphasize that NOT1 proteins typically act as scaffolds, with the recruited effector proteins acting to hasten the degradation and/or to preserve associated transcripts. We believe that studying these effector proteins is the next important effort to undertake. In fact, we hypothesized that these antagonistic effector proteins would be analogous to TTP and ELAV/HuR-family proteins as are found in other eukaryotes, and that the critical interaction with PyNOT1-G would be via its putative TTP-binding domain. It was for that reason that we interrogated the TTP-binding domain itself, and were surprised that its deletion did not phenocopy the complete gene deletion. Ongoing work will be focused on identifying these antagonistic effector proteins that likely are expressed in a stage-enriched manner, and to define how they interact with PyNOT1-G in order to direct specific mRNAs to their fates. Additionally, it would be very important and exciting to directly test if PyNOT1 and PyNOT1-G are functionally opposed. However, this would be exceptionally challenging to study from a technical standpoint. While we were able to delete the pynot1 gene after many repeated attempts, these parasites are very sickly and grow very slowly. Because of this, we believe that assessing direct versus indirect effects of PyNOT1 in these cells would not be feasible or robust. Given this, comparing functions between PyNOT1 and PyNOT1-G could not be done in a conclusive manner.** In our revised manuscript, we have expanded our descriptions of the mechanisms by which we believe PyNOT1-G and its complex affects mRNA fates. In particular, we have expanded our Discussion section to incorporate the results that indicate that the TTP-binding domain is not required for the essential functions of PyNOT1-G.

      • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

        The authors describe the role of NOT1-G as 'preserving' mRNA. The lower abundance of many transcripts in the NOT1-G knockout suggest this, but experimental proof is not provided (see suggestions below). Maybe rephrase to 'putatively preserved/stabilized' or 'has a potentially stabilizing function'. The same is true for the mutually exclusive association of the two paralogs with CCR4/CAF1. The authors refer to a protein co-IP of CCR4 showing that CCR4 can interact with both NOT1 and NOT1-G, but a reciprocal experiment is lacking.

      Response**: In our first publication on the deadenylase members of this complex, we also saw a similar effect on specific mRNAs when pyccr4-1 was deleted: the abundance of specific mRNAs went up in pyccr4-1- parasites. In that work and here in this manuscript, we have carefully decided to apply the word “preserved” to the fate of these mRNAs as it describes in a general way what is happening. In order to robustly state that mRNAs are stabilized by PyNOT1-G (directly or indirectly) would require additional experiments designed to test this (more description on this is provided on a response below). Second, as described above, we agree that doing a reciprocal IP for mass spectrometry-based proteomics would be ideal, we attempted four different approaches to do this to no avail. However, the composite proteomics data that is already available in the literature and via the Liu et al. preprint from the Bozdech Lab all indicate that these interactions occur, and perhaps that NOT1 and NOT1-G are mutually exclusive as expected. In our revised manuscript, we have provided further explanation in the Discussion for our use of the descriptor “preserve” instead of “stabilize”, and as noted above, and we have expanded our Discussion to more comprehensively define the interaction network depicted in Figure 7.

      In both cases, the conclusions of the authors are very likely (e.g. downregulation of many genes as seen by RNA-seq), but the final experimental evidence is not provided and a network such as in Figure 7 is not fully supported. If the authors would like to maintain these statements, then they should be rephrased and made clear or the additional experimental evidence suggested below is necessary.

      Response**: We hold that the published proteomic datasets do support such a network, with further support offered from the preliminary proteomic evidence from the Liu et al preprint. Therefore, we have not modified our manuscript beyond the additional text now provided in the Discussion as noted above.

      • 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.

        The essential claim that NOT1-G is important for gametocytogenesis and early mosquito development is well presented and fully supported by the experiments. As for the role of NOT1-G in 'preserving' mRNA, an mRNA half-life experiment would be necessary (or the text should be adjusted as mentioned above). In a short-term in vitro culture, pynot1-g- and WT parasites could be treated with ActD and abundances of select transcripts are measured by RT-qPCR.

      Response**: We appreciate that Reviewer 2 considers the rigor of our experiments to be high. Regarding the use of the term “preserve” vs “stabilize”, we agree that to shift from our more general descriptor (preserve) to one that has specific connotations (stabilize) would require additional experimentation. To correctly and most robustly make the claim of stabilization would require work on par with that done by Painter et al. (PMID: 29985403) that uses a thiol-containing nucleotide (4-TU) along with a yeast-derived fusion enzyme (yFCU) to convert it for use by Plasmodium. Previously we have shown that an associated deadenylase (PyCCR4-1) also acted to preserve mRNAs, and moreover that deletion of its gene resulted in no discernable effect upon the poly(A) tail or 3’ UTR of an mRNA that is bound by this complex (p28).

      While understanding mRNA stability is an exciting area of study, this 4-TU labeling experiment alone warranted a standalone, high impact publication for Painter et al. As this has not been adapted for any rodent-infectious Plasmodium species to date, and as adaptation of this labeling approach took several years for Dr. Painter while in the Llinas Laboratory (personal communication), we believe this work is beyond the scope of this study. Moreover, the additional information that it would provide to understand NOT1-g functions (preserve vs stabilize) would be incremental beyond the major storyline presented in this manuscript. In our revised manuscript, we have added text to ensure that our choice of “preserve” is well defined and explained.

      To support the idea that NOT-1 and NOT1-G associate in a mutually exclusive way or to just show that they act in distinct complexes despite their similar expression patterns, an IFA with a double stained NOT1/NOT-1G cell line could be performed. Alternatively, the authors could perform a protein co-IP using the already existing NOT1/NOT1-G-GFP cell line and show that the proteins don't interact with each other or even have certain distinct interaction partners.

      Response**: We agree, and these studies were attempted but were unsuccessful (described in our responses above). In our revised manuscript, we have included this information as noted above.

      • Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

        All necessary cell lines for a NOT1/NOT1-G co-IP and the ActD experiment are already present. The authors already present a ring to schizont in vitro culture (for ActD) and also have substantial experience in protein co-IP and proteomics.

        I am not sure about the cost for a proteomics experiment at the author's institute and I don't want to make a guess on time investment given the still on-going COVID situation.

      Response**: We agree that these experiments would be interesting, and would be costly to do at a transcriptome-wide scale and would require substantial time to conduct. We believe that the 4-TU approach noted above is the most rigorous, but is well beyond the scope of this study as it has not yet been adapted to rodent-infectious malaria parasites. As noted above, we have attempted four different proteomics approaches to provide reciprocal evidence for the complex composition which were unsuccessful. In our revised manuscript, we have added text to ensure that our choice of “preserve” is well defined and explained, and have noted the unsuccessful reciprocal proteomics approaches.

      • Are the data and the methods presented in such a way that they can be reproduced?

        The MM section is well structured and presented and the supplemental material includes all data.

      Response**: Thank you. We want to ensure that our work is clearly described and can be reproduced with the information reported.

      • Are the experiments adequately replicated and statistical analysis adequate?

        There is hardly any test of significance presented in the main text of the manuscript (e.g. Figure 3B and 4A). Please show the individual data points for these graphs and make sure the n= and the statistical test is described in the figure legend. If you use the term significant in the text, then just add the p-value behind it. This is also true for the RNA-seq data: Genes are sorted by fold-changes, leaving it unclear if these changes are significant. These data are however presented in Table S1 and could be incorporated in the main text.

      Response**: We agree. In our revised manuscript, we have incorporated additional details about the statistical tests used, p-values for noteworthy comparisons, and have included more panels for our comparative RNA-seq datasets (heatmap, PCA, MA plots). We have also made adjustments to our plots to make individual data points more readily observed, especially when error bars may block them (e.g. Figure 3B). And as in the original submission, all of the pertinent values, including fold changes, statistics and more are provided in our comprehensive supplementary files. We have structured the Supplementary Tables to flow from one tab to the next with the filtering/threshold applied noted both in the tab name and in the README tab that is found first among the tabs.

      **Minor comments:**

      • Specific experimental issues that are easily addressable.

        One idea that is also not discussed but could be added is for example that NOT1-G itself doesn't even have a stabilizing effect itself, but act as a decoy for other components of the CCR4/Caf1 complex, keeping them from associating with NOT1. In the NOT1-G knockout, the decrease in RNA abundance might then be just a result of an 'overactivity' of CCR4/Caf1/NOT1.

      Response**: This hypothesis proposed by Reviewer 2, that PyNOT1-G is acting as a decoy or a binding partner sponge, is certainly feasible. For this scenario to be effective, PyNOT1-G would need to be in excess of PyNOT1 and/or would need to be able to bind to the critical effector protein(s) better than does PyNOT1. However, our microscopy data, along with the transcriptomic data presented here and previously published proteomic data would indicate that these two gene products are in approximately balanced proportions and are similarly localized. This does not exclude the possibility that PyNOT1-G could act as a sponge for relevant binding partners. In our revised manuscript, we have raised this possibility as an alternate explanation for the phenotype in the Discussion section.

      • Are prior studies referenced appropriately?

        Throughout the manuscript, the authors should make clear what results come from which organism. Just as an example, the genome wide KO screens were performed in P. berghei and P. falciparum, CCR4/CAF1 experiments were performed in P. yoelii, whereas the original DDX6 work was done in P. berghei.

      Response**: We agree. In our revised manuscript, we have added additional text to further clarify what data comes from which Plasmodium species.

      • Are the text and figures clear and accurate?

        The Introduction is a bit long and partially turns into a minireview of eukaryotic RNA degradation. In the main text on page 13, the authors introduce a model for proteins involved in translational repression. This in not fully accurate, since for many of the proteins in this network, an effect on translation has actually not been shown. This includes NOT1-G characterized in the present work that most likely has an effect on mRNA stability, but for which a role in regulating translation is not presented.

      Response**: We believe the length and content of this Introduction is appropriate to provide the context that some readers outside of the parasitology field will need to appreciate these findings. Regarding designations for these proteins as being related to translational repression, we think that the ample proteomic evidence tying them to translationally repressive complexes warrants this. In our revised manuscript, we have made it more clear that these proteins themselves have not been directly implicated in translational repression.

      • Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

        Overall the RNA-seq is underrepresented and Figure 5 could easily be expanded by adding several panels that would help the future reader getting a better idea of the data:

      1. Summary graphs such as PCA/MDS plots of the different replicates and MA-plots (all of which can be easily generated in DESeq2)
      2. Heatmaps comparing the expression patterns of pynot1-g-, pbdozi-, pbcith-, pyalba4- highlighting some key gametocyte genes mentioned in the text
      3. Alternatively to 2., a simple Venn Diagram would already be very informative

        An informative representation might also be to sort the differentially expressed genes as predominant male and/or female. The P. berghei data by Yeoh et al (PMID: 28923023) could be a starting point.

      Response**: We agree. In our revised manuscript, we have expanded Figure 5 to include additional plots that speak the rigor of these datasets. Specifically, we have added a comprehensive heatmap and PCA plots, as well as MA plots as recommended. We have chosen not to include a Venn diagram for the overlap of affected mRNAs across these transgenic parasite lines, as we hold that this information is best provided in the text (high level observations) and the Supplement (details).

      Reviewer #2 (Significance (Required)):

      **Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.**

      Technically this manuscript builds on standard methods of the field that are well executed. There is no direct clinical advancement, although one might argue that a unique adaptation of the parasite could always be a novel therapeutic target. Conceptually this is great advancement for the parasitology field as it is, providing additional evidence for the importance of post-transcriptional regulation for parasite transmission. With the two experiments suggested above and the additional evidence gained from it, this manuscript could also gain great interest to readers outside the field by clearly showing how alternative ways to regulate RNA stability evolved.

      Response**: We are grateful for your careful review of our work and for the recommendations that you provided. We have incorporated many of them into the revised manuscript to make it even more rigorous and comprehensive. We also appreciate hearing that this work would be of great interest to a broader community. We feel that this is already the case, as the duplication of NOT1 and the dedication of one paralogue to an essential function is exciting and novel among eukaryotes.

      **Place the work in the context of the existing literature (provide references, where appropriate)**

      The work builds on the early reports of the particular RNA metabolism in gametocytes performed in the groups of Andy Waters. Since then, the authors themselves have published a great set of manuscripts extending our knowledge of the proteins involved in gametocytogenesis and nicely place the current work into this framework.

      Response**: We appreciate this positive feedback. This is a fascinating topic to study.

      **State what audience might be interested in and influenced by the reported findings.**

      The manuscript as it stands is particularly interesting for the parasitology and potentially the evolutionary biology field. For a broader readership for example in the RNA field, the possibly antagonistic roles and mutually exclusive association with CAF1/CCR4 are likely most interesting.

      Response**: We agree that this should be interesting to readers beyond our own field, as the duplication and specialization of NOT1, and the finding that the “canonical” PyNOT1 can be deleted, are both of general interest to how eukaryotes have adapted and deployed a highly conserved and essential RNA metabolic complex.

      **Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.**

      **Expertise:**

      RNA biology, Plasmodium falciparum, Bioinformatics

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

      In this manuscript, the authors investigate the requirement of two possible Not1 paralogs for the development of asexual blood stages and for the sexual transmission stages of Plasmodium yoelii. While Not1 is critical for asexual blood stages, its putative paralog, Not1G is important for the development of sexual transmission stages. In the absence of Not1G, male gametes are not formed while female gametes are formed and can be fertilised by wt male gametes. However, the resulting zygote cannot develop further into ookinete. The in vitro genetic cross assay to show this is elegant! A transcriptomic analysis further indicates that the transcriptomes of Not1G deficient parasites are significantly different from their WT counterpart.

      Response**: We are thrilled that you found our evidence and approaches to be rigorous and compelling. Thank you.

      **Major comments:**

      The discussion section is very nice and the authors describe well what is speculative and should be further confirmed by additional experiments. However, I did find this was not the case in the results section where the authors are proposing conclusions that are not supported by the results. I think the reading of this manuscript would be much more enjoyable if the authors only describe the results shown and move all the discussions to the dedicated section. Below are some examples. The data presented in this manuscript is not showing a nexus, this is a suggestion based on the results of other articles, the word should thus be removed from the title (and kept for a future review!). The last two sentences of the localisation section should be moved to the discussion because they do refer to results not shown in this manuscript. The last sentence of the second paragraph of the zygote development section should also be moved to the discussion. For the transcriptomic analysis there is also no formal comparison with transcriptomes of other previously analysed mutants: the results of the comparisons should either be shown or not discussed in the result section. Finally, the discussions mentioning interactors of the complex should be removed from the result section and moved to the discussion unless the results are formally analysed.

      Response**: We again thank you for the complement. In our original manuscript, we opted to provide some limited interpretations and context within the Results section in order to help guide readers along our train-of-thought and line-of-experimentation. While a more traditional split of keeping essentially all discussion and interpretation for the Discussion is a tried-and-true approach, we prefer this more narrative method and have opted to keep these short sections in the Results section.

      I would strongly suggest the author the better present and describe their transcriptomic results. There is only one volcano plot indicating the overall defect in mixed gametocytes in the main figure. Apart from this, the results are only described in the main text or in supplementary tables. It is therefore difficult to understand the subtilities of the analysis. For example, the authors frequently mention dysregulated genes, but without specifying whether it is up or down-regulated in the mutant. To address this issue, I would suggest the authors to better describe their results in the figures. They could show the GO term enrichment analysis they mention and show how they assign GO term or transcripts to male and female parasites. It would also be nice to discuss some of the results a bit more in details. For example, it is not surprising to see a reduction in transcripts that are under the control of AP2-O in retort-arrested ookinetes as the parasite do not reach this stage. It is thus highly speculative to specifically link this observation with ALBA4 without further detailed analysis. On the other hand, it is more surprising to see a decrease in ap2g transcripts, while the authors observe an increased gametocytaemia. Could the authors comment this observation? It may also be nice to better present the comparison between gametocytes and schizonts to possibly speculate on the early requirement of Not1G in committed schizonts.

      Response**: We (and Reviewer 2) agree. In our revised manuscript, we have expanded Figure 5 to include additional plots that speak the rigor of these datasets. Specifically, we have added a heatmap, and PCA and MA plots as recommended. We have chosen not to include a Venn diagrams for the overlap of affected mRNAs across these transgenic parasite lines for the reasons stated above in our response to Reviewer 2. Similarly, we have opted to keep the specifics of the GO Term analyses in the Supplement as we believe these should always be taken with a grain of salt (especially high level GO Terms, as many choose to report). Finally, we have expanded our discussion on our observation that pyapiap2-g transcript levels are lower in the pynot1-g- line, despite seeing a slight increase in gametocytemia.

      The conclusion regarding the similar localisation of Not1 and Not1G with other members of the CAF1/CCR4/NOT complex is not really convincing for two reasons. First, there is not colocalization shown and, second, the distribution is not very peculiar so it is difficult to draw any conclusion with this level of resolution. The presence of alpha-tubulin in the nucleus of male gametocytes is also very surprising as it is rather nucleus-excluded in both P. falciparum and P. berghei, could the authors comment this peculiar localisation?

      Response**: We agree and disagree here. First, we agree that no colocalization data is presented here to place NOT1-G within the limit of resolution of fluorescence microscopy. What we can (and do) state is that these proteins are all localized to cytosolic puncta, which matches what is observed for essentially all other studied eukaryotes. In further support of this, our published, quantitative proteomic data indicates that the bioinformatically predictable members of the CAF1/CCR4/NOT complex do associate as anticipated. In the same vein, the micrographs presented were not captured by confocal microscopy, and thus the apparent localization of alpha tubulin “in” the nucleus is most likely attributed to being above and/or below the nucleus. Taken together, we do feel that the combined evidence is convincing. As we have already made all of these points in the original manuscript, we have not adjusted the revised manuscript further.

      One of my major frustration when reading this manuscript was that the authors are not trying to discriminate between an early role of Not1G during gametocytogenesis or later in gametogenesis. The fact that the transcriptomes of gametocytes and schizonts seem to show similarities suggests that the phenotype observed during both male gametogenesis or ookinete development are probably linked to early knock-on defects during gametocytogenesis. Could the authors test whether male gametocytes replicate DNA or female activate translation? These are of course non-essential experiments as the authors are careful with their conclusions and mention possible defects during both gametocytogenesis or gametogenesis. Addressing this question may however add significant insights into the requirement for Not1G.

      Response**: We are sorry for the frustration. We wrote the manuscript so as to state what we feel we could robustly say, and where we are drawn to speculate, we made that speculation clear. As Reviewer 3 notes, we have not attempted to discriminate between functions that PyNOT1-G may be playing in different stages or substages of development because we do not believe the experiments allow that discrimination. While we could investigate finer and finer aspects of possible defects in both male and female gametocyte development, the most impactful take home messages remain the same. We continue to address questions related to translational repression and its release, and anticipate that PyNOT1-G will play a substantial and essential role in this. As Reviewer 3 noted, we have already discussed these possibilities in the original manuscript, and thus have not added anything further about this in our revised manuscript.

      **Minor comments:**

      Please use page and line numbering for your next submissions! Please describe what "bioinformatics" was used. I would show the nice localisation in oocyst and sporozoite in the main section. The conclusions drawn from the genetic cross seem to come from a single biological replicate, if this is the case please indicate it clearly.

      Response**: We apologize for these oversights. In our revised manuscript, we have provided page and line numbering, have expanded on what bioinformatic processes were done in the manuscript, and have made it more clear that the genetic crosses come from multiple biological replicates (biological triplicate for the transmission-based genetic cross, biological duplicate for the in vitro culture genetic cross). However, we have opted to retain the oocyst and sporozoite IFA data in the Supplement, as the rest of the story is focused on blood stage and early mosquito stage.

      Reviewer #3 (Significance (Required)):

      This manuscript highlights the requirement of a Not1 paralog in the transmission stages of a Plasmodium parasite. More specifically it describes a new player in the control of RNA biology during this process where our knowledge is scarce. It will be a valuable manuscript for molecular parasitologists interested in transmission or RNA biology.

      Response**: We agree and are grateful that our colleagues find this study to be a valuable addition in our efforts to understand how malaria parasites have adapted classic eukaryotic mechanisms to suit their purposes.

      Our expertise is largely in molecular and cellular parasitology.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this manuscript, the authors investigate the requirement of two possible Not1 paralogs for the development of asexual blood stages and for the sexual transmission stages of Plasmodium yoelii. While Not1 is critical for asexual blood stages, its putative paralog, Not1G is important for the development of sexual transmission stages. In the absence of Not1G, male gametes are not formed while female gametes are formed and can be fertilised by wt male gametes. However, the resulting zygote cannot develop further into ookinete. The in vitro genetic cross assay to show this is elegant! A transcriptomic analysis further indicates that the transcriptomes of Not1G deficient parasites are significantly different from their WT counterpart.

      Major comments:

      The discussion section is very nice and the authors describe well what is speculative and should be further confirmed by additional experiments. However, I did find this was not the case in the results section where the authors are proposing conclusions that are not supported by the results. I think the reading of this manuscript would be much more enjoyable if the authors only describe the results shown and move all the discussions to the dedicated section. Below are some examples. The data presented in this manuscript is not showing a nexus, this is a suggestion based on the results of other articles, the word should thus be removed from the title (and kept for a future review!). The last two sentences of the localisation section should be moved to the discussion because they do refer to results not shown in this manuscript. The last sentence of the second paragraph of the zygote development section should also be moved to the discussion. For the transcriptomic analysis there is also no formal comparison with transcriptomes of other previously analysed mutants: the results of the comparisons should either be shown or not discussed in the result section. Finally, the discussions mentioning interactors of the complex should be removed from the result section and moved to the discussion unless the results are formally analysed.

      I would strongly suggest the author the better present and describe their transcriptomic results. There is only one volcano plot indicating the overall defect in mixed gametocytes in the main figure. Apart from this, the results are only described in the main text or in supplementary tables. It is therefore difficult to understand the subtilities of the analysis. For example, the authors frequently mention dysregulated genes, but without specifying whether it is up or down-regulated in the mutant. To address this issue, I would suggest the authors to better describe their results in the figures. They could show the GO term enrichment analysis they mention and show how they assign GO term or transcripts to male and female parasites. It would also be nice to discuss some of the results a bit more in details. For example, it is not surprising to see a reduction in transcripts that are under the control of AP2-O in retort-arrested ookinetes as the parasite do not reach this stage. It is thus highly speculative to specifically link this observation with ALBA4 without further detailed analysis. On the other hand, it is more surprising to see a decrease in ap2g transcripts, while the authors observe an increased gametocytaemia. Could the authors comment this observation? It may also be nice to better present the comparison between gametocytes and schizonts to possibly speculate on the early requirement of Not1G in committed schizonts.

      The conclusion regarding the similar localisation of Not1 and Not1G with other members of the CAF1/CCR4/NOT complex is not really convincing for two reasons. First, there is not colocalization shown and, second, the distribution is not very peculiar so it is difficult to draw any conclusion with this level of resolution. The presence of alpha-tubulin in the nucleus of male gametocytes is also very surprising as it is rather nucleus-excluded in both P. falciparum and P. berghei, could the authors comment this peculiar localisation?

      One of my major frustration when reading this manuscript was that the authors are not trying to discriminate between an early role of Not1G during gametocytogenesis or later in gametogenesis. The fact that the transcriptomes of gametocytes and schizonts seem to show similarities suggests that the phenotype observed during both male gametogenesis or ookinete development are probably linked to early knock-on defects during gametocytogenesis. Could the authors test whether male gametocytes replicate DNA or female activate translation? These are of course non-essential experiments as the authors are careful with their conclusions and mention possible defects during both gametocytogenesis or gametogenesis. Addressing this question may however add significant insights into the requirement for Not1G.

      Minor comments:

      Please use page and line numbering for your next submissions! Please describe what "bioinformatics" was used. I would show the nice localisation in oocyst and sporozoite in the main section. The conclusions drawn from the genetic cross seem to come from a single biological replicate, if this is the case please indicate it clearly.

      Significance

      This manuscript highlights the requirement of a Not1 paralog in the transmission stages of a Plasmodium parasite. More specifically it describes a new player in the control of RNA biology during this process where our knowledge is scarce. It will be a valuable manuscript for molecular parasitologists interested in transmission or RNA biology.

      Our expertise is largely in molecular and cellular parasitology.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary

      The manuscript by Hart et al. builds upon a fascinating finding presented in a previous manuscript by the same authors, in which they show that CCR4 seems to be able to associate with two members of the NOT1 family. In this work, the authors first re-annotate the two NOT1 paralogs in Plasmodium yoelii and then perform an in depth characterization of the role of NOT1-G during gametocytogenesis and early mosquito development. Using gene knockout and different genetic crosses, the authors show that NOT1-G is essential for male gametocyte development and leads to an arrest of development in zygotes arising from female gametocytes. Using RNA-seq the authors show that NOT1-G leads to lower transcript abundances, leading to the hypothesis that NOT1-G might be involved in preserving mRNAs in a larger RNA-binding complex. Lastly, the authors characterize a NOT1-G defining TPP domain and find that it is not essential for either male/female phenotype observed for the whole gene KO.

      Major comments:

      • Are the key conclusions convincing?

      The phenotypic characterization of NOT1-G during gametocytogenesis / early mosquito development is nicely presented and the experiments are well performed. Because a duplication of NOT1 with possibly opposing roles of the paralogs is a very unique feature with broad implication on RNA metabolism, it would have been great to see two select experiments on the molecular level adding evidence that 1) NOT1/NOT1-G are mutually exclusive in a complex with CCR4/CAF1 and 2) NOT1-G acts post-transcriptionally in an antagonistic way to NOT1 (i.e. as a mRNA 'stabilizer' as proposed by the authors).

      • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

      The authors describe the role of NOT1-G as 'preserving' mRNA. The lower abundance of many transcripts in the NOT1-G knockout suggest this, but experimental proof is not provided (see suggestions below). Maybe rephrase to 'putatively preserved/stabilized' or 'has a potentially stabilizing function'. The same is true for the mutually exclusive association of the two paralogs with CCR4/CAF1. The authors refer to a protein co-IP of CCR4 showing that CCR4 can interact with both NOT1 and NOT1-G, but a reciprocal experiment is lacking.

      In both cases, the conclusions of the authors are very likely (e.g. downregulation of many genes as seen by RNA-seq), but the final experimental evidence is not provided and a network such as in Figure 7 is not fully supported. If the authors would like to maintain these statements, then they should be rephrased and made clear or the additional experimental evidence suggested below is necessary.

      • 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.

      The essential claim that NOT1-G is important for gametocytogenesis and early mosquito development is well presented and fully supported by the experiments. As for the role of NOT1-G in 'preserving' mRNA, an mRNA half-life experiment would be necessary (or the text should be adjusted as mentioned above). In a short-term in vitro culture, pynot1-g- and WT parasites could be treated with ActD and abundances of select transcripts are measured by RT-qPCR.

      To support the idea that NOT-1 and NOT1-G associate in a mutually exclusive way or to just show that they act in distinct complexes despite their similar expression patterns, an IFA with a double stained NOT1/NOT-1G cell line could be performed. Alternatively, the authors could perform a protein co-IP using the already existing NOT1/NOT1-G-GFP cell line and show that the proteins don't interact with each other or even have certain distinct interaction partners.

      • Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

      All necessary cell lines for a NOT1/NOT1-G co-IP and the ActD experiment are already present. The authors already present a ring to schizont in vitro culture (for ActD) and also have substantial experience in protein co-IP and proteomics.

      I am not sure about the cost for a proteomics experiment at the author's institute and I don't want to make a guess on time investment given the still on-going COVID situation.

      • Are the data and the methods presented in such a way that they can be reproduced?

      The MM section is well structured and presented and the supplemental material includes all data.

      • Are the experiments adequately replicated and statistical analysis adequate?

      There is hardly any test of significance presented in the main text of the manuscript (e.g. Figure 3B and 4A). Please show the individual data points for these graphs and make sure the n= and the statistical test is described in the figure legend. If you use the term significant in the text, then just add the p-value behind it. This is also true for the RNA-seq data: Genes are sorted by fold-changes, leaving it unclear if these changes are significant. These data are however presented in Table S1 and could be incorporated in the main text.

      Minor comments:

      • Specific experimental issues that are easily addressable.

      One idea that is also not discussed but could be added is for example that NOT1-G itself doesn't even have a stabilizing effect itself, but act as a decoy for other components of the CCR4/Caf1 complex, keeping them from associating with NOT1. In the NOT1-G knockout, the decrease in RNA abundance might then be just a result of an 'overactivity' of CCR4/Caf1/NOT1.

      • Are prior studies referenced appropriately?

      Throughout the manuscript, the authors should make clear what results come from which organism. Just as an example, the genome wide KO screens were performed in P. berghei and P. falciparum, CCR4/CAF1 experiments were performed in P. yoelii, whereas the original DDX6 work was done in P. berghei.

      • Are the text and figures clear and accurate?

      The Introduction is a bit long and partially turns into a minireview of eukaryotic RNA degradation. In the main text on page 13, the authors introduce a model for proteins involved in translational repression. This in not fully accurate, since for many of the proteins in this network, an effect on translation has actually not been shown. This includes NOT1-G characterized in the present work that most likely has an effect on mRNA stability, but for which a role in regulating translation is not presented.

      • Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

      Overall the RNA-seq is underrepresented and Figure 5 could easily be expanded by adding several panels that would help the future reader getting a better idea of the data:

      1. Summary graphs such as PCA/MDS plots of the different replicates and MA-plots (all of which can be easily generated in DESeq2)
      2. Heatmaps comparing the expression patterns of pynot1-g-, pbdozi-, pbcith-, pyalba4- highlighting some key gametocyte genes mentioned in the text
      3. Alternatively to 2., a simple Venn Diagram would already be very informative

      An informative representation might also be to sort the differentially expressed genes as predominant male and/or female. The P. berghei data by Yeoh et al (PMID: 28923023) could be a starting point.

      Significance

      Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

      Technically this manuscript builds on standard methods of the field that are well executed. There is no direct clinical advancement, although one might argue that a unique adaptation of the parasite could always be a novel therapeutic target. Conceptually this is great advancement for the parasitology field as it is, providing additional evidence for the importance of post-transcriptional regulation for parasite transmission. With the two experiments suggested above and the additional evidence gained from it, this manuscript could also gain great interest to readers outside the field by clearly showing how alternative ways to regulate RNA stability evolved.

      Place the work in the context of the existing literature (provide references, where appropriate)

      The work builds on the early reports of the particular RNA metabolism in gametocytes performed in the groups of Andy Waters. Since then, the authors themselves have published a great set of manuscripts extending our knowledge of the proteins involved in gametocytogenesis and nicely place the current work into this framework.

      State what audience might be interested in and influenced by the reported findings.

      The manuscript as it stands is particularly interesting for the parasitology and potentially the evolutionary biology field. For a broader readership for example in the RNA field, the possibly antagonistic roles and mutually exclusive association with CAF1/CCR4 are likely most interesting.

      Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

      Expertise:

      RNA biology, Plasmodium falciparum, Bioinformatics

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      The manuscript „The Plasmodium NOT1-G paralogue acts as an essential nexus for sexual stage maturation and parasite transmission" investigates the two forms of NOT1 in rodent malaria parasites. The authors found out that the original NOT1 is crucial for gametocyte induction as well as transmission to the mosquito, they therefore renamed it NOT1-G. The paralogous proteins, on the other hand, appears to be crucial for intraerythrocytic growth, since it cannot be knocked out. The authors then investigated NOT1-G in more detail, using standard phenotyping assays. They found a slightly increased gametocytemia and a minor effect on transmission to the mosquito.

      Significance

      If the authors are able to provide convincing data that NOT1-G is indeed important for gametocyte induction and transmission to the mosquito, then the report would be of high significance for the malaria and molecular cell biology fields.

      My expertise: molecular cell biology of gametocytes, translational regulation, parasite transmission

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

      Learn more at Review Commons


      Reply to the reviewers

      We thank all four reviewers for their positive and constructive comments! We have carefully considered these comments and provided a point-by-point response below.

      Reviewer #1 (Evidence, reproducibility and clarity):

      This paper explores an interesting problem of SHP1/SHP2 preferences of inhibitory immunoreceptors. The author are quick to point out that many of their individual data points confirm published results at some level, but the power of the paper is in the parallel analysis of both PD1, which is strongly biased towards SHP2 and BTLA, which is biased towards SHP1. This gives them the opportunity to test the predictions of descriptive experiment by making simple mutated receptors with swapped ITIM or ITSM domains.

      The work is very well done and generally the authors are quite careful and precise about the language used to describe results, in general.

      The results are quite striking in that the find plenty of evidence for transient interaction of SHP1 with PD1 based on the biophysical measurements, but don't detect the interactions in pull down or in "in cell" microcluster recruitment experiments. In describing the pull-downs they discuss the issue of dissociation during washing potentially missing interactions that are taking place. I would prefer that the pull down is fine evidence for binding, but lack of pull down is not evidence for lack of binding. They should double check that this language is consistent. Also, unless something has changed in the microcluster binding experiments, this in situ recruitment of SHP2 to PD1 is only observed or a 2-3 minutes and then can't be detected, the situation for SHP2 becoming the same as it is for SHP1. If the kinetics are different in the cleaner systems that have now developed they should show this in a primary figure as this would be then different when what is reported previously.

      We agree with the reviewer that pull down is evidence for binding. Indeed, in most, if not all of our assays, our results with pull down were consistent with those in the microcluster imaging. As suggested by the reviewer, we will check through the manuscript and ensure the language is accurate and consistent. In our recent study (Xu et al., JCB, 2020, PMID: 32437509), we conducted a side-by-side comparison of SHP2 and SHP1 recruitment kinetics to PD-1 in a similar system as the current study. Both microcluster imaging and co-IP assays showed that PD-1:SHP2 association lasted at least 10 minutes, whereas PD-1:SHP1 recruitment was nearly undetectable. The duration of PD- 1:SHP2 association was in good agreement with Takashi Saito’s finding in CD4+ mouse T cells (Yokosuka et al., JEM, 2012, PMID: 22641383). Regardless the somewhat different kinetics in different studies, SHP2 recruitment was transient, as pointed out by the reviewer. We believe that some other effectors contribute to PD-1 inhibitory signaling. In supportive of this notion, we recently found that PD-1 remains partially inhibitory in CD8+ T cells deficient in both SHP1 and SHP2 (Xu et al., JCB, 2020).

      The gap in this study is lack of any functional analysis. The Jurkat model could be quite useful as they have a relatively clean system for asking if the transient binding of SHP1 to PD1 has any functional impact, which they have not yet followed through on. Does PD-1 recruited SHP2 have any impact on function after the 5 minutes? Furthermore, the authors need to keep in mind that mice deficient in SHP2 respond to anti-PD1 checkpoint therapies (Rota, G., Niogret, C., Dang, A. T., Barros, C. R., Fonta, N. P., Alfei, F., Morgado, L., Zehn, D., Birchmeier, W., Vivier, E., & Guarda, G. (2018). Shp-2 Is Dispensable for Establishing T Cell Exhaustion and for PD-1 Signaling In Vivo. Cell Rep, 23(1), 39-49. https://doi.org/10.1016/j.celrep.2018.03.026). This is an important issue to discuss in light the the very interesting binding analysis the authors have performed. But I think the functional analysis can be part of a future paper.

      In our recent publication (Xu et al. JCB, 2020, PMID: 32437509), we found that deletion of SHP1 from Jurkat cells had little, if any effect on PD-1 mediated suppression of IL-2 production. As the reviewer alluded to, we did observe SHP2 dissociation from PD-1 after 10 minutes, so the question of whether and how PD-1:SHP2 complex influence T cell function in a longer term is a great one. We currently are pursuing a hypothesis that there is a SHP2-independent mechanism of PD-1 inhibitory function, and indeed, in our recent study (Xu et al. JCB, 2020, PMID: 32437509), we found that PD-1 retains its partial inhibitory function in SHP1/SHP2 double knockout murine primary T cells. These results are consistent with the in vivo data by Rota et al. cited by the reviewer. We will also briefly discuss this point in a revised manuscript.

      I would suggest that the title be modified slightly from "SHP1/SHP2 discrimination" to "differential SHP1/SHP2 interaction" and leave discussion of discrimination until they have the functional data integrated over times that are relevant to T cell transcriptional regulation (1-2 hrs). The functional analysis can be in another paper, but it would be interesting to have a paragraph in the discussion raising the outstanding issues beyond stable binding detected by the pull-down and microcluster recruitment experiments- what are the implications for function. Could the transient interactions in the noise of the steady state and equilibrium measurements be functional?

      We thank the reviewer for the suggestion, even though reviewer #3 felt that our current title is appropriate. We will be happy to change the title at the editors’ discretion.

      I would summarise that the work is outstanding as biochemistry and biophysics and it should be published nearly as is. I'm suggesting minor revisions in that the changes are just to text, but I think this is important and somewhat nuanced aspect of the paper that will make it even more helpful to readers.

      We appreciate the positive and insightful comments!

      Reviewer #1 (Significance):

      The authors generate a detailed descriptive data set about the component interaction of SHP1 and SHP2 SH2 domains with PD1 and BTLA intracellular domains. They then test hypotheses generated from the descriptive data set to better define the nature of the interactions and why PD1 recruits primarily SHP2, while BTLA mainly recruits SHP1. PD1 is a major driver or the cancer immunotherapy revolution and SHP2 is the major candidate for a signalling effector of PD1. This paper can become the reference paper for the specificity and engineering of this interaction, which will make it highly significant in a very active and still expanding field.

      Referee Cross-commenting

      I still feel that "discrimination" has a functional/activity connotation that is not addressed at all in this paper, but can be addressed. I'm happy to have the suggestion stand and let the authors decide. They need to live with it once its published. Another suggestion- the citations on regulation are mostly old. A good recent paper is Pádua, R. A. P., Sun, Y., Marko, I., Pitsawong, W., Stiller, J. B., Otten, R., & Kern, D. (2018). Mechanism of activating mutations and allosteric drug inhibition of the phosphatase SHP2. Nature Communications, 9(1),

      1. https://doi.org/10.1038/s41467-018-06814-w .

      We believe that some of the functional questions raised by this reviewer, including the SHP1 and SHP2 contribution in PD-1 signaling, was addressed in our recent publication (Xu et al., JCB, 2020). Using SHP1 KO and SHP2 KO T cells, we showed that PD-1 inhibitory function is contributed by SHP2, but very little if any by SHP1. Thus in the current study, we focus on the mechanism behind the striking SHP2 preference by PD-1. We thank this reviewer for suggesting this excellent reference. We will cite this reference in the revised manuscript.

      Reviewer #2 (Evidence, reproducibility and clarity):

      In this study, Xu and co-workers investigate the biophysical nature of the interaction between the structurally-related non-transmembrane PTPs Shp1 and Shp2 with the ITIM/ITSM-containing inhibitory receptors PD-1 and BTLA using cell-based, biochemical, biophysical and domain swapping assays. The primary aim being to better understand how these receptors discriminate between binding Shp1 and/or Shp2, and the orientation of Shp1 and Shp2 engagement. These are major unresolved questions in the field that the authors go some way to addressing in a methodical, rigorous, clear and concise manner. Findings are convincing, correlate well with previous findings and internally, and are complemented with excellent schematics, making it easy to comprehend.

      Major comments

      The authors focus primarily on binding affinities to explain differential binding of Shp1 and Shp2 by PD-1 and BTLA ITIMs and ITSMs, but this is only part of the story. Avidity, compartmentalization, stoichiometry of kinases, and relative abundance of Shp1 and Shp2 are also important aspects of the discriminatory mechanism that are not addressed. Competition assays would go some way to addressing the latter point and should be at least be considered and discussed.

      We agree that various parameters mentioned by this reviewers, such as compartmentalization and relative expression levels would be a concern for purely cell-free assays such as SPR, however, we feel that our cell-based assays already integrate these parameters. This is also precisely the reason why we chose to examine the recruitments of Shp1/2 in a cellular context instead of a purely cell-free system.

      Regarding the competition, we have confirmed our key results in both WT and SHP2 KO background, with or without the potential competition from endogenous SHP2, suggesting that competition might not be a dominant mechanism for the recruitment specificity we observed.

      Similarly, authors do not address how distortion of the pY binding pocket of Shp1 and Shp2 nSH2 domains in the auto-inhibited conformation is released, allowing the domain to engage with phopho-ITIM/ITSM. Again, this should be at least discussed. Current binding studies do not address this issue.

      We feel that the overall recruitment to the PD-1 microclusters as we observed in cells already integrate this auto-inhibition mechanism of Shp1 and Shp2, because we used full length proteins. We do agree with the reviewer that future studies are warranted to address the contributions of each mechanism, including auto-inhibition, concentration, competition, etc., to the overall recruitment. This might require careful and extensive biophysical analyses coupled with mathematical modeling.

      Minor comments:

      Phosphorylation should be indicated in schematic representations in Figures 3, 6 b, c.

      We thank the reviewer for this advice, we will indicate phosphorylation in the revised figure 3.

      Cellular and physiological significance should be further discussed, as well as broader implications of findings to other ITIM/ITSM-containing receptors in other lineages.

      We will further discuss this as suggested.

      Reviewer #2 (Significance)

      Findings from this study advance our knowledge of how inhibitory checkpoint regulatory receptors discriminate between Shp1 and Shp2, which has important implications for understanding how the unique biochemical, cellular and physiological functions of these receptors and phosphatases are dictated. Indeed, findings lay the foundation for a universal mechanism, that may apply to all ITIM/ITSM receptors in other cell lineages, and perhaps novel ways of targeting these interactions therapeutically.

      Compare to existing published knowledge

      Although largely correlative with previous studies, findings from this study start to fill major gaps in our knowledge of these biochemical processes, in a highly rigorous, concise and clear manner. Findings from previous studies were more 'piecemeal', whereas this study consolidates and advances important nuances of these interactions. Moreover, it lays the foundation for further structural, physiological and therapeutic studies.

      Audience

      The immune receptor signaling community and beyond, including any lineage in which ITIM/ITSM-containing receptors play a major role in regulating cellular responses.

      Your expertise

      ITIM/ITSM-containing receptors, kinase-phosphatase molecular switches, cellular reactivity to extracellular matrix proteins

      Referee Cross-commenting

      Generally agree with reviewer's comments. Constructive overall and fair. Although I was thinking additional competition experiments, I do not think necessary. Over the top for this study. Hence, 1 month should suffice to revise accordingly.

      We thank this reviewer for the excellent comments and understanding!

      Reviewer #3 (Evidence, reproducibility and clarity):

      Summary:

      Inhibitory immune receptors containing ITIMs function through recruiting the phosphatases SHP-1 and SHP-2. SHP-1 and SHP-2 are remarkably similar yet have different roles in vivo. How can ITIM-containing immune receptors specifically recruit SHP-1 or SHP-2? In this paper, Xu et al ask how SHP-1 vs SHP-2 specificity is achieved. They use very thorough biochemical assays to measure the affinity of SHP-1 and SHP-2 for various ITIM/ITSMs and finally pin point some key amino acids that switch an ITIM/ITSM from SHP-2 to SHP-1 specificity. The in vitro biochemical assays are augmented by in cell assays that support their conclusions. Overall, this paper is an incredibly elegant and straight forward paper addressing how SHP-1/SHP-2 specificity is achieved.

      Major Comments: none

      Minor Comments:

      • Could the western blots in Figure 1 be quantified as the western blots in other figures?

      We will quantify the western blots in Figure 1 as suggested in the revised manuscript.

      • The data that the y+1 reside is essential for SHP-1/2 specificity is very convincing. We are curious if the other residues of the ITIM/ITSM also contribute to this specificity, albeit less potently. The PD-1 G224A mutant is still less potent than the PD-1 BTLA ITIM swap, suggesting that while the y+1 position is most important, the other residues contribute some specificity. The authors also included data on a PD-1 variant with the BTLA ITIM A224G mutation (8f), which is slightly better at recruiting SHP-1 than the PD-1 ITIM. It may be worth mentioning this data in the text of the paper as well as displaying it in the figure.

      The reviewer raised an excellent point, yes, our data does suggest that other pY-flanking residues within the ITIM also contribute to SHP1 binding. However, the pY+1 residue replacement produced the strongest effect as the reviewer noted. In the revised manuscript, we will acknowledge the potential contributions of other residues.

      • A brief introduction to ITIM vs ITSM in the introduction of the paper may be helpful background for readers. For example, ITIM receptors are reasonably well known but how ITSM functionally differs is probably less well known.

      We will rewrite the introduction about ITIM and ITSM for better clarity.

      • Although not the major focus of the paper, broadening out this SHP-1/2 specificity to other immune receptors in the discussion is fascinating. (a) The authors find that a Valine, Leucine, or Isoleucine in place of the Alanine in y+1 is very close to equivalent, yet the A is highly conserved. The authors speculate that there may be an advantage to sub-maximal SHP-1 affinity because it is more easy to regulate. I think this is reasonable speculation but a little unsatisfying given the very small observed difference in SHP-1 binding. If the authors have additional thoughts, I would be interested to hear them. (b) The authors note that PD-1 is the only ITIM with a glycine in the Y+1 position. Are there other receptors that function primarily through SHP-2, and how might they achieve this specificity?

      Response to a: Even though valine, leucine or isoleucine did not produce a striking enhancement in Shp1 recruitment over alanine, the differences were statistically significant. In fact, when we performed these point mutations at a BTLA ITIM background, valine, leucine or isoleucine markedly enhanced the SHP1 recruitment (see unpublished data below). We speculate that other pY-flanking residues in BTLA, as this reviewer alluded to above, creates an environment that amplifies the differences. The strong sensitivity on pY+1 residue, as observed in BTLA, might be true for other SHP1-recruiting receptors too. If they were to have leucine or isoleucine at the pY+1 position of ITIM, they may recruit too much SHP1 that presumably decreases the fitness/growth of the cells. We propose to show this unpublished data as a supplemental figure in the revised manuscript. We will also discuss the potential contributions of other pY-flanking residues as this reviewer suggested.

      {{images cannot be rendered at this time in reply letters}}

      Response to b: Among the several receptors that we tested, PD-1 is the only receptor that exhibited no recruitment of SHP1. The lack of SHP1 recruitment is also true for murine PD-1, which has a glutamate residue (charged) at Y+1 position. In addition, earlier work reported that PECAM1 also selectively recruits SHP2, but not SHP1. We have noted that PECAM1 contain a threonine (polar) at the pY+1 position of their ITIMs. Thus, their inability to recruit SHP1 is consistent with our model that a nonpolar residue at Y+1 position is required for strong SHP1 recruitment. We will discuss these points in the revised manuscript.

      • Figure 9 b Val not Vla, Figure 3a - a legend for the color code may be nice (ie, 20-1000 nM) Thanks for catching this, we will fix the error in Figure 9b and provide the color code in Figure 3a in the revised manuscript.

      Reviewer #3 (Significance):

      Significance:

      SHP-1 and SHP-2 play a critical role in regulating immune system function. In addition, the receptors recruiting these phosphatases (like PD-1) are important immunotherapy targets. Previously, the question of SHP-1/SHP-2 specificity has been primarily described for ITIM bearing receptors individually. Other studies have predicted consensus sequences for the tSH2 domains of SHP-1 or SHP-2, but not addressed the defining molecular characteristics of these consensus sites or how these could be combined on ITIM receptors to generate selectivity between these related phosphatases. This paper represents a significant step forward because it provides a unifying mechanism explaining how ITIM-bearing immune receptors specifically recruit SHP-1 or SHP-2. I expect this paper will be broadly interesting to biochemists, immunologists and cancer biologists.

      Referee Cross-commenting

      I generally think the other reviewers comments are reasonable and insightful. Together, they suggest no new experiments are necessary. As for the proposed title change, I prefer the authors title and find it to be justified given their data.

      Reviewer #4 (Evidence, reproducibility and clarity):

      In this manuscript, Xu and college performed an elaborate study to investigate the molecular basis of Shp1 and Shp2 discrimination by immune checkpoints PD-1 and BTLA. The paper is original, clear, and well written. I only have a few minor comments:

      1. Please label the molecular weights to all the western blots/IPs results.

      We will label the molecular weights to all the blots in the revised manuscript.

      1. Please add scale bars to all the microscopy pictures.

      We will add scale bars to all the microcopy images in the revised manuscript.

      1. For the SPR data, please add the fitting curves.

      We thank the reviewer for the suggestion. However, we did not use the fitting curve to calculate the Kd, we plotted the maximum response as a function of concentration to determine the Kd. This is another well accepted method for Kd calculation. In fact, some of the SPR curves fit poorly with the existing algorithm. Thus, showing the fitting curve might distract the readers.

      Reviewer #4 (Significance):

      The strength of this paper relies on the details they dissected by using a series of mutagenesis screening experiments, which should be interesting to cell biologists and cancer immunologists.

      Referee Cross-commenting

      I think the other reviewer's comments are insightful and constructive, the suggested experiments are necessary and will improve the paper.

      We thank this reviewer for the positive comments!

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

      Learn more at Review Commons


      Referee #4

      Evidence, reproducibility and clarity

      In this manuscript, Xu and college performed an elaborate study to investigate the molecular basis of Shp1 and Shp2 discrimination by immune checkpoints PD-1 and BTLA. The paper is original, clear, and well written. I only have a few minor comments:

      1. Please label the molecular weights to all the western blots/IPs results.
      2. Please add scale bars to all the microscopy pictures.
      3. For the SPR data, please add the fitting curves.

      Significance

      The strength of this paper relies on the details they dissected by using a series of mutagenesis screening experiments, which should be interesting to cell biologists and cancer immunologists.

      Referee Cross-commenting

      I think the other reviewer's comments are insightful and constructive, the suggested experiments are necessary and will improve the paper.

    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:

      Inhibitory immune receptors containing ITIMs function through recruiting the phosphatases SHP-1 and SHP-2. SHP-1 and SHP-2 are remarkably similar yet have different roles in vivo. How can ITIM-containing immune receptors specifically recruit SHP-1 or SHP-2? In this paper, Xu et al ask how SHP-1 vs SHP-2 specificity is achieved. They use very thorough biochemical assays to measure the affinity of SHP-1 and SHP-2 for various ITIM/ITSMs and finally pin point some key amino acids that switch an ITIM/ITSM from SHP-2 to SHP-1 specificity. The in vitro biochemical assays are augmented by in cell assays that support their conclusions. Overall, this paper is an incredibly elegant and straight forward paper addressing how SHP-1/SHP-2 specificity is achieved.

      Major Comments:

      none

      Minor Comments:

      • Could the western blots in Figure 1 be quantified as the western blots in other figures?
      • The data that the y+1 reside is essential for SHP-1/2 specificity is very convincing. We are curious if the other residues of the ITIM/ITSM also contribute to this specificity, albeit less potently. The PD-1 G224A mutant is still less potent than the PD-1 BTLA ITIM swap, suggesting that while the y+1 position is most important, the other residues contribute some specificity. The authors also included data on a PD-1 variant with the BTLA ITIM A224G mutation (8f), which is slightly better at recruiting SHP-1 than the PD-1 ITIM. It may be worth mentioning this data in the text of the paper as well as displaying it in the figure.
      • A brief introduction to ITIM vs ITSM in the introduction of the paper may be helpful background for readers. For example, ITIM receptors are reasonably well known but how ITSM functionally differs is probably less well known.
      • Although not the major focus of the paper, broadening out this SHP-1/2 specificity to other immune receptors in the discussion is fascinating. (a) The authors find that a Valine, Leucine, or Isoleucine in place of the Alanine in y+1 is very close to equivalent, yet the A is highly conserved. The authors speculate that there may be an advantage to sub-maximal SHP-1 affinity because it is more easy to regulate. I think this is reasonable speculation but a little unsatisfying given the very small observed difference in SHP-1 binding. If the authors have additional thoughts, I would be interested to hear them. (b) The authors note that PD-1 is the only ITIM with a glycine in the Y+1 position. Are there other receptors that function primarily through SHP-2, and how might they achieve this specificity?
      • Figure 9 b Val not Vla, Figure 3a - a legend for the color code may be nice (ie, 20-1000 nM)

      Significance

      SHP-1 and SHP-2 play a critical role in regulating immune system function. In addition, the receptors recruiting these phosphatases (like PD-1) are important immunotherapy targets. Previously, the question of SHP-1/SHP-2 specificity has been primarily described for ITIM bearing receptors individually. Other studies have predicted consensus sequences for the tSH2 domains of SHP-1 or SHP-2, but not addressed the defining molecular characteristics of these consensus sites or how these could be combined on ITIM receptors to generate selectivity between these related phosphatases. This paper represents a significant step forward because it provides a unifying mechanism explaining how ITIM-bearing immune receptors specifically recruit SHP-1 or SHP-2. I expect this paper will be broadly interesting to biochemists, immunologists and cancer biologists.

      Referee Cross-commenting

      I generally think the other reviewers comments are reasonable and insightful. Together, they suggest no new experiments are necessary. As for the proposed title change, I prefer the authors title and find it to be justified given their data.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this study, Xu and co-workers investigate the biophysical nature of the interaction between the structurally-related non-transmembrane PTPs Shp1 and Shp2 with the ITIM/ITSM-containing inhibitory receptors PD-1 and BTLA using cell-based, biochemical, biophysical and domain swapping assays. The primary aim being to better understand how these receptors discriminate between binding Shp1 and/or Shp2, and the orientation of Shp1 and Shp2 engagement. These are major unresolved questions in the field that the authors go some way to addressing in a methodical, rigorous, clear and concise manner. Findings are convincing, correlate well with previous findings and internally, and are complemented with excellent schematics, making it easy to comprehend.

      Major comments

      The authors focus primarily on binding affinities to explain differential binding of Shp1 and Shp2 by PD-1 and BTLA ITIMs and ITSMs, but this is only part of the story. Avidity, compartmentalization, stoichiometry of kinases, and relative abundance of Shp1 and Shp2 are also important aspects of the discriminatory mechanism that are not addressed. Competition assays would go some way to addressing the latter point and should be at least be considered and discussed.

      Similarly, authors do not address how distortion of the pY binding pocket of Shp1 and Shp2 nSH2 domains in the auto-inhibited conformation is released, allowing the domain to engage with phopho-ITIM/ITSM. Again, this should be at least discussed. Current binding studies do not address this issue.

      Minor comments:

      Phosphorylation should be indicated in schematic representations in Figures 3, 6 b, c.

      Cellular and physiological significance should be further discussed, as well as broader implications of findings to other ITIM/ITSM-containing receptors in other lineages.

      Significance

      Findings from this study advance our knowledge of how inhibitory checkpoint regulatory receptors discriminate between Shp1 and Shp2, which has important implications for understanding how the unique biochemical, cellular and physiological functions of these receptors and phosphatases are dictated. Indeed, findings lay the foundation for a universal mechanism, that may apply to all ITIM/ITSM receptors in other cell lineages, and perhaps novel ways of targeting these interactions therapeutically.

      Compare to existing published knowledge

      Although largely correlative with previous studies, findings from this study start to fill major gaps in our knowledge of these biochemical processes, in a highly rigorous, concise and clear manner. Findings from previous studies were more 'piecemeal', whereas this study consolidates and advances important nuances of these interactions. Moreover, it lays the foundation for further structural, physiological and therapeutic studies.

      Audience

      The immune receptor signaling community and beyond, including any lineage in which ITIM/ITSM-containing receptors play a major role in regulating cellular responses.

      Your expertise

      ITIM/ITSM-containing receptors, kinase-phosphatase molecular switches, cellular reactivity to extracellular matrix proteins

      Referee Cross-commenting

      Generally agree with reviewer's comments. Constructive overall and fair. Although I was thinking additional competition experiments, I do not think necessary. Over the top for this study. Hence, 1 month should suffice to revise accordingly.

    5. Referee #1

      Evidence, reproducibility and clarity

      This paper explores an interesting problem of SHP1/SHP2 preferences of inhibitory immunoreceptors. The author are quick to point out that many of their individual data points confirm published results at some level, but the power of the paper is in the parallel analysis of both PD1, which is strongly biased towards SHP2 and BTLA, which is biased towards SHP1. This gives them the opportunity to test the predictions of descriptive experiment by making simple mutated receptors with swapped ITIM or ITSM domains.

      The work is very well done and generally the authors are quite careful and precise about the language used to describe results, in general.

      The results are quite striking in that the find plenty of evidence for transient interaction of SHP1 with PD1 based on the biophysical measurements, but don't detect the interactions in pull down or in "in cell" microcluster recruitment experiments. In describing the pull-downs they discuss the issue of dissociation during washing potentially missing interactions that are taking place. I would prefer that the pull down is fine evidence for binding, but lack of pull down is not evidence for lack of binding. They should double check that this language is consistent. Also, unless something has changed in the microcluster binding experiments, this in situ recruitment of SHP2 to PD1 is only observed or a 2-3 minutes and then can't be detected, the situation for SHP2 becoming the same as it is for SHP1. If the kinetics are different in the cleaner systems that have now developed they should show this in a primary figure as this would be then different when what is reported previously.

      The gap in this study is lack of any functional analysis. The Jurkat model could be quite useful as they have a relatively clean system for asking if the transient binding of SHP1 to PD1 has any functional impact, which they have not yet followed through on. Does PD-1 recruited SHP2 have any impact on function after the 5 minutes? Furthermore, the authors need to keep in mind that mice deficient in SHP2 respond to anti-PD1 checkpoint therapies (Rota, G., Niogret, C., Dang, A. T., Barros, C. R., Fonta, N. P., Alfei, F., Morgado, L., Zehn, D., Birchmeier, W., Vivier, E., & Guarda, G. (2018). Shp-2 Is Dispensable for Establishing T Cell Exhaustion and for PD-1 Signaling In Vivo. Cell Rep, 23(1), 39-49. https://doi.org/10.1016/j.celrep.2018.03.026). This is an important issue to discuss in light the the very interesting binding analysis the authors have performed. But I think the functional analysis can be part of a future paper.

      I would suggest that the title be modified slightly from "SHP1/SHP2 discrimination" to "differential SHP1/SHP2 interaction" and leave discussion of discrimination until they have the functional data integrated over times that are relevant to T cell transcriptional regulation (1-2 hrs). The functional analysis can be in another paper, but it would be interesting to have a paragraph in the discussion raising the outstanding issues beyond stable binding detected by the pull-down and microcluster recruitment experiments- what are the implications for function. Could the transient interactions in the noise of the steady state and equilibrium measurements be functional?

      I would summarise that the work is outstanding as biochemistry and biophysics and it should be published nearly as is. I'm suggesting minor revisions in that the changes are just to text, but I think this is important and somewhat nuanced aspect of the paper that will make it even more helpful to readers.

      Significance

      The authors generate a detailed descriptive data set about the component interaction of SHP1 and SHP2 SH2 domains with PD1 and BTLA intracellular domains. They then test hypotheses generated from the descriptive data set to better define the nature of the interactions and why PD1 recruits primarily SHP2, while BTLA mainly recruits SHP1. PD1 is a major driver or the cancer immunotherapy revolution and SHP2 is the major candidate for a signalling effector of PD1. This paper can become the reference paper for the specificity and engineering of this interaction, which will make it highly significant in a very active and still expanding field.

      Referee Cross-commenting

      I still feel that "discrimination" has a functional/activity connotation that is not addressed at all in this paper, but can be addressed. I'm happy to have the suggestion stand and let the authors decide. They need to live with it once its published. Another suggestion- the citations on regulation are mostly old. A good recent paper is Pádua, R. A. P., Sun, Y., Marko, I., Pitsawong, W., Stiller, J. B., Otten, R., & Kern, D. (2018). Mechanism of activating mutations and allosteric drug inhibition of the phosphatase SHP2. Nature Communications, 9(1), 4507. https://doi.org/10.1038/s41467-018-06814-w .

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

      Learn more at Review Commons


      Reply to the reviewers

      Reviewer #1 (Evidence, reproducibility and clarity):

      This study reveals the role of WW-PLEKHAs (PLEKHA5, 6 and 7) in the basolateral targeting of copper (Cu) transporter ATP7A. The Authors suggest that the WW-PLEKHAs/PDZD11/ATP7A interaction directs Cu-induced trafficking of ATP7A to the basolateral surface of epithelial cells. Suppression of WW-PLEKHAs impairs basolateral delivery of ATP7A and causes increased intracellular Cu levels. On the contrary, WW-PLEKHAs do not seem to participate in the retrieval of ATP7A back to the Golgi once the Cu levels return to basal values. To support these notions the manuscript provides a substantial set of the data, which were achieved with a wide repertoire of methods. In my view, this manuscript could be of interest to a broad readership, ranging from cells biologists to medical doctors. However, further revision should address the concerns outlined below.

      Major points:

      1. The Authors claim that at basal Cu conditions ATP7A resides in the TGN regardless of PDZD11 or WW-PLEKHAs depletion (Figs. 3, 4 and Fig. S6, S7). However, colocalization with TGN marker and its quantification are not shown. Thus, the colocalization of ATP7A with TGN marker (Golgin 97 should work in all cell types) has to be shown and its quantification (Pearson coefficient) has to be provided for control and all KO cells.

      Response: We thank the Reviewer for this comment. We plan to carry out the IF colocalization of ATP7A with Golgin97 and quantifications for WT and KO clonal lines at basal Cu conditions.

      1. Along the same line, ATP7A colocalization with TGN marker and its quantification also has to be conducted for the Cu washout experiments.

      Response: We plan to carry out the IF colocalization of ATP7A with Golgin97 and quantifications for WT and KO clonal lines for the Cu washout conditions.

      1. The authors say that upon addition of Cu ATP7A labeling was detected along lateral contacts, and near the apical and basal plasma membranes (Fig. 3B, WT). Here again "near apical" localization of ATP7A has to be clarified. This could either represent the ATP7A pool that still remains in the Golgi (which is usually close to apical surface in polarized epithelial cells) or the ATP7A pool delivered to the apical membrane of the cells. However, apical targeting of ATP7A would be odd considering previously published data that shows basolateral localization in polarized epithelial cells. Thus, the authors have to show whether "apical" ATP7A overlaps with TGN marker or with an apical marker (Gp135).

      Response. This Reviewer is correct. Effectively we believe that the localization of ATP7A that we observe in cysts is not apical, but sub-apical, as shown for the localization of PLEKHA5, where colocalization with the apical marker gp135 clearly shows a different localization (Fig. 2I).

      Therefore, we will carry out co-localization of ATP7A with gp135 in MDCK cells (the monoclonal antibody does not work on mCCD cells) and the labeling of the micrographs in Fig. 3B and 4B will be revised (sub-apical instead of apical). The labeling of PLEKHA5 sub-apical pool will also be revised (sub-apical and not apical) in Fig. 2.

      1. PDZD11 or PLEKHA6/7 KOs lead to an ATP7A pattern, which looks like pretty large scattered vesicles that do not overlap with basolateral marker. What are these round ATP7A structures, endosomes? Colocalization assessments with EEA1 (early endosomes), VPS35 (sorting endosome) and LAMP1 (late endosomes) would be needed to clarify this. Alternatively, these vesicles could represent a fragmented Golgi with ATP7A inside. To establish this, labelling with TGN marker at these conditions is required.

      Response: We thank the Reviewer for this comment. To clarify the nature (endosomes, Golgi, etc) of the membrane vesicles where ATP7A is localized in KO lines we will carry out double IF colocalization of ATP7A with either Golgin97 or early/sorting (which are mostly overlapped) endosome or the late endosome markers.

      1. Biotinylation experiments. The Authors say that KO of either PDZD11, or PLEKHA7, or both PLEKHA6 and PLEKHA7, but not PLEKHA6 alone, decreased ATP7A levels at the basolateral surface of mCCD cells (Fig. 3G), while a small decrease in the basolateral levels of ATP7A is observed in PLEKHA5-KO, but not PLEKHA6-KO MDCK cells (Fig. 4G). Honestly, it is tough to see this. In Fig. 4G all ATP7A bands in the biotinylated fraction look similar. In Fig. 3G, the P11 and P6/7 KO bands of biotinylated ATP7A might be a bit less intense than in WT, while the P6 KO signal looks even more intense that WT. More convincing blots with quantification have to be provided for both figures.

      Response: We will carry out additional immunoblots and quantifications of the biotinylation experiments results.

      1. Along the same line. Why was apical biotinylation of ATP7A not included? It absolutely should be done to understand whether any KO induces apical mistargeting of ATP7A.

      Response: The levels of ATP7A at the apical surface upon basal or elevated copper are negligible and not physiologically relevant, as established by previous biotinylation studies (for example Greenough et al AJP 2004, and Nyasae et al AJP-Gastrointest Liver Physiol 2007). We will carry out IF analysis of WT and KO cells with ATP7A and apical markers (ex. gp135) to clarify if the subapical labeling for ATP7A is or not on the apical membrane. Importantly, LESS, and not more subapical labeling is detected in KO lines (Fig. 3B, Fig. 4B), as we pointed out in the results section. Therefore, the KO lines do not show increased apical (mistargeting of) ATP7A.

      1. Copper metabolism. The authors say that KO of either PDZD11 or PLEKHA6/7 results in higher Cu levels. What does this mean in terms of physiology and pathology? In the context of Menkes disease one has to show that this intracellular Cu increase is due to a reduction in Cu release from the cells. So, Cu release from the cells into medium has to be measured by ICP-MS or Cu64. On the other hand, it would be important to understand whether Cu accumulation in KO cells is toxic. To this end viability of KO cells should be tested in Cu dose-response experiments.

      Response: The focus of this paper is the molecular mechanisms of ATP7A targeting to the BL plasma membrane, rather than a quantitative analysis of copper transport by and analysis of physiology/pathology of copper homeostasis ATP7A in our WT and KO cell lines. Our measurement of the intracellular copper using the CF4 probe was designed as a physiological readout to confirm that altered localization at the BL plasma membrane correlates with reduced copper extrusion, as it can be hypothesized. This said, to address this point, we plan to carry out an ICP-MS analysis of intracellular copper in selected WT and KO lines, after loading cells with different amounts of copper, and at different times after return to basal copper levels. CF4 and ICP-MS generally track, but they do measure distinct copper pools: CF4 measures exchangeable Cu pools while ICP-MS measures total Cu pools. We will also carry out a crystal violet analysis (see Gudekar et al, Scient. Reports, 2020) of the viability of WT and KO cells in the absence and presence of low or elevated copper levels, as suggested by the Reviewer.

      1. How critical is WW-PLEKHAs or PDZD11 deficiency in terms of Cu metabolism? Are there genetic disorders or mouse phenotypes associated with their loss of function? If yes, do these phenotypes include any impairment of Cu metabolism?

      Response: To our knowledge, no genetic study has addressed the role of WW-PLEKHAs and PDZD11 in Cu metabolism in vivo. PLEKHA7-KO mice are viable and were not reported to display any phenotype consistent with grossly altered Cu metabolism (Popov et al 2015). Mice KO for either PLEKHA5 or PLEKHA6 or PDZD11 have not been described. However, if WW- PLEKHAs have redundant functions in the trafficking of ATP7A, one would expect that mutation/KO of only one of them may not yield a significant phenotype. Furthermore, we cannot exclude that additional PDZ-containing proteins may participate in the trafficking of ATP7A, compensating a pathological or experimental loss of PDZD11. So, answering this question will require to generate single, double and triple KO mice for WW-PLEKHAs, and carry out a detailed analysis of in vivo Cu metabolism. This is beyond the scope of this paper. The text of the Discussion will be revised to address this comment.

      1. Discussion. Could PH domains of WW-PLEKHAS be involved in their basolateral localization, thereby generating a targeting patch for ATP7A? Some publications suggest that the basolateral membrane might be enriched in specific PIPs, which in turn generates a favorable environment for some PH domains. Is this the case for PH domains of WW-PLEKHAS?

      Response: This is an interesting hypothesis that should be investigated in future studies (lipidomic analysis of KO lines, overexpression studies, etc), but is outside of the scope of the present manuscript.

      Minor points:

      1. Fig. 6C. CFP-HA is a negative control but still gives a band (although of lower intensity). So how can one be sure that other interactions are specific? This is particularly worrying because the quantification shows a very minor (less than 1.5) increase in the intensity of bands corresponding to specific interactors.

      Response: CFP-HA is used as a “negative control” 3_rd _protein, added to bait (GST-PDZD11) and prey (GFP-ATP7A-Cter) (Fig. 6C). The IB shows that in the presence of CFP-HA the bait binds the prey, which is in agreement with the previously reported interaction between PDZD11 and the C-terminal region of ATP7A (Stephenson et al JBC, 2005). The point of the Figure is to show that the interaction between bait and prey is enhanced in the presence of HA-tagged WW- PLEKHAs (again, CFP-HA is the negative control). We agree that the increase is not huge, but it is nevertheless statistically significant, based on several experiments (Fig. 6E).

      1. Page 11. The result section title "WW-PLEKHAs promote PDZD11 binding to ATP7A through PDZD11 (Figure 6)" does not sound right and has to be corrected.

      Response: The text was revised ("WW-PLEKHAs promote PDZD11 binding to ATP7A”).

      Reviewer #1 (Significance):

      Delivery of copper transporter ATP7A to the basolateral surface of epithelial cells is of great importance for maintenance of copper metabolism and, hence for human health in general. Impairment of this process in enterocytes causes fatal Menkes disease. However, the mechanisms driving basolateral targeting of ATP7A remained poorly characterized. This study provides a significant advance in our understanding of these mechanisms and opens new avenues for investigation of how WW-PLEKHAs/PDZD11-mediated targeting of ATP7A might be affected in the context of inherited disorders of copper metabolism.

      Reviewer #2 (Evidence, reproducibility and clarity):

      This manuscript uncovers new PDZD11 interactors that participate in trafficking of the copper transporter ATP7A from the Golgi/TGN to the cell periphery in response to high copper concentrations. These interactors named PLEKHA5, PLEKHA6, and PLEKHA7 interact with the N-terminal Pro-rich domain of PDZD11 through their WW domains. As PDZD11 interacts with the C-terminal region of ATP7A, the authors investigated the hypothesis that WW-PLEKHAs are required for copper-induced relocalization of ATP7A from the TGN to the plasma membrane where it functions in copper efflux. In vitro pull down experiments verified the formation of ATP7A-, PLEKHAs-, and PDZD11-containing complexes. Using using CRISPR/Cas9 technology, the authors have generated PDZD11-, PLEKHA5-, PLEKHA6-, and PLEKHA6/7-KOs cell lines.

      Cells lacking one (or more) of these proteins were examined by microscopy with respect to their ability of targeting ATP7A to the cell periphery in response to copper. Abnormal trafficking of ATP7A in these mutant cell lines (PDZD11-, PLEKHA5-, PLEKHA6-, PLEKHA7-, and PLEKHA6/7-KOs) presumably prevented copper efflux since elevated intracellular copper was detected using the fluorescent copper probe CF4.

      Although it is difficult to read across the article's figures and supporting figure files (going back- and-forth repeatedly), the manuscript is generally clear and well written, and the results seem well documented accompanied by a tremendous amount of work.

      Comments.

      1. Two-hybrid screen occurs in the nucleus. How the authors could explain the fact that the use of PDZD11 as a bait exhibited an interaction with PLEKHA5 and PLEKHA6 (as well as PLEKHA7) in this system? Microscopic analysis of PLEKHA5 showed a cytoplasmic submembrane localization with E-cadherin, whereas PLEKHA6 exhibited a localization along the plasma membrane at apical junctions. In the case of PLEKHA7, it is an adherens junction protein. Furthermore, these three proteins are quite big (1116, 1297, and 1121 AAs, respectively) with their WW regions at their N termini, which involved the expression of very long cDNAs fused to the TA domain. As truly membrane-associated proteins, isn't surprising that a two hybrid approach worked?

      Response: We carried out several Y2H screens with a number of different baits, and we have always validated the physiological significance of the high score interactions (Pulimeno et al JBC 2011, Guerrera et al, 2016 JBC, and other unpublished data). So, it is an approach that reliably works very well. The Hybrigenics human placenta library that we used contains fragments of proteins, not the full-length proteins. Fig. 1A shows the preys identified with the Y2H using PDZD11 as a bait. The preys that were found comprise only the N-terminal regions of WW- PLEKHAs, not the FL proteins.

      1. Fig. 1C, what are the 4 bands seen for the second blot (anti-HA) in lines 1, 2, 9 and 10? The blot was cut in a way that not enough of the membrane can be evaluated. Why using Ponceau for GST-baits and not using anti-GST antibodies? It would be much better having an uniform method (Western blot assays) to show the data.

      This latter comment is true for Fig 1D, E, and Fig 6.

      Response: The 4 bands seen for the second blot (anti-HA) in lanes 1, 2, 9 and 10 are non- specific cross-reaction of the antibodies with the baits, that are present in high concentration (and present also where there is no CFP-HA, in lanes 1 and 9). The preys can be identified on the basis of their molecular size. For example, no CFP-HA prey is detected, since its size is intermediate between the baits, thus the negative control is validated. We use Ponceau for 2 reasons: 1) Ponceau can detect very well baits on nitrocellulose membranes; 2) to use GST antibodies we would need to cut the membranes. But cutting membranes is not possible when the size of some preys (in this case, the negative control) is in the same range of sizes as the baits. Thus, if we used anti-GST antibodies we would have to strip and re-probe the membranes, which is not optimal in our experience to elicit good signals.

      1. Fig 1B, why Caco-2 cells? All the other experiments were conducted with other cell lines such as mCCD and MDCK. Using different cell lines could give different results.

      Response: We used Caco2 cells because the Y2H was carried out with a human bait on a human placental library, and Caco2 are human cells. We also tried to use MDCK cells, but the efficiency of the IP was lower.

      1. Fig 1D, it is unclear whether the GST-PDZD11 fusion protein (bait) was present or not when used in pull down assays with GFP alone. This is a clear disadvantage of Ponceau, immunoblot would be much better to use.

      Response: The labeling by Ponceau is not optimal in one image (Fig. 1D), probably due to a problem of transfer. But a clearer image for the same pulldown with the same bait is shown in the bottom panel of Fig. 1E (where we show 3 PDZD11 baits, FL, N-term and delta-24), and it clearly shows good normalization of baits. We stain baits with Ponceau for normalization.

      1. In Fig 3A, under basal copper conditions, microscopic image of the PLEKHA6/7-KO seems indicate a distinct pattern of localization for ATP7A in comparison to that of WT. However, this difference does not seem to be highlighted in Fig 3E.

      Response: We will re-examine all the micrographs used for the quantification and integrate the data with the results of the colocalization between ATP7A and TGN marker. This should allow us to establish whether there is a dissociation of ATP7A labeling from TGN marker labeling in KO cells, or else a fragmentation of the TGN in the double-KO mCCD cells.

      In Figs 3F and 4F, what was the method for quantification?

      Response: The methods for quantifications are described in the “Image quantification” section of the Methods. We will add new data about the quantification of co-localization of ATP7A with TGN and endosomal markers.

      Along these lines, what is the copper concentration under basal conditions? How much copper was used for elevated copper conditions and what was the time of treatment?

      Response: Basal conditions refers to normal cell culture medium (“Cell culture” section of the Methods), and elevated copper is 315 µM of CuCl2 dissolved in culture medium. Cells were treated for 4hr (MDCK) or 5hr (mCCD) when cells were cultured on Transwells, overnight in the case of cysts.

      1. Is there any evidence for Atp7A-PDZD11-PLEKHAs association in vivo? Do the authors have assessed these protein-protein interactions using methods such as bimolecular fluorescence in cells?

      Response: We have attempted co-IP experiments with endogenous proteins, but they were inconclusive, probably due to the different extraction conditions required to solubilize membrane (ATP7A) and cytoplasmic (WW-PLEKHAs, PDZD11) proteins, and a disassembly of the complex under the conditions required to solubilize ATP7A. We have not tried bimolecular fluorescence, but for the revision we plan to carry out Proximity Ligation Assay (PLA) experiments, which in our hands are very effective in assessing physiological proximity of proteins in cells. Our pulldown experiments however provide evidence that the three proteins form a complex, and that WW- PLEKHAs enhance the interaction between PDZD11 and ATP7A (Fig. 6C-E). This is a mechanism that we have shown occurs also for the complex between PLEKHA7, PDZD11 and Tspan33 (Shah et al, 2018 Cell Rep, Rouaud et al, 2020 JBC).

      1. In Fig 5, do the authors have verified the mRNA (or/and protein) steady-state levels of metallothioneins? Probing whether metallothioneins are induced would strongly reinforced their conclusion as to whether an increase intracellular copper levels occurred in PDZD11-, PLEKHA5-, PLEKHA6-, PLEKHA7-, and PLEKHA6/7-Kos cell lines.

      Response: We thank the Reviewer for this comment. In the revision we will carry out RT-PCR analysis of the levels of expression of mRNAs for Metallothioneins I and II.

      1. In Fig 6E, what was the method for quantitative immunoblot assays? Have you used an Odyssey infrared imaging system (Li-Cor). What was the loading (internal) control under the same analytical method?

      Response: The Li-Cor imaging system was used to capture the signals, and intensities were measured in Image Studio Lite program (Li-cor). Signals from the prey (C-terminus of ATP7A) were normalized to signals from the bait (GST-PDZD11) which were used as loading control.

      1. In the case of the manuscript section entitled " PLEKHA5, PLEKHA6 and PLEKHA7 show distinct localizations in cells and tissues and define cytoplasmic...", (pages 5 to 7) the reader would benefit having a Table that would summarize all the data. It would be more understandable.

      Response: We thank the Reviewer for this suggestion. We will include a Table in the revision.

      1. Do PDZD11, PLEKHA5, PLEKHA6, and PLEKHA7 proteins exist as multiple isoforms? If that is the case, for each of them, are they exhibiting the same tissue-specific expression profiles as shown in Fig S3? For each protein, if different isoforms exist, perhaps some of them participate in a different way for the targeting of ATP7A?

      Response: No PDZD11 isoforms are known, but 15, 5, and 9 different protein-coding transcripts are reported (ensemble.org) for PLEKHA5, PLEKHA6 and PLEKHA7, respectively, the largest ones being the WW-containing transcripts. We focused exclusively on the WW-containing isoforms of PLEKHAs because PDZD11 binds to the WW domains, and the Y2H identified only the WW-containing isoforms of PLEKHA5, PLEKHA6 and PLEKHA7. The observation that the phenotype of PDZD11-KO cells is similar to that of either PLEKHA6-KO, PLEKHA7-KO or double-KO mCCD cells suggests that PLEKHA5, PLEKHA6 and PLEKHA7 WW-containing isoforms act in a complex with PDZD11. This is consistent with the previous observations that highlight a role of the C-terminal region of ATP7A in regulating its traffic, and the binding of the same region to PDZD11. However, we cannot exclude that PLEKHA5/6/7 isoforms that lack the WW domains could participate in the regulation of the targeting of ATP7A, through other, PDZD11-independent mechanisms. The text of the Discussion will be revised to clarify this point.

      1. Is it known whether PDZD11, PLEKHA5, PLEKHA6, and PLEKHA7 proteins participate in the copper-regulated trafficking of the ATP7B (Wilson) protein? In Fig S3, it is shown that they are expressed in liver, with PLEKHA7 exhibiting a slower migration (protein modification?). Alternatively, are they strictly involved in the regulation of ATP7A (Menkes)? Could the authors discuss about it?

      Response: ATP7B lacks the PDZ-binding motif that is responsible for PDZD11 binding, and the C-terminus of ATP7B does not interact with PDZD11 (AIPP1) by beta-galactosidase assays in yeast, unlike ATP7A (Stephenson et al, JBC 2005). For this reason, ATP7B is not expected to be regulated by PDZD11 and WW-PLEKHAs. However, analysis of the localization of ATP7B in our cell lines could be done in future studies. The text of the Discussion will be revised to make this point.

      1. The proposed model in Fig 7 is unclear illustrating a nucleus that consumes a lot of space while it is not involved in the proposed mechanism. Cellular proteins that are involved in the proposed mechanism should be bigger and their interactions that lead to formation of protein complexes must be better illustrated as a function of copper availability.

      Response: The model of Fig. 7 will be re-drawn to take into account these suggestions.

      1. Typo. Line 320: remove "or" and replace it by "and" : ...both PLEKHA6 and PLEKHA7 (Fig. 5A-D).

      2. Typo. Line 329: remove (Figure 6) in the title.

      Response: The typos were corrected.

      Reviewer #2 (Significance ):

      This study represents a significant advanced in the copper field.

      Reviewer #3 (Evidence, reproducibility and clarity):

      Summary

      The authors identified some major interesting findings including the key role of WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) in the recruitment of PDZD11 targeting ATP7A to the cell periphery in response to elevated copper. Generating the antibodies against PLEKHAs and PDZD11 and various knock out cell lines and validating their expression in these cell lines and tissues is innovative. Further, the authors showed that copper dependent WW-PLEKHAs and PDZD11 regulate the localization and function of ATP7A to modulate cellular copper homeostasis.

      Major comments:

      We are in agreement with the manuscript conclusions. Based on the presented studies, the authors propose the in-vivo role of WW-PLEKHAs and PDZD11 in ATP7A trafficking, and how microtubule dynamics and trafficking machinery regulate ATP7A localization. Additionally, investigating the effects of the cell membrane trimolecular complexes ATP7A-PDZD11-WW- PLEKHA on elevated copper would be impactful.

      Additional notes:

      1. Figure 4, the authors provide excellent data and images showing localization of PLEKHA, PDZD11 and ATP7A within different cell lines. Nevertheless, showing PLEKHA, PDZD11 and ATP7A localization on membrane of the cell surfaces at elevated copper condition with cell fractionation technique and their interaction through co-immunoprecipitation (co-IP) could validate author's hypothesis. At least the authors should comment on this.

      Response: As stated in the response to comment n.6 from Reviewer #2, we attempted co-IP experiments with endogenous proteins, which were inconclusive. Our pulldown experiments provide evidence that the three proteins form a complex, and that WW-PLEKHAs enhance the interaction between PDZD11 and ATP7A (Fig. 6C-E). This is a mechanism that we have shown occurs also for the complex between PLEKHA7, PDZD11 and Tspan33 (Shah et al, 2018 Cell Rep, Rouaud et al, 2020 JBC). We plan for the revision to carry out Proximity Ligation Assay (PLA) experiments, which in our hands are very effective in assessing proximity of proteins in cells, when co-IPs are technically difficult or impossible.

      1. Figure 5, Alternatively, intracellular copper levels by ICPMS in the cell lines would strengthen the results. As author's treated the cell lines with very high copper concentration, copper concentration dependent studies would be appreciated to verify how PLEKHA's and PDZD11 response depends on copper concentration

      Also, the authors should clearly mention the number of replicates for each experiment and indicate in the figure legends.

      Response. We will carry out ICP-MS to evaluate intracellular copper levels as a function of genotype. Depending on the results, we will carry out studies about the dose-dependence of the effects of copper. The number of replicates of the experiment will be mentioned in the Figure legend in the revised text.

      Minor comments:

      1. Figure1B, co-IP efficiency is lower in Caco-2 cells, therefore endogenous levels of PLEKHA5, PLEKHA6 and PDZD11 in Caco-2 should be checked and shown. Mention the number of replicates for the experiments.

      Response. Endogenous levels of proteins are shown in the Input lanes. The low levels of PLEKHA5 in Caco2 cells are consistent with the IB analysis of tissue lysates, showing relatively low levels in intestine (Fig. S3D). The number of replicates of the experiment will be mentioned in the Figure legend in the revised text.

      1. Figure 2A, as per result of 2A, E-cadherin labeling is missing. Figure 2M and 2N, author analyzed the co-localization of PLEKHA-5 in presence of nocodazole but not PDZD11. It would be interesting to see the PDZD11 as well after nocodazole treatment.

      Response. E-cadherin-labelled panel will be added to Fig. 2A in the revision. We will also show the effect of nocodazole on the localization of PDZD11.

      1. The result section title for figure 6 (line 329) is misleading. Also, trimolecular complex PLEKHA's, PDZD11 and ATP7A membrane localization at elevated copper concentration could be shown by immunofluorescence, if possible.

      Response. The title of the section was revised, to reflect more accurately the results of Figure 6. It is now “WW-PLEKHAs promote the binding of the C-terminal region of ATP7A to PDZD11”.

      Triple IF colocalization of endogenous PLEKHAs, PDZD11 and ATP7A is not possible for 2 reasons: 1) PDZD11 antibodies can only reveal endogenous junctional (clustered) labeling (Guerrera et al JBC2016); the lateral and cytoplasmic labeling is too weak, and can only be appreciated upon overexpression of PDZD11, as shown in Fig. 2B-E (co-expression with selected WW-PLEKHAs highlights how each PLEKHA directs PDZD11 to a different pool). 2) Both antibodies against PDZD11 and ATP7A were raised in rabbits, which makes it technically impossible to do triple labeling. We will address the question of the existence of the ATP7A- containing trimolecular complex by PLA analysis (ATP7A+PDZD11 and ATP7A+WW-PLEKHAs)._

      1. General comment: it would be interesting to see the hypothesis and finding in mice model with copper accumulation (for example Atp7b KO mice) as PLEKHA's and PDZD11 are sensitive to copper concentration. Or at least the authors can comment on this future possibility.

      Response. We agree with the Reviewer that mouse models could be useful to test the relevance of WW-PLEKHAs and PDZD11 as targets or effectors of copper-sensing mechanisms in vivo.

      The text of the Discussion will be modified to envisage these possible future studies

      Reviewer #3 (Significance):

      In conditions including Menkes disease, occipital horn syndrome (OHS), and ATP7A-related distal motor neuropathy (DMN), characterized by altered intestinal copper metabolism, the new knowledge ATP7A associates with WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) and PDZD11 is an important finding for the study of copper homeostasis.

      As ATP7A is structurally similar to ATP7B (60% homology), the current study opens the area of the research where WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) and PDZD11 could also play role in ATP7B trafficking to address not only Menkes disease but also Wilson disease and other diseases related to altered copper levels.

      This is a well written and presented manuscript with excellent mechanistic work utilizing molecular imaging techniques and several confirmatory experiments. I recommend the manuscript to be accepted for publication with minor modifications.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Summary

      The authors identified some major interesting findings including the key role of WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) in the recruitment of PDZD11 targeting ATP7A to the cell periphery in response to elevated copper. Generating the antibodies against PLEKHAs and PDZD11 and various knock out cell lines and validating their expression in these cell lines and tissues is innovative. Further, the authors showed that copper dependent WW-PLEKHAs and PDZD11 regulate the localization and function of ATP7A to modulate cellular copper homeostasis.

      Major comments:

      We are in agreement with the manuscript conclusions. Based on the presented studies, the authors propose the in-vivo role of WW-PLEKHAs and PDZD11 in ATP7A trafficking, and how microtubule dynamics and trafficking machinery regulate ATP7A localization. Additionally, investigating the effects of the cell membrane trimolecular complexes ATP7A-PDZD11-WW-PLEKHA on elevated copper would be impactful.

      Additional notes:

      1. Figure 4, the authors provide excellent data and images showing localization of PLEKHA, PDZD11 and ATP7A within different cell lines. Nevertheless, showing PLEKHA, PDZD11 and ATP7A localization on membrane of the cell surfaces at elevated copper condition with cell fractionation technique and their interaction through co-immunoprecipitation (co-IP) could validate author's hypothesis. At least the authors should comment on this.
      2. Figure 5, Alternatively, intracellular copper levels by ICPMS in the cell lines would strengthen the results. As author's treated the cell lines with very high copper concentration, copper concentration dependent studies would be appreciated to verify how PLEKHA's and PDZD11 response depends on copper concentration Also, the authors should clearly mention the number of replicates for each experiment and indicate in the figure legends.

      Minor comments:

      1. Figure1B, co-IP efficiency is lower in Caco-2 cells, therefore endogenous levels of PLEKHA5, PLEKHA6 and PDZD11 in Caco-2 should be checked and shown. Mention the number of replicates for the experiments.
      2. Figure 2A, as per result of 2A, E-cadherin labeling is missing. Figure 2M and 2N, author analyzed the co-localization of PLEKHA-5 in presence of nocodazole but not PDZD11. It would be interesting to see the PDZD11 as well after nocodazole treatment.
      3. The result section title for figure 6 (line 329) is misleading. Also, trimolecular complex PLEKHA's, PDZD11 and ATP7A membrane localization at elevated copper concentration could be shown by immunofluorescence, if possible.
      4. General comment: it would be interesting to see the hypothesis and finding in mice model with copper accumulation (for example Atp7b KO mice) as PLEKHA's and PDZD11 are sensitive to copper concentration. Or at least the authors can comment on this future possibility.

      Significance

      In conditions including Menkes disease, occipital horn syndrome (OHS), and ATP7A-related distal motor neuropathy (DMN), characterized by altered intestinal copper metabolism, the new knowledge ATP7A associates with WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) and PDZD11 is an important finding for the study of copper homeostasis.

      As ATP7A is structurally similar to ATP7B (60% homology), the current study opens the area of the research where WW-PLEKHAs (PLEKHA5, PLEKHA6, PLEKHA7) and PDZD11 could also play role in ATP7B trafficking to address not only Menkes disease but also Wilson disease and other diseases related to altered copper levels.

      This is a well written and presented manuscript with excellent mechanistic work utilizing molecular imaging techniques and several confirmatory experiments. I recommend the manuscript to be accepted for publication with minor modifications.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      This manuscript uncovers new PDZD11 interactors that participate in trafficking of the copper transporter ATP7A from the Golgi/TGN to the cell periphery in response to high copper concentrations. These interactors named PLEKHA5, PLEKHA6, and PLEKHA7 interact with the N-terminal Pro-rich domain of PDZD11 through their WW domains. As PDZD11 interacts with the C-terminal region of ATP7A, the authors investigated the hypothesis that WW-PLEKHAs are required for copper-induced relocalization of ATP7A from the TGN to the plasma membrane where it functions in copper efflux. In vitro pull down experiments verified the formation of ATP7A-, PLEKHAs-, and PDZD11-containing complexes. Using using CRISPR/Cas9 technology, the authors have generated PDZD11-, PLEKHA5-, PLEKHA6-, and PLEKHA6/7-KOs cell lines. Cells lacking one (or more) of these proteins were examined by microscopy with respect to their ability of targeting ATP7A to the cell periphery in response to copper. Abnormal trafficking of ATP7A in these mutant cell lines (PDZD11-, PLEKHA5-, PLEKHA6-, PLEKHA7-, and PLEKHA6/7-KOs) presumably prevented copper efflux since elevated intracellular copper was detected using the fluorescent copper probe CF4.

      Although it is difficult to read across the article's figures and supporting figure files (going back-and-forth repeatedly), the manuscript is generally clear and well written, and the results seem well documented accompanied by a tremendous amount of work.

      Comments.

      1. Two-hybrid screen occurs in the nucleus. How the authors could explain the fact that the use of PDZD11 as a bait exhibited an interaction with PLEKHA5 and PLEKHA6 (as well as PLEKHA7) in this system? Microscopic analysis of PLEKHA5 showed a cytoplasmic submembrane localization with E-cadherin, whereas PLEKHA6 exhibited a localization along the plasma membrane at apical junctions. In the case of PLEKHA7, it is an adherens junction protein. Furthermore, these three proteins are quite big (1116, 1297, and 1121 AAs, respectively) with their WW regions at their N termini, which involved the expression of very long cDNAs fused to the TA domain. As truly membrane-associated proteins, isn't surprising that a two hybrid approach worked?
      2. Fig. 1C, what are the 4 bands seen for the second blot (anti-HA) in lines 1, 2, 9 and 10? The blot was cut in a way that not enough of the membrane can be evaluated. Why using Ponceau for GST-baits and not using anti-GST antibodies? It would be much better having an uniform method (Western blot assays) to show the data. This latter comment is true for Fig 1D, E, and Fig 6.
      3. Fig 1B, why Caco-2 cells? All the other experiments were conducted with other cell lines such as mCCD and MDCK. Using different cell lines could give different results.
      4. Fig 1D, it is unclear whether the GST-PDZD11 fusion protein (bait) was present or not when used in pull down assays with GFP alone. This is a clear disadvantage of Ponceau, immunoblot would be much better to use.
      5. In Fig 3A, under basal copper conditions, microscopic image of the PLEKHA6/7-KO seems indicate a distinct pattern of localization for ATP7A in comparison to that of WT. However, this difference does not seem to be highlighted in Fig 3E.

      In Figs 3F and 4F, what was the method for quantification?

      Along these lines, what is the copper concentration under basal conditions? How much copper was used for elevated copper conditions and what was the time of treatment?

      1. Is there any evidence for Atp7A-PDZD11-PLEKHAs association in vivo? Do the authors have assessed these protein-protein interactions using methods such as bimolecular fluorescence in cells?
      2. In Fig 5, do the authors have verified the mRNA (or/and protein) steady-state levels of metallothioneins? Probing whether metallothioneins are induced would strongly reinforced their conclusion as to whether an increase intracellular copper levels occurred in PDZD11-, PLEKHA5-, PLEKHA6-, PLEKHA7-, and PLEKHA6/7-KOs cell lines.
      3. In Fig 6E, what was the method for quantitative immunoblot assays? Have you used an Odyssey infrared imaging system (Li-Cor). What was the loading (internal) control under the same analytical method?
      4. In the case of the manuscript section entitled " PLEKHA5, PLEKHA6 and PLEKHA7 show distinct localizations in cells and tissues and define cytoplasmic...", (pages 5 to 7) the reader would benefit having a Table that would summarize all the data. It would be more understandable.
      5. Do PDZD11, PLEKHA5, PLEKHA6, and PLEKHA7 proteins exist as multiple isoforms? If that is the case, for each of them, are they exhibiting the same tissue-specific expression profiles as shown in Fig S3? For each protein, if different isoforms exist, perhaps some of them participate in a different way for the targeting of ATP7A?
      6. Is it known whether PDZD11, PLEKHA5, PLEKHA6, and PLEKHA7 proteins participate in the copper-regulated trafficking of the ATP7B (Wilson) protein? In Fig S3, it is shown that they are expressed in liver, with PLEKHA7 exhibiting a slower migration (protein modification?). Alternatively, are they strictly involved in the regulation of ATP7A (Menkes)? Could the authors discuss about it?
      7. The proposed model in Fig 7 is unclear illustrating a nucleus that consumes a lot of space while it is not involved in the proposed mechanism. Cellular proteins that are involved in the proposed mechanism should be bigger and their interactions that lead to formation of protein complexes must be better illustrated as a function of copper availability.
      8. Typo. Line 320: remove "or" and replace it by "and" : ...both PLEKHA6 and PLEKHA7 (Fig. 5A-D).
      9. Typo. Line 329: remove (Figure 6) in the title.

      Significance

      This study represents a significant advanced in the copper field.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      This study reveals the role of WW-PLEKHAs (PLEKHA5, 6 and 7) in the basolateral targeting of copper (Cu) transporter ATP7A. The Authors suggest that the WW-PLEKHAs/PDZD11/ATP7A interaction directs Cu-induced trafficking of ATP7A to the basolateral surface of epithelial cells. Suppression of WW-PLEKHAs impairs basolateral delivery of ATP7A and causes increased intracellular Cu levels. On the contrary, WW-PLEKHAs do not seem to participate in the retrieval of ATP7A back to the Golgi once the Cu levels return to basal values. To support these notions the manuscript provides a substantial set of the data, which were achieved with a wide repertoire of methods. In my view, this manuscript could be of interest to a broad readership, ranging from cells biologists to medical doctors. However, further revision should address the concerns outlined below.

      Major points:

      1. The Authors claim that at basal Cu conditions ATP7A resides in the TGN regardless of PDZD11 or WW-PLEKHAs depletion (Figs. 3, 4 and Fig. S6, S7). However, colocalization with TGN marker and its quantification are not shown. Thus, the colocalization of ATP7A with TGN marker (Golgin 97 should work in all cell types) has to be shown and its quantification (Pearson coefficient) has to be provided for control and all KO cells.
      2. Along the same line, ATP7A colocalization with TGN marker and its quantification also has to be conducted for the Cu washout experiments.
      3. The authors say that upon addition of Cu ATP7A labeling was detected along lateral contacts, and near the apical and basal plasma membranes (Fig. 3B, WT). Here again "near apical" localization of ATP7A has to be clarified. This could either represent the ATP7A pool that still remains in the Golgi (which is usually close to apical surface in polarized epithelial cells) or the ATP7A pool delivered to the apical membrane of the cells. However, apical targeting of ATP7A would be odd considering previously published data that shows basolateral localization in polarized epithelial cells. Thus, the authors have to show whether "apical" ATP7A overlaps with TGN marker or with an apical marker (Gp135).
      4. PDZD11 or PLEKHA6/7 KOs lead to an ATP7A pattern, which looks like pretty large scattered vesicles that do not overlap with basolateral marker. What are these round ATP7A structures, endosomes? Colocalization assessments with EEA1 (early endosomes), VPS35 (sorting endosome) and LAMP1 (late endosomes) would be needed to clarify this. Alternatively, these vesicles could represent a fragmented Golgi with ATP7A inside. To establish this, labelling with TGN marker at these conditions is required.
      5. Biotinylation experiments. The Authors say that KO of either PDZD11, or PLEKHA7, or both PLEKHA6 and PLEKHA7, but not PLEKHA6 alone, decreased ATP7A levels at the basolateral surface of mCCD cells (Fig. 3G), while a small decrease in the basolateral levels of ATP7A is observed in PLEKHA5-KO, but not PLEKHA6-KO MDCK cells (Fig. 4G). Honestly, it is tough to see this. In Fig. 4G all ATP7A bands in the biotinylated fraction look similar. In Fig. 3G, the P11 and P6/7 KO bands of biotinylated ATP7A might be a bit less intense than in WT, while the P6 KO signal looks even more intense that WT. More convincing blots with quantification have to be provided for both figures.
      6. Along the same line. Why was apical biotinylation of ATP7A not included? It absolutely should be done to understand whether any KO induces apical mistargeting of ATP7A.
      7. Copper metabolism. The authors say that KO of either PDZD11 or PLEKHA6/7 results in higher Cu levels. What does this mean in terms of physiology and pathology? In the context of Menkes disease one has to show that this intracellular Cu increase is due to a reduction in Cu release from the cells. So, Cu release from the cells into medium has to be measured by ICP-MS or Cu64. On the other hand, it would be important to understand whether Cu accumulation in KO cells is toxic. To this end viability of KO cells should be tested in Cu dose-response experiments.
      8. How critical is WW-PLEKHAs or PDZD11 deficiency in terms of Cu metabolism? Are there genetic disorders or mouse phenotypes associated with their loss of function? If yes, do these phenotypes include any impairment of Cu metabolism?
      9. Discussion. Could PH domains of WW-PLEKHAS be involved in their basolateral localization, thereby generating a targeting patch for ATP7A? Some publications suggest that the basolateral membrane might be enriched in specific PIPs, which in turn generates a favorable environment for some PH domains. Is this the case for PH domains of WW-PLEKHAS?

      Minor points:

      1. Fig. 6C. CFP-HA is a negative control but still gives a band (although of lower intensity). So how can one be sure that other interactions are specific? This is particularly worrying because the quantification shows a very minor (less than 1.5) increase in the intensity of bands corresponding to specific interactors.
      2. Page 11. The result section title "WW-PLEKHAs promote PDZD11 binding to ATP7A through PDZD11 (Figure 6)" does not sound right and has to be corrected.

      Significance

      Delivery of copper transporter ATP7A to the basolateral surface of epithelial cells is of great importance for maintenance of copper metabolism and, hence for human health in general. Impairment of this process in enterocytes causes fatal Menkes disease. However, the mechanisms driving basolateral targeting of ATP7A remained poorly characterized. This study provides a significant advance in our understanding of these mechanisms and opens new avenues for investigation of how WW-PLEKHAs/PDZD11-mediated targeting of ATP7A might be affected in the context of inherited disorders of copper metabolism.

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

      Learn more at Review Commons


      Reply to the reviewers

      Editor comments:

      Thank you for sending your manuscript entitled "In situ imaging of bacterial membrane projections and associated protein complexes using electron cryo-tomography" to Review Commons. We have now completed the peer review of the manuscript. Please find the full set of reports below.

      We thank the editors of Review Commons and all the reviewers for their insightful comments which helped us to improve our manuscript. We have now modified our manuscript based on the Reviewers’ comments and would like to ask you to consider our revised manuscript for publication.

      Reviewer #1:

      This manuscript by the Jensen lab surveys a plethora of bacterial outer-membrane projections captured over the years by in situ cryo-tomography under near-native conditions. The authors classify the different visualized structures, highlighting both similarities and differences among them. They further describe molecular complexes that are associated with these projections. The manuscript highlights the abundance of such understudied structures in nature, indicating the need to deepen our exploration into their biological functions and mechanisms of action.

      We thank the reviewer for her/his insightful comments that allowed us to improve our manuscript.

      The authors should state in the Abstract and Introduction that only diderm bacteria and outer- membrane extensions are included in the study.

      Done. We have modified the title, the abstract and the introduction to explicitly highlight this point.

      In the Introduction or Discussion the authors should mention the limits of the in situ cryo-tomography, such as the difficulty to observe regions in between neigbouring bacterial cells, and into the thick bacterial cell body.

      Done. We have added the following to our revised manuscript:

      “Currently, only electron cryo-tomography (cryo-ET) allows visualization of structures in a near-native state inside intact (frozen-hydrated) cells with macromolecular (~5 nm) resolution. However, this capability is limited to thin samples (few hundred nanometers thick, like individual bacterial cells of many species) while thicker samples like the central part of eukaryotic cells, thick bacterial cells, or clusters of bacterial cells are not amenable for direct cryo-ET imaging. Such thick samples can be rendered suitable for cryo-ET experiments by thinning them first using different methods including focused ion beam milling and cryosectioning [30]. Cryo-ET has already been invaluable in revealing the structures of several membrane extensions, including Shewanella oneidensis nanowires [6], Helicobacter pylori tubes [15], Delftia acidovorans nanopods [25], Vibrio vulnificus OMV chains [16], and more recently cell-cell bridges in the archaeon Haloferax volcanii [31].” (Lines 108-118)

      Please provide a legend to Table S1 explaining the numbers (organelles?), how many cells were viewed? I think that at least part of it should be included in the main text. Also, there are examples of vesicles emanating from H. pylori. This information is missing from Table S1.

      Done. We added a column to the table indicating the number of cells available for each species. We also added the information about the vesicles in H. pylori to the table. This table is now incorporated into the main text of the manuscript as Table 1.

      Please provide an ordered list including all the strains (and IDs of the specific isolates) used in this study and their genotypes.

      Done. We added Table S1 to the revised manuscript that contains this information. This table also includes relevant references to all the published papers where these strains were previously used.

      The authors describe in detail the H. pylori tubes that seem to be flagellum-core independent. However, the authors found previously (ref 15) that during infection, these structures are dependent on CagA T4SS, and they visualized T4SS sub-complexes in proximity to the point of tube emanation. This should be described and discussed in the text. Also, please indicate if the "host-independent" tubes are similarly dependent on T4SS.

      Done. We added the following to the revised manuscript:

      “The scaffolded uniform tubes of H. pylori that we observed were formed in samples not incubated with eukaryotic cells, indicating that they can also form in their absence. However, the tubes we found had closed ends and no clear lateral ports, while some of the previously-reported tubes (formed in the presence of eukaryotic host cells) had open ends and prominent ports [15]. It is possible that such features are formed only when H. pylori are in the vicinity of host cells. Moreover, while it was previously hypothesized that the formation of membrane tubes in H. pylori (when they are in the vicinity of eukaryotic cells) is dependent on the cag T4SS [15], we could not identify any clear correlation between the emanation of membrane tubes and cag T4SS particles in our samples where H. pylori was not incubated with host cells. We also show that the tubes of H. pylori are CORE-independent, indicating that they are different from the CORE-dependent nanotubes described in other species.” (Lines 303-313)

      Is there any difference in the frequency or length of the tubes in the mutants presented in Figure 4? The flgS mutant in the image exhibits a very short filament; is that typical?

      We did not see any significant statistical difference in the number or lengths of the tubes in these different mutants. We added Table S2 to the revised manuscript which details the number of cells we visualized for each mutant and the number of the tubes seen there. In all these mutants the lengths of the tubes ranged between few tens to hundreds of nanometers. In addition, we added Fig. S2 to show more examples of these tubes in each of these mutants.

      Minor points:

      -Please check full bacterial names that are sometimes missing (e.g., lines 110-112).

      Done.

      -There is no reference to panel 2G. Please check the references to all panels.

      Done. Please see lines 154 and 183 in the main text.

      -Lines 181-184: There is no figure related to the formation of teardrop-like extensions from C. pinensis. Please review the text accordingly.

      Done. Corrected.

      -Line 235, not clear to what "as these" refers to.

      Done. We modified the text as the following:

      “As these MEs/MVs from S. oneidensis were purified” (Lines 246-247)

      -Line 241, not clear what "a secretin-like complex" is, and no reference is provided.

      Done. We modified the text as the following:

      “In the third category, we observed a secretin-like complex in many tubes and vesicles of F. johnsoniae. Secretins are proteins that form a pore in the outer membrane and are associated with many secretion systems like type IV pili and type II secretion systems (T2SS) [39–41]” (Lines 252-254)

      Reviewer #1 (Significance)

      As described in this manuscript, even in model bacteria these structures are generated (e.g., Caulobacter forms the hardly studied nanopod extensions). The manuscript also provides visual categories of these structures, defining "extension types" that are likely to be used by the scientific community for years to come, similar to the initial pili classification during the 1960s-70s. It is a "descriptive study," in the positive sense of the term, as it significantly contributes to the field of bacteriology.

      We thank the reviewer for her/his kind words and enthusiasm about our work. It is an honor to have our work compared to the seminal pili classification work done in the 1960s-70s by pioneers in the field of bacteriology.

      Reviewer #2:

      The manuscript "In situ imaging of bacterial membrane projections and associated protein complexes using electron cryo-tomography" by Kaplan et al., identifies and catalogues membrane extensions (MEs) and membrane vesicles (MVs) from 13 different species using cryo-electron tomography. Furthermore, they identify and discuss several protein complexes observed in these membrane projections.

      The manuscript is beautifully written, interesting, and genuinely got this reviewer excited about the biology. I applaud the authors on their manuscript and have only minor comments and a few thoughts that the authors may wish to think on and discuss.

      We thank the reviewer for her/his kind words and insightful comments that allowed us to improve our manuscript.

      Some schematics throughout the introduction would be useful to readers new to the field/ outside the field who are not used to these different membrane structure features.

      We thank the Reviewer for this suggestion. First, we made an extra figure with schematics showing the cell body and membrane tubes but that was rather redundant with Figure 8. For this reason, we added explicit labels to figure 1 highlighting the cell body and the tubes in these examples to help the reader following that figure and the subsequent ones. However, if the Reviewer has an explicit suggestion/view about the schematics then we would be very happy to do that.

      The size of scale bars should be indicated on the figure panels themselves rather than in the figure legend to assist the reader.

      Done.

      In reference to lines 193-196 - what was the extracellular environment like in these micrographs? Were other cells present? Could it be the extracellular environment/surrounding cells that stimulate pearling? Have the authors considered this? Please discuss if relevant/insightful.

      This is a good point. The cells were usually plunge-frozen in their standard growth media (except in H. pylori where the cells were resuspended in PBS and subsequently plunge-frozen). Yes, there are other cells present in the sample, however, usually, only one cell is present in the field of view of the tomogram as areas with multiple cells have thick ice and therefore not amenable for cryo-ET imaging. We added the following to the revised manuscript:

      “As usually only one (or part of a) cell is present in the cryo-tomogram, we can’t exclude that differences in the extracellular environments, like the presence of a cluster of cells in the vicinity of the individual cells with pearling tubes, might play a role in this observation” (Lines 198-201).

      "Randomly-located complexes" in this reviewers opinion should actually be described "seemingly randomly-located complexes" given there may be an organization present that is beyond the resolution limit of this study.

      The is a good point. Indeed, we can’t exclude that these complexes have a preferred localization in specific lipid patches that we can’t detect in our cryo-tomograms. We added the following statement to the revised manuscript:

      “These complexes, which were also found in the OM of intact cells, did not exhibit a preferred localization or regular arrangement within the tube at least within the fields of view provided by our cryo- tomograms (Fig. 5a & b).” (lines 227-230).

      In reference to lines 287-292 - is it possible this has to do with lipid composition? Have the authors considered this? Please discuss if relevant/insightful.

      Done. We added the following to the revised manuscript:

      “In addition, differences in the lipid compositions among the various species investigated here might also play a role in the formation of these different forms of projections” (Lines 299-301).

      Reviewer #2 (Significance ):

      These results advance the field by shedding new light on bacterial membrane extension morphologies. The authors use a cryo-ET to catalogues membrane extensions and membrane vesicles which has not been done before.

      This paper is likely to be of interest to structural biologists, biophysicist, membrane protein biologists, virologists and microbiologists.

      This reviewer is a single-particle cryo-EM structural biologist with interest in membrane proteins._

      We thank the reviewer for her/his enthusiasm about our work described here.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      The manuscript "In situ imaging of bacterial membrane projections and associated protein complexes using electron cryo-tomography" by Kaplan et al., identifies and catalogues membrane extensions (MEs) and membrane vesicles (MVs) from 13 different species using cryo-electron tomography. Furthermore, they identify and discuss several protein complexes observed in these membrane projections.

      The manuscript is beautifully written, interesting, and genuinely got this reviewer excited about the biology. I applaud the authors on their manuscript and have only minor comments and a few thoughts that the authors may wish to think on and discuss.

      • Some schematics throughout the introduction would be useful to readers new to the field/ outside the field who are not used to these different membrane structure features.
      • The size of scale bars should be indicated on the figure panels themselves rather than in the figure legend to assist the reader.
      • In reference to lines 193-196 - what was the extracellular environment like in these micrographs? Were other cells present? Could it be the extracellular environment/surrounding cells that stimulate pearling? Have the authors considered this? Please discuss if relevant/insightful.
      • "Randomly-located complexes" in this reviewers opinion should actually be described "seemingly randomly-located complexes" given there may be an organization present that is beyond the resolution limit of this study.
      • In reference to lines 287-292 - is it possible this has to do with lipid composition? Have the authors considered this? Please discuss if relevant/insightful.

      Significance

      These results advance the field by shedding new light on bacterial membrane extension morphologies. The authors use a cryo-ET to catalogues membrane extensions and membrane vesicles which has not been done before.

      This paper is likely to be of interest to structural biologists, biophysicist, membrane protein biologists, virologists and microbiologists.

      This reviewer is a single-particle cryo-EM structural biologist with interest in membrane proteins.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      This manuscript by the Jensen lab surveys a plethora of bacterial outer-membrane projections captured over the years by in situ cryo-tomography under near-native conditions. The authors classify the different visualized structures, highlighting both similarities and differences among them. They further describe molecular complexes that are associated with these projections. The manuscript highlights the abundance of such understudied structures in nature, indicating the need to deepen our exploration into their biological functions and mechanisms of action.

      Comments:

      1. The authors should state in the Abstract and Introduction that only diderm bacteria and outer-membrane extensions are included in the study.
      2. In the Introduction or Discussion the authors should mention the limits of the in situ cryo-tomography, such as the difficulty to observe regions in between neigbouring bacterial cells, and into the thick bacterial cell body.
      3. Please provide a legend to Table S1 explaining the numbers (organelles?), how many cells were viewed? I think that at least part of it should be included in the main text. Also, there are examples of vesicles emanating from H. pylori. This information is missing from Table S1.
      4. Please provide an ordered list including all the strains (and IDs of the specific isolates) used in this study and their genotypes.
      5. The authors describe in detail the H. pylori tubes that seem to be flagellum-core independent. However, the authors found previously (ref 15) that during infection, these structures are dependent on CagA T4SS, and they visualized T4SS sub-complexes in proximity to the point of tube emanation. This should be described and discussed in the text. Also, please indicate if the "host-independent" tubes are similarly dependent on T4SS.
      6. Is there any difference in the frequency or length of the tubes in the mutants presented in Figure 4? The flgS mutant in the image exhibits a very short filament; is that typical?

      Minor points:

      • Please check full bacterial names that are sometimes missing (e.g., lines 110-112).
      • There is no reference to panel 2G. Please check the references to all panels.
      • Lines 181-184: There is no figure related to the formation of teardrop-like extensions from C. pinensis. Please review the text accordingly.
      • Line 235, not clear to what "as these" refers to.
      • Line 241, not clear what "a secretin-like complex" is, and no reference is provided.

      Significance

      As described in this manuscript, even in model bacteria these structures are generated (e.g., Caulobacter forms the hardly studied nanopod extensions). The manuscript also provides visual categories of these structures, defining "extension types" that are likely to be used by the scientific community for years to come, similar to the initial pili classification during the 1960s-70s. It is a "descriptive study," in the positive sense of the term, as it significantly contributes to the field of bacteriology.

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

      Learn more at Review Commons


      Reply to the reviewers

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

      In this manuscript, Mishima et al., designed a reporter system (dubbed PACE, for Parallel Analysis of Codon Effects) to assess the effect of codon usage in regulating mRNA stability in a controlled sequence context. This reporter corresponds to a stretch of 20 repetitions of a given codon (to be tested for its effect on mRNA stability), each repetition being separated by one codon corresponding to each of the 20 canonical amino acids. This stretch is inserted at the 3' end of the coding sequence of a superfolder GFP flanked with fixed 5' and 3' untranslated regions. In vitro transcribed capped and polyadenylated RNAs are then produced from these reporters (each with a specific stretch of repetitions of a given codon), pooled together and injected into zebrafish zygotes to monitor their relative abundance at different time points upon injection.

      Using the PACE reporter, the authors were able to obtain a quantitative estimation of the impact of 58 out of the 61 sense codons on modulating mRNA stability. Their results are in agreement with a previous report that estimated the effect of codon usage on mRNA stability using endogenous mRNAs and an ORFeome library (Bazzini et al., 2016). However, contrary to relying on endogenous mRNAs and ORFeome reporters, the advantage of the PACE strategy is that the effect of the codon to be studied can be probed in a defined context, thus avoiding the presence of other motifs or transcript features that could also regulate mRNA stability. Similarly to results from Bazzini et al., 2016, the authors show that blocking translation completely abrogates the effect of codon usage, indicating that translation is required to drive codon-dependent mRNA degradation from their reporters. Also, the extent of codon-dependent mRNA decay is correlated with tRNA abundance and occurs through a process involving mRNA deadenylation as previously described in the zebrafish (Mishima et al., 2016 and Bazzini et al., 2016).

      Having validated their PACE protocol, the authors performed ribosome profiling to test whether ribosome occupancy on tested codons is correlated with their capacity to drive mRNA degradation. Their results indicate that, at least for polar amino acids, there is indeed an inverse correlation between ribosome occupancy at tested codons and mRNA stability thus suggesting that slow decoding of codons due to low levels of available cognate tRNA can induce mRNA degradation. The authors further validate this finding by reducing the levels of aminoacylated tRNAAsn (corresponding a polar amino acid) and showing that stability of the reporter RNA carrying a stretch of AAC codons (decoded by tRNAAsnGUU) is reduced. To test whether codon-dependent mRNA degradation in the context of slow ribosome decoding lead to ribosome stalling and collisions, the authors generated a mutant zebrafish strain with impaired expression of ZNF598 (an essential factor of the No-Go decay (NGD) pathway in yeast). They also integrated a known ribosome stalling sequence from hCMV (and a mutant version that does not trigger ribosome stalling) in their sfGFP reporter construct as a positive control for NGD in their assays. Their results indicate that although ZNF598 depletion impairs degradation of the hCMV reporter (as expected), it does not affect codon-dependent mRNA degradation, which appears to occur for most codons through a NGD-independent manner. Finally, through the use of a tandem ORF reporter assay separated by codon tags to be tested, the authors show that destabilizing codons do not stall ribosomes but only lead to their transient slowdown which induces mRNA deadenylation and degradation in a ZNF598-independent manner.

      Overall, the manuscript is very well written and pleasant to read. The introduction is well documented and relevant to the study as it allows readers to place the study in the current context of the field while highlighting open questions that have not been addressed yet. The results are clearly presented, the technical approaches are elegant and the conclusions convincing.

      Below you will find some major and minor points that, in my opinion, should be addressed by the authors.

      **Major point:**

      • One interesting aspect of the PACE reporter assay is the possibility to monitor ribosome occupancy in parallel for all codon-tags tested, which the authors did in Figure 3. However, instead of using RNA-seq data to normalize ribosome footprints and obtain ribosome occupancy, the authors used an alternative normalization approach consisting, for each codon-tag, to calculate the number of ribosome footprints with test codons in the A site divided by the number of ribosome footprints with spacer codons in the A site. This approach is elegant and appears to work with codons corresponding to polar amino acids. However, it might have its limitations for other codons.

      Indeed, ribosome dwell times (in yeast and mammals) have been shown to respond both to tRNA availability but also to other features such as the nature of the pair of adjacent codons, and the nature of the amino acid within the exit channel (Gobet C et al., 2020 PNAS; Gamble CE et al., 2016 Cell; Pavlov MY et al., 2009 PNAS). However, based on the work of "Buschauer R et al., 2020 Science", only ribosomes lacking an accommodated tRNA at the A site are able to recruit Ccr4-Not to mediate mRNA deadenylation and degradation. Other events that increase ribosome dwell time (and thus occupancy), such as slow peptidyl-transfer, do not lead to Ccr4-Not recruitment and are resolved by eIF5A. It is therefore possible that depending on the nature of the codon that is being tested, ribosome occupancy at test and spacer codons can be biased by the nature of codon-pairs and "dilute" the effects of tRNA availability.

      If the authors performed RNA-seq together with the ribosome profiling experiment, it might be interesting to use the RNA-seq data to calculate ribosome occupancy on "tested" and "spacer" codons to check whether using this normalization, they do find a negative correlation between ribosome occupancy and PACE stability. A different approach would be to perform ribosome run-off experiments using harringtonine and estimate the elongation speed across the codon tag. However, I am aware that this experiment could be tedious an expensive.

      • Figure 6: Insertion of the Lys x8 AAA stretch in the tandem ORF reporter leads to a decrease in HA-DsRedEx expression compared to that of Myc-EGFP. However, results from "Juszkiewicz and Hedge, 2017" using a similar reporter in mammalian cells indicate that stretches of Lys AAA below 20 repetitions only elicit poor RQC (less than 10% of true ribosome stalling for 12 repetitions of the AAA codon). Instead, most of the loss in RFP signal results from a change in the reading frame of ribosomes due to the "slippery" translation of the poly(A) stretch. I therefore think that it could be important to perform the experiment in ZNF598 KO embryos to validate that the observed reduction in HA-dsRedEx does indeed result from stalling and RQC and not from a change in the reading frame of ribosomes.

      On a similar note, how do the authors explain the decrease in signal of the Flag-EGFP and HA-DsRedEx observed when using the Flag-EGFP with non-optimal codons? I understand that RQC occurring through NGD leads to ribosome disassembly at the stalling site and possibly mRNA cleavage (thus explaining the decrease in HA-DsRedEx signal compared to Myc-EGFP). However, I would assume that codon-mediated mRNA decay (even for ORF longer than 200 of non-optimal codons) should trigger mRNA deadenylation, followed by decapping and co-translational 5'to3' mRNA degradation, following the last translating ribosome. I would therefore expect not to see any change in the HA-DsRedEx/Myc-EGFP ratio even for the non-optimal Flag-EGFP reporter. Could the 200 non-optimal codons trigger some background RQC through NGD? Or could there be some ribosome drop-off? It might be interesting to test the optimal and non-optimal Flag-EGFP reporters in the ZNF598 KO background to check whether the observed decrease in the relative amount of HA-DsRedEx results from stalling-dependent RQC.

      **Minor comments:**

      • The color-coded CSC results from "Bazzini et al., 2016" presented at the bottom of panel B in figure 2 are misleading because many codons (such as PheUUU, AsnAAU, TyrUAC...) are lacking information. I have the impression that the authors used the combined data from the rCSCI (obtained from the reporter RNAs) and CSC (obtained from endogenous transcripts) corresponding to Figure 1F from Bazzini et al., 2016. This data set excluded all codons that were not concordant between the endogenous and reporter CSCs (which are those that are lacking a color code in Figure 2B from this study). However, in the scatter-plot of PACE Vs CSC (from Supplemental Figure 1D of this study), the authors used the complete set of CSC values from Bazzini et al .,2016. Could the authors please use the complete set of CSC values from Bazzini et al., 2016 to color code codons in their Figure 2B?

      • Figure 4B. The charged tRNA measurements seem to have been done in a single biological replicate (there aren't any error bars in the chart). I understand that the procedure is tedious and requires a large amount of total RNA to begin with, but it would be preferable to perform it in three biological replicates.

      • Supplementary Figure 2B. I do not understand what the figure represents. The legend is quite cryptic and states that the panel corresponds to the information content of each reading frame. More information should be given so that readers can understand how to interpret de figure and extract periodicity information.

      Reviewer #1 (Significance (Required)):

      Since the seminal work from Jeff Coller's laboratory in 2015 (Presnyak et al., 2015 Cell) showing a global and major role for codon optimality in determining mRNA half-lives in yeast, the role of codon usage in modulating translation and stability of mRNAs has been widely studied in different organisms (including zebrafish and mammals). As stated by the authors in the introduction, most studies have relied on correlation analyses between codon usage and mRNA half-lives from endogenous transcripts or from ORF libraries with fixed 5'UTR and 3'UTRs. This approaches could suffer from the presence of transcript features that can participate in other mRNA degradation pathways, which could limit their use when performing further mechanistic studies.

      The work by Mishima and collaborators presents an original reporter assay that allows to evaluate the role of codon usage on regulating mRNA stability in a defined context, thus avoiding the impact of confounding factors that could bias the measurement of mRNA stability. Results obtained using this reporter are in good agreement with previous reports from Zebrafish (Bazzini et al 2016., and Mishima et al., 2016). From this validated reporter approach, the authors further show that codon-dependent mRNA degradation is directly related to tRNA availability and (at least partially) to ribosome occupancy (two factors already suggested as being important for codon-mediated decay in zebrafish, although they were based on correlation analyses). Furthermore, the authors show that codon-mediated mRNA decay occurs during productive mRNA translation and that it is functionally distinct from RQC induced by ribosome stalling. As a consequence, codon-mediated mRNA degradation is independent from the RQC factor ZNF598 (which they also validate for the first time as an important RQC factor in zebrafish). This information is new within metazoans since only in yeast it has been clearly shown that codon-mediated mRNA decay is distinct from RQC induced by ribosome stalling and collisions.

      Taken together, the reported findings will be of interest to the community working on mRNA metabolism and translation. It could also interest, more broadly, scientists working on translational selection and genome evolution.

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

      In this manuscript, Mishima et al aim to determine if the RNA-mediated decay determined by codon optimality is part of the ribosome quality control pathway, triggered by slowed codon decoding and ribosome stalling or it is an independent pathway.

      To this end, the authors capitalize on their previous work to design a very elegant high-throughput reporter system that can analyze individually codon usage, ribosome occupancy and tRNA abundance. This reporter system, called PACE, is rigorously validated throughout the manuscript, because blocking translation with a morpholino blocking the AUG codon demonstrated that the effects no RNA stability are translation dependent.

      When most of the available codons are tested using the PACE system, the authors recapitulate codon optimality profiles similar to the ones previously uncovered using transcriptome-wide approaches.

      Thanks to the design of the reporter, which alternates repeats of a test codon with random codons, the authors can calculate how quickly a ribosome decodes the test codon on average. With this approach, the authors uncover a negative correlation between RNA stability and ribosome density on codons for polar amino acids and suggest that codon optimality is related to a slower decoding of the codons.

      With the PACE reporter validated, the authors can interrogate the system to gain mechanistic insights of codon optimality. First, they test if RNA decay and deadenylation mediated by codon optimality is determined, in part, by the levels of aminoacylated tRNAs available. The authors use a very elegant approach, as they overexpress a bacterial enzyme (AnsB) in zebrafish that degrades asparagine, effectively reducing the levels of tRNA-Asn. The authors demonstrate that AnsB turns a previously optimal Asn codon, AAC, into a non-optimal one. This effect is translated into RNA destabilization and deadenylation, but this effect in not extended to other codons encoding amino acids not affected by Asn. These results provide a direct experimental validation of the previously published observation of tRNA levels and codon optimality.

      Finally, the authors interrogate the relationship between the codon optimality pathways and the ribosome quality control pathways, that takes care of stalled ribosomes. The authors generate a zebrafish mutant of Znf598, a vertebrate homolog of the yeast protein in charge of resolving stalled ribosomes. Using a maternal-and-zygotic mutant, the authors demonstrate that in these mutant's codon optimality proceeds as usual but ribosome stalling is not resolved, providing evidence for first time that Znf598 is involved in ribosome quality control in vertebrates.

      Altogether, this manuscript presents work that builds on the previous findings of the authors and other labs but it is a qualitative leap forward rather than a marginal increment, because the body of work in the current manuscript i) establishes a reporter to dissect the mechanisms of codon optimality, ii) demonstrates that ribosome slow-down but not stalling is part of the trigger of RNA decay mediated by codon optimality, iii) demonstrates that this pathway is independent of ribosome quality control pathway and finally iv) demonstrates that vertebrate Znf598 is involved in the RNA decay mediated by ribosome stalling.

      Due to these novel findings, and the rigor of the experimental design, this manuscript should be accepted for publication. The authors should first address the following comments:

      **Major comment:**

      1. The authors very elegantly demonstrate the impact of AnsB on the stability of the RNA reporter, and it is precisely the simplicity of the reporter that allows the authors to draw clear conclusions. Nevertheless, it would be interesting to determine if the reporter results in embryos injected with AnsB also translate to endogenous genes rich in AAC codons. The authors could perform a polyA-selected RNA-Seq in embryos treated with AnsB to determine if the transcripts rich in AAC codons are destabilized compared to wild-type, thus validating the reporter results in endogenous genes. **Minor comments:**

      In figure 5J the authors plot the normalized codon tag levels of the PACE reporter run in the MZznf598 mutant. The authors color code the labels in the x-axis following the PACE results in wild-type (figure 2B). The authors should also plot the wild-type values to have a direct visual comparison of the results trend in both genotypes. The authors focus in the title on the role of Znf598 or the lack thereof in RNA decay induced by codon optimality. However, for the non-aficionados in codon-optimality, ZnF598 is an unknown protein and adds little information to the title. The authors should provide a more informative title, directly pinpointing that codon-optimality is independent of the ribosome quality control pathway.

      Reviewer #2 (Significance (Required)):

      This manuscript presents work that builds on the previous findings made by the authors and other laboratories but it is a qualitative leap forward rather than a marginal increment, because the body of work in the current manuscript i) establishes a reporter to dissect the mechanisms of codon optimality, ii) demonstrates that ribosome slow-down but not stalling is part of the trigger of RNA decay mediated by codon optimality, iii) demonstrates that this pathway is independent of ribosome quality control pathway and finally iv) demonstrates that vertebrate Znf598 is involved in the RNA decay mediated by ribosome stalling.

      In addition to the conceptual findings, the authors establish a new high-throughput reporter system to evaluate the influence of codon optimality in RNA decay.

      The work its done in zebrafish embryos, an in vivo model system where codon optimality has been extensively tested by the authors and others, following the stability of reporter and endogenous genes.

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

      Mishima et al. address a very timely topic of how the codon composition of the ORF and the associated translation elongation speed affect mRNA stability. Several studies have already shown a strong correlation between codon optimality and mRNA stability - meaning the more "optimal" the codons, the faster supposedly the elongation speed and the more stable the mRNA. Most of these studies were done by analyzing global expression data, with limited follow up, therefore being also impacted by other co-translational mRNA decay pathways and in addition these studies could also not test directly the effect of each single codon on mRNA stability. The authors took a systematic reporter-based assay approach, called PACE, which allowed them to test systematically the effect of codon composition on mRNA decay. By integrating also ribosome profiling data, the authors could nicely show that the speed of translation (measured by ribosome density) correlates with their determined mRNA stability effect of each codon and also the corresponding tRNA levels. However, interestingly this seems to be the case only for codons encoding polar amino acids, but not the ones that encode charged or non-polar amino acids. It will be very interesting to find out why that is? Finally, the authors address if some of the effects they see might be due to ribosome collisions and associated no go decay (NGD). For this they generated a Znf598 mutant by CRISPR-Cas9. Znf598 is the proposed homolog of Hel2, the protein in yeast that is essential for NGD. The authors go on to show that NGD is defective in this mutant, but that codon mediated decay, which is elongation dependent, is not to a large part not dependent on Znf598.

      **All minor comments:**

      1. It is intriguing why only polar AAs show a tRNA amount specific effect in the ribosome footprint data. Some hypothesis/discussion about this could be expanded further in the discussion or results.
      2. On the same token some additional analysis might be helpful. For example, in Figure 2E, the authors group codons in weak, neutral and strong based on PACE measurements and then look at the tRNA expression range for each of the three groups. Could the authors do this also separately for the codons of polar, non-polar and charged amino acids? What do you see - still the same pattern as for all the codons or do again only polar amino acids show the trend?
      3. Can the authors elaborate on the development of their PACE system? Why is it designed the way it is? What parameters did they test? For example, why the 20 amino acids tail, did you you test shorter sequences of the amino acid, spacer repeats, etc?
      4. The next few questions are a bit more of a technical nature regarding the reporter construct used for PACE.
      5. Did all AA pairs (Codon of interest + spacer codon) behave the same in the footprint assay? Does the data have enough information and resolution for this?
      6. Was the order of the spacer codons always the same in all the constructs? Could the specific order, if it is consistent, have any unseen consequences (ie. interaction with the exit tunnel)? Did the authors test other orders?
      7. Are the spacer codons optimized?
      8. Are the codons affected in the NGD mutant the ones that are most different in the Bazzini data?
      9. The authors inject directly mRNA into the embryos, therefore avoiding that the reporter mRNA is ever in the nucleus. However, there could be nuclear events (e.g. loading of particular proteins) that might affect the fate of an mRNA in the cytosol, among these the translation efficiency and also stability. Maybe some comment in the discussion as to the effect of missing nuclear factors would be welcome. This is not a criticism; it would just be nice to hear the authors' thoughts on that.
      10. Page 6; final paragraph: "Finally, we compared the speed of the ribosome translating mRNA destabilizing codons to that of an aberrantly stalled ribosome." Not sure the authors did that actually. They tested the effect of ribosome slowing down on protein production and mRNA levels and compared that to stalling ribosomes, but did not compare the "speed" directly and I am not even sure what they mean by that in this context. Probably good to rephrase.

      Page 7, upper half: ".....by taking the positional effect of codon-mediated decay into account (Mishima and Tomari, 2016)."

      This is my limited knowledge of the literature, but I think you should mention what this positional effect is and not just cite a paper.

      Very minor, but on page 8 when PACE is introduced, the authors show the different destabilizing effects of the three Ile codons. While that is ok, in the section before, when the authors tested their construct by qRT-PCR, they focused on the two Leu codons. I would also mention them here and do a direct comparison of the qRT-PCR results with the pooled PACE result for these two codons. Based on the figure the two codons seem to behave qualitatively like expected, but I am not sure how good the quantitative behavior matches. The AnsB experiment - the authors only mention data about one of the two Asn codons (AAC), but what about the second Asn codon (AAU) - do you also see an effect on that codon upon overexpression of AnsB as well? AAU is already a quite destabilizing codon and you might not see a further increase in destabilization, but it would be great to know if there was or not. Page 13, second paragraph: More out of interest, but it is quite intriguing that GCG turned into a destabilizing codon (opposite of what one would expect if NGD would play a bit a role). Any speculation why? Page 14, end of page and related to Figure 6C: AAU seems much more destabilizing than AAC. Therefore, I would have expected that the inserted sequence with the AAU codons would lead actually to downregulation of the mRNA and therefore the EGFP and DsRFP total protein signal relative to the construct with the AAC inserted in between, even if the ratio of EGFP/DsRed seems unchanged. However, based on the western blot in 6C the total protein levels seem very similar. Isn't that surprising? Although, AAU obviously allows translation to proceed it should still induce a stronger mRNA decay than AAC and therefore result in less total mRNA (and protein level as a consequence). Did the authors quantify the exact levels of the reporter proteins and mRNA and compared them between the two constructs? Page 15, last sentence: Somehow for me the word "transient" is a bit hard to grasp in this context. What do you mean by that - do you really mean "impermanent" or "lasting only for a short amount of time"? Don't you simply mean "weaker", "less strong"? Page 17, second sentence: I think the authors want to reference here Figure 2E and not Figure 2D.

      Reviewer #3 (Significance (Required)):

      All in all, I have to say that it was a real pleasure to read this manuscript. The authors were extremely thorough with their experiments and did nearly never overstate any of their conclusions. It is a very insightful story, which in my opinion will contribute greatly to the field of gene expression and posttranscriptional gene expression regulation in particular. The PACE assay, although a bit artificial, gave very clean results, which agree with the previous literature and could be very useful for future studies. Generating the Znf598 mutant and showing that the codon-dependent decay is independent from NGD is a great addition to this paper. Although it is a bit of a pity that we do not see more of a characterization of the Znf598 mutant in this paper, I do agree with the authors that this might take away a bit of the focus of this manuscript and that the mutant deserves actually its own story. I only have very minor comments/questions for the authors that they should be able to address easily. Finally, I can only repeat myself by saying: congrats on this great paper and I fully support publication.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Mishima et al. address a very timely topic of how the codon composition of the ORF and the associated translation elongation speed affect mRNA stability. Several studies have already shown a strong correlation between codon optimality and mRNA stability - meaning the more "optimal" the codons, the faster supposedly the elongation speed and the more stable the mRNA. Most of these studies were done by analyzing global expression data, with limited follow up, therefore being also impacted by other co-translational mRNA decay pathways and in addition these studies could also not test directly the effect of each single codon on mRNA stability. The authors took a systematic reporter-based assay approach, called PACE, which allowed them to test systematically the effect of codon composition on mRNA decay. By integrating also ribosome profiling data, the authors could nicely show that the speed of translation (measured by ribosome density) correlates with their determined mRNA stability effect of each codon and also the corresponding tRNA levels. However, interestingly this seems to be the case only for codons encoding polar amino acids, but not the ones that encode charged or non-polar amino acids. It will be very interesting to find out why that is? Finally, the authors address if some of the effects they see might be due to ribosome collisions and associated no go decay (NGD). For this they generated a Znf598 mutant by CRISPR-Cas9. Znf598 is the proposed homolog of Hel2, the protein in yeast that is essential for NGD. The authors go on to show that NGD is defective in this mutant, but that codon mediated decay, which is elongation dependent, is not to a large part not dependent on Znf598.

      All minor comments:

      1. It is intriguing why only polar AAs show a tRNA amount specific effect in the ribosome footprint data. Some hypothesis/discussion about this could be expanded further in the discussion or results.
      2. On the same token some additional analysis might be helpful. For example, in Figure 2E, the authors group codons in weak, neutral and strong based on PACE measurements and then look at the tRNA expression range for each of the three groups. Could the authors do this also separately for the codons of polar, non-polar and charged amino acids? What do you see - still the same pattern as for all the codons or do again only polar amino acids show the trend?
      3. Can the authors elaborate on the development of their PACE system? Why is it designed the way it is? What parameters did they test? For example, why the 20 amino acids tail, did you you test shorter sequences of the amino acid, spacer repeats, etc?
      4. The next few questions are a bit more of a technical nature regarding the reporter construct used for PACE. a. Did all AA pairs (Codon of interest + spacer codon) behave the same in the footprint assay? Does the data have enough information and resolution for this? b. Was the order of the spacer codons always the same in all the constructs? Could the specific order, if it is consistent, have any unseen consequences (ie. interaction with the exit tunnel)? Did the authors test other orders? c. Are the spacer codons optimized?
      5. Are the codons affected in the NGD mutant the ones that are most different in the Bazzini data?
      6. The authors inject directly mRNA into the embryos, therefore avoiding that the reporter mRNA is ever in the nucleus. However, there could be nuclear events (e.g. loading of particular proteins) that might affect the fate of an mRNA in the cytosol, among these the translation efficiency and also stability. Maybe some comment in the discussion as to the effect of missing nuclear factors would be welcome. This is not a criticism; it would just be nice to hear the authors' thoughts on that.
      7. Page 6; final paragraph: "Finally, we compared the speed of the ribosome translating mRNA destabilizing codons to that of an aberrantly stalled ribosome." Not sure the authors did that actually. They tested the effect of ribosome slowing down on protein production and mRNA levels and compared that to stalling ribosomes, but did not compare the "speed" directly and I am not even sure what they mean by that in this context. Probably good to rephrase.
      8. Page 7, upper half: ".....by taking the positional effect of codon-mediated decay into account (Mishima and Tomari, 2016)." This is my limited knowledge of the literature, but I think you should mention what this positional effect is and not just cite a paper.
      9. Very minor, but on page 8 when PACE is introduced, the authors show the different destabilizing effects of the three Ile codons. While that is ok, in the section before, when the authors tested their construct by qRT-PCR, they focused on the two Leu codons. I would also mention them here and do a direct comparison of the qRT-PCR results with the pooled PACE result for these two codons. Based on the figure the two codons seem to behave qualitatively like expected, but I am not sure how good the quantitative behavior matches.
      10. The AnsB experiment - the authors only mention data about one of the two Asn codons (AAC), but what about the second Asn codon (AAU) - do you also see an effect on that codon upon overexpression of AnsB as well? AAU is already a quite destabilizing codon and you might not see a further increase in destabilization, but it would be great to know if there was or not.
      11. Page 13, second paragraph: More out of interest, but it is quite intriguing that GCG turned into a destabilizing codon (opposite of what one would expect if NGD would play a bit a role). Any speculation why?
      12. Page 14, end of page and related to Figure 6C: AAU seems much more destabilizing than AAC. Therefore, I would have expected that the inserted sequence with the AAU codons would lead actually to downregulation of the mRNA and therefore the EGFP and DsRFP total protein signal relative to the construct with the AAC inserted in between, even if the ratio of EGFP/DsRed seems unchanged. However, based on the western blot in 6C the total protein levels seem very similar. Isn't that surprising? Although, AAU obviously allows translation to proceed it should still induce a stronger mRNA decay than AAC and therefore result in less total mRNA (and protein level as a consequence). Did the authors quantify the exact levels of the reporter proteins and mRNA and compared them between the two constructs?
      13. Page 15, last sentence: Somehow for me the word "transient" is a bit hard to grasp in this context. What do you mean by that - do you really mean "impermanent" or "lasting only for a short amount of time"? Don't you simply mean "weaker", "less strong"?
      14. Page 17, second sentence: I think the authors want to reference here Figure 2E and not Figure 2D.

      Significance

      All in all, I have to say that it was a real pleasure to read this manuscript. The authors were extremely thorough with their experiments and did nearly never overstate any of their conclusions. It is a very insightful story, which in my opinion will contribute greatly to the field of gene expression and posttranscriptional gene expression regulation in particular. The PACE assay, although a bit artificial, gave very clean results, which agree with the previous literature and could be very useful for future studies. Generating the Znf598 mutant and showing that the codon-dependent decay is independent from NGD is a great addition to this paper. Although it is a bit of a pity that we do not see more of a characterization of the Znf598 mutant in this paper, I do agree with the authors that this might take away a bit of the focus of this manuscript and that the mutant deserves actually its own story. I only have very minor comments/questions for the authors that they should be able to address easily. Finally, I can only repeat myself by saying: congrats on this great paper and I fully support publication.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this manuscript, Mishima et al aim to determine if the RNA-mediated decay determined by codon optimality is part of the ribosome quality control pathway, triggered by slowed codon decoding and ribosome stalling or it is an independent pathway.

      To this end, the authors capitalize on their previous work to design a very elegant high-throughput reporter system that can analyze individually codon usage, ribosome occupancy and tRNA abundance. This reporter system, called PACE, is rigorously validated throughout the manuscript, because blocking translation with a morpholino blocking the AUG codon demonstrated that the effects no RNA stability are translation dependent.

      When most of the available codons are tested using the PACE system, the authors recapitulate codon optimality profiles similar to the ones previously uncovered using transcriptome-wide approaches.

      Thanks to the design of the reporter, which alternates repeats of a test codon with random codons, the authors can calculate how quickly a ribosome decodes the test codon on average. With this approach, the authors uncover a negative correlation between RNA stability and ribosome density on codons for polar amino acids and suggest that codon optimality is related to a slower decoding of the codons.

      With the PACE reporter validated, the authors can interrogate the system to gain mechanistic insights of codon optimality. First, they test if RNA decay and deadenylation mediated by codon optimality is determined, in part, by the levels of aminoacylated tRNAs available. The authors use a very elegant approach, as they overexpress a bacterial enzyme (AnsB) in zebrafish that degrades asparagine, effectively reducing the levels of tRNA-Asn. The authors demonstrate that AnsB turns a previously optimal Asn codon, AAC, into a non-optimal one. This effect is translated into RNA destabilization and deadenylation, but this effect in not extended to other codons encoding amino acids not affected by Asn. These results provide a direct experimental validation of the previously published observation of tRNA levels and codon optimality.

      Finally, the authors interrogate the relationship between the codon optimality pathways and the ribosome quality control pathways, that takes care of stalled ribosomes. The authors generate a zebrafish mutant of Znf598, a vertebrate homolog of the yeast protein in charge of resolving stalled ribosomes. Using a maternal-and-zygotic mutant, the authors demonstrate that in these mutant's codon optimality proceeds as usual but ribosome stalling is not resolved, providing evidence for first time that Znf598 is involved in ribosome quality control in vertebrates.

      Altogether, this manuscript presents work that builds on the previous findings of the authors and other labs but it is a qualitative leap forward rather than a marginal increment, because the body of work in the current manuscript i) establishes a reporter to dissect the mechanisms of codon optimality, ii) demonstrates that ribosome slow-down but not stalling is part of the trigger of RNA decay mediated by codon optimality, iii) demonstrates that this pathway is independent of ribosome quality control pathway and finally iv) demonstrates that vertebrate Znf598 is involved in the RNA decay mediated by ribosome stalling.

      Due to these novel findings, and the rigor of the experimental design, this manuscript should be accepted for publication. The authors should first address the following comments:

      Major comment:

      1. The authors very elegantly demonstrate the impact of AnsB on the stability of the RNA reporter, and it is precisely the simplicity of the reporter that allows the authors to draw clear conclusions. Nevertheless, it would be interesting to determine if the reporter results in embryos injected with AnsB also translate to endogenous genes rich in AAC codons. The authors could perform a polyA-selected RNA-Seq in embryos treated with AnsB to determine if the transcripts rich in AAC codons are destabilized compared to wild-type, thus validating the reporter results in endogenous genes.

      Minor comments:

      1. In figure 5J the authors plot the normalized codon tag levels of the PACE reporter run in the MZznf598 mutant. The authors color code the labels in the x-axis following the PACE results in wild-type (figure 2B). The authors should also plot the wild-type values to have a direct visual comparison of the results trend in both genotypes.
      2. The authors focus in the title on the role of Znf598 or the lack thereof in RNA decay induced by codon optimality. However, for the non-aficionados in codon-optimality, ZnF598 is an unknown protein and adds little information to the title. The authors should provide a more informative title, directly pinpointing that codon-optimality is independent of the ribosome quality control pathway.

      Significance

      This manuscript presents work that builds on the previous findings made by the authors and other laboratories but it is a qualitative leap forward rather than a marginal increment, because the body of work in the current manuscript i) establishes a reporter to dissect the mechanisms of codon optimality, ii) demonstrates that ribosome slow-down but not stalling is part of the trigger of RNA decay mediated by codon optimality, iii) demonstrates that this pathway is independent of ribosome quality control pathway and finally iv) demonstrates that vertebrate Znf598 is involved in the RNA decay mediated by ribosome stalling.

      In addition to the conceptual findings, the authors establish a new high-throughput reporter system to evaluate the influence of codon optimality in RNA decay.

      The work its done in zebrafish embryos, an in vivo model system where codon optimality has been extensively tested by the authors and others, following the stability of reporter and endogenous genes.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      In this manuscript, Mishima et al., designed a reporter system (dubbed PACE, for Parallel Analysis of Codon Effects) to assess the effect of codon usage in regulating mRNA stability in a controlled sequence context. This reporter corresponds to a stretch of 20 repetitions of a given codon (to be tested for its effect on mRNA stability), each repetition being separated by one codon corresponding to each of the 20 canonical amino acids. This stretch is inserted at the 3' end of the coding sequence of a superfolder GFP flanked with fixed 5' and 3' untranslated regions. In vitro transcribed capped and polyadenylated RNAs are then produced from these reporters (each with a specific stretch of repetitions of a given codon), pooled together and injected into zebrafish zygotes to monitor their relative abundance at different time points upon injection.

      Using the PACE reporter, the authors were able to obtain a quantitative estimation of the impact of 58 out of the 61 sense codons on modulating mRNA stability. Their results are in agreement with a previous report that estimated the effect of codon usage on mRNA stability using endogenous mRNAs and an ORFeome library (Bazzini et al., 2016). However, contrary to relying on endogenous mRNAs and ORFeome reporters, the advantage of the PACE strategy is that the effect of the codon to be studied can be probed in a defined context, thus avoiding the presence of other motifs or transcript features that could also regulate mRNA stability. Similarly to results from Bazzini et al., 2016, the authors show that blocking translation completely abrogates the effect of codon usage, indicating that translation is required to drive codon-dependent mRNA degradation from their reporters. Also, the extent of codon-dependent mRNA decay is correlated with tRNA abundance and occurs through a process involving mRNA deadenylation as previously described in the zebrafish (Mishima et al., 2016 and Bazzini et al., 2016). Having validated their PACE protocol, the authors performed ribosome profiling to test whether ribosome occupancy on tested codons is correlated with their capacity to drive mRNA degradation. Their results indicate that, at least for polar amino acids, there is indeed an inverse correlation between ribosome occupancy at tested codons and mRNA stability thus suggesting that slow decoding of codons due to low levels of available cognate tRNA can induce mRNA degradation. The authors further validate this finding by reducing the levels of aminoacylated tRNAAsn (corresponding a polar amino acid) and showing that stability of the reporter RNA carrying a stretch of AAC codons (decoded by tRNAAsnGUU) is reduced. To test whether codon-dependent mRNA degradation in the context of slow ribosome decoding lead to ribosome stalling and collisions, the authors generated a mutant zebrafish strain with impaired expression of ZNF598 (an essential factor of the No-Go decay (NGD) pathway in yeast). They also integrated a known ribosome stalling sequence from hCMV (and a mutant version that does not trigger ribosome stalling) in their sfGFP reporter construct as a positive control for NGD in their assays. Their results indicate that although ZNF598 depletion impairs degradation of the hCMV reporter (as expected), it does not affect codon-dependent mRNA degradation, which appears to occur for most codons through a NGD-independent manner. Finally, through the use of a tandem ORF reporter assay separated by codon tags to be tested, the authors show that destabilizing codons do not stall ribosomes but only lead to their transient slowdown which induces mRNA deadenylation and degradation in a ZNF598-independent manner.

      Overall, the manuscript is very well written and pleasant to read. The introduction is well documented and relevant to the study as it allows readers to place the study in the current context of the field while highlighting open questions that have not been addressed yet. The results are clearly presented, the technical approaches are elegant and the conclusions convincing.

      Below you will find some major and minor points that, in my opinion, should be addressed by the authors.

      Major point:

      • One interesting aspect of the PACE reporter assay is the possibility to monitor ribosome occupancy in parallel for all codon-tags tested, which the authors did in Figure 3. However, instead of using RNA-seq data to normalize ribosome footprints and obtain ribosome occupancy, the authors used an alternative normalization approach consisting, for each codon-tag, to calculate the number of ribosome footprints with test codons in the A site divided by the number of ribosome footprints with spacer codons in the A site. This approach is elegant and appears to work with codons corresponding to polar amino acids. However, it might have its limitations for other codons.

      Indeed, ribosome dwell times (in yeast and mammals) have been shown to respond both to tRNA availability but also to other features such as the nature of the pair of adjacent codons, and the nature of the amino acid within the exit channel (Gobet C et al., 2020 PNAS; Gamble CE et al., 2016 Cell; Pavlov MY et al., 2009 PNAS). However, based on the work of "Buschauer R et al., 2020 Science", only ribosomes lacking an accommodated tRNA at the A site are able to recruit Ccr4-Not to mediate mRNA deadenylation and degradation. Other events that increase ribosome dwell time (and thus occupancy), such as slow peptidyl-transfer, do not lead to Ccr4-Not recruitment and are resolved by eIF5A. It is therefore possible that depending on the nature of the codon that is being tested, ribosome occupancy at test and spacer codons can be biased by the nature of codon-pairs and "dilute" the effects of tRNA availability.

      If the authors performed RNA-seq together with the ribosome profiling experiment, it might be interesting to use the RNA-seq data to calculate ribosome occupancy on "tested" and "spacer" codons to check whether using this normalization, they do find a negative correlation between ribosome occupancy and PACE stability. A different approach would be to perform ribosome run-off experiments using harringtonine and estimate the elongation speed across the codon tag. However, I am aware that this experiment could be tedious an expensive.

      • Figure 6: Insertion of the Lys x8 AAA stretch in the tandem ORF reporter leads to a decrease in HA-DsRedEx expression compared to that of Myc-EGFP. However, results from "Juszkiewicz and Hedge, 2017" using a similar reporter in mammalian cells indicate that stretches of Lys AAA below 20 repetitions only elicit poor RQC (less than 10% of true ribosome stalling for 12 repetitions of the AAA codon). Instead, most of the loss in RFP signal results from a change in the reading frame of ribosomes due to the "slippery" translation of the poly(A) stretch. I therefore think that it could be important to perform the experiment in ZNF598 KO embryos to validate that the observed reduction in HA-dsRedEx does indeed result from stalling and RQC and not from a change in the reading frame of ribosomes. On a similar note, how do the authors explain the decrease in signal of the Flag-EGFP and HA-DsRedEx observed when using the Flag-EGFP with non-optimal codons? I understand that RQC occurring through NGD leads to ribosome disassembly at the stalling site and possibly mRNA cleavage (thus explaining the decrease in HA-DsRedEx signal compared to Myc-EGFP). However, I would assume that codon-mediated mRNA decay (even for ORF longer than 200 of non-optimal codons) should trigger mRNA deadenylation, followed by decapping and co-translational 5'to3' mRNA degradation, following the last translating ribosome. I would therefore expect not to see any change in the HA-DsRedEx/Myc-EGFP ratio even for the non-optimal Flag-EGFP reporter. Could the 200 non-optimal codons trigger some background RQC through NGD? Or could there be some ribosome drop-off? It might be interesting to test the optimal and non-optimal Flag-EGFP reporters in the ZNF598 KO background to check whether the observed decrease in the relative amount of HA-DsRedEx results from stalling-dependent RQC.

      Minor comments:

      • The color-coded CSC results from "Bazzini et al., 2016" presented at the bottom of panel B in figure 2 are misleading because many codons (such as PheUUU, AsnAAU, TyrUAC...) are lacking information. I have the impression that the authors used the combined data from the rCSCI (obtained from the reporter RNAs) and CSC (obtained from endogenous transcripts) corresponding to Figure 1F from Bazzini et al., 2016. This data set excluded all codons that were not concordant between the endogenous and reporter CSCs (which are those that are lacking a color code in Figure 2B from this study). However, in the scatter-plot of PACE Vs CSC (from Supplemental Figure 1D of this study), the authors used the complete set of CSC values from Bazzini et al .,2016. Could the authors please use the complete set of CSC values from Bazzini et al., 2016 to color code codons in their Figure 2B?
      • Figure 4B. The charged tRNA measurements seem to have been done in a single biological replicate (there aren't any error bars in the chart). I understand that the procedure is tedious and requires a large amount of total RNA to begin with, but it would be preferable to perform it in three biological replicates.
      • Supplementary Figure 2B. I do not understand what the figure represents. The legend is quite cryptic and states that the panel corresponds to the information content of each reading frame. More information should be given so that readers can understand how to interpret de figure and extract periodicity information.

      Significance

      Since the seminal work from Jeff Coller's laboratory in 2015 (Presnyak et al., 2015 Cell) showing a global and major role for codon optimality in determining mRNA half-lives in yeast, the role of codon usage in modulating translation and stability of mRNAs has been widely studied in different organisms (including zebrafish and mammals). As stated by the authors in the introduction, most studies have relied on correlation analyses between codon usage and mRNA half-lives from endogenous transcripts or from ORF libraries with fixed 5'UTR and 3'UTRs. This approaches could suffer from the presence of transcript features that can participate in other mRNA degradation pathways, which could limit their use when performing further mechanistic studies.

      The work by Mishima and collaborators presents an original reporter assay that allows to evaluate the role of codon usage on regulating mRNA stability in a defined context, thus avoiding the impact of confounding factors that could bias the measurement of mRNA stability. Results obtained using this reporter are in good agreement with previous reports from Zebrafish (Bazzini et al 2016., and Mishima et al., 2016). From this validated reporter approach, the authors further show that codon-dependent mRNA degradation is directly related to tRNA availability and (at least partially) to ribosome occupancy (two factors already suggested as being important for codon-mediated decay in zebrafish, although they were based on correlation analyses). Furthermore, the authors show that codon-mediated mRNA decay occurs during productive mRNA translation and that it is functionally distinct from RQC induced by ribosome stalling. As a consequence, codon-mediated mRNA degradation is independent from the RQC factor ZNF598 (which they also validate for the first time as an important RQC factor in zebrafish). This information is new within metazoans since only in yeast it has been clearly shown that codon-mediated mRNA decay is distinct from RQC induced by ribosome stalling and collisions.

      Taken together, the reported findings will be of interest to the community working on mRNA metabolism and translation. It could also interest, more broadly, scientists working on translational selection and genome evolution.

  2. Jul 2021
    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      We thank the reviewers for their thoughtful comments. We were delighted the reviewers found our results “compelling”, “striking”, “well presented”, “implications exciting”, “excellent results! really nice!”, “this microscopy is beautiful!” and “translational-dependence (of mRNA localization) in a transcript-specific way without perturbing translation globally”, which is a “complete surprise, and opens exciting doors to investigate how translation leads to mRNA organization and its connection to **tissue development” and “may represent a new pathway of mRNA transport”.

      We also appreciated the comments regarding the “wide appeal”, “broad readership of readers”, and “broad interest” the reviewers gave to our manuscript regarding its impact, and also the comments of “well-written (and) well-cited”.

      We can address all the concerns raised by the reviewers. In addition to textual changes, we will add the following to the Results section:

      1. Additional quantitation of smFISH beyond Figure 2;
      2. Addition of a negative (uniformly distributed) mRNA control and its quantitation;
      3. Western blots for our ΔATG lines to determine what and how much protein is made.
      4. Unbiased nuclear masking. Our specific responses are shown below, in blue.

      Reviewer #1

      **Major comments**

      Fig. 1: Main and supplementary figures present smFISH signals for eight localized mRNAs, while in the results section authors describe that they analyzed twenty-five transcripts. Authors should explain the choice of transcripts presented in the paper.

      We will include a panel in Fig. S1E to show every mRNA that we tested, and we will edit Table 1 to describe the observed subcellular localization.

      We will edit the text, adding a few sentences to clarify, along the lines of: “O**ur survey revealed mRNAs with varying degrees of localization within epithelia that we divided into three classes: CeAJ/membrane localized, perinuclearly localized, and unlocalized (Fig. 1 and S1 and Table 1).” and “The rest of our tested mRNAs did not possess any evident subcellular localization at any of the analyzed embryonic stages/tissues and were not further investigated (Fig. S1E and Table 1).

      Moreover, smFISH signal of different localized mRNAs in epidermal cells was visualized at different stages (bean, comma or late comma), and authors did not comment what was the reason of such conditions. This may make transcripts localization results difficult to interpret, as further analysis showed that mRNA localization varied in a stage-specific manner.

      We have clarified this point now in Figure legend 1: “Specific embryonic stages were selected for each transcript based on the highest degree of mRNA localization they exhibited.

      Did author used smFISH probes designed against endogenous mRNAs for all tested transcripts?

      We did not. We clarify this point now in Materials and methods: “All probes were designed against the endogenous mRNA sequences except dlg-1 (some constructs), pkc-3, hmp-2, spc-1, let-805, and vab-10a, whose mRNA were detected with gfp probes in their corresponding transgenic lines (Table S2). An exception to this is Fig. S1A where we used probes against the endogenous dlg-1 mRNA.”.

      Marking dlg-1 mRNA as dlg-1-gfp suggests that smFISH probe was specific for gfp transcript. Is it true? If yes, authors should compare localization of wild-type endogenous dlg-1 mRNA with that of the transcript encoding a fusion protein, to confirm that fusion does not affect mRNA localization.

      Yes, in Fig. 1C we show smFISH for GFP (i.e., the tagged dlg-1 only). In Fig. S1A, we show smFISH against endogenous dlg-1. Tagged and endogenous dlg-1 mRNAs are both localized. We clarified this point in the main text: “Five of these transcripts were enriched at specific loci at or near the cell membrane: laterally and at the CeAJ for dlg-1 (Fig. 1C for endogenous/GFP CRISPR-tagged dlg-1::gfp mRNA and S1A for endogenous/non-tagged dlg-1 mRNA), (…)”. And in the Supplemental figure legend (Fig. S1A): “Endogenous/non-tagged dlg-1 mRNA shows CeAJ/membrane localization like its endogenous/GFP CRISPR-tagged counterpart.

      Fig. 2B: Authors conclude that at later stages of pharyngeal morphogenesis mRNA enrichment at the CeAJ decreased gradually in comparison to comma stage. Data do not show statistically significant decrease in ratio of localized mRNAs - for dlg-1: bean: 0.39{plus minus}0.09, comma: 0.29{plus minus}0.08, 1.5-fold: 0.30{plus minus}0.09; for ajm-1: bean: 0.36{plus minus}0.08, comma: 0.30{plus minus}0.05, 1.5-fold: 0.28{plus minus}0.09.

      t-test (one-tailed) analysis revealed a significant difference between bean and comma stages for both dlg-1 and ajm-1 mRNAs. Statistical analysis and data will be provided.

      Fig. 4: What was the difference between the first and the second __ΔATG transgenic line? Authors should analyze the size of the truncated DLG-1 protein that is expressed from the second Δ__ATG transgenic line that localizes to CeAJ. Knowing alternative ATGs and protein size may suggest domain composition of the truncated protein. This will allow to confront truncated protein localization with the results from.

      We will perform a Western blot to determine the size and levels of proteins produced.

      Fig. 5. Moreover, to prove that the localization of dlg-1 mRNA at the CeAJ is translation-dependent, additional experiment should be performed where transcripts localization will be analyzed in embryos treated with translation inhibitors such as cycloheximide (translation elongation inhibitor) and puromycin (that induces premature termination).

      We believe this comment might refer to Fig. 4. If this is the case: drugs like cycloheximide and puromycin affect the translation of the whole transcriptome, whereas with our ΔATG experiment, we aimed to target the translation of one specific transcript and avoid secondary effects. Nevertheless, we understand Reviewer #1’s concern and will include a second experiment. In our hands, cycloheximide and puromycin have never worked in older embryos (it’s hard to get past the eggshell and into the embryo). Instead, we will use stress conditions, which induce a “ribosome drop-off” (Spriggs et al., 2010). Heat stress has been shown to decrease polysome occupancy (Arnold et al., 2014). We, therefore, have used heat-shock at 33°C for 30’, and the results are now shown in Fig. S4. These show the loss of RNA localization upon heat shock.

      **Minor comments**

      In the introduction section authors should emphasize the main goal and scientific significance of the paper.

      We added this sentence to state the significance before summarizing the results: “To investigate the impact of mRNA localization during embryonic development, we conducted a single molecule fluorescence in situ hybridization (smFISH)-based survey (…)” and “Our data demonstrate that the dlg-1 UTRs are dispensable, whereas translation is required for localization, therefore providing an example of a translation-dependent mechanism for mRNA delivery in C. elegans.” To state the significance.

      Fig 1A: It's hard to distinguish different colors on the schematics. Schematics presents intermediate filaments that are not included in the Table 1.

      We modified Table 1 based on this and other reviewers’ comments.

      Fig. 1C: dlg-1 transcript is marked as dlg-1-gfp on the left panel and dlg-1 on the right panel.

      Corrected.

      Fig. 2B: Axis labels and titles are not visible, larger font size should be used.

      We will modify the graph (following Reviewer #2’s suggestion) and axes label and title sizes will be taken into account.

      Fig. 5C: Enlarge the font size.

      Will do.

      Fig. S2: Embryonic stages should be marked on the figure for easier interpretation.

      Added.

      Reviewer #2

      Major comments

      Figure 2 requires a negative (or uniformly distributed) mRNA control for comparison. Figure 2C should be quantified. The plot quality should be improved, and appropriate statistical tests should be employed to strengthen the claimed findings.

      We will add a negative control (jac-1 mRNA), and quantify Fig. 2C as well. Plots will be changed accordingly to the suggestion.

      Most claims of perinuclear mRNA localization are difficult to see and not well supported visually or statistically. The usage of DAPI markers, membrane markers, 3D rendering, or a quantified metric would bolster this claim. Also, sax-7 is claimed to be perinuclear and elsewhere claimed to be uniform then used as a uniform control. Please explain or resolve these discrepancies more clearly.__

      Regarding perinuclear mRNAs:

      We are not trying to make a big statement out of these data as perinuclear (ER) localization of mRNAs coding for transmembrane/secreted proteins is well known. The aim of our study was to describe transcript localized at or in the proximity of the junction. However, we thought it was worth mentioning these examples of perinuclearly localized mRNAs (hmr-1, sax-7, and eat-20) for two reasons: scientific correctness – show accessory results that might be interesting for other scientists – and use as positive controls for our smFISH survey – these mRNAs were expected to localize perinuclearly for the reasons mentioned above. We will rewrite the text to make these points clearer.

      Regarding sax-7 mRNA:

      sax-7 mRNA localizes perinuclearly in sporadic instances (Fig S1C), but it is predominantly scattered throughout the cytoplasm (i.e., unlocalized). It presumably localizes perinuclearly in a translation-dependent manner as sax-7 codes for a transmembrane protein that would be targeted to the ER. We have described this ER-type of localization in the introduction and reiterated it partially in the first paragraph of the results. sax-7 UTRs are therefore presumably not responsible for subcellular localization, which would instead depend on a signal sequence. We will better clarify this point in the main text.

      The major concern about the paper is the data display and interpretation of Figure 5C. I'm not comfortable with the approach the authors took of blurring out the nucleus. A more faithful practice would be to use an automated mask over DAPI staining or to quantify the entirety of the cell. If the entirety of the cell were quantified, one could still focus analysis on specific regions of relevance. The interpretations distinguishing membrane versus cytoplasmic localization (or mislocalization) are hard to differentiate in these images especially since they are lacking a membrane marker. The ability to make these distinctions forms the basis of Tocchini et al's two pathways of dlg-1 mRNA localization. These interpretations also heavily rely on how the image was processed through the different Z-stacks, and it's not clear to me how that was done. For example, the diffusion of mRNA in figure 5F and 5I are indistinguishable to my eye but are claimed to be different.

      In the images, the nuclei have been blurred to allow the reader to focus on the cytoplasmic signal and not on the nuclear (transcriptional) signal as it is not meaningful for this study. In the quantitation, the nuclear signal has been unbiasedly and specifically removed from the analysis by cropping out the DNA signal from the other channels. The frontal plane views of the seam cells in Fig. 5 show maximum intensity projections (MIPs) of 3 Z-stacks (0.54 µm total) that each contain nuclei and, therefore, the transcriptional signal (schematics in Fig. 5B). We will clarify these points in the text.

      Regarding cytoplasmic versus membrane-associated mRNAs, although we did not have a membrane marker, we relied on the brightness of the DLG-1::GFP signal to identify the cell borders (i.e., membranes) after over-exposure. This approach allowed us to discern apicobasal and apical sides for the intensity profile analyses. We will clarify this point as well in the text and, in parallel, we will try a different approach using transverse sections on top views to clarify our data.

      To my eye, it seems that Figure 5 could be more faithfully interpreted to state that DGL-1 protein localization depends on the L27-SH3 domains. The Huk/Guk domains are dispensable for DLG-1 protein localization; however, through other studies, we know they are important for viability. In contrast, dlg-1 mRNA localization requires all domains of the protein (L27-Guk). It is exceptionally interesting to find a mutant condition in which the mRNA and protein localizations are uncoupled. It would be very interesting to explore in the discussion or by other means what the purpose of localized translation may be. Because, in this instance, proper mRNA localization and protein function are closely associated, it may suggest that DLG-1 needs to be translated locally to function properly.

      We will rewrite the Results and Discussion to clarify our model. We agree that L27 and SH3 domains are critical, but we also detected effects of the HooK/GuK domains. We have refined our model to describe functions of the N and C termini for membrane or junctional localization.

      The manuscript requires an improve materials & methods description of the quantification __procedures and statistics employed.__

      We will add these points.

      Minor & Major comments together - text

      Summary statement: Is "adherent junction" supposed to be "adherens junction?"

      Corrected.

      Abstract: Sentence 1, I think they should add a caveat word to this sentence. Something like "...phenomenon that can facilitate sub-cellular protein targeting." In most instances this isn't very well characterized or known.

      Corrected.

      In the first paragraph, it might be good to mention that Moor et al also showed that mRNA localize to different regions to alter their level of translation (to concentrate them in high ribosome dense regions of the cell).

      Added as follows: “For example, a global analysis of localized mRNAs in murine intestinal epithelia found that 30% of highly expressed transcripts were polarized and that their localization coincided with highly abundant regions in ribosomes **(Moor, 2017).”

      There are some new studies of translation-dependent mRNA localization - that might be good to highlight - Li et al., Cell Reports (PMID: 33951426) 2021; Sepulveda et al., 2018 (PCM), Hirashima et al., 2018; Safieddine, et al 2021. Also, Hughes and Simmonds, 2019 reviews membrane associated mRNA localization in Drosophila. And a new review by Das et al (Nat Rev MCB) 2021 is also nice.

      We will add them to the text.

      Parker et al. did not show that the 3'UTR was dispensable for mRNA localization. They showed the 3'UTR was sufficient for mRNA localization.

      Quoting from the paper Parker et al.: “3′UTRs of erm-1 and imb-2 were not sufficient to drive mRNA subcellular localization. Endogenous erm-1 and imb-2 mRNAs localize to the cell or nuclear peripheries, respectively, but mNeonGreen mRNA appended with erm-1 or imb-2 3′UTRs failed to recapitulate those patterns (Fig. 4A-D).” We will make this point clearer in the rewritten text.

      In the second paragraph, the sentence about bean stages is missing one closing parenthesis.

      Corrected.

      Last paragraph: FISH is fluorescence, not fluorescent.

      Corrected.

      Both "subcellular" and "sub-cellular" are used.

      Corrected.

      Minor comments – Figures

      Figure 1

      o Figure 1A is confusing. It's not totally clear what the rectangles and circles signify. There are many acronyms within the figure. Which of the cell types depicted in the figure are shown here? For example, for the dorsal cells, which is the apical v. basal side?

      We tried to simplify the cartoon for a general C. elegans epithelial cell. We followed schematics already shown in previous publications to maintain consistency. Acronyms and color-codes are listed in the corresponding figure legend and have been better clarified.

      o Some of the colors are difficult to distinguish, particularly when printed out or for red/green colorblind readers. Is erm-1 meant to be a cytoskeletal associated or a basolateral polarity factor?

      We understand the issue, but unfortunately, with 8 classes of factors, shades of gray might not solve the problem. We tried to circumvent the red-green issue changing red to dark grey. Furthermore, we added details about shapes to the figure legends. We will work to make the colors work better.

      ERM-1 is a cytoskeletal-associated factor.

      o The nomenclature for dlg-1 is inconsistent within "C".

      Corrected.

      o Please specify what the "cr" is in "cr.dlg-1:-gfp" in the legend.

      Added.

      Figure 2

      o Can Figure 2C be quantified in a similar manner to 2A/2B?

      Currently our script cannot do that, but we will try to optimize it to be able to quantify this type of images.

      o 2B - please jitter the dots to better visualize them when they land on top of one another

      Yes, we will.

      o Please include a negative control example, a transcript that is not peripherally localized for comparison.

      Yes, we will.

      o There is no place in the text of the document where Fig 2C is referenced

      Corrected (it was wrongly referred to as “2B”).

      o I can't see any discernable ajm-1 localization in Fig 2A.

      We added some arrowheads to point at specific examples and increased the intensities of the corresponding smFISH signal for better visualization.

      o I can't see any dlg-1 pharyngeal localization in Fig2C.

      We added some arrowheads to point at specific examples and increased the intensities of the corresponding smFISH signal for better visualization.

      o More details on how the quantification was performed would be welcome. Particularly, in 2B, what is the distance from the membrane in which transcripts were called as membrane-associated? What statistics were used to test differences between groups?

      We will add a full description of the script used as well as the statistic details.

      Figure 3

      o Totally optional but might be nice: can you make a better attempt to approximate the scale of the cartoon depiction?

      The UTRs, especially the 5’ one, are much smaller than the dlg-1 gene sequence. A proper scaling of the cartoon to the actual sequences, would draw the attention away from the main subjects of this figure, the UTRs. Nevertheless, we made sure it is clear in the corresponding figure legend that the cartoon is not in scale: “The schematics are not in scale with the actual size of the corresponding sequences. UTR lengths: dlg-1 5’UTR: 61 nucleotides; sax-7 5’UTR: 63 nucleotides; dlg-1 3’UTR: 815 nucleotides; unc-54 3’UTR: 280 nucleotides.”

      o The GFP as an asterisk illustration may be confusing for some readers. Could you add another rectangular box to depict the gfp coding sequence?

      Corrected.

      o This microscopy is beautiful!

      Thanks Reviewer #2!

      o Were introns removed? Is the endogenous copy still present?

      All the transgenes were analyzed in a wild-type background, therefore, yes, the endogenous copy was still present. All the transgenes possessed introns. We will change the corresponding text as follows: “To test whether the localization of one of the identified localized mRNAs, dlg-1, relied on zip codes, we generated extrachromosomal transgenic lines carrying a dlg-1 gene whose sequence was fused to an in-frame GFP and to exogenous UTRs.”. In the figure “dlg-1 ORF” has been replaced with “dlg-1 gene”.

      o The wording in the legend "CRISPR or transgenic" may be confusing as Cas9 genome editing is still a form of transgenesis.

      We added “extrachromosomal” to clarify the nature of the mRNA.

      o The authors state that the 5'-3'UTR construct produces perinuclear dlg-1 transcripts but in the absence of DAPI imaging, it's not clear that this is the case.

      We could not find such a statement, but we tried to clarify the localization of these mRNAs in the text: “The mRNA localization patterns of the two UTR reporters were compared to the localization of dlg-1 transcripts from the CRISPR line (“wild-type”, Fig. 3A; Heppert et al., 2018), described in Fig. 2. Both reporter strains showed enrichment at the CeAJ and localization dynamics of their transcripts that were comparable to the wild-type cr.dlg-1 (Fig. 3B). These results indicate that the UTR sequences of dlg-1** mRNA are not required for its localization.”

      o Which probe set was used? The gfp probe?

      Yes, please see the main text: “Given that the transgenic constructs were expressed in a wild-type background, smFISH experiments were conducted with probes against GFP RNA sequences to focus on the transgenic dlg-1::GFP mRNAs (cr.dlg-1 and tg.dlg-1).”

      o Here, sax-7 is used as a uniform control, but sax-7 is claimed in Fig S1B-D as being perinuclear. This is a bit confusing.

      sax-7 mRNA localizes perinuclearly in sporadic instances (Fig S1C), but it is predominantly scattered throughout the cytoplasm (i.e., unlocalized). It presumably localizes perinuclearly in a translation-dependent manner as sax-7 codes for a transmembrane protein that would be targeted to the ER. We have described this ER-type of localization in the introduction and reiterated it partially in the first paragraph of the results. sax-7 UTRs are therefore presumably not responsible for any subcellular localization, which would instead rely on a signal sequence. We will better clarify this point in the main text.

      Figure 4

      o Excellent results! Really nice!

      Thanks Reviewer #2!

      o Fig 4A. The GFP depicted as a circle is strange.

      We changed it into a rectangle.

      o Fig 4A. Can you include the gene/protein name for easy skimming?

      Added.

      o Fig 4B. the color here is too faint and it is unclear what is being depicted. Overall, this part of the figure could be improved.

      We are optimizing the coloring and simplifying the schematics.

      o Were the introns removed?

      No, the introns were maintained in this and in all our transgenic lines. We described our transgenic lines in the materials and methods section (now with more detail). What we depict in the scheme (Fig. 4A) is the mature RNA (now specified in the figure), therefore no introns depicted. We will also specify this in the main text.

      Figure 5

      o Fig 5A. can you add the gene/protein name

      Added.

      o Fig 5B. Can you make the example apicobasal (non-apical) mRNA more distinctive? If it had its own peak in the lower trace, the reader would more clearly understand that this mRNA will be excluded from apical measurements whereas it will be included in apicobasal measurements.

      We actually wanted to show this specific example: a cytoplasmic mRNA and a junctional mRNA may seem close from the apicobasal analysis (partially overlapping peaks that Reviewer #2 mentioned). With the apical analysis, instead, we can show that these mRNAs are actually not close, and they belong to two different compartments (cytoplasm and junction). We would therefore like to keep the current scheme, while better clarifying this point in the corresponding figure legend.

      o D' - I' The grey font is too light.

      Noted. We will change it.

      o D' - I' The inconsistent y-axis scaling makes it difficult to compare across these samples. Can you set them to the same maximum number?

      The values are indeed quite different. We tried to use the same scale, but this would make some of the data unappreciable. The idea was to evaluate, within each graph, how mRNA and protein are localized relative to the junctional marker. We will make this clearer in the text.

      o D' - I' The x-axis labels are formatted incorrectly

      Corrected.

      o The practice of masking out the nucleus appears to remove potentially important mRNAs that are not nuclear localized. This could really impact the findings and interpretation. Instead, consider an automated DAPI mask.

      The masking on the images is not the same used for the analysis: in the images, a shaded circle has been drawn on the DNA channel and moved onto its corresponding location in the other channels or merges. For the analysis, the DNA signal has been specifically removed in the channel with the smFISH signal. Given that the analysis has been performed on maximum intensity projections of 3 Z-stacks, we believe we did not remove any non-nuclear mRNA. We will clarify this point in Materials and methods.

      o I can't see what the authors are calling membrane diffuse versus cytoplasmic. This is making it hard for me to see their "two step" pathway to localization.

      We will add in Fig. 5B-C an example of a membrane localized mRNA. Furthermore, we will add transverse sections of membrane and cytoplasm to make the date clearer to the reader.

      o Can more details of the quantification be included? How were Z-sections selected, chosen for inclusion? Which Z-sections and how many were selected?

      We will add the details to Materials and methods.

      o Also, why do these measurements focus on what I think are the seam cells when Lockwood et al., 2008 show the entire epithelium that is much easier to see?

      We are focusing on the seam cells at the bean stage as these are the cells and the embryonic stage where we see the highest localization of dlg-1 mRNA in the wild-type.

      o Please name these constructs to correlate the text more explicitly to the figures.

      Added.

      o How many embryos were analyzed for each trace? How many embryos showed consistent patterns?

      We will add the details of the analysis to Materials and methods.

      o Why were these cells used for study here? Lockwood et al., 2008 use a larger field of epithelial cells for visualization.

      As stated before: we are focusing on the seam cells at the bean stage as these are the cells and the embryonic stage where we see the highest localization of dlg-1 mRNA in the wild-type.

      Figure 6

      There are major discrepancies between what this figure is depicting graphically and what is described in the text. Again, I'm not comfortable making the "two step" claims this figure purports given the data shared in Figure 5.

      We are planning to re-write the last part of the results to better clarify our two-step model. A two-step model had been previously suggested in McMahon et al., 2001, where they could show that DLG-1 and AJM-1 (referred to in that publication as JAM-1) are initially localized laterally and only later in development are then enriched apically. Our data agree with McMahon very well, so we used the earlier study as a start. We will cite and explain this paper in greater depth during the rewriting.

      **Minor comments - Tables & Supplemental Figures**

      Table 1

      I think this table could be improved to more clearly illustrate which mRNAs were tested and what their mRNA localization patterns were (for example, gene name identifiers included, etc). Could the information that is depicted by gray shading instead be added as its own column? For example, have a column for "Observed mRNA localization"

      We modified Table 1 based on these and the other reviewers’ comments.

      Can you add distinct column names for the two columns that are labeled as "protein localization - group"

      We modified Table 1 based on these and the other reviewers’ comments.

      Can you also add which of these components are part of ASI v. ASII (as described in the introduction?)

      A new table has been added with the factors belonging to the two adhesion systems (same color code as in Table 1).

      Supplemental Figure 1

      It is hard to see that some of these spots are perinuclear. More information (membrane marker, 3D rendering, improved metrics) is required to support this claim.

      We are not trying to make a big statement out of these data as perinuclear localization for mRNAs coding for transmembrane/secreted proteins is well known. The aim of our study was to describe transcript localized at or in the proximity of the junction. We thought it was worth mentioning these examples of perinuclearly localized mRNAs (hmr-1, sax-7, and eat-20) for two reasons: scientific correctness – show accessory results that might be interesting for other scientists – and use as positive controls for our smFISH survey – these mRNAs were expected to have a somewhat perinuclear localization for the reasons mentioned above.

      What do these images look like over the entire embryo, not just in the zoomed in section?

      We added a column with the zoom-out embryos.

      sax-7 localization in S4 looks similar but a different localization claim is made.

      sax-7 mRNA can localize perinuclearly in sporadic instances (Fig S1C), but is predominantly scattered throughout the cytoplasm (i.e., unlocalized). It presumably localizes perinuclearly in a translation-dependent manner as sax-7 codes for a transmembrane protein that would be targeted to the ER. We have described this ER-type of localization in the introduction and reiterated it partially in the first paragraph of the results. sax-7 UTRs are therefore presumably not responsible for any subcellular localization, which would instead rely on a signal sequence. We will better clarify this point in the main text.

      Supplemental Figure 2

      Before adherens junctions even exist dlg-1 go to the membrane - this is really neat!

      Thanks Reviewer #2!

      Supplemental Figure 3

      Technical question: If either 5 or 3 stack images are used, how does this work? Do they have different z-spacings? Or do they do 5-stack images represent a wider Z-space?

      This is the sentence under question: “Maximum intensity projections of 5 (1.08 µm) (A) and 3 (0.54 µm) (B) Z-stacks”. The space between each Z-stack image is constant in all our imaging and its value is 270 nm. When we consider 5 planes, the distance from the 1st to the 5th is 4 x 270 nm = 1.08 µm, whereas for 3 planes will be 2 x 270 nm = 0.54 µm.

      Supplemental Figure 4

      Line #2 retains translation and keeps mRNA localization.

      Totally optional, but consider showing both lines in the main figure to illustrate the two possibilities.

      Noted.

      Materials and methods - how did they created the ATG mutations? Is it an array? - why does one translate, and one doesn't?

      We will clarify this point in Materials and methods: “dlg-1 deletion constructs ΔATG (SM2664 and SM2663) and ΔL27-PDZs (SM2641) were generated by overlap extension PCR using pML902 as a template.”.

      We will perform a Western blot to clarify Reviewer #2’s last point. Currently we do not know what peptide is translated, but the comparison with our full-length control will probably shed some light on the issue.

      Reviewer #3

      Major comments

      The smFISH results are striking and implications exciting. The conclusions made from the smFISH results reported in all Figures will be strengthened considerably by quantifying the mRNA localized to the defined specific subcellular regions. At the very least, localization to the cytoplasm versus the plasma membrane should be determined as performed in Figure 2B, but quantifying finer localization will enhance the conclusions made about regional localization (e.g. CeAJ versus plasma membrane mRNA localization in Figure 5). Inclusion of a non-localizing control in Figures 1-4 will enable statistical comparisons between mRNA localizing and non-localizing groups.

      We will add more quantitation, statistics, and negative controls.

      The script used for smFISH quantitation should be included in the methods or published in an accessible forum (Github, etc). Criteria for mRNA "dot" calling should be defined in the methods. All raw smFISH counts should also be reported.

      We will add the full description of the script in Materials and methods, and we will provide the raw data in an additional supplementary table.

      Figure 2: What is the localizing ratio of a non-localizing control mRNA (e.g. jac-1)? Including an unlocalized control with quantitation would strengthen the localization arguments presented.

      Yes, we will add quantitation for an unlocalized mRNA.

      Figure 5: Quantifying colocalization of mRNA and protein (+/- AJM-1) will strengthen the arguments made about mRNA/protein localization.

      Yes, we will quantify Fig. S5 to have a full picture of the cells (the images in Fig. 5 represent only a portion of the cell).

      Discussion of the CeAJ mRNA localization mechanism is warranted. Do the authors speculate that the newly translated protein drives localization during translation, similar in concept to SRP-mediated localization to the ER, or ribosome association is a trigger to permit a secondary factor to drive mRNA localization, or another model?

      Unfortunately, this is hard to say at the moment as we do not have any data regarding where translation actually occurs. We will add a conjecture to the Discussion.

      Minor comments

      Please complete the following sentence: "We identified transcripts enriched at the CeAJ in a stage- and cell type-specific."

      Corrected.

      It would be helpful to provide reference(s) for the protein localization summary in Table 1.

      Added.

      Figure 2B: Did dlg-1 and ajm-1 localize at similar ratios? Appropriate statistics comparing the different ratios may be informative.

      We will modify the graph (following Reviewer #2’s suggestion) and add the requested details.

      Figure 2: In the paragraph that begins, "Morphogenesis of the digestive track," the text should refer to Figure 2C? If not, the text requires further clarification.

      Corrected.

      Figure 2: Reporting the smFISH localizing ratios of 8E and 16E will be informative.

      We will add the information.

      Please include citations when summarizing the nonsense-mediated decay NMD mechanism and AJM-1 identifying the CeAJ.

      Added.

      The sentence, "Embryos from our second __Δ__ATG transgenic line displayed a little GFP protein and some dlg-1::gfp mRNA," should refer to Figure S4.

      Added.

      An immunoblot of this reporter versus wild type may be informative regarding the approximate position of putative alternative start codon.

      We will perform a Western blot to verify the size of the protein product produced.

      Figure 5: N's and repetitions performed should be included for localization experiments.

      Yes, we will add them here and in all the other quantifications we will add to the manuscript.

      Please clarify that the "the mechanism of UTR-independent targeting is unknown in any species" refers to dlg-1 mRNA localization.

      Added.

      "Our findings suggest..." discussion paragraph should reference Figure 6.

      Added.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Subcellular localization of mRNAs plays a critical role in gene regulation and ultimately cellular function. While mRNA untranslated regions often serve as key regulatory codes for expression, mRNA translation can also have a significant effect, a notable example being secretory peptides delivering translating transcripts to the endoplasmic reticulum. A complete understanding of the signals that organize mRNAs in the cell remains an open question. Here, Tocchini, et al. use the C. elegans embryo and single molecule FISH (smFISH) to determine the subcellular localization of key mRNAs involved in epithelial morphogenesis. This survey identifies several mRNAs that appear to localize to specific regions of the cell, such as the plasma membrane or apical junction, and in a developmental stage-specific manner. Dissection of the mRNA of dlg-1/discs large, an apical junction component, provides evidence that mRNA localization requires active translation, but surprisingly the untranslated regions are dispensable. Further mRNA truncation mapping supports the model that the N-terminal coding region helps target mRNAs to the apical junctions, but the C-terminal coding regions are sufficient to localize dlg-1 mRNA to the plasma membrane. The manuscript describes a two-step model for dlg-1 localization and recruitment to the apical junction that depends on translation.

      MAJOR:

      1. The smFISH results are striking and implications exciting. The conclusions made from the smFISH results reported in all Figures will be strengthened considerably by quantifying the mRNA localized to the defined specific subcellular regions. At the very least, localization to the cytoplasm versus the plasma membrane should be determined as performed in Figure 2B, but quantifying finer localization will enhance the conclusions made about regional localization (e.g. CeAJ versus plasma membrane mRNA localization in Figure 5). Inclusion of a non-localizing control in Figures 1-4 will enable statistical comparisons between mRNA localizing and non-localizing groups.
      2. The script used for smFISH quantitation should be included in the methods or published in an accessible forum (Github, etc). Criteria for mRNA "dot" calling should be defined in the methods. All raw smFISH counts should also be reported.
      3. Figure 2: What is the localizing ratio of a non-localizing control mRNA (e.g. jac-1)? Including an unlocalized control with quantitation would strengthen the localization arguments presented.
      4. Figure 5: Quantifying colocalization of mRNA and protein (+/- AFM-1) will strengthen the arguments made about mRNA/protein localization.
      5. Discussion of the CeAJ mRNA localization mechanism is warranted. Do the authors speculate that the newly translated protein drives localization during translation, similar in concept to SRP-mediated localization to the ER, or ribosome association is a trigger to permit a secondary factor to drive mRNA localization, or another model?

      MINOR:

      1. Please complete the following sentence: "We identified transcripts enriched at the CeAJ in a stage- and cell type-specific."
      2. It would be helpful to provide reference(s) for the protein localization summary in Table 1.
      3. Figure 2B: Did dlg-1 and ajm-1 localize at similar ratios? Appropriate statistics comparing the different ratios may be informative.
      4. Figure 2: In the paragraph that begins, "Morphogenesis of the digestive track," the text should refer to Figure 2C? If not, the text requires further clarification.
      5. Figure 2: Reporting the smFISH localizing ratios of 8E and 16E will be informative.
      6. Please include citations when summarizing the nonsense-mediated decay NMD mechanism and AJM-1 identifying the CeAJ.
      7. The sentence, "Embryos from our second ΔATG transgenic line displayed a little GFP protein and some dlg-1::gfp mRNA," should refer to Figure S4. An immunoblot of this reporter versus wild type may be informative regarding the approximate position of putative alternative start codon.
      8. Figure 5: N's and repetitions performed should be included for localization experiments.
      9. Please clarify that the "the mechanism of UTR-independent targeting is unknown in any species" refers to dlg-1 mRNA localization.
      10. "Our findings suggest..." discussion paragraph should reference Figure 6.

      Significance

      This well-written, well-cited manuscript describes the striking subcellular localization pattern of a critical, conserved gene involved in both animal development and human disease. The observation that the start codon, and thus translation, is necessary for transcript localization is a complete surprise, and opens exciting doors to investigate how translation leads to mRNA organization and its connection to tissue development. As such, this manuscript will be of broad interest to RNA, cell and developmental biologists, particularly those who investigate post-transcriptional gene regulation and protein complex assembly. However, while the images are indeed supportive of the manuscript's claims, the conclusions will be markedly strengthened by quantifying the subcellular localization of mRNAs in the smFISH experiments, paired with negative controls (e.g. non-localizing, cytoplasmic mRNA). Addition of more quantitative smFISH analyses will enhance the experimental reproducibility, rigor, and statistical significance. The text, figures, and methods should also be revised to include more details about the smFISH analyses, in particular the inclusion of n's, descriptions of how spots were identified, descriptions of scripts used, and the raw mRNA counts. Regardless, the reporter genes tested were well conceived and dlg-1 shows promise to be a fantastic model to further investigate the mechanisms underlying translation-dependent mRNA localization.

      My expertise covers post-transcriptional gene regulation, the C. elegans model organism, and fluorescent imaging with smFISH.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      Tocchini et al. screened apical junction and cell membrane proteins for mRNA localization. They identified multiple proteins that are translated from localized mRNAs. Of these, dlg-1 (Discs large) mRNA localizes to cell cortices of dorsal epithelial cells, endoderm cells, and epidermal (seam) cells and is dependent on active translation for transport. The manuscript dissects the contributions of different DLG-1 protein domains to mRNA localization.

      A major strength of the paper is the way it assesses translational-dependence in a transcript-specific way without perturbing translation globally. The authors cleverly combine mutations in ATG start sites with a knock down of the non-sense mediated decay pathway. This allows Tocchini et al to examine whether dlg-1 mRNA depends on active translation for localization, which it does. The authors observe an interesting finding, that the domains required for protein localization can be separated from those required for mRNA localization. Namely, mRNA localization (but not protein localization) requires C-terminal domains of the protein.

      My major points of concern focus on the presentation and interpretation of Figure 5. In this figure, the blocking approach used seems confounding, the observations described by the authors are not visible, the quantification is confusing, and the interpretations seem like an over-reach. The

      Major comments:

      • Figure 2 requires a negative (or uniformly distributed) mRNA control for comparison. Figure 2C should be quantified. The plot quality should be improved, and appropriate statistical tests should be employed to strengthen the claimed findings.

      • Most claims of perinuclear mRNA localization are difficult to see and not well supported visually or statistically. The usage of DAPI markers, membrane markers, 3D rendering, or a quantified metric would bolster this claim. Also, sax-7 is claimed to be perinuclear and elsewhere claimed to be uniform then used as a uniform control. Please explain or resolve these discrepancies more clearly.

      • The major concern about the paper is the data display and interpretation of Figure 5C. I'm not comfortable with the approach the authors took of blurring out the nucleus. A more faithful practice would be to use an automated mask over DAPI staining or to quantify the entirety of the cell. If the entirety of the cell were quantified, one could still focus analysis on specific regions of relevance. The interpretations distinguishing membrane versus cytoplasmic localization (or mislocalization) are hard to differentiate in these images especially since they are lacking a membrane marker. The ability to make these distinctions forms the basis of Tocchini et al's two pathways of dlg-1 mRNA localization. These interpretations also heavily rely on how the image was processed through the different Z-stacks, and it's not clear to me how that was done. For example, the diffusion of mRNA in figure 5F and 5I are indistinguishable to my eye but are claimed to be different.

      • To my eye, it seems that Figure 5 could be more faithfully interpreted to state that DGL-1 protein localization depends on the L27-SH3 domains. The Huk/Guk domains are dispensable for DLG-1 protein localization; however, through other studies, we know they are important for viability. In contrast, dlg-1 mRNA localization requires all domains of the protein (L27-Guk). It is exceptionally interesting to find a mutant condition in which the mRNA and protein localizations are uncoupled. It would be very interesting to explore in the discussion or by other means what the purpose of localized translation may be. Because, in this instance, proper mRNA localization and protein function are closely associated, it may suggest that DLG-1 needs to be translated locally to function properly.

      • The manuscript requires an improve materials & methods description of the quantification procedures and statistics employed.

      Minor & Major comments together:

      Text

      • Summary statement: Is "adherent junction" supposed to be "adherens junction?"

      • Abstract: Sentence 1, I think they should add a caveat word to this sentence. Something like "...phenomenon that can facilitate sub-cellular protein targeting." In most instances this isn't very well characterized or known.

      • In the first paragraph, it might be good to mention that Moor et al also showed that mRNA localize to different regions to alter their level of translation (to concentrate them in high ribosome dense regions of the cell).

      • There are some new studies of translation-dependent mRNA localization - that might be good to highlight - Li et al., Cell Reports (PMID: 33951426) 2021; Sepulveda et al., 2018 (PCM), Hirashima et al., 2018; Safieddine, et al 2021. Also, Hughes and Simmonds, 2019 reviews membrane associated mRNA localization in Drosophila. And a new review by Das et al (Nat Rev MCB) 2021 is also nice.

      • Parker et al. did not show that the 3'UTR was dispensable for mRNA localization. They showed the 3'UTR was sufficient for mRNA localization.

      • In the second paragraph, the sentence about bean stages is missing one closing parenthesis.

      • Last paragraph: FISH is fluorescence, not fluorescent.

      • Both "subcellular" and "sub-cellular" are used. Minor comments - Figures

      • Figure 1

      o Figure 1A is confusing. It's not totally clear what the rectangles and circles signify. There are many acronyms within the figure. Which of the cell types depicted in the figure are shown here? For example, for the dorsal cells, which is the apical v. basal side? o Some of the colors are difficult to distinguish, particularly when printed out or for red/green colorblind readers. Is erm-1 meant to be a cytoskeletal associated or a basolateral polarity factor? o The nomenclature for dlg-1 is inconsistent within "C". o Please specify what the "cr" is in "cr.dlg-1:-gfp" in the legend.

      • Figure 2

      o Can Figure 2C be quantified in a similar manner to 2A/2B? o 2B - please jitter the dots to better visualize them when they land on top of one another o Please include a negative control example, a transcript that is not peripherally localized for comparison. o There is no place in the text of the document where Fig 2C is referenced o I can't see any discernable ajm-1 localization in Fig 2A. o I can't see any dlg-1 pharangeal localization in Fig2C. o More details on how the quantification was performed would be welcome. Particularly, in 2B, what is the distance from the membrane in which transcripts were called as membrane-associated? What statistics were used to test differences between groups?

      • Figure 3

      o Totally optional but might be nice: can you make a better attempt to approximate the scale of the cartoon depiction? o The GFP as an asterisk illustration may be confusing for some readers. Could you add another rectangular box to depict the gfp coding sequence? o This microscopy is beautiful! o Were introns removed? Is the endogenous copy still present? o The wording in the legend "CRISPR or transgenic" may be confusing as Cas9 genome editing is still a form of transgenesis. o The authors state that the 5'-3'UTR construct produces perinuclear dlg-1 transcripts but in the absence of DAPI imaging, it's not clear that this is the case. o Which probeset was used? The gfp probe? o Here, sax-7 is used as a uniform control, but sax-7 is claimed in Fig S1B-D as being perinuclear. This is a bit confusing.

      • Figure 4

      o Excellent results! Really nice! o Fig 4A. The GFP depicted as a circle is strange. o Fig 4A. Can you include the gene/protein name for easy skimming? o Fig 4B. the color here is too faint and it is unclear what is being depicted. Overall, this part of the figure could be improved. o Were the introns removed?

      • Figure 5

      o Fig 5A. can you add the gene/protein name o Fig 5B. Can you you make the example apicobasal (non-apical) mRNA more distinctive? If it had its own peak in the lower trace, the reader would more clearly understand that this mRNA will be excluded from apical measurements whereas it will be included in apicobasal measurements. o D' - I' The grey font is too light. o D' - I' The inconsistent y-axis scaling makes it difficult to compare across these samples. Can you set them to the same maximum number? o D' - I' The x-axis labels are formatted incorrectly o The practice of masking out the nucleus appears to remove potentially important mRNAs that are not nuclear localized. This could really impact the findings and interpretation. Instead, consider an automated DAPI mask. o I can't see what the authors are calling membrane diffuse versus cytoplasmic. This is making it hard for me to see their "two step" pathway to localization. o "F" looks the same as "I" to me, but the authors claim they represent different patterns and use these differences as the basis for their claim that X. o Can more details of the quantification be included? How were Z-sections selected, chosen for inclusion? Which Z-sections and how many were selected? o Also, why do these measurements focus on what I think are the seam cells when Lockwood et al., 2008 show the entire epithelium that is much easier to see? o Please name these constructs to correlate the text more explicitly to the figures. o How many embryos were analyzed for each trace? How many embryos showed consistent patterns? o Why were these cells used for study here? Lockwood et al., 2008 use a larger field of epithelial cells for visualization.

      • Figure 6

      o There are major discrepancies between what this figure is depicting graphically and what is described in the text. Again, I'm not comfortable making the "two step" claims this figure purports given the data shared in Figure 5.

      Minor comments - Tables & Supplemental Figures

      Table 1

      • I think this table could be improved to more clearly illustrate which mRNAs were tested and what their mRNA localization patterns were (for example, gene name identifiers included, etc). Could the information that is depicted by gray shading instead be added as its own column? For example, have a column for "Observed mRNA localization"

      • Can you add distinct column names for the two columns that are labeled as "protein localization - group"

      • Can you also add which of these components are part of ASI v. ASII (as described in the introduction? Supplemental Figure 1

      • It is hard to see that some of these spots are perinuclear. More information (membrane marker, 3D rendering, improved metrics) is required to support this claim.

      • What do these images look like over the entire embryo, not just in the zoomed in section?

      • sax-7 localization in S4 looks similar but a different localization claim is made.

      Supplemental Figure 2

      • Before adherens junctions even exist dlg-1 go to the membrane - this is really neat! Supplemental Figure 3

      • Technical question: If either 5 or 3 stack images are used, how does this work? Do they have different z-spacings? Or do they do 5-stack images represent a wider Z-space?

      Supplemental Figure 4

      • Line #2 retains translation and keeps mRNA localization.

      • Totally optional, but consider showing both lines in the main figure to illustrate the two possibilities.

      • Materials and methods - how did they created the ATG mutations? Is it an array? - why does one translate, and one doesn't?

      Significance

      The authors discover that dlg-1, ajm-1, and hmr-1 mRNAs (among others) are locally translated, and this represents an important conceptual advance in the field as these are well studied proteins and important markers. This is the first study to illustrate translation-dependent mRNA localization in C. elegans, to my knowledge. The mechanisms transporting these mRNAs and their associated translational complexes to the membrane may represent a new pathway of mRNA transport and is therefore significant. The authors identify domains within DLG-1 responsible which is a nice advance. If they are unable to order the events of association as they claim in Figure 5 (and that I dispute), this doesn't detract from the impact of the paper.

      Other high-profile studies have recently been published that echo how mRNA localization to membranes can be observed for transcripts that encode membrane-associated proteins (Choaib et al., Dev Cell, 2020; Li et al., Cell Reports, 2021 (PMID: 33951426); and Reviewed in Hughes & Simmonds, Front Gen, 2019). These recent findings underscore the impact of Tocchini et al.'s paper. Similar studies have identified mRNAs localizing through translation dependent mechanisms to a variety of different regions of the cell (Sepulveda et al., eLife, 2018; Hirashima et al., Sci Reports, 2018; Safieddine, et al., Nat Comm, 2021; and reviewed in Ryder et al., JCB 2020). Given the timely nature of these findings and the recent interest in these concepts, a broad readership of readers should be interested in this paper.

      My field of expertise is in mRNA localization imaging and quantification. I feel sufficiently qualified to evaluate the manuscript on all its merits.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      In the current study Tocchini et al analyze mRNA localization during development of Caenorhabditis elegans embryonic epithelia. Using smFISH-based method they have identified mRNAs associated with the cell membrane or cortex, and with apical junctions. They showed that most of mRNAs involved in AS-II cell adhesion system localize to the membrane. To examine how epithelial morphogenesis affects mRNA localization, authors studied two transcripts encoding DLG-1 and AJM-1 that form a complex. Data showed that studied mRNAs enrichment at the CeAJ varies at distinct stages and cell types of embryogenesis. Then the study was focused on one of the identified transcripts - dlg-1/discs large. Using transgenic lines authors demonstrated that dlg-1 localization to the CeAJ is UTRs-independent, but requires active translation. Moreover, authors mapped protein domains involved in that process.

      Major comments:

      Fig. 1: Main and supplementary figures present smFISH signals for eight localized mRNAs, while in the results section authors describe that they analyzed twenty-five transcripts. Authors should explain the choice of transcripts presented in the paper. Moreover, smFISH signal of different localized mRNAs in epidermal cells was visualized at different stages (bean, comma or late comma), and authors did not comment what was the reason of such conditions. This may make transcripts localization results difficult to interpret, as further analysis showed that mRNA localization varied in a stage-specific manner. Did author used smFISH probes designed against endogenous mRNAs for all tested transcripts? Marking dlg-1 mRNA as dlg-1-gfp suggests that smFISH probe was specific for gfp transcript. Is it true? If yes, authors should compare localization of wild-type endogenous dlg-1 mRNA with that of the transcript encoding a fusion protein, to confirm that fusion does not affect mRNA localization.

      Fig. 2B: Authors conclude that at later stages of pharyngeal morphogenesis mRNA enrichment at the CeAJ decreased gradually in comparison to comma stage. Data do not show statistically significant decrease in ratio of localized mRNAs - for dlg-1: bean: 0.39{plus minus}0.09, comma: 0.29{plus minus}0.08, 1.5-fold: 0.30{plus minus}0.09; for ajm-1: bean: 0.36{plus minus}0.08, comma: 0.30{plus minus}0.05, 1.5-fold: 0.28{plus minus}0.09.

      Fig. 4: What was the difference between the first and the second ΔATG transgenic line? Authors should analyze the size of the truncated DLG-1 protein that is expressed from the second ΔATG transgenic line that localizes to CeAJ. Knowing alternative ATGs and protein size may suggest domain composition of the truncated protein. This will allow to confront truncated protein localization with the results from Fig. 5. Moreover, to prove that the localization of dlg-1 mRNA at the CeAJ is translation-dependent, additional experiment should be performed where transcripts localization will be analyzed in embryos treated with translation inhibitors such as cycloheximide (translation elongation inhibitor) and puromycin (that induces premature termination).

      Minor comments:

      In the introduction section authors should emphasize the main goal and scientific significance of the paper. Fig 1A: It's hard to distinguish different colors on the schematics. Schematics presents intermediate filaments that are not included in the Table 1.

      Fig. 1C: dlg-1 transcript is marked as dlg-1-gfp on the left panel and dlg-1 on the right panel.

      Fig. 2B: Axis labels and titles are not visible, larger font size should be used.

      Fig. 5C: Enlarge the font size.

      Fig. S2: Embryonic stages should be marked on the figure for easier interpretation.

      Significance

      This study provides a few contributions into understanding mRNA localization in Caenorhabditis elegans during embryo development. Firstly, it identifies adhesion system II mRNAs associated with epithelial cells. Secondly, it demonstrates a case study of translation-dependent dlg-1/DLG-1 mRNA localization mechanism that does not involve zip codes. Finally, it provides a model showing the roles of different DLG-1 domains in dlg-1 localization. The results are compelling and experiments are well presented, although in my opinion authors should provide a stronger evidence to support the idea that active translation is essential for dlg-1 localization.

      Overall, I believe the work will have a wide appeal covering areas such as mRNA localization, developmental biology and embryogenesis.

      My field of expertise is in the RNA-protein interactions and mRNA turnover using biochemical methods as well as in vivo studies in C. elegans and mammalian cell lines. I do not have an expertise in smFISH-based methods.

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

      Learn more at Review Commons


      Reply to the reviewers

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


      Reply to the Reviewers

      We thank the Referees for their evaluation and their useful comments.


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

      The MS from Bonaventure and colleagues used a CRISPR to identify novel IFN-induced antiviral effectors targeting HIV-1. One hit, the DEAD Box helicase DDX42, while not itself part of the IFN response, exerts a substantial inhibitory effect on HIV-1 replication when over expressed, and gives a several fold boost to viral replication when knocked down in cells. The effect of DDX42 KO or O/E is manifest at reverse transcription and PLA analysis suggests and interaction with incoming virions. Moreover, DDX42 appears to exert an inhibitory effect generally against retroviruses and retroelements, with evidence that it associates with viral/transposon RNA. The authors further show that DDX42 has antiviral against a range (but not all) RNA viruses, with very striking phenotypes seen especially with Zika, CHIKV and SARS CoV2, with DDX42 associating with dsRNA in infected cells. These data suggest DDX42 is a constitutively expressed a broad-spectrum inhibitor of a range of mammalian RNA viruses. The manuscript is very well written, the data is of good quality and clearly DDX42 is having a general effect on viral replication. The results are novel, important and potentially of wide interest. Where the MS is somewhat lacking is understanding whether DDX42 has direct antiviral activity or is globally affecting cellular RNA metabolism. Some important areas for the authors to consider are:

      • DDX42 has a potential role in splicing and/or RNA metabolism so I think it would be important to see whether there is any clear global change in gene expression in knockout or knockdown cells cells vs control that might be suggestive of a generalized effect.

      Responses

      We thank the reviewer for this important question. Indeed, DDX42 didn’t impact the replication of 2 negative strand RNA viruses and this suggested that DDX42 didn’t have a global impact on the target cells, but we could not formally exclude a generalized effect. Therefore, we have performed RNA-seq analysis in order to evaluate the impact of DDX42 depletion (using 3 different siRNAs targeting DDX42 in comparison to a CTRL siRNA in U87-MG cells, and 2 different siRNA in comparison to a CTRL siRNA in A549-ACE2 cells, in samples obtained in 3 independent silencing experiments). The RNA-seq data (See Supplemental File 1 and Figure S5) showed that only 63 genes are commonly differentially expressed by the 3 siRNAs targeting DDX42 in U87-MG cells and only 23 of these genes were also found differentially expressed in A549-ACE2 cells depleted for DDX42. Importantly, the identity of these genes could not explain the observed antiviral phenotypes. These data are in favor of the absence of generalized effect on the target cells, which could have explained the antiviral phenotypes of the sensitive viruses.

      • The HIV experiments in primary cells are only one round at present. Does the DDX42 knockdown enhance viral replication in multiround? Does it lead to more viral PAMPs for PRRs to induce IFN?

      Responses

      We agree with the reviewer that it would have been very informative to measure the impact of DDX42 knockdown in multiround infections in primary T cells. However, we tried several times to do this experiment (with primary T cells from several donors) and we were not successful: indeed, DDX42 KO appeared to slow down cell division, which could be taken into account for a short, one-cycle experiment (i.e. 24 h) 3 days post-Cas9/sgRNA electroporation by adjusting the number of cells at the time of infection. However, DDX42 KO appeared quite toxic in longer experiments, with cells stopping to grow.

      The question regarding the generation of more viral PAMPs for PRRs to induce IFN is also very interesting. We know from published work (including ours) that primary T cells don’t normally produce IFN following HIV-1 infection (see for instance Bauby and Ward et al, mBio 2021). However, one can indeed hypothesize that as more viral DNAs are produced in the absence of DDX42, perhaps the primary T cells could detect them and produce IFN. To address this question in primary T cells, we would have needed to be able to perform multiround infections, which was not possible, as mentioned above. Moreover, we could not test this hypothesis in the cell lines that we used, such as U87-MG/CD4/CXCR4 cells, as they are unable to produce IFN following HIV-1 infection.

      • More could be made mechanistically of the lack of sensitivity of Flu and VSV to DDX42. In particular showing whether or not DDX42 interacts with the RNA of the insensitive virus, or whether DDX42/virus or dsRNA interactions by PLA occur with Flu would highlight the relevance of these observations to the antiviral mechanism.

      Responses

      This is an excellent remark. We have now performed RNA immunoprecipitation experiments using 2 viruses targeted by DDX42 (CHIKV and SARS-CoV-2) and 1 virus that is insensitive to DDX42 (IAV) (See New Figure 4J-L): whereas CHIKV and SARS-CoV-2 RNAs could be specifically pulled-down with DDX42 immunoprecipitation, this was not the case for IAV RNA. This strongly argues for a direct mechanism of action of DDX42 helicase on viral RNAs.

      Reviewer #1 (Significance (Required)):


      __ The role of helicases in host defence are of wide interest and importance. This has the potential to be a very important study that deserves a wide audience. However in my opinion it needs some further mechanistic insight along the lines I have suggested.

      Responses

      As mentioned above, we have now added important data: First, DDX42 is able to interact with RNAs from targeted viruses (and not from an insensitive virus); Second, we have checked that DDX42 didn’t have a substantial impact on the cell transcriptome. Taken together, these data are clearly in favour of a direct mode of action of DDX42.

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

      In this brief report, the authors use a CRISPR screening approach to identify cellular proteins that limit HIV infection. The screen itself is elegantly designed and most of the top hits are components of the interferon signaling pathway that would be expected to emerge from such a screen, thus providing confidence in the results. The authors followed up on DDX42 as a new hit identified in their screen and confirmed that targeting DDX42 with distinct guide RNAs resulted in increased HIV infection in at least 3 cell lines. Conversely, DDX42 overexpression inhibited infection. They also confirmed a role for DDX42 in inhibiting HIV infection in primary macrophages and CD4 T cells using siRNA and CRISPR KO strategies, respectively. They also demonstrate that DDX42 inhibits several other divergent lentiviruses as well as Chikungunya virus and SARS-CoV-2, but not influenza virus. These data convincingly show that DDX42 plays a role in inhibiting many lentivirus and positive sense RNA virus infections. Using PCR assays for reverse transcription products they conclude that DDX42 inhibits an early process in the HIV life cycle occurring after virus entry, though the statistical significance of these differences is not clear. They further use proximity ligation assays to suggest that DDX42 is in proximity to HIV-1 and SARS-CoV-2 replication complexes. Mechanistically, these data are largely unsatisfying as they do not provide specific insight into how DDX42 so broadly inhibits virus replication. Overall, the manuscript presents a significant advance, it also has some weaknesses as listed below.

      1. Statistical analysis is not included in any of the figures.

      Response

      Statistical analyses have now been included.

      Many of the figure legends do not state how many independent biological replicates the figures are based on.

      Response

      The number of biological replicates for each panel is stated at the very end of each figure legend.

      Detailed mechanistic understanding of DDX42 effects on virus replication is not provided by the manuscript.


      Response

      As mentioned in response to Reviewer 1, we have now added data showing that DDX42 could interact with RNAs from targeted viruses but not from an insensitive virus, arguing for a direct antiviral mode of action of this Dead-Box helicase.

      Reviewer #2 (Significance (Required)):

      DDX42 is a new antiviral protein identified and confirmed in this manuscript. It was also identified as one of many hits in a genome wide CRISPR screen for cellular proteins that regulate SARS-CoV-2 infections, but was not followed up. Thus, the identification and confirmation of DDX42 antiviral activity is highly significant for both the HIV and SARS-CoV-2 fields. This high significance may compensate to some extent for the lack of mechanistic insight contained in this initial report.

      **Referees Cross-commenting**

      I find the comments of the other reviewers to be fair and reasonable, and I concur that the work is overall important and novel. It seems that reviewers generally agreed that some additional mechanistic insights would be desirable for publication in a high impact journal. Reviewer 1 makes some good suggestions in this regard. As for mouse experiments, I would reserve these for a follow up manuscript.

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


      __In this manuscript, Bonaventure et al report the results of a screen to identify cellular inhibitors of HIV-1 infection in IF treated cells. They identify DDX42 as such a factor though, unexpectedly, DDX42 did not turn out to be an ISG. Strikingly, DDX42 turns out to inhibit a wide range of retroviruses as well as retrotransposons and + sense, but not - sense, RNA viruses among which SARS-CoV2 turns out to be especially sensitive to DDX42, with siRNAs specific for SARS-CoV2 DDX42 increasing viral RNA expression by a startling 3 orders of magnitude, compared to only an 2-5 fold positive effect with HIV-1.

      Response

      We agree with the reviewer that DDX42’s impact on HIV-1 may appear as somewhat modest, however, it is highly reproducible across cell lines and primary cells and, more importantly, it is observed upon depletion of the endogenous protein (either by KO or silencing) in target cells that are highly permissive to viral replication, such as activated primary CD4+ T cells. We therefore believe that these findings, combined with the findings that other positive-strand RNA viruses are targeted, are of high interest.

      Reviewer #3 (Significance (Required)):


      __I found this paper generally convincing and technically sound though the emphasis was odd and clearly driven more by the history of how this work was done than by the actual results obtained. Specifically, the emphasis is on HIV-1 yet the most interesting data are the dramatic effects seen with Chikungunya and SARS2. If I was writing this paper, I would delete figure 4 and focus this paper entirely on retroviruses and retrotransposons. In that form, I think it would be competitive at PLoS Pathogens or perhaps EMBO Journal. The RNA virus work shown in figure 4 could then be figure 1 of a new, high impact, paper looking at the mechanism of action of DDX42 as an inhibitor of + sense, but not - sense, viral gene expression. Though Wei et al do mention DDX42 in their SARS-CoV2 screening paper this is certainly not a major theme of that paper so I don't think that would be a problem.

      Responses

      We thank the reviewer for this comment. We had hesitated to present the manuscript as suggested by the reviewer (i.e. focusing only on HIV-1, retroviruses and retroelements) and prepare a second manuscript with the remaining data. We’ve finally decided against it, as we believe that showing a broad antiviral effect of DDX42 on +strand RNA viruses increases the impact of our findings.

      On another note, a conditional DDX42 KO mouse has been generated by the Wellcome trust Sanger institute and it would greatly improve this manuscript if they could show an in vivo a result similar to figure 3F using MLV.

      Responses

      We thank the reviewer for this information. We completely agree that in vivo work would be a massive plus and we will be planning to explore this in the future, but not at this stage as it would require specific funding and resources.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this manuscript, Bonaventure et al report the results of a screen to identify cellular inhibitors of HIV-1 infection in IF treated cells. They identify DDX42 as such a factor though, unexpectedly, DDX42 did not turn out to be an ISG. Strikingly, DDX42 turns out to inhibit a wide range of retroviruses as well as retrotransposons and + sense, but not - sense, RNA viruses among which SARS-CoV2 turns out to be especially sensitive to DDX42, with siRNAs specific for SARS-CoV2 increasing viral RNA expression by a startling 3 orders of magnitude, compared to only an 2-5 fold positive effect with HIV-1.

      Significance

      I found this paper generally convincing and technically sound though the emphasis was odd and clearly driven more by the history of how this work was done than by the actual results obtained. Specifically, the emphasis is on HIV-1 yet the most interesting data are the dramatic effects seen with Chikungunya and SARS2. If I was writing this paper, I would delete figure 4 and focus this paper entirely on retroviruses and retrotransposons. In that form, I think it would be competitive at PLoS Pathogens or perhaps EMBO Journal. The RNA virus work shown in figure 4 could then be figure 1 of a new, high impact, paper looking at the mechanism of action of DDX42 as an inhibitor of + sense, but not - sense, viral gene expression. Though Wei et al do mention DDX42 in their SARS-CoV2 screening paper this is certainly not a major theme of that paper so I don't think that would be a problem. On another note, a conditional DDX42 KO mouse has been generated by the Wellcome trust Sanger institute and it would greatly improve this manuscript if they could show an in vivo a result similar to figure 3F using MLV.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this brief report, the authors use a CRISPR screening approach to identify cellular proteins that limit HIV infection. The screen itself is elegantly designed and most of the top hits are components of the interferon signaling pathway that would be expected to emerge from such a screen, thus providing confidence in the results. The authors followed up on DDX42 as a new hit identified in their screen and confirmed that targeting DDX42 with distinct guide RNAs resulted in increased HIV infection in at least 3 cell lines. Conversely, DDX42 overexpression inhibited infection. They also confirmed a role for DDX42 in inhibiting HIV infection in primary macrophages and CD4 T cells using siRNA and CRISPR KO strategies, respectively. They also demonstrate that DDX42 inhibits several other divergent lentiviruses as well as Chikungunya virus and SARS-CoV-2, but not influenza virus. These data convincingly show that DDX42 plays a role in inhibiting many lentivirus and positive sense RNA virus infections. Using PCR assays for reverse transcription products they conclude that DDX42 inhibits an early process in the HIV life cycle occurring after virus entry, though the statistical significance of these differences is not clear. They further use proximity ligation assays to suggest that DDX42 is in proximity to HIV-1 and SARS-CoV-2 replication complexes. Mechanistically, these data are largely unsatisfying as they do not provide specific insight into how DDX42 so broadly inhibits virus replication. Overall, the manuscript presents a significant advance, it also has some weaknesses as listed below.

      1. Statistical analysis is not included in any of the figures.
      2. Many of the figure legends do not state how many independent biological replicates the figures are based on.
      3. Detailed mechanistic understanding of DDX42 effects on virus replication is not provided by the manuscript.

      Significance

      DDX42 is a new antiviral protein identified and confirmed in this manuscript. It was also identified as one of many hits in a genome wide CRISPR screen for cellular proteins that regulate SARS-CoV-2 infections, but was not followed up. Thus, the identification and confirmation of DDX42 antiviral activity is highly significant for both the HIV and SARS-CoV-2 fields. This high significance may compensate to some extent for the lack of mechanistic insight contained in this initial report.

      Referees Cross-commenting

      I find the comments of the other reviewers to be fair and reasonable, and I concur that the work is overall important and novel. It seems that reviewers generally agreed that some additional mechanistic insights would be desirable for publication in a high impact journal. Reviewer 1 makes some good suggestions in this regard. As for mouse experiments, I would reserve these for a follow up manuscript.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      The MS from Bonaventure and colleagues used a CRISPR to identify novel IFN-induced antiviral effectors targeting HIV-1.

      One hit, the DEAD Box helicase DDX42, while not itself part of the IFN response, exerts a substantial inhibitory effect on HIV-1 replication when over expressed, and gives a several fold boost to viral replication when knocked down in cells. The effect of DDX42 KO or O/E is manifest at reverse transcription and PLA analysis suggests and interaction with incoming virions. Moreover, DDX42 appears to exert an inhibitory effect generally against retroviruses and retroelements, with evidence that it associates with viral/transposon RNA. The authors further show that DDX42 has antiviral against a range (but not all) RNA viruses, with very striking phenotypes seen especially with Zika, CHIKV and SARS CoV2, with DDX42 associating with dsRNA in infected cells. These data suggest DDX42 is a constitutively expressed a broad-spectrum inhibitor of a range of mammalian RNA viruses.

      The manuscript is very well written, the data is of good quality and clearly DDX42 is having a general effect on viral replication. The results are novel, important and potentially of wide interest. Where the MS is somewhat lacking is understanding whether DDX42 has direct antiviral activity or is globally affecting cellular RNA metabolism. Some important areas for the authors to consider are:

      • DDX42 has a potential role in splicing and/or RNA metabolism so I think it would be important to see whether there is any clear global change in gene expression in knockout or knockdown cells cells vs control that might be suggestive of a generalized effect.

      • The HIV experiments in primary cells are only one round at present. Does the DDX42 knockdown enhance viral replication in multiround? Does it lead to more viral PAMPs for PRRs to induce IFN?

      • More could be made mechanistically of the lack of sensitivity of Flu and VSV to DDX42. In particular showing whether or not DDX42 interacts with the RNA of the insensitive virus, or whether DDX42/virus or dsRNA interactions by PLA occur with Flu would highlight the relevance of these observations to the antiviral mechanism.

      Significance

      The role of helicases in host defence are of wide interest and importance. This has the potential to be a very important study that deserves a wide audience. However in my opinion it needs some further mechanistic insight along the lines I have suggested.

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

      Learn more at Review Commons


      Reply to the reviewers

      General Statements

      We appreciate the thoughtful and constructive comments provided by the reviewers and the opportunity to submit our revision plan for consideration. We have copied the reviewers’ comments below and have detailed our proposed revisions and/or clarifications after each comment (or set of comments). We also provide a partially revised manuscript with editorial changes highlighted in red.

      Reviewer 1:

      In this work, the authors Titialii-Torres and Morris assess how hyperglycemia affects the development of the neural retina using a genetic and a nutritional approach in the model organism zebrafish. This is important as diabetes can contribute to retinal degeneration in during the progression of diabetic retinopathy which often leads to blindness in adults. The authors examine how different cell types in the neural retina are affected in a genetic hyperglycemic model, the pdx1 mutant embryos, and in a nutritional model, in which hyperglycemia is induced by glucose and dexamethasone exposure. Titialii-Torres and Morris show that in both models, photoreceptor rods and cones, as well as horizontal cells, are reduced in number. Additionally, they report a delay in retinal cell differentiation accompanied by increased ROS production in the hyperglycemic retina. Altered expression of metabolism related genes and effects on visual function were also found in their hyperglycemic models. Overall, the assessment of the different retinal cell types impacted by hyperglycemia and examination of potential molecular mechanisms contributes important and novel data to the field. However, the data as presented falls short in supporting the conclusions of the authors.

      **Major comments**

      Overall, the conclusions would be more strongly supported by improving the clarity of the images, and by additional analyses.

      Figure1:

      Referring to figure 1 E' the text states that an arrowhead points to the shorter and thinner outer segment of a rod. In the figure there is an arrow pointing to a cell without a visible outer segment, making it hard to make the same conclusion. Additionally the GFP signal is very weak in D and E in the dorsal retina. Therefore it is not possible to see if there is also a decreased amount of rods in the dorsal retina as claimed. In the text it is mentioned that cones in the ventral region are affected. Is there also a difference in the dorsal region?

      Response: In our revised manuscript, we will include higher magnification panels for better visualization of the morphological differences between photoreceptor outer segments; we will also revise the graphs to show separate quantification of photoreceptors in the dorsal vs ventral retina.

      Figure 3: Rods and cones might be better displayed in close-ups from sections rather than from projections of the whole eye.

      Response: We will make this change

      The authors write about a reduction of cones upon glucose treatment. In the graph this is not highlighted as significant.

      Response: the change is significant; the graph will be edited to indicate this

      Figure 4: as the overall number of cones was already assessed before, focusing on a smaller region might help the reader to see the Zpr3 staining showing that the outer segments of the cones are stunted (as stated in the main text). In the figure panels presented, outer segments cannot be clearly seen.

      Response: we agree, and will make this change

      Figure 6: scale bar is missing. Please clarify what the red and the green is. Why is there a red signal from Mitosox outside of the embryo (panel C)? The fluorescence of the superoxide probe should be displayed in a more convincing way. For example, in sections to enable assignment of signal to tissues and cells, as shown in Supplemental Figure 5.

      Response: for the revised manuscript we will replace this figure with one containing analysis of tissue sections, with appropriate figure annotation and scale bar

      Figure 8: Is the coincubation with methylene blue leading to a significant increase in photoreceptors? If yes, this should be indicated in the graph.

      Response: for methylene blue treatment alone, the increase was not statistically significant; we have added text in the Results to clarify this. For the revised manuscript, we are also performing additional experiments with a methylene blue + SOD treatment group and with other ROS inhibitors, so this figure will be updated with those data.

      Supplemental Figure 2: The authors assert that TUNEL+ cell labeling coincides with Müller glial cells. This would be better supported with a magnified view of the INL, optimally by applying TUNEL staining to hyperglycemic, GFAP:GFP transgenic samples.

      Response: we will repeat this experiment using the GFAP:GFP line as suggested

      It would be of interest to determine if an incubation with methylene blue also affects photoreceptors in pdx1 mutants. Is it possible to confirm that Methylene blue treatment reduces ROS in the retina ? Can changes in ROS response gene expression be demonstrated by qPCR ? The assumptions about ROS should be either strengthened by additional experiments or less emphasized in the discussion.

      Response: for the revised version we will include the ROS inhibitor experiments on pdx1 mutants as suggested, as well as imaging with the Mitosox probe to confirm the efficacy of the ROS inhibitors; we are also testing additional ROS inhibitors as described above.

      For completeness, glucose metabolism in the genetic model should be also addressed and compared to the nutritional model.

      Response: While we agree that it would be helpful to have these data, it would take a very long time to collect the necessary number of pdx1 mutant individuals needed for this experiment due to the small numbers of homozygous mutants recovered in each clutch. As an alternative approach, for the revised manuscript we will use qRT-PCR to test a subset of the genes on the pdx1 mutants that showed significant changes in the nutritional model.

      The authors talk about a "long term" return to normoglycemia and long term effects of hyperglycemia. Analysis at 7 dpf after a 2 day return to normoglycemic conditions can hardly be called long term. To make these statements, an assessment after a longer time period (one week or more if possible?) would be more convincing.

      Response: for our revised manuscript, we are adding an additional time point for analysis at one week post hyperglycemia

      The claims of 'reactive gliosis' in glucose-treated larvae is overstated. Biologically meaningful differences in cell shape between control and treated samples are not evident from the images (Fig. 5A-F). This should at least be quantitated by shape analysis. The Glucose+Dex samples do not show increased number of Müller glial cells, and glucose treatment alone leads to highly variable glucose levels. This complicates and weakens a correlation with hyperglycemia.

      Response: we will add the suggested shape quantification of these images; we are also performing Western blots with an anti-GFAP antibody to further strengthen our conclusions – this is a well-accepted method for demonstrating gliosis.

      **Minor comments**

      Some figures would benefit if they would follow the sequence of the text. Eg: figure 1 and 3, the text addresses first the rods and then the cones. In several places the panels referred to in the text do not match the figures or figure panels are not mentioned at all. For example: Pg 3 "Quantification revealed a significant decrease in both rod and cone photoreceptors in pdx1 mutants at 5 dpf (Fig. 1C)." - the quantification is in panels C and F. The main text does not mention or explain Figure 2A.

      Pg 5 "The results confirmed that rods and cones from hyperglycemic larvae have shorter outer segments compared to wild type larvae at 5 dpf (Fig. 4A-C)." - panel C is a graph of Saccades. Fig. S3 - only panel Y is referred to in the text.

      Response: the text has been edited to correct these issues

      Supplemental figure 2: the authors claim a significant increase of apoptotic cells in the genetic model. In the corresponding graph significance is not indicated.

      Response: the increase in apoptotic cells was significant for the nutritional but not the genetic model; the text has been corrected to reflect this.

      Figure 5: scale bars are missing, the figure text and the numbering of the figure do not fit.

      The suggested corrections will be made to this figure and the corresponding text

      Supplemental figures 4 and 5: The Prox1 staining is hard to see and it is unclear what was counted as cells.

      Response: annotations will be added to Sup Figs 4 and 5 to clarify which cells are being quantified

      In Supp Fig. 4E the PKC staining looks increased compared to the controls.

      Response: the variability in staining intensity is within the normal range of what we have observed across all treatments and genotypes

      The graphs could have similar y axes, especially because in Supp Fig. 5 the amount of cells/µm is also different. Why not always use per 50µm? Shouldn't the amount of cells in wild types and untreated embryos be the same per 50µm? Also the labelling of the y axes could be made coherent in the two figures.

      Response: The denominator will be standardized for all graphs. The scale of the y-axes varies by cell type because some retinal cell classes are significantly more abundant than others.

      Supplemental figure 6: K is not mentioned in the legend.

      Response: this has been corrected

      2-NDBG treatment is not explained in material and methods

      Response: this information has been added to the Methods

      • *

      Reviewer #1 (Significance (Required)):

      **Significance**

      Titialii-Torres et al. characterize the impaired development of neural retinal cells under hyperglycemic conditions in zebrafish larvae and also show evidence of impaired visual function. This work will be of interest for researchers in the field of diabetes, especially those focused on diabetic retinopathy, and for developmental biologists interested in pathologies that impact human development. While the manuscript provides insights into the development of the retina under hyperglycemic conditions, a revision addressing weaknesses of figure presentation and some additional confirmatory experiments would be of great benefit.

      Response: we appreciate the reviewer’s assessment that our work will be of interest to various research communities, and agree that the suggested revisions to the figures and confirmatory experiments will greatly strengthen the impact.

      Reviewer 2:

      **Summary:**

      This paper uses immersion of embryonic zebrafish in high glucose solution to model the effects of hyperglycemia on retinal development. The paper finds that high glucose causes a reduction in the number of photoreceptors and horizontal cells, abnormalities in the morphology of photoreceptors and Müller glia, increased retinal cell apoptosis, a change in the timing of neuronal cell birth, and a defect in the optokinetic response. The mechanistic link between high glucose and changes in retinal development is not well described but may involve an increase in reactive oxygen species.

      **Major comments:**

      1. Is the photoreceptor phenotype a degenerative rather a developmental phenotype? In embryos treated with high glucose, photoreceptors in the periphery of the retina near the ciliary margin, which are younger in age, seem to be structurally more normal than those at the center, away from the ciliary margin, which are older in age. Could this reflect the fact that photoreceptor development proceeds normally followed by degenerative changes?

      Response: this is certainly a possibility, given the increase in TUNEL positive cells we detected in hyperglycemic retinas. However, we did not detect many apoptotic cells in the ONL at 3 and 4 dpf, suggesting that there is not widespread degeneration among differentiated photoreceptors at that stage. This result, in combination with the altered differentiation timing data shown in Figure 7, is what led us to favor a developmental phenotype. In the revised manuscript, we will add text to the Discussion that more thoroughly explores these alternative interpretations.

      1. For many or most phenotypes the main examined treatment is glucose + dexamethasone. The authors state this combination achieves more uniform glucose concentrations in the embryos as compared to glucose alone. However, dexamethasone may have effects independently of glucose and the dexamethasone only control is not used in some or most experiments. For example in Fig. 3, could dexamethasone alone causes changes in photoreceptor morphology? In the combo treatment, is it possible that some effects are simply due to a synergism of glucose+dex and not because dex causes a more uniformly high intraembryonic glucose?

      Response: we have evaluated photoreceptor number and morphology in the dex alone treatment group and found no significant differences. We will add these results to the main text and the supplemental figures.

      1. It is interesting that hyperglycemic retinas show more neurons born between 2-5 days post fertilization in the RGC layer than in the outer nuclear layer (Fig 7). One interpretation is delayed birth of RGCs after hyperglycemia as the authors suggest. Another interpretation is that non-RGC cell types are in now in the RGC layer; or that some proliferating progenitors persist at 5dpf. Co-localization of EdU with differentiation markers, and EdU analysis after a short pulse of 2 hours would help to nail down if there is developmental delay or something else going on here.

      Response: we appreciate the suggestion, and will perform this experiment for the revision

      1. Do Müller cells go into cycle after high glucose treatment?

      Response: this is a great question – we will do a co-localization experiment and add these results to the revised manuscript.

      1. The increase in ROS in Fig. 6 does not seem very convincing. Is the difference between untreated and glucose or glucose/dex treatments statistically significant? I would avoid making too much of this unless some type of phenotype rescue with N-acetylcysteine or vitamin C, or Trolox, can be shown. Methylene blue is a bit non-specific as an antioxidant.

        Response: for methylene blue treatment alone, the increase was not statistically significant; we have added text in the Results to clarify this. For the revised manuscript, we are also performing additional experiments with a methylene blue + SOD treatment group and with other ROS inhibitors, so this figure will be updated with those data

      Reviewer #2 (Significance (Required)):

      The translational significance of the findings is that they might provide a model to study how embryonic hyperglycemia due to maternal diabetes changes embryonic development. Pitfalls include the fact that its relevance to humans is unclear. Is maternal diabetes known to cause visual abnormalities due to abnormal retinal development in newborns? The basic biology significance may be to provide a model to investigate how glucose metabolism is connected to developmental decisions. However it is unclear whether glucose metabolism within retinal cells mediates the observed effects; and the high glucose used here is likely unphysiological as at these developmental stages zebrafish embryos feed from the yolk sac.

      Response: yes, maternal diabetes is associated with retinal abnormalities in humans, although there are not many published studies on this topic. In the Discussion, we talked about how our results align with prior clinical studies which documented reduced inner and outer macular thickness in children of diabetic pregnancies. At the suggestion of Reviewer 3, we have added this information to the Introduction as well to highlight the relevance of our study to humans. With respect to the comment about physiological relevance, we feel that the inclusion of the genetic model, which does not rely on high levels of exogenous glucose and yet exhibits a similar photoreceptor phenotype, speaks to this issue.

      Reviewer 3: **Summary:**

      The authors use a combination of genetic and pharmacological immersion approaches to investigate the effects of hyperglycemia on development of the retina in zebrafish larvae. They demonstrate a rather mild phenotype (though still convincing) such that photoreceptor maturation is delayed/impaired and the Muller glia are also affected. Visual function is modestly impacted, as measured with an assay that can be influenced by motor as well as sensory defects. The authors conclude that altered timing of the differentiation of retinal cells, together with accumulation of reactive oxygen species (ROS) underly the photoreceptor defects and reduced visual function in the hyperglycemic larvae.

      **Major comments**

      The retinal phenotype related to hyperglycemia is quite subtle, but sufficiently consistent. This phenotype would be more convincing, and lead to more definitive conclusions, if the authors could include some ultrastructural (TEM) information, or even high-resolution/magnification color images of thinner sections processed using conventional histological methods, such as H&E, or toluidine blue/pyronin B. It is difficult to appreciate the features of the apical projections of the photoreceptors in the fluorescently-labeled images.

      Response: we are adding higher magnification images to the photoreceptor figures (also suggested by Reviewer 1) and will incorporate an H&E stain as well.

      Comparison of zpr1 labeling with the TaC:eGFP transgenic is unfortunate. Ideally the authors would use the pdx1 mutant on this transgenic background. Alternatively, the authors could perform TaC in situ hybridizations.

      Response: we have crossed the pdx1 line onto the TaC:eGFP transgenic background and will have this experiment completed for the revision

      The visual function defect is also quite mild. The authors should mention that the OKR assay also relies upon motor function, and so the defect may be related to sensory deficit, motor deficit, or both. Larval ERGs would address this issue.

      Response: we will add this alternative explanation for the OKR results to the text.

      The "reactive gliosis" phenotype is also mild/subtle, and not entirely convincing. More information should be provided regarding what the authors considered an "abnormal shape" of an MG cell body. Ideally, there is an at least somewhat objective means to score normal vs. abnormal and then quantify.

      Response: for the revision, we are adding shape quantification and Western blots (please see our response to the similar comment made by Reviewer 1)

      In Figure 6 legend, the authors state that superoxide production is increased, but the graph does not appear convincing in this regard, and no statistical evaluation is provided.

      Response: for methylene blue treatment alone, the increase was not statistically significant; we have added text in the Results to clarify this. For the revised manuscript, we are also performing additional experiments with a methylene blue + SOD treatment group and with other ROS inhibitors, so this figure will be updated with those data

      The authors do not indicate whether they checked datasets for having a normal distribution prior to the selection of a t-test (or ANOVA) for analysis vs. nonparametric tests.

      Response: a more thorough description of our statistical analyses will be added to the methods

      The model and accompanying text in the Discussion seem overly wordy and speculative. This discussion also does not acknowledge that the effects upon the retina may be indirect, mediated by other tissues that are impacted by hyperglycemia. For example, ocular vascular defects have been described to result from hyperglycemia, over a similar time frame of analysis, and the effects on the retina may be downstream of these defects.

      Response: we will revise the Discussion to remove extraneous information and to incorporate alternative mechanisms that could explain the retinal phenotypes induced by hyperglycemia

      **Minor comments:**

      Introduction - the statement appearing in the Discussion (offspring of diabetic pregnancies had significantly thinner inner and outer macula as well as lower macular volume [43].) should appear in the Introduction to better capture the interest of the reader.

      Response: this change has been made

      Page 1. (..in nearly 10% of US pregnancies) - citation needed.

      Response: this has been added

      Page 2. Pdx1 mutation should be briefly described when first mentioned.

      Response: this has been added

      Legend for Figure 2 could benefit from a definition of 2-NDBG.

      Response: the figure legend has been revised

      Figure 2B does not show Whole Body [Glucose] because the heads were removed for histological analysis.

      Response: this correction will be made to the figure

      It is this reviewer's experience and opinion that zpr-3 labels rods and the RH2 members of the double cones, due to the sequence similarity of RH1 (rhodopsin) and RH2. The cited paper (Yin et al., 2012) hints at this as well, but is slightly unclear due to the terminology used in the paper in describing cone subtypes.

      We have edited the Results to clarify that Zpr3 labels the Rh2-expressing member of the double cones

      Page 8. The marker used to detect bipolar neurons should be mentioned within the Results section.

      Response: this information has been added to the Results

      Pages 12-13. The localization of TUNEL+ profiles may be related to microglia extending processes into the ONL, engulfing photoreceptors, and then internally transporting the bits to their cellular "eating stations" within other retinal layers.

      Response: we will add text to the Results including this as a possibility. We are also (at the suggestion of Reviewer 1) adding an experiment to determine whether some of the TUNEL+ cells co-localize with Muller glia markers.

      Reviewer #3 (Significance (Required)):

      **Significance/Comparison to published knowledge:**

      The zebrafish model(s) are sufficiently novel, versatile, and interesting to constitute an advance in this field. A literature search by this reviewer revealed that the focus upon retinal cells in hyperglycemic, larval zebrafish appears novel. However, the phenotype is remarkably mild, and there is concern that follow-up studies in pursuit of more mechanistic insights will be challenging to perform by the authors and by others in the field. This paper does lay some key groundwork, but what comes next sounds like a lot of fishing expeditions.

      Response: we appreciate the reviewer’s assessment that our work represents a novel advance in this field, and lays “key groundwork” for future studies. Although our nutritional and genetic models do not present with photoreceptor loss so severe that it causes complete blindness at the timepoints we tested, the photoreceptor reductions we observe are consistent, readily scoreable, and are associated with demonstrable defects in visual behavior. Given that we also provide evidence of both altered cell differentiation kinetics and increased oxidative stress in embryonic hyperglycemic retinas, we feel that these are excellent starting places for future work to uncover more mechanistic insights. Finally, our results have implications for human visual system development under hyperglycemic conditions. Timely vision development in infancy is required for attainment of a host of developmental milestones. Even mild delays in this process could have long term consequences for intellectual and social development, and due to the difficulty of measuring visual acuity in infants, subtle but significant impairments may go undetected at this critical stage. Therefore, having a reliable animal model for embryonic hyperglycemia will facilitate efforts to better understand this condition with the goal of developing appropriate intervention and treatment strategies.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Summary:

      The authors use a combination of genetic and pharmacological immersion approaches to investigate the effects of hyperglycemia on development of the retina in zebrafish larvae. They demonstrate a rather mild phenotype (though still convincing) such that photoreceptor maturation is delayed/impaired and the Muller glia are also affected. Visual function is modestly impacted, as measured with an assay that can be influenced by motor as well as sensory defects. The authors conclude that altered timing of the differentiation of retinal cells, together with accumulation of reactive oxygen species (ROS) underly the photoreceptor defects and reduced visual function in the hyperglycemic larvae.

      Major comments:

      The retinal phenotype related to hyperglycemia is quite subtle, but sufficiently consistent. This phenotype would be more convincing, and lead to more definitive conclusions, if the authors could include some ultrastructural (TEM) information, or even high-resolution/magnification color images of thinner sections processed using conventional histological methods, such as H&E, or toluidine blue/pyronin B. It is difficult to appreciate the features of the apical projections of the photoreceptors in the fluorescently-labeled images.

      Comparison of zpr1 labeling with the TaC:eGFP transgenic is unfortunate. Ideally the authors would use the pdx1 mutant on this transgenic background. Alternatively, the authors could perform TaC in situ hybridizations.

      The visual function defect is also quite mild. The authors should mention that the OKR assay also relies upon motor function, and so the defect may be related to sensory deficit, motor deficit, or both. Larval ERGs would address this issue.

      The "reactive gliosis" phenotype is also mild/subtle, and not entirely convincing. More information should be provided regarding what the authors considered an "abnormal shape" of an MG cell body. Ideally, there is an at least somewhat objective means to score normal vs. abnormal and then quantify.

      In Figure 6 legend, the authors state that superoxide production is increased, but the graph does not appear convincing in this regard, and no statistical evaluation is provided.

      The authors do not indicate whether they checked datasets for having a normal distribution prior to the selection of a t-test (or ANOVA) for analysis vs. nonparametric tests.

      The model and accompanying text in the Discussion seem overly wordy and speculative. This discussion also does not acknowledge that the effects upon the retina may be indirect, mediated by other tissues that are impacted by hyperglycemia. For example, ocular vascular defects have been described to result from hyperglycemia, over a similar time frame of analysis, and the effects on the retina may be downstream of these defects.

      Minor comments:

      Introduction - the statement appearing in the Discussion (offspring of diabetic pregnancies had significantly thinner inner and outer macula as well as lower macular volume [43].) should appear in the Introduction to better capture the interest of the reader.

      Page 1. (..in nearly 10% of US pregnancies) - citation needed.

      Page 2. Pdx1 mutation should be briefly described when first mentioned.

      Legend for Figure 2 could benefit from a definition of 2-NDBG.

      Figure 2B does not show Whole Body [Glucose] because the heads were removed for histological analysis.

      It is this reviewer's experience and opinion that zpr-3 labels rods and the RH2 members of the double cones, due to the sequence similarity of RH1 (rhodopsin) and RH2. The cited paper (Yin et al., 2012) hints at this as well, but is slightly unclear due to the terminology used in the paper in describing cone subtypes.

      Page 8. The marker used to detect bipolar neurons should be mentioned within the Results section.

      Pages 12-13. The localization of TUNEL+ profiles may be related to microglia extending processes into the ONL, engulfing photoreceptors, and then internally transporting the bits to their cellular "eating stations" within other retinal layers.

      Significance

      Significance/Comparison to published knowledge:

      The zebrafish model(s) are sufficiently novel, versatile, and interesting to constitute an advance in this field. A literature search by this reviewer revealed that the focus upon retinal cells in hyperglycemic, larval zebrafish appears novel. However, the phenotype is remarkably mild, and there is concern that follow-up studies in pursuit of more mechanistic insights will be challenging to perform by the authors and by others in the field. This paper does lay some key groundwork, but what comes next sounds like a lot of fishing expeditions.

      Interested audiences:

      Investigators in vision science/ophthalmology, developmental biology, toxicology/teratology, and diabetes.

      Reviewer keywords:

      Developmental biology, genetics, retina, zebrafish, photoreceptors.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary:

      This paper uses immersion of embryonic zebrafish in high glucose solution to model the effects of hyperglycemia on retinal development. The paper finds that high glucose causes a reduction in the number of photoreceptors and horizontal cells, abnormalities in the morphology of photoreceptors and Müller glia, increased retinal cell apoptosis, a change in the timing of neuronal cell birth, and a defect in the optokinetic response. The mechanistic link between high glucose and changes in retinal development is not well described but may involve an increase in reactive oxygen species.

      Major comments:

      1. Is the photoreceptor phenotype a degenerative rather a developmental phenotype? In embryos treated with high glucose, photoreceptors in the periphery of the retina near the ciliary margin, which are younger in age, seem to be structurally more normal than those at the center, away from the ciliary margin, which are older in age. Could this reflect the fact that photoreceptor development proceeds normally followed by degenerative changes?
      2. For many or most phenotypes the main examined treatment is glucose + dexamethasone. The authors state this combination achieves more uniform glucose concentrations in the embryos as compared to glucose alone. However, dexamethasone may have effects independently of glucose and the dexamethasone only control is not used in some or most experiments. For example in Fig. 3, could dexamethasone alone causes changes in photoreceptor morphology? In the combo treatment, is it possible that some effects are simply due to a synergism of glucose+dex and not because dex causes a more uniformly high intraembryonic glucose?
      3. It is interesting that hyperglycemic retinas show more neurons born between 2-5 days post fertilization in the RGC layer than in the outer nuclear layer (Fig 7). One interpretation is delayed birth of RGCs after hyperglycemia as the authors suggest. Another interpretation is that non-RGC cell types are in now in the RGC layer; or that some proliferating progenitors persist at 5dpf. Co-localization of EdU with differentiation markers, and EdU analysis after a short pulse of 2 hours would help to nail down if there is developmental delay or something else going on here.
      4. Do Müller cells go into cycle after high glucose treatment?
      5. The increase in ROS in Fig. 6 does not seem very convincing. Is the difference between untreated and glucose or glucose/dex treatments statistically significant? I would avoid making too much of this unless some type of phenotype rescue with N-acetylcysteine or vitamin C, or Trolox, can be shown. Methylene blue is a bit non-specific as an antioxidant.

      Significance

      The translational significance of the findings is that they might provide a model to study how embryonic hyperglycemia due to maternal diabetes changes embryonic development. Pitfalls include the fact that its relevance to humans is unclear. Is maternal diabetes known to cause visual abnormalities due to abnormal retinal development in newborns? The basic biology significance may be to provide a model to investigate how glucose metabolism is connected to developmental decisions. However it is unclear whether glucose metabolism within retinal cells mediates the observed effects; and the high glucose used here is likely unphysiological as at these developmental stages zebrafish embryos feed from the yolk sac.

      Audience interested in this work may include zebrafish developmental biologists. My expertise: metabolism, retinal development.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary

      In this work, the authors Titialii-Torres and Morris assess how hypergycemia affects the development of the neural retina using a genetic and a nutritional approach in the model organism zebrafish. This is important as diabetes can contribute to retinal degeneration in during the progression of diabetic retinopathy which often leads to blindness in adults.

      The authors examine how different cell types in the neural retina are affected in a genetic hyperglycemic model, the pdx1 mutant embryos, and in a nutritional model, in which hyperglycemia is induced by glucose and dexamethasone exposure. Titialii-Torres and Morris show that in both models, photoreceptor rods and cones, as well as horizontal cells, are reduced in number. Additionally, they report a delay in retinal cell differentiation accompanied by increased ROS production in the hyperglycemic retina. Altered expression of metabolism related genes and effects on visual function were also found in their hyperglycemic models. Overall, the assessment of the different retinal cell types impacted by hyperglycemia and examination of potential molecular mechanisms contributes important and novel data to the field. However, the data as presented falls short in supporting the conclusions of the authors.

      Major comments

      Overall, the conclusions would be more strongly supported by improving the clarity of the images, and by additional analyses.

      Figure1:

      o Referring to figure 1 E' the text states that an arrowhead points to the shorter and thinner outer segment of a rod. In the figure there is an arrow pointing to a cell without a visible outer segment, making it hard to make the same conclusion.

      o Additionally the GFP signal is very weak in D and E in the dorsal retina. Therefore it is not possible to see if there is also a decreased amount of rods in the dorsal retina as claimed.

      o In the text it is mentioned that cones in the ventral region are affected. Is there also a difference in the dorsal region?

      Figure 3:

      o Rods and cones might be better displayed in close-ups from sections rather than from projections of the whole eye.

      o The authors write about a reduction of cones upon glucose treatment. In the graph this is not highlighted as significant.

      Figure 4: as the overall number of cones was already assessed before, focusing on a smaller region might help the reader to see the Zpr3 staining showing that the outer segments of the cones are stunted (as stated in the main text). In the figure panels presented, outer segments cannot be clearly seen.

      Figure 6: scale bar is missing. Please clarify what the red and the green is. Why is there a red signal from Mitosox outside of the embryo (panel C)? The fluorescence of the superoxide probe should be displayed in a more convincing way. For example, in sections to enable assignment of signal to tissues and cells, as shown in Supplemental Figure 5.

      Figure 8: Is the coincubation with methylene blue leading to a significant increase in photoreceptors? If yes, this should be indicated in the graph.

      Supplemental Figure 2: The authors assert that TUNEL+ cell labeling coincides with Müller glial cells. This would be better supported with a magnified view of the INL, optimally by applying TUNEL staining to hyperglycemic, GFAP:GFP transgenic samples.

      It would be of interest to determine if an incubation with methylene blue also affects photoreceptors in pdx1 mutants. Is it possible to confirm that Methylene blue treatment reduces ROS in the retina ? Can changes in ROS response gene expression be demonstrated by qPCR ? The assumptions about ROS should be either strengthened by additional experiments or less emphasized in the discussion.

      For completeness, glucose metabolism in the genetic model should be also addressed and compared to the nutritional model.

      The authors talk about a "long term" return to normoglycemia and long term effects of hyperglycemia. Analysis at 7 dpf after a 2 day return to normoglycemic conditions can hardly be called long term. To make these statements, an assessment after a longer time period (one week or more if possible?) would be more convincing.

      The claims of 'reactive gliosis' in glucose-treated larvae is overstated. Biologically meaningful differences in cell shape between control and treated samples are not evident from the images (Fig. 5A-F). This should at least be quantitated by shape analysis. The Glucose+Dex samples do not show increased number of Müller glial cells, and glucose treatment alone leads to highly variable glucose levels. This complicates and weakens a correlation with hyperglycemia.

      In addition, there are many minor inconsistencies and awkwardness in the presentation of the figures, as detailed below.

      Minor comments

      Some figures would benefit if they would follow the sequence of the text. Eg: figure 1 and 3, the text addresses first the rods and then the cones.

      In several places the panels referred to in the text do not match the figures or figure panels are not mentioned at all. For example:

      Pg 3 "Quantification revealed a significant decrease in both rod and cone photoreceptors in pdx1 mutants at 5 dpf (Fig. 1C)."

      • the quantification is in panels C and F.

      The main text does not mention or explain Figure 2A Pg 5 "The results confirmed that rods and cones from hyperglycemic larvae have shorter outer segments compared to wild type larvae at 5 dpf (Fig. 4A-C)."

      • panel C is a graph of Saccades.

      Fig. S3 - only panel Y is referred to in the text.

      Supplemental figure 2: the authors claim a significant increase of apoptotic cells in the genetic model. In the corresponding graph significance is not indicated.

      Figure 5: scale bars are missing, the figure text and the numbering of the figure do not fit.

      Supplemental figures 4 and 5:

      o The Prox1 staining is hard to see and it is unclear what was counted as cells.

      o In Supp Fig. 4E the PKC staining looks increased compared to the controls.

      o The graphs could have similar y axes, especially because in supplemental figure 5 the amount of cells/µm is also different. Why not always use per 50µm? Shouldn't the amount of cells in wild types and untreated embryos be the same per 50µm?

      o Also the labelling of the y axes could be made coherent in the two figures.

      Supplemental figure 6: K is not mentioned in the legend.

      2-NDBG treatment is not explained in material and methods

      Significance

      Significance

      Titialii-Torres et al. characterize the impaired development of neural retinal cells under hyperglycemic conditions in zebrafish larvae and also show evidence of impaired visual function. This work will be of interest for researchers in the field of diabetes, especially those focused on diabetic retinopathy, and for developmental biologists interested in pathologies that impact human development. While the manuscript provides insights into the development of the retina under hyperglycemic conditions, a revision addressing weaknesses of figure presentation and some additional confirmatory experiments would be of great benefit.

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

      Learn more at Review Commons


      Reply to the reviewers

      The authors do not wish to provide a response at this time.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Kleijn et al. measured transcript and protein abundance in fission yeast cultures growing on different nutrient sources (and thus at different growth rates) in turbidostats. Their experimental design is sound and the data quality appears good. The authors focus on analyzing their data from the vantage point of previously reported ideas on principles of proteome allocation, and expand beyond this framework with interesting analyses, e.g., on the stoichiometry of translation complexes changes with the growth rate.

      Generally I find the paper well written and the conclusions well substantiated. Below are specific recommendations that may help the authors improve their study:

      • Your data allow investigating the extend of transcriptional and post-transcriptional regulation in fission yeast, and I think this analysis will be very interesting. PMID: 28481885 provides one simple approach to such analysis, and the authors may use another. Importantly, they authors must account for measurement noise.
      • Your analysis of the ribosomal proteins (RP), the ribosome biogenesis regulon (RiBi), and the translation initiation, elongation and termination factors (IET) is interesting. I would love to know whether there changes within these groups of proteins, e.g., different RP in budding yeast change differently with growth rate (PMID: 24767987, PMID: 26565899) and I would love to know if this is the case with fission yeast.
      • The Z score ranges on some of the heatmaps (e.g. fig 2A) are so wide that the changes in protein / RNA abundance are difficult to see.
      • It will be very useful to perform unbiased gene set enrichment analysis of the functions that show significant growth rate dependent and nutrient dependent effects, e.g., as in Fig 11 of PMID: 21525243

      Significance

      I am an expert in this field, and I think that this study represents a significant advance.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      In this paper Kleijn et al study global gene expression profiles in S. pombe grown in different nitrogen sources using a turbidostat that result in variation in growth rate. The authors use both RNAseq to quantify RNA expression and mass spectrometry to quantify protein expression. They find that the expression of many genes is correlated with growth rate. This finding builds on prior work performed by other groups in S. cerevisiae and bacteria that show organismal growth rate is a primary determinant of gene expression state for a large fraction of genes. The findings in this paper confirm and extend those results.

      Significance

      One surprising aspect of this manuscript is that the authors do not seem to have made the most of their experimental design. The acquisition of both protein and mRNA expression across these conditions provides a unique dataset for looking at how these two levels of expression agree with respect to each other. A simple plot showing the strength of the growth rate response for a gene at the level of mRNA and protein would already be interesting, but I would think that there is the opportunity to look more quantitatively at whether the ratio of mRNA to protein remains constant across growth rates or whether there systematic deviations that are biologically interesting. I would encourage the authors to address this question with their unique dataset.

      Prior to publication the authors should address the following points.

      At what point in the turbidostat cycle was the sampling performed? At steady state or during the dilution phase?

      It is unclear in the text what transcripts are included in the category ncRNA. Does this include tRNA and rRNA?

      The basis for the abbreviations for positive (R) negative (P) and not significant (Q) are obscure. Why not P, N, NS?

      In Brauer et al., the fraction of cells in G1 is correlated with growth rate. Is that the case in S pombe? Is there any relationship between cell cycle gene expression and growth rate related gene expression?

      Is there anything unique to the set of ~100 genes that are anticorrelated between mRNA and protein in response to growth rate variation?

      A clearer explanation of the FC metric and the rationale for its use should be made in the results. What is FC an abbreviation for? It is unclear why this metric is needed, when the strength of the response to growth rate is captured by the slope.

      Airoldi et al., 2016 and Airoldi et al., 2009 looked at methods for normalizing gene expression to growth rate and may be relevant sources.

      The contrast in the experimental rationale between using chemostats and turbidostats is interesting, but I am left unclear about whether the result is really that different. What is the key distinction in the observed data in comparing gene expression response to growth rate in the chemostat and turidostat?

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

      Learn more at Review Commons


      Reply to the reviewers

      Reviewer #1 (Evidence, reproducibility and clarity (Required)): In this manuscript, using in vivo infection of Zebrafish embryos with Mycobacterium marinum and THP1-derived macrophages infected with Mycobacterium tuberculosis, the authors show that these pathogenic mycobacteria trigger an increase of K+ concentration through the expression of OXSR1. The ESX1 secretion system that is essential for the virulence of M. marinum is required for the expression of OXSR1 and SPAK. OXSR1 and SPAK are involved in the WNK signaling pathway and are cytoplasmic serine/threonine protein kinases that regulate the function of a series of sodium, potassium and chloride co-transporters via phosphorylation. Given that K+ efflux is now accepted as the main inducer of NLRP3 inflammasome, the authors report that this infection-induced OXSR1 expression restrains the protective NLRP3 inflammasome response leading to IL-1b maturation and secretion. Il-1b as a very potent pro-inflammatory triggers TNF-a production and the authors demonstrate that infection-induced OXSR1 expression suppressed host protective TNF-a and cell death early in fection. It appears therefore that virulent mycobacteria induce OXSR1 expression to reduce inflammasome activation by maintaining high intracellular K+. The results presented by the authors are convincing and the conclusions raised by the authors are well supported by the data. In zebrafish embryos, OXSR1 knockdown nicely reduces mycobacteria burden. Based on their conclusions that infection-induced OXSR1 expression reduces NLRP3 inflammasome activation, NLRP3 inflammasome activation has therefore a protective effect against bacterial infection. My main concern is that surprisingly, nlrp3 or il1b knockdown has no effect on bacterial burden in comparison to control embryos. Lane 256, as an explanation, the authors wrote "This may have been because we were using mosaic F0 CRISPR knockout, which is not a complete removal". The removal using mosaic F0 CRISPR knockout is nevertheless sufficient to observe a decrease in bacterial burden following OXSR1 knockdown. Would it be possible that OXSR1 also regulates immunity independently of NLRP3 inflammasome?

      Yes, we will add text to the discussion to address potential NLRP3-independent mechanisms that connect OXSR1 to immunity against mycobacterial infection.

      The lack of effect of il1b knockdown on M. marinum burden has been corroborated by independent laboratories including a publication from the Elks lab in Journal of Immunology: Ogryzko et al 2019. The Ogryzko study found no effect of il1b knockout on M. marinum burden.

      **Other comments:** OXSR1 WB in extended Data 3 is really poor quality so that it is hard to see the increased expression of OXSR1 following infection.

      The western blot will be repeated for cleaner images.

      Figure 2C. It is not shown but I guess that similar results should be obtain using M. tuberculosis.

      Material leaving our BSL3 facility must be decontaminated which makes this suggested analysis impossible in our facility.

      Figures 5D and 5E. To confirm the involvement of NLRP3, in addition of using MCC950, NLRP3 knock down using siRNA should be also performed. NLRP3-deficient THP-1 cells are also commercially available if the siRNA-mediated knock down of NLRP3 is not convincing enough.

      We will purchase NLRP3 deficient THP-1 cells and use our existing shRNA vector to create NLRP3 and OXSR1 deficient cells. We will repeat the experiments in 5D and 5E in these cells to confirm NLRP3 involvement.

      **Minor comments:** How do the authors think that mycobacterium induces OXSR1 expression following infection? It has not been investigated and it is not discussed.

      In Fig1A we showed upregulation of oxsr1a transcription and in Fig2A we showed upregulation of OXSR1 protein. In line 204 of the discussion we described our hypothesis that oxsr1a transcription is responsive to the mycobacterial ESX1 secretion system.*

      *

      Reviewer #1 (Significance (Required)): The observations reported in this manuscript are interesting since for the first time, it is described that virulent mycobacteria induce OXSR1 expression to reduce NLRP3 inflammasome activation by maintaining high intracellular K+. This is quite a significant advance in the field. To escape immune control, many successful intracellular pathogens have evolved methods to limit inflammasome activation. While it is known that potassium efflux is a trigger for inflammasome activation, the interaction between mycobacterial infection, potassium efflux and inflammasome activation was not explored. My field of expertise is the regulation of inflammasome activation. As far as I remember, I've never reviewed a paper using zebrafish embryos but here, the explanations and data are clear so that it was easy to understand and to evaluate. Likewise, I did not know the WNK signaling pathway but the literature clearly shows that it is involved in intracellular ionic balance.

      Reviewer #2 (Evidence, reproducibility and clarity (Required)): Hortle et al, in this study evaluated the role of WNK kinases SPAK and OXSR1 during M. marinum and M. tuberculosis infection. These two kinases inhibit the KCC channels which have a tendency to export potassium out of the cell. Since potassium efflux is a known stimulator of NLRP3 inflammasome activation, this raises the possible role of these kinases in inflammation and infection. Authors showed that inhibiting OXSR1 genetically and chemically reduced the mycobacterium survival in cells and zebrafish model, thus proposing OXSR1 as a host-directed therapeutic candidate. They showed that knockdown of OXSR1a leads to NLRP3 inflammasome mediated IL1B induction, which results in increase in TNFa and suppression of mycobacterium growth. Furthermore, reduction in mycobacterium growth in OXSR1a KD zebrafish embryos was found to be dependent on ESX1 machinery of Mycobacterium. The role of potassium in regulating Mycobacterium host response is novel. However there are few things which are missing from this interesting work. **Main comments**

      1. Since OXSR1 is known to inhibit KCC channels, which will lead to increase intracellular potassium. Why in infected control cells there is no increase potassium, Fig 2C. What would be the role of potassium in OXSR1 mediated control of Mtb growth?

      We will perform more experiments with altered levels of extracellular potassium to determine if infected control cells have increased intracellular potassium compared to OXSR1 knockdown cells.

      Does addition of extracellular potassium restricts mycobacterium in OXSR1-KD cells?

      We will perform additional experiments with the addition of potassium to the cell culture medium to address this concern.

      Since OXSR1 is known to inhibit KCC channels, What happens to the activity of these channels in OXSR1 KD cells? This is important, because authors could not find any difference in intracellular potassium between uninfected control and uninfected OXSR1 KD cells (Fig 2C). It will be good to add the flowcytometric histogram or dot plots of potassium staining in the main figure or in extended figures.

      We have data showing that although there is minimal difference in basal K+ level in OXSR1 KD cells, there is significantly lower K+ level when the cells are placed in High K+ media, or osmotic shock. We will include this data in the revised manuscript. We will amend the figures to include Flow plots.

      Acquisition of potassium stained cells - In methodology it has been mentioned that ion K+ Green stained undifferentiated THP1 cells were acquired using PE channel while differentiated THP1 cells were acquired using FITC channel. Furthermore in methods its mentioned that Leica Sp8 microscope was used to acquire images, however I do not see any of this data in the manuscript.

      Ion K+ green emits into both the PE and FITC channels. Our choice to use the FITC or PE channel depended on whether the cells were also infected with red fluorescent bacteria which “contaminates” the PE channel.

      Fig 2E and 3D - Meaning of "Normalized CFU/ml"? Each dot represents what? How many times this experiment was performed, please add in the legend.

      Normalized CFU/ml means that the CFU at 3 day post infection were normalized to the 0 day post infection intracellular bacterial burden, to adjust for any differences in phagocytosis of bacteria. Each dot represents the CFU from an infected well in a single representative experiment and the experiment was repeated 3 times. This information will be added to the figure legend.


      Fig 1D - What could be the reason of no statistical significant difference between wild type and homozygous oxsr1a-KO fish?

      This data is from two experimental replicates. We are currently growing more breeding fish to generate embryos for experimental replicates.

      Good to have a schematic model showing the finding s of the study

      We will add a schematic model to the manuscript.

      TNFa is double edge sword and can lead to pathology. Hence treatment of chronically infected animals (say mice) by Compound B, will be needed to confirm the HDT activity of OXSR1.

      Yes, we will add discussion of this point as a caveat to our future direction of using OXSR1 inhibition as a HDT.

      Reviewer #2 (Significance (Required)): This study showed role of kinases, which regulate trafficking of potassium, in mycobacterium-host interaction. Since kinases are draggable, so this opens a new area for developing host-directed therapies for TB. Reviewer #3 (Evidence, reproducibility and clarity (Required)): In this study, the authors suggest to have evidence for OXSR1 to inhibit NLRP3 inflammasome activation by limiting potassium efflux during mycobacterial infection. To my opinion, the study lacks important results supporting their main conclusions. In many instances, the authors have over-interpreted their data and I therefore do not support publication of this study. **Main comments:** Activation of the NLRP3 inflammasome upon OXSR1 knockdown was not convincingly demonstrated.

      We will address the activation state of the NLRP3 inflammasome with NLRP3 KO and OXSR1 KD cells as also suggested by reviewer 1: We will purchase NLRP3 deficient THP-1 cells and use our existing shRNA vector to create NLRP3 and OXSR1 deficient cells. We will repeat the experiments in 5D and 5E in these cells to confirm NLRP3 involvement.

      Clearance of bacteria in an organism, herein zebrafish, involves mechanisms in different cell types including downstream of inflammasome activation. Thus, bacterial clearance experiments in THP-1 cells might not necessarily be related to in vivo experiments in an organismal context. Finally, a mechanism as to how mycobacteria enhance OXSR1 expression to block a NLRP3-mediated response has not been addressed.

      We are not able to perform in depth analysis of the bacterial side of this host-pathogen interaction as my lab will close in the next 4 months. We have shown that transcriptional upregulation of oxsr1a is ESX1-dependent. We will include data on OXSR1 protein expression with WT and ESX1 mutant bacteria when we repeat the western blots in Extended data 3.

      **Specific comments:**

      1. The author showed that the M. marinum ESX1 secretion system induced OXSR1 expression to inhibit the NLRP3 inflammasome activation. This is contradictory to another recent study (PMID: 18852239), which showed that the ESX1 secretion system activated the NLRP3 inflammasome. These effects are not mutually exclusive. The ESX1 secretion system has a “deliberate” purpose in exporting mycobacterial effector proteins to subvert cellular immunity while also having an “accidental” role in exposing the host cell cytosol to vacluolar contents that can activate cellular immunity. We do not assert that mycobacteria completely inhibit all NLRP3 activation – rather that attempts to stop full activation via inducing the expression of host OXSR1. This can be seen in the IL-1b data in figure 3E, where infected WT cells release more IL-1b than MCC950 treated cells, but less than OXSR1 KD cells.

      In line 102, based on Data shown in Fig 1D, the authors concluded that homozygous, but not heterozygous, oxsr1asyd5 embryos showed reduced bacterial burden. However, in Fig 1D, the difference among the genotypes is not significant.

      This concern will be addressed with additional replicates.

      In line 196, the authors stated that "We present evidence that pathogenic mycobacteria increase macrophage K+ concentration by inducing expression of OXSR1." However, the authors did not provide evidence for this.

      We will soften this phrase in the discussion to replace “by inducing” with “and induce”.

      Based on Extended data 3, the authors concluded that infection increases the expression of OXSR1. However, this is not evidenced in the Western Blot. In addition, in panel B, the OXSR1 blot showed many non-specific bands with decreased intensity in OXSR1 knockdown conditions suggesting that there is unequal protein loading making it impossible to interpret these results.

      We will repeat the western blots as per Reviewer 1’s comment as well.

      The authors concluded that infection-induced OXSR1 expression suppressed inflammasome activity to aid mycobacterial infection. Experiments with Compound B, that inhibits OXSR1 phosphorylation, are used in support of the above conclusion. I do not really see a connection between OXSR1 expression and the inhibitor experiment.

      We will reword “expression” to “activity” in regards to the inhibitor experiment.

      In line 187, "Knockdown of tnfa reduced the amount of infection-induced tnfa promoter-driven GFP produced around sites of infection ....". How can a knockdown of tnfa affect the GFP expression driven by the tnfa promoter ?

      The promoter fragment used in the TgBAC construct contains target sites for two of our guide RNAs. We will also include qPCR validation of the knockdown.

      Reviewer #3 (Significance (Required)): Mechanism underlying decreased intracellular potassium level is of great interest in the inflammasome field. However, their observation is not in line with published studies. Audience in the pathogen-host interaction field will be interested. Expertise: dissection of signalling pathway regulation, molecular and cellular mechanism underlying NLRP3 inflammasome activation. We are not using zebrafish model.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      In this study, the authors suggest to have evidence for OXSR1 to inhibit NLRP3 inflammasome activation by limiting potassium efflux during mycobacterial infection. To my opinion, the study lacks important results supporting their main conclusions. In many instances, the authors have over-interpreted their data and I therefore do not support publication of this study.

      Main comments:

      Activation of the NLRP3 inflammasome upon OXSR1 knockdown was not convincingly demonstrated. Clearance of bacteria in an organism, herein zebrafish, involves mechanisms in different cell types including downstream of inflammasome activation. Thus, bacterial clearance experiments in THP-1 cells might not necessarily be related to in vivo experiments in an organismal context. Finally, a mechanism as to how mycobacteria enhance OXSR1 expression to block a NLRP3-mediated response has not been addressed.

      Specific comments:

      1. The author showed that the M. marinum ESX1 secretion system induced OXSR1 expression to inhibit the NLRP3 inflammasome activation. This is contradictory to another recent study (PMID: 18852239), which showed that the ESX1 secretion system activated the NLRP3 inflammasome.
      2. In line 102, based on Data shown in Fig 1D, the authors concluded that homozygous, but not heterozygous, oxsr1asyd5 embryos showed reduced bacterial burden. However, in Fig 1D, the difference among the genotypes is not significant.
      3. In line 196, the authors stated that "We present evidence that pathogenic mycobacteria increase macrophage K+ concentration by inducing expression of OXSR1." However, the authors did not provide evidence for this.
      4. Based on Extended data 3, the authors concluded that infection increases the expression of OXSR1. However, this is not evidenced in the Western Blot. In addition, in panel B, the OXSR1 blot showed many non-specific bands with decreased intensity in OXSR1 knockdown conditions suggesting that there is unequal protein loading making it impossible to interpret these results.
      5. The authors concluded that infection-induced OXSR1 expression suppressed inflammasome activity to aid mycobacterial infection. Experiments with Compound B, that inhibits OXSR1 phosphorylation, are used in support of the above conclusion. I do not really see a connection between OXSR1 expression and the inhibitor experiment.
      6. In line 187, "Knockdown of tnfa reduced the amount of infection-induced tnfa promoter-driven GFP produced around sites of infection ....". How can a knockdown of tnfa affect the GFP expression driven by the tnfa promoter ?

      Significance

      Mechanism underlying decreased intracellular potassium level is of great interest in the inflammasome field. However, their observation is not in line with published studies.

      Audience in the pathogen-host interaction field will be interested.

      Expertise: dissection of signalling pathway regulation, molecular and cellular mechanism underlying NLRP3 inflammasome activation. We are not using zebrafish model.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Hortle et al, in this study evaluated the role of WNK kinases SPAK and OXSR1 during M. marinum and M. tuberculosis infection. These two kinases inhibit the KCC channels which have a tendency to export potassium out of the cell. Since potassium efflux is a known stimulator of NLRP3 inflammasome activation, this raises the possible role of these kinases in inflammation and infection. Authors showed that inhibiting OXSR1 genetically and chemically reduced the mycobacterium survival in cells and zebrafish model, thus proposing OXSR1 as a host-directed therapeutic candidate. They showed that knockdown of OXSR1a leads to NLRP3 inflammasome mediated IL1B induction, which results in increase in TNFa and suppression of mycobacterium growth. Furthermore, reduction in mycobacterium growth in OXSR1a KD zebrafish embryos was found to be dependent on ESX1 machinery of Mycobacterium.

      The role of potassium in regulating Mycobacterium host response is novel. However there are few things which are missing from this interesting work.

      Main comments

      1. Since OXSR1 is known to inhibit KCC channels, which will lead to increase intracellular potassium. Why in infected control cells there is no increase potassium, Fig 2C. What would be the role of potassium in OXSR1 mediated control of Mtb growth? Does addition of extracellular potassium restricts mycobacterium in OXSR1-KD cells?
      2. Since OXSR1 is known to inhibit KCC channels, What happens to the activity of these channels in OXSR1 KD cells? This is important, because authors could not find any difference in intracellular potassium between uninfected control and uninfected OXSR1 KD cells (Fig 2C). It will be good to add the flowcytometric histogram or dot plots of potassium staining in the main figure or in extended figures.
      3. Acquisition of potassium stained cells - In methodology it has been mentioned that ion K+ Green stained undifferentiated THP1 cells were acquired using PE channel while differentiated THP1 cells were acquired using FITC channel. Furthermore in methods its mentioned that Leica Sp8 microscope was used to acquire images, however I do not see any of this data in the manuscript.
      4. Fig 2E and 3D - Meaning of "Normalized CFU/ml"? Each dot represents what? How many times this experiment was performed, please add in the legend.
      5. Fig 1D - What could be the reason of no statistical significant difference between wild type and homozygous oxsr1a-KO fish?
      6. Good to have a schematic model showing the finding s of the study
      7. TNFa is double edge sword and can lead to pathology. Hence treatment of chronically infected animals (say mice) by Compound B, will be needed to confirm the HDT activity of OXSR1.

      Significance

      This study showed role of kinases, which regulate trafficking of potassium, in mycobacterium-host interaction. Since kinases are draggable, so this opens a new area for developing host-directed therapies for TB.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      In this manuscript, using in vivo infection of Zebrafish embryos with Mycobacterium marinum and THP1-derived macrophages infected with Mycobacterium tuberculosis, the authors show that these pathogenic mycobacteria trigger an increase of K+ concentration through the expression of OXSR1. The ESX1 secretion system that is essential for the virulence of M. marinum is required for the expression of OXSR1 and SPAK. OXSR1 and SPAK are involved in the WNK signaling pathway and are cytoplasmic serine/threonine protein kinases that regulate the function of a series of sodium, potassium and chloride co-transporters via phosphorylation. Given that K+ efflux is now accepted as the main inducer of NLRP3 inflammasome, the authors report that this infection-induced OXSR1 expression restrains the protective NLRP3 inflammasome response leading to IL-1b maturation and secretion. Il-1b as a very potent pro-inflammatory triggers TNF-a production and the authors demonstrate that infection-induced OXSR1 expression suppressed host protective TNF-a and cell death early in fection. It appears therefore that virulent mycobacteria induce OXSR1 expression to reduce inflammasome activation by maintaining high intracellular K+.

      The results presented by the authors are convincing and the conclusions raised by the authors are well supported by the data.

      In zebrafish embryos, OXSR1 knockdown nicely reduces mycobacteria burden. Based on their conclusions that infection-induced OXSR1 expression reduces NLRP3 inflammasome activation, NLRP3 inflammasome activation has therefore a protective effect against bacterial infection. My main concern is that surprisingly, nlrp3 or il1b knockdown has no effect on bacterial burden in comparison to control embryos. Lane 256, as an explanation, the authors wrote "This may have been because we were using mosaic F0 CRISPR knockout, which is not a complete removal". The removal using mosaic F0 CRISPR knockout is nevertheless sufficient to observe a decrease in bacterial burden following OXSR1 knockdown. Would it be possible that OXSR1 also regulates immunity independently of NLRP3 inflammasome?

      Other comments:

      OXSR1 WB in extended Data 3 is really poor quality so that it is hard to see the increased expression of OXSR1 following infection.

      Figure 2C. It is not shown but I guess that similar results should be obtain using M. tuberculosis.

      Figures 5D and 5E. To confirm the involvement of NLRP3, in addition of using MCC950, NLRP3 knock down using siRNA should be also performed. NLRP3-deficient THP-1 cells are also commercially available if the siRNA-mediated knock down of NLRP3 is not convincing enough.

      Minor comments:

      How do the authors think that mycobacterium induces OXSR1 expression following infection? It has not been investigated and it is not discussed.

      Significance

      The observations reported in this manuscript are interesting since for the first time, it is described that virulent mycobacteria induce OXSR1 expression to reduce NLRP3 inflammasome activation by maintaining high intracellular K+. This is quite a significant advance in the field. To escape immune control, many successful intracellular pathogens have evolved methods to limit inflammasome activation. While it is known that potassium efflux is a trigger for inflammasome activation, the interaction between mycobacterial infection, potassium efflux and inflammasome activation was not explored.

      My field of expertise is the regulation of inflammasome activation. As far as I remember, I've never reviewed a paper using zebrafish embryos but here, the explanations and data are clear so that it was easy to understand and to evaluate. Likewise, I did not know the WNK signaling pathway but the literature clearly shows that it is involved in intracellular ionic balance.

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

      Learn more at Review Commons


      Reply to the reviewers

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

      In this project, authors develop a colorimetric and luminescence assay for the detection of SARS-CoV-2 RNA in vitro. They design an RNA based sensor that will be triggered by target RNA then release the ribosome binding site and a translation start site followed by a reporter gene. The released sequence will then trigger the production of reporter protein by transcription-translation coupled assay. Authors also introduce an RNA amplification step in order to increase the sensitivity of this assay.

      **Strengths:**

      This assay provides a simple, rapid way to detect SARS-CoV2 and it is an elegant way to incorporate transcription-translation coupled assay for SARS-CoV-2 RNA detection and identify SARS-CoV-2 patient samples. It is a nice assay and the performance is comparable with the existing method.

      **Weaknesses:**

      However, the positioning of this assay is not very clear. The readout of this assay could be recorded by camera whereas it includes several steps such as RNA extraction, amplification, transcription-translation coupled assay and reporter reaction. The limitations of the existing methods (RT-PCR, paper strip) and the advantages of this assay haven't been demonstrated by the experiments. The stability of RNA may also restrict the application of the proposed assay on site.

      **Major comments:**

      Authors are suggested to design an experiment to show the advantage of this assay compared with the existing method.

      Response: We thank the reviewer for pointing this out. In Fig 5, we show a comparison of our assay with the bench mark in COVID-19 diagnostics, which is the RT-qPCR assay. We specifically correlate the Ct- values obtained for RT-qPCRs with the amount of color or luminescence obtained through our assay. From these experiments we note that the sensitivity of our assay is a lttle less than the RT-qPCRs where our assay does not detect Ct-values in the 36 to 38 range (very low viral loads). This comparative experiment highlights that our assay bears clear advantages over the RT-qPCR in terms of ease of assay set up, ease of color detection, amenability to cell-phone imaging and no requirement of sophisticated equipment or technical training to interpret results. The full details of these comparisons are discussed in the manuscript.

      This is consistent with the literature on COVID-19 diagnostics where new assays are routinely bench-marked against the “gold-standard” RT-qPCR assay ((Corman et al., 2020; Pearson et al., 2021).

      What is the limit of detection of this assay using LacZ and Luciferase reporter respectively?

      Response: The limit of detection of the assay as shown in Fig 4B and Fig 4C-D, was found to be 100 copies of RNA, which translates to a concentration of 8 attomolar RNA. In this case, we find the limit of detection to be the same for both LacZ (Fig 4B) and Luciferase (Fig 4C-D) reporter.

      The calculations of copy number and sensitivity were made using a commercial source of synthetic CoV-2 RNA (Twist Biosciences) that is used in several studies about COVID-19 diagnostics (Joung et al., 2020; Rabe & Cepko, 2020; Wu et al., 2021). The RNA copy numbers are taken from the product details provided by the manufacturer. These details are now clearly stated in the manuscript. The commercial RNA is provided at 106 copies per ul. From this we take as low as 100 copies per 20ul of NASBA reaction, which we are able to detect using our assay. Hence our sensitivity comes to 8 attoMolar. We have clarified this in the manuscript. We noticed a typo in the original submission where we refer to a sensitivity of 80 attomolar in the Discussion. This is corrected to 8 attomolar. With this sensitivity we are within the range to detect RNA in patient samples, as confirmed by our patient data.

      Authors have not examined the selectivity of this assay. What is the specificity, selectivity for each of these variants? Does altering target RNA change the specificity?

      Response: We thank the reviewer for raising this point. As recommended by the reviewer, we have now examined the selectivity of this assay through new data (See new Fig S3, new Fig S4 and new Fig S8, also shown below).

      We have examined selectivity in 3 different ways.

      1. Is our sensor selective to the said region of the SARS-CoV-2 genome? To address this, we generated 19 different Target (Trigger) RNAs spread across the SARS-CoV-2 genome. These were tested against Sensor 12 to examine for their ability to trigger the sensor. We find that our sensor is highly selective for its target RNA and does not show any detectable response to the other regions of SARS-CoV-2 (see new Fig S3).

      Next, we asked if our assay is selective to SARS-CoV-2 versus other related human corona viruses. For this, we first examined the sequence of the target RNA (Amplicon RNA 12) that is sensed by Sensor 12. We selected equivalent regions of RNA from a different coronavirus, the HKU1 human coronavirus family. We generated these RNA sequences in vitro and performed IVTT. These new data are shown in new Fig S4 and below. We find that the human coronavirus (HKU1) RNAs are not able to turn on our sensor, whereas the cognate SARS-CoV-2 RNA is able to.

      We then asked if our assay can detect a current prominent variant of SARS-CoV-2. A major cause of concern is the ability of SARS-CoV-2 to accumulate mutations in its genome, resulting in different variant strains of SARS-CoV-2. Of these variants, the Delta variant (B.1.617.2) is not only highly contagious but has been noted as a possible vaccine breakthrough mutant of SARS-CoV-2. For this, we obtained RNA from the patient nasopharyngeal swab samples from the NCBS-inStem Covid-19 testing Center, Bangalore, India. RNA was isolated in the BSL-3 facility at the testing center. RNA samples were sequenced and confirmed to be the Delta variant- B.1.617.2 (sequences deposited in GASIAD). RNA extracted from these patient samples were tested against Sensor 12 using NASBA followed by IVTT. We find that our assay can efficiently detect the Delta variant SARS-CoV-2 RNA from patient samples with a build up of color, but no color was observed from control samples. These new data are shown below and in new Fig 5F and new Fig S8. The ability to detect the Delta variant of SARS-CoV-2 is an important feature of our sensor since this variant is now of global concern and extensively found in the population, even becoming the dominant variant in several countries (Callaway, 2021; O’Dowd, 2021; Torjesen, 2021).

      In Figure 2C-F, sensor 17 showed higher fold change and sensitivity. Why was sensor 12 selected for further study in Figure 3

      Response: The reviewer rightly notes that sensor 17 responds to 1012 copies of RNA and hence appears to be inherently more sensitive than sensor 12, which responds to 1013 copies of RNA. However, neither of these sensitivities are good enough to detect the levels of viral RNA found in patient samples. Hence we coupled these sensors with a step of NASBA amplification. The screen to identify pairs of NASBA primers gave us great hits for sensor 12 right off the bat, where we could detect down to 100 copies of RNA. Hence we moved forward with sensor 12 for further experiments. This has now been clarified in the manuscript.

      Authors should show the error bar in all plots. Authors should also indicate what the error bar means (SD, S.E.M. etc.) throughout the manuscript.

      Response: This is an important point. We have added the error bars and statistical analyses to all relevant plots. We have included the description of these statistical parameters in the figure legends throughout the manuscript, where relevant. Alternatively, experimental replicates are indicated and shown in the revised manuscript. Specifically in Figures 2 and 3 and 4D we have performed statistical analysis to include p-values to show significance of the data. For the data in Figure 4 B-C we include the experimental replicates as a new Supplementary Figure (see new Fig S5). Data in Figure S5 is now updated to include the experimental replicates. For the patient data in Figure 5, we have included details of specificity and sensitivity analysis for clinical samples (see new Fig 5C).

      **Minor comments:**

      "This method is relatively faster but may generate false positives due to non-specific amplification and primer interactions." Reference is needed.

      Response: We have now added the following references in support of this statement. (Gadkar, Goldfarb, Gantt, & Tilley, 2018; Sahoo, Sethy, Mohapatra, & Panda, 2016)

      "using the softwares Primer 3 and NUPACK." Reference is needed.

      Response: We have now added the following references (Untergasser et al., 2012; Zadeh et al., 2011)

      Reference 15 belongs to CRISPR-CAS based assay but it was cited under RT-LAMP assay.

      Response: This has now been corrected. We thank reviewer for this.

      Reviewer #1 (Significance (Required)):

      This paper will be of interest to scientists interested in developing diagnostic tools for the detection of SARS-CoV2 in viral and host pathogenic sequences; genetic disorders and development of precision medicine.

      Reviewer works in the field of Chemical Biology and Nanotechnology including sensor development and the application in diagnosis, cell physiological studies.

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

      In this manuscript, Charkravarthy et al. report a new method for detecting SARS-CoV-2 RNA in both in vitro and human saliva and nasal samples. The new detection method, PHANTOM, is capable of detecting as few as 100 copies of the SARS-CoV-2 genome. The method is demonstrated to reproducible over a large range of viral titers and results in a binary report on CoV-2 infection. From my perspective the results are strong and fairly convincing (please see comments below). There is clear, logical, flow to the experiments and engineering of the PHANTOM system. The collaborative work is well organized and logical. The work is clearly of high significance and certainly merits expedited review and publication. I would like to unambiguously state that support publication of this manuscript in its current form in the non-peer reviewed context of this journal, would be more than happy to provide further peer review of this manuscript upon submission to another journal, and would be more than happy to provide further comments if requested by the authors.

      My personal background is broad in range, however, I have a long track record of research in RNA folding, structural biology, biosensor development, and bioinformatics. Given this knowledge base, I found the manuscript rather easy to read and digest. The manuscript is well written and clear. In order to expedite the process of review I will not give a detailed review which would include grammatical errors (there are are very few). Rather, I will touch on the most pressing issues I see.

      **Major concerns:**

      1) There a number of figures that do not show a statistical measure of significance (e.g. error bears, ANOVA, etc.). It is essential that these be included in the final peer reviewed publication. (See Figure 2A, Figure 3D, Figure 4B, Figure 4C, Figure 5A, Figure 5C, Figure 5D).

      Response: This is an important point. We have added the error bars and statistical analyses to all relevant plots. We have included the description of these statistical parameters in the figure legends throughout the manuscript, where relevant. Alternately, experimental replicates are indicated.

      Specifically in Figures 2 and 3 and 4D we have performed statistical analysis to include p-values to show significance of the data. For the data in Figure 4 B-C we include the experimental replicates as a new Supplementary Figure (see new Fig S5). Data in Figure S5 is now updated to include the experimental replicates. For the data in Figure 5, we have included details of specificity and sensitivity analysis for clinical samples (see new Fig 5C).

      2) There are some important points that do not include references within the manuscript. I believe that the authors should reference Abdolahzadeh et al. RNA 2019 in the introduction. This manuscript describes another NASBA viral detection system using fluorescent RNA reporters (also see Trachman et al. Q. Rev. Biophys 2019, for reference on fluorescent aptamers). Also see the ROSALIND method (Jung et al. 2020 Nature Biotechnology) for detecting water contaminants using visual identification by fluorescent aptamers.

      Response: We have added the above mentioned references to the manuscript as suggested by the reviewer.

      3) The discussion states that "The overall sensitivity in the attomolar range ensures detection of infection in the majority of Covid-positive patients in a population". Please provide a reference to support this and explicitly state the concentration of viral RNA in patient samples. There are a number of times that the copy number of viral genomes and sensitivity of the measurement is stated throughout the manuscript. There should also be a reference and statement about concentration.

      Response: The reviewer has raised multiple connected points here, which we address in the revised manuscript.

      1. Concentration of RNA in patient samples: We have added the references (Pujadas et al., 2020; Wyllie et al., 2020) where the authors report that the typical concentration of viral RNA in patient nasopharyngeal swab samples lies in the range of 104 to 105 copies of RNA per ml. This translates to a concentration range of 10 to 100 attoMolar. This reference is now added to the manuscript. For the patient samples used on our study, we refer to the Ct- values obtained from the RT-PCR tests and correlate Ct values to the readout from our assay, consistent with other reports on COVID-19 diagnostics ((Joung et al., 2020; Vogels et al. 2020; Wu et al., 2021).

      Copy number and sensitivity: As the reviewer notes, we refer to viral genome copy number and sensitivity of our assay in the manuscript. These calculations of copy number and sensitivity were made using a commercial source of synthetic CoV-2 RNA (Twist Biosciences) that is used in several studies about COVID-19 diagnostics (Joung et al., 2020; Rabe & Cepko, 2020; Wu et al., 2021). The RNA copy numbers are taken from the product details provided by the manufacturer. These details are now clearly stated in the manuscript. The commercial RNA is provided at 106 copies per ul. From this, we take as low as 100 copies per 20ul of NASBA reaction, which we are able to detect using our assay. Hence our sensitivity comes to 8 attoMolar. We have clarified this in the manuscript. We noticed a typo in the original submission where we refer to a sensitivity of 80 attomolar in the Discussion. This is corrected to 8 attomolar. With this sensitivity we are within the range to detect RNA in patient samples, as confirmed by our patient data.

      Reviewer #3 (Significance (Required)):

      I think this is a significant advancement in the field. The introduction of smartphone technology to this robust diagnostic is very attractive. The work is of high significance since the researchers demonstrated robust responses against SARS-CoV-2 variants. As well all now know these are on the rise and cheap robust detection methods are essential for containing this virus.

      Response: We thank the reviewers for the positive comments.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      In this manuscript, Charkravarthy et al. report a new method for detecting SARS-CoV-2 RNA in both in vitro and human saliva and nasal samples. The new detection method, PHANTOM, is capable of detecting as few as 100 copies of the SARS-CoV-2 genome. The method is demonstrated to reproducible over a large range of viral titers and results in a binary report on CoV-2 infection. From my perspective the results are strong and fairly convincing (please see comments below). There is clear, logical, flow to the experiments and engineering of the PHANTOM system. The collaborative work is well organized and logical. The work is clearly of high significance and certainly merits expedited review and publication. I would like to unambiguously state that support publication of this manuscript in its current form in the non-peer reviewed context of this journal, would be more than happy to provide further peer review of this manuscript upon submission to another journal, and would be more than happy to provide further comments if requested by the authors.

      My personal background is broad in range, however, I have a long track record of research in RNA folding, structural biology, biosensor development, and bioinformatics. Given this knowledge base, I found the manuscript rather easy to read and digest. The manuscript is well written and clear. In order to expedite the process of review I will not give a detailed review which would include grammatical errors (there are are very few). Rather, I will touch on the most pressing issues I see.

      Major concerns:

      1) There a number of figures that do not show a statistical measure of significance (e.g. error bears, ANOVA, etc.). It is essential that these be included in the final peer reviewed publication. (See Figure 2A, Figure 3D, Figure 4B, Figure 4C, Figure 5A, Figure 5C, Figure 5D).

      2) There are some important points that do not include references within the manuscript. I believe that the authors should reference Abdolahzadeh et al. RNA 2019 in the introduction. This manuscript describes another NASBA viral detection system using fluorescent RNA reporters (also see Trachman et al. Q. Rev. Biophys 2019, for reference on fluorescent aptamers). Also see the ROSALIND method (Jung et al. 2020 Nature Biotechnology) for detecting water contaminants using visual identification by fluorescent aptamers.

      3) The discussion states that "The overall sensitivity in the attomolar range ensures detection of infection in the majority of Covid-positive patients in a population". Please provide a reference to support this and explicitly state the concentration of viral RNA in patient samples. There are number of times that the copy number of viral genomes and sensitivity of the measurement is stated throughout the manuscript. There should also be a reference and statement about concentration.

      Significance

      I think this is a significant advancement in the field. The introduction of smart phone technology to this robust diagnostic is very attractive. The work is of high significance since the researchers demonstrated robust reposes against SARS-CoV-2 variants. As well all now know these are on the rise and cheap robust detection methods are essential for containing this virus.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      In this project, authors develop a colorimetric and luminescence assay for the detection of SARS-CoV-2 RNA in vitro. They design an RNA based sensor that will be triggered by target RNA then release the ribosome binding site and a translation start site followed by a reporter gene. The released sequence will then trigger the production of reporter protein by transcription-translation coupled assay. Authors also introduce an RNA amplification step in order to increase the sensitivity of this assay.

      Strengths:

      This assay provides a simple, rapid way to detect SARS-CoV2 and it is an elegant way to incorporate transcription-translation coupled assay for SARS-CoV-2 RNA detection and identify SARS-CoV-2 patient samples. It is a nice assay and the performance is comparable with the existing method.

      Weaknesses:

      However, the positioning of this assay is not very clear. The readout of this assay could be recorded by camera whereas it includes several steps such as RNA extraction, amplification, transcription-translation coupled assay and reporter reaction. The limitations of the existing methods (RT-PCR, paper strip) and the advantages of this assay haven't been demonstrated by the experiments. The stability of RNA may also restrict the application of the proposed assay on site.

      Major comments:

      Authors are suggested to design an experiment to show the advantage of this assay compared with the existing method.

      What is the limit of detection of this assay using LacZ and Luciferase reporter respectively?

      Authors have not examined the selectivity of this assay. What is the specificity, selectivity for each of these variants? Do altering target RNA change the specificity?

      In Figure 2C-F, sensor 17 showed higher fold change and sensitivity. Why was sensor 12 selected for further study in Figure 3?

      Authors should show the error bar in all plots. Authors should also indicate what the error bar means (SD, S.E.M. etc.) throughout the manuscript.

      Minor comments:

      "This method is relatively faster but may generate false positives due to non-specific amplification and primer interactions." "using the softwares Primer 3 and NUPACK." Reference is needed.

      Reference 15 belongs to CRISPR-CAS based assay but it was cited under RT-LAMP assay.

      Significance

      This paper will be of interest to scientists interested in developing diagnostic tools for the detection of SARS-CoV2 in viral and host pathogenic sequences; genetic disorders and development of precision medicine.

      Reviewer works in the field of Chemical Biology and Nanotechnology including sensor development and the application in diagnosis, cell physiological studies.

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

      Learn more at Review Commons


      Reply to the reviewers

      Dear Editor,

      We appreciate the critical and constructive comments that Reviewers made on our paper entitled “The structure and flexibility analysis of the Arabidopsis Synaptotagmin 1 reveal the basis of its regulation at membrane contact sites” by Benavente et al.. We are very pleased that three reviewers from Review commons appreciated the work, found it interesting and suggested constructive modifications to the paper. We have now responded to their suggestions and enclose a modified manuscript.

      As suggested by the reviewers we now provide additional biochemical evidence about the role of calcium on C2 activation, more controls in the biochemical assays, an improved protocol for the lipid binding assays as well as details in the molecular dynamics. In addition, following reviewers’ recommendations, we have rephrased some of the headings and statements in the discussion section, to adjust them better to our experimental results. Altogether, we believe that the manuscript is more solid and interesting than the previous version for the general readers of eLife.

      Specifically, we have made the following changes in response to the particular comments of the referees:

      Reviewer 1

      Major points

      The conclusion that SYT1C2A determines protein behavior by binding calcium and likely switching its function from a pure tether to a lipid transporter is accurately and convincingly portrayed. However, there are some problematic statements in the way this information is interpreted and used to propose a model.

        • p7: "This means that site I will be occupied depending on the physiological Ca2+ concentration to activate the protein and trigger Ca2+-dependent SYT1C2A lipid binding." Please exchange "means" by "suggests"* We agree with the reviewer and we have exchanged “means” by “suggests”
        • There is not enough data to determine whether SYT1C2A interacts with the SYT1SMP domain, much less to propose that it is the SYTC2A that loads the SMP with DAG. They may also operate independently by means of the unstructured hinge sequence between them. *
        • It is implied that SYTC2A would have to release its calcium atom for it to leave the PM and go to the ER to complete a single transport cycle. It is unclear whether SYT1C2A could bind and release calcium repeatedly and fast enough to cycle between membranes efficiently. It appears more plausible for it to remain at the PM until intracellular calcium concentrations decrease. **

        These statements should be marked as speculative. *

      We understand that the reviewer comments are related to steps 3 and 4 of our model for the function of SYT1 (figure7; pages 17 and 18). We have now included these scenarios suggested by the reviewer (page 17 and 18). In addition, we have discussed the work from Vennekate et al, PNAS, 2012, related to the cis and trans membrane interactions of the two C2 domains of human synaptotagmin (Syt1). The authors demonstrated that they are driven by the balance in the concentration of anionic lipids between target membranes and by Ca2+ and protein concentration. Likewise, the changes in the local composition of PIP and/or Ca2+ may drive the transfer of SYT1C2A from the PM to de ER and vice versa while the C2B remains at the PM.

      Most of the graphs shown (except Fig. 4A) appear to show only one experiment. Please include results from at least three independent experiments in all graphs, and analyze them statistically. Please describe only differences between samples that are statistically significant (for all graphs and tables).

      Following reviewed recommendation, we have assured that all the biochemical data arises from three independent measurements. Indeed, as indicated in the figure legends and methods section, this was the case for the Ca2+ binding experiments at figure 2C and for the lipid binding data at figure 4. We have now repeated the ITC experiments to provide three independent measurements of the dissociation constant of SYT1C2A for Ca2+.

      Minor comments

      Please include the PDB codes for the available structures used. It is not clear where the C2C domain structure comes from. We guess it may correspond to PDB entry 2DMG. A word of caution, this 2006 structure has been released but not published and its quality is suboptimal. After its appearance in Idevall-Hagren et al. 2015 supplementary images it has been re-cited in other publications. I would advise against making topology claims based on this structure (Figure S1), and only use it as an example for polybasic patches in C2 domains.

      We have included the PDB codes for the available structures used at the corresponding figure captions 1, 3 and S1.

      We agree that the stereochemistry of the refined structure of the E-Syt2 C2C domains (2DMG) is suboptimal. However, we have sought for the advice of an NMR expert (M.A. Jimenez, https://scholar.google.es/citations?hl=es&user=iAF0ymYAAAAJ) to evaluate the NMR restrains available alongside the entry at PDB. She concluded that the Halpha-Halpha distance restrains indicate that the connectivity of the beta strands is correct. The confusion may arise from the fact that the PDB entry has been updated twice since 2006, the last one in 2011.

        • p4: Schauder et al. 2014. Only the "shuttle" model is mentioned in the text. Although this may be the most plausible scenario, the "tunnel" model cannot be completely ruled out*. We have now stated that both shuttle or tunnel modes are possible for lipid transfer at page 4.
        • p5: Fernandez-Busnadiego et al. 2015 does not make a reference to the claim in the text.* Fernandez-Busnadiego et al. 2015 has been replaced by Collado et. al 2019
        • p18: Collado et. al 2019 don't assign the peak-forming function to the N-terminal hairpin domain of tricalbins. As the reviewer indicates, Collado et. al 2019 just suggests that the peak forming function could be due to the N-terminal hairpin domain of tricalbins: “This phenomenon may rely on the hairpin sequence that anchors Tcbs to the ER membrane. Tcb hairpin sequences could sense and/or generate membrane curvature as in reticulons and other ER morphogenetic proteins (Hu et al., 2011).” (Taken from Collado et al. 2019).*

      Accordingly, we have indicated that “It has been suggested that the insertion of the N-terminal hydrophobic end of Tcbs may induce the formation of peaks of strong curvature at the ER region facing the PM. These structures shorten the distance between the ER and the PM by ~7 nm and facilitate lipid transport (Collado et al., 2019)”

      Fig. 1A: the notation for the domains C2C-C2E of E-Syt1 is confusing and not described in the legend. Also, please mark which beta sheet is which more clearly in Fig 1B, it is very difficult to understand the labelling** Fig. 1C: several elements (e.g. types of boxes, "T" on the top) used in the figure are not described in the legend. Both beta sheets and mutations are described as "arrows" in the legend, please differentiate between vertical and horizontal.

      We have modified the Fig. 1A legend to explain the notation of C2 domains of human E-Syts “Human E-Syt1 displays five C2 domains while E-Syt2 and E-Syt3 display three”.

      Labeling of beta strands has been modified to make the panel clearer.

      We have rewritten the caption of figure 1C to describe the types of boxes and the “T” and to distinguish between vertical and horizontal arrows.

      Fig. 2A: indicate Lys 286, which is mentioned in the text

      The reference to Lys 286 in the text was not correct. We have modified the text to indicate that Lys275 replaces the Ca I as it is shown in figure 2.

      Fig. 2B: Inset too small to read.

      Labels for inset at Fig 2B have been enlarged.

      Fig. 3: please show the scale and explain the blue-red color code. Please mark the polybasic patch.

      We have explained the meaning of the blue and red code, marked the polybasic patch of SYT1C2A and E-Syt1C2C and indicated that all of them are scaled to the same value.

      Fig. 4A should show both mutants for both types of liposomes.

      We have included new liposome binding data including both mutants for both types of liposomes.

      Table 1: what is "control"?

      We have indicated that the unbound wild-type protein is taken as a control

      Fig. S5: Difficult to read, increase font size. The text talks about experiments with calcium and EGTA that are not shown in the figure.

      We have increased figure size to facilitate reading. We have also indicated that C2AB-Ca and C2A stand for the experiments carried out in presence of Ca and EGTA, respectively.

      Reviewer 2

      Minor comments

      Typos: please re-read carefully through the manuscript to remove them. We advise the authors to have the manuscript corrected by a native english-speaker.

      We have done a thorough revision of the manuscript to correct typos and to improve the English style of the manuscript

      Reviewer 2 considers that “The authors discuss their findings within the frame of experimental observations that are already published but these remain speculative”

      We are glad that the reviewer appreciates our work in “decrypting the roles of C2 individual operating modes” as it is a “central issue for providing functional specificity but also plasticity in response to developmental /environmental clues”. The reviewer also considers our work “important and identifies a number of very interesting features”. As already mentioned to the editor, the present version of the manuscript includes new biochemical data and analysis to support further the functions of C2A-C2B tandem. In addition, we have included new references and rephrased some of the headings and statements in the discussion section, to adjust them better to our experimental results.

      Reviewer 3

      Major points

      1)The authors investigated SYT1C2A calcium-binding sites using two different methods, ITC and differential scanning fluorimetry. By using ITC, they described the first binding site coordinating calcium ions in the nanomolar range. The second calcium-binding site was then characterized by differential scanning fluorimetry. The second calcium-binding site binds calcium with Kd of 277 µM. The authors then mutated SYT1C2A at two positions and performed again differential scanning fluorimetry. In this case, they did not observe any blue shift in intrinsic fluorescence concluding that "calcium-binding is mediated by the calcium-binding site". It is not clear which binding site the authors mean. In the structure of SYT1C2A, the mutated residues (D276 and D282) are shared by both calcium-binding sites. It is, therefore, difficult to interpret the data. The authors should generate a unique mutation for each site and perform both ITC and differential scanning fluorimetry

      The SYT1C2A-DADA double mutant was prepared to abolish Ca2+ binding to both site I and site II and to discard an unspecific effect of Ca2+ on intrinsic fluorescence that accounts for the low affinity Kd. Our data showed that the addition of Ca2+ to SYT1C2A-DADA does not produce a shift in intrinsic fluorescence of the protein; indicating that the observed Ca2+-binding activity is specifically mediated by the Ca2+ dependent lipid binding site. We clarify this point in the present version of the manuscript.

      In addition, following reviewer’s recommendation, we have prepared two additional point mutant proteins, SYT1C2A D276A and SYT1C2A E340A, to investigate the Ca2+ binding properties at site II and I, respectively. As expected, the reduction of one carboxylate ligand at the structural site II produces a drastic decrease in the Ca2+ binding affinity (Kd = 1.8± 0.5 mM) which is coupled with a decrease in thermal stability of the protein (Ti = 55°C) and a red-shift in fluorescence emission with respect to the wild type protein that resembles those effects observed for the SYT1C2A-DADA mutant (Figure S2A). Differentially, SYT1C2A Glu334Ala doubled the Kd (534 ± 60 mM) while reducing slightly its thermal stability (page 7 figure S2A)

      2)The authors described an increase in the inflexion point temperature with increasing calcium concentration. They noted that the effect was measurable from 30 µM to 300 µM. Looking at the plot with the first derivative ratio (Figure 2C), there is also an apparent change in the inflexion point temperature from 300 µM to 3 mM. Does this mean that the SYT1C2A domain binds more than two calcium ions?

      The ligand induced stabilization of proteins results in changes of the thermally induced melting curves for the ligand complexed relative to the uncomplexed proteins. This effect is used to unequivocally identify ligand hits for a particular protein from large libraries of compounds and to provide an initial estimation of the binding affinity. However, a deeper analysis of the ligand binding affinities using this technique is discouraged as the increase of melting temperature with the ligand concentration does not saturate to a particular value. This is why the effect of Ca2+ addition to SYT1C2A was also measurable from 300 µM to 3 mM, and it does not necessarily imply the binding of Ca2+ to another site. Consequently, we used other techniques such as ITC or the analysis of the change of intrinsic fluorescence upon Ca2+ addition to precisely characterize the Ca2+ binding affinities of SYT1C2A. In the present version of the manuscript, we have indicated that the change in Ti vs Ca2+ concentration is measurable when moving from 30 mM to 300 mM, thus demonstrating a Ca2+-binding event “is initiated” in this concentration range (page 7)

      3)To address lipid-binding properties of the SYT1 C2 domains, the authors used two independent methods, lipid co-sedimentation assay and BLI. In the Material and Methods section, the authors wrote that a solution of liposomes was sonicated for seven minutes to achieve homogeneity. How was homogeneity checked? Standard protocols for the liposome co-sedimentation assay use the extruder to achieve a homogeneous population of the liposomes. Also, the authors noted that the liposomes were resuspended in the buffer containing 50 mM Tris/HCl, 80 mM KCl, and 5 mM NaCl. The liposomes are usually loaded with sugar molecules, like raffinose at this step to allow subsequent co-sedimentation using centrifugation.

      Following the reviewer recommendation, we have used an extruder to achieve a homogeneous population of the liposomes (see the Methods section). This has produced an improvement of the data from the statistical point of view. However, we did not employ any sugar to facilitate lipid sedimentation as we found that 1 hour centrifugation at 58,000 rpm using a TLA100 rotor (Beckman) was enough to separate adequately soluble and precipitated fractions.

      The authors wrote that the samples were centrifuged at 58,000 rpm. The information does not allow reproducibility without rotor specification.

      We have now included the rotor specification in the methods section.

      The bound protein was estimated by subtracting the supernatant from the total amount of protein used in the assay. Direct estimation of the protein amounts in the bound fraction via e.g. SDS-PAGE would be more suitable.

      We respectfully disagree with the reviewer in this issue. The reviewer may note that the differences in the fraction of bound protein to the liposomes are at maximum around 25%. Such values are statistically significant using spectrophotometric techniques but there will be less accurately determined by the analysis of an SDS-PAGE. In addition, the later would require several washing steps of the insoluble fraction that may induce additional errors. This is well documented in a previous work from our group (Diaz et al, PNAS 2015). There, the lipid binding properties of WT and mutant C2 domain are compared using both approaches; in this work it is shown that the spectrophotometric techniques showed significant differences with the SDS-PAGE, which resulted just indicative.

      Nevertheless, we have prepared the requested SDS-PAGE for reviewer evaluation (see below) and if required we will include it as a new panel in figure 4 or include it as supplementary material. The SDS-PAGE shows the amount of soluble protein after the incubation with liposomes. It is clearly shown a reduction in the amount of WT and DADA soluble protein upon incubation with PCPSPI liposomes with respect to the sample incubated with PCPS liposomes. Differentially, no effect is observed for the PolyB and WT in presence of EGTA. Sample migrates abnormally producing a double band in presence of EGTA, probably due to the effect of removing the structural Ca site.

      Critical controls for the liposome co-sedimentation assay are missing:

      Do the SYT1 C2 domains bind liposomes without negatively charged lipids (i.e. PC-only liposomes)?

      Does the SYT1C2A-DADA mutant domain bind the PC/PS/PI liposomes?

      Does the SYT1C2A-PolyB mutant interact with the PC/PS liposomes?

      We have included all the controls suggested by the referee in the present version of the manuscript, in the results section and in figure 4A.

      The authors wrote both in the main text and the figure 4 legend that they used a lipid monolayer in the BLI method. However, in the Material and Methods section, they wrote that they used small unilamellar vesicles.

      The starting material for lipid monolayer immobilization at the biosensor tip is a solution of small unilamellar vesicles. We have clarified this issue in the methods section.

      4)The authors beneficially used all-atom MD simulations to address mechanistic details of the SYT1 C2 domains with two different lipid bilayers. However, several issues need to be addressed.

      Is there a particular reason to include sitosterol in the MD simulations? Does sitosterol contribute to protein binding?

      We clarify this issue in the present version of the manuscript. The use of sitosterol or stigmasterol in research involving plant membranes and membrane-associated proteins is recommended (DOI: 10.1063/1.4983655 PMID: 28595398) to recapitulate the fluidity and thermotropic properties of model membranes (DOI: 10.1016/j.colsurfb.2019.110422 PMID: 31437609). Sitosterol, as is the case for cholesterol in mammal membranes, contributes to packing the lipid bilayer more tightly into a liquid ordered phase (DOI: 10.1016/j.chemphyslip.2017.01.003 PMID: 28088325) and this aspect is important in molecular dynamics simulations to attain equilibration more effectively (DOI: 10.1016/j.jcis.2011.02.048 PMID: 21429500).

      We observed one hydrogen bond contact between Asn 338 at loop L3 and sitosterol in PSPI-M. We clarify this point in the results section.

      Why was not the simulation ran for a longer time? What was a criterion to determine that the system reached a stable state? Results of the MD simulations are presented as static snapshots. The manuscript would benefit from a more detailed analysis, e.g. the number of hydrogen bonds between the protein and phospholipid molecules over time, development of the tilt angle over time, etc.

      Following the reviewer’s recommendations, we now provide data showing the dynamic features of our MD calculations. In particular, we have compared the overlay of the different structures along the simulation of the SYT1C2A domain attached to the PS-Membrane and to the PSPIP-Membrane highlighting those amino acids hydrogen bonded to phospholipids (Figure S4A and S4B). This picture illustrates well that the pattern of interactions is conserved along the simulation. In addition, it also shows that the orientation of the domain with respect to the plane of the membrane is conserved. In this respect, as the reviewer requested, we have also included, a picture illustrating the change in the tilt angle of the C2A with respect to the plane of the membrane along the simulation. This analysis reveals that the tilt angle with respect to the membrane is 20 degrees larger for the PSPIP-Membrane than for the PS membrane due to the interaction of PIP molecules with the polybasic site (Figure S4E).

      In addition, we now present time traces illustrating the course of the simulations. In particular, those corresponding to the RMSD of the individual SYT1C2A, SYT1C2B and the linker between them. They clearly show that the simulations reached equilibrium within the time sampled (Figure S5H).

      The authors performed the MD simulations for both SYT1 C2 domains. It would be informative to include an electrostatic potential mapped on the surface of the SYT1C2B domain similarly to figure 3. The experimental results showed that SYT1C2B binds liposomes containing PC/PS. How does SYT1C2B interact with the PC/PS membrane? The authors noted that SYT1C2 inserts loop L3 into the lipid bilayer and that this loop adopts a β-hairpin structure. This is the case for the simulation with the SYT1C2AB fragment, but not for the SYT1C2B domain alone. Is the β-hairpin formed during the MD simulation? Or is it a result of the template-based modelling?

      To clarify reviewers’ concerns, we have included a supplementary figure illustrating the RMSD per residue along the MD simulation for the SYT1C2AB protein fragment in solution, and attached to PSPI-M. The representation illustrates an overall reduction of the RMSD as a result of the protein stabilization at the membrane, which is more significant at the membrane binding loops. In particular, the highly flexible SYT1C2B L3 loop in solution becomes highly stabilized upon membrane interaction and it folds as a beta hairpin.

      We have calculated the electrostatic potential mapped on the surface of the SYT1C2B. As it does not come from an experimental structure, we decided not to include in figure 3. The map, which is depicted below in the orientation shown in figure 3, shows that SYT1C2B might not display a polybasic site in accordance with our experimental results.

      How is the proposed binding model of the SYT1C2AB affected by the data obtained using SAXS?

      SAXS data showed that SYT1C2AB fragment may adopt a V-shaped compact structure and an extended conformation in which there is no interaction between the C2A and C2B domains. Interestingly, the V-shaped structure in solution resembles the proposed binding mode for SYT1C2AB to the membrane. We clarify this point in page 15.

      Minor comments

      1) SYT1CB construct is not listed in figure 1A along with the other constructs used in the study.

      We have modified Figure 1B to include SYT1C2B

      2) Phospholipids contain a phosphate group rather than a phosphoryl group.

      We have modified the text to correct this.

      3) In the text, the authors sometimes used the ratio 350 nm/330 nm and sometimes 330 nm/350 nm.

      This has been corrected in Figure 2C

      4) Figure S2, displayed curves look strikingly similar, different line representation does not allow a proper comparison. There are no units for y-axes.

      The figure has been corrected according to the reviewer’s recommendation for proper comparison.

      5) Figure 4B, y-axes do not have scales.

      The Y-axis from the BlitZ system represents the thickening of the layer of proteins attached to the biosensor tip and it is given in nm. However, this parameter could be meaningless when comparing different membranes and/or proteins. Hence, we scaled the plots to the maximum thickness for comparison purposes.

      6) Figure 5B, IP3 or I3P?

      We have corrected the labels corresponding to inositol triphosphate IP3 at figure 5B

      7) The authors noted that based on their SAXS experiments, the SYT1-SMPC2A has a maximum size of 176 Å and they wrote that this value is in accordance with the size of the previously characterized E-Syt2-SMPC2AB protein. However, as the authors reported, E-Syt2-SMPC2AB is two times smaller.

      We agree with the reviewer, the size of the SMP tunnel of E-Syt2-SMPC2AB is two times smaller than the Dmax size of SYT1 SMPC2A. We clarify this point in the text.

      Yours sincerely

      Armando Albert

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      The focus of this manuscript is the structural characterization of Arabidopsis synaptotagmin 1 (SYT1), which is a plant ortholog of the mammalian extended synaptotagmins (E-Syts). E-Syts play a principal role in the non-vesicular lipid transfer at the endoplasmic reticulum - plasma membrane contact sites. The function of the extended synaptotagmins is given by a unique combination of the evolutionarily conserved domains. The authors combined both experimental and computational methods to characterize different domains of Arabidopsis SYT1.

      The authors solved the structure of one of the SYT1 C2 domains (SYT1C2A) using X-ray crystallography. The obtained structure is highly similar to the previously characterized C2A domain of mammalian E-Syt2. In contrast to the orthologous C2A domain of E-Syt2, only two calcium ions are coordinated by the SYT1C2A domain. Next, the authors aimed to characterize the identified calcium-binding sites by two different experimental techniques, namely isothermal titration calorimetry (ITC) and differential scanning fluorimetry. To study lipid interactions of SYT1, the authors used a combination of the experimental approach (liposome-binding assays and biolayer interferometry, BLI) and computational methods (ligand-protein docking and molecular dynamics simulations). The authors found that both SYT1 C2 domains can interact with the lipid bilayer containing anionic phospholipids. Intriguingly, the authors described two lipid-interacting sites in the SYT1C2A domain. One site is involved in the coordination of phosphatidylserine (PS) in a calcium-dependent manner, and the second one coordinates molecules of phosphatidylinositol 4,5-bisphosphate (PIP2) through mostly electrostatic interactions. Last, the authors study the flexibility of SYT1 using partially overlapping protein fragments via small-angle X-ray scattering. Three hinge points are suggested to be responsible for the high flexibility of SYT1.

      The notion that SYT1C2A domain confers two independent lipid-binding sites, one unique for PIP2 molecules, and the second one regulated by calcium ions, is very compelling. In addition, the results describing the potential SYT1 flexibility provide novel insight into the function of extended synaptotagmins. However, my enthusiasm is mitigated by several points listed below:

      1)The authors investigated SYT1C2A calcium-binding sites using two different methods, ITC and differential scanning fluorimetry. By using ITC, they described the first binding site coordinating calcium ions in the nanomolar range. The second calcium-binding site was then characterized by differential scanning fluorimetry. The second calcium-binding site binds calcium with Kd of 277 µM. The authors then mutated SYT1C2A at two positions and performed again differential scanning fluorimetry. In this case, they did not observe any blue shift in intrinsic fluorescence concluding that "calcium-binding is mediated by the calcium-binding site". It is not clear which binding site the authors mean. In the structure of SYT1C2A, the mutated residues (D276 and D282) are shared by both calcium-binding sites. It is, therefore, difficult to interpret the data. The authors should generate a unique mutation for each site and perform both ITC and differential scanning fluorimetry.

      2)The authors described an increase in the inflexion point temperature with increasing calcium concentration. They noted that the effect was measurable from 30 µM to 300 µM. Looking at the plot with the first derivative ratio (Figure 2C), there is also an apparent change in the inflexion point temperature from 300 µM to 3 mM. Does this mean that the SYT1C2A domain binds more than two calcium ions?

      3)To address lipid-binding properties of the SYT1 C2 domains, the authors used two independent methods, lipid co-sedimentation assay and BLI. In the Material and Methods section, the authors wrote that a solution of liposomes was sonicated for seven minutes to achieve homogeneity. How was homogeneity checked? Standard protocols for the liposome co-sedimentation assay use the extruder to achieve a homogeneous population of the liposomes. Also, the authors noted that the liposomes were resuspended in the buffer containing 50 mM Tris/HCl, 80 mM KCl, and 5 mM NaCl. The liposomes are usually loaded with sugar molecules, like raffinose at this step to allow subsequent co-sedimentation using centrifugation. The authors wrote that the samples were centrifuged at 58,000 rpm. The information does not allow reproducibility without rotor specification. The bound protein was estimated by subtracting the supernatant from the total amount of protein used in the assay. Direct estimation of the protein amounts in the bound fraction via e.g. SDS-PAGE would be more suitable. Critical controls for the liposome co-sedimentation assay are missing. Do the SYT1 C2 domains bind liposomes without negatively charged lipids (i.e. PC-only liposomes)? Does the SYT1C2A-DADA mutant domain bind the PC/PS/PI liposomes? Does the SYT1C2A-PolyB mutant interact with the PC/PS liposomes? The authors wrote both in the main text and the figure 4 legend that they used a lipid monolayer in the BLI method. However, in the Material and Methods section, they wrote that they used small unilamellar vesicles.

      4)The authors beneficially used all-atom MD simulations to address mechanistic details of the SYT1 C2 domains with two different lipid bilayers. However, several issues need to be addressed. Is there a particular reason to include sitosterol in the MD simulations? Does sitosterol contribute to protein binding? Why was not the simulation ran for a longer time? What was a criterion to determine that the system reached a stable state? Results of the MD simulations are presented as static snapshots. The manuscript would benefit from a more detailed analysis, e.g. the number of hydrogen bonds between the protein and phospholipid molecules over time, development of the tilt angle over time, etc. The authors performed the MD simulations for both SYT1 C2 domains. It would be informative to include an electrostatic potential mapped on the surface of the SYT1C2B domain similarly to figure 3. The experimental results showed that SYT1C2B binds liposomes containing PC/PS. How does SYT1C2B interact with the PC/PS membrane? The authors noted that SYT1C2 inserts loop L3 into the lipid bilayer and that this loop adopts a β-hairpin structure. This is the case for the simulation with the SYT1C2AB fragment, but not for the SYT1C2B domain alone. Is the β-hairpin formed during the MD simulation? Or is it a result of the template-based modelling? How is the proposed binding model of the SYT1C2AB affected by the data obtained using SAXS?

      Minor comments:

      1) SYT1CB construct is not listed in figure 1A along with the other constructs used in the study.

      2) Phospholipids contain a phosphate group rather than a phosphoryl group.

      3) In the text, the authors sometimes used the ratio 350 nm/330 nm and sometimes 330 nm/350 nm.

      4) Figure S2, displayed curves look strikingly similar, different line representation does not allow a proper comparison. There are no units for y-axes.

      5) Figure 4B, y-axes do not have scales.

      6) Figure 5B, IP3 or I3P?

      7) The authors noted that based on their SAXS experiments, the SYT1-SMPC2A has a maximum size of 176 Å and they wrote that this value is in accordance with the size of the previously characterized E-Syt2-SMPC2AB protein. However, as the authors reported, E-Syt2-SMPC2AB is two times smaller.

      Significance

      Nature and significance of the advance:

      By a combination of experimental and computational methods, the manuscript provides novel insight into the structure-function relationship of plant SYT1.

      Work in the context of the existing literature:

      The submitted manuscript deals with Arabidopsis SYT1 protein, which belongs to an evolutionarily conserved family of proteins functioning at the ER-PM contact sites. Several recent reports described a principal role of Arabidopsis SYT1 in the regulation of lipid homeostasis, plasmodesmata functions and the response to various stresses (Lee et al. 2019. Ionic stress enhances ER-PM connectivity via phosphoinositide-associated SYT1 contact site expansion in Arabidopsis. PNAS 116, 1420-1429. https://doi.org/10.1073/pnas.1818099116; Ishikawa et al. 2020. Structural and functional relationships between plasmodesmata and plant endoplasmic reticulum-plasma membrane contact sites consisting of three synaptotagmins. New Phytologist 226, 798-808. https://doi.org/10.1111/nph.16391; Ruiz-Lopez et al. 2021. Synaptotagmins at the endoplasmic reticulum-plasma membrane contact sites maintain diacylglycerol homeostasis during abiotic stress. The Plant Cell. https://doi.org/10.1093/plcell/koab122). However, in contrast to mammalian orthologs, mechanistic details of the plant SYT1 function are largely missing.

      Audience:

      The manuscript might be of interest to the community of plant molecular biologists, structural biologists dealing with peripheral membrane proteins and computational biologists.

      Reviewer's expertise:

      Reviewer's field of expertise is plant molecular biology, plant biochemistry, protein-protein, protein-membrane and ion-membrane interactions, molecular dynamics simulations, integrative structural biology, structural bioinformatics.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      This work focuses on Synaptotagmin1 (SYT1), a regulatory element of plant membrane contact site, that act as a tether and physically and functionally connects the endoplasmic reticulum (ER) to the plasma membrane (PM). Like many membrane contact site proteins, SYT1 harbours lipid transfer activity through its SMP domain. Plant SYTs proteins are orthologous to the Extended Synaptotagmin and Tricalbin ER-PM tethers from animal and yeast. SYT1 presents two C2 (C2A and C2B) lipid-binding domains at their C-terminus that are determinant for PM binding, presumably regulating SYT1 function at ER-PM membrane contact sites.

      In this paper, Benavente et al. aimed at investigating the molecular mechanisms of SYT1 binding to the PM and specificity of function of C2A (previous work has shown that SYT1 C2A, but not SYT1 C2B, binds membrane lipids in a Ca2+ dependant manner). The authors combined a wide range of approaches from X-ray crystallography to biophysics and in silico molecular modelling to understand the mechanisms of SYT1 C2A interaction with lipids, at the molecular level. From their study, Benavente et al. shows that C2A display dual lipid binding activity interacting with PS in a Ca2+-dependant manner and with phosphoinositide in a Ca2+-independent manner. These interactions involve two distinct sites; a polybasic amino acid site for phosphoinositides binding and a Ca2+-dependant lipid binding site for PS. They propose that this two-steps binding mechanism confers plasticity in membrane docking under low and high intracellular calcium concentration. They also show that SYT1 full length protein displays three flexible hinges, which they propose confers SYT1 a high degree of conformational freedom

      Minor comments

      • Typos: please re-read carefully through the manuscript to remove them.
      • We advise the authors to have the manuscript corrected by a native english-speaker.

      Significance

      C2- domains are central functional elements of SYT/E-SYT/Tricalbin ER-PM tethers. They regulate PM docking through their lipids binding activity, which can be calcium-dependant or calcium-independent, but also presents lipid specificity (PS, phosphoinositides...). Beside their lipid-binding activity, C2 domains have also been shown to be involved in protein-protein interaction, including intramolecular interaction to inhibit lipid-transfer activity (E-SYT). Multiple C2s are present in SYT/E-SYT/Tricalbin tethers and their diversity of function is likely central for providing functional specificity but also plasticity in response to developmental /environmental clues (together with changes membrane lipid composition and intracellular calcium levels). Decrypting C2 individual operating mode is therefore central. This work is important as it investigates SYT1A docking mechanisms at the molecular level and identifies a number of very interesting features. However, as it stands, the paper does not make the link with SYT1 functionality. The authors discuss their findings within the frame of experimental observations that are already published but these remain speculative.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). The authors obtain a high-resolution structure of SYT1C2A and characterize its calcium and lipid-binding capabilities and those of SYT1C2B. They also study the influence of these interactions on domain topology and the overall architecture of SYT1. They uncover a calcium-dependent lipid binding mechanism that allows them to propose an interesting model in which SYT1C2A binds to the PM to help shuttle DAG to the ER upon abiotic stress.

      Major comments:

      • Are the key conclusions convincing? - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? The conclusion that SYT1C2A determines protein behavior by binding calcium and likely switching its function from a pure tether to a lipid transporter is accurately and convincingly portrayed. However, there are some problematic statements in the way this information is interpreted and used to propose a model.

      • p7: "This means that site I will be occupied depending on the physiological Ca2+ concentration to activate the protein and trigger Ca2+-dependent SYT1C2A lipid binding." Please exchange "means" by "suggests".

      • There is not enough data to determine whether SYT1C2A interacts with the SYT1SMP domain, much less to propose that it is the SYTC2A that loads the SMP with DAG. They may also operate independently by means of the unstructured hinge sequence between them.

      • It is implied that SYTC2A would have to release its calcium atom for it to leave the PM and go to the ER to complete a single transport cycle. It is unclear whether SYT1C2A could bind and release calcium repeatedly and fast enough to cycle between membranes efficiently. It appears more plausible for it to remain at the PM until intracellular calcium concentrations decrease. These statements should be marked as speculative.

      • 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. No new experiments suggested.
      • Are the data and the methods presented in such a way that they can be reproduced? Yes, to the best of my knowledge.
      • Are the experiments adequately replicated and statistical analysis adequate? No. Most of the graphs shown (except Fig. 4A) appear to show only one experiment. Please include results from at least three independent experiments in all graphs, and analyze them statistically. Please describe only differences between samples that are statistically significant (for all graphs and tables).

      Minor comments:

      • Specific experimental issues that are easily addressable. No issues with experiments.
      • Are prior studies referenced appropriately? Generally yes, but there are a few things to change:

      • Please include the PDB codes for the available structures used. It is not clear where the C2C domain structure comes from. We guess it may correspond to PDB entry 2DMG. A word of caution, this 2006 structure has been released but not published and its quality is suboptimal. After its appearance in Idevall-Hagren et al. 2015 supplementary images it has been re-cited in other publications. I would advise against making topology claims based on this structure (Figure S1), and only use it as an example for polybasic patches in C2 domains. • p4: Schauder et al. 2014. Only the "shuttle" model is mentioned in the text. Although this may be the most plausible scenario, the "tunnel" model cannot be completely ruled out. • p5: Fernandez-Busnadiego et al. 2015 does not make a reference to the claim in the text. • p18: Collado et. al 2019 don't assign the peak-forming function to the N-terminal hairpin domain of tricalbins.

      • Are the text and figures clear and accurate? Generally yes, although the figures and legends should contain more information. For example:
      • Fig. 1A: the notation for the domains C2C-C2E of E-Syt1 is confusing and not described in the legend. Also, please mark which beta sheet is which more clearly in Fig 1B, it is very difficult to understand the labelling.
      • Fig. 1C: several elements (e.g. types of boxes, "T" on the top) used in the figure are not described in the legend. Both beta sheets and mutations are described as "arrows" in the legend, please differentiate between vertical and horizontal.
      • Fig. 2A: indicate Lys 286, which is mentioned in the text.
      • Fig. 2B: Inset too small to read.
      • Fig. 3: please show the scale and explain the blue-red color code. Please mark the polybasic patch.
      • Fig. 4A should show both mutants for both types of liposomes.
      • Table 1: what is "control"?
      • Fig. S5: Difficult to read, increase font size. The text talks about experiments with calcium and EGTA that are not shown in the figure.

      Text:

      • p5, first Results paragraph: reference the loop numbers marked in the figure to facilitate the description. The description of the loop between b6 and b7 is not clear: the text implies it is not present in SYT1C2A, although it should say it is unstructured.
      • Please indicate more prominently throughout the text and methods the plant species from which the C2 domains were studied, rather than just saying "plant". For example, are all the other SYTs shown in Fig 1C from Arabidopsis as well? To generalize the findings reported here as "plants", the conservation in other species need to be shown. Alternatively, please rephrase the relevant statements.
      • "Data" is used both in singular and plural.
      • There are several typos (it would be much easier to point to them having line numbers in the manuscript!), e.g. p6 title says "C2+" and the next paragraph "SYTCA2", Fig. 4D is mentioned in the text but it doesn't appear in the figures, p16: "DAG form the PM" -> from, p17 "overexpression of Syt1" -> E-Syt1.
      • Do you have suggestions that would help the authors improve the presentation of their data and conclusions? References to abiotic stress leading to DAG accumulation can be found throughout the paper, but little is explained about how this correlates with the increase in calcium concentration suggested for SYT1 activation. It would be an improvement to explain how this happens.

      Significance

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. Multi-C2 domain proteins perform multiple different functions through mechanisms not yet fully understood. One important reason for this lack of knowledge is the presence of multiple C2 domains and their similarity to each another. The key to understanding how a C2 domain operates can hinge on a single calcium-binding site, which often escape accurate prediction by homology modelling and other computational methods. This is also the case for lipid binding affinity and stereophysical properties of any protein. This work highlights the importance of obtaining reliable, high-resolution structures of these domains as well as characterizing their binding partners and the dynamics of their interactions. Through the identification of the properties of each C2 domain and adjacent sequences the authors provide a coherent mechanistic model for SYTC2AB and a reproducible workflow for the study of other C2 domains.
        • Place the work in the context of the existing literature (provide references, where appropriate). This work is an important step in the study of plant synaptotagmins, in which previously reported SYT1 biochemical properties (Schapire 2018, Perez-Sancho 2015, among others) and the study of its orthologs in other organisms are brought together with novel structural and biophysical data to create a compelling mechanistic model. It also has important implications for the mechanisms of functions of this family of proteins in other organisms.
        • State what audience might be interested in and influenced by the reported findings. Researchers in the broad membrane traffic community will enjoy this work, and particularly those working with orthologs of SYT1 might find insights that apply to their research. In a broader sense, this work will be appealing to researchers working in plant cellular stress response.
        • Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Structure of ER-PM contact sites. I have no expertise in titration calorimetry or SAXS and cannot properly evaluate those results.
  3. Jun 2021
    1. Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      Responses to reviewers’ comments

      We thank the reviewers for their encouraging comments and helpful suggestions.

      Reviewer #1

      (Evidence, reproducibility and clarity (Required)):

      Sanchez et al report several new findings about the adhesive protrusions on Plasmodium falciparum infected erythrocytes. Using super resolution microscopy and correlation analysis, they tracked associations between the knob protein KAHRP and erythrocyte membrane cytoskeleton proteins. They have expanded on and improved previous work on the unusual spiral structure of the knobs, which appears to be a spiral ribbon or blade and have shown a developmental pathway for the association of KAHRP with the cytoskeleton. They have localised KAHRP close to the spiral and determined its abundance in the knobs. They have also used cryo electron tomography and subtomogram averaging to get an improved 3D view of the knob structure.

      The work appears to be carefully and thoroughly done, and the paper is clearly written, though non specialists in the optical methods may find it challenging to navigate through the many super resolution images and correlation plots.

      Comment 1: The writing needs minor editing to fix a variety of small linguistic errors and typos. For example, line 97 "sideway positions" (they presumably mean lateral location), line 980 typo overlay, line 366 "then could reorganizes", line 435, "a predict volume".

      We apologize for the linguistic errors and typos. These have been corrected in the revised manuscript.

      (Significance (Required)):

      Comment 2: The study provides a distinct advance on the previous state of knowledge of the structure and biochemistry of the knobs. The knobs play a key role in virulence of P. falciparum and they are quite poorly understood. Although this paper does not represent a major breakthrough in determining the molecular structure or mechanistic role of the knobs, e.g. the biochemical identity of the spiral remains unknown, the new information is valuable and likely to be important in understanding the pathogenic actions of P falciparum.

      We thank the reviewer for appreciating the importance of our study. We believe that our first-time observations on the dynamics of KAHRP are a very important advance in the field and that revealing the mechanistic basis is a great challenge that at the current stage has to be left to future work.

      Comment 3: The interpretation shown in Figure 7 seems fine, except for the proposal that the actin cytoskeleton is reorganised. There is no evidence for that. The cryo tomograms of the cytoskeleton in Watermeyer et al addressed this point and did not find any evidence for reorganisation of the cytoskeleton other than the insertion of the knobs.

      In two previous studies we could show that actin is indeed reorganized by the parasite. It is mined from the protofilaments to generate long actin filaments that connect the knobs with the Maurer’s clefts and which are used for trafficking of cargo vesicles from the Maurer’s clefts to the erythrocyte plasma membrane (Cyklaff et al. Hemoglobins S and C interfere with actin remodeling in Plasmodium falciparum-infected erythrocytes. Science. 2011 334:1283-1286; Cyrklaff et al. Oxidative insult can induce malaria-protective trait of sickle and fetal erythrocytes. Nat Commun. 2016 7:13401). Moreover, a life-cycle resolved AFM-study of the cytoplasmic side of iRBCs by the group of CT Lim has demonstrated dramatic coarsening of the spectrin network, which must be accompanied by changes to the actin component of the skeleton (Shi, Hui, et al. "Life cycle-dependent cytoskeletal modifications in Plasmodium falciparum infected erythrocytes." PLoS One 8.4 (2013): e61170). Coarsening of the actin-spectrin network would imply a decrease of the amount of actin in the network, which is consistent with its use in the parasite-derived long actin filaments.

      \*Referee Cross-commenting***

      I also agree with the other 2.

      Reviewer #2

      (Evidence, reproducibility and clarity (Required)):

      Malaria parasites replicate within circulating red blood cells (RBC). During parasite maturation, the parasite coordinates extensive modification of the host cell, including structural modifications of the RBC cytoskeleton and surface membrane. These host cell alterations play crucial roles in the pathology of malaria, including vascular adhesion by parasitised cells and avoidance of splenic clearance, and so are of great interest. This interesting manuscript describes a detailed examination of the role in these RBC modifications of a well-described parasite protein called KAHRP. Using a combination of cutting-edge super-resolution microscopy, cryo-electron tomography, immuno-EM, SEM and parasite mutagenesis, the authors provide evidence that KHARP localisation alters during parasite maturation but eventually becomes closely associated with the previously-described spiral structures that underlie infected RBC surface membrane protrusions called knobs. The authors provide improved resolution of the spiral formations, generate a quantitative estimate of the number of KAHRP molecules per knob, and provide a model for the role of KAHRP in attaching other proteins to the spirals based on their observations.

      In general, this study is thorough and well-performed, and the conclusions drawn are well-supported by the data. Although the work does not advance understanding of knob function or the parasite components that form the bulk of the spirals, it provides an interesting and useful contribution to understanding of the manner in which this important pathogen manipulates its host cell.

      We thank the reviewer for appreciating the importance of our study and in acknowledging that it is an important intermediate step towards a complete understanding of skeleton remodelling by the parasite.

      I have just a few minor suggestions that should improve the manuscript.

      Comment 1: Line 91 (Page 2 paragraph 2). It would be greatly helpful here if the authors could provide a more detailed background on the makeup of the RBC cytoskeleton, and in particular the interactions between beta-spectrin and the actin protofilaments of the junctional complexes. The authors should make it clear that the actin-binding domain of beta-spectrin comprises 2 calponin like domains, and that these are attached to the end of the tandem spectrin repeat domains that make up the bulk of the molecule.

      We thank the reviewer for this helpful suggestion and have added a new paragraph to the results section providing detailed background information on the makeup of the RBC membrane skeleton. The new text reads as follows:

      “Major components of the red blood cell membrane skeleton are spectrin and actin filaments (Fig. 1B). The spectrin filaments consist of α- and ß-spectrin, which form α2ß2 heterotetramers by head-to-head association of two αß dimers (Lux, 2016; Machnicka et al., 2014). The N-termini of the ß-spectrin subunits are positioned at the tail ends of the heterotetramer and contain two calponin homology (CH) domains for binding to actin protofilaments consisting of 6 to 8 actin monomers in each of the two strands (Lux, 2016; Machnicka et al., 2014). Protein 4.1R strengthens the spectrin actin interaction (Lux, 2016; Machnicka et al., 2014). Groups of up to six spectrin heterotetramers can attach to an actin protofilaments, resulting in a pseudohexagonal meshwork (Lux, 2016). Ankyrin binds to the C-terminal domain of ß-spectrin and connects integral membrane proteins with the actin spectrin network in an ankyrin complex (Lux, 2016; Machnicka et al., 2014).”

      Comment 2: Line 97 "These values are slightly larger than the reported physical dimension of the protofilament...". Please provide these reported dimensions here, as well as relevant references.

      The requested information is now provided. The sentence now reads as follows:

      “These values are slightly larger than the reported physical dimension of the protofilaments of ~37 nm (Lux, 2016) and might be explained by the lateral localization of the spectrin binding sites and the additional sizes of the primary and secondary antibody trees used to detect the two targets.”

      Comment 3: Line 366 "reorganize"

      The spelling mistake has been corrected.

      (Significance (Required)):

      Comment 4: This is a useful technical advance in understanding of the structure of the P. falciparum-infected red blood cell, and builds on the work of Watermeyer et al. (2016). The study should certainly be of interest to most malaria researchers, particularly those interested in the pathobiology of the organism.

      We thank the reviewer for supporting our study.

      \*Referee Cross-commenting***

      I fully agree with and endorse the comments of the other 2 reviewers.

      Reviewer #3

      (Evidence, reproducibility and clarity (Required)):

      The binding of P. falciparum infected erythrocyte (iRBCs) to the endothelium is mediated by protuberances (knobs). These knobs are assembled by a multi-protein complex at the iRBC surface. It acts as a scaffold for the presentation of the major virulence antigen, P. falciparum Erythrocyte Membrane Protein-1 (PfEMP1). The knob-associated histidine-rich protein (KAHRP) is an essential component of the knobs and therefore essential for the binding of iRBC to the endothelium under physiological conditions. This manuscript focusses on the knob architecture and KAHRP localization.

      Comment 1: It is, at least for this reviewer - hard to assess how the "preparation of exposed membranes by hypotonic shock" and the analysis of the "inverted erythrocyte membrane ghosts" is i) reflective of the physiological architecture within the iRBC and ii) how the authors exclude remnants from Maurers clefts (MCs) in their preparation. The latter appears especially important for the interpretation of dynamic KAHRP repositioning, as MCs are mobile in early stages and non-mobile later on (e.g. McMilian et al. 2013, Grüring et al. 2011) and the authors observed at least some MAHRP1 signal (Figure S8), which is hard to interpret by the single representative image provided.

      We understand the reviewer’s concerns, but are convinced that we have done the necessary controls to evaluate our approaches. For example, we evaluated the exposed membrane approach by investigating uninfected erythrocytes and comparing the findings with literature reports (see Figure 1). A high degree of agreement was observed. We further would like to point out that the exposed membrane approach has been successfully used by several other studies referenced in the manuscript (Dearnley et al., 2016; Looker et al., 2019; Shi et al., 2013). Please also allow us to explain why we have used exposed membranes instead of whole cells. The reason is that the hemozoin produced by the parasite interferes with STED microscopy, resulting in a quick and strong build-up of resonance energy in the specimen and, eventually, in the disruption of the cell.

      With regard to the question of whether remnants of Maurer’s clefts are present in our preparations, we do not think so, at least we never observed membrane profiles reminiscent of Maurer’s clefts in SEM images of exposed membranes (see figure at the end of the response letter). Irrespectively, we will double check this result using STED imaging of exposed membranes treated with an antibody against the established Maurer’s clefts marker SBP1. These data could be added to a revised manuscript.

      Comment 2: line 173: Please provide a detailed description about parasite synchronization (also absent in the methods section).

      A detailed description including references are now added to the methods section:

      “For synchronization of cultures, schizont-infected erythrocytes were sterile purified using a strong magnet (VarioMACS, Miltenyi Biotec) (Staalsoe et al., 1999) and mixed with fresh erythrocytes to high parasitaemia. 5000 heparin units (Heparin-sodium 25000, Ratiopharm) were added and the cells were returned to culture for 4 hrs (Boyle et al., 2010). Following the treatment with heparin, cells were washed with pre-warmed supplemented RMPI 1640 medium and then returned to culture for 2 hrs to allow for re-invasion of erythrocytes. Subsequently, cells were treated with 5% sorbitol to remove late parasite stages (Lambros and Vanderberg, 1979).”

      Comment 3: line 136: Please re-check nomenclature of "PHIS1605w" (mixed nomenclature used throughout the manuscript). I suggest to use either LyMP or the up-to-date ID PF3D7_0532400.

      We apologize for the oversight and now consistently use the ID PF3D7_0532400.

      Comment 4: Please provide source and references for PfEMP1, MAHRP1 and "PHIS1605w" antibodies that are used. I cannot find them in the methods section or in Table S1.

      We apologize for the oversight and now provide the requested information in the amended Table S1.

      Comment 5: line 165: Warncke et al. (2016) appears to be misplaced as an appropriate MAHRP1 reference.

      We now cite the original MAHRP1 publication by Spycher et al. 2003.

      Comment 6: line 159: the sentence "The strong cross-correlation between KAHRP and actin is consistent with previous cryo-electron tomographic analysis showing long actin filaments connecting the knobs with Maurer's clefts in trophozoites (Cyrklaff et al., 2012; Cyrklaff et al., 2011; Cyrklaff et al., 2016)" could be moved to the discussion section.

      The sentence was indeed redundant with a section in the discussion and was removed.

      Comment 7: line 199: The text refers to Fig. 9AB - but should refer to 4AB or suppl. 11.

      We are sorry for this mistake and now refer to the correct figures in the revised manuscript.

      Comment 8: Fig. 4: A solid average for the number of subtomograms, but please provide information about what the arrowheads (4E) indicate.

      Thank you for this comment. The arrowheads indicate peripheral crown-like densities. We have updated the figure legend to clarify this issue.

      Comment 9: The "flexible periphery" is likely a combination of flexibility and occupancy as the average was made from subtomograms with varying number of turns in the spiral. As occupancy is likely a significant contributing factor to the average that should be discussed or at least mentioned.

      Thank you for this important comment. Indeed, a significant variation was observed between the individual knobs. The spirals have variable diameter, and the number of peripheral proteins also varied. We added measurements to the supplementary figure 11D. In addition, we update the text and extended the discussion.

      Comment 10. On that note, did the authors try and classify based on number of turns prior to averaging and if so did the authors see any differences in structures between few turn and many turn spirals?

      We attempted several classifications on the full knobs with variable masks. However due to a limited number of particles in the dataset we could not converge to stable solutions. Instead, we decided to adopt the subboxing strategy where locally ordered segments at the periphery could be analyzed. This showed several structural snapshot at the periphery of the knobs.

      Comment 11. What size mask was used? Was it a soft sphere around the core or big enough for the knobs with multiple spiral turns?

      While we attempted several alignments and classifications with variable masks, the final refinement and measurement of FSC was performed with a soft contour mask mask. We overlaid it with the structure in Figure S11F and uploaded it as a part of the EMDB deposition. We further show the masks used in this study in a new Figure S14.

      Comment 12. It might be useful for readers who are not familiar with Dynamo to provide a little bit more information about how the initial reference was produced. Additionally more information about the sub-boxing strategy ie: spacing etc. would helpful.

      Thank you very much for the suggestion. For the initial reference we manually aligned all the particles, summed them up and low-pass filtered them. We now describe it in the methods section.

      For the subboxing procedure we added more description to the main text:

      “40 segments were extracted at the radius of the 2nd and 3rd spirals followed by their classification into structural classes.”

      We further extended and simplified the description in the results section (line ~221).

      Comment 13: Fig. 5 Additional (earlier) maturation stages of the iRBC with Ni2+NAT-gold-labelling would be a nice add on - this could help confirm the model and would itself be a control for the later stage labelling.

      We thank the reviewer for this insightful suggestion. We are currently performing the proposed experiment and will include it in a revised version of the manuscript.

      Comment 14: line 637: DMSI typo and please provide the supplier for DMSI (DSM1).

      We corrected the typographic error and now provide the name of the supplier.

      Comment 15: Figure 7: Please provide what the purple arrows indicate.

      The figure legend has been updated.

      Comment 16: Fig S11D: The labels X, Y and Z are confusing, describing the slicing axis as "XZ, YZ and XY" view is more intuitive.

      Done as suggested by the reviewer.

      Comment 17: Figure S13 B: WBs are cropped. Please provide un-cropped WB.

      Uncropped Western blots will be provided in the revised manuscript.

      (Significance (Required)):

      In general, I highly appreciated the solid data and its thorough analysis of the microscopy data. The authors investigate the structural organization of knobs in iRBCs using high-resolution imaging techniques including STED and PALM super-resolution microscopy-based approaches and electron tomography. The beauty of this paper is that it does nicely re-investigate knob architecture in iRBC (e.g. Watermeyer et al., 2016, Cutts et al., 2017, Looker et al., 2019, McHugh et al., 2020) and provides some intriguing KHARP co-localization with cytoskeleton components. The downside of it is that - by nature - it is descriptive (and the data rather confirmative) and as it stands does not provide us with a deeper molecular dissection of the knob associated structure and its cellular function.

      We thank the reviewer for appreciating our study and would like to emphasize the following novelties in our study:

      • We show that the association of KAHRP with membrane skeletal components is highly dynamic and changes as the parasite matures. Our results on the dynamics of KAHRP organization reconciles conflicting reports in the literature, and establish for the first time a dynamical model for KAHRP organization.
      • We further show that KAHRP finally assembles at remnant actin-junctional complexes devoid of the actin-capping factors adducin and tropomodulin.
      • We further quantified the number of KAHRP molecules per knob and show that KAHRP is present as 60 copies per knob, a number one order of magnitude greater than previously thought.
      • Last but not least, we provide a 35 Å map of the spiral scaffold underlaying knobs and show that KAHRP associates with the spiral scaffold.
      • We conclude by providing a novel model on the biological function of KAHRP by proposing that KAHRP acts as a glue that connects spectrin and parasite-remodeled actin filaments with the knob spiral.

        \*Referee Cross-commenting***

      Fully agreed.

      Boyle, M.J., Wilson, D.W., Richards, J.S., Riglar, D.T., Tetteh, K.K., Conway, D.J., Ralph, S.A., Baum, J., and Beeson, J.G. (2010). Isolation of viable Plasmodium falciparum merozoites to define erythrocyte invasion events and advance vaccine and drug development. Proc Natl Acad Sci U S A 107, 14378-14383.

      Lambros, C., and Vanderberg, J.P. (1979). Synchronization of Plasmodium falciparum erythrocytic stages in culture. J Parasitol 65, 418-420.

      Lux, S.E.t. (2016). Anatomy of the red cell membrane skeleton: unanswered questions. Blood 127, 187-199.

      Staalsoe, T., Giha, H.A., Dodoo, D., Theander, T.G., and Hviid, L. (1999). Detection of antibodies to variant antigens on Plasmodium falciparum-infected erythrocytes by flow cytometry. Cytometry 35, 329-336.

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      Summary:

      The binding of P. falciparum infected erythrocyte (iRBCs) to the endothelium is mediated by protuberances (knobs). These knobs are assembled by a multi-protein complex at the iRBC surface. It acts as a scaffold for the presentation of the major virulence antigen, P. falciparum Erythrocyte Membrane Protein-1 (PfEMP1). The knob-associated histidine-rich protein (KAHRP) is an essential component of the knobs and therefore essential for the binding of iRBC to the endothelium under physiological conditions. This manuscript focusses on the knob architecture and KAHRP localization.

      General point:

      It is, at least for this reviewer - hard to assess how the "preparation of exposed membranes by hypotonic shock" and the analysis of the "inverted erythrocyte membrane ghosts" is i) reflective of the physiological architecture within the iRBC and ii) how the authors exclude remnants from Maurers clefts (MCs) in their preparation. The latter appears especially important for the interpretation of dynamic KAHRP repositioning, as MCs are mobile in early stages and non-mobile later on (e.g. McMilian et al. 2013, Grüring et al. 2011) and the authors observed at least some MAHRP1 signal (Figure S8), which is hard to interpret by the single representative image provided.

      Specific points:

      • line 173: Please provide a detailed description about parasite synchronization (also absent in the methods section)
      • line 136: Please re-check nomenclature of "PHIS1605w" (mixed nomenclature used throughout the manuscript). I suggest to use either LyMP or the up-to-date ID PF3D7_0532400.
      • Please provide source and references for PfEMP1, MAHRP1 and "PHIS1605w" antibodies that are used. I cannot find them in the methods section or in Table S1 -line 165: Warncke et al. (2016) appears to be misplaced as an appropriate MAHRP1 reference.
      • line 159: the sentence "The strong cross-correlation between KAHRP and actin is consistent with previous cryo-electron tomographic analysis showing long actin filaments connecting the knobs with Maurer's clefts in trophozoites (Cyrklaff et al., 2012; Cyrklaff et al., 2011; Cyrklaff et al., 2016)" could be moved to the discussion section
      • line 199: The text refers to Fig. 9AB - but should refer to 4AB or suppl. 11.
      • Fig. 4: A solid average for the number of subtomograms, but please provide information about what the arrowheads (4E) indicate. A few additional comments on this section:

      1: The "flexible periphery" is likely a combination of flexibility and occupancy as the average was made from subtomograms with varying number of turns in the spiral. As occupancy is likely a significant contributing factor to the average that should be discussed or at least mentioned.

      1. On that note, did the authors try and classify based on number of turns prior to averaging and if so did the authors see any differences in structures between few turn and many turn spirals?
      2. What size mask was used? Was it a soft sphere around the core or big enough for the knobs with multiple spiral turns?
      3. It might be useful for readers who are not familiar with Dynamo to provide a little bit more information about how the initial reference was produced. Additionally more information about the sub-boxing strategy ie: spacing etc. would helpful.
      • Fig. 5 Additional (earlier) maturation stages of the iRBC with Ni2+NAT-gold-labelling would be a nice add on - this could help confirm the model and would itself be a control for the later stage labelling.
      • line 637: DMSI typo and please provide the supplier for DMSI (DSM1).
      • Figure 7: Please provide what the purple arrows indicate.
      • Fig S11D: The labels X, Y and Z are confusing, describing the slicing axis as "XZ, YZ and XY" view is more intuitive.
      • Figure S13 B: WBs are cropped. Please provide un-cropped WB.

      Significance

      In general, I highly appreciated the solid data and its thorough analysis of the microscopy data. The authors investigate the structural organization of knobs in iRBCs using high-resolution imaging techniques including STED and PALM super-resolution microscopy-based approaches and electron tomography. The beauty of this paper is that it does nicely re-investigate knob architecture in iRBC (e.g. Watermeyer et al., 2016, Cutts et al., 2017, Looker et al., 2019, McHugh et al., 2020) and provides some intriguing KHARP co-localization with cytoskeleton components. The downside of it is that - by nature - it is descriptive (and the data rather confirmative) and as it stands does not provide us with a deeper molecular dissection of the knob associated structure and its cellular function.

      Referee Cross-commenting

      Fully agreed.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Malaria parasites replicate within circulating red blood cells (RBC). During parasite maturation, the parasite coordinates extensive modification of the host cell, including structural modifications of the RBC cytoskeleton and surface membrane. These host cell alterations play crucial roles in the pathology of malaria, including vascular adhesion by parasitised cells and avoidance of splenic clearance, and so are of great interest. This interesting manuscript describes a detailed examination of the role in these RBC modifications of a well-described parasite protein called KAHRP. Using a combination of cutting-edge super-resolution microscopy, cryo-electron tomography, immuno-EM, SEM and parasite mutagenesis, the authors provide evidence that KHARP localisation alters during parasite maturation but eventually becomes closely associated with the previously-described spiral structures that underlie infected RBC surface membrane protrusions called knobs. The authors provide improved resolution of the spiral formations, generate a quantitative estimate of the number of KAHRP molecules per knob, and provide a model for the role of KAHRP in attaching other proteins to the spirals based on their observations. In general, this study is thorough and well-performed, and the conclusions drawn are well-supported by the data. Although the work does not advance understanding of knob function or the parasite components that form the bulk of the spirals, it provides an interesting and useful contribution to understanding of the manner in which this important pathogen manipulates its host cell.

      I have just a few minor suggestions that should improve the manuscript.

      Line 91 (Page 2 paragraph 2). It would be greatly helpful here if the authors could provide a more detailed background on the makeup of the RBC cytoskeleton, and in particular the interactions between beta-spectrin and the actin protofilaments of the junctional complexes. The authors should make it clear that the actin-binding domain of beta-spectrin comprises 2 calponin like domains, and that these are attached to the end of the tandem spectrin repeat domains that make up the bulk of the molecule.

      Line 97 "These values are slightly larger than the reported physical dimension of the protofilament...". Please provide these reported dimensions here, as well as relevant references.

      Line 366 "reorganize"

      Significance

      This is a useful technical advance in understanding of the structure of the P. falciparum-infected red blood cell, and builds on the work of Watermeyer et al. (2016). The study should certainly be of interest to most malaria researchers, particularly those interested in the pathobiology of the organism.

      Referee Cross-commenting

      I fully agree with and endorse the comments of the other 2 reviewers.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Sanchez et al report several new findings about the adhesive protrusions on Plasmodium falciparum infected erythrocytes. Using super resolution microscopy and correlation analysis, they tracked associations between the knob protein KAHRP and erythrocyte membrane cytoskeleton proteins. They have expanded on and improved previous work on the unusual spiral structure of the knobs, which appears to be a spiral ribbon or blade) and have shown a developmental pathway for the association of KAHRP with the cytoskeleton. They have localised KAHRP close to the spiral and determined its abundance in the knobs.

      They have also used cryo electron tomography and subtomogram averaging to get an improved 3D view of the knob structure.

      The work appears to be carefully and thoroughly done, and the paper is clearly written, though non specialists in the optical methods may find it challenging to navigate through the many super resolution images and correlation plots. The writing needs minor editing to fix a variety of small linguistic errors and typos. For example, line 97 "sideway positions" (they presumably mean lateral location), line 980 typo overlay, line 366 "then could reorganizes", line 435, "a predict volume".

      Significance

      The study provides a distinct advance on the previous state of knowledge of the structure and biochemistry of the knobs. The knobs play a key role in virulence of P. falciparum and they are quite poorly understood. Although this paper does not represent a major breakthrough in determining the molecular structure or mechanistic role of the knobs, e.g. the biochemical identity of the spiral remains unknown, the new information is valuable and likely to be important in understanding the pathogenic actions of P falciparum.

      The interpretation shown in Figure 7 seems fine, except for the proposal that the actin cytoskeleton is reorganised. There is no evidence for that. The cryo tomograms of the cytoskeleton in Watermeyer et al addressed this point and did not find any evidence for reorganisation of the cytoskeleton other than the insertion of the knobs.

      I am generally familiar with the area of this work, but not expert in the details of the optical methods and localisation analysis.

      Referee Cross-commenting

      I also agree with the other 2.

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

      Learn more at Review Commons


      Reply to the reviewers

      Dear Editor,

      We appreciate the critical and constructive comments that Reviewers made on our paper entitled “The structure and flexibility analysis of the Arabidopsis Synaptotagmin 1 reveal the basis of its regulation at membrane contact sites” by Benavente et al.. We are very pleased that three reviewers from Review commons appreciated the work, found it interesting and suggested constructive modifications to the paper. We have now responded to their suggestions and enclose a modified manuscript.

      As suggested by the reviewers we now provide additional biochemical evidence about the role of calcium on C2 activation, more controls in the biochemical assays, an improved protocol for the lipid binding assays as well as details in the molecular dynamics. In addition, following reviewers’ recommendations, we have rephrased some of the headings and statements in the discussion section, to adjust them better to our experimental results. Altogether, we believe that the manuscript is more solid and interesting than the previous version for the general readers of eLife.

      Specifically, we have made the following changes in response to the particular comments of the referees:

      Reviewer 1

      Major points

      The conclusion that SYT1C2A determines protein behavior by binding calcium and likely switching its function from a pure tether to a lipid transporter is accurately and convincingly portrayed. However, there are some problematic statements in the way this information is interpreted and used to propose a model.

        • p7: "This means that site I will be occupied depending on the physiological Ca2+ concentration to activate the protein and trigger Ca2+-dependent SYT1C2A lipid binding." Please exchange "means" by "suggests"* We agree with the reviewer and we have exchanged “means” by “suggests”
        • There is not enough data to determine whether SYT1C2A interacts with the SYT1SMP domain, much less to propose that it is the SYTC2A that loads the SMP with DAG. They may also operate independently by means of the unstructured hinge sequence between them. *
        • It is implied that SYTC2A would have to release its calcium atom for it to leave the PM and go to the ER to complete a single transport cycle. It is unclear whether SYT1C2A could bind and release calcium repeatedly and fast enough to cycle between membranes efficiently. It appears more plausible for it to remain at the PM until intracellular calcium concentrations decrease. **

        These statements should be marked as speculative. *

      We understand that the reviewer comments are related to steps 3 and 4 of our model for the function of SYT1 (figure7; pages 17 and 18). We have now included these scenarios suggested by the reviewer (page 17 and 18). In addition, we have discussed the work from Vennekate et al, PNAS, 2012, related to the cis and trans membrane interactions of the two C2 domains of human synaptotagmin (Syt1). The authors demonstrated that they are driven by the balance in the concentration of anionic lipids between target membranes and by Ca2+ and protein concentration. Likewise, the changes in the local composition of PIP and/or Ca2+ may drive the transfer of SYT1C2A from the PM to de ER and vice versa while the C2B remains at the PM.

      Most of the graphs shown (except Fig. 4A) appear to show only one experiment. Please include results from at least three independent experiments in all graphs, and analyze them statistically. Please describe only differences between samples that are statistically significant (for all graphs and tables).

      Following reviewed recommendation, we have assured that all the biochemical data arises from three independent measurements. Indeed, as indicated in the figure legends and methods section, this was the case for the Ca2+ binding experiments at figure 2C and for the lipid binding data at figure 4. We have now repeated the ITC experiments to provide three independent measurements of the dissociation constant of SYT1C2A for Ca2+.

      Minor comments

      Please include the PDB codes for the available structures used. It is not clear where the C2C domain structure comes from. We guess it may correspond to PDB entry 2DMG. A word of caution, this 2006 structure has been released but not published and its quality is suboptimal. After its appearance in Idevall-Hagren et al. 2015 supplementary images it has been re-cited in other publications. I would advise against making topology claims based on this structure (Figure S1), and only use it as an example for polybasic patches in C2 domains.

      We have included the PDB codes for the available structures used at the corresponding figure captions 1, 3 and S1.

      We agree that the stereochemistry of the refined structure of the E-Syt2 C2C domains (2DMG) is suboptimal. However, we have sought for the advice of an NMR expert (M.A. Jimenez, https://scholar.google.es/citations?hl=es&user=iAF0ymYAAAAJ) to evaluate the NMR restrains available alongside the entry at PDB. She concluded that the Halpha-Halpha distance restrains indicate that the connectivity of the beta strands is correct. The confusion may arise from the fact that the PDB entry has been updated twice since 2006, the last one in 2011.

        • p4: Schauder et al. 2014. Only the "shuttle" model is mentioned in the text. Although this may be the most plausible scenario, the "tunnel" model cannot be completely ruled out*. We have now stated that both shuttle or tunnel modes are possible for lipid transfer at page 4.
        • p5: Fernandez-Busnadiego et al. 2015 does not make a reference to the claim in the text.* Fernandez-Busnadiego et al. 2015 has been replaced by Collado et. al 2019
        • p18: Collado et. al 2019 don't assign the peak-forming function to the N-terminal hairpin domain of tricalbins. As the reviewer indicates, Collado et. al 2019 just suggests that the peak forming function could be due to the N-terminal hairpin domain of tricalbins: “This phenomenon may rely on the hairpin sequence that anchors Tcbs to the ER membrane. Tcb hairpin sequences could sense and/or generate membrane curvature as in reticulons and other ER morphogenetic proteins (Hu et al., 2011).” (Taken from Collado et al. 2019).*

      Accordingly, we have indicated that “It has been suggested that the insertion of the N-terminal hydrophobic end of Tcbs may induce the formation of peaks of strong curvature at the ER region facing the PM. These structures shorten the distance between the ER and the PM by ~7 nm and facilitate lipid transport (Collado et al., 2019)”

      Fig. 1A: the notation for the domains C2C-C2E of E-Syt1 is confusing and not described in the legend. Also, please mark which beta sheet is which more clearly in Fig 1B, it is very difficult to understand the labelling** Fig. 1C: several elements (e.g. types of boxes, "T" on the top) used in the figure are not described in the legend. Both beta sheets and mutations are described as "arrows" in the legend, please differentiate between vertical and horizontal.

      We have modified the Fig. 1A legend to explain the notation of C2 domains of human E-Syts “Human E-Syt1 displays five C2 domains while E-Syt2 and E-Syt3 display three”.

      Labeling of beta strands has been modified to make the panel clearer.

      We have rewritten the caption of figure 1C to describe the types of boxes and the “T” and to distinguish between vertical and horizontal arrows.

      Fig. 2A: indicate Lys 286, which is mentioned in the text

      The reference to Lys 286 in the text was not correct. We have modified the text to indicate that Lys275 replaces the Ca I as it is shown in figure 2.

      Fig. 2B: Inset too small to read.

      Labels for inset at Fig 2B have been enlarged.

      Fig. 3: please show the scale and explain the blue-red color code. Please mark the polybasic patch.

      We have explained the meaning of the blue and red code, marked the polybasic patch of SYT1C2A and E-Syt1C2C and indicated that all of them are scaled to the same value.

      Fig. 4A should show both mutants for both types of liposomes.

      We have included new liposome binding data including both mutants for both types of liposomes.

      Table 1: what is "control"?

      We have indicated that the unbound wild-type protein is taken as a control

      Fig. S5: Difficult to read, increase font size. The text talks about experiments with calcium and EGTA that are not shown in the figure.

      We have increased figure size to facilitate reading. We have also indicated that C2AB-Ca and C2A stand for the experiments carried out in presence of Ca and EGTA, respectively.

      Reviewer 2

      Minor comments

      Typos: please re-read carefully through the manuscript to remove them. We advise the authors to have the manuscript corrected by a native english-speaker.

      We have done a thorough revision of the manuscript to correct typos and to improve the English style of the manuscript

      Reviewer 2 considers that “The authors discuss their findings within the frame of experimental observations that are already published but these remain speculative”

      We are glad that the reviewer appreciates our work in “decrypting the roles of C2 individual operating modes” as it is a “central issue for providing functional specificity but also plasticity in response to developmental /environmental clues”. The reviewer also considers our work “important and identifies a number of very interesting features”. As already mentioned to the editor, the present version of the manuscript includes new biochemical data and analysis to support further the functions of C2A-C2B tandem. In addition, we have included new references and rephrased some of the headings and statements in the discussion section, to adjust them better to our experimental results.

      Reviewer 3

      Major points

      1)The authors investigated SYT1C2A calcium-binding sites using two different methods, ITC and differential scanning fluorimetry. By using ITC, they described the first binding site coordinating calcium ions in the nanomolar range. The second calcium-binding site was then characterized by differential scanning fluorimetry. The second calcium-binding site binds calcium with Kd of 277 µM. The authors then mutated SYT1C2A at two positions and performed again differential scanning fluorimetry. In this case, they did not observe any blue shift in intrinsic fluorescence concluding that "calcium-binding is mediated by the calcium-binding site". It is not clear which binding site the authors mean. In the structure of SYT1C2A, the mutated residues (D276 and D282) are shared by both calcium-binding sites. It is, therefore, difficult to interpret the data. The authors should generate a unique mutation for each site and perform both ITC and differential scanning fluorimetry

      The SYT1C2A-DADA double mutant was prepared to abolish Ca2+ binding to both site I and site II and to discard an unspecific effect of Ca2+ on intrinsic fluorescence that accounts for the low affinity Kd. Our data showed that the addition of Ca2+ to SYT1C2A-DADA does not produce a shift in intrinsic fluorescence of the protein; indicating that the observed Ca2+-binding activity is specifically mediated by the Ca2+ dependent lipid binding site. We clarify this point in the present version of the manuscript.

      In addition, following reviewer’s recommendation, we have prepared two additional point mutant proteins, SYT1C2A D276A and SYT1C2A E340A, to investigate the Ca2+ binding properties at site II and I, respectively. As expected, the reduction of one carboxylate ligand at the structural site II produces a drastic decrease in the Ca2+ binding affinity (Kd = 1.8± 0.5 mM) which is coupled with a decrease in thermal stability of the protein (Ti = 55°C) and a red-shift in fluorescence emission with respect to the wild type protein that resembles those effects observed for the SYT1C2A-DADA mutant (Figure S2A). Differentially, SYT1C2A Glu334Ala doubled the Kd (534 ± 60 mM) while reducing slightly its thermal stability (page 7 figure S2A)

      2)The authors described an increase in the inflexion point temperature with increasing calcium concentration. They noted that the effect was measurable from 30 µM to 300 µM. Looking at the plot with the first derivative ratio (Figure 2C), there is also an apparent change in the inflexion point temperature from 300 µM to 3 mM. Does this mean that the SYT1C2A domain binds more than two calcium ions?

      The ligand induced stabilization of proteins results in changes of the thermally induced melting curves for the ligand complexed relative to the uncomplexed proteins. This effect is used to unequivocally identify ligand hits for a particular protein from large libraries of compounds and to provide an initial estimation of the binding affinity. However, a deeper analysis of the ligand binding affinities using this technique is discouraged as the increase of melting temperature with the ligand concentration does not saturate to a particular value. This is why the effect of Ca2+ addition to SYT1C2A was also measurable from 300 µM to 3 mM, and it does not necessarily imply the binding of Ca2+ to another site. Consequently, we used other techniques such as ITC or the analysis of the change of intrinsic fluorescence upon Ca2+ addition to precisely characterize the Ca2+ binding affinities of SYT1C2A. In the present version of the manuscript, we have indicated that the change in Ti vs Ca2+ concentration is measurable when moving from 30 mM to 300 mM, thus demonstrating a Ca2+-binding event “is initiated” in this concentration range (page 7)

      3)To address lipid-binding properties of the SYT1 C2 domains, the authors used two independent methods, lipid co-sedimentation assay and BLI. In the Material and Methods section, the authors wrote that a solution of liposomes was sonicated for seven minutes to achieve homogeneity. How was homogeneity checked? Standard protocols for the liposome co-sedimentation assay use the extruder to achieve a homogeneous population of the liposomes. Also, the authors noted that the liposomes were resuspended in the buffer containing 50 mM Tris/HCl, 80 mM KCl, and 5 mM NaCl. The liposomes are usually loaded with sugar molecules, like raffinose at this step to allow subsequent co-sedimentation using centrifugation.

      Following the reviewer recommendation, we have used an extruder to achieve a homogeneous population of the liposomes (see the Methods section). This has produced an improvement of the data from the statistical point of view. However, we did not employ any sugar to facilitate lipid sedimentation as we found that 1 hour centrifugation at 58,000 rpm using a TLA100 rotor (Beckman) was enough to separate adequately soluble and precipitated fractions.

      The authors wrote that the samples were centrifuged at 58,000 rpm. The information does not allow reproducibility without rotor specification.

      We have now included the rotor specification in the methods section.

      The bound protein was estimated by subtracting the supernatant from the total amount of protein used in the assay. Direct estimation of the protein amounts in the bound fraction via e.g. SDS-PAGE would be more suitable.

      We respectfully disagree with the reviewer in this issue. The reviewer may note that the differences in the fraction of bound protein to the liposomes are at maximum around 25%. Such values are statistically significant using spectrophotometric techniques but there will be less accurately determined by the analysis of an SDS-PAGE. In addition, the later would require several washing steps of the insoluble fraction that may induce additional errors. This is well documented in a previous work from our group (Diaz et al, PNAS 2015). There, the lipid binding properties of WT and mutant C2 domain are compared using both approaches; in this work it is shown that the spectrophotometric techniques showed significant differences with the SDS-PAGE, which resulted just indicative.

      Nevertheless, we have prepared the requested SDS-PAGE for reviewer evaluation (see below) and if required we will include it as a new panel in figure 4 or include it as supplementary material. The SDS-PAGE shows the amount of soluble protein after the incubation with liposomes. It is clearly shown a reduction in the amount of WT and DADA soluble protein upon incubation with PCPSPI liposomes with respect to the sample incubated with PCPS liposomes. Differentially, no effect is observed for the PolyB and WT in presence of EGTA. Sample migrates abnormally producing a double band in presence of EGTA, probably due to the effect of removing the structural Ca site.

      Critical controls for the liposome co-sedimentation assay are missing:

      Do the SYT1 C2 domains bind liposomes without negatively charged lipids (i.e. PC-only liposomes)?

      Does the SYT1C2A-DADA mutant domain bind the PC/PS/PI liposomes?

      Does the SYT1C2A-PolyB mutant interact with the PC/PS liposomes?

      We have included all the controls suggested by the referee in the present version of the manuscript, in the results section and in figure 4A.

      The authors wrote both in the main text and the figure 4 legend that they used a lipid monolayer in the BLI method. However, in the Material and Methods section, they wrote that they used small unilamellar vesicles.

      The starting material for lipid monolayer immobilization at the biosensor tip is a solution of small unilamellar vesicles. We have clarified this issue in the methods section.

      4)The authors beneficially used all-atom MD simulations to address mechanistic details of the SYT1 C2 domains with two different lipid bilayers. However, several issues need to be addressed.

      Is there a particular reason to include sitosterol in the MD simulations? Does sitosterol contribute to protein binding?

      We clarify this issue in the present version of the manuscript. The use of sitosterol or stigmasterol in research involving plant membranes and membrane-associated proteins is recommended (DOI: 10.1063/1.4983655 PMID: 28595398) to recapitulate the fluidity and thermotropic properties of model membranes (DOI: 10.1016/j.colsurfb.2019.110422 PMID: 31437609). Sitosterol, as is the case for cholesterol in mammal membranes, contributes to packing the lipid bilayer more tightly into a liquid ordered phase (DOI: 10.1016/j.chemphyslip.2017.01.003 PMID: 28088325) and this aspect is important in molecular dynamics simulations to attain equilibration more effectively (DOI: 10.1016/j.jcis.2011.02.048 PMID: 21429500).

      We observed one hydrogen bond contact between Asn 338 at loop L3 and sitosterol in PSPI-M. We clarify this point in the results section.

      Why was not the simulation ran for a longer time? What was a criterion to determine that the system reached a stable state? Results of the MD simulations are presented as static snapshots. The manuscript would benefit from a more detailed analysis, e.g. the number of hydrogen bonds between the protein and phospholipid molecules over time, development of the tilt angle over time, etc.

      Following the reviewer’s recommendations, we now provide data showing the dynamic features of our MD calculations. In particular, we have compared the overlay of the different structures along the simulation of the SYT1C2A domain attached to the PS-Membrane and to the PSPIP-Membrane highlighting those amino acids hydrogen bonded to phospholipids (Figure S4A and S4B). This picture illustrates well that the pattern of interactions is conserved along the simulation. In addition, it also shows that the orientation of the domain with respect to the plane of the membrane is conserved. In this respect, as the reviewer requested, we have also included, a picture illustrating the change in the tilt angle of the C2A with respect to the plane of the membrane along the simulation. This analysis reveals that the tilt angle with respect to the membrane is 20 degrees larger for the PSPIP-Membrane than for the PS membrane due to the interaction of PIP molecules with the polybasic site (Figure S4E).

      In addition, we now present time traces illustrating the course of the simulations. In particular, those corresponding to the RMSD of the individual SYT1C2A, SYT1C2B and the linker between them. They clearly show that the simulations reached equilibrium within the time sampled (Figure S5H).

      The authors performed the MD simulations for both SYT1 C2 domains. It would be informative to include an electrostatic potential mapped on the surface of the SYT1C2B domain similarly to figure 3. The experimental results showed that SYT1C2B binds liposomes containing PC/PS. How does SYT1C2B interact with the PC/PS membrane? The authors noted that SYT1C2 inserts loop L3 into the lipid bilayer and that this loop adopts a β-hairpin structure. This is the case for the simulation with the SYT1C2AB fragment, but not for the SYT1C2B domain alone. Is the β-hairpin formed during the MD simulation? Or is it a result of the template-based modelling?

      To clarify reviewers’ concerns, we have included a supplementary figure illustrating the RMSD per residue along the MD simulation for the SYT1C2AB protein fragment in solution, and attached to PSPI-M. The representation illustrates an overall reduction of the RMSD as a result of the protein stabilization at the membrane, which is more significant at the membrane binding loops. In particular, the highly flexible SYT1C2B L3 loop in solution becomes highly stabilized upon membrane interaction and it folds as a beta hairpin.

      We have calculated the electrostatic potential mapped on the surface of the SYT1C2B. As it does not come from an experimental structure, we decided not to include in figure 3. The map, which is depicted below in the orientation shown in figure 3, shows that SYT1C2B might not display a polybasic site in accordance with our experimental results.

      How is the proposed binding model of the SYT1C2AB affected by the data obtained using SAXS?

      SAXS data showed that SYT1C2AB fragment may adopt a V-shaped compact structure and an extended conformation in which there is no interaction between the C2A and C2B domains. Interestingly, the V-shaped structure in solution resembles the proposed binding mode for SYT1C2AB to the membrane. We clarify this point in page 15.

      Minor comments

      1) SYT1CB construct is not listed in figure 1A along with the other constructs used in the study.

      We have modified Figure 1B to include SYT1C2B

      2) Phospholipids contain a phosphate group rather than a phosphoryl group.

      We have modified the text to correct this.

      3) In the text, the authors sometimes used the ratio 350 nm/330 nm and sometimes 330 nm/350 nm.

      This has been corrected in Figure 2C

      4) Figure S2, displayed curves look strikingly similar, different line representation does not allow a proper comparison. There are no units for y-axes.

      The figure has been corrected according to the reviewer’s recommendation for proper comparison.

      5) Figure 4B, y-axes do not have scales.

      The Y-axis from the BlitZ system represents the thickening of the layer of proteins attached to the biosensor tip and it is given in nm. However, this parameter could be meaningless when comparing different membranes and/or proteins. Hence, we scaled the plots to the maximum thickness for comparison purposes.

      6) Figure 5B, IP3 or I3P?

      We have corrected the labels corresponding to inositol triphosphate IP3 at figure 5B

      7) The authors noted that based on their SAXS experiments, the SYT1-SMPC2A has a maximum size of 176 Å and they wrote that this value is in accordance with the size of the previously characterized E-Syt2-SMPC2AB protein. However, as the authors reported, E-Syt2-SMPC2AB is two times smaller.

      We agree with the reviewer, the size of the SMP tunnel of E-Syt2-SMPC2AB is two times smaller than the Dmax size of SYT1 SMPC2A. We clarify this point in the text.

      Yours sincerely

      Armando Albert

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

      Learn more at Review Commons


      Referee #3

      Evidence, reproducibility and clarity

      The focus of this manuscript is the structural characterization of Arabidopsis synaptotagmin 1 (SYT1), which is a plant ortholog of the mammalian extended synaptotagmins (E-Syts). E-Syts play a principal role in the non-vesicular lipid transfer at the endoplasmic reticulum - plasma membrane contact sites. The function of the extended synaptotagmins is given by a unique combination of the evolutionarily conserved domains. The authors combined both experimental and computational methods to characterize different domains of Arabidopsis SYT1.

      The authors solved the structure of one of the SYT1 C2 domains (SYT1C2A) using X-ray crystallography. The obtained structure is highly similar to the previously characterized C2A domain of mammalian E-Syt2. In contrast to the orthologous C2A domain of E-Syt2, only two calcium ions are coordinated by the SYT1C2A domain. Next, the authors aimed to characterize the identified calcium-binding sites by two different experimental techniques, namely isothermal titration calorimetry (ITC) and differential scanning fluorimetry. To study lipid interactions of SYT1, the authors used a combination of the experimental approach (liposome-binding assays and biolayer interferometry, BLI) and computational methods (ligand-protein docking and molecular dynamics simulations). The authors found that both SYT1 C2 domains can interact with the lipid bilayer containing anionic phospholipids. Intriguingly, the authors described two lipid-interacting sites in the SYT1C2A domain. One site is involved in the coordination of phosphatidylserine (PS) in a calcium-dependent manner, and the second one coordinates molecules of phosphatidylinositol 4,5-bisphosphate (PIP2) through mostly electrostatic interactions. Last, the authors study the flexibility of SYT1 using partially overlapping protein fragments via small-angle X-ray scattering. Three hinge points are suggested to be responsible for the high flexibility of SYT1.

      The notion that SYT1C2A domain confers two independent lipid-binding sites, one unique for PIP2 molecules, and the second one regulated by calcium ions, is very compelling. In addition, the results describing the potential SYT1 flexibility provide novel insight into the function of extended synaptotagmins. However, my enthusiasm is mitigated by several points listed below:

      1)The authors investigated SYT1C2A calcium-binding sites using two different methods, ITC and differential scanning fluorimetry. By using ITC, they described the first binding site coordinating calcium ions in the nanomolar range. The second calcium-binding site was then characterized by differential scanning fluorimetry. The second calcium-binding site binds calcium with Kd of 277 µM. The authors then mutated SYT1C2A at two positions and performed again differential scanning fluorimetry. In this case, they did not observe any blue shift in intrinsic fluorescence concluding that "calcium-binding is mediated by the calcium-binding site". It is not clear which binding site the authors mean. In the structure of SYT1C2A, the mutated residues (D276 and D282) are shared by both calcium-binding sites. It is, therefore, difficult to interpret the data. The authors should generate a unique mutation for each site and perform both ITC and differential scanning fluorimetry.

      2)The authors described an increase in the inflexion point temperature with increasing calcium concentration. They noted that the effect was measurable from 30 µM to 300 µM. Looking at the plot with the first derivative ratio (Figure 2C), there is also an apparent change in the inflexion point temperature from 300 µM to 3 mM. Does this mean that the SYT1C2A domain binds more than two calcium ions?

      3)To address lipid-binding properties of the SYT1 C2 domains, the authors used two independent methods, lipid co-sedimentation assay and BLI. In the Material and Methods section, the authors wrote that a solution of liposomes was sonicated for seven minutes to achieve homogeneity. How was homogeneity checked? Standard protocols for the liposome co-sedimentation assay use the extruder to achieve a homogeneous population of the liposomes. Also, the authors noted that the liposomes were resuspended in the buffer containing 50 mM Tris/HCl, 80 mM KCl, and 5 mM NaCl. The liposomes are usually loaded with sugar molecules, like raffinose at this step to allow subsequent co-sedimentation using centrifugation. The authors wrote that the samples were centrifuged at 58,000 rpm. The information does not allow reproducibility without rotor specification. The bound protein was estimated by subtracting the supernatant from the total amount of protein used in the assay. Direct estimation of the protein amounts in the bound fraction via e.g. SDS-PAGE would be more suitable. Critical controls for the liposome co-sedimentation assay are missing. Do the SYT1 C2 domains bind liposomes without negatively charged lipids (i.e. PC-only liposomes)? Does the SYT1C2A-DADA mutant domain bind the PC/PS/PI liposomes? Does the SYT1C2A-PolyB mutant interact with the PC/PS liposomes? The authors wrote both in the main text and the figure 4 legend that they used a lipid monolayer in the BLI method. However, in the Material and Methods section, they wrote that they used small unilamellar vesicles.

      4)The authors beneficially used all-atom MD simulations to address mechanistic details of the SYT1 C2 domains with two different lipid bilayers. However, several issues need to be addressed. Is there a particular reason to include sitosterol in the MD simulations? Does sitosterol contribute to protein binding? Why was not the simulation ran for a longer time? What was a criterion to determine that the system reached a stable state? Results of the MD simulations are presented as static snapshots. The manuscript would benefit from a more detailed analysis, e.g. the number of hydrogen bonds between the protein and phospholipid molecules over time, development of the tilt angle over time, etc. The authors performed the MD simulations for both SYT1 C2 domains. It would be informative to include an electrostatic potential mapped on the surface of the SYT1C2B domain similarly to figure 3. The experimental results showed that SYT1C2B binds liposomes containing PC/PS. How does SYT1C2B interact with the PC/PS membrane? The authors noted that SYT1C2 inserts loop L3 into the lipid bilayer and that this loop adopts a β-hairpin structure. This is the case for the simulation with the SYT1C2AB fragment, but not for the SYT1C2B domain alone. Is the β-hairpin formed during the MD simulation? Or is it a result of the template-based modelling? How is the proposed binding model of the SYT1C2AB affected by the data obtained using SAXS?

      Minor comments:

      1) SYT1CB construct is not listed in figure 1A along with the other constructs used in the study.

      2) Phospholipids contain a phosphate group rather than a phosphoryl group.

      3) In the text, the authors sometimes used the ratio 350 nm/330 nm and sometimes 330 nm/350 nm.

      4) Figure S2, displayed curves look strikingly similar, different line representation does not allow a proper comparison. There are no units for y-axes.

      5) Figure 4B, y-axes do not have scales.

      6) Figure 5B, IP3 or I3P?

      7) The authors noted that based on their SAXS experiments, the SYT1-SMPC2A has a maximum size of 176 Å and they wrote that this value is in accordance with the size of the previously characterized E-Syt2-SMPC2AB protein. However, as the authors reported, E-Syt2-SMPC2AB is two times smaller.

      Significance

      Nature and significance of the advance:

      By a combination of experimental and computational methods, the manuscript provides novel insight into the structure-function relationship of plant SYT1.

      Work in the context of the existing literature:

      The submitted manuscript deals with Arabidopsis SYT1 protein, which belongs to an evolutionarily conserved family of proteins functioning at the ER-PM contact sites. Several recent reports described a principal role of Arabidopsis SYT1 in the regulation of lipid homeostasis, plasmodesmata functions and the response to various stresses (Lee et al. 2019. Ionic stress enhances ER-PM connectivity via phosphoinositide-associated SYT1 contact site expansion in Arabidopsis. PNAS 116, 1420-1429. https://doi.org/10.1073/pnas.1818099116; Ishikawa et al. 2020. Structural and functional relationships between plasmodesmata and plant endoplasmic reticulum-plasma membrane contact sites consisting of three synaptotagmins. New Phytologist 226, 798-808. https://doi.org/10.1111/nph.16391; Ruiz-Lopez et al. 2021. Synaptotagmins at the endoplasmic reticulum-plasma membrane contact sites maintain diacylglycerol homeostasis during abiotic stress. The Plant Cell. https://doi.org/10.1093/plcell/koab122). However, in contrast to mammalian orthologs, mechanistic details of the plant SYT1 function are largely missing.

      Audience:

      The manuscript might be of interest to the community of plant molecular biologists, structural biologists dealing with peripheral membrane proteins and computational biologists.

      Reviewer's expertise:

      Reviewer's field of expertise is plant molecular biology, plant biochemistry, protein-protein, protein-membrane and ion-membrane interactions, molecular dynamics simulations, integrative structural biology, structural bioinformatics.

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

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      This work focuses on Synaptotagmin1 (SYT1), a regulatory element of plant membrane contact site, that act as a tether and physically and functionally connects the endoplasmic reticulum (ER) to the plasma membrane (PM). Like many membrane contact site proteins, SYT1 harbours lipid transfer activity through its SMP domain. Plant SYTs proteins are orthologous to the Extended Synaptotagmin and Tricalbin ER-PM tethers from animal and yeast. SYT1 presents two C2 (C2A and C2B) lipid-binding domains at their C-terminus that are determinant for PM binding, presumably regulating SYT1 function at ER-PM membrane contact sites.

      In this paper, Benavente et al. aimed at investigating the molecular mechanisms of SYT1 binding to the PM and specificity of function of C2A (previous work has shown that SYT1 C2A, but not SYT1 C2B, binds membrane lipids in a Ca2+ dependant manner). The authors combined a wide range of approaches from X-ray crystallography to biophysics and in silico molecular modelling to understand the mechanisms of SYT1 C2A interaction with lipids, at the molecular level. From their study, Benavente et al. shows that C2A display dual lipid binding activity interacting with PS in a Ca2+-dependant manner and with phosphoinositide in a Ca2+-independent manner. These interactions involve two distinct sites; a polybasic amino acid site for phosphoinositides binding and a Ca2+-dependant lipid binding site for PS. They propose that this two-steps binding mechanism confers plasticity in membrane docking under low and high intracellular calcium concentration. They also show that SYT1 full length protein displays three flexible hinges, which they propose confers SYT1 a high degree of conformational freedom

      Minor comments

      • Typos: please re-read carefully through the manuscript to remove them.
      • We advise the authors to have the manuscript corrected by a native english-speaker.

      Significance

      C2- domains are central functional elements of SYT/E-SYT/Tricalbin ER-PM tethers. They regulate PM docking through their lipids binding activity, which can be calcium-dependant or calcium-independent, but also presents lipid specificity (PS, phosphoinositides...). Beside their lipid-binding activity, C2 domains have also been shown to be involved in protein-protein interaction, including intramolecular interaction to inhibit lipid-transfer activity (E-SYT). Multiple C2s are present in SYT/E-SYT/Tricalbin tethers and their diversity of function is likely central for providing functional specificity but also plasticity in response to developmental /environmental clues (together with changes membrane lipid composition and intracellular calcium levels). Decrypting C2 individual operating mode is therefore central. This work is important as it investigates SYT1A docking mechanisms at the molecular level and identifies a number of very interesting features. However, as it stands, the paper does not make the link with SYT1 functionality. The authors discuss their findings within the frame of experimental observations that are already published but these remain speculative.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      Summary:

      Provide a short summary of the findings and key conclus