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

Review: In this manuscript the authors generated macrophages derived from the THP-1 cell line or human peripheral blood mononuclear cells stimulated with MCSF and infected them with alphaviruses some containing GFP expression cassettes. In Figure 1, they demonstrate that CHIKV infected these cells more robustly than RRV, SINV or the related ONNV. The authors generated an extensive array of CHIKV/ONNV chimeras to identify the viral proteins that dictate release from infected macrophages and narrowed it down to the envelop proteins E1 and E2. Fine mapping identified a couple of single mutations that affected macrophage infection outcomes. The authors then shifted their approach to identifying env protein interactors using a myc-tag pulldown methods followed by mass spectrometry. The assay identified a number of proteins including those involved in vesicular transport and interferon pathways. siRNA knockdown experiments were performed to identify interactors and many of them were shown to improve virus output.

Critique: Overall, the manuscript is well written but in its current state it is more like two different stories because the effects of envelop proteins and list of interactors are not brought together in on one story. A possible fix to this problem would be inclusion of ONNV and CHIKV containing env mutations that do and do not restore viral release from macrophages into the pulldown/association experiments shown in Figure 6. The other major issue is the lack of protein data for the viral mutants relative to WT ONNV and CHIKV and assessment of viral RNA in the supernatants to determine whether the block is release or an earlier event since viral RNA levels in the cell seems to be the same or at least normalized. Lastly, knockdown experiments indicate an effect of things like OAS3 or other innate immune modulators. There are no controls to demonstrate that these are specific to CHIKV infection or if knockdown would assist growth of ONNV as well.

Other points to consider:

1. The title does not fit the manuscript findings and should be modified.
2. It is unclear why the authors show results for SINV and RRV in Figure 1. Either these should be removed or the viruses should be carried throughout the experiments described in the Figure. Better yet would be to add additional alphaviruses to this analysis to determine if there are additional viruses that act similarly to CHIKV.
3. Is the data presented in Figure 1A significant?
4. The justification for inclusion of Figure 4A is lacking. It is unclear what this panel is supposed to be demonstrating.
5. There is little justification for the candiates assessed in
6. Extended data Figure 3 is very difficult to read due to the small font size.
7. Just to be clear, the blots shown in Figure 6D are different from those depicted in Extended data Figure 4b, because some of them look very similar.

Significance

The study provides a fresh look at Alphavirus replication in macrophages. There are a number of issues that should be worked out that would enhance impact and interpretation of this study.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary: The authors utilize: 1) chimeric arthritogenic alphaviruses; evolution selection analyses with virus sequences isolated from human patients; and 3) mass spectrometry and proteomics to interrogate determinants of chikungunya virus (CHIKV) permissiveness in primary human macrophages and the human macrophage cell line, THP-1. The authors find that the vaccine strain, CHIKV 181/clone 25 replicates the most efficiently in primary monocyte-derived macrophages compared to other arthritogenic alphaviruses. Using o'nyong o'nyong (ONNV) as a comparison, the authors generate several chimeric viruses with CHIKV structural proteins and ONNV non-structural proteins (and vice versa) and perform a series of E1 and E2 domain swap experiments. They determine that both CHIKV structural proteins, E2 and E1, are necessary to confer efficient virus production over ONNV in the absence of a difference in viral RNA production. The authors also identify a specific residue in E1 that appears to be important for efficient virus production in THP-1 macrophage cell lines. Finally, using mass spectrometry, the authors identify two host proteins, SPCS3 and eIF3k, that bind to CHIKV E1 structural protein and appear to act as antiviral host factors.

Major comments: The authors elegantly demonstrate that CHIKV structural proteins confer an advantage over ONNV structural proteins in a step in the replication cycle downstream of virus RNA synthesis, possibly virion assembly. This point would be strengthened determining the particle-to-PFU ratio of the parental viruses and the chimeras. Presumably, the ratio would increase in the chimeras containing CHIKV structural proteins. Additionally, the authors should consider performing virion assembly blocking assays with a small molecule inhibitor to determine if this abrogates the virus production advantage of CHIKV structural proteins within the ONNV backbone. Finally, the authors should perform competition experiments with the chimeric viruses and ONNV in macrophages to determine if the chimeras can outcompete the parental ONNV strain. Based on their data, the chimeric viruses should outcompete. These experiments would likely take 3-4 weeks to complete.

The authors use both primary macrophages and macrophage cell lines as their in vitro model system and make one of their major points (listed in the title) that the determinants they identified in the CHIKV structural proteins convert macrophages into dissemination vessels; however, they do not show: 1) an in vivo model that the CHIKV-ONNV chimeras disseminate more efficiently than the parental ONNV; and 2) that these chimeras generate virus more efficiently specifically in macrophages. It would be useful to show that ONNV and CHIKV have equivalent virion production in other cell lines and that the advantage conferred by CHIKV structural proteins in the ONNV backbone is specific to macrophages. The authors should also change their title to reflect that dissemination is not directly being addressed in their study; the implications of their in vitro experimentation in a mammalian host would be more appropriate for the discussion.

OPTIONAL: The authors use CHIKV-ONNV chimeras but it would be interesting to test other chimeras to determine if CHIKV structural proteins confer the same advantage in the backbone of other arthritogenic alphaviruses. The study would also be strengthened by using a pathogenic strain of CHIKV instead of the vaccine strain, as this is significantly attenuated in vivo. In Figure 4, the authors identify residues in the CHIKV structural proteins that appear to be under positive selection in human subjects and generate point mutants in these residues with the corresponding ONNV residues. They find that one mutation, V1029I located in E1, completely abolishes virion production in THP-1 macrophage cell lines. However, in their previous chimeric experiments, they find that neither CHIKV E1 or E2 was sufficient to increase virus production in the ONNV backbone. The authors should address this discrepancy, otherwise they should consider moving the data in their point mutation experiments to a supplementary figure. While worthy of reporting, especially given the patient data, these experiments do not buttress the points made in the previous figures.

The authors conclude their manuscript with an assessment of several host proteins, namely SPCS3 and eIF3k, that were identified by mass spectrometry and whose knockdown results in increased virion production. The authors speculate about the role of these proteins but do not provide any mechanistic detail on how they might be playing a role. It is unclear that the putative antiviral role of these proteins involves steps downstream of virus replication, especially given that the authors speculate translation might be affected by eIF3k which, if the case, RNA synthesis should also be expected to be affected.

Overall, while the initial chimeric virus and domain swap approach is strong, the manuscript would benefit with a more thorough examination of virion assembly steps and a mechanistic link to virion production. Otherwise, the authors should revise the structure of their manuscript by de-emphasizing points about virion assembly and leave room for other mechanistic explanations of their chimeric data that more clearly link the host antiviral factor/E1 binding studies.

Minor comments: In Figure 3e, the line under "with CHIKV E1" should be moved over to include the E2-II+E1 virus.

Figure 5a, 5b, and 6a should be replaced with higher resolution images.

Significance

Strengths of the study include the initial chimeric virus and domain swap approach to determine factors that allow for the productive replication of chikungunya virus in macrophages compared to other arthritogenic alphaviruses. This approach yielded useful insights and could be adapted to other viruses. The study is limited, however, by the lack of mechanistic detail linking the antiviral host factors identified which bind to the E1 structural protein, and the advantage conferred by CHIKV structural proteins in the ONNV backbone. The study would be greatly improved by structural studies of the chimeric viruses that directly demonstrate more efficient virion production and that knockdown of the identified factors specifically affects virion production. This point could be addressed either through additional experimentation or tempering of the authors' conclusions about the mechanism by which CHIKV structural proteins provide an advantage over those of ONNV.

The study advances knowledge in the field on what might advantage different pathogenic alphaviruses and explain differences in disease pathology. Additionally, the authors devise a simple and clever strategy that could be applied across different alphaviruses and would be useful to test in vivo in future studies. This study would be useful to a virology-specific audience.

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 Yao et al. show CHIK is able to infect macrophages in contrast to other arthritogenic alphaviruses RRV, ONNV, and SINV. They use a series to chimeric viruses made with ONNV, the closest species to CHIK, and determine the E2-E1 proteins are important viral determinants which allow CHIK to replicate in machophages compared to ONNV. By comparing 397 CHIK sequences from infected patients, they identified 14 residues under pervasive and positive selection. Of these, 3 residues in E2 and 3 residues in E1 (amino acids) were different between CHIK and ONNV suggesting these residues contributed to the difference in macrophage tropism of CHIK compared to ONNV. The authors go on to determine what host factors the CHIK E2 protein is interacting with to presumably connect the viral and host determinants for CHIK infection in macrophages.

Major concerns:

1. The authors show one configuration of the E1-E2 heterodimer in Figure 4d. As shown, the E1 protein is exterior to the E2 protein and would suggest E1 is on the surface on the spike complex and virus surface. However, another configuration of the glycoproteins has E2 on the exterior of E1 and also on the exterior of the virus. The latter conformation is what has been observed in cryoEM studies of alphaviruses. The first configuation represents the E1-E2 between the three heterodimers which are important for spike assembly. The reason the orientation of the E2-E1 dimer is important is the authors speculate on the importance of the 6 CHIK residues not found in ONNV based on the structure, but the structural interpretation is, in my opinion, not correct.
2. Validation of E1 interaction with SPSC3 and eIF3k needs to be stronger. Some concerns/questions are listed below. A myc tag was inserted between E3 and E2. How efficeintly does furin cleave E3 from E2 in this virus and how are viral titers of the myc-tagged virus compared to the non-tagged virus? I ask because is the IP looking at what is being pulled down by E2 or E3-myc-E2 that could be part of the spike polyprotein? The authors found E2 interacts with E3, E1 and a list of other host proteins. These results suggest several interactions including E2-host factor, E2-E1, E2-E3, E2-E1-host factor, E2-E3-E1, E2-E3-host factor. In figure 6d, and the subsequent conclusions, the authors suggest E1 is interacting with the host facor and do not see E2 alone and very low amounts of E3-E2-6K-E1. based on how the IP was performed I am not sure how an interaction between E1 and SPCS3 alone, without E2, would be detected. I would also like to see a reciprocal pull down using E1 and also E2 to see if these host factors are pulled down.
3. If CHIK E1 is interacting with the host factors and that is antagonizing the antiviral response of SPSC3 (as one example), then what do pull downs using ONNV structural proteins look like? One would expect reduced interactions because the different amino acid causes a different E2-E1 dimer or attenuates the E1-host factor binding site.
4. E1 and E2 are thought to interact during polyprotein translation and the initial dimer forms in the ER. If E1 is interacting wht SPSC3 in the ER, is E2 also present? Or is a population of E1 not interacting with E2 in order to inhibit SPSC3? I would love a model of how the authors see all these factors coming together for this new role of E1.

Minor concerns:

1. In Figure 1c, (-) RNA is shown but in the rest of the figures (+) RNA is shown. Show both or select one. I do find it interesting the (-) RNA levels are similar over time, even at 4 hours post transfection (early time). Related to this, ONNV has higher levels of (-) RNA but what is known about structural protein levels in ONNV and CHIK in macrophages? Are there comparable levels of CP and GP being produced?
2. Figure 2e and figure 3 have ONNV has the first bar followed by CHIK. In figure 1 and 2b, CHIK is first and then ONNV. helps the reader to have the controls in the same order.
3. Line 143-145 the authors discuss that when ONNV is the backbone and CHIK proteins are inserted the infection is more attenuated because of the E2 and E1 are from CHIK and ONNV, not the same virus (could also be E2-CP interactions are disrupted). However the chimeras made witht he CHIK backbone (in Figure 2) have a mismatch between E2 and E1 as well.
4. When discussing the residues that were found in the FEL and MEME analysis, the authors start the amino acid numbering from CP and continue along the polyprotein. Usually when discussing amino acids in the structural proteins, each protein starts at amino acid 1. So V460 would be E2-V135. It would also be useful to know what the residues in ONNV were at these positions to see if amino acids changed in charge, size, bond forming potential, etc. Showing these residues in the E2-E1 conformation found in the virion would also allow one to find adjeacent residues that could explain differences in spike assembly and potentially where/how E1 is binding to a host protein.
5. How effective is a non-attenuated CHIK strain in infecting macrophages? Could you make a SINV-La Reunion chimeric virus (which is BSL2) to see if a higher percentage of macrophages are infected and is this potentially contributing to the increased pathogenesis of La Reunion? Also how different is 181/25 with a pathogenic strain in the E2 and E1 resdiues? and compared to ONNV?
6. When describing the last results section, "CHIK E1 binding proteins exhibit potent anit-CHIV activities" the authors use macrophages. In the rest of the text they consistently use THP-1 macrophages or human primary monocyte derived macrophages. The details of the cell type are extremely useful to the reader and having those in the last results section would be great.
7. The paper is well-written. There is a slight disconnect as the authors go from discussing results in Figure 4 to Figure 5.

Referees cross-commenting

I agree with R#2 that having some Particle:PFU data would add some data to determine why such differences in titers/infectivity.

I also see how this m/s could be split into two different m/s. One that focuses on the chimeric viruses and another that identifies the host factors important and goes in more depth with mechanism

Significance

Strengths:

The authors have tackeled an intriguing question: why do some alphaviruses infect macrophages and others do not. They have used a chimeric approached to very systematically identify the viral determinants E2 and E1 as being important in macrophage infection. Using AP-MS they identify host factors that interact with E2 (possibly E2 and E1, see comments above) but if their findings that E1 has a role in attenuating a host antiviral factor, this would be fantastic.

More and more examples of viral proteins having multiple roles during infection are in the literature. The idea that structural proteins also attenutate host antivirals is a developing field and vastly understudied. By fleshing out the results some more the authors might be onto something ery important in alphavirus virology.

Limitations:

The study has it is presented is limited in the validation of host factors and their interacting partners. I have many questions about the methodology, validation, and model from this last section.

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 careful reading of the document and feedback which will help us to improve our manuscript. We will go through their comments one by one.

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

This study would be much convincing if additional line of eukaryotic cells can be used to demonstrate the GEF-GAP synergy tis important for cell physiology. In addition, it would be best to demonstrate the spatiotemporal interaction of GEF-GAP using high-resolution live cell imaging.

Response from the authors:

The reviewer requests additional in vivo data to support our in vitro findings:

(1) The reviewer requests in vivo data showing that GEF-GAP synergy is important for cell physiology. We believe that in order to show GEF-GAP synergy in vivo, Cdc42 cycling rates would need to be measured in vivo. For that single-molecule resolution is required – to track a single Cdc42 molecule and measure its GTPase cycling. We agree that such data would indeed be interesting, but are unaware of established techniques that would facilitate measurements of Cdc42 cycling rates in vivo.

(2) The reviewer requests in vivo data showing the spatiotemporal interaction of GEF-GAP. Cdc24 and Rga2 are shown to interact (direct or mediated by another protein) (McCusker et al. 2007, Breitkreutz et al. 2010, Chollet et al. 2020). Cdc24 and Rga2 share 11 binding partners (https://thebiogrid.org/31724/table/saccharomyces-cerevisiae-s288c/cdc24.html, https://thebiogrid.org/32438/table/saccharomyces-cerevisiae-s288c/rga2.html) and have been found at the polarity spot (Gao et al. 2011). Live cell imaging of fluorescently tagged Cdc24 and Rga2 will show that they exhibit some interaction, but not specify the role of the interaction nor if the interaction is direct or mediated by one of the shared binding partners. In order to show a direct interaction between Cdc24 and Rga2, one could consider (A) super-resolution imaging or (B) FRET experiments: For both fluorescently tagged Cdc24 and Rga2 cell lines would need to be constructed.

(A) Super-resolution imaging could show direct interaction between Cdc24 and Rga2, but even with the techniques available this would be on the limit. Further, it is usually done in fixed cells, and not in live cells (as requested from the reviewer).

(B) To show a direct interaction of Cdc24 and Rga2 using FRET, suitable protein constructs would need to be engineered. We believe that the main obstacle in showing direct binding of Cdc24 and Rga2 using FRET is to design the fluorophore linker. The linker would need to be designed in such a way that it is flexible enough to give a FRET signal even if the two large proteins bind on the opposite sites of the fluorophore, but also is stiff/short enough to not show binding if both proteins are in close proximity through binding to a common binding partner.

__We believe that an investigation of GEF GAP binding in vivo is beyond the scope of this study. Instead, we will further explore one possible mechanism underlying GEF GAP synergy - Cdc24 Rga2 binding - through conducting Size-Exclusion Chromatography Multi-Angle Light Scattering experiments with purified Cdc24 and Rga2 (alone and in combination). __

Reviewer #1 (Significance (Required)):

The revised study would provide first line evidence that GEF-GAP synergy to be general regulatory property in eukaryotic kingdom.

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

The study entitled, "The GEF Cdc24 and GAP Rga2 synergistically regulate Cdc42 GTPase cycling" by Tschirpke et al., uses an in vitro GTPase assay to examine the GTPase cycle of Cdc42 in combination with its GEF and GAP effectors. The authors find that the Cdc24 GEF activity scales non-linearly with its concentration and the GAP Rga2 has substantially weaker effect on stimulating Cdc42 GTPase activity. Not surprisingly, the combined addition of Cdc24 and Rga2 lead to a substantial increase in Cdc42 GTPase activity.

**Referees cross-commenting**

In Zheng, Y., Cerione, R., and Bender, A. (1994) J. Biol. Chem. 269: 2369-2372 (Fig. 3C), the authors show that Cdc24 combined with the GAP Bem3 lead to a large synergy in boosting Cdc42 GTPase activity.

Reviewer #2 (Significance (Required)):

There is very little new information in this manuscript. Previous studies (Rapali et al. 2017) have shown that the scaffold protein Bem1 enhances the GEF activity of Cdc24. It is expected that the reconstitution of a GEF and GAP protein promote the GTPase cycle and indeed Zheng et al. (1994) showed that that Cdc24 combined with the GAP Bem3 lead to a large synergy in boosting Cdc42 GTPase activity. Hence the only potentially interesting finding in this work is that, in solution Cdc24 activity scales non-linearly with its concentration. However as this GEF and Cdc42 are associated with the membrane, the relevance of solution studies are less clear and furthermore the mechanistic basis for the non-linearity is not explored in detail. Given the limited new information from this work, the findings are, in their current form, too preliminary.

Response from the authors:

__We appreciate the reviewer recognizing our work on the non-linear concentration-dependence of Cdc24’s activity. We disagree that this is the only new finding in our study: __

We explore the effect of Cdc24 and Rga2 on Cdc42’s entire GTPase cycle and show that Cdc24 and Rga2 synergistically upregulate Cdc42 cycling. So-far Cdc42 effectors were only characterized in isolation (with the exception of Cdc24-Bem1 (Rapali et al. 2017)) and through how they affect a specific GTPase cycle step. The regulation of single GTPase cycle steps through an effector yields mechanistic insight into this specific GTPase cycle step. However, it does not show how the effector affects overall GTPase cycling of Cdc42 – a process Cdc42 constantly undergoes in vivo. Our approach allows us to study synergistic effects between proteins affecting different GTPase cycle steps. Synergies are another regulatory layer of the polarity system, adding further complexity: Which polarity proteins exhibit synergy, to which extend? The assay employed here, which studies the entire GTPase cycle, enables studying the effect of any GTPase cycle regulator, alone and in combination with another regulator.

The reviewer states that the GEF GAP synergy is to be expected, as it was already shown in Zheng et al. 1994. In Fig. 3C Zheng et al. shows the time course of the GTPase activity of Cdc42 in presence of Cdc24, Bem3, and Cdc24 plus Bem3. Fig. 3C is the only data in which the combined effect of a GEF (Cdc24) and a GAP (Bem3) is investigated. The data indicates synergy, but is neither discussed as such in the text of the publication, nor analyzed quantitatively. Further, only one concentration of each effector (GEF/GAP) is used and the study uses a Bem3 peptide containing codons 751-1128 (30%) of the full-length BEM3 gene. Zheng et al. 1994 gives an early indication of GEF GAP synergy, but does not claim, discuss, or further investigate the synergy as such. In contrast, we use full-length Rga2 (not Bem3) as GAP, conduct several concentration-dependent assays, and analyze them quantitatively. We thank the reviewer for pointing out the pioneering character of Zheng et al.‘s study and will mention it more prominently in our report. However, we disagree that Zheng et al. sufficiently studied the GEF GAP interaction. To our awareness no theoretical studies include a GEF GAP synergy term, which we would expect if GEF GAP synergy is well-established in the field.

The reviewer criticizes the relevance of bulk in vitro studies (lacking membranes) of proteins that bind to membranes in vivo. We agree that the presence of a membrane can affect the protein’s property, and we can not exclude that membrane-binding could alter the magnitude of a GEF GAP synergy. However, we believe that membrane-binding does not impede the GEF GAP synergy altogether. If membrane binding would influence GTPase properties that strongly, other studies on Cdc42’s GTPase activity and GEF and GAP activity, that do not include a membrane, would be inconclusive as well (e.g. Zheng et al. 1993, Zheng et al. 1994, Zheng et al. 1995, Zhang et al. 1997, Zhang et al. 1998, Zhang et al. 1999, Zhang et al. 2000, Zhang et al. 2001, Smith et al. 2002, Rapali et al. 2017). Both studies mentioned by the reviewer (Zheng et al. 1994, Rapali et al. 2017) were also conducted without membranes present.

We believe that an inclusion of membrane-binding into reconstituted Cdc42 systems will enhance our understanding of Cdc42 and recognize it as a next step, which could be enabled by the assay used in our study.

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

This work reports a biochemical analysis of the effects of a recombinant yeast GEF (Cdc24) and GAP (Rga2) on Cdc42 GTPase cycling in vitro. The central conclusion is that the GEF and GAP act "synergistically", which occurs "due to proteins enhancing each other's effects". By this they appear to mean that the GEF enhances the GAP's activity and vice versa. I was not persuaded that this is correct, and was confused by many aspects of the approach and interpretation, as outlined below.

1. GEF and GAP are expected to accelerate GTPase cycle synergistically even with no effect on each other's activity:

The Cdc42 GTPase cycle is understood to occur via distinct steps (GDP release, GTP binding, and GTP hydrolysis): GDP release and GTP hydrolysis are intrinsically slow steps that are accelerated by GEFs (GDP release) and GAPs (GTP hydrolysis). This fundamental biochemistry was established in the 1990s using biochemical assays that measure each step independently. Here instead the authors use an assay that measures [GTP] decline in a mix with 5 uM starting GTP, 1 uM Cdc42, plus or minus some amount of GEF or GAP. They assume exponential decline of [GTP] with time, yielding a cycling "rate". If that is so, then one would expect that added GEF would accelerate only the first step, leaving a slow GTP hydrolysis step that limits the overall cycling rate, while added GAP would accelerate only the last step, leaving a slow GDP release step that limits the overall cycling rate. Adding both together would speed up both steps, and should therefore "synergistically" accelerate cycling. This would be expected based on previous work and does not imply that GEF or GAP are affecting each other's action (except trivially by providing substrate for the next reaction). If the authors wish to demonstrate that something more complex is indeed happening, they need to use assays that directly measure the sub-reaction of interest, as done by prior investigators.

Response from the authors:

The reviewer raises the point that we do not consider a simpler, rate-limiting model and that this rate-limiting model could explain our synergy between GAP and GEF in accelerating the GTPase cycle.

We very much welcome this consideration of the reviewer! We will add a clarification to our manuscript to explain why a rate-limiting model/interpretation does not match our data.

Intuitively, the rate-limiting model is appealing, as it permits interpretation of cycle rate increases in terms of individual biochemical steps. So, a consideration of this model is indeed relevant. However, as also noted by the reviewer in the next points, data from e.g., figure 3e are not compatible with a simple rate-limiting model with two steps (hydrolysis and nucleotide exchange). We will explain how the acceleration of the total rate by both GAP and GEF individually does not match the rate-limiting model, even if we assume maximal effects of adding GAPs and GEF to the cycle. For this purpose, we consider the rate-limiting model scenario where the sensitivity of the GTPase cycle to adding GAP/GEF is maximized, so the best case-scenario for the rate limiting step-model.

In the rate-limiting step model, we assume that we have a GTPase cycle in which at least one of the three GTPase cycle steps is rate-limiting: (A) GTP binding, (B) GTP hydrolysis, and (C) GDP release.

We assume that the addition of a GEF and GAP only accelerates GDP release and GTP hydrolysis respectively. Biochemically, all three steps in the GTPase cycle are expected to be relevant. However, here we will consider only the final two steps, as sensitivity to rate limitation by GAP/GEF is maximized when time spent in the GAP/GEF-independent step in the cycle (step A: GTP) is negligible (i.e. never rate-limiting). The two-step model thus consists of (1) a nucleotide exchange step (step C+A) which is dominated by GDP release (step C) and assumed to be accelerated exclusively by the GEF, and (2) a GTP hydrolysis step (step B) exclusively enhanced by the GAP.

In the rate limiting step model GEF-GAP synergy can appear if one of the conditions applies:

1. the addition of a GAP speeds up the GTP hydrolysis step so much that the hydrolysis step stops (or almost stops) being the rate-limiting step, or
2. the addition of a GEF speeds up the GDP release step so much that the release step stops (or almost stops) being the rate-limiting step. In these conditions, the acceleration of the GTPase cycle, accomplished by adding only a GAP or adding only a GEF, is interdependent. Therefore, we consider the possible acceleration of the GTPase cycle by GAP and GEF individually, and compare these to our observations to determine whether the rate-limiting step model can explain our data.

The GTPase cycle time Tc is thus composed of hydrolysis Th and nucleotide exchange time Te, and the rates r are connected through:

1/rc=1/rh + 1/re

If we compare the ratio of the rates with protein (GAP/GEF) added in the assay (index 1) with the basal rate without protein added (index 0), we obtain the cycle acceleration factor alpha:

alpha=rc1/rc0=(1/rh0 + 1/re0)/(1/rh1 + 1/re1)=(re0 + rh0)/(re0*rh0/rh1 + rh0*re0/re1)

Here, rc1 and rc0 are the total GTPase cycle rate with and without effector respectively, rh1 and rh0 are the GTP hydrolysis rate with and without effector respectively, and re1 and re0 are the nucleotide exchange rate with and without effectors respectively.

There is indeed an interdependence created between how much the GAP and GEF can both accelerate the total cycle, if the GAP and GEF are assumed to only accelerate GTP hydrolysis and nucleotide exchange respectively. E.g., how much the total GTPase cycle rate rc is accelerated by an increase in GTP hydrolysis rate rh depends on and can be limited by the current nucleotide exchange rate re. However, this interdependence is too strict to match the data in Figure 3e, as we will explain in the next paragraphs:

When we only add a GAP and the GAP accelerates only the GTP hydrolysis rate (re1=re0), then the maximal total GTPase cycle rate acceleration alphaGAP that the GAP can accomplish is when rh1>>rh0,re0:

alphaGAP=rc1/rc0=(1/rh0 +1/re0)/(1/rh1+1/re0)=(re0+rh0)/(re0*rh0/rh1+rh0)

~(re0+rh0)/rh0=1+ re0/rh0

We thus assume the GAP accelerates the cycle so much that the hydrolysis step is much faster than the exchange step, at which point the effect of adding more GAP would saturate. We note that we do not consider the GAP concentration regime where we see saturation, thus in reality the acceleration by the GAP is more restricted than predicted here.

Analogously, if the GEF accelerates only the nucleotide exchange rate (rh1=rh0), then the maximum GTPase cycle rate ratio will be when re1>>re0,rh0 , yielding acceleration factor alphaGEF :

alphaGEF= rc1/rc0=1+ rh0/re0

Again, note we assume the GEF accelerates the cycle so much that the exchange step is much faster than the hydrolysis step, at which point the effect of adding more GEF would saturate. We note that we do not observe the GEF concentration regime where we see saturation, thus in reality the acceleration by the GEF is more restricted than predicted here.

We see that the maximum gain in rates for GAP-only and GEF-only assays is limited by the same basal GTP hydrolysis and nucleotide exchange rates (rh0 and re0), leading to the following interdependence:

alphaGAP=1+ 1/(alphaGEF -1)=alphaGEF/(AlphaGEF -1)

In our GAP-only and GEF-only assays (Fig. 3e, Tab. 2), we see both a 2-fold and 100-fold increase in the total rate respectively. A 100-fold acceleration factor of the GEF would maximize the GAP acceleration factor to 1.01 (or alternatively, the 2-fold GAP acceleration would maximize the GEF acceleration to 2), which are both significantly lower than what we observe. So even though we made favorable assumptions for the rate-limiting model to maximize rate sensitivity to GAP/GEF, namely neglecting nucleotide binding and assuming GAP/GEF concentrations that saturate in their effects, we still cannot reproduce the acceleration factors in our GAP-only and GEF-only assays.

Moreover, a rate-limiting step model would also imply saturation effects as stated in the next point of the reviewer. While we observe saturation in total rate acceleration for certain GAP concentrations, we use GEF and GAP concentrations in the combined protein assays for which no saturation effects were observed. Absence of saturation in both cycle steps simultaneously is also not reconcilable with the rate-limiting step model, as will be further discussed in the next point of the reviewer.

In summary, this means that the rate-limiting model is not sufficient to explain our results: the GAP/GEF synergy we observe is not simply resulting from GEF and GAP independently lifting two different rate-limiting steps.

Model-based interpretation of the GTPase assay is poorly supported:

The assay employed measures overall GTP concentration with time. It is assumed (but not well documented-see below) that [GTP] declines exponentially, and that the rate constant for a particular condition can be fit by the sum of a series of terms that are linear or quadratic in the concentrations of Cdc42, GEF, and GAP. There is no theoretical derivation of this model from the elementary reactions, and the assumptions involved are not well articulated.

As discussed in point 1 above, one would expect that a GEF or GAP alone could only accelerate the cycle to a certain point, where the other (slow) reaction becomes rate limiting. But that does not appear to be true for their phenomenological model, where slow steps (small terms in the sum) will always be overwhelmed by fast steps. This is not the traditional understanding of how GTPases operate.

Response from the authors:

The reviewer expresses the concern that because we do not derive our coarse-grained model from elementary reactions, we miss important effects that can occur when adding GAP and GEFs, particularly saturation.

We understand the concern of the reviewer that if a rate-limiting step model is considered, saturation effects of GAP/GEF will limit the amount with which these effectors can speed up the total cycle. Our coarse-grained model indeed does not account for this saturation. However, as discussed in the previous point of the reviewer, we do not opt for the rate-limiting model interpretation, as the GAP and GEF effects are not compatible with the rate-limiting step model.

Secondly, we agree that for high enough concentrations of GEF and GAPs, we would experience a saturation in the effect of adding the effectors. We are aware of this possibility, and we verify that we are not in saturation regimes with our added proteins by checking the plots of the individual protein titrations (see Figure 3a-d). If we enter the saturation regime, we expect a negative second derivative in the rate as function of protein concentration (the curve shallows off). We do not see this for any protein except for Rga2 at some point, as discussed in our main text of the manuscript. However, for this protein we only use the data in the linear regime for further analysis. In short, we understand the concern of the author but we empirically check that we are not in the saturation regime.

Data that do not conform to expectation are not explained: Strangely, the data (as interpreted by the model assumptions) also appear inconsistent with the expectation of rate-limiting steps. GEF addition (alone) is said to accelerate cycling 100-fold, while GAP addition (alone) accelerates it 2-fold. But that would seem to imply that GDP release takes up >99% of the basal cycle (so accelerating that step alone reduces cycling time 100-fold), while GTP hydrolysis takes up >50% of the basal cycle (so accelerating that step alone reduces cycling time 2-fold). In the conventional understanding of GTPase cycles, these cannot both be be true (as the steps would then add to >100% of the basal cycle). There is no attempt to reconcile these findings with previous work.

Response from the authors:

The reviewer raises the point that our findings do not match the expectations of the rate-limiting model perspective.

We fully agree with the reviewer that our data is not compatible with the rate-limiting step model. The 100-fold and 2-fold gain of the total cycle rates for GEF-only and GAP-only assays are one of our arguments against the rate-limiting model view, as described in the first point of the reviewer. Also, our lack of saturation as described in the previous point of the reviewer provides another argument against using expectations based on rate-limiting steps to interpret our findings.

Lack of detailed timecourse data:

The decline in [GTP] with time is stated to be exponential, allowing extraction of an overall cycling "rate". But this claim is supported only weakly (S3 Fig. 1 uses only 3 timepoints, is not plotted on semi-log axis, and does not report fit to exponential vs other models) and only for the Cdc42-alone scenario: no data at all are presented to support exponential decline in reactions with GEF or GAP. Most assays seem to measure only a single timepoint, so extraction of a "rate" is very heavily influenced by the unsupported assumption of exponential decline. And if the decline is not exponential, it becomes extremely difficult to interpret what a single timepoint means.

Response from the authors:

The reviewer requests additional timeseries data with GEF and GAP to support the assumption of an exponential decline of GTP in the assay and requests to plot it on a semi-log axis.

We will add data for Cdc42 + Cdc24 and for Cdc42 + Rga2 with two to three time points, and plot it as requested on a semi-log axis.

Other issues with interpretation of the data:

(i) It is unclear why the authors chose to employ an assay that is much harder to interpret than the biochemical assays used by others. In biochemical studies, assays that report an output of multiple reactions are always harder to interpret than assays targeting a single reaction. As well-established assays are available for each individual step in GTPase cycles, any conclusions must be supported using such assays.

Response from the authors:

The reviewer wonders why an assay that investigates several GTPase steps at once was chosen over assays that investigate sub-steps of the GTPase cycle, given that these give more mechanistic insights.

We agree that assays investigating GTPase cycle substeps can give more mechanistic insights into these specific steps. However, they do not allow to study how proteins affecting different steps act together. We were interested in investigating the overall GTPase cycle of Cdc42 and a possible interplay of GEFs and GAPs. Cdc42 GTPase cycling was found to be a requirement for polarity establishment (Wedlich-Soldner et al. 2004) and Cdc42 GTPase cycling is physiologically relevant. Ultimately, we hope that in vitro results provide stepping stones towards understanding the complex and less controlled in vivo environment. The in vivo environment often entails the output of many reactions combined, so there is every incentive to study aggregated effects of a full cycle which are not necessarily the sum of individual outputs.

__We believe that both assay types – assays that investigate sub-steps and yield mechanistic details, and assays that investigate the entire cycle – are important and disagree that one assay type is superior to the other. Instead, we believe they complement each other. __

(ii) The reported basal (and GEF/GAP-accelerated) rates are very slow, perhaps due to poor folding of recombinant proteins. This raises the possibility that much of the Cdc42 is inactive. If so, then accelerated GTP hydrolysis could come from increasing the active fraction of Cdc42, rather than catalyzing a specific step.

Response from the authors:

The reviewer wonders whether the reported rates are slow due to poor folding of recombinant Cdc42. We used S. cerevisae Cdc42, for which it has been shown that it has a significantly lower basal GTPase activity than Cdc42 of other organisms (see Zhang et al. 1999). Many other studies on Cdc42 were conducted with human Cdc42, which has a significantly higher basal GTPase activity (Zhang et al. 1999). We assessed the activity of several recombinantly expressed Cdc42 constructs previously (Tschirpke et al. 2023). We there observed that most constructs had a similar GTPase activity, only some purification batches and constructs had a significantly reduced GTPase activity (which might be linked to poor folding). The Cdc42 construct used here shows a similar activity as the active Cdc42 constructs in Tschirpke et al. 2023, and we therefore believe that it exhibits proper folding. If recombinant Cdc42 folds poorly, we would expect greater variations between Cdc42 constructs and purification batches (caused by different levels of folding/ a different fraction of active Cdc42) than what we observed previously (see Tschirpke et al. 2023).

Tschirpke et al. 2023:

Tschirpke et al. A guide to the in vitro reconstitution of Cdc42 activity and its regulation (2023) BioRxiv. (https://doi.org/10.1101/2023.04.24.538075) (in submission at Current Protocols)

(iii) The GEF and GAP preparations include multiple partial degradation products and it is unclear whether the measured activities come from full-length proteins or more active fragments.

Response from the authors:

We agree with the reviewer that the Cdc24 and Rga2 preparations contain degradation products.

It would be more ideal if the protein purifications were entirely pure, but this is experimentally very difficult to achieve for the used proteins (which are large and partially unstructured, making them prone to partial degradation). Further, it is not uncommon to use protein preparations where some degradation products were present (e.g. Zheng et al. 1993, Zheng et al. 1994). Other studies did not show their purified preparations.

The vast majority of the Cdc24 preparation is the full-length protein. We therefore expect that the degradation fragments only contribute in a small extend to the overall protein behavior.

The Rga2 preparation contains a higher amount of degradation product, but only larger size protein fragments (> 60kDa), suggesting that the fragments contain at least and more than 1/3 of the full-length protein (the protein fragments are thus the size or larger than of the GAP peptides used previously). The fragments could in principle have a higher or lower activity. We account for fragments of no/lower activity by comparing our cycling rates to those of BSA/Casein, which has no specific effect on Cdc42. The cycling rate Rga2 is almost an order of magnitude greater than that of BSA/Casein, suggesting that the effect of the full-length protein dominates. We could only imagine that a Rga2 fragment has a higher GAP activity if the fragment consists mainly of the GAP domain and if in Rga2 the activity of the GAP domain is downregulated. Nevertheless, we will do an additional experiment using a purified GAP domain peptide to assess that if a GAP domain by itself has a higher GAP activity than our Rga2 preparation. Using that data, we will discuss possible implication of the GAP fragments in our manuscript.

(iv) Cdc42 cycling is also accelerated by BSA and casein, suggesting that there are poorly understood aspects of the assay and that GEF and GAP actions may (like BSA and casein) involve non-canonical effects on Cdc42. As GEF and GAP are expected to interact better with Cdc42 than BSA or casein, these effects could dominate the observed changes in GTP levels.

Response from the authors:

The reviewer raises the concern that the effects of the added effector proteins on the rates could be caused by non-canonical effects. We do not believe non-canonical effects play a relevant role in our assays. While BSA and casein accelerate the GTPase cycle in our assays, the GAP effect and GEF effect are orders of magnitude stronger.

(v) Cdc42-alone cycling assays are said to be reproducible. However, assays with added GEF/GAP/BSA/Casein yield rates that vary almost an order of magnitude between replicates. This poor reproducibility further reduces confidence in the findings.

Response from the authors:

The reviewer is concerned about the variations in Cdc42 effector rates.

__We disagree that the variations are concerning and believe to have accounted for them in our analysis: __The Cdc42 (Cdc42 alone) data is very reproducible (see Tschirpke et al. 2023). The GTPase assay is generally sensitive to small concentration changes and errors introduced through pipetting small volumes (as required for the assay). We believe that the small variation observed for Cdc42 alone is because Cdc42 has such a low basal rate and therefore the small concentration changes due to pipetting have a smaller effect. Once other effectors are added, especially highly GTPase stimulating ones as Cdc24, small concentration changes due to pipetting can lead to larger variations between assays (small variations in Cdc24 concentration lead to larger changes in remaining GTP due to Cdc24’s strong and non-linear effect on Cdc42). We conduct the assays multiple times to account for these variations. In our analysis we do not compare single rate numbers but the orders of magnitude of the rate, and report the variations present. Even given the present variations, the differences in effect sizes are still significant. We map and discuss assay variation in (Tschirpke et al. 2023), to which we refer to several times throughout the manuscript.

Tschirpke et al. 2023:

Tschirpke et al. A guide to the in vitro reconstitution of Cdc42 activity and its regulation (2023) BioRxiv. (https://doi.org/10.1101/2023.04.24.538075) (in submission at Current Protocols)

(vi) It is unclear what timepoint was used for the different assays. 1.5 h at 30 degrees seems to be the standard here for the Cdc42-alone assays, but I assume that cannot be what was measured to assess GTP decline for GEF-containing assays as there would be very little GTP left at 1.5 h.

Response from the authors:

We used 60-100 min as incubation times for all assays. The assay data will be published on a data server, where all these numbers can be checked. We further added a clarification to the materials and methods section. In order to still have remaining GTP for the Cdc42 GEF mixtures after 60-100 min, we lowered the used protein concentrations.

(vii) The graph reporting GEF activity is plotted only for [GEF]Response from the authors:

The graphs show the full range of protein concentrations used.

In order to calculate K1, K2, K3,Cdc24, K3,Rga2, K3,Cdc24,Rga2 from k1, k2, k3,Cdc24, k3,Rga2, k3,Cdc24,Rga2, …, a protein concentration has to be included in the term (as K1 = k1 [Cdc42], ….). In order to make K comparable, we chose to use 1uM for all protein concentrations. This was done to compare the cycling rate values of different proteins. 1uM was a choice, in the same fashion 0.2uM could have been chosen.

__We will further discuss in the manuscript how the choices in protein concentration affect the effector strength on Cdc42. __

(viii) S8 Data with casein seems very noisy and it is no longer at all clear that the quadratic fit for [Cdc24] is justified. Also, the symbol colors are very similar so it is hard to tell what data corresponds to what condition. The synergy between Cdc24 and Rga2 is also very noisy and the fits seem arbitrary.

Response from the authors:

The reviewer is concerned with (1) the noise in the S8 data, and (2) the Cdc42-Cdc24-Rga2 fits.

(1) We acknowledge in the manuscript that the S8 data is noisy and should be viewed with caution. We do not put much emphasis on these data sets and their interpretation and show them only in the supplement.

(2) We disagree that the Cdc42-Cdc24-Rga2 fits are arbitrary. The fits contain several data points per protein, and reproduce the rate values from Cdc42-Cdc24 and Cdc42-Rga2 assays well.

The reviewer is concerned with the color scheme choice in the fits.

__We will adapt the color scheme of the fits to make the colors more distinguishable. __

(ix) It is disturbing that different Cdc42 constructs behave quite differently (S4). This suggests that protein behavior is influenced by the various added epitope tags and protease cleavage sites (they also leave the C-terminal CAAX box rather than removing the AAX as would happen in vivo). These features raise the concern that these findings may not be directly relevant to the situation with endogenous yeast Cdc42. Of course, it is also the case that relevant Cdc42 biochemistry occurs with prenylated Cdc42 on membranes.

Response from the authors:

The reviewer is concerned that the behavior of the Cdc42 constructs is influenced by their tags. In a previous manuscript (Tschirpke et al. 2023) we explored the effect of various N- and C-terminal tags on Cdc42, by comparing it to Cdc42 that is not tagged in that position. We found that most tags, including the tags present in the Cdc42 construct used here, do not affect Cdc42’s properties.

Instead, we found a general, tag independent, heterogeneity in Cdc42 behavior (which can occur between purification batches and between constructs (but not between different assays)): in some batches GTPase activity depended quadratically on its concentration, others showed a linear relationship. Most batches exhibited a mixed behavior. The differences between the batches are generally small, and only visible in the activity to concentration plots and because of the assay’s high accuracy. We use a two-parameter fit (k1 [Cdc42] + k2 [Cdc42]2) to phenomenologically account for this heterogeneity, and to estimate the basal Cdc42 GTPase activity. We do not interpret this heterogeneity, as more research is needed. We believe that Cdc42 still has unexplored properties, of which this heterogeneous behavior can be one. We speculate in Tschirpke et al. 2023 that it is linked to Cdc42 dimerization mediated by its polybasic region, a relationship that is far from being fully understood yet. __We believe that it is of scientific interest to point out heterogeneous behaviors to encourage more research. __

Tschirpke et al. 2023:

Tschirpke et al. A guide to the in vitro reconstitution of Cdc42 activity and its regulation (2023) BioRxiv. (https://doi.org/10.1101/2023.04.24.538075) (in submission at Current Protocols)

The reviewer is concerned that our findings are biologically not relevant, as our experiments (1) included Cdc42 that was not prenylated and (2) did not include membranes.

(1) We here used recombinantly purified proteins, which do not contain posttranslational modifications, such as prenylations. So-far Cdc42’s prenyl group, which is responsible for binding it to membranes, has not been linked to its GTPase properties. We therefore believe that unprenylated Cdc42 is an equal choice to prenylated Cdc42 when studying Cdc42’s GTPase cycle. Further, the use of recombinantly purified proteins can be of advantage: when proteins are purified from their native host, the post-translationally modified protein is purified. However, many proteins contain a multitude of post-translational modifications (PTMs). Thus, the purified protein is a mixture of protein with different PTMs. For example, S. cerevisae Cdc42 undergoes ubiquitinylation (Swaney et al. 2013, Back, Gorman, Vogel, & Silva 2019), phosphorylation (Lanz et al. 2021), farnesylation and geranyl-geranylation (Caplin, Hettich, & Marshall 1994). We here used protein preparations that do not contain PTMs, and show how they behave. Natively purified proteins would be mixtures of various PTMs, and the observed protein behavior would be that of the mixture. If Cdc42’s PTMs affect it’s GTPase behavior, the observed behavior of natively purified Cdc42 would represent the average behavior of the mixture. It then would require additional work to disentangle which PTMs affect the GTPase cycling in which way. The use of recombinantly expressed Cdc42 does not require this work, and can set the baseline for how Cdc42 without PTMs behaves. If in the future a link between Cdc42’s GTPase behavior and PTMs are found, the work here could be used as a baseline for Cdc42’s behavior when it is without PTMs.

(2) The concern about missing membranes was also raised by reviewer 2 (significance), and we like to refer to our response there.

Reviewer #3 (Significance (Required)):

The basic biochemistry of Cdc42 cycles was figured out about 30 years ago. However, those studies did not examine how combinations of Cdc42 regulators (as opposed to individual regulators) might interact to produce effects not expected from combining their individual actions. Recently, this combination approach did lead to interesting findings by Rapali et al. This approach is worthwhile and addresses a major question of interest to the broader field of GTPase biochemistry.

One main limitation of this study is technical: the main assay is less informative (though perhaps easier) than traditional assays, and it is unclear whether the recombinant proteins employed retain their normal activities. Another limitation is the model-based interpretation of the assay that does not include the potential for rate-limiting steps.

Response from the authors:

We thank the reviewer for the detailed comments.

One important point of confusion originated from our lack of discussion concerning a rate-limiting step model, which is an obvious starting point for modelling the GTPase cycle. We thank the reviewer for pointing this out, and we will include an explanation in our manuscript why we reject this model and instead opt for a coarse-grained model.

Firstly, a rate-limiting model would generate saturation effects that we would observe when adding GEF and/or GAPs. In assays exploring GEF GAP synergy we use GEF and GAP concentrations for which no saturation effects were observed.

Secondly, in our data we observed a two-fold increase of the total GTPase cycling rate when adding a GAP and a 100-fold rate increase when a GEF is added. These increases are not compatible with a model where either hydrolysis or nucleotide exchange limits the GTPase cycle. While a synergy could arise from the rate-limiting model perspective, the incompatibility of the rate-limiting model with the GAP-only and GEF-only assay data excludes this synergy explanation. Finally, through coarse-graining our model we avoid using single step parameters from literature which are incompatible in terms of proteins/buffers used. (For example; the mayor studies that kinetically characterized the individual GTPase steps of Cdc42 used human Cdc42 (Zhang et al. 1997, Zhang et al. 2000). Because human Cdc42 exhibits a higher basal GTPase activity (Zhang et al. 1999) we are skeptical how useful it is to transfer these parameters to S. cerevisae Cdc42.)

At the same time, coarse-graining our model permits absorbing unidentified molecular details which is essential when we wish to incorporate BSA and casein rate contributions.

The reviewer finds our assay, which investigates the GTPase cycle as a whole, less informative. Assays investigating single GTPase cycle sub-steps give more mechanistic insights into these steps. We opted for an assay that studies GTPase cycling as a whole instead, as we were interested in studying how proteins effecting different steps act together. We believe that both assay types are important as they complement each other.

The reviewer is concerned about our use of recombinant proteins, and whether they retain their normal activities. We assessed Cdc42’s GTPase activity and the influence of added purification tags extensively (Tschirpke et al. 2023), and found that added tags do not affect Cdc42’s GTPase properties. We checked Cdc24’s GEF activity using the GTPase assay and found that it bound strongly to Bem1, as expected (Tschirpke et al. 2023). The Cdc24 concentrations needed to affect Cdc42’s GTPase activity were similar to those used previously (Rapali et al. 2017), suggesting that it is fully active. A similar comparison for Rga2 was not possible, as so-far only domains of Rga2 were used (Smith et al. 2002). We here used recombinantly purified proteins, which do not contain posttranslational modifications (PTMs). To our knowledge the PTMs of the herein used proteins are not linked to their GTPase/GEF/GAP properties. Thus, a lack of PTMs does not diminish our findings. Further, when proteins are purified from their native host, the post-translationally modified protein is purified. However, many proteins contain a multitude of post-translational modifications in vivo. Natively purified proteins would be mixtures of various PTMs, and the observed protein behavior would be that of the mixture. We here used protein preparations that do not contain PTMs, and show how they behave, setting the baseline for proteins without PTMs behaves. If in the future a link between GTPase behavior and PTMs are found, the work here could be used as a baseline for the proteins behavior when it is without PTMs.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

Summary

The GTPase cdc42 is a key determinant of yeast polarization. Its activity is amplified at the site of polarization through a poorly defined positive feedback mechanism, and depends on numerous GAPs regulating GTP hydrolysis and the GEF cdc24 that regulates GDP release. These components have previously been evaluated for their quantitative effects on the individual steps in the GTPase cycle that they modulate, but potential interactions between the cdc24 GEF and any GAP could not be examined based on these assays. The authors validate and employ a bulk assay of the total GTPase cycle based on GTP consumption to study the activities of and potential interactions between cdc24 and the GAP Rga2. Fitting their data to a mathematical model, they come to three central conclusions: (1) the activating activity of cdc24 to activate cdc42 GTPase activity is nonlinear, showing a quadratic relationship, (2) Rga2 shows a much lower activating activity that is linear at low levels before saturating, and (3) there is a strongly synergistic interaction between the activating activities of cdc24 and Rga2. Some hypotheses for the mechanistic bases of these findings are hypothesized, but not further investigated. Their conclusions are well supported by the data which appears to be of sufficient rigor.

Major comments

The three main conclusions of the manuscript are well supported by the data and associated modeling.

One unresolved issue is the discrepancy between the authors' conclusion that the non-linear activation by cdc24 is likely a result of oligomerization, whereas Mionnet et al 2008 reach the opposite conclusion. It seems that the authors wish to discount the Mionnet results because they used truncated constructs to test deficient oligomerization and an engineered construct to test induced oligomerization. If the authors are correct, then a relatively easy test would be to introduce the oligomerization deficient mutants defined by Mionnet into their fuill length construct and compare to wild type protein. While the authors' measured results don't depend on the offered mechanism and this experiment is therefore optional, their explanation is quite unsatisfying, especially since an experiment to resolve the difference is entirely feasible and not very strenuous.

Response from the authors:

__The reviewer suggests to conduct experiments with oligomerization deficient Cdc24 mutants to test our hypothesis that the non-linear concentration dependence of Cdc24’s activity is due to Cdc24 oligomerization. __

We agree that this is an insightful experiment, and will conduct it. In order to observe the effect in our GTPase assays, we require a mutant that is oligomerizes substantially less than wild-type protein. Mionnet et al. constructed several Cdc24 mutants, but none were entirely oligomerization deficient. However, the DH5 (L339A/E340A) mutant showed a 10-fold reduction in oligomerization and the DH3 (F322A) mutant exhibited 2.5-fold reduction in oligomerization. We will therefore use the DH5 and DH3 mutant for two additional experiments.

Minor comments

The results in Fig S4 serve as assay validation, and this should be pointed out early in the Results section. I was initially concerned when the assay was described as based on consumption of GTP that a significantly diminished pool would alter the rate and thereby distort results, and being made aware of the S4 result would have alleviated that concern as I read further.

Response from the authors:

We believe that the reviewer refers to S3 (not S4). We appreciate this suggestion and now mention it earlier.

On page 4 and Fig S4 the authors mention several cdc42 constructs, some of which show linear activity curves and others slightly non-linear curves. I was unable to find where these constructs or their differences are discussed. The authors should also tell us if the construct used for the remaining experiments was one of the two shown in S4, or a different one.

Response from the authors:

We added the requested information and explanations to the manuscript.

It seems that in Fig 4 and Fig S8, some points are missing from the graphs. Were all concentrations for each condition not always assayed, or is some data omitted for some reason? For example, for the 0.125 microM Rga2 condition, only two points are shown vs 4 for some other conditions, and the two missing ones are expected to not be excluded by the >5% GTP remaining criterion.

Response from the authors:

The reviewer wonders whether Fig.4 and Fig. S8 miss data points. This is not the case, and __we added clarifying information to the manuscript. __

In detail: Not all assays contain the same amount of data points/ concentrations for each protein. We first assessed Cdc42 alone using several Cdc42 concentration. We then examined the individual Cdc42 – effector mixtures, using a larger number of effector concentrations. We included a reduced number of effector concentrations in the assays containing two effectors and Cdc42. It would be ideal to include more concentrations, but this is not always feasible: The assay involves a multitude of pipetting steps and is sensitive to any pipetting errors. Further, assays can vary slights from each other, therefore all samples that ought to be compared need to be included in each assay.

Each three-protein assay contains samples shown (Cdc42, Cdc42 + effector 1, Cdc42 + effector 2, Cdc42 + effector 1 + effector 2) and additional ‘buffer’ wells used for normalization. Each data point shown corresponds to the average of 3-4 replica samples per assay. We therefore did not include all concentrations in all conditions. As pointed out, Fig. 4a only shows two data points for the 0.125uM Rga2 axis (Rga2 + Cdc42 and Rga2 + Cdc24 + Cdc42). The rational was the following: We included three Cdc24 concentrations (for proper fitting for K3,Cdc24), three Rga2 concentrations (for proper fitting for K3,Rga2), and 5 mixtures of the used Cdc24 and Rga2 concentrations (for proper fitting for K3,Cdc24,Rga2).

The Cdc42-Rga2-BSA and Cdc42-Rga2-Casein data is rather sparse and would benefit from additional data points. However, we only use those as control experiments and are cautious in their interpretation.

In these graphs, a diamond symbol of slightly varying color is used for the different conditions. The different colors are hard to distinguish. Please use different shape symbols for the different conditions, and choose colors that are more distinct.

Response from the authors:

We will adapt the color scheme of the fits to make the colors more distinguishable.

There are a few sentences that are of unclear meaning, for example on page 10, "It was suggested that each GAP plays a distinct role in Cdc42 regulation, of which the level of GAP activity could be a part of [Smith et al., 2002]." There are also typos and grammatical errors that should be fixed.

Response from the authors:

__We will further check the document for potentially unclear sentences and will try to clarify them, as well as further check for grammatical and spelling errors. __

Reviewer #4 (Significance (Required)):

Significance

The most novel and important finding is the strong synergy observed between cdc24 and Rga2 in activating cdc42 GTPase activity. This is undoubtedly an important mechanism underlying positive feedback in polarization. The measured non-linear activity of cdc24 alone is also quite important given that availability of cdc24 is thought to be a critical in vivo stimulus for polarization. However, the unexplained discrepancy between this result and that of Mionnet leaves one to wonder which result is more reliable. Only Mionnet attempts to directly test whether oligomerization is important in cdc24 activity.

The conclusions are of importance to a broad audience of cell biologists, though the lack of any mechanism for the synergy or the non-linearity of cdc24 activity somewhat diminishes significance.

Note that my expertise and that of my co-reviewer is in the biology, and while we are able to follow the contributions of the modeling, we do not have the expertise to critically evaluate for potential errors or weaknesses in the modeling itself.

The reviewer wonders whether our data or the data of Mionnet et al. on the link between Cdc24 oligomerization and its GEF activity is more reliable and suggests to conduct experiments with oligomerization deficient Cdc24 mutants.

We thank the reviewer for this recommendation and we will do the suggested experiments to resolve the seemingly contradicting observations by us and Mionnet et al..

The reviewer would find mechanistic insights into (2) the non-linear concentration dependence of Cdc24’s activity and (2) the Cdc24-Rga2 synergy useful.

(1) We will conduct experiments with partially oligomerization deficient Cdc24 mutants, as suggested by the reviewer.

(2) We speculate that Cdc24-Rga2 binding could lead to the synergy. ____We will add data on Cdc24 – Rga2 binding (in vitro: Size-Exclusion Chromatography Multi-Angle Light Scattering) to this study.

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

Summary

The GTPase cdc42 is a key determinant of yeast polarization. Its activity is amplified at the site of polarization through a poorly defined positive feedback mechanism, and depends on numerous GAPs regulating GTP hydrolysis and the GEF cdc24 that regulates GDP release. These components have previously been evaluated for their quantitative effects on the individual steps in the GTPase cycle that they modulate, but potential interactions between the cdc24 GEF and any GAP could not be examined based on these assays. The authors validate and employ a bulk assay of the total GTPase cycle based on GTP consumption to study the activities of and potential interactions between cdc24 and the GAP Rga2. Fitting their data to a mathematical model, they come to three central conclusions: (1) the activating activity of cdc24 to activate cdc42 GTPase activity is nonlinear, showing a quadratic relationship, (2) Rga2 shows a much lower activating activity that is linear at low levels before saturating, and (3) there is a strongly synergistic interaction between the activating activities of cdc24 and Rga2. Some hypotheses for the mechanistic bases of these findings are hypothesized, but not further investigated. Their conclusions are well supported by the data which appears to be of sufficient rigor.

Major comments

The three main conclusions of the manuscript are well supported by the data and associated modeling.

One unresolved issue is the discrepancy between the authors' conclusion that the non-linear activation by cdc24 is likely a result of oligomerization, whereas Mionnet et al 2008 reach the opposite conclusion. It seems that the authors wish to discount the Mionnet results because they used truncated constructs to test deficient oligomerization and an engineered construct to test induced oligomerization. If the authors are correct, then a relatively easy test would be to introduce the oligomerization deficient mutants defined by Mionnet into their fuill length construct and compare to wild type protein. While the authors' measured results don't depend on the offered mechanism and this experiment is therefore optional, their explanation is quite unsatisfying, especially since an experiment to resolve the difference is entirely feasible and not very strenuous.

Minor comments

The results in Fig S4 serve as assay validation, and this should be pointed out early in the Results section. I was initially concerned when the assay was described as based on consumption of GTP that a significantly diminished pool would alter the rate and thereby distort results, and being made aware of the S4 result would have alleviated that concern as I read further.

On page 4 and Fig S4 the authors mention several cdc42 constructs, some of which show linear activity curves and others slightly non-linear curves. I was unable to find where these constructs or their differences are discussed. The authors should also tell us if the construct used for the remaining experiments was one of the two shown in S4, or a different one.

It seems that in Fig 4 and Fig S8, some points are missing from the graphs. Were all concentrations for each condition not always assayed, or is some data omitted for some reason? For example, for the 0.125 microM Rga2 condition, only two points are shown vs 4 for some other conditions, and the two missing ones are expected to not be excluded by the >5% GTP remaining criterion.

In these graphs, a diamond symbol of slightly varying color is used for the different conditions. The different colors are hard to distinguish. Please use different shape symbols for the different conditions, and choose colors that are more distinct.

There are a few sentences that are of unclear meaning, for example on page 10, "It was suggested that each GAP plays a distinct role in Cdc42 regulation, of which the level of GAP activity could be a part of [Smith et al., 2002]." There are also typos and grammatical errors that should be fixed.

Significance

The most novel and important finding is the strong synergy observed between cdc24 and Rga2 in activating cdc42 GTPase activity. This is undoubtedly an important mechanism underlying positive feedback in polarization. The measured non-linear activity of cdc24 alone is also quite important given that availability of cdc24 is thought to be a critical in vivo stimulus for polarization. However, the unexplained discrepancy between this result and that of Mionnet leaves one to wonder which result is more reliable. Only Mionnet attempts to directly test whether oligomerization is important in cdc24 activity.

The conclusions are of importance to a broad audience of cell biologists, though the lack of any mechanism for the synergy or the non-linearity of cdc24 activity somewhat diminishes significance.

Note that my expertise and that of my co-reviewer is in the biology, and while we are able to follow the contributions of the modeling, we do not have the expertise to critically evaluate for potential errors or weaknesses in the modeling itself.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This work reports a biochemical analysis of the effects of a recombinant yeast GEF (Cdc24) and GAP (Rga2) on Cdc42 GTPase cycling in vitro. The central conclusion is that the GEF and GAP act "synergistically", which occurs "due to proteins enhancing each other's effects". By this they appear to mean that the GEF enhances the GAP's activity and vice versa. I was not persuaded that this is correct, and was confused by many aspects of the approach and interpretation, as outlined below.

1. GEF and GAP are expected to accelerate GTPase cycle synergistically even with no effect on each other's activity:

The Cdc42 GTPase cycle is understood to occur via distinct steps (GDP release, GTP binding, and GTP hydrolysis): GDP release and GTP hydrolysis are intrinsically slow steps that are accelerated by GEFs (GDP release) and GAPs (GTP hydrolysis). This fundamental biochemistry was established in the 1990s using biochemical assays that measure each step independently. Here instead the authors use an assay that measures [GTP] decline in a mix with 5 uM starting GTP, 1 uM Cdc42, plus or minus some amount of GEF or GAP. They assume exponential decline of [GTP] with time, yielding a cycling "rate". If that is so, then one would expect that added GEF would accelerate only the first step, leaving a slow GTP hydrolysis step that limits the overall cycling rate, while added GAP would accelerate only the last step, leaving a slow GDP release step that limits the overall cycling rate. Adding both together would speed up both steps, and should therefore "synergistically" accelerate cycling. This would be expected based on previous work and does not imply that GEF or GAP are affecting each other's action (except trivially by providing substrate for the next reaction). If the authors wish to demonstrate that something more complex is indeed happening, they need to use assays that directly measure the sub-reaction of interest, as done by prior investigators. 2. Model-based interpretation of the GTPase assay is poorly supported:

The assay employed measures overall GTP concentration with time. It is assumed (but not well documented-see below) that [GTP] declines exponentially, and that the rate constant for a particular condition can be fit by the sum of a series of terms that are linear or quadratic in the concentrations of Cdc42, GEF, and GAP. There is no theoretical derivation of this model from the elementary reactions, and the assumptions involved are not well articulated.

As discussed in point 1 above, one would expect that a GEF or GAP alone could only accelerate the cycle to a certain point, where the other (slow) reaction becomes rate limiting. But that does not appear to be true for their phenomenological model, where slow steps (small terms in the sum) will always be overwhelmed by fast steps. This is not the traditional understanding of how GTPases operate. 3. Data that do not conform to expectation are not explained: Strangely, the data (as interpreted by the model assumptions) also appear inconsistent with the expectation of rate-limiting steps. GEF addition (alone) is said to accelerate cycling 100-fold, while GAP addition (alone) accelerates it 2-fold. But that would seem to imply that GDP release takes up >99% of the basal cycle (so accelerating that step alone reduces cycling time 100-fold), while GTP hydrolysis takes up >50% of the basal cycle (so accelerating that step alone reduces cycling time 2-fold). In the conventional understanding of GTPase cycles, these cannot both be be true (as the steps would then add to >100% of the basal cycle). There is no attempt to reconcile these findings with previous work. 4. Lack of detailed timecourse data:

The decline in [GTP] with time is stated to be exponential, allowing extraction of an overall cycling "rate". But this claim is supported only weakly (S3 Fig. 1 uses only 3 timepoints, is not plotted on semi-log axis, and does not report fit to exponential vs other models) and only for the Cdc42-alone scenario: no data at all are presented to support exponential decline in reactions with GEF or GAP. Most assays seem to measure only a single timepoint, so extraction of a "rate" is very heavily influenced by the unsupported assumption of exponential decline. And if the decline is not exponential, it becomes extremely difficult to interpret what a single timepoint means. 5. Other issues with interpretation of the data:

(i) It is unclear why the authors chose to employ an assay that is much harder to interpret than the biochemical assays used by others. In biochemical studies, assays that report an output of multiple reactions are always harder to interpret than assays targeting a single reaction. As well-established assays are available for each individual step in GTPase cycles, any conclusions must be supported using such assays.

(ii) The reported basal (and GEF/GAP-accelerated) rates are very slow, perhaps due to poor folding of recombinant proteins. This raises the possibility that much of the Cdc42 is inactive. If so, then accelerated GTP hydrolysis could come from increasing the active fraction of Cdc42, rather than catalyzing a specific step.

(iii) The GEF and GAP preparations include multiple partial degradation products and it is unclear whether the measured activities come from full-length proteins or more active fragments.

(iv) Cdc42 cycling is also accelerated by BSA and casein, suggesting that there are poorly understood aspects of the assay and that GEF and GAP actions may (like BSA and casein) involve non-canonical effects on Cdc42. As GEF and GAP are expected to interact better with Cdc42 than BSA or casein, these effects could dominate the observed changes in GTP levels.

(v) Cdc42-alone cycling assays are said to be reproducible. However, assays with added GEF/GAP/BSA/Casein yield rates that vary almost an order of magnitude between replicates. This poor reproducibility further reduces confidence in the findings.

(vi) It is unclear what timepoint was used for the different assays. 1.5 h at 30 degrees seems to be the standard here for the Cdc42-alone assays, but I assume that cannot be what was measured to assess GTP decline for GEF-containing assays as there would be very little GTP left at 1.5 h.

(vii) The graph reporting GEF activity is plotted only for [GEF]<0.2 uM, but the rates used in the subsequent experiments are reported for mixtures with 1 uM GEF. The full range of GEF data should be plotted.

(viii) S8 Data with casein seems very noisy and it is no longer at all clear that the quadratic fit for [Cdc24] is justified. Also, the symbol colors are very similar so it is hard to tell what data corresponds to what condition. The synergy between Cdc24 and Rga2 is also very noisy and the fits seem arbitrary.

(ix) It is disturbing that different Cdc42 constructs behave quite differently (S4). This suggests that protein behavior is influenced by the various added epitope tags and protease cleavage sites (they also leave the C-terminal CAAX box rather than removing the AAX as would happen in vivo). These features raise the concern that these findings may not be directly relevant to the situation with endogenous yeast Cdc42. Of course, it is also the case that relevant Cdc42 biochemistry occurs with prenylated Cdc42 on membranes.

Significance

The basic biochemistry of Cdc42 cycles was figured out about 30 years ago. However, those studies did not examine how combinations of Cdc42 regulators (as opposed to individual regulators) might interact to produce effects not expected from combining their individual actions. Recently, this combination approach did lead to interesting findings by Rapali et al. This approach is worthwhile and addresses a major question of interest to the broader field of GTPase biochemistry.

One main limitation of this study is technical: the main assay is less informative (though perhaps easier) than traditional assays, and it is unclear whether the recombinant proteins employed retain their normal activities. Another limitation is the model-based interpretation of the assay that does not include the potential for rate-limiting steps.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The study entitled, "The GEF Cdc24 and GAP Rga2 synergistically regulate Cdc42 GTPase cycling" by Tschirpke et al., uses an in vitro GTPase assay to examine the GTPase cycle of Cdc42 in combination with its GEF and GAP effectors. The authors find that the Cdc24 GEF activity scales non-linearly with its concentration and the GAP Rga2 has substantially weaker effect on stimulating Cdc42 GTPase activity. Not surprisingly, the combined addition of Cdc24 and Rga2 lead to a substantial increase in Cdc42 GTPase activity.

Referees cross-commenting

In Zheng, Y., Cerione, R., and Bender, A. (1994) J. Biol. Chem. 269: 2369-2372 (Fig. 3C), the authors show that Cdc24 combined with the GAP Bem3 lead to a large synergy in boosting Cdc42 GTPase activity.

Significance

There is very little new information in this manuscript. Previous studies (Rapali et al. 2017) have shown that the scaffold protein Bem1 enhances the GEF activity of Cdc24. It is expected that the reconstitution of a GEF and GAP protein promote the GTPase cycle and indeed Zheng et al. (1994) showed that that Cdc24 combined with the GAP Bem3 lead to a large synergy in boosting Cdc42 GTPase activity. Hence the only potentially interesting finding in this work is that, in solution Cdc24 activity scales non-linearly with its concentration. However as this GEF and Cdc42 are associated with the membrane, the relevance of solution studies are less clear and furthermore the mechanistic basis for the non-linearity is not explored in detail. Given the limited new information from this work, the findings are, in their current form, too preliminary.

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 would be much convincing if additional line of eukaryotic cells can be used to demonstrate the GEF-GAP synergy tis important for cell physiology. In addition, it would be best to demonstrate the spatiotemporal interaction of GEF-GAP using high-resolution live cell imaging.

Significance

The revised study would provide first line evidence that GEF-GAP synergy to be general regulatory property in eukaryotic kingdom.

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 would like to thank Review Commons for their innovative approach to scientific peer-review and publishing. We thank all the Reviewers for their positive, highly complementary assessment of the manuscript and for highlighting the high quality and reproducibility of the work and the novelty and significance of the results: “The experiments are well-designed and perfectly executed and presented”; “I felt that this is a strong manuscript for peer-review as it serves diversified interests in modern cell biology.”; “The manuscript would be of interest to basic researchers working on epithelial development. Also potentially to basic researchers working on cancer, due to the mitotic errors described.”. We are grateful for the Reviewers’ comments and suggestions that have contributed to improving the revised manuscript. We have addressed all the Reviewers’ concerns, as detailed below in the point-by-point response to the Reviewers. Textual changes in the revised manuscript are marked in Blue.

__Reviewer #1 (Evidence, reproducibility and clarity (Required)): __

*The manuscript "Crosstalk between the plasma membrane and cell-cell adhesion maintains epithelial identity for correct polarised cell divisions" by Dr. Hosawi and colleagues reports the characterisation of the mitotic connection between plasma membrane dynamics and division orientation in polarised mammalian epithelial cells in culture. The authors start from the comparison of mitotic events of human mammary MCF10A cells grown at optimal density or at low density. They observed that only at optimal density MCF10A cells polarise by E-cadherin mediated cell-cell contacts, and display uniform membrane enrichment at the cortex, whereas cells grown at low density do not show cortical E-Cadherin enrichment, and distribute aberrantly the plasma membrane at one side and in cytoplasmic vesicles, generating daughter cells with unequal size. Consistently, further analyses revealed that low-density MCF10A cells undergo misoriented mitosis, with chromosome congression and misegregetion defects. Mechanistically, low density MCF10A cells fail to organise a symmetric mitotic spindle and center it in metaphase. This is due to an increased cortical actomyosin thickness coupled to abnormal astral microtubule stability. Building on previous data from the Elias lab, the authors uncover a role of the membrane-associated S100A11 protein in maintaining correct plasma membrane dynamics and E-cadherin localisation in mitosis. Further dissection of the molecular mechanism underlying this mitotic function od S10011A revealed that it enriches at the cortex only in optimal-density MCF10A cells, and promotes spindle orientation by association with LGN and E-cadherin, upstream of E-cadherin. This evidence depicts the plasma membrane and S100A11 proteins as a key mechanical sensors of cell-cell adhesion orchestrating the recruitment of E-cadherin and LGN-dependent force generators to ensure correct division orientation. *

*Major points: *

*- Important information is presented in Supplementary Figure S3. I suggest to move these panels in the main figures. Specifically, I would replace figure 4A with S3A showing the distribution of endogenous S100A11 in MCF10A cells, rather than the one of the GFP-tagged version which is over-expressed. *

__Authors response: __We thank the Reviewer for this suggestion. We have now moved Figure S3A to Figure 4, to replace Figure 4A and show the localisation of endogenous S100A11 during mitosis and included new quantifications in new Figure 4B. We have moved Figure 3A to supplementary figures (new Figure S4A). We have amended the text of the results section and the Source Data file accordingly.

*- The mechanisms of division orientation governed by S100A11 seems to impinge on the control of cortical F-actin and astral microtubule dynamics. This is illustrated in figure S3C, which in my opinion should be shown in the main figures with some more explanation / experiments. The authors mention the " tight actin F-actin bundles at the cell-cell contacts" that are lost in S100A11-depleted cells, and that interact with astral microtubules. However this is not fully clear in figure S3C. I think the authors should find a way to present better these evidence which is key in supporting their molecular model. *

__Authors response: __As requested by the Reviewer we have now moved Figure S3C to the main manuscript, as new Figure 5. To clarify further the effect of S100A11 depletion on the tight actin bundle formation at the cell-cell contacts, we have now included a new illustration in new Figure 5C. Additionally, we have clarified further these findings in the results section (page 11). While we agree with the Reviewer that additional experiments, for example using live imaging of MCF-10A cells co-labelled for F-actin and tubulin, would help assess further the crosstalk between cortical actin and astral microtubules, based on our experience these live imaging experiments are challenging and can take up to several months to optimise and may not warrant successful outcome.

*- I think the discussion would benefit from the addition of a graphical cartoon model illustrating the role of S100A11 in controlling plasma membrane dynamics in mitosis and spindle orientation. *

__Authors response: __We thank the Reviewer for this suggestion. We have now added a graphical cartoon (new Figure 7), summarising the role of S100A11-mediated regulation of plasma membrane dynamics in polarised cell division orientation, progression and outcome. We hope this new illustration clarifies further the mechanisms described in this study.

*- Finally, to understand the relevance of S100A11 in the context of 3D polarised mammary epithelia, it would be very interesting to analyse the effect of S100A11 knock-downn in mouse mammary epithelial acini grown in matrigel. This is not essential for the proposed studies, but would add biological relevance to the mechanisms characterised in 2D colture. *

__Authors response: __We agree with the Reviewer that validating our findings in 3D cultures of mammary epithelial cells will be important to determine the influence of S100A11-mediated regulation of plasma membrane dynamics during mitosis on lumen formation and tissue morphogenesis. This is exactly the direction where the follow-up of these findings will go. While the first author who led this work has graduated and left our lab, we have recently recruited a new PhD student to address this important question, which will need a few years of investigation to provide important insights, similarly to what we did in our previous work (Fankhaenel et al., 2023 Nat Commun).

*Minor comments: *

*- It would be preferable to mention the known functions of S100A11 in the introduction rather than at the beginning of the paragraph at pg. 9. *

__Authors response: __In response to the Reviewer’s suggestion, we have now moved the paragraph describing known functions of S100A11 to the introduction of the revised manuscript (see page 5).

*- at pg 10, beginning of paragraph, I find it a weird phrasing that "LGN interacts with F-actin". As reported in the reference cited here, this is through Afadin, which binds simultaneously LGN and cortical F-actin. I would rephrase it. *

__Authors response: __We thank the Reviewer for clarifying this point, which we have now rectified in the revised manuscript (see page 11).

__Reviewer #1 (Significance (Required)): __

*The description of cell adhesion as key factor instructing correct mitotic progression and execution of oriented division of vertebrate epithelial cells by controlling plasma membrane dynamics is novel and interesting for scientist in the spindle orientation/polarity field. The experiments are well-designed and perfectly executed and presented. I am in favour of publication of the manuscript, providing that a few points are addressed. *

Authors response: We thank the Reviewer for their very positive evaluation of our work.

__Reviewer #2 (Evidence, reproducibility and clarity (Required)): __

*Establishment and maintenance of cell polarity are fundamental processes for physiology in multi-cellular organism given the fact that more than 380 million epithelial cell renewal for every second in human adults. However, the precise mechanisms linking plasma membrane polarity and cortical cytoskeleton dynamics of epithelial cells during mitotic exit and interphase remain ill-illustrated. Salah Elias and her colleagues experimentally manipulated the density of mammary epithelial cells in culture, which led to several mitotic defects. Specifically, they found that perturbation of cell-cell adhesion integrity impairs the dynamics of the plasma membrane during mitosis, affecting the shape and size of mitotic cells and resulting in defects in mitosis progression and generating daughter cells with aberrant cytoarchitecture. In these conditions, F-actin-astral microtubule crosstalk is impaired leading to mitotic spindle misassembly and misorientation, which in turn contributes to chromosome mis-segregation. Mechanistically, they identified the S100 Ca2+-binding protein A11 as a key membrane-associated regulator that forms a complex with E-cadherin and LGN to coordinate plasma membrane remodelling with E-cadherin-mediated cell adhesion and LGN-dependent mitotic spindle machinery. I felt that this is a strong manuscript for peer-review as it serves diversified interests in modern cell biology. *

Authors response: We thank the Reviewer for their overall very positive feedback on our manuscript.

__Reviewer #2 (Significance (Required)): __

Several key cellular experiments should be repeated using a second line of epithelial cells such as RPE1.

__Authors response: __We agree with the Reviewer it will be interesting to test our findings in other epithelial cells, including RPE1 cells, a widely used epithelial cell model to study the mechanisms controlling cell division. Nonetheless, we would like to emphasise that while our work demonstrates the importance of the interplay between plasma membrane dynamics and cell-cell adhesion for correct execution of polarised cell divisions in mammary epithelial cells, our aim is not to generalise the role of these S100A11-mediated mechanisms. An elegant study has shown that the mechanisms controlling plasma membrane remodelling and elongation during mitosis to ensure correct positioning of the mitotic spindle and symmetric division differ between HeLa cells and RPE1 cells (Kiyomitsu and Cheeseman, 2013 Cell). Additional experiments in a second cell line will require a thorough characterisation of the expression and localisation of S100A11 during the cell cycle, as well as the use of extensive and time-consuming knockdown and imaging experiments over several months and may lead to different observations requiring further mechanistic investigation, which is beyond the initial scope of this study. Additionally, the PhD student who led this study has graduated and left the lab and presently we don’t have capacity or resources to conduct these suggested experiments. Finally, to precisely address the Reviewer’s concern, we have now amended the revised manuscript to make our statements more specific to mammary epithelial cells throughout the text.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)): __

*Summary: your understanding of the study and its conclusions. *

*The scope of the study is to understand the links between cell-cell adhesion integrity, plasma membrane dynamics and mitotic spindle in mammalian epithelial tissues. To test this, the authors cultured mammary epithelial cells at optimal or low density as a way of perturbing cell-cell adhesion. The authors conclude that perturbing cell-cell adhesion alters plasma membrane dynamics, causing mitotic defects and that S100A11 coordinates this link via E-cadherin. Whilst this is an interesting manuscript, illustrating the differences of mitotic success in optimal density vs. low density cell cultures, I do not think that the conclusions are supported by the evidence presented for the reasons stated below. *

*Major comments: major issues affecting the conclusions. *

*- The manuscript clearly shows that culturing cells at a lower density results in a higher incidence of asymmetric division (figure 1) and mitosis defects (figure 2). Cells round more and faster and there is more actin at the cortex during rounding (figure 3). However, whilst differences in cell-cell adhesion are likely to play a role in mediating these effects, I don't think that it is possible to claim from the data presented that these defects are specifically due to cell-cell adhesion differences. This is because the morphology of cells at low density is also very different - cells appear more mesenchymal, with migratory front-rear polarity instead of apical-basal polarity. These cells will therefore have many differences between them, cell-adhesion being just one. The data is also not showing a 'loss' of cell-cell adhesion integrity but are rather illustrating the differences between cells that have formed cell-cell adhesions and those that have not. To really test the specific role of cell-cell adhesions, the authors would need to inhibit adhesions directly but without altering the cell density - for example via chemical or genetic perturbation within a confined environment. I suggest that the authors either need to do these experiments or to requalify what their data is telling us. *

__Authors response: __We thank the Reviewer for their insightful discussion of the proposed mechanisms described in our manuscript. Several of the Reviewer’s comments pinpoint and exactly match the messages that we would like to convey to the scientific community. Therefore, to address the Reviewer’s comments, we have carefully requalified our statements in several places in the revised manuscript, to ensure they are more clear and more precise.

We agree with the Reviewer’s comment that our experiments using sub-optimal density of mammary epithelial cells rather prevents the formation of cell-cell adhesions than disturbing them. The Reviewer is right, our experiments in low-density cultures suggest that perturbation of cell-cell adhesion formation impairs mammary epithelial identity, where cells lose their polarity and adopt a more mesenchymal phenotype, associated with plasma membrane remodelling defects. This affects the dynamics and progression of cell division. Nonetheless, our observations suggest an interplay between cell-cell adhesion and the plasma membrane to maintains correct cell shape during mitosis. To test this hypothesis, we explored the function of S100A11 which we have identified in the LGN interactome in mitotic mammary epithelial cells (Fankhaenel et al., 2023 Nat Commun), and which has been shown to interact with E-cadherin at adherens junctions in MDCK cells (Guo et al., 2014 Sci Signal). This, together with the fact that S100A11 controls plasma membrane repair (Jaiswal et al., 2014 Nat Commun), suggested S100A11 as an interesting candidate to investigate the interplay between cell-cell adhesion and membrane remodelling during mitosis. The data presented here suggest that we were right and the perturbation of our membrane-bound target, S100A11, indeed leads to the same mitotic phenotypes. S100A11 RNAi-mediated knockdown (48h) affects E-cadherin localisation at the plasma membrane and impairs cell-cell adhesion formation with effects on plasma membrane dynamics that phenocopy the defects observed in our low-density culture experiments. Remarkably, perturbation of cell-cell adhesions persisted in cell treated with si-S100A11 for 72h (see Figure S3). Of note, all our siRNA experiments have been carried out in cells cultured at optimal density to establish cell-cell adhesions. Thus, S100A11 knockdown allows genetic perturbation of E-cadherin-mediated cell-cell adhesion and recapitulates the plasma membrane and mitotic defects observed in sub-optimal cultures of mammary epithelial cells. Future experiments will be key to dissect these S100A11-mediated mechanisms to further understand how plasma membrane remodelling and cell-cell adhesions are coordinated during mitosis. Finally, as suggested by the Reviewer, we have now requalified our conclusions as appropriate in the revised manuscript.

*- The current manuscript also demonstrates that cell adhesion is affected when S100A11 is knocked down (figure 4). It shows binding between and colocalization of S100A11 and E-cadherin, and shows that LGN cortical distribution is affected when S100A11 is knocked down (Figure 5). The results presented are suggestive of S100A11 being upstream of E-cadherin. However, I don't understand how the data shows "crosstalk between the plasma membrane, cell-cell adhesion, and the cell cortex during mitosis". For example, on P9: "We observed unequal distribution of CellMaskTM in a vast majority of S100A11-depleted cells (si-S100A11#1: ~79% versus si-Control: ~26%), indicating defects in plasma membrane remodelling (Figures 4B and 4C)." I don't agree that this demonstrates a defect in PM remodelling. Rather the cells in the representative images are less adherent and have adopted a more migratory cell state similar to that seen in figure 1 when seeded at low density. The fluidity of the much larger cells shown in knock down cells in panel F also appears higher, again suggesting an adhesion defect. *

• *

__Authors response: __The Reviewer has raised very important points here, which we would like to clarify.

We agree with the Reviewer that our results in S100A11-depleted cells indicate impaired cell adhesions which generates cells displaying an invasive/migratory behaviour. However, our analysis of S100A11-depleted mitotic cells labelled with CellMaskTM reveals abnormal plasma membrane elongation generating two daughter cells displaying defective geometry as compared to control cells. These defects in the plasma membrane and cell shape were not noticeable upon E-cadherin knockdown (see previous Figures 5K and 5L; now new Figures 6K and 6L). Thus, our results strongly suggest that S100A11 acts as an upstream cue that coordinates plasma membrane dynamics with E-cadherin-mediated cell adhesions, and that additional mechanisms may be regulated by S100A11 to coordinate cell-cell adhesion with plasma membrane remodelling. How S100A11 ensures such a dynamic interplay between the plasma membrane and E-cadherin during mitosis remains a key question that we have not fully addressed in this initial study. An attractive mechanism could be mediated by the function of S100A11 in regulating the dynamic interaction between F-actin and the plasma membrane, as previously reported (Jaiswal et al., 2014 Nat Commun). Increasing evidence shows the importance of the crosstalk between the plasma membrane, the cortex and cell shape for correct execution of mitosis (Rizzelli et al., 2020 Open Biol). In our experiments, we show that impaired plasma membrane remodelling and cell shape are associated with defects in F-actin and astral microtubule organisation. Thus, our findings reinforce a model whereby S100A11 is a key membrane-associated protein that coordinates the crosstalk between the plasma membrane, cell-cell adhesion, and the cell cortex during mitosis. It will be key to characterise the interactome of S100A11 during mitosis to provide important mechanistic insights into this new role of S100A11; it is our intention to investigate this in the future.

To address the points raised by the Reviewer, we have changed and clarified the statements they highlighted above, in the revised manuscript (pages 10 and 11).

*- An earlier paper from the same lab this year identified Annexin A1 as directing mitotic spindle orientation via localising LGN at lateral cortex. During this earlier paper they also identified S100A11, which is a partner for Annexin A1. The authors could more clearly explain what S100A11 is in the current manuscript and how the current study builds on this earlier study. *

__Authors response: __We thank the Reviewer for highlighting our previous work characterising the interactome of LGN in mitotic mammary epithelial cells (Fankhaenel et al., 2023 Nat Comms), and identifying Annexin A1 (ANXA1) as a polarity cue regulating the localisation and function of the evolutionarily conserved mitotic spindle orientation LGN complex. We also showed that ANXA1 direct partner S100A11 co-purifies with LGN and that perturbation of the ANXA1-S100A11 complex impairs the localisation of the LGN complex at the cell cortex during mitosis. Thus, as rightly pointed out by the Reviewer, this work builds on our previous work discussed above, but also on previous studies establishing S100A11 as a key regulator of plasma membrane repair by regulating the dynamic interplay between F-actin and the plasma membrane (Jaiswal et al., 2014 Nat Commun), and studies showing that S100A11 interacts with E-cadherin at adherens junctions (Guo et al., 2014 Sci Signal). To address the Reviewer’s point (also raised by Reviewer 1), we have now included a paragraph in the introduction (page 5) and results (page 10) of the revised manuscript describing these and other functions of S100A11 to provide a strong rational to our decision to investigate this protein.

---

*- Based on the data presented, I suggest that the authors should requalify their data. I suggest that the conclusions that can be drawn from the data are that cellular state is important for regulating mitosis orientation and fidelity (i.e. adherent epithelia cells vs. less adherent more migratory cells). S100A11 is important for promoting cell-cell adhesions and might be upstream of the known role of E-cadherin in regulating spindle orientation. Whilst I suggest that more quantified experiments would need to be included in order to assess possible effects on plasma membrane remodelling, the manuscript could be generally improved by a clearer explanation of the open question that they are addressing and what specific advance this manuscript has made in relation to the current literature, including their own. I do not currently feel that the title of the manuscript is appropriate since I don't think that a crosstalk between the plasma membrane and cell-cell adhesion has been shown here. *

__Authors response: __We would like to reiterate our agreement with the Reviewer’s suggestion about the conclusions drawn from our data. In the initial submission we proposed that perturbation of S100A11-mediated regulation of cell adhesion and plasma membrane impairs the identity of mammary epithelial cells, which affects their shape during mitosis leading to aberrant mitotic progression and outcome. While we have not checked the migratory behaviour of cells not forming cell-cell adhesions, we suggested that the cells adopted a mesenchymal phenotype. Furthermore, we discussed the implication of epithelial-to-mesenchymal transition on chromosome segregation fidelity and execution of mitosis, and how precisely they link with our study (see initial submission’s pages 14 and 19). As suggested by the Reviewer, we have now clarified further these observations in the results (pages 7 and 11) and discussion (pages 15 and 19) of the revised manuscript.

We have quantified several aspects of the changes in plasma membrane dynamics and remodelling throughout, in the initial manuscript (Figure 1D-H; Figure 4C). To address the Reviewer’s point, we have now added quantifications of membrane blebbing (new Figure 1I).

We would like to emphasise that the introduction of the initial manuscript has included the open questions that led to this study. These questions have been addressed further in the discussion, where we have also formulated new hypotheses and discussed what we think are the important outstanding questions for future investigations, in light of our findings. In this study we demonstrate that maintaining epithelial identity is essential for correct execution of polarised cell divisions. Our findings indicate that mammary epithelial cells grown at sub-optimal density lose their epithelial identity, which results in several mitotic defects. We propose a novel mechanism in which S100A11 acts as a molecular sensor of external cues coordinating the interplay between plasma membrane dynamics and cell-cell adhesion to maintain epithelial identity and integrity, thereby ensuring correct progression, orientation, and outcome of cell division. As suggested by the Reviewer, we have clarified further the advances made in this study, in the revised Results and Discussion sections.

To address the Reviewer’s final point, we would like to suggest the following revised title “Interplay between the plasma membrane and cell-cell adhesion maintains epithelial identity for correct polarised cell divisions”, which we hope reflects better the results described in our studies.

*Minor comments: important issues that can confidently be addressed. *

- P3: I wouldn't describe the junctional proteins listed as polarity proteins.

__Authors response: __We have now made this rectification in page 3, as suggested by the Reviewer.

*- Figure 1 - can the membrane blebbing phenotype by quantified? At the moment this part is observational so can't really be used to determine the role of plasma membrane remodelling. *

• *

__Authors response: __We have now included quantifications of blebbing in the revised manuscript, as suggested by the Reviewer (new Figure 1I).

*- Figure 3. I'm not sure what the 'subcortical actin cloud' measurement is. Figure 3G suggests it may be the distance from the cortex to the spindle pole but how does this relate to actin? *

__Authors response: __The Reviewer is right, the subcortical actin cloud includes a pool of dynamic subcortical actin that extends from the cortex (excluding the stiff cortical actin) to the cytoplasm, interacting with the centrosomes and concentrating near the retraction fibres. The subcortical actin cloud has been shown to mediate cortical forces and to concentrate force-generating proteins at the retraction fibres acting on centrosome dynamics and pulling on astral microtubules to orient the mitotic spindle (for example, please see Kwon et al., 2015 Dev Cell). We have now included this clarification in the revised manuscript (page 10).

*- Figure 4A. I can't see GFP-S100A11 accumulating at the cell surface. To me these images suggest that it is relatively ubiquitously expressed throughout the cytoplasm and surface, which is different to the later antibody stains, that show localisation at the cell surface. *

__Authors response: __A similar point has been raised by Reviewer 1. Although our retroviral-mediated transduction allows to avoid excessive expression of GFP-S100A11, the ectopic S100A11 is expressed at higher levels as compared to its endogenous counterpart. Our live images show an accumulation of the protein at the cell surface, but relatively high levels are also visible in the cytoplasm (previous Figure 4A, new Figure S4A). By contrast immunolabelling for endogenous S100A11 shows an obvious accumulation of the protein at the plasma membrane. This difference could also be due to a dynamic behaviour of the protein translocation of GFP-S100A11 between the cell surface and cytoplasm that is captured in our live imaging. Similar slight differences between immunofluorescence and live imaging of cortical proteins involved in mitosis, such as Dynein, NuMA, LGN and CAPZB, have been reported in several studies (to cite a few: di Pietro et al., 2017 Curr Biol; Elias et al., 2014 Stem Cell Rep; Fankhaenel et al., 2023). To address this point, we have now moved the panel showing S100A11 immunofluorescence in Figure S3A to new Figure 4A (also see response to Reviewer 1 Major Point 1).

*- Fig 4H doesn't show an active process of translocation of E-Cadherin to the cytoplasm. It shows representative images with slightly higher levels of E-Cadherin in the cytoplasm. This could be due to translocation or it could be to do with lack of E-Cadherin assembly. *

• *

__Authors response: __We thank the Reviewer for pointing this out. We have rectified this statement accordingly (page 11).

*- 4I I don't understand where the line profile is derived from - where is apical and where is basal in the images? Could a diagram be included? *

__Authors response: __We have now included an illustration of this quantification, in the revised manuscript (new Figure 4J).

- The discussion could be shortened and more clearly written - perhaps with subheadings of the main findings.

__Authors response: __We have clarified several ideas and statements, based on the specific points addressed above. While it is challenging to reduce the size of this section, given that the study addresses several mechanisms of mitosis, we have now shortened the discussion in the revised manuscript.

*- Methods: Why is cholera toxin used in the cell culture medium? *

• *

__Authors response: __Cholera toxin is a key component of MCF-10A medium, which has been shown to stimulate cAMP activation promoting cell proliferation in culture. This culture protocol is a gold standard in the field (Debnath et al., 2023 Methods). Given that cholera toxin is a highly regulated chemical and takes several months to purchase, we have tried culturing MCF-10A without the toxin, but this negatively affected proliferation and passage of this cells. Therefore, we concluded that adding it to the culture medium is important.

__Reviewer #3 (Significance (Required)): __

*In general, this is an interesting paper about the fidelity of mitosis in cells in adherent monolayers vs. in more migratory, non-adherent states. There is existing literature on this topic (some cited in the manuscript, alongside reviews of the topic). *

• *

*The main conceptual advance, as far as I can see, is that S100A11 is important for promoting cell-cell adhesions and might be upstream of the known role of E-cadherin in regulating spindle orientation via LGN. The main limitation is that plating cells at different densities is not a direct 'perturbation' of cell-cell adhesion. This means that the phenotypes seen could be due to many factors, not just cell adhesion. Assessment of plasma membrane and cytoskeletal dynamics are also often observational and not conclusive. *

• *

*The manuscript would be of interest to basic researchers working on epithelial development. Also potentially to basic researchers working on cancer, due to the mitotic errors described. *

• *

*I have expertise in epithelial cell biology. *

I estimate the authors would need between 3 and 6 months for revisions if they decide to do further experiments and between 1 and 3 months if they decide to re-qualify their claims.

• *

__Authors response: __We thanks the Reviewer for their overall positive feedback on our work and its broader importance for researchers in epithelial development and cancer biology.

We would like to reiterate our agreement with the Reviewer’s assessment of the conceptual advances of our work. We show that S100A11 complexes with E-cadherin and LGN during mitosis to control cell-cell adhesion assembly and the mitotic spindle machinery, respectively, which in turn ensures faithful chromosome segregation. Our results also suggest that S100A11 lies upstream of E-cadherin in the regulation of the LGN-mediated mitotic spindle machinery. We also agree with the Reviewer that plating epithelial cells at low density does not directly affect cell-cell adhesion, because, in these culture conditions, cells are not dense enough to establish cell-cell contacts necessary to assemble stable adherens junctions. Rather, and as rightly pointed out by the Reviewer, cells grown at low density fail to maintain their epithelial identity and adopt a more mesenchymal and elongated behaviour, which is accompanied by dramatic changes in plasma membrane remodelling throughout mitosis. Interestingly, our results show that both S100A11 and E-cadherin do not localise at the plasma membrane in these sub-optimal culture conditions. This along with our results showing that depletion of S100A11 phenocopies the effect of low-density culture conditions on plasma membrane remodelling and E-cadherin mediated cell-cell adhesion assembly, allow us to propose a mechanism whereby the membrane-associated S100A11 protein acts as a molecular sensor of external cues bridging plasma membrane remodelling to E-cadherin-dependent cell adhesion to coordinate correct progression and outcome of mammary epithelial cell divisions.

We are grateful for the Reviewer’s insightful discussion of our findings. As we discussed above in our responses to their specific points, we have requalified many of our statements to clarify further our main findings and conclusions throughout the revised manuscript. We have also added new quantifications in response to the Reviewer’s suggestions. We believe, that together, these revisions have advanced further the initial manuscript.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary: your understanding of the study and its conclusions.

The scope of the study is to understand the links between cell-cell adhesion integrity, plasma membrane dynamics and mitotic spindle in mammalian epithelial tissues. To test this, the authors cultured mammary epithelial cells at optimal or low density as a way of perturbing cell-cell adhesion. The authors conclude that perturbing cell-cell adhesion alters plasma membrane dynamics, causing mitotic defects and that S100A11 coordinates this link via E-cadherin. Whilst this is an interesting manuscript, illustrating the differences of mitotic success in optimal density vs. low density cell cultures, I do not think that the conclusions are supported by the evidence presented for the reasons stated below.

Major comments: major issues affecting the conclusions.

The manuscript clearly shows that culturing cells at a lower density results in a higher incidence of asymmetric division (figure 1) and mitosis defects (figure 2). Cells round more and faster and there is more actin at the cortex during rounding (figure 3). However, whilst differences in cell-cell adhesion are likely to play a role in mediating these effects, I don't think that it is possible to claim from the data presented that these defects are specifically due to cell-cell adhesion differences. This is because the morphology of cells at low density is also very different - cells appear more mesenchymal, with migratory front-rear polarity instead of apical-basal polarity. These cells will therefore have many differences between them, cell-adhesion being just one. The data is also not showing a 'loss' of cell-cell adhesion integrity but are rather illustrating the differences between cells that have formed cell-cell adhesions and those that have not. To really test the specific role of cell-cell adhesions, the authors would need to inhibit adhesions directly but without altering the cell density - for example via chemical or genetic perturbation within a confined environment. I suggest that the authors either need to do these experiments or to requalify what their data is telling us. The current manuscript also demonstrates that cell adhesion is affected when S100A11 is knocked down (figure 4). It shows binding between and colocalization of S100A11 and E-cadherin, and shows that LGN cortical distribution is affected when S100A11 is knocked down (Figure 5). The results presented are suggestive of S100A11 being upstream of E-cadherin. However, I don't understand how the data shows "crosstalk between the plasma membrane, cell-cell adhesion, and the cell cortex during mitosis". For example, on P9: "We observed unequal distribution of CellMaskTM in a vast majority of S100A11-depleted cells (si-S100A11#1: ~79% versus si-Control: ~26%), indicating defects in plasma membrane remodelling (Figures 4B and 4C)." I don't agree that this demonstrates a defect in PM remodelling. Rather the cells in the representative images are less adherent and have adopted a more migratory cell state similar to that seen in figure 1 when seeded at low density. The fluidity of the much larger cells shown in knock down cells in panel F also appears higher, again suggesting an adhesion defect. An earlier paper from the same lab this year identified Annexin A1 as directing mitotic spindle orientation via localising LGN at lateral cortex. During this earlier paper they also identified S100A11, which is a partner for Annexin A1. The authors could more clearly explain what S100A11 is in the current manuscript and how the current study builds on this earlier study.

Based on the data presented, I suggest that the authors should requalify their data. I suggest that the conclusions that can be drawn from the data are that cellular state is important for regulating mitosis orientation and fidelity (i.e. adherent epithelia cells vs. less adherent more migratory cells). S100A11 is important for promoting cell-cell adhesions and might be upstream of the known role of E-cadherin in regulating spindle orientation. Whilst I suggest that more quantified experiments would need to be included in order to assess possible effects on plasma membrane remodelling, the manuscript could be generally improved by a clearer explanation of the open question that they are addressing and what specific advance this manuscript has made in relation to the current literature, including their own. I do not currently feel that the title of the manuscript is appropriate since I don't think that a crosstalk between the plasma membrane and cell-cell adhesion has been shown here.

Minor comments: important issues that can confidently be addressed.

P3: I wouldn't describe the junctional proteins listed as polarity proteins. Figure 1 - can the membrane blebbing phenotype by quantified? At the moment this part is observational so can't really be used to determine the role of plasma membrane remodelling.

Figure 3. I'm not sure what the 'subcortical actin cloud' measurement is. Figure 3G suggests it may be the distance from the cortex to the spindle pole but how does this relate to actin?

Figure 4A. I can't see GFP-S100A11 accumulating at the cell surface. To me these images suggest that it is relatively ubiquitously expressed throughout the cytoplasm and surface, which is different to the later antibody stains, that show localisation at the cell surface.

Fig 4H doesn't show an active process of translocation of E-Cadherin to the cytoplasm. It shows representative images with slightly higher levels of E-Cadherin in the cytoplasm. This could be due to translocation or it could be to do with lack of E-Cadherin assembly.

4I I don't understand where the line profile is derived from - where is apical and where is basal in the images? Could a diagram be included?

The discussion could be shortened and more clearly written - perhaps with subheadings of the main findings.

Methods: Why is cholera toxin used in the cell culture medium?

Significance

In general, this is an interesting paper about the fidelity of mitosis in cells in adherent monolayers vs. in more migratory, non-adherent states. There is existing literature on this topic (some cited in the manuscript, alongside reviews of the topic).

The main conceptual advance, as far as I can see, is that S100A11 is important for promoting cell-cell adhesions and might be upstream of the known role of E-cadherin in regulating spindle orientation via LGN. The main limitation is that plating cells at different densities is not a direct 'perturbation' of cell-cell adhesion. This means that the phenotypes seen could be due to many factors, not just cell adhesion. Assessment of plasma membrane and cytoskeletal dynamics are also often observational and not conclusive.

The manuscript would be of interest to basic researchers working on epithelial development. Also potentially to basic researchers working on cancer, due to the mitotic errors described.

I have expertise in epithelial cell biology.

I estimate the authors would need between 3 and 6 months for revisions if they decide to do further experiments and between 1 and 3 months if they decide to re-qualify their claims.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Establishment and maintenance of cell polarity are fundamental processes for physiology in multi-cellular organism given the fact that more than 380 million epithelial cell renewal for every second in human adults. However, the precise mechanisms linking plasma membrane polarity and cortical cytoskeleton dynamics of epithelial cells during mitotic exit and interphase remain ill-illustrated. Salah Elias and her colleagues experimentally manipulated the density of mammary epithelial cells in culture, which led to several mitotic defects. Specifically, they found that perturbation of cell-cell adhesion integrity impairs the dynamics of the plasma membrane during mitosis, affecting the shape and size of mitotic cells and resulting in defects in mitosis progression and generating daughter cells with aberrant cytoarchitecture. In these conditions, F-actin-astral microtubule crosstalk is impaired leading to mitotic spindle misassembly and misorientation, which in turn contributes to chromosome mis-segregation. Mechanistically, they identified the S100 Ca2+-binding protein A11 as a key membrane-associated regulator that forms a complex with E-cadherin and LGN to coordinate plasma membrane remodelling with E-cadherin-mediated cell adhesion and LGN-dependent mitotic spindle machinery. I felt that this is a strong manuscript for peer-review as it serves diversified interests in modern cell biology.

Significance

Several key cellular experiments should be repeated using a second line of epithelial cells such as RPE1.

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 "Crosstalk between the plasma membrane and cell-cell adhesion maintains epithelial identity for correct polarised cell divisions" by Dr. Hosawi and colleagues reports the characterisation of the mitotic connection between plasma membrane dynamics and division orientation in polarised mammalian epithelial cells in culture. The authors start from the comparison of mitotic events of human mammary MCF10A cells grown at optimal density or at low density. They observed that only at optimal density MCF10A cells polarise by E-cadherin mediated cell-cell contacts, and display uniform membrane enrichment at the cortex, whereas cells grown at low density do not show cortical E-Cadherin enrichment, and distribute aberrantly the plasma membrane at one side and in cytoplasmic vesicles, generating daughter cells with unequal size. Consistently, further analyses revealed that low-density MCF10A cells undergo misoriented mitosis, with chromosome congression and misegregetion defects. Mechanistically, low density MCF10A cells fail to organise a symmetric mitotic spindle and center it in metaphase. This is due to an increased cortical actomyosin thickness coupled to abnormal astral microtubule stability. Building on previous data from the Elias lab, the authors uncover a role of the membrane-associated S100A11 protein in maintaining correct plasma membrane dynamics and E-cadherin localisation in mitosis. Further dissection of the molecular mechanism underlying this mitotic function od S10011A revealed that it enriches at the cortex only in optimal-density MCF10A cells, and promotes spindle orientation by association with LGN and E-cadherin, upstream of E-cadherin. This evidence depicts the plasma membrane and S100A11 proteins as a key mechanical sensors of cell-cell adhesion orchestrating the recruitment of E-cadherin and LGN-dependent force generators to ensure correct division orientation.

Major points:

• Important information is presented in Supplementary Figure S3. I suggest to move these panels in the main figures. Specifically, I would replace figure 4A with S3A showing the distribution of endogenous S100A11 in MCF10A cells, rather than the one of the GFP-tagged version which is over-expressed.
• The mechanisms of division orientation governed by S100A11 seems to impinge on the control of cortical F-actin and astral microtubule dynamics. This is illustrated in figure S3C, which in my opinion should be shown in the main figures with some more explanation / experiments. The authors mention the " tight actin F-actin bundles at the cell-cell contacts" that are lost in S100A11-depleted cells, and that interact with astral microtubules. However this is not fully clear in figure S3C. I think the authors should find a way to present better these evidence which is key in supporting their molecular model.
• I think the discussion would benefit from the addition of a graphical cartoon model illustrating the role of S100A11 in controlling plasma membrane dynamics in mitosis and spindle orientation.
• Finally, to understand the relevance of S100A11 in the context of 3D polarised mammary epithelia, it would be very interesting to analyse the effect of S100A11 knock-downn in mouse mammary epithelial acini grown in matrigel. This is not essential for the proposed studies, but would add biological relevance to the mechanisms characterised in 2D colture.

Minor comments:

• It would be preferable to mention the known functions of S100A11 in the introduction rather than at the beginning of the paragraph at pg. 9.
• at pg 10, beginning of paragraph, I find it a weird phrasing that "LGN interacts with F-actin". As reported in the reference cited here, this is through Afadin, which binds simultaneously LGN and cortical F-actin. I would rephrase it.

Significance

The description of cell adhesion as key factor instructing correct mitotic progression and execution of oriented division of vertebrate epithelial cells by controlling plasma membrane dynamics is novel and interesting for scientist in the spindle orientation/polarity field. The experiments are well-designed and perfectly executed and presented. I am in favour of publication of the manuscript, providing that a few points are addressed.

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

Learn more at Review Commons

---

Reply to the reviewers

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

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This MS contains carefully carried out and well controlled experiments describing a new pFFAT in ELYS. There is a similarly convincing demonstration of functionally relevant colocalisation by proximity ligation assay (PLA), particularly that both ELYS and VAP are nuclear envelope proteins in interphase without interacting (neg control in Fig 4D).

Major Issue: Functional significance

A key conclusion is that experiments prove that "ELYS serves as the crucial initiation factor for post-mitotic NPC-assembly" (p5). However, evidence for this is lacking as this would require reconstitution of NPC assembly with a mutant form of ELYS carefully changing the FFAT motif (e.g. 1321A 1324E) and exclusion of other probable VAP targets in experiments with mutant VAP. VAPs are among the proteins with the highest number of documented interactors (see Huttlin 2015/7 etc, e.g. PMID 26186194), so knocking down VAP may have pleiotropic effects and quite indirect read-outs in many aspects of cell function. In addition, for this work specifically there are other NE proteins that are known interactors of VAP: Emerin (EMD) and LBR both interact with VAP (high-throughput data, VAPA and VAPB). EMD has a motif similar to the canonical phospho-FFAT: 98 SYFTTRT 104. LBR has no motif. These findings should not be overlooked in this work. For example, was the interaction with emerin (page 4) sensitive to mutating VAP or ELYS? Could the effect seen in Figure 5 result from interactions with proteins other them ELYS?

Further experiments should be carried out to justify all statements in the current MS of functional significance. Instead of doing more experiments, an alternative for the authors would be to describe the current set of results more cautiously. However, that would require changing much of the impact of the current MS, from the title onwards.

Moderate Issue: VAPA

From the start of the Introduction and some elements of the Discussion, include VAPA in equal measure with VAPB. When describing interactions of ELYS with VAP note that Huttlin et al., reported interactions twice for each of VAPA and VAPB. When describing own results (James et al. 2019) and those of others (Saiz-Ros et al., 2019) that focused on VAPB, clarify if the authors' view is that VAPA would (or would not) have the same interaction.

Is there any evidence that only VAPB is on NE? Note that some refs in the Introduction relate to VAPA: Mesmin (not VAPB); ACBD5: although article titles refer to VAPB, early work (10.1083/jcb.201607055) showed almost identical involvement of VAPA. Also, this redundancy likely explains "function of VAPB in mitosis is not essential," (in Discussion). The lack of effect of VAPA knock-down may indicate that in these cells VAPB is dominant, but does not exclude a role for VAPA when VAPB is reduced. That might be tested by depleting both. Even following that, there is MOSPD2 to consider

Other aspects of the writing

"two amino acid residues are crucial for the interaction (VAPB K87 and M89)." This is wrong. Many residues are critical, these are merely 2 of possibly >10 that were chosen by Kaiser et al (2005) to create their non-binder.. Others have used different mutations to block FFAT binding.

"They may exhibit a certain binding preference to specific members of the VAP ... family...". I cannot think of any example. I note no citation is given.

When listing many or all MSP proteins, the text should state that MOSPD2 is uniquely close to VAPA/B. CFAP65 is typically not mentioned in the VAP-like lists as it does not have any of the conserved sequence that binds FFAT. If however the authors wish to include all human MSP domain protein, they should also include Hydin.

Slightly wrong to cite De Vos et al., 2012 about PTPIP51's FFAT as that paper makes no mention of the motif. Better pick Di Mattia (again)

On VAPB (and also A) on INM: there are references to be cited esp. relating to intranuclear Scs2 in yeast (Brickner et al 2004, Ptak et al 2021)

Citations for VAP at ER-mito contacts "De Vos et al., 2012; Gómez-Suaga et al., 2019; Stoica et al., 2014)". These all refer to the same bridging protein, PTPIP51. Reduce to one citation. Then mention other proteins at the same site VPS13A, mitoguardin(MIGA)-2 ...

"The domain interacts with characteristic peptide sequences ..." add citation to this sentence

"Several variants of such motifs have been described: (i)" ... "(ii)": (i) and (ii) are entirely unlinked. Delete these and also "Several variants of such motifs have been described." Which is repeated later

"FFAT-like motifs come in different flavors and may even lack the two phenylalanine residues (Murphy and Levine, 2016)": while motifs can tolerate variation at both positions, this text is misleading as it implies much more variation than is known. The 1st F can only be conservatively substituted (Y).

Minor aspects in Results:

ORP1L peptide as positive control: cite Kaiser 2005

Was phosphoproteomics done in such a way as to find peptides that have both S1314 and S1326?

Figure 4D, row 2: Comment on intranuclear staining in Prophase (at approx 4 o'clock) of both ELYS & VAP that is PLA positive

Referees cross-commenting

I agree with this point from Reviewer #1. We all agree that the main issue can be resolved experimentally to determine the effect of subtle point mutations in ELYS. Both other reviewers have done a good job in finding issues with the experiments that can also be addressed.

Significance

This work documents an interaction between the protein ELYS, that is involved in the reformation of nuclear pore complexes after mitosis, and the ER membrane protein VAPB. The interactions was previously known through high-throughput studies, along with many 100's of others for VAP, but here it is studied in detail and with care, identifying how the motif is induced by phosphorylation of ELYS. The two proteins are co-localised using convincing proximity ligation assays. This biochemistry and cell biological localisation is well done.

Functional experiments then show that VAP (in this case VAPB) knock-down affects mitosis and chromosome segregation. While the result is incontrovertible, it has many possible interpretations, mainly because VAP has hundreds of interactions, including with multiple proteins involved in mitosis beyond just ELYS. This means that there are major limitations on how the interaction and co-localisation should be interpreted, reducing the advance associated with the current manuscript to incremental, and the limiting the audience to specialized.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this study, James et al. follow up on their prior discovery that the ER contact site protein VAPB localizes to the nuclear envelope and is a putative binding partner of the nucleoporin ELYS, which coordinates nuclear envelope reformation (NER) with nuclear pore complex (NPC) biogenesis at mitotic exit and is also a constituent of the nuclear facing Y-complexes of mature NPCs. Using a series of complementary biochemical approaches the authors 1) demonstrate that VAPB and ELYS directly interact, 2) map the binding sites on ELYS that are sufficient to bind VAPB, 3) show that mutations that disrupt VAPB-FFAT motif binding also abrogate binding to ELYS including of the full-length protein, 4) define mitotic phosphorylation sties on VAPB-bound ELYS, 5) demonstrate that phosphorylation of ELYS, specifically at the FFAT2 motif, is required for binding to VAPB, and 6) demonstrate that the phosphorylation of ELYS that regulates VAPB binding occurs in mitosis. Turning to cell biology, the authors find that VAPB, which is an established ER protein, has some preference for non-core regions during NER (like ELYS). In addition, PLA analysis suggests that the interaction of VAPB and ELYS is most robust during anaphase and is somewhat disrupted when the binding of VAPB to FFAT motifs is lost due to targeted mutation. Last, the authors demonstrate that depletion of VAPB leads to metaphase delay and lagging chromosomes.

Major comments:

The data supporting direct binding of peptides encoding the FFAT 1 and FFAT2 motif derived from ELYS to VAPB in a manner similar to other FFAT sequences is strong, as is the effect of phosphorylation of FFAT1 on the strength of this interaction.

The evidence supporting the mitotic-specific nature of the ELYS-VAPB interaction is strong, and that this interaction is direct, is also strong, and was rigorously tested using a combination of endogenous expression, heterologous expression, and recombinant protein approaches. Moreover, the sensitivity of this interaction to established mutations in VAPB abrogating FFAT interactions reinforces the outlined underlying biochemical interaction mechanism. The essentiality of ELYS phosphorylation (and therefore the mechanism underlying the mitotic specificity of the interaction) is also strongly supported by the data using phosphatase treatment. Although it is an intuitive model, whether the cell biological evidence support the simplest view that the ELYS-VAPB complex bridges the nuclear envelope to chromatin during NER in late anaphase / at mitotic exit is far less solid and, at a minimum, alternative models should be considered/discussed. For example, how a delay in metaphase in the siVAPB condition is consistent with a role in NER, which occurs exclusively post-metaphase, is unclear. Is it not possible that the VAPB-ELYS complex is regulated by phosphorylation during mitotic progression such that VAPB and/or ELYS can only exert their biological effects when released from the complex? In other words, might ELYS be licensed to act in NER only when it is released from VAPB, which could prevent premature NER/NPC biogenesis? Subtleties of when during mitosis the phosphorylation occurs is challenging, and it could be that the anaphase A to anaphase B transition, when many mitotic entry phosphorylation events begin to be reversed, could be relevant here. Along these lines, in Fig. 5 how the VAPB knock-down does or does not recapitulate the phenotype of an ELYS knock-down in this cell type (and the effect of the combination, to address epistasis) is needed for context, as is whether VAPB knock-down affects ELYS distribution in mitosis. ELYS knock-down would also be very beneficial for the PLA analysis to establish the "floor" of measurable signal. Last, it is also possible that VAPB has other roles in mitosis that should be acknowledged - for example although it is in yeast, it is relevant that a VAPB orthologue Scs2 is required for normal nuclear envelope expansion in mitosis by regulating SUMOylation (Ptak, Saik et al., JCB, 2021 and Saik et al., JCB, 2023) - this work should be referenced as well. Of course, the ideal experiment would be one in which an ELYS knock-down is complemented with a resistant form that encodes the S to A mutations in the FFAT2 region to assess its localization and to see if it can complement the knock-out function of ELYS in post-mitotic NPC assembly or, as suggested by a sequestration model, it can drive the same metaphase delay seen upon VAPB knock-down. This is technically challenging for sure, particularly given the size of the ELYS gene, but it would address the cell biological function of this interaction in the most direct manner. Several other observations that could warrant further comment or study include 1) is there a VAPB signal at the metaphase poles as suggested by Fig. 4A and, if so, could this represent aa distinct mitotic function?; 2) Does the HA-VAPB KD/MD mutant localize differently in mitosis compared to the WT - it appears that it might be less enriched in non-core regions (Fig. 4E)?; 3) does VAPB alter post-mitotic NPC biogenesis/number?

Minor comments:

I would suggest avoiding the use of "novel" when describing newly assembled NPCs or post-mitotic nuclear envelope reformation, as its other meaning of "non-standard" makes this wording confusing. It is unclear whether when the authors state that ELYS localizes "to the nuclear side of the nuclear envelope" they are referring to the nuclear aspect of the NPC and/or a separate pool at the INM - please edit to clarify. More descriptive y-axes for the plots in Fig. 4F and 5F and related legends would be useful; although the details are in the methods section, it would be nice not to have to hunt them down. Also, please clarify the meaning of blue and orange points in Fig. 4F.

Significance

General assessment: The biochemical analysis is rigorous and compelling and establishes the mitotic-specific interaction of VAPB and ELYS including detailed information about the binding interface and its regulation by phosphorylation. The new insight provided into the function of this VAPB-ELYS interaction is somewhat less well developed as the current manuscript, in its current form, does not yet mechanistically define the function of the VAPB-ELYS interaction in mitosis.

Advance: Conceptually, to the best of my knowledge, the idea that VAPB contributes to mitosis in mammalian cells is novel and is therefore impactful and will motivate further work. As the authors connect VAPB biochemically to ELYS, an established factor that promotes the coordination of NER and NPC biogenesis, this interaction is likely to be mechanistically important, although the specific details by which this interaction facilitate normal mitotic progression is not yet clear.

Audience: This work will be of interest to a broad swath of cell biologists including those interested in NPCs, the nuclear envelope, the ER, membrane remodeling, and chromosome segregation.

My expertise is in nuclear envelope dynamics, nuclear pore complexes, and chromatin organization.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The VAP proteins are well established as tail anchored proteins of the ER membrane. VAPs mediates co-operation between the ER and other organelles by creating a transient molecular tether with binding partners on opposing organelles to form a membrane contact site over which lipids and metabolites are exchanged. Proteins which bind VAPs generally contain a short FFAT motif, of varying sequence which binds the MSP domain of VAP. More recently the FFAT motif has been more extensively analysed in multiple different proteins and differential phosphorylation of the FFAT motif has been shown to either enhance or block VAP binding depending on the position of the phosphosite.

Recent work conducted by the authors demonstrated that a small population of VAPB is not exclusively localised to the ER and can also reach the inner nuclear membrane. They also identified ELYS as a potential interaction partner of VAPB in a screening approach. ELYS is a nucleoporin that can be found at the nuclear side of the nuclear envelope where it forms part of nuclear pore complexes. During mitosis, ELYS serves as an assembly platform that bridges an interaction between decondensing chromosomes and recruited nucleoporin subcomplexes to generate new nuclear pore complexes for post-mitotic daughter cells. In this manuscript, James et al seek to explore this enigmatic potential interaction between ELYS and VAPB to address why VAPB may be found at the inner nuclear membrane.

Peptide binding assays and some co-immunoprecipitation experiments are used to demonstrate that interactions occur via the MSP-domain of VAPB and FFAT-like motifs within ELYS. In addition, it is demonstrated that, for the ELYS FFAT peptides, the interaction is dependent on the phosphorylation status of serine residues of a particular FFAT-motif that can either promote or reduce its affinity to VAPB. Of most relevance is a serine in the acidic tract (1314) which, when phosphorylated increases VAPB binding. This is completely in line with what is already known about the FFAT motif and so is not surprising, in particular when using a peptide in an in vitro assay.

The authors then utilise cell synchronisation techniques to provide evidence that both phosphorylation of ELYS and its binding to VAPB are heightened during mitosis. Immunofluorescence and proximity ligation assays are used to demonstrate that the proteins co-localise specifically during anaphase and at the non-core regions of segregating chromosomes.

The manuscript is concluded by investigating the effect of VAPB depletion on mitosis with some evidence to suggest that transition from meta-anaphase is delayed and defects such as lagging chromosomes are observed.

Major comments

Overall, this manuscript is well written and the data presented in Figures 1-3 convincingly show the nature of the interaction between ELYS and VAPB. Clearly the proteins interact via FFAT motifs and this interaction appears to be enhanced during mitosis. However, the work as is, relies heavily on peptide binding assays and would benefit from additional experiments to further support the results. The authors need to more clearly show that this specific phosphorylation happens during mitosis, they may have this data but it is not clearly explained. In addition, the data that VAPB-ELYS interaction contributes to temporal progression of mitosis (as per the title) is not sufficiently clear. VAPB silencing appears to have some impact on mitosis but this is not the same thing. So this section needs to be strengthened before this statement can be made.

The authors claim that the study "suggests an active role of VAPB in recruiting membrane fragments to chromatin and in the biogenesis of a novel nuclear envelope during mitosis". Given the data presented in Figures 4 and 5, this appears to be rather speculative with little evidence to support it, so data should be provided or this statement toned down. Currently, without additional supporting data the authors may wish to revise the overarching conclusions of the study and change the title.

Specific points.

Peptide pull down assays clearly show which FFAT-like motifs are important in facilitating binding. The co-immunoprecipitation systems used in Figure 2 also provide useful information on the interaction in a cell context. The authors should combine these findings by introducing full length ELYS mutants with altered FFAT-like motifs into their stably expressing GFP-VAPB HeLa cell line and then performing Co-IPs to help identify which FFAT motif/s drive the mitotic interaction. Other mutants of ELYS harbouring either phosphomimetic or phospho-resistant residues may also be introduced to further investigate mechanisms of the molecular switch in a cellular environment to support the work currently done with peptides alone. This is an obvious gap in the work which, based on the other data the authors have shown, should presumably be straightforward and would also lead directly into the next major point.

• Whilst silencing VAPB does appear to delay mitosis, no reference is made to ELYS throughout Figure 5 nor as part of its associated discussion. Given that VAPB has more than 250 proposed binding partners, the observed aberration of mitotic progression could result from a huge number of indirect processes. Further work is needed to link the experiment specifically to the VAPB-ELYS interaction and not just loss of VAPB. We would suggest generating a complementation system where ELYS is either knocked out or silenced and then wild-type ELYS and an ELYS FFAT mutant (which cannot interact with VAPB),and/or a phospho mutant (whose interaction cannot be regulated during mitosis) are introduced. Then the observed effects can be better attributed to the VAPB-ELYS interaction and not just loss of VAPB.
• The immunofluorescence and PLA results in Figure 4 could be strengthened by including other ER markers. This would show that co-localisation of ELYS at the non-core region is specific to VAPB protein, not any ER protein or rather than an artefact of the ER being pushed out of the organelle exclusion zone during mitosis and therefore 'bunching' at the periphery of the nuclear envelope. It would be worthwhile repeating these experiments with candidates such as VAPA, other ER membrane proteins or at least GFP-KDEL, to make this phenomenon more convincing. As part of this the authors should ideally generate a complemented ELYS KO (see point above) to avoid the residual activity attributed to endogenous background in the PLA Figure 4E.
• Authors should clarify if the phosphorylation events (in particular S1314) only occur or are increased during mitosis. This may be data they have from the MS experiment in Figure 3 or it could also be shown using a phospho-antibody (although this can be challenging if a suitable antibody cannot be made).
• The authors should clarify why they need to do these semi in-vitro assays with purified GST-VAPB-MSP on beads and then lysates added and not just a standard co-IP. If this is simply signal intensity due to a very small proportion of VAPB binding to ELYS then this is fine but this should be stated and it should be made clear that ELYS is not a major binding partner - most of VAPB is on the ER. Otherwise, this is misleading.

I estimate that the suggested alterations above would incur approximately 3-6 months of additional experimental work, depending on if KO cell lines were required.

Minor comments

• To show that the observed interactions and potential role of VAPB-ELYS interaction is universal it would be useful to have at least a subset of experiments also shown in another cell line or system - this is now also a requirement for some journals.
• Consider re-wording the title of the manuscript to better reflect the data presented within the study. Alternatively, provide further evidence that VAPB-ELYS interactions directly affect temporal progression of mitosis to validate this claim, as discussed above.
• Quantification of blots in Figure 2A could allow measurement of relative binding affinities between VAPB-ELYS throughout the cell cycle. The same could be applied to the effect of phosphorylation on binding affinity in Figure 2D.
• The cells used are never clearly mentioned in the text - I assume this is always in HeLa but this should be added in all cases for clarity
• Page 8: "As shown in Fig. 2A,a large proportion of GFP-VAPB was precipitated under our experimental conditions." - I don't understand how this is shown in this figure as the non-bound fraction is not shown?
• Please provide some controls to demonstrate the extent to which the samples used are asyn, G1/M or M.
• Page 9 - why are Phos-tag gels not shown as this would make this result more convincing?
• Figure 3A - I find the SDS-PAGE gel confusing. Why not show the whole gel and why is the band size apparently reduced in the mitotic fraction when previously it was increased (by phosphorylation)? It would also be useful to see if there were any other band shifts.
• "FFAT-2 of ELYS is regulated by phosphorylation" The way you have setup the experiment leads the reader to think you are going to show which sites are differentially phosphorylated in mitosis, but then this is not the case - so there seems no purpose to doing the experiment this way. If you used TMT MS approach you would be able to potentially quantify the change in phosphorylation at the FFAT motif sites in mitosis. Otherwise what is the purpose of using these 2 samples, mitotic and AS?
• For all of the antibodies used, in particular for the PLA, please provide evidence of validation of the antibodies.
• Just a minor point to consider - In the methods for your lysis buffer you use 400mM NaCl - might this slightly reduce the VAPB-FFAT interaction? Worth considering reducing this?
• "The rather small difference observed between the wild-type and the mutant protein observed in this experiment probably results from the presence of endogenous VAPB in the stable cell lines, which could form dimers with the exogeneous HA-tagged versions." If this is the case then please demonstrate that this is happening, or use the KO approach in the major points above.
• "we now show that the proteins can indeed interact with each other, without the need for additional bridging factors (Figs. 1 and 3)." You show that the peptides can bind - but this is not the same thing as the peptide in the full context of the protein - so this should be toned down or removed.
• "Remarkably, this region is highly conserved between species, suggesting that it is important for protein functions (data not shown)". Please show the alignments so the reader can judge for themselves. It is conserved in ALL species and the phosphosites are also conserved??
• "In our experiments, knockdown of VAPA alone did not lead to a delay in mitosis (data not shown). " Why not show this data - as this is a very interesting and potentially important observation? Also add the validation of knockdown of VAPA.
• I find the end to the discussion to the paper rather abrupt. It would be interesting to discuss further how VAPB, but not apparently VAPA reaches the INM and if so why this function is required of an ER adaptor and not another more obvious adaptor protein. In short - why would VAPB be performing this role?

Referees cross-commenting

I agree with the comments of the other reviewers, and they are very much in line with my own review. We all seem convinced that VAPB binds ELYS via a pFFAT, and that this interaction is enhanced during mitosois. However the role of this interaction in mitotic progression remains unclear and based on this data should not be claimed in the title or discussion of the paper.

Significance

Overall, if the manuscript could be improved with the suggested changes, then this could be a considerable conceptual advance in how we understand the VAP proteins, showing functions beyond those as an ER adaptor. This would be significant for the field.

In the context of the existing literature the work does not advance our knowledge of FFAT-VAP interactions, this has already been shown, but it would give a nice example of how this can be regulated during mitosis and how VAP can contribute beyond just as an ER adaptor at membrane contact sites.

There would be a wide audience in the cell biology field and more widely as mutations in VAPB cause a form of ALS, and many people are working in this area.

My field of expertise is in organelle cell biology and membrane contact sites.

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.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This work uses state-of-the art cell imaging and careful image quantification to study early secretory pathway dynamics during budding yeast gametogenesis. The work builds on previous findings of roles for Sec16 in ER exit site formation, Sec-body formation during specific developmental stages, and for Vps13 in lipid homeostasis. The work appears to be carefully conducted and is nicely presented.

Much of the early work - here relating to meiosis in budding yeast, reflects that from studies of mitosis in other systems. This work therefore adds nicely to our understanding of membrane dynamics during cell division. Figures 1 and 2 are useful additions to the literature in this regard. I would have preferred images to be presented in magenta/green rather than red/green for wider accessibility.

The advance is therefore not conceptual but functional. It would, in my view, be unfair to dismiss this as incremental.

Major

My major comment here relates to the FRAP data - the difference in half-life of recovery is clear but there are also substantial differences in the immobile fraction. It is vital that this is expanded on and discussed - it has direct relevance to the conclusions relating to the ongoing functional activity of ERES and the comparison to Sec bodies. Is it not possible that the immobile element here is a functional "reserve" like with Sec bodies? This might be consistent with multiple pools of COPII proteins acting at different stages to maintain then promote secretory activity. Some consideration needs to be given to expanding this and possibly including these data in the main figure. Further analysis and controls should also be included here, other COPII proteins and other markers that one might predict would not alter dynamics in these conditions.

The key mechanistic advance in the manuscript relates to the role of Gip1 with clearly defined outcomes showing its role in ERES remodelling in nascent spores, regeneration of the Golgi and PSM elongation. The context of this part of the work is most important. Specifically, the comparison to VPS13 mutant needs to be expanded on and better explained. The analysis in Fig.4 needs a clearer explanation within the figure of how localization to the PSM is defined. The detail in the methods is also insufficient and the "custom R script" should be published with the work (or on a publicly accessible repository such as Git/Zenodo etc).

The development of this work with the delta-sep mutant gives useful insight and the analysis of Sec16 does indeed support a model where this is an early marker for the process. Despite the link to septins no direct analysis of YSW1 is included (which suppresses the sporulation defect in gip1 ts alleles.

Minor:

Figure 3 introduces new data on reticulons and their impact on ER membrane shape. Again, this reflects findings in other systems but does not add much to the specific narrative of this story but is useful for those in the field. Similar to this, the data on Sec4 are of interest to the specialist but add little to the overall story.

The discussion is quite lengthy and speculative dealing with themes and ideas that are not addressed directly by this work. My comments on the FRAP data relate directly to the models in Figure 8 and this discussion. Given the emphasis on nutrient starvation in the final discussion more detail is needed on the relative experimental conditions used here and in flies/mammals.

Some relevant prior work should also be cited e.g. on the role of Sec16 on exit from mitosis PMID: 21045114, other work relating to gip1 mutants (PMID: 19465564).

Consider presenting images as magenta/green.

Referees cross-commenting

I agree broadly with the other reviewers comments.

While there are elements that could be developed much further. I am not familiar with the role of GIP1 in transcriptional regulation - is this from work in yeast or solely Arabidopsis (is GIP1 here - GBF1 interacting protein, a true equivalent?).

I agree with the comments on the need for further - and well explained - statistical analyses.

Significance

Overall, the work is solid and adds nicely to our understanding. It is likely to be of most interest to a quite specialist audience. The work on PSM formation and spore formation is a clear advance with significant sections of the work being of interest to a wider audience working on early secretory pathway (notably COPII dynamics). Deeper mechanistic insight is missing but non-trivial. More depth would be added by studying further deletion mutants but I am not entirely convinced that this will rapidly advance the field further than this current presentation.

My expertise is in early secretory pathway function.

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

Suda et al. conducted an in-depth investigation of gametogenesis in budding yeast, focusing on the formation of the prospore membrane (PSM) through membrane trafficking rearrangement. They made an interesting observation that the number of endoplasmic reticulum exit sites (ERES) fluctuates during PSM formation, transitioning from decreasing to increasing. The study proposed that ERES regeneration, facilitated by protein phosphatase-1 and its specific subunit, Gip1, plays a crucial role in this process. However, the mechanism by which Gip1 regulates ERES numbers remains unclear, and the authors primarily used Gip1 mutants that may affect transcriptional regulation through Glc7, raising concerns about potential indirect effects. It is essential for the authors to experimentally validate the key mechanisms underlying their findings to strengthen their conclusions.

Major Points:

1. The conclusion that the loss of ERES causes a transient stall in membrane trafficking and leads to Golgi loss is based on the phenotype of GIP1 KO and SED4 KO cells. However, how Gip1 regulates ERES numbers remains unclear. The authors need to define whether Gip1 mediates this regulation through Glc7 dephosphorylation or via transcriptional regulation.
2. The claim of ERES fluctuation during gametogenesis lacks statistical validation (Figure 1D). Since the difference is very small, the authors should perform a statistical analysis to determine if there is a significant difference in ERES numbers during different stages of gametogenesis.
3. The conclusion regarding the loss and regeneration of the Golgi apparatus is based on qualitative observations of Mnn9, Sys1, and Sec7 signals. A quantitative analysis is necessary to strengthen these findings, as some cells may retain these signals despite their disappearance in representative images.
4. Based on phenotypic similarity between GIP1 KO and SED4 KO cells, they concluded that Gip1 regulates the ERES number required for PSM expansion. They demonstrated that the number of ERES and Golgi dramatically decreased in GIP1 KO cells. The authors also need to do this experiment in SED4 KO cells? Since Sed4 affects ER function in general, the authors should demonstrate that SED4 KO cells are appropriate to make a conclusion about ERES regulation and PSM expansion.

Referees cross-commenting

Consistent with the other two reviewers, we feel our comments should be addressed prior to publication of this manuscript.

Significance

Overall, the study presents a high-quality imaging analysis of gametogenesis in budding yeast. However, the authors should experimentally validate the mechanisms underlying ERES regulation by Gip1 and conduct rigorous statistical analyses to support their observations. Additionally, since gametogenesis and Gip1 are yeast specific, the significance of this study might be limited.

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: Yeast gametogenesis requires major membrane reorganization to ensure proper spore formation for survival during starvation, but many questions remain for how this process occurs. This current manuscript by Suda et al. uses fluorescence live imaging to visualize the dynamics of secretory pathway components which are critical for contributing lipids to the prospore membrane (PSM) in the developing spore. The authors find that ER exit sites (ERES) initially decrease and then gradually increase, coinciding with their appearance inside the PSM, suggesting that new lipids for PSM growth are trafficked through the secretory pathway from within the PSM. By screening through known genetic mutants that cause meiosis defects, the authors identify Gip1p, an adaptor for protein phosphatase 1, as a master upstream factor for prospore-associated ERES formation. Interestingly, the authors additionally identify a non-essential component of ERES in vegetative cells, sed4, to be important for sporulation and ERES PSM localization.

Overall, the biological question is interesting and the imaging quality is appropriate. The general conclusion that ERES foci localize inside developing PSMs in a gip1- and sed4- dependent manner is supported by the data. However, the manuscript is purely descriptive; not much molecular insight is gleaned into how secretory pathway components localize to the inside of the PSM, nor is it clear how important this localization is in contributing new lipids to the PSM. Additionally, there are multiple points within the writing and presentation of the results, some specified below, that require clarification; more details in the quantifications also need to be included to ascertain whether the data robustly support the authors' current conclusions.

Major comments:

• The loss of gip1 affects multiple aspects of sporulation and leads to an early termination of spore formation, giving little insight into how ERES are established inside the PSM. The most intriguing result is that loss of sed4, a nonessential paralog of the membrane-bound Sar1 GEF, leads not only to sporulation defects, but also affects the localization of Sec13/ERES to the spores. Given that some spores still form in the sed4 cells, more experiments detailing ERES and golgi localization within the forming spores could be done. Does the golgi no longer localize within the PSM in sed4 cells? Is there are a PSM size difference between those that do and do not have ERES foci in this genetic background? Where does Sed4 localize in gip1 cells?
• While Vps13 is introduced as an additional pathway for supplying lipids, this manuscript does not address the relative contribution of vesicular trafficking versus vps13 lipid transport in PSM formation. Where does Vsp13 localize in the sed4 cells? Are they enriched around/within those spores that do form?
• The clarity of writing in the results and discussion section could be improved, some of which I point out below. The discussion could also be shortened.

Specific comments:

• For all quantifications, more information is necessary, including sample sizes, mean/median values, and number of biological replicates. It may be helpful to include these values in a separate supplemental table.
• Relatedly, for 1D, 3C, and 4C graphs: It is difficult to judge whether the changes of ERES # are significantly different across the various genetic backgrounds as displayed, and given the large spread and small changes, statistical analyses are required to make such conclusions. Could the authors comment on why there is a minor yet noticeable percent of cells with very high ERES numbers?
• To make specific conclusions that ERES 'regenerate' inside PSMs, more detailed quantifications of ERES foci # inside the developing prospores should be included, with appropriate statistical analysis.
• Figure 6A, B: The localization of Glc7 does not look different to me, as claimed. The septin-like cable localization presumably occurs during elongation, as seen in 6A, and gip1D cells do not enter this phase, then it should be expected that there would be no septin-like localization. In 6B, the lower panels seem to show mature, closed PSMs; can the authors label the phases and explain why this is?
• Figure 8, Top panels, indicate the purple coverage is PSM. It is unclear why the authors suddenly say that ERES are 'transiently inactivated' here and in the discussion to describe the lower # of ERES foci, whereas the appearance of PSM-associated ERES foci is considered 'regenerated' (which implies de novo assembly). In general, from the present data, one cannot conclude inactivation vs. formation/regeneration, so some caution in terminology is warranted.

Minor comments:

• A schematic showing the different stages of meiosis and of PSM formation would be useful.
• Scale bar dimensions are missing for most of the figures.
• It may be helpful to use an alternate color combination for merged images (i.e. cyan/yellow, red/cyan, or magenta/green), to accommodate colorblind readers.
• For Figure 1C, authors should show orthogonal views along the z plane at timepoint 8 to show that ERES foci are indeed inside the PSM.
• Figure 1E legend, define closed arrowhead; additionally, include an explanation in the main text of what the Spo20(51-91) marker is.
• For kymograph displays (Figures 1E, 5A, 7E), please include time points in each frame.
• Figure 4D, 4E legend, the multiple terms describing PSM circumference length is confusing: 'cell perimeter, 'PSM perimeter' and 'PSM length'. Please choose one term and describe this fully in the text.
• Fig 5: The images in Fig 5 are dim and the gain should be adjusted accordingly. Figure 5B, is this an intensity trace of one punctum, or punctae from multiple cells as implied in the text?
• More details, not just references, in methods for sporulation induction and image analysis should be included.

Referees cross-commenting

I also agree that our comments should be addressed prior to publication. Of the existing data, the need for further statistical analysis is a high priority.

Significance

The data and conclusions presented here are for a specialized, basic audience interested in yeast meiosis, especially focused on how membranes and the secretory pathway are remodeled during this process. The paper's results have some implications for the reproductive aging field, but this area is not directly investigated in this current manuscript. The paper uses mostly established organelle markers and gene mutants previously known to be involved. The finding of Sed4's involvement in sporulation is, to my knowledge, novel and intriguing.

Reviewer Expertise: organelle morphology, the secretory pathway, protein aggregation, stress responses, aging, fluorescence microscopy, yeast, C. elegans, mammalian cells, biochemistry

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Learn more at Review Commons

---

Reply to the reviewers

I - General criticisms

Reviewer #1: My main criticism is unfortunately inherent to the approach: comparative studies are absolutely critical, but they can only provide a very sparse sampling of diversity. Fortunately, thanks to high-throughput sequencing, bioinformatic analyses can now be performed on a large number of species, but experimental validation is typically restricted to two or three species. The consequence of this for the present manuscript is that while the functional conservation of the Gwl site is convincingly shown, the exact mechanisms responsible for the reduced effect of PKA phosphorylation remain relatively vaguely defined. Indeed, in their Discussion the authors list a number of experimental approaches to address this - but I understand that these would all involve substantial efforts to address. In particular, testing chimeric constructs around the consensus PKA site and from multiple species could be very informative.

We completely agree with the reviewer that comparative approaches are critical to understanding biological mechanisms, and are excited by the increasing possibilities to perform not only sequence and descriptive comparisons but functional studies across a range of emerging model organisms. We hope that more and more researchers in cell and molecular biology will profit from experimental tools and techniques now available in such species, and to pioneer new ones. Of course, and he/she rightly points out, conclusions are currently limited by the number of species studied, but comparisons between two judiciously chosen species can already be very informative. Thus, in our study, the use of Xenopus and Clytia allowed us to make significant progress towards our main objective of understanding the cAMP-PKA paradox in the control of oocyte maturation; specifically by showing both that PKA phosphorylation of Clytia ARPP19 is lower in efficiency and that the phosphorylated protein has a lower effect on oocyte maturation than the Xenopus protein. As the reviewer points out, unravelling the exact mechanisms underlying these differences will require a large amount of additional work and is beyond the scope of the current study. Actually, we have embarked on several series of experiments to this end using some of the approaches listed in the Discussion. Specifically, we are testing the biochemical and functional properties of chimeric constructs containing the consensus PKA site from various species. This is a substantial undertaking which will require one to two years to complete, but is already giving some very interesting findings.

Reviewer #1: The figures and text could be slightly condensed down to about 6 figures.

We have reduced the number of figure panels but we prefer to maintain the number of figures, because the experimental data presented in them is essential to the interpretation of our results and the overall conclusions of the article. If the journal editor would like us to reduce the number of figures, we could do this by displacing Figure 4 and some panels of other figures (to then fuse some of them) to supplementary material, but this would be a pity.

____________II - Abstract

As recommended by Reviewer #2, we have reworked the Abstract to make it more accessible to new readers, attempting to bring out more clearly and simply the main results and conclusions of the study. We correspondingly simplified and shortened the title of the article. Changes: Page 2.

____________III- Introduction points

Reviewer #2: I believe that it would be interesting to include some time-references when introducing the prophase arrest of Clytia and Xenopus oocytes. How long is prophase arrest in Xenopus compared to Clytia or other organisms? How can this affect the prophase arrest mechanisms? It seems that the prophase arrest in Xenopus oocytes is found to be significantly more prolonged compared to Clytia and various other organisms, and also meiotic maturation proceeds much more rapidly in Clytia than in Xenopus. This should be indicated in the introduction with a short introduction of why, and not others, were these species chosen for this study.

Differences in timing of oocyte prophase arrest and in maturation kinetics across animals are indeed highly relevant in relation to the underlying biochemical mechanisms. Unfortunately, not enough information is currently available concerning the duration of the successive phases of oocyte prophase arrest across species to make any meaningful correlations with PKA regulation of maturation initiation. We have nevertheless expanded the Introduction to cover this issue as follows:

• We start the introduction by mentioning how the length of the prophase arrest varies across species. Changes: Page 3, lines 5-11.
• We have added examples of species which likely have similar durations of prophase arrest but show cAMP-stimulated vs cAMP-inhibited release. Changes: Page 4, lines 28-35.
• We have specified the temporal differences in meiotic maturation in Xenopus (3-7 hrs) and Clytia (10-15 min). Changes: Page 5, lines 32-33.

Reviewer #2: why, and not others, were these species [Xenopus, Clytia] chosen for this study. A brief justification is included in lines 1-page 5 "..a laboratory model hydrozoan species well suited to oogenesis studies", but it does not explain why this and not other hydrozoan species like Hydra, that has also been used for meiosis studies.

As requested by Reviewer #2, fuller details are now included about the advantages of Clytia compared to other hydrozoan species, citing several articles and recent reviews here and also in the Discussion. Changes: Page 5, lines 21-32 & 37-39.

Hydra is a classic cnidarian experimental species and has proved an extremely useful model for regeneration and body patterning, but is not suitable for experimental studies on oocyte maturation because spawning is hard to control and fully-grown oocytes cannot easily be obtained, manipulated or observed. In contrast many hydromedusae (including Clytia, Cytaeis, and Cladonema) have daily dark/light induced spawning and accessible gonads, so provide great material for studying oogenesis and maturation. Of these, Clytia has currently by far the most advanced molecular and experimental tools.

Reviewer #2: The proteins MAPK is not introduced properly, as it is first mentioned in the results section in line 12. Given the importance of the results provided with it, it should be presented in the introduction prior to the results section.

As requested by Reviewer #2, the involvement of MAPK activation during Xenopus oocyte meiotic maturation is now introduced, explaining how its phosphorylation serves as a marker of Cdk1 activation. Changes: Page 5, lines 1-5.

Reviewer #2: These sentences need a more elaborate explanation: Page 4 Lines 16-17 "... no role for cAMP has been detected in meiotic resumption, which is mediated by distinct signaling pathways" Which pathways?

We now give the example of the well-characterized pathway Gbg-PI3K pathway for oocyte maturation initiation in the starfish. Changes: Page 4, lines 1-15.

Reviewer #2: Page 4 line 34-39. Introduction indicates that the phosphorylation of ARPP19 on S67 by Gwl is a poorly understood molecular signaling cascade (line 34). However, the positive role of ARPP19 on Cdk1 activation, through the S67 phosphorylation by Gwl, appears to be widespread across all eukaryotic mitotic and meiotic divisions studied (lines 36-37). These two sentences seem a little contradictory. If the general pathway has been identified but the signaling cascade is still not well described, please indicate that in a clearer way.

We apologise that the wording we used was not clear and implied that the mechanisms of PP2A inhibition by Gwl-phosphorylated ARPP19 were poorly understood. On the contrary, they are very well studied. The part that remains mysterious concerns the upstream mechanisms. We have reworded the paragraph to make this point unambiguous. Changes: Page 5, lines 1-8.

____________IV - Results

Reviewer #2: The text of the results is generally well described; however, all the sections start with a long introductory paragraph. I believe this facilitates the contextualization of the experiments, but please try to summarize when possible. For example, in page 5 lines 12-25, or page 7 lines 30-37, are all introduction information.

As requested by Reviewer #2, we have shortened or removed the introductory passages of the Results section paragraphs, which were redundant with the information given in the introduction. We did not restrict to the two examples cited by the reviewer, but have shortened all the Results passages that repeat information already provided in the Introduction. Changes: Page 7, lines 3-4 & 14-16 & 36-37 - Page 8, lines 12-15 - Page 8, lines 37-40 & Page 9, lines 1-6.

Reviewer #2: Page 7, Lines 14-19 present a general conclusion of the findings explained in lines 20-27. I think these results are important and they should be explained better, in my opinion they are slightly poorly described.

We have followed the reviewer's recommendation. The explanation of the experiments and the results are more detailed and the paragraph ends with a general conclusion which came too early in the previous version. Changes: Page 8, lines 22-24 & 32-34.

Reviewer #2: Page 8, lines 16-17: "It was not possible to increase injection volumes or protein concentrations without inducing high levels of non-specific toxicity". What are the non-specific toxicity effects? How was this addressed? What fundaments this conclusion?

Clytia oocytes are relatively fragile. Sensitivity of oocytes to injection varies between batches, while in general increasing injection volumes or protein concentrations increases the levels of lysis observed. We do not know exactly what causes this but lysis can happen either immediately following injection or during the natural exaggerated cortical contraction waves that accompany meiotic maturation, suggesting that it relates to mechanical trauma. We have expanded this paragraph and the legend of Fig. 3C to explain these injection experiments more fully in the text and to clarify these issues. Changes: Page 9, lines 16-29 - Page 32, lines 34-41 & Page 33, lines 1-11 - Supplementary Table 1.

Same paragraph: Lines 25-27 of page 8. Text reads, "These results suggest that PP2A inhibition is not sufficient to induce oocyte maturation in Clytia, although we cannot rule out that the quantity of OA or Gwl thiophosphorylated ARPP proteins delivered was insufficient to trigger GVBD.". Please provide evidence if higher concentrations of OA or Gwl were tested to state this conclusion.

As explained above, we could not increase the concentrations of ARPP19 protein beyond 4mg/ml. It is important to note that at the same concentration, both Clytia and Xenopus proteins induce activation of Cdk1 and GVBD in the Xenopus oocyte.

Concerning OA, it is well documented in many systems including Xenopus, starfish and mouse oocytes as well as mammalian cell cultures, that high concentrations lead to cell lysis/apoptosis as a result of a massive deregulation of protein phosphorylation (Goris et al, 1989; Rime & Ozon, 1990; Alexandre et al, 1991; Boe et al, 1991; Gehringer, 2004; Maton el al, 2005; Kleppe et al, 2015). Specific tests in Xenopus oocytes, have shown that injecting 50 nl of 1 or 2 mM OA specifically inhibits PP2A, while injecting 5 mM also targets PP1 and higher OA concentrations inhibit all phosphatases. For these reasons, we did not increase OA concentrations over 2 mM. When injected in Xenopus oocyte at 1 or 2 mM, OA induces Cdk1 activation, GVBD but then the cell dies because PP2A has multiple substrates essential for cell life. When injected at 2 mM in Clytia oocytes, OA does not induce Cdk1 activation nor GVBD but promotes cell lysis. This supports the conclusion that 2 mM OA is sufficient to inhibit PP2A (and possibly other phosphatases) but that PP2A inhibition is not sufficient to induce oocyte maturation in Clytia.

We have reworded the relevant text to make these points clearer. The previous statement that “we cannot rule out that the quantity of OA or Gwl thiophosphorylated ARPP proteins delivered was insufficient to trigger GVBD” has been removed because it was unnecessarily cautious in the context of the literature cited above, as now fully explained_._ Changes: Page 9, lines 31-35 - Page 32, lines 34-41 & Page 33, lines 1-11 - Supplementary Table 1.

References: Alexandre et al, 1991, doi: 10.1242/dev.112.4.971; Boe et al, 1991, doi: 10.1016/0014-4827(91)90523-w; Gehringer, 2004, doi: 10.1016/s0014-5793(03)01447-9; Goris et al, 1989, doi: 10.1016/0014-5793(89)80198-x; Kleppe et al, 2015, doi: 10.3390/md13106505; Maton el al, 2005, doi: 10.1242/jcs.02370; Rime & Ozon, 1990, doi: 10.1016/0012-1606(90)90106-s

Reviewer #2: Lines 12-13: the sentence "This in vitro assay thus places S81 as the sole residue in ClyARPP19 for phosphorylation by PKA." is overstated. As not all residues had been tested, please indicate that "it is likely that" or "among the residues tested", in contrast to "the sole residue in ClyARPP19".

We realise that we had not explained clearly enough how the thiophosphorylation assay works. In this assay, γ-S-ATP will be incorporated into any amino acid of ClyARPP19 phosphorylatable by PKA. The observed thiophosphorylation of the wild-type protein, demonstrates that one or more residues are phosphorylated by PKA. This thiophosphorylation was completely prevented by mutation of a single residue, S81. This experiment thus shows that S81 is entirely responsible for phosphorylation by PKA in this assay. We have rewritten this section more clearly. Changes: Page 10, lines 18-28.

____________V - Figures and text related to the figures

Figure 1A

Reviewer #2: Why is mouse not included in Figure 1A? Although it might be very similar to human, given that mouse is the species that is most commonly use as a mammalian model, I believe it could be included. However, this is optional upon decision by the authors.

We have replaced the human sequence in Figure 1A with the mouse sequence as suggested. The sequences of each of the mouse and human ENSA/ARPP19 proteins are indeed virtually identical across mammals. Changes: Fig. 1A.

Figure 1C

Reviewer #2: There should be a better explanation in the text of the results sections for the image included in in Fig1 C. Note that Clytia is not a commonly used species, therefore images should be properly explained for general readers. Please indicate in the text that ClyARPP19 mRNA is expressed in previtellogenic oocytes and not in vitellogenic, plus any additional information needed to understand the image. In addition, the detection of ARPP19 in the nerve rings is intriguing. This is mentioned in the discussion section, any idea of its function there? Please include some additional information or additional references, if they exist.

We have expanded the explanations of Fig. 1C in the text and in the figure legend. We have also added cartoons to the figure to help readers understand the organisation of the Clytia jellyfish and gonad. As now explained, ClyARPP19 mRNA is detected in oocytes at all stages, but the signal is much stronger in pre-vitellogenic oocytes because all cytoplasmic components including mRNAs are significantly diluted by high quantity of yolk proteins as the oocytes grow to full size. Changes: page 7, line 40 & page 8, lines 1-9 - Fig. 1C - Legend page 31, lines 19-31.

Nothing is known about the function of ARPP19 in the Clytia nervous system. The only data linking ARPP19 and the nervous system concerns mammalian ARPP16, an alternatively spliced variant of ARPP19. ARPP16 is highly expressed in medium spiny neurons of the striatum and likely mediates effects of the neurotransmitter dopamine acting on these cells (Andrade et al, 2017; Musante et al, 2017). This point is included in the Discussion in relation to the hypothesis that PKA phosphorylation of ARPP19 proteins in animals first arose in the nervous system and only later was coopted into oocyte maturation initiation. Changes: page 16, lines 12-13 & 17-20 - page 19, lines 6-9.

Figure 2A

Reviewer #1: Fig. 2A (and similar plots in subsequent figures): is it really necessary to cut the x axis? Would it be possible to indicate the number of oocytes for each experiment (maybe in the legend in brackets)?

As requested by reviewer #1, the x-axis is no longer cut. The number of oocytes for each experiment is now provided in the legend of Fig. 2A and in similar plots of Fig. 5A and 5D. Changes: Fig. 2A - Legends page 31, lines 37-38 (Fig. 2A), page 33, line 25 (Fig. 5A) - page 33, line 34 (Fig. 5D).

Figure 2D-E (as well as Figure 6C-D and Figure 8B-C)

Reviewer #1: Fig. 2D (and all similar plots below): I am lacking the discrete data points that were measured. Without these it is impossible to evaluate the fits. The half-times shown in 2E are somewhat redundant, and the information could be combined on a single plot.

We added all the data points to the concerned plots: 2D, 6C and 8B. As recommended by reviewer #1, we combined on a single plot the phosphorylation levels and the half-times. 2D-E => 2D, 6C-D => 6C and 8B-C => 8B. Changes: Figs 2D, 6C and 8B - Legends page 32, lines 9-14 (Fig. 2D), page 34, lines 24-30 (Fig. 6C) - page 35, lines 13-18 (Fig. 8B).

Figure 3A and 3B

Reviewer #1: Fig. 3: why is the blot for PKA substrates cut into 3 pieces? It would be clearer to show the entire membrane.

In western blot experiments using Clytia oocytes, the amount of material was limited so the membranes were cut into three parts. The central part was incubated sequentially in distinct antibodies. We finally incubated all three parts of the membrane with the anti-phospho-PKA substrate antibody to reveal the full spectrum of proteins recognized by this antibody. The 3 pieces in Fig. 3A therefore together make up the same original membrane. We had separated them on the figure to make it clear that the membrane had been cut. In the new presentation, the 3 pieces are shown next to each other, making it clear that all the membrane is present, with dotted lines indicating the cut zone as explained in the legend. Changes: Fig. 3A and 3B - Legend page 32, lines 22-25 (Fig. 3A), lines 30-33 (Fig. 3B) - Page 24, lines 3-6 (Methods).

Figure 3C

Reviewer #2: Fig. 3C needs a better explanation in the text. The way these graphs are presented is somehow confusing. The meaning of the dots is not self-explanted in the graph, and it seems that each experiment was done independently but then the complete set of results is presented. Legend says that "each dot represents one experiment" but this is difficult to read as in every analysis the figure also indicates the average and the total number of oocytes. If authors wish so, they can keep the figure as it is, but then please explain this graph better in the text, and please include statistical analysis. These results are very robust, but a comparison between the number of oocytes that go through spontaneous GVBD of lysis in the different conditions will benefit their understanding.

This figure is intended to provide an overview of all the Clytia oocyte injection experiments that we performed, for which full details are given in Supplementary Table 1. Since these experiments were not equivalent in terms of exact timing and types of observation (or films) made and oocyte sensitivity to injection -as ascertained by buffer injections-, it is not justified to make statistical comparisons between groups. We apologise that the presentation was misleading in this respect and hope that the new version is easier to understand. We removed from the figure the average percentage of maturation for each condition between experiments to avoid any misunderstanding of the nature of the data, and rather represent the values of each experiment independently. We also now explain the data included in the figure fully in the text and figure legend. Changes: Page 9, lines 16-39 - Fig. 3C and Supplementary Table 1 - Legend page 32, lines 34-41 & page 33, lines 1-11.

Reviewer #2: Also, please provide in the text a plausible explanation for the cause of oocyte lysis for all experimental conditions (Fig 3C). Given that in the control experiments with buffer this effect is also observed in some oocytes, please explain if this is caused by a mechanical disruption of the oocyte during the injection. In contrast, okadaic acid induces the lysis in all the 14/14 oocytes analyzed, is this due also to the mechanical approach? Or is there other reason more related to the PP2A inhibition? Please explain.

These points are treated above in the response to this reviewer concerning the Results section.

Figure 5

Reviewer #2: In Figure 5 D-F, cited in page 9 lines 35-35. Can you provide an explanation of why the time course of meiosis resumption was delayed?

The binding partners/effectors of XeARPP19-S109D that are involved in maintaining the prophase arrest have not yet been identified. The most probable explanation of the delay in meiotic maturation induced by ClyARPP19-S109D is that Clytia protein recognizes less efficiently these unknown ARPP19 effectors that mediate the prophase arrest. As a result, maturation would be delayed, but not blocked. This explanation was provided in the Discussion (page 17, lines 14-17) and is now mentioned in the Results section. Changes: page 11, lines 16-19.

____________VI - Discussion

Reviewer #2: Although it presents highly interesting suggestions, discussion may border on being overly speculative, especially from line 37 of page 15 till the end.

We agree and have reduced the speculation in this part of the discussion, in particular regrouping and reformulating ideas about evolutionary scenarios in a single paragraph. Changes: page 17, lines 37-41 - page 18, lines 1-41 - page 19, lines 1-18.

SUMMARY - Point by point responses to individual reviewers’ comments in their order of appearance.

Reviewer 1

• The figures and text could be slightly condensed down to about 6 figures.

We have reduced the number of figure panels but we prefer to maintain the number of figures, because the experimental data presented in them is essential to the interpretation of our results and the overall conclusions of the article. If the journal editor would like us to reduce the number of figures, we could do this by displacing Figure 4 and some panels of other figures (to then fuse some of them) to supplementary material, but this would be a pity.

• The exact mechanisms responsible for the reduced effect of PKA phosphorylation remain relatively vaguely defined. Indeed, in their Discussion the authors list a number of experimental approaches to address this - but I understand that these would all involve substantial efforts to address. In particular, testing chimeric constructs around the consensus PKA site and from multiple species could be very informative.

As the reviewer points out, unravelling these exact mechanisms will require a large amount of additional work and is beyond the scope of the current study.

• 2A (and similar plots in subsequent figures): is it really necessary to cut the x axis? Would it be possible to indicate the number of oocytes for each experiment (maybe in the legend in brackets)?

Fig. 2A has been changed in line with the reviewer's request (as well as similar plots in Fig. 5A and 5D). Changes: Fig. 2A - Legends page 31, lines 37-38 (Fig. 2A), page 33, line 25 (Fig. 5A) - page 33, line 34 (Fig. 5D).

• 2D (and all similar plots below): I am lacking the discrete data points that were measured. Without these it is impossible to evaluate the fits. The half-times shown in 2E are somewhat redundant, and the information could be combined on a single plot.

Fig. 2D has been changed in line with the reviewer's request (as well as similar plots in Figs 6C-D and 8B-C). Changes: Fig. 2D, 6C and 8B - Legends page 32, lines 9-14 (Fig. 2D), page 34, lines 24-30 (Fig. 6C) - page 35, lines 13-18 (Fig. 8B).

• 3: why is the blot for PKA substrates cut into 3 pieces? It would be clearer to show the entire membrane.

In western blot experiments using Clytia oocytes, the amount of material was limited so the membranes were cut into three parts. The central part was incubated sequentially in distinct antibodies. We finally incubated all three parts of the membrane with the anti-phospho-PKA substrate antibody to reveal the full spectrum of proteins recognized by this antibody. The 3 pieces in Fig. 3A therefore together make up the same original membrane. In the new presentation, the 3 pieces are shown next to each other, making it clear that all the membrane is present, with dotted lines indicating the cut zone as explained in the legend. Changes: Fig. 3A and 3B - Legend page 32, lines 22-25 (Fig. 3A), lines 30-33 (Fig. 3B) - Page 24, lines 3-6 (Methods).

Reviewer 2

• Abstract needs to be simplified if wants to reach a broader range of readers.

We have reworked the Abstract to make it more accessible to new readers. Changes: Page 2.

• It would be interesting to include some time-references when introducing the prophase arrest of Clytia and Xenopus oocytes. This should be indicated in the introduction with a short introduction of why, and not others, were these species chosen for this study.

We have expanded the Introduction to cover the issue of time-references. Fuller details are now included about the advantages of Clytia compared to other hydrozoan species. Changes: Page 3, lines 5-11, page 4, lines 28-35, page 5, lines 32-33, page 5, lines 21-32 & 37-39.

• The proteins MAPK is not introduced properly, as it is first mentioned in the results section.

The involvement of MAPK activation during Xenopus oocyte meiotic maturation is now introduced. Changes: Page 5, lines 1-5.

• Page 4 Lines 16-17 "... no role for cAMP has been detected in meiotic resumption, which is mediated by distinct signaling pathways" Which pathways?

We now give the example of the well-characterized pathway Gbg-PI3K pathway for oocyte maturation in starfish, also mentioning that in many species the pathways are still unknown. Changes: Page 4, lines 1-15.

• Page 4 line 34-39. Introduction indicates that the phosphorylation of ARPP19 on S67 by Gwl is a poorly understood molecular signaling cascade (line 34). However, the positive role of ARPP19 on Cdk1 activation, through the S67 phosphorylation by Gwl, appears to be widespread across all eukaryotic mitotic and meiotic divisions studied (lines 36-37). These two sentences seem a little contradictory.

The mechanisms of PP2A inhibition by Gwl-phosphorylated ARPP19 are very well studied. The part that remains mysterious concerns the upstream mechanisms. We have reworded the paragraph to make this point unambiguous. Changes: Page 5, lines 1-8.

• Why is mouse not included in Figure 1A?

We have replaced the human sequence in Figure 1A with the mouse sequence. Changes: Fig. 1A.

• 1C: There should be a better explanation in the text of the results sections for the image included in in Fig1 C. Please indicate in the text that ClyARPP19 mRNA is expressed in previtellogenic oocytes and not in vitellogenic.

We have expanded the explanations of Fig. 1C in the text. We have also added cartoons to the figure to help readers understand the organisation of the Clytia jellyfish and gonad. As now explained, ClyARPP19 mRNA is detected in oocytes at all stages, but the signal is much stronger in pre-vitellogenic oocytes because all cytoplasmic components are significantly diluted by high quantity of yolk proteins. Changes: page 7, line 40 & page 8, lines 1-9 - Fig. 1C - Legend page 31, lines 19-31.

• In addition, the detection of ARPP19 in the nerve rings is intriguing. Any idea of its function there?

The only data linking ARPP19 and the nervous system concerns a mammalian variant of ARPP19 that is highly expressed in the striatum. This point is included in the Discussion_. Changes: page 16, lines 12-13 & 17-20 - page 19, lines 6-9._

• Figure 3C. The way these graphs are presented is somehow confusing. If authors wish so, they can keep the figure as it is, but then Also, please provide in the text a plausible explanation for the cause of oocyte lysis for all experimental conditions. please explain this graph better in the text, and please include statistical analysis.

This figure is intended to provide an overview of all the Clytia oocyte injection experiments, for which full details are given in Supplementary Table 1. We have modified the figure and now clarified this fully in the text and figure legend. Clytia oocytes are relatively fragile. Sensitivity of oocytes to injection varies between batches, while in general increasing injection volumes or protein concentrations increases the levels of lysis observed. We do not know exactly what causes this but it probably relates to mechanical trauma. We now explain these injection experiments more fully in the text. Changes: Page 9, lines 16-39 - Fig. 3C and Supplementary Table 1 - Legend page 32, lines 34-41 & page 33, lines 1-11.

• In Figure 5 D-F, cited in page 9 lines 35-35. Can you provide an explanation of why the time course of meiosis resumption was delayed?

The most probable explanation is that Clytia protein recognizes less efficiently the unknown ARPP19 effectors that mediate the prophase arrest in Xenopus. This explanation is provided in the Results section. Changes: page 11, line 16-19.

• All the sections start with a long introductory paragraph. I believe this facilitates the contextualization of the experiments, but please try to summarize when possible.

As requested, we have shortened or removed the introductory passages of the Results section paragraphs, which were redundant with the information given in the introduction. Changes: Page 7, lines 3-4 & 14-16 & 36-37 - Page 8, lines 12-15 - Page 8, lines 37-40 & Page 9, lines 1-6.

• Page 7, Lines 14-19 present a general conclusion of the findings explained in lines 20-27. I think these results are important and they should be explained better, in my opinion they are slightly poorly described.

The explanation of the experiments and the results are now more detailed and the paragraph ends with a general conclusion which came too early in the previous version. Changes: Page 8, lines 22-24 & 32-34.

• Page 8, lines 16-17: "It was not possible to increase injection volumes or protein concentrations without inducing high levels of non-specific toxicity". What are the non-specific toxicity effects? How was this addressed? What fundaments this conclusion?

As explained above, increasing injection volumes or protein concentrations increases the levels of lysis observed due probably to mechanical trauma. But it is important to note that at the same concentration, both Clytia and Xenopus proteins induce activation of Cdk1 and GVBD in the Xenopus oocyte. Changes: Page 9, lines 16-29 - Page 32, lines 34-41 & Page 33, lines 1-11 - Supplementary Table 1.

• Lines 25-27 of page 8. "These results suggest that PP2A inhibition is not sufficient to induce oocyte maturation in Clytia, although we cannot rule out that the quantity of OA or Gwl thiophosphorylated ARPP proteins delivered was insufficient to trigger GVBD." Please provide evidence if higher concentrations of OA or Gwl were tested to state this conclusion.

High OA concentrations lead to cell lysis/apoptosis as a result of a massive deregulation of protein phosphorylation. For these reasons, we cannot increase OA concentrations over 2 µM. When injected in Xenopus oocyte at 1 or 2 µM, OA induces Cdk1 activation, but then the cell dies because PP2A has multiple substrates essential for cell life. When injected at 2 µM in Clytia oocytes, OA does not induce Cdk1 activation but promotes cell lysis. This supports the conclusion that 2 µM OA is sufficient to inhibit PP2A but that PP2A inhibition is not sufficient to induce oocyte maturation in Clytia. We have reworded the relevant text. Changes: Page 9, lines 31-35 - Page 32, lines 34-41 & Page 33, lines 1-11 - Supplementary Table 1.

• Lines 12-13: the sentence "This in vitro assay thus places S81 as the sole residue in ClyARPP19 for phosphorylation by PKA." is overstated. As not all residues had been tested, please indicate that "it is likely that" or "among the residues tested", in contrast to "the sole residue in ClyARPP19".

The observed thiophosphorylation of the wild-type protein demonstrates that one or more residues are phosphorylated by PKA. This thiophosphorylation was completely prevented by mutation of a single residue, S81. This experiment thus shows that S81 is entirely responsible for phosphorylation by PKA in this assay. We have rewritten this section more clearly. Changes: Page 10, lines 18-28.

• Some parts of the discussion are a bit speculative.

We have reduced the speculation in this part of the discussion, in particular regrouping and reformulating ideas about evolutionary scenarios into a single paragraph. Changes: page 17, lines 37-41 - page 18, lines 1-41 - page 19, lines 1-18.

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 of the main findings of the study.

This work presents very interesting data about the maintenance and release of the prophase arrest of oocytes during sexual reproduction. Authors approach some of the remaining questions about oocyte maturation in animals by taking a comparative approach between two species (Clytia and Xenopus) that use opposing cAMP/PKA signaling pathways to trigger oocyte maturation. To do it they focused on phosphorylation characteristics and function of the regulatory protein ARPP19 from the amphibian Xenopus and its orthologue in the hydrozoan Clytia. Results suggest that the low capacity of Clytia ARPP19 to be phosphorylated by PKA. Moreover, Clytia ARPP19 is inherently a poorer PKA substrate than Xenopus ARPP109 both in vivo and in vitro, despite the presence of a functional PKA site. In addition, the absence of functional interactors mediating its negative effects on Cdk1 activation may provide a double security allowing induction of meiosis resumption in Clytia by elevated PKA activity despite the presence of ARPP19, while additional and yet unidentified mechanisms ensure the Clytia oocyte prophase arrest.

Minor comments: read detailed review below. Figure 1 and Figure 3 need a better explanation of the results. Abstract needs to be simplified if wants to reach a broader range of readers. Some parts of the discussion are a bit speculative.

Overall, this work used a robust set of molecular experiments that strongly support the conclusions of the study.

Significance

Strengths and limitations of this work:

The primary strength of this work lies in its innovative use of two distinct species and the integration of molecular experiments to extract conclusions from their different signaling pathways. The well-designed and executed experiments, particularly those of figures 5-9, contribute to an elaborated exploration of the topic, elucidating the underlying mechanisms with clarity. The explanation of each experiment in the results section further adds to the clarity and depth of the study.

The abstract requires improvement, particularly from lines 10 to 21, as it becomes fully understood only after reading the entire manuscript. To make the work more accessible to new readers, it would be good to present the abstract in a more approachable manner. Figures 1C and 3C need a better explanation in the text. Additionally, some sentences would benefit from citations or further clarification in the results or discussion section. Although is presents highly interesting suggestions, discussion may border on being overly speculative, especially from line 37 of page 15 till the end.

Detailed review

Introduction:<br /> I believe that it would be interesting to include some time-references when introducing the prophase arrest of Clytia and Xenopus oocytes. How long is prophase arrest in Xenopus compared to Clytia or other organisms? How can this affect the prophase arrest mechanisms? It seems that the prophase arrest in Xenopus oocytes is found to be significantly more prolonged compared to Clytia and various other organisms, and also meiotic maturation proceeds much more rapidly in Clytia than in Xenopus. This should be indicated in the introduction with a short introduction of why, and not others, were these species chosen for this study. A brief justification is included in lines 1-page 5 "..a laboratory model hydrozoan species well suited to oogenesis studies", but it does not explain why this and not other hydrozoan species like Hydra, that has also been used for meiosis studies.<br /> The proteins MAPK is not introduced properly, as it is first mentioned in the results section in line 12. Given the importance of the results provided with it, it should be presented in the introduction prior to the results section.

These sentences need a more elaborate explanation:<br /> Page 4 Lines 16-17 "... no role for cAMP has been detected in meiotic resumption, which is mediated by distinct signaling pathways" Which pathways?

Page 4 line 34-39. Introduction indicates that the phosphorylation of ARPP19 on S67 by Gwl is a poorly understood molecular signaling cascade (line 34). However, the positive role of ARPP19 on Cdk1 activation, through the S67 phosphorylation by Gwl, appears to be widespread across all eukaryotic mitotic and meiotic divisions studied (lines 36-37). These two sentences seem a little contradictory. If the general pathway has been identified but the signaling cascade is still not well described, please indicate that in a clearer way.

Results section: this review will first comment the figures, and then the text.<br /> Figure 1<br /> Why is mouse not included in Figure 1A? Although it might be very similar to human, given that mouse is the species that is most commonly use as a mammalian model, I believe it could be included. However, this is optional upon decision by the authors.<br /> There should be a better explanation in the text of the results sections for the image included in in Fig1 C. Note that Clytia is not a commonly used species, therefore images should be properly explained for general readers. Please indicate in the text that ClyARPP19 mRNA is expressed in previtellogenic oocytes and not in vitellogenic, plus any additional information needed to understand the image. In addition, the detection of ARPP19 in the nerve rings is intriguing. This is mentioned in the discussion section, any idea of its function there? Please include some additional information or additional references, if they exist.

Figure 3<br /> The way these graphs are presented is somehow confusing. The meaning of the dots is not self-explanted in the graph, and it seems that each experiment was done independently but then the complete set of results is presented. Legend says that "each dot represents one experiment" but this is difficult to read as in every analysis the figure also indicates the average and the total number of oocytes. If authors wish so, they can keep the figure as it is, but then please explain this graph better in the text, and please include statistical analysis. These results are very robust, but a comparison between the number of oocytes that go through spontaneous GVBD of lysis in the different conditions will benefit their understanding.

Also, please provide in the text a plausible explanation for the cause of oocyte lysis for all experimental conditions (Fig 3C). Given that in the control experiments with buffer this effect is also observed in some oocytes, please explain if this is caused by a mechanical disruption of the oocyte during the injection. In contrast, okadaic acid induces the lysis in all the 14/14 oocytes analyzed, is this due also to the mechanical approach? Or is there other reason more related to the PP2A inhibition? Please explain.

Figure 5<br /> In Figure 5 D-F, cited in page 9 lines 35-35. Can you provide an explanation of why the time course of meiosis resumption was delayed?

• The text of the results is generally well described; however, all the sections start with a long introductory paragraph. I believe this facilitates the contextualization of the experiments, but please try to summarize when possible. For example, in page 5 lines 12-25, or page 7 lines 30-37, are all introduction information.<br /> Page 7, Lines 14-19 present a general conclusion of the findings explained in lines 20-27. I think these results are important and they should be explained better, in my opinion they are slightly poorly described.

Page 8, lines 16-17: "It was not possible to increase injection volumes or protein concentrations without inducing high levels of non-specific toxicity". What are the non-specific toxicity effects? How was this addressed? What fundaments this conclusion?

Lines 25-27 of page 8. Text reads, "These results suggest that PP2A inhibition is not sufficient to induce oocyte maturation in Clytia, although we cannot rule out that the quantity of OA or Gwl thiophosphorylated ARPP proteins delivered was insufficient to trigger GVBD.". Please provide evidence if higher concentrations of OA or Gwl were tested to state this conclusion.

Lines 12-13: the sentence "This in vitro assay thus places S81 as the sole residue in ClyARPP19 for phosphorylation by PKA." is overstated. As not all residues had been tested, please indicate that "it is likely that" or "among the residues tested", in contrast to "the sole residue in ClyARPP19".

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 their present manuscript Meneau and coworkers investigate the evolutionary conserved functions of ARPP19 in regulation of meiotic maturation of oocytes. During meiotic maturation, the maturation hormone induces a signaling cascade ultimately leading to the activation of the master regulator, cdk1-cyclin B. within this signaling network, the phosphatase PP2A prevents cdk1 activation in immature oocytes. Upon the action of the maturation hormone, ARPP19 is activated through phosphorylation by the kinase Gwl, and then functions as a potent inhibitor of PP2A, thereby contributing to cdk1 activation. Additionally, ARPP19 is subject to a second layer of regulation: a second site is phosphorylated by the kinase PKA. Interestingly, in vertebrates this cAMP/PKA pathway prevents maturation, while in many other species the same pathway has an opposite effect and cAMP/PKA is indeed sufficient to drive maturation -- referred to as the cAMP paradox.

The authors' major aim was to reveal the molecular basis of these diverse functions of ARPP19 in triggering meiotic maturation. Firstly, they show that the Gwl site is extremely well-conserved all across eukaryotes. They then functionally validate this by comparing the functions of Xenopus ARPP19 to its orthologue in the jellyfish Clytia hemisphaerica. They confirm that the jellyfish ARPP19 is phosphorylated on the conserved Gwl site in vitro and in frog and jellyfish oocytes, acting as a PP2A inhibitor and contributing to cdk1 activation. However, while this is sufficient to drive maturation in Xenopus, PP2A inhibition alone is not sufficient to trigger entry to meiosis in Clytia oocytes, indicating the existence of additional mechanisms. Secondly, they show that the PKA site exists and is phosphorylated both in Xenopus and Clytia. However, the Clytia protein appears to be a much worst substrate for PKA and other interactors, which explains why PKA-phosphorylated ARPP19 does not inhibit maturation either in jellyfish oocytes or when exogenously injected into Xenopus oocytes.

I find the manuscript well-written and easy to follow. The experiments are carefully performed, well-controlled and well-documented. The data shown on the figures fully supports the conclusions drawn -- although the figures and text could be slightly condensed down to about 6 figures. Overall, I would highly recommend the manuscript for publication.

My main criticism is unfortunately inherent to the approach: comparative studies are absolutely critical, but they can only provide a very sparse sampling of diversity. Fortunately, thanks to high-throughput sequencing, bioinformatic analyses can now be performed on a large number of species, but experimental validation is typically restricted to two or three species. The consequence of this for the present manuscript is that while the functional conservation of the Gwl site is convincingly shown, the exact mechanisms responsible for the reduced effect of PKA phosphorylation remain relatively vaguely defined. Indeed, in their Discussion the authors list a number of experimental approaches to address this - but I understand that these would all involve substantial efforts to address. In particular, testing chimeric constructs around the consensus PKA site and from multiple species could be very informative.

In addition, I would have a few small suggestions for improving the figures:

• Fig. 2A (and similar plots in subsequent figures): is it really necessary to cut the x axis? Would it be possible to indicate the number of oocytes for each experiment (maybe in the legend in brackets)?
• Fig. 2D (and all similar plots below): I am lacking the discrete data points that were measured. Without these it is impossible to evaluate the fits. The half-times shown in 2E are somewhat redundant, and the information could be combined on a single plot.
• Fig. 3: why is the blot for PKA substrates cut into 3 pieces? It would be clearer to show the entire membrane.

Significance

Overall, I find this study extremely important, because it is only possible to entangle the diversity of cellular mechanisms though such comparative studies. Oocyte maturation perfectly exemplifies this issue: without doubt, oocyte maturation is a fundamental process and its detailed understanding is critical. However, researchers are often discouraged by diversity across species, which indeed complicates and hinders progress, well-reflected by the name "cAMP paradox". Combined with careful bioinformatic analyses, comparative studies can elegantly resolve such "paradoxes" through resolving the evolutionary history of molecular mechanisms.

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: Centrosomes separate early in mitosis to allow faithful spindle assembly and chromatin segregation. In the current study the author show that in RPE-1 cells the separated centrosomes typically position each other along the shorter axis of the nucleus while in cancer derived U2OS and MDA-MB-486 this is rather random. Mitotic cell rounding is not causal for this effect. Rather, the LINC (linker of nucleoskeleton and cytoskeleton) complex, a protein complex spanning both membranes of the nuclear envelope, is required for this. The data indicate that this is dynactin1 recruitment to the nuclear envelope. The work suggest that proper arrangement of the centrosomes along the short nuclear axis via the LINC complex contributes to chromatin segregation fidelity in RPE-1 cells.

Major comments: The data, derived mostly by life cell imaging of cell culture lines, are of very high quality, carefully controlled and analyzed. They fully support the claims of the study and are well presented, both in text and figures. Statistical analysis seems adequate, but since the authors show different kinds of data sets including time series and use several kinds of statistical tests, it would make sense to indicate the test used for each p-value in all the figure legends. I have no major criticism or experiments to suggest.

Minor comments:

1. Figure legends are quite repetitive and could be shortened. E.g. in Fig. 1 the description for E, F, H and I repeats what has been explained for B and C. Same applies between figure legends. The authors might refer to previous legends if the analysis was done in a similar way.
2. How is nuclear solidity defined and analyzed in Fig S3D?
3. The references to Fig S3 in figure legend 3 ("see Fig S3") do not enlighten the message and could be removed. The same applies to Fig5 - here it is not clear why the author refer to Fig S4.
4. Fig. 3: I suggest to quantify the lamin B1 and LBR overexpression levels.
5. Fig. 5: Consider reordering the panel: Start with the current panel C (as in the text) as it is the necessary control prior to the experimental data.
6. Fig 5 I: what means "before"? Can the authors give a time window they use for analysis.
7. Page 20: "... shortest nuclear axis (Fig. 1C, 5D-G; n.s. - not significant). However, DN-KASH-expressing cells showed compromised separation and positioning of centrosome (Fig. 5D-G, * p=0.0155 and * p=0.0237, respectively). - rather point to the specific panels, i.e. Fig. 1C, 5D and F as well as and Fig. 5E and G.
8. Fig 6B. The DN- KASH bars are on my pdf not visible - use a darker grey
9. Fig S6, albeit mentioned in the text, is not included in the supplementary info.
10. Material and Methods: in general very clear and carefully written
11. a. GlutaMAX instead of GlutaMAXE (page 29)
12. b. What means "as described previously"? No reference is given. Do you refer to the upper part of the method section? (page 30)
13. c. 20 nM HEPES should most probably read 20 mM (page 32)
14. d. "1:50 protease inhibitor; 1:100 Phenylmethylsulfonul fluoride" - which protease inhibitor (mixture)? Rather phenylmethylsulfonyl fluoride.
15. e. exact composition of the cytoskeleton buffer used to prepare 4% paraformaldehyde could be given

Referees cross-commenting

I also mentioned in teh significance section the two weak points (only one non-cancer cell line (RPE-1); the precise role of the LINC complex). I thus think all three reviewers come to a similar conclusion: technically well done albeit some improvements are possible (reviewer 2). Manuscript is interesting but whether the findings can be generalized remains open and the overall impact is limited. Personally, I think a good strategy for the authors might be to stay with the three cell lines and avoid too general statements.

Significance

General assessment: This is a very elaborate analysis of centrosome positioning at the entry of mitosis. The experiments are carefully controlled and the findings supported by multiplied experiments, e.g. the aspect of mitotic cell rounding by analysis of unperturbed cells but also by manipulation accelerating and inhibiting cell rounding. Contribution of the LINC complex is evaluated by shRNA against SUN1/2, i.e. main LINC components, but also by the KASH-DN fragment, which acts as dominant negative. On the downside the study is limited to one untransformed cell line. Given that the treatments interfering with LINC complex function most likely affect all aspects of LINC-centromere interplay, it remains open what precise function of the LINC-complex contributes to chromatin segregation fidelity

Advance: The work clearly shows that at least in RPE-1 cells the separated centrosomes arrange each other along the shorter axis of the nucleus and that the LINC complex is required for this.

Audience: The work is certainly interesting for researches interested in mitosis, most precisely in spindle assembly. It enlightens a very specific aspect of spindle assembly but this very convincing. The work is basic research.

Our experience is basic research of mitosis, nuclear structure and function both using biochemical assays and life cell imaging.

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

A nuclear signal in prophase determines centrosome positioning and ensures efficient mitotic spindle assembly.<br /> Lima and Ferreira investigate in this manuscript the regulation of centrosome positioning in early mitosis. The authors first analyze the position of the two centrosomes either relative to the cell length axis or the shortest or longest axis of the nucleus and describe differences between RPE1, U2OS, and MDA-MB cells. Next, they analyze whether mitotic cell rounding determines the position of centrosomes; however, delayed cortical retraction (Rho-kinase inhibition), adhesion disassembly inhibition (Rap1Q63E), and inducing premature rounding (CalA) did not impact centrosome positioning in RPE1, U2OS, and MDM-MB cells. In addition, the nuclear lamina and LBR were also not required for centrosome positioning on the shortest nuclear axis. In contrast, depletion of SUN1 or SUN2 and overexpression of a dominant-negative DN-KASH affected the nuclear positioning of centrosomes in RPE1 cells. Finally, the authors analyze whether the LINC complex impacts mitotic fidelity. This is indeed the case when SUN1 is depleted, but it is not the case for SUN2 depletion or DN-KASH overexpression. This difference between LINC complex components is not discussed in the manuscript. Since SUN1, SUN2, and DN-KASH affect centrosome positioning in a similar way (Figs. 4 and 5), the chromosome segregation defect in SUN1-depleted cells is most likely not caused by a centrosome position defect but probably by another defect caused by SUN1 depletion.

Major comments

1. Figure 1 is insufficiently explained. The authors have to describe in an understandable way how they measured centrosome-centrosome angle and centrosome-nucleus angle. They should show a cartoon in which these angles are clearly shown. The small cartoons in Fig. 1C are not helpful at all; they are also not explained. The authors should explain the meaning of the black dots (are these centrosomes?) and the even smaller dots. The short nuclear axis should be indicated, e.g., by a red line.
2. On the first page of the manuscript: "Consequently, at the NEP, centrosomes are positioned on the shortest nuclear axis (Fig. 1C) as can be seen in Fig. 1A. This means that the centrosome-nucleus angle relative to the shortest nuclear axis should be 0. However, in Fig. 1C, this angle is between 45 and 90 degrees. This is also the case for Fig. 1D. Please clarify.
3. I find it confusing that in Fig. 1, depending on the subfigure, the short or longest nuclear axis is used as a reference point: Fig. 1C: shortest; D: shortest; F: shortest; G: longest; I: shortest; J: longest. Thus, even within the same cell line, the reference point is changing. What is the rational for this variation?
4. Fig. 4K, L, M: in figure, y-axis: "shortest nuclear axis". In legend: "relative to the longest nuclear axis". I guess the longest nuclear axis is correct. Same in Fig. 5D and E. Fig. 5C lacks the WT control.
5. The cells in Fig. 5J are not comparable: one has a monopolar spindle, the other a bipolar. The authors need some other NE protein as a control to show that the reduction of dynein by DN-KASH is a specific defect and not a broad impact on the NE. The dynein data in Figs. 5J-L need to be extended to SUN1/2.
6. The title of the paper is misleading: the authors do not provide any indication for a nuclear signal in prophase that determines centrosome positioning.

Minor comment

1. It would make sense to use the same time scale in Figs. 1A and B (either min.sec. or sec.) to allow direct comparison.
2. 2nd section: Mitotic cell rounding "The authors state: Given that cancer cells failed... I would be careful with this generalization; only one cancer cell was used in this study.
3. The authors say: "However, they did not place the centrosomes at the shortest nuclear axis (Figure 4K-M)." Centrosomes are still on the shortest nuclear axis but not as frequent as in control.
4. The white color in Fig. 6B cannot be seen and needs to be changed to something else.
5. The paper has neither line nor page numbers.

Referees cross-commenting

My comments are more or less reflected by the comments and concerns of reviewer 1 (only one cancer cell line; the role of the LINC complex). This reduced the impact of this manuscript that is certainly intresting and has novel aspects.

Significance

The manuscript analysis an early step in spindle assembly: the positioning of the two centrosomes on the NE. As such, the paper is interesting and important. They exclude cell rounding and lamin disassembly as mechanisms for centrosome positioning. The SUN1/2 and KASH data on centrosome positioning are convincing, and they provide a novel finding on the function of the LINC complex in centrosome positioning, probably via dynein recruitment to the NE. It remains unclear whether LINC recruits dynein directly or functions via one of the two known dynein/NE recruitment pathways. LINC-dynein at the NE binds centrosome microtubules and dynein pulls them towards the NE. However, how LINC-dynein spatially positions centrosomes relative to the short axis of the nucleus remains unclear (dynein uniformly decorates the NE (Fig. 5J)). The data on chromosome missegregation are not so clear because the defect only occurs in SUN1-depleted cells. Thus, this phenotype indicates most likly a function of SUN1 but not the LINC complex and is probably not related to centrosome positioning since all LINC components affect centrosome positioning. The paper falls short in explaining how parameters were measured and contains mistakes in the figures, as outlined above. The paper lacks a coherent story (a little bit on cancer, some negative data, LINC-dynein, but it stops on the surface).

It will be relatively easy to improve some aspects of the manuscript (explaining the angles, correcting the figures: one week). Measuring dynein at the NE in SUN1/2-depleted cells is also easy to do (1-2 months). To get more mechanistic insides into how LINC-dynein positions centrosomes probably will not be possible during revision time.

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 study, Lima and colleagues, investigate the mechanisms controlling the position of the two centrosomes at nuclear envelope breakdown. The authors show that in the non-cancerous human epithelial RPE1-cell line, the centrosomes are generally positioned in the short axis of the nucleus; in contrast in two cancer cell lines, they did not find an equivalent pattern. When the authors set out to identify potential molecular players required for this positioning, they find that the LINC complex is required , possibly by recruiting dynein to the nuclear membrane. Finally, the authors show that disruption of the LINC complex is associated with chromosome segregation errors.

Major Comments:

In general, the presented experiments are of excellent technical quality. The main conclusions of the manuscript, are however, not always well supported by the experimental data. They should be either interpreted more cautiously or supported by additional experimental evidence. I highlight these here, using the main conclusions of the abstract.

1. "We show that in untransformed cells, centrosome positioning is regulated by a nuclear signal, independently of external cues. »

The authors conclude based on three cell lines that the centrosome positioning mechanisms is present in non-transformed cells and not in cancerous cells. The authors have, however, only analysed 1 non-cancerous cell line, and they compare cells originating from vastly different tissues (retina, bones and breast) and origins (epithelial vs. mesenchymal cartilage cells). Such a general statement is not possible, without a systematic comparison of several healthy cells vs cancerous cells from the same tissue.<br /> 2. "This nuclear mechanism relies on the linker of nucleoskeleton and cytoskeleton (LINC) complex that controls the loading of dynein on the nuclear envelope (NE), providing spatial cues for robust centrosome positioning on the shortest nuclear axis, prior to nuclear envelope permeabilization (NEP). »

While the data showing that centrosome positioning depends on the LINC complex is solid and robust, some of the "negative" examples identified by the authors are less convincing. One the process the authors study is cell rounding. Based on the fact that Rap1 transfection or treatment with Calyculin A does not lead to differences that are statistically different, the authors conclude that cell rounding is not involved. However, absence of statistical difference does not mean that there is no difference. Indeed, when comparing the raw data in Figure 2L and 2Q to the positive hit shSun2 in Figure 4J, one could conclude that cell rounding does make a difference, and that this statistical difference would emerge if the authors would count a high number of cells. Therefore the authors should interpret these results in a more differentiated manner, and also instead of just stating non-significant, state also the real p-values for the different experiment.<br /> The second major concerns emerges when looking at the data in Figure 5, when the authors test for the abundance of the dynein complex on the nuclear envelope in cells treated with DPPPL-KASH or DN-KASH. Yes, there is a statistical difference, but the absolute difference is tiny (I estimated a normalized intensity of 1.44 vs 1.35). This is a difference of less than 10%. How do the authors think that such a small change in dynein could have such a strong effect on centrosome positioning? Would a partial dynactin depletion by 10% give an equivalent result? Does the depletion of other proteins involved in the late recruitment of dynein at the NE also affect centrosome positioning?<br /> 3. « Moreover, we demonstrate this mechanism is altered in cancer cells, leading to increased chromosome segregation errors. »

Here the authors infer that the identified mechanism is absent in cancer cells and that its absence contributes to chromosome segregation errors. Both conclusions are not supported by the presented data. First, the authors did not test whether any members of the LINC complex or dynactin is present at lower levels on the nuclear membranes of the cancer cells. Such a direct validation would be essential to make such a strong statement. Second, the authors conclude that this mechanism prevents chromosome segregation errors, based on the fact that depletion or impairment of the LINC complex (shSUN1, shSUN2, DN-KASH) results in chromosome segregation errors. These perturbations lead ,however, as noted by the authors themselves to pleiotropic effects, including insufficient retraction of nuclear membrane, which will can all contribute to chromosome segregation errors. It is therefore impossible to estimate the contribution of the centrosome positioning mechanism to these segregation errors using this type of pertubrations. One could even argue that this mechanism might not be that important, since depletion of SUN2, which also impairs centrosome positioning has no significant effect on chromosome segregation.

Minor comments:

The author state in the Material and methods that all the figure legends contain the number of replicates. This is, however, not the case, the authors only indicate the total number of analyzed cells.

Referees cross-commenting

I agree that all three reviewers come to similar conclusions - strong technical quality, novel results and concepts, but some limitations due to lack of precise tools or the limited number of model cell lines investigated.<br /> I recommend that the authors prioritize which are the suggested experiments that could be done within a few months, and otherwise rephrase their conclusions in less general terms.

Significance

This study establishes for the first time that some cell lines set up the mitotic spindle at predefined positions of the nucleus and they identify a first molecular complex controlling this complex.

The strength of this study is the high technical quality of the data - a limitation is the over-interpretation of the current data (see major comments), and the fact that the authors do not have a tool that specifically only disrupts centrosome positioning, which would allow them to probe the importance of this mechanism.

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.

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 interesting paper Bideau et al. report an antero-posterior gradient in the plasticity of tail tissues during tail regeneration in the annelid worm Platynereis dumerilii. The experiments are well designed and thoroughly quantified, the figures are of high quality.

Major comments:

The microscopic images lack scale bars. These should be added to all figures.

The authors should provide the source data for all quantifications as txt files. They should also provide as supplement representative confocal stacks for the various stainings.

The authors use LatrunculinB treatment to investigate the role of cell migration in regeneration. However, since LatB inhibits f-actin, it could also interfere with cell proliferation and other processes. The authors should check if this is the case and provide control data.

Minor comments:

Sometimes the language is a bit quite cryptic. For example, the title of Figure 4 is "Cell proliferation and migration, as well as tissue maturity modulate the plasticity of posteriorized gut progenitors through regeneration"<br /> in short: 'cell migration modulates the plasticity of progenitors' - this is just to say that inhibiting cell migration reduces regeneration

The authors should attempt to simplify the language.

Language:

"is an tremendous and essential process in animals" - not clear what 'tremendous' means here - please revise

"Those regeneration processes have been studied from a long" - for a long time

"more EdU+ cells in S1 than in S6 or S7 regardless the EdU incubation time" - regardless of the

"It showed that the gut is composed of" - The stainings showed that

"Indeed, cell labelled with a rather short EdU" - cells labelled

"tissues plays a major role on the reformation" - in the formation

The paper will be of interest to animal developmental biologists and scientists working on the plasticity of tissues during regeneration.

Referees cross-commenting

I agree with the comments made by the other reviewers. The authors need to be more clear and careful in interpreting their data. I don't think that new data are needed (unless they would like to demonstrate a 'gradient' with more positions) and the comments could be addressed by substantially rewriting the text and revising the claims.

Significance

The paper will be of interest to animal developmental biologists and scientists working on the plasticity of tissues during regeneration.

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

Bideau et al. studied the origin, plasticity and fate of the cells participating in blastema formation during posterior regeneration in the annelid Platynereis dumerilii. To label and track the fate of proliferative cells, the authors applied EdU/BrDU incorporation coupled with mRNA in situ hybridization and fluorescent beads labelling, on a wide array of regeneration assays. They also performed drug treatments to assess the role of proliferation and cell migration during posterior regeneration. Interestingly, the Authors showed that some proliferative gut cells can participate in the formation of ectodermal and mesodermal tissues during regeneration, in case of two successive posterior regeneration events, suggesting that gut cycling cells residing in a regenerating segment are more plastic than those located in a non-regenerating segment. They also suggested the existence of two different cell populations - "slowly" and "quickly" cycling cells - acting during regeneration. Overall, the experiments are well presented, and methods clearly described. The Authors concluded that posterior regeneration in Platynereis relies on a gradient of cell plasticity and cell proliferation, along the antero-posterior axis of the animal.

Major comments

Two of the main conclusions of this study are, in my opinion, not supported by the data:

As indicated in the title of the manuscript, the Authors put forward a cellular model for posterior regeneration relying on gradients of cell proliferation, cell differentiation and cell plasticity along the the antero-posterior axis of the animal. I am not convinced that the Authors have provided strong enough evidence to prove any of these gradients. They showed that there are differences between the region directly adjacent the most posterior segment and a region located more anteriorly (6 or 7 segments from the posterior end). However, by comparing only two positions, they cannot distinguish between graded or clearly regionalized contexts. To prove the existence of a gradient along the animal's antero-posterior axis, the authors would need to compare cell proliferation, cell dynamics and cell differentiation between multiple regions at increasing anterior positions, and show that their responses are indeed graded. This would represent a quite substantial amount of work. Instead, I would suggest removing the reference to a gradient in the paper entirely.

Using "short" (5h) and "long" (48h) EdU pulses, the Authors claim they have established the existence of two cell populations, namely "slowly-cycling cells" and "quickly-cycling cells" (first paragraph of the result section - pages 5/6 "We exposed uninjured worms to EdU, either for 5 or 48h to discriminate quickly-cycling cells from cells harboring a slower replication rate"). I am not convinced that the Authors provide strong enough evidence to demonstrate the existence of two such cell populations. Given that about 20% of cells incorporated EdU after 5h of exposure, that almost all of them have done so after 48h, and that only a fraction of proliferative cells are in S-phase at any given time, it is well possible that a majority of cells stained after 5h and 48h are from the same cell cycling population. To show the existence of different populations of cells, cycling at different rates, the Authors would need to compare staining after an equal EdU exposure time, following a period of chase of different duration. Without this set of experiments, I would refrain from distinguishing between several slower and faster cell cycling populations.

Minor comments

Page 1. Please correct "is an tremendous" into "is a tremendous".

Page 7. "The huge majority of the EdU+ cells colocalize with FoxA". Please provide quantification.

Page 11 "Quickly-cycling gut progenitors.... cannot give rise to neural progenitors and probably not to stem cells from the ectodermal growth zone"; Page 12 "cannot regenerate neural tissues"; Page 20 "posterior gut progenitors cannot produce nervous system or putative posterior stem cells". What the authors show in their experiments, is that labeled gut cycling cells likely do not generate neural cells or stem cells, in the assessed context. However, the Authors do not show that those cells 'cannot' do so. Please rephrase.

Page 11. "migration, through actin polymerization (LatrunculinB or LatB) widely used inhibitors". Please add a reference to justify the use of LatB as a cell migration inhibitor.

Significance

This is a thorough, well executed and interesting study on a tractable annelid regeneration model. The experiments are neatly performed and the manuscript reads well. As stated in my major comments, two of the main conclusions of the study (gradient of cell proliferation/plasticity/differentiation; identification of two types of progenitors differing in their cell cycle rates) have not been demonstrated properly and would need to be either strengthened or deleted from the manuscript. Several other findings, notably the increased plasticity of cells that recently participated in posterior regeneration (notably gut cells) are well demonstrated and of interest. Overall, this manuscript significantly advances our understanding of the cellular mechanisms that occur during posterior regeneration in Platynereis. It will be of interest to anyone working on comparative regeneration, but may be of lesser interest to researchers working outside this field.

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

Bideau et al. describe posterior regeneration in the annelid Platynereis. The authors aimed to identify patterns of proliferative cells. Pulse-chase and double labeling by EdU/BrdU was used to track cells. Finally, they attempted to reveal the identity of cycling cells and their contribution to regeneration. Platynereis is a relatively new regeneration model. Understanding the cellular source of regeneration in this annelid would be of considerable interest.

The authors performed EdU labeling of intact worms to find a posterior-anterior decreasing gradient of S-phase cells. Histological sections showed that proliferative cells are located in all tissues in the posterior-most segment, but mostly restricted to the gut epithelium in more anterior segments. Using fluorescent beads that are taken up specifically by gut epithelial cells, they show that gut epithelial cells of the intact animal contribute only to gut regeneration, i.e., they are lineage restricted. The authors also performed immunostaining and mRNA in situ hybridization experiments to better understand the tissue identity of proliferative cells.

The following are my specific comments:

1. I am not sure I understand how the authors identify slow- or fast-cycling cells. EdU gets incorporated in S-phase; the longer the incubation time the more cells will be labeled until saturation is reached. The length of the cell cycle and the number of different populations cannot be directly derived from this experiment. I think it would be fair to conclude that there are more cycling cells in posterior segments and in the gut of anterior segments but the conclusion of two distinct populations is unsupported in my opinion.
2. The Results and Discussion sections will have to be revised to address the above issue. The two supported conclusions are (1) the gradient of proliferative cells (but w/o reference to the number of distinct populations); (2) the fate-restricted nature of gut epithelial cells. The plasticity gradient is unsupported because worms can regenerate if amputated at segment #5. This suggests the presence of either resident stem cells (with broad potential or lineage-specific), cells that can dedifferentiate, or a combination of both. The authors' experiments cannot discriminate between the alternatives.
3. The text would benefit from copy editing to improve the language and making it more accessible. In its current form, it is rather difficult to read, with descriptions of experiments that are not easy to follow.
4. The figures and their legends can be improved. Re the legends, one has to read the full text to understand what each panel shows. The figures are very complex. It would already be easier if usage of A', A', A' was avoided. The figures could also be improved by direct annotation. Finally, consider simplifying the main figures and moving some material to the supplement.
5. How do the authors know that proliferative cells in the gut are "gut progenitors"? They might simply be proliferative gut epithelial cells.
6. The conclusions drawn from the drug experiments are overstated.

Referees cross-commenting

The comments made by the two other reviewers are complementary to mine. Either the authors extensively revise the text to remove unsupported conclusions or they perform additional experiments.

Significance

Little is known about the cellular basis of Platynereis posterior regeneration.

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 helpful and constructive comments. In response to these suggestions we have performed new experiments and amended the manuscript, as we describe in our detailed response below.

Reviewer #1:

1. The Reviewer notes that while our analysis of centrosome size was comprehensive, we provided no analysis of centrosomal MTs, pointing out that while centrosome size declines as the embryos enter mitosis, the ability of centrosomes to organise MTs might not. This is a good point, and we now provide an analysis of centrosomal-MT behaviour (Figure 2). We find that there is a dramatic decline in centrosomal MT fluorescence at NEB, although the pattern of centrosomal MT recruitment prior to NEB is surprisingly complex.

2. The Reviewer questions how PCM client proteins can be recruited in different ways by the same Cdk/Cyclin oscillator. We apologise for not explaining this properly. It is widely accepted that Cdk/Cyclins drive cell cycle progression, in part, by phosphorylating different substrates at different activity thresholds (e.g. Coudreuse and Nurse, Nature, 2010; Swaffer et al., Cell, 2016). Moreover, it is also clear that Cdk/Cyclins can phosphorylate the same protein at different sites at different activity thresholds (e.g. Koivomagi et al., Nature, 2011; Asafa et al., Curr. Biol., 2022; Ord et al., Nat. Struct. Mol. Biol., 2019). Thus, we hypothesise that rising Cdk/Cyclin cell cycle oscillator (CCO) activity phosphorylates multiple proteins at different times and/or at different sites to generate the complicated kinetics of centrosome growth. We now explain this point more clearly throughout the manuscript.

3. The Reviewer is puzzled as to how we conclude that Cdk/Cyclins phosphorylate Spd-2 and Cnn at all the potential Cdk/Cyclin phosphorylation sites we mutate in our study. The Reviewer is right that we cannot make this conclusion, and we did not intend to make this claim. As we now clarify (p11, para.1), although it is unclear if Cdk/Cyclins phosphorylate Spd-2 or Cnn on all, some, or none of these sites, if either protein can be phosphorylated by Cdk/Cyclins, then these mutants should not be able to be phosphorylated in this way—allowing us to address the potential significance of any such phosphorylation. We now also note that several of these sites have been shown to be phosphorylated in embryos in Mass Spectroscopy screens (Figure S6).

4. The Reviewer highlights differences in how Spd-2 and Cnn help recruit γ-tubulin to centrosomes (Figure 6). They ask for a more detailed description, and are puzzled as to how this is compatible with direct regulation by a single oscillator. We now explain our thinking on this important point in much more detail. It appears that Spd-2 helps recruit γ-tubulin throughout S-phase, while Cnn has a more prominent role in late S-phase (Figure 6). This is consistent with our overall hypothesis of CCO regulation, as we postulate that low-level CCO activity promotes the Spd-2/γ-tubulin interaction in early S-phase, while higher CCO activity promotes the Cnn/γ-tubulin interaction in late-S-phase, potentially explaining the increase in the rate of γ-tubulin (but not γ-TuRC) recruitment we observe at this point (see minor comment #1, below, for an explanation of the various γ-tubulin complexes in flies). This is consistent with recent literature showing that CCO activity promotes γ-tubulin (but not γ-TuRC) recruitment by Cnn/SPD-5 in worms and flies (Ohta et al., 2021; Tovey et al., 2021).

5. The Reviewer was not convinced by our model (Figure 8, now Figure 9), raising two major concerns. First, they were unsure how a single oscillator could generate different patterns of protein recruitment. We addressed this in point #2 and #4, above, where we explain how different thresholds of CCO activity trigger different events, so there is no expectation that we should observe steady changes in recruitment over time as CCO activity rises. Second, they questioned how modest levels of Cdk/Cyclin activity can promote recruitment, while high levels of activity can inhibit recruitment. In point #1, above, we cite several examples where such positive and negative regulation by different Cdk/Cyclin activity levels have been described. We also now explain throughout the manuscript why this hypothesis provides a plausible explanation for our results: with moderate CCO activity promoting Spd-2-dependent PCM-client recruitment in early S-phase; higher CCO activity promoting a decrease in Spd-2 recruitment in mid-late-S-phase (so centrosomal Spd-2 levels decline); and even higher levels of CCO activity leading to a decrease in the interactions between the client proteins and the Spd-2/Cnn scaffold as the embryos enter mitosis (so the client proteins are rapidly released from the centrosome).

The Reviewer also raised the important point here that our model does not explain why the mutant forms of Spd-2 and Cnn accumulate to higher levels at the start of S-phase, and not just at the end of S-phase/entry into mitosis. We apologise for not explaining this properly. The accumulation of the mutant proteins (particularly Spd-2, Figure 5C) in early-S-phase occurs because the excess mutant protein that accumulates at centrosomes in _late-_S-phase/mitosis is not removed properly from centrosomes during mitosis (presumably because there is insufficient time). Thus, centrosomes still have too much mutant Spd-2 at the start of the next S-phase. We show this in Reviewer Figure 1 (attached to this letter), which tracks Spd-2 behaviour further into mitosis, and now explain this in more detail in the text (p12, para.1).

1. The Reviewer questions how the CCO can both induce centrosome growth and also switch it off, as it is unclear how an oscillator that only phosphorylates sites to decrease centrosome binding could also promote growth. They ask if we can identify and mutate any Cdk/Cyclin sites in centrosome proteins that promote centrosome recruitment. As we now clarify, we did not intend to claim that the CCO only phosphorylates sites that decrease the centrosome binding of proteins, although we do hypothesise that such phosphorylation is important for switching off centrosome growth in mitosis. In addition, we hypothesise that moderate levels of CCO initially promote centrosome growth, and our data suggests that the CCO does this, at least in part, by promoting Polo recruitment (Figure 8). We speculate that the CCO phosphorylates specific Polo-box-binding sites in Ana1 and Spd-2, the main proteins that recruit Polo to centrioles. We agree that identifying these sites is an important next step, but it is complicated as our studies indicate that multiple sites contribute in a complex manner. Importantly, it is well established that the CCO triggers centrosome growth as cells prepare to enter mitosis, so our hypothesis that moderate levels of CCO activity initiate centrosome growth is not new or controversial.

Minor Comments

1. The reviewer asks how we explain the different incorporation profiles we observe for the different subunits of the γ-tubulin ring complex. We apologise for not discussing this point. In flies there is a “core” γ-tubulin-small complex (γ-TuSC) and a larger γ-tubulin-ring complex (γ-TuRC) that contains the Grip71, Grip75 and Grip128 subunits we analyse here (Oegema et al., JCB, 1999). The γ-TuSC functions independently of the γ-TuRC so γ-tubulin and γ-TuRC components can behave differently.

2. The Reviewer questions why we claim an “inverse-linear” relationship between S-phase length and the centrosome growth rate when the relationship is not linear (Figure 3, now Figure S3). I was originally confused by this as well but, mathematically, a linear relationship means y is proportional to x, whereas an inverse-linear relationship means y is proportional to 1/x. Thus, an inverse-linear relationship between x and y does not plot as a straight line, but rather as the curves we show on the graphs. We now explain this in text (p9, para.2).

Reviewer #2:

This Reviewer found the manuscript hard to follow, so we are very grateful that they took the time to try to understand it. We agree that the subject matter is complicated, and that our presentation was not always helpful. The Reviewer’s comments have been very useful in helping us to identify (and hopefully improve) areas of particular difficulty.

Major points:

1. The Reviewer highlights that the two experimental approaches underpinning our main conclusions are problematic: (1) Experiments with mutants of Spd-2 and Cnn that theoretically cannot be phosphorylated by Cdk/Cyclins are hard to interpret as these mutations may have other effects; (2) It is unclear whether reducing Cyclin B levels reduces peak CDK activity or simply slows the time it takes to reach peak levels. They suggest a more direct test of our model would be to analyse PCM recruitment in embryos arrested in S-phase or mitosis. (1) We agree that the mutations designed to prevent Cdk/Cyclin phosphorylation could perturb function in other ways, but this is true for any such mutation, and there are many papers that infer a function for Cdk/Cyclin phosphorylation from such experiments. Importantly, the centrosomal accumulation of the phospho-null mutants actually slightly increases compared to WT (Figure 5C and I), and we now show that the centrosomal accumulation of a phosphomimicking Spd-2-Cdk20E mutant slightly decreases (Figure S8). We now acknowledge the potential caveat of a non-specific perturbation of protein function, but feel that the reciprocal behaviour of the phospho-null and phospho-mimicking mutants somewhat mitigates this concern (p12, para.2). (2) Fortunately, and as we now clarify, it has recently been shown that reducing Cyclin levels does not reduce peak Cdk activity, but rather slows the time it takes to reach peak activity (Figure 2A, Hayden et al., Curr. Biol., 2022). Thus, the cyclin half-dose experiments provide an excellent alternative test of our hypothesis as they show that the WT proteins can exhibit similar behaviour to the mutants if the rate of Cdk/Cyclin activation is slowed. We feel the evidence supporting our hypothesis is strong enough that it warrants serious consideration.

The suggestion to look at PCM recruitment in embryos arrested in either S-phase or M-phase is a good one, but these experiments produce complicated data. In M-phase arrested embryos, for example, Cnn levels continue to rise (see Figure 1G, Conduit et al., Dev. Cell, 2014), but the other PCM proteins do not (unpublished); in S-phase arrested embryos (arrested by mitotic cyclin depletion) centrosomes continue to duplicate, but now do so asynchronously, greatly complicating the analysis (McCleland and O’Farrell, Curr. Biol.., 2008; Aydogan et al., Cell, 2020). The centrosomes that don’t duplicate, however, reach a constant steady-state size (where the rate of centrosome protein addition is balanced by the rate of loss). These observations are consistent with our recent mathematical modelling of mitotic PCM assembly (Wong et al., 2022) if we additionally account for cell cycle regulation (which was not considered in our original model). We believe such analyses are beyond the scope of the current paper and we plan to publish a second paper incorporating our new hypothesis into our mathematical modelling.

1. The Reviewer questions whether our methods accurately measure centrosomal protein accumulation, pointing out that γ-tubulin and Grip128 occupy different centrosomal areas—which should not be possible if they are part of the same complex. They suspect that our use of different transgenes with different promotors could explain these differences. As we should have described (see point #1 in our response to the minor comments of Reviewer #1), γ-tubulin exists in two complexes in flies, only one of which contains Grip128, so γ-tubulin and Grip128 exhibit different localisations. Moreover, as we now show (Figure S2), using different promotors does not seem to make a difference to overall recruitment kinetics. Thus, we are confident that our methods measure centrosome protein recruitment dynamics accurately.

2. The Reviewer is concerned that our measurements of centrosome size based on fluorescence intensity (Figure 1) and centrosomal area (Figure S1) do not always match. They suggest a potential reason for this is that proteins are not uniformly distributed within centrosomes, and this may impact our ability to measure protein accumulation based on 2D projections (noting, for example, that Polo and Spd-2 are concentrated at centrioles and in the PCM, potentially explaining the different shape of their growth curves compared to the client proteins). When the centrosome-fluorescence-intensity and centrosome-area recruitment profiles of a protein do not match, the average “centrosome-density” of that protein must be changing over time. In some cases, we understand why density changes. Cnn, for example, stops flaring outwards on the centrosomal MTs during mitosis so its centrosomal area decreases even as its fluorescence intensity increases (leading to an increase in its centrosomal-density). We agree (and now discuss—p19, para.3) that the prominent accumulation of Spd-2 and Polo at centrioles could help to explain why Spd-2 and Polo accumulation dynamics differ from the client proteins.

Other points:

1. The Reviewer suggests it would be good to know how much Polo at the centrosome is active. We agree, but although commercial antibodies against PLK1 phosphorylated in its activation loop work in cultured fly cells, we cannot get them to work in embryos. Moreover, the recruitment of Polo/PLK1 to its site of action by its Polo-Box Domain is sufficient to partially activate the kinase independently of phosphorylation (Xu et al., NSMB, 2013). Thus, it seems likely that all the Polo/PLK1 recruited to centrosomes will be at least partially activated, even if it is not necessarily phosphorylated on its activation loop.

2. The Reviewer asks if it is clear that less Spd-2 and Cnn are recruited to centrosomes in the half gene-dosage embryos. We apologise for not mentioning that this is indeed the case. We showed this previously for Cnn (Conduit et al., Curr. Biol., 2010) and we now state that this is also the case for Spd-2. We do not show the Spd-2 data as we plan to publish a comprehensive dose-response curve of Spd-2 (and Cnn) recruitment in our next modelling paper.

3. Would it not be relevant to examine Polo ½ dosage embryos? We do have this data (Reviewer Figure 2), attached to this letter, but it is quite complicated to interpret (as we explain in the legend). We feel it would be more appropriate to include this in our next modelling paper where we can properly explain the behaviours we observe. Publishing this data here would distract from our main message without changing any of our conclusions.

4. The Reviewer asks why the non-phosphorylatable Spd-2 protein is also present at higher levels on centrosomes at the start of S-phase (not just the end of S-phase). This was also raised by Reviewer #1 (point #5), so please see the second paragraph of our response there.

Minor/Discussion Points:

1. We thank the Reviewer for highlighting that absolute and relative centrosome size control are different things and we have amended the manuscript accordingly.

2. The Reviewer questions whether it is accurate to describe Spd-2 and Polo as scaffold proteins, noting that only Cnn has been shown to have scaffolding properties. There is strong evidence that Spd-2 has Cnn-independent scaffolding properties in flies (e.g. Conduit et al., eLife, 2014), but this is a fair point for Polo. We think it is justified to separate Polo from other client proteins as Polo is essential for scaffold assembly, whereas other client proteins are not. We now define our scaffold/client terminology to avoid confusion (p4, para.3).

3. The Reviewer highlights several points related to differences in recruitment kinetics (also touched on in points #2 and #3, above), noting we don’t discuss properly the idea of two different modes of PCM recruitment. These are all good points, largely addressed in our response to points #2 and #3, above. We now discuss much more prominently the two different modes of client protein recruitment throughout the manuscript.

4. As we now clarify, in all our experiments we use centrosome separation and nuclear envelope breakdown (NEB) to define the start and end of S-phase, respectively.

5. The Reviewer quotes the landmark Woodruff paper (Cell, 2017) as showing that the ability to concentrate client proteins (including ZYG-9, the worm homologue of Msps) is an intrinsic property of the PCM scaffold, so how do we explain that Msps departs prior to NEB while Cnn continues to accumulate? It is indeed a striking observation of our study that all PCM client proteins (not just Msps) start to leave the centrosome prior to NEB, even as Cnn levels continue to accumulate. Our hypothesis is that this ‘leaving’ event is triggered by a threshold level of Cdk/Cyclin activity—explaining why these client proteins all start to leave the PCM at the same time (just prior to NEB) irrespective of nuclear cycle length. This is not incompatible with the Woodruff paper, which did not attempt to reconstitute any potential regulation by Cdk/Cyclins in their in vitro studies.

6. The Reviewer questions why Spd-2 that cannot be phosphorylated by Cdk/Cyclins (Spd-2-Cdk20A) accumulates abnormally at centrosomes in late S-phase, yet γ-tubulin (which is recruited by Spd-2) seems to leave centrosomes more slowly in the presence of the mutant protein. As we now explain more clearly, there is no contradiction here. Spd-2-Cdk20A accumulates to abnormally high levels in late-S-phase/early mitosis (Figure 5C), and this reduces the γ-tubulin dissociation rate, as we would predict (Figure 7B, right most graph). It does not “prevent” dissociation, however, (as the Reviewer seems to suggest it should?), but this is probably because these experiments have to be performed in the presence of large amounts of the WT Spd-2 (Figure 5A).

7. The referencing error has been corrected.

8. The Reviewer asks why in Figure 1 not all of the centrosome proteins could be followed for the full time period (as we mention in the legend, but do not explain). There are different reasons for different proteins: (1) Polo cannot be followed in mitosis as it binds to the kinetochores, making it impossible to accurately track centrosomes (so the data for mitosis is missing for Polo); (2) Cnn exhibits extensive flaring at the end of mitosis/early S-phase (Megraw et al., JCS, 1999), so we cannot track individual separating centrosomes labelled with NG-Cnn in early S-phase until they have moved sufficiently far-apart (so the early S-phase time-points are missing for Cnn); (3) In addition, several of the client proteins bind to the mitotic spindle, so although we can still track and measure the centrosomes in late mitosis in the graphs, we don’t show pictures of these late mitosis centrosomes in the montage in Figure 1A as the images look a bit odd. We now explain these reasons in the Materials and Methods.

9. We now indicate that nuclear cycle 12 (NC12) is being analysed in Figures 4-8.

10. The reviewer questions why we don’t show the decrease rate for γ-tubulin in Figure 6 (the Spd-2 and Cnn half-dose experiments), when we do show it in Figure 7 (the Spd-2 and Cnn Cdk-mutant experiments), suspecting that it is slowed in both cases. The reviewer is correct and we now show this data for both sets of experiments.

11. We have corrected the labelling error in Figure S1.

12. The Reviewer suggest moving some of the data from the main Figures, and the entirety of Figures 2 and 3 to the Supplemental Information. We understand this point, and agree that the amount of data presented in Figures 1-3 is somewhat overwhelming. We have played around with the Figures a lot—in particular trying to show a few examples of the data and moving the rest to Supplementary—but it is hard to pick a “typical” example, and the power of comparing the behaviour of so many different centrosome proteins is somewhat lost. We have tidied up several Figures and, as a compromise, we keep Figure 2 (now Figure 3) in the main text, but have moved Figure 3 to Supplementary (now Figure S5).

13. The Reviewer suggests that we should repeat the analysis of Spd-2, Polo and Cnn dynamics that we show here, as we already presented this data in a previous publication (Wong et al., EMBO. J, 2022). We understand this point, but feel this would be a less accurate comparison, as essentially all of the data shown in Figure 1 was obtained several years ago during a contiguous ~6month period. Since then, the lasers and software on our microscope system have been updated, so it would probably be less fair of a comparison to obtain new data for a subset of these proteins (and it seems overkill to perform the entire analysis again). We clearly state that this data has been presented previously, so we hope the Reviewer will agree that it is acceptable to present it again here so readers can more easily compare the data.

Reviewer #3:

This Reviewer is broadly supportive of the manuscript, but to publish in a prestigious journal they think additional experimental evidence will be required to support our hypothesis.

The Reviewer notes that our only evidence that Cdk/Cyclins directly phosphorylate Spd-2 comes from our analysis of the Spd-2-Cdk20A mutant, as the effect of reducing Cyclin B dosage on WT Spd-2 behaviour is very modest. They request that we analyse the behaviour of a Spd-2-Cdk20E phospho-mimicking mutant. The effect of halving the dose of Cyclin B on Spd-2 behaviour is modest, but this is what we would predict as all we are doing in this experiment is slowing S-phase by ~15%, so Spd-2 should accumulate at centrosomes for a slightly longer time and to a slightly higher level (as we observe, Figure 5E). A great advantage of the early fly embryo system is that we can compare the behaviour of many hundreds of centrosomes, so even subtle differences like this are usually meaningful. To illustrate this point, we have now repeated the Spd-2 analysis in WT and CycB1/2 embryos (but now using a CRISPR/Cas9 Spd-2-NG knock-in line) and we see the same subtle differences (Figure S9). In addition, as requested, we have now analysed the behaviour of a Spd-2Cdk20E mutant protein using an mRNA injection assay (as it would have taken too long to generate and test new transgenic lines). In this assay we injected embryos with mRNA encoding either WT Spd-2-GFP, Spd-2-Cdk20A-GFP or Spd-2-Cdk20E-GFP. The mRNA is quickly translated, and we computationally measured the fluorescence intensity of the centrosomes in mid-S-phase (i.e. at the Spd-2 peak) (Figure S8). This analysis confirms that Cdk20A accumulates to slightly higher levels, and reveals that Cdk20E accumulates to slightly lower levels, than the WT protein. Together, these new experiments strongly support our original conclusions.

The Reviewer notes that we propose that the CCO initially promotes centrosome growth by stimulating Polo recruitment to centrosomes, but states that we only provide indirect evidence for this by showing that centrosomal Polo levels are strongly reduced in Cyclin B half-dose embryos. They suggest we determine Spd-2 levels in Polo half-dose embryos, and/or the centrosome levels of mutant forms of Spd-2 that cannot be phosphorylated by Polo. We believe the Cyclin B half-dose experiment provide direct support for our hypothesis that Cdk/Cyclin activity influences Polo recruitment (Figure 8), although, clearly, we have not identified the mechanism. We do, however, suggest a plausible mechanism: Ana1 and Spd-2 are largely responsible for recruiting Polo to centrosomes, and we have previously shown that several of the potential phosphorylation sites in these proteins that help recruit Polo to centrosomes are Cdk/Cyclin or Polo phosphorylation sites (Alvarez-Rodrigo et al., eLife, 2020 and JCS, 2021; Wong et al., EMBO J., 2022). We are currently testing this hypothesis, but progress is slow as it is clear that multiple sites in both proteins can influence this process.

As the Reviewer requests, we have now also examined how Spd-2 and Cnn behave in Polo half-dose embryos (Reviewer Figure 2, attached to this letter). As we describe in the Figure legend, this data is informative, but is complicated. With relatively minor, but mechanistically important, tweaks to our previous mathematical modelling we can explain these behaviours, but introducing such a significant mathematical modelling element would be beyond the scope of this paper. As described above, these findings will form the basis of a follow-up paper that is more mathematically oriented.

It is a great idea to look at mutant forms of Spd-2 that cannot be phosphorylated by Polo, but the consensus Polo phosphorylation site (N/D/E-X-S, with the N/D/E at -2 and the S at 0 being preferences, rather than a strict rule) is less well-defined than the consensus Cdk/Cyclin phosphorylation site (where the Pro at -1 is essentially invariant). Thus, we cannot accurately predict which sites would need to be mutated to generate such a mutant.

The Reviewer requests that we analyse the behaviour of TACC in embryos expressing the Spd-2-Cdk20A and Cnn-Cdk6A (as we do in Figure 7 for γ-tubulin). This is a reasonable request, but we prefer not to show this data as we have recently identified an interesting interaction between TACC, Spd-2 and Aurora A that will be the subject of another paper we hope to submit shortly. This data is hard to interpret without explaining these interactions properly, which is beyond the scope of the current manuscript.

We hope the Reviewers will agree that these changes have improved the manuscript substantially, and that it is now suitable for publication. We would like to thank them again for taking the time to read this rather complicated paper so thoroughly.

We look forward to hearing from you.

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 investigated growth control of PCM at the mitotic centrosomes in late stages of the Drosophila syncytial embryos. They observed that mitotic centrosomes reach to the correct sizes through 13 rounds of nuclear division by reciprocally slowing their growth rate and increasing their growth period. They assumed that the Cdk/Cyclin cell cycle oscillator (CCO) is a main controller, based on their previous works (Aydogan et al., 2018, 2020; 2022). They determined the recruitment dynamics of the key mitotic PCM scaffolding proteins (Spd-2, Polo and Cnn) and PCM-client proteins (γ-tubulin, Msps, TACC, GFP, Grip75, Grip128 and Aurora A) in living embryos, and proposed that moderate levels of the CCO activity promote centrosome growth by stimulating Polo recruitment to centrosomes, while higher levels of activity subsequently inhibit centrosome growth by phosphorylating centrosome proteins, such as Spd-2, to decrease their centrosome recruitment and/or maintenance as the embryos enter mitosis.

Experiments were cleverly designed and carefully executed. The results are nicely presented, the manuscript is clearly written, and their proposal draws a strong attention. However, in order to publish the manuscript in a prestigious journal, the authors may provide additional experimental evidence to support their proposal.

• It is very significant that the centrosome levels of Spd-2-Cdk20A-NG is stronger than Spd-2-NG throughout the cell cycle (Figure 5B,C). However, this is only an experimental evidence to support that Cdk/Cyclins directly phosphorylate Spd-2 in the run-up to mitosis to help reduce Spd-2's centrosome recruitment and/or maintenance. As the authors confessed, recruitment of Spd-2-NG to the centrosomes in CycB1/2 embryos (Figure 5D,E) may be moderate or not significant at least in this reviewer's eyes. It is worth to perform the same experiments with a phospho-mimetic Spd2-Cdk20E-NG mutant.
• The authors proposed that moderate levels of CCO activity promote centrosome growth by stimulating Polo recruitment to centrosomes. They provided an indirect evidence that centrosome levels of polo were strongly reduced in CycB1/2 embryos (Figure 4E,F). It is worth to determine the centrosome levels of Spd-2 in the Polo1/2 embryos and/or the centrosome levels of Polo phospho-resistant Spd-2 (Spd-2-Polo#A-NG).
• TACC may be an ideal PCM-client protein, apart from its importance in spindle formation in comparison to γ-tubulin (Figure 4C,D). Therefore, it is worth to perform the Figure 7 experiments with TACC.

Significance

Experiments were cleverly designed and carefully executed. The results are nicely presented, the manuscript is clearly written, and their proposal draws a strong attention. However, in order to publish the manuscript in a prestigious journal, the authors may provide additional experimental evidence to support their proposal.

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

Control of organelle size has been an active field of research for many years for a large variety of cellular structures and in a range of experimental models. Here, Jordan Raff and colleagues examine the mechanisms underlying centrosome (PCM) size control in Drosophila syncytial embryos, building on their previous work (Wong, EMBOJ 2022) to propose a role for CDK in both promoting (at intermediate levels) and inhibiting (at high levels) PCM expansion.

I found this a difficult manuscript to review, not only because the subject matter is complicated, but so is the writing. Having read and re-read the manuscript some clarity eventually emerges, but it shouldn't be that inaccessible. As for the authors' model I find it intriguing, but not fully supported by the data currently presented.

Major points

1. Central to the authors' model is the proposed dual function of Cdk (or CCO in the authors' terminology) in both promoting and inhibiting centrosomal protein accumulation. This the authors test by reducing the gene dosage of cyclin B and using putatively non-phosphorylatable versions of Spd-2 and Cnn. Both approaches to me appear quite problematic. The latter perturbation is hard to interpret given that whether these are indeed Cdk phosphosites that they have mutated is unknown and there are plenty of other possibilities how this might perturb protein function, as the authors' lack of success doing the same for gamma-tubulin illustrates. The former perturbation also lacks context. Does reducing cyclin B gene dosage reduce peak CDK activity or does it merely take longer to reach the same maximum, as appears to occur naturally as the cell cycle slows between embryonic cycles 11 and 13 (Edgar, Genes Dev 1994)? A more direct way to test their model would be to arrest the embryo in S phase (which in their model should lead to indefinite growth) or mitosis using suitable drugs/genetic perturbations. Is this not feasible in the fly system?
2. Similarly critical is that centrosomal protein accumulation is accurately measured. I am not entirely convinced that this is so. If one takes their estimations of centrosome size at face value, then the space occupied by gamma-tubulin (slightly over 1 µm2 peak area according to Fig. S3) is significantly smaller than that occupied by Grip128 (4µm2). How is this possible if these form part of the same gamma-tubulin complex? This likely reflects the fact that the dynamics of many proteins is being assessed using transgenic reporters under the control of heterologous regulatory sequences (not all of which are fully functional, eg Polo), which could result in wildly inappropriate centrosomal protein levels. It may then not be a coincidence that the centrosomal domain for Grip128 (endogenously tagged) is larger than that for gamma-tubulin (transgene).
3. Another concern is that centrosome size and integated signal intensity do not always match, as demonstrated by Grip71 (increasing as expected during centrosome maturation in cycle 13 based on fluorescence intensity but not area
4. compare Figs. 1B and S1). A potential reason for this is that proteins are not uniformly distributed within centrosomes. For example, Polo and Spd2 are highly concentrated at centrioles. This impacts the ability to accurately measure protein accumulation based on 2D projections. Such inaccuracies likely will not affect estimation of when peak protein accumulation occurs, but may explain apparent differences in the kinetics of recruitment/dissociation of different components. Thus, the differences in the shape of the PCM client growth curves compared to those of Polo and Spd-2 (p6) may simply reflect the centriole concentration of the latter.

Other points<br /> 4. In C. elegans much of Polo at centrosomes is apparently inactive, particularly in the vicinity of centrioles (Cabral, Dev Cell 2019). Knowing whether this is also the case in flies would seem like important information to have, particularly when comparing signal intensities across the cell cycle.<br /> 5. Is it clear that there is less Spd2/Cnn at centrosomes in Spd-2/Cnn 1/2 gene dosage embryos, as the authors assume?<br /> 6. Would it not be relevant to also examine Polo 1/2 dosage embryos?<br /> 7. Based on the authors model, Cdk phosphorylation first drives PCM accumulation, then at higher levels inhibits. Yet, their non-phosphorylatable Spd2 mutant exhibits not only a delayed decline in centrosomal levels, but also higher initial levels (Fig. 5B). If Cdk initially promotes Spd2 activity what is their explanation for this?

Minor/discussion points

1. p4 "In typical somatic cells the two mitotic centrosomes need to grow to approximately the same size, as mitotic centrosome size asymmetry can lead to asymmetric spindle assembly and so to defective chromosome segregation. How centrosome growth is regulated in somatic cells is unclear, but in early C. elegans embryos, mitotic centrosome size appears to be set by a limiting pool of the PCM-scaffolding protein SPD-2."<br /> The authors here conflate absolute and relative size. Relative size matters to avoid spindle asymmetries, and centriole involvement in PCM recruitment helps to prevent this (Zwicker et al., PNAS 2014). Absolute size, which is what the authors are concerned with in this manuscript, may be important for spindle scaling, but this is not the same thing.
2. p5 "The centrosomal levels of Polo, Spd-2 and Cnn all started to increase at the start of S-phase, but whereas Cnn levels continued to rise and/or plateau as the embryos entered mitosis, the centrosomal levels of Polo and Spd-2 started to decrease before the entry into mitosis (Wong et al, 2021) (Figure 1A,B). Thus, the components of the mitotic PCM scaffold exhibit different growth kinetics, making it hard to use these proteins to define centrosome "size" at any particular point in the cell cycle."<br /> It is misleading and confusing for the reader to describe Polo and Spd2 as scaffold proteins as opposed to regulators of scaffold assembly. Presently Cnn is the only PCM protein demonstrated to have self-assembly/scaffolding properties based on the authors' own work (conduit, Dev Cell 2014; Feng, Cell 2017). There is little evidence that Polo and Spd2 form anything other than a nucleus for PCM growth.
3. p7 "The centrosomal levels of Grip71, Grip75, Grip128, and Aurora A tended to increase steadily through most of NC13, whereas TACC, Msps and γ-tubulin exhibited a noticeable increase in their recruitment rate towards the end of S-phase, shortly before their recruitment levels peaked (compare NC13 graphs in Fig. 1B). This difference was also obvious if we used centrosome area as a measure of centrosome size (Fig. S1). We conclude that PCM client proteins can be recruited to centrosomes in at least two different ways."<br /> As discussed above apparent differences in kinetics may reflect limitations in the way protein accumulation is measured. It is hard to conceive of a reason why the Grips would display a different mode of protein accumulation from gamma-tubulin, nor is the idea of two different modes of protein accumulation picked up again later in the manuscript.
4. Since the authors mention that the duration of S phase increases between cycles 11 and 13 (p9), are there any measures for the timing of the beginning/end of S phase in each cycle?
5. One of the main findings in the landmark Woodruff paper from 2017 Cell paper was that PCM scaffold polymer could dynamically concentrate client proteins in the absence of any other factors, to an extent similar to that observed in vivo. This list did not include gamma-tubulin, which was later shown to require PLK1 phosphorylation of SPD-5 (Ohta, JCB 2021). However, it did include ZYG-9, the C. elegans ortholog of Msps. If client protein accumulation is an intrinsic property of the PCM scaffold, how do the authors explain that Msps departs prior to NEBD while Cnn continues to accumulate?
6. p13 "The expression of the mutant proteins did not appear to dramatically perturb the centrosomal recruitment of γ-tubulin-GFP, except that the rate at which γ-tubulin-GFP left the centrosome as the embryos entered mitosis was reduced in both mutants compared to WT (Figure 7). This phenotype was subtle, but it was statistically significant, and it seems likely that the presence of large amounts of WT Spd-2 and Cnn in the mutant embryos (Figure 5A,F) would help to mask the potential severity of this phenotype."<br /> This does not quite make sense. Fig. 5 shows that Spd2 dissociation is significantly slowed in the mutant condition. If Spd2 drives gamma-tubulin accumulation (as Fig 6 shows), then the continued presence of Spd2 should prevent dissociation. Yet it apparently does not. Why?

Other

1. p3 and following. The reference for the authors' prior work on PCM recruitment (Wong et al, 2021) should probably be for the final, published article in EMBO J, not the 2021 preprint.
2. Fig. 1. legend "Note that for technical reasons not all of the centrosome proteins could be followed for the full time period." Why not?
3. Figs 4-6. Which cycle is being assessed here?
4. Fig 6. Not plotted here is the rate of dissociation of gamma-tubulin, unlike eg in Fig 7. It is notable that both accumulation and dissociation appear to be slowed in the Spd2 1/2 gene dosage condition.
5. Fig S1B. Some of the graphs in this figure are not labeled (based on Fig.1 presumably gamma-tubulin and Msps).
6. Some of the data in the main figures, including the entirety of Figs. 2 and 3, could be moved to Supplemental to present a more crisp and accessible manuscript.
7. While I sympathize with the authors needing to repeat entire sets of experiments I am not entirely sure it is appropriate to recycle entire sets of data from a previous publication of theirs (Cnn, Spd-2 and Polo recruitment kinetics, reproduced from Wong et al., EMBOJ 2022), since this manuscript is largely concerned with apparent differences between the kinetics of those components and the PCM client proteins now being analysed.

Significance

Control of organelle size has been an active field of research for many years for a large variety of cellular structures and in a range of experimental models. Here, Jordan Raff and colleagues examine the mechanisms underlying centrosome (PCM) size control in Drosophila syncytial embryos, building on their previous work (Wong, EMBOJ 2022) to propose a role for CDK in both promoting (at intermediate levels) and inhibiting (at high levels) PCM expansion.

I found this a difficult manuscript to review, not only because the subject matter is complicated, but so is the writing. Having read and re-read the manuscript some clarity eventually emerges, but it shouldn't be that inaccessible. As for the authors' model I find it intriguing, but not fully supported by the data currently presented.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Wong et al. investigates how cells regulate the increase in the size of the centrosomes (more specifically the size of the pericentriolar material or PCM) that occurs during preparation for mitosis. They use the Drosophila syncytial embryo as a model, focusing on nuclear cycles 11-13, during which cell cycle progression gradually slows. The authors find that centrosomes grow to a consistent size at each cycle by adjusting to the slowed cell cycle, reducing the growth rate and increasing the growth period. This adjustment is proposed to be regulated by the Cdk/Cyclin cell cycle oscillator. Curiously, Cdk/Cyclin activity seems to both promote and inhibit the increase in centrosome size, depending on whether its activity is moderate or very high, respectively. Both effects are proposed to depend on the phosphorylation of centrosome proteins by Cdk/Cyclin.

1. While being comprehensive in the number and type of markers that are being analyzed, there is no analysis of the centrosome's MTOC activity. In my opinion this is missing since centrosome size alone is not necessarily indicative of its MTOC activity, but MTOC activity is what ultimately matters for its role during mitosis. For example, it was observed that centrosome size declines already before mitotic entry, but it is possible that centrosome MTOC activity does not (similar to differences in the timing of the decline of PCM scaffold vs PCM client proteins). While not strictly related to size control, centrosome activity is biologically more relevant than solely size. I would consider it optional, if the authors decide to talk only about centrosome size, but then it should be made clear that size here may not be the most relevant factor.
2. The authors say that during NC13 PCM client proteins can be recruited in "at least two different ways" (p. 7), including a way (rapid increase before peak) that does not resemble PCM scaffold recruitment (steady increase during NC13). How can these two different ways and kinetics be determined by the same Cdk/Cyclin oscillator?
3. I am puzzled by the conclusion that Cdk/Cyclin directly phosphorylates Spd-2 or Cnn at the sites used for mutagenesis. This cannot be concluded based on the presented data.
4. Fig. 6: Doesn't the data show that Cnn does not affect the initial rate of g-tub recruitment, but only the later rapid recruitment shortly before mitosis? In contrast Spd-2 seems to affect the initial phase. This should be described more precisely. Again, I am wondering how this is compatible with direct regulation by a single oscillator, as suggested by the authors (see also point 2 above.
5. I don't find the proposed model very convincing and not fully supported by the presented data.<br /> First, the recruitment kinetics of different centrosome proteins are not all the same, arguing against a simple relationship based on phosphorylation by Cdk/Cyclin. For example, kinases (or phosphatases) may be recruited (or displaced) by Cdk/Cylclin at the centrosome and then locally regulate binding or maintenance of certain centrosome proteins. This could explain profiles that do not display a steady change over time, as would be expected by direct regulation by Cdk/Cyclin.<br /> Second, it is not clear from the description in the text or from Fig. 8 how moderate Cdk/Cyclin activity can promote recruitment and high activity induce loss of proteins at centrosomes. In fact, the experiments with Spd-2 and Cnn phospho-mutants suggest that phosphorylations at the mutated sites also reduce centrosome binding during S phase (at moderate activity) and not only shortly before mitosis (at high activity), since alanine mutants of both Spd-2 and Cnn are increased at centrosomes also during S phase. The model seems to ignore this observation. If these sites are already phosphorylated to decrease centrosome binding in S phase, then what triggers the rapid decrease shortly before mitosis?
6. Can the authors identify and mutate CdK/Cyclin dependent phospho-sites in centrosome proteins that promote centrosome recruitment at moderate Cdk/Cyclin activity? As an alternative to the "protein availability" model for regulation of centrosome size, the proposed model needs to explain how a steadily increasing activity (Cdk/Cyclin) can first induce growth and then turn growth off, when the desired size is reached. This is obvious in the "protein availability" model, where the available protein steadily decreases as centrosomes grow, but this is not at all obvious for an oscillator that behaves in the opposite way during the same period and that can only phosphorylate sites that decrease centrosome binding.

Minor:

1. The authors observe differences in the intensity profiles for different subunits of the gamma-tubulin complex. How do they explain this? Are they not in the same complex? The authors should mention and comment on this.
2. The authors refer at various points in the manuscript to an "inverse-linear" relationship between S phase length and centrosome growth rate, but according to the graphs the rate does not change linearly.

Significance

This is an interesting manuscript that reaches somewhat different conclusions regarding centrosome size control when compared to previous studies in other organisms. In particular, work in C. elegans has proposed that centrosome growth regulation is controlled by the limited cytoplasmic availability of PCM building blocks, whereas the current study proposes a different model based on the activity of a cell cycle oscillator. The model system and approaches are well presented and the data is of good quality. The authors monitor a large number of centrosome markers, each with detailed quantifications of intensity and distribution over time during the different cycles. They also employ two different ways of quantifying centrosome size with similar results, making their quantifications more robust. While the authors include phospho-mutants in their analyses that presumably cannot be phosphorylated by Cdk/Cyclin, the study is largely descriptive. Still, the authors present interesting observations and propose the "oscillator model" an alternative to the "limited availability model" for the regulation of centrosome size, and perhaps that of other organelles. Assuming the authors can clarify inconsistencies and/or provide additional data to support the proposed model, this could be an important finding that expands cell biologists' understanding of organellar size control.

I have expertise in centrosome biology and the role of centrosomes as MTOCs, as well as more general expertise regarding the function of the microtubule cytoskeleton in cell division and differentiation.

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):

In their manuscript „Live-cell super-resolution nanoscopy reveals modulation of cristae<br /> dynamics in bioenergetically compromised mitochondria", Golombek et al. tested the effects of different mitochondrial toxins on cristae dynamics. The main focus of their work lies on live STED imaging, which they use to visualize cristae merging and splitting. They found swelling of mitochondria and reduced cristae density in response to most toxins, but cristae dynamics remained largely unaffected. Depletion of the membrane potential by administration of CCCP increased cyristae dynamics, while inhibition of ANT had a negative effect on cristae dynamics at least in a subset of mitochondria.

1. The authors state that the used concentrations of mitochondrial toxins commonly result in a change in oxygen consumption. While this is believable, it is not guaranteed that the specific chemicals used for the experiments were working properly (freeze/thawing or simply incorrect storage or aliquotation may have an effect on the compounds). This is even more important in the case of results where no significant change after the administration of the toxins is seen. In Figure 5, the authors report no change in membrane potential after oligomycin administration, this is unexpected. I therefore suggest to include a supplementary figure, in which the functionality of the compounds is verified. This could be done by respiratory measurements (e.g. Seahorse). A Mito Stress Test was performed for Figure 6, but this was done using the Seahorse kit chemicals, which were probably different from the chemicals used in the microscopy experiments.

Response: We appreciate the valid concerns of the reviewer in this point.

A) In order to show the functionality of compounds which were used for performing our experiments including STED imaging, we now performed respiratory measurements employing the concentrations of mitochondrial toxins (Oligomycin A, CCCP, rotenone/antimycin A) which were used during imaging conditions as well as commercially available mitochondrial toxins (Oligomycin A, FCCP, rotenone/antimycin A) with respective concentrations used as a standard for the Mito stress Kit. The new figures are included in Fig S1A & B. HeLa cells treated with seahorse compounds or those used during imaging conditions showed similar results including basal, maximal and spare respiratory capacity. Further, in order to overcome the inefficiency of mitochondrial toxins employed, due to freeze/thaw cycles, we used fresh aliquots (stored at -20°C) as a general strategy. This is clearly observed by a drastic reduction of ΔΨm upon treating HeLa cells with CCCP, antimycin A as well as rotenone (Fig S6A & B). A reduction of mitochondrial ATP levels was also observed upon employing rotenone, antimycin A and oligomycin A confirming that active mitochondrial toxins were used. These experiments demonstrate that the mitochondrial toxins employed throughout our manuscript are functional as expected.

New Figure S1A & B

B) The Fig 6 (now Fig 5 due to Reviewer # 2, Point 7) respirometry experiments which initially employed seahorse compounds and BKA has now been replaced with new experiments where we used mitochondrial toxins similar to STED imaging. Needless, to say, the results are similar to what were observed with seahorse compounds. The new figures are replaced in Fig 5A & 5B.

New Figure 5A & B

C) Oligomycin A inhibits ATP synthase which results in decreased ATP synthesis as observed (Fig 4A & B). Further, oligomycin A is expected to hyperpolarise mitochondria (2). In Fig S6, despite some cells having more ΔΨm, there was no overall significant change when compared to untreated cells. Previous publications also show that there is no significant difference in ΔΨm upon treatment with oligomycin (1) demonstrating that the ΔΨm depends on the concentration of oligomycin, treatment time and cell type.

1. Figure 1 would benefit from a more detailed description of merging/splitting events. Maybe a cartoon plus a zoomed in image of an exemplary event?

Response: Thank you for the suggestion. In order to clearly explain/simplify the understanding of cristae merging and splitting events, we added a cartoon in Fig 1B. The green and magenta arrows show sites of imminent merging and splitting with the green and magenta asterisks representing them respectively in the subsequent frames. The zoomed in images in Fig1A (leftmost panel) are shown to the right as time-lapse images.

New Figure 1B

1. Could the reduced cristae density be an effect of mitochondrial swelling? It is curious that all toxins appear to have the same effect on mitochondrial architecture. What is the fait of an enlarged mitochondrion over time? Mitophagy? And does the percentage of enlarged mitochondria change with increasing treatment time?

Response: Thank you for the comment.

A) We agree that the reduced cristae density is due to mitochondrial swelling. We added the relevant text in the results section ‘Cristae structure is altered in a subset of mammalian cells treated with mitochondrial toxins’. Treatment of HeLa cells, with all the mitochondrial toxins mentioned, uniformly result around 50 % of mitochondria undergoing enlargement (Fig 2B). In enlarged mitochondria where the mitochondrial width is ≥ 650 nm, there is no change in cristae area occupied per mitochondria (Fig S3C & D) and as a result reduced cristae density (Fig 2H). Therefore, it indicates that reduced cristae density occurs due to mitochondrial enlargement.

Figure 2B-F

Figure S3C and D

B) In order to address the fate of mitochondria with increasing time upon treatment with various mitochondrial toxins, we treated the HeLa cells for 4 hrs with mitochondrial toxins. Untreated cells maintained normal mitochondrial morphology while cells treated with various mitochondrial toxins displayed fragmented and swollen mitochondrial morphology. The new Fig S5 is included in the supplementary. Cristae morphology was abnormal displaying interconnected cristae in swollen mitochondria. Since mitochondrial fragmentation is already observed at 4 hours and accompanied by interconnected cristae, the number of cristae merging and splitting were severely reduced.

Our imaging performed within 30 mins of addition of respective toxins overcomes the additional aberrancy of mitochondrial fragmentation which would not allow a reliable analysis of cristae dynamics as too few cristae would be visible within one mitochondrion.

New Figure S5

1. Figure 4C: How was the mitochondrial width determined in the LSM images? Especially in the perinuclear area it will be difficult to determine this parameter without the super-resolution provided by STED. Was this parameter determined manually for selected mitochondria? In the methods part it says that only a maximum of two mitochondria per cell were analyzed. How were these chosen? Was the process blinded?

Response: Thank you for the comment. We could imagine the reason for the ambiguity in understanding.

A) For LSM confocal images involving FRET-based microscopy to determine the ATP levels, we calculated the cell population as belonging to either normal or enlarged category. The confocal images of HeLa cells displayed clear separation of mitochondria even in the perinuclear area (representative images are shown in Fig 4A) and thus it was possible to measure the width of individual mitochondria. The methods section ‘FRET-based microscopy to measure ATP levels’ describes that ‘the cut off for swollen mitochondria was set to 650 nm in congruence with STED SR nanoscopy. If 85% of the mitochondrial population featured enlarged mitochondria, the cells were designated as swollen. Similarly, if 85% of the mitochondrial population featured mitochondria whose width was less than 650 nm, the cell was considered as having normal mitochondria’.

Figure 4A

B) The cristae morphology of various mitochondria is fairly uniform in individual cells. Thus, the mitochondria are representative of the individual cells. Therefore, in order to increase the coverage of various cells, we considered a maximum of two mitochondria from each cell which were randomly chosen. This part is modified in the methods section ‘Quantification of various parameters related to cristae morphology’ to make it clear. Thus, while the quantification of various parameters including dynamics involved individual mitochondria, various cells were classified as belonging to normal or enlarged category while measuring ATP levels.

1. What is the average size of all mitochondria per cell? Is this addressed in Figure 2B or are only analyzed mitochondria included? Please carify. Were the mitochondria chosen for analysis representative for the respective cell?

Response: The data obtained by super-resolution imaging of mitochondria is used for quantifying cristae dynamics which is a very challenging and time-consuming method done in a blind-manner. As mentioned in response 4B, the cristae morphology is fairly uniform in individual cells, therefore, we only included the mitochondria which were analysed for various cristae parameters in our analysis which are really huge data-sets already. Thus, the average size of individual mitochondria per cell are not represented while analysing images obtained with STED SR imaging. Please also check response 4B.

1. explain the mt-Go-AT team2, what is GFP (green fluorescent protein) and OTP (?)

Response: GFP is Green Fluorescent Protein and OFP is Orange Fluorescent protein and included in the revised text.

1. the graphs show in principle, e.g. Fig.1B, 3B-E show events/mitochondrion as far as I understand, not per cristae.

Response: Thank you for pointing this out. It is actually the average number of events per cristae per mitochondria. We have changed the Y-axis to events/cristae/mito in Fig 1C (previous 1B), Fig 3B-E and wherever applicable for other figures throughout the manuscript.

Figure 1C

Figure 3B-E

1. I would recommend changing the legend of the x-axis of Fig.2B-F to mito-width (y-axis could be probability density function, PDF).

Response: We have now changed the X-Axis to mito width (originally width) in Fig 2B-F. The Y-axis are still retained as percentage mitochondria where cells treated with few mitochondrial toxins do not show a gaussian distribution of mitochondrial width.

Figure 2B-F

Referees cross-commenting

both expert opinions address similar concerns and therefore a revision should be requested

Reviewer #1 (Significance):

The study is thorough and the experiments and results are well described. Overall, however, it remains a descriptive study and does not provide mechanisms. There is also no discussion of how MMP-dependent proteins, such as Opa1, which was previously studied by the Reichert group, might be affected. For swelling mechanisms, the opening of the mitochondrial permeability transition pore was discussed. This could be tested using inhibitors, but perhaps not within the scope of this publication. Nevertheless, the information provided by the study is of interest to the bioenergetics community and should be made available.

Response: Thank you for the overall inputs.

We tested the processing of OPA1 forms and found that after 30 mins, only CCCP treatment led to the processing of long isoforms to short forms (Fig S6C). We now included in the discussion that it is possible that short OPA1-forms are correlative to increased cristae merging as well as splitting events upon treatment with CCCP.

New Figure S6C

Reviewer #2 (Evidence, reproducibility and clarity):

Summary:<br /> The authors investigated cristae merging and splitting events using ultra-resolution STED. The goal was to test if cristae membrane remodeling is dependent on OXPHOS complexes, mitochondrial membrane potential (ΔΨm), and the ADP/ATP nucleotide translocator. To do this the authors utilized several mitochondrial toxins with known mechanisms of action. Interestingly, many changed overall cristae density but did not change the cristae remodeling events. Inhibition of ANT did change cristae morphology and cristae dynamics.

Major Concerns

1. Many conclusions and concepts need more clarification. For example, a major take home from the abstract is that various ETC inhibitors and protonophores reduce cristae density but not did not change cristae remodeling events. If cristae density is reduced, how can this occur without cristae remodeling events? Remodeling events need to be clearly defined in the introduction and abstract.

Response: Thank you for pointing out this lack of sharpness in our terminology which indeed can cause a misunderstanding. To avoid this, we have now included ‘changes in cristae morphology’ as well as ‘dynamic merging and splitting events of cristae’ under the broader term cristae remodelling. Thus, we had changed the wording ‘cristae remodeling’ to cristae dynamics in the abstract and wherever appropriate in the manuscript text.

The cristae morphology analysis showed no change in cristae area (Fig S3C) which was accompanied by mitochondrial enlargement. Therefore, cristae density was reduced. For the purpose of clarity, we added a sentence in the introduction section while giving a peek into our results that ‘cristae dynamic events are ongoing despite reduced cristae density’. In addition, we have now included in the results section the following statement: ‘Cristae membrane remodeling has been used to describe cristae dynamic events (i.e. cristae merging and splitting) as well as overall changes in cristae morphology within a single mitochondrion in this manuscript’.

Figure S3C and D

1. Other interpretations are also unclear such as how ETC inhibitors which reduce ATP levels did not impact cristate remodeling events, yet inhibiting ATP/ADP exchange did greatly impact this phenomenon. It seems likely that the inhibition of ANT has nothing to do with ATP/ADP exchange since most of the ETC inhibitors no doubt greatly impact overall ATP/ADP exchange. This interpretation needs clarification.

Response: We agree that further clarification is needed, in particular to explain why ATP/ADP exchange is actually ongoing even when OXPHOS inhibitors are applied and to explain why reduced ATP levels do not mean that there is no ATP/ADP exchange occuring. Treatment of HeLa cells with various mitochondrial toxins inhibiting the function of OXPHOS complexes leads to decreased ATP levels due to ongoing ATP consumption within the cell (Fig 4). One should also consider that two things can and do happen when most of these toxins are applied regarding ATP exchange. First, the ATPase can act in reverse mode which is a (partial) compensatory mechanism to restore ΔΨm and which will further decrease ATP levels (Note: not in the presence of oligomycin). Second, under these conditions ADP/ATP exchange is still ongoing in order to transport ATP derived from glycolysis in the cytosol to the mitochondrial matrix which also causes an (partial) compensatory increase in membrane potential. After ATP import ATP is hydrolysed to ADP for reverse proton pumping via the F1FO-ATPase or alternatively by the F1-part alone without proton pumping. In all these cases it is essential and possible to exchange ADP with ATP constantly. Therefore, the overall exchange of ADP and ATP is not necessarily grossly expected to be different when compared to untreated cells (due to compensatory glycolysis and subsequent ATP import and hydrolysis in the matrix). On the other hand, BKA treatment which clearly impairs the exchange of ADP and ATP will lead to a completely different situation compared to only treating with OXPHOS inhibitors. With BKA the mitochondrial matrix cannot anymore be resupplemented with ATP derived from glycolysis and metabolite flux is grossly hampered. Consistent with this a strong reduction in ΔΨm and oxygen consumption is accompanied with BKA treatment (Fig. 5AB & SFig 7F). Thus, w.r.t cristae dynamic events, in the time-frame we used for imaging, a reduction of ATP levels does not impede occurrence of cristae merging and splitting events while BKA treatment does (Fig S7). We discuss this indeed interesting and unexpected finding in the discussion section. We propose that rather ongoing metabolite flux (ATP/ADP exchange) is critical for maintaining cristae dynamics and blocking it is detrimental for it. We adapted the discussion in this direction to make it more clear.

Figure S7A, B and D

1. Why did the authors wait 30 min to image after the addition of mitochondrial toxins? I would have guessed there is a more rapid change in response to these inhibitors. Is there is a chance he authors missed the most dramatic events?

Response: Since we were inclined to observe early responses, cells were imaged within the first 30 mins after addition of the respective mitochondrial toxins (Please see methods ‘cell culture transfection and mitochondrial toxin treatment’). Thus, to answer this question we want to emphasize that we did not wait 30 minutes but we restricted our time frame of analysis to 30 min. Therefore, we think that we did not miss out on any rapid changes occurring early on. Regarding this point, Reviewer #1 (Query 3) asked for responses at a later time-point. Please read the Reviewer #1, response 3B.

1. How do these mitochondrial toxins that are known to cause mitochondrial swelling not induce changes in cristate density?

Response: Thank you for the question. Probably, there is a misunderstanding. In Fig S3E, we clearly show that as the mitochondrial width increases in cells after treatment with mitochondrial toxins, there is a clear decrease in cristae density. In fact, the reduced cristae density is observed exclusively in enlarged mitochondria. Figure S3E-I

5. It's interesting that inhibition of the ANT translocator by BKA treatment led to increased percentage of mitochondria with abnormal cristae morphology. It's accepted that inhibition of ANT profoundly reduces mitochondrial swelling. Do the authors have any data suggesting that abnormal cristae morphology actually is a mechanism for reducing cell death events such as permeability transition? Did the authors utilize cyclosporin A concomitantly with any of the mitochondrial toxins?

Response: This is a very interesting question! As the reviewer might be aware, there is evidence connecting cristae remodelling to induction of apoptosis (3). Cristae transitioned to a highly interconnected state after tBID treatment within minutes. However, it is unclear what is the contribution of cristae dynamics in this regard. Within 30 mins, there were no visual signs of cell death in our experiments as observed under a microscope. Hence, we did not use cyclosporin A in our experiments. In our opinion, this question will form part of a very interesting future study and is currently beyond the scope of this manuscript.

1. Are the authors confident in the data given many of the experiments utilized quantification of 10-20 mitochondria? How are you sure this sampling is sufficient for phenomenon being studied?

Response: Please see Reviewer 1, Response 4B. As pointed in the response to reviewer #1, the cristae morphology is fairly uniform in individual cells. Therefore, in order to maximise the cell population covered, we randomly used a maximum of two mitochondria from each cell. In addition, we included cristae analysis from at least three biological replicates in order to observe the reproducibility of the data. Taking these factors into consideration, we are confident that our results reflect a sufficient sample size. Further, we would like to point out while our group performs STED super-resolution imaging routinely, the quantification of cristae merging and splitting events done in a blind yet manual manner is a really laborious and time-consuming process. In the future, we are also looking to optimise this at least in a semi-automated manner.

1. Figure 4 and 5 merely confirm current dogma and don't really contribute to the overall conclusions and can be moved to supplemental data.

Response: We agree that Fig 5 is confirming to the current dogma. Therefore, we moved it to Fig S6. Regarding Fig 4, we would like to highlight that there is a decrease of ATP levels before mitochondria enlarge. Thus, we would like to retain it as part of the main figure.

1. It's interesting that BKA dose dependently decreased ATP-linked respiration and all doses limited maximal respiratory capacity. It would be interesting to know if the BKA normal vs. abnormal mitochondria have differential membrane potential?

Response: Thank you for the interesting question. Overall, BKA treatment leads to a significant decrease of ΔΨm in the whole cell population (Fig S7). Further, the abnormal cristae morphology is only seen in one-third of the population of mitochondria (Fig shown in Response 2). Thus, a drop in ΔΨm seems to be a very early response upon exposure to BKA and independent of cristae morphology. An ideal experiment to address this question would be to image cristae dynamics and ΔΨm using super-resolution imaging which is challenging according to the state-of-art and available chemicals.

Figure S7E and F

1. Overall, this is an interesting study and seems appropriately performed but the results and conclusions are unclear. More discussion should include physiological relevance and impact and how this data influences previous work. Some physiological perturbations beyond the mitochondrial toxins and or utilization of genetic models would strengthen the interpretation and overall impact.

Response: Thank you. We added an OPA1 blot showing the different L-OPA1 and S-OPA1. (Reviewer #1, response in significance section) where we observed that S-OPA1cleavage is selectively enhanced in CCCP-treated cells which could be correlated with enhanced cristae dynamics. We also included these results in the main text.

New Figure S6C

Referees cross-commenting

Yes, I conclude that given the significant overlap in reviwer comments and general need for clarification of concepts and data that a revision is in order.

Reviewer #2 (Significance):

Overall, a highly specialized study with audience limited to mitochondriacs. Although, I'll note tis is a hot area of study and there is high interest in the field. Some of the data interpretation is difficult to understand and overall more context is needed to explain the results, impact and relevance. Defining exactly what a cristae remodeling event is and how this differs from cristae density and how the two aren't directly connected is unclear.

Review by a mitochondrial biologist specializing in mitochondrial signaling and connection to physiology.

References:

1. Baker MJ, Lampe PA, Stojanovski D, Korwitz A, Anand R, et al. 2014. Stress-induced OMA1 activation and autocatalytic turnover regulate OPA1-dependent mitochondrial dynamics. EMBO J 33: 578-93
2. Farkas DL, Wei MD, Febbroriello P, Carson JH, Loew LM. 1989. Simultaneous imaging of cell and mitochondrial membrane potentials. Biophys J 56: 1053-69
3. Scorrano L, Ashiya M, Buttle K, Weiler S, Oakes SA, et al. 2002. A distinct pathway remodels mitochondrial cristae and mobilizes cytochrome c during apoptosis. Dev Cell 2: 55-67

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The authors investigated cristae merging and splitting events using ultra-resolution STED. The goal was to test if cristae membrane remodeling is dependent on OXPHOS complexes, mitochondrial membrane potential (ΔΨm), and the ADP/ATP nucleotide translocator. To do this the authors utilized several mitochondrial toxins with known mechanisms of action. Interestingly, many changed overall cristae density but did not change the cristae remodeling events. Inhibition of ANT did change cristae morphology and cristae dynamics.

Major Concerns

1. Many conclusions and concepts need more clarification. For example, a major take home from the abstract is that various ETC inhibitors and protonophores reduce cristae density but not did not change cristae remodeling events. If cristae density is reduced, how can this occur without cristae remodeling events? Remodeling events need to be clearly defined in the introduction and abstract.
2. Other interpretations are also unclear such as how ETC inhibitors which reduce ATP levels did not impact cristate remodeling events, yet inhibiting ATP/ADP exchange did greatly impact this phenomenon. It seems likely that the inhibition of ANT has nothing to do with ATP/ADP exchange since most of the ETC inhibitors no doubt greatly impact overall ATP/ADP exchange. This interpretation needs clarification.
3. Why did the authors wait 30 min to image after the addition of mitochondrial toxins? I would have guessed there is a more rapid change in response to these inhibitors. Is there is a chance he authors missed the most dramatic events?
4. How do these mitochondrial toxins that are known to cause mitochondrial swelling not induce changes in cristate density?
5. It's interesting that inhibition of the ANT translocator by BKA treatment led to increased percentage of mitochondria with abnormal cristae morphology. It's accepted that inhibition of ANT profoundly reduces mitochondrial swelling. Do the authors have any data suggesting that abnormal cristae morphology actually is a mechanism for reducing cell death events such as permeability transition? Did the authors utilize cyclosporin A concomitantly with any of the mitochondrial toxins?
6. Are the authors confident in the data given many of the experiments utilized quantification of 10-20 mitochondria? How are you sure this sampling is sufficient for phenomenon being studied?
7. Figure 4 and 5 merely confirm current dogma and don't really contribute to the overall conclusions and can be moved to supplemental data.
8. It's interesting that BKA dose dependently decreased ATP-linked respiration and all doses limited maximal respiratory capacity. It would be interesting to know if the BKA normal vs. abnormal mitochondria have differential membrane potential?
9. Overall, this is an interesting study and seems appropriately performed but the results and conclusions are unclear. More discussion should include physiological relevance and impact and how this data influences previous work. Some physiological perturbations beyond the mitochondrial toxins and or utilization of genetic models would strengthen the interpretation and overall impact.

Referees cross-commenting

Yes, I conclude that given the significant overlap in reviewer comments and general need for clarification of concepts and data that a revision is in order.

Significance

Overall, a highly specialized study with audience limited to mitochondriacs. Although, I'll note tis is a hot area of study and there is high interest in the field. Some of the data interpretation is difficult to understand and overall more context is needed to explain the results, impact and relevance. Defining exactly what a cristae remodeling event is and how this differs from cristae density and how the two aren't directly connected is unclear.

Review by a mitochondrial biologist specializing in mitochondrial signaling and connection to physiology.

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 their manuscript „Live-cell super-resolution nanoscopy reveals modulation of cristae<br /> dynamics in bioenergetically compromised mitochondria", Golombek et al. tested the effects of different mitochondrial toxins on cristae dynamics. The main focus of their work lies on live STED imaging, which they use to visualize cristae merging and splitting. They found swelling of mitochondria and reduced cristae density in response to most toxins, but cristae dynamics remained largely unaffected. Depletion of the membrane potential by administration of CCCP increased cyristae dynamics, while inhibition of ANT had a negative effect on cristae dynamics at least in a subset of mitochondria.

Major comments

• The authors state that the used concentrations of mitochondrial toxins commonly result in a change in oxygen consumption. While this is believable, it is not guaranteed that the specific chemicals used for the experiments were working properly (freeze/thawing or simply incorrect storage or aliquotation may have an effect on the compounds). This is even more important in the case of results where no significant change after the administration of the toxins is seen. In Figure 5, the authors report no change in membrane potential after oligomycin administration, this is unexpected. I therefore suggest to include a supplementary figure, in which the functionality of the compounds is verified. This could be done by respiratory measurements (e.g. Seahorse). A Mito Stress Test was performed for Figure 6, but this was done using the Seahorse kit chemicals, which were probably different from the chemicals used in the microscopy experiments.
• Figure 1 would benefit from a more detailed description of merging/splitting events. Maybe a cartoon plus a zoomed in image of an exemplary event?
• Could the reduced cristae density be an effect of mitochondrial swelling? It is curious that all toxins appear to have the same effect on mitochondrial architecture. What is the fait of an enlarged mitochondrion over time? Mitophagy? And does the percentage of enlarged mitochondria change with increasing treatment time?
• Figure 4C: How was the mitochondrial width determined in the LSM images? Especially in the perinuclear area it will be difficult to determine this parameter without the super-resolution provided by STED. Was this parameter determined manually for selected mitochondria? In the methods part it says that only a maximum of two mitochondria per cell were analyzed. How were these chosen? Was the process blinded?
• What is the average size of all mitochondria per cell? Is this addressed in Figure 2B or are only analyzed mitochondria included? Please carify. Were the mitochondria chosen for analysis representative for the respective cell?

Minor comments

1. explain the mt-Go-AT team2, what is GFP (green fluorescent protein) and OTP (?)
2. the graphs show in principle, e.g. Fig.1B, 3B-E show events/mitochondrion as far as I understand, not per cristae.
3. I would recommend changing the legend of the x-axis of Fig.2B-F to mito-width (y-axis could be probability density function, PDF).

Referees cross-commenting

both expert opinions address similar concerns and therefore a revision should be requested

Significance

The study is thorough and the experiments and results are well described. Overall, however, it remains a descriptive study and does not provide mechanisms. There is also no discussion of how MMP-dependent proteins, such as Opa1, which was previously studied by the Reichert group, might be affected. For swelling mechanisms, the opening of the mitochondrial permeability transition pore was discussed. This could be tested using inhibitors, but perhaps not within the scope of this publication. Nevertheless, the information provided by the study is of interest to the bioenergetics community and should be made available.

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):

Major comments<br /> In the paper "Microtubules under mechanical pressure can breach dense actin networks", the authors showed clear evidence that pressure plays an important role in microtubule breaching into dense actin networks using elegant in vitro reconstitution assays. They have argued that the pressure results from polymerization force of microtubules, which builds up when microtubules are immobilized in the opposite end of breaching, by the means of actin microtubule crosslinking factor Tau.

Authors answer:

We thank the reviewer for his/her positive comments on our manuscript.

It would definitely be interesting to see lack of breaching in the presence of crosslinking deficient Tau construct in order to rule out the off -target effect of Tau on microtubule and actin architecture which may possibly facilitate breaching.

Authors answer:

This is an interesting suggestion. Unfortunately, we do not have in hand such crosslinking deficient Tau construct. However, please note that we showed two independent ways to demonstrate the role of pressure. One is indeed by crosslinking microtubule to actin bundle with Tau, but the other is by blocking the two opposite ends of microtubules with two dense actin networks. So, we think our conclusion about the role of pressure is solid.

The authors have also observed microtubule breaching into dense actin networks in living cells. However, in Figure 1C, better cell/ image processing might have been chosen to increase the visibility of actin structures that microtubules encounter on their way to breaching. In Figure S1D, for example, the similar actin structures in lamellipodia are very nicely visible.

Authors answer:

We apologize but we don’t understand reviewer’s comment. In figure 1C images of actin networks are shown in black and white and are more visible than in figure S1D where they are shown in magenta and overlaid with microtubules. In any case, we increased the contrast of images to make fine actin structures at the cell edge clearer.

It is also interesting that on Figure 6A, actin bundles look different than the rest of the figures on the paper. It almost looks like actin bundles become branched, whereas in the other Figures actin bundles are either singular or two-three bundles joined together at the point very close to the edge of micropatterned lipid bilayer.

Authors answer:

This is correct. In this experiment several bundles co-aligned. As mentioned by the reviewer this could also be visible in other conditions without Tau (such as in Figure 4E), and, as shown below, this structure of bundle was not visible in all fields we looked at. So we don’t think this structure is responsible for the changes we measured in the ability of microtubules to penetrate the actin network in the presence of Tau.

Minor comments<br /> In the legend of Figure 4E, it should be written "white arrow" instead of "yellow arrow".<br /> In the Results section "crosslinking between microtubules and actin bundles increase piercing frequency", in the sentence number 7, it should be written "backwards" instead of "reaward".

Authors answer: We modified the text and legend according to the reviewer suggestions.

Reviewer #1 (Significance):

The experimental setup of the paper is quite significant in the field, given the difficulty of observing dynamics of dense cytoskeletal structures in living cells. Moreover, the paper gives insight into how microtubule behavior can vary depending on different morphological states of actin network.

Authors answer: We thank the reviewer for his/her overall very positive feedback on our manuscript.

Reviewer #2 (Evidence, reproducibility and clarity):

The authors developed a novel in vitro system to investigate the interaction of dynamic microtubules with the F-actin network. While this system does produce some interesting results, it is unclear how exactly this replicates or explains what might happen near a cell's leading edge. There is a limited characterization of the produced F-actin networks. For example, it is unclear to what extent the F-actin networks are similar or different to cell lamellipodial networks. What is the density / expected mesh size of these networks and could that be varied / manipulated? The bottomline observation that microtubules can grow into F-actin networks if they have nowhere else to go does not seem particularly ground-breaking, and the discussion is very shallow. Overall writing could be improved; there are lots of typos and grammatical inconsistencies. The second paragraph of the introduction is a bit convoluted.

Authors answer:

We thank the reviewer for his/her comments. Figure 1 was used to illustrate the behavior of microtubules encountering actin networks in cells and the fact that they struggle to penetrate actin network. This is only a way to argue that the penetration of actin network is a relevant question, that cannot be easily addressed in cells. However, it is correct that our in vitro systems, as it is the case for all in vitro reconstituted systems, cannot tend exactly to reproduce a lamellipodial cellular network. But it offers a better way to modulate actin network architecture. We have used in vitro systems to characterize the different behavior of microtubules when they encounter dense actin networks in different conditions, guided or not by actin bundles, constraint or not at the two ends.

The observation that microtubule can penetrate actin network when pressurized might not be “ground breaking”, still it contradicts previous works showing that microtubule under pressure tend to depolymerize (Janson et al, J Cell Biol, 2003), which would obviously prevent them from penetrating actin networks. So, our conclusion was somehow unexpected.

We found important to discuss the fact that although the microtubule polymerizing forces is sufficient to breach dense actin network, it must be counteracted by another mechanism immobilizing microtubules. This means that in cells, expression level of actin-microtubule crosslinker modulate the penetration of microtubule into the lamellipodium.

However, we agree that the second paragraph of the introduction is not absolutely necessary and removed it.

Specific comments:

Fig. 1 seems a bit anecdotal. The authors revisit an observation that has been made before. I can see how it is used as rationale for the in vitro system, but not sure that this adds much to the overall story. Clearly different cell types are different, but without some sort of quantification this remains meaningless. It should also be noted in the discussion maybe that there are large differences between cells in 2D and 3D. Microtubules much more frequently grow to the cell edge compared with 2D (see Akhmanova SLAIN2 paper from some years ago).

Authors answer:

We agree with these comments. Indeed, Figure 1 is used only as an illustration of the behavior of microtubules encountering actin network in cells. As the reviewer said, microtubule penetration and actin architectures will both vary a lot from one cell type to another. So we believe that quantification for these particular cases will not extend the illustrative purpose of this figure where it is already clear that some microtubules can penetrate and other can’t.

Fig. 2: While Arp2/3 certainly promotes branched F-actin networks, from the data provided it is not clear to me to what extent the produced F-actin networks replicate F-actin organization at the cell edge. If this a the point the authors are trying to make, the ultrastructure of their in vitro networks needs some additional characterization. As far as it is possible to discern from the data provided, the F-actin meshwork on the stripes in E looks pretty much identical in both panels (and not really like a dendritic network that in a cell also would have a certain polarity with barbed ends facing out), and the bundles on the left don't look like anything that normally occurs in a cell.

Authors answer:

We also agree with these comments. The networks we assembled are not lamellipodial-like networks, there are branched network of various densities, with or without bundles. It is true that bundles of filaments do not grow out of lamellipodial network in cells. However, bundles of aligned and linear filaments exist in cells, in the form of radial fibers or transverse arcs, along which microtubule tend to align. And these structures might guide microtubules toward cell protrusions, as it is the case in growth cone for example.

Fig. 4 It is unclear what is going on here. Given that without F-actin bundles, polymerizing microtubules are freely moving around, it does not come as a surprise that they would never penetrate the F-actin network because as the authors correctly state the growing end will push back from the barrier. So, then why do they sometimes penetrate when bundles are present? In 4A it appears that microtubule growth into f-actin only happens once the microtubule minus ends gets stuck between F-actin bundles on the other side. 4D is the same as 4A; so that makes me think this really does not occur that often. Does the microtubule plus end only penetrate the F-actin meshwork when the minus end gets stuck on the other side? This seems important and also means microtubule penetration may not have anything to do with the F-actin network architecture at the plus end. This needs to be quantified.

Authors answer:

This is perfectly correct. In figure 4 the two actin networks are distant, and the microtubules only rarely penetrate them because they are rarely in contact with them at both ends. This occurs only when bundles orient microtubules perpendicular to the edges of the actin network, since in this configuration the distance between the two actin networks is shorter. Hence our motivation to bring actin networks closer to each other in figure 5.

Fig. 5 I guess that sort of solves my confusion with Fig. 4. The quantification graphs in 5B and 5C are flipped with respect to the figure legend (?).

Authors answer:

Indeed, in this figure we distinguished the role of pressure (when both microtubule ends are in contact with actin networks) and the role of alignment with actin bundles. And found that the presence of bundles is useless and that only pressure matters.

I understand the rationale for not considering microtubules that grow at a shallow angle, but there does not seem to be that much of a difference between 5B and 5D. Wouldn't a better quantification simply compare microtubules that contact F-actin at both ends compared with microtubules were the minus end is free. In this case, I would expect a very large difference in penetration.

Authors answer:

This is also correct. The difference is so important that when one end is free the microtubule never penetrate. We mention it in the text but did not plot these data. This is why we measured only microtubule with both ends contacting an actin network and did not consider the one at shallow angles.

We added the illustration of the condition with short distance and actin bundles (shown below) to make this more clear in the figure.

The small difference between 5B and 5D shows that by eliminating those microtubules there is no more difference between the conditions with or without bundles. And thus that their contribution in favoring microtubule penetration was to favor optimal orientation to get pressurized at the two ends rather than offering a sort of favorable network organization at their base. However, we agree with the reviewer that the absence of difference between the two populations, with or without actin bundle, when considering only microtubule interacting with actin at angles higher than 30° is not quite striking. We tested all angles (see below) and found that actually the absence of difference is more obvious when considering microtubules interacting with more than 60°. And the analysis of angle distribution, now reported in Figure 5D, showed that in both conditions most microtubules interact with more than 60°, so we only exclude few outliers by considering those that interact with more than 60°. So we changed the presentation of our data in Figure 5C by changing the threshold from 30 to 60°.

Do microtubules under pressure ever bend/buckle in this in vitro situation. As the authors state, in cells, that happens frequently. This difference is interesting. Why?

In vitro microtubules buckle homogeneously between their two ends. These long buckling wavelengths are not very spectacular. In cells, microtubules are crosslinked to actin filaments or other structures over shorter distances (see quantification below). This leads to buckling with shorter wavelength, which is more striking.

It is customary to refer to polymerized actin as F-actin.

The supplementary videos are not referenced in the manuscript.

Authors answer:

We apologize and have now referenced the supplementary video in the manuscript.

Reviewer #2 (Significance):

The manuscript describes results from a novel assay to study interactions between F-actin networks and dynamic microtubules in vitro. While of interest to a specialized audience, the overall finding that microtubules can grow into an F-actin meshwork is somewhat incremental especially because of the limited characterization of the F-actin networks used. It remains unclear to what extent this is relevant to a physiological context in cells.

My field of expertise is related to cytoskeleton dynamics and quantitative microscopy in live cells.

Authors answer:

Although intuitive, the demonstration that the density of actin network can prevent microtubule penetration is novel. More importantly, the demonstration that anchoring of microtubule is sufficient to increase the pressure to such a point that microtubule can then penetrate those networks is also novel and significant to appreciate when and how they do so in cells.

Reviewer #3 (Evidence, reproducibility and clarity):

In this paper, the authors present an in vitro assay designed to explore the dynamic interaction between growing microtubules and pre-existing actin networks. Notably, the presence of linear actin bundles facilitated the movement of polymerizing microtubules along actin filaments. When microtubules were immobilized to two spatially separated actin networks, they exhibited the ability to breach and penetrate dense actin meshworks. This penetration was attributed to the mechanical pressure generated by microtubule polymerization. The authors tested tau as a microtubule-actin crosslinking protein in this process and found that tau promoted microtubule penetration into dense actin meshwork. Although the findings in this paper are potentially significant, the work is still in its preliminary stage and the scope is limited.

Authors answer:

We thank the reviewer to summarize properly the main findings of our manuscript.

1. The authors observed that the inclusion of tau, a microtubule-associated protein known for its role in promoting microtubule polymerization, significantly facilitated microtubule penetration into dense actin meshworks. This enhancement is likely attributed to tau's ability to promote microtubule polymerization, generating stronger forces within the microtubules that enable them to breach the actin meshworks. To validate the involvement of the crosslinking function in the process, the authors should explore the effects of other microtubule-actin crosslinking proteins in their assay.

Authors answer:

We thank the reviewer for this interesting suggestion regarding the role of Tau in our experiments. To address this comment, we have analyzed the rate of growth in our experiments in presence and absence of Tau (see quantification below). We found that the construction of Tau we used reduced microtubule growth rate. Therefore, we believe that microtubule growth was not responsible for the improved penetration of microtubule in dense actin networks in our assay, and that it was rather the crosslinking ability of Tau that played a significant role.

1. The paper highlights the importance of anchoring both ends of microtubules to two adjacent actin networks for successful penetration into the actin meshworks. However, the precise mechanisms by which these microtubule ends are anchored to actin filaments are not elaborated upon. Providing detailed insights into this anchoring process would enhance the readers' comprehension of the experimental setup and its relevance to the observed results.

Authors answer:

We apologize for this lack of clarity. We don’t think that microtubule ends are “anchored” to the actin network. They are simply embedded into it. This embedding prevents them from moving rearward and thus lead to pressure increase as they polymerize.

1. Additional information on the experimental methods is warranted to improve the reproducibility and clarity of the study. Specifically, the authors should elucidate the process through which nucleation-promoting factors were grafted onto lipid bilayers. This detail is crucial for researchers seeking to replicate or build upon the study's findings.

Authors answer:

We apologize for this lack of clarity. There was indeed an error in our description of SUV preparation with lipid-biotin. We have now revised our material and method section. In particular we have described more accurately the various steps we used to micropattern WA-streptavidin onto lipid-biotin.

1. In Fig. 5D, the authors observed no significant difference in the breaching probability between microtubules that contacted the actin meshwork at an angle higher than 30°, with or without actin bundles. To ensure a better comparison, it is advisable to focus on quantifying the microtubules that are contacting two actin meshworks at both ends (the immobilized microtubules), as they would have similar probabilities of being pressurized by their growth. Moreover, further justification is required to explain the choice of 30° as the threshold angle and its significance in the context of microtubule behavior.

Authors answer:

We thank the reviewer for this comment. We apologize for the confusion. The quantification we made is precisely the one described by the reviewer. We made this more clear by adding further illustration of the two conditions and the measurement made.

1. Fig. 5C appears to depict the "Distribution of the angle of the interaction of microtubules in the presence (10nM of Arp2/3 complex) or absence (100 nM of Arp2/3 complex) of actin bundles" instead of the "proportion of microtubules piercing the branched actin network." The alphabet labels in the figure should be updated accordingly. Additionally, the authors should clarify whether a comparison was conducted between the means of the angles in the two conditions and whether any observed differences were statistically significant.

Authors answer:

We apologize for this confusion. We updated the figure legend in which 5C and 5D were inverted.

1. Investigating the potential significant difference in the mean interaction angles between the absence and presence of actin bundles would be intriguing. The presence of actin bundles might indeed influence the interaction angle or contact position, potentially increasing penetration frequency. This insight would further enrich the findings and provide valuable context for understanding the interplay between microtubules and actin networks.

Authors answer:

We apologize for this confusion. We now report the statistical difference. And indeed, it accounts for the difference it the penetration frequency, as shown by the absence of difference when we consider only microtubules that are more or less perpendicular to the network. This is indeed one of the most significant conclusion of our work. We added some schematics to make this clearer.

1. More comprehensive information about the statistical analyses should be provided. This'd be important for the validity and reliability of the study's conclusions.

Authors answer:

We apologize for this lack of clarity. The statistical analysis we performed were not described in the Materials and Methods section but in each figure legend.

Reviewer #3 (Significance):

The work represents an advance in understanding the mechanism by which microtubules navigate dense actin meshworks.

Authors answer:

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

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 paper, the authors present an in vitro assay designed to explore the dynamic interaction between growing microtubules and pre-existing actin networks. Notably, the presence of linear actin bundles facilitated the movement of polymerizing microtubules along actin filaments. When microtubules were immobilized to two spatially separated actin networks, they exhibited the ability to breach and penetrate dense actin meshworks. This penetration was attributed to the mechanical pressure generated by microtubule polymerization. The authors tested tau as a microtubule-actin crosslinking protein in this process and found that tau promoted microtubule penetration into dense actin meshwork. Although the findings in this paper are potentially significant, the work is still in its preliminary stage and the scope is limited.

1. The authors observed that the inclusion of tau, a microtubule-associated protein known for its role in promoting microtubule polymerization, significantly facilitated microtubule penetration into dense actin meshworks. This enhancement is likely attributed to tau's ability to promote microtubule polymerization, generating stronger forces within the microtubules that enable them to breach the actin meshworks. To validate the involvement of the crosslinking function in the process, the authors should explore the effects of other microtubule-actin crosslinking proteins in their assay.
2. The paper highlights the importance of anchoring both ends of microtubules to two adjacent actin networks for successful penetration into the actin meshworks. However, the precise mechanisms by which these microtubule ends are anchored to actin filaments are not elaborated upon. Providing detailed insights into this anchoring process would enhance the readers' comprehension of the experimental setup and its relevance to the observed results.
3. Additional information on the experimental methods is warranted to improve the reproducibility and clarity of the study. Specifically, the authors should elucidate the process through which nucleation-promoting factors were grafted onto lipid bilayers. This detail is crucial for researchers seeking to replicate or build upon the study's findings.
4. In Fig. 5D, the authors observed no significant difference in the breaching probability between microtubules that contacted the actin meshwork at an angle higher than 30{degree sign}, with or without actin bundles. To ensure a better comparison, it is advisable to focus on quantifying the microtubules that are contacting two actin meshworks at both ends (the immobilized microtubules), as they would have similar probabilities of being pressurized by their growth. Moreover, further justification is required to explain the choice of 30{degree sign} as the threshold angle and its significance in the context of microtubule behavior.
5. Fig. 5C appears to depict the "Distribution of the angle of the interaction of microtubules in the presence (10nM of Arp2/3 complex) or absence (100 nM of Arp2/3 complex) of actin bundles" instead of the "proportion of microtubules piercing the branched actin network." The alphabet labels in the figure should be updated accordingly. Additionally, the authors should clarify whether a comparison was conducted between the means of the angles in the two conditions and whether any observed differences were statistically significant.
6. Investigating the potential significant difference in the mean interaction angles between the absence and presence of actin bundles would be intriguing. The presence of actin bundles might indeed influence the interaction angle or contact position, potentially increasing penetration frequency. This insight would further enrich the findings and provide valuable context for understanding the interplay between microtubules and actin networks.
7. More comprehensive information about the statistical analyses should be provided. This'd be important for the validity and reliability of the study's conclusions.

Significance

The work represents an advance in understanding the mechanism by which microtubules navigate dense actin meshworks.

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 authors developed a novel in vitro system to investigate the interaction of dynamic microtubules with the F-actin network. While this system does produce some interesting results, it is unclear how exactly this replicates or explains what might happen near a cell's leading edge. There is a limited characterization of the produced F-actin networks. For example, it is unclear to what extent the F-actin networks are similar or different to cell lamellipodial networks. What is the density / expected mesh size of these networks and could that be varied / manipulated? The bottomline observation that microtubules can grow into F-actin networks if they have nowhere else to go does not seem particularly ground-breaking, and the discussion is very shallow. Overall writing could be improved; there are lots of typos and grammatical inconsistencies. The second paragraph of the introduction is a bit convoluted.

Specific comments:

Fig. 1 seems a bit anecdotal. The authors revisit an observation that has been made before. I can see how it is used as rationale for the in vitro system, but not sure that this adds much to the overall story. Clearly different cell types are different, but without some sort of quantification this remains meaningless. It should also be noted in the discussion maybe that there are large differences between cells in 2D and 3D. Microtubules much more frequently grow to the cell edge compared with 2D (see Akhmanova SLAIN2 paper from some years ago).

Fig. 2: While Arp2/3 certainly promotes branched F-actin networks, from the data provided it is not clear to me to what extent the produced F-actin networks replicate F-actin organization at the cell edge. If this a the point the authors are trying to make, the ultrastructure of their in vitro networks needs some additional characterization. As far as it is possible to discern from the data provided, the F-actin meshwork on the stripes in E looks pretty much identical in both panels (and not really like a dendritic network that in a cell also would have a certain polarity with barbed ends facing out), and the bundles on the left don't look like anything that normally occurs in a cell.

Fig. 4 It is unclear what is going on here. Given that without F-actin bundles, polymerizing microtubules are freely moving around, it does not come as a surprise that they would never penetrate the F-actin network because as the authors correctly state the growing end will push back from the barrier. So, then why do they sometimes penetrate when bundles are present? In 4A it appears that microtubule growth into f-actin only happens once the microtubule minus ends gets stuck between F-actin bundles on the other side. 4D is the same as 4A; so that makes me think this really does not occur that often. Does the microtubule plus end only penetrate the F-actin meshwork when the minus end gets stuck on the other side? This seems important and also means microtubule penetration may not have anything to do with the F-actin network architecture at the plus end. This needs to be quantified.

Fig. 5 I guess that sort of solves my confusion with Fig. 4. The quantification graphs in 5B and 5C are flipped with respect to the figure legend (?). I understand the rationale for not considering microtubules that grow at a shallow angle, but there does not seem to be that much of a difference between 5B and 5D. Wouldn't a better quantification simply compare microtubules that contact F-actin at both ends compared with microtubules were the minus end is free. In this case, I would expect a very large difference in penetration. Do microtubules under pressure ever bend/buckle in this in vitro situation. As the authors state, in cells, that happens frequently. This difference is interesting. Why?<br /> It is customary to refer to polymerized actin as F-actin.<br /> The supplementary videos are not referenced in the manuscript.

Significance

The manuscript describes results from a novel assay to study interactions between F-actin networks and dynamic microtubules in vitro. While of interest to a specialized audience, the overall finding that microtubules can grow into an F-actin meshwork is somewhat incremental especially because of the limited characterization of the F-actin networks used. It remains unclear to what extent this is relevant to a physiological context in cells.

My field of expertise is related to cytoskeleton dynamics and quantitative microscopy in live cells.

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

Major comments

In the paper "Microtubules under mechanical pressure can<br /> breach dense actin networks", the authors showed clear evidence that pressure plays an important role in microtubule breaching into dense actin networks using elegant in vitro reconstitution assays. They have argued that the pressure results from polymerization force of microtubules, which builds up when microtubules are immobilized in the opposite end of breaching, by the means of actin microtubule crosslinking factor Tau.<br /> It would definitely be interesting to see lack of breaching in the presence of crosslinking deficient Tau construct in order to rule out the off -target effect of Tau on microtubule and actin architecture which may possibly facilitate breaching.

The authors have also observed microtubule breaching into dense actin networks in living cells. However, in Figure 1C, better cell/ image processing might have been chosen to increase the visibility of actin structures that microtubules encounter on their way to breaching. In Figure S1D, for example, the similar actin structures in lamellipodia are very nicely visible.

It is also interesting that on Figure 6A, actin bundles look different than the rest of the figures on the paper. It almost looks like actin bundles become branched, whereas in the other Figures actin bundles are either singular or two-three bundles joined together at the point very close to the edge of micropatterned lipid bilayer.

Minor comments

In the legend of Figure 4E, it should be written "white arrow" instead of "yellow arrow".<br /> In the Results section "crosslinking between microtubules and actin bundles increase piercing frequency", in the sentence number 7, it should be written "backwards" instead of "reaward".

Significance

The experimental setup of the paper is quite significant in the field, given the difficulty of observing dynamics of dense cytoskeletal structures in living cells. Moreover, the paper gives insight into how microtubule behavior can vary depending on different morphological states of actin network.

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 are happy to resubmit our manuscript “The protective function of an immunity protein against the cis-toxic effects of a Xanothomonas Type IV Secretion System Effector” by Gabriel Oka et al. This paper shows that the cohort of immunity proteins associated with the cocktail of toxic effectors secreted by the Xanthomonas citri T4SS are not required to protect against toxins injected by neighboring cells but rather provide protection against endogenous toxins of the cell in which they were produced. To our knowledge, this the first description of an antibacterial secretion system in which the immunity proteins are dedicated to protecting cells against cis-intoxication, a point we emphasize in the revised introduction.

We thank the reviewers for their thorough revision of the manuscript. Two of the three reviewers clearly expressed the opinion that the manuscript would be of general interest and should be published. We have carried out a number of new experiments and data analyses to respond to most of the suggestions of all three reviewers and believe that the manuscript is significantly improved as a result.

Reviewer #1 (Evidence, reproducibility and clarity):

The manuscript by Oka et al. shows that one effector of X. citri is likely translocated into the periplasm where it cleaves PG unless inhibited by its cognate immunity protein. Interestingly, this effector is required for killing of target cells like E. coli in T4SS-dependent manner but it does not seem to be delivered into X. citri cells by T4SS. Authors show using various assays that cells lacking the immunity protein have various phenotypes including lysis and defect in biofilm formation, however, despite "cis-intoxication" the ability to kill other bacteria or infect plants remains unaffected. The manuscript is well written and in general all the experiments have proper controls and thus the conclusions seem solid. The results described here are novel and interesting as they are unexpected.

Major issues that should be addressed:

- Test various deletion variants of the toxin to identify which part of the protein is responsible for its translocation into the periplasm. This may help to identify the possible mechanism of translocation of the toxin into the periplasm. Alternatively, the authors may attempt to select for non-toxic point mutants of the toxin. This could be done by a random PCR mutagenesis of the toxin and a selection of the surviving mutants in the absence of the immunity protein.

Thank you for your insightful experimental suggestions using PCR mutagenesis to investigate the molecular mechanisms of the alternative translocation of X-Tfes to the periplasm. However, I regret to inform you that the first five authors of this manuscript are no longer a member of my lab. Therefore, please consider accepting the results shown in Figure 5A where we observe that the N-terminal domain of X-TfeXAC2609 that lacks the XVIPCD domain still abolishes biofilm formation in the absence of X-TfiXAC2610. Also, note that the E48A point mutation in the active site of the GH19 domain that abolishes the in vitro activity of X-TfeXAC2609 (Souza et al. 2015), also abolishes X-TFEXAC2609 toxicity in vivo in the absence of X-TfiXAC2610(Figure 5). Furthermore, in the predicted structure of the X-TfiXAC2610(54-267)-X-TfeXAC2609(1-194) complex (Figure 6), A tyrosine side chain in the conserved loop in X-TfiXAC2610 interacts directly with Glu48 in the X-TfeXAC2609 active site. One possibility for further investigation is the remaining region of X-TfeXAC2609(195-306) as a putative translocation domain. Sequence analysis of this region indicates that it encodes a canonical peptidoglycan binding domain. Another possibility is the existing intrinsic leakage of cytoplasmic proteins to the periplasm. As we understand it, the leakage of cytoplasmic proteins to the periplasm is not a well-documented phenomenon, although there is some evidence that suggests it may occur (PMID: 28808000, PMC3016450 references cited in the revised manuscript). This poorly characterized T4SS-independent pathway of translocation as indicated in Figure 1B (pathway 2).

- Test if localization of the immunity protein to the cytoplasm blocks its activity. An immunity protein mutant that lacks its secretion signal should not protect against cis-intoxication.

To address the question, we conducted new assays on colony opacity, as shown in Figure S2C. The X. citri ∆X-TfiXAC2610 strain was transformed with a plasmid expressing a cytoplasmic version of X-TfiXAC2610 that lacks the signal peptide and lipobox (X-TfiXAC2610(His-22-267)). Figure S2C shows that this cytosolic version of X-TfiXAC2610 protects X. citri from the toxic effects of X-TfeXAC2609 in the ∆X-TfiXAC2610 background. This suggests that X-TfiXAC2610(His-22-267) may directly interact with X-TfeXAC2609 in the cytoplasm, leading to the inhibition of X-TfeXAC2609 hydrolase activity and/or inhibition of its translocation into the periplasm. This is now mentioned in the results section of the revised manuscript

While many experiments support the conclusion that the toxin is responsible for "cis-intoxication, the test of "trans-intoxication" should be done again but with the same setup as was used for testing of killing of E. coli. The CPRG based assay is far more sensitive than counting survival by plating to count CFUs. This test should be done at a relatively high initial OD so that there is an immediate contact between the "killer" and the "prey" bacteria (lacking immunity/effector). If needed, LacZ should be over-expressed in X. citri to make use of the CPRG based assay. In addition, such assay could be used also for "cis-intoxication" to supplement the potentially hard to quantify biofilm experiments shown in Fig. 4 (e.g. test all the T4SS mutants for "cis-intoxication").

We are confident that the X-Tfis do not play a role in protecting against T4SS-mediated trans-intoxication since we continue to observe X-TfeXAC2609-dependent intoxication even in the absence of a functional XT4SS (see experiments using strains lacking X-T4SS subunits in Figures 2, S2, 3, 4 and 5. This is not to say that trans-intoxication does not occur. In fact, it does, and there is an independent mechanism that protects against it. We will provide details of the mechanism that protects against trans-intoxication in a forthcoming manuscript. In the present manuscript, we are addressing the phenomenon of cis-intoxication. To support our conclusion that the immunity proteins are not involved in the prevention of trans-intoxication does not occur in X. citri, we have included one additional supplementary video: Movie S7 shows that wild-type Xanthomonas citri does not kill and X. citri Δ8Δ2609-GFP. The absence of killing events in these experiments indicates that the X-T4SS-associated X-Tfi immunity proteins are not required for protection against X-T4SS-mediated sibling attack.

In addition, such assay could be used also for "cis-intoxication" to supplement the potentially hard to quantify biofilm experiments shown in Fig. 4 (e.g. test all the T4SS mutants for "cis-intoxication").

- Fig. 2A needs a positive control. For example, test killing of E. coli under the same conditions.

Figure 2A of the revised manuscript now shows a CPRG assay competition assay that clearly demonstrates X-T4SS-dependent killing of E.coli MG by X. citri. We have now included the results of CFU experiments of X. citri vs E. coli competitions in a new Supplementary Figure (Figure S1) that are consistent with the CPRG assays. We note that our group has published similar results in the past (Souza et al. 2015; Oliveira et al. 2016; Oka et al. 2022). CFU measurements of X. citri vs E. coli competition assays are performed under slightly different conditions from the X. citri vs X. citri assays shown in Fig 2B. This is because E. coli grows at a significantly faster rate than X. citri so the initial cell ratios in these experiments have to be modified.

- Authors should look at the paper by Ho et al. PNAS 2017, which describes trafficking of VgrG of V. cholerae into the periplasm of E. coli without an obvious secretion signal. The effector of X. citri may behave similarly.

We thank the reviewer for this observation and now mention the paper by Ho et al. in the Discussion of the revised manuscript. Using a number of different algorithms (TatP, SignalP 6.0) we do not find any evidence of putative signal sequences. In the Discussion, we also mention the manuscript by Dong et al., 2013 that showed that the immunity protein TsiV3 that neutralizes VgrG3 is critical to prevent trans-intoxication.

- Provide some form of quantification of the phenotypes (cell rounding and cell death) observed using live-cell imaging.

As suggested by the reviewer, we performed a quantitative analysis of the propidium iodide (PI) permeability by calculating the percentage of PI permeable cells observed in movies S1-S5. This data is now presented in Figure 3 and Table S4 of the revised manuscript.

- Provide quantification of biofilm related phenotypes as well as of the citrus canker development assay

As suggested by the reviewer, we have carried out experiments to quantify the amount of biofilm using a crystal violet assay (absorbance at 570 nm). The results are presented in Figure S5 of the revised manuscript.

Reviewer #1 (Significance):

The study provides an interesting insight into immunity proteins against anti-bacterial toxins. It points to a need to protect against "cis-intoxication". This is novel and interesting to a wide audience of microbiologists interested in bacterial competition as this could be true also for other toxins.

We thank the reviewer for his/her positive recommendation.

It would be however important to identify how is the toxin translocating to the periplasm of the producing bacterium. Some insight into the mechanism would vastly improve the study. My expertise is in understanding bacterial interactions and competition but I lack a direct experience with assays specific for X. citri.

We agree that an understanding of the mechanism of translocation into the periplasm would be interesting but is beyond the scope of the present manuscript. However, we do point out that this has been observed previously by other groups in the fourth paragraph of the Discussion of the revised manuscript: “... In the case of X-TfeXAC2609, the toxin somehow makes its way into the cell periplasm where, in the absence of X-TfiXAC2610, it degrades the peptidoglycan layer. Analysis of the X-TfeXAC2609 sequence by the SignalP 6.0 (Teufel et al., 2022) and other algorithms failed to detect any putative N-terminal signal peptide. Although the mechanism responsible for X-TfeXAC2609 transfer into the periplasm is at the moment unknown, we have shown that it is independent of a functional X-T4SS and of the XVIPCD secretion signal. Other bacterial proteins have been shown to transfer into the periplasm without any obvious secretion signal, for example VgrG3 from Vibrio cholerae (Ho et al. 2017) and recombinant forms of HdeA and chymotrypsin inhibitor 2 (Banes and Pielak, 2011).”

Reviewer #2 (Evidence, reproducibility and clarity):

This manuscript explores the role of an immunity protein of the Xanthomonas type IV secretion system (X-T4SS). In contrast to most T4SSs that conjugate plasmids or transfer effectors into host cells, this system is able to kill other bacteria similar to the role of T6SSs. Here, the authors tested whether the immunity protein XAC2610 functions to prevent cis-intoxication (by self) and/or trans-intoxication (by sister cells). They provide data that the XAC2610 immunity protein functions to protect cis intoxication, but not trans-intoxication, by the T4SS effector XAC2609 (which functions as a peptidoglycan hydrolase). Based on AlphaFold modeling, they went on to identify a residue in XAC2610 that is critical for inhibiting the activity of the XAC2609 toxin. Overall the data is fairly solid and generally support the conclusions the authors made.

Major comments:

One of the major conclusions of the manuscript is that XAC2610 does not prevent trans-intoxication and the data in the manuscript support this conclusion. However, I wonder if this is an oversimplification. Notably, the authors observed that wild type Xantho was unable to kill a target cell lacking 8 different toxin/immunity systems (Fig. 1A). One could conclude that none of these immunity proteins function in preventing trans-intoxication ... or ... perhaps it appears that none perform this role because wild-type Xantho never attacks its siblings? For example, it is conceivable that Xantho uses a general mechanism, perhaps somewhat similar to phage exclusion or plasmid incompatibility, to prevent sibling attack? To me this seems more likely than none of the eight immunity proteins play a role in preventing trans-intoxication. Moreover, the phenotype observed for the ∆2610 mutant in preventing cis-intoxication is somewhat subtle, likely because the toxin and the immunity protein are topologically restricted to the cytoplasm and the periplasm, respectively. This would make sense if this were not the primary role for 2610.

Ideally the authors will be able to test this theory by demonstrating that a wild-type Xantho strain can attack (but likely not kill) its siblings. Alternatively, could the authors test if related, but not identical, Xantho strains that express 2609/2610 are able to kill their ∆2610 mutant, i.e. do "cousins" attack each other? Not sure about the semantics but this could be described as preventing trans-intoxication. If they are unable to do either experiment, that is ok but they should at least describe this concept in their discussion (assuming they agree).

We thank the reviewer for his insightful comments. Indeed, this manuscript is focussed solely on the role of X-Tfi immunity proteins which we show to be principally involved in avoiding cis-intoxication (self-intoxication). The question of trans-intoxication will be left to an upcoming manuscript by our group. In fact we have identified a key factor (not an X-Tfi) that is responsible for inhibiting trans-intoxication. As suggested by the reviewer, we have now added the following text to the end of the third paragraph of the Discussion: “Nevertheless, the fact that wild-type X. citri is unable to kill strains lacking immunity proteins is intriguing. That cells in some way avoid trans-intoxication is revealed by the fact that X. citri wild-type cells carrying an X-T4SS and full cohort of X-Tfes do not kill the X. citri Δ8Δ2609-GFP, the X. citri ∆X-TfeXAC2609∆X-TfiXAC2610, or any other X-T4SS-deficient strain tested points to a still-to-be-characterized mechanism of protection against trans-intoxication (fratricide) that will be addressed in future studies by our group.”

Minor comments:

1. Figure 1 is a bit confusing in terms of the layout. It would be beneficial if the authors separated parts A and B by a few spaces.

As suggested, we have modified the layout of Figure 1 to more clearly distinguish between the two mechanisms tested.

1. Figure 2A should start off by showing that the Xantho T4SS can kill other bacteria (e.g. Fig S4A). This would set up the paper better.

As suggested, old Figure S4A has now been transferred to Figure 2A in the revised manuscript.

1. Fig. 2A should include p values.

As suggested, p values have been provided in the legend of Figure 2B (old Figure 2A).

1. Fig. 2B is really hard to see and should be removed from the manuscript (although I do appreciate the novelty of the technique using Marilyn Monroe).

We have transferred the old Fig. 2B to the Supplementary Material (Fig S2) of the revised manuscript. We agree that the effect is subtle, but we want to maintain the figure since the transparency of the ΔXAC2610 X. citri colonies over time were the first observations that led us to investigate this phenomenon. Additionally, to reduce potential human bias and to enhance the objectivity of the assay, we employed a Convolutional Neural Network (CNN) to analyze all the colonies presented in Fig S2. This method provides a confidence tendency index for opacity and transparency variations. A detailed description of this new methodology is in the "Materials and Methods" section (Convolutional Neural Network (CNN) analysis).

1. Instead all of the data in Fig. 2B should be shown in a new version of Fig. 2C. Fig. 2C should include additional controls including:
2. A wild type strain containing 2609 and 2610 mutants
3. A complete virB operon deletion in combination with 2609 and 2610 mutants
4. ∆8 strain
5. 2609 lacking its T4SS signal sequence
6. 2609 targeted to the periplasm with a sec signal sequence
7. etc.

We sincerely value the comprehensive suggestions for improving what was previously presented as Fig. 2D. (Current version of Figure 2B is the Fig S2 as mentioned in the previous observation). We encounter a practical challenge here: the primary authors responsible for these experiments, especially the first five, have since departed from our lab. This situation limits our immediate capacity to execute the extensive set of experiments you've proposed.

Recognizing the significance of the controls you've outlined for a quantitative analysis of the colony phenotypes (Fig. 2C (current version)) we have instead supplemented our study with a rigorous quantitative analysis of the microscopy assays referenced in Movies S1-S5, Figure 3, Table S4. These analyses further emphasize our observations concerning colony transparency (Fig S2).

1. Figure 2C. The VirB7 western band looks like in the 2610 complemented strain.

Thank you for pointing out the discrepancy in our previous manuscript at line 366, which pertains to the description of the mutants in old Fig. 2. The double mutant, ΔX-TfiXAC2610ΔvirB7 strain, was actually complemented with X-TfiXAC2610 (as stated in the current version (Fig S2B), and not with VirB7. Additionally, we have corrected the legend of the figure (line 684 previous version) from (∆X-TfiXAC2610∆VirB7c) to (∆X-TfiXAC2610c∆VirB7). We apologize for the mix-up in our earlier description and are grateful for your meticulous review and feedback in this matter. Furthermore, we agree that, in this particular experiment, the VirB7 band seems weaker but it is clearly visible in the 2610 complemented strain.

1. Figure 3C should include a comparison of exponential vs. stationary phase cells. In addition, the results for the ∆2610 mutant and the ∆2610 ∆B7 double mutant appear to be different(?). P values should be provided. If it is statistically significant, then this should be explained in the manuscript. It was not clear how the % damaged cells were calculated? # of cells? Stats?

The statistical analysis that the reviewer suggested has been provided in the new version of the Figure 4C and its legend. In addition, we have also included a supplementary Table S5 that presents the total number of cells analyzed in these experiments.

1. The majority of Figure 4 should be replaced by assaying the effect of a virB operon deletion rather than showing the individual mutants.

We believe that retaining old Figure 4 (Figure 5 of the revised manuscript) is important. By showcasing results from this specific set of single mutants, we are able to rule out the possibility that X-TfeXAC2609 translocation into the periplasm is mediated by a distinct X-T4SS subunit or subcomplex. We've expanded on this rationale at the start of the paragraph to provide a more comprehensive justification for our approach.

1. Discussion:
2. The last one to two paragraphs of the results belong in the Discussion.
3. A more detailed description of cis-intoxication would be useful.

As suggested, the last two paragraphs the Results section of the original manuscript have now been moved to the end of the Discussion.

As suggested by the reviewer, third paragraph of the Discussion describes cis-intoxication in more detail.

Reviewer #2 (Significance):

This work provides a conceptual advance in understanding the protective function of a T4SS immunity protein, X-Tfe XAC2610, against the cis-toxic effects of the T4SS effector, X-Tfi XAC2610. It will likely be of interest to scientists interested in T4SSs & T6SSs and interbacterial competition. Overall this is a thought-provoking manuscript and should be published in a respectable journal.

We sincerely thank Reviewer #2 for the thoughtful appraisal and positive feedback regarding our work. We are gratified to hear that the reviewer recognizes the conceptual advance our research brings to the understanding of T4SS immunity proteins and are encouraged by the acknowledgment that this manuscript will be of interest to our peers. We truly appreciate the endorsement for publication in a reputable journal.

Reviewer #3 (Evidence, reproducibility and clarity):

In this study, the authors suggest that TfeXAC2609-TfiXAC2610 represent a novel deviation from the established paradigm in contact-dependent interbacterial secretion systems. X. citri strains lacking the predicted immunity protein, TfiXAC2610, do not suffer a competitive disadvantage when grown in T4SS-inducing conditions against a wild-type strain. Furthermore, cells lacking the immunity develop aberrant morphology and auto-lyse. The mechanism for self-intoxication by TfeXAC2609 is independent of a functional T4SS, and intoxication is exacerbated when the toxin's T4SS-signal sequence is removed.

Major Points

1. The authors of the study do not provide sufficient evidence that TfeXAC2609 contributes to T4SS mediated killing. Does the toxin behave in a synergistic way, rather than mediate killing independently? Does removing the toxin and immunity change the competitive advantage of X. citri?

We have shown in a previous publication that X-TfeXAC2609 does contribute to X-T4SS mediated killing (Oka et al, 2022). In that published paper we show that even in the absence of seven other toxin/antitoxin pairs, X-T4SS mediated transfer of only one effector (X-TfeXAC2609 or X-TfeXAC3634) can kill E. coli cells.

Removing only X-TfeXAC2609 and X-TfiXAC2610 does not significantly reduce the ability of X. citri cells to kill E. coli (Fig. 2A of the revised manuscript). This is expected since this double mutant still retains seven other toxin/immunity pairs.

Suggested Experiments: Competing, against E. coli, both WT X. citri and X. citri ΔXAC2609 ΔXAC2610, and determining whether there is a change in relative competitive advantage, or expressing TfeXAC2609 in a heterologous system and marking any observed toxic phenotype.

The results of the experiment suggested by the reviewer have now been included in part A of the revised version of Figure 2. The effect of deleting only one toxin such as X-TfeXAC2609 results in no detectable difference in killing efficiency, most likely due to the presence of the eight other X-Tfes, three of which have been shown (XAC3634) or are predicted (XAC0466 and XAC1918) to have pedptidoglycan hydrolase activity (Oka et al, 2022, Souza, 2015, Sgro et al, 2019).

1. Authors should directly answer where the toxin is active and localized in the cell.

Suggested Experiments: Western blot subcellular fractionation (cytoplasm, periplasm, etc) to determine the localization of each protein.

In response to the query about the toxin's activity and localization within the cell, we acknowledge the importance of such experiments to shed light on these aspects. However, I would like to highlight that the five first authors of this work are no longer affiliated with our lab. Consequently, we are facing constraints in terms of manpower and expertise to undertake comprehensive experiments such as the suggested subcellular fractionation.

Also, our earlier work demonstrated the importance of the XVIPCD for secretion via X-T4SS (Souza, 2015) and in vivo activity of X-TfeXAC2609 (Oka et al., 2022). Moreover, using heterologous proteins expressed in E. coli (Souza, 2015) and our current observation that the absence of X-TfiXAC2610 induces spheroplast formation (Fig 4A-B, Movie S6) strongly suggest that the peptidoglycan glycohydrolase activity of the N-terminal domain of X-TfeXAC2609 acts in the periplasm.

1. There is no evidence that TfeXAC2609 plays any role in inter-bacterial killing besides that is predicted from its genetic arrangement and in vitro assays from a previous publication.

Suggested Experiments: Again, with the available antibodies, detecting whether TfeXAC2609 is being secreted, either in competition settings against X. citri or E. coli; given that there is no killing observed in Fig. 2B, it may also be a suitable control for this experiment.

We have published in vivo evidence in the past:

Souza et al, 2015 showed that X-TfeXAC2609 is secreted when in contact with E. coli cells.

Oka et al, 2022 showed that an X. citri strain expressing X-TfeXAC2609/X-TfiXAC2610 but lacking seven other toxin/antitoxin pairs can still kill E. coli.

1. The structural and co-evolutionary analysis seems to miss an essential point - that the lack of fratricide protection is not due to a novel protein-protein interaction.

We do not understand this comment. As we point out in the manuscript, X-TfiXAC2610 does not protect against fratricide (trans-intoxication) but instead does protect against suicide (cis-intoxication). This protection requires a X-TfeXAC2609-X-TfiXAC2610 protein-protein interaction supported by the structural and co-evolutionary analysis as well as the experimental data using the X-TfiXAC2610 Y170A mutant (Fig. 6D of the revised manuscript). Moreover, we believe that the structural and sequence analysis significant expand the knowledge of the broader family of immunity proteins to which X-TfiXAC2610 belongs (Fig. S10 and Fig. S11 of the revised manuscript).

1. The role of the immunity in biofilm formation is confusing. Cells lacking the immunity die within 96 hours (the auto-lysis phenotype). Given that the immunity is required for viability in this time frame, wouldn't it also be required for viability after five days?

Suggested Amendments: Remove or de-emphasize.

In the manuscript we use several different techniques to show that cells lacking the X-TfiXAC2610 immunity protein are less viable than the wild-type strain under certain conditions (growth on LB agar plates, biofilm formation) but perhaps not under others (ie in direct short-term competition experiments against E. coli and in long-term (2 week) in planta citrus canker assays). This is consistent with the fact that ultrastructural analysis by transmission electron microscopy shows that when grown in liquid media, only around 2% of X. citri cells lacking X-TfiXAC2610 present significant damage to their cell envelope (only 0.1% of wild-type cells show damage).

1. Why does cell permeability increase with the loss of the T4SS signal sequence? Without there being greater evidence to support that an alternative secretion system is secreting or transporting the toxin into the periplasm, which may compete with the T4SS, additional hypotheses should be experimentally probed.

The reviewer is comparing the propidium iodide permeability results observed for the ΔX-TfiXAC2610 mutant (carrying an empty pBRA plasmid) that expresses full-length X-TfeXAC2609 from its chromossomal gene with the ΔX-TfeXAC2609/ΔX-TfiXAC2610 double mutant carrying the pBRA-X-TfeXAC2609NT plasmid that expresses the X-TfeXAC2609 protein lacking the T4SS signal sequence from a very strong inducible promoter. Therefore, it can be expected that the levels of the truncated effector could be significantly greater than that of the full-length effector, leading to more damage.

Note that, in the absence of X-TfiXAC2610, cell permeability increases only if X-TfeXAC2609 is present, with or without its XVIPCD T4SS signal sequence. This is consistent with a cis-intoxication mechanism which is independent of the X-T4SS-mediated transfer of the toxin from one cell to another. As we mention in the revised manuscript, and as pointed out by reviewer 1, Ho et al have also observed that when a lysozyme-containing domain of the T6SS effector VgrG3 is expressed in E. coli or in Vibrio cholerae, it can be detected in the periplasm in spite of the lack of a detectable signal sequence and in the absence of a functional T6SS. Ho et al attributed this observation to a “cryptic” secretion mechanism.

1. Unclear if the the loss of cell envelope integrity is a direct effect of TfeXAC2609 activity and not an artifact of cell death. The microscopy also does not show a consistent change in morphology amongst intoxicated cells as there are healthy cells adjacent to lysed cells. This needs to be investigated in much more mechanistic detail.

We observed that the X-TfeXAC2609 toxicity is dependent on its lysozyme domain since a point mutation in the active site residue (E48A) abolishes the toxicity-related phenotype in the biofilm assay (Figure 5).

1. The role for immunity proteins in cis-intoxication is not novel as proposed by the authors. For example, see PMID:22511866 and PMID:26456113 where the authors used an inducible degradation system to show that in a T6SS null strain, cis-intoxication occurs when immunity is depleted.

We thank the reviewer for pointing out these observations which are now mentioned and cited in the Introduction and in the Discussion of the revised manuscript.

Minor Revisions

1. Inconsistent use of the term "self-killing"; either refers to the killing of kin cells, or self (interchangeably used to refer to trans and cis killing).

The term “self-killing” no longer appears in the manuscript.

1. Terms trans-intoxication and cis-intoxication are convoluted and not constructive to the points being communicated. Self-killing vs kin-killing seem more intuitive and clearer. We prefer to maintain the use of the terms cis-intoxication and trans-intoxication which we defined in the Introduction, at the beginning of the Results section and in the Discussion as well as in Figure 1.
2. Readability would be improved by the removal of double negatives.

We have tried to avoid these whenever possible.

1. Bacterial competition assay in methods only refers to the E. coli competition, not the one between the different genotypes of X. citri.

Both methods were described in the same paragraph in the original manuscript. For clarity, this has now been divided into two sections in the revised manuscript: “X. citri vs E. coli competition assays” and “X. citri vs X. citri competition assays”.

1. Strain naming scheme presented on pg. 16 doesn't conform to traditional, and clearer, nomenclature typically used.

We have checked the manuscript to make sure that strain naming was consistent throughout the manuscript.

1. On Pg 25, there is a typo "X-TfiXAC2609" as opposed to X-TfeXAC2609

Thank you for the observation. This has now been corrected.

1. Line 619 - "or several other immunity proteins in competition assays"... where was this data shown? No immediate connection to any figures from this paper nor are there any references.

This is shown in Figure 2B and in Movie S7 which is now cited directly in the revised manuscript.

Reviewer #3 (Significance):

Overall it is difficult to take paradigm-conflicting conclusions at face-value when they are not presented alongside concrete experimental evidence. Without directly showing that the toxin localizes to the periplasm, the explanation that "the toxin somehow makes its way into the cell periplasm [independent of the T4SS] where it degrades the peptidoglycan layer" hinders the other conclusions presented by the authors. Consequently, my enthusiasm for this work is minimal.

We deeply appreciate the insightful feedback from Reviewer #3, particularly regarding the concerns about paradigm-conflicting conclusions. We are steadfast in our commitment to ensuring that our findings are both rigorous and scientifically relevant.

Evidence for Toxin Localization: We understand the criticality of concrete experimental evidence for toxin localization to the periplasm. While our data suggest an yet to be discovered translocation pathway of X-TfeXAC2609 from the cytoplasm to the periplasm, we recognize the importance of providing direct evidence. We are actively working on methodologies to understand this phenomenon. However, we do not believe that answering this question is absolutely necessary to understand the main conclusions of the present manuscript.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this study, the authors suggest that TfeXAC2609-TfiXAC2610 represent a novel deviation from the established paradigm in contact-dependent interbacterial secretion systems. X. citri strains lacking the predicted immunity protein, TfiXAC2610, do not suffer a competitive disadvantage when grown in T4SS-inducing conditions against a wild-type strain. Furthermore, cells lacking the immunity develop aberrant morphology and auto-lyse. The mechanism for self-intoxication by TfeXAC2609 is independent of a functional T4SS, and intoxication is exacerbated when the toxin's T4SS-signal sequence is removed.

Major Points

1. The authors of the study do not provide sufficient evidence that TfeXAC2609 contributes to T4SS mediated killing. Does the toxin behave in a synergistic way, rather than mediate killing independently? Does removing the toxin and immunity change the competitive advantage of X. citri?<br /> Suggested Experiments: Competing, against E. coli, both WT X. citri and X. citri ΔXAC2609 ΔXAC2610, and determining whether there is a change in relative competitive advantage, or expressing TfeXAC2609 in a heterologous system and marking any observed toxic phenotype.
2. Authors should directly answer where the toxin is active and localized in the cell.<br /> Suggested Experiments: Western blot subcellular fractionation (cytoplasm, periplasm, etc) to determine the localization of each protein.
3. There is no evidence that TfeXAC2609 plays any role in inter-bacterial killing besides that is predicted from its genetic arrangement and in vitro assays from a previous publication.<br /> Suggested Experiments: Again, with the available antibodies, detecting whether TfeXAC2609 is being secreted, either in competition settings against X. citri or E. coli; given that there is no killing observed in Fig. 2B, it may also be a suitable control for this experiment.
4. The structural and co-evolutionary analysis seems to miss an essential point - that the lack of fratricide protection is not due to a novel protein-protein interaction.
5. The role of the immunity in biofilm formation is confusing. Cells lacking the immunity die within 96 hours (the auto-lysis phenotype). Given that the immunity is required for viability in this time frame, wouldn't it also be required for viability after five days?<br /> Suggested Amendments: Remove or de-emphasize.
6. Why does cell permeability increase with the loss of the T4SS signal sequence? Without there being greater evidence to support that an alternative secretion system is secreting or transporting the toxin into the periplasm, which may compete with the T4SS, additional hypotheses should be experimentally probed.
7. Unclear if the the loss of cell envelope integrity is a direct effect of TfeXAC2609 activity and not an artifact of cell death. The microscopy also does not show a consistent change in morphology amongst intoxicated cells as there are healthy cells adjacent to lysed cells. This needs to be investigated in much more mechanistic detail.
8. The role for immunity proteins in cis-intoxication is not novel as proposed by the authors. For example, see PMID:22511866 and PMID:26456113 where the authors used an inducible degradation system to show that in a T6SS null strain, cis-intoxication occurs when immunity is depleted.

Minor Revisions

1. Inconsistent use of the term "self-killing"; either refers to the killing of kin cells, or self (interchangeably used to refer to trans and cis killing).
2. Terms trans-intoxication and cis-intoxication are convoluted and not constructive to the points being communicated. Self-killing vs kin-killing seem more intuitive and clearer
3. Readability would be improved by the removal of double negatives.
4. Bacterial competition assay in methods only refers to the E. coli competition, not the one between the different genotypes of X. citri.
5. Strain naming scheme presented on pg. 16 doesn't conform to traditional, and clearer, nomenclature typically used.
6. On Pg 25, there is a typo "X-TfiXAC2609" as opposed to X-TfeXAC2609
7. Line 619 - "or several other immunity proteins in competition assays"... where was this data shown? No immediate connection to any figures from this paper nor are there any references.

Significance

Overall it is difficult to take paradigm-conflicting conclusions at face-value when they are not presented alongside concrete experimental evidence. Without directly showing that the toxin localizes to the periplasm, the explanation that "the toxin somehow makes its way into the cell periplasm [independent of the T4SS] where it degrades the peptidoglycan layer" hinders the other conclusions presented by the authors. Consequently, my enthusiasm for this work is minimal.

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 explores the role of an immunity protein of the Xanthomonas type IV secretion system (X-T4SS). In contrast to most T4SSs that conjugate plasmids or transfer effectors into host cells, this system is able to kill other bacteria similar to the role of T6SSs. Here, the authors tested whether the immunity protein XAC2610 functions to prevent cis-intoxication (by self) and/or trans-intoxication (by sister cells). They provide data that the XAC2610 immunity protein functions to protect cis intoxication, but not trans-intoxication, by the T4SS effector XAC2609 (which functions as a peptidoglycan hydrolase). Based on AlphaFold modeling, they went on to identify a residue in XAC2610 that is critical for inhibiting the activity of the XAC2609 toxin. Overall the data is fairly solid and generally support the conclusions the authors made.

Major comments:

One of the major conclusions of the manuscript is that XAC2610 does not prevent trans-intoxication and the data in the manuscript support this conclusion. However, I wonder if this is an oversimplification. Notably, the authors observed that wild type Xantho was unable to kill a target cell lacking 8 different toxin/immunity systems (Fig. 1A). One could conclude that none of these immunity proteins function in preventing trans-intoxication ... or ... perhaps it appears that none perform this role because wild-type Xantho never attacks its siblings? For example, it is conceivable that Xantho uses a general mechanism, perhaps somewhat similar to phage exclusion or plasmid incompatibility, to prevent sibling attack? To me this seems more likely than none of the eight immunity proteins play a role in preventing trans-intoxication. Moreover, the phenotype observed for the ∆2610 mutant in preventing cis-intoxication is somewhat subtle, likely because the toxin and the immunity protein are topologically restricted to the cytoplasm and the periplasm, respectively. This would make sense if this were not the primary role for 2610.

Ideally the authors will be able to test this theory by demonstrating that a wild-type Xantho strain can attack (but likely not kill) its siblings. Alternatively, could the authors test if related, but not identical, Xantho strains that express 2609/2610 are able to kill their ∆2610 mutant, i.e. do "cousins" attack each other? Not sure about the semantics but this could be described as preventing trans-intoxication. If they are unable to do either experiment, that is ok but they should at least describe this concept in their discussion (assuming they agree).

Since the cis-intoxication phenotype of the ∆2610 mutant is subtle, it would strengthen the authors' conclusions on cis-intoxication if they artificially targeted XAC2609 to the periplasm with a sec signal sequence. If the authors are correct, this should be a lethal event in the absence of the 2610 immunity protein. This might be useful in terms of figuring out how the 2609 toxin normally gets into the periplasm, a major unanswered question in this manuscript.

Minor comments:

1. Figure 1 is a bit confusing in terms of the layout. It would be beneficial if the authors separated parts A and B by a few spaces.
2. Figure 2A should start off by showing that the Xantho T4SS can kill other bacteria (e.g. Fig S4A). This would set up the paper better.
3. Fig. 2A should include p values.
4. Fig. 2B is really hard to see and should be removed from the manuscript (although I do appreciate the novelty of the technique using Marilyn Monroe).
5. Instead all of the data in Fig. 2B should be shown in a new version of Fig. 2C. Fig. 2C should include additional controls including:
6. a. A wild type strain containing 2609 and 2610 mutants
7. b. A complete virB operon deletion in combination with 2609 and 2610 mutants
8. c. ∆8 strain
9. d. 2609 lacking its T4SS signal sequence
10. e. 2609 targeted to the periplasm with a sec signal sequence
11. f. etc.
12. Figure 2C. The VirB7 western band looks like in the 2610 complemented strain.
13. Figure 3C should include a comparison of exponential vs. stationary phase cells. In addition, the results for the ∆2610 mutant and the ∆2610 ∆B7 double mutant appear to be different(?). P values should be provided. If it is statistically significant, then this should be explained in the manuscript. It was not clear how the % damaged cells were calculated? # of cells? Stats?
14. The majority of Figure 4 should be replaced by assaying the effect of a virB operon deletion rather than showing the individual mutants.
15. Discussion:
16. a. The last one to two paragraphs of the results belong in the Discussion.
17. b. A more detailed description of cis-intoxication would be useful.

Significance

This work provides a conceptual advance in understanding the protective function of a T4SS immunity protein, X-Tfe XAC2610, against the cis-toxic effects of the T4SS effector, X-Tfi XAC2610. It will likely be of interest to scientists interested in T4SSs & T6SSs and interbacterial competition. Overall this is a thought-provoking manuscript and should be published in a respectable journal.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Oka et al. shows that one effector of X. citri is likely translocated into the periplasm where it cleaves PG unless inhibited by its cognate immunity protein. Interestingly, this effector is required for killing of target cells like E. coli in T4SS-dependent manner but it does not seem to be delivered into X. citri cells by T4SS. Authors show using various assays that cells lacking the immunity protein have various phenotypes including lysis and defect in biofilm formation, however, despite "cis-intoxication" the ability to kill other bacteria or infect plants remains unaffected. The manuscript is well written and in general all the experiments have proper controls and thus the conclusions seem solid. The results described here are novel and interesting as they as unexpected.

Major issues that should be addressed:

• Test various deletion variants of the toxin to identify which part of the protein is responsible for its translocation into the periplasm. This may help to identify the possible mechanism of translocation of the toxin into the periplasm. Alternatively, the authors may attempt to select for non-toxic point mutants of the toxin. This could be done by a random PCR mutagenesis of the toxin and a selection of the surviving mutants in the absence of the immunity protein.
• Test if localization of the immunity protein to the cytoplasm blocks its activity. An immunity protein mutant that lacks its secretion signal should not protect against cis-intoxication.
• While many experiments support the conclusion that the toxin is responsible for "cis-intoxication, the test of "trans-intoxication" should be done again but with the same setup as was used for testing of killing of E. coli. The CPRG based assay is far more sensitive than counting survival by plating to count CFUs. This test should be done at a relatively high initial OD so that there is an immediate contact between the "killer" and the "prey" bacteria (lacking immunity/effector). If needed, LacZ should be over-expressed in X. citri to make use of the CPRG based assay. In addition, such assay could be used also for "cis-intoxication" to supplement the potentially hard to quantify biofilm experiments shown in Fig. 4 (e.g. test all the T4SS mutants for "cis-intoxication").
• Fig. 2A needs a positive control. For example, test killing of E. coli under the same conditions.
• Authors should look at the paper by Ho et al. PNAS 2017, which describes trafficking of VgrG of V. cholerae into the periplasm of E. coli without an obvious secretion signal. The effector of X. citri may behave similarly.
• Provide some form of quantification of the phenotypes (cell rounding and cell death) observed using live-cell imaging.
• Provide quantification of biofilm related phenotypes as well as of the citrus canker development assay.

Significance

The study provides an interesting insight into immunity proteins against anti-bacterial toxins. It points to a need to protect against "cis-intoxication". This is novel and interesting to a wide audience of microbiologists interested in bacterial competition as this could be true also for other toxins. It would be however important to identify how is the toxin translocating to the periplasm of the producing bacterium. Some insight into the mechanism would vastly improve the study. My expertise is in understanding bacterial interactions and competition but I lack a direct experience with assays specific for X. citri.

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 have addressed all the queries and suggestions put forth by the reviewers. Major changes include:

1. Expansion of PILOT Functionality and Analysis: We have substantially extended the functionality and analysis capabilities of PILOT, particularly in relation to sample clustering. This enhancement now encompasses the incorporation of statistical tests aimed at identifying cell types and genes associated with distinct patient groups. We applied this expanded feature in an exploratory analysis of sub-clusters within pancreas ductal adenocarcinoma data (PDAC).
2. Clarification of Benchmarking Methods: We have provided clear elucidations of the methods employed for benchmarking PILOT alongside competing methodologies. Our benchmarking approach is notably comprehensive, encompassing twelve different datasets and evaluating four to five competing methods through statistical assessment across three problem domains: clustering, distance measurement, and trajectory estimation. The outcomes of these evaluations consistently demonstrate the superior performance of PILOT's Wasserstein metric across all three problem domains. It is noteworthy that previous studies have often limited their analyses to exploratory evaluations on individual datasets, lacking the level of comprehensive benchmarking undertaken in this study.
3. Examination of Experimental Factors: We have conducted a thorough investigation into the impacts of batch correction, cluster/cell type resolution, and parameter choices used within the PILOT framework.
4. Enhancement of Text Description: We have enhanced the textual descriptions to provide a high-level overview of the PILOT methodology, along with justifications for the methodological decisions made.
5. Improvement of Code and GitHub Repository: To enhance accessibility and promote reproducibility, we have made improvements to the codebase and the associated GitHub repository.

In summary, PILOT stands as a distinctive and all-encompassing framework. It holds the unique distinction of being the sole method offering comprehensive tools for both clustering and trajectory analysis of samples within multiscale single-cell and pathomics data. Moreover, it incorporates statistical methodologies for the interpretation of results. The effectiveness of these tools has been thoroughly validated through the most extensive benchmarking study performed to date on sample-level analysis. The versatility of PILOT is demonstrated through its successful application in exploratory analyses of three distinct datasets: elucidating trajectories in myocardial infarction single-cell RNA-seq data, uncovering trajectories within pathomics data from kidney IgNA patients, and facilitating the clustering of pancreas adenocarcinoma samples. We firmly believe that these contributions hold significant value for the fields of bioinformatics, single-cell genomics, and pathology.

Reviewer #1 (Evidence, reproducibility and clarity):

The paper describes a computational method, PILOT, that uses optimal transport to compute the Wasserstein distance between two individual single-cell samples. It uses PILOT to detect sample (patient) level trajectories and clusters associated with diseases. The method was applied separately to single-cell genomics data and to digital pathology data. The method was applied to several datasets and compared against other tools.

Major comments:

The paper is not easy to follow and should be improved considerably to make it readable and reproducible. Consequently, I was not convinced that the PILOT method is much better than other methods.

We extend our appreciation to the reviewer for their valuable suggestion. We have further refined the manuscript by incorporating a comprehensive and high-level description of our method. This expansion encompasses methodological justifications and clarifications to enhance the overall clarity. Additionally, we wish to emphasize that, to the best of our knowledge, our benchmarking analysis stands as the most comprehensive within the current literature. The results of this analysis unequivocally demonstrate that PILOT surpasses all competing methods in at least one of the various computational analysis tasks, namely clustering, trajectory estimation, and distance evaluation.

Furthermore, we have undertaken significant enhancements in the codebase of PILOT, coupled with a reorganization of the associated GitHub repository. This effort includes the development of in-depth and improved tutorials that faithfully replicate the analyses conducted on datasets related to myocardial infarction, pancreas adenocarcinoma, and pancreas pathomics (https://pilot.readthedocs.io/en/latest/). This changes guarantee the reproducibility of the PILOT framework.

See below for specific changes and additional clarifications.

At first read of the title and abstract, I got the impression that the method analyzes single cell and pathomics data concurrently rather than separately. This should be fixed.

We have changed the text of the abstract and introduction to make clear that PILOT is either applied to single cell or pathomics data independently.

The usage of Wasserstein distance to compute distance between single-cell samples is elegant and is the main strength of this study. Given that PILOT is the main achievement, it should be described more carefully and in a detailed manner.

For example, in the first Results paragraph, "The test indicates for features explaining the predicted pseudotime by fitting either linear or quadratic models" - I could not understand this sentence. Also, which test do the authors refer to? A few sentences down, there is a reference to a Wald test, is that it?

PILOT has three major parts: (1) a method for measuring distance of samples with optimal transport; (2) an patient level unsupervised analysis part (clustering or trajectory analysis) and (3) a part for explaining predicted trajectories/clustering. The sentence mentioned before, refers to the interpretation approach after trajectory analysis. Here, we fit linear, quadratic or linear quadratic models to find association of predicted sample pseudo-time with data features (gene expression values in scRNA or morphological features in pathomics data). This fit can be done for all cells in the data or only for cells from a specific type. In the case of a cell specific fit, we use a Wald test to check if the cell type fit differs from all other cell types in the data, i.e. the gene is associated with the trajectory and the expression changes are specific to the cluster at hand.

While these details were found in the method section, we agree with the referee that they can be better introduced in the main manuscript. We have therefore improved the first subsection of the results and Figure 1 to reflect this.

One of the key aspects of the Wasserstein distance is the cost metric. The determination of the cost metric should be detailed as part of the Results. Have the authors considered and estimated other ways to define the distance?

This is an interesting question. Currently, PILOT uses the Cosine metric. In our revision, we evaluate other metrics (Euclidean, Manhattan, and Chebyshev). This benchmarking indicates that the Cosine and Manhattan performed best regarding the clustering problem (ARI), while Cosine was better than all metrics for the Silhouette statistic; and Cosine and Euclidean performed best regarding AUPR. Therefore, we adopt the Cosine metric in the paper. We include these results in the revised manuscript and in Sup. Fig. 5F-H.

Figure 1 provides a schematic view of PILOT. However, there is no explanation of the notation, which makes it confusing rather than helpful. Also, what is the relationship between J and j, if any?

We understand that the figure 1 was problematic, as it did not introduce the formulation. We have now improved the first sub-section of the results page and figure 1 to improve this.

The motivation and usage of adjusted Rand index (ARI) and Friedman-Nemenyi tests should be provided. Currently, they are not clear, including why those tests are suitable in the cases shown.

The adjusted Rand index is a well known metric to evaluate clustering results when labels are known. Among others this metric has many interesting features as it does not require an association of clusters with class labels. Moreover, it has a correction for random clustering solutions, therefore values lower than zero indicate poor solutions and values of 1 a perfect solution.

The Friedman-Nemenyi test allows us to compare the performance of several algorithms whenever evaluated in the same data sets. Here, the null hypothesis is that all algorithms have the same performance (same ARI statistic). The test is nonparametric and is based on the rank of the algorithm at each data set. This is important, as ARI values (or any other evaluation statistic) are data set specific, e.g. some clustering problems are more difficult than others. By evaluating the rank, the test indicates which methods perform relatively better than others. Moreover, it follows a rigorous statistical framework including correction for multiple testing. This test has an increasing adoption in the machine learning community (Demsar et al., JMLR, https://jmlr.org/papers/v7/demsar06a.html).

We have added phrases with these justifications in the main text (subsection Evaluation of patient-level clustering and trajectory analysis) and included a new section in the materials and methods with more information in the experimental design of the benchmarking analysis.

Fig. 2 the use of method colors should be constant across panels.

We have changed the colors of panels in figure 2A-C (and equivalent panels everywhere else) to avoid confusions.

The proportions method works at least as well as PILOT in 2B and 2C (silhouette and AUPR). Explain why PILOT is better.

The benchmarking analysis shows that PILOT has the highest ARI value (clustering performance) at absolute and ranking levels (Fig. 2A). Moreover the Friedman-Neymeni test indicates this PILOT has significantly higher ranking than all evaluated methods. Regarding Silhouette (distance evaluation) and AUPR (trajectory evaluation) both proportion and PILOT have similar absolute values (Fig. 2B and 2C; panel left), while PILOT has a superior ranking in both cases (Fig. 2B and 2C panel right). Friedman-Neymeni test indicates higher ranking of PILOT than PhEMD for Silhouette and higher ranking of PILOT than PhEMD and pseudo-Bulk regarding trajectory evaluation. The difference in the results on absolute and ranking values can be understood by looking at the statistics in table Table S1. PILOT has highest AUPR in 8 out 12 data sets; proportion has highest values in 5 (including 4 ties with PILOT); proportion-PHATE had 3 best results (including 3 ties with both PILOT and proportions), while PhEMD is best in one data set and Pseudo-bulk in 3 (including 1 tie with PILOT). Altogether, PILOT obtained a higher or equal AUCPR in 9 out of the 12 data sets. We have also changed Fig.2A, 2B and 2C to include all data points and to show the mean, as this provides a better visualization of the previously reported results.

Altogether, these results indicate that PILOT outperforms all competing methods in at least one of the evaluated problems (clustering, trajectory and distance estimation) and ranks favorably in all evaluated scenarios. We have changed the manuscript text to reflect these results.

Likewise, Figure 2C,D and Figures S1 and S2 don't show a clear and consistent advantage for PILOT over other methods. Explain what advantage of PILOT do the fraction panels highlight in Fig. 2E and Fig. S3. Fig. 2C is not mentioned in the text.

Figure 2D, 2E, and now figures S2 and S3 represent visualizations of the results, which were statistically evaluated in panels of Fig.2A-2C. As discussed in the previous point, PILOT does perform better than all methods for the clustering problem and performs better or as good as the proportion test on 9 of the 12 evaluated data sets in the trajectory problem. We also have improved the text to include references to all figures in the main text.

I assume Kidney IgAN (text) and Kidney IgA (fig. 2) are the same.

The correct name is IgAN and this has been corrected in Figure 2.

Fig. 3B fix the p-value notation (what is p=1.05E?) and R2 (R square?). Nrte tha both this problem also occurs in other figs. Fig. 3B shows the major cellular changes.

We now adopt the term “R-squared” in the figures. Also, the previous version did not display p-values properly. We apologize for this. This has been fixed now.

Are these changes consistent with known ones? Explain and provide references. Are there cell types that were expected to show a change and did not? Same questions for Fig. 3C wrt genes. Is this an exploratory analysis highlighting interesting candidate genes? If so, it should be described as such.

Cardiac remodeling after myocardial infarction is characterized by loss of cardiomyocytes, infiltration by immune cells (myeloid and lymphocytes) and increase in myofibroblast populations (doi.org/10.1038/s41392-022-00925-z;doi.org/10.3389/fcvm.2019.00026). PILOT indicates these populations, with the exception of lymphocytes, are most relevant at both clustering levels (see Sup. Fig, 6). Particularly important are results from the low granularity analysis, as this indicates particular macrophage/fibroblast sub-populations (SPP1+ Mac. and Myofibroblast) with increase in disease. PILOT could not detect changes in lymphocyte cells, but this is explained by the poor coverage of these cells in the data set (>3%). We have updated the main manuscript to reflect this.

We also explicitly mention that the analysis of genes and cells are exploratory analysis.

The point of Fig. S6 and its major findings should be mentioned in the text (or it can be removed).

We now make the reference to the gene ontology analysis presented in the new Figure S7 more explicit in the text.

Fig. 4B legend - eGFR not GFR. What do the high-low values of Fig. 2B refer to?

We have fixed these points.Hhigh and low values of panel 4B refer to the eGFR.

Fig. S12 is out of order in supp file.

This has been fixed.

AUCPR - explain.

The AUCPR stands for area under the curve of the precision recall (AUCPR) curve. We have now improved the explanation of the evaluation metric in the main text and methods section.

The github looks like work in progress with many internal comments (eg, add ,edit, etc). I could not find the tutorials.

We have removed all the comments, improved the repository organization and code. The tutorials are explicitly mentioned in the main github page (https://github.com/CostaLab/PILOT/) and in readthedocs webpage (https://pilot.readthedocs.io/). It include tutorials replicating analysis with trajectory inference and clustering problems, which are discussed in the manuscript.

In the process of code review, we have noticed that while we could replicate all the analysis, the procedure for selection of healthy cardiomyocyte genes was distinct (gene were ranked by regression model fit p-value) than the analysis of the myofibroblast genes (genes were ranked by the Wald test p-value). As explained before, the Wald test, which compares the expression of the regression model fits across samples, is a more appropriate criteria, as it finds cluster and trajectory specific genes. We have changed the analysis of the cardiomyocyte to make the gene selection to be based on the Wald-test p-value. New results recover other sarcomere related genes (MYBPC3 and MYOM1) as being dysregulated during disease progression. These findings are in accordance with observations made in the original study presenting the data (Kuppe et al. 2022). We have updated Fig.3 and respective genes accordingly.

Minor comments:

Introduction: "Alternatively, trajectory analysis can be performed to uncover disease progression allowing the characterization of early disease events." Citations should be added (some appear later in the text).

We included a reference to PhEMD.

"Currently, there are no analytical methods to compare two single cell experiments from the same tissue from two distinct individuals." There have been several comparisons among data from patients, (e.g. Cain et al, 2023), so the authors should be more careful/accurate in their statements.

We assume that the referee mentions https://www.nature.com/articles/s41593-023-01356-x. Indeed, we were not aware of this recently published study. The manuscript focuses on comparing cell proportion changes (estimated by deconvolution) between distinct phenotypes and does not provide any approach for sample level analysis of single cell data. This is in our view a different problem than the one addressed by PILOT or PhEMD. We refer to it in our manuscript, as its cell community based analysis is an interesting approach for the interpretation of PILOT results.

"Except for PhEMD, all related methods9, 11, 12 require labels of patients for their analysis and cannot be used in the unsupervised analysis " - this sentence comes immediately after describing ref 13, which can be used in unsupervised analysis and accordingly is not cited in this sentence. The authors did well in describing ref 13 (a bioRxiv paper), and its description should come after this sentence.

We changed the text to reflect this.

"These can be clusters", clustered?

Done.

" acquire an injury cell states" remove an.

Done.

"As for scRNA-seq, there is no analytical method which is able to compare two or more histological slides based on morphometric properties of their structures." The sentence seems to refer to pathomics, not to sc data as suggested in "As for scRNA-seq"

This has been rephrased.

"Thus PILOT represents the first approach to detect unknown patient trajectories and clusters" patient clusters were also observed by others (eg ref 13, Cain et al).

This has been rephrased.

Equation 7 - Cosine(Mi,Mi) should be Cosine(Mi,Mj)

Done.

In the beginning of the Results, PILOT is not referred to as a package but as a researcher ("PILOT explores").

This has been rephrased.

Reviewer #1 (Significance):

In general, the paper is a Methods paper. Hence, likely audience includes computational biologists interested in methodologies, not to biologists interested in the actual findings.

Although I am among the likely audience, I was not convinced by the merits of the method, potentially due to the way the paper was written.

I do not have sufficient expertise to check the math.

In this revision, we have significantly enhanced the text to incorporate high-level descriptions of methods tailored for non-computational experts. Additionally, we have refined the description of the benchmarking process, which, as far as our knowledge extends, stands as the most comprehensive in the literature. This comprehensive analysis strongly underscores the statistical superiority of PILOT when compared to other methods. Lastly, PILOT presents an unique framework, encompassing methods for trajectory analysis, clustering, and interpretation of sample-level analyses within the realm of multiscale single-cell genomics and pathomics data.

Reviewer #2 (Evidence, reproducibility and clarity):

Joodaki et al. propose PILOT, a computational method for analysing single-cell genomics and pathomics data. PILOT is a method that enable clustering, trajectory analysis, and feature detection at a patient level using scRNA-seq data. This is an important task and represent the growing application of scRNA-seq to understand diseases and other perturbations to other biological systems. In particular, PILOT enables unsupervised analysis which alleviate the need of patient labels required by many alternative methods. We have the following comments for the authors' consideration.

1. A key consideration in dealing with scRNA-seq data at a patient level is the batch effect in the data. Typically, each patient sample may be treated as a "batch" especially when they are processed separately to obtain a scRNA-seq dataset that are subsequently combined with scRNA-seq datasets from other patients to form a single dataset. Analysing these data without batch correction may lead to the identification of "cell types" and "states" that are driven by batch effect. In Figure 1. PILOT takes a clustered and integrated scRNA-seq data as input for analysis. I wonder how PILOT would behave if there is a strong batch effect in the data and how would the authors propose to handle them?

This is an interesting question. Currently, PILOT is using the batch procedure used in the paper proposing the original data. We evaluate now the impact of batch correction methods implemented in scanpy (Harmony, bknn and Scanorama). We focus here on single cell data, which we have access to the original count matrix (Lupus, COVID, and Diabetes). We observe no impact of the batch correction algorithm in these data sets (see Sup. Fig. 5C-E). These results are now included in the manuscript.

We have noticed however that strong batch effects in the lung cell atlas or the kidney cell atlas.For the lung cell atlas, we observed that single cell data measured from distinct techniques (Seq-well, Drop-seq, 10x 5’ and 10x 3’) or distinct tissue sampling approaches confounded results for all evaluated approaches. Therefore, we restricted the analysis to the technology with more samples (10x genomics 3’) and to lung tissues. This sample selection was previously described in the material and methods. Of note, the use of samples from distinct 10x genomic version kits (v1, v2 or v3) did not impact results. For the kidney cell atlas, we also observed a strong batch between single nuclei and single cell protocols. Here, we opted to focus on the largest cohort of single cell RNA experiments (see Review Fig. 1). Altogether, PILOT and other evaluated methods do require samples to be analyzed with an uniform technique and sampling approaches. We now include a brief discussion about this open point in the “Discussion” section. This is an important topic of future research.

Review Fig. 1. - Data of the Kidney Precision Medicine Project was measured using either single cell or single nucleus protocols. All evaluated methods were affected by the differences in these technologies and could not separate disease status in this data.

1. It appears that the Wasserstein distance (W) matrix of the samples was used for patient clustering and also trajectory analysis. However, most of the figures presented in the manuscript are for trajectory analysis. Since the patient clustering were performed prior to trajectory analysis, could the authors visualize the patient data based on the W before performing disease trajectory estimation?

Indeed, despite the clustering-based analysis (ARI statistics; Fig. 2A) the current manuscript focuses on results of the trajectory analysis. We now include additional features for clustering analysis. This includes heatmap visualizations of the OT distance matrices together with Leiden clustering (Sup. Fig. 1). See points 4 and 5 below for further changes regarding clustering analysis.

1. In trajectory analysis in figure 2D and E, why not use Multi-scale PHATE which appears to be specifically designed for trajectory analysis? The authors also mentioned SCANCell. While these methods require labels of patients for their analysis, it would be interesting to know how well they perform in comparison to PILOT if such information is available.

This is an interesting point. Multiscale-PHATE is based on doing a multi-resolution clustering of the cells. It then applies PHATE (instead of diffusion map) to find a non-linear embedding on the cell proportions across samples and resolutions. While this analysis is presented at Multiscale-PHATE manuscript (Fig. 5), we could not find any code or functionality in their github to replicate this (https://github.com/KrishnaswamyLab/Multiscale_PHATE). Moreover, we were not able to find a function to find the cluster/resolution associations of cells to reimplement the above mentioned analysis following the descriptions of the manuscript. We also contacted authors, but obtained no reply. It is also important to state that Multi-PHATE used a supervised filter to select cell types for further analysis.

Alternatively, we now include an evaluation of the use of cell proportions followed by a PHATE embedding in the trajectory based evaluation, which is close to the method proposed in Multiscale-PHATE. Our benchmarking indicates that Multiscale-PHATE is the third best ranked method being overpassed by proportion and PILOT. Regarding SCANCell, it focuses on the interpretation of cell communities and it uses embedding/distances by exploring PhEMD. Therefore, its performance in the trajectory or clustering performance problem is the same as PhEMD. We refer to these points in the text now.

1. The current design of PILOT appears to assume that there is always a "smooth" trajectory in the data. Is this going to be the case in reality? What if we have a well separated and distinct groupings of the patients and controls data? In the latter case, imposing a trajectory seems artificial. I am also not sure how meaningful the trajectory analysis would be if, biologically, such a smooth transition is not present in the data.

The EMD based distance can be used both for clustering or trajectory analysis. Also, PILOT performed quite well in the clustering problem benchmarking (Fig. 2A). The choice of application lies on the problem at hand. In our view, both the kidney pathomics and the myocardial infarction data (explored in Fig. 3 and Fig. 4) represent medical data with potential disease trajectories. We now expand the PILOT framework to include new visualizations and statistical methods to improve the interpretation of the clusterings (see point 2; Fig. S1; and point 5 and new Fig. 5).

1. The feature analysis is also built on trajectory analysis using regression models. Again, how would this work out if there isn't a smooth trajectory/transition in the data (e.g. the data are obtained from a discrete case-control study)?

We expanded the PILOT framework to also include statistical tests for accessing changes in cell populations and markers for the clustering problem. First, we use a Welch’s t-test to evaluate cell proportion changes associated with detected clusters. Next, we use a differential analysis test from limma to find genes within a cluster, whose gene expression is changing between the two groups of samples for a given cluster of cells. While these are standard approaches in the literature, this further improves the functionality of PILOT as a general framework for patient level analysis. This is now described in PILOT manuscript (Results subsection “Patient level distance with Optimal Transport” and methods).

We also include in the manuscript an explorative analysis of a sub-cluster found in the PDAC data. This analysis could find a population of PDAC patients displaying higher levels of malignant cells and marked by both increase in hypoxia and fibrosis pathways. This example highlights how PILOT can be used to find potentially interesting groups of samples. These are implemented in the main manuscript (Fig. 5). We also include a new tutorial of PILOT with this analysis (see https://pilot.readthedocs.io/).

1. It is not clear from the formulation of PILOT (and also Figure 1) if the cell type labels is required/used or the cluster id of a clustering algorithm was used instead. The author also mentioned that the clustering output does not have much impact on the downstream analysis. I wonder why and if so can we group the data in any way we want for downstream analysis? This can be useful when one would like to focus on certain grouping of cells.

Clustering at the cells (or structure level) is required. For the benchmarking analysis, we have used the cell annotation reported in the original paper, which were derived via clustering analysis. The use of annotated clusters is crucial for interpretation. We also included in the original manuscript an analysis on the impact of the clustering resolution of the Leiden algorithm. This indicates that the change in resolution did not have a high impact in the clustering (ARI) of the samples (Sup. Fig. 5A-B). However, this analysis could not consider any interpretation of results, as cluster labels were not present.

We believe, however, that the granularity of the clustering will impact the interpretation of the sample analysis. To investigate this, we evaluate how using higher level annotation/clustering of the heart myocardial infarction (also reported in Kuppe et al. 2022) impacts our ability to find cell specific changes. We observe similar changes whenever using low resolution clustering (decrease of cardiomyocytes, increase in fibroblasts and myeloid cells). However, this analysis loses a lot of important nuances found in the high resolution clustering (see Sup. Fig. 6). For example, it does not recover the fact that damaged cardiomyocyte populations have a slower decay than healthy myocytes. Or the fact that myofibroblasts has an increase in the latter disease trajectory stage, while progenitor fibroblast cells (Fibro_Scara5) have an increase previous to myofibroblasts. These results show how low resolution clustering can lead to loss of interesting information contained in cellular sub-states or cell sub-populations. This is now discussed in the results subsection ‘PILOT trajectories detect events associated with cardiac remodeling in myocardial infarction’.

Reviewer #2 (Significance):

PILOT is designed for analyzing scRNA-seq data at a patient level. There is a growing application of scRNA-seq to diseases and the development of computational tools for analyzing such data at phenotype level is critical. The key aspect of PILOT compared to other currently available tools is that it enables unsupervised analysis which alleviate the need of patient labels required by many alternative methods.

Thanks for this very positive feedback and constructive comments.

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

Joodaki et al. propose PILOT, a computational method for analysing single-cell genomics and pathomics data. PILOT is a method that enable clustering, trajectory analysis, and feature detection at a patient level using scRNA-seq data. This is an important task and represent the growing application of scRNA-seq to understand diseases and other perturbations to other biological systems. In particular, PILOT enables unsupervised analysis which alleviate the need of patient labels required by many alternative methods. We have the following comments for the authors' consideration.

1. A key consideration in dealing with scRNA-seq data at a patient level is the batch effect in the data. Typically, each patient sample may be treated as a "batch" especially when they are processed separately to obtain a scRNA-seq dataset that are subsequently combined with scRNA-seq datasets from other patients to form a single dataset. Analysing these data without batch correction may lead to the identification of "cell types" and "states" that are driven by batch effect. In Figure 1. PILOT takes a clustered and integrated scRNA-seq data as input for analysis. I wonder how PILOT would behave if there is a strong batch effect in the data and how would the authors propose to handle them?
2. It appears that the Wasserstein distance (W) matrix of the samples was used for patient clustering and also trajectory analysis. However, most of the figures presented in the manuscript are for trajectory analysis. Since the patient clustering were performed prior to trajectory analysis, could the authors visualize the patient data based on the W before performing disease trajectory estimation?
3. In trajectory analysis in figure 2D and E, why not use Multi-scale PHATE which appears to be specifically designed for trajectory analysis? The authors also mentioned SCANCell. While these methods require labels of patients for their analysis, it would be interesting to know how well they perform in comparison to PILOT if such information is available.
4. The current design of PILOT appears to assume that there is always a "smooth" trajectory in the data. Is this going to be the case in reality? What if we have a well separated and distinct groupings of the patients and controls data? In the latter case, imposing a trajectory seems artificial. I am also not sure how meaningful the trajectory analysis would be if, biologically, such a smooth transition is not present in the data.
5. The feature analysis is also built on trajectory analysis using regression models. Again, how would this work out if there isn't a smooth trajectory/transition in the data (e.g. the data are obtained from a discrete case-control study)?
6. It is not clear from the formulation of PILOT (and also Figure 1) if the cell type labels is required/used or the cluster id of a clustering algorithm was used instead. The author also mentioned that the clustering output does not have much impact on the downstream analysis. I wonder why and if so can we group the data in any way we want for downstream analysis? This can be useful when one would like to focus on certain grouping of cells.

Significance

PILOT is designed for analysing scRNA-seq data at a patient level. There is a growing application of scRNA-seq to diseases and the development of computational tools for analysing such data at phenotype level is critical. The key aspect of PILOT compared to other currently available tools is that it enables unsupervised analysis which alleviate the need of patient labels required by many alternative methods.

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 paper describes a computational method, PILOT, that uses optimal transport to compute the Wasserstein distance between two individual single-cell samples. It uses PILOT to detect sample (patient) level trajectories and clusters associated with diseases. The method was applied separately to single-cell genomics data and to digital pathology data. The method was applied to several datasets and compared against other tools.

Major comments:

The paper is not easy to follow and should be improved considerably to make it readable and reproducible. Consequently, I was not convinced that the PILOT method is much better than other methods.

At first read of the title and abstract, I got the impression that the method analyzes single cell and pathomics data concurrently rather than separately. This should be fixed.

The usage of Wasserstein distance to compute distance between single-cell samples is elegant and is the main strength of this study. Given that PILOT is the main achievement, it should be described more carefully and in a detailed manner.

For example, in the first Results paragraph, "The test indicates for features explaining the predicted pseudotime by fitting either linear or quadratic models" - I could not understand this sentence. Also, which test do the authors refer to? A few sentences down, there is a reference to a Wald test, is that it?<br /> One of the key aspects of the Wasserstein distance is the cost metric. The determination of the cost metric should be detailed as part of the Results. Have the authors considered and estimated other ways to define the distance?

Figure 1 provides a schematic view of PILOT. However, there is no explanation of the notation, which makes it confusing rather than helpful. Also, what is the relationship between J and j, if any?

The motivation and usage of adjusted Rand index (ARI) and Friedman-Nemenyi tests should be provided. Currently, they are not clear, including why those tests are suitable in the cases shown.

Fig. 2 the use of method colors should be constant across panels. The proportions method works at least as well as PILOT in 2B and 2C (silhouette and AUPR). Explain why PILOT is better. Likewise, Figure 2C,D and Figures S1 and S2 don't show a clear and consistent advantage for PILOT over other methods. Explain what advantage of PILOT do the fraction panels highlight in Fig. 2E and Fig. S3. Fig. 2C is not mentioned in the text.

I assume Kidney IgAN (text) and Kidney IgA (fig. 2) are the same.

Fig. 3B fix the p-value notation (what is p=1.05E?) and R2 (R square?). Norte tha both this problem also occurs in other figs. Fig. 3B shows the major cellular changes. Are these changes consistent with known ones? Explain and provide references. Are there cell types that were expected to show a change and did not?<br /> Same questions for Fig. 3C wrt genes. Is this an exploratory analysis highlighting interesting candidate genes? If so, it should be described as such.<br /> The point of Fig. S6 and its major findings should be mentioned in the text (or it can be removed).

Fig. 4B legend - eGFR not GFR. What do the high-low values of Fig. 2B refer to?<br /> Fig. S12 is out of order in supp file.

AUCPR - explain.

The github looks like work in progress with many internal comments (eg, add ,edit, etc). I could not find the tutorials.

Minor comments:

Introduction: "Alternatively, trajectory analysis can be performed to uncover disease progression allowing the characterization of early disease events." Citations should be added (some appear later in the text).<br /> "Currently, there are no analytical methods to compare two single cell experiments from the same tissue from two distinct individuals." There have been several comparisons among data from patients, (e.g. Cain et al, 2023), so the authors should be more careful/accurate in their statements.<br /> "Except for PhEMD, all related methods9, 11, 12 require labels of patients for their analysis and cannot be used in the unsupervised analysis " - this sentence comes immediately after describing ref 13, which can be used in unsupervised analysis and accordingly is not cited in this sentence. The authors did well in describing ref 13 (a bioRxiv paper), and its description should come after this sentence.

"These can be clusters", clustered?

" acquire an injury cell states" remove an.

"As for scRNA-seq, there is no analytical method which is able to compare two or more histological slides based on morphometric properties of their structures." The sentence seems to refer to pathomics, not to sc data as suggested in "As for scRNA-seq"

"Thus PILOT represents the first approach to detect unknown patient trajectories and clusters" patient clusters were also observed by others (eg ref 13, Cain et al).

Equation 7 - Cosine(Mi,Mi) should be Cosine(Mi,Mj)

In the beginning of the Results, PILOT is not referred to as a package but as a researcher ("PILOT explores").

Significance

In general, the paper is a Methods paper. Hence, likely audience includes computational biologists interested in methodologies, not to biologists interested in the actual findings.

Although I am among the likely audience, I was not convinced by the merits of the method, potentially due to the way the paper was written.

I do not have sufficient expertise to check the math.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary:<br /> In this study, the authors delineate the association of paralog dispensability with the frequency of homozygous deletions (HDs) and thereby show that paralog dispensability can play a significant role in shaping tumor genomes. The authors analyzed the strength of negative selection on the paralogs relative to the singletons using frequencies of the homozygous deletions (HD). The study focused on HDs because they ensure a complete loss of function, unlike other mutational aberrations that can be masked because of haplo-sufficiency. While accounting for potential confounding factors, authors find that paralogs tend to have a relatively high frequency of HDs, suggesting a relaxed negative selection. Furthermore, the authors specifically attribute this association to the dispensable paralogs by analyzing gene inactivation data generated from multiple experimental systems. Overall, the findings of this study can potentially have significant implications in cancer biology field and specifically to the researchers studying cancer genome evolution.

We thank the reviewer for the careful reading and positive assessment of our manuscript

Major comments:

1. To dissect further which dispensable paralogs are more likely to be associated with a high HD frequency, synthetic lethal paralogs could be compared with non-synthetic lethal ones.

In the section titled 'Homozygous deletion frequency of paralog passengers is influenced by paralog properties' (begins from line #289), authors have shown that paralogs with a high frequency of HDs are more likely to have the properties of dispensability (in Figure 4). It seems that all of those properties are also associated with synthetic lethality as the authors identified in their previous study (DeKegel et al. 2021). Furthermore, as shown in the subsequent section ('Essential paralogs are less frequently homozygously deleted than non-essential paralogs', begins from line #344), the high HD is associated with the dispensable paralogs. Some of those dispensable paralogs are expected to be synthetic lethal. Therefore, the association of paralogs with a high frequency of HDs with experimentally validated or predicted sets of synthetic lethal paralogs could be tested. This may help authors to contextualize their findings in terms of genetic interactions between paralogs.

We thank the reviewer for highlighting the potential relationship with our previous work. We agree that many of these properties are associated with synthetic lethality, but we note that they are also associated with single gene essentiality. This makes the relationship between synthetic lethality, essentiality, and deletion frequency somewhat difficult to dissect.

Nonetheless we have tested, in a number of ways, whether there is a relationship between a paralog having a reported/predicted synthetic lethality and being homozygously deleted. We find no obvious connection between the two.

We first tested using a set of synthetic lethal interactions identified by integrating molecular profiling data with genome wide CRISPR screens in a large panel of cancer cell lines (the data used to train the classifier in De Kegel et al, 2021). As there is an ascertainment bias in this dataset (paralogs must have frequent loss of function alterations / silencing to be tested) we restricted our analysis to only those paralog pairs tested for synthetic lethality. We identified no clear pattern (p>0.05, Fisher's Exact Test).

We next tested using an integrated set of four combinatorial CRISPR screens (aggregated in De Kegel et al) where we considered a pair to be synthetic lethal if it was a hit in any screen and not synthetic lethal if it was screened at least once and never identified as a hit. Again we restricted our analysis to paralogs that were present in this dataset to prevent issues with ascertainment bias. We found no clear association.

We further tested using a consensus dataset derived from the same combinatorial screens, where a pair were marked as synthetic lethal if they were identified as a hit in at least two screens and not synthetic lethal if they were screened at least twice and never identified as a hit. Again we restricted our analysis to paralogs that were present in this dataset and found no clear association.

We finally tested using our predicted synthetic lethal interactions – annotating the top 3% of predictions as synthetic lethal and the remainder as non-synthetic lethal. The 3% threshold is similar to the observed frequency of synthetic lethality in the training set. In this case, as this dataset covers all paralogs considered, no restriction was necessary.

None of the above analyses revealed a clear relationship between deletion frequency and synthetic lethality. A caveat of these analyses is that none of the experimental datasets are complete (covering only a minority of all paralog pairs) and they are all somewhat noisy. Furthermore, as we show in our modelling analysis (Fig S3) the observed homozygous deletions are far from saturating.

However we think there may be a simpler explanation, beyond limitations of the data, for why we do not observe a relationship between HDs and synthetic lethality.

As the reviewer notes, there is evidence in cell lines that one reason paralogs are more dispensable than singletons is because of buffering / redundant relationships as revealed by synthetic lethal interactions. These relationships therefore provide an explanation for why some paralogs are dispensable. As our primary claim is that paralogs are more frequently deleted because they are more dispensable we might anticipate a relationship between deletion frequency and synthetic lethality. However, by definition, synthetic lethal interactions can only be observed for non-essential (dispensable) genes. Therefore when analysing the overlap with synthetic lethal interactions we are primarily restricting our analyses to genes that are already individually dispensable. Consequently we might not anticipate observing any enrichment. The buffering relationship revealed by synthetic lethality provides an explanation for why a paralog is dispensable but once we are restricting our analysis to dispensable paralogs we do not necessarily expect to see further enrichment.

We think that an ideal way to explore this question further would be to look at selection on deletions of pairs of paralogs – we anticipate that if a gene is dispensable because of paralog buffering then both paralogs should not be deleted simultaneously. However, the current copy number datasets are too small to evaluate such pairwise relationships. This is discussed in manuscript as follows:

Analyzing the frequency with which two members of a paralog family are lost would provide more direct insight into the contribution of paralog redundancy, but due to the overall rarity of passenger gene HDs, we cannot make a comprehensive assessment of co-deletions here – e.g. among paralog pairs where both genes are non-drivers, and not on the same chromosome, only two pairs are co-deleted in at least one TCGA sample. Larger cohorts would also allow us to search for patterns of mutual exclusivity of HDs to identify genetic interactions – this has been applied for identifying interactions between driver genes [57,58]__, but is more challenging for interactions between non-driver genes, which are much less frequently altered.

Minor comments:<br /> 1. The number of TCGA and ICGC tumor samples analyzed:<br /> As mentioned in the Results section (line #106), 9966 tumor samples were analyzed. However, the sample size mentioned in Figure 2A is 9951. Is the lower number shown in the figure due to the filtering procedure mentioned in the Methods section (line #455)? The change in sample sizes could be explained. A similar difference in sample sizes exists for the ICGC data also.

The difference was indeed due to filtering process, but numbers were only provided in the methods. We have now addressed this in the main text :

After removing a small number of ‘hyper-deleted’ samples (see Methods) we retained 9,951 samples for further analysis.

1. The rationale behind setting the threshold at 100 HD genes to classify 'hyper-deleted' samples for TCGA (line #462) and ICGC data (line #473) could be explained.

We excluded hyper-deleted samples to avoid any individual sample having undue influence on the genes observed to be ever deleted or indeed to influence the overall patterns observed. It is also common in analyses of selection in tumours that make use of mutational profiles (rather than copy number profiles) to exclude hypermutated samples (e.g. Martincorena et al, Cell 2017; Lopez et al, Nature 2020). However the exact threshold of 100 samples was somewhat arbitrary and this query prompted us to assess whether it had any significant impact on the results.

We therefore repeated all analyses using a more stringent threshold (50 samples) and also without thresholding. Although the exact percentages and odds-ratios vary somewhat with the different thresholds, all major conclusions are still supported.

We appreciate that this was minor comment and that reviewer did not request this new analysis, but in the absence of a strong justification for a single threshold we felt it appropriate to assess multiple thresholds (and none).

1. Citation for DepMap is missing (caption of Figure 5). We have added the text below to the legend for Figure 5 :

Essential genes for the DepMap dataset (Meyers et al, 2017) are obtained from a version of the data reprocessed in (De Kegel et al, 2021) to reduce off-target sgRNA effects (see Methods).

CROSS-CONSULTATION COMMENTS<br /> Along the lines of Reviewer #3's second major comment, I have a suggestion, the possible benefits of which would depend on the target audience to which the authors intend to communicate their study.

I would suggest including a brief comparison of the findings of this study which deal with human paralogs, with the findings in model organisms such as yeast, perhaps in the discussion section. To facilitate such a comparison, authors could try measuring the enrichments of, for example, molecular functions, gene families, types of genetic interactions, etc., among the paralogs associated with a high frequency of HDs and then discussing their comparison with what is known in the literature for paralogs in other model organisms that tend to be frequently deleted.

Such a comparison could be of interest to the community of researchers working on other model organisms and put this study in a much broader context. However, as I said before, this would depend on the authors' intended target audience.

We thank the reviewer for the suggestion. We have added an additional section to the discussion highlighting differences and similarities to the observations from yeast as follows:

Much of our understanding of the factors that influence gene dispensability comes from studies in model organisms, in particular the budding yeast Saccharomyces cerevisiae [3,9,10,43,44]__. Analyses of the yeast gene deletion collection, a set of gene deletion mutants systematically generated in a single S. cerevisiae strain, revealed that paralogs were less likely to be essential than singleton genes [3,45]__. Furthermore, more detailed analyses of yeast paralogs revealed that paralogs from large families were less likely to be essential as were genes with highly sequence similar paralogs [43,44]__. Previous analyses, including our own, demonstrated that many of these trends are also evident when analyzing gene essentiality from CRISPR screens in cancer cell lines [12,13,15,35]__. Our results here are also consistent with these findings – many of the features that are associated with paralog dispensability in yeast are also associated with gene deletion frequency in tumor genomes.

The connection between the budding yeast observations and those in cancer is less clear when it comes to the relative dispensability of WGDs and SSDs. Analyses of the yeast gene deletion collection revealed that SSDs are more likely to be essential than WGDs in the single genetic background studied [43,44]__. In our previous analyses of gene essentiality in hundreds of cancer cell lines we found that SSDs were more likely to be broadly essential (essential in most cell lines) than WGDs but that WGDs were less likely to be never essential (i.e. more likely to be essential in at least one cell line)__[13]__. As the analyses of gene essentiality in budding yeast were generated in a single genetic background the concordance with our cancer cell line results was difficult to assess, but as gene deletion collections are now being generated in additional yeast strains it should become possible to perform a more direct comparison__[46–48]__.

Here we found that WGDs are less likely to be deleted than SSDs in tumors. This is surprising in light of the yeast gene deletion collection results, where SSDs were more likely to be essential than WGDs in the strain studied, but less so in light of the cancer cell line results, where WGDs were less likely to be never essential. It is also worth noting that experimental evolution studies in yeast found that SSDs accumulate protein-altering mutations at a higher rate than WGDs [49,50]__. These results are perhaps especially relevant when analyzing the influence of paralog features on selection in tumors.

We note that there are many additional differences in the features of WGDs and SSDs in budding yeast that may alter their relative dispensability in tumors. An obvious large scale difference is that in the ancestor of humans there were two rounds of whole genome duplication compared to a single duplication event in yeast__[51,52]__. Less obvious, but potentially of importance for cancer, is that the two classes of paralogs are enriched in pathways in humans that do not have obvious counterparts in yeast. For example, WGDs are highly enriched in signaling pathways involved in development while SSDs are enriched in immune response genes__[53]__. How the membership of these pathways influences the dispensability and selection of genes in tumors and cancer cell lines warrants further study.

Reviewer #1 (Significance):

As the authors note in their manuscript, it is expected that paralog dispensability could be associated with the relaxed negative selection in tumor genomes because (1) paralogs are prevalent in the human genome, and (2) many of them are dispensable, as apparent from the large-scale gene inactivation screens in hundreds of cancer cell lines (Blomen et al. 2015, Wang et al. 2015, Dandage and Landry 2019, De Kegel and Ryan 2019). However, direct mapping of this association, while importantly accounting for potential confounding factors, has been lacking.<br /> As a researcher with prior experience in the research topics such as gene duplication and genetic interactions, it appears to me that this study presents formal proof of the important association between paralog dispensability and tumor genome evolution which could be of major implication for the research community of cancer biology field and specifically to the researchers dealing with the topics such as cancer evolution, copy number alterations in cancer genomes, and synthetic lethality-based precision oncology therapeutics.

Thank you again for the positive assessment.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary

Here, De Kegel & Ryan analyse thousands of tumour samples from the TCGA and ICGC projects to identify homozygously deleted genes, finding that about 40% of protein-coding genes are deleted in at least one sample. They find homozygously deleted genes to be enriched for paralogous genes, and further, more frequently deleted genes are increasingly likely to be paralogs. The authors then test the influence of several factors on the likelihood of being deleted, including gene length, distance to a fragile site or chromosomal region, and distance to a recurrently deleted tumour suppressor gene (TSG). They find that proximity of a TSG, telomere, centromere, and fragile site all increase likelihood of being deleted in a sample, as does gene length. Having a paralog also remains an important predictor of deletion after accounting for these other factors. Additionally, the more similar in sequence the closest paralog is to the gene and having a larger gene family size are also predictive of deletion. Conversely, if a gene is a whole genome duplicate as opposed to a small-scale duplicate, it is less likely to be deleted. Finally, the authors test the hypothesis that paralogs that are deleted in cancer are less likely to be essential and find that this is indeed the case.

Comments

The authors have done a good job of identifying trends of paralog deletion in cancer samples and the factors influencing them. The results are well described and presented and support the conclusions. I like the inclusion of the saturation analysis as an estimate of what to expect given current and potential future sample sizes, and I appreciate the inclusion of a WGD/SSD paralog distinction. The data and methods are sufficiently detailed. I have a few minor comments below.

We thank the reviewer for the careful reading and positive assessment of our manuscript

1. Around line 160 in the text and supplemental figure 4A, the authors test if the trends they see are observed across individual cancer types. With 9 of 33 cancer types reaching a sample size threshold, 8 of 9 comparisons are significant. The authors do not state correcting for multiple testing.

We have now also assessed the significance of the results after performing a Holm-Bonferroni correction for multiple hypothesis testing and find that all 8/9 cancer types remain significant.

1. I initially misunderstood the hemizygously deletion analysis, thinking the analysis in supplement figure 4B/C was asking if a sample has any singleton or any paralog deleted and comparing the number of samples with any deletion of either - given the number of genes deleted per sample this wouldn't make sense as a good test. I think the authors are actually comparing the number of loss-of-hemizygosity events per gene and grouping by paralog/singleton. I think this is a good analysis, but I think it would be helpful to clarify this in the text and figure legend e.g. "Samples w/ gene LOH" could be "LOH events per gene" or something similar.

As suggested we have now updated the y-axis label in these charts to ‘LOH events per gene’. We note that there are now two additional panels in this figure to address copy neutral LOH, per Reviewer 3’s request.

1. Occasionally, I wanted some more detail in the text for context, which was sometimes later provided - e.g. I noted when reading about line 125 that I was curious at this point how often TSGs occurred on segments, and this was later provided on line 241. Similarly, around line 114 I was curious how many genes are typically deleted per HD segment, for which the median value was provided on line 206 (and distribution in supplemental figure 1), and again for hemizygous deletions. I think sometimes it would be helpful to provide this context earlier in the text to aid interpretation of the results.

We thank the reviewer for these suggestions which we have now incorporated into the text.

On line 115 (previously 114) the relevant sentence now reads:

Typically an HD that results in the loss of a protein coding gene also results in the loss of several chromosomally adjacent genes – in the TCGA dataset a median of three genes are lost per gene-deleting HD segment

On line 124 the relevant sentence now reads:

We found that almost half (49%) of the HDs that result in the loss of at least one protein coding gene overlap a known tumor suppressor.

1. In the discussion, on line 420, the authors include the point that a paralog might not be required at all in a tumour cell and therefore easily deleted. I think this possibility could be expanded on here and in the introduction/results section, as it is an important point. I think it would be helpful to include more about the possibility that a paralog might be deleted in a tumour cell because it is simply not required or that is more likely to have less of a phenotypic impact compared to a singleton, and that this could be a reason for the observed enrichment of paralogs in deleted genes. A citation to support this point could be Áine N O'Toole, Laurence D Hurst, Aoife McLysaght, Faster Evolving Primate Genes Are More Likely to Duplicate, Molecular Biology and Evolution, Volume 35, Issue 1, January 2018, Pages 107-118, https://doi.org/10.1093/molbev/msx270. Duplicate genes can be duplicates because copy number variation of them has minimal impact.

We thank the reviewer for raising this important point.

We have briefly addressed this in the introduction as follows:

In multiple model organisms, paralogs have been demonstrated to be more dispensable than singletons (genes without a paralog) [3–5]__. There are a number of reasons for why a paralog might be more dispensable than a singleton gene, including preferential retention of duplications of non-essential genes [6,7]__, but perhaps the most obvious explanation is buffering between paralogs.

Where references 6 and 7 are as follows:

1. O’Toole ÁN, Hurst LD, McLysaght A. Faster Evolving Primate Genes Are More Likely to Duplicate. Mol Biol Evol. 2018;35: 107–118.
2. He X, Zhang J. Higher duplicability of less important genes in yeast genomes. Mol Biol Evol. 2006;23: 144–151.

We discuss this more comprehensively in the discussion as follows:

In both yeast and cancer there are a number of reasons for why paralogs might be more dispensable than singleton genes. Perhaps the most obvious is the existence of buffering relationships between paralog pairs, such that when one paralog is lost the other paralog can compensate for this loss. Such buffering relationships between paralogs can be revealed through synthetic lethality screens and a number of recurrently deleted paralogs in cancer have already been reported to display synthetic lethal interactions with their paralog (recently reviewed in [54]__). Supporting this model, in previous work analysing essentiality in cancer cell lines we found that buffering relationships between paralogs could explain 13-17% of cases where a paralog was essential in some cell lines but not others__[13]__. This suggests that at least some of the increased dispensability of paralogs in cancer cells can be attributed to buffering relationships between paralog pairs. However this is not the only explanation for paralogs displaying increased dispensability in tumour cells. An additional explanation is that paralogs may perform essential functions in specific contexts (e.g. within specific tissues or at specific developmental stages) but are not required within the specific context of a tumour. Consistent with this model, human paralogs are more likely to display tissue-specific expression patterns [55]__. Finally we note that there is evidence to suggest that genes whose perturbation has a lower phenotypic impact may more ‘duplicable’ – i.e. rather than paralogs being under weaker selection because they are duplicated, their duplication was tolerated because they were already under weaker selection__[6,7]__. Teasing apart the relative contributions of these factors to the increased dispensability of paralogs in cancer will require further research and potentially new data resources such as gene essentiality profiles in diverse non-cancer cell types [56]__.

CROSS-CONSULTATION COMMENTS<br /> I agree, that's a helpful suggestion from reviewer 1.

Reviewer 3's suggestion regarding age of the two whole genome duplication events is quite difficult to unpick as the duplication events seem to have happened relatively close in time to each other while rediploidisation of the first was occurring. Additionally, paralogs from SSDs tend to be more similar in sequence simply because the two WGD events are relatively old while SSDs can occur at any time up to present. They're therefore biased by young duplicates that have not had the opportunity to diverged much and decrease in sequence similarity.

We appreciate these comments.

Reviewer #2 (Significance):

This is a novel study as it examines the frequency of paralog deletion in cancer samples and the factors influencing it, building upon work already conducted in cancer cell lines. This study extends the knowledge of the field confirming previous trends observed in cell lines, this time in actual cancer samples. It confirms that paralogs are more dispensable than singletons, likely because they have a similar counterpart that can provide some level of functional redundancy. The more similar the closest paralog, the more likely it is to be deleted provides support for this.<br /> It is certainly limited by the number of samples currently available in the two cancer sample projects included but the authors attempt to quantify how limiting this sample size is by conducting a saturation analysis using down-sampling to estimate how many gene deletions one can expect from different numbers of samples. This is important as the lack of observance of many gene deletions is likely due to the limited sample size and not due to negative selection. This low observance of gene deletions disappointingly limits further testing beyond single paralogs to consider the deletion effects of multiple gene family members and more directly test evidence of functional redundancy between paralogs. The authors provide a good discussion of the limitations of their study.

The results are of interest to evolutionary biologists and cancer biologists. Those with an interest in duplicate genes, and/or factors affecting gene loss in tumours will be interested in this work.

My field of expertise is molecular evolution, gene duplication and copy number variation.

We thank the reviewer for the positive assessment of the significance of our work.

Reviewer #3 (Evidence, reproducibility and clarity):

Thank you review "Paralog dispensability shapes homozygous deletion patterns in tumor genomes" by DeKegel et al. This manuscript uses TCGA and ICGC tumor data to show evidence for paralog dispensability. They analyze the rate of homozygous deletions and show that it is higher for paralogs compared to singletons. Their findings are robust to a number of confounding variables that they take into account e.g. distance to tumor suppressor, telomere, centromere or fragile site. They show that paralogs that belong to large families and have higher sequence identity tend to show more dispensability and these dispensable paralogs are less likely to be WGD.

We thank the reviewer for the time taken to review our manuscript.

Major comments.<br /> 1. Does the finding pertaining to lack of enrichment of paralogs in regions LOH take into account whether LOH is copy neutral or not i.e. how does dosage affects this finding? Is it possible that there is a difference in paralog rate in LOH that results in total copy 1 and that the presence of copy neutral LOH masks the effect? Also, Integration of gene expression dataset would be helpful to resolve the difference between dosage paralog that compensate of the lack of their sister by upregulating their gene expression.

In the submitted manuscript we focussed solely on LOH events where the copy number of one allele was 0 and the other allele was ≥1. These include copy loss events (total copy number = 1), copy neutral events (total copy = 2), as well as amplifications (total copy number > 2). The rationale for this approach was that we were interested in understanding whether the mechanism that was generating deletions was preferentially generating deletions in paralog-rich regions.

However, we agree that understanding the influence of dosage is of interest here. We have therefore expanded the analysis in the paper to separately assess the enrichment of paralogs in copy neutral LOH regions (total copy number = 2) and copy loss LOH regions (total copy number = 1).

As shown in the new updated Figure S4B we do not find an enrichment of paralogs in genes subject to either copy neutral LOH or copy loss LOH.

The relevant section of the text on page 6 now reads :

We do not find that paralogs are more frequently subject to LOH than singletons in either the TCGA or ICGC cohort (Fig. S4B-C); when considering all LOH segments we even see that singletons are slightly more frequently subject to LOH in the ICGC cohort (Fig. S4C, left), but when considering only focal LOH segments – i.e. segments whose length is less than half of the chromosome arm’s length, which is the case for all HD segments – there is no significant difference between paralog and singleton LOH frequency in either cohort. To assess whether gene dosage influenced the observed LOH frequency we further restricted our analysis to copy neutral LOH events (total copy number = 2) and copy loss LOH events (total copy number = 1) and again found no significant increase in deletion frequency of paralogs compared to singletons (Fig. S4B-C).

Regarding the integration of gene expression to identify dosage compensation between paralogs – we agree that this is extremely interesting. However, it is quite challenging to address properly. Most paralogs are only observed to be homozygously deleted a single time and so statistically identifying how loss of one gene impacts the mRNA abundance of another is challenging. In the minority of cases where a paralog is recurrently deleted, often these deletions occur in samples from different cancer types and so integrating transcriptomic data still presents some technical challenges. Given this complexity, and as the question of dosage compensation is not central to our key observations, we have not integrated transcriptomic data here.

1. Is the finding that paralogs are depleted among WGD is influenced by the age of WGD since there are 2 WGD events? Do SSD tend to be more or less similar by seq than WGD? This should be explored further since this observation is the opposite of what is observed in model organisms such as yeast whereby SSD are less functionally similar than WGD and often show properties similar to singletons than WGD.

As noted by reviewer 2 in the cross commentary, it is extremely challenging to age the duplicates that arose from the WGD due to the close temporal proximity of the two whole genome duplication events. In the dataset of paralogs analysed used here, SSDs have lower average sequence identity than WGDs. However we note that both sequence identity and duplication type are included in our regression analysis (Figure 4D) and both are significantly associated with homozygous deletion frequently.

This should be explored further since this observation is the opposite of what is observed in model organisms such as yeast whereby SSD are less functionally similar than WGD and often show properties similar to singletons than WGD.

We do not actually think that our results are in opposition to the findings from model organisms. The bulk of studies on the functional consequences of deletions of SSDs/WGDs in model organisms are derived from analyses of the budding yeast gene deletion collection, which is generated in a single strain and grown in lab conditions. Consequently these studies report on which genes can be lost in a single genetic background when grown in rich media. We think it is not fully clear how these findings will apply in the context of a panel of genetically heterogenous tumours derived from multiple different cell types. We note that there are additional complexities when analysing human genes (tissue types, two rounds of WGD, metazoan specific pathways enriched in either WGDs/SSDs) that make a straightforward comparison with yeast challenging. We also note that although the results of analyses of the yeast gene deletion collection suggest that SSDs are more likely to be essential than WGDs, experimental evolution studies have demonstrated that SSDs are more likely to accumulate protein altering mutations than SSDs (Keane et al, Genome Research 2014; Fares et al, PLoS Genetics 2013). This is not what would expect based on the analyses of the yeast gene deletion collection, but is closer to what we observe for tumour genomes where SSDs are more likely to be homozygously deleted.

We agree that we did not adequately discuss these issues in the previous version of our manuscript and so have added a new section to the discussion where we compare our results here with those from budding yeast:

Much of our understanding of the factors that influence gene dispensability comes from studies in model organisms, in particular the budding yeast Saccharomyces cerevisiae [3,9,10,43,44]__. Analyses of the yeast gene deletion collection, a set of gene deletion mutants systematically generated in a single S. cerevisiae strain, revealed that paralogs were less likely to be essential than singleton genes [3,45]__. Furthermore, more detailed analyses of yeast paralogs revealed that paralogs from large families were less likely to be essential as were genes with highly sequence similar paralogs [43,44]__. Previous analyses, including our own, demonstrated that many of these trends are also evident when analyzing gene essentiality from CRISPR screens in cancer cell lines [12,13,15,35]__. Our results here are also consistent with these findings – many of the features that are associated with paralog dispensability in yeast are also associated with gene deletion frequency in tumor genomes.

The connection between the budding yeast observations and those in cancer is less clear when it comes to the relative dispensability of WGDs and SSDs. Analyses of the yeast gene deletion collection revealed that SSDs are more likely to be essential than WGDs in the single genetic background studied [43,44]__. In our previous analyses of gene essentiality in hundreds of cancer cell lines we found that SSDs were more likely to be broadly essential (essential in most cell lines) than WGDs but that WGDs were less likely to be never essential (i.e. more likely to be essential in at least one cell line)__[13]__. As the analyses of gene essentiality in budding yeast were generated in a single genetic background the concordance with our cancer cell line results was difficult to assess, but as gene deletion collections are now being generated in additional yeast strains it should become possible to perform a more direct comparison__[46–48]__.

Here we found that WGDs are less likely to be deleted than SSDs in tumors. This is surprising in light of the yeast gene deletion collection results, where SSDs were more likely to be essential than WGDs in the strain studied, but less so in light of the cancer cell line results, where WGDs were less likely to be never essential. It is also worth noting that experimental evolution studies in yeast found that SSDs accumulate protein-altering mutations at a higher rate than WGDs [49,50]__. These results are perhaps especially relevant when analyzing the influence of paralog features on selection in tumors.

We note that there are many additional differences in the features of WGDs and SSDs in budding yeast that may alter their relative dispensability in tumors. An obvious large scale difference is that in the ancestor of humans there were two rounds of whole genome duplication compared to a single duplication event in yeast__[51,52]__. Less obvious, but potentially of importance for cancer, is that the two classes of paralogs are enriched in pathways in humans that do not have obvious counterparts in yeast. For example, WGDs are highly enriched in signaling pathways involved in development while SSDs are enriched in immune response genes__[53]__. How the membership of these pathways influences the dispensability and selection of genes in tumors and cancer cell lines warrants further study.

Minor comments<br /> 1. There is a missing reference on line 55.

We thank the reviewer for catching this oversight. We have now added a reference to Zerbino et al, NAR 2018 on this line.

CROSS-CONSULTATION COMMENTS<br /> That's a good suggestion by reviewer 1. Homozygous deletion collection is available in yeast so these data can be used directly in addition tot he haploid gene deletion collection data.

Since authors of this manuscript included in their analysis the comparison of WGD and SSD then they should do it more thoroughly. It is not sufficient what they presented here especially given that it contradicts the findings from model organisms.

As noted above we have now added a significant discussion of the yeast findings and also of the SSD/WGD observations

Reviewer #3 (Significance):

This work provides the first systematic assessment of paralog dispensability specifically looking at homozygous deletions of paralogs across primary tumor samples and builds on the existing findings in cancer cell lines. It will be broadly interesting to those studying duplicated gene evolution and genome robustness. My expertise is in complex genetic networks in yeast and human cancer as well as genome evolution.

We thank the reviewer for the positive assessment of our manuscript.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Thank you review "Paralog dispensability shapes homozygous deletion patterns in tumor genomes" by DeKegel et al. This manuscript uses TCGA and ICGC tumor data to show evidence for paralog dispensability. They analyze the rate of homozygous deletions and show that it is higher for paralogs compared to singletons. Their findings are robust to a number of confounding variables that they take into account e.g. distance to tumor suppressor, telomere, centromere or fragile site. They show that paralogs that belong to large families and have higher sequence identity tend to show more dispensability and these dispensable paralogs are less likely to be WGD.

Major comments.

1. Does the finding pertaining to lack of enrichment of paralogs in regions LOH take into account whether LOH is copy neutral or not i.e. how does dosage affects this finding? Is it possible that there is a difference in paralog rate in LOH that results in total copy 1 and that the presence of copy neutral LOH masks the effect? Also, Integration of gene expression dataset would be helpful to resolve the difference between dosage paralog that compensate of the lack of their sister by upregulating their gene expression.
2. Is the finding that paralogs are depleted among WGD is influenced by the age of WGD since there are 2 WGD events? Do SSD tend to be more or less similar by seq than WGD? This should be explored further since this observation is the opposite of what is observed in model organisms such as yeast whereby SSD are less functionally similarthan WGD and often show properties similar to singletons than WGD.

Minor comments

1. There is a missing reference on line 55.

Referees cross-commenting

That's a good suggestion by reviewer 1. Homozygous deletion collection is available in yeast so these data can be used directly in addition tot he haploid gene deletion collection data.

Since authors of this manuscript included in their analysis the comparison of WGD and SSD then they should do it more thoroughly. It is not sufficient what they presented here especially given that it contradicts the findings from model organisms.

Significance

This work provides the first systematic assessment of paralog dispensability specifically looking at homozygous deletions of paralogs across primary tumor samples and builds on the existing findings in cancer cell lines. It will be broadly interesting to those studying duplicated gene evolution and genome robustness. My expertise is in complex genetic networks in yeast and human cancer as well as genome evolution.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

Here, De Kegel & Ryan analyse thousands of tumour samples from the TCGA and ICGC projects to identify homozygously deleted genes, finding that about 40% of protein-coding genes are deleted in at least one sample. They find homozygously deleted genes to be enriched for paralogous genes, and further, more frequently deleted genes are increasingly likely to be paralogs. The authors then test the influence of several factors on the likelihood of being deleted, including gene length, distance to a fragile site or chromosomal region, and distance to a recurrently deleted tumour suppressor gene (TSG). They find that proximity of a TSG, telomere, centromere, and fragile site all increase likelihood of being deleted in a sample, as does gene length. Having a paralog also remains an important predictor of deletion after accounting for these other factors. Additionally, the more similar in sequence the closest paralog is to the gene and having a larger gene family size are also predictive of deletion. Conversely, if a gene is a whole genome duplicate as opposed to a small-scale duplicate, it is less likely to be deleted. Finally, the authors test the hypothesis that paralogs that are deleted in cancer are less likely to be essential and find that this is indeed the case.

Comments

The authors have done a good job of identifying trends of paralog deletion in cancer samples and the factors influencing them. The results are well described and presented and support the conclusions. I like the inclusion of the saturation analysis as an estimate of what to expect given current and potential future sample sizes, and I appreciate the inclusion of a WGD/SSD paralog distinction. The data and methods are sufficiently detailed. I have a few minor comments below.

1. Around line 160 in the text and supplemental figure 4A, the authors test if the trends they see are observed across individual cancer types. With 9 of 33 cancer types reaching a sample size threshold, 8 of 9 comparisons are significant. The authors do not state correcting for multiple testing.
2. I initially misunderstood the hemizygously deletion analysis, thinking the analysis in supplement figure 4B/C was asking if a sample has any singleton or any paralog deleted and comparing the number of samples with any deletion of either - given the number of genes deleted per sample this wouldn't make sense as a good test. I think the authors are actually comparing the number of loss-of-hemizygosity events per gene and grouping by paralog/singleton. I think this is a good analysis, but I think it would be helpful to clarify this in the text and figure legend e.g. "Samples w/ gene LOH" could be "LOH events per gene" or something similar.
3. Occasionally, I wanted some more detail in the text for context, which was sometimes later provided - e.g. I noted when reading about line 125 that I was curious at this point how often TSGs occurred on segments, and this was later provided on line 241. Similarly, around line 114 I was curious how many genes are typically deleted per HD segment, for which the median value was provided on line 206 (and distribution in supplemental figure 1), and again for hemizygous deletions. I think sometimes it would be helpful to provide this context earlier in the text to aid interpretation of the results.
4. In the discussion, on line 420, the authors include the point that a paralog might not be required at all in a tumour cell and therefore easily deleted. I think this possibility could be expanded on here and in the introduction/results section, as it is an important point. I think it would be helpful to include more about the possibility that a paralog might be deleted in a tumour cell because it is simply not required or that is more likely to have less of a phenotypic impact compared to a singleton, and that this could be a reason for the observed enrichment of paralogs in deleted genes. A citation to support this point could be Áine N O'Toole, Laurence D Hurst, Aoife McLysaght, Faster Evolving Primate Genes Are More Likely to Duplicate, Molecular Biology and Evolution, Volume 35, Issue 1, January 2018, Pages 107-118, https://doi.org/10.1093/molbev/msx270. Duplicate genes can be duplicates because copy number variation of them has minimal impact.

Referees cross-commenting

I agree, that's a helpful suggestion from reviewer 1.

Reviewer 3's suggestion regarding age of the two whole genome duplication events is quite difficult to unpick as the duplication events seem to have happened relatively close in time to each other while rediploidisation of the first was occurring. Additionally, paralogs from SSDs tend to be more similar in sequence simply because the two WGD events are relatively old while SSDs can occur at any time up to present. They're therefore biased by young duplicates that have not had the opportunity to diverged much and decrease in sequence similarity.

Significance

This is a novel study as it examines the frequency of paralog deletion in cancer samples and the factors influencing it, building upon work already conducted in cancer cell lines. This study extends the knowledge of the field confirming previous trends observed in cell lines, this time in actual cancer samples. It confirms that paralogs are more dispensable than singletons, likely because they have a similar counterpart that can provide some level of functional redundancy. The more similar the closest paralog, the more likely it is to be deleted provides support for this.

It is certainly limited by the number of samples currently available in the two cancer sample projects included but the authors attempt to quantify how limiting this sample size is by conducting a saturation analysis using down-sampling to estimate how many gene deletions one can expect from different numbers of samples. This is important as the lack of observance of many gene deletions is likely due to the limited sample size and not due to negative selection. This low observance of gene deletions disappointingly limits further testing beyond single paralogs to consider the deletion effects of multiple gene family members and more directly test evidence of functional redundancy between paralogs. The authors provide a good discussion of the limitations of their study.

The results are of interest to evolutionary biologists and cancer biologists. Those with an interest in duplicate genes, and/or factors affecting gene loss in tumours will be interested in this work.

My field of expertise is molecular evolution, gene duplication and copy number variation.

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 study, the authors delineate the association of paralog dispensability with the frequency of homozygous deletions (HDs) and thereby show that paralog dispensability can play a significant role in shaping tumor genomes. The authors analyzed the strength of negative selection on the paralogs relative to the singletons using frequencies of the homozygous deletions (HD). The study focused on HDs because they ensure a complete loss of function, unlike other mutational aberrations that can be masked because of haplo-sufficiency. While accounting for potential confounding factors, authors find that paralogs tend to have a relatively high frequency of HDs, suggesting a relaxed negative selection. Furthermore, the authors specifically attribute this association to the dispensable paralogs by analyzing gene inactivation data generated from multiple experimental systems. Overall, the findings of this study can potentially have significant implications in cancer biology field and specifically to the researchers studying cancer genome evolution.

Major comments:

1. To dissect further which dispensable paralogs are more likely to be associated with a high HD frequency, synthetic lethal paralogs could be compared with non-synthetic lethal ones.<br /> In the section titled 'Homozygous deletion frequency of paralog passengers is influenced by paralog properties' (begins from line #289), authors have shown that paralogs with a high frequency of HDs are more likely to have the properties of dispensability (in Figure 4). It seems that all of those properties are also associated with synthetic lethality as the authors identified in their previous study (DeKegel et al. 2021). Furthermore, as shown in the subsequent section ('Essential paralogs are less frequently homozygously deleted than non-essential paralogs', begins from line #344), the high HD is associated with the dispensable paralogs. Some of those dispensable paralogs are expected to be synthetic lethal. Therefore, the association of paralogs with a high frequency of HDs with experimentally validated or predicted sets of synthetic lethal paralogs could be tested. This may help authors to contextualize their findings in terms of genetic interactions between paralogs.

Minor comments:

1. The number of TCGA and ICGC tumor samples analyzed:<br /> As mentioned in the Results section (line #106), 9966 tumor samples were analyzed. However, the sample size mentioned in Figure 2A is 9951. Is the lower number shown in the figure due to the filtering procedure mentioned in the Methods section (line #455)? The change in sample sizes could be explained. A similar difference in sample sizes exists for the ICGC data also.
2. The rationale behind setting the threshold at 100 HD genes to classify 'hyper-deleted' samples for TCGA (line #462) and ICGC data (line #473) could be explained.
3. Citation for DepMap is missing (caption of Figure 5).

Referees cross-commenting

Along the lines of Reviewer #3's second major comment, I have a suggestion, the possible benefits of which would depend on the target audience to which the authors intend to communicate their study.

I would suggest including a brief comparison of the findings of this study which deal with human paralogs, with the findings in model organisms such as yeast, perhaps in the discussion section. To facilitate such a comparison, authors could try measuring the enrichments of, for example, molecular functions, gene families, types of genetic interactions, etc., among the paralogs associated with a high frequency of HDs and then discussing their comparison with what is known in the literature for paralogs in other model organisms that tend to be frequently deleted.

Such a comparison could be of interest to the community of researchers working on other model organisms and put this study in a much broader context. However, as I said before, this would depend on the authors' intended target audience.

Significance

As the authors note in their manuscript, it is expected that paralog dispensability could be associated with the relaxed negative selection in tumor genomes because (1) paralogs are prevalent in the human genome, and (2) many of them are dispensable, as apparent from the large-scale gene inactivation screens in hundreds of cancer cell lines (Blomen et al. 2015, Wang et al. 2015, Dandage and Landry 2019, De Kegel and Ryan 2019). However, direct mapping of this association, while importantly accounting for potential confounding factors, has been lacking.<br /> As a researcher with prior experience in the research topics such as gene duplication and genetic interactions, it appears to me that this study presents formal proof of the important association between paralog dispensability and tumor genome evolution which could be of major implication for the research community of cancer biology field and specifically to the researchers dealing with the topics such as cancer evolution, copy number alterations in cancer genomes, and synthetic lethality-based precision oncology therapeutics.

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 would like to thank all reviewers for taking the time to evaluate our manuscript. Many helpful suggestions and discussion points were raised. These comments were instrumental to provide more data that strengthen our conclusion about the relevance of centrin condensation in vivo, expand our findings to other organisms, and improve the manuscript in general. Details are given in the following individual replies.

Reviewer #1 (Evidence, reproducibility and clarity):

Voss and colleagues show calcium-dependent assembly of Plasmodium falciparum centrins in vitro and in parasites. This assembly is dependent on the EF-hands of centrin and an N-terminal disordered region.

Major concerns:

1. The very definitive title is not wholly supported by the data. This should be qualified by specifying the conditions under which the centrins can accumulate in this way.

We understand this comment by the reviewer. There are multiple dimensions to the potential of centrins to condensate, such as the specific centrin family member, in vivo vs in vitro situation, and media conditions. Naturally it is difficult to represent these various conditions in a concise and compelling title but in line with the suggestion by Reviewer 2 we are changing the title to “Malaria parasite centrins can assemble by Ca2+-inducible condensation” to reflect the conditionality of this process.

1. A major concern is whether this behaviour of centrins represents a biologically relevant mechanism in centriolar plaque formation. Is this limited to high overexpression conditions or in vitro high concentrations? Or is it a result of the tagging of the P. falciparum centrins?...

Centrin accumulation at the centriolar plaque and assembly of the centriolar plaque itself must be differentiated. Although compelling we are already very careful in the text about extrapolating our findings about centrin accumulation in cells to centriolar plaque or centrosomal assembly in general. We, however, thank the reviewer for this important comment and now have carried out hexanediol treatment of wild type parasites to test the effect on centrin in a native context. After IFA staining we failed to detect any centrin foci at the centriolar plaques, suggesting that they can be resolved by inhibiting weak hydrophobic interactions that are typical for phase separation (now Fig. 6, lines 283ff).

Concerning the effect of tagging we have generated new data of cells overexpressing an untagged version of PfCen1 in parasites, which still shows formation of ECCAs as revealed by IFA (now Fig. 4H-K, lines 243ff). This significantly alleviates the concern that the observed phenomenon is only a consequence of GFP-tagging. Our in vitro data already showed that native and tagged PfCentrin1 & 3 can undergo condensation.

Concerning the critical concentration of our in vitro assay we find it to be around 10-15 µM without the addition of crowding agents such as PEG (now Fig. S3C, lines 120ff). To our understanding it is challenging to select an in vitro concentration that is adequate to define a threshold for “biological relevance” due to so many additional factors playing a role in vivo. Those factors can also favor a phase separation locally when total saturation concentration is not reached as we now discuss in more detail (lines 440ff). For reference the critical concentration of FUS, which is one of the most studied phase separating proteins in model system, is around 2 µM, but concentrations below 15 µM are well within the range of what is observed for in vitro LLPS. Additionally, it is important to consider that we find Cen1/3 and HsCen2 LLPS is inducible and reversible and that very homologous proteins i.e. Cen2 and 4 serve as an adequate internal control.

… A convincing approach to addressing this issue would be to knock-in a fluorescent tag to the centrin loci. Roques et al. (ref. 12 in this submission) report the GFP tagging of centrin-4 in P. berghei, although they note that centrins-1 to -3 were refractory to tagging in this organism. It is unclear whether Voss et al. attempted this tagging in P. falciparum. This should be clarified and relevant data presented.

We indeed attempted several unsuccessful iterations of tagging Cen1/3 with HA and GFP tag and now explain this in the text more clearly (lines 81ff). We did not attempt tagging Cen2 and 4 as they do not display phase separation in vitro or carry IDRs.

If the tagged molecules used in the biochemical parts of this study are functional, it is challenging to understand why the centrins cannot be tagged in P. falciparum. If the tags render the P. falciparum centrins dysfunctional, the study becomes significantly less useful.

Our data shows that in vitro Cen1-GFP can undergo Ca2+-inducible and reversible LLPS and that GFP-tagged centrins can still localize to the centriolar plaque. Centrin function, however, certainly goes beyond its ability to condensate and localize. It is easily conceivable that interaction with critical binding partners at the centriolar plaque is inhibited by tagging a protein as small as centrin, which prohibits tagging the endogenous version, while its ability to phase separate remains unaltered. To dynamically study a protein in cells tagging is, however, unavoidable. Even though tagging affects any proteins function to highly variable degree we are still convinced that studying those proteins still provides useful information. Our mutant versions of PfCen1 in vivo shows that non-condensating version display different localization. Importantly, as mentioned above, we now provide images of cells overexpressing an untagged Cen1 version, which still causes ECCA formation (Fig. 5H-K). Ultimately, even though tagged versions might not be fully functional, our observations are compatible with the ability of centrins to condensate in vivo.

1. If a knock-in cannot be achieved, it must be shown that the transgenic expression of tagged Plasmodium centrins does not confound the analysis of centrin behaviour. It is known that these proteins can behave anomalously when overexpressed (Yang et al. 2010, PMID: 20980622; Prosser et al. 2009, PMID: 19139275), at least in other species.

Thank you for this comment. Transgenic expression of proteins can in principle influence their behavior. In the context of this study the overexpression is, however, used intentionally since protein concentration correlates with the phase separation. Here, transgenic overexpression is used as a tool, rather than being a confounding factor, and ECCA formation can be used as quantifiable phenotype. The observation that ECCAs appear significantly earlier the higher they are expressed is in our opinion one of the stronger points of evidence that this result from phase separation in vivo. Yet centrins maintain their centriolar plaque localization and no significant impact on growth is observed. To definitely answer whether phase separation of endogenous centrin is occurring during centriolar plaque accumulation is challenging. These challenges and limitations are now addressed in the significantly extended discussion. As explained above untagged Cen1 also forms ECCAs.

A previous description of centriolar plaque from the authors' lab (Simon et al. 2021, PMID: 34535568) shows an organized structure of an established size. It should be demonstrated whether the structures formed with the GFP tagged centrins show the same dimensions and dynamics as those in wild-type parasites. The extent of the overexpression of the GFP-tagged centrins should also be demonstrated.

We thank the reviewer for this suggestion. We have now added spatial measurements of the centrin signal dimensions at the centriolar plaque of mitotic spindle containing nuclei in PfCen1-GFP overexpressing vs non-induced cell lines. We found that the width of the centrin-signal at the centriolar plaque was unaltered while the height only increased by 11% (Fig. S9). Further, we found no significant growth phenotype in overexpressing parasites, which indicates that the centriolar plaque is functional.

Due to several confounding factors, we were, unfortunately, unable to clearly quantify the extent of overexpression. Most notably the induction of overexpression only works in about 50% of the cells (Fig. S6). The mean intensity after induction further displays quite some variability. Furthermore, the expression kinetics along the IDC of endogenous centrin and our overexpression system that we use as a tool differ. Lastly, our centrin antibodies display crossreactivity (see also Fig. S12) making it impossible to identify how much of the endogenous pool we are labeling in comparison to the GFP- tagged Cen1 protein.

1. It would also be useful to remove the His tag from the recombinantly expressed and purified centrins for the in vitro analyses, particularly if concern remains about the impact of tags on Plasmodium centrin behaviour.

Based on the published in vitro studies on other centrins, we did not anticipate the His-tag to change LLPS properties. Also, Cen1 and 3 and Cen2 and 4 would need to be differentially affected by the tag. We further have experimented with N-terminally tagged 6His-Cen3 protein and found no significant differences in our turbidity assays. Nevertheless, we expressed new versions of the recombinant PfCen1-4 proteins with a TEV cleavage site inserted after the His-tag to purify untagged proteins and found no fundamental differences in our LLPS assay aside some slight variation in the kinetics (Fig. S3E).

1. The discussion is very short and does not consider the findings presented here in the context of the literature, with respect to centrins, Plasmodium MTOC assembly mechanisms, or to general considerations around biological condensates. Andrea Musacchio's recent commentary (ref. 44 in the current submission) advocates caution in ascribing phase separation as an assembly mechanism for organelles in vivo, particularly on the basis of in vitro experiments with high concentrations of homogeneous protein. It is not clear that the concentration dependence of extracentrosomal centrin accumulations (ECCAs) at the onset of schizogony provides sufficient justification of a phase separation model in vivo. The authors' recent description of the involvement of an SFI1-like protein, SIp (Wenz et al. 2023 PMID: 37130129), in the centriolar plaque makes a case for non-homotypic interactions also driving assembly and alternative models for ECCA are not convincingly excluded. The absence of a robust discussion of such considerations is unhelpful to the reader.

We very much thank the reviewer for this suggestion, which helped to significantly improve the manuscript. We have purposefully included the commentary by Andrea Musacchio to highlight a different (possibly the most antipodal) point of view on the role of biomolecular condensation in membraneless organelle formation for the unfamiliar readers that might be just getting to know the field of phase separation. In the absence of word limitations, the reviewer is right to point out the lack of more extensive discussion. We now have significantly extended this section and address the suggested points including the potential role of the novel centriolar plaque protein Slp, which was not published upon submission of our previous version (lines 450ff.)

1. It is also unclear whether the analysis of human centrin is suggested to indicate a phase separation mechanism for centrins in human cells. As this is readily testable, this notion could be considered further. Although its experimental examination may lie outside the theme of this study, one would expect some discussion of the significance of the data presented in the study.

Since it is the first description of phase separation of centrin, it would indeed be interesting to explore the functional relevance in other organisms such as humans. We are considering approaching this in the future. We have, as requested above, significantly extended the discussion and now also include this aspect. Earlier reports have e.g. shown centriole overduplication in human cells upon centrin overexpression.

Minor points

1. There are only three centrins in humans. Centrin 4 is a pseudogene (Gene ID: 729338 on NCBI).

Thank you for detecting this error, which we now corrected (line 60). Centrin 4 seems only to be an expressed gene in mice.

1. Line 175 should say 'temporally', rather than 'temporarily. The Abstract should say 'evolutionarily conserved', rather than 'evolutionary conserved'. 'To condensate' is not ideal as a phrase- 'to form a condensate' would be clearer.

Thank you for those suggestions. The text has been modified accordingly.

Referees cross-commenting

I think the other 2 reviewers have made fair, cogent and constructive points. There is good convergence between the reviewers on the significant issues around the study. These concern in vivo and in vitro effects of tagging and of high concentrations.

Reviewer #1 (Significance):

The biology of the Plasmodium centriolar plaque is of great interest as an alternative MTOC structure, with obvious additional interest deriving from the role of this organism in malaria. Much remains to be learned about this structure, so the topic of this paper is likely to attract a broad readership. Furthermore, the centrins are a widely-expressed and evolutionarily conserved family of eukaryotic proteins, with multiple roles; a new model for their behaviour, such as is suggested here, would be of interest to many cell biologists.

With that in mind, significant additional data should be provided to substantiate the model proposed by the authors.

We appreciate that the reviewer considers our manuscript of interest for a broad audience. We feel that our modifications of the text including a more thorough contextualization and addition of some new experimental data now sufficiently supports our claims.

Reviewer #2 (Evidence, reproducibility and clarity):

The authors analyzed the properties of the four Centrin proteins of the malaria parasite using a combination of in vitro and in vivo approaches. Their findings indicate that two of the four Plasmodium Centrin proteins, PfCen1 and PfCen3, as well as the human Centrin protein HsCen2, exhibit features of biomolecular condensates. Moreover, analysis of cells overexpressing PfCen1 indicates that such biomolecular condensates become more numerous as cells approach mitosis and are dissolved thereafter.

Major comments

A) A critical point that requires clarification is how the protein concentrations used in the in vitro and in vivo assays (20-200 microM in vitro, and not estimated in vivo) compare to that of the endogenous components. This is important because it may well be that 6His-tagged PfCen1, PfCen3 and HsCen2 can form biomolecular condensates when present in vast excess, but not when present in physiological concentrations. The authors should report the estimated cellular concentration of PfCen1-4, as well as that achieved upon PfCen1-GFP overexpression (on top of endogenous PfCen1), for instance using semi-quantitative immunoblotting analysis. Given this limitation, the authors may also want to temper their title by introducing the word "can" after "centrins".

In the context of phase separation, protein concentration is of course a critical metric. However, in vitro and in vivo concentrations cannot be directly compared as the composition of the surrounding solute has a significant impact on the effective saturation concentration. In vitro we find a saturation concentration for Cen1 of 10-15 µM (Fig. S3C), which is within a range that is frequently found other in vitro studies as listed in the in vitro LLPS data base (PMID: 35025997). We now more explicitly discuss this in the text (lines 422ff). At this point, unfortunately, we have no means of investigating the absolute concentrations of centrin in vivo and to our knowledge no such data is available for apicomplexan. Additionally, one has to keep in mind the presence of other centrin family members in the cell which can interact and co-condensate as well as other centriolar plaque proteins, like PfSlp, but are difficult to separate through analysis. Further we now discuss several contexts that modify the saturation concentration in vivo (lines 440ff).

As explained above in a response to Reviewer 1, we were not able to produce a satisfactory quantification of the overexpression levels. We are repasting the previous response here:

“Due to several confounding factors we were, unfortunately, unable to clearly quantify the extent of overexpression. Most notably the induction of overexpression only works in about 50% of the cells (Fig. S6). The mean intensity after induction further displays quite some variability. Lastly the expression kinetics along the IDC of endogenous centrin and our overexpression system that we use as a tool differ. Lastly, our centrin antibodies display crossreactivity (see also Fig. S12) making it impossible to identify how much of the endogenous pool we are labeling in comparison to the GFP- tagged Cen1 protein. “

Concerning the title, as explained above, we followed the suggestion and added the word “can”.

B) Movies S1 and S2 (and the related Fig. 1D and 1E) are not the most convincing to support the notion that the observed assemblies are biomolecular condensates, as not much activity is going on during the recordings. Likewise, Movies S3, and even more so Movie S4, as out of focus for a large fraction of the time, making it difficult to assess what happens at the beginning of the process. Moreover, it appears that fusion events, while occurring, are rather rare. The movies should be exchanged for ones that are in focus, and ideally a rough quantification of fusion events as a function of biomolecular condensate size provided.

We thank the reviewer for requesting clarification. Movies S1 and S2 are by no means direct evidence for biomolecular condensation and we do not claim them to be but rather say that they are “…reminiscent of biomolecular condensates…”. We think that this is an appropriate entry into the subsequent analyses. For Movie S1 it is noteworthy that the shape of the accumulation, which can only be resolved by super-resolution microscopy in live cells, is round as would be expected for a liquid condensate in the absence of forces and on these short time scales. Nevertheless, the centriolar plaque must be duplicated which might be the process partly depicted in Movie S2. The observation that centrin can be still change its shape at least suggests that it is not a solid aggregate. In the context of centriolar plaque biology and the technological advance of applying live cell STED in P. falciparum, we think these data are still worth reporting.

Concerning Movies S3 and S4 we have carefully selected the focal plane to highlight all the hallmarks of LLPS. Since the protein droplets freely move in 3D throughout the entire imaged liquid volume there is no z-plane that is in focus. Our positioning of the focal plane presents the best compromise between showing round droplet shape, droplet fusion events, and surface wetting. All those observations demonstrate the liquid nature of the condensates. Fusion events are indeed relatively rare, and we do not go beyond this qualitative statement that it can be seen.

C) An important control is missing from Fig. 2, namely assaying PfCen1-4 without the 6His tag, to ensure that the tag does not contribute to the observed behavior (although it can of course not be sufficient as evidenced by the lack of biomolecular condensates for PfCen2 and PfCen4).

Thank you for this suggestion. Since reviewer 1 made a similar comment, I’m reiterating our previous reply here: Generally speaking, and based on the published in vitro studies on other centrins, we didn’t anticipate the very small His-tag to change LLPS properties. Also, Cen1 and 3 and Cen2 and 4 would need to be differentially affected by the tag. We further have experimented with N-terminally tagged 6xHis-Cen3 protein and found no significant differences in our turbidity assays. However, we expressed new versions of the recombinant PfCen1-4 proteins with a TEV cleavage site inserted after the His-tag to purify untagged proteins and found no significant differences in our LLPS assay (Fig. S3E).

D) The authors should test whether the assemblies formed by PfCen1 and PfCen3 are sensitive to 1,6-hexanediol treatment, as expected for biomolecular condensates.

This is an interesting and helpful suggestion. We now tested 1,6-hexanediol addition to recombinant PfCen1 and wildtype parasites (now Fig. 6). Interestingly the dissolving effect of hexanediol on PfCen1 in vitro was moderate, which we attribute to the polar component in centrin assembly, which has been documented earlier (Tourbez et al. 2004). In vivo, however, only 5 min of treatment caused a striking dissolution of most centrin foci in wild type parasites, which is compatible with the interpretation that centrin or centriolar plaque assembly could be driven by biomolecular condensation.

E) The fact that HsCen2 also forms biomolecular condensates is very intriguing, but further investigation would be needed to assess the generality of these findings. For instance, the authors could test in vitro also S. cerevisiae Cdc31, the founding member of the Centrin family of proteins to further enhance the impact of their study.

We thank the reviewer for this suggestion. It would of course be exciting to investigate in more detail how widely this biochemical property of some centrins is conserved. To take a first step in that direction, we have recombinantly expressed centrins containing some N-terminal IDRs from C. reinhardtii, T. brucei and S. cerevisiae to represent organism of significant evolutionary distance. Using our in vitro phase separation assays, we found a very similar behavior to PfCen1 for two centrins while yeast Cdc31, although forming droplets, had a much higher saturation concentration, which could be explained by the significantly lower intrinsic disorder in its sequence (now new Fig. 3).

Minor comments

1) For the experiments reported in Fig. 3D, the same concentrations as those used in Fig. 3A-C (namely 10 microM, and not 30 microM as in Fig. 3D) should be used. Moreover, it would be informative to test whether PfCen2 and PfCen4 as PfCen3 when added to PfCen1.

Unfortunately, this experiment is not feasible since Cen3 does not produce droplets at 10 µM. Hence, in Fig. 3D we aimed to test if Cen1 is incorporated into preformed droplets i.e. whether there is still some interaction between them. We have, however, tested the addition of Cen2 to Cen1 and Cen3 and as expected from the inability PfCen2 to condensate we did not find the same synergistic effect as for Cen1 and 3 together (now Fig. S6). The combination of Cen1/2/3 still enabled co-condensation while addition of Cen4 did not further improve droplet formation. Taken together this strongly suggests that only Cen1 and 3 contribute to the phase separation in vitro (lines 184ff).

2) The authors mention that the effect of Calcium in inducing biomolecular condensates is specific, as Magnesium was not effective (lines 94-95). However, an examination of Fig. S3B indicates that the Magnesium also exhibits some activity, albeit less potent than Calcium. The authors should discuss this point and rectify the wording in the main text.

Thank you for pointing this out. While PfCen1 is not reactive to Magnesium, PfCen3 and HsCen2 do display a small reaction, which we now more clearly mention in the text (lines 118ff). Of note Mg2+ and other divalent cation are known to generally promote phase separation.

3) Do the authors think that PfCen2 and PfCent4 localize to the centriolar plaque in vivo using another mechanism that deployed by PfCen1 and PfCent3? It would be good to discuss this point.

This is indeed a point worth discussing. Centrins can of course still interact in the absence of biomolecular condensation and their localization to the centriolar plaque is not dependent on their ability to phase-separate as seen for PfCen2 and 4. We have recently described a novel centriolar plaque protein PfSlp that interacts with centrins and might assist recruitment (Wenz et al. 2023). Cellular condensates are, however, often separated into scaffold proteins, which actually phase separate and client protein which get recruited into those condensates. It is easily conceivable that Cen1 and 3 participate in formation of the biomolecular condensate into which Cen2 and 4 as well as other centriolar plaque proteins might be recruited. Unfortunately, we were not yet able to establish a recruitment hierarchy by e.g. dual-labeling of centrins to test whether PfCen1 and 3 might appear prior to PfCen2 and 4. We now include those aspects in the extended discussion.

4) Given that the EFh-dead mutant exhibits no activity in vitro and fails to localize in vivo, one potential concern is that the protein is misfolded. The authors should conduct a CD spectrum to investigate this.

Thank you for suggesting this relevant control experiment. We have carried out CD spectroscopy of wild type and EFh-dead PfCen1 and find no difference in secondary structure distribution. We now added these data to the supplemental information (now Fig. S14).

5) It is not entirely clear from the main text in lines 103-104, as well as from the legend, what Fig. S3B shows. When was EDTA added in this case?

Thank you for requesting clarification. We will assume the reviewer is referring to Fig S4B. We wanted to show that contrary to PfCen3 that PfCen1 droplets can still be resolved after an elongated period of incubation with calcium but forgot to mark the timepoint of EDTA addition at 180 min in the graph. We have now corrected this and further reworded the sentence for more clarity (lines 132ff).

6) Fig. S7: the correlation between PfCen1-GFP expression levels and ECCA appearance is modest at best. What statistical test was applied? This should be spelled out. Moreover, the authors should combine the two data sets, as this will provide further statistical power to assess whether a correlation is truly present.

Indeed, the correlation is modest but statistically significant, which is why we decided to place this data in the supplemental information. The used statistical test was an F-test provided by Prism, which compares two competing regression models, which we now mention in the legend. Combining the two data sets is unfortunately not possible since they arise from two independent sets of measurements where different imaging settings had to be used to adjust for the very different fluorescent protein levels in both lines after induction.

7) The authors may want to discuss how their findings can be reconciled with the notion that Centrin assemble into a helical polymer on the inside of the centriole (doi: 10.1126/sciadv.aaz4137).

This is an interesting point. Although centrin does localize to the inside of the centriole (https://doi.org/10.15252/embj.2022112107), more precisely one pool at the distal part and one pool at the core, there is no evidence that it is itself part of the helical inner scaffold described by the authors even though it might localize in close proximity to it. Further, there are several examples where polymers such as microtubules act as seeding point for biomolecular condensates or the other way around, and our work suggest this could be a potential working model for centrins. We have discussed our results extensively with the two corresponding authors of the aforementioned study (i.e. Virginie Hamel and Paul Guichard) and agreed that our data are not conflicting. Nevertheless, we include the inner centriole localization and potential association with polymer structures of centrin in our extended discussion.

9) Likewise, the authors may want to speculate regarding what their findings signify for the role of Centrin proteins in detection of nucleotide excision repair (doi: 10.1083/jcb.201012093).

We appreciate the comment by the reviewer. Centrins seem to have many different potential roles that remain to be clarified. While we are excited about this, we think it is too early to speculate about the impact of centrin condensation on less well studied aspects of centrins such as nucleotide excision repair. We, however, now cite this study in the discussion to highlight the functional diversity of centrins.

Small things

• Fig. 1A: change color for microtubules as red on red is difficult to discern.

Throughout our publications we use this shade of magenta to label microtubules in schematics and have therefore opted to use a slightly brighter shade of red for the RBCs instead to improve visibility.

• Fig. 1C: the indicated boxes in the top row do not seem to correspond exactly to the insets shown in the bottom row.

We have verified the position of the boxes and found them to be accurate. Possibly the different imaging modality used for both panels (confocal vs STED) creates this impression.

• line 266: typo, promotor > promoter.

Has been corrected.

• line 360: a reference should be provided for the GFP-booster, including the concentration at which it was used.

Has been added.

• line 363: "an" missing before "HC".

Has been corrected.

• line 428: it would be best to deposit the macros on Github or an analogous repository.

Macros have been deposited on https://github.com/SeverinaKlaus/ImageJ-Macros (line 737)

• line 461: "to the" is duplicated.

Has been corrected.

• Fig. S5A: maybe draw the lines in red (as red in Fig. S5B correspond to the proteins that do not have IDRs).

Since we cannot easily change the line colors of the IDR graphs, we have inverted the font color for Fig. S5B instead.

• Movie S7, legend: left frames shows PfCen1-GFP, not microtubules as currently stated.

Has been corrected.

Reviewer #2 (Significance):

This is a provocative study that extends initial observations regarding self-assembly properties of Centrin proteins, and posits that some members of this evolutionarily conserved family can form biomolecular condensates. After the above outstanding issues have been properly addressed, these data could have important implications for understanding Centrin function in centriole biology and DNA repair. Therefore, these findings will be of interest to a cell biology audience.

Field of expertise: cell biology.

Reviewer #3 (Evidence, reproducibility and clarity):

Summary:

The authors have provided a comprehensive characterisation of centrin proteins in Plasmodium falciparum. Through expression of episomal GFP-tagged centrin for in vitro, they were able to observe co-localisation of centrin with centriolar plaques during the replicative stage of the parasite. They also utilised live cell STED microscopy to track dynamic changes in centrin morphology. They have also demonstrated calcium-dependent phase separation dynamics in bacterially-expressed P. falciparum centrin and human centrin 2. The formation of liquid-liquid phase separation in PfCen1, 3 and HsCen2 tied well with IUPred3 predictions of intrinsically disordered regions in these proteins. Using an inducible DiCre overexpression system with two promoters of varying strengths, the authors have shown accumulation of centrin1 outside of centrosomes and premature appearance of centriolar plaques. Finally, changes on the centrin1 protein, i.e., N-terminal deletion, and mutations in calcium binding sites in the EFh domains, have shown a reduction in the formation of ECCAs during overexpression and inability to form LLPS in vitro, respectively.

Major comments:

1. Given that parasites cannot tolerate endogenous C-terminal tagging of some centrins (but not all, as PbCen4 was successfully tagged), has N-terminal tagging been attempted either by the authors or in previous publications? Note that this is not a request for further experimentation; rather, maybe this can be noted in the manuscript; and line 62 can be rephrased for transparency.

We have not attempted N-terminal tagging ourselves but through personal communication with Rita Tewari we were informed that neither N- nor C-terminal tagging for PbCen1-3 was successful in the context of the study published by Roques et al 2018. We have only unsuccessfully attempted C-terminal tagging in several iterations. Due to importance of N-terminus for interaction and function in other organisms it is plausible that N-terminal tagging is even more unlikely to work. Since we have not exhaustively attempted every tagging strategy on every centrin we, as suggested, rephrased the text accordingly (lines 81ff).

1. Is there a possibility that by adding a C-terminal tag, centrin may lose a specific function or cause change in the physicochemical properties of the protein (thus making C-terminal tagging lethal)? Was His tag removal attempted so the native protein can be used in the LLPS experiments? IUPred3 analysis showed potential IDR at the C-terminal end of PfCen4. Could the C-terminal tag have caused the protein to not form droplets in the presence of Ca2+?

As we could show for PfCen1-GFP, the tag did not impair its ability to undergo LLPS which is at least partly mediated by the N-terminus, and that it could still properly localizes to the centriolar plaque. The fact that some endogenous centrins cannot be tagged suggest that there is a functional relevance to the C-terminus that could e.g. be an interaction with other essential centriolar plaque components. As suggested in a reply to Reviewer 1, we consider a substantial and centrin-specific effect of the small His-tag on phase separation unlikely. To be sure, we have repeated our turbidity assays with tag-free versions of PfCen1-4 and found no change in phase separation properties (now Fig. S3E).

1. It has been shown by the authors that different tagged centrins co-condense which may support the localisation data (Figure 1C). However, is there a way to show that the episomally- and endogenously-expressed centrin co-localise with each other (e.g., confocal microscopy with anti-centrin vs anti-gfp in PfCen-GFP lines, that is if the authors have access to anti-centrin antibodies)? Has endogenous centrin been demonstrated to form ECCAs (in previous publications or by the authors)?

These are important questions by the reviewer. Due to the high sequence homology centrin antibodies, even if raised against a specific centrin (such as PfCen3 in this study), will likely cross-react with other centrins. So far, we have not been able to produce a staining were the anti-GFP-positive foci are devoid of anti-centrin3 staining, which limits the interpretation of these data. The outer centriolar plaque compartment containing centrin is, however, well defined by now and the localization pattern of endogenous centrin and Centrin1 and 4-GFP seems identical. In a more recent study from our lab Cen1-GFP IP has identified other endogenous centrins as interaction partners (Wenz et al 2023), like the Roques et al. 2018 study did for PbCen4-GFP indicating that the tag does not abolish interaction between centrins. So far, we have never detected any ECCAs, nor have we identified any similar structure in the literature. This suggest that this is indeed a consequence of excessive centrin concentration. Importantly we now have added data from a new parasite line overexpressing untagged PfCen1 using the T2A skip peptide (pFIO+_GFP-T2A-Cen1) which displays ECCAs upon induction, showing that this effect is not a mere consequence of tagging (now Fig. 5H-K).

Minor comments:

1. How were the times (post addition of Ca2+) presented in Figure 2A determined?

We noted down the time of calcium addition and cross-referenced it with the timestamps available in the metadata of the movie files (e.g. file creation timepoint marks the start of the movie). We now mention this in the legend.

1. Line 126: Figure 1B should be Figure 1C

2. Line 145: Figure 1C-D should be Figure 1D-E

3. Line 151: Figure 3A should be Figure 4A

Thank you for spotting these mistakes, which now have been corrected.

1. Line 152: Suggest rephrasing "placing the gene of interest in front of the promoter" to "placing the gene of interest immediately downstream of the promoter" or something similar

Thank you for this good suggestion.

1. Any growth phenotype changes observed in the overexpressors?

The parasite lines seem to silence the Cen1-4-GFP expression plasmids readily, which suggest that there might be a growth disadvantage. However, repeated attempts to quantify a growth phenotype were unsuccessful due to high variability in the data, which might be partly connected to the fact that the fraction of GFP positive cells after induction can vary between lines and replicas.

1. How often are ECCAs observed in pARL strains, or are they not observed at all? This might be good to mention.

ECCAs in the pArl strains have been observed on very limited instances but are too rare to be quantified. We now mention this in the text (lines 217ff).

1. Line 192 and Figure S8: n {less than or equal to} 33 (either a typographical error and should have been {greater than or equal to}, otherwise, it may be expressed as a range)

It was indeed a typographical error that was now corrected.

1. Line 258: Methods on the generation of FIO/FIO+ was a bit difficult to understand. Maybe a simple plasmid schematic with the restriction sites (at least for the original plasmid) in the supplementary may help clarify this.

Cloning strategy has been expanded with additional information for clarity.

1. Line 295: include abbreviation of cRPMI here rather than in Line 303

Has been corrected.

1. Line 322: typographical error on WR99210 working concentration?

Has been corrected.

1. Line 372: Last sentence on area and raw integrated density measurement is unclear.

We have reformulated the sentence for more clarity.

1. Line 461: typographical error in last sentence

Has been corrected.

1. Line 532: Figure 4E should be Figure 4F

Has been corrected.

Reviewer #3 (Significance):

DNA replication is vital to the survival of malaria parasites. A deeper understanding on their unusual form of replication may be exploited to find drug targets uniquely directed to the parasite. Biological insights from this work can also provide a jump-off point for unravelling unusual replication in other organisms. Data on the physicochemical analysis of centrin is not just of great interest for those in the field of parasitology, but also for those in the much wider fields of biology, physics and chemistry. Techniques presented in this work (e.g., DiCre overexpression with different promoters) can definitely be utilised for the elucidation of protein function within and outside the field of parasitology.

My field of expertise is in Plasmodium spp., particularly in parasite replication, molecular and cellular biology, and epigenetics.

We thank the reviewer for the appreciation of our work in terms of insight and technology development.

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 have provided a comprehensive characterisation of centrin proteins in Plasmodium falciparum. Through expression of episomal GFP-tagged centrin for in vitro, they were able to observe co-localisation of centrin with centriolar plaques during the replicative stage of the parasite. They also utilised live cell STED microscopy to track dynamic changes in centrin morphology. They have also demonstrated calcium-dependent phase separation dynamics in bacterially-expressed P. falciparum centrin and human centrin 2. The formation of liquid-liquid phase separation in PfCen1, 3 and HsCen2 tied well with IUPred3 predictions of intrinsically disordered regions in these proteins. Using an inducible DiCre overexpression system with two promoters of varying strengths, the authors have shown accumulation of centrin1 outside of centrosomes and premature appearance of centriolar plaques. Finally, changes on the centrin1 protein, i.e., N-terminal deletion, and mutations in calcium binding sites in the EFh domains, have shown a reduction in the formation of ECCAs during overexpression and inability to form LLPS in vitro, respectively.

Major comments:

1. Given that parasites cannot tolerate endogenous C-terminal tagging of some centrins (but not all, as PbCen4 was successfully tagged), has N-terminal tagging been attempted either by the authors or in previous publications? Note that this is not a request for further experimentation; rather, maybe this can be noted in the manuscript; and line 62 can be rephrased for transparency.
2. Is there a possibility that by adding a C-terminal tag, centrin may lose a specific function or cause change in the physicochemical properties of the protein (thus making C-terminal tagging lethal)? Was His tag removal attempted so the native protein can be used in the LLPS experiments? IUPred3 analysis showed potential IDR at the C-terminal end of PfCen4. Could the C-terminal tag have caused the protein to not form droplets in the presence of Ca2+?
3. It has been shown by the authors that different tagged centrins co-condense which may support the localisation data (Figure 1C). However, is there a way to show that the episomally- and endogenously-expressed centrin co-localise with each other (e.g., confocal microscopy with anti-centrin vs anti-gfp in PfCen-GFP lines, that is if the authors have access to anti-centrin antibodies)? Has endogenous centrin been demonstrated to form ECCAs (in previous publications or by the authors)?

Minor comments:

1. How were the times (post addition of Ca2+) presented in Figure 2A determined?
2. Line 126: Figure 1B should be Figure 1C
3. Line 145: Figure 1C-D should be Figure 1D-E
4. Line 151: Figure 3A should be Figure 4A
5. Line 152: Suggest rephrasing "placing the gene of interest in front of the promoter" to "placing the gene of interest immediately downstream of the promoter" or something similar
6. Any growth phenotype changes observed in the overexpressors?
7. How often are ECCAs observed in pARL strains, or are they not observed at all? This might be good to mention.
8. Line 192 and Figure S8: n {less than or equal to} 33 (either a typographical error and should have been {greater than or equal to}, otherwise, it may be expressed as a range)
9. Line 258: Methods on the generation of FIO/FIO+ was a bit difficult to understand. Maybe a simple plasmid schematic with the restriction sites (at least for the original plasmid) in the supplementary may help clarify this.
10. Line 295: include abbreviation of cRPMI here rather than in Line 303
11. Line 322: typographical error on WR99210 working concentration?
12. Line 372: Last sentence on area and raw integrated density measurement is unclear.
13. Line 461: typographical error in last sentence
14. Line 532: Figure 4E should be Figure 4F

Significance

DNA replication is vital to the survival of malaria parasites. A deeper understanding on their unusual form of replication may be exploited to find drug targets uniquely directed to the parasite. Biological insights from this work can also provide a jump-off point for unravelling unusual replication in other organisms. Data on the physicochemical analysis of centrin is not just of great interest for those in the field of parasitology, but also for those in the much wider fields of biology, physics and chemistry. Techniques presented in this work (e.g., DiCre overexpression with different promoters) can definitely be utilised for the elucidation of protein function within and outside the field of parasitology.

My field of expertise is in Plasmodium spp., particularly in parasite replication, molecular and cellular biology, and epigenetics.

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 authors analyzed the properties of the four Centrin proteins of the malaria parasite using a combination of in vitro and in vivo approaches. Their findings indicate that two of the four Plasmodium Centrin proteins, PfCen1 and PfCen3, as well as the human Centrin protein HsCen2, exhibit features of biomolecular condensates. Moreover, analysis of cells overexpressing PfCen1 indicates that such biomolecular condensates become more numerous as cells approach mitosis and are dissolved thereafter.

Major comments

• A) A critical point that requires clarification is how the protein concentrations used in the in vitro and in vivo assays (20-200 microM in vitro, and not estimated in vivo) compare to that of the endogenous components. This is important because it may well be that 6His-tagged PfCen1, PfCen3 and HsCen2 can form biomolecular condensates when present in vast excess, but not when present in physiological concentrations. The authors should report the estimated cellular concentration of PfCen1-4, as well as that achieved upon PfCen1-GFP overexpression (on top of endogenous PfCen1), for instance using semi-quantitative immunoblotting analysis. Given this limitation, the authors may also want to temper their title by introducing the word "can" after "centrins".
• B) Movies S1 and S2 (and the related Fig. 1D and 1E) are not the most convincing to support the notion that the observed assemblies are biomolecular condensates, as not much activity is going on during the recordings. Likewise, Movies S3, and even more so Movie S4, as out of focus for a large fraction of the time, making it difficult to assess what happens at the beginning of the process. Moreover, it appears that fusion events, while occurring, are rather rare. The movies should be exchanged for ones that are in focus, and ideally a rough quantification of fusion events as a function of biomolecular condensate size provided.
• C) An important control is missing from Fig. 2, namely assaying PfCen1-4 without the 6His tag, to ensure that the tag does not contribute to the observed behavior (although it can of course not be sufficient as evidenced by the lack of biomolecular condensates for PfCen2 and PfCen4).
• D) The authors should test whether the assemblies formed by PfCen1 and PfCen3 are sensitive to 1,6-hexanediol treatment, as expected for biomolecular condensates.
• E) The fact that HsCen2 also forms biomolecular condensates is very intriguing, but further investigation would be needed to assess the generality of these findings. For instance, the authors could test in vitro also S. cerevisiae Cdc31, the founding member of the Centrin family of proteins to further enhance the impact of their study.

Minor comments

1. For the experiments reported in Fig. 3D, the same concentrations as those used in Fig. 3A-C (namely 10 microM, and not 30 microM as in Fig. 3D) should be used. Moreover, it would be informative to test whether PfCen2 and PfCen4 as PfCen3 when added to PfCen1.
2. The authors mention that the effect of Calcium in inducing biomolecular condensates is specific, as Magnesium was not effective (lines 94-95). However, an examination of Fig. S3B indicates that the Magnesium also exhibits some activity, albeit less potent than Calcium. The authors should discuss this point and rectify the wording in the main text.
3. Do the authors think that PfCen2 and PfCent4 localize to the centriole plaque in vivo using another mechanism that deployed by PfCen1 and PfCent3? It would be good to discuss this point.
4. Given that the EFh-dead mutant exhibits no activity in vitro and fails to localize in vivo, one potential concern is that the protein is misfolded. The authors should conduct a CD spectrum to investigate this.
5. It is not entirely clear from the main text in lines 103-104, as well as from the legend, what Fig. S3B shows. When was EDTA added in this case?
6. Fig. S7: the correlation between PfCen1-GFP expression levels and ECCA appearance is modest at best. What statistical test was applied? This should be spelled out. Moreover, the authors should combine the two data sets, as this will provide further statistical power to assess whether a correlation is truly present.
7. The authors may want to discuss how their findings can be reconciled with the notion that Centrin assemble into a helical polymer on the inside of the centriole (doi: 10.1126/sciadv.aaz4137).
8. Likewise, the authors may want to speculate regarding what their findings signify for the role of Centrin proteins in detection of nucleotide excision repair (doi: 10.1083/jcb.201012093).

Small things

• Fig. 1A: change color for microtubules as red on red is difficult to discernn.
• Fig. 1C: the indicated boxes in the top row do not seem to correspond exactly to the insets shown in the bottom row.
• line 266: typo, promotor > promoter.
• line 360: a reference should be provided for the GFP-booster, including the concentration at which it was used.
• line 363: "an" missing before "HC".
• line 428: it would be best to deposit the macros on Github or an analogous repository.
• line 461: "to the" is duplicated.
• Fig. S5A: maybe draw the lines in red (as red in Fig. S5B correspond to the proteins that do not have IDRs).
• Movie S7, legend: left frames shows PfCen1-GFP, not microtubules as currently stated.

Significance

This is a provocative study that extends initial observations regarding self-assembly properties of Centrin proteins, and posits that some members of this evolutionarily conserved family can form biomolecular condensates. After the above outstanding issues have been properly addressed, these data could have important implications for understanding Centrin function in centriole biology and DNA repair. Therefore, these findings will be of interest to a cell biology audience.

Field of expertise: cell biology.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Voss, Reinert and colleagues show calcium-dependent assembly of Plasmodium falciparum centrins in vitro and in parasites. This assembly is dependent on the EF-hands of centrin and an N-terminal disordered region.

Major concerns:

1. The very definitive title is not wholly supported by the data. This should be qualified by specifying the conditions under which the centrins can accumulate in this way.
2. A major concern is whether this behaviour of centrins represents a biologically relevant mechanism in centriolar plaque formation. Is this limited to high overexpression conditions or in vitro high concentrations? Or is it a result of the tagging of the P. falciparum centrins? A convincing approach to addressing this issue would be to knock-in a fluorescent tag to the centrin loci. Roques et al. (ref. 12 in this submission) report the GFP tagging of centrin-4 in P. berghei, although they note that centrins-1 to -3 were refractory to tagging in this organism. It is unclear whether Voss et al. attempted this tagging in P. falciparum. This should be clarified and relevant data presented.

If the tagged molecules used in the biochemical parts of this study are functional, It is challenging to understand why the centrins cannot be tagged in P. falciparum. If the tags render the P. falciparum centrins dysfunctional, the study becomes significantly less useful.<br /> 3. If a knock-in cannot be achieved, it must be shown that the transgenic expression of tagged Plasmodium centrins does not confound the analysis of centrin behaviour. It is known that these proteins can behave anomalously when overexpressed (Yang et al. 2010, PMID: 20980622; Prosser et al. 2009, PMID: 19139275), at least in other species.

A previous description of centriolar plaque from the authors' lab (Simon et al. 2021, PMID: 34535568) shows an organized structure of an established size. It should be demonstrated whether the structures formed with the GFP tagged centrins show the same dimensions and dynamics as those in wild-type parasites. The extent of the overexpression of the GFP-tagged centrins should also be demonstrated.<br /> 4. It would also be useful to remove the His tag from the recombinantly expressed and purified centrins for the in vitro analyses, particularly if concern remains about the impact of tags on Plasmodium centrin behaviour.<br /> 5. The discussion is very short and does not consider the findings presented here in the context of the literature, with respect to centrins, Plasmodium MTOC assembly mechanisms, or to general considerations around biological condensates. Andrea Musacchio's recent commentary (ref. 44 in the current submission) advocates caution in ascribing phase separation as an assembly mechanism for organelles in vivo, particularly on the basis of in vitro experiments with high concentrations of homogeneous protein. It is not clear that the concentration dependence of extracentrosomal centrin accumulations (ECCAs) at the onset of schizogony provides sufficient justification of a phase separation model in vivo. The authors' recent description of the involvement of an SFI1-like protein, SIp (Wenz et al. 2023 PMID: 37130129), in the centriolar plaque makes a case for non-homotypic interactions also driving assembly and alternative models for ECCA are not convincingly excluded. The absence of a robust discussion of such considerations is unhelpful to the reader.<br /> 6. It is also unclear whether the analysis of human centrin is suggested to indicate a phase separation mechanism for centrins in human cells. As this is readily testable, this notion could be considered further. Although its experimental examination may lie outside the theme of this study, one would expect some discussion of the significance of the data presented in the study.

Minor points

1. There are only three centrins in humans. Centrin 4 is a pseudogene (Gene ID: 729338 on NCBI).
2. Line 175 should say 'temporally', rather than 'temporarily. The Abstract should say 'evolutionarily conserved', rather than 'evolutionary conserved'. 'To condensate' is not ideal as a phrase- 'to form a condensate' would be clearer.

Referees cross-commenting

I think the other 2 reviewers have made fair, cogent and constructive points. There is good convergence between the reviewers on the significant issues around the study. These concern in vivo and in vitro effects of tagging and of of high concentrations.

Significance

The biology of the Plasmodium centriolar plaque is of great interest as an alternative MTOC structure, with obvious additional interest deriving from the role of this organism in malaria. Much remains to be learned about this structure, so the topic of this paper is likely to attract a broad readership. Furthermore, the centrins are a widely-expressed and evolutionarily conserved family of eukaryotic proteins, with multiple roles; a new model for their behaviour, such as is suggested here, would be of interest to many cell biologists.

With that in mind, significant additional data should be provided to substantiate the model proposed by the authors.

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.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The manuscript by Schauer et al. uses embryonic explants to study the coordination of Nodal and BMP signaling for embryo morphogenesis. They show that Nodal signaling triggers explant elongation by inducing mesendodermal progenitors that undergo cell intercalation. Looking at the role of BMP signaling, they show that BMP overactivation ventralizes the explants, reducing cell intercalation and therefore explant elongation. Looking at pSmad5, they then establish that BMP signaling in the explant is attenuated by Nodal signaling, through activation of chordin expression, and through some unidentified chordin-independent mechanisms. Moving to the entire embryo, using combinations of BMP overexpression and Nodal inhibition, authors show that Nodal signaling limits BMP signaling on the dorsal side of the embryo, which is key to proper embryo elongation.

Major comments:

• The authors used the sebox::EGFP line to show that the growing region of the explant consists mostly of mesendodermal cells. Although this transgenic line had not been used to do so, the authors and others, had previously demonstrated that the extending part of the explant is mostly made of mesoderm and even shows some patterning (1,2). This should be stated and not presented as a new finding.
• Explant elongation is driven by cell intercalation. The authors analyzed the shape of the mesendodermal tissue to conclude that cells intercalate. While I do not question this conclusion, as it is well known in the embryo, direct observation of cell intercalation, as was done in the embryo (3), would be a better demonstration.
• Explant elongation is driven by mesendodermal cell intercalation. I certainly agree from the movies and images that the extending region is mostly made of mesendoderm. However, it seemed clear to me that in Movie 1, starting at about 140 minutes, most of the convergence movement is taking place in a non-green region of the explant, fueling the extension of the mesendodermal region. Also, to demonstrate that cell intercalation is occurring in the mesendoderm, the authors performed clone dispersal analysis, comparing clones of mesendodermal and ectodermal cells. However, the selected ectodermal clone is very far from the extending region. To show that the cell intercalation is specific to mesendoderm, I think the authors should try to compare the behavior of mesendodermal and non-mesendodermal cells that are located at the same distance from the extending region. For example, from the image in Figure 1E (235 mpe), it appears that the right side of the base of the extended region is not green and could be compared to the left side. Currently, the quantification shown in 1G mostly demonstrates that the extending region is extending, and that the non-extending region is not.
• Based on their observations in explants, the authors propose that Nodal signaling maintains an area of low BMP signaling on the dorsal side of the gastrula for robust axis elongation. While I acknowledge that the experiments performed by the authors have not been previoulsy reported, I did not understand how this differs from the very well established fact that Nodal inhibits BMP signaling, in particular through chordin expression. Von der Hardt for instance already reported that overexpression of bmp and inhibition of chordin leads to severe elongation defects (4). More insight could probably be gained by analyzing the effect in more detail: Is the elongation defect due to cell intercalation defects? How are cell fates affected? Is this Nodal effect mediated by Chordin?...

Minor comments:

• Fig6B. Are the curves significantly different? If so, how were they compared?
• Fig6D-E, I found the quantification a bit confusing. The reader is left with the impression of an all-or-nothing answer (effect only with BMP overexpression and strong Nodal inhibition), whereas the effect on the pSmad5 gradient is gradual. Plotting and comparing the pSmad5 intensity gradients would be better.
• Fig6G. 'Axis length/embryo height' should appear on the x-axis, not the y-axis.

Referees cross-commenting

I feel that the three reviews are very much in agreement, recognising that the experiments carried out are well done and calling for a reasonable amount of additional data. The three reviews also agree that the results obtained here in explants were already known from intact embryos, limiting the relevance to ex vivo research.

Significance

Overall, the experiments appear carefully carried out, and very precisely quantified. The paper is well written and easy to read. The results add to our understanding of the morphogenetic events occurring in embryonic explants. I therefore support their publication.

My main concern is with the significance of the results. I am convinced that embryonic explants are great tools, to reduce the complexity of the embryo and to address questions that cannot be addressed in the embryo, as the authors and others have done, for instance, to address the role of extraembryonic tissues and patterning by maternal contributions. Here, however, I felt that most, if not all, of the experiments essentially demonstrate in embryonic explants, results that have been known for years in the intact embryo. While gathering detailed information on what happens in embryonic explants will certainly prove useful in further understanding the self-organizing abilities of these explants, and is worth publishing, the significance of the results reported here seems limited. Specifically, that elongation is driven by cell intercalation, that BMP-mediated dorsoventral patterning affects cell intercalation, that BMP signaling is attenuated by Nodal through Chordin, that Chordin is required for elongation, has been well established in the embryo over the last 20 years. Again, showing that it works the same way in embryonic explants is of interest, but at this point, does not add to our understanding of embryonic development.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This paper from the Heisenberg lab takes a reductionist approach to understanding how BMP and Nodal signaling interact to coordinate morphogenesis. They mostly use blastoderm explants that they culture in vitro. These explants elongate over time, with Nodal signaling that induces mesendoderm driving the cell intercalations that explain the elongation. They show that increased BMP signaling inhibits this process, but reducing BMP signaling has no effect. They see that reducing Nodal signaling results in an upregulation of BMP activity as read out by phosphorylated Smad5 staining and increasing Nodal signaling has the opposite effect. They explain this mostly by the observation that Nodal induces the expression of the BMP antagonist, Chordin, and validate this idea by demonstrating that a reduction in Chordin expression reduces explant elongation. Returnign to the embryo, the authors show that manipulation of Nodal signaling levels influences the size of the BMP activity gradient as expected from the in vitro results. Finally, they show that reduction of Nodal signaling with SB505124 sensitises the embryos the effects of bmp2b overexpression, and that BMP overeactivation at 90% epiboly reduced C&E movements.

Major comments

In general I think the work is well done and the data justify the conclusions.

I have several suggestions for additional experiments and discussion that I think would improve the paper.

1. In Figure S1 they present data on elongation of explants treated with a Nodal inhibitor. It would be good to show some examples of images of the explants.
2. In Figure 1G and 3A, the same wildtype images are shown. This is mentioned and I assume therefore that the results were all part of the same experiment. How many times were these experiments performed? It would be much better to use different biological replicates in the two figures.
3. It is important for the authors to make clear how many biological replicates each of the experiments correspond to.
4. In Figure 4E, it would be good to show the levels of P-Smad2 in the Oep and MZ lefty1, 2 explants.
5. On page 11 the authors mention chordin-independent inhibition of BMP signaling. The most likely candidate would be noggin as it too is expressed dorsally and is at least in part activated by Nodal. This should be tested in their model.
6. The authors focus on Chordin as downstream of Nodal signaling, and discuss the role of Nodal signaling in inducing chordin as being due to peak Nodal signaling. However, Chordin has been shown to also be downstream of Fgf signaling and Bozozok (PMIDs 23499658 and 16873584), which likely explains its dorsal expression domain. Furthermore, Rogers et al, (PMID 33174840) who the authors refer to, also show that to disrupt BMP signaling in embryos, inhibition of Nodal and Fgf is required. These issues need to be discussed in more detail. It is the combinatorial signaling that is thought to be responsible for the dorsal location of the chordin (and noggin) expression domains.

Minor comments

I think in general the manuscript is well written and the figures are clear. Previous data is generally well cited. My only comment is that there is a wealth of data from Xenopus and zebrafish that BMP antagonists are induced as a result of combinatorial Nodal signaling and other pathways (dorsal wnt and fgf) that inhibit BMP signaling. I think this could be better referenced.

Significance

The paper is well done and provides important information about the interactions between Nodal and BMP signaling to induce axis elongation. I think the work would be improved if the authors revise it along the lines suggested above. In terms of novelty, many of the component parts of the paper are known (Nodal signaling is important for elongation via cell intercalation and Nodal and BMP can antagonize on another by the induction of BMP antagonists by Nodal), but it is novel to put them together to investigate axis extension using explants. The paper will be of interest to those interested in how these signaling pathways operate in early vertebrate development and to those interested in morphogenesis.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The authors presented intriguing observations on the molecular mechanisms regulating morphogenic cell movement, with a particular focus on convergent-extension (CE) movement associated with cell type specification in the zebrafish blastoderm explant. In this manuscript, Schauer et al. identified the CE movement of the mesendoderm as triggering the elongation of the zebrafish embryonic explant. In this process, the Nodal signal represses the BMP signal, which negatively regulates the movement of the mesendoderm precursors, through the induction of its inhibitor chordin. This suggests that the Nodal signal is the key factor coordinating cell fate specification and morphogenesis in the zebrafish blastoderm explant. Finally, suppression of Nodal signalling increases sensitivity to BMP signalling in the CE movement of intact embryos. This suggests that promotion of mesendoderm cell intercalation via BMP suppression by Nodal may be involved in conferring robustness to morphogenic cell movement in vivo.

Major comments

1. While one of the main conclusions of this manuscript is that "Nodal signaling regulates CE movement of mesendodermal cells by promoting their intercalation through inhibition of BMP signaling". However, this was predicted by changes in individual cell morphology and cell dispersal, and the authors didn't directly examine the behavior of individual cells. It would be better to confirm intercalation during the process of explant elongation by cell tracking analysis.
2. Although the authors discuss that Nodal signaling inhibits BMP signaling in the later gastrulation stage, this has not been experimentally tested. If possible, the time window in which Nodal signaling acts should be investigated by temporal inhibition of Nodal signaling using chemical inhibitors.
3. Only the signal gradient of pSmad5 and axis elongation were examined in the intact embryo part of the study (Fig. 6 and Fig. S7). The information on the domain of pSmad2 and the expression of chordin would be helpful for the comparison of the blastoderm explant and the intact embryos.

Minor concerns

The first letter of a gene name should be in lowercase. ( ex. Fig.S3C; Smad5 MO)

Significance

The zebrafish blastoderm explant assay has the potential to elucidate the molecular mechanisms regulating the complex processes of morphogenesis during vertebrate gastrulation, as the authors demonstrate in this paper. In this manuscript, the authors addressed the molecular mechanism coordinating cell fate specification and morphogenic cell movement in the blastoderm explant. All of the experiments are well-designed, the interpretation of the results is convincing and the paper is well-written. Also, the conclusion is very clear and well supported by the presented data. These findings provide fundamental and important insights for studying morphogenic cell movements in early vertebrate embryos using zebrafish blastoderm explants. On the other hand, most of the molecular mechanisms reported in this manuscript are already predicted by previous studies using intact embryos. Therefore, the impact of this work may be limited to ex vivo research.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary: This manuscript describes molecular mechanisms by which ACBD3 is recruited to the Golgi complex. ACBD3 recruits PI4KIIIb which is required to generate PI4P, a phosphoinositide which is key for the recruitment of essential Golgi proteins and hence is key to Golgi identity. The authors have used a combination of mass spectrometry, high quality fluorescence imaging, transient CRISPR knockdowns, and biochemical approaches such as IPs to identify the key determinant for recruitment of ACBD3 to the Golgi complex. They map the interaction between ACBD3 and the Golgi as a unique region (UR) upstream of its GOLD domain, identifying, in particular, an MWT motif as key for this recruitment. Using mass spectrometry they identify several novel interactors of ACBD3 as well as some established binding partners. Knockdown of these interactors reveal a key role for the SNARE, SCFD1, where reduced levels lead to complete loss of ACBD3 localisation to the Golgi without apparent disruption of Golgi structure. They further validate this interaction and that of another SNARE (Sec22b), which is part of the same SNARE complex as SCFD1, mapping the interaction to the longin domain of Sec22b. Surprisingly however they demonstrate that the UR domain does not mediate the interaction between ACBD3 and these SNAREs suggesting an alternative mechanism of recruitment. Previously identified ACBD3 interactors, Golgi proteins giantin and golgin-45 were also identified in the mass spectrometry screen and the authors demonstrate that these two proteins can recruit ACBD2 to the Golgi and this is dependent on the MWT motif identified in the UR domain. By knocking down SCFD1, they show reduced recruitment of ACBD3 leading them to propose a model of sequential recruitment of ACBD3 by SCFD1 followed by interactions with the golgins.

Major points: This study is a well-executed and rigorous study of the molecular requirements for the recruitment of ACBD3 to the Golgi. The experimental approaches are state-of-the-art and the data are clean and convincing. The only caveat, raised by the authors themselves, is their interpretation that there are two sequential steps for Golgi recruitment of ACBD3. While they show that loss of SCFD1 reduces the interaction of ACBD3 with giantin and golgin 45, their model depends on doing the reverse experiment, i.e. assessing the effects of knocking down either giantin or golgin-45. This is especially relevant given the demonstration that golgin-45 is sufficient to recruit ACBD3 to mitochondria. It may well be that recruitment involves a tripartite complex, which is not uncommon in vesicular transport mechanisms Giantin is not an essential protein do it should be feasible to perform this experiment. The authors are equipped in the quantitative fluorescence microscopy which would be required and which would help resolve whether sequential or redundant mechanisms are required for ACBD3 recruitment.

We thank the reviewer for the positive comments and are glad that they consider our study "well-executed and rigorous". We totally agree with the reviewer that our conclusions regarding the sequential aspect of the recruitment of ACBD3 in the original submission could be better supported. We have worked to strengthen this in our resubmission. As the reviewer states, this limitation was already discussed in the original submission. To further support our model, we have performed the experiment suggested by the reviewer, in which we test the effects of knocking down both giantin and golgin45 (double knockdown) on the binding of ACBD3 to SCFD1.

The results of this experiment further support our sequential model with little to no effect of loss of the Golgins on ACBD3. As we already knew, a large effect of SCFD1 KO on the binding of the Golgins to ACBD3 was also observed here. We should note that this was performed in a different cell line than before (HeLa cells rather than HEK cells), as the efficiency of multiple knockdowns was much lower in HEK cells, as determined by qPCR. Taken together, the new data in Figure 7 supports a sequential model for Golgi recruitment. We also agree that other, less likely models could explain our data and have included this openly in the discussion. In conclusion, we thank the reviewer for their comments and have revised the manuscript with a new experiment with the relevant repeats, which supports our model.

Reviewer #1 (Significance):

Significance PI4P is a phosphoinositide that is important for the recruitment of Golgi proteins. As with most PIs it is likely to act by coincidence detection in that Golgi associated proteins will recognise PI4P as well as other factors on Golgi membranes. This results in different local membrane environments which will be specific for particular functions. PI4KIII__b_ is key for PI4P production although the absolute levels of PI4P are likely to be determined by a balance of lipid kinases and phosphatases. However, since ACBD3 is key for the recruitment of PI4KIII__b, it is important to understand the molecular mechanisms by which it is recruited. The manuscript thus makes a significant contribution to understanding one of the underlying mechanisms for PI4KIII__b _recruitment although, as indicated above, stops short of establishing a clear model for the roles SCDF1 and Sec22b versus golgin 45 and giantin. For the future it will be of interest to determine why either a sequential or a redundant mechanism is required for the recruitment of ACBD3 as a scaffold protein.

We thank the reviewer for this set of positive comments on the manuscript and for agreeing that this is a significant contribution. Our revised version further supports our sequential model of ACBD3 recruitment to the Golgi apparatus, and the comments here have helped us further to strengthen the quality and clarity of the manuscript.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary This is a very interesting and potentially important paper for the field of membrane biology and membrane trafficking, in which the authors have studied the molecular mechanisms by which ACBD3 (and consequently PI4KIIIb) is recruited to the cis-Golgi membranes. The authors suggest that this recruitment is based on a two-step process, mediated by interactions to, on the one hand, SCFD-1 (SLY1) and, on the other hand, two redundant golgins (golgin-45 and giantin).

We once again thank the reviewer for the positive comments and are glad that they consider our manuscript important.

Comments:- Pg.1 : arfaptins, as far as I know, have not been shown to be involved in intra-golgi trafficking but rather in Golgi export (see e.g. ref. 12)

We thank the reviewer for pointing this out. We have corrected the text accordingly.

• Pg. 1: reigon --> region

We thank the reviewer for noticing this typo. We have corrected the text accordingly.

• Arf1 also recruits PI4KIIIb right?

This is correct. The De Matteis lab has shown that PI4KIIIβ associates with the Golgi complex in an Arf1-dependent manner (Godi et al. 1999). We think this is excellent work. However, Arf1 is somewhat of a master regulator of the Golgi, affecting the recruitment and localisation of many different Golgi proteins. It has also previously been reported that Arf1 does not directly interact with PI4KIIIβ (Klima et al. 2016). Overall, the molecular relationship between Arf1 and the kinase remains unclear. We do not exclude, however, that there are factors other than ACBD3 important for recruiting and regulating PI4KIIIβ levels at the Golgi. We have changed the wording in the manuscript to reflect that there are multiple ways that PI4KIIIβ is recruited to the Golgi apparatus.

Fig. S1: the information about the number of cells per experiment is missing. Also, please add the information about what exactly is represented in the box plots (is it the distribution of the mean value of R per experiment? or the total distribution on a cell-by-cell basis of a representative experiment?)

For each experiment, a minimum of 100 cells per condition were imaged. The Pearson's correlation was then calculated, and the average was taken for each biological repeat. The plot in Fig. S1B represents 3 independent biological repeats. We have included this information in the revised manuscript.

• The definition of Avg. Golgi int/avg. cell int. (a.u.) in Fig 1E,F is a bit difficult to understand to me. If I understand correctly, the total fl. int in the Golgi mask was computed and divided by the area of the Golgi mask (this is the av. Golgi intensity). A similar computation is done for the entire cell (including the Golgi), i.e., total fl. intensity in the cell mask is computed and divided by the area of the cell mask. Then the two av. intensities are divided (ratio = av. Golgi int / av. cell int.). This ratio, for a protein that is enriched in the Golgi area, should be larger than 1. For a protein that is equally distributed all over the cell, it should be 1, and for a protein that is excluded from the Golgi area, smaller than 1. Then to this value, the authors subtract the value of the ratio found for an inert construct (GFP of Halo alone), which I imagine should have an original ratio value of the order of 1, and hence, after this subtraction, norm. ratio values larger than 0 mean that they are more enriched at the Golgi area than GFP/HaloTag themselves. Is this correct? In principle, I don't see anything entirely wrong with this way of thought, but I just found it a bit difficult to understand, and in general one has to be careful when computing rations (quotients) and then subtract another ratio. Also, the units are not a.u., the value is dimensionless, what is "arbitrary" is the definition of 0 value and the based on this definition, also the actual value. I think it would probably be much clearer for the readers to compute somthing like the relative enrichment in the Golgi area as compared to the rest of the cell (excluding the Golgi area). That is, a value r'=(Int. Golgi mask / Area Golgi mask) / [(Int. Cell mask - Int. Golgi mask)/(Area cell mask - Area Golgi mask)]. This can be computed directly or defining a mask that is the cell mask - the Golgi mask. Also, some maths (unless I made a mistake) give that this r'= r (1-aG)/(1-r aG); where r is the ratio (before subtraction) defined by the authors, and aG=Area cell mask/Area Golgi mask. In any case, I'd suggest the authors to either adopt this other quantitation (without subtraction of the GFP/HAloTAG), which gives directly the fold-enrichment in the intensity density in the Golgi area with respect to the rest of the cell; or explain in more detail the maths of the value they are plotting now.

We thank the reviewer for these well-reasoned and thoughtful suggestions for our imaging analysis. These are issues that we have also considered when quantifying this dataset. At the heart of it, the second method of calculation (Golgi/outside of Golgi), results in a non-linear distribution, as the pool of proteins re-distribute from inside the Golgi to the cytosol. This is why we have chosen to use the first method of Golgi/total, as it provides a linear distribution.

The reviewer is also correct that the GFP (inert protein) ratio is 1 without adjustment. We have chosen to normalise to GFP/HaloTag (inert protein) as we think this is the clearest way of conveying our conclusions from these experiments. We have included the non-normalised graph here for the reviewer to see; however we thought that this conveys the key result less clearly. Overall, we agree this was poorly communicated in the manuscript and we have clarified it in the revised version.

• Fig. 1C&F: Besides the MWT mutant, the FKE mutant also seems to have a somewhat compromised Golgi localization. Have the authors followed on that, or what is the reason that they have just focused on the MWT mutant?

In contrast to the MWT mutant, the FKE mutant does not affect ACBD3 localisation significantly. In addition, when having a close look at the pdb structure of the GOLD domain of ACBD3 with 3A protein of Aichivirus A (5LZ3), the MWT patch, in particular residues M and T, make clear contact with protein 3A, which is not the case for FKE residues. Therefore we focused on the MWT residues, which we hypothesised to interact with a Golgi resident protein which competes with protein 3A to interact with ACBD3.

• Very minor point, and without wanting to sound pedant at all, but I think (I might be wrong of course, so apologies if I am) that the plural of apparatus in latin is not apparati, but apparatus (fourth declination). So, I'd change the word in page 2 (or just rephrase the sentence: e.g. "resulting in Golgi fragmentation"). But of course, I'd leave this to the authors' discretion.

We thank the reviewer for this precision, do not consider it pedantic, and have made the suggested change to the text.

• Fig. 3A: have the authors tried or been able to perform IF of the endogenous SCFD1 protein?

As suggested by the reviewer, we attempted to perform IF of endogenous SCFD1, as shown below. Despite trying several different antibodies, we were not satisfied that we were detecting real SCFD1 signal as there was no change in this staining upon SCFD1 CRISPR KO. Please see an example of this IF below (ProteinTech, 12569-1-AP). We have contacted the antibody manufacturers to inform them of this issue.

• Similarly to what has been done for other panels, could you quantify Fig. 3C? Are PI4KIIIb protein levels affected upon the different KOs?

As suggested by the reviewer, we are now showing in Figure S2D the percentage of cells with a partial or total loss of PI4KIIIβ at the Golgi in CRISPR-Cas9 KO cells of either PI4KIIIβ, ACBD3 or SCFD1. 3 independent biological repeats were performed and approximately 150 cells were quantified (~50 cells per condition). The results show that the PI4KIIIβ antibody used (BD Bioscience, 611816) is specific (93.22% of cells lose the antibody signal) and that ACBD3 and SCFD1 KO affects PI4KIIIβ recruitment to the Golgi in 88% and 73% of the cells, respectively._-

The last paragraph of the "SCFD1 and ACBD3 interact upstream of PI4KIIIβ recruitment to the Golgi apparatus" section reads a bit odd placed there. I think it is more appropriate for the discussion or for the intro part on SCFD1.

Many thanks to the reviewer for pointing this out. We simplified that paragraph to describe the relationship between SCFD1 and SEC22B.

• I am confused on Fig. 5A/B. The labels in the blots show that 390-528 (without UR) does not bind sec22 or scfd1, but the 368-529 does? Or I guess, judging by the MW seen in the middle blots, that there's some error in the labelling?

Many thanks to the reviewer for noticing this, which was clearly a labelling error. We corrected this accordingly in Figures 5A and B. We apologise for this oversight.

also, the IP efficiency of the MWT mutant in the panel A blot is quite low, still sec22 seems to be very efficiently pulled down. Can the authors comment on that please? Would co-IPing against endogenous sec22 and scfd1 would work (so you don't need to rely on HaloTag+ligand?)

We know that the MWT residues of ACBD3 are important for recruiting ACBD3 to the Golgi (Figure 1C and F). We also know that ACBD3 interacts with SEC22B and SCFD1 (Figure 3B and 4A) and that SCFD1 is important for ACBD3 Golgi recruitment. Therefore we initially speculated that ACBD3 interacts with SEC22B and SCFD1 through the MWT residues. However, as the reviewer points out, Figure 5 shows the opposite. Mutating MWT residues makes the interaction of ACBD3 with SEC22B and SCFD1 stronger. For this reason, we hypothesised that another player(s) also contributes to ACBD3 recruitment through interactions with the MWT residues. We have shown that the second recruitment factors are the 2 golgins, golgin-45 and giantin (Figure 6C). In short, whilst we agree that the IP efficiency is low, the binding is actually stronger, supporting our conclusions. No interaction of ACBD3 with endogenous SEC22B could be detected due to a lack of a sufficiently sensitive antibody (we tried Abcam ab181076 and ProteinTech 14776-1 AP).

• I really like the experiment 6B. Have the authors tested whether SEC22 is also recruited to mitochondria in those conditions? But not SCFD1?

We thank the reviewer for the positive comment. We have performed the suggested experiment and are now including this as an additional figure (Figure S3). Ectopic expression of golgin-45 targeted to the mitochondria is not sufficient to redistribute SCFD1-HaloTag or HaloTag-SEC22B to the mitochondria (Figure S3A and B, respectively). We, therefore, speculate that the fraction of ACBD3 that gets redirected in Figure 6B must be the small fraction of ACBD3 that is spontaneously in an open conformation and compatible for interaction with golgin-45.

• The results shown in Fig 7 might show a partial depletion in the interactions, but to be fully trusted they would need to be quantified and a statistical test used to compare the values. I think this part is important to show very clearly, because even with low binding to golgins (remember, single knockouts do not prevent Golgi localization of ACBD3), one could expect that ACBD3 still localized to the Golgi but it does not in the absence of SCFD1 as shown in this paper. A prediction of the proposed model is that in cells depleted of the two Golgins, SCFD1 and ACBD3 should still bind to one another, right? Did the authors test this?

We fully agree with the reviewer. As discussed in the replies to reviewer 1, we have repeated this experiment, including both sets of KO. This was not trivial, as a double transient KO is technically challenging and involves validation with qPCR and switching cell types (HEK cells to HeLa). The new data supports our current model and suggests some additional regulatory mechanisms at play.

• The model presented here (fig 8) seems to suggest that only the conformational variation of ACBD3 that binds Golgins is able to recruit (bind) PI4KIIIb. Is this known, or is there any experimental evidence for that?

HDX-MS experiments show that the ACBD and GOLD domains undergo conformational changes in the presence of 3A proteins (McPhail et al. 2017). Demonstrating this would require a complicated reconstitution experiment which is technically very challenging and would involve purifying various complex proteins, including SNAREs, SM proteins and golgins. This could perhaps be the subject of several future studies.

• Have the authors thought about testing the FKE mutant in the experiemnts shown in Fig. 5?

As mentioned above, since the FKE residues are not making any contact with the protein 3A and since the loss of ACBD3 recruitment to the Golgi is not statistically significant (Figure 1F), we haven't tested the FKE mutant for the binding to SEC22B and SCFD1. We do, however, agree with the reviewer that there might be something interesting happening here. We would like to experimentally interrogate this in future studies and develop more sensitive assays to test if there is a significant effect with the FKE mutant.

In general, I think the title might be a bit misleading because of the use of PI4Kiiib. I understand what the authors mean, but because they have not thoroughly tested PI4Kiiib recruitment in their experiments, I think they should focuse rather on the mechanism of recruitment of ACBD3 the authors have found.

We thank the reviewer for their advice regarding the manuscript title, and this is something that we have discussed internally. We chose that title as it highlights the key mechanistic impact of our findings and note that we did include a figure on the recruitment of PI4KIIIβ. However, we remain open to discussing this with advice from the journal editorial team.

Reviewer #2 (Significance):

I think, as said above, that this is potentially an important paper for the field of membrane trafficking and membrane biology. Most of the experiments are in general well performed and well controlled, and the paper is clearly written and follows a logical line.

We once again thank the reviewer for their comments and overall thoughtful and considered review. We believe that the suggestions here have improved the manuscript.

Reviewer #3 (Evidence, reproducibility and clarity):

Stalder and colleagues report experiments designed to identify interactors of the Golgi-localized protein ACBD3 (a.k.a. GCP60), and to delineate mechanisms that allow ACBD3 to localize at Golgi compartments. ACBD3 is a 528aa protein with diverse previously reported interactions and functions, both in normal physiology and as a host factor in viral assembly processes. Stalder et al. first map which domains of ACBD3 are required for Golgi localization in HeLa cells, concluding that residues 368-528 are sufficient for localization. This region includes a GOLD (GOLgi Dynamics) domain previously reported to interact with Golgin tethering proteins. Alanine scanning identifies the motif MWT just upstream of the GOLD motif as necessary for Golgi localization. Acute CRISPR knockout identifies two Golgins, Golgin45 and Giantin, as necessary for ACBD3 Golgi localization, and IP indicates that the MWT motif breaks this interaction. These data are a bit scattered around the paper but taken together are reasonably persuasive, particularly when viewed in context with published work. This reader would have found the manuscript easier to follow had the Golgin and MWT motif data been presented en bloc.

We thank the reviewer for these comments and have considered presenting and rewriting the data as the reviewer suggested. On reflection, we have decided to present it in the original order. We feel that this allows us to highlight the two independent mechanisms individually, bringing them together at the end. In addition, as the experiments were performed in the order presented, it allows for more appropriate controls for each experiment rather than trying to combine them. We hope the reviewer accepts our preferred order.

In a second set of experiments, IP-mass spec is used to identify ACBD3 interactors that might assist in the protein's localization. The MS data presented are filtered to exclude proteins not already identified as Golgi-localized. This is, I think, a mistake. Even if the authors choose to focus on known Golgi interactors as candidates for a localization function, the biological functions of ACBD3 are far from fully understood, and the full dataset would be of value to both cell biologists and virologists.

We agree with the reviewer that there are many interesting mysteries surrounding ACBD3 and have therefore included an additional table (table S1) in the revised manuscript, showing the dataset of newly identified ACBD3 interactors before applying the Golgi localisation filter.

Hits in the filtered dataset include the R-SNARE Sec22B, and the SNARE chaperone Sly1/SCFD1. Acute CRISPR inactivation of Sec22 decreases ACBD3 localization to the Golgi and SCFD1 inactivation more or less abolishes localization. Co-IP experiments are used to argue that ACBD3 interacts with the N-terminal regulatory Longin domain of SEC22B, as well as with SCFD1. The Sec22 data are more detailed and persuasive. No experiments with purified proteins are presented to establish that the detected interactions are direct rather than mediated through a bridging factor or factors. Importantly, SCFD1 is likely to have multiple different client SNARE complexes that operate at different stages of ER and Golgi traffic. Hence its inactivation is likely to be pleiotropic and consequently phenotypes arising must be interpreted with caution.

We completely agree that studying membrane trafficking in an interconnected system is challenging. We also agree that direct binding experiments in reconstituted systems would be key to proving our model. Our data uses multiple different experimental approaches, including co-localisation, co-immunoprecipitation, CRISPR-KO, and biochemistry, to support our model. In the future, we agree full reconstitution would be necessary to examine this further, and we hope that either ourselves or others can do this in further studies.

Lastly, the authors perform IP experiments which show that ACBD3-Golgin co-IP efficiency is lower in cells with acute inactivation of SCFD1. This epistatic relationship is used to argue for a sequential model of recruitment with SCFD1 and perhaps client SNARE proteins operating upstream of ACBD3-Golgin interaction. This argument is not persuasive because we do not know whether SCFD1 and its downstream activities increase the rate of ACBD3-Golgin complex asssembly, or alternatively stabilizes ACBD3-Golgin complexes, decreasing the rate of their dissociation.

We agree with this weakness in our original submission, and it is a comment shared among all reviewers. Overall, we feel that we have chosen the model that best summarises our data. We, of course, accept that there are still components of this pathway that need clarification and are open for further study. This includes the issue raised here by the reviewer, as well as the intriguing observation that both golgins are transcriptionally upregulated upon SCFD1 KO in HeLa cells. In the revised manuscript, we have more clearly laid out the weaknesses of our model in the discussion and suggested future experiments to help clarify some of these issues. We have also modified the model to reflect some of these potential additional regulatory mechanisms.

In general the methods are fairly clear but that there is room for improvement. The "high throughput" imaging pipeline is not clearly described.

We agree with the reviewer, and apologise for not clearly explaining this. We feel that this unbiased approach of quantification is particularly rigorous and we have clarified this in the methods section of the updated manuscript.

Each figure legend should specify the microscopy methods used, and for each result the number of biological replicates and cells analyzed should be specified.

We agree with the reviewer and have included these details appropriately in the revised manuscript.

The statistical methods (Student, Tukey, etc.) used for each experiment should be specified. Saying that statistics were calculated using Python 3.7 is useless without additional details. e.g. at least the libraries and codebase used should be indicated or deposited.

We agree with the reviewer and have updated the manuscript accordingly. In short, all comparisons were made using either Student's t-test or Multiple Comparison of Means - Tukey HSD, FWER=0.05. These were conducted in Python 3.9 using pandas, matplotlib, seaborn and scipy. We used the MultiComparison function in scipy, and the comp.tukeyhsd for the post-hoc adjustment.

Many figure labels (e.g. Fig. 2) use absurdly small fonts.

We apologise for this. We believe that this is because we submitted it with in-line formatting. Our resubmission has full-page figures, and we feel the text is clearer now.

The mass spec hits obtained should be provided both with and without exclusion of non-Golgi-localized proteins.

We agree with the reviewer. Please see the new Table S1.

Reviewer #3 (Significance):

In general I think this is a useful and well controlled set of experiments producing useful insights. However, the interpretations need to be more carefully considered, and alternative interpretations must laid out as clearly as possible. Specifying the limitations of the study will make it more, not less, useful to the field. If the authors want to make the case more robustly that the interactions described are mediated through direct binding, or that the operation of SCFD1 and Golgins operate sequentially to recruit ACBD3, additional wet bench work will be required which will of course take time to complete.

We once again thank the reviewer for the thoughtful and critical comments. These have helped to strengthen the manuscript. We have performed the additional bench work requested by the reviewer, which has further supported the paper and our model.

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

Stalder and colleagues report experiments designed to identify interactors of the Golgi-localized protein ACBD3 (a.k.a. GCP60), and to delineate mechanisms that allow ACBD3 to localize at Golgi compartments. ACBD3 is a 528aa protein with diverse previously reported interactions and functions, both in normal physiology and as a host factor in viral assembly processes. Stalder et al. first map which domains of ACBD3 are required for Golgi localization in HeLa cells, concluding that residues 368-528 are sufficient for localization. This region includes a GOLD (GOLgi Dynamics) domain previously reported to interact with Golgin tethering proteins. Alanine scanning identifies the motif MWT just upstream of the GOLD motif as necessary for Golgi localization. Acute CRISPR knockout identifies two Golgins, Golgin45 and Giantin, as necessary for ACBD3 Golgi localization, and IP indicates that the MWT motif breaks this interaction. These data are a bit scattered around the paper but taken together are reasonably persuasive, particularly when viewed in context with published work. This reader would have found the manuscript easier to follow had the Golgin and MWT motif data been presented en bloc.

In a second set of experiments, IP-mass spec is used to identify ACBD3 interactors that might assist in the protein's localization. The MS data presented are filtered to exclude proteins not already identified as Golgi-localized. This is, I think, a mistake. Even if the authors choose to focus on known Golgi interactors as candidates for a localization function, the biological functions of ACBD3 are far from fully understood, and the full dataset would be of value to both cell biologists and virologists. Hits in the filtered dataset include the R-SNARE Sec22B, and the SNARE chaperone Sly1/SCFD1. Acute CRISPR inactivation of Sec22 decreases ACBD3 localization to the Golgi and SCFD1 inactivation more or less abolishes localization. Co-IP experiments are used to argue that ACBD3 interacts with the N-terminal regulatory Longin domain of SEC22B, as well as with SCFD1. The Sec22 data are more detailed and persuasive. No experiments with purified proteins are presented to establish that the detected interactions are direct rather than mediated through a bridging factor or factors. Importantly, SCFD1 is likely to have multiple different client SNARE complexes that operate at different stages of ER and Golgi traffic. Hence its inactivation is likely to be pleiotropic and consequently phenotypes arising must be interpreted with caution.

Lastly, the authors perform IP experiments which show that ACBD3-Golgin co-IP efficiency is lower in cells with acute inactivation of SCFD1. This epistatic relationship is used to argue for a sequential model of recruitment with SCFD1 and perhaps client SNARE proteins operating upstream of ACBD3-Golgin interaction. This argument is not persuasive because we do not know whether SCFD1 and its downstream activities increase the rate of ACBD3-Golgin complex asssembly, or alternatively stabilizes ACBD3-Golgin complexes, decreasing the rate of their dissociation.

In general the methods are fairly clear but that there is room for improvement. The "high throughput" imaging pipeline is not clearly described. Each figure legend should specify the microscopy methods used, and for each result the number of biological replicates and cells analyzed should be specified. The statistical methods (Student, Tukey, etc.) used for each experiment should be specified. Saying that statistics were calculated using Python 3.7 is useless without additional details. e.g. at least the libraries and codebase used should be indicated or deposited. Many figure labels (e.g. Fig. 2) use absurdly small fonts. The mass spec hits obtained should be provided both with and without exclusion of non-Golgi-localized proteins.

Significance

In general I think this is a useful and well controlled set of experiments producing useful insights. However, the interpretations need to be more carefully considered, and alternative interpretations must laid out as clearly as possible. Specifying the limitations of the study will make it more, not less, useful to the field. If the authors want to make the case more robustly that the interactions described are mediated through direct binding, or that the operation of SCFD1 and Golgins operate sequentially to recruit ACBD3, additional wet bench work will be required which will of course take time to complete.

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 is a very interesting and potentially important paper for the field of membrane biology and membrane trafficking, in which the authors have studied the molecular mechanisms by which ACBD3 (and consequently PI4KIIIb) is recruited to the cis-Golgi membranes. The authors suggest that this recruitment is based on a two-step process, mediated by interactions to, on the one hand, SCFD-1 (SLY1) and, on the other hand, two redundant golgins (golgin-45 and giantin).

Comments:

• Pg.1 : arfaptins, as far as I know, have not been shown to be involved in intra-golgi trafficking but rather in Golgi export (see e.g. ref. 12)
• Pg. 1: reigon --> region
• Arf1 also recruits PI4KIIIb right?
• Fig. S1: the information about the number of cells per experiment is missing. Also, please add the information about what exactly is represented in the box plots (is it the distribution of the mean value of R per experiment? or the total distribution on a cell-by-cell basis of a representative experiment?)
• The definition of Avg. Golgi int/avg. cell int. (a.u.) in Fig 1E,F is a bit difficult to understand to me. If I understand correctly, the total fl. int in the Golgi mask was computed and divided by the area of the Golgi mask (this is the av. Golgi intensity). A similar computation is done for the entire cell (including the Golgi), i.e., total fl. intensity in the cell mask is computed and divided by the area of the cell mask. Then the two av. intensities are divided (ratio = av. Golgi int / av. cell int.). This ratio, for a protein that is enriched in the Golgi area, should be larger than 1. For a protein that is equally distributed all over the cell, it should be 1, and for a protein that is excluded from the Golgi area, smaller than 1. Then to this value, the authors subtract the value of the ratio found for an inert construct (GFP of Halo alone), which I imagine should have an original ratio value of the order of 1, and hence, after this subtraction, norm. ratio values larger than 0 mean that they are more enriched at the Golgi area than GFP/HaloTag themselves. Is this correct? In principle, I don't see anything entirely wrong with this way of thought, but I just found it a bit difficult to understand, and in general one has to be careful when computing rations (quotients) and then subtract another ratio. Also, the units are not a.u., the value is dimensionless, what is "arbitrary" is the definition of 0 value and the based on this definition, also the actual value. I think it would probably be much clearer for the readers to compute somthing like the relative enrichment in the Golgi area as compared to the rest of the cell (excluding the Golgi area). That is, a value r'=(Int. Golgi mask / Area Golgi mask) / [(Int. Cell mask - Int. Golgi mask)/(Area cell mask - Area Golgi mask)]. This can be computed directly or defining a mask that is the cell mask - the Golgi mask. Also, some maths (unless I made a mistake) give that this r'= r (1-aG)/(1-r aG); where r is the ratio (before subtraction) defined by the authors, and aG=Area cell mask/Area Golgi mask. In any case, I'd suggest the authors to either adopt this other quantitation (without subtraction of the GFP/HAloTAG), which gives directly the fold-enrichment in the intensity density in the Golgi area with respect to the rest of the cell; or explain in more detail the maths of the value they are plotting now.
• Fig. 1C&F: Besides the MWT mutant, the FKE mutant also seems to have a somewhat compromised Golgi localization. Have the authors followed on that, or what is the reason that they have just focused on the MWT mutant?
• Very minor point, and without wanting to sound pedant at all, but I think (I might be wrong of course, so apologies if I am) that the plural of apparatus in latin is not apparati, but apparatus (fourth declination). So, I'd change the word in page 2 (or just rephrase the sentence: e.g. "resulting in Golgi fragmentation"). But of course, I'd leave this to the authors' discretion.
• Fig. 3A: have the authors tried or been able to perform IF of the endogenous SCFD1 protein?
• Similarly to what has been done for other panels, could you quantify Fig. 3C? Are PI4KIIIb protein levels affected upon the different KOs?
• The last paragraph of the "SCFD1 and ACBD3 interact upstream of PI4KIIIβ recruitment<br /> to the Golgi apparatus" section reads a bit odd placed there. I think it is more appropriate for the discussion or for the intro part on SCFD1.
• I am confused on Fig. 5A/B. The labels in the blots show that 390-528 (without UR) does not bind sec22 or scfd1, but the 368-529 does? Or I guess, judging by the MW seen in the middle blots, that there's some error in the labelling? also, the IP efficiency of the MWT mutant in the panel A blot is quite low, still sec22 seems to be very efficiently pulled down. Can the authors comment on that please? Would co-IPing against endogenous sec22 and scfd1 would work (so you don't need to rely on HaloTag+ligand?)
• I really like the experiment 6B. Have the authors tested whether SEC22 is also recruited to mitochondria in those conditions? But not SCFD1?
• The results shown in Fig 7 might show a partial depletion in the interactions, but to be fully trusted they would need to be quantified and a statistical test used to compare the values. I think this part is important to show very clearly, because even with low binding to golgins (remember, single knockouts do not prevent Golgi localization of ACBD3), one could expect that ACBD3 still localized to the Golgi but it does not in the absence of SCFD1 as shown in this paper.
• A prediction of the proposed model is that in cells depleted of the two Golgins, SCFD1 and ACBD3 should still bind to one another, right? Did the authors test this?
• The model presented here (fig 8) seems to suggest that only the conformational variation of ACBD3 that binds Golgins is able to recruit (bind) PI4KIIIb. Is this known, or is there any experimental evidence for that?
• Have the authors thought about testing the FKE mutant in the experiemnts shown in Fig. 5?
• In general, I think the title might be a bit misleading because of the use of PI4Kiiib. I understand what the authors mean, but because they have not thoroughly tested PI4Kiiib recruitment in their experiments, I think they should focuse rather on the mechanism of recruitment of ACBD3 the authors have found.

Significance

I think, as said above, that this is potentially an important paper for the field of membrane trafficking and membrane biology. Most of the experiments are in general well performed and well controlled, and the paper is clearly written and follows a logical line.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

This manuscript describes molecular mechanisms by which ACBD3 is recruited to the Golgi complex. ACBD3 recruits PI4KIII which is required to generate PI4P, a phosphoinositide which is key for the recruitment of essential Golgi proteins and hence is key to Golgi identity. The authors have used a combination of mass spectrometry, high quality fluorescence imaging, transient CRISPR knockdowns, and biochemical approaches such as IPs to identify the key determinant for recruitment of ACBD3 to the Golgi complex. They map the interaction between ACBD3 and the Golgi as a unique region (UR) upstream of its GOLD domain, identifying, in particular, an MWT motif as key for this recruitment. Using mass spectrometry they identify several novel interactors of ACBD3 as well as some established binding partners. Knockdown of these interactors reveal a key role for the SNARE, SCFD1, where reduced levels lead to complete loss of ACBD3 localisation to the Golgi without apparent disruption of Golgi structure. They further validate this interaction and that of another SNARE (Sec22b), which is part of the same SNARE complex as SCFD1, mapping the interaction to the longin domain of Sec22b. Surprisingly however they demonstrate that the UR domain does not mediate the interaction between ACBD3 and these SNAREs suggesting an alternative mechanism of recruitment. Previously identified ACBD3 interactors, Golgi proteins giantin and golgin-45 were also identified in the mass spectrometry screen and the authors demonstrate that these two proteins can recruit ACBD2 to the Golgi and this is dependent on the MWT motif identified in the UR domain. By knocking down SCFD1, they show reduced recruitment of ACBD3 leading them to propose a model of sequential recruitment of ACBD3 by SCFD1 followed by interactions with the golgins.

Major points:

This study is a well-executed and rigorous study of the molecular requirements for the recruitment of ACBD3 to the Golgi. The experimental approaches are state-of-the-art and the data are clean and convincing. The only caveat, raised by the authors themselves, is their interpretation that there are two sequential steps for Golgi recruitment of ACBD3. While they show that loss of SCFD1 reduces the interaction of ACBD3 with giantin and golgin 45, their model depends on doing the reverse experiment, i.e. assessing the effects of knocking down either giantin or golgin-45. This is especially relevant given the demonstration that golgin-45 is sufficient to recruit ACBD3 to mitochondria. It may well be that recruitment involves a tripartite complex, which is not uncommon in vesicular transport mechanisms Giantin is not an essential protein do it should be feasible to perform this experiment. The authors are equipped in the quantitative fluorescence microscopy which would be required and which would help resolve whether sequential or redundant mechanisms are required for ACBD3 recruitment.

Significance

PI4P is a phosphoinositide that is important for the recruitment of Golgi proteins. As with most PIs it is likely to act by coincidence detection in that Golgi associated proteins will recognise PI4P as well as other factors on Golgi membranes. This results in different local membrane environments which will be specific for particular functions. PI4KIII is key for PI4P production although the absolute levels of PI4P are likely to be determined by a balance of lipid kinases and phosphatases. However, since ACBD3 is key for the recruitment of PI4KIII it is important to understand the molecular mechanisms by which it is recruited. The manuscript thus makes a significant contribution to understanding one of the underlying mechanisms for PI4KIII recruitment although, as indicated above, stops short of establishing a clear model for the roles SCDF1 and Sec22b versus golgin 45 and giantin. For the future it will be of interest to determine why either a sequential or a redundant mechanism is required for the recruitment of ACBD3 as a scaffold protein.

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 would like to thank all reviewers for taking the time to evaluate our manuscript fairly and critically. Many helpful suggestions and discussion points were raised. One important group of comments raised concerns whether our proposed timer and counter models were the appropriate conceptual framework to discuss nuclear multiplication in schizogony, whether they were mutually exclusive, and whether other alternatives should be considered. These comments were instrumental for us to uncover some inconsistencies in our previous modeling approach. In the new manuscript, we now define the counter and timer models much more rigorously in the context of Plasmodium cell division. Based on these refined models we now provide a new statistical analysis that goes beyond the previous analysis, significantly improving the statistical support for our conclusions. Details are given in the following individual replies.

Reviewer #1 (Evidence, reproducibility and clarity):

Summary

Malaria parasites replicating in human red blood cells show a striking diversity in the number of progeny per replication cycle. Variation in progeny number can be seen between different species of malaria parasites, between parasite isolates, even between different cells from the same isolate. To date, we have little understanding of what factors influence progeny number, or how mechanistically it is controlled. In this study, the authors try to define how the mechanism that determines progeny number works. They propose two mechanisms, a 'counter' where progeny number is determined by the measurement of some kind of parasite parameter, and a 'timer' where parasite lifecycle length would be proportional to progeny number. Using a combination of long-term live-cell microscopy and mathematical modelling, the authors find consistent support for a 'counter' mechanism. Support for this mechanism was found using both Plasmodium falciparum, the most prominent human malaria parasite, and P. knowlesi, a zoonotic malaria parasite. Of the parameters measured in this study, the only thing that seemed to predict progeny number was parasite size around the onset of mitosis. The authors also found that during their replication inside red blood cells, malaria parasites drastically increase their nuclear to cytoplasmic ratio, a cellular parameter remains consistent in the vast majority of cell-types studied to date.

Major Comments

It is stated a few times in this study that P. knowlesi has an ~24 hour lifecycle, and while this is the case for in vivo P. knowlesi, it was established in the study when P. knowlesi A1-H1 was adapted to human RBCs (Moon et al., 2013) that this significantly extended the lifecycle to ~27 hours, which should be made clear in the text. As much of this study revolves around lifecycle length and timing, the authors should consider some of their findings with the context that in vitro adaption can significantly alter lifecycle length.

The reviewer raises an important point that we didn’t discuss for P. knowlesi. We now mention this directly in the introduction chapter (line 67) and in the discussion (lines 470ff). We are aware that P. knowlesi takes about 27 hours in the lab, which was also communicated by the Moon lab. We now cite relevant studies again in this context. We further address the issue of modified cell cycle time in vitro in the discussion in the sense that absolute values must be taken with caution and the focus of this study is about the relative ratio and correlation between the different cell cycle metrics.

• The dichotomous distinction between 'timer' and 'counter' as mutually exclusive mechanisms seems to be a drastic oversimplification. Considering the drastic variation we see in merozoite number across species, between isolates, and between cells, it seems much more likely that there are factors controlled by both time-sensed and counter-sensed mechanisms that both influence progeny number.

The study of progeny regulation in malaria parasites is very much in the early stages. We can agree that our models are simplifications, as is the case with all models. Our choice of just the two models timer and counter was driven by the number of cellular parameters we measure, i.e., duration of division phase and progeny number. These data essentially allow us to test the two competing models we presented. As we quantify more and more cellular parameters, based on the quantitative live cell imaging protocols established here, we will be able to test more complex cell cycle models. With our current data, we believe more complex models are not warranted.

However, this valuable criticism, in conjunction with related remarks by other reviewers, made us reevaluate the constraints of our model more precisely. We noticed that the criteria used in the previous version in the manuscript contained unnecessary additional assumptions. Briefly, the previous counter model also required that final merozoite number was tightly controlled, while the previous timer model required the growth rate to be tightly controlled. These side assumptions were not made explicit in the manuscript and could bias the support towards one or the other model.

We now improved the modeling approach substantially by removing implicit side assumptions, and clearly defining timer and counter models in terms of their correlations. The refined formulation of the timer posits that between individual parasites the target duration and the nuclear multiplication rate vary in a statistically independent way; while in a counter, target number and nuclear multiplication rate are statistically independent. We now explain this extended analysis in more detail in the introduction (lines 86ff). We also now more clearly state the dichotomous nature of the model (line 488). A new results paragraph (lines 213ff) and an entirely new Fig. 2 (and Fig. S4) contains the model predictions and statistical comparison between the models.

This more rigorous treatment showed that including the variance of the multiplication rate was critical to allow a clean discrimination between the models. Also, with the sole exception of P.knowlesi H2B, where no model was clearly favored (Fig. 2G-H,K), the timer model was found to be inconsistent with the data, while the counter was clearly favored. Our new goodness-of-fit analysis also showed that although the counter is strongly simplified, it produced adequate fits, demonstrating that potential model refinements would need to be justified by new, more extensive data.

It is also important to consider that the degree of variation in merozoite number could rather be an expression of varying growth conditions and does not directly predict which of the proposed models are true. For instance, a counter where the target merozoite number varies strongly depending on growth conditions, would be consistent with all available data. It is an interesting question for future work whether a counter would indeed describe growth across different isolates.

The biological reality of growth regulation is certainly complex, and the counter model will likely need to be refined in the future, which we acknowledge in a corresponding statement in the discussion (lines 491ff). Nevertheless, we find it encouraging that a simple model can explain the vast majority of our data very well.

Additionally, the only parasite parameter measured in this study, size at time of first nuclear division, explained only a small proportion of the variance observed in merozoite number.

It is indeed the case that amongst the measured parasite parameters i.e. schizont stage duration, nuclear volume, and cell size we only found the latter to correlate with the final progeny number. We did not aim to imply that all variation in progeny number is explained by cell size. It is likely that a putative counter relies on a set of factors, which are somehow linked to cell size. In addition, intrinsic stochasticity in nuclear growth is likely to contribute to final merozoite number variability, which is included in our models via a variable growth rate. Defining the actual limiting factor or combination of factors will be an exciting challenge for the future studies building on this one.

• For modelling of a timer-based mechanism, the designation of t0 is subjective. The authors chose the time of first nuclear division as their t0. It is possible that a timer-based mechanism could not be supported based on this model the chosen t0 differs from when the "parasite's timer" starts. For example, t could also have been designated as the time from merozoite invasion (t0) to egress (tend). It would be unreasonable to suggest the authors repeat experiments with a longer time-frame to address this, but this possibility should be discussed as a limitation of the model. It may also be possible to develop a different model where t0 = merozoite invasion and tend = egress, and test this model against the data already collected in this study.

This is a valid point. We indeed, considered the time point of invasion as the other relevant time point in the IDC for a possible timer. Due to necessary compromises in imaging protocols between acquisition length, temporal, and spatial resolution we have not been able yet to combine full-length IDC measurements with quantification of progeny number. Given the choice, however, between time point of invasion and the onset of nuclear division as starting point for a potential timer we would still favor the latter: An argument can be made that a timer that regulates offspring number would be more accurate when activated at the moment of the relevant cellular events rather than “running” for a very prolonged growth phase before any “decision” concerning parasite replication. We are still convinced that the entry into the schizont stage, which we analyze here, marks an important cell cycle transition point that has been highlighted in many different studies. As suggested, we now discuss the limitations of our selection of t0 in the text (lines 146ff).

• The calculation of the multiplication rate is confusingly defined. In Figure 1 it is stated that it is "...based on t and n", which would imply that the multiplication rate is the number of merozoites formed per hour of schizogony, which would give an average value of ~2 for P. falciparum and ~1.5 for P. knowlesi. The averages rate values shown, however, are in the range of 0.15-3. The authors should clarify how these values were determined.

Thank you for pointing out the need for more clarity. Since the nuclear multiplication, similar to e.g. cell population growth, follows an exponential law, the multiplication rate used (lambda) is in fact a logarithmic growth rate. Therefore, it occurs in the exponent (not as a coefficient) in the exponential growth function ( ), which explains the range. We now mention this more explicitly in the results (lines 163ff).

• In Figure 2, the time from tend until egress is calculated, and this is interpreted as the time required for segmentation. In the Rudlaff et al., 2020 study cited in this paper, it is shown that segmentation starts before the final round of nuclear divisions are complete. Considering this, the time from tend until egress is not an appropriate proxy for segmentation time. The authors should consider rewording to something akin to "time from final nuclear division until egress" to more accurately reflect these data.

Thank you for indicating our imprecise use of the nomenclature. Indeed, some essential segmentation-associated structures such as rhoptries and subpellicular microtubules are clearly forming before the last division. We were referring to “segmentation” as the time window where actual ingression of the plasma membrane occurs between nuclei with the concurrent formation of more prominent IMC-associated sub-pellicular microtubules between nuclei (as in Fig. 1A last panel). We can, however, agree that consistently using the term “merozoite formation” is more adequate here. We have now corrected the terminology according to the suggestions of the reviewer (lines 271ff).

• There is a significant discrepancy between the data in Figure 5 and Supplementary Figure 8. In Supplementary Figure 8, the authors establish that culturing parasites in media diluted 0.5x has a marginal effect on parasite growth, with no discernible change in parasitaemia over 96 hours. By contrast, in Figure 5a the parasitaemia of parasites cultured in 0.5x diluted media is approximately 5-fold lower than those in 1x media. The authors should explain the significant discrepancy between these results.

The reviewer correctly points out a difference in parasitaemia between two parasite culture experiments, shown in Figs 5a (now 6A) and S8 (now S11), respectively. There were several differences in the experimental setup used in the two experiments that could explain this discrepancy. In Fig. 5a the parasites were synchronized to early ring stages while in Fig. S8 we used asynchronous cultures (maybe with a slight majority of late stages). One could speculate that by the time the synchronized ring stage culture reached egress the effect of nutrient depletion, which started at t = 0 h is more pronounced. This effect could have been exacerbated by the more frequent media change of 24 h in Fig. 5a vs 48h in Fig. S8. Lastly, the starting parasitemia was differently set being higher at around 0.5% in the Fig. 5a while only 0.2% in Fig. S8. Possibly a lack of nutrient is “felt less” by the culture at lower parasitemias. Generally, in Fig. S8 we were more focused on highlighting the difference between 1x/0.5x and the more diluted conditions on the long-term culture and to show that continuous culture is actually possible in 0.5x medium. We have now expanded the legends to highlight those differences more clearly.

• In Supplementary Figure 4, the mask on the cell at t0 shows two distinct objects, but it seems very unlikely that they are two distinct nuclei as they vary approximately 5-fold in diameter. The authors should provide more detail on how their masking was performed for their volumetric analysis. Specifically, whether size thresholds were also applied during object detection.

Thank you for requesting clarification here. Fig S4 (now S7) shows only one z-slice (not a projection) of the entire image stack, to illustrate how the thresholding approach was performed on every single image slice. The two objects in the shown cell are indeed two nuclei, but because they are not in the same z-plane appear to be of different size. In particular, only a slice of the upper part of the nucleus on the lower right is visible in the shown slice. Throughout the study, volume determination was realized by adding up the individual slices, as is explained in detail in the Materials and Methods sections. We have now added a more explanation in the figure legend to clarify the procedure.

Minor Comments

• Line 45-48 mentions that merozoite number influences growth rate and virulence, but the corresponding reference (Mancio-Silva et al., 2013) only discusses the relationship between merozoite number and growth rate, not virulence.

We thank the reviewer for requesting this distinction. Merozoite number and virulence have not been correlated in vivo so far. Certainly, because one can’t retrieve late-stage P. falciparum parasites from patients, but maybe partly because merozoite number has not gotten significant attention as a metric in the previous decades. Even if merozoite number is intuitively connected to growth rate which might causes higher parasitemia which is in turn linked to more severe disease outcome it is important to emphasize that those are certainly not equivalent. We have therefore removed the statement about virulence (line 48).

• Line 59 states that a 48 hour lifecycle is a baseline from which in vitro cultured parasites deviate. Clinical isolates also show variation in lifecycle length and so it is more accurate to just say that 48 hours is an average, rather than a baseline.

The word “baseline” has been changed to “average” (line 61).

• Line 63 cites a study for the lifecycle length of P. knowlesi (Lee et al., 2022), but there seems to be no mention of lifecycle length in this reference

This reference was meant to serve as an introductory review article to research in P. knowlesi. Actually, to the knowledge of the authors, there is no study presenting quantitative data showing that the in vitro cycle of P. knowlesi is actually around 27 h. Our lab experience is however coherent with a 27 h cycle, which was confirmed by personal communication by the Moon lab. We now also cite in the next sentence the inaugural P. knowlesi adaptation publication (Moon et al. 2013) showing some time course data indicating the duration of the IDC to be around ~27h (lines 67ff).

• If I am interpreting Figure 3B correctly, this is essentially a paired analysis where the same erythrocytes are measured twice, once at t0 and once at tend. If this is the case, this data may be better represented with lines that connect the t0 and tend values.

Yes, these are the same erythrocytes measured twice. We have modified Figure 3 (now Fig. 4) accordingly.

• Figure 3A seems to imply that to calculate diameter of the erythrocytes, three measurements were made and averaged for each cell. I think this is a nice way to get a more accurate erythrocyte diameter, but if this is the case, it should be specified in the figure legend or methods.

This is already described in the figure legend (line 305).

• In Figure 4I it is shown that in P. falciparum merozoite number doesn't correlate with nucleus size, but for P. knowlesi in Supplementary Figure 7c, a significant anticorrelation is observed. The authors should state this in the text and discuss this discrepancy.

Contrary to all other graphs, visual inspection of the distribution of data points in Fig. S10C shows that it contains two outlier data points at the bottom right. Those two specific points are also responsible for the significant anticorrelation. We did not filter or remove any quantification results but also didn’t have sufficient confidence in this data distribution (which is further based on the segmentation of the Histone2B not on an NLS mCherry signal) to make substantial claims about anticorrelation. Because we considered it informative we still decided to show it in the supplements. We now briefly mention the issues with the data set and its interpretation in the text (lines 350ff).

• The authors show that merozoite number roughly correlates with cell size at t0 but it would be interesting to see whether cell size at tend also corresponds with cell size at t0. This might help answer whether the cell is larger because it has more merozoites, or whether it has more merozoites because it is larger.

Plotting parasite cell volume at t0 against cell volume at tend (as well as between t-2 and tend) indeed shows a positive correlation (see below). While it is an interesting thought we concluded after some discussion that no convincing causal relationship between cell size and merozoite number can be inferred based on this analysis. Since we consider the possible statement that cells that are bigger in the beginning are also bigger in the end unavailing, we decided not to include the data.

• I don't feel that "nearly identical" is an appropriate summary of erythrocyte indices in Supplementary Figure 9, considering there is a statistically significant increase in mean cell volume. I think it is unlikely that this change is consequential, and performing these haematology analyses is a nice quality control step, but this change should be stated in the text.

In the modified text we now express the significant change in MCV in terms of percentage, which is around 1.2% (line 381).

• In Supplementary Figure 8, parasitaemia only increases ~2-fold compared to >5-fold the previous two cycles. It seems likely that at the final timepoint on this graph the parasites are starting to crash, and therefore it may be best to end the graph with the 96 hour timepoint.

The reviewer suggests that cultures at those parasitemias might not be in perfect health. Our Giemsa stains did not show signs of an unhealthy culture and kept growing. It was, however, important for us to show that cultures can be maintained in culture over a prolonged period of time in 0.5x medium, even when resulting in reduced growth, while this was not possible with lower dilutions. Therefore, we would like to keep the data point. We have added a cautionary comment in the legend.

• The error bars in Figure 5C aren't easily visible, moving them in front of the datapoints may help their visibility.

Error bars were moved in front of the data points.

• In Figure 6D & E, the y-axis labels should be changed to whole integers as all the values in the graph are whole numbers.

We have changed the y-axis labels accordingly.

• My interpretation of Figure 6 C-E, is that these are the same cells measured at three time points (t-2, t0 and tend). If this is the case, 6C is missing the cell that has a merozoite number of 8, which is presumably why the y-axes are not equalised for the three graphs.

It is correct that the same cells are displayed in all three plots, with the exceptions of three cells in 6C (for the timepoint t-2), which are missing for the following reasons: 1) it was not possible to determine the volume at this respective timepoint due to technical issues or 2) the cell was already just before t0 at the start of the movie so that t-2 had already passed. We now note this in the figure legend and have also equalized the y-axes (now Fig. 7C-E).

Reviewer #1 (Significance):

In the asexual blood-stage of their lifecycle, malaria parasites replicate through a process called schizogony. During schizogony an initially mononucleated parasite undergoes multiple asynchronous rounds of mitosis followed by nuclear division without cytokinesis, producing a variable number of daughter nuclei. Parasites then undergo a specialised cytokinesis, termed segmentation to where nuclei are packaged into merozoites that go on to invade new host cells. While nucleus, and therefore merozoite, number are known to be varied between cells, across isolates, and across species, little is known about the mechanisms regulating merozoite number. In this study, the authors use live-cell microscopy to understand how parasites determine their progeny number. They suggest that parasites regulate their progeny number using a 'counter' mechanism, which would respond to the size or concentration of a cellular parameter, as opposed to a 'timer' mechanism. Long-term live-cell microscopy experiments using malaria parasites are extremely technically challenging, and the authors should be commended for their efforts in this regard. While I agree that the data generated from these experiments are technically sound, I have some reservations expressed above about the interpretation of some of these results. I would strongly encourage the authors to consider rewording some of their interpretations taking into account some of the caveats listed above. I would also consider fitting/testing an additional mathematical model where the time-frame proposed for the 'timer' mechanism begins following merozoite invasion.

We thank the reviewer for the appreciation of our work and hope we have sufficiently reworked the manuscript based on the comments listed above. Furthermore, we think the improved model statement and analysis improves the clarity of our conclusions. Indeed, we would like to test additional models including the full IDC once, as mentioned above, we are technically able to generate these data.

This work is of specific interest to anybody who grows malaria parasites, as the dynamics of their growth is obviously important to understand. Further, this work is of interest more generally to cell biologists who study the regulation of progeny number or cell size. I have no experience with the application of mathematical modelling to understand biological systems, and so I cannot comment on the interest of this work to that field.

Reviewer #2 (Evidence, reproducibility and clarity):

This is a solid study that further characterises the dynamics of nuclear division in Plasmodium falciparum and P. knowlesi. Of two, among potentially several, models for how the number of daughter nuclei, and thus parasites - (called merozoites in this genus), are one that posits nuclei divide until a fixed timer ends, and one that posits that nuclei divide to reach a fixed number that is defined by a cellular counter. I find some practical difficulties in definitive measurement of either model, one issue with the former is that experimental definition of the start of the timer is problematic - we may define the starter's gun (eg by the first nuclear division) but it isn't necessary that the cell is using that same start time.

We are pleased that the Reviewer found our study ‘solid’. Concerning the timer model, we agree that the selection of the starting point is a critical aspect of this study, as also Reviewer 1 pointed out. We selected this particular “t0” because the entry into the mitotic phase marks an important cell cycle transition. Several studies have suggested a “schizogony entry checkpoint” might be active just before (Matthews et al, 2018; Voß et al, 2023; van Biljon et al, 2018; McLean & Jacobs-Lorena, 2020). Once cells are committed to the schizont stage they are less responsive to stimuli. Alternatively, the timepoint of erythrocyte invasion could be a legitimate starting point. Due to necessary compromises in our imaging protocol between acquisition length, temporal, and spatial resolution we have not been able yet to combine full-length IDC measurements with quantification of progeny number, and therefore we leave exploration of an earlier timer start for future work. Within the confines of the model comparison in the current study, we think the selected t0 is already highly informative. We now explain the selection and limitations more explicitly in the text (line 144ff).

Additionally, as the authors confirm here, being sure when that first nuclear division has occurred is particularly tricky with Plasmodium parasites, in part because the first few nuclei seem to clump together, preventing one from unambiguously calibrating the first division.

The Reviewer is concerned about difficulties with precise reporting of the time point of first nuclear division. We suspect there was a misunderstanding here. In the text (line 137) we had written the following:

“Although separating individual nuclei after the first two rounds of division was challenging due to their spatial proximity, the improvements in resolution and 3D image analysis allowed us to count the final number of nuclei routinely and reliably at the transition into the segmenter stage.”

To clarify, when analyzing 3D image stacks produced by the LSM900 Airyscan the first nuclear division can consistently and unambiguously be detected. In anaphase the nuclei are pushed apart quite substantially before getting a bit closer together afterwards (see e.g. Fig. 1B and C). Hence the precision of the detection is only limited by the 30 min interval of the time lapse. Later, at the four nuclei stage, crowding makes distinction more difficult. In the final segmenter stage, the reorganization and condensation of nuclei makes reliable counting possible again. We have now reformulated the quoted sentence for more clarity (lines 137ff).

Furthermore, getting decent replicate numbers is hard because of the difficulties of time lapse microscopy, and most Plasmodium studies (including this one) suffer from low enough numbers that it isn't always clear whether the numbers support one model over another.

The reviewer points out the difficulty of obtaining enough replicates in Plasmodium time-lapse studies. We agree that depending on technology, sufficient replicates can be challenging. In the present study we obtained Ns between 25 and 35 for all conditions in P. falciparum and P. knowlesi from three independent replicas. To gain confidence in the conclusions from a limited, but not austere, data, it is essential to 1) reduce model complexity to a minimum and 2) perform stringent statistical analysis including accounting for small-sample variation. Motivated by this concern of the Reviewer and a similar point raised by Reviewer 1, we have revisited our modeling approach in the revised manuscript. This led us to a corrected, more rigorous definition of what precisely we mean by ‘counter’ and ‘timer’ models: The timer posits that between individual parasites the target duration and the nuclear multiplication rate and vary in a statistically independent way, while in a counter target number and nuclear multiplication rate are statistically independent. With no further adjustable parameters, the two models are thus both mutually exclusive and minimal. Although biological reality is likely to be more complex, we feel that these minimal models are adequate for the amount and resolution of our current, state-of-the art data. The general result remained the same: The counter model is strongly preferred in almost all our experiments data (new Fig. 2), with the sole exception of P. knowlesi H2B, where indeed more data may be needed to come to a clear conclusion. Furthermore, we have taken care to scrutinize these conclusions accounting for goodness-of-fit for the respective sample size N. This analysis showed, surprisingly, that the counter model was sufficient to account for the data: the real dataset was as similar to the counter prediction as synthetic, counter-generated data. We hope that this improved statistical analysis can help the reader judge the robustness of our conclusions.

Nonetheless, several recent studies, particularly a study from the same institute (Klaus et al., 2022) employing timelapse imaging of nuclei, and timing the nuclear division of parasites, finds poor correlation between the duration of "schizogeny" (although perhaps using a different definition to the one used by the parasite) and the final number or merozoites. They therefore argue that there is poor evidence for a timer, and conclude by elimination that a counter must exist instead. A review by some of the authors of that study and some of this current study (Voß et al 2023), also concludes that the data from Klaus and colleagues "strongly support" a counter model. This current study also concludes that a counter model controls final nuclear/merozoite number in P. falciparum and P. knowlesi. This much at least is not particularly novel given the recent work on this topic, although the addition of the P. knowlesi data is interesting and consistent with the prior P. falciparum work.

Our present work, indeed, does confirm the previous report of a counter over a timer, through a more targeted approach. While Klaus et al. used timing data of first nuclear cycle vs. the full duration, we now provide, thanks to an improvement microscopy setup and protocol, simultaneous measurements of timing and final progeny number, i.e. counting of merozoites/nuclei. While the preference for a counter model is not fundamentally novel, the additional information that the counter model holds in different strains, conditions and species is, in our opinion, not trivial and points to some degree of evolutionary conservation. We also demonstrate here that the counter model is not only preferred over the timer, it also fits the data adequately, so that it can be considered ‘correct’ at this level of complexity. Another, possibly more important, value of this study lies in the quantitative and time-resolved assessment of multiple important parasite metrics such a cell volume and nuclear volume together with merozoite number at the single cell level. Although descriptive, this has not been achieved in Plasmodium until now.

As above, the authors concede that it is difficult to determine with strong confidence when the first nuclear division has occurred, so it may well be that there is substantial noisiness in the time that they define schizogeny to commence. If that were the case, this would contribute to the poor correlation observed between schizogeny duration and number of merozoites produced, so this could be an important confounding experimental factor. This deserves some more discussion by the authors.

Concerning the confidence with which we identify the first nuclear division we could hopefully clarify in the section above that our precision is only limited by the time resolution of the acquired time-lapse. Therefore, the uncertainty about the start time is not particularly high, and moreover, can expected to affect timer and counter (via the growth rate) to a similar degree. We see no unfair advantage for the counter for this reason.

Alternative methods to count absolute DNA content (rather than trying to count individual nuclei) might be useful ways of independently confirming this phenomenon. Alternative possibilities for what constitutes the "start" of a possible timer are also warranted - it could be for example, the first division of one of the other organelles.

This is an interesting suggestion. Next generation fluorogenic DNA dyes have been used by us and the Ganter group (Simon et al. 2021, Klaus et al. 2022, Wenz et l. 2023) to assess DNA content of single cells over time. Our experience shows that there are some caveats to using these Hoechst based dyes, some of which we discussed in the aforementioned publications. While they allow some reasonable absolute quantification of DNA content for the very first S-Phase (and subsequent nuclear division), in later stages only relative quantification can be achieved. One underlying reason is the apparent increase of dye permeability, and therefore higher intensity, at late schizont stages. This issue is exacerbated by the asynchronous DNA replication of multiple nuclei. Further, nuclear division itself can be delayed or even inhibited when increasing the concentration of the dye, which suggest an impact on cell physiology (well documented for Hoechst based dyes in other organisms). When reaching the segmenter stage, the resulting variance in fluorescent intensity would make it challenging to assign a reliable number of nuclei required for analysis, a problem that does not occur when counting individual nuclei. Taken together, unfortunately, all these confounding factors make DNA content analysis in live single cells for the entire schizont stage unachievable at this point.

These and previous authors in any case conclude that a counter model must exist through exclusion of a timer model. I am less convinced that the evidence discounting the timer is conclusive, and that a straight counter model is the only alternative. Indeed I am unconvinced by the suitability of this strictly dichotomous two-model system to categorise the division of unicellular eukaryotes, and these theories are not universally held to be sufficient to describe division.

We thank the Reviewer for this insightful comment. As already detailed above, we have clarified and corrected our model definitions in the revised manuscript. Further, we want to make the important distinction between organisms, including unicellular ones that undergo binary fission and the ones like Plasmodium that use schizogony. Our model, although inspired by model organisms, is tailored to a multinucleated division mechanism, and clearly defined within those boundaries. The timer and counter models we consider are defined by their correlation structures. They are at two extremes of a continuum of models which could be characterized, for instance, by the ratio of correlations (growth rate - nuclear number) vs. (growth rate – duration) as an additional parameter. As the reviewer points out, excluding the timer model is not equivalent to proving the counter model, and indeed a partially correlated model, or a more complex model entirely, could yield a better fit. However, within the realm of models without additional parameters, and which are testable with the available data, only timer and counter remain, as different timer start points are not experimentally accessible. Importantly and somewhat surprisingly, the counter model also gave a fit that is as good as can be reasonably expected for the experimental sample size (new Fig. 2). So, we maintain that within the current experimental constraints, the counter model is the only viable option for almost all our tested conditions. The observation that in H2B-GFP expressing P. knowlesi parasites no clear distinction can be made between the models, indeed, suggest that the reality of multiplication rate regulation is more complex and may be limited by different constraints in different growth regimes. We now state these limitations and the room for further model adjustments with more data in the Discussion section.

Nonetheless, if a counter exists, what is being counted that determines the final number? The authors consider that this might be a physical object or resource inside the parasite, or an extrinsic/extracellular resource. They investigate this by comparing the final cell number to a number of factors. First, the authors investigate the size of the RBC (by musing the diameter as an indicator)- little information is given about the source of the blood used, but it appears to be from a single donor of unknown age, who has approximately typical variance in RBC diameter (at least, after manipulation and storage). The authors observe little correlation between these variables.

We share the curiosity of the reviewer about what might be “counted” by the parasite. This shall be the subject of future studies, and our present study provides the necessary basis for asking this question and defines a framework to investigate it. Concerning the size of the host cell, the blood used was from a different donor for each of the replicas, which we now specify in the figure legend (line 302). No significant difference between the RBC diameters between the donors was observed. A correlation between RBC diameter and progeny number was indeed not observed.

Second the authors measure parasite size at the onset of schizogeny, and find that bigger parasites result in more daughter merozoites early in schizogeny (perhaps not surprising, given the earlier mentioned technical problems with measuring the first few steps of schizogeny), but that this different initial cell size doesn't result in a different final merozoite number, or as they describe it "not quite significant anymore". Previous p values were taken as cause for rejecting the timer hypothesis and the timer model. In this case the authors instead interpret the data as suggesting "that the setting of the counter might correlate with parasite cell size". This is inconsistent statistical and analytical handling, and highlights the earlier potential pitfall of rejecting timer-based models based on not gathering data that statistically show a correlation. This needs reworking to highlight that these data are inherently noisy, difficult to measure accurately, and aren't necessarily going strongly reveal a trend even where one biologically exists, and that this ought not be used as grounds for confident rejection of a model.

The Reviewer raises concerns about the consistency of the statistical interpretation of our data. We care deeply about the well-foundedness of our conclusions and hope to eliminate these concerns in the following. First, we hope that the issue about the “technical problems” in measuring the first division has been solved in our response to previous comments. Next, to clarify an apparent misunderstanding: As stated in the text (lines 329ff) and shown in now Fig. 5D-E, cell size at onset of nuclear division or 2 hours prior does significantly correlate with final merozoite number. The lack of significant p-value (0.08) only pertains to the correlation of cell size at the end of the schizont stage (tend) with merozoite number (now Fig. 5F). We have removed the unfortunate wording “not quite significant anymore” in that context. Finally, regarding potential mechanisms, a potential counter must be set before the first nuclear division is completed because only that way it can be set independent of the speed of nuclear multiplication. This observation gives the statistically significant correlation of volume at the onset of division and progeny number its relevance. We have reformulated the marked sentence for more clarity (lines 331ff). Furthermore, we point out that our rejection of the timer is now based on a revisited statistical analysis (Fig. 2), which is no longer based on a simple correlation between final number and duration, as detailed above.

Finally, the authors grow the parasites in dilute media, and find that they produce fewer daughter parasites. This is anecdotally unsurprising, as most Plasmodium laboratories are aware that sub-optimal growth conditions result in less healthy schizonts with fewer viable merozoites (and lower magnitudes of single-cycle expansion), but is nonetheless an important result that highlights explicitly how much this occurs in the specific conditions of dilute media. Given the lack of investigation of exactly which nutrient, carbon source, or combination thereof leads to the reduced merozoite number, it is unclear if or how much this is relevant to the scenario of a natural infection and realistic levels of that nutrient in a human or primate parasite environment.

As rightfully pointed out by the reviewer suboptimal growth conditions affecting parasite growth and multiplication rate have been shown in many instances. The number of studies that actually quantify a reduction in merozoite number under different growth conditions is certainly much lower (Brancucci et al. 2017 (lipids), Mancio-Silva et al. 2017 (calorie-restriction in mice), Tinto-Font et al. 2022 (temperature) come to mind). What our study adds to this body of literature is to which extent duration of the schizont stage and cell volume are affected in relation to progeny number at the single cell level. Importantly, we wanted to test whether the counter model still holds under these more adverse conditions, which we found to be the case. Along the lines of the work on calorie restriction and the likely implication of isoleucine in the process investigated in the laboratory of Maria Mota, it will be exciting to identify a “limiting factor” in future studies. Indeed, any study done in complete RPMI culture medium can be questioned regarding its physiological relevance and we added a sentence addressing this aspect in the discussion (lines 514ff). Yet, our medium dilution experiments suggest that at least to some degree an extracellular resource is implicated, which makes sense from a biological function point-of-view.

Minor issues

The manuscript confuses the terms "less" and "fewer". Fewer should be used for countable nouns (fewer daughter cells, fewer nuclei, fewer merozoites), less for uncountable nouns (e.g. less speed, less volume).

Thank you for pointing this out. The words have been replaced accordingly.

I didn't understand lines 93-95; "This excluded a timer and thereby confirmed a counter as the mechanism regulating termination of nuclear multiplication (Klaus et al., 2022). A direct correlation between duration of schizont stage and merozoite number is, however, still missing." If I understand the first sentence concludes that there ought not be a direct correlation between schizont duration and merozoite number, but the second sentence, says that that correlation is "however" missing. Isn't this expected? Perhaps reword for clarity?

Thank you for requesting clarification here. The exclusion of the timer by Klaus et al. 2022 was based on the correlation between duration of the first nuclear division cycle and the total duration of all nuclear replication phases. At no point did Klaus et al. count merozoites in live single cells, which was mainly due to lower spatial resolution of their images (M. Ganter, personal communication). Therefore, they could not directly assess the relation between progeny number and schizont stage duration, which we now report for the first time. The sentence was supposed to convey that this type of data was missing and was now reformulated for more clarity (line 114).

Lines 104

"We further uncover that throughout schizogony P. falciparum infringes on the otherwise ubiquitously constant N/C-ratio (Cantwell and Nurse, 2019)" This seems obvious to me, and not something uncovered by this study. In most of the numerous apicomplexans that divide by endoschizogeny, the cells achieve a near final size considerably before the final rounds of nuclear division so the N/C ratio must not remain constant - this is a direct corollary of many previous descriptions and not a novel finding of this study, and this claim here should be made more modest.

We understand the point raised by the reviewer but still think that our claim is justified due to several aspects. There are examples of eukaryotic cells that undergo multinucleated stages during division were the N/C-ratio is constant (Dundon et al. 2016, Cantwell and Nurse, 2019), while we are not aware of any counter-example in the literature. Studies have also shown that e.g. certain mutant yeast that fail to undergo cytokinesis will increase their volume by factor of up to 16 alongside the still replicating and growing nucleus maintain the N/C-ratio (Neumann et al. 2007, Jorgensen et al. 2007). This demonstrates the tremendous plasticity that cells can reveal with respect to nucleus and cell size regulation. Until the contrary was shown, it was conceivable that nuclear compaction, which does occur (Fig. 5H), compensates for the increase in nuclear number while the cell volume is only increasing slightly. Importantly, we are not aware of any literature where nuclear volume has been quantified for blood stage Plasmodium. Cell volume quantifications remain limited to modelling and the study by Waldecker et al., which provides a few datapoints throughout the IDC. Whether this finding is expected or not, formally speaking, our claim is justified, but for more clarity we replace “uncover” with “demonstrate”. We also introduce the N/C-ratio as cellular parameter in P. falciparum pointing out another divergent aspect of its biology and might in the future understand the functional implication of this usually constant ratio, which is still unclear.

Dundon SE, Chang SS, Kumar A, Occhipinti P, Shroff H, Roper M, Gladfelter AS. Clustered nuclei maintain autonomy and nucleocytoplasmic ratio control in a syncytium. Mol Biol Cell. 2016 Jul 1;27(13):2000-7.

Neumann FR, and Nurse P. Nuclear size control in fission yeast. J. Cell Biol. 2007; 179: 593–600. pmid:17998401

Jorgensen P, Edgington NP, Schneider BL, Rupeš I, Tyers M & Futcher B Molecular Biology of the Cell 18 (2007) The size of the nucleus increases as yeast cells grow.

Helena Cantwell, Paul Nurse; A homeostatic mechanism rapidly corrects aberrant nucleocytoplasmic ratios maintaining nuclear size in fission yeast. J Cell Sci; 132 (22)

I lack specialist statistical knowledge to comment on the statistical analyses performed on the correlation data, and in particular, whether the high p values for t-Tests for correlation are sufficient to support the argument that there is not a correlation, and whether these observations are sufficiently powered to robustly test that hypothesis.

We are confident that our reworked model analysis, as explained above, now sufficiently supports our hypotheses.

Reviewer #2 (Significance):

The manuscript purports to find a counting mechanism that determines parasite merozoite numbers, and that this coutner is set by an externally provided and diffusible resource. Many nutrients are in excess in normal culture media, but not all. If that counted nutrient(s) were normally in excess in the bloodstream, it could hardly be said to be the factor that is counted and that therefore defines merozoite number. Conversely, if the amount of that nutrient were increased in normal media, would parasites make even more merozoites? Further, if the "counted" item is a freely diffusible compound in the media, it should be equally accessible to each parasite in a culture condition, and isn't a reasonable explanation for the variable merozoite numbers in the normal media conditions. To me, it is unsurprising that parasites that are healthy and well fed are able to produce more merozoites, but I don't see this as being the same as support for a counter model where the parasite senses and counts a set number of merozoites to produce in response to a specific external counter. I think the shoehorning of this phenomenon into a paradigm used to describe some other eukaryotes may not be appropriate, and that the rejection of one overly simplistic timer model should not automatically lead to us dichotomously accepting a simple counter method as the alternative. The authors need to do more to either identify a countable input whose gradual increase leads to a predictable and gradual increase in merozoite number, to show that they do use a counter, or provide substantially more caveats to their argument that the parasites are using a counter based on an externally provided resource to determine merozoite number.

The reviewer comments on the feasibility of a counter mechanism based on an externally provided and diffusible resource. In fact this is a limited view of how a counter may arise and not the one we subscribe to. Rather, while a resource may be diffusible in the medium, it would need to be consumed during schizogony, and insufficiently replenished, in order to enable counting by dilution in the host cell. Furthermore, the reviewer has doubts that the fact that “healthy and well fed […] produce more merozoites” implies “support for a counter model”. We fully agree, and we argue in the manuscript that it is the correlations between schizogony durations and merozoite counts that support a counter model.

As we have argued above, the two alternative models we consider are inspired by paradigm from other eukaryotes, but their definitions in the present context are simple enough for them to be considered natural minimal models of schizogony. As the simplest imaginable phenomenological models of multiplication control, we find it natural to compare them, and we hope our new introductory section introduces them appropriately now. Naturally, we hope to expand on this simple model in the future and identify more precisely the limiting resources and describe a more direct response.

Audience - relatively specialised - likely interested audience would combine apicomplexan cell biologists, as well as theorists of cell division mechanism

Advance - limited - confirms phenomenon also described by other researchers in their institute, and extends to another related organism.

We would like to add that the present data are the first quantitative joint measurements of schizogony dynamics and outcome in P.falciparum and knowlesi. They allowed for the first time a direct correlation of duration and merozoite number, thereby accessing the question of growth control head on. Further they provide a quantitative reference of several key cellular parameters for anybody studying asexual blood stage parasites.

Reviewer #3 (Evidence, reproducibility and clarity):

Summary:

Stürmer and colleagues used super-resolution time-lapse microscopy to probe the mechanism regulating the number of merozoites produced by a single cell in Plasmodium falciparum and P. knowlesi. The authors conclude the followings-

1. P. knowlesi has similar duration of schizont stage to P. falciparum, although having a 24 h intraerythrocytic developmental cycle (IDC) to 48 h of P. falciparum.
2. Nuclear multiplication dynamics suggests a counter mechanism of division- which is further suggested by a significant relation of merozoite numbers with schizont size at the onset of division.
3. Nutritional deprivation caused increase in nuclear volume and decrease in merozoite number. For the most part, the experiments that are presented in this manuscript support the conclusion of the authors. The data are presented in a concise and clear manner. However, some clarification and a couple of experiment (listed below) would improve this manuscript.

Major comments:

1. The authors generated at least 3 transgenic lines for this study, But the did not present any genetic validation of the lines in the manuscript. For completeness, I recommend to provide genetic validation (either pcr genotyping or whole genome sequencing) of the lines that were generated and used in this study in the supplement.

Our study exclusively used episomal expression of the respective fluorescent reporter (H2B-GFP, NLS-mCherry, and cytoplasmic GFP). As is customary in the field resistance to selection drugs and distinct fluorescent signals are assumed to sufficiently validate the presence of the plasmids. We now added the schematic maps of the plasmids in a new Fig. S1 to make our approach more visually clear.

1. In the H2B-GFP lines, the authors episomally GFP-tagged histone 2B to label the nuclear chromatin for both P. falciparum and P. knowlesi. This provides a very useful parasite line which enables the live time-lapse microscopy. Using these parasite lines, the authors first show that despite having a 24 h IDC in P. knowlesi vs 48 h in P. falciparum, both these parasites have a similar duration of the schizont stage (8.s vs 9.4 h). My concern here is whether this GFP-tagging is influencing the growth dynamics as in slowing down the P. knowlesi parasites. However, if that was the case authors should have seen that for P. falciparum too. Also, for the P. falciparum parasites that episomally express cytosolic GFP and Nuclear mCherry have a higher number of merozoites compared to the H2B-GFP P. falciparum and the authors speculate this is probably because of not tagging Histone 2B. Given this, it is important to show that none of the H2B-GFP parasites show any significant fitness cost due to GFP tagging of histone. I recommend a simple experiment to compare the multiplication rate of H2B-GFP lines to the parental lines in identical growth conditions. This suggested experiment was described in PMID: 35164549 to determine fitness cost of knockout lines. This experiment is vital for validation of the H2B-GFP lines and subsequent interpretation of the data that were presented in this manuscript.

We thank the reviewer for this excellent suggestion. To validate our lines further we now have carried out multiplication rate measurements similar to the one described in the designated publication for all the used lines alongside their parental strains (Fig. S2). We found no significant differences in between the wild type and the episomally expressing parasite lines (lines 131ff), which gives us confidence that episomal expression of tagged proteins do not significantly alter growth dynamics in these cases.

1. The authors used the microtubule live cell dye SPY555-Tubulin in P. falciparum to validate the findings presented in 1D and 1E. They did not do that for P. knowlesi. If there is no unsurmountable technical difficulty, I suggest doing the same with P. knowlesi. This will also address the concern that I have pointed out in #1.

Thank you for this suggestion. We have now generated the requested data with P. knowlesi, added it to what is now Supplemental Figure 3 and included it in our new analysis (Fig. 2I-J). The numerical values align well with the observations made when measuring schizont stage dynamics with the H2B-GFP expressing P. knowlesi line (line 158). A notable difference is that the Tubulin data strongly support the (refined) counter model, while the H2B data alone allow no distinction.

1. The data in Figure 3 shows that merozoite number does not depend on host cell diameter. My question here is, were these data collected using different donor blood? Or were this measured from different biological replicate? These are not clear from the writing. I am not sure about whether blood from various donor would have on the data, however, different preparation of the cells across various biological replicate will have some effect on host cell diameter hence on data. State if these were collected from independent biological replicates and about the donor blood.

The data results where indeed collected from three independent biological replicates using different donor blood batches. This is now stated in the figure legend. The batches displayed no difference in RBC diameter.

1. It is interesting to see that nutrient-limited conditions increase average nuclear volume but less merozoite numbers. In this experiment, as I understand, complete media was diluted 0.5x, which basically diluted every component of the media by half. From this experiment I can see nutritional deprivation as a whole having an effect and supports the counter mechanism, it would be intriguing to see if there is any effect of a particular nutrient have any effect on progeny division. For example, parasites can be grown in amino acid deprived media (except isoleucine) which makes the parasites fully dependent on host cell amino acids. This sort of specific nutrient deprivation will probably allow the authors to probe for specific nutrients that plays role as counter mechanism factor.

This is indeed a very exciting direction we would like to investigate in more detail in follow-up studies. Our aim for this study was to confirm that nutrient deprivation actually affects “counting” and to provide a workflow to investigate individual nutrients. In the meantime the Mota group, in a study we now cite in the discussion (lines 507ff), actually reported that isoleucine (and possibly methionine) levels are linked to progeny number. A follow-up on this topic using our strains and methodology is certainly worthwhile but requires more detailed analysis in the future.

Minor comments:

1. P. knowlesi is sometimes just written as knowlesi. Please, write P. Knowlesi.

Has been corrected.

1. Supplemental figure 1D, missing x-axis label.

We added the x-axis label.

1. In line 105, define N/C.

Done.

1. In line 205, I assume the authors mean episomally, not episomally.

Thank you for pointing this out. We have replaced “ectopically” with “episomally” throughout the text.

1. In line 275, Duration of Schizont stage was slightly....

Has been corrected.

1. All 'ml' or 'µl' should be 'mL' or 'µL'.

Changes have been made.

1. Define iRPMI.

We added a definition (line 610).

1. In line 475, replace 'as' with 'and'.

Done.

Reviewer #3 (Significance):

The factors that regulate the number of progenies in malaria parasites remain unknown. While there are few previous studies attempting to answer the question, those studies were done on fixed stained cells. In this study, the authors used genetically modified fluorescent P. falciparum and P. knowlesi parasites that enable live microscopy. These parasites coupled with super-resolution time-lapse microscopy the authors attempt to investigate the mechanism(s) at play in regulating progeny division. This manuscript provides data to suggest that external resources might have some role in progeny division and supports the counter mechanism. More careful validation of the transgenic lines that were used to collect data presented needs to be more systematic and rigorous.

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:

Stürmer and colleagues used super-resolution time-lapse microscopy to probe the mechanism regulating the number of merozoites produced by a single cell in Plasmodium falciparum and P. knowlesi. The authors conclude the followings:<br /> - a. P. knowlesi has similar duration of schizont stage to P. falciparum, although having a 24 h intraerythrocytic developmental cycle (IDC) to 48 h of P. falciparum.<br /> - b. Nuclear multiplication dynamics suggests a counter mechanism of division- which is further suggested by a significant relation of merozoite numbers with schizont size at the onset of division.<br /> - c. Nutritional deprivation caused increase in nuclear volume and decrease in merozoite number.<br /> For the most part, the experiments that are presented in this manuscript support the conclusion of the authors. The data are presented in a concise and clear manner. However, some clarification and a couple of experiment (listed below) would improve this manuscript.

Major comments:

1. The authors generated at least 3 transgenic lines for this study, But the did not present any genetic validation of the lines in the manuscript. For completeness, I recommend to provide genetic validation (either pcr genotyping or whole genome sequencing) of the lines that were generated and used in this study in the supplement.
2. In the H2B-GFP lines, the authors ectopically GFP-tagged histone 2B to label the nuclear chromatin for both P. falciparum and P. knowlesi. This provides a very useful parasite line which enables the live time-lapse microscopy. Using these parasite lines, the authors first show that despite having a 24 h IDC in P. knowlesi vs 48 h in P. falciparum, both these parasites have a similar duration of the schizont stage (8.s vs 9.4 h). My concern here is whether this GFP-tagging is influencing the growth dynamics as in slowing down the P. knowlesi parasites. However, if that was the case authors should have seen that for P. falciparum too. Also, for the P. falciparum parasites that episomally express cytosolic GFP and Nuclear mCherry have a higher number of merozoites compared to the H2B-GFP P. falciparum and the authors speculate this is probably because of not tagging Histone 2B. Given this, it is important to show that none of the H2B-GFP parasites show any significant fitness cost due to GFP tagging of histone. I recommend a simple experiment to compare the multiplication rate of H2B-GFP lines to the parental lines in identical growth conditions. This suggested experiment was described in PMID: 35164549 to determine fitness cost of knockout lines. This experiment is vital for validation of the H2B-GFP lines and subsequent interpretation of the data that were presented in this manuscript.
3. The authors used the microtubule live cell dye SPY555-Tubulin in P. falciparum to validate the findings presented in 1D and 1E. They did not do that for P. knowlesi. If there is no unsurmountable technical difficulty, I suggest doing the same with P. knowlesi. This will also address the concern that I have pointed out in #1.
4. The data in Figure 3 shows that merozoite number does not depend on host cell diameter. My question here is, were these data collected using different donor blood? Or were this measured from different biological replicate? These are not clear from the writing. I am not sure about whether blood from various donor would have on the data, however, different preparation of the cells across various biological replicate will have some effect on host cell diameter hence on data. State if these were collected from independent biological replicates and about the donor blood.
5. It is interesting to see that nutrient-limited conditions increase average nuclear volume but less merozoite numbers. In this experiment, as I understand, complete media was diluted 0.5x, which basically diluted every component of the media by half. From this experiment I can see nutritional deprivation as a whole having an effect and supports the counter mechanism, it would be intriguing to see if there is any effect of a particular nutrient have any effect on progeny division. For example, parasites can be grown in amino acid deprived media (except isoleucine) which makes the parasites fully dependent on host cell amino acids. This sort of specific nutrient deprivation will probably allow the authors to probe for specific nutrients that plays role as counter mechanism factor.

Minor comments:

1. P. knowlesi is sometimes just written as knowlesi. Please, write P. Knowlesi.
2. Supplemental figure 1D, missing x-axis label.
3. In line 105, define N/C.
4. In line 205, I assume the authors mean episomally, not ectopically.
5. In line 275, Duration of Schizont stage was slightly....
6. All 'ml' or 'µl' should be 'mL' or 'µL'.
7. Define iRPMI.
8. In line 475, replace 'as' with 'and'.

Significance

The factors that regulate the number of progenies in malaria parasites remain unknown. While there are few previous studies attempting to answer the question, those studies were done on fixed stained cells. In this study, the authors used genetically modified fluorescent P. falciparum and P. knowlesi parasites that enable live microscopy. These parasites coupled with super-resolution time-lapse microscopy the authors attempt to investigate the mechanism(s) at play in regulating progeny division. This manuscript provides data to suggest that external resources might have some role in progeny division and supports the counter mechanism. More careful validation of the transgenic lines that were used to collect data presented needs to be more systematic and rigorous.

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 is a solid study that further characterises the dynamics of nuclear division in Plasmodium falciparum and P. knowlesi. Of two, among potentially several, models for how the number of daughter nuclei, and thus parasites - (called merozoites in this genus), are one that posits nuclei divide until a fixed timer ends, and one that posits that nuclei divide to reach a fixed number that is defined by a cellular counter. I find some practical difficulties in definitive measurement of either model, one issue with the former is that experimental definition of the start of the timer is problematic - we may define the starter's gun (eg by the first nuclear division) but it isn't necessary that the cell is using that same start time. Additionally, as the authors confirm here, being sure when that first nuclear division has occurred is particularly tricky with Plasmodium parasites, in part because the first few nuclei seem to clump together, preventing one from unambiguously calibrating the first division. Furthermore, getting decent replicate numbers is hard because of the difficulties of time lapse microscopy, and most Plasmodium studies (including this one) suffer from low enough numbers that it isn't always clear whether the numbers support one model over another.

Nonetheless, several recent studies, particularly a study from the same institute (Klaus et al., 2022) employing timelapse imaging of nuclei, and timing the nuclear division of parasites, finds poor correlation between the duration of "schizogeny" (although perhaps using a different definition to the one used by the parasite) and the final number or merozoites. They therefore argue that there is poor evidence for a timer, and conclude by elimination that a counter must exist instead. A review by some of the authors of that study and some of this current study (Voß et al 2023), also concludes that the data from Klaus and colleagues "strongly support" a counter model. This current study also concludes that a counter model controls final nuclear/merozoite number in P. falciparum and P. knowlesi. This much at least is not particularly novel given the recent work on this topic, although the addition of the P. knowlesi data is interesting and consistent with the prior P. falciparum work. As above, the authors concede that it is difficult to determine with strong confidence when the first nuclear division has occurred, so it may well be that there is substantial noisiness in the time that they define schizogeny to commence. If that were the case, this would contribute to the poor correlation observed between schizogeny duration and number of merozoites produced, so this could be an important confounding experimental factor. This deserves some more discussion by the authors. Alternative methods to count absolute DNA content (rather than trying to count individual nuclei) might be useful ways of independently confirming this phenomenon. Alternative possibilities for what constitutes the "start" of a possible timer are also warranted - it could be for example, the first division of one of the other organelles.

These and previous authors in any case conclude that a counter model must exist through exclusion of a timer model. I am less convinced that the evidence discounting the timer is conclusive, and that a straight counter model is the only alternative. Indeed I am unconvinced by the suitability of this strictly dichotomous two-model system to categorise the division of unicellular eukaryotes, and these theories are not universally held to be sufficient to describe division. Nonetheless, if a counter exists, what is being counted that determines the final number? The authors consider that this might be a physical object or resource inside the parasite, or an extrinsic/extracellular resource. They investigate this by comparing the final cell number to a number of factors. First, the authors investigate the size of the RBC (by musing the diameter as an indicator)- little information is given about the source of the blood used, but it appears to be from a single donor of unknown age, who has approximately typical variance in RBC diameter (at least, after manipulation and storage). The authors observe little correlation between these variables. Second the authors measure parasite size at the onset of schizogeny, and find that bigger parasites result in more daughter merozoites early in schizogeny (perhaps not surprising, given the earlier mentioned technical problems with measuring the first few steps of schizogeny), but that this different initial cell size doesn't result in a different final merozoite number, or as they describe it "not quite significant anymore". Previous p values were taken as cause for rejecting the timer hypothesis and the timer model. In this case the authors instead interpret the data as suggesting "that the setting of the counter might correlate with parasite cell size". This is inconsistent statistical and analytical handling, and highlights the earlier potential pitfall of rejecting timer-based models based on not gathering data that statistically show a correlation. This needs reworking to highlight that these data are inherently noisy, difficult to measure accurately, and aren't necessarily going strongly reveal a trend even where one biologically exists, and that this ought not be used as grounds for confident rejection of a model.

Finally, the authors grow the parasites in dilute media, and find that they produce fewer daughter parasites. This is anecdotally unsurprising, as most Plasmodium laboratories are aware that sub-optimal growth conditions result in less healthy schizonts with fewer viable merozoites (and lower magnitudes of single-cycle expansion), but is nonetheless an important result that highlights explicitly how much this occurs in the specific conditions of dilute media. Given the lack of investigation of exactly which nutrient, carbon source, or combination thereof leads to the reduced merozoite number, it is unclear if or how much this is relevant to the scenario of a natural infection and realistic levels of that nutrient in a human or primate parasite environment.

Minor issues

The manuscript confuses the terms "less" and "fewer". Fewer should be used for countable nouns (fewer daughter cells, fewer nuclei, fewer merozoites), less for uncountable nouns (e.g. less speed, less volume).

I didn't understand lines 93-95;<br /> "This excluded a timer and thereby confirmed a counter as the mechanism regulating termination of nuclear multiplication (Klaus et al., 2022). A direct correlation between duration of schizont stage and merozoite number is, however, still missing."<br /> If I understand the first sentence concludes that there ought not be a direct correlation between schizont duration and merozoite number, but the second sentence, says that that correlation is "however" missing. Isn't this expected? Perhaps reword for clarity?

Lines 104<br /> "We further uncover that throughout schizogony P. falciparum infringes on the otherwise 105 ubiquitously constant N/C-ratio (Cantwell and Nurse, 2019)" This seems obvious to me, and not something uncovered by this study. In most of the numerous apicomplexans that divide by endoschizogeny, the cells achieve a near final size considerably before the final rounds of nuclear division so the N/C ratio must not remain constant - this is a direct corollary of many previous descriptions and not a novel finding of this study, and this claim here should be made more modest.

I lack specialist statistical knowledge to comment on the statistical analyses performed on the correlation data, and in particular, whether the high p values for t-Tests for correlation are sufficient to support the argument that there is not a correlation, and whether these observations are sufficiently powered to robustly test that hypothesis.

Significance

The manuscript purports to find a counting mechanism that determines parasite merozoite numbers, and that this coutner is set by an externally provided and diffusible resource. Many nutrients are in excess in normal culture media, but not all. If that counted nutrient(s) were normally in excess in the bloodstream, it could hardly be said to be the factor that is counted and that therefore defines merozoite number. Conversely, if the amount of that nutrient were increased in normal media, would parasites make even more merozoites? Further, if the "counted" item is a freely diffusible compound in the media, it should be equally accessible to each parasite in a culture condition, and isn't a reasonable explanation for the variable merozoite numbers in the normal media conditions. To me, it is unsurprising that parasites that are healthy and well fed are able to produce more merozoites, but I don't see this as being the same as support for a counter model where the parasite senses and counts a set number of merozoites to produce in response to a specific external counter. I think the shoehorning of this phenomenon into a paradigm used to describe some other eukaryotes may not be appropriate, and that the rejection of one overly simplistic timer model should not automatically lead to us dichotomously accepting a simple counter method as the alternative. The authors need to do more to either identify a countable input whose gradual increase leads to a predictable and gradual increase in merozoite number, to show that they do use a counter, or provide substantially more caveats to their argument that the parasites are using a counter based on an externally provided resource to determine merozoite number.

Audience - relatively specialised - likely interested audience would combine apicomplexan cell biologists, as well as theorists of cell division mechanism

Advance - limited - confirms phenomenon also described by other researchers in their institute, and extends to another related organism.

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

Malaria parasites replicating in human red blood cells show a striking diversity in the number of progeny per replication cycle. Variation in progeny number can be seen between different species of malaria parasites, between parasite isolates, even between different cells from the same isolate. To date, we have little understanding of what factors influence progeny number, or how mechanistically it is controlled. In this study, the authors try to define how the mechanism that determines progeny number works. They propose two mechanisms, a 'counter' where progeny number is determined by the measurement of some kind of parasite parameter, and a 'timer' where parasite lifecycle length would be proportional to progeny number. Using a combination of long-term live-cell microscopy and mathematical modelling, the authors find consistent support for a 'counter' mechanism. Support for this mechanism was found using both Plasmodium falciparum, the most prominent human malaria parasite, and P. knowlesi, a zoonotic malaria parasite. Of the parameters measured in this study, the only thing that seemed to predict progeny number was parasite size around the onset of mitosis. The authors also found that during their replication inside red blood cells, malaria parasites drastically increase their nuclear to cytoplasmic ratio, a cellular parameter remains consistent in the vast majority of cell-types studied to date.

Major Comments

• It is stated a few times in this study that P. knowlesi has an ~24 hour lifecycle, and while this is the case for in vivo P. knowlesi, it was established in the study when P. knowlesi A1-H1 was adapted to human RBCs (Moon et al., 2013) that this significantly extended the lifecycle to ~27 hours, which should be made clear in the text. As much of this study revolves around lifecycle length and timing, the authors should consider some of their findings with the context that in vitro adaption can significantly alter lifecycle length.
• The dichotomous distinction between 'timer' and 'counter' as mutually exclusive mechanisms seems to be a drastic oversimplification. Considering the drastic variation we see in merozoite number across species, between isolates, and between cells, it seems much more likely that there are factors controlled by both time-sensed and counter-sensed mechanisms that both influence progeny number. Additionally, the only parasite parameter measured in this study, size at time of first nuclear division, explained only a small proportion of the variance observed in merozoite number.
• For modelling of a timer-based mechanism, the designation of t0 is subjective. The authors chose the time of first nuclear division as their t0. It is possible that a timer-based mechanism could not be supported based on this model the chosen t0 differs from when the "parasite's timer" starts. For example, t could also have been designated as the time from merozoite invasion (t0) to egress (tend). It would be unreasonable to suggest the authors repeat experiments with a longer time-frame to address this, but this possibility should be discussed as a limitation of the model. It may also be possible to develop a different model where t0 = merozoite invasion and tend = egress, and test this model against the data already collected in this study.
• The calculation of the multiplication rate is confusingly defined. In Figure 1 it is stated that it is "...based on t and n", which would imply that the multiplication rate is the number of merozoites formed per hour of schizogony, which would give an average value of ~2 for P. falciparum and ~1.5 for P. knowlesi. The averages rate values shown, however, are in the range of 0.15-3. The authors should clarify how these values were determined.
• In Figure 2, the time from tend until egress is calculated, and this is interpreted as the time required for segmentation. In the Rudlaff et al., 2020 study cited in this paper, it is shown that segmentation starts before the final round of nuclear divisions are complete. Considering this, the time from tend until egress is not an appropriate proxy for segmentation time. The authors should consider rewording to something akin to "time from final nuclear division until egress" to more accurately reflect these data.
• There is a significant discrepancy between the data in Figure 5 and Supplementary Figure 8. In Supplementary Figure 8, the authors establish that culturing parasites in media diluted 0.5x has a marginal effect on parasite growth, with no discernible change in parasitaemia over 96 hours. By contrast, in Figure 5a the parasitaemia of parasites cultured in 0.5x diluted media is approximately 5-fold lower than those in 1x media. The authors should explain the significant discrepancy between these results.
• In Supplementary Figure 4, the mask on the cell at t0 shows two distinct objects, but it seems very unlikely that they are two distinct nuclei as they vary approximately 5-fold in diameter. The authors should provide more detail on how their masking was performed for their volumetric analysis. Specifically, whether size thresholds were also applied during object detection.

Minor Comments

• Line 45-48 mentions that merozoite number influences growth rate and virulence, but the corresponding reference (Mancio-Silva et al., 2013) only discusses the relationship between merozoite number and growth rate, not virulence.
• Line 59 states that a 48 hour lifecycle is a baseline from which in vitro cultured parasites deviate. Clinical isolates also show variation in lifecycle length and so it is more accurate to just say that 48 hours is an average, rather than a baseline.
• Line 63 cites a study for the lifecycle length of P. knowlesi (Lee et al., 2022), but there seems to be no mention of lifecycle length in this reference
• If I am interpreting Figure 3B correctly, this is essentially a paired analysis where the same erythrocytes are measured twice, once at t0 and once at tend. If this is the case, this data may be better represented with lines that connect the t0 and tend values.
• Figure 3A seems to imply that to calculate diameter of the erythrocytes, three measurements were made and averaged for each cell. I think this is a nice way to get a more accurate erythrocyte diameter, but if this is the case, it should be specified in the figure legend or methods.
• In Figure 4I it is shown that in P. falciparum merozoite number doesn't correlate with nucleus size, but for P. knowlesi in Supplementary Figure 7c, a significant anticorrelation is observed. The authors should state this in the text and discuss this discrepancy.
• The authors show that merozoite number roughly correlates with cell size at t0 but it would be interesting to see whether cell size at tend also corresponds with cell size at t0. This might help answer whether the cell is larger because it has more merozoites, or whether it has more merozoites because it is larger.
• I don't feel that "nearly identical" is an appropriate summary of erythrocyte indices in Supplementary Figure 9, considering there is a statistically significant increase in mean cell volume. I think it is unlikely that this change is consequential, and performing these haematology analyses is a nice quality control step, but this change should be stated in the text.
• In Supplementary Figure 8, parasitaemia only increases ~2-fold compared to >5-fold the previous two cycles. It seems likely that at the final timepoint on this graph the parasites are starting to crash, and therefore it may be best to end the graph with the 96 hour timepoint.
• The error bars in Figure 5C aren't easily visible, moving them in front of the datapoints may help their visibility.
• In Figure 6D & E, the y-axis labels should be changed to whole integers as all the values in the graph are whole numbers.
• My interpretation of Figure 6 C-E, is that these are the same cells measured at three time points (t-2, t0 and tend). If this is the case, 6C is missing the cell that has a merozoite number of 8, which is presumably why the y-axes are not equalised for the three graphs.

Significance

In the asexual blood-stage of their lifecycle, malaria parasites replicate through a process called schizogony. During schizogony an initially mononucleated parasite undergoes multiple asynchronous rounds of mitosis followed by nuclear division without cytokinesis, producing a variable number of daughter nuclei. Parasites then undergo a specialised cytokinesis, termed segmentation to where nuclei are packaged into merozoites that go on to invade new host cells. While nucleus, and therefore merozoite, number are known to be varied between cells, across isolates, and across species, little is known about the mechanisms regulating merozoite number. In this study, the authors use live-cell microscopy to understand how parasites determine their progeny number. They suggest that parasites regulate their progeny number using a 'counter' mechanism, which would respond to the size or concentration of a cellular parameter, as opposed to a 'timer' mechanism. Long-term live-cell microscopy experiments using malaria parasites are extremely technically challenging, and the authors should be commended for their efforts in this regard. While I agree that the data generated from these experiments are technically sound, I have some reservations expressed above about the interpretation of some of these results. I would strongly encourage the authors to consider rewording some of their interpretations taking into account some of the caveats listed above. I would also consider fitting/testing an additional mathematical model where the time-frame proposed for the 'timer' mechanism begins following merozoite invasion.

This work is of specific interest to anybody who grows malaria parasites, as the dynamics of their growth is obviously important to understand. Further, this work is of interest more generally to cell biologists who study the regulation of progeny number or cell size. I have no experience with the application of mathematical modelling to understand biological systems, and so I cannot comment on the interest of this work to that field.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity):

Summary:<br /> In this original article from Mutscher et al., the authors developed a compartmentalized dissociated mouse myelinating SC-DRG coculture system to investigate the distinct roles of Schwann cells in axon protection and degeneration after injury. The innovation of this approach relies on (i) the use of mouse SCs and mouse DRGS neurons instead of rat cells; (ii) the use of microfluidic chambers, seeded by axons and SCs in different compartment; (iii) the possibility to perform a traumatic injury in vitro. While this novel approach offers new ways to study peripheral nerve regeneration and SC-axon interaction, and technical the study is robust, the paper is currently limited by the exploration of their model.

Major points:<br /> Reviewer 1. It is unclear is this approach will ever lead to the identification of key mechanism or key candidates. This is a major miss in the current manuscript form. In short: the authors should demonstrate that their in vitro system can lead to significant leap in our understanding of peripheral nerve regeneration by identifying novel targets/pathways or mechanisms.

Author response: We agree with the reviewer that cell culture approaches have limitations however we would disagree that it is not a viable approach given that a number of seminal studies in the field have already helped identified key cellular and molecular steps using rat SC-DRG cocultures or using mouse DRGs and rat SCs in combination with in vivo study. We have added the following to the introduction to highlight this point in more detail:

Introduction.

Dissociated myelinating SC-DRG cocultures from rats were first developed by the Bunge laboratory in the 1980’s to investigate PNS myelination in a more dynamic way (Bunge et al, 1989; Eldridge et al, 1987)__. These cultures have been used to make seminal discoveries in uncovering the cellular and molecular mechanisms of SC myelination alongside in vivo investigation. These include how the inner SC membrane (mesaxon) advances to myelinate axons, and the role of b__-neuregulin-1 (__b__NRG1) and polarity proteins in SC myelination (Bunge et al, 1989; Shen et al, 2014; Chan et al, 2006; Taveggia et al, 2005)__. Similarly, SC-DRG cocultures have been useful in demonstrating how SCs proliferate after axon injury, transfer metabolites, such as pyruvate, to delay axon degeneration, how placental growth factor (Plgf) regulates axon fragmentation by SCs and how SC JUN promotes axon outgrowth after injury (Arthur-Farraj et al, 2011; Babetto et al, 2020; Vaquié et al, 2019; Salzer & Bunge, 1980)__. The use of a coculture system to study axon-SC interactions during axon degeneration and regeneration offers some advantages over in vivo approaches as both neurons and SCs can be genetically manipulated separately and live imaged with ease.

Discussion.

Most importantly SCs and DRG neurons from various transgenic mice can be used to perform in vitro analysis to complement findings from in vivo transgenic mouse studies.

Author response: Furthermore, as this is a methods paper, demonstrating novel molecular mechanisms is outside the scope of this article. However, we have already used this technique with a collaborator to study the role of cdk7 in myelination (see link to conference abstract below) and this manuscript is under preparation to be submitted soon. Additionally, we have ongoing projects within the lab using this technique to help characterise novel molecular targets in nerve injury. https://scholar.google.com/citations?view_op=view_citation&hl=en&user=uEtwAd8AAAAJ&sortby=pubdate&citation_for_view=uEtwAd8AAAAJ:_Qo2XoVZTnwC

Author response: Despite this, we have shown that the axo-protective effect of SCs is independent of myelination status, which is an advance on what is known in the literature.

Author response: We have realised through all of the reviewers’ comments that the title and the aims of the manuscript were confusing. We have made this clearer by removing the word novel in the title changing the title to the following:

A method for mouse myelinating Schwann cell-DRG neuron compartmentalised cocultures: myelination status does not influence the axo-protective effect of Schwann cells after injury.

Author response: We have also made it much clearer what the purpose of our study is and where and how it fits in with the previous literature by adding the following paragraph to the introduction.

Introduction.

Indeed there has only ever been one laboratory detailing convincing myelin formation in dissociated mouse myelinating SC-DRG neuron cocultures, however this was never published as a step by step detailed protocol (Stevens et al, 1998; Stevens & Fields, 2000)__. In the last twenty years, there have been no published studies demonstrating myelination in fully dissociated mouse SC-mouse DRG cocultures. This has largely prevented the use of cells, particularly SCs, from transgenic mice in cocultures and thus restricted the ability to study SC-axon interactions in a system that can be readily manipulated and live imaged and results directly applied back to in vivo findings in the same species.

Reviewer 1. The use of embryonic DRG neurons or SC isolated from P2 animals are arguably physiologically not the same cells that are affected by traumatic nerve injury, which happen most often than not in adult. This is a problem in the long-term reliance on this approach to study axotomy peripheral nerve regeneration.

Author response: We agree with the reviewer that one should always be cautious with the use of embryonic/neonatal cells to directly refer them to adult cellular mechanisms. We have added discussion of this point to the discussion:

Discussion.

One limitation of our coculture model and indeed all coculture and cell culture models that are used to investigate cellular and molecular mechanisms in nerve injury is that the cells are obtained from embryonic or neonatal animals. This is an important caveat when applying results from cell culture to adult in vivo nerve injury. However, while we would argue that cell culture approaches should always be used in combination with in vivo study it is important to remember that nerve injury is not restricted to adults and brachial plexus injury secondary to birth trauma is unfortunately a significant clinical problem (Pondaag et al, 2007)__. Furthermore, neonatal SCs replicate many of key cellular and molecular mechanisms seen in adult SCs after injury, including JUN upregulation, myelinophagy, promotion of axon growth and expression of key repair program transcripts (Arthur-Farraj et al, 2012; Gomez-Sanchez et al, 2015; Arthur-Farraj et al, 2017; Parkinson et al, 2008)__. A future development would be to try to adapt this protocol to make a coculture model with adult mouse or even human cells.

Author response: Additionally, we already know Schwann cells in P5 neonatal mice in vivo after nerve transection demyelinate in a similar way to Schwann cells in adult mice and that neonatal cells in vivo and in vitro require the transcription factor c-Jun to do so (Parkinson et al., 2008 JCB Fig.7).

Moderate points:<br /> Reviewer 1"there are no established protocols in the field describing the use of mouse SCs with mouse DRG neurons in dissociated myelinating cocultures". The use of mouse cells is laudable, but it is not necessarily a technical innovation, or at least the current manuscript does not explain why their approach particularly suitable to mouse Schwann cells.

Author response: We feel that a detailed working protocol for compartmentalised dissociated mouse myelinating cocultures showing convincing and extensive myelination has been missing from our field for a long time. We agree that it is an incremental technical advance, but it is an important one. We have modified the title as we explained above. We have explained this point more clearly in the introduction, results, and discussion with the following additions:

Introduction.

Indeed there has only ever been one laboratory detailing convincing myelin formation in dissociated mouse myelinating SC-DRG neuron cocultures, however this was never published as a step by step detailed protocol (Stevens et al, 1998; Stevens & Fields, 2000)__. In the last twenty years, there have been no published studies demonstrating myelination in fully dissociated mouse SC-mouse DRG cocultures. This has largely prevented the use of cells, particularly SCs, from transgenic mice in cocultures and thus restricted the ability to study SC-axon interactions in a system that can be readily manipulated and live imaged and results directly applied back to in vivo findings in the same species.

The consensus within the field is that inducing myelination in dissociated mouse SCs is challenging. Certainly, induction of myelin differentiation with cyclic adenosine monophosphate (cAMP) analogues or elevating agents, such as forskolin, is more difficult in mouse SC monocultures compared to rat SC cultures. This is because mouse SCs require additional exogenous b__-neuregulin-1 (__b__NRG1), plating on poly-L-lysine (PLL) instead of poly-D-lysine (PDL), and low concentration horse serum as opposed to foetal calf serum (Stevens et al, 1998; Arthur-Farraj et al, 2011; Päiväläinen et al, 2008)__.

Author response: We have now explained more clearly that without plating on Matrigel and the regular addition of Matrigel to the myelination medium that mouse cocultures do not myelinate with ascorbic acid or indeed addition of NRG1 nor forskolin. Please see NEW DATA in Supplemental figure 1. We have added the following paragraph to the results section.

Results

Importantly, we found that L-ascorbic acid was insufficient to induce substantial myelination in our cultures, unlike in rat SC-DRG cocultures, and in the one previously published dissociated mouse SC-DRG protocol (Stevens et al, 1998)__. In fact, plating cocultures on laminin, adding ascorbic acid (50 m__g ml−1), b__NRG1 (10 ng ml−1) and forskolin (10 m__M) induced very few myelin sheaths (Supp. Fig. 1). Only when cultures were plated on Matrigel__â and further Matrigel__â was added to the myelination medium for each medium change, were we able to visualise robust reproducible myelination in our cocultures (Supp. Fig. 1).

Author response: We have also added further discussion on how our protocol differs from the Stevens 1998 and other protocols in the discussion.

Discussion

Our protocol differs somewhat from the one used by Stevens et al., 1998 to induce myelination in dissociated mouse SC-DRG cocultures__, as they used ascorbic acid and 10% horse serum and presumably plated their cultures on laminin, though they do not explicitly detail this (Stevens et al, 1998)__. In our preliminary experiments we were unable to visualise much myelination with use of laminin, ascorbic acid or indeed if b__NRG1 and high concentration forskolin was added to the medium for up to four weeks. However, if we plated cocultures on Matrigel® and continuously added it to the myelination medium then we saw comparable levels of myelination in our mouse cocultures to that of rat cocultures (Eldridge et al, 1987)__. This approach of using Matrigel® to enhance myelination has previously been successfully employed in cultures of human iPSC sensory neurons with rat SCs and in non-dissociated mouse DRG explant cultures (Clark et al, 2017; Päiväläinen et al, 2008)__. Importantly, we used growth factor depleted Matrigel® as standard Matrigel® preparations contain substantial amounts of Transforming growth factor b (TGF__b__) which is a known inhibitor of myelination (Einheber et al, 1995)__. Additionally, the majority of rat and mouse coculture protocols plate cells on glass whereas we found cultures were healthier and myelinated better when cultured on plastic Alcar® coverslips.

Discussion

Furthermore, as this is a dissociated and compartmentalised purely mouse cell culture system, one can utilise the vast array of transgenic and knockout lines available to study neuron-SC interactions in more detail, without concern of contaminating endogenous SCs and other non-neuronal cells that remains a drawback of current mouse dissociated or non-dissociated DRG explant models.

Reviewer 1: The figures in the paper are largely descriptive. They are very little quantitative measurement. Thus, the readers will have a hard to determine, if they replicate the proposed approach, whether their efficient is on par with the current authors.

Author response: We have now added quantification of myelin segments per mm2, percentage of SCs that myelinate and quantification of the interperiodic distance of the myelin formed. This is all included in a new version of TABLE 1. We have discussed this data in the results section as follows.

Results

Quantifying the number of PRX positive myelin segments we found that there were 325.33 ± 12.3 sheaths per mm2, comparable to what has been originally described in rat SC/DRG cocultures and two fold more extensive myelination than recently described compartmentalised rat cocultures models (n=3; Table 1; Eldridge et al, 1987; Vaquié et al, 2019)__. Furthermore, 25.47 ± 1% of Schwann cells were myelinating in our cultures (n=3; Table 1). To confirm that cocultured myelinated SCs formed compact myelin we performed electron microscopy (EM), which revealed compact myelin formation with multiple myelin wraps and formation of readily visible major dense (MDL) and intraperiod lines (IPL; Fig. 2D). Additionally, we measured the periodicity, i.e., the distance between two adjacent major dense lines, to make sure myelin was compacted. Interperiodic distance was 12.16 ± 0.28 nm, in line with previous reports (n=3; Table 1; Boutary et al, 2021; Fernando et al, 2016; García-Mateo et al, 2018; Giese et al, 1992; Perrot et al, 2007)__.

Author response: Interestingly, after we performed this analysis, we realised we have double the level of myelination in our mouse cultures (325.33 ± 12.3 per mm2) than in the compartmentalised myelinating rat cocultures in Vaquie et al., 2019 (147 ± 27 internodes per mm2 (n = 3)).

Author response: We have also quantified the JUN upregulation after injury in both myelinating and aligned cocultures. See Fig. 3B-E).

Results

Additionally, we noted a strong upregulation of JUN protein in SCs 12 hours after axotomy (Fig. 3B and C). We also saw significant JUN upregulation 12 hours after axotomy in cocultures with aligned SCs (Fig. 3D and E).

Author response: The above quantification is in addition to the quantification of the rate of axon degeneration in the presence and absence of aligned and myelinating Schwann cells in Fig. 4B. We have also quantified the % of Schwann cells that contained axonal debris after injury – this data is now quoted in the text as we removed Fig. 4E.

We thank the reviewer for asking for additional quantification as this has improved the manuscript.

Minor points:<br /> Fig.4B and E should show individual data points.

Author response: We have added the individual data points to Fig.4B. We have removed Fig.4E and instead quoted the data in the results section as follows:

Results

When we quantified this phenomenon, we found that 97.84 ± 1.462% (n=2) of SCs in our cocultures contained mCherry-labelled axonal fragments.

Reviewer #1 (Significance):

In addition, to the demonstration of feasibility of this in vitro approach, the main finding by the authors is that SCs have a role in neuronal protection and support is key for peripheral nerve regeneration. Thus while in vitro approach does not add key information that do not already exists in the field, it somewhat confirms that the effect is SC autonomous. Overall the approach is interesting and has potential, but the study currently lack a demonstration of its usefulness to the community.

It would have been interesting to have the authors discuss the advantages of their approach in comparison to other innovative approaches to study SC-axon interactions that have been developed in the last decade (i.e., 3D environment, microfluidic approach, transwell systems). There is also a lack of citations about similar studies in the field.

Author response: We direct the reviewer to the following paragraphs in the introduction and the discussion, which we have now elaborated on further post peer review. We discuss all relevant cocultures studies in mouse as well as all the relevant microfluidic studies and 3D coculture studies as well as the one human nerve organoid study. We found two additional studies, Numata-Uematasu 2023 using DRG mouse explant cultures and Park et al., 2021 using motorneuron-SC cocultures which we have now added to the discussion. We also briefly discuss transwell studies to assess migration as the reviewer requested.

We also discuss in detail the two microfluidic coculture injury studies Babetto et al., 2020 and Vaqiue et al., 2019 extensively throughout the manuscript. We have added further discussion of the similarities and differences between theirs and our approach in the discussion.

Introduction.

Protocols exist where endogenous mouse SCs are used to myelinate dissociated or non-dissociated DRG explant cultures. (Shen et al, 2014; Harty et al, 2019; Sundaram et al, 2021; Stettner et al, 2013; Numata-Uematasu et al, 2023)__. Furthermore, another protocol seeded exogenous SCs onto non-dissociated DRG explant cultures (Päiväläinen et al, 2008)__. Other laboratories seed cultured rat SCs onto dissociated mouse DRG axons (Taveggia & Bolino, 2018)__. Use of dissociated or non-dissociated DRG explants cultures preclude many experimental uses, such as using SCs from different transgenic animals and separate transfection of SCs and neurons with viruses for live imaging or genetic manipulation, and easy use of microfluidic chambers to allow injury studies and separate drug treatments to neurons or SCs. The reason for this is that antimitotics cannot be used in dissociated or non-dissociated DRG explant cultures as this depletes SCs, and the culture quickly becomes contaminated with other non-neuronal cell types, such as satellite cells and fibroblasts migrating out of the DRG. Furthermore, use of exogenous SCs in a non-dissociated DRG explant culture risks, after a period of antimitotic exposure, which was developed by Päiväläinen et al., 2008 still risks potential contamination from endogenous SCs and satellite glia migrating out of the DRG explant over time. This occurs because antimitotic treatment is unlikely to fully penetrate the whole DRG without prior dissociation. Additionally, a compartmentalised culture system cannot be readily used with non-dissociated DRG explant cultures (Päiväläinen et al, 2008)__.

Discussion.

Furthermore, our protocol is complementary to the recently described 3D mouse myelinating SC-motor neuron coculture system using collagen hydrogels (Hyung et al, 2021; Park et al, 2021)__. It will be interesting in the future to up titrate the concentration of Matrigel®, which is similar to collagen hydrogels, in our cultures to see whether further increasing extracellular matrix viscosity and stiffness improves our myelination efficiency even further. While it is possible to study cell migration in microfluidic cell culture devices, transwell models offer significant advantages to study this cellular phenomenon (Negro et al, 2022)__. To date, there have been no published studies of successful myelination in human SC-neuron coculture systems. Despite this rat SCs have been shown to readily myelinate human-induced pluripotent stem cell (iPSC)-derived sensory neurons and an iPSC-derived peripheral nerve organoid system which does contain myelinating SCs has recently been described (Clark et al, 2017; Van Lent et al, 2022)__.

These findings confirm the observations of both Babetto et al., 2020 and Vaquié et al., 2019 who used rat SCs in similar microfluidic culture systems (Vaquié et al, 2019; Babetto et al, 2020)__. We have shown that the axo-protective observation seen by Babetto et al., 2020 does not rely on myelination status, which was an outstanding question from that study (Babetto et al, 2020)__. Furthermore, in an advance from previous studies, we have visualised the axo-protective and axon clearance phenomena in the same culture and shown that they are temporally separated, with axon fragmentation and debris clearance by SCs occurring at much later timepoints after axotomy. Several different culture and experimental conditions preclude direct comparison of our study with both those of Vaquié et al., 2019 and Babetto et al., 2020. These include the use of rat SCs in both prior studies and that Babetto et al., 2020 mixed rat SC with mouse DRG axons; length of time in culture, time points quantified after injury and distance from injury and site of analysis. Babetto et al., 2020 performed axotomy on relatively short term cocultures (6 days in vitro) whereas Vaiquié et al., 2019 cultured for at least 4 weeks and in our case 6 weeks prior to axotomy. Vaiquié et al., 2019 removed nerve growth factor (NGF) prior to laser axotomy and quantified proximally (though also imaged distally) whereas both our study and Babetto et al., 2020 kept NGF in the medium, performed axotomy with a scalpel and quantified more distally and in our case extremely distal, where only individual neurites and no axon bundles were visible. Vaquié et al., 2019 had SCs on both sides of the barrier in the microfluidic chambers whereas we seeded SCs only in the axonal/bottom compartment. Additionally, our cocultures had both forskolin and b__NRG1 added to help induce myelination, whereas these factors are not required in rat myelinating cocultures. Finally, it is important to permeabilise myelinated cultures with acetone after fixation__, as we did, to visualise the entire axon through heavily myelinated segments otherwise axon integrity cannot be reliably assessed in a quantitative manner (Vaquié et al, 2019; Babetto et al, 2020)__.

Because of the lack of key novel mechanisms, and lack of discussion on what this approach is superior to others in vitro approach limits the impact of the study and the excitement of the reader, even from the SC-axon community.

Author response: We have developed the first compartmentalised fully dissociated mouse myelinating coculture system in over twenty years. Thanks to the reviewer’s suggestions, we have now shown that myelination is comparable to the original rat cocultures from the Bunge lab, which is the gold standard in the field, and superior to recently described compartmentalised rat coculture system by Vaquie et al., 2019. We have provided a detailed step by step protocol to allow other researchers to use our technique.

Additionally, thanks to the reviewer, we have now described in detail exactly how our protocol differs from others and why we succeeded to get mouse SCs to myelinate so robustly in a fully dissociated coculture (see previous answers). This is an incremental but important advance given that studies currently use a coculture system using entirely cells from rat, or where rat Schwann cells are seeded on mouse axons, or dissociated or non-dissociated mouse explant cultures are used which abrogates using neurons and SCs from different transgenic mice.

As this is a methods paper, we did not intend to describe novel molecular mechanisms though our method is already being used for such purposes by ourselves and a collaborator as outlined above in prior answers. We did not make this clear and we hope the extensive revision of the manuscript now addresses this point. Despite this, we have shown that the axo-protective effect of SCs is independent of myelination status, which is an advance on what is known in the literature.

Finally, in addition to myelination, we have demonstrated that one can study all the key components of the SC and axonal response to injury in a quantifiable way in addition to demonstrating that these cocultures can be live imaged and used for drug studies. None of the prior mouse studies looked at injury responses of axons nor SCs. We believe this will be of use to the community.

Reviewer #2 (Evidence, reproducibility and clarity):

In the manuscript entitled "Distinct axo-protective and axo-destructive roles for Schwann cells after injury in a novel compartmentalised mouse myelinating coculture system", Arthur-Farraj and colleagues detail a method of dissociated coculture of mouse-DRG neurons and Schwann cells in microfluidic chambers. In this system, neurons and Schwann cells harvested from the same or from different animals are grown in different compartments that are connected by microgrooves, thereby allowing for spatial and diffusive separation. Neurons are shown to extent their axons across the microgroove barrier to the glial compartment where Schwann cells align with the axons and myelination can be induced. Detailed analysis of myelination and axon injury/degradation are presented as use cases, including the capability to genetically and pharmacologically manipulate neurons and Schwann cells independently, which also enabled fluorescent life cell imaging. The authors then examine the effect of immature/premyelinating and myelinating Schwann cells on the rate of axon degeneration. Upon axotomy Schwann cells significantly delayed degeneration, with no difference between non-myelinating and myelinating Schwann cells. Finally, live imaging during axon degeneration with fluorescent proteins separately expressed in neurons and Schwann cells demonstrated that Schwann cells ingest axonal fragments.

Major comment:<br /> In establishing a much needed in-vitro system for PNS myelination and injury research the paper represents a valuable contribution to the PNS community. However, I find the presentation of aspects concerning a protective/destructive role of Schwann cells somewhat inconclusive. That these roles exist has been known, as the authors discuss. Then what does this study contribute concerning the open question that was raised by the discrepancies between Babetto et al., 2020 and Vaquié et al., 2019, i.e. how Schwann cells contribute to axon survival/regeneration after injury? Essentially, the only significant conclusion in this regard is that myelinating and non-myelinating mouse Schwann cells do not differ in their capability to protect axons from degeneration. The manuscript, including the title, would benefit from focusing more on this aspect. In particular, the discussion of the factors that lead to the still remaining apparent discrepancies between Babetto et al., 2020 and Vaquié et al., 2019 and this study should be revised. The authors state that "The study by Vaiquié et al., 2019 quantified axon fragmentation proximally in the microgrooves at timepoints starting from 12 hours after axotomy." (Discussion). While this observation is accurate, Jacob and colleagues also show accelerated, obviously distal axon degeneration in the presence of Schwann cells (Figure 3C in Vaquié et al., 2019). It is therefore unlikely that the discrepancies stem from analysis of more proximal vs. more distal axons, or the timepoints of analyses. In my opinion, a further study (using the coculture system presented in this manuscript) that compares the role of Schwann cells from rats and mice, and that includes analysis of more proximal and distal axon degeneration as well analysis of axon regeneration is needed. In a rework of the manuscript, the authors may therefore elaborate more on the shortcomings of the present study, or alternatively soften the aims of the study in the first place.

Author response: We thank the reviewer for their comments, and we agree that the title and the aims of the manuscript were confusing. We didn’t make it explicit that this was a methods paper, and that we didn’t intend to show entirely novel findings but instead thoroughly characterise our mouse system so that it is comparable to what has been done for rat cocultures. We have now made this clearer by removing the word novel in the title and changing the title to the following:

A method for mouse myelinating Schwann cell-DRG neuron compartmentalised cocultures: myelination status does not influence the axo-protective effect of Schwann cells after injury.

Author response: We have decided to focus the manuscript more on the comparison of myelinating versus non-myelinating cocultures, given that we have shown that the axo-protective effect of SCs is independent of myelination status, which is an advance on what is known in the literature. We have added further characterisation of our aligned cocultures with p75NTR immuno and EM images (Fig.1D and E). We have added the following summary of how our study relates to findings of Babetto et al., 2020 and Vaquie et al., 2019 in the discussion, in line with the reviewer’s suggestions.

Discussion.

These findings confirm the observations of both Babetto et al., 2020 and Vaquié et al., 2019 who used rat SCs in similar microfluidic culture systems (Vaquié et al, 2019; Babetto et al, 2020)__. We have shown that the axo-protective observation seen by Babetto et al., 2020 does not rely on myelination status, which was an outstanding question from that study (Babetto et al, 2020)__. Furthermore, in an advance from previous studies, we have visualised the axo-protective and axon clearance phenomena in the same culture and shown that they are temporally separated, with axon fragmentation and debris clearance by SCs occurring at much later timepoints after axotomy.

Author response: We have now added more in-depth discussion of the similarities and differences between Babetto et al., 2020 and Vaquie et al., 2019 and our approach in the discussion.

Discussion.

Several different culture and experimental conditions preclude direct comparison of our study with both those of Vaquié et al., 2019 and Babetto et al., 2020. These include the use of rat SCs in both prior studies and that Babetto et al., 2020 mixed rat SC with mouse DRG axons; length of time in culture, time points quantified after injury and distance from injury and site of analysis. Babetto et al., 2020 performed axotomy on relatively short term cocultures (6 days in vitro) whereas Vaiquié et al., 2019 cultured for at least 4 weeks and in our case 6 weeks prior to axotomy. Vaiquié et al., 2019 removed nerve growth factor (NGF) prior to laser axotomy and quantified proximally (though also imaged distally) whereas both our study and Babetto et al., 2020 kept NGF in the medium, performed axotomy with a scalpel and quantified more distally and in our case extremely distal, where only individual neurites and no axon bundles were visible. Vaquié et al., 2019 had SCs on both sides of the barrier in the microfluidic chambers whereas we seeded SCs only in the axonal/bottom compartment. Additionally, our cocultures had both forskolin and b__NRG1 added to help induce myelination, whereas these factors are not required in rat myelinating cocultures. Finally, it is important to permeabilise myelinated cultures with acetone after fixation__, as we did, to visualise the entire axon through heavily myelinated segments otherwise axon integrity cannot be reliably assessed in a quantitative manner (Vaquié et al, 2019; Babetto et al, 2020)__.

Author response: We would like to add that we showed Claire Jacob, senior author of the Vaquie et al., 2019 study, our manuscript prior to peer review and she offered helpful comments, which we incorporated into the manuscript, which is why she is acknowledged. We have also discussed our findings with Elisabetta Babetto as well.

While it would be a great future study to compare both axon degeneration rates in rat and mouse cocultures this was not the original intention of our study. We believe we have included enough detail of our experimental procedures, including the distance from the barrier we imaged axon degeneration, a crucial bit of information missing from the other studies, should others want to perform a comparative study between rats and mice.

Methods

Five images at a distance of between 1.2-1.4mm from the microgroove barrier (the most distal part of the culture that could be imaged) were quantified per culture, taken in comparable locations in each culture.

Author response: I would add that one of our previous studies (Arthur-Farraj et al., 2012 Neuron, Fig5I) has already looked at axon outgrowth/regeneration in dissociated non-myelinating mouse SC-DRG co-cultures. We showed the presence of Schwann cells accelerates axon regeneration/outgrowth and this relies upon Schwann cell c-JUN.

We have now added quantification of the extent of myelination in our cocultures and it is comparable to the original Bunge lab rat cocultures and more extensive than the Vaquie et al., 2019 coculture.

Results

Quantifying the number of PRX positive myelin segments we found that there were 325.33 ± 12.3 sheaths per mm2, comparable to what has been originally described in rat SC/DRG cocultures and two fold more extensive myelination than recently described compartmentalised rat cocultures models (n=3; Table 1; Eldridge et al, 1987; Vaquié et al, 2019)__. Furthermore, 25.47 ± 1% of Schwann cells were myelinating in our cultures (n=3; Table 1).

Methods

To quantify the number of myelin segments per area, we counted the number of myelin segments for five areas per culture for three cultures and normalised this per mm2. To quantify the percentage of Schwann cells in myelinating cocultures that are actively myelinating, we quantified the number of myelin segments and the number of DAPI-positive nuclei for five areas per culture for three cultures.

Minor comments:<br /> - Figure 2D: From the electron micrograph there is no doubt that compact myelin is formed, however to me it seems the compaction is not complete. A rough estimation with the aid of the provided scale bar resulted in an interperiodic distance of about 17 nm, which contrasts with the remarkably well reproduced values reported in multiple reports using conventional specimen preparation (like in this study), of which I am citing just a few: about 13 nm in rat ex vivo nerve (Peterson and Pease, J. Ultrastruct Res 1972; Fledrich et al., Nat Commun 2018), 12 nm (Giese et al., Cell 1992), 12.2 nm (Perot et al., J Neurosci 2007), about 12 (Fernando et al., Nat Commun 2016), about 13 nm (García-Mateo et al., Glia 2017) or about 13 nm in mouse ex vivo (Boutary et al., Commun Biol 2021), which was also reproduced with about 13 nm in rat in vitro (Taveggia and Bolino, Methods Mol Biol 2018). The authors should acknowledge this deviation and might discuss possible reasons.

Author response: We have now provided a more representative EM image of our myelination (Fig. 2D). Additionally, thanks to the reviewer’s comments we have now quantified the interperiodic distance and find it is comparable to the studies the reviewer suggested. We have added the data to the new Table 1 and added the references the reviewer advised. Please see the additions to the methods and the results section regarding this new data below.

Results

To confirm that cocultured myelinated SCs formed compact myelin we performed electron microscopy (EM), which revealed compact myelin formation with multiple myelin wraps and formation of readily visible major dense (MDL) and intraperiod lines (IPL; Fig. 2D). Additionally, we measured the periodicity, i.e., the distance between two adjacent major dense lines, to make sure myelin was compacted. Interperiodic distance was 12.16 ± 0.28 nm, in line with previous reports (n=3; Table 1; Boutary et al, 2021; Fernando et al, 2016; García-Mateo et al, 2018; Giese et al, 1992; Perrot et al, 2007)__.

Methods

To measure interperiodic distance, we measured at least 10 periods per myelinated fibre for at least three fibres per sample for three separate samples.

• Figure 4A,B: The result of a slowed axon degeneration in coculture relies on the accurate assessment of continuity of the NFL staining. While the authors report that acetone permeabilization was necessary to afford complete penetration of the used antibodies in myelinating cultures, I cannot see why the authors have not used the same staining protocol for all cultures, as it is detailed in the method section. While I consider it unlikely that the staining conditions have led to an apparent delay of degeneration in coculture, experiments should generally be performed under identical conditions, unless there are good reasons not to do so. If this is not the case, it will be reassuring to see the same effect when identical staining conditions are employed. On the same note, do the compared cultures have the same age, i.e. have the neuron monocultures been in vitro for the same time as the cocultures?

Author response: We thank the reviewer for picking up this inaccuracy in the manuscript. We can confirm that for the purposes of the axon degeneration experiment all cultures were stained using exactly the same staining protocol. Additionally, we were very careful to maintain all cultures for exactly the same time in culture – 6 weeks. Additionally, axon only cultures were maintained in myelination medium to make sure medium constituents were not responsible for the observed differences in degeneration rate. We have added the following elaboration to the methods section to clarify these points.

Methods

Axon only cultures related to Figure 1 were permeabilised in PBS + 0.5% Triton (Merck) + 5% HS + 5% donkey serum (DS, Merck - D9663) at RT for 1 hour. For the purposes of quantifying the rate of axon degeneration (Figure 4) both axon only cultures and cocultures with SCs were permeabilised in 50% Acetone for 2 minutes, 100% Acetone for 2 minutes, 50% Acetone for 2 minutes (all at RT), and then blocked in PBS + 0.5% Triton + 5% HS + 5% DS at RT for 1 hour.

Methods

All cultures (axon only, aligned SCs and myelinating SCs) were cultured for 6 weeks prior to axotomy experiments. To minimise the possibility that medium constituents were responsible for differences in axon degeneration rates, axonal compartments of axon only cultures were cultured in medium containing 10 ng ml-1 b__NRG1 and 10 m__M forskolin (axon only medium, extended methods section D6) once SCs were seeded on other cultures, and then switched into myelination medium (additional Matrigel__â and 50 m__g ml−1 L-Ascorbic Acid), 24 hours before axotomy. Bottom compartments of aligned SC cultures, 24 hours before axotomy, were switched into DRG/SC medium containing 10 ng ml-1 b__NRG-1, 10 m__M forskolin and 50 m__g ml−1 L-Ascorbic Acid, which is insufficient to induce myelination in mouse cultures. Bottom compartments of myelinating cocultures were medium changed into fresh myelination medium (Extended methods section D7) 24 hours prior to axotomy.

• In several instances of the manuscript, the term "transfection" is used to refer to lentiviral gene transfer. I advise to use the more appropriate term "transduction" instead
• I could not seem to find a meaningful reference to the microfluidic chambers that were used in the study. The protocol should contain details on the device and source of supply in order to enable potential readers to execute the protocol

Author response: We thank the reviewer for this comment. We have replaced transfection with transduction throughout the manuscript. Please see the track changes manuscript for all instances.

Reviewer #2 (Significance):

The paper presents a convincing establishment of a dissociated coculture derived exclusively from mouse that leads to robust myelination. As the manuscript correctly states, Schwann cell culture and especially coculture with neurons has been experienced difficult in the field, and by providing a detailed protocol as well as demonstrating how the coculture system can be used to address important questions of PNS myelination and repair, the paper fills an important gap. However, the experiments directed to the role of Schwann cells in axon degeneration do not clarify much, which should be better addressed in the discussion and also by modifying the title accordingly.<br /> The paper will be of high value for basic researchers that are interested in performing studies addressing cellular and molecular mechanisms of myelination and repair in the PNS. Importantly, the paper can pave the way to usage of transgenic or knockout mouse models in coculture. Thereby it might spark interest also in those researchers that use transgenic and knockout mouse models and who have so far refrained from using coculture models.

Field of expertise of the reviewer: Cellular and molecular mechanisms of myelination and growth signaling in the PNS; in-depth experience with DRG coculture models from rats and mice

Author response: We thank the reviewer for their kind comments. We have now modified the title, aims and discussion of the manuscript in line with the reviewer’s suggestions.

Reviewer #3 (Evidence, reproducibility and clarity):

SUMMARY<br /> The authors present a detailed protocol for co-cultures of mouse DRGs with mouse SCs using microfluidics. In this model, cells of interest grow in different compartments while allowing for axons to grow in between, thereby making them accessible to injury induction. Using this experimental system, the authors show that myelination occurs, myelin gets compacted and acquires nodal organization. The authors then show that such a system allows for compartment-specific lentivirus transduction and live imaging. Next, they perform physical and chemical axonal injury and show that at early time point pos- injury the presence of SCs protects from axonal degeneration regardless of the myelination status, and helps with clearing of damaged axons at later time points.<br /> Major comments:

The novelty of the study is questionable.<br /> While the model is well described and appears to be useful for the proposed applications (live imaging, transduction, injury model), the arguments provided regarding its novelty are not fully convincing. The main argument from the authors of this paper is that there are no established protocols describing the use of mouse SCs with mouse DRG neurons in dissociated myelinating cocultures. However, this appears to be inaccurate, as the model described in Stevens et al., 1998 (cited in the paper) uses mouse DRG neurons dissected at E13.5 with mouse SCs dissected at P3 to study myelination. Also, in Päiväläinen et al., 2008, mouse DRGs and SCs are cultured from transgenic mice at different developmental ages, thereby arguing that coculture models have been previously successfully implemented. The main difference appears to be rather the compartmentalization of SCs and DRGs which appears to be a mouse adaptation of the rat model described by Vaquie et al,2019. Based on the above, it seems imperative for the authors to tone down the novelty aspect and provide a more thorough discussion on how the current novel differs from protocols in published study, highlighting advantages and caveats for each.

Author response: We agree with the reviewer that we did not make the case clear enough for how our coculture model adds to what is currently described in the literature. We have now changed the title and removed the word novel. New title:

A method for mouse myelinating Schwann cell-DRG neuron compartmentalised cocultures: myelination status does not influence the axo-protective effect of Schwann cells after injury.

Author response: We have now added the following paragraph to the introduction:

Introduction.

Indeed there has only ever been one laboratory detailing convincing myelin formation in dissociated mouse myelinating SC-DRG neuron cocultures, however this was never published as a step by step detailed protocol (Stevens et al, 1998; Stevens & Fields, 2000)__. In the last twenty years, there have been no published studies demonstrating myelination in fully dissociated mouse SC-mouse DRG cocultures. This has largely prevented the use of cells, particularly SCs, from transgenic mice in cocultures and thus restricted the ability to study SC-axon interactions in a system that can be readily manipulated and live imaged and results directly applied back to in vivo findings in the same species.

Author response: Regarding the study by Päiväläinen et al, 2008_,_ they did not fully dissociate their DRGs (see Fig.1 which demonstrates a DRG explant) and thus it is a non-dissociated DRG explant model. While they demonstrated convincing myelination due to the use of Matrigel which we acknowledge them for, their model is not perfectly suited for the use of neurons and SCs from different transgenic animals as the use of a DRG explant, even with temporary use of an antimitotic, risks contamination by endogenous SCs and satellite glia over time, especially as their model is not compartmentalised. We discuss the caveats of their protocol and those using dissociated mouse explant cocultures in a revised paragraph in the introduction.

Introduction.

Protocols exist where endogenous mouse SCs are used to myelinate dissociated or non-dissociated DRG explant cultures. (Shen et al, 2014; Harty et al, 2019; Sundaram et al, 2021; Stettner et al, 2013; Numata-Uematasu et al, 2023)__. Furthermore, another protocol seeded exogenous SCs onto non-dissociated DRG explant cultures (Päiväläinen et al, 2008)__. Other laboratories seed cultured rat SCs onto dissociated mouse DRG axons (Taveggia & Bolino, 2018)__. Use of dissociated or non-dissociated DRG explants cultures precludes many experimental uses, such as using SCs from different transgenic animals and separate transfection of SCs and neurons with viruses for live imaging or genetic manipulation, and easy use of microfluidic chambers to allow injury studies and separate drug treatments to neurons or SCs. The reason for this is that antimitotics cannot be used in dissociated or non-dissociated DRG explant cultures as this depletes SCs, and the culture quickly becomes contaminated with other non-neuronal cell types, such as satellite cells and fibroblasts migrating out of the DRG. Furthermore, use of exogenous SCs in a non-dissociated DRG explant culture risks, after a period of antimitotic exposure, which was developed by Päiväläinen et al., 2008 still risks potential contamination from endogenous SCs and satellite glia migrating out of the DRG explant over time. This occurs because antimitotic treatment is unlikely to fully penetrate the whole DRG without prior dissociation. Additionally, a compartmentalised culture system cannot be readily used with non-dissociated DRG explant cultures (Päiväläinen et al, 2008)__.

Author response: We have also added further discussion on how our protocol differs from the Stevens 1998 and other protocols in the discussion.

Discussion

Our protocol differs somewhat from the one used by Stevens et al., 1998 to induce myelination in dissociated mouse SC-DRG cocultures__, as they used ascorbic acid and 10% horse serum and presumably plated their cultures on laminin, though they do not explicitly detail this (Stevens et al, 1998)__. In our preliminary experiments we were unable to visualise much myelination with use of laminin, ascorbic acid or indeed if b__NRG1 and high concentration forskolin was added to the medium for up to four weeks. However, if we plated cocultures on Matrigel® and continuously added it to the myelination medium then we saw comparable levels of myelination in our mouse cocultures to that of rat cocultures (Eldridge et al, 1987)__. This approach of using Matrigel® to enhance myelination has previously been successfully employed in cultures of human iPSC sensory neurons with rat SCs and in in non-dissociated mouse DRG explant cultures (Clark et al, 2017; Päiväläinen et al, 2008)__. Importantly, we used growth factor depleted Matrigel® as standard Matrigel® preparations contain substantial amounts of Transforming growth factor b (TGF b__) which is a known inhibitor of myelination (Einheber et al, 1995)__. Additionally, the majority of rat and mouse coculture protocols plate cells on glass whereas we found cultures were healthier and myelinated better when cultured on plastic Alcar® coverslips.

Author response: We have added further discussion of comparable models in the literature in the discussion.

Discussion.

Furthermore, our protocol is complementary to the recently described 3D mouse myelinating SC-motor neuron coculture system using collagen hydrogels (Hyung et al, 2021; Park et al, 2021)__. It will be interesting in the future to up titrate the concentration of Matrigel®, which is similar to collagen hydrogels, in our cultures to see whether further increasing extracellular matrix viscosity and stiffness improves our myelination efficiency even further. While it is possible to study cell migration in microfluidic cell culture devices, transwell models offer significant advantages to study this cellular phenomenon (Negro et al, 2022)__. To date, there have been no published studies of successful myelination in human SC-neuron coculture systems. Despite this rat SCs have been shown to readily myelinate human-induced pluripotent stem cell (iPSC)-derived sensory neurons and an iPSC-derived peripheral nerve organoid system which does contain myelinating SCs has recently been described (Clark et al, 2017; Van Lent et al, 2022)__.

Next, the authors emphasize the conflicting results of two articles, Babetto et al., 2020 and Vaquie et al., 2019, as the basis to use their newly developed model in the same species and testing two ages corresponding to distinct myelination states. However, both studies reach the same conclusion as the current study, that SCs have a protective role, although at two different developmental time points. As such, it is likely that multiple mechanisms may account for the protective effect of SC on axonal damage, and therefore the different studies do not seem conflicting but rather complementary. Yet, it is interesting that this manuscript shows that the myelination status of SCs does not impact their ability to slow down degeneration and yet it confirms that different timing after injury elicits different behaviors in SCs, as suggested by the studies of Babetto et al., 2020 and Vaquie et al., 2019. In other words, a more accurate description of the results of these two studies is needed and a better explanation of what the authors consider to be conflicting and why (there could be more differences than species and myelination, for instance, such as the method used for axotomy - laser vs cut with scalpel which tear and pull membranes).

Author response: We would like to humbly correct the reviewer that the studies by Babetto et al., 2020 and Vaquie et al., 2019 do not reach the same conclusion that Schwann cells have a protective role. Instead, they describe axon protection (Babetto et al., 2020) and axon fragmentation (Vaquie et al., 2019). Our studies now visualise both phenomena in the same culture system. We have now made this point more explicit as well as highlighted the one conceptual advance our methods paper makes on the current literature, which is that myelination status does not influence the SC axo-protection, as the reviewer suggested.

Discussion.

These findings confirm the observations of both Babetto et al., 2020 and Vaquié et al., 2019 who used rat SCs in similar microfluidic culture systems (Vaquié et al, 2019; Babetto et al, 2020)__. We have shown that the axo-protective observation seen by Babetto et al., 2020 does not rely on myelination status, which was an outstanding question from that study (Babetto et al, 2020)__. Furthermore, in an advance from previous studies, we have visualised the axo-protective and axon clearance phenomena in the same culture and shown that they are temporally separated, with axon fragmentation and debris clearance by SCs occurring at much later timepoints after axotomy.

Author response: We have now added more in-depth discussion of the similarities and differences between Babetto et al., 2020 and Vaquie et al., 2019 and our approach in the discussion.

Discussion.

Several different culture and experimental conditions preclude direct comparison of our study with both those of Vaquié et al., 2019 and Babetto et al., 2020. These include the use of rat SCs in both prior studies and that Babetto et al., 2020 mixed rat SC with mouse DRG axons; length of time in culture, time points quantified after injury and distance from injury and site of analysis. Babetto et al., 2020 performed axotomy on relatively short term cocultures (6 days in vitro) whereas Vaiquié et al., 2019 cultured for at least 4 weeks and in our case 6 weeks prior to axotomy. Vaiquié et al., 2019 removed nerve growth factor (NGF) prior to laser axotomy and quantified proximally (though also imaged distally) whereas both our study and Babetto et al., 2020 kept NGF in the medium, performed axotomy with a scalpel and quantified more distally and in our case extremely distal, where only individual neurites and no axon bundles were visible. Vaquié et al., 2019 had SCs on both sides of the barrier in the microfluidic chambers whereas we seeded SCs only in the axonal/bottom compartment. Additionally, our cocultures had both forskolin and b__NRG1 added to help induce myelination, whereas these factors are not required in rat myelinating cocultures. Finally, it is important to permeabilise myelinated cultures with acetone after fixation__, as we did, to visualise the entire axon through heavily myelinated segments otherwise axon integrity cannot be reliably assessed in a quantitative manner (Vaquié et al, 2019; Babetto et al, 2020)__.

Author response: We would like to add that we showed Claire Jacob, senior author of the Vaquie et al., 2019 study, our manuscript prior to peer review and she offered helpful comments, which we incorporated into the manuscript, which is why she is acknowledged. We have also discussed our findings with Elisabetta Babetto as well.

Overall, the title does not appear to be the most appropriate because the content rather proposes a detailed protocol and gives examples of applications, rather than focusing on the protective versus destructive role of SCs on axons. It also appears to be misleading, as "axo-destructive" appears to suggest a negative role of Schwann cells on axons, whereas SC are rather helpful in clearing degenerative axons, a step which facilitates regeneration.

Author response: We have now changed the title and the focus of the manuscript in line with the reviewer’s comments. New title:

A method for mouse myelinating Schwann cell-DRG neuron compartmentalised cocultures: myelination status does not influence the axo-protective effect of Schwann cells after injury.

Author response: We have removed the phrase axo-destructive throughout the manuscript and instead referred to axon fragmentation and axon debris clearance roles of SCs in line with the reviewer’s suggestion. Please see track changes manuscript for all instances where this was modified.

The number of biological replicates for each experiment is not always indicated, and if the "n=" represent cultures prepared independently/passaged or wells/cell. It is essential to be rigorous and clearly indicate the number of technical replicates and biological samples throughout the manuscript and provide a thorough description of them. One example is Fig. 4 E were only 10 cells from a single culture appeared to have been imaged. Is this accurate? This aspect is essential to evaluate reproducibility, especially in view of the technical and biological variability.

Author response: We have now added quantification of myelin segments per mm2, percentage of SCs that myelinate and quantification of the interperiodic distance of the myelin formed. This is all included in a new version of TABLE 1. We have discussed this data in the results section as follows.

Quantifying the number of PRX positive myelin segments we found that there were 325.33 ± 12.3 sheaths per mm2, comparable to what has been originally described in rat SC/DRG cocultures and two fold more extensive myelination than recently described compartmentalised rat cocultures models (n=3; Table 1; Eldridge et al, 1987; Vaquié et al, 2019)__. Furthermore, 25.47 ± 1% of Schwann cells were myelinating in our cultures (n=3; Table 1). To confirm that cocultured myelinated SCs formed compact myelin we performed electron microscopy (EM), which revealed compact myelin formation with multiple myelin wraps and formation of readily visible major dense (MDL) and intraperiod lines (IPL; Fig. 2D). Additionally, we measured the periodicity, i.e., the distance between two adjacent major dense lines, to make sure myelin was compacted. Interperiodic distance was 12.16 ± 0.28 nm, in line with previous reports (n=3; Table 1; Boutary et al, 2021; Fernando et al, 2016; García-Mateo et al, 2018; Giese et al, 1992; Perrot et al, 2007)__.

Author response: We have discussed the number of cultures used for each quantification in the methods section. See below.

Quantification of Myelination in cocultures

To quantify the number of myelin segments per area, we counted the number of myelin segments for five areas per culture for three cultures and normalised this per mm2. To quantify the percentage of Schwann cells in myelinating cocultures that are actively myelinating, we quantified the number of myelin segments and the number of DAPI-positive nuclei for five areas per culture for three independently prepared cultures. To measure interperiodic distance, we measured at least 10 periods per myelinated fibre for at least three fibres per sample for three separate samples.

Quantification of Degeneration

Five images at a distance of between 1.2-1.4mm from the microgroove barrier (the most distal part of the culture that could be imaged) were quantified per culture, taken in comparable locations in each culture. A line was drawn across each image, and each axon crossing this line was either scored as degenerated or intact. Images were blinded prior to quantification. A minimum of three independently prepared cultures were assessed per timepoint for each condition.

Author response: We have removed Fig.4E and instead quoted the data in the results section as follows:

When we quantified this phenomenon, we found that 97.84 ± 1.462% (n=2) of SCs in our cocultures contained mCherry-labelled axonal fragments.

Author response: We apologise as the n number for this experiment was 2 (not 10), with cells in 10 areas quantified throughout all imaging timepoints from each independently prepared culture. We have included the following description in the methods section:

To quantify number of SCs with fragments, each cell was defined as a region of interest and checked for the presence of mCherry positive fragments at all timepoints. Two separate independently prepared cultures and cells in 10 areas per culture were analysed.

Author response: Additionally for Fig. 4B we have now included individual data points from independently prepared cultures.

N numbers are included in all figure legends and always refers to independently prepare cultures/biological replicates.

We have added to the relevant figure legends (Fig.3 and 4 and Table 1) the phrase:

n number refers to independently prepared cultures from separate litters of mice.

Minor comments:

• Does myelination reach axons in the microgrooves (it seems to from 2C, but up to where)? Where is axotomy performed and are axons myelinated where the cut was performed?

Author response: Myelination occasionally reaches the beginning of the microgrooves. We didn’t visualise myelination in the DRG cell body compartment. We have added the following detail to the methods section:

Traumatic axotomies were carried out by carefully removing the microfluidic chamber (SND150 and RND150, Xona Microfluidics__Ò__) from the Aclar__â coverslip using sterile forceps and severing axons with a surgical blade under a light microscope. Axotomies were carried out at the level of the microgroove barrier. To confirm all axons were severed, a second higher cut was performed and axons between the cut sites removed using the surgical blade.

Author response: Given this, we cannot exclude that the odd proximal myelin segment is cut, but the vast majority of axons are not myelinated at the site of cut (lower cut).

• Since the model allows for comparison of aligned vs myelinating SCs, and that both aligned and myelinating SCs seem to slow down degeneration, and that c-JUN is upregulated after in vivo injury, have the authors measured if c-JUN levels increase similarly in both myelinating vs aligned SCs?

Author response: We thank the reviewer for this suggestion. We have now quantified the JUN upregulation after injury in both myelinating and aligned cocultures as well as adding images of JUN upregulation in aligned cocultures. See Fig. 3B-E).

Additionally, we noted a strong upregulation of JUN protein in SCs 12 hours after axotomy (Fig. 3B and C). We also saw significant JUN upregulation 12 hours after axotomy in cocultures with aligned SCs (Fig. 3D and E).

Author response: We have decided to focus the manuscript more on the comparison of myelinating versus non-myelinating cocultures, given that we have shown that the axo-protective effect of SCs is independent of myelination status, which is an advance on what is known in the literature. In addition to changing the title, as we have mentioned previously, we have added further characterisation of our aligned cocultures with p75NTR immuno and EM images (Fig.1D and E).

We have

• On clarity:<br /> - In the step-by-step protocol, wording needs to be improved.

Author response: We have substantially edited the step-by-step protocol. Please see track changes document for all specific changes in wording.

• Temperatures for centrifugations are missing.

Author response: We have added temperatures for all centrifugation steps. Please see track changes document

• The MOI described for lentivirus is 2-10 in the protocol but 200 in the legend of Figure 3F.

Author response: The MOI for DRGs was 2-10 and SCs was 200 in Figure 3F. This is described similarly in the extended methods section. DRGs are transduced much more easily than SCs.

We have added the following sentence to the results section to emphasise this point:

Importantly dissociated mouse SCs required a much higher multiplicity of infection (MOI) than dissociated mouse DRGs (see extended methods section).

• Certain citations in the references list are incomplete (i.e. Babetto et al.; Catenaccio et al.,).

Author response: We have updated the reference list.

Reviewer #3 (Significance):

The advance for the field proposed by this paper is mostly technical, as it details a new model to be used by the field, of mouse SCs-mouse DRGs in dissociating myelinating cultures. The tested applications allowed the authors to also confirm a protective role for SCs on axonal damage, which was independent from myelination status.

Being a method paper, it is essential that the authors provide clear statements on the number of biological replicates, and technical repeats, as well as a very thorough and accurate description of the methodology.

The model described has similarities with existing models in the field such as Stevens et al., 1998 and Vaquié et al., 2019. To place it in context in a more helpful way, the authors should emphasize on the novelty brought by their protocol compared to existing models. The authors compare their findings to results from Vaquié et al., 2019 and Babetto et al., 2020 that they describe as conflicting, when it seems they rather address different mechanisms of SCs in protection and repair, occurring at different time points.

Audience might be interested in the detailed step by step protocol to use this in vitro model for the applications described, and investigate further why SCs myelination status does not influence their ability to protect from neurodegeneration early on or how to make use of this for neuroprotection studies.

Author response: We have now rephrased the description of Vaquié et al., 2019 and Babetto et al., 2020 studies in line with the reviewer’s suggestions. We have now added further discussion of our model in the context of all other models in the field as we have outlined in detail in above responses.

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 present a detailed protocol for co-cultures of mouse DRGs with mouse SCs using microfluidics. In this model, cells of interest grow in different compartments while allowing for axons to grow in between, thereby making them accessible to injury induction. Using this experimental system, the authors show that myelination occurs, myelin gets compacted and acquires nodal organization. The authors then show that such a system allows for compartment-specific lentivirus transduction and live imaging. Next, they perform physical and chemical axonal injury and show that at early time point pos- injury the presence of SCs protects from axonal degeneration regardless of the myelination status, and helps with clearing of damaged axons at later time points.

Major comments:

The novelty of the study is questionable.<br /> While the model is well described and appears to be useful for the proposed applications (live imaging, transduction, injury model), the arguments provided regarding its novelty are not fully convincing. The main argument from the authors of this paper is that there are no established protocols describing the use of mouse SCs with mouse DRG neurons in dissociated myelinating cocultures. However, this appears to be inaccurate, as the model described in Stevens et al., 1998 (cited in the paper) uses mouse DRG neurons dissected at E13.5 with mouse SCs dissected at P3 to study myelination. Also, in Päiväläinen et al., 2008, mouse DRGs and SCs are cultured from transgenic mice at different developmental ages, thereby arguing that coculture models have been previously successfully implemented. The main difference appears to be rather the compartmentalization of SCs and DRGs which appears to be a mouse adaptation of the rat model described by Vaquie et al,2019. Based on the above, it seems imperative for the authors to tone down the novelty aspect and provide a more thorough discussion on how the current novel differs from protocols in published study, highlighting advantages and caveats for each.

Next, the authors emphasize the conflicting results of two articles, Babetto et al., 2020 and Vaquie et al., 2019, as the basis to use their newly developed model in the same species and testing two ages corresponding to distinct myelination states. However, both studies reach the same conclusion as the current study, that SCs have a protective role, although at two different developmental time points. As such, it is likely that multiple mechanisms may account for the protective effect of SC on axonal damage, and therefore the different studies do not seem conflicting but rather complementary. Yet, it is interesting that this manuscript shows that the myelination status of SCs does not impact their ability to slow down degeneration and yet it confirms that different timing after injury elicits different behaviors in SCs, as suggested by the studies of Babetto et al., 2020 and Vaquie et al., 2019. In other words, a more accurate description of the results of these two studies is needed and a better explanation of what the authors consider to be conflicting and why (there could be more differences than species and myelination, for instance, such as the method used for axotomy - laser vs cut with scalpel which tear and pull membranes).

Overall, the title does not appear to be the most appropriate because the content rather proposes a detailed protocol and gives examples of applications, rather than focusing on the protective versus destructive role of SCs on axons. It also appears to be misleading, as "axo-destructive" appears to suggest a negative role of Schwann cells on axons, whereas SC are rather helpful in clearing degenerative axons, a step which facilitates regeneration.

The number of biological replicates for each experiment is not always indicated, and if the "n=" represent cultures prepared independently/passaged or wells/cell. It is essential to be rigorous and clearly indicate the number of technical replicates and biological samples throughout the manuscript and provide a thorough description of them. One example is Fig. 4 E were only 10 cells from a single culture appeared to have been imaged. Is this accurate? This aspect is essential to evaluate reproducibility, especially in view of the technical and biological variability.

Minor comments:

• Does myelination reach axons in the microgrooves (it seems to from 2C, but up to where)? Where is axotomy performed and are axons myelinated where the cut was performed?
• Since the model allows for comparison of aligned vs myelinating SCs, and that both aligned and myelinating SCs seem to slow down degeneration, and that c-JUN is upregulated after in vivo injury, have the authors measured if c-JUN levels increase similarly in both myelinating vs aligned SCs?
• On clarity:
• In the step-by-step protocol, wording needs to be improved.
• Temperatures for centrifugations are missing.
• The MOI described for lentivirus is 2-10 in the protocol but 200 in the legend of Figure 3F.
• Certain citations in the references list are incomplete (i.e. Babetto et al.; Catenaccio et al.,).

Significance

The advance for the field proposed by this paper is mostly technical, as it details a new model to be used by the field, of mouse SCs-mouse DRGs in dissociating myelinating cultures. The tested applications allowed the authors to also confirm a protective role for SCs on axonal damage, which was independent from myelination status.

Being a method paper, it is essential that the authors provide clear statements on the number of biological replicates, and technical repeats, as well as a very thorough and accurate description of the methodology.

The model described has similarities with existing models in the field such as Stevens et al., 1998 and Vaquié et al., 2019. To place it in context in a more helpful way, the authors should emphasize on the novelty brought by their protocol compared to existing models. The authors compare their findings to results from Vaquié et al., 2019 and Babetto et al., 2020 that they describe as conflicting, when it seems they rather address different mechanisms of SCs in protection and repair, occurring at different time points.

Audience might be interested in the detailed step by step protocol to use this in vitro model for the applications described, and investigate further why SCs myelination status does not influence their ability to protect from neurodegeneration early on or how to make use of this for neuroprotection studies.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In the manuscript entitled "Distinct axo-protective and axo-destructive roles for Schwann cells after injury in a novel compartmentalised mouse myelinating coculture system", Arthur-Farraj and colleagues detail a method of dissociated coculture of mouse-DRG neurons and Schwann cells in microfluidic chambers. In this system, neurons and Schwann cells harvested from the same or from different animals are grown in different compartments that are connected by microgrooves, thereby allowing for spatial and diffusive separation. Neurons are shown to extent their axons across the microgroove barrier to the glial compartment where Schwann cells align with the axons and myelination can be induced. Detailed analysis of myelination and axon injury/degradation are presented as use cases, including the capability to genetically and pharmacologically manipulate neurons and Schwann cells independently, which also enabled fluorescent life cell imaging. The authors then examine the effect of immature/premyelinating and myelinating Schwann cells on the rate of axon degeneration. Upon axotomy Schwann cells significantly delayed degeneration, with no difference between non-myelinating and myelinating Schwann cells. Finally, live imaging during axon degeneration with fluorescent proteins separately expressed in neurons and Schwann cells demonstrated that Schwann cells ingest axonal fragments.

Major comment:

In establishing a much needed in-vitro system for PNS myelination and injury research the paper represents a valuable contribution to the PNS community. However, I find the presentation of aspects concerning a protective/destructive role of Schwann cells somewhat inconclusive. That these roles exist has been known, as the authors discuss. Then what does this study contribute concerning the open question that was raised by the discrepancies between Babetto et al., 2020 and Vaquié et al., 2019, i.e. how Schwann cells contribute to axon survival/regeneration after injury? Essentially, the only significant conclusion in this regard is that myelinating and non-myelinating mouse Schwann cells do not differ in their capability to protect axons from degeneration. The manuscript, including the title, would benefit from focusing more on this aspect. In particular, the discussion of the factors that lead to the still remaining apparent discrepancies between Babetto et al., 2020 and Vaquié et al., 2019 and this study should be revised. The authors state that "The study by Vaiquié et al., 2019 quantified axon fragmentation proximally in the microgrooves at timepoints starting from 12 hours after axotomy." (Discussion). While this observation is accurate, Jacob and colleagues also show accelerated, obviously distal axon degeneration in the presence of Schwann cells (Figure 3C in Vaquié et al., 2019). It is therefore unlikely that the discrepancies stem from analysis of more proximal vs. more distal axons, or the timepoints of analyses. In my opinion, a further study (using the coculture system presented in this manuscript) that compares the role of Schwann cells from rats and mice, and that includes analysis of more proximal and distal axon degeneration as well analysis of axon regeneration is needed. In a rework of the manuscript, the authors may therefore elaborate more on the shortcomings of the present study, or alternatively soften the aims of the study in the first place.

Minor comments:

• Figure 2D: From the electron micrograph there is no doubt that compact myelin is formed, however to me it seems the compaction is not complete. A rough estimation with the aid of the provided scale bar resulted in an interperiodic distance of about 17 nm, which contrasts with the remarkably well reproduced values reported in multiple reports using conventional specimen preparation (like in this study), of which I am citing just a few: about 13 nm in rat ex vivo nerve (Peterson and Pease, J. Ultrastruct Res 1972; Fledrich et al., Nat Commun 2018), 12 nm (Giese et al., Cell 1992), 12.2 nm (Perot et al., J Neurosci 2007), about 12 (Fernando et al., Nat Commun 2016), about 13 nm (García-Mateo et al., Glia 2017) or about 13 nm in mouse ex vivo (Boutary et al., Commun Biol 2021), which was also reproduced with about 13 nm in rat in vitro (Taveggia and Bolino, Methods Mol Biol 2018). The authors should acknowledge this deviation and might discuss possible reasons.
• Figure 4A,B: The result of a slowed axon degeneration in coculture relies on the accurate assessment of continuity of the NFL staining. While the authors report that acetone permeabilization was necessary to afford complete penetration of the used antibodies in myelinating cultures, I cannot see why the authors have not used the same staining protocol for all cultures, as it is detailed in the method section. While I consider it unlikely that the staining conditions have led to an apparent delay of degeneration in coculture, experiments should generally be performed under identical conditions, unless there are good reasons not to do so. If this is not the case, it will be reassuring to see the same effect when identical staining conditions are employed. On the same note, do the compared cultures have the same age, i.e. have the neuron monocultures been in vitro for the same time as the cocultures?
• In several instances of the manuscript, the term "transfection" is used to refer to lentiviral gene transfer. I advise to use the more appropriate term "transduction" instead
• I could not seem to find a meaningful reference to the microfluidic chambers that were used in the study. The protocol should contain details on the device and source of supply in order to enable potential readers to execute the protocol

Significance

The paper presents a convincing establishment of a dissociated coculture derived exclusively from mouse that leads to robust myelination. As the manuscript correctly states, Schwann cell culture and especially coculture with neurons has been experienced difficult in the field, and by providing a detailed protocol as well as demonstrating how the coculture system can be used to address important questions of PNS myelination and repair, the paper fills an important gap. However, the experiments directed to the role of Schwann cells in axon degeneration do not clarify much, which should be better addressed in the discussion and also by modifying the title accordingly.

The paper will be of high value for basic researchers that are interested in performing studies addressing cellular and molecular mechanisms of myelination and repair in the PNS. Importantly, the paper can pave the way to usage of transgenic or knockout mouse models in coculture. Thereby it might spark interest also in those researchers that use transgenic and knockout mouse models and who have so far refrained from using coculture models.

Field of expertise of the reviewer: Cellular and molecular mechanisms of myelination and growth signaling in the PNS; in-depth experience with DRG coculture models from rats and mice

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 original article from Mutscher et al., the authors developed a compartmentalized dissociated mouse myelinating SC-DRG coculture system to investigate the distinct roles of Schwann cells in axon protection and degeneration after injury. The innovation of this approach relies on (i) the use of mouse SCs and mouse DRGS neurons instead of rat cells; (ii) the use of microfluidic chambers, seeded by axons and SCs in different compartment; (iii) the possibility to perform a traumatic injury in vitro. While this novel approach offers new ways to study peripheral nerve regeneration and SC-axon interaction, and technical the study is robust, the paper is currently limited by the exploration of their model.

Major points:

It is unclear is this approach will ever lead to the identification of key mechanism or key candidates. This is a major miss in the current manuscript form. In short: the authors should demonstrate that their in vitro system can lead to significant leap in our understanding of peripheral nerve regeneration by identifying novel targets/pathways or mechanisms.

The use of embryonic DRG neurons or SC isolated from P2 animals are arguably physiologically not the same cells that are affected by traumatic nerve injury, which happen most often than not in adult. This is a problem in the long-term reliance on this approach to study axotomy peripheral nerve regeneration.

Moderate points:

"there are no established protocols in the field describing the use of mouse SCs with mouse DRG neurons in dissociated myelinating cocultures". The use of mouse cells is laudable, but it is not necessarily a technical innovation, or at least the current manuscript does not explain why their approach particularly suitable to mouse Schwann cells.

The figures in the paper are largely descriptive. They are very little quantitative measurement. Thus, the readers will have a hard to determine, if they replicate the proposed approach, whether their efficient is on par with the current authors.

Minor points:

Fig.4B and E should show individual data points.

Significance

In addition, to the demonstration of feasibility of this in vitro approach, the main finding by the authors is that SCs have a role in neuronal protection and support is key for peripheral nerve regeneration. Thus while in vitro approach does not add key information that do not already exists in the field, it somewhat confirms that the effect is SC autonomous. Overall the approach is interesting and has potential, but the study currently lack a demonstration of its usefulness to the community.

It would have been interesting to have the authors discuss the advantages of their approach in comparison to other innovative approaches to study SC-axon interactions that have been developed in the last decade (i.e., 3D environment, microfluidic approach, transwell systems). There is also a lack of citations about similar studies in the field.

Because of the lack of key novel mechanisms, and lack of discussion on what this approach is superior to others in vitro approach limits the impact of the study and the excitement of the reader, even from the SC-axon community.

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):

In this manuscript the author is presenting a deep-learning model used to predict the development stage of zebrafish embryo. A robust method that can accurately classify a zebrafish into different development stages is highly relevant for many researchers working with zebrafish and hence the importance in developing methods like this is high.

The manuscript is overall ok. However, more data is needed to convince the reader that the method is robust enough to work with other samples in other labs. This would greatly improve the impact of the publication.

We agree with the reviewer and have included in our revised manuscripts additional test data that was acquired at a different laboratory to the training data (Figures 5 - 7).

Page 6.<br /> - How is the data acquired?

Images used to do initial model training are the same as those used in a previous study - the details of image acquisition are contained in the relevant publication (doi: 10.12688/wellcomeopenres.18313.1). However, we have now added “Zebrafish Husbandry” and “Live Imaging” for newly-acquired images. We have added a table (Table 1) listing details of all image data used in the study.

Page 8.<br /> "This indicates that whileKimmelNet can be used successfully with noisier test data than that on which it was trained,there is an upper limit to how noisy the data can be."<br /> - This is an obvious statement there will always be an upper limit on noise.

We agree with the reviewer that this statement is not terribly informative. This section (“KimmelNet’s prediction accuracy is not significantly impacted by moderate levels of additive noise”) has been removed from the revised manuscript in favour of incorporating a section detailing testing of the model on newly-acquired images (“KimmelNet can generalise to previously unseen data”).

Page 9.<br /> - Are only wildtype embryos used? How would this work on different mutants. To evaluate the robustness of the method this it would be valuable to test on some mutant line with known developmental difference from the wild type.

We agree with the reviewer that testing on a mutant line would lend more weight to our findings. For example, the p53-/- zebrafish has a reported, published developmental delay, but we did not have access to that line. However, the developmental delay reported for the p53-/- mutant is virtually indistinguishable from that effected by a temperature change. We therefore focussed our efforts on developmental delay affected by altering incubation temperature only.

Image data.<br /> - I would strongly suggest that the author should include a description of the data in the manuscript. A description of how the data is acquired, microscope, different batches, type of embryos used.

The image data used in the first draft of the manuscript is the same as that used in a previous publication (Jones et al. 2022), which contains all the relevant details the reviewer seeks. However, we have now added the relevant information for the newly-acquired image data.

"Random160translation in the y-direction was avoided due to the aspect ratio of the images (width>161height) - any artifacts introduced by translation in the x-direction would be removed by the162centre crop layer, but this would not be the case for translation in the y-direction."<br /> - Could this be solved by using some border method reflection, repetition or fixed value?

The reviewer is correct that some form of image reflection or repetition could be utilised. However, given the nature of our images, if an embryo is located close to the image boundary, reflection/repetition can result in some odd artefacts, so we minimised the use of such fill methods (also used by the random zoom augmentation layer). We could instead use an arbitrary fixed value, as the reviewer suggested, but finding a value suitable for all images is difficult.

Page 10.<br /> Addition of Noise to Image Data<br /> - This should be added in the training phase. This would probably improve the robustness of the network and also improve the results on the test data.

We agree with the reviewer and have now added a random Gaussian noise layer for data augmentation purposes during model training (see Figure 1).

• Supplementary 3 images with high noise. It is worrying that the network is not able to handle the noise in this figure. Looks like the features that is used to distinguish the developmental stage of the embryo is still clearly seen with this high noise level? Retrain the model with noise as an augmentation to improve this.

As the reviewer suggested, addition of random noise is now incorporated into model training. The new version of the manuscript does not include the same supplemental figures, but instead includes additional testing on newly-acquired data instead.

Reviewer #1 (Significance):

The development of methods like this is highly relevant in the zebrafish community. Staging and evaluating the developmental stage for zebrafish is common and is of interest to the broad community. A lot of this work today is done manually, limiting the throughput, and adding human bias.

The limit of this study is the dataset used for training and evaluation. Firstly, it is not clear about the structure of the data and how it is acquired, different types of fish or imaging setup etc. For a method to be useful to the community it needs to be robust enough to handle different types of fish (transgenic lines). The manuscript would be greatly improved by adding this to the training and evaluation.

We have now added additional datasets for the purposes of evaluating the model.

My expertise is image analysis and machine learning for quantification of biological samples, with focus on zebrafish screening.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary<br /> The paper "Automated staging of zebrafish embryos with KimmelNet" by Barry et al., presents a method to automatically stage developmental timepoints of zebrafish embryos based on convolutional neural networks (CNN). The authors show that a CNN trained on ~20k images can determine time post fertilization on test-image sets with an accuracy on the range of a few hours. This technique undoubtedly has the potential to become very useful for any zebrafish researchers interested in developmental timing as it eases analysis and removes potential subjective bias.

Major comments<br /> In its current form the paper lacks sufficient graph annotations and method descriptions. This makes it hard in places to judge the validity of the claims. I do believe that the presented method can be useful and is likely valid but to be convincing, the authors need to spend more time expanding the methods, justifying their choices of analysis and clarifying figure annotations.

We believe that we have addressed the reviewer’s concerns in this revised manuscript, as detailed in response to the specific points below.

Specific points:<br /> 1) The annotation of the training data is not described and specifically it is unclear how valid the staging of the training data itself is. The authors state in the introduction "the hours post fertilization (hpf) [...] provides only and approximation of the actual developmental stage". It is therefore critical to know how this was accounted for in the annotation of the training data. Since the quality of the training data will ultimately limit the best-case quality of Kimmel Net. The authors need to go into some detail here even though the training data appears to be from another published dataset.

The reviewer raises a valid point – two individual zebrafish embryos that are x hours post-fertilisation are not necessarily at the same developmental stage. However, we believe it is reasonable to assume that two populations of embryos x hours post-fertilisation are, on average, at the same developmental stage and it is this assumption that forms the basis for our approach. Given the inherent variability the reviewer refers to, we are not suggesting that our model would be particularly accurate for staging individual embryos. However, we are very confident (and we believe the data in the manuscript supports this) that given a population of embryos, our model will provide an accurate rate of development. We have added a paragraph (lines 131-141) to address this point.

2) Why were "test predictions fit to a straight line through the origin". On the one hand this makes sense (since the slope would indicate the correspondence) but it should be clarified why an intercept was omitted in the fit. After all it is unclear if Kimmel net correctly identifies 0Hpf embryos.

The reviewer makes a valid point – we do not know what predictions KimmelNet would produce for images of embryos closer to 0 hpf. However, we felt an equation of the form y=mx was a reasonable choice for two reasons. First of all, it matches the form of the Kimmel equation, which, despite its flaws, we are using as a benchmark of sorts – the absence of a y intercept makes comparisons with the Kimmel equation straightforward. Secondly, a “perfect” model would produce a straight line fit with y=x, so the lack of a y intercept seemed a reasonable constraint to impose. We have added some brief text (lines 103-105) to clarify this choice.

3) The methods do not list how the mean of the absolute error was calculated from 3B/C. I think this should be the mean of the absolute error (not the mean of the error) but in that case the numbers listed in the text appear rather small given the histograms in 3 B/C. A clear statement in the methods would clarify this issue.

We have now added a “Statistical Analysis” section under Materials & Methods to detail exactly what was used to calculate the values given for error analysis. We have calculated the mean of the error, not the mean of the absolute error, as we wish to illustrate that the mean is close to zero. We have used the standard deviation of the errors to illustrate that there is a significant spread in the error values, as depicted in Figure 3C and D.

Minor comments<br /> 1) The Y-axis in Figure 2B is simply labeled "Loss" - what is the unit of this loss? HPF? Or HPF^2? This is important for judging the quality of the fit

We thank the reviewer for drawing our attention to this omission. The loss is hpf2 (mean squared error) and we have updated the plot to reflect this.

2) Figure 3 B. I would suggest changing the labels of the confidence intervals in the legend. "Inner and outer" is already clear from the figure itself, so labels that state that these are derived from n=100 vs. n=20 test image sized samples would be more useful to the reader

We thank the reviewer for this suggestion – we have updated the figure legend accordingly.

Referees cross-commenting

I concur with comments issued by the other reviewers. I think it will be especially important to address the comments related to testing the method on mutants (Reviewer #1) and training the model in the presence of noise to increase its robustness (Reviewers #1 and #3) as well as addressing the overall annotation/generation of the training data (Reviewers #1 and #2).

We believe we have now addressed all of these concerns. The model has been retrained with additional data augmentation incorporating random noise, tested on newly-acquired data and we have added tables summarising the details of all image data used in this study.

I think these points will be critical to make the paper useful by increasing transparency and ensuring reproducibility in other labs with different imaging conditions, strains, mutants, etc.

Reviewer #2 (Significance):

Developmental delay is a common occurrence that can be caused by genetic and environmental background effects. It is therefore highly desirable to properly quantify this variable. The work presented here makes an important step in this direction, by allowing to quantify developmental timepoints independent of subjective staging. This speeds up analysis, increases reproducibility and enhances rigor. However, as my comments above indicate, the significance also depends on the ability of potential users to judge the quality of the work. Once those issues have been addressed, I think the work will be of broad interest to the developmental biology community, first and foremost labs utilizing the zebrafish model. However, as the authors state, the presented model architecture could be trained with the data from other species as well.

Expertise: Zebrafish, quantitative analysis, behavior, neuroscience

We thank the reviewer for their positive feedback.

Reviewer #3 (Evidence, reproducibility and clarity):

Summary:

Properly staging embryos of zebrafish embryos is important, yet provides challenging since it can depend on many factors, such as temperature, water quality, fish population density, etc. Here, the authors provide a deep-learning-based model called KimmelNet that allows the prediction of the age of zebrafish embryos, using 2D brightfield images. The technique is robust to weak measurement noise and can also be used to identify developmental delays from a very small number of experimental data.

The code is accessible to the reader, open-source and should be useable by experimentalists without huge effort.

Major comments:

I suggest retraining the model and application of the model to additional data for the following reasons:<br /> • Why did the authors not train for (high) measurement noise and heterogeneous background illumination? Would that not make the model more robust? In principle, creating training should not be considerably harder than before, since the manipulation of the images has been already shown in the manuscript and the authors just need to run it again on the HPC cluster. If there are no technical or administrative constraints (access to the cluster, computational effort, high costs, etc.), the authors should retrain their model.

We thank the reviewer for this suggestion. As detailed in Figure 1, with a view to making the model more robust, we have now added several more layers of data augmentation, including the addition of random noise, and retrained our model.

• For Fig. S2 and S3 it is not clear if there is such a strong deviation from the Kimmel equation due to measurement noise or due to the background illumination. The saliency maps appear as if they are mainly affected by the illumination, and maybe less by the noise. Would it be possible to apply the model to a case without artificial noise, but with heterogeneous background illumination to identify what has a bigger impact?

We thank the reviewer for this suggestion. We have now replaced the “artificial” examples used in the previous version of the manuscript with newly-acquired data (Figure 5), which exhibits different characteristics to that used for training.

Additionally, the authors need to clarify what exactly they are comparing in this manuscript and rework their interpretation of their findings:<br /> • When comparing the predictions between KimmelNet and the Kimmel equation, the authors use an equation of the form y=mx. Could the authors please elaborate on why they introduce the constraint of y(0)=0? It might be naturally given by the so-called Kimmel equation, but by looking at Fig 3a, it seems like an equation of the form y=mx+a would be more appropriate and it appears like KimmelNet introduces an offset of around a=2h for 25 Celsius. The authors need to discuss this.

The main rationale for using an equation of the form y=mx is to be consistent with the Kimmel equation (see lines 103-105). The reviewer is correct that an equation of the form y=mx+c may well produce a better fit to the data, but omitting a y intercept makes comparison with the Kimmel Equation trivial.

• In lines 5-8 the authors say that developmental stage progression does not only depend on temperature, but also on population density, water quality etc. and they explain that usually not only hpf, but also staging guides based on morphological criteria are used! If that is true, how good is their training data set that only uses hpf and not the other important guides? How did the authors test that these factors have no impact on their training data?

We have now added a paragraph (lines 131-141) to address this point.

Since this tool has the potential to have a big impact on zebrafish research, it would be nice to provide some examples of how exactly this could be achieved:<br /> • Could the authors discuss how exactly their tool is useful to experimentalists? Is it the idea that if an experimentalist wants to investigate an embryo of a particular stage, they apply KimmelNet to images of embryos to identify the stage of the embryo and only then undertake their planned experiment? Is that a realistic undertaking?

As is evidenced by the errors presented in Figure 3C & D, testing KimmelNet on individual images of embryos may well result in a large error in the predicted hpf. As such, it is not appropriate to use the tool in such a manner. However, to modify the example provided by the reviewer, should an experimentalist have a population of embryos they wished to stage, then yes, KimmelNet would certainly be an appropriate tool for doing so.

• Would it be possible to provide a tutorial (or even video tutorial if such skills are available in the group of authors) that provides real examples of how to apply the technique? This would make it easier for people without advanced Python/Deep-Learning skills to use the tool, hence improving the impact of KimmelNet.

A lack of user-friendly interfaces for applying deep learning methods in biology is well-documented – basic knowledge of python and tools like jupyter notebooks are often necessary (https://doi.org/10.1038/s41592-023-01900-4). However, we have endeavoured to make the running of KimmelNet as easy as possible for new users. A jupyter notebook instance can be run on Binder with absolutely no set-up required. To run KimmelNet on their own data, biologists just need to download the Git repo and replace the test images with their own data. We have updated the landing page on the GitHub repo to provide more specific step-by-step instructions for each of these tasks. We will also endeavour to upload our model to the BioImage Model Zoo (https://bioimage.io/#/) to further increase accessibility.

I am very critical towards equation 1. Please note that I don't think this has any impact on the quality of the technique provided in this manuscript and the significant flaws can already be found in Kimmel 1995 (which is not under review here). That is why I suggest rewriting of this manuscript to not support an over-interpretation of this equation.<br /> • I do not think that the Kimmel equation is an established term. At least a Google Scholar Search for "Kimmel equation" only gives one result: the preprint of this manuscript.<br /> • The equation has no mathematical meaning regarding its units (subtracting temperature and a unitless value). I also very rarely see equations with Degrees Celsius and not Kelvin.<br /> • Additionally, the equation provides a linear relationship between the development time and temperature h(T) and in Kimmel et al, it is shown that this is only true for 25-33 Celsius. Such a linearisation is not very surprising for a small temperature range, but I am not sure how true it is for higher temperature differences. Hence, I feel that it is very bold to give a specific name to such an equation, giving it an importance that it does not deserve.

We appreciate the reviewer’s concerns and have removed explicit references to “The Kimmel Equation”, without substantively changing the content of the manuscript.

Minor comments:

• For the measurement noise cases it would be nice to have some example images of fish with the specific noise levels in Fig S1 and Fig S2.

We have now removed the “synthetic” additive noise test data, previously depicted in Figures S1-3, in favour of newly-acquired images in Figures 5-7.

• Could the authors hypothesize why they predict a slower dynamic for 25 Celsius than predicted by the Kimmel equation?

Referring to Figure 2 in Kimmel et al (1995), it is apparent that the straight lines are by no means perfect fits to the datapoints. In Fig 2A in particular, some datapoints for the 25C data fall well below the line fit. While the published equation suggests a rate of development 80.5% of the rate at 28.5C, according to Fig 2A, an alternative line fit could give a developmental rate as low as 70-75%, which would be in agreement with our data.

Reviewer #3 (Significance):

Strengths of the study:

An easy-to-use method to automatically stage zebrafish embryos and identify differences in the developmental stage is very important for the zebrafish community and the technique in this manuscript definitely novel. The tool is can be used without large effort and the authors suggest that it can also find applications beyond zebrafish embryos. Hence, it is not only interesting to the zebrafish community, but to a broader developmental biology audience.

Weakness of the study:<br /> The main weakness of the manuscript is in the training data used for the deep-learning model and the apparent large impact of heterogeneous background illumination. If that is not solved, it is unclear if this technique will find an application in the zebrafish community.

We believe this weakness has now been addressed by incorporating additional data augmentation measures in the training process and testing the model on newly-acquired data.

Field of expertise of the reviewer: Image Analysis, Mathematical Modelling, Biological Physics. While I have limited experience in deep learning, I cannot evaluate the specific details of the network architecture. I also have no experience in zebrafish research.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Properly staging embryos of zebrafish embryos is important, yet provides challenging since it can depend on many factors, such as temperature, water quality, fish population density, etc. Here, the authors provide a deep-learning-based model called KimmelNet that allows the prediction of the age of zebrafish embryos, using 2D brightfield images. The technique is robust to weak measurement noise and can also be used to identify developmental delays from a very small number of experimental data.

The code is accessible to the reader, open-source and should be useable by experimentalists without huge effort.

Major comments:

I suggest retraining the model and application of the model to additional data for the following reasons:<br /> - Why did the authors not train for (high) measurement noise and heterogeneous background illumination? Would that not make the model more robust? In principle, creating training should not be considerably harder than before, since the manipulation of the images has been already shown in the manuscript and the authors just need to run it again on the HPC cluster. If there are no technical or administrative constraints (access to the cluster, computational effort, high costs, etc.), the authors should retrain their model.<br /> - For Fig. S2 and S3 it is not clear if there is such a strong deviation from the Kimmel equation due to measurement noise or due to the background illumination. The saliency maps appear as if they are mainly affected by the illumination, and maybe less by the noise. Would it be possible to apply the model to a case without artificial noise, but with heterogeneous background illumination to identify what has a bigger impact?

Additionally, the authors need to clarify what exactly they are comparing in this manuscript and rework their interpretation of their findings:<br /> - When comparing the predictions between KimmelNet and the Kimmel equation, the authors use an equation of the form y=mx. Could the authors please elaborate on why they introduce the constraint of y(0)=0? It might be naturally given by the so-called Kimmel equation, but by looking at Fig 3a, it seems like an equation of the form y=mx+a would be more appropriate and it appears like KimmelNet introduces an offset of around a=2h for 25 Celsius. The authors need to discuss this.<br /> - In lines 5-8 the authors say that developmental stage progression does not only depend on temperature, but also on population density, water quality etc. and they explain that usually not only hpf, but also staging guides based on morphological criteria are used! If that is true, how good is their training data set that only uses hpf and not the other important guides? How did the authors test that these factors have no impact on their training data?

Since this tool has the potential to have a big impact on zebrafish research, it would be nice to provide some examples of how exactly this could be achieved:<br /> - Could the authors discuss how exactly their tool is useful to experimentalists? Is it the idea that if an experimentalist wants to investigate an embryo of a particular stage, they apply KimmelNet to images of embryos to identify the stage of the embryo and only then undertake their planned experiment? Is that a realistic undertaking?<br /> - Would it be possible to provide a tutorial (or even video tutorial if such skills are available in the group of authors) that provides real examples of how to apply the technique? This would make it easier for people without advanced Python/Deep-Learning skills to use the tool, hence improving the impact of KimmelNet.

I am very critical towards equation 1. Please note that I don't think this has any impact on the quality of the technique provided in this manuscript and the significant flaws can already be found in Kimmel 1995 (which is not under review here). That is why I suggest rewriting of this manuscript to not support an over-interpretation of this equation.<br /> - I do not think that the Kimmel equation is an established term. At least a Google Scholar Search for "Kimmel equation" only gives one result: the preprint of this manuscript.<br /> - The equation has no mathematical meaning regarding its units (subtracting temperature and a unitless value). I also very rarely see equations with Degrees Celsius and not Kelvin.<br /> - Additionally, the equation provides a linear relationship between the development time and temperature h(T) and in Kimmel et al, it is shown that this is only true for 25-33 Celsius. Such a linearisation is not very surprising for a small temperature range, but I am not sure how true it is for higher temperature differences. Hence, I feel that it is very bold to give a specific name to such an equation, giving it an importance that it does not deserve.

Minor comments:

• For the measurement noise cases it would be nice to have some example images of fish with the specific noise levels in Fig S1 and Fig S2.
• Could the authors hypothesize why they predict a slower dynamic for 25 Celsius than predicted by the Kimmel equation?

Significance

Strengths of the study:

An easy-to-use method to automatically stage zebrafish embryos and identify differences in the developmental stage is very important for the zebrafish community and the technique in this manuscript definitely novel. The tool is can be used without large effort and the authors suggest that it can also find applications beyond zebrafish embryos. Hence, it is not only interesting to the zebrafish community, but to a broader developmental biology audience.

Weakness of the study:

The main weakness of the manuscript is in the training data used for the deep-learning model and the apparent large impact of heterogeneous background illumination. If that is not solved, it is unclear if this technique will find an application in the zebrafish community.

Field of expertise of the reviewer:

Image Analysis, Mathematical Modelling, Biological Physics. While I have limited experience in deep learning, I cannot evaluate the specific details of the network architecture. I also have no experience in zebrafish research.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The paper "Automated staging of zebrafish embryos with KimmelNet" by Barry et al., presents a method to automatically stage developmental timepoints of zebrafish embryos based on convolutional neural networks (CNN). The authors show that a CNN trained on ~20k images can determine time post fertilization on test-image sets with an accuracy on the range of a few hours. This technique undoubtedly has the potential to become very useful for any zebrafish researchers interested in developmental timing as it eases analysis and removes potential subjective bias.

Major comments

In its current form the paper lacks sufficient graph annotations and method descriptions. This makes it hard in places to judge the validity of the claims. I do believe that the presented method can be useful and is likely valid but to be convincing, the authors need to spend more time expanding the methods, justifying their choices of analysis and clarifying figure annotations.

Specific points:

1. The annotation of the training data is not described and specifically it is unclear how valid the staging of the training data itself is. The authors state in the introduction "the hours post fertilization (hpf) [...] provides only and approximation of the actual developmental stage". It is therefore critical to know how this was accounted for in the annotation of the training data. Since the quality of the training data will ultimately limit the best-case quality of Kimmel Net. The authors need to go into some detail here even though the training data appears to be from another published dataset.
2. Why were "test predictions fit to a straight line through the origin". On the one hand this makes sense (since the slope would indicate the correspondence) but it should be clarified why an intercept was omitted in the fit. After all it is unclear if Kimmel net correctly identifies 0Hpf embryos.
3. The methods do not list how the mean of the absolute error was calculated from 3B/C. I think this should be the mean of the absolute error (not the mean of the error) but in that case the numbers listed in the text appear rather small given the histograms in 3 B/C. A clear statement in the methods would clarify this issue.

Minor comments

1. The Y-axis in Figure 2B is simply labeled "Loss" - what is the unit of this loss? HPF? Or HPF^2? This is important for judging the quality of the fit
2. Figure 3 B. I would suggest changing the labels of the confidence intervals in the legend. "Inner and outer" is already clear from the figure itself, so labels that state that these are derived from n=100 vs. n=20 test image sized samples would be more useful to the reader

Referees cross-commenting

I concur with comments issued by the other reviewers. I think it will be especially important to address the comments related to testing the method on mutants (Reviewer #1) and training the model in the presence of noise to increase its robustness (Reviewers #1 and #3) as well as addressing the overall annotation/generation of the training data (Reviewers #1 and #2).

I think these points will be critical to make the paper useful by increasing transparency and ensuring reproducibility in other labs with different imaging conditions, strains, mutants, etc.

Significance

Developmental delay is a common occurrence that can be caused by genetic and environmental background effects. It is therefore highly desirable to properly quantify this variable. The work presented here makes an important step in this direction, by allowing to quantify developmental timepoints independent of subjective staging. This speeds up analysis, increases reproducibility and enhances rigor. However, as my comments above indicate, the significance also depends on the ability of potential users to judge the quality of the work. Once those issues have been addressed, I think the work will be of broad interest to the developmental biology community, first and foremost labs utilizing the zebrafish model. However, as the authors state, the presented model architecture could be trained with the data from other species as well.

Expertise: Zebrafish, quantitative analysis, behavior, neuroscience

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript the author is presenting a deep-learning model used to predict the development stage of zebrafish embryo. A robust method that can accurately classify a zebrafish into different development stages is highly relevant for many researchers working with zebrafish and hence the importance in developing methods like this is high.

The manuscript is overall ok. However, more data is needed to convince the reader that the method is robust enough to work with other samples in other labs. This would greatly improve the impact of the publication.

Page 6.<br /> - How is the data acquired?

Page 8.<br /> "This indicates that whileKimmelNet can be used successfully with noisier test data than that on which it was trained,there is an upper limit to how noisy the data can be."<br /> - This is an obvious statement there will always be an upper limit on noise.

Page 9.<br /> - Are only wildtype embryos used? How would this work on different mutants. To evaluate the robustness of the method this it would be valuable to test on some mutant line with known developmental difference from the wild type.

Image data.<br /> - I would strongly suggest that the author should include a description of the data in the manuscript. A description of how the data is acquired, microscope, different batches, type of embryos used.

"Random160translation in the y-direction was avoided due to the aspect ratio of the images (width>161height) - any artifacts introduced by translation in the x-direction would be removed by the162centre crop layer, but this would not be the case for translation in the y-direction."<br /> - Could this be solved by using some border method reflection, repetition or fixed value?

Page 10.<br /> Addition of Noise to Image Data<br /> - This should be added in the training phase. This would probably improve the robustness of the network and also improve the results on the test data.

• Supplementary 3 images with high noise. It is worrying that the network is not able to handle the noise in this figure. Looks like the features that is used to distinguish the developmental stage of the embryo is still clearly seen with this high noise level? Retrain the model with noise as an augmentation to improve this.

Significance

The development of methods like this is highly relevant in the zebrafish community. Staging and evaluating the developmental stage for zebrafish is common and is of interest to the broad community. A lot of this work today is done manually, limiting the throughput, and adding human bias.

The limit of this study is the dataset used for training and evaluation. Firstly, it is not clear about the structure of the data and how it is acquired, different types of fish or imaging setup etc. For a method to be useful to the community it needs to be robust enough to handle different types of fish (transgenic lines). The manuscript would be greatly improved by adding this to the training and evaluation.

My expertise is image analysis and machine learning for quantification of biological samples, with focus on zebrafish screening.

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

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements

The authors greatly appreciate Review Commons’ innovative approach to scientific review and publishing. We thank the reviewers for their kind words regarding the manuscript overall quality and for highlighting the quantitative approach and reproducibility of this work. We further thank the concerns raised and suggestions made that have contributed to improving the manuscript. Below is a point-by-point response to the reviewers, organized into sections that discriminate the alterations already made and plans for further experiments and revisions. We hope that they appropriately address the reviewers' concerns.

2. Description of the planned revisions

Reviewer #2: Figure 6 in particular, the number of analyzed embryos is small, given the fact that there is a lot of inter-individual heterogeneity in this process it could well be that the authors got, by chance, two embryos out of three having the same pattern of Hairy1 expression.

R: The authors appreciate the concern raised by Reviewer#2. This experiment is very time consuming and difficult to execute, which is why the number of samples is limited. Overall, we analyzed 7 embryos and 5 recapitulated the pattern of gene expression. We were particularly interested in the occipital somites and, in this time window, 3 out of 4 showed the same expression pattern. Nevertheless, further experiments will be performed to increase the number of analyzed individuals. We are confident this will contribute to strengthening the conclusions of our work.

Reviewer #2: I believe that an additional shorter time point (+15 or 30 min) with a different pattern of the oscillatory gene would also add to the characterization of the dynamics (same for Fig 5c). This is particularly true given that the domain of expression of Hairy 1 analyzed in Figure 6 is localized quite rostral which might be interpreted as a phase 1 or a phase 2 as well (as initially described in Palmeirim et al 1997).

R: We thank Reviewer#2 for this suggestion. We have preliminary data (20-40 min) evidencing different patterns of expression and will perform more experiments to complement these results. Figures will be modified to include samples incubated for shorter time intervals, to evidence different expression patterns obtained in these conditions.

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

Reviewer #1 | Major points:

1. In the RESULTS session, "Occipital somites are formed faster than cervical and trunk somites," the authors argue that the occipital somites form with greater temporal variability than the neck and trunk somites. Judging from Figures 3C and 3D, I feel it is the case. However, the authors should demonstrate it through statistical analysis.

2. In the RESULTS session, "Anterior-posterior length of rostral somites," the authors concluded that "the large variability in measurements of somites 17-20 most probably results from the rotation of the embryo body in these developmental stages." Probably they mentioned data of the length of #17-#20 somites in Table1. They should demonstrate it through statistical analysis to show the large variability in the specific area. I understand that embryo rotation could be a reason for the variability. The authors should show evidence. Or they should discuss various possibilities from a broad perspective.

R: The authors thank Reviewer#1 for suggesting a statistical approach to better characterize the data variability obtained. We performed a Brown-Forsythe test and found that, indeed, there is a statistically significant difference between the temporal variability (period) of somites 1-7 and 8-20 (P-value = 0.02319). Application of the Brown-Forsythe test also found a non-equal variability in the length of the somites #17-20 (P-value = 0.005403). A new Supplementary Figure 4 was added, displaying these results.

The Brown-Forsythe test is a statistical method used to assess the equality of variances in a dataset across different groups, in this case, the period or the length of early and late somites. It is a robust alternative to the traditional Levene's test, particularly useful when the assumption of homogeneity of variances is not met, such as when the data distributions are skewed or contain outliers (which is the case with our data). It calculates the absolute deviations of individual observations from their respective group medians, which makes it less sensitive than the Levene's test to extreme values. By comparing these deviations between groups, the Brown-Forsythe test helps determine whether the variance differs significantly across the groups. We are confident that this result confers robustness to our claims, and hope that it appropriately addresses the reviewer's concerns.

Regarding the reasons underlying the variability in the length of #17-#20 somites, we believe it is mainly due to technical constraints. In early developmental stages the chicken embryo is flat, and measurements are easily performed along the anterior-posterior (A-P) somite axis. When somites 17-20 are formed, the embryonic axis starts undergoing rotation, meaning that in some cases we may be measuring along a rotated somite axis. In our work somite length is determined as soon as the posterior intersomitic cleft is formed, so an alternative explanation could be that each somite is formed with a variable length, that is soon after consolidated, resulting in the characteristic consistent metameric organization of somites along the embryo body axis. This is highly unlikely because the length of somites 17-20 long after they are formed (Herrmann et al., 1951) is within the same value range we observed.

The manuscript has been altered to include the above-mentioned information, as follows:

• Methods section, under Statistical analysis (Line 153): “To assess the homogeneity of variances between early and late somites, we applied the Brown-Forsythe test on both the period and length measurements. This method involves computing the absolute deviations of individual observations from their respective group medians, rendering it less sensitive to extreme values (outliers). Through the comparison of these deviations across groups, the Brown-Forsythe test aids in determining the statistical significance of variance disparities.”
• Results section, under Anterior-posterior length of rostral somites (Line 201): “A larger variability was obtained for measurements of somites 17-20 (Supplementary Figure 4A), although this most probably results from the rotation of the embryo body in these developmental stages, hindering precise length measurements of the somite A-P axis.”
• Results section, under Occipital somites are formed faster than cervical and trunk somites (Line 216): “Remarkably, there is substantial variability in the time of formation of the early-most somites (Figure 3C; Supplementary Figures 3), which gradually stabilizes until somite 8 onwards, where both somite formation time and variability is equivalent to that observed for somites 15-20 (Figure 3C, D; Table 1; Supplementary Figure 4B).”

1. The authors do not describe the expression patterns of hairy1 in the PSM in the manuscript, but they merely judged whether they are different or the same (recapitulate). The description of the expression pattern needs to be revised totally. The authors should describe the expression patterns of hairy1 in the PSM of each sample carefully and in detail. Fortunately, the previous report (Pourquie and Tam, Developmental Cell, 1, 619-620, 2001) categorized the expression patterns of the EC genes into three phases. The authors should at least categorize each sample according to the criterion by Pourque and Tam. If arrows of brackets indicate the area of expression, it is reader-friendly.

R: A thorough characterization of segmentation clock gene expression (including hairy1) in the PSM of early somitogenesis chick embryos has been previously described (Rodrigues et al, 2006). For this reason, the authors focused mainly on a comparative analysis of the hairy1 expression patterns obtained in explants incubated for different periods of time. The authors acknowledge, however, the need for further description of the expression patterns obtained in early gastrulation stages, which haven’t been previously documented. Overall, the following alterations were made to the manuscript text:

Line 231: “The regions with greater variability of hairy1 expression included the neural plate, anterior to the node, the epiblast posterior to the node encompassing the precursors of the paraxial mesoderm (Psychoyos & Stern, 1996) and the caudal-most epiblast. hairy2 expression was also very dynamic along the embryo A-P axis (n=20) (Figure 4B), evidencing chevron-like expression domains, that appear at different levels of the primitive streak, as previously described by Jouve and collaborators (Jouve et al., 2002).”

Line 245: “As somitogenesis takes place, hairy1 and hairy2 expression patterns retain their dynamic properties in the PSM (Figure 5A, B), as previously described (Rodrigues et al., 2006).”

The authors further thank Reviewer#1’s suggestion to include brackets indicating the areas of hairy1 expression. Figures 4, 5 and 6 have been altered accordingly, which, indeed, makes figure interpretation more reader-friendly. The gene expression phases presented in Palmeirim et al (1997) and then by Pourquié and Tam (2001) mean to summarize a dynamic expression, with continuous intermediate phases, making it difficult to clearly categorize each pattern obtained. Since our purpose was to evaluate if the entire expression pattern was recapitulated (irrespective of the specific phase), we believe that categorizing each sample in phases is not paramount for result interpretation.

Reviewer #1 | Minor points.

1. In the RESULTS session, "Anterior-posterior length of rostral somites," the authors described that the average length of somites ranges from 118 - 191 μ__m. But according to Table-1, the lowest length of somite is 115.92 μ__m (__∼__116 μ__m). So, the lower limit should be corrected here.

R: The authors thank the reviewer for pointing this out. The text has been appropriately modified in Line 199 of the revised manuscript.

1. In the DISCUSSION section, the authors mentioned that they presented a thorough characterization of the size and time of formation of the first ten somites in the chicken embryo. But based on the tables and figures, it will only be the first nine somites, not the ten.

R: The authors agree with the reviewer’s comment. The text has been appropriately modified in Lines 128, 176 and 279 of the revised manuscript.

1. In the GRAPHICAL ABSTRACT, if the color of the oscillation line is the same as the corresponding somites, it is intuitive.

R: The authors thank the reviewer for this suggestion and have modified the graphical abstract accordingly. We employed multiple shades of the same color (corresponding to different positions along the body axis) to represent different embryos.

1. It would be helpful if the manuscript contained both page and line numbers.

R: Page and line numbers were added to the revised manuscript.

Reviewer #2

Minor comment: In 5 c similar patterns and newly formed somites should be pointed out by arrows on the figure to help the readers.

R: We thank Reviewer#2 for this suggestion. Arrows and brackets have been added to the figure to highlight the newly formed somites and gene expression domains, respectively.

Minor comment: To be more specific the term segmentation clock should be used instead of embryonic clock as I believe there are other embryonic clocks (cell cycle, circadian, etc..)

R: The authors appreciate the suggestion of Reviewer#2 regarding the term used to identify the molecular oscillator in our work. The term “segmentation clock” or “somitogenesis clock” is commonly used to refer to oscillations in hairy1/2 gene expression because their discovery and subsequent study has mainly focused on the somitogenesis process. Oscillations of hairy1/2 expression (Hes1/7 in mouse), however, have also been described in cells and developmental stages that are not associated with somite formation, and herein we describe dynamic expression in epiblast regions containing precursors that don’t give rise to segmented structures. As discussed in our recent paper (Carraco et al., Front. Cell Dev. Biol, 2022), the broader term Embryo Clock may be used to refer to molecular oscillations in embryonic cells, controlled by negative feedback regulation, that play a role in temporally controlled morphogenetic processes and/or cell fate specification.

In the beginning of our manuscript (Line 75), we clearly state that we are referring to the embryo clock operating during somitogenesis: “(…) somitogenesis embryo clock (EC), comprising genes with cell-autonomous oscillatory expression in the PSM driven by negative feedback loops (reviewed in Carraco et al., 2022)”, so we believe that the term used will be clearly perceived by the reader. For further clarification, however, the subtitle The Embryo Clock in early somitogenesis in the Discussion section has been modified to (Line 321): “The Embryo Segmentation Clock in early somitogenesis”

Reviewer #3 | Minor comments:

The authors did not consider the fact that the first formed somite is the second somite. After the formation of the second somites, the real first somite forms anterior to the second somite. Furthermore, the real first and the third somite seems to be formed simultaneously. It is worthy for the authors to re-examine the data, whether the real first somite and the third somite are formed at the same time. And to check whether the first somite was counted to the segmented region. And this point should be at least discussed.

R: The authors thank Reviewer#3 for the opportunity to clarify this important issue in our manuscript. It was previously described that the first morphological somite formed is, in fact, the second somite, while the “real” first somite is formed later, anteriorly to this one (Hamburger and Hamilton, 1951). This rostral-most somite-like structure is not anteriorly delimited by a fissure and has thus been termed an “incomplete” or “rudimentary” somite (Hinsch & Hamilton, 1956). Since the methodology used in our work relies on measuring the length between the rostral-most and the posterior-most intersomitic clefts, the “rudimentary” somite is not included in our data, and we considered somite #1 as the first somite delimited both anteriorly and posteriorly by intersomitic clefts. This was stated in the Methods section, under Embryo measurements, and has now also been made explicit under Somite nomenclature (Line 120): “Only structures delimited both anteriorly and posteriorly by intersomitic clefts were counted as somites.”

We, indeed, observe the formation of the “rudimentary” somite anteriorly to somite #1, when somites 3-4 and formed. This information was included in the Discussion section, under Spatio-temporal properties of the rostral somite segmentation (Line 311): “Note that our analysis did not consider the “rudimentary somite”, as defined by Hinsch and Hamilton (Hinsch & Hamilton, 1956) since it does not possess an anterior somitic cleft. We found that this structure becomes clearly visible, rostrally to somite 1, as somites 3-4 are formed.”

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

Reviewer #1 (Significance (Required)):

General assessment: The results are not conceptually new or surprising. However, their careful quantitative analysis is informative and worthwhile to be published because it has yet to be done in the early somite formation. In this manuscript, the authors focused on the early stage of somitogenesis. It will be more informative if they complete this analysis for the whole somite area including the end of somitogenesis.

Reviewer #3 | Referees cross commenting****

I find all criticisms are justified. The most advance, as stated by other both reviewers, is the quantitative assay of the somite formation, since this is no yet done previously. As suggested by the first reviewer, it will be more informative if the authors complete this analysis for all regions. For all somites would be too much work, but they can select some representative somites of each region in addition to occipital region, such as 3 somites for one region, including the cervical, thoracic, lumbal, sacral and caudal region. Thus, the dynamic of the temporal somite formation of the whole embryo can be analysed using the same method. This will provide much more impact for this work.

R: The authors thank the Reviewers #1 and #3 for the kind words and for highlighting the quantitative approach taken. Regarding completing the analysis for the whole somite area including the end of somitogenesis, the authors agree that this would be interesting for the community. The focus of this work, however, was a detailed understanding of early somite segmentation, where measurements of somites 14-20 were performed for validation purposes alone of the technical approach developed, since their time of formation has been previously well established. Characterization of somite formation dynamics along the entire embryonic axis, while informative, would entail significant technical challenges, which are beyond the focus of this work. Briefly, we performed live imaging using the EC culture system (Chapman et al, 2001). This appropriately reproduces in ovo development of early embryos but imposes significant constraints on embryo development in older developmental stages, including the ones corresponding to the formation of the last somites. A possible alternative to perform these measurements would be to apply the tissue explant culture system developed by Palmeirim et al., 1997 to different portions of the embryo body, and optimize it for real-time imaging. However, we believe that the time and effort required are beyond the scope of this work and would not significantly contribute to elucidating the main questions addressed in this manuscript.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Somites form consecutively along the anterior to posterior (AP) axis. The time of the formation of a somite is controlled by the segmentation clock, oscillation of cyclic genes in the presomitic mesoderm. The length of an oscillation cycle differs between species and should also differ between the axial levels. In chicken embryos, one cycle for a trunk somite requires 90 minutes, while it is much slower (150 minutes) for a posterior-most somite. Is this quicker or slower for an anterior-most somite? Andrade' group addressed this question and measured the time of the formation of each occipital somites (somite 1-5). They found that the formation of an occipital somite requires only 75 minutes, while somites from somite 6 onwards takes as long as the trunk somites (about 90 minutes). The faster formation of occipital somites is correlated with the time of the cyclic expression of hairy1 and hairy2.

Major comments:

The conclusion is well supported by the data. The measurement of the length increments of the segmented region and then assay using algorithm are well established. Thus, the data are well reproducible.

Minor comments:

The authors did not consider the fact that the first formed somite is the second somite. After the formation of the second somites, the real first somite forms anterior to the second somite. Furthermore, the real first and the third somite seems to be formed simultaneously. It is worthy for the authors to re-examine the data, whether the real first somite and the third somite are formed at the same time. And to check whether the first somite was counted to the segmented region. And this point should be at least discussed.

Referees cross commenting

I find all criticisms are justified. The most advance, as stated by other both reviewers, is the quantitative assay of the somite formation, since this is no yet done previously. As suggested by the first reviewer, it will be more informative if the authors complete this analysis for all regions. For all somites would be too much work, but they can select some representative somites of each region in addition to occipital region, such as 3 somites for one region, including the cervical, thoracic, lumbal, sacral and caudal region. Thus, the dynamic of the temporal somite formation of the whole embryo can be analysed using the same method. This will provide much more impact for this work.

Significance

Significance: The measurement of the length increments of the segmented region and then assay using algorithm are the novelty and strengths of this study. So, the data are reproducible and objective.

The results of this study extend our understanding about the dynamic process of the somitogenesis. Especially, the most interesting point is that based on this result, we can see that the segmentation clock runs faster in the head region, and then slow down gradually along the AP axis.

Audience: specialized, basic research<br /> The developmental biologist will be interested in this topic.

My expertise is the somite development, somite differentiation, mesoderm development.

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, Maia-Fernandez et al used time-lapse imaging of chicken embryos to analyze the formation of the first formed somite, by doing in situ hybridization, they checked the dynamic nature of well-known segmentation clock genes during this process. They found that the segmentation clock period is faster for the formation of the first five somites (most anterior) and that this process is underlain by dynamic/cyclic expression of Hairy 1 and Hairy 2 as it has been described for more posterior somites.

• I believe that there few issues that should be addressed to strengthen the conclusions of the manuscript:<br /> Figure 6 in particular, the number of analyzed embryos is small, given the fact that there is a lot of inter-individual heterogeneity in this process it could well be that the authors got, by chance, two embryos out of three having the same pattern of Hairy1 expression.
• I believe that an additional shorter time point (+15 or 30 min) with a different pattern of the oscillatory gene would also add to the characterization of the dynamics (same for Fig 5c). This is particularly true given that the domain of expression of Hairy 1 analyzed in Figure 6 is localized quite rostral which might be interpreted as a phase 1 or a phase 2 as well (as initially described in Palmeirim et al 1997).

Minor comments:

In 5 c similar patterns and newly formed somites should be pointed out by arrows on the figure to help the readers.<br /> To be more specific the term segmentation clock should be used instead of embryonic clock as I believe there are other embryonic clocks (cell cycle, circadian, etc..)

Significance

In this study, the authors address the question of the formation of the first-formed somites using bird a model system; this is a conceptual advance in the sense that our knowledge of the dynamics of these critical morphological events is minimal. The technical advances (time-lapse, image analysis, dissection) made by the authors are quite remarkable and allow for filling the gap of knowledge the community has in this particular domain. The article is well written, data are well presented and it is interesting for a large community of developmental biologists.

My expertise is in cell and tissue morphogenesis of amniote embryos

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

Maia-Fernandes et al. investigated somite formation dynamics in the chick embryo's early stage in this manuscript. They found that the cranial most somites (1-5) form faster than the trunk. They also show that the oscillatory expression pattern of hairy1, regarded as the somitogenesis embryo clock (EC), is coupled to the somite segmentation in the occipital somites. The results are not conceptually new or surprising; they merely show what has been widely believed. However, their careful quantitative analysis is informative and worthwhile to be published because it has yet to be done in the early somite formation. To improve the manuscript, I have several concerns to be addressed.

Major points.

1. In the RESULTS session, "Occipital somites are formed faster than cervical and trunk somites," the authors argue that the occipital somites form with greater temporal variability than the neck and trunk somites. Judging from Figures 3C and 3D, I feel it is the case. However, the authors should demonstrate it through statistical analysis.
2. The authors do not describe the expression patterns of hairy1 in the PSM in the manuscript, but they merely judged whether they are different or the same (recapitulate). The description of the expression pattern needs to be revised totally. The authors should describe the expression patterns of hairy1 in the PSM of each sample carefully and in detail. Fortunately, the previous report (Pourquie and Tam, Developmental Cell, 1, 619-620, 2001) categorized the expression patterns of the EC genes into three phases. The authors should at least categorize each sample according to the criterion by Pourque and Tam. If arrows of brackets indicate the area of expression, it is reader-friendly.
3. In the RESULTS session, "Anterior-posterior length of rostral somites," the authors concluded that "the large variability in measurements of somites 17-20 most probably results from the rotation of the embryo body in these developmental stages." Probably they mentioned data of the length of #17-#20 somites in Table1. They should demonstrate it through statistical analysis to show the large variability in the specific area. I understand that embryo rotation could be a reason for the variability. The authors should show evidence. Or they should discuss various possibilities from a broad perspective.

Minor points.

1. In the RESULTS session, "Anterior-posterior length of rostral somites," the authors described that the average length of somites ranges from 118 - 191 μm. But according to Table-1, the lowest length of somite is 115.92 μm (∼116 μm). So, the lower limit should be corrected here.
2. In the DISCUSSION section, the authors mentioned that they presented a thorough characterization of the size and time of formation of the first ten somites in the chicken embryo. But based on the tables and figures, it will only be the first nine somites, not the ten.
3. In the GRAPHICAL ABSTRACT, if the color of the oscillation line is the same as the corresponding somites, it is intuitive.
4. It would be helpful if the manuscript contained both page and line numbers.

Significance

General assessment: The results are not conceptually new or surprising. However, their careful quantitative analysis is informative and worthwhile to be published because it has yet to be done in the early somite formation. In this manuscript, the authors focused on the early stage of somitogenesis. It will be more informative if they complete this analysis for the whole somite area including the end of somitogenesis.

Advance: Previously no one provided quantitative data for somite formation. In this viewpoint, this manuscript has an advantage.

Audience: Their data could be helpful in the generation of mathematical models.

My field: Developmental Biology

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1:

1. The authors claim PA-induced mTORC1 activation is dependent on acetyl-CoA derived from mitochondrial fatty acid oxidation. Using isotope tracing to determine the contribution of PA to acetyl-CoA, would improve.

Response: We thank the reviewer for this valuable comment. As suggested, we measured the contribution of PA to the total cellular acetyl-CoA pool using metabolic flux assays. The results showed that the incorporation of 13C from [U-13C16]-palmitate into acetyl-CoA was exceeding 60% (Page 40, Figure 4D in the revised version), indicating that exceeding 60% of the acetyl-CoA pool was PA derived. Likewise, we also found that the incorporations of 13C from [U-13C16]-palmitate into 6:0-CoA, 8:0-CoA, 10:0-CoA and 12:0-CoA were all exceeding 50% (Page 40, Figure 4D in the revised version). Thus, these results suggested that a large portion of the PA was used for fatty acid oxidation upon entering the cell. Moreover, we found that fatty acid β oxidation blocked by perhexiline maleate inhibited PA-induced increase of acetyl-CoA, suggesting that the induction of acetyl-CoA content was largely dependent on the fatty acid oxidation of PA. Furthermore, we also demonstrated that inhibition of mitochondrial fatty acid β oxidation by pharmacological inhibitor or genetic knockdown abrogated PA-induced activation of mTORC1. However, using sodium acetate treatment to elevate cellular acetyl-CoA levels rescued impaired mTORC1 activity induced by the inhibition of fatty acid β oxidation under PA condition. Together, these results revealed that PA-induced mTORC1 activation is dependent on acetyl-CoA derived from mitochondrial fatty acid oxidation.

1. They showed PA increases fatty acid oxidation related gene expression and acetyl-CoA level, while OA and LA could not. why only PA could increases fatty acid oxidation and acetyl-CoA level, considering both of these lipids could be oxidation in mitochondria? Is there any differences in mitochondria among treatment of PA, OA and LA? It is better to monitor fatty acid oxidation in real time using seahorse. And add discussions.

Response: We thank the reviewer for this constructive question. As suggested, we performed the seahorse real-time cell metabolic analysis and the results showed that PA treatment enhanced mitochondrial OCR and elevated maximal oxygen consumption rates compared with OA or LA treatment in fish myocytes (Page 37, Figure 3B in the revised version). Likewise, we also found that PA-induced increase of fatty acid oxidation-related gene expressions was more robust than OA or LA in vivo and in vitro. Thus, these results indicated that the induction of mitochondrial fatty acid oxidation by pa treatment was stronger than OA or LA treatment.

In this study, using LC–MS, we showed that PA treatment increased the contents of short/medium-chain acyl-CoA and acylcarnitine in comparison with OA or LA treatment. Thus, these results suggested that although all three fatty acids can be oxidation in mitochondria, PA may be preferred to enter the mitochondria for fatty acid β oxidation, compared with OA or LA. Previous studies have found that OA is more inclined to synthesize triglycerides to induce the formation of lipid droplets than PA (Chen et al., 2023; Plötz et al., 2016). Likewise, we also found that OA significantly increased the contents of 18:1-CoA in comparison with PA. Thus, we speculate that, after entering the cell, OA is more preferentially synthesized to triglyceride for storage than fatty acid oxidation. Moreover, LA is considered to be a precursor of arachidonic acid, and can be converted to a myriad of bioactive compounds called eicosanoids (Whelan & Fritsche, 2013). Similarly, we found that LA markedly elevated the contents of 18:2-CoA/18:3-CoA. Thus, we conjecture that LA preferentially synthesizes functional lipids compared to entering mitochondria for fatty acid oxidation. Together, differences in the levels of acetyl-CoA produced by these three fatty acids may be related to their metabolic pathway preferences.

There may be two reasons for why PA prefers to enter mitochondrial for fatty acid oxidation. On one hand, due to differences in the structure of PA, OA and LA, the substrate affinity of CPT1B to these fatty acyl-CoAs may be different, that may contribute to the different rates of fatty acid to enter into mitochondria. On the other hand, in contrast to the β-oxidation of SFAs, the β-oxidation of UFAs requires the involvement of 2,4-dienoyl-CoA reductase (You et al., 1989), and thus the β-oxidation of SFAs may be more efficient.

At present, the understanding of differences in fatty acid oxidation between SFAs and UFAs is insufficient, so more studies are needed in the future to further explore the underling mechanisms behind these differences. The reviewers have raised a very important direction for research, and so we will continue to address this issue in future.

We have expanded this section of the Discussion (Page 14, line 388-412 in the revised version).

1. The authors present lots of western blot images, suggest to provide quantification data of these blots.

Response: We thank the reviewer for their careful assessment of our study. We apologize for not providing quantification data for western blot images in our initial manuscript. To support our conclusions, we have now added a densitomentric and statistical analysis of all western blots in the revised version (Page 33, Figure 1 in the revised version; Page 35, Figure 2 in the revised version; Page 37, Figure 3 in the revised version; Page 40, Figure 4 in the revised version; Page 43, Figure 5 in the revised version; Page 45, Figure 6 in the revised version; Page 47, Figure 7 in the revised version; Page 51, Figure S1 in the revised version; Page 53, Figure S2 in the revised version; Page 54, Figure S3 in the revised version; Page 56, Figure S4 in the revised version; Page 57, Figure S5 in the revised version).

1. It is better to discuss the relationship between fatty acid oxidation and mTOR signaling.

Response: The reviewer’s comments are very valuable. We apologize for not discussing enough for the relationship between fatty acid oxidation and mTOR signaling in our initial manuscript. We have now expanded this section of the Discussion (Page 13, line 366-387 in the revised version, see below).

Growing lines of evidence suggested a strong link between mitochondrial fatty acid oxidation and mTORC1 signaling (Ricoult & Manning, 2013). As a central regulator of anabolism, mTORC1 is considered to inhibit fatty acid β oxidation pathway for energy storage or ketogenesis (Aguilar et al., 2007). Several studies revealed that restrained mTORC1 by rapamycin induced fatty acid β oxidation in rat hepatocytes through increasing expression of fatty acid β oxidation related enzymes (Brown et al., 2007; Peng et al., 2002). Likewise, mice with whole-body knockout of S6K1 showed enhanced fatty acid β oxidation and increased expression levels of CPT1 in isolated adipocytes (Um et al., 2004), and S6K1/S6K2 double-knockout mice also exhibited elevated fatty acid β oxidation of fatty acids in isolated myoblasts by activating AMPK (Aguilar et al., 2007). Furthermore, a recent study has established that FOXK1 can mediate the inhibition of fatty acid β oxidation by mTORC1 (Fujinuma et al., 2023). Thus, these collective data revealed that fatty acid β oxidation was restrained by mTORC1. However, conversely, the role of mitochondrial fatty acid oxidation in the regulation of mTORC1 is still controversy. A study in prostate cancer cells suggested that inhibited fatty acid β oxidation by etomoxir reduced mTORC1 activity (Schlaepfer et al., 2014), and another study found that deleting CPT1B specifically in skeletal muscle of mice suppressed mTORC1 by provoking AMPK activation (Vandanmagsar et al., 2016). Consistent with these studies, our results showed that acetyl-CoA derived from mitochondrial fatty acid β oxidation induced mTORC1 activation under PA treatment, indicating that acetyl-CoA may be a novel insight linking fatty acid β oxidation and mTORC1 signaling. Paradoxically, unlike other studies, a recent study found that mice with heart-specific CPT2-deficient exhibited induction of mTORC1 pathway. Thus, the effects of fatty acid β oxidation on mTORC1 pathway are complicated and may differ under variable physiological and pathological conditions. Further studies are needed to determine the sophisticated mechanisms underlying the regulation of fatty acid β oxidation on mTORC1 signaling.

Reviewer #2:

Major comments:

Initial experiment: Among several fatty acid-rich diets, fish were fed a palmitic acid (PA) rich (PO) diet for 10 weeks, and the PO diet significantly raised fasting blood glucose levels compared to control diet (fish oil of equal lipid content). The PO diet also impaired the fish's glucose and insulin tolerance. The PO diet also led to decreased phosphorylation levels of AKT, which regulates glucose metabolism. Therefore, the researchers initially concluded that a palmitic acid-rich diet leads to systemic insulin resistance in fish.

1. I have a couple of questions on this initial experiment on which all the subsequent studies are based. In Figure 1A, the body weight was identical in control and PO group. Don't you expect PO feeding lead to obesity in fish, as HFD induces obesity in mice?

Response: We thank the reviewer for this constructive question. In our study, we found that dietary PO diet for 10 weeks failed to affect the body weight of fish, compared with CON diet. Similar to our results, a study in human also found that there was no significant differences in the body weight and body mass index (BMI) between saturated fat diet and monounsaturated fat diet (Vessby et al., 2001). Unlike high-fat diet, the lipid content level of PO diet was not elevated, but only the fatty acid composition was altered, with palmitic acid composition being significantly increased in comparison with CON diet (Page 60, Table S1). Thus, this may be the reason of why the PO diet did not induce weight gain.

Although accumulating evidence showed that the onset of insulin resistance was often accompanied by weight gain and obesity (Kahn & Flier, 2000; Shoelson et al., 2007), some studies also found that insulin resistance occurred without obesity. A recent study found that mice with liver knockout of Lpcat3 exhibited improved insulin sensitivity without a change in the body weight (Tian et al., 2023). Moreover, another study in mice showed that dietary phenylalanine-rich diet induced insulin resistance, but had no effects on the body weight (Zhou et al., 2022). Likewise, our study also found that dietary PO diet provoked systemic insulin resistance, while did not affect the body weight in fish. Thus, these studies indicated that the development of insulin resistance may not always be entirely accompanied by obesity.

1. Figure 1G and 1H show glucose and insulin tolerance after PO feeding for 10 weeks. The area under curve (AUC) should be compared to determine if GTT and ITT were statistically different. The ITT curve is particularly interesting as the control fish did not seem to respond to insulin, while the PO-fed fish responded more robustly. The only difference is the initial glucose level. Are the GTT and ITT done after fasting? How long is the fasting? The curves suggest that even though PO increased (fasting) blood glucose levels, it improved insulin sensitivity - therefore the premise that PO induces insulin resistance is not supported here. The lack of insulin induced response in the control group is worrisome. I suggest that the measures should be retaken, and AUC should be used to support if there are any differences in GTT and ITT.

Response: We thank the reviewer for this valuable comment. We apologize for making this confusion in the initial manuscript and we thank the reviewer for providing this opportunity to correct our manuscript. As suggested, to further investigate whether dietary PO could cause impairment of insulin sensitivity, we have re-performed the GTT and ITT assays. Considering that fish have a poor capacity to utilize glucose, we extended the assay time to 8 h. To make the results more accurate, we also added the biological replicates. Moreover, before injection of glucose or insulin, fish were fasted for 24 h. Furthermore, we added area under curve (AUC) of GTT and ITT, and performed statistical analyses of the AUC data.

Our results showed that dietary PO diet reduced glucose tolerance and insulin tolerance in fish (Page 33, Figure 1G and 1H in the revised version). Moreover, compared with CON diet, the AUC of GTT and ITT were significantly increased in PO diet (Page 33, Figure 1G and 1H in the revised version). Similarly, we found that dietary PO diet elevated fasting blood glucose levels and plasma insulin concentrations. Furthermore, we showed that dietary PO diet decreased the phosphorylation levels of AKT in the liver and skeletal muscle. In addition, we also demonstrated that PA treatment could induce cellular insulin resistance in fish myocytes and C2C12 myotubes. Thus, in our opinion, the above results could indicate that dietary PO induced insulin resistance in fish.

1. Based on the assumption that PO induces IR (which needs to be confirmed based on the previous comments), the researchers attempted to understand how PA triggers IR through a series of experiments, predominantly western blot analysis. All the Western blots should be quantified. The model is that PA activates FAO in mitochondrial that elevates cytosolic acetyl-coA, which acetylates Rheh to activate mTORC1. mTORC1 on one hand alters IRS1 phosphorylation and on the other hand inhibits transcriptional activity of TFEB to reduce Irs1 mRNA level. Together reduces IRS1 leads to Insulin Resistance.

Response: The reviewer’s comments were very important to verify the validity of our findings. We have now added a densitomentric and statistical analysis of all western blots in the revised version (Page 33, Figure 1 in the revised version; Page 35, Figure 2 in the revised version; Page 37, Figure 3 in the revised version; Page 40, Figure 4 in the revised version; Page 43, Figure 5 in the revised version; Page 45, Figure 6 in the revised version; Page 47, Figure 7 in the revised version; Page 51, Figure S1 in the revised version; Page 53, Figure S2 in the revised version; Page 54, Figure S3 in the revised version; Page 56, Figure S4 in the revised version; Page 57, Figure S5 in the revised version).

1. Figure 1. PA reduces basal and insulin stimulated AKT phosphorylation in fish liver and muscle, as well as in culture fish and murine myocytes (Fig. 1I-M). The results appear to be solid but need to be quantified.

Response: We thank the reviewer for this kind suggestion. We have now added a densitomentric and statistical analysis of all western blots in Figure 1 (Page 33, Figure 1 in the revised version).

1. Figure 2 shows that PA provoked hyperactivation of mTORC1 (indicated by elevated phosphorylated S6K levels. This effect was abolished by Rapamycin treatment (an mTORC1 inhibitor) and also abolished by insulin stimulation (2F). Again, the western blots should be quantified.

Response: We thank the reviewer for this excellent suggestion. We have now added a densitomentric and statistical analysis of all western blots in Figure 2 (Page 35, Figure 2 in the revised version).

1. Figure 6: the researchers measured the effect of PA treatment on IRS1 phosphorylation in order to understand the mechanism of insulin resistance induced by mTORC1 activation under PA treatment. A PO diet intensified S636/S639 phosphorylation in fish muscle. In fish myocytes and C2C12 myotubes, PA treatment elevated S636/S639 phosphorylation but decreased the Y612 phosphorylation of IRS1 in a dose-dependent manner. Treatment of fish myocytes and C2C12 myotubes with an mTOR inhibitor blocked increased IRS1 S636/S639 phosphorylation levels under PA treatment. Also, PA specifically reduced mRNA levels of Irs1. This indicates that PA-induced, mTOR-dependent alteration of IRS1 phosphorylation and transcription may have contributed to insulin resistance. It is unclear how mTORC induces either increase or decrease in IRS1 phosphorylation depending on the residuals.

Response: We appreciate the reviewers for this important question. In fact, previous studies have clearly explored how mTORC1 pathway affects S636/S639 phosphorylation of IRS1. On one hand, as a kinase complex, mTORC1 could directly induce S636/S639 phosphorylation of IRS1 in vitro (Ozes et al., 2001). On the other hand, mTORC1 could activate S6K to promote S636/S639 phosphorylation of IRS1 (Shah & Hunter, 2006; Um et al., 2004). In addition, considering that the serine/threonine phosphorylation status of IRS has been shown to affect its tyrosine phosphorylation and protein degradation (Copps & White, 2012), we speculate that the decrease of Y612 phosphorylation of IRS1 is dependent on the induction of IRS1 S636/S639 phosphorylation.

In this study, we found that PA could induce S636/S639 phosphorylation of IRS1 in a mTORC1-dependent manner. Considering that previous studies have explored the mechanism by which mTORC1 induced IRS1 S636/S639 phosphorylation, we did not conduct further studies on this issue. Notably, we found that mTORC1 could also regulate the transcription of IRS1, so we subsequently investigated the mechanism by which mTORC1 inhibited IRS1 transcription.

1. Figure 7 shows that PA inhibits nuclear translocation of TFEB to suppress IRS1 transcription. The EMSA in 7D is not convincing.

Response: We thank the reviewer for this valuable comment and we apologize for providing unclear blots in the initial manuscript. To support our conclusions, we have now re-performed the EMSA assays. The results suggested that TFEB can directly bind to the IRS1 promoter at these two sites (Page 47, Figure 7D in the revised version).

Minor comments:

1. Some data appears to weaken the results and/or contradictory. For example, the paper initially showed reduced AKT phosphorylation to support PA induced IR, but shouldn't a lower level of pAKT reduces mTORC activation? But then the rest of the manuscript explores how PA activates mTOR. Part of the IR is manifested by impaired mTORC1 activation, yet the PA activates mTORC1. The authors should present the rationale and flow of the ideas in a better way.

Response: We thank the reviewer for this excellent suggestion. We appreciate the points that in some insulin resistance conditions, as a downstream of the insulin pathway, mTORC1 activity is manifested to be inhibited. However, mTORC1 activity showed different under other insulin resistance conditions.

In fact, multiple negative feedback signals exist in cells to maintain cellular homeostasis under diverse environmental challenges and stimulations (Kearney et al., 2021). However, aberrant of negative feedback can lead to impaired intracellular signaling pathway and induce a variety of diseases (Nguyen & Kholodenko, 2016). Similarly, numerous negative feedback mechanisms also exist in insulin signaling to prevent the development of cancers that may be induced by hyperactivation of insulin pathway. The negative feedback of insulin pathway is mainly mediated by mTORC1, which has been found to inhibit insulin signaling transduction by directly or indirectly affecting IRS1 phosphorylation (Copps & White, 2012; Shah & Hunter, 2006; Um et al., 2004). However, under some pathological or stress conditions, mTORC1 is over-activated, resulting in the amplification of the negative feedback of insulin pathway and the development of insulin resistance. A recent study found that imidazole propionate, a metabolite produced by the gut microbiota, provoked insulin resistance through inducing mTORC1 activation and phosphorylation of IRS1 (Koh et al., 2018). Other studies also showed that elevated abundance of branched-chain amino acids (BCAAs) or branched-chain α-keto acid (BCKA) could cause insulin resistance by boosting mTORC1 pathway (Zhou et al., 2019). Thus, mTORC1 activation induced-negative feedback inhibition of insulin pathway may be a critical factor in the development of insulin resistance.

Consistently, our study found that PA could activate mTORC1 in an acetylation modification-dependent manner. Moreover, activation of mTORC1 inhibited the phosphorylation of AKT and caused insulin resistance by affecting the phosphorylation and transcription of IRS1.Indeed, AKT is considered to activate mTORC1 in multiple manners, and inhibition of AKT results in the reduction of mTORC1 activity. However, mTORC1 activity is not only affected by AKT, but is also regulated by a diverse set of upstream signals (Saxton & Sabatini, 2017). Thus, we considered that the activating effect of PA on mTORC1 activity is higher than the negative effect of mTORC1 activity produced by AKT inhibition. This also led to the fact that mTORC1 remained in an activated state despite the inhibition of AKT in PA condition.

1. There are also many run-on sentences and grammar issues, making it very hard to read. The writing can be improved.

Response: We thank the reviewer for this valuable comment and we apologize for these grammar mistakes in the initial manuscript. Following the reviewer’s suggestion, we have invited native speaker to guide the English writing and carefully corrected these run-on sentences and grammar issues. We thank the reviewer for this careful evaluation of our manuscript.

Reviewer #3:

Major issues affecting the conclusions:

1. The conclusions are supported by the data. However, I suggest to perform a densitomentric and statistical analysis of western blots, especially when the authors report a representative blot, showing samples loaded in single.

Response: We thank the reviewer for this excellent suggestion. We agree that it would be important to verify the validity of our findings and we apologize for not providing quantification data for western blot images in our initial manuscript. We have now added a densitomentric and statistical analysis of all western blots in the revised version (Page 33, Figure 1 in the revised version; Page 35, Figure 2 in the revised version; Page 37, Figure 3 in the revised version; Page 40, Figure 4 in the revised version; Page 43, Figure 5 in the revised version; Page 45, Figure 6 in the revised version; Page 47, Figure 7 in the revised version; Page 51, Figure S1 in the revised version; Page 53, Figure S2 in the revised version; Page 54, Figure S3 in the revised version; Page 56, Figure S4 in the revised version; Page 57, Figure S5 in the revised version).

1. The methods are clear and reproducible. The authors should better explain how they have dissolved all the powders (i.e. fatty acids) to obtain the stock solutions next diluted (from what concentration?) in the media

Response: We thank the reviewer for this valuable comment. We apologize for not explaining how to dissolved all the powders in the media. As suggested, we have now provided a detailed explanation of how to dissolved all the powders to obtain the stock solutions in the Methods (Page 19, line 537-586 in the revised version, see below).

For PA, OA or LA in vitro treatment, fatty acid free BSA (Equitech-Bio, USA) was dissolved in FBS-free DMEM at room temperature according the ratio 1:100 (1 g fatty-acid free BSA: 100 ml FBS-free DMEM). 500 mg PA (Merck, Cat#P0500), OA (Merck, Cat#O1008) or LA (Merck, Cat#L1376) was dissolved in 10 ml ethanol to obtain PA, OA or LA stock solution respectively. Then PA, OA or LA stock solution was blow-drying with nitrogen gas and was dissolved in 0.1 M NaOH and warming at 75°C until clear to obtain 100 mM PA, OA or LA solution. Subsequently, 100 mM PA, OA or LA solution was added to 1% BSA solution according the ratio 1:100 (100 mM PA:1% BSA, v/v) at 50°C. Finally, the mixture was filtered using a 0.45 μM filter and stored at -20°C. For insulin in vitro treatment, insulin powder (Merck, USA) was dissolved in hydrochloric acid (pH=2) to obtain 1 mg/ml stock solution. For rapamycin or Torin1 in vitro treatment, rapamycin (Med Chem Express, #HY-10219, USA) or Torin1 (Med Chem Express, #HY-13003, USA) was dissolved in dimethyl sulfoxide (DMSO, Solarbio, China) to obtain 1 mM stock solution respectively. For MHY1485 in vitro treatment, MHY1485 (Med Chem Express, #HY-B0795, USA) was dissolved in DMSO (Solarbio, China) to obtain 10 mM stock solutions.

For etomoxir or perhexiline maleate in vitro treatments, etomoxir (Med Chem Express, #HY-50202, USA) or perhexiline maleate (Med Chem Express, #HY-B1334A, USA) was dissolved in DMSO (Solarbio, China) to obtain 50 mM stock solution respectively. For BMS-303141 treatment, BMS-303141 (Med Chem Express, #HY-16107, USA) was dissolved in DMSO (Solarbio, China) to obtain 25 mM stock solutions. For sodium acetate treatment, sodium acetate (Merck, #S2889, USA) was dissolved in ultrapure water from a Milli-Q water system to obtain 5M stock solution. For C646, spermidine or MB-3 treatment, C646 (Med Chem Express, #HY-13823, USA), spermidine (Med Chem Express, #HY-B1776, USA) or MB-3 (Merck, #M2449, USA) was dissolved in DMSO (Solarbio, China) to obtain 50 mM stock solution respectively. For MG149 treatment, MG149 (Med Chem Express, #HY-15887, USA) was dissolved in DMSO (Solarbio, China) to obtain 150 mM stock solution. For TFEB activator 1 treatment, TFEB activator 1 (Med Chem Express, #HY-135825) was dissolved in DMSO (Solarbio, China) to obtain 10 mM stock solution.

1. Anova analysis should be performed to analyze western blot densitometries.

Response: The reviewer raises an important point and we appreciate this comment. As suggested, we have now added statistical analyses of all western blot densitometries in the revised version. The data are presented as the means ± SEM and were analyzed using independent t-tests for two groups and one-way ANOVA with Tukey’s test for multiple groups (Page 33, Figure 1 in the revised version; Page 35, Figure 2 in the revised version; Page 37, Figure 3 in the revised version; Page 40, Figure 4 in the revised version; Page 43, Figure 5 in the revised version; Page 45, Figure 6 in the revised version; Page 47, Figure 7 in the revised version; Page 51, Figure S1 in the revised version; Page 53, Figure S2 in the revised version; Page 54, Figure S3 in the revised version; Page 56, Figure S4 in the revised version; Page 57, Figure S5 in the revised version).

Minor comments:

Prior studies are referenced appropriately, text and figures are clear. I suggest to add in the abstract all the model systems used. HEK293 also should be inserted in the description of the results. Please add the reference to figure 8 in the text. Please, describe cell origin.

Response: We thank the reviewer for this careful assessment of our study. We apologize for not making this clearer in our initial manuscript. We have now added all the model systems used in the Abstract (Page 2, line 24-28 in the revised version, see below).

Here, using a croaker model, we report that dietary palmitic acid (PA), but not oleic acid or linoleic acid, leads to dysregulation of mTORC1 signaling which provokes systemic insulin resistance and glucose intolerance. Mechanistically, using croaker primary myocytes, mouse C2C12 myotubes and HEK293T cells, we show that PA-induced mTORC1 activation is dependent on mitochondrial fatty acid β oxidation.

Moreover, we have now added the description of HEK293T cells in the Results (Page 10, line 261-265 in the revised version; Page 11-12, line 309-315 in the revised version, see below).

To further investigate whether the regulation of mTORC1 by Tip60 is dependent on the acetylation of Rheb, the interaction between Tip60 and Rheb was analyzed via co-immunoprecipitation assays, and the results showed that Tip60 can interact with Rheb in HEK293T cells (Figure 5E). Moreover, overexpressed Tip60 reinforced the acetylation of Rheb and phosphorylation levels of S6K in HEK293T cells (Figure 5F).

Dual luciferase experiments in HEK293T cells showed that TFEB had the strongest ability to elevate the luciferase activity of the IRS1 promoter among the crucial downstream transcription factors of mTORC1 (Figure 7A). Moreover, TFEB enhanced the promoter activity of IRS1 in a dose-dependent manner (Figure 7B) and mutations of the predicted TFEB binding site 4 and site 6 in the IRS1 promoter significantly reduced the promoter activity of IRS1 in HEK293T cells (Figure 7C). Furthermore, ChIP and EMSA experiments in HEK293T cells verified that TFEB can directly bind to the IRS1 promoter at site 4 and site 6 (Figures 7D and 7E).

As suggested, we have added the reference to figure 8 in the Discussion (Page 17, line 483-487 in the revised version, see below).

In summary, our work unveils an evolutionarily conserved mechanism by which mitochondrial fatty acid β oxidation flux of acetyl-CoA induces mTORC1 activation through enhancing Tip60-mediated Rheb acetylation under PA condition. Subsequently, hyperactivation of mTORC1 boosted serine phosphorylation of IRS1 and inhibited TFEB-mediated transcription of IRS1, leading to insulin resistance (Figure 8).

As suggested, we have added the description of cell origin in the Methods (Page 19, line 526-527 in the revised version; Page 19, line 533-534 in the revised version, see below).

Mouse C2C12 myoblast cells were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China).

HEK293T cells were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China).

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Zhao et al aimed to elucidate the mechanisms by which palmitic acids drives insulin resistance. The authors performed experiments in croaker, fish myocytes and mouse differentiated C2C12. They manage in vivo metabolic assays, enzymatic assays, immunoblot procedures, double luciferase assays, RNA analysis, pharmacological inhibitions and genetic knockdown. By using these model systems and procedures, the authors demonstrate that palmitic acid, but not oleic and linoleic acids, induces systemic and cellular insulin resistance through the hyper activation of mTORC1. They show that palmitic acid stimulates the mitochondrial fatty acid β oxidation, increasing the acetyl-CoA levels which enhances the acetylation of Rheb, a well known activator of mTORC1, by Tip60. Moreover, the authors show that mTORC1, beside reinforcing IRS1 phosphorylation, inhibits nuclear translocation of TFEB, thus preventing IRS1 transcription.

Major issues affecting the conclusions:

The conclusions are supported by the data. However, I suggest to perform a densitomentric and statistical analysis of western blots, especially when the authors report a representative blot, showing samples loaded in single.<br /> The methods are clear and reproducible. The authors should better explain how they have dissolved all the powders (i.e. fatty acids) to obtain the stock solutions next diluted (from what concentration?) in the media<br /> Anova analysis should be performed to analyze western blot densitometries.<br /> Minor comments:<br /> Prior studies are referenced appropriately, text and figures are clear. I suggest to add in the abstract all the model systems used. HEK293 also should be inserted in the description of the results. Please add the reference to figure 8 in the text. Please, describe cell origin.

Referee Cross-commenting

All reviewers have requested densitomentric (and statistical) analysis of western blot to prove the strength of the results. This is the major point to be addressed. Other points should also be only discussed.

Significance

The study extends the knowledge in the field of fatty acids-induced insulin resistance, that is a field studied by many researchers from many years, but with a lot of unclear mechanisms yet. Thus, the nature of the advance is conceptual and mechanistic. The only limitation is the lack of evidence in human samples/cells.

Basic researchers and experts in translational medicine will be interested by this research.<br /> This is the point of view of a basic researcher, mainly interested in the molecular mechanisms underlining type 2 diabetes/obesity and cancer.

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 goal of this study was to investigate how saturated fat induce insulin resistance through activating mTOR. More specifically, the researchers show that palmitate (saturated fat) activates mTORC1 to induce insulin via transcriptional and posttranslational suppression of IRS1. Specifically, the researchers show that PA stimulates FAO to raise cytosolic acetyl-coA levels, promoting Tip60-mediated acetylation of Rheb to activate mTORC1 activity. To study the relationship between mTORC1 activity, fatty acid stimulation and insulin resistance, the researchers decided to conduct their study in fish, as fish are known to be glucose intolerant by nature, making them an appropriate model organism for studying insulin resistance. Mouse C2C12 cell line and fish primary myocytes were also used to validate key results.

Major comments:

Initial experiment: Among several fatty acid-rich diets, fish were fed a palmitic acid (PA) rich (PO) diet for 10 weeks, and the PO diet significantly raised fasting blood glucose levels compared to control diet (fish oil of equal lipid content). The PO diet also impaired the fish's glucose and insulin tolerance. The PO diet also led to decreased phosphorylation levels of AKT, which regulates glucose metabolism. Therefore, the researchers initially concluded that a palmitic acid-rich diet leads to systemic insulin resistance in fish.

I have a couple of questions on this initial experiment on which all the subsequent studies are based. In Figure 1A, the body weight was identical in control and PO group. Don't you expect PO feeding lead to obesity in fish, as HFD induces obesity in mice?<br /> Figure 1G and 1H show glucose and insulin tolerance after PO feeding for 10 weeks. The area under curve (AUC) should be compared to determine if GTT and ITT were statistically different. The ITT curve is particularly interesting as the control fish did not seem to respond to insulin, while the PO-fed fish responded more robustly. The only difference is the initial glucose level. Are the GTT and ITT done after fasting? How long is the fasting? The curves suggest that even though PO increased (fasting) blood glucose levels, it improved insulin sensitivity - therefore the premise that PO induces insulin resistance is not supported here. The lack of insulin induced response in the control group is worrisome. I suggest that the measures should be retaken, and AUC should be used to support if there are any differences in GTT and ITT.

Based on the assumption that PO induces IR (which needs to be confirmed based on the previous comments), the researchers attempted to understand how PA triggers IR through a series of experiments, predominantly western blot analysis. All the Western blots should be quantified. The model is that PA activates FAO in mitochondrial that elevates cytosolic acetyl-coA, which acetylates Rheh to activate mTORC1. mTORC1 on one hand alters IRS1 phosphorylation and on the other hand inhibits transcriptional activity of TFEB to reduce Irs1 mRNA level. Together reduces IRS1 leads to Insulin Resistance.

Figure 1. PA reduces basal and insulin stimulated AKT phosphorylation in fish liver and muscle, as well as in culture fish and murine myocytes (Fig. 1I-M). The results appear to be solid but need to be quantified.

Figure 2 shows that PA provoked hyperactivation of mTORC1 (indicated by elevated phosphorylated S6K levels. This effect was abolished by Rapamycin treatment (an mTORC1 inhibitor) and also abolished by insulin stimulation (2F). Again, the western blots should be quantified.

Figure 3: PA treatment increases mRNA expression levels of fatty acid beta oxidation genes in fish myocytes and C2C12 myotubes, and subsequent suppression of CPT1B and CPT2 (rate-limiting enzymes of FAO) inhibited mTORC1 activity and signaling in muscle, C2C12 myotubes and fish myocytes under PA treatment. This suggests PA-induced mTORC1 activation is dependent on mitochondrial FAO. Inhibition of CPT1 improved suppression of insulin stimulated phosphorylation of AKT under PA treatment in fish myocytes and C2C12 myotubes, indicating that mitochondrial FAO is heavily involved in PA-induced mTORC1 activation that contributes to insulin resistance.

Figure 4: PO diet increases acetyl-CoA levels in muscle and PA treatment increases intracellular acetyl-CoA in a dose-dependent manner in fish myocytes. Inhibition of FAO by perhexiline maleate diminished induction of acetyl-CoA under PA treatment. In vivo dsRNA knockdown of ATP citrate lyase (ACLY, catalyze acetyl-CoA synthesis from mitochondrial citrate) decreased mTORC1 activity in muscle, and inhibition of ACLY in fish myocytes and C2C12 myotubes decreased induction of mTORC1 activity under PA treatment. This indicates that palmitic acid promotes mTORC1 activation through acetyl-CoA that is derived from mitochondrial FAO. PA treatment elevates acetylation of Rheb in a dose-dependent manner. Inhibition of FAO by perhexiline maleate attenuated PA-stimulated Rheb acetylation, while fish myocytes and C2C12 myotubes treated with sodium acetate (which can enhance acetyl-CoA) exhibited enhanced Rheb acetylation. The data indicate that acetyl-CoA produced by FAO activates mTORC1 signaling through increased Rheb acetylation. Phosphorylation of AKT were enhanced in muscle with dsACLY knockdown injection, and sodium acetate addition blocked recovery of insulin-stimulated glucose uptake and phosphorylation levels of AKT by perhexiline maleate under PA treatment. ACLY inhibition promoted insulin stimulated phosphorylation of AKT under PA treatment. So, in terms of acetyl-CoA's role in PA-induced insulin resistance, the data suggest that acetyl-CoA derived from FAO mediates PA-induced mTORC1 activation and insulin resistance.<br /> Figure 5: Acetyl-CoA can activate lysine acetyltransferases, and the researchers found that mRNA expression of tip60 was elevated in fish myocytes and C2C12 myotubes under PA treatment. Cultured fish myocytes and C2C12 myotubes treated with a Tip60 inhibitor prevented the induction of mTORC1 activity under PA treatment. Tip60 knockdown also blocked PA-induced mTORC1 activation in C2C12 myotubes. The researchers then determined that Tip60 regulation of mTORC1 is dependent on the acetylation of Rheb by studying the interaction between the two via a CoIP assay, which indeed indicated that Tip60 and Rheb interact. Additionally, Tip60 knockdown impaired PA-induced acetylation of Rheb, supporting the notion that Tip60 mediates the acetylation of Rheb under PA treatment. Inhibition of Tip60 attenuated PA-induced suppression of insulin-stimulated glucose uptake in C2C12 myotubes, and inhibition of Tip60 also restored insulin-stimulated phosphorylation of AKT under PA treatment. These data suggest that Tip60 mediates the regulation of Rheb acetylation under PA treatment and may be a novel therapeutic target for insulin resistance.

Figure 6: the researchers measured the effect of PA treatment on IRS1 phosphorylation in order to understand the mechanism of insulin resistance induced by mTORC1 activation under PA treatment. A PO diet intensified S636/S639 phosphorylation in fish muscle. In fish myocytes and C2C12 myotubes, PA treatment elevated S636/S639 phosphorylation but decreased the Y612 phosphorylation of IRS1 in a dose-dependent manner. Treatment of fish myocytes and C2C12 myotubes with an mTOR inhibitor blocked increased IRS1 S636/S639 phosphorylation levels under PA treatment. Also, PA specifically reduced mRNA levels of Irs1. This indicates that PA-induced, mTOR-dependent alteration of IRS1 phosphorylation and transcription may have contributed to insulin resistance. It is unclear how mTORC induces either increase or decrease in IRS1 phosphorylation depending on the residuals.

Figure 7 shows that PA inhibits nuclear translocation of TFEB to suppress IRS1 transcription. The EMSA in 7D is not convincing.

Minor comments:

Some data appears to weaken the results and/or contradictory. For example, the paper initially showed reduced AKT phosphorylation to support PA induced IR, but shouldn't a lower level of pAKT reduces mTORC activation? But then the rest of the manuscript explores how PA activates mTOR. Part of the IR is manifested by impaired mTORC1 activation, yet the PA activates mTORC1. The authors should present the rationale and flow of the ideas in a better way.

There are also many run-on sentences and grammar issues, making it very hard to read. The writing can be improved.

Referee cross-commenting

Other than lacking quantification of western blots, my major concern is the ITT curve in Figure 1G, which does not support the conclusion that PO induces insulin resistance and therefore the rest of the study is based on a faulty premise. The curve shows that insulin reduced blood glucose much more robustly in the PO group than in the control group, suggesting PO increased insulin sensitivity. Area above curve should be calculated to quantify the difference.

Significance

Overall, this research was trying to show that PA-induced IR is dependent on hyper activation of mTORC1. More specifically, acetyl-CoA induces mTORC1 activation under a palmitic acid diet, and this is achieved through Tip60-mediated Rheb acetylation, which ultimately leads to insulin resistance through IRS1 suppression. The study is very mechanistic and important for understanding IR that is associated with diets high in saturated fatty acids and could potentially leads to therapeutic targets for combating insulin resistance and glucose intolerance.

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 work, the authors showed PA induces hyperactivation of mTORC1 and insulin resistance. They found acetyl-CoA derived from mitochondrial fatty acid oxidation is required for PA-induced mTORC1 activation and insulin resistance by increasing TIP60-mediated Rheb acetylation. They also showed PA induced mTORC1 activation enhances IRS1 phosphorylation and inhibits transcription of IRS1 by impeding TFEB nuclear translocation. Overall, the authors did a lot of experiments to prove that PA causes mTORC1 activation and insulin resistance. The results are generally convincing, and the finding is novel and instructive.

1. The authors claim PA-induced mTORC1 activation is dependent on acetyl-CoA derived from mitochondrial fatty acid oxidation. Using isotope tracing to determine the contribution of PA to acetyl-CoA, would improve.
2. They showed PA increases fatty acid oxidation related gene expression and acetyl-CoA level, while OA and LA could not. why only PA could increases fatty acid oxidation and acetyl-CoA level, considering both of these lipids could be oxidation in mitochondria? Is there any differences in mitochondria among treatment of PA, OA and LA? It is better to monitor fatty acid oxidation in real time using seahorse. And add discussions.
3. The authors present lots of western blot images, suggest to provide quantification data of these blots.
4. It is better to discuss the relationship between fatty acid oxidation and mTOR signaling.

Significance

The finding is interesting and significant.

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.

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 present manuscript presents a thorough description of the relative localization (in space and time) of a number of proteins of the early secretory pathway. To that aim, the authors used by their custom-made 3D live cell super-resolution microscope (SCLIM) and the yeast S. cerevisiae as a model system. The main claim of these data is that the early secretory pathway in S. cerevisiae is organized by maturation from a newly proposed yeast ERGIC compartment all the way to the trans-Golgi network (TGN).

Major comments:

I have two major comments regarding this manuscript:

1. It is not clear to me how the presented data shows the existence of an ERGIC in the yeast S. cerevisiae. I understand, and appreciate from this text too, that a clear definition of ERGIC, even in a mammalian system, is unclear. For this reason, I would first suggest that the authors provide a clear definition of what ERGIC means to them. Next, the experiments herein presented are all based on a very careful, thorough and nicely organized spatio-temporal mapping of a large number of early secretory pathway proteins, including the ERGIC53 yeast "counterpart" Emp46 (it would help to add, even as a supplementary figure, an alignment/sequence comparison between the human ERGIC53 and the S. cerevisiae emp46). However, the data presented here does not clearly indicate to me that there is a bona fide ERGIC in yeast. Couldn't it just be that what the authors call ERGIC is a cis-Golgi cisterna? I understand that the BFA experiments show a different behavior for some proteins, which fits with what the authors previously names GECCO in plants, so why not calling this GECCO? Again, it will be important to provide definitions of these compartments for the audience. Next, my main concern here is that this is all based on SCLIM, which is a very nice technique, but the resolution is limited in both space and time (by the way, it would be nice to explicitly measure of quantify the spatial resolution in x-y-z). Hence, it is not possible to discern whether an "independent" ERGIC is formed as compared to cis-Golgi cisterna. Electron microscopy (possibly CLEM) could help somehow resolve that and massively increase the strength of the claims, but I do understand this might be difficult for this group and very time consuming, so it might be important to clearly state the limitation of the herein presented data. A possible alternative to test if protein that are seen segregated are within the same membrane (as claimed here) would be to do trapping experiments where a reagent induces dimerization between the two proteins (when tagged with specific tags, such as FKBP/FRB).
2. I could not find any details (maybe I have missed them) about how many times experiments were replicated and the statistical significance of the findings herein reported. In most figures, examples of microscopy images/videos are shown, and selected lines profiles are presented. However, it is not clear how robust these experiments are. Some ideas:

2.1.) The major source of quantification is the peak-to-peak time distance between two proteins. In Table S1 some stdev is presented, but not clear how it is find (it is the sted of all n number of puncta? or of the mean duration per cell? or of the mean duration per experiment? I would suggest that the authors provide the results shown in Table S1 plotted as a histogram or superplot (see e.g. https://rupress.org/jcb/article/219/6/e202001064/151717/SuperPlots-Communicating-reproducibility-and) and clearly explain how statistics is performed.

2.2) Also, the time-lapse movies are acquired with a 5s gap between time points. How is this included in the incertainty of the peak-to-peak duration in Table S1?

2.3) In pg. 7 the authors write "Although experimental variation was high, the two zones appeared to be spatially segregated". Can the authors provide quantitative and statistical support of this claim?

2.4) It is not clear to me how the puncta for analysis are selected. For example, in Fig. 1C, the punctum shown already shows some initial co-localization (it could be e.g. that a peak value was prior or after the duration of the time lapse movie, thereby biassing the computation of the peak-to-peak duration). So, if one would consider those spots e.g., positive for Emp46 that do not contain Mnn9 signal, how often do you see conversion (that is, appearance of Mnn9 signal)? Along the same lines, in pg. 8 the authors write "... signal appeared first and then mnn9-mCherry came up". Details on how this quantification is done and statistical analysis would be needed, to my opinion, to support the claim.

Minor comments:

1. The color code for the 3 color microscopy images is nice, however, the use of green and red for the 2 color images is a bit unfortunate for some people (like myself) who suffer from color blindness. I'd suggest to use green and magenta instead.
2. Pg. 8: have the authors tested Rer1 vs Emp46?
3. Pg. 8: I was of the impression that GRASP65 (GORASP1) is considered to be a cis-Golgi protein (see e.g., Tie et al eLife 2018). Then, what the authors call "ERGIC" couldn't it simply be a cis-Golgi cisterna?
4. pg. 13: "propose to define Grh1, Rer1, and Sed5 as yeast ERGIC/GECCO...". What about Emp46?
5. The first part of the manuscript (up to mid page 13) is clearly focused on defining ERGIC in yeast, then the paper appears as a set of experiments aimed at adding more components in their spatio-temporal mapping. This is ok, but is should be clearly motivated and explained in the Title, abstract and intro.
6. The visualization of colocalization according to the opacity (as said in the methods) is somehow confusing to me. Are the 3D images projections or 3D renderings (no axes are seen)? In e.g. Fig. 6G or 8L, regions where green and magenta (or green and red) are colocalized do not appear white (or yellow), which visually suggests to the inattentive reader that there is no colocalization, when there is.
7. I have not understood what this sentence in pg. 18 means: Similar segregation patterns are also observed during the Golgi-TGN maturation process (Tojima et al., 2019). "We propose that the ERGIC, Golgi, and TGN can coexist as structurally and functionally distinct zones within a single, maturing cisterna." Are they referring to ERGIC, Golgi, and TGN steady state components (proteins) or the structures themselves?
8. The introduction of new data (mammalian data) in the discussion is odd. It might be ok, but I would frame it within a results section and use it later in the discussion.
9. Fig.9: the arrows should go from protein to protein (some seem to go from in between proteins, such as the bottom-most arrow with 87.8 s time duration. Also in panel B, bottom part, some proteins are missing (Erd2 ad Chs5).
10. Fig. 1 and many other: in the line profiles the distance in the x axis has no units or labels. Please add this and the direction of the line profile (an arrowhead would suffice).

Significance

General assessment:

The experiments herein presented are based on a very careful, thorough and nicely organized spatio-temporal mapping of a large number of early secretory pathway proteins, including the ERGIC53 yeast "counterpart" Emp46. However, the data presented here does not convincingly show that there is a bona fide ERGIC in yeast. A major limitation is that the experiments are all based on a state-of-the-art, but still with a limited resolution, fluorescence microscopy technique. Ultrastructural data (e.g., CLEM) would massively help support or revisit the claims presented in this manuscript regarding the existence of an ERGIC compartment in yeast. Also, adding the information about the number of biological replicates and proper statistical analyses on the presented results would be needed to further support the claims.

Advance:

This manuscript builds on the authors' custom build 4D super-resolution microscope (SCLIM) and on previous results (e.g., Tojima et al., J. Cell Sci. 2019). The main novelty is in the study of a number of new early secretory pathway proteins and in the proposal of the existence of a non-stable, maturing ERGIC compartment in S. cerevisiae.

Audience:

This paper might be attractive for a broad audience of cell biologists, especially those interested in membrane biology, cell compartmentalization, and intracellular trafficking and secretion.

Describe your expertise:

I am an expert in membrane trafficking.

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

Tojima and colleagues present a very exciting 4D SCLIM analysis of 20 key-proteins controlling or occupying different stages of Golgi-mediated protein trafficking, taking the reader on a trip from the ER export sites to the ERGIC/GECCO, cis-, medial- and trans-Golgi towards the TGN/recycling endosome. The choice of molecular markers to be patiently time-resolved in 3D allowed the authors to assemble a temporal roadmap for the molecular players studied. This is most impressive.

Major comments

There is an enormous amount of patient systematic analysis packed in this paper, following a punctate organelle as it emerges from the dark, evolves over time (following a combination of 2 markers) with fluorescence peaks at specific time points, after which the signal disappears again. I am certain other cell biologists will be impressed, as I was, viewing the individual images and graphs presented, culminating ultimately in figure 9 that could go straight into a textbook to form a starting point for anybody who wishes to study a particular protein of interest and chose the most appropriate markers to compare it with. The authors propose that the ERGIC/GECCO/Golgi-remnants compartment is an evolutionary conserved structure even though it has a different subcellular distribution/morphology in different classes of eukaryotes. The data presented here and in earlier work seem to support this notion. In particular, the authors demonstrate that the yeast GRASP 65 homologue Grh1 is the earliest to appear closely followed by Ypt1 and Emp46. The fact that RER1 and ERD2 come slightly later is in line with a proposed gate-keeper function, because if they were instead to recycle continuously they should appear first in line. I agree with the authors that the simple model of ER-derived COPII vesicles fusing with each other and thus creating an ERGIC/GECCO de novo is probably too simplistic. The idea of a more permanent structure, pulsating between cargo-loading and cargo-releasing events, possibly associated with creating zones/subdomains within a single cisterna seems very attractive given the data shown here. This work is descriptive, but it is of very high importance to anybody engaged with experimental approaches to study protein sorting from the Golgi-apparatus back to the ER, or on to the plasma membrane or the lytic compartments. The 5 functional stages proposed for Golgi-maturation is an attractive starting point for future research, and I very much like the notion that ERGIC and cis-Golgi cisternae may start as zones/subdomains within a single cisternae, possibly formed via phase separations involving both protein-protein and protein-lipid interactions.

Minor comments

The title strongly focusses on the ERGIC and therefore the earliest sorting steps in the ER-Golgi system, but this manuscripts offers so much more. I was fascinated to learn that Ypt1 appears twice during cisternal maturation in yeast. This may be a yeast-specific phenomenon but it is very interesting. The same can be said about the proposal that Gea1 and Gea2 have different roles in the Golgi and act in different cisternae, and the localisation of AP-3 at the trans-Golgi rather than the TGN. The functional distinction between trans-Golgi and TGN and the differences in their origin are important points and it will be a shame if readers don't realise that this manuscript offers further insight into later steps in Golgi-mediated transport. There may be a case to add something to the abstract and/or modify the title accordingly, but then I also feel that long titles are not ideal and the ERGIC/GECCO portion is the more important take-home message. This is a case for the editorial team and the authors to make the most of the findings.

Given the importance of ERD2 in sorting soluble proteins to be returned back to the ER, the authors may consider using the biological active XFP-TM-ERD2 fusion instead of ERD2-GFP, but this may be kept for future work. In plants, ERD2-GFP is mainly at the Golgi when overexpressed, its erroneous leakage to the ER is only observed at low expression or when K/HDEL proteins are co-expressed. The XFP-TM-ERD2 construct may be better confined to the ERGIC-GECCO and may have a different temporal pattern.<br /> The authors may consider citing Stornaiuolo et al., Mol Biol Cell 2003 Mar;14(3):889-902 who compared the trafficking of KDEL and KKXX pathways and concluded that KDEL proteins are retrieved prior to KKXX proteins....as this fits nicely into the current findings showing that ERD2 and RER1 appear sooner than COPI markers.

Significance

The significance of the work is high because it basically allows to add facts to models. We use models to explain an elusive process because it escapes direct observation. Once we can observe directly, a model can become fact. In simple terms, the authors allow us to see things that we could only speculate about in the past. Therefore, the results present a very significant advance and will be highly relevant to the entire cell biology community. This paper is an important landmark and will help the field to formulate new experimental approaches and models to understand the origins of the Golgi apparatus, the core of the secretory pathway that defines being a eukaryote.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

It is generally believed that budding yeast does not have the ER-Golgi intermediate compartment (ERGIC). In this study, the authors attempt to prove the existence of the ERGIC. They measured the kinetics of a few Golgi markers involved in the early secretory pathway in live cell imaging and observed the recruitment of Grh1 precedes that of Mnn9, suggesting the presence of pre-early cisternae. The authors propose that the Grh1-positive cisternae, assembling at the ER exit site (ERES) and progressing to become the early Golgi cisternae, represent the equivalent of the mammalian ERGIC in yeast.

Major comments:

The concept of the ERGIC in mammalian cells was initially proposed based on the protein ERGIC-53 in the 1990s. However, recent nanoscopy imaging data from the Lippincott-Schwartz lab challenges the conventional view of ERGIC by revealing the "ERGIC" is a membrane domain of the ERES (Wegel et al., Cell, 2021; PMID: 33852913), suggesting it might not be appropriate to adopt this concept.

The authors' observations could be interpreted differently. Since the ERGIC is not molecularly defined in their study, the authors cannot prove its existence in yeast unequivocally. Their data indicate the presence of Golgi cisternae, characterized by Grh1, that precede the earliest known cisternae. Although the authors refer to these Grh1-positive cisternae as the "ERGIC", they are essentially "pre-early" cisternae that progress to become the early Golgi cisternae. Nevertheless, their findings could extend the budding yeast Golgi cisternal progression unit further upstream to include the ERES as the starting point for Golgi cisternal maturation. To further explore this, it would be interesting to investigate the kinetics of COPII subunits in cisternal progression along with Grh1 or Mnn9 and to plot COPII components in the Figure 9 map.

The second half of the manuscript appears to deviate from the main focus on identifying the ERGIC. This section primarily presents the Golgi localization of four Golgi proteins (Ypt1, Gea1, Gea2, and Alp6) deduced from kinetics. However, it lacks functional studies to substantiate the authors' claims on their cellular functions. As a result, this part of the study remains purely speculative and might not support the authors' claims. Given that Figure 9 provides a highly informative summary of all kinetics and localization data, I recommend the authors keep but significantly abridge this section.

The manuscript also has a few major concerns.

1. The analysis of only one fluorescent particle or Golgi cisternal punctate structure is insufficient for a Golgi marker, considering the substantial variation of Golgi cisternae. To improve statistical robustness, the authors should select multiple fluorescent particles from multiple cells, displaying plots with averaged intensities, error bars, and sample sizes (n).
2. In Fig. 5A, the BFA-induced lumps positive for Grh1, Rer1, and Sed5 may potentially represent the ERES, as observed in mammalian cells (Ward et al., JCB, 2002; PMID: 11706049). To verify this, the authors should co-label these lumps with COPII subunits.
3. The authors previously reported the "hug-and-kiss" model for cargo transport from the ERES to the early Golgi cisternae. As the current study is highly relevant to the "hug-and-kiss" model, it is disappointing that the authors did not provide further data and comment on it. The "hug-and-kiss" and "ERGIC" transport modes are two distinct ways for secretory cargo transport from the ERES to the early Golgi cisternae. The authors should verify the "hug-and-kiss" transport and report the relative frequency of the two transport modes.
4. The current version has minimal background knowledge of ERGIC in mammalian and yeast cells. Therefore, the authors should provide a comprehensive introduction to ERGIC.

Significance

General assessment:

The data presented in the manuscript is novel and appears to be convincing. However, one of the major concerns is the lack of statistical robustness, which requires to be addressed. Furthermore, the manuscript's data could be interpreted differently, as elaborated in the major comments.

Advance:

While the interpretation of the manuscript's data requires reconsideration, it can contribute to our understanding of secretory trafficking at the Golgi. The manuscript could fill a crucial gap in our knowledge in this field by addressing the major comments.

Audience:

Cell biologists in the membrane trafficking field, particularly those working on the Golgi, would find the manuscript interesting.

My expertise:

My research focuses on membrane trafficking at the ER, Golgi, and endosome.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>