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 explored the function of the protein kinase KIS in splicing regulation associated with neuronal differentiation in vitro. KIS is a serine threonine kinase known to phoshorylate splicing factors such as SF1 and SUGP1, and to be preferentially expressed in adult brain in mammals. Using an shRNA based approach, the authors characterize cassette exon usage upon partial KIS depletion in cultured mouse cortical neurons. In parallel using mass spectrometry of proteins in KIS overexpressing HEK293 cells, they identify potential KIS substrates including the splicing regulator PTBP2. They confirm that recombinant KIS can phosphorylates PTBP2 in vitro. They show a correlation between KIS-activated and PTBP2-inhibited exons using published data for this factor. They report opposite effects of KIS and PTBP2 on CamKIIB splicing and Finally, coimmunoprecipitation and FRET experiments suggest that KIS inhibits the interactions of PTBP2 with known protein binders, hnRNPM and Matrin3 as well as with RNA. Altogether these data suggest that KIS downregulates PTBP2 during neuronal differentiation.

Major comments:

Overall the manuscript is well written and the data are interesting. However several points could have been more extensivelly studied or discussed to achieve a stronger demonstration of the role of KIS in PTBP2 phosphorylation and neuronal differentiation.

1. To minimize possible off target problems, the RNAseq analysis would be more convincing if replicated with a second shRNA to knockdown KIS.
2. Part of the reported splicing changes might reflect an indirect consequence of an altered differentiation contributing to the correlation observed in figure 1F. It would be interesting to confirm splicing changes using shorter incubation times with the shRNA compared to the 11 days used in this study.
3. Standard deviation is more relevant to describe data dispersion in all figures.
4. Previous papers of the group described a function of KIS in translation (Cambray et al 2009, Pedraza et al 2014). This is not discussed here. For example, the possibility that RBPs are regulated by KIS at the translation level is not excluded by the analysis in Fig EV2a.

Minor comments:

Figure 1:

The authors state that : "KIS...accumulates in nuclear sub-structures adjacent to those formed by splicing factors". As the figure presents in fact GFP-KIS, it should be mentioned, and how this localisation is relevant for endogenous KIS should be adressed.

Fig EV1: SI range in pannel D is very different from that in pannel C and Fig1E.

On page 4 "KIS expression reached maximal levels in hippocampal cultures (Fig 1B)." However the figure legend indicate that this analysis was performed with cortical neurons. The use of cortical or hippocampal neurons along the manuscript should be clarified.

page 4 " KISK54A, a point mutant without kinase activity" The authors should indicate the reference.

Figure EV2C: It is not clear whether the Coomassie staining and autoradiography do correspond to the same gel.

Figure 3C The authors use a dual fluorescence reporter to analyse PSD95 exon 18 splicing. However the well to well variability in such experiments might be elevated. Not only the cells number in a single well but also the number of replicates should be indicated and well to well variability reported.

Figure 3D. The precise timing for the transfection and culture of cells before staining is unclear

Figure 4A. The input should be loaded to evaluate the coIP efficiencies and ascertain that KIS does not downregulate Matrin3 and hnRNPM levels.

Figure EV4A. No difference of Matrin3 binding is to be seen on the gel. In addition, the authors should confirm that PTBP2 or binders are phosphorylated by recombinant KIS. The preparation of GST-KIS is not described. Page 6: "We found that PTBP2-inhibited exons are significantly (FDR=0.001) enriched in KIS knockdown neurons, supporting the notion that KIS acts on AS, at least in part, by inhibiting PTBP2 activity." This should be rephrased as in fact PTBP2-inhibited exons are enriched among KIS activated exons. Page 10: "SUGP1 is one of the most enriched proteins in our KIS phosphoproteome (see Fig 2A)". Phosphorylation and interaction with KIS was already reported by Arfelli and coll. 2023 supplementary figure 2.

" It forms part of the spliceosome complex, interacts with the general splicing factor U2AF2 and has been reported to play an important role in branch recognition by its association with SF3B1." A reference is needed there.

The authors previously reported a differentiation defect in cultured neurons 'Cambray et al, 2008' that was not observed by another group (Manceau et al., PLOS One 2012). This should be discussed in view of these more recent results. Is there any differentiation defect in the experiments reported there?

Statistical values are difficult to read in the figures. Please use larger fonts.

Significance

This manuscript brings new elements supporting the function of the protein kinase KIS in splicing regulation in neurons. In particular it identifies for the first time the splicing regulator PTBP2 as a substrate for KIS.

It will be of interest to a specialized audience of researchers interested in splicing regulators in neuronal differentiation.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The manuscript by Moerno-Aguilera et al. shows that the brain enriched protein kinase KIS targets the well known neuronal splicing regulator PTBP2 and several of its interaction partners. As a consequence, PTBP2 activity is down-regulated. Using cultured primary immature neurons they show that KIS expression increases during differentiation and that shRNA knockdown of KIS alters the splicing of many alternative exons. Phosphoproteomic anlaysis of HEK293 cells transfected with KIS or a kinase dead mutant (K545A) show that it phosphoryates both PTBP2 as well as a cluster of proteins that are known to interact with PTBP2 or its paralog PTBP1. By comparing the new data on KIS-dependent splicing with previous data-sets on PTBP2-dependent splicing targets they show that KIS appears to act antagoniostically with PTBP2 when it acts as a repressive regulator, but not when it is an activator. Using combinations of wt and kinase-dead KIS with PTBP2 mutants in the 3 main phsophorylation sites (3SA - non-phosphorylatable, S3D - phosphomimetic) to look at the effects on a known PTBP2 functional target, PSD95, they show that the likely effect of KIS is to antagonise PTBP2 function by phosphorylation at one or more of three residues (S178, S308, S434). Finally, they show that transfected KIS (but not K54A) reduces known protein-protein interactions of PTBP2 and that the triple phosphomimetic PTBP2 mutant shows reduced binding to RNA. Alphafold2 predictions show that the S178 phosphomimetic mutant might alter the conformation of the RRM2 domain, in particular altering the environment of Y244, which has been shown in PTBP1 to be critical for interaction with MATR3 and other coregulators.

Major points

In general, the conclusions drawn are consistent with the data. I have a few suggestions where the authors could either extend their findings with a few straightforward additional experiments, or clarify some of the existing data.

FigEV4 (also introductory text on p3): RRMs 3 and 4 of PTBP1/2/3 fold as a single back to back packed didomain - with the so-called linker contributing to the didomain fold (e.g. PMID: 24688880, PMID: 16179478) and also extending the RNA binding surface by creating a positive patch (e.g. PMID:20160105 PMID: 24957602). AlphaFold successfully predicts the didomain in full length PTBP2 (https://alphafold.ebi.ac.uk/entry/A0A7I2RVZ4). The authors should therefore use AlphaFold2 to predict the RRM3-4 di-domain structure of wt and phosphomimetic mutant PTBP2s. Phosphorylation of S434 or S434D, which is on the C-terminal end of RRM3 may have no predicted effect on RRM3 alone (FigEV4), but it could conceivably disrupt didomain packing, which could itself have important knock-on consequences for RNA binding. In addition, the inrtoduction of negative charges at S434 might affect the ability of R438, K440 & K441 to interact with RNA. An image of the didomain charge density of WT and mutant PTBP2 would be useful to address this.

Figure 4 could also easily go further in experimentally testing the effects of individual phosphomimetic mutations upon protein-protein interactions (Alphafold predicts that S178D, but not S308 or S434D, should affect Y244 mediated interactions, such as MATR3). The co-IP approach in Fig 4A could readily be used with FLAG-PTBP2 mutants. Likewise, consequences of individual mutations upon RNA binding (Fig 4D) could be tested. The use of a Y244N mutant here would test whether the loss of RNA binding is a consequence of the loss of protein-protein interactions. Such experiments are not essential, but they are readily carried out and have the potential to unravel the consequences of the individual phorphoryation events (more correctly of phosphomimetic mutants).

Minor

Do KIS regulated exons show enrichment of motifs associated with PTBP2, consistent with the proposed model - particularly CU-rich motifs upstream of exons that are more repressed upon KIS shRNA treatment.

For the splicing analysis pipeline, how were exon-exon junction reads treated? If "only exons with more than 5 reads in all samples" were considered, will this not exclude highly regulated exons that are completely skipped under one condition?

The Introduction mentions U2AF homology (UHM) domains, but neglects to discuss their known binding partners - ULMs (UHM ligand motifs), which contain an essential tryptophan. It would be useful for the discussion to highlight whether any direct KIS interactors possess ULMs and how this relates to the phospho-targets identified here. The authors may wish to draw the parallel with the structurally analagous way that PTBP1 (and presumably PTBP2) interact with their short peptide ligand motifs.

Figure EV2C. The S3A and S308A mutations clearly reduce phosphorylation. However, the effects of S178A and S434A are far less clear. Presumably the quantitation shown in the lower panel of EV2C relies on normalization to PTBP2 protein input, which appears quite variable in the Coomassie gel. It might be better to repeat the experiment with uniform protein inputs. Minimally, details of the quantitation approach should be added to Materials and Methods.

Fig 3D shows PTBP2 overexpression, but the main text (p7) states KIS overexpression.

Fig 4B should have a scale bar for the FRET signal

Fig 4E should indicate the location of S178

Significance

This interesting, clear and concise manuscript provides important new insights into the way that a neuron specific kinase can regulate neuronal splicing networks by phosphorylating and thereby downregulating the known neuronal splicing regulator PTBP2. Alternative splicing is known to play a particularly important role in neurons, so this demonstration of an additional layer of regulation by post-translational modification should make the manuscript of wide interest to investigators of splicing regulation, neuronal differentiation and maturation.

Issues that are not addressed in the manuscript include; i) how does KIS specifically target PTBP2 and related proteins? The UHM domain can mediate interaction with ULM containing splicing factors (such as U2AF2, SF3B1), but none of the identified targets have known ULMs. ii) the consequences of individual phoshomimetic mutants upon protein-protein interactions and RNA binding could readily be explored further using computational and experimental methods already used in the manuscript.

For context, this reviewer has a direct interest in the mechanisms of regulation of alternative splicing, but not in the context of neurons (though I am familiar with a lot of the relevant literature), and I do not have expertise in neuronal 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

Summary

In this manuscript, the authors characterized the molecular function of the brain-enriched kinase KIS by combining transcriptome-wide approaches with molecular and functional studies. They uncover that KIS regulates isoform selection of genes involved in neuronal differentiation and inhibits through phosphorylation the capacity of the splicing regulator PTB2 to interact with both target RNAs and protein partners.

Major comments

• This is a very clear and well-written manuscript presenting high-quality and carefully controlled experimental results. The authors used an impressive range of approaches (transcriptome-wide exon usage, phospho-proteomic, imaging, biochemical assays..) to profile exon usage alterations upon KIS knock down and provide a mechanistic understanding of how KIS regulate the splicing activity of PTBP2. Specifically, they convincingly demonstrate that the phosphorylation of PTBP2 by KIS leads to both dismantling of PTBP2 protein complexes and impaired RNA binding. My only main concerns relate to the understanding of the biological context in which the mechanism studied may be at play. That KIS can counteract PTB2 activity through direct phosphorylation has been very clearly shown by the authors using overexpression of KIS and /or PTB constructs in different contexts (HEK293T cells, N2A cell line, hippocampal neurons). Whether this occurs endogenously in the context of neuronal differentiation, and how much this contributes to the overall phenotypes induced by KIS inactivation, is less clear. While fully investigating the interplay between KIS and PTB2 in the context of neuronal differentiation is beyond the scope of this study, the three following points could be addressed to provide some evidence in this direction.

• Building on the experiments they perform in a KIS knock-down context (e.g. Fig. 3B, or previously described spine phenotype), the authors should investigate whether inhibiting PTBP2 in this context (through shRNA or expression of a phospho-mimetic construct) might suppress the phenotypes observed when inactivating KIS.

• Based on Figures 1E and 3A, it seems that KIS downregulation affects both exon inclusion and exon skipping, and that its function in exon usage is only partly explained by modulation of PTBP2-dependent exons. Have the authors analyzed the populations of PTBP2-dependent exons that are regulated by KIS in an opposite manner? This may point to specific classes of transcripts (in terms of expression pattern, function, molecular signature) important in the context of endogenous neuronal differentiation.
• The authors should better discuss when and where they think PTBP2 phosphorylation by KIS might be relevant. Is there evidence that this process (or PTBP2 complex assembly) might be regulated upon differentiation or plasticity?

Minor comments

1. Figures and associated legends are overall very clear and well-organized. Addressing the following points would however help improving the clarity of some Figures:
   • In Figure 2EV2C legend, the characteristics of the 3SA constructs are not described
   • the difference between Figure EV1A and Figure 1H classifications is unclear, nor the interpretation regarding the different GO classes identified
2. Whether PTBP2 is endogenously the major target of KIS explaining transcriptome-wide changes in exon selection is a possibility that remains to be demonstrated. Thus, the authors should correct and tune down the following sentences: "KIS phosphorylation counteracts PTBP2 activity and thus alters isoform expression patterns ..." (end of introduction) "PTBP2 being one of the most relevant phosphotargets" (results, end of the second section)

Significance

• The splicing regulator PTBP2 is a known master regulator of neuronal fate whose tightly controlled expression drives the progenitor-to-neuron transition as well as the establishment of neuronal differentiation programs. How this protein is regulated at the post-translational level has so far remains poorly investigated. In this manuscript, the authors provide a thorough mechanistic understanding of how KIS-mediated phosphorylation of PTB2 impacts on its regulation of exon usage. They also provide a transcriptome-wide view on the function of the brain-enriched KIS kinase in exon usage, uncovering its broad functions in alternative splicing. If the physiological context in which KIS-mediated phosphorylation of PTB2 is induced remains to be precisely defined, this work opens interesting new perspectives on regulatory mechanisms at play during neuronal differentiation. Providing extra lines of evidence indicating that KIS acts on neuronal functions through PTBP2 phosphorylation will help further strengthen this aspect.
• This manuscript will be of interest to different large communities interested on one hand on the regulation of gene expression programs underlying neuronal differentiation and on the other hand on the molecular regulation of major complexes involved in alternative splicing and isoform selection. It opens new perspectives related to the spatiotemporal regulation of neuronal isoform selection.

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

Learn more at Review Commons

---

Reply to the reviewers

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

* Srinivasan et al. present a comprehensive study on systematizing the structure-dynamics-function relation of lipid transfer proteins (LTPs), combining extensive molecular simulations and complementary experiments. Indeed, the current state-of-the-art in the field is quite chaotic and fractional, and such systematic studies are necessary to advance our general and conceptual understanding of the mechanisms of action of LTPs. The selected techniques and research strategies are all suitable, their description is sufficient and enables reproducibility; the obtained results are carefully presented and discussed; the conclusions are adequately supported by the data.

Given my primarily computational background, I evaluated mainly the simulation part of the manuscript. Considering experiments, I do not see any significant flows or deficiencies that could diminish the value of the data and following conclusions given in the manuscript. I would even suggest improving the abstract by more explicitly saying that this work includes experimental measurements because it currently reads like purely computational work was performed. *

We thank Reviewer #1 for the positive evaluation of our work. The abstract has now been updated to include that our work allows us to interpret existing data but also to design and perform new experimental measurements.

* Major comments: *

1) Although I like the central message of the paper and have no objections, I am curious whether the conclusion "a more "dynamic" or/and "mobile" part of the protein interacts with the membrane or any other (macro)(bio)molecule" makes sense globally and is not limited to LTPs. For example, it is a reasonable assumption that a more flexible part of the protein, i.e., capable of adopting necessary binding configurations, would be a more likely interacting spot. Locking in a less flexible and more specific configuration upon binding with a target molecule is also anticipated and quite typical, e.g., when ligands interact with target proteins, thereby blocking their function. The authors themselves recognize this paradigm as referring to the enzymes' dynamics. It would be great if authors could comment more on dynamics-function relation, referring to the existing literature, where such observations were/were not observed for different protein families. Performing simulations on proteins that do not exhibit such a feature and do not belong to LTPs, but, e.g., structurally similar to some of the studied LTPs, would be an excellent addition too, highlighting this signature characteristic of LTPs.

We have now added a discussion comparing the mechanism we observe with those described for other proteins such as membrane transporters and receptors. Since those proteins are very different and have been already thoroughly characterized (including with molecular simulations) we don’t think that additional simulations are required. Also, concerning protein binding dynamics, we refer to the excellent review of Wade and coworkers: "Acc. Chem. Res. 2016, 49, 5, 809–815"

"____Notably, the conformational plasticity we observe for LTPs is reminiscent of other, previously described, functional protein mechanisms, including enzyme dynamics during catalysis (____DOI: 10.1126/science.1066176____), the alternating-access model of membrane transporters (____https://doi.org/10.1038/nsmb.3179____) or GPCR dynamics (____https://doi.org/10.1021/acs.chemrev.6b00177____). In all these cases, protein dynamics is strongly coupled to ligand binding (____https://doi.org/10.1021/acs.accounts.5b00516____) and protein function, be it for signaling, transport or enzymatic activity. Unlike for these fields, however, the contribution of structural and spectroscopic studies to uncover LTP dynamics remains quite limited, and our simulations provide an important contribution to fill this gap. We hope that our results will motivate researchers to increase efforts to experimentally quantify LTPs conformational plasticity, e.g. by structural determination of LTPs in different states (or bound to different lipids) or by single-molecule spectroscopy studies."

*Minor comments: *

*

1) Fig 1d. What is so special in Lysine compared to Arginine? Is there any disbalance in their presence in studied proteins? Any correlations between the binding affinity of certain amino acids and their overall presence on the protein surface? *

Indeed, there is disbalance in the presence of lysine and arginine residues in our proteins. The relation between the number of these residues in our dataset is Lys:Arg = 1.6:1. On top of that, and as described in (Tubiana T et al PLoS Comput Biol. 2022 ;18(12):e1010346) lysine is preferred over arginine in peripheral membrane proteins, likely because it induces fewer perturbations in the lipid bilayer. Our data also agree with Tubiana et al, concerning the correlation between abundance of specific residues on the protein surface and membrane binding.

* 2) Fig S1. GM2A and TTPA seem to be irreversibly adsorbed to the membrane on the microsecond timescale in most replicas. Is anything special in these proteins? Did this affect the sampling of a claimed membrane-binding interface?*

Our interpretation of the different adsorption profile of GM2A and TTPA is that these two proteins appear to have higher membrane affinity in our computational assay in comparison with the other proteins in our dataset. However, this has no effect on the membrane-binding interface as the proteins are still able to undergo significant tumbling before binding to the lipid bilayer, as demonstrated by the angle between the two main protein axes and the bilayer normal before membrane binding (Fig. S8 in Supplementary Information).

* 3) A related follow-up question. Multiple replicas were performed to identify the membrane-binding interface. However, if I understand well, the initial orientation of the protein with respect to the membrane was always the same. I found it a pity since performing multiple replicas starting from different initial geometries (e.g., rotating the protein in a somewhat systematic way) would likely result in a more efficient exploration of the conformation space. Can the authors comment on whether this predefined initial configuration could negatively affect the results? Performing a few additional simulations for the most problematic proteins I mentioned earlier (GM2A and TTPA) could be a nice opportunity to apply this strategy. *

In our protocol, all proteins start from the same initial orientation but undergo significant tumbling in solution before interacting with the lipid bilayer, including for the two most extreme cases, GM2A and TTPA (Fig. R1). Hence, we think that there is no bias for what pertains to the final membrane interacting region. We have added the Fig. R1 in Supplementary Information (Fig. S8) and added the following text in the Methods Section:

"____Despite starting from a single orientation, all proteins undergo extensive tumbling before binding to the bilayer, as illustrated by the angle between the two principal protein axes and the membrane normal for the two proteins that display the highest binding propensity, GM2A and TTPA (Fig. S8)."

* 4) How was the volume of the cavity affected by mutations in STARD11 and Mdm12? Do these data somehow correlate with the experimentally observed reduced efficiency of the lipid transfer? *

Our data on the volume of the cavity in STARD11 and Mdm12 are inconclusive. However, we caution from such a simplistic interpretation, since it completely neglects the lipid-bound conformation that normally has a much larger cavity than the apo form (Fig. 3).

*5) I would appreciate it if the authors considered playing with the templates of the main Figures at later stages because in the current version, and when printed on A4 paper, the readability of certain graphs and pictures is uncomfortable and sometimes even impossible. Obviously, the final schematics would depend on the journal and its formatting. *

We will modify the templates of the main Figures to improve readability according to journal formatting.

* **Referees cross-commenting** *

* I would like to acknowledge the thoughtful and detailed reviews provided by other reviewers. I do like their reports, and I believe that by addressing the reviewers' comments and incorporating their revisions, the article will significantly improve in terms of scientific rigor and contribution to the field. *

*Reviewer #1 (Significance (Required)):

This manuscript is a solid scientific work addressing gaps in our knowledge about Lipid Transfer Proteins by employing state-of-the-art methods. It advances the field on conceptual and fundamental levels. This study is of interest to both computational biophysicists and physical chemists (to whom I belong myself) as well as experimentalists, who seek a rational explanation of the experimental observations. *

We thank the reviewer again for the positive evaluation of the significance of our work.

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

* Summary:

In a combined computational and experimental study, the authors provide insights into general features of lipid transfer proteins (LTPs), which play key roles in lipid trafficking: Through molecular dynamics simulations of a diverse set of 12 shuttle-like LTPs, they demonstrate that LTPs consistently exist in an equilibrium between two or more conformations, whose populations are modulated by a bound lipid, and that residues significantly involved in these collective conformational changes typically interact with a membrane. Their simulations indicate that conformational plasticity is a general feature of LTPs, leading them to suggest that the ability to change conformations is essential for LTP function. They test the generality of this hypothesis through in cellulo assays of two LTPs (STARD11 and Mdm12) that were not originally simulated. While experiments of STARD11 support their hypothesis, those presented for Mdm12 provide ambiguous results. *

*

Major comments: *

* Throughout the manuscript, it's stated that common 'dynamical features' correlate with LTP function. The accuracy of this statement is unclear since 'dynamical features' are never precisely defined and, while equilibrium conformational ensembles are characterized, dynamics (ie kinetics or time-dependent observables) are not. Please clarify.*

We plan to improve the scholarly presentation of our article to clarify this issue. In short, two distinct properties modulate protein function: 1. Conformational plasticity, i.e. the (thermodynamic) ability of the protein to adopt different conformations (and with different populations depending on the bound substrate). 2. Conformational “dynamics”, i.e. the propensity to exchange between these different thermodynamic states. This ability depends on the free energy barriers between different states and it is intrinsically a kinetic (rather than thermodynamic) property.

*More importantly, further evidence is needed to determine a correlation with *function*. LTPs are suggested to have faster transfer rates (a measure of function) if the apo form adopts a substantial population of holo-like conformations, akin to enzyme preorganization. This is further tested by rationally mutating STARD11 and Mdm12. However, the support for this conclusion and if these mutations alter the LTPs conformational ensembles as desired is unclear: *

In our opinion, the interpretation suggested by Reviewer #2 that there is a “correlation” between transfer rates and the overlap of apo-like and holo-like conformations, though fascinating, cannot be derived from the available data at this stage, and we did not mean to imply as such. Rather, lipid transport is a complex phenomenon that involves several steps (membrane binding/unbinding, lipid uptake/release,…). Our simulations indicate that protein conformational plasticity, including potentially the overlap between apo-like and holo-like conformations, also influences lipid transfer rates. We will clarify this aspect in the text.

* Is there a quantitative correlation between the overlap of apo and holo conformational distributions (as could be quantified by KL divergence or Wasserstein distance, for example) and difference in transfer rates as suggested by Fig S6?*

We plan to compute quantitative correlation between apo and holo conformational distribution for Fig.S6 and for mutant simulations (see answer below) but, as discussed above, we are skeptical that we will observe a clear correlation.

* The conclusion and the generality of the findings would be greatly strengthened if a correlation can be shown for other LTPs through additional simulations of mutants whose transfer rates have been previously characterized experimentally in the literature. (For example: Ryan 2007 PMID 17344474, Grabon 2017 PMID 28718450, Iaea 2015 PMID 26168008, among many others)*

We are currently running simulations of several mutants to address this point and provide additional data/context.

* While differences in the apo conformational ensembles of the WT and mutants are observed in Fig S7b and d, if these mutations reduce overlap with holo-like conformations is not determined. Simulations of the WT holo forms are needed to properly test this hypothesis. *

We are currently performing these simulations.

• For Mdm12, mutations are specifically made to "lock the protein in the apo-like state;" however, the mutant adopts conformations distinct from the apo form as show in Fig S7d. How do the authors interpret the results of the cellular assays considering this and could it help explain why the mutant has similar kinetics to WT? What may explain the puzzling results of similar transfer kinetics but differing mitochondrial morphology? *

As discussed above, interpretation of lipid transport rates based exclusively on apo and holo conformational population is premature, as this is a complex mechanism that depends on many variables. For what concerns the experimental results, we think three explanations are possible: 1. Mitochondrial morphology could be more sensitive to small variations in lipid composition than our METALIC assay. 2. Our assay only quantifies transport of unsaturated PC and PE species, and we can’t quantify variations in transport of other lipid species that are likely to also be transported by ERMES, such as PS and PA. 3. According to a recent structural model (Wozny et al, Nature 618, 88–192, 2023), Mdm12 might be part of a tunnel-like LTP complex in which it doesn't establish direct interactions with nearby organellar membranes. As such, its mechanism might be different from the one described here for other shuttle-like lipid transport domains. We will discuss these possibilities in the main text.

• Confounding factors potentially complicate the interpretation of the in cellulo experiments. Simpler in vitro experiments may be better suited to determine if altering LTP's biophysical properties, namely rationally altering the population of apo- vs holo-like configurations, quantitatively affects transport rates as suggested.*

We agree with Reviewer #2 that this information could be useful. However, this is beyond our technical abilities, and it would require lengthy and expensive experiments that are unlikely to be completed within a reasonable time framework for a revision (3 months). We have rather opted to better discuss our model in the context of published in vitro lipid transport experiments.

• The abstract, intro, and title highlight that the manuscript's findings are indicative of and correlated with *function* but on p. 12 it's foreseen "that future studies will focus on the functional consequence of such observation." Please reconcile these conflicting statements and ensure connections to function are accurately described. The current title is rather bold. *

We will rewrite and clarify the extent of our hypotheses and validations.

* All mentions of "correlation" throughout the manuscript need to be quantitatively evaluated or properly qualified. In addition to that mentioned above regarding Fig S6, what is the correlation coefficient between residues' contribution to PC1 and membrane interaction frequency (Fig 2)? *

To address this point, we will quantify the correlation between residues' contribution to PC1 and membrane interaction frequency. However, we expect a low correlation between residues' contribution to PC1 and membrane interaction frequency for at least two main reasons. __ First, not all residues contributing to PC1 interact with membranes, but only a subset, as discussed above. Second, our methodology to compute membrane binding, based on the geometric distance between residues and bilayer, is intrinsically quite noisy (since residues in proximity of bona fide membrane binding regions will also appear as involved in membrane binding), thus making quantification of correlations somewhat inaccurate. Rather, we will try to explain in the text that our observations are not of "correlation" but rather of dependence/association, and we will use quantitative measures to quantify these properties (such as rank correlation coefficients or multivariate analyses).__

* Residue's contributions to collective conformational changes are found to be indicative of membrane binding. Yet, membrane interacting residues are identified from CG simulations that cannot capture such collective conformational changes due to the use of an elastic network. Given that the CG simulations agree with previous experimental findings, this suggests that collective conformational changes are not important for membrane binding. *

We disagree with this interpretation by Reviewer #2 of our data: we do not claim that residue's contributions to collective conformational changes is indicative of membrane binding. Rather, membrane binding happens at protein regions displaying high contribution to collective conformational changes. This distinction is subtle but important: protein motion does not determine membrane binding regions. Rather, it appears that, for LTPs, membrane binding regions are also characterized by collective motions (suggesting function). We will clarify this in the main text.

*Are similar conclusions drawn from residues' RMSFs? In other words, are local conformational fluctuations just as indicative of membrane binding? *

We will compute protein residues’ RMSFs and compare it with the membrane binding data. However, given that RMSF is representative of thermal fluctuations, we again expect a bad correlation between RMSF and membrane binding. On the other hand, we indeed observe that most membrane binding regions are protein loops, but this is not unexpected (e.g. Tubiana et al, PLoS Comput Biol. 2022 Dec; 18(12): e1010346.). However, such observation does not provide any information on lipid transport, but only on the mechanism of membrane binding. Rather, the observation of a relationship between membrane binding and global motion is more interesting, since the latter is often indicative of protein function.

*The stated correlation may in fact be spurious and instead arise because residues at the entrance to LTP's hydrophobic cavities need to be positioned at the membrane surface for productive lipid uptake and these same residues must undergo significant conformational changes to allow lipid entry. *

This is exactly what we think it is happening and what our data suggest. However, one must remember that our simulations allow us to predict the membrane binding interface, that is often difficult to determine experimentally (and often via indirect evidence). Hence our data provide novel evidence in this direction.

*Is proximity to cavity entrance more or less correlated with membrane binding than 'dynamics'? *

If we consider that, as discussed before, dynamics does not correlate with membrane binding (there are many dynamical regions that are not at the membrane interface), it is safe to assume that proximity to cavity entrance would correlate more with membrane binding. However, we have to consider that often we do not know where the cavity entrance in LTPs is located simply based on structure alone, and hence our approach provides important clues into this process.

p.12 speculatively suggests "the high degree of protein dynamics we observed in membrane proximal regions could potentially facilitate the energetically unfavorable reaction that involves the extraction of a lipid from a membrane." Yet, the logic behind this idea does not make sense since a free energy barrier, an equilibrium thermodynamic quantity, cannot be lowered by changes in dynamics. Please explain.*

Our current understanding of the mechanism of lipid extraction is quite poor. However, both using chemical intuition and following a recent MD study on one LTP (Rogers et al, 2023, Plos Comp Biol), it is safe to assume that the hydrophobic environment around the lipid is important for its stabilization in the lipid bilayer. Hence, reducing the number of hydrophobic contacts between the lipid and its environment could facilitate transport. A highly dynamic protein, by cycling between different conformations, could “stir” the bilayer, and hence decrease the number of contacts between the lipid and its environment favoring transport. We will clarify this point in the text.

*Examining how the LTPs impact membrane properties would offer insight into the functional relevance of such residues for lipid extraction. *

Indeed, our point above is connected to this one. We are performing simulations to compute hydrophobic contacts in bilayer as proposed in (Rogers et al, 2023, Plos Comp Biol).

The authors highlight that a bound lipid alters LTPs' conformational ensembles akin to "conformational selection" or "induced fit." How sensitive are these findings to the bound lipid species? Do LTPs with multiple known substrates exhibit an increasing diversity of holo conformations and are different conformations stabilized by different substrates? Would similar observations (Fig 3) be made with a lipid that is not known to be transferred by a given LTP? An interesting future direction would be to examine if lipid substrate specificity could be assessed by comparing conformational ensembles to that of a known substrate and/or by overlap with the apo ensemble.

We deem that the role of lipid specificity on LTP conformational plasticity is beyond the scope of the current work. While this topic is certainly worth future investigations, we must point out that (i) not all proteins bind/transport multiple lipids (at least according to current knowledge) and (ii) only few LTPs have been structurally characterized bound to different lipids (Osh4, Osh6, …). This limitation prevents a wide generalization, and we prefer not to speculate on this topic. So far, we have tested our approach for Osh4 bound to cholesterol or PI(4)P and found that indeed the protein exhibits different holo conformations (in agreement with the experimental data) when bound to different substrates. We have added a short comment on this topic in the Discussion section.

"____We foresee that future studies will focus on the functional consequence of such observation, and most notably to the characterization of the extent to which such conformational changes affect multiple steps of protein function, including membrane binding or lipid extraction and release, and whether these are further modulated when different lipids are being transported."

For LTPs to transfer lipids between membranes, transitions between apo and holo forms ought to occur when LTPs are membrane bound. How does membrane binding influence the conformational ensembles observed in solution? Does it promote conformational changes between apo- and holo-like structures, as suggested to regulate lipid uptake and release by previous studies of Osh/ORP, Ups/PRELI, and START family members? (For example: Miliara 2019 PMID 30850607, Watanabe 2015 PMID 26235513, Grabon 2017 PMID 28718450, Iaea 2015 PMID 26168008, Kudo 2008 PMID 18184806, Dong 2019 PMID 30783101) While answering these questions would require further computational effort, doing so will allow more accurate assessment of the role of conformational changes in LTP function.

We can’t unfortunately currently quantify how membrane binding influences the conformational ensembles observed in solution, as the slowdown in diffusion at the water-membrane interface makes this task computationally challenging (and certainly not feasible within the time framework of a review). We have so far tested two different proteins and have not succeeded in converging their conformational distribution when membrane-bound despite long MD simulations that lasted several months (even though the non-converged data indicate sampling of both “open” and “closed” conformations). Interestingly, our observations are in qualitative agreement with a recent study on CPTP (Rogers et al, PLOS Comp Biol, 2023), where membrane-bound CPTP is able to sample different conformations (“open” and “closed”) but not to transition between the two states in 300 ns-long MD simulations.

* The authors motivate the study with the *assumption* that a common molecular mechanism of LTP function exists. Yet LTPs have evolved diverse sequences, structures, and substrate preferences; thus there seems to be no a priori requirement (or even necessarily a benefit) for a single molecular mechanism. What evidence then supports this premise? While previous studies are limited to individual LTPs, when viewed altogether retrospectively, they suggest features that could be shared among LTPs. Synthesizing previous studies and more thoroughly referencing them (only 5 are cited in the intro on p. 3) would strengthen both the premise and findings of the manuscript. *

Indeed, despite having different structures, substrates and the ability to target distinct organelles, previous evidence on LTPs seem to suggest a potential role for protein conformational plasticity for function, e.g. for Osh/ORP (Jun Im et al, Nature 2005; Canagarajah et al, JMB 2008; Moser von Filseck et al, Nat Comm, 2015; Lipp et al, Nat Comm. 2019,...), StART (Arakane et al, PNAS, 1996; Feng et al, Biochemistry, 2000; Grabon et al, JBC, 2017; Khelashvili et al, eLife, 2019;...) and PITP domains (Tremblay et al, Archives of Biochemistry and Biophysics, 2005; Ryan et al, MBOC, 2007; …). Our simulations provide additional evidence in this direction and allow for generalizing these observations, allowing to draw parallelisms with “enzyme-like” or transporter-like” features that could be exploited for further design of testable hypotheses. We will rewrite our text to better contextualize/acknowledge previous findings and to clarify these points.

*The LTPs investigated are known to target distinct membranes. Should they then be expected to share structural or sequence-based features predictive of membrane binding interfaces, as motivates the analysis in Fig 1d, 1e, and S3? Or is it beneficial for LTPs to recognize membranes in different ways? *

Since membrane binding is membrane/organelle-specific, it is possible that residue’s diversity in membrane binding interfaces could indeed be beneficial for this diversity. We will add this comment as a potential explanation of our finding of a lack of conserved sequence-based features for membrane binding interfaces.

*

Minor comments:*

* 2 "making lipid transfer across the cytoplasm a potentially energetically favorable process": Is it meant that it is less energetically costly than transfer without a LTP? Why it would be energetically favorable is unclear (and would indicate that the LTP sequesters lipids away from membranes instead of transferring them between membranes). *

Yes, this is what we meant. We will rewrite this appropriately.

* 3 "The excellent agreement between the membrane interface determined from the simulations and the experimentally-proposed one available for... Osh6" is missing a citation. *

We have now added the relevant citation.

* The plots in Fig 1d and S3 are difficult to interpret. Bar plots, for example, would allow easier comparison and evaluation. Currently, it seems that most proteins individually exhibit some of the same trends observed among the whole set, counter to the conclusion on p 5. *

We will improve the presentation of our Figures.

* Negatively charged residues engage in a number of membrane interactions (Fig 1d and S3). What is a potential explanation for this unconventional observation? *

One possible interpretation is that negatively charged residues could interact with positively charged moieties (ethanolamine, choline) of PC and PE lipids.

* How much variance is captured by PC1, and how many PCs are needed to capture most of the variance in the conformations? *

PC1 explains 38 % of the total variance, by average, whereas PC2 accounts for 17 % of it. Therefore, PC1 and PC2 capture most of the variance in almost all cases.

We have also added this to the text:

"____We specifically focused on PC1 as it explains most of the variance in the dynamics (38% on average for all the proteins in our dataset, see Supplementary Table 2).____ "

We have computed this variance and we have added this analysis in Supplementary Information.

* Plots in Fig 3, especially panels c and d are difficult to see. Please make the panels larger (perhaps a 3 x 4 layout instead of 2 x 6 would work better). *

We will improve the presentation of our Figures.

* 8 "these conformational changes are localized in protein regions that interact with the lipid bilayer" is contradicted by the results in Fig 2b showing that all residues with large contributions to PC1 do not interact with the membrane and discussed on p 5. *

As discussed above, we don’t observe “correlation” between membrane binding and conformational plasticity, but we rather observe that membrane binding regions display high conformational plasticity (the opposite is not true). We will further clarify in the text.

*

8 "in the absence of bound lipids, it is able to sample multiple conformations" is not supported by the orange distributions in Fig 3d that appear unimodal. Is it instead meant that the apo form exhibits larger variance in cavity volume? *

Yes, this is what we meant. We’ll clarify.

*

Please clarify if the elastic network was constructed to maintain the holo or apo structures of each protein and if a bound lipid was used in the CG simulations. *

For membrane binding CG simulations, we used the apo structure and no bound lipid was used in the simulations. However, analogous simulations in the holo form (not shown) have essentially identical membrane binding interfaces.

*

Was *CHARMM* TIP3P used? *

Yes.

* Please clarify how membrane interacting residues were defined and how interaction frequency was calculated from the longest duration of interaction. *

We will add this explanation in the Methods. The method is identical to (Srinivasan et al, Faraday Discussion, 2021).

* Refs 16 and 45 refer to the same paper. *

Thanks, it is now corrected!

* Reviewer #2 (Significance (Required)): *

* General assessment: *

* The work aims to tackle a grand question regarding membrane homeostasis mechanisms-what are universal principles underlying LTP function-and offers initial insights; however, further evidence is needed to support the conclusions as written, and some key results require further investigation and explanation. *

*Advance and audience: *

*

By concurrently investigating the largest number of lipid transfer proteins to-date, the authors provide data invaluable for uncovering general mechanisms of non-vesicular lipid transport and advancing our understanding of membrane homeostasis mechanisms. By illuminating the wide-spread importance of conformational plasticity among lipid transfer proteins, the work presents a conceptual advance in our understanding of lipid transfer mechanisms and unifies previous studies. Because the manuscript emphasizes common biophysical principles and draws connections to enzyme biophysics, it ought to be of interest not only to membrane biologists but biochemists and molecular biologists more broadly.*

We thank Reviewer #2 for the very positive evaluation of the significance of our work and for the in-depth analysis provided that will certainly help improve the quality of our work.

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

*The article "Conformational dynamics of lipid transfer domains provide a general framework to decode their functional mechanism." by Sriraksha Srinivasan, Andrea DiLuca, Arun Peter, Charlotte Gehin, Museer Lone, Thorsten Hornemann, Giovanni D'Angelo and Stefano Vanni study the interaction of Lipid transport Domains with membranes. This is done mainly by molecular modelling but also with selected experimental validations. *

* Major comments: *

* - The key conclusions are generally well supported by the analysis. - The authors could however analyze in more details some aspects in which specific cases appear. For example, p3 "multiple binding and unbinding events, as shown by the minimum distance curves" does not give an entire description of the variability seen in Fig S1, e.g. LCN1 versus GM2A.*

We now discuss in more detail the variability seen in Fig. S1 and attribute it to different membrane binding affinities of the proteins in our dataset. We also discuss how this variability could reflect the diversity of organellar membranes to which these proteins bind in vivo.

"____Notably, the proteins in our dataset display distinct binding affinities, with some proteins showing very transient binding while others remain membrane-bound for most of the simulation trajectory (Fig. S1). This behavior could be, in part, attributed to the wide diversity of organellar membranes to which the LTDs in our dataset bind to in vivo, and to the comparative simplicity of our in silico model DOPC lipid bilayers."

• Later the "excellent agreement" for the data in Fig S2 is not quantified which does not allow the reader to know whether it better than would have been with other methods (SASA, OPM, DREAM). *

We have explicitly quantified this agreement by providing a direct comparison between the experimental results and our in silico assay, and we further compared it against two alternative methods: OPM and DREAMM. In detail, we have identified 12 experimentally-characterized spots suggested to be involved in membrane binding in our protein dataset (see shaded blue regions in Fig. S2). Of those 12, our method identifies all of them (100%), while DREAMM identifies 7 of them (58 %) and OPM 4 out of 8 (50 %), since of the 12 proteins we tested, only 7 are available in the OPM database. Overall, even if our approach is much noisier than the others, and thus suggesting multiple binding regions that are not currently supported by experimental observations, using physics-based methodologies appears to remain a preferable strategy to characterize the binding of peripheral proteins to lipid bilayers. Given the limited size of our dataset, we prefer not to make a direct comparison between our assay and OPM/DREAMM in the main text as this won't be representative of the various methodologies.

*p5 commenting on Fig2b the case of Osh6 that appears to disagree should probably be mentioned. *

We now discuss this case, and attribute to this disagreement to insufficient sampling for the peculiar case of Osh6:

"____One interesting exception in our database appears to be Osh6, where the experimentally determined membrane-binding region at the N-terminus (https://doi.org/10.1038/s41467-019-11780-y) is only marginally binding to the lipid bilayer in silico and it also appears to have limited contribution to PC1. However, our simulations are unable to sample the large conformational changes that the N-terminal lid of Osh6 has been proposed to undergo from its lipid-bound to its apo state, indicating that insufficient sampling could be the reason for this apparent discrepancy."

*

-The data and the methods are generally well presented allowing to be reproduced.

• The experiments adequately replicated with adequate statistical analysis. *

* Minor comments: *

* - When presenting the dataset the authors could probably detail a bit more the protocol undertaken to chose the cases. In particular it is unclear whether the chosen proteins have any membrane selectivity, which in principle could be affected by the choice of lipid used here.*

We have now added in Table 1 a column with a list of potential organelles the different LTPs have been shown to localize to (source: UniProt). As model membrane bilayer, we opted to use a pure DOPC bilayer, for both simplicity and to compare membrane binding in a uniform setting. We foresee that future studies investigating the membrane specificity of the various proteins will shed further light into the molecular mechanism of LTPs. Finally, we also indicate that our choice of proteins was mainly driven by the availability of lipid-bound structures in the protein data bank. We have added the following sentences in the main text:

"____Specifically, we selected all LTPs for which a crystallographic structure in complex with a lipid was available at the start of our project, plus two additional proteins (GM2A and LCN1) to increase the structural diversity of our dataset (Fig. 1a)"

and

"____Notably, the proteins in our dataset display distinct binding affinities, with some proteins showing very transient binding while other remain membrane-bound for most of the simulation trajectory (Fig. S1). This behavior could be, in part, attributed to the wide diversity of organellar membranes to which the LTDs in our dataset bind to in vivo, and to the comparative simplicity of our in silico model DOPC lipid bilayers."

*- The authors could probably give some indication of how much of the variance is explained by PC1 and comment briefly on the choice to ignore other PCs. *

PC1 explains 38 % of the total variance, on average. This means that PC1 has a large contribution to the variance, especially in comparison to the other PCs. For instance, PC2 only accounts for 17 % of the total variance. This is the reason we limited our discussion to PC1. We have added a table in supplementary Information quantifying the variance explained by PC1 and PC 2 and added the following sentence in the main text:

"____We specifically focused on PC1 as it explains most of the variance in the dynamics (38% on average for all the proteins in our dataset)____. "

* - When analyzing the residues involved in the interaction with the membrane the results could probably be compared with that of the systematic analysis performed recently: Tubiana, T., Sillitoe, I., Orengo, C., & Reuter, N. (2022). Dissecting peripheral protein-membrane interfaces. PLOS Computational Biology, 18(12), e1010346. *

We have added in the text a reference to the work by Tubiana et al and we have further stressed that our results agree with previous observations (including theirs). This includes the preference for Lys over Arg and the importance of protruding hydrophobes:

"____Concomitant analysis of all LTDs (Fig. 1d) indicates that the membrane binding interface of LTDs is enriched in the positively charged amino acid Lysine, as this amino acid is less membrane-disruptive than Arginine22, and aromatic/hydrophobic ones (Phe, Leu, Val, Ile). This confirms previous observations, as (i) binding of negatively charged lipids via positively charged residues and (ii) hydrophobic insertions are two of the main mechanisms involved in membrane binding by peripheral proteins22-27."

* - In the discussion on allostery/conformational selection might not be centered so much on enzymes. *

We thank the reviewer for this important observation. We have now included in the Discussion the following paragraph that provides additional references and discussion of membrane transporters and receptors.

"____Notably, the conformational plasticity we observe for LTPs is reminiscent of other, previously described, functional protein mechanisms, including enzyme dynamics during catalysis (____DOI: 10.1126/science.1066176____), the alternating-access model of membrane transporters (____https://doi.org/10.1038/nsmb.3179____) or GPCR dynamics (____https://doi.org/10.1021/acs.chemrev.6b00177____). In all these cases, protein dynamics is strongly coupled to ligand binding and protein function, be it for signaling, transport or enzymatic activity. Unlike for these fields, however, the contribution of structural and spectroscopic studies to uncover LTP dynamics remains quite limited, and our simulations provide an important contribution to fill this gap. We hope that our results will motivate researchers to increase efforts to experimentally quantify LTPs conformational plasticity, e.g. by structural determination of LTPs in different states (or bound to different lipids) or by single-molecule spectroscopy studies."

* Reviewer #3 (Significance (Required)): *

*

The article shows convincing results on the debated issue of the mechanism of lipid transport by lipid transfer proteins. *

First the study employs molecular modelling to allow a rather large test on 12 cases. The molecular dynamics experiments allow the authors to draw clear hypotheses on role of protein dynamics on the interaction with membranes and the effect on bound lipids on the modification of this dynamics.

*Then the authors use this knowledge to design experiments that largely confirm those hypotheses. The results should therefore be interesting for a large audience of biochemists and cell biologists interested in lipid transport in the cell. *

We thank Reviewer #3 for its very positive evaluation and contextualization 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

The article "Conformational dynamics of lipid transfer domains provide a general framework to decode their functional mechanism." by Sriraksha Srinivasan, Andrea DiLuca, Arun Peter, Charlotte Gehin, Museer Lone, Thorsten Hornemann, Giovanni D'Angelo and Stefano Vanni study the interaction of Lipid transport Domains with membranes. This is done mainly by molecular modelling but also with selected experimental validations.

Major comments:

• The key conclusions are generally well supported by the analysis.
• The authors could however analyze in more details some aspects in which specific cases appear. For example, p3 "multiple binding and unbinding events, as shown by the minimum distance curves" does not give an entire description of the variability seen in Fig S1, e.g. LCN1 versus GM2A. Later the "excellent agreement" for the data in Fig S2 is not quantified which does not allow the reader to know whether it better than would have been with other methods (SASA, OPM, DREAM). p5 commenting on Fig2b the case of Osh6 that appears to disagree should probably be mentioned.
• The data and the methods are generally well presented allowing to be reproduced.
• The experiments adequately replicated with adequate statistical analysis.

Minor comments:

• When presenting the dataset the authors could probably detail a bit more the protocol undertaken to chose the cases. In particular it is unclear whether the chosen proteins have any membrane selectivity, which in principle could be affected by the choice of lipid used here.
• The authors could probably give some indication of how much of the variance is explained by PC1 and comment briefly on the choice to ignore other PCs.
• When analyzing the residues involved in the interaction with the membrane the results could probably be compared with that of the systematic analysis performed recently: Tubiana, T., Sillitoe, I., Orengo, C., & Reuter, N. (2022). Dissecting peripheral protein-membrane interfaces. PLOS Computational Biology, 18(12), e1010346.
• In the discussion on allostery/conformational selection might not be centered so much on enzymes.

Significance

The article shows convincing results on the debated issue of the mechanism of lipid transport by lipid transfer proteins.

First the study employs molecular modelling to allow a rather large test on 12 cases. The molecular dynamics experiments allow the authors to draw clear hypotheses on role of protein dynamics on the interaction with membranes and the effect on bound lipids on the modification of this dynamics. Then the authors use this knowledge to design experiments that largely confirm those hypotheses.

The results should therefore be interesting for a large audience of biochemists and cell biologists interested in lipid transport in the cell.

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 a combined computational and experimental study, the authors provide insights into general features of lipid transfer proteins (LTPs), which play key roles in lipid trafficking: Through molecular dynamics simulations of a diverse set of 12 shuttle-like LTPs, they demonstrate that LTPs consistently exist in an equilibrium between two or more conformations, whose populations are modulated by a bound lipid, and that residues significantly involved in these collective conformational changes typically interact with a membrane. Their simulations indicate that conformational plasticity is a general feature of LTPs, leading them to suggest that the ability to change conformations is essential for LTP function. They test the generality of this hypothesis through in cellulo assays of two LTPs (STARD11 and Mdm12) that were not originally simulated. While experiments of STARD11 support their hypothesis, those presented for Mdm12 provide ambiguous results.

Major comments:

Throughout the manuscript, it's stated that common 'dynamical features' correlate with LTP function. The accuracy of this statement is unclear since 'dynamical features' are never precisely defined and, while equilibrium conformational ensembles are characterized, dynamics (ie kinetics or time-dependent observables) are not. Please clarify.

More importantly, further evidence is needed to determine a correlation with function. LTPs are suggested to have faster transfer rates (a measure of function) if the apo form adopts a substantial population of holo-like conformations, akin to enzyme preorganization. This is further tested by rationally mutating STARD11 and Mdm12. However, the support for this conclusion and if these mutations alter the LTPs conformational ensembles as desired is unclear:

• Is there a quantitative correlation between the overlap of apo and holo conformational distributions (as could be quantified by KL divergence or Wasserstein distance, for example) and difference in transfer rates as suggested by Fig S6?
• The conclusion and the generality of the findings would be greatly strengthened if a correlation can be shown for other LTPs through additional simulations of mutants whose transfer rates have been previously characterized experimentally in the literature. (For example: Ryan 2007 PMID 17344474, Grabon 2017 PMID 28718450, Iaea 2015 PMID 26168008, among many others)
• While differences in the apo conformational ensembles of the WT and mutants are observed in Fig S7b and d, if these mutations reduce overlap with holo-like conformations is not determined. Simulations of the WT holo forms are needed to properly test this hypothesis.
• For Mdm12, mutations are specifically made to "lock the protein in the apo-like state;" however, the mutant adopts conformations distinct from the apo form as show in Fig S7d. How do the authors interpret the results of the cellular assays considering this and could it help explain why the mutant has similar kinetics to WT? What may explain the puzzling results of similar transfer kinetics but differing mitochondrial morphology?
• Confounding factors potentially complicate the interpretation of the in cellulo experiments. Simpler in vitro experiments may be better suited to determine if altering LTP's biophysical properties, namely rationally altering the population of apo- vs holo-like configurations, quantitatively affects transport rates as suggested.
• The abstract, intro, and title highlight that the manuscript's findings are indicative of and correlated with function but on p. 12 it's foreseen "that future studies will focus on the functional consequence of such observation." Please reconcile these conflicting statements and ensure connections to function are accurately described. The current title is rather bold.

All mentions of "correlation" throughout the manuscript need to be quantitatively evaluated or properly qualified. In addition to that mentioned above regarding Fig S6, what is the correlation coefficient between residues' contribution to PC1 and membrane interaction frequency (Fig 2)?

Residue's contributions to collective conformational changes are found to be indicative of membrane binding. Yet, membrane interacting residues are identified from CG simulations that cannot capture such collective conformational changes due to the use of an elastic network. Given that the CG simulations agree with previous experimental findings, this suggests that collective conformational changes are not important for membrane binding. Are similar conclusions drawn from residues' RMSFs? In other words, are local conformational fluctuations just as indicative of membrane binding? The stated correlation may in fact be spurious and instead arise because residues at the entrance to LTP's hydrophobic cavities need to be positioned at the membrane surface for productive lipid uptake and these same residues must undergo significant conformational changes to allow lipid entry. Is proximity to cavity entrance more or less correlated with membrane binding than 'dynamics'?

p. 12 speculatively suggests "the high degree of protein dynamics we observed in membrane proximal regions could potentially facilitate the energetically unfavorable reaction that involves the extraction of a lipid from a membrane." Yet, the logic behind this idea does not make sense since a free energy barrier, an equilibrium thermodynamic quantity, cannot be lowered by changes in dynamics. Please explain. Examining how the LTPs impact membrane properties would offer insight into the functional relevance of such residues for lipid extraction.

The authors highlight that a bound lipid alters LTPs' conformational ensembles akin to "conformational selection" or "induced fit." How sensitive are these findings to the bound lipid species? Do LTPs with multiple known substrates exhibit an increasing diversity of holo conformations and are different conformations stabilized by different substrates? Would similar observations (Fig 3) be made with a lipid that is not known to be transferred by a given LTP? An interesting future direction would be to examine if lipid substrate specificity could be assessed by comparing conformational ensembles to that of a known substrate and/or by overlap with the apo ensemble.

For LTPs to transfer lipids between membranes, transitions between apo and holo forms ought to occur when LTPs are membrane bound. How does membrane binding influence the conformational ensembles observed in solution? Does it promote conformational changes between apo- and holo-like structures, as suggested to regulate lipid uptake and release by previous studies of Osh/ORP, Ups/PRELI, and START family members? (For example: Miliara 2019 PMID 30850607, Watanabe 2015 PMID 26235513, Grabon 2017 PMID 28718450, Iaea 2015 PMID 26168008, Kudo 2008 PMID 18184806, Dong 2019 PMID 30783101) While answering these questions would require further computational effort, doing so will allow more accurate assessment of the role of conformational changes in LTP function.

The authors motivate the study with the assumption that a common molecular mechanism of LTP function exists. Yet LTPs have evolved diverse sequences, structures, and substrate preferences; thus there seems to be no a priori requirement (or even necessarily a benefit) for a single molecular mechanism. What evidence then supports this premise? While previous studies are limited to individual LTPs, when viewed altogether retrospectively, they suggest features that could be shared among LTPs. Synthesizing previous studies and more thoroughly referencing them (only 5 are cited in the intro on p. 3) would strengthen both the premise and findings of the manuscript.

The LTPs investigated are known to target distinct membranes. Should they then be expected to share structural or sequence-based features predictive of membrane binding interfaces, as motivates the analysis in Fig 1d, 1e, and S3? Or is it beneficial for LTPs to recognize membranes in different ways?

Minor comments:

p. 2 "making lipid transfer across the cytoplasm a potentially energetically favorable process": Is it meant that it is less energetically costly than transfer without a LTP? Why it would be energetically favorable is unclear (and would indicate that the LTP sequesters lipids away from membranes instead of transferring them between membranes).

p. 3 "The excellent agreement between the membrane interface determined from the simulations and the experimentally-proposed one available for... Osh6" is missing a citation.

The plots in Fig 1d and S3 are difficult to interpret. Bar plots, for example, would allow easier comparison and evaluation. Currently, it seems that most proteins individually exhibit some of the same trends observed among the whole set, counter to the conclusion on p 5.

Negatively charged residues engage in a number of membrane interactions (Fig 1d and S3). What is a potential explanation for this unconventional observation?

How much variance is captured by PC1, and how many PCs are needed to capture most of the variance in the conformations?

Plots in Fig 3, especially panels c and d are difficult to see. Please make the panels larger (perhaps a 3 x 4 layout instead of 2 x 6 would work better).

p. 8 "these conformational changes are localized in protein regions that interact with the lipid bilayer" is contradicted by the results in Fig 2b showing that all residues with large contributions to PC1 do not interact with the membrane and discussed on p 5.

p. 8 "in the absence of bound lipids, it is able to sample multiple conformations" is not supported by the orange distributions in Fig 3d that appear unimodal. Is it instead meant that the apo form exhibits larger variance in cavity volume?

Please clarify if the elastic network was constructed to maintain the holo or apo structures of each protein and if a bound lipid was used in the CG simulations.

Was CHARMM TIP3P used?

Please clarify how membrane interacting residues were defined and how interaction frequency was calculated from the longest duration of interaction.

Refs 16 and 45 refer to the same paper.

Significance

General assessment:

The work aims to tackle a grand question regarding membrane homeostasis mechanisms-what are universal principles underlying LTP function-and offers initial insights; however, further evidence is needed to support the conclusions as written, and some key results require further investigation and explanation.

Advance and audience:

By concurrently investigating the largest number of lipid transfer proteins to-date, the authors provide data invaluable for uncovering general mechanisms of non-vesicular lipid transport and advancing our understanding of membrane homeostasis mechanisms. By illuminating the wide-spread importance of conformational plasticity among lipid transfer proteins, the work presents a conceptual advance in our understanding of lipid transfer mechanisms and unifies previous studies. Because the manuscript emphasizes common biophysical principles and draws connections to enzyme biophysics, it ought to be of interest not only to membrane biologists but biochemists and molecular biologists more broadly.

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

Srinivasan et al. present a comprehensive study on systematizing the structure-dynamics-function relation of lipid transfer proteins (LTPs), combining extensive molecular simulations and complementary experiments. Indeed, the current state-of-the-art in the field is quite chaotic and fractional, and such systematic studies are necessary to advance our general and conceptual understanding of the mechanisms of action of LTPs. The selected techniques and research strategies are all suitable, their description is sufficient and enables reproducibility; the obtained results are carefully presented and discussed; the conclusions are adequately supported by the data.

Given my primarily computational background, I evaluated mainly the simulation part of the manuscript. Considering experiments, I do not see any significant flows or deficiencies that could diminish the value of the data and following conclusions given in the manuscript. I would even suggest improving the abstract by more explicitly saying that this work includes experimental measurements because it currently reads like purely computational work was performed.

Major comments:

1. Although I like the central message of the paper and have no objections, I am curious whether the conclusion "a more "dynamic" or/and "mobile" part of the protein interacts with the membrane or any other (macro)(bio)molecule" makes sense globally and is not limited to LTPs. For example, it is a reasonable assumption that a more flexible part of the protein, i.e., capable of adopting necessary binding configurations, would be a more likely interacting spot. Locking in a less flexible and more specific configuration upon binding with a target molecule is also anticipated and quite typical, e.g., when ligands interact with target proteins, thereby blocking their function. The authors themselves recognize this paradigm as referring to the enzymes' dynamics. It would be great if authors could comment more on dynamics-function relation, referring to the existing literature, where such observations were/were not observed for different protein families. Performing simulations on proteins that do not exhibit such a feature and do not belong to LTPs, but, e.g., structurally similar to some of the studied LTPs, would be an excellent addition too, highlighting this signature characteristic of LTPs.

Minor comments:

1. Fig 1d. What is so special in Lysine compared to Arginine? Is there any disbalance in their presence in studied proteins? Any correlations between the binding affinity of certain amino acids and their overall presence on the protein surface?
2. Fig S1. GM2A and TTPA seem to be irreversibly adsorbed to the membrane on the microsecond timescale in most replicas. Is anything special in these proteins? Did this affect the sampling of a claimed membrane-binding interface?
3. A related follow-up question. Multiple replicas were performed to identify the membrane-binding interface. However, if I understand well, the initial orientation of the protein with respect to the membrane was always the same. I found it a pity since performing multiple replicas starting from different initial geometries (e.g., rotating the protein in a somewhat systematic way) would likely result in a more efficient exploration of the conformation space. Can the authors comment on whether this predefined initial configuration could negatively affect the results? Performing a few additional simulations for the most problematic proteins I mentioned earlier (GM2A and TTPA) could be a nice opportunity to apply this strategy.
4. How was the volume of the cavity affected by mutations in STARD11 and Mdm12? Do these data somehow correlate with the experimentally observed reduced efficiency of the lipid transfer?
5. I would appreciate it if the authors considered playing with the templates of the main Figures at later stages because in the current version, and when printed on A4 paper, the readability of certain graphs and pictures is uncomfortable and sometimes even impossible. Obviously, the final schematics would depend on the journal and its formatting.

Referees cross-commenting

I would like to acknowledge the thoughtful and detailed reviews provided by other reviewers. I do like their reports, and I believe that by addressing the reviewers' comments and incorporating their revisions, the article will significantly improve in terms of scientific rigor and contribution to the field.

Significance

This manuscript is a solid scientific work addressing gaps in our knowledge about Lipid Transfer Proteins by employing state-of-the-art methods. It advances the field on conceptual and fundamental levels. This study is of interest to both computational biophysicists and physical chemists (to whom I belong myself) as well as experimentalists, who seek a rational explanation of the experimental observations.

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

Chakraborty et al describe the biochemical and structural characterization of Spiroplasma FtsZ and report that the protein has unusual properties compared to other FtsZ. Sedimentation and GTPase measurements showed that whereas the wild-type protein has a high critical concentration and low GTPase activity, a mutant predicted to facilitate FtsZ cleft opening (F224M) exhibited lower critical concentration and higher GTPase activity. In addition, the crystal structures of both wild-type and F224M SmFtsZ revealed a unique domain-swapped dimer configuration in which one of the monomers in each dimer exhibited an R/T hybrid or intermediate conformation, with the NTD in the T state and the CTD in the R state. The T state of FtsZ has only been observed before when the protein crystallizes as filaments. Thus, the crystal structure of SmFtsZ - which is not assembled in filaments - was interpreted as capturing a conformational state that could explain the kinetic polarity of FtsZ (preferential addition of subunits to the CTD-exposed end of FtsZ filament).

This is a good quality manuscript overall, but which could still be improved by the suggestions below. In terms of significance, it provides new data to support current models for FtsZ assembly mechanism but no major new insights. The findings are interesting for a more specialized audience.

Major points

1. The peculiar biochemical properties of SmFtsZ (high CC, low GTPase) are well documented and interesting but deserve further critical assessment to rule out artifacts. The EM in Fig 1B suggests abundant aggregated protein (not monomers), in addition to filament bundles, which suggests that SmFtsZ is not stable under the experimental conditions used. There are reports that some FtsZ will lose nucleotide during purification and become partially unfolded and unstable (doi.org/10.1111/febs.15235). Figure S1E suggests that the same may be happening here, as the amount of GDP released by SmFtsZ seems to be lower than expected if all the protein had nucleotide. Perhaps the authors should repeat their experiments with SmFtsZ purified in the presence of GDP, which should stabilize the protein, to confirm that the biochemical properties of the protein stay the same.
2. Another unexpected observation is that the SmFtsZ bundles are quite short despite the low GTPase activity of the protein, whereas mutant F224M forms much longer bundles and is a stronger GTPase. In general, filament length correlates inversely with GTPase activity, if measurements are being made at steady state. However, no kinetic (light scattering or fluorescence) experiments seem to have been done to ensure measurements were done in steady state. The authors do try to explain the odd behavior of SmFtsZ but the idea that the increase in GTPase reflects a faster kinetics of nucleation and elongation is not necessarily true. GTP turnover is usually limited by the kinetics of filament disassembly not by assembly. However, it is possible that in reactions with a mutant that is much better at nucleation there will be many more filaments than with a poorly nucleating protein and, thus, more filament ends for subunit turnover. A complicator to these experiments is that they were carried out in high magnesium and at pH 6.5 which favor bundling, and bundling affects subunit and GTP turnover in ways that are hard to account for. Ideally, experiments aimed at properly determining the kinetic properties of FtsZ should be carried out under conditions that avoid bundling (pH 7.4-7.7, 2-5 mM Mg2+) and include proper kinetic measurements, such as light scattering. Thus, before any hard conclusions can be drawn about the properties of SmFtsZ, the authors may wish to revisit some of their biochemical experiments in light of the caveats pointed out here.
3. A central part of the paper is the description of the intermediate R/T conformation but that was a bit confusing and perhaps could be improved. The first thing would be to more clearly define what are the structural changes of the NTD in the T conformation. From other publications, it seems that the NTD undergoes little alteration upon switching to the T conformation, the main one being the flipping of the guanine of the bound nucleotide. But if the NTD structure remains essentially the same, what causes the flipping of the guanine? My impression was that guanine flipping was caused by the downward movement of H7 but if H7 and its attached elements (H6, S6) are moving, why is this not manifested as a significant structural change in the NTD in the T state? Moreover, from Figs. 3C and 5A we conclude that the relative position of H7 in the R/T structures is the same as in R structures. If H7 has not changed in the R/T structure, can you call this a T structure? Also, if there is no H7 movement, what caused the change in guanine angle?
4. The observation that the intermediate conformation was detected in a swapped-dimer is always a matter of some concern, as domain swapping imposes additional constraints on the conformational freedom of a protein and generates structures that are often different from their non-swapped counterparts. This seems to be the case for other FtsZ domain-swapped structures, which were outliers in the extensive comparisons made by Wagstaff et al (doi.org/10.1128/mBio.00254-17) and also stand out in the analysis in Fig. 3BC. Perhaps the authors should discuss more thoroughly why this structure must reflect a natural conformation of FtsZ.
5. Still regarding the structural basis of kinetic polarity, it would be desirable to present a more complete view of the debate in the field about this issue. For example, Ruiz et al, (doi.org/10.1371/journal.pbio.3001497) recently provided structural arguments for the NTD being the face used for monomer addition without detecting the same intermediate form reported in this manuscript. How do their data and arguments differ from your findings? More generally, isn´t the fact that the NTD does not change substantially as FtsZ transitions from R to T already an argument for the NTD being the surface used for monomer addition?
6. l. 74 "led us to propose a structural basis for the kinetic polarity of FtsZ, where transition of the NTD to the T-state conformation driven by GTP binding is sufficient to add a GTP-bound monomer to the bottom interface of the FtsZ filament." This statement suggests that GTP is necessary for the intermediate conformation but this is not supported by the data, as the GDP bound 7YSZ structure also has one monomer in the intermediate conformation. As far as I can tell, there is no structural evidence to suggest that the nucleotide gamma phosphate plays any role in the R-T transition. Even the role of the gamma phosphate in organizing the T3 loop in an assembly-conducive conformation seems to still be a controversial matter in the field. According to Matsui 2014 (doi.org/10.1074/jbc.M113.514901) "based on the results of the present study as well as on the structures deposited previously by other groups (PDB codes 2RHL, 2RHO, 2Q1X, and 2Q1Y) (43, 44), nucleotide exchange appears not to directly induce a structural change in the monomer, including the T3 loop."
7. The experiments with the reciprocal cleft mutation in E. coli are not very informative as it is difficult to correlate the division defect in vivo with specific kinetic defects of the mutant FtsZ. The authors should have at least done a basic characterization of the E. coli mutant in vitro to demonstrate that it is altered in its CC alone. In fact, the dominant negative effect of the mutation in vivo is not something one expects from a poorly nucleating protein, which, if anything, should have a hard time poisoning the endogenous protein. The effect on ring compaction also suggests that the mutation must affect the protein in a broader way, perhaps including filament geometry. I would suggest that this part of the manuscript could be excluded without any loss for the SmFtsZ conclusions.

Minor points

1. l. 14 "CTD of the nucleotide-bound monomer cannot bind to the NTD-exposed end of the filament unless relative rotation of the domains leads to cleft opening." This is not accurate. There is no steric impediment to this reaction. Monomers in R conformation should be able to add to the NTD end of the filament as well, even if this is slower than the opposite reaction. The absence of growth from the NTD end is because the rate of addition/conformational change is slower than the rate of GTP hydrolysis.
2. The comparison between the 7YOP (B) structure and the S. aureus 3WGN structure to show the effect of the gamma phosphate on T3 loop structure should be presented in a single figure, instead of being split between Fig. 4 and Fig. S2, and preferably using similar poses of the two structures. In the current state, it is quite hard to visualize the similarities mentioned by the authors.
3. In contrast to what´s in the main text (l. 130), the chain with continuous density in Figure 2 is assigned as B, not A. Please clarify which is correct.
4. l. 271 pBAD is the plasmid name, not the promoter. The promoter is PBAD(subscript).

Significance

This is a good quality manuscript overall, but which could still be improved by the suggestions below. In terms of significance, it provides new data to support current models for FtsZ assembly mechanism but no major new insights. The findings are interesting for a more specialized 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 #2

Evidence, reproducibility and clarity

Summary

This is a study of cell division protein SmFtsZ from Spiroplasma melliferum, a cell wall-less Mollicutes bacterium where FtsZ may provide the primary force for division. Using X-ray crystallography, biochemical and microbiological experiments, the authors provide insight into how FtsZ's relaxed (R) to tense (T) conformational switching and its GTPase activity explain the kinetic polarity of FtsZ filament treadmilling. They propose: 1) an intermediate R/T state of FtsZ that facilitates preferential binding of N-terminal domain (NTD) of a monomer to C-terminal domain (CTD) of the terminal subunit at the filament bottom end; 2) that R to T switching is the rate-limiting step of FtsZ polymerization; 3) a T3 loop mechanism for GTP gamma phosphate triggering FtsZ polymerization.

All comments and criticisms below are made for the sake of this interesting study.

General Comments

The study is well thought, carefully executed, and the manuscript is well written. However, the first conclusion is not convincing, because it is based on a misleading analysis; and the third conclusion is complicated by the use of an unqualified analog of GTP. SmFtsZ crystallizes as a dimer with the NTD and CTD domains swapped and a NTD-NTD contact. Mechanistic conclusions drawn from this unusual structural context are largely speculative, as they may not hold for normal FtsZ assembly. The NTD structure changes really very little between R-FtsZ and T-FtsZ, so that it is probably incorrect to define a T-NTD/R-CTD intermediate conformation from the guanine ring angle, as will be reasoned below. In addition, the GTP analog GMPPNP employed to investigate the effects of the gamma phosphate is not demonstrated to promote FtsZ assembly as GTP does; in fact, its beta-gamma phosphate geometry in the SmFtsZ structure is clearly different from GTP in other FtsZ structures. This raises the concern that GMPPNP may be not a bona fide functional analog of GTP for FtsZ.

The authors should moderate the first and third claims, keeping speculations for discussion. Additional experiments required to assess the activity of GMPPNP inducing FtsZ assembly should be reported, even if the result was negative. Authors may consider partially refocusing the manuscript, including the title, towards the mechanism of the R to T transition, with Phe224Met modulating the opening of the cleft between NTD and CTD. Careful discussion of the structural basis of the kinetic polarity of FtsZ filaments in the light of previous and current results should be fine. In addition, SmFtsZ is now the third FtsZ that has been crystallized as a domain swapped dimer, suggesting a tendency for intramolecular dissociation of NTD and CTD with potential mechanistic implications, as the authors point out by the attractive end of the discussion.

Specific comments (major and minor)

Line 37 should read "...connected via H7 helix (15), with divergent C-terminal extensions".

Line 55 Please note that the kinetic polarity of FtsZ has been deduced from mutational analysis (rather than observed as in the case of microtubules)

Line 75 ""..transition of the NTD to the T-state conformation driven by GTP binding is sufficient..." This sentence appears conceptually wrong, because the R of T conformation of FtsZ is deemed independent of GTP or GDP binding in the literature (for example, Ref 23)

Line 89 should read " in the presence of GTP and Mg2+ and not with GDP and Mg2+"

Line 90-Figure 1A. The greyish gel electrophoresis image and those in SI require improving staining or photos. Standard Coomassie staining typically gives less background and better contrast.

Line 90. Were the SmFtsZ filaments single or multiple in EM?

Line 92. "...other characterized bacterial FtsZs" Some references should be cited

Line 107 suggesting -> indicating

Lines 120-122 " a truncated construct....SmFtsZdeltaCt showed similar GTPase activity as the wild type" is repeated from lines 110-111 above

Line 124. Why the choice of GMPPNP, rather than GTP or GMPCPP?. Have SmFtsZ structures with GTP or GMPCPP been attempted?

Line 124. It would be helpful to the reader to explain here that structure 7YSZ has two GDP-bound chains whereas structure 7YOP has GDP in chain A and GMPPNP in chain B.

Lines 131 and 133. The names of chain A and chain B are swapped in the text Figure 2A-D. Consider enhancing the nucleotide tracing for easier visualization

Line 141 and Figure 2F. Why change from the refraction detector in panel E to the absorbance detector in panel F? Importantly, how to know whether the shoulder corresponds to a dimer or to an extended monomer, and was the column calibrated?. In any case, do extended monomers and domain swapped dimers exist in solution? Additional crosslinking experiments, and analytical ultracentrifugation if available, could provide interesting results, although this is not a strict requirement for this manuscript.

Lines 155-157 and Figure 3. The "GTP-bound T-state" 3WGN structure is not GTP but GTP-gamma S-bound, which makes a difference. 3WGM is GTP-bound SaFtsZ, although with a truncated loop T7. There is an unnecessary mix of FtsZs from different species in structures 2RHL and 3VOA; using instead 5H5G-molecule A (T-state, GDP) and 5H5G-molecule B (R-state, GDP) would simplify the structures employed for comparison in Figure 3 to a single species, SaFtsZ. In fact, 5H5G is employed as a reference in Figure 3C, although the distinction between 5H5G molecules A and B is not mentioned.

Lines 157-160. The guanine ring angle depends on a stacking interaction with Phe183 form helix H7, which shifts in known FtsZ R and T structures. But this part of the structure is actually missing from the so called "T-state GDP" and "T-state GTP" SmFtsZ swapped domain structures. Instead, the guanine ring interacts with the main chain carbonyl of Phe137, an interaction which is not observed in the standard R or T FtsZ structures employed for comparison. This makes using the guanine ring angle alone misleading for conformational classification of SmFtsZ. In addition, both SmFtsZ "T-state" structures show a R-like Arg29 disengaged from interacting with the guanine (Figure 3), contrary to the interaction observed in the FtsZ T conformation. The overall conformation of the SmFtsZ structures does correspond to R-FtsZ. However, the swapped domain context of the SmFtsZ structure hampers meaningful comparisons with other FtsZ structures at a detailed local level around the guanine ring.

Lines 175-177. "We concluded that in B chain the nucleotide-bound NTD is in T-state...". Importantly, the structure of the NTD of FtsZ, not including helix H7, is known to be very similar in the R and T conformations; differences are the position of helix H7, the position of the CTD relative to the NTD and the opening of the interdomain cleft (refs 23 and 24). The guanine ring angle is clearly related to the H7-Phe183 shift. Therefore, distinguishing R and T-conformations of the NTD in FtsZs and in SmFtsZ in particular seems unsupported by experimental data.

Lines 197-199. Checking known FtsZ structures shows that Gly71 in loop T3 can be flipped out or in with GDP in both R and T conformation, whereas it is out with GTP or its analogs, making room for the gamma phosphate. It is interesting that the authors now observe this change with SmFtsZ, comparing the structures of GDP-bound and GMPPNP-bound protein. However, they should analyze and mention the precedents in the PDB, not only the GTP-gamma-S-bound 3WGN, and draw their conclusion very carefully due to the swapped domain context. There are known interactions made by the nucleotide gamma phosphate (PDB 3WGM) and one analog (PDB 7OHK) across the association interface in FtsZ filaments that explain FtsZ polymerization. In addition, is loop T3 really stabilized by the gamma phosphate of by filament formation?

Lines 202-210. Tyr145 is not part of loop T5 but of helix H5. The observed interplay between loop T3 Pro73 and H5 Tyr145 is an attractive feature (apparently reminiscent of the tubulin T3-T5 story, but see Discussion). Please indicate if this has not been pointed out before in other FtsZs with the residue corresponding toTyr145, and consider analyzing existing FtsZ structures for T3-H5/T5 cross talk in different nucleotide states.

Lines 212-324. The last three sections of Results convincingly demonstrate how residue 224 Phe/Met in the cleft between CTD and NTD modulates SmFtsZ assembly, EcFtsZ assembly, and E. coli cell division. In addition to this study, is it known whether SmFtsZ can replace EcFtsZ for E. coli cell division?

Line 220 and Figure S3A. Please explain the color code in this Figure.

Line 243. How can it be proposed that SmFtsZF224M could not be crystallized with GMPPNP probably due to efficient filament formation, if the activity of GMPPNP inducing filament formation has not been documented?

Figure 6 panel F. The NeonGreen Z-ring microscopy images need enlargement to be properly appreciated.

Discussion Line 342. Please notice that loop T3 is not always disordered with GDP. The proposal lacks an analysis of other FtsZ structures, in addition to 3WGN, and ignores intermolecular interactions of the nucleotide gamma phosphate and the coordinated Mg2+ ion (Matsui et al, 2014 J Biol Chem; Ruiz et al, 2022 PLoS Biol).

Discussion Lines 356-370. The similarities to the classical GTP/GDP-dependent T3-T5 cross talk in the tubulin-RB3 complex (reviewed in Ref 27) is appealing, but notice that this was curved R-state tubulin with an accessory protein. But maybe the nucleotide dependent T3-T5 cross talk does not take place in T-tubulin from cryoEM microtubule structures with GDP and GTP (LaFrance et al and Nogales 2022 PNAS)?. And the authors should carefully check the tubulin T3 and T5 loop GDP/GTP-dependent conformations in the recently available cryoEM structures of free tubulin heterodimers (R-state) bound to GDP (PDB 7QUC) and GTP (PDB 7QUD) without any accessory proteins, which differ from the classical view.

Discussion Lines 371-379. It should be noticed that a simpler interpretation of the results is that SmFtsZ is in the R-state, with R-CTD and R-H7, whereas the NTD is practically the same in both R and T states, as for other FtsZs (Ref 23). The T-like guanine angle may result from anomalous interactions of the swapped domains in SmFtsZ.

Discussion Lines 381-384. There is really no need to postulate a NTD transition from R- to T-state in order to propose a kinetic polarity for the FtsZ filament from structure. In fact, having the NTD conformation constant results in a monomer top interface that is pre-formed for association and with the help of GTP should bind to the filament bottom subunit, as already proposed in Ref 35.

Referees cross-commenting

In addition to the concerns shared by the reviewers, especially those related to the existance or the role of distinct R and T conformations of the NTD of FtsZ, as welll as the individual reviewer concerns, we would like to highlight the relevance of:

The comment of reviewer 1, requiring more information on the biological role of FtsZ in cell division of Spiroplasma and whether it forms a ring.

Comment 2 of reviewer 3, requiring time-dependance of SmFtsZ polymerization and GTPase data, which are essential for properly analyzing the GTPase activity.

Significance

This interesting work, if successfully revised, will provide valuable insight into how the FtsZ polymerization switch and the nucleotide binding loops work for assembly of polar filaments, employing FtsZ from a wall-less bacterium. Please see the comments above for the existing literature context of the manuscript. This paper will be possibly suitable for a general biological audience, in addition to microbiologists and cytoskeletal researchers.

This review has been prepared by biochemistry and structural biology experts familiar with FtsZ, hoping that it may be useful to the authors.

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 paper reports the biochemical and structural characterization of FtsZ from a wall-less bacterial organism, Spiroplasma melliferum. From the analysis of the crystal structures (and comparison to known structures) the authors propose a model to explain the kinetic polarity of FtsZ treadmilling that was derived from a mostly genetic analysis (ref 26). In that model FtsZ with GTP bound adds to the bottom end of a filament with the C-terminal domain then shifting to the T-state. The status of the N-ter (whether in the T or R-state was not considered). Here the authors have proposed that they have captured an intermediate with the N-ter in the T-state and the C-ter in the R-state. It is proposed that this form adds to the bottom of a filament with the C-ter now adopting the T-state. I am not convinced this model is supported by the data as it is not clear when the N-ter domain switches to the T-state (before or after addition the end of a filament).

Significance

Note: the authors define the N-ter being in the T-state vs R-state based on the orientation of the guanine ring. The C-ter domain is in the T vs R-state based on whether the cleft is open or closed respectively.

The basis of the authors' model comes mostly from the analysis of the crystal structure of FtsZ from the wall-less bacterial that was obtained in this study. The crystal structure revealed an unusual swapped dimer. Although in solution this FtsZ is a monomer, it crystallized as a swapped dimer indicating that during crystallization FtsZ domains came apart and reassociated with the opposite domains from another monomer. Careful analysis reveals that in one monomer the N-ter domain is in the T-state whereas in the other the N-ter domain is in the R-state (independent of nucleotide; GDP or GMPCPP). Although the N-ter domain is in the T-state in both monomers there are some differences - with GMPCPP the T3 loop is ordered whereas it is disordered with GDP. Also, they propose that the orientation of Gly71 is such in the GTP state that it favors interaction with the bottom end of a filament.

Is it known whether FtsZ assembles into a Z ring and is required for cell division in this organism?

In the dimer both C-terminal domains are in the R-state. From this the authors propose that the one monomer in the dimer is in the R-state whereas the other is in transition state (T-for N-ter and R-for C-terminal domain).

The authors analyze sequences of FtsZ from different bacteria and notice that position 242 is a Phe in their organism whereas it is a Met in other bacteria. They wonder whether this residue influences the C-ter transitioning to the T-state so they swap residues - putting a Met in Sm and a Phe in Ecoli at this position. Interestingly, they notice that Sm-FtsN-met results in increased GTPase activity but longer filaments - this seems contradictory as higher GTPase is usually associated with shorter filaments - e.g. increased Mg slows GTPase activity and increased filament length and bundling. The Ec-FtsZ-Phe mutant displays increased cell length but not sure this can be ascribed to an effect on the GTPase activity.

Overall, the work is well done and it comes to interpretation and whether the data support the model. The emphasis is on the N-ter getting to the T-state, but I am not sure that is the important step. It seems to me that rate-limiting step in FtsZ assembly is the C-ter getting into the T-state, which happens when a subunit is added to the end of a filament. Obviously the N-ter has to get to the T-state as well but how that happens is not clear. Presumably, it happens as a GTP-bound monomer in the R-state adds to the end of a filament resulting in the N-ter adopting the T-state followed by the C-ter adopting the T-state. In other words the T-state is only achieved by addition of a subunit to the end of the filament.

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-Localization of ESYT1 and SYNJ2BP

The claim of a localization at ER-mitochondria contacts relies on two type of assays. Light microscopy and subcellular fractionation. Concerning microscopy, while the staining pattern is obviously colocalizing with the ER (a control of specificity of staining using KO cells would nevertheless be desirable)

the idea that ESYT1 foci "partially colocalized with mitochondria" is either trivial or unfounded

Every cellular structure is "partially colocalized with mitochondria" simply by chance at the resolution of light microscopy

If the meaning of the experiment is to show that ESYT1 'specifically' colocalizes with mitochondria, then this isn't shown by the data

There is no quantification that the level of colocalization is more than expected by chance

nor that it is higher than that of any other ER protein

Moreover, the author's model implies that ESYT1 partial colocalization with mitochondria is, at least partially, due to its interaction with SYNJ2BP. This is not tested.

• To analyze and measure MERCs parameters and functions, we used a set of validated methods described in the following specialized review articles (Eisenberg-Bord, Shai et al. 2016, Scorrano, De Matteis et al. 2019).
• To support and confirm the localization of ESYT1-SYNJ2BP complex at MERCs, we performed supplementary BioID analysis using ER target BirA*, OMM targeted BirA* and ER-mitochondria tether BirA* (Table S1, Figure S1 and Figure 1 A and B). These results confirmed the specificity of the interaction of the 2 partners. ESYT1 is not identified as a prey in OMM BioID and SYNJ2BP is not identified in ER BioID, on the other hand both partners are identified in the ER-mitochondria tether BioID.
• To improve our description of the partial localization of ESYT1 at mitochondria, we performed a quantitative analysis using confocal microscopy on control human fibroblasts stably overexpressing SEC61B-mCherry as an ER marker which were labelled with ESYT1 and TOMM40 for mitochondria. We measured the % of ESYT1 signal colocalizing with mitochondria and the % of mitochondria positive for ESYT1 (Figure 1E).

• To demonstrate than ESYT1 partial colocalization with mitochondria is, at least partially, due to its interaction with SYNJ2BP, we performed a quantitative analysis using confocal microscopy. Human control fibroblasts, KO SYNJ2BP fibroblasts and SYNJ2BP overexpressing fibroblasts were labelled with ESYT1, TOMM40 for mitochondria and CANX for ER. We measured the % of ESYT1 signal colocalizing with mitochondria in each condition (Figure 3C). Membranes (MAM) can be purified and are enriched for proteins that localize at ER-mitochondria contacts. This idea originated in the early 90's and since then, myriad of papers has been using MAM purification, and whole MAM proteomes have been determined. Yet the evidence that MAM-enriched proteins represent bona fide ER-mitochondria-contact-enriched proteins (as can nowadays be determined by microscopy techniques) remain scarce. Here, anyway, ESYT1 fractionation pattern is identical to that of PDI, a marker of general ER, with no indication of specific MAM accumulation.

• To highlight the enrichment of ESYT1 in the MAM fraction, we quantified the ESYT1 signal in each fraction. Those results show a similar fractionation pattern than the MAM resident protein SIGMAR1 (Figure 1F). For SYNJ2BP, it is different as it is more enriched in the MAM than the general mitochondrial marker PRDX3. However, PRDX3 is a matrix protein, making it a poor comparison point, since SYNJ2BP is an OMM protein.

• To confirm the partial enrichment of SYNJ2BP in the MAM fraction compared to another outer mitochondrial membrane protein, we added the signal of the well characterized OMM protein CARD19 (Rios, Zhou et al. 2022). Again, the model implies that ESYT1 and SYNJ2BP accumulation in the MAM should be dependent on each other. This is not tested.

• As describe above, we demonstrated in Figure 3C than the accumulation of ESYT1 at mitochondria is, at least partially, dependent on the quantity of SYNJ2BP.

• We moreover showed a reciprocal effect in Figure 3E. A quantitative analysis using confocal microscopy demonstrated that the effect of SYNJ2BP overexpression on MERCs formation is partially dependent of the presence of ESYT1. 2-ESYT1-SYNJ2BP interaction.

The starting point of the paper is a BioID signal for SYNJ2BP when BioID is fused to ESYT1. One confirmation of the interaction comes in figure 4, using blue native gel electrophoresis and assessing comigration. Because BioID is promiscuous and comigration can be spurious, better evidence is needed to make this claim. This is exemplified by the fact that, although SYNJ2BP is found in a complex comigrating with RRBP1, according to the BN gel, this slow migrating complex isn't disturbed by RRBP1 knockdown, but is somewhat disturbed by ESYT1 knockdown. More than a change in abundance, a change in migration velocity when either protein is absent would be evidence that these comigrating bands represent the same complex.

• We showed in Figure 4C that the presence of SYNJ2BP in a complex of a similar molecular weight that ESYT1 (410KDa) is totally dependent of the presence of ESYT1, suggesting an interaction of the 2 proteins.

• To confirm this interaction, in figure 4A we analyzed on BN cells overexpressing SYNJ2BP together with a 3xFlag tagged version of ESYT1. As a result of the addition of the Flag tag, the complex positive for ESYT1 shifted to a higher molecular weight. The complex positive for SYNJ2BP shifted to a similar the molecular weight, demonstrating the interaction and dependence of the 2 partners. ESYT1-SYNJ2BP interaction needs to be tested by coimmunoprecipitation of endogenous proteins, yeast-2-hybrid, in vitro reconstitution or any other confirmatory methods.

• To confirm the interaction of the 2 partners, we performed co-immunoprecipitation of the ESYT1-3xFlag protein that we showed in Figure 1H to form complexes similar to the endogenous protein. SYNJ2BP is found as the strongest prey, followed by ESYT2 and SEC22B two described interactors of ESYT1, confirming the quality of the analysis (Table S2) (Giordano, Saheki et al. 2013, Gallo, Danglot et al. 2020). 3-Tethering by ESYT1- SYNJ2BP.

This is assessed by light and electron microscopy. Absence of ESYT1 decreases several metrics for ER-mitochondria contacts (whether absence of SYNJ2BP has the same effect isn't tested).

• Using PLA (proximity ligation assay) we demonstrated that the loss of SYNJ2BP leads to a decrease in MERCs (Figure 7 H and I), confirming previous studies (Ilacqua, Anastasia et al. 2022, Pourshafie, Masati et al. 2022). This interesting phenomenon could be due to many things, including but not limited to the possibility that "ESYT1 tethers ER to mitochondria".

This statement and the respective subheading title are therefore clearly overreaching and should be either supported by evidence or removed.

Indeed, absence of ESYT1 ER-PM tethering and lipid exchange could have knock-on effects on ER-mito contacts, therefore strong statements aren't supported.

Moreover, the effect on ER-mitochondria contact metrics could be due to changes in ER-mitochondria contact indeed but may also reflect changes in ER and/or mitochondria abundance and/or distribution, which favour or disfavour their encounter. Abundance and distribution of both organelles are not controlled for.

• The mitochondrial phenotypes caused by the loss of ESYT1 are all rescued by the introduction of an artificial mitochondrial-ER tether, demonstrating that they are due to loss of the tethering function of ESYT1. Finally, the authors repeat a finding that SYNJ2BP overexpression induces artificial ER-mitochondria tethering. Again, according to the model, this should be, at least in part, due to interaction with ESYT1. Whether ESYT1 is required for this tethering enhancement isn't tested.

• As described above, we demonstrated in Figure 3C that the accumulation of ESYT1 at mitochondria is, at least partially, dependent on the quantity of SYNJ2BP.

• We moreover showed a reciprocal effect in Figure 3F. A quantitative analysis using confocal microscopy demonstrated that the effect of SYNJ2BP overexpression on MERC formation is partially dependent of the presence of ESYT1. 4-Phenotypes of ESYT1/SYNJ2BP KD or KO.

The study goes in details to show that downregulation of either protein yields physiological phenotypes consistent with decreased ER-mitochondria tethering. These phenotypes include calcium import into mitochondria and mitochondrial lipid composition.

Figure 5 shows that histamine-evoked ER-calcium release cause an increase in mitochondrial calcium, and this increase is reduced in absence of ESYT1, without detectable change in the abundance of the main known players of this calcium import. This is rescued by an artificial ER-mitochondria tether. However, Figure 5D shows that the increase in calcium concentration in the cytosol upon histamine-evoked ER calcium release is equally impaired by ESYT1 deletion, contrary to expectation. Indeed, if the impairment of mitochondrial calcium import was due to improper ER-mitochondria tethering in ESYT1 mutant cells, one would expect more calcium to leak into the cytosol, not less.

The remaining explanation is that ESYT1 knockout desensitizes the cells to histamine, by affecting GPCR signalling at the PM, something unexplored here.

In any case, a decreased calcium discharge by the ER upon histamine treatment, explains the decreased uptake by mitochondria.

The authors argue that ER calcium release is unaffected by ESYT1 KO, but crucially use thapsigargin instead of histamine to show it. Thus, the most likely interpretation of the data is that ESYT1 KO affects histamine signalling and histamine-evoked calcium release upstream of ER-mitochondria contacts.

• Silencing ESYT1 impairs SOCE efficiency in Jurkat cells (Woo, Sun et al. 2020), but not in HeLa cells (Giordano, Saheki et al. 2013, Woo, Sun et al. 2020). Analysis of the role of ESYT1 in HeLa cells prevents confounding effects due to the loss of ESYT1 at ER-PM. In this model, knock-down of ESYT1 led to a decrease of mitochondrial Ca2+ uptake from the ER upon histamine stimulation, as monitored by genetically encoded Ca2+ indicator targeted to mitochondrial matrix (Figure 5A and B). ESYT1 silencing in HeLa cells did not impact ER Ca2+ store measured by the ER-targeted R-GECO Ca2+ probe (Figure 5C and D). The expression of the artificial mitochondria-ER tether was able to rescue mitochondrial Ca2+ defects observed in ESYT1 silenced cells (Figure 5B), confirming that the observed anomalies are specifically due to MERC defects.
• In contrast loss of ESYT1 impaired SOCE efficiency in fibroblasts (Figure 6 A and B). This phenotype was fully rescued by re-expression of ESYT1-Myc but not the artificial tether. We therefore investigated the influence of ESYT1 loss on cytosolic Ca2+ concentration following ATP (Figure 6F to H) or histamine stimulation (Figure S3 D to F), both of which showed a reduced cytosolic Ca2+ concentration and uptake in ESYT1 KO cells. This phenotype was fully rescued by the re-expression of ESYT1-Myc but not the artificial tether. Measurment of cytosolic Ca2+ after tharpsigargin treatment in Ca2+-fee media, an inhibitor of the sarco/endoplasmic reticulum Ca2+ ATPase SERCA that blocks Ca2+ pumping into the ER, showed that ESYT1 KO does not influence the total ER Ca2+ pool (Figure 6K and L). However, ER-Ca2+ release capacity upon histamine stimulation (Figure 6I and J) is decreased in ESYT1 KO cells. This phenotype was fully rescued by the re-expression of ESYT1-Myc but not the artificial tether. Loss of ESYT1 decreased the Ca2+ uptake capacities of mitochondria after activation with histamine (Figure S3 A to C) or ATP (Figure 6 C to E). This phenotype was rescued by re-expression of ESYT1-Myc and also the engineered ER-mitochondria tether. Thus, despite the ER-Ca2+ release defect observed after ESYT1 loss, the artificial tether fully rescued the mitochondrial phenotype.

• These results highlight the distinct and dual roles of ESYT1 in Ca2+ regulation at the ER-PM and at MERCs. The data with SYNJ2BP deletion are more compatible with decreased ER-mito contacts, as no decreased in cytosolic calcium is observed. This is compatible with the previously proposed role of SYNJ2BP in ER-mitochondria tethering, but the difference with ESYT1 rather argue that both proteins affect calcium signaling by different means, meaning they act in different pathways.

• We explain the different results concerning cytosolic calcium by the fact that ESYT1 is a bi-localized protein with dual functions on cellular calcium. Implicated both in SOCE at ER-PM and in mitochondrial calcium uptake at MERCs. On the other hand, SYNJ2BP is only present at MERCs and its loss do not influence PM-ER signaling or ER-Ca2+ release. Finally, the study delves into mitochondrial lipids to "investigated the role of the SMP-domain containing protein ESYT1 in lipid transfer from ER to mitochondria". In reality, it is not ER-mitochondria lipid transport that is under scrutiny, but general lipid homeostasis, and changes in ER-PM lipids could have knock-on effects on mitochondrial lipids without the need to invoke disruptions in ER-mitochondria transfer activity.

• The fact that the artificial tether, which specifically rescue MERCs, fully rescue the lipid phenotype argue for a direct loss of MERCs tethering function when ESYT1 is missing. The changes observed are interesting but could be due to anything. Surprisingly, PCA analysis shows that the rescue of the knockout by the ESYT1 gene clusters with the rescue by the artificial tether, and not with the wildtype. This indicates that overexpressing either ESYT1 or a tether cause similar lipidomic changes. These could be due, for instance, to ER stress caused by protein overexpression, and not to a rescue.

• In order to verify if the overexpression of ESYT1 or the artificial tether induces ER stress, we performed a WB analysis to compare markers of ER stress in control fibroblasts, KO ESYT1 fibroblasts, KO ESYT1 fibroblasts overexpressing ESYT1-Myc or the tether (Figure S4C). This showed no changes in the levels of several different markers of ER stress or cell death. __Reviewer 2____: __

1) the interaction between those proteins is direct,

2) if SYNJ2BP is necessary and sufficient to localize E-Syt1 at MERC, and

3) if MERCs extension induced by SYNJ2BP is dependent on E-Syt1.

Those points are important to investigate because SYNJ2BP has already been shown to induce MERCs by interacting with the ER protein RRBP1. In addition, some experiments need to be better quantified.

Major comments: E-syt1/SYNJ2BP in MERCs formation: the authors provide several convincing lines of evidence that both proteins are in the same complex (proximity labelling, localization in the same complex in BN-PAGE, localization in MAM) but it is not clear in which extent the direct interaction between both proteins regulates ER-mitochondria tethering. 1- Pull down experiments or BiFC strategy could be performed to show the direct interaction between both proteins.

• We showed in Figure 4C that the presence of SYNJ2BP in a complex of a similar molecular weight to that ESYT1 (410KDa) is totally dependent of the presence of ESYT1, suggesting an interaction of the 2 proteins.
• To confirm this interaction, in figure 4A we analyzed on BN cells overexpressing SYNJ2BP together with a 3xFlag tagged version of ESYT1. As a result of the addition of the Flag tag, the complex positive for ESYT1 shifted to a higher molecular weight. Significantly, the complex positive for SYNJ2BP shifted to a similar the molecular weight, demonstrating the interaction and dependence of the 2 protein partners.

• To confirm the interaction of the 2 partners, we performed co-immunoprecipitation of the ESYT1-3xFlag protein (Table S2). SYNJ2BP was found as the strongest prey, followed by ESYT2 and SEC22B two described interactors of ESYT1, confirming the quality of the analysis (Giordano, Saheki et al. 2013, Gallo, Danglot et al. 2020). 2- SYNJ2BP OE has already been demonstrated to increase MERCs and this being dependent on the ER binding partners RRBP1 (10.7554/eLife.24463). Therefore, it would be of interest to perform OE of SYNJ2BP in KO Esyt1 to address the question of whether ESyt1 is also required to increase MERCs.

• A quantitative analysis using confocal microscopy demonstrated that the effect of SYNJ2BP overexpression on MERCs formation is partially dependent of the presence of ESYT1 (Figure 3F). 3- The authors show that Esyt1 punctate size increases when SYNJ2BP is OE (Fig3C), but this can be indirectly linked to the increase of MERCs in the OE line. Thus, it could be interesting to test if the number/shape of E-syt1 punctate located close to mitochondria decreases in KO SYNJ2B. This could really show the dependence of SYNJ2BP for E-syt1 function at MERCs.

• To improve our description of the partial localization of ESYT1 at mitochondria, we performed a quantitative analysis using confocal microscopy on control human fibroblasts stably overexpressing SEC61B-mCherry as an ER marker which were labelled with ESYT1 and TOMM40 for mitochondria. We measured the % of ESYT1 signal colocalizing with mitochondria and the % of mitochondria colocalizing with ESYT1 (Figure 1E).

• To demonstrate than ESYT1 partial colocalization with mitochondria is, at least partially, due to its interaction with SYNJ2BP, we performed a quantitative analysis using confocal microscopy. Human control fibroblasts, KO SYNJ2BP fibroblasts and SYNJ2BP overexpressing fibroblasts were labelled with ESYT1, TOMM40 for mitochondria and CANX for ER. We measured the % of ESYT1 signal colocalizing with mitochondria in each condition (Figure 3C). Lipid analyses: the results of MS on isolated mitochondria clearly show that mitochondrial lipid homeostasis is affected on KO-Syt1 and rescued by expression of Syt1-Myc and artificial mitochondria-ER tether. However, p.15, the authors wrote "The loss of ESYT1 resulted in a decrease of the three main mitochondrial lipid categories CL, PE and PI, which was accompanied by an increase in PC ». As the results are expressed in mol%, this interpretation can be distorted by the fact that mathematically, if the content of one lipid decreases, the content of others will increase. I would suggest to express the results in lipid quantity (nmol)/mg of mitochondria proteins instead of mol%. This will clarify the role of E-Syt1 on mitochondrial lipid homeostasis and which lipid increase and decrease.

• We changed the sentence in the text as suggested. Also it could be of high interest to have the lipid composition of the whole cells to reinforce the direct involvement of E-Syt1 in mitochondrial lipid homeostasis and verify that the disruption of mitochondrial lipid homeostasis is not linked to a general perturbation of lipid metabolism as this protein acts at different MCSs.

• This is beyond the scope of the project and we would argue that the results of such an experiment would be difficult to interpret. To better understand the impact of Esyt1 of mitochondria morphology, the author could analyze the mitochondria morphology (size, shape, cristae) on their EM images of crt, KO and OE lines. Indeed, on OE (Fig3A), the mitochondria look bigger and with a different shape compared to crt.

• As we do not observe obvious differences in mitochondrial morphology between control, KO and OE fibroblasts we do not think that quantitative analysis would add to the understanding of the effect of ESYT1 on mitochondrial function. Also, they performed a lot of BN-PAGE. Is it possible to check whether the mitochondrial respiratory chain super-complexes are affected on Esyt1 KO line compared to crt?

• We decided to remove the data on the metabolic consequences of ESYT1 loss since it was too preliminary and required deeper investigations, focusing instead on the effect of ESYT1 loss on calcium homeostasis. Quantifications: some western blots needs to be quantified (Fig 5K, 6J, S3E);

• We did not observe obvious differences in the protein levels so we think that quantitation would not add significantly to the understanding of the differences in calcium dynamics that we report. Fig1A: Can the author provide a higher magnification of the triple labeling and perform quantification about the proportion of E-Syt1 punctate located close to mitochondria?

• We added higher magnification of the same area in all channels and arrows that point to the foci of ESYT1 colocalizing with both ER and mitochondria (Figure 1D).

• To improve our description of the partial localization of ESYT1 at mitochondria, we performed a quantitative analysis using confocal microscopy on control human fibroblasts stably overexpressing SEC61B-mCherry as an ER marker which were labelled with ESYT1 and TOMM40 for mitochondria. We measured the % of ESYT1 signal colocalizing with mitochondria and the % of mitochondria colocalizing with ESYT1 (Figure 1E). Minor comments:

• Fig1E + text: according to the legend, the BN-PAGE has been performed on Heavy membrane fraction. Why the authors speak about complexes at MAM in the text of the corresponding figure? Is-it the MAM or the heavy fraction (MAM + mito + ER...)? If BN have been performed from heavy membranes, it is not a real proof that E-syt1 is in MAMs.

• Heavy membranes have been used in this experiment. The text and conclusions have been changed accordingly.

• On fig3C (panel crt): it seems like SYNJ2BP dots are not co-localizaed with mito. Is this protein targeted to another organelle beside mitochondria?

• It is not described that SYNJ2BP would be targeted to another organelle beside mitochondria. It is possible that those dots outside of mitochondria could be non-specific signals from the antibody we used.

• Fig4A: can the author provide a control of protein loading (membrane staining as example) to confirm the decrease of E-Syt1 in siSYNJ2BP?

• As we performed this experiment only once we have removed the statement suggesting a decrease in ESYT1 protein in response to the siSYNJ2BP.

• Fig5E/F: it is not clear to me why the expression of E-Syt1 in the KO is not able to complement the KO phenotype for cytosolic Ca++. Can the authors comment this?

• We performed further analysis using ATP to trigger calcium release from the ER (figure 6 F to H). In those conditions, expression of ESYT1 in the KO is able to complement the KO phenotype for cytosolic Ca2+. __Reviewer 3____: __

Main points 1. Confirming the MERC localization of ESYT1 should include some more of tethering factors as demonstrated interactors (some are mentioned above) and should not be limited to lipid homeostasis.

• As shown in Figure 1B, VAPB, PDZD8 and BCAP31 are found as preys in the ESYT1 bioID analysis. Those proteins have been described as MERC tethers, their loss leading to mitochondrial calcium defects. To support and confirm the specificity of ESYT1-SYNJ2BP complex at MERCs, we performed a supplementary BioID analysis using ER targeted BirA* and OMM targeted BirA* (Table S1, Figure S1 and Figure 1 A and B). These results confirmed the specificity of the interaction of the 2 partners. ESYT1 is not identified as a prey in OMM BioID and SYNJ2BP is not identified in ER BioID. Additional ER-mitochondria tether BirA* analyses showed that tether-BirA* identified both ESYT1 and SYNJ2BP as a prey at MERCs, confirming the localisation of this interaction. Interestingly, a large majority of the known MERCs tethers VAPB-PTPIP51, MFN2, ITPRs, BCAP31 are also found as preys in the tether-BirA* (Figure 1B), confirming the quality of these data.
• To confirm the interaction of the 2 partners, we performed co-immunoprecipitation of the ESYT1-3xFlag protein. SYNJ2BP is found as the strongest prey, followed by ESYT2 and SEC22B two described interactors of ESYT1, confirming the quality of the analysis (Table S2) (Giordano, Saheki et al. 2013, Gallo, Danglot et al. 2020).

The fact that in ESYT1 KO cells both mitochondrial calcium transfer and cytosolic calcium accumulation are accompanied by decreased ER-cepia1ER signal decay upon histamine addition suggest that the main reason for ER-mitochondria calcium transfer defects are due to impaired SOCE. Calcium-free medium and histamine are used to show that ESYT1 does not affect ER calcium content. However, if it affects SOCE, then the absence of extracellular calcium would abolish such an effect; moreover, histamine does not test for leak effects. As additional information, the authors should investigate whether ER calcium content is affected by the presence of extracellular calcium in the ko scenario using thapsigargin. The authors should inhibit SOCE to test whether this mechanism is affected in ESYT1 KO and could account for observed signal differences. Excluding SOCE is critical, since any change in calcium entry from the outside would potentially negate a role of ESYT1 in mitochondrial calcium uptake.

• Silencing ESYT1 impairs SOCE efficiency in Jurkat cells (Woo, Sun et al. 2020), but not in HeLa cells (Giordano, Saheki et al. 2013, Woo, Sun et al. 2020). Analysis of the role of ESYT1 in HeLa cells prevents confounding effects due to the loss of ESYT1 at ER-PM. In this model, knock-down of ESYT1 led to a decrease of mitochondrial Ca2+ uptake from the ER upon histamine stimulation, as monitored by genetically encoded Ca2+ indicator targeted to mitochondrial matrix (Figure 5A and B). ESYT1 silencing in HeLa cells did not impact ER Ca2+ store measured by the ER-targeted R-GECO Ca2+ probe (Figure 5C and D). The expression of the artificial mitochondria-ER tether was able to rescue mitochondrial Ca2+ defects observed in ESYT1 silenced cells (Figure 5B), confirming that the observed anomalies are specifically due to MERC defects.
• In contrast loss of ESYT1 impaired SOCE efficiency in fibroblasts (Figure 6 A and B). This phenotype was fully rescued by re-expression of ESYT1-Myc but not the artificial tether. We therefore investigated the influence of ESYT1 loss on cytosolic Ca2+ concentration following ATP (Figure 6F to H) or histamine stimulation (Figure S3 D to F), both of which showed a reduced cytosolic Ca2+ concentration and uptake in ESYT1 KO cells. This phenotype was fully rescued by the re-expression of ESYT1-Myc but not the artificial tether. Measurment of cytosolic Ca2+ after tharpsigargin treatment in Ca2+-fee media, an inhibitor of the sarco/endoplasmic reticulum Ca2+ ATPase SERCA that blocks Ca2+ pumping into the ER, showed that ESYT1 KO does not influence the total ER Ca2+ pool (Figure 6K and L). However, ER-Ca2+ release capacity upon histamine stimulation (Figure 6I and J) is decreased in ESYT1 KO cells. This phenotype was fully rescued by the re-expression of ESYT1-Myc but not the artificial tether. Loss of ESYT1 decreased the Ca2+ uptake capacities of mitochondria after activation with histamine (Figure S3 A to C) or ATP (Figure 6 C to E). This phenotype was rescued by re-expression of ESYT1-Myc and also the engineered ER-mitochondria tether. Thus, despite the ER-Ca2+ release defect observed after ESYT1 loss, the artificial tether fully rescued the mitochondrial phenotype.
• These results highlight the distinct and dual roles of ESYT1 in Ca2+ regulation at the ER-PM and at MERCs.

The authors claim that ER-Geco measurements show that no change of ER calcium was observed. However, they use thapsigargin treatment and then get a peak, when the signal should show a decrease due to leak. This suggests they did not use ER-Geco in Figure S3C. What was measured and what does it mean?

• We used R-GECO (not ER-GECO) which measures the cytosolic calcium.
• We measured total ER Ca2+ store using the cytosolic-targeted R-GECO Ca2+ probe upon thapsigarin treatment, an inhibitor of the sarco/endoplasmic reticulum Ca2+ ATPase SERCA that blocks Ca2+ pumping into the ER (Figure 5C and D) and observed no difference in our different conditions.

The findings on growth in galactose medium are intriguing but are not accompanied by respirometry to confirm mitochondria are compromised upon ESYT1 KO.

• We decided to remove the data on the metabolic consequences of ESYT1 loss since it was to preliminary and required deeper investigations, focusing instead on the effect of ESYT1 loss on calcium homeostasis

Minor points: 1. The authors mention they measure mitochondrial uptake of "exogenous" calcium by applying histamine. They should specify that these measures transferred calcium from the ER rather than uptake of calcium from the exterior (directly at the plasma membrane).

• The text was clarified as suggested.

• Expression levels of IP3Rs are not very indicative of any change of their activity. The authors should discuss how ESYT1 could affect their PTMs.

• A large numer of post translational modifications are known to regulate IP3R activity (Hamada and Mikoshiba 2020), and it is possible that the loss of ESYT1 could interfere with these modifications, but an exploration of this issue is beyond the scope of this study. The text was clarified as suggested. Eisenberg-Bord, M., N. Shai, M. Schuldiner and M. Bohnert (2016). "A Tether Is a Tether Is a Tether: Tethering at Membrane Contact Sites." Dev Cell 39(4): 395-409.

Gallo, A., L. Danglot, F. Giordano, B. Hewlett, T. Binz, C. Vannier and T. Galli (2020). "Role of the Sec22b-E-Syt complex in neurite growth and ramification." J Cell Sci 133(18).

Giordano, F., Y. Saheki, O. Idevall-Hagren, S. F. Colombo, M. Pirruccello, I. Milosevic, E. O. Gracheva, S. N. Bagriantsev, N. Borgese and P. De Camilli (2013). "PI(4,5)P(2)-dependent and Ca(2+)-regulated ER-PM interactions mediated by the extended synaptotagmins." Cell 153(7): 1494-1509.

Hamada, K. and K. Mikoshiba (2020). "IP(3) Receptor Plasticity Underlying Diverse Functions." Annu Rev Physiol 82: 151-176.

Ilacqua, N., I. Anastasia, D. Aloshyn, R. Ghandehari-Alavijeh, E. A. Peluso, M. C. Brearley-Sholto, L. V. Pellegrini, A. Raimondi, T. Q. de Aguiar Vallim and L. Pellegrini (2022). "Expression of Synj2bp in mouse liver regulates the extent of wrappER-mitochondria contact to maintain hepatic lipid homeostasis." Biol Direct 17(1): 37.

Pourshafie, N., E. Masati, A. Lopez, E. Bunker, A. Snyder, N. A. Edwards, A. M. Winkelsas, K. H. Fischbeck and C. Grunseich (2022). "Altered SYNJ2BP-mediated mitochondrial-ER contacts in motor neuron disease." Neurobiol Dis: 105832.

Rios, K. E., M. Zhou, N. M. Lott, C. R. Beauregard, D. P. McDaniel, T. P. Conrads and B. C. Schaefer (2022). "CARD19 Interacts with Mitochondrial Contact Site and Cristae Organizing System Constituent Proteins and Regulates Cristae Morphology." Cells 11(7).

Scorrano, L., M. A. De Matteis, S. Emr, F. Giordano, G. Hajnoczky, B. Kornmann, L. L. Lackner, T. P. Levine, L. Pellegrini, K. Reinisch, R. Rizzuto, T. Simmen, H. Stenmark, C. Ungermann and M. Schuldiner (2019). "Coming together to define membrane contact sites." Nat Commun 10(1): 1287.

Woo, J. S., Z. Sun, S. Srikanth and Y. Gwack (2020). "The short isoform of extended synaptotagmin-2 controls Ca(2+) dynamics in T cells via interaction with STIM1." Sci Rep 10(1): 14433.

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

Janer et al. have identified ESYT1 as a novel tether between the ER and mitochondria (MERCs) with roles in lipid and calcium homeostasis. They discovered extended synaptotagmin (ESYT1) in a BioID screen, where it interacts with SYNJ2BP and forms a high molecular weight complex. The study addressed a lack of information at the level of mammalian cell system, where a key protein complex known from yeast (ERMES) is absent, suggesting other proteins take over this critical role. These proteins then control the production of cardiolipin and PE, two lipid types essential for the functioning of mitochondria. They contain SMP motifs as a signature domain required for lipid transport. ESYT1 had previously been found to mediate lipid transfer at the plasma membrane and at peroxisomes, but the authors found it also localizes to MERCs. In a BioID screen, they have found numerous ER proteins with known roles in MERC tethering (e.g., EMC complex, BAP31, VAPB or TMX1). They have decided to focus on the aforementioned pair, which they demonstrate is enriched on MERCs (ESYT1) and mitochondria (SYNJ2BP), respectively, forming high molecular weight complexes, as detected by BN gels. Unlike RRBP1-SYNJ2BP, this complex is not dependent on ongoing protein synthesis. Upon generation of ESYT1 KO fibroblasts, they show that this SMP protein compromises MERC formation through electron microscopy. SYNJ2BP overexpression specifically increases contacts, as again shown by EM, independent of mitochondrial dynamics.

In its present form, the manuscript accurately describes the role of the ESYT1-SYNJ2BP complex for MERCs. The study contains nice lipidomics that reinforce this point and suggest a metabolic consequence. This latter observation is, however, very basic and requires some extension by assaying respirometry. The calcium phenotype is currently not fully characterized either. Interference with SOCE remains a possibility and if true, this would compromise the statement that the complex also controls calcium signaling. Both would need to be investigated better to either confirm or reject these roles, in my opinion, an important question. Overall, the manuscript contains interesting characterization of a tether that could have important consequences for calcium signaling, which would be an exciting finding.

Main points

1. Confirming the MERC localization of ESYT1 should include some more of tethering factors as demonstrated interactors (some are mentioned above) and should not be limited to lipid homeostasis.
2. The fact that in ESYT1 KO cells both mitochondrial calcium transfer and cytosolic calcium accumulation are accompanied by decreased ER-cepia1ER signal decay upon histamine addition suggest that the main reason for ER-mitochondria calcium transfer defects are due to impaired SOCE. Calcium-free medium and histamine are used to show that ESYT1 does not affect ER calcium content. However, if it affects SOCE, then the absence of extracellular calcium would abolish such an effect; moreover, histamine does not test for leak effects. As additional information, the authors should investigate whether ER calcium content is affected by the presence of extracellular calcium in the ko scenario using thapsigargin.
3. The authors should inhibit SOCE to test whether this mechanism is affected in ESYT1 KO and could account for observed signal differences. Excluding SOCE is critical, since any change in calcium entry from the outside would potentially negate a role of ESYT1 in mitochondrial calcium uptake.
4. The authors claim that ER-Geco measurements show that no change of ER calcium was observed. However, they use thapsigargin treatment and then get a peak, when the signal should show a decrease due to leak. This suggests they did not use ER-Geco in Figure S3C. What was measured and what does it mean?
5. The findings on growth in galactose medium are intriguing but are not accompanied by respirometry to confirm mitochondria are compromised upon ESYT1 KO.

Minor points:

1. The authors mention they measure mitochondrial uptake of "exogenous" calcium by applying histamine. They should specify that this measures transferred calcium from the ER rather than uptake of calcium from the exterior (directly at the plasma membrane).
2. Expression levels of IP3Rs are not very indicative of any change of their activity. The authors should discuss how ESYT1 could affect their PTMs.

Significance

The study is certainly of high interest due to its implications for cell metabolism and calcium signaling. It contains very strong data on MERC formation and lipidomics. However, the calcium and metabolic aspects are currently not well developed and require improvements.

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 work of Janer and al. investigates the role of E-Syt1, a well known lipid transfer protein tethering ER and PM and ER and peroxisome, at ER-mitochondria contact sites (MERCs). E-Syt1 was identified has a putative MERCs component by proximity labeling performed from four SMP domain containing proteins. They identified the mitochondrial SYNJ2BP as a binding partner of E-Syt1 only. By different biochemical and microscopy approaches, they show that 1) E-Syt1 is located at MERCs and is involved in MERCs formation, 2) SYNJ2BP is located at MERCs and regulate the extent of MERCs in cells, 3) E-Syt1 and SYNJ2BP are located in MAM and in the same high molecular weight complex. Then, they show that both proteins impaired ER-mitochondria Ca++ exchange and that E-Syt1 influences mitochondrial lipid homeostasis, both phenotypes being rescued by artificial tether showing that only the tethering function of E-Syt1 is required. The proximity labelling experiments suggests SYNJ2BP as the mitochondrial partners of E-Syt1, however, from the data, it is not clear whether 1) the interaction between those proteins is direct,2) if SYNJ2BP is necessary and sufficient to localize E-Syt1 at MERC, and 3) if MERCs extension induced by SYNJ2BP is dependent on E-Syt1. Those points are important to investigate because SYNJ2BP has already been shown to induce MERCs by interacting with the ER protein RRBP1. In addition, some experiments need to be better quantified.

Major comments:

E-syt1/SYNJ2BP in MERCs formation: the authors provide several convincing lines of evidence that both proteins are in the same complex (proximity labelling, localization in the same complex in BN-PAGE, localization in MAM) but it is not clear in which extent the direct interaction between both proteins regulates ER-mitochondria tethering.

1. Pull down experiments or BiFC strategy could be performed to show the direct interaction between both proteins;
2. SYNJ2BP OE has already been demonstrated to increase MERCs and this being dependent on the ER binding partners RRBP1 (10.7554/eLife.24463). Therefore, it would be of interest to perform OE of SYNJ2BP in KO syt1 to address the question of whether Syt1 is also required to increase MERCs.
3. The authors show that Syt1 punctate size increases when SYNJ2BP is OE (Fig3C), but this can be indirectly linked to the increase of MERCs in the OE line. Thus, it could be interesting to test if the number/shape of E-syt1 punctate located close to mitochondria decreases in KO SYNJ2B. This could really show the dependence of SYNJ2BP for E-syt1 function at MERCs. Lipid analyses: the results of MS on isolated mitochondria clearly show that mitochondrial lipid homeostasis is affected on KO-Syt1 and rescued by expression of Syt1-Myc and artificial mitochondria-ER tether. However, p.15, the authors wrote "The loss of ESYT1 resulted in a decrease of the three main mitochondrial lipid categories CL, PE and PI, which was accompanied by an increase in PC ». As the results are expressed in mol%, this interpretation can be distort by the fact that mathematically, if the content of one lipid decreases, the content of others will increase. I would suggest to express the results in lipid quantity (nmol)/mg of mitochondria proteins instead of mol%. This will clarify the role of E-Syt1 on mitochondrial lipid homeostasis and which lipid increase and decrease. Also it could be of high interest to have the lipid composition of the whole cells to reinforce the direct involvement of E-Syt1 in mitochondrial lipid homeostasis and verify that the disruption of mitochondrial lipid homeostasis is not linked to a general perturbation of lipid metabolism as this protein acts at different MCSs.

Role of Syt1 in mitochondria: the authors show a perturbation of ER-mito Ca exchange and mitochondrial lipid homeostasis in KO-Syt1 as well as a growth defect of cells grown on galactose media. Modification of lipid mitochondrial lipid homeostasis often leads to defect in mitochondria morphology and mitochondria respiration, usually because of defects in supercomplexes assembly. To better understand the impact of Syt1 of mitochondria morphology, the author could analyze the mitochondria morphology (size, shape, cristae) on their EM images of crt, KO and OE lines. Indeed, on OE (Fig3A), the mitochondria look bigger and with a different shape compared to crt. Also, they performed a lot of BN-PAGE. Is it possible to check whether the mitochondrial respiratory chain super-complexes are affected on Syt1 KO line compared to crt? <br /> Quantifications: some western blots needs to be quantified (Fig 5K, 6J, S3E); Fig1A: Can the author provide a higher magnification of the triple labeling and perform quantification about the proportion of E-Syt1 punctate located close to mitochondria?

Minor comments:

• Fig1E + text: according to the legend, the BN-PAGE has been performed on Heavy membrane fraction. Why the authors speak about complexes at MAM in the text of the corresponding figure? Is-it the MAM or the heavy fraction (MAM + mito + ER...)? If BN have been performed from heavy membranes, it is not a real proof that E-syt1 is in MAMs.
• On fig3C (panel crt): it seems like SYNJ2BP dots are not co-localizaed with mito. Is this protein targeted to another organelle beside mitochondria?
• Fig3C: can the author show each channel alone and not only the merge to better appreciate mito and ER shape in control vs OE lines (as in fig S2)
• Fig4A: can the author provide a control of protein loading (membrane staining as example) to confirm the decrease of E-Syt1 in siSYNJ2BP?
• Fig5E/F: it is not clear to me why the expression of E-Syt1 in the KO is not able to complement the KO phenotype for cytosolic Ca++. Can the authors comment this.

Significance

Sevral mitochondrial-ER tethers as well as some proteins involved in Ca and/or lipid exchanges have been identified in mammals. E-Syt1 is well known to be located at ER-PM contact sites as well as ER-peroxisomes, and the presence of E-Syt1 at MERCs and its role in Ca++ and lipid exchange are new exciting results further showing the versatility of this protein. The results concerning E-Syt1 in Ca++ and lipid exchange are very convincing. In addition, the proximity labeling performed from four different SMP domain containing proteins is a highly valuable source of information for future work about interaction networks of those proteins. What is less in the study is the involvement of E-Syt1 interaction with SYNJ2BP for localization and function at MERCs and vice versa. Indeed, SYNJ2BP has already been shown to promote MERCs extension and to interact with the ER protein RRBP1. Thus, it will be of interest to further investigate E-Syt1/SYNJ2BP interaction at MERCs.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript reports the results of a study of the potential involvement of the SMP-domain-containing protein ESYT1 in ER-mitochondria tethering, and Ca+ and lipid exchange between the two organelles. SMP-domain proteins have been shown to localize to membrane contact site and have lipid transport activity. Esyt proteins have thus far been found at ER-plasma-membrane (PM) contacts. Here, starting from a BioID screen for partners of various SMP-domain proteins, the study focuses on a potential new interaction between ER-resident ESYT-1 and the mitochondrial outer-membrane protein SYNJ2BP. Then using a host of different approaches, the study concludes with a model in which ESYT-1-SYNJ2BP interaction tethers ER and mitochondria to regulate ion and lipid exchange between the two organelles.

This model would be very novel and interesting, as ESYT proteins have thus far only been detected at ER-PM contacts. However, the data supporting it are not unambiguous, are subject to alternative interpretation, and are sometimes contrary to the interpretation that the authors make of them. A lot of the reasoning behind the interpretation seems to be based on the fact that the authors have a hypothesis of what the effect of impacting ER-mitochondria should be, a priori, and when they observe such effects, they take it as evidence that they have indeed impacted tethering, disregarding alternative hypotheses and the possibility that the same effects can be wrought by entirely different mechanisms. Thus, the manuscript takes a few steps to involve ESYT1 in ER-mitochondria contacts but fails to make a decisive point.

Here are major points:

1. Localization of ESYT-1 and SYNJ2BP. The claim of a localization at ER-mitochondria contacts relies on two type of assays. Light microscopy and subcellular fractionation. Concerning microscopy, while the staining pattern is obviously colocalizing with the ER (a control of specificity of staining using KO cells would nevertheless be desirable), the idea that ESYT1 foci "partially colocalized with mitochondria" is either trivial or unfounded. Every cellular structure is "partially colocalized with mitochondria" simply by chance at the resolution of light microscopy. If the meaning of the experiment is to show that ESYT1 'specifically' colocalizes with mitochondria, then this isn't shown by the data. There is no quantification that the level of colocalization is more than expected by chance, nor that it is higher than that of any other ER protein. Moreover, the author's model implies that ESYT1 partial colocalization with mitochondria is, at least partially, due to its interaction with SYNJ2BP. This is not tested.

The subcellular fractionation assays are grounded on the idea that Mitochondria-Associated (ER) Membranes (MAM) can be purified, and are enriched for proteins that localize at ER-mitochondria contacts. This idea originated in the early 90's and since then, myriad of papers has been using MAM purification, and whole MAM proteomes have been determined. Yet the evidence that MAM-enriched proteins represent bona fide ER-mitochondria-contact-enriched proteins (as can nowadays be determined by microscopy techniques) remain scarce. Here, anyway, ESYT1 fractionation pattern is identical to that of PDI, a marker of general ER, with no indication of specific MAM accumulation. For SYNJ2BP, it is different as it is more enriched in the MAM than the general mitochondrial marker PRDX3. However, PRDX3 is a matrix protein, making it a poor comparison point, since SYNJ2BP is an OMM protein.

Again, the model implies that ESYT1 and SYNJ2BP accumulation in the MAM should be dependent on each other. This is not tested. 2. ESYT1-SYNJ2BP interaction. The starting point of the paper is a BioID signal for SYNJ2BP when BioID is fused to ESYT1. One confirmation of the interaction comes in figure 4, using blue native gel electrophoresis and assessing comigration. Because BioID is promiscuous and comigration can be spurious, better evidence is needed to make this claim. This is exemplified by the fact that, although SYNJ2BP is found in a complex comigrating with RRBP1, according to the BN gel, this slow migrating complex isn't disturbed by RRBP1 knockdown, but is somewhat disturbed by ESYT1 knockdown. More than a change in abundance, a change in migration velocity when either protein is absent would be evidence that these comigrating bands represent the same complex.

ESYT1-SYNJ2BP interaction needs to be tested by coimmunoprecipitation of endogenous proteins, yeast-2-hybrid, in vitro reconstitution or any other confirmatory methods. 3. Tethering by ESYT1- SYNJ2BP. This is assessed by light and electron microscopy. Absence of ESYT1 decreases several metrics for ER-mitochondria contacts (whether absence of SYNJ2BP has the same effect isn't tested). This interesting phenomenon could be due to many things, including but not limited to the possibility that "ESYT1 tethers ER to mitochondria".This statement and the respective subheading title are therefore clearly overreaching and should be either supported by evidence or removed. Indeed, absence of ESYT1 ER-PM tethering and lipid exchange could have knock-on effects on ER-mito contacts, therefore strong statements aren't supported. Moreover, the effect on ER-mitochondria contact metrics could be due to changes in ER-mitochondria contact indeed, but may also reflect changes in ER and/or mitochondria abundance and/or distribution, which favour or disfavour their encounter. Abundance and distribution of both organelles are not controlled for.

Finally, the authors repeat a finding that SYNJ2BP overexpression induces artificial ER-mitochondria tethering. Again, according to the model, this should be, at least in part, due to interaction with ESYT1. Whether ESYT1 is required for this tethering enhancement isn't tested. 4. Phenotypes of ESYT1/SYNJ2BP KD or KO. The study goes in details to show that downregulation of either protein yields physiological phenotypes consistent with decreased ER-mitochondria tethering. These phenotypes include calcium import into mitochondria and mitochondrial lipid composition.

Figure 5 shows that histamine-evoked ER-calcium release cause an increase in mitochondrial calcium, and this increase is reduced in absence of ESYT1, without detectable change in the abundance of the main known players of this calcium import. This is rescued by an artificial ER-mitochondria tether.

However, Figure 5D shows that the increase in calcium concentration in the cytosol upon histamine-evoked ER calcium release is equally impaired by ESYT1 deletion, contrary to expectation. Indeed, if the impairment of mitochondrial calcium import was due to improper ER-mitochondria tethering in ESYT1 mutant cells, one would expect more calcium to leak into the cytosol, not less. The remaining explanation is that ESYT1 knockout desensitizes the cells to histamine, by affecting GPCR signalling at the PM, something unexplored here. In any case, a decreased calcium discharge by the ER upon histamine treatment, explains the decreased uptake by mitochondria. The authors argue that ER calcium release is unaffected by ESYT1 KO, but crucially use thapsigargin instead of histamine to show it. Thus, the most likely interpretation of the data is that ESYT1 KO affects histamine signalling and histamine-evoked calcium release upstream of ER-mitochondria contacts.

The data with SYNJ2BP deletion are more compatible with decreased ER-mito contacts, as no decreased in cytosolic calcium is observed. This is compatible with the previously proposed role of SYNJ2BP in ER-mitochondria tethering, but the difference with ESYT1 rather argue that both proteins affect calcium signalling by different means, meaning they act in different pathways.

Finally, the study delves into mitochondrial lipids to "investigated the role of the SMP-domain containing protein ESYT1 in lipid transfer from ER to mitochondria". In reality, it is not ER-mitochondria lipid transport that is under scrutiny, but general lipid homeostasis, and changes in ER-PM lipids could have knock-on effects on mitochondrial lipids without the need to invoke disruptions in ER-mitochondria transfer activity. The changes observed are interesting but could be due to anything. Surprisingly, PCA analysis shows that the rescue of the knockout by the ESYT1 gene clusters with the rescue by the artificial tether, and not with the wildtype. This indicates that overexpressing either ESYT1 or a tether cause similar lipidomic changes. These could be due, for instance, to ER stress caused by protein overexpression, and not to a rescue.

In any case the data here do not support the strong statement "Together these results demonstrate that ESYT1 is required for lipid transfer from ER to mitochondria [...]".

Significance

This model would be very novel and interesting, as ESYT proteins have thus far only been detected at ER-PM contacts. However, the data supporting it are not unambiguous, are subject to alternative interpretation, and are sometimes contrary to the interpretation that the authors make of them. A lot of the reasoning behind the interpretation seems to be based on the fact that the authors have a hypothesis of what the effect of impacting ER-mitochondria should be, a priori, and when they observe such effects, they take it as evidence that they have indeed impacted tethering, disregarding alternative hypotheses and the possibility that the same effects can be wrought by entirely different mechanisms. Thus, the manuscript takes a few steps to involve ESYT1 in ER-mitochondria contacts but fails to make a decisive point.

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

Learn more at Review Commons

---

Reply to the reviewers

We are grateful to both reviewers for reviewing our manuscript, and for providing very helpful feedback as to how we can improve this work. We have now implemented nearly all of the changes as recommended, and provide responses to these points below.

In terms of novelty, while recent pre-prints and publications have suggested that the application of multi-omics analysis improves GRN inference, there has yet to be a systematic comparison of linear and non-linear machine learning methods for GRN prediction from single cell multi-omic data. here are many computational and statistical challenges to such a study, and we therefore believe that others in the field will be especially interested in our systematic comparison of network inference methods, especially given the increased interest and utility of multi-omic data.

In addition, we report the first comprehensive inference of GRNs in early human embryo development. This is a particularly challenging to study developmental context given genetic variation, limitations of sample size due to the precious nature of the material and regulatory constraints. We anticipate that the methodology we developed and datasets we generated will be informative for computational, developmental and stem cell biologists.

We have uploaded all the network predictions on FigShare and these can be accessed using the following link: https://doi.org/10.6084/m9.figshare.21968813. In addition, we anticipate that the computational and statistical codes and pipelines we developed (available on https://github.com/galanisl/early_hs_embryo_GRNs) will be applied to other cellular and developmental contexts, especially in challenging contexts such as human development, non-typical model organisms and in clinically relevant samples.

Reviewer 1

Major comments

- The proposed strategy (i.e. combining gene expression-based regulatory inference with cis-*regulatory evidence) have been well developed (and implemented) by multiple published works like SCENIC and CellOracle, which is also properly acknowledged by the authors in the discussion section too. This leads to a serious concern on the major methodological contribution of this work. *

We would like to note that our study is the first to comprehensively evaluate machine learning linear or non-linear gene regulatory network prediction strategies from single-cell transcriptional datasets combined with available multi-omic data. We also apply these methods to a challenging to study context of human early embryogenesis. There are specific methodological challenges arising in this context that other published work has not yet addressed. In particular, the precious nature of the source material means that sample sizes are limited, unlike the contexts where SCENIC and CellOracle were applied. Notably, the numbers of cells available for downstream analysis is typically several orders of magnitude fewer than when scRNA-seq data are collected from adult human tissue or from cell culture. This restriction on sample sizes places corresponding restrictions on statistical power, and is therefore likely to mean that different statistical network inference methodologies are optimal in specific contexts. Furthermore, the inclusion of multi-omic data from complementary platforms (such as ATAC-seq data) becomes even more important in this context to mitigate the effect of reduced sample sizes. These issues are very important for choice of gene regulatory network inference methodology in relation to studies of human embryo development, and ours is the first study to address these issues directly in any context. We have further clarified the novelty of our work in the manuscript in the abstract, introduction and discussion sections.

- Most of the compared network reconstruction methods involve hyper-parameters setup (e.g., *sparsity regularization weights of the regression methods). The authors did not discuss how these hyper-parameters were chosen. *

For sparse regression, the hyperparameter controlling sparsity was set by cross-validation (CV), using the internal CV function of the R package. All default settings for GENIE3 were used. This information has now been added to the manuscript (in the Methods section), along with a description of the implementation of the mutual information method we use.

- For the real-world blastocyst data, the network prediction methods were compared in terms of their reproducibility across validation folds (Fig. 3, Fig. S4-6). However, reproducibility does not necessarily imply accuracy. In fact, statistical learning methods are generally subject to the bias-variance tradeoff, where lower variance (i.e., higher reproducibility) could imply higher bias in model prediction. While there is a lack of gold-standard ground truth to evaluate network accuracy in real biological systems, silver-standards like the ranking of known regulatory interactions in the predictions could be employed as an indirect estimate.

We thank the reviewer for the opportunity to clarify this point. We would like to avoid any misunderstanding of the reproducibility statistic R, as follows. A higher value of R indicates that the fitted model would generalise well to new data; i.e., R=1 indicates that the model is robust (stable) to perturbations of the data-set. We note that this is not the same as analysing the residual variance of the data after model fitting and related over-fitting (i.e., bias-variance trade-off). The variance that is referred to when discussing bias-variance trade-off is the mean-squared error (of data compared to model), which is not the same as what is assessed by reproducibility statistic R . Specifically, R is a Bayesian estimate of the posterior probability of observing a gene regulation given the data. R is calculated by taking a random sample of the data, doing the network inference again, checking if each gene regulation still appears in the GRN, and then recording (as the R statistic) the average fraction of inclusions over many repetitions. So when we have R close to 1, this indicates that our model predictions generalise well to new data, which is the opposite of what is suggested in this comment. In summary, the accuracy quantified by the reproducibility statistic R relates to the stability of the model predictions to perturbation of the data. We thank the reviewer for the helpful comment to draw our attention to this point, and have now clarified this point in the manuscript on page 6 line 252.

- The gene set enrichment results were reported only on EPI and TE cell types (Fig. 4C and Fig. *S12), due to the reason that CA data is only available for TE and ICM. However, many of the other results presented in Fig. 3-6 did include the PE cell type albeit using the same CA data. It is not particularly convincing why the cell type inclusion standard for gene set enrichment is different from the other results. *

We thank the reviewer for noting this and would like to clarify that we restricted the analysis to the EPI and TE, because similar lists of gene-sets were not available for primitive endoderm, where it is currently unclear which pathways are most relevant to this cell type. This has now been clarified in the manuscript on page 8, line 337.

- The authors cited TF binding in cis-regulatory regions as supporting evidence of several MICA-inferred regulatory interactions (e.g., NANOG -> ZNF343). However, the same cis-regulatory *evidence has already been used in the CA filtering step. All interactions passing CA filtering should in principle have TF-binding support. It would be more convincing if the authors provided other types of evidence as independent support, such as genetic associations like eQTL, experimental perturbations like gene knockdown/knockout, etc. *

We appreciate the reviewer’s point. We address this by describing published ChIP-seq validation in human pluripotent stem cells which is widely used as a proxy for the study of the epiblast. We feel that the ChIP-seq validation in this context is an appropriate independent validation to support the MICA-inferred cis regulatory interactions predicted from the human embryo datasets we analysed. Our inferences from ATAC-seq data cannot identify TF-DNA binding directly. ChIP-seq data is a widely accepted independent methods to support the inferred interactions from ATAC-seq data.

We agree that knockdown/knockout would provide further evidence suggesting gene regulation, and indeed these are experiments we would like to conduct systematically in the future, but such perturbations are difficult to achieve at genome-wide scale, especially with very restricted quantities of human embryo material. Notably, these studies would not be evidence of direct regulation and the gold-standard in our opinion is to perturb the cis regulatory region to demonstrate its functional importance in gene regulation. These are important experiments to conduct systematically in the future. We also note that assessing quantitative trait loci in the context of human pre-implantation embryos is extremely challenging due to the restricted sample sizes and genetic variance in the samples collected.

*- Many of the MICA-inferred regulatory interactions do not exhibit Spearman correlation (Fig. 5, Fig. S17), which could probably be explained by the ability of mutual information to capture complex non-monotonic dependencies. It would be interesting to provide further investigation on these "uncorrelated" edges, which may help demonstrate the superiority of mutual information over Spearman correlation. *

This has been added as a new Fig.S18.

- The authors conducted immunostaining experiments to validate the MICA-inferred regulatory *interaction between TFAP2C and JUND. While the identified protein co-localization is a step further than RNA co-expression, it is still correlation rather than causality. Additional evidence like the effect of knockout/knockdown perturbations would be more convincing. *

We agree with Reviewer 1 that experimental perturbations of TFAP2C and JUND to determine what consequence this has for interactions between these proteins would be informative. However due to the complexity of such an investigation in human embryos, we feel that this is beyond the scope of the current study. One option is to conduct the perturbations in human pluripotent stem cells, however it is unclear if the GRN in this context reflects the same interactions as human embryos and is a distinct question to address in the future. Moreover, while knockdown/knockout studies would be suggestive of up-stream regulation, it will not address the question of whether this is a direct or indirect effect without systematic further analysis including transcription factor-DNA binding (such as CUT&RUN, CUT&Tag or ChIP-seq) analysis as well as perturbations of the putative cis regulatory regions. These are all exciting future experiments and our study provides us and others with hypotheses to functionally test in the future. These are future directions and we have clarified this in the discussion section on page 16, line 576.

__Minor comments __

• *The γ symbols in AP-2γ are not correctly rendered. *

We note that this applies only to the way AP-2γ appears on the Review Commons website, and we are trying to fix this issue. We hope this transformation after the manuscript upload will not apply to a subsequent transfer to a journal.

• The UMAP figures (Fig. 4A, Fig. S7) are of low resolution compared to other figures.

We thank the reviewer for noting this. These figures have now been added as vector graphics files to overcome this issue.

• As the authors are focused on studying the blastocyst regulatory network, the inferred regulatory interactions should be provided as supplementary data.

We have included all of the inferred gene regulatory interactions as a supplementary folder for the MICA predictions using FigShare: doi.org/10.6084/m9.figshare.21968813. We have included code to reproduce the inferred gene regulatory interactions for the other methods which we compared to MICA. Because this includes 100,000 regulatory interactions per method, we feel that it would be impractical to include the alternative inferred interaction as supplementary data.

Reviewer 2

Minor comments

*- In the abstract, it would be adequate to already mention which normalisation method works the best. *

This has now been added to the abstract and we appreciate this suggestion.

*- In Fig. 1: *

* Describe what are squares and circles

This information has been included in the figure 1 legend.

** In the GRNs refined by keeping CA-predicted regulations only, mention that this are Cis interactions *

We have modified the figure 1 legend and the text on page 5, line 224 to clarify that these are putative cis-regulatory interactions.

* The ATAC seq shows KRT8, GATA3, RELB motifs, while the rest of the figure is very general. Maybe make the ATAC-seq peaks panel also as a sketch and relate it to the square/circles graphs on the right hand side to showcase how the filtering of the network is performed.

We appreciate this suggestion and modified figure 1 accordingly.

** The caption says Five GRN inference approaches, while abstract and text say 4. If is clear after reading that the 5th is a random approach. However, it was a surprise at first. *

We have modified the figure 1 legend to clarify that we also compared random prediction in addition to the 4 GRN inference approaches.

*- How the Simulation study was performed is not understandable for non experts as it is described in the Methods section. This is an important approach in general, and I think the audience would benefit if the authors add a full section about it in their supplementary data. *

Further details have now been added to the subsection ‘simulation study’ in the Methods section.

*- Fig. 2: *

** As it is, it is hard to tell the difference between GRN inference methods for a given sample size and number of regulators. Could the authors add a comparative panel for this (maybe some scatter plots would be enough)? MI by itself looks worse here? *

We thank the reviewer for this helpful suggestion. This comparative plot has now been included in figure 2 and indicates that MI is on par with the other GRN inference methods using simulation RNA-seq data.

*- When mentioning "samples" (e.g. last paragraph of section 1 in results), do the authors refer to "cells"? *

We appreciate the reviewer pointing this out and have amended the text throughout to state that these are cells.

*- What about normalisation effects in the simulated data? *

With regards to the simulated data, normalisation effects are not relevant as we are generating data that are idealised and therefore not subject to unwanted sources of variation such as read depth. However, in future work, this could be investigated with an expanded simulation study and we appreciate the reviewer’s suggestion.

*- Figure S7 should be cited in the first paragraph of section 2 in results. *

This has now been cited.

*Could the authors add a panel to indicate whether the data is SMART-seq2 or 10X. *

We thank the reviewer for the suggestion to clarify this, which we think is an important point. We have included a statement that all data used was generated using the SMART-seq2 sequencing technique in the figure legend. The choice of sequencing method/depth of sequencing will likely impact on the choice of GRN inference method and we have also clarified this in the discussion section on page 13, line 516.

*- In the association of inferred GRNs to human blastocyst cell lineages, the authors find the GRN edges predicted that overlap between the 4 inference methods in each cell type. Do they, therefore, recommend to always use more than one GRN inference method? *

Identifying overlapping inferences by comparing more than one GRN inference method may be a strategy to identify network edges with more confidence due to the agreement between several inference methodologies. However, this strategy may also miss some edges which can only be detected by one method and not another. We have included a statement in the discussion section to clarify this point on page 15, line 571.

- If the CA data used was only generated for the TE and ICM only, how do the authors use it to perform MICA on PE?

We appreciate that this is confusing and have since revised the manuscript on page 5, line 223 to state that the inner cell mass (ICM), comprises EPI (epiblast) and PE (primitive endoderm) cells. It may be that we miss putative cis-regulatory interactions if the ICM CA data does not reflect developmentally progressed PE and EPI cells and we have noted this caveat in the discussion section on page 15, line 561.

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 work, Alanis-Lobato et al apply different GRN inference methods on scRNA-seq data from human blastocysts. By integrating the data with ATACseq, they manage to address the small sample size challenge and predict novel TF-gene interactions that they later validate with immunofluorescence. Main take-home-messages from this work are that proper GRN inference methods work better upon integration of different omic technologies (here RNA and ATAC seq) and proper data normalisation strategies (logTPM or logFPKM).

Hereby I present some minor concerns and questions that I have after reading the manuscript, that I hope the authors can address.

• In the abstract, it would be adequate to already mention which normalisation method works the best.
• In Fig. 1:
  • Describe what are squares and circles
  • In the GRNs refined by keeping CA-predicted regulations only, mention that this are Cis interactions
  • The ATAC seq shows KRT8, GATA3, RELB motifs, while the rest of the figure is very general. Maybe make the ATAC-seq peaks panel also as a sketch and relate it to the square/circles graphs on the right hand side to showcase how the filtering of the network is performed.
  • The caption says Five GRN inference approaches, while abstract and text say 4. If is clear after reading that the 5th is a random approach. However, it was a surprise at first.
• How the Simulation study was performed is not understandable for non experts as it is described in the Methods section. This is an important approach in general, and I think the audience would benefit if the authors add a full section about it in their supplementary data.
• Fig. 2:
  • As it is, it is hard to tell the difference between GRN inference methods for a given sample size and number of regulators. Could the authors add a comparative panel for this (maybe some scatter plots would be enough)? MI by itself looks worse here?
• When mentioning "samples" (e.g. last paragraph of section 1 in results), do the authors refer to "cells"?
• What about normalisation effects in the simulated data?
• Figure S7 should be cited in the first paragraph of section 2 in results. Could the authors add a panel to indicate whether the data is SMART-seq2 or 10X.
• In the association of inferred GRNs to human blastocyst cell lineages, the authors find the GRN edges predicted that overlap between the 4 inference methods in each cell type. Do they, therefore, recommend to always use more than one GRN inference method?
• If the CA data used was only generated for the TE and ICM only, how do the authors use it to perform MICA on PE?

Significance

In this paper, one main message is that to infer GRN one should combine different omic datasets. This does not come as a surprise and has been published before. What it is very well addressed in this study is the problem of the sample size: the authors decide to test GRN inference methods in the human blastocyst, for which currently we do not have a lot of sequencing data available. Interestingly, they find that 1k cells should be enough to infer relevant GRN. Maybe the manuscript would benefit if the authors emphasize this more in their text.

Interestingly, and despite the fact that the sample size here is below 1k, the authors identify novel regulatory relationships between TFs for different cell types, that they also validate.

This paper will be relevant to a wide audience of scientists interested in human developmental biology, or in the development of computational approaches to analyse single cell sequencing data.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary

The authors proposed MICA strategy as an attempt to infer gene regulatory network at the blastocyst stage of early embryo development which features limited sample size. While the motivation seems reasonable to me and the results showed several interesting insights, the methodological novelty and significance of this study need further elaboration, and the evaluation/benchmark part is largely insufficient.

Major comments

• The proposed strategy (i.e. combining gene expression-based regulatory inference with cis-regulatory evidence) have been well developed (and implemented) by multiple published works like SCENIC and CellOracle, which is also properly acknowledged by the authors in the discussion section too. This leads to a serious concern on the major methodological contribution of this work.
• Most of the compared network reconstruction methods involve hyper-parameters setup (e.g., sparsity regularization weights of the regression methods). The authors did not discuss how these hyper-parameters were chosen.
• For the real-world blastocyst data, the network prediction methods were compared in terms of their reproducibility across validation folds (Fig. 3, Fig. S4-6). However, reproducibility does not necessarily imply accuracy. In fact, statistical learning methods are generally subject to the bias-variance tradeoff, where lower variance (i.e., higher reproducibility) could imply higher bias in model prediction. While there is a lack of gold-standard ground truth to evaluate network accuracy in real biological systems, silver-standards like the ranking of known regulatory interactions in the predictions could be employed as an indirect estimate.
• The gene set enrichment results were reported only on EPI and TE cell types (Fig. 4C and Fig. S12), due to the reason that CA data is only available for TE and ICM. However, many of the other results presented in Fig. 3-6 did include the PE cell type albeit using the same CA data. It is not particularly convincing why the cell type inclusion standard for gene set enrichment is different from the other results.
• The authors cited TF binding in cis-regulatory regions as supporting evidence of several MICA-inferred regulatory interactions (e.g., NANOG -> ZNF343). However, the same cis-regulatory evidence has already been used in the CA filtering step. All interactions passing CA filtering should in principle have TF-binding support. It would be more convincing if the authors provided other types of evidence as independent support, such as genetic associations like eQTL, experimental perturbations like gene knockdown/knockout, etc.
• Many of the MICA-inferred regulatory interactions do not exhibit Spearman correlation (Fig. 5, Fig. S17), which could probably be explained by the ability of mutual information to capture complex non-monotonic dependencies. It would be interesting to provide further investigation on these "uncorrelated" edges, which may help demonstrate the superiority of mutual information over Spearman correlation.
• The authors conducted immunostaining experiments to validate the MICA-inferred regulatory interaction between TFAP2C and JUND. While the identified protein co-localization is a step further than RNA co-expression, it is still correlation rather than causality. Additional evidence like the effect of knockout/knockdown perturbations would be more convincing.

Minor comments

• The γ symbols in AP-2γ are not correctly rendered.
• The UMAP figures (Fig. 4A, Fig. S7) are of low resolution compared to other figures.
• As the authors are focused on studying the blastocyst regulatory network, the inferred regulatory interactions should be provided as supplementary data.

Significance

Given the concerns listed above, I still hold doubts on the significance of the manuscript in its current form. In particular, the major contribution of this work, in methodological senses, seems to be the specific choice of mutual information for regulatory inference in the low-data regime, which may have a limited audience and impact.

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. Point-by-point description of the revisions

Reviewer #1

Evidence, reproducibility and clarity (Required):

In this paper by Wideman et al, the authors seek to determine the role of cellular iron homeostasis in the pathogenesis of murine malaria.

The authors to attempt to disentangle the effects of anemia from that of cellular iron deficiency. The authors elegantly make use of a murine model of a rare human mutation in the transferrin receptor. This mutation leads to decreased receptor internalization and decreased cellular iron, but otherwise healthy mice. Using this model, the authors use a P. chabaudi infection model and show an increase in pathogen burden and a decrease in pathology. They show in some detail that the immune response to P. chabaudi infection is blunted, both T and B-cell responses are attenuated in the TfRY20H/Y20H model, and the block in proliferation can be rescued by exogenous iron supplementation. They also show that decreased cellular iron attenuates liver pathology through potentially multiple mechanisms.

Minor comments:

• The peak of parasitemia is relatively low (approx..3%) compared to other published studies (e.g. PMID: 22100995, 16714546, 31110285) where the peak in C57BL/6 mice reached 25 - 40%. Can the authors account for this low parasitemia?

Response: We thank the reviewer for their constructive comments and appreciate that they are highlighting this important point. It has previously been shown (PMID: 23217144, 23719378) that mosquito-transmission of P. chabaudi leads to significantly lower parasitaemia (“Recently mosquito-transmitted parasites were used to mimic a natural infection more closely, as vector transmission is known to regulate Plasmodium virulence and alter the host’s immune response (48-50). Consequently, parasitaemia is expected to be significantly lower upon infection with recently mosquito-transmitted parasites, compared to infection with serially blood-passaged parasites that are more virulent (48,49).”

• Figure 1K - At homeostasis, serum iron is low in TfR mice however increases to significantly higher than the WT mice at 8 days post infection. Do the authors have an explanation on why these dramatic changes in serum iron are seen?

Response: During malaria infection, RBC lysis releases haem and iron into circulation, which leads to an increase in serum iron levels. This effect is observed in both wild-type and TfrcY20H/Y20H mice infected with P. chabaudi (Supplementary Figure 1F & Figure 1K). However, the significantly higher serum iron levels observed in infected TfrcY20H/Y20H mice can likely be explained by their decreased capacity for transferrin receptor-1 mediated iron uptake, leading to relatively slower uptake and storage of circulating transferrin-bound iron into tissues. This has been clarified in the manuscript (line 142-143):

“The elevated serum iron observed in infected TfrcY20H/Y20H mice was consistent with their restricted capacity to take up transferrin-bound circulating iron into tissues.”

• Figure S3 - Is it surprising that no effects on splenic neutrophils are seen? Were neutrophils quantified at any other point? These would also be expected to have a role in both the control of malaria infection and on any pathology.

Response: We thank the reviewer for raising this interesting question. It is known that neutrophils can be sensitive to cellular iron deficiency (PMID: 36197985) and that neutrophils can play an important part in malaria infection (PMID: 31628160). However, the magnitude and significance of the neutrophil response to recently mosquito-transmitted P. chabaudi parasites has not been thoroughly investigated. A recent study demonstrated that monocytes and macrophages may be more important than granulocytes in the early response to recently mosquito-transmitted P. chabaudi infection (PMID: 34532703).

Moreover, we performed neutrophil quantifications in our initial experiments and found that the splenic neutrophil response was not altered in TfrcY20H/Y20H mice eight days after infection. Additionally, no neutrophil infiltration was observed in the liver of either genotype upon P. chabaudi infection. In light of these findings, we did not characterise the neutrophil response further, as it appeared unlikely that neutrophils were the principal causal agent of either the altered immunity or pathology, in this context. However, we agree with the reviewer that larger question of whether neutrophil iron plays a role in the pathology of malaria is an interesting open question which we hope future studies can elucidate.

A section was added to the discussion to address the role of innate immune cells in our model (line 354-363):

“The inhibited innate immune response to P. chabaudi in TfrcY20H/Y20H mice likely contributed to both the increased pathogen burden and the decreased liver pathology. Splenic MNPs are important for controlling parasitaemia (34,35,72), but MNPs are also vital for maintaining tissue homeostasis and preventing tissue damage in malaria (43,73). Although other innate cells, such as neutrophils, NK cells and γδT cells are an important part of the immune response to malaria, only the MNP response was distinctly impaired in TfrcY20H/Y20H mice. Notably, neutrophils are known to be sensitive to iron deficiency (16,74) and to affect both immunity and pathology in malaria (75,76). However, in the context of recently mosquito-transmitted P. chabaudi it appears that monocytes and macrophages, rather than granulocytes, may be particularly important for parasite control and tissue homeostasis (43,72).”

Changes to the text:

• Fig S1EandF - Please add to the figure legend that these were measured at homeostasis.

Response: This clarification has been added to the legend of Supplementary Figure 1 (line 954-957).

• Figure 3 - In the legend, H and I are the wrong way around.

Response: The legend of Figure 3 has been corrected accordingly (line 888-890).

• Figure 4 - please add the units of concentration of FeSO4 to all panels

Response: The units of concentration for FeSO4 and AFeC have been added to all panels of Figure 4 and 6, respectively.

• Line 246 - The authors state: "there was some evidence of decreased malaria-induced hepatomegaly" however there is no significant difference between WT and TfR mice and both show significant hepatomegaly. I feel that this line should be reworded.

Response: The sentence (line 252-254) has been reworded as follows:

“Furthermore, while both genotypes developed malaria-induced hepatomegaly, there was a trend toward less severe hepatomegaly in TfrcY20H/Y20H mice (Figure S5C).”

Significance (Required):

This work is one of the first to attempt to define the requirements for cellular iron in malaria infection. This is a difficult topic, as infection and associated inflammation and the red blood cell destruction caused by malaria all have complex effects on iron within the body. This study fits well with previous observations showing that anemia can be protective as it both prevents parasite growth and limit immunopathology. This work advances the field by demonstrating a cell intrinsic role for iron in malaria infection. There is a broad possible audience for this work, including malaria researchers, immunologists and people interested in the role or iron, both at a cellular level and systemically.

Reviewer #2

Evidence, reproducibility and clarity (Required):

In this manuscript, the authors have studied the role of iron deficiency in the host response to Plasmodium infection using a transgenic mouse model that carries a mutation in the transferrin receptor. They show that restricted cellular iron acquisition attenuated P. chabaudi infection- induced splenic and hepatic immune responses which in turn mitigated the immunopathology, even though the peak parasitemia was significantly high in the mutant mice. Interestingly, the course of parasite infection doesn't seem to be affected in the mutant mice compared to the wildtype mice despite the induction of poor immune responses. The authors show that the decreased cellular iron uptake broadly impact both innate and adaptive components of the immune system. Conversely, free iron supplementation restored the immune cell functions.

• The study is well performed, and the manuscript is well written. However, the authors should show how conserved the role of cellular iron is across other rodent malaria parasite species at least with * yoelii or P. berghei* blood stage infection models. This question becomes critical to address in order to understand broad relevance to human malaria infections where both the host and parasites are genetically diverse.

Response: We thank the reviewer for appreciating our study and for the thoughtful comments. We agree with the reviewer that the diverse genetic background of both parasites and hosts makes it difficult to draw broad conclusions about human malaria infection from animal studies performed in a laboratory setting. The recently mosquito-transmitted P. chabaudi chabaudi AS blood-stage infection model replicates many key features of mild to moderate malaria infection in humans, such as low parasitaemia, anaemia, cyto-adhesive sequestration in microvasculature, and self-resolving immunopathology. Importantly, the immune response elicited by recently mosquito-transmitted parasites also more closely mimics the immune response to a natural infection (PMID: 23719378). Therefore, we consider the recently mosquito-transmitted P. chabaudi chabaudi AS model as the most relevant to answer our particular research questions.

Furthermore, specific pathogen-free parasitised erythrocyte stabilates made from recently mosquito-transmitted P. berghei or P. yoelii parasites are unfortunately not readily accessible (e.g. through the European Malaria Reagent Repository), in contrast to P. chabaudi. Consequently, preparing and characterising recently mosquito-transmitted strains to perform the experiments suggested by the reviewer would require a substantial amount of additional time and labour, which we deem out of scope for this study.

In the design of our model we have also taken care to minimise the effects of anaemia, something which would be difficult or impossible to achieve using serially blood passaged P. yollii or P. berghei parasites. Both P. yoelii and P. berghei merozoites preferentially invade immature RBCs (PMID: 34322397) making readouts such as parasitaemia far more sensitive to small variations in erythropoietic output. In addition, the extensive RBC destruction caused by most serially blood-passaged murine Plasmodium strains would likely exaggerate any erythropoietic impairment caused by the TfrcY20H/Y20H mutation.

Although we strongly believe that the chosen mouse model of malaria is the most appropriate for our study, ultimately, no mouse model can replicate all features of human malaria infection. Inevitably, the direct relevance of animal studies for human infection will always be somewhat opaque. Hence, we respectfully disagree with the reviewer that repeating the experiments with additional murine malaria parasite species would allow us to extrapolate conclusions about human malaria infection. Such experiments would also conflict with the 3Rs principles that govern work with animals in the UK (https://nc3rs.org.uk/). Especially, because most strains of P. yoelii and P. berghei cause severe or non-resolving infections and have a significant negative impact on animal welfare.

In our opinion, the logical continuation of this study must be to utilise the insights from our research to inform future human studies on the relationships between iron deficiency and malaria-related immunopathology. However, we agree that this is an important topic and have added a section addressing the broad relevance of our findings to the discussion (line 393-396):

“It remains to be seen what the broader importance of cellular iron is in human malaria infection, in particular within the diverse genetic context of both humans and parasites found in malaria endemic regions. Murine models of malaria are useful in providing hypothesis-generating results, but such findings ultimately ought to be confirmed and developed further through studies in human populations.”

• Since, restricted cellular iron uptake mitigates the immunopathology, the authors should explore whether this could also relieve the cerebral malaria condition that is caused by the hyper inflammation in the brain. They should use the * berghei* ANKA parasite strain which causes t cerebral malaria in mice. I think would increase impact of the paper.

Response: Although we agree that this would be an interesting line of inquiry, we think that it is outside of the scope of this study, which predominantly aims to characterise and study the effects of cellular iron deficiency in host cells, particularly immune cells, during mild to moderate malaria infection. The severe pathology underlying cerebral malaria differs greatly from that of a self-resolving blood-stage infection. Furthermore, the relevance to human cerebral malaria of the P. berghei ANKA model is controversial within the field (PMID: 21288352) and as a severe infection its use would again conflict with the 3Rs principles.

Minor comments:

• Line 222: repeating word, "iron iron-supplemented...."

Response: The sentence has been corrected (line 228).

• Figure 3C, S4C & S5F: Why Mann-Whitney test is performed in these particular graphs, whereas rest of the two groups comparison were done using Welch's test? The authors should clearly mention this in the methods section.

Response: We apologise if this was unclear in the manuscript. We routinely tested all our datasets for normality to identify the appropriate tests for each dataset. In case of the graphs shown in figure 3C, S4C and S5F, the dataset did not pass the D’Agostino-Pearson normality test and we therefore applied a non-parametric test (i.e. Mann-Whitney), in contrast to the other datasets that passed the test for normal or lognormal distribution. This has been further clarified in the method section (line 581-586):

“The D’Agostino-Pearson omnibus normality test was used to determine normality/lognormality. Parametric statistical tests (e.g. Welch’s t-test) were used for normally distributed data. For lognormal distributions, the data was log-transformed prior to statistical analysis. Where data did not have a normal or lognormal distribution, or too few data points were available for normality testing, a nonparametric test (e.g. Mann-Whitney test) was applied.“

• Have authors explored whether gamma-delta T cell responses are affected in the mutant mouse strain compared to wildtype mice as they are one of the early responders and the key cytokine producing cells against the Plasmodium blood stage infection.

Response: __We thank the reviewer for this valuable comment. We briefly explored the role of γδT cells, but did not observe a significant difference in splenic γδT cell numbers between wild-type and TfrcY20H/Y20H mice, eight days post-infection (__Reviewer Figure 1). It is of course possible that γδT cell numbers were affected at an earlier stage, or that γδT cell function (e.g. cytokine production) was affected by cellular iron deficiency during P. chabaudi infection. However, γδT cells may also be less sensitive to cellular iron deficiency than conventional T cells, as has been previously demonstrated for developing T cells (PMID: 7957580).

A section was added to the discussion to address the role of innate immune cells in our model (line 354-363):

“The inhibited innate immune response to P. chabaudi in TfrcY20H/Y20H mice likely contributed to both the increased pathogen burden and the decreased liver pathology. Splenic MNPs are important for controlling parasitaemia (34,35,72), but MNPs are also vital for maintaining tissue homeostasis and preventing tissue damage in malaria (43,73). Although other innate cells, such as neutrophils, NK cells and γδT cells are an important part of the immune response to malaria, only the MNP response was distinctly impaired in TfrcY20H/Y20H mice. Notably, neutrophils are known to be sensitive to iron deficiency (16,74) and to affect both immunity and pathology in malaria (75,76). However, in the context of recently mosquito-transmitted P. chabaudi it appears that monocytes and macrophages, rather than granulocytes, may be particularly important for parasite control and tissue homeostasis (43,72).”

Significance (Required):

Overall, the study provides novel insights into the role of iron in the immune response to Plasmodium blood stage infection using a rodent malaria model and the interplay of infection, immunity and the development of pathology. As such it is an important study.

Reviewer #3

Evidence, reproducibility and clarity (Required):

Herein Wideman provide novel and important evidence on the role of iron availability for mounting an efficient immune response in a malaria infection model. They employed TfRC Y201H/Y201H mice which develop iron deficiency due to impaired cellular ingestion of transferrin bound iron. They found that those mice develop higher peak parasitemia after vector borne exposure to Pl. chabaudi chabaudi which was paralleled by an impaired immune response as reflected by altered CD4 cell activation, reduced IFN-g formation or reduced B-cell responsiveness. Those deficiencies could be re-covered upon ex vivo iron supplementation pointing to the importance of iron availability for mounting-CD4+ and B-cell specific anti-plasmodial immune responses at the initial phase of infection. However, TFRC mutated mice were able to clear infection over time in a comparable fashion to wt mice.

This excellent study is important in convincingly showing (by employing high quality immunological analyses) the importance of cellular iron deficiency on immune responses in an infection model of general interest. It also indicates that overwhelming immune response as seen in wt mice is associated with organ damage over time.

Minor comments:

• The authors should discuss why and how TFRC mutated mice were able to control infection over time in a comparable fashion as wt mice although peak parasitemia was significantly higher?

__Response: __We thank the reviewer for the helpful feedback on our study and for posing this interesting question. It does indeed appear as if the immune response, while significantly inhibited in the TfrcY20H/Y20H mice, is still sufficient to clear the infection. It is plausible that the early cell-mediated immune response is inhibited to the degree that parasite control is impaired, resulting in higher peak parasitaemia in TfrcY20H/Y20H mice. In contrast, parasite clearance is comparable and contemporary in both genotypes. Based on the fact that parasite clearance occurs at a time when a substantial adaptive immune response is expected to emerge, we hypothesize that this significantly contributes to pathogen clearance. Thus, it seems likely that the humoral response in TfrcY20H/Y20H mice, even if inhibited, may still be effective enough to clear the parasites and prevent recrudescence.

As malaria infection progresses, RBC loss and increasing anaemia also contributes to limiting exponential parasite growth. This occurs more or less equally in both genotypes, but it could be particularly important for parasite control in the TfrcY20H/Y20H mice that have an inhibited immune response.

We have added a section to the discussion to address this (line 380-386):

“Despite the higher peak parasitaemia in TfrcY20H/Y20H mice, both genotypes were able to clear P. chabaudi parasites at a comparable rate and prevent recrudescence. It follows that even a weakened humoral immune response appears to be sufficient to control P. chabaudi infection. However, our study did not investigate the effects of immune cell iron deficiency on the formation of long-term immunity, which may have been more severely affected. The impaired GC response, in particular, suggests that iron deficiency could counteract the formation of efficient immune memory to subsequent malaria infections.”

• The authors and others have previously shown (Frost J et al. Sci Adv 2022, Hoffmann et al. EBioMedicine 2021) that iron deficiency results in reduced neutrophil numbers in different infection models. This could also have contributed to the observed effect in initial infection control but may have also been linked altered histopathology seen in Figure 7. However, no mention of neutrophil numbers in this model is made. It would be important if the authors could provide information on neutrophil numbers (only if this analysis has been already performed) and discuss this issue in association with their observation.

Response: We appreciate that the reviewer has brought attention to this important topic. As they mention, iron deficiency can have a negative impact on the neutrophil response (PMID: 36197985, 34488018) but it can also cause a maladaptive excessive neutrophil response due to failed adaptive immunity (PMID: 33665641). In this study, we show that there is no difference in splenic neutrophil numbers between wild-type and TfrcY20H/Y20H mice, eight days after P. chabaudi infection (Figure S3B). Moreover, the histopathologists detected no liver neutrophil infiltration in either genotype, but rather observed infiltration of mononuclear leukocytes upon P. chabaudi infection. Hence, it appears unlikely that neutrophils were a major contributor to differences in either immunity or pathology in this specific context. However, we cannot definitively rule out that neutrophil numbers were affected earlier in the infection or that neutrophil function was impaired due to cellular iron deficiency.

A section was added to the discussion to address the role of innate immune cells in our model (line 354-363):

“The inhibited innate immune response to P. chabaudi in TfrcY20H/Y20H mice likely contributed to both the increased pathogen burden and the decreased liver pathology. Splenic MNPs are important for controlling parasitaemia (34,35,72), but MNPs are also vital for maintaining tissue homeostasis and preventing tissue damage in malaria (43,73). Although other innate cells, such as neutrophils, NK cells and γδT cells are an important part of the immune response to malaria, only the MNP response was distinctly impaired in TfrcY20H/Y20H mice. Notably, neutrophils are known to be sensitive to iron deficiency (16,74) and to affect both immunity and pathology in malaria (75,76). However, in the context of recently mosquito-transmitted P. chabaudi it appears that monocytes and macrophages, rather than granulocytes, may be particularly important for parasite control and tissue homeostasis (43,72).”

• In addition, alternative mechanism leading to immune tolerance and reduced tissue damage such as induction of heme oxygenase-1, which is also affected by systemic iron availability, should be discussed.

Response: __An addition was made to the results section and to Figure S5 to address this reviewer comment (line __269-274):

“In addition, we measured the expression of two genes that are known to have a hepatoprotective effect in the context of iron loading in malaria: Hmox1 (encodes haemoxygenase-1) and Fth1 (encodes ferritin heavy chain). Liver gene expression of Hmox1 was higher in TfrcY20H/Y20H mice, while the expression of Fth1 did not differ between genotypes, eight days after infection (Figure S5H-I). Thus, the higher expression of Hmox1 may have contributed to the hepatoprotective effect in TfrcY20H/Y20H mice.”

A relevant sentence was also added to the discussion (line 313-318):

“For example, HO-1 plays an important role in detoxifying free haem that occurs as a result of haemolysis during malaria infection, thus preventing liver damage due to tissue iron overload, ROS and inflammation (62). Interestingly, infected TfrcY20H/Y20H mice had higher expression of Hmox1, but levels of liver iron and ROS comparable to that of wild-type mice. Consequently, this may be indicative of increased haem processing that could have a tissue protective effect”

Significance (Required):

Important and intersting study highlighting the central role of iron homeostasis for immune repsonse to infection. General interest because iron deficiency has high prevalence in areas with high enedemic burden of infection

Reviewer's expertise: infectious disease, immunity, iron homeostasis-- both basic science and clincal expertise (more than 300 peer reviewed publications on these topcis)

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

Herein Wideman provide novel and important evidence on the role of iron availability for mounting an efficient immune response in a malaria infection model. They employed TfRC Y201H/Y201H mice which develop iron deficiency due to impaired cellular ingestion of transferrin bound iron. They found that those mice develop higher peak parasitemia after vector borne exposure to Pl. chabaudi chabaudi which was paralleled by an impaired immune response as reflected by altered CD4 cell activation, reduced IFN-g formation or reduced B-cell responsiveness. Those deficiencies could be re-covered upon ex vivo iron supplementation pointing to the importance of iron availability for mounting-CD4+ and B-cell specific anti-plasmodial immune responses at the initial phase of infection. However, TFRC mutated mice were able to clear infection over time in a comparable fashion to wt mice. This excellent study is important in convincingly showing (by employing high quality immunological analyses) the importance of cellular iron deficiency on immune responses in an infection model of general interest. It also indicates that overwhelming immune response as seen in wt mice is associated with organ damage over time.

Minor points:

The authors should discuss why and how TFRC mutated mice were able to control infection over time in a comparable fashion as wt mice although peak parasitemia was significantly higher? The authors and others have previously shown (Frost J et al. Sci Adv 2022, Hoffmann et al. EBioMedicine 2021) that iron deficiency results in reduced neutrophil numbers in different infection models. This could also have contributed to the observed effect in initial infection control but may have also been linked altered histopathology seen in Figure 7. However, no mention of neutrophil numbers in this model is made. It would be important if the authors could provide information on neutrophil numbers (only if this analysis has been already performed) and discuss this issue in association with their observation. In addition, alternative mechanism leading to immune tolerance and reduced tissue damage such as induction of heme oxygenase-1, which is also affected by systemic iron availability, should be discussed.

Significance

Important and intersting study highlighting the central role of iron homeostasis for immune response to infection General interest because iron deficiency has high prevalence in areas with high enedemic burden of infection

Reviewer's expertise: infectious disease, immunity, iron homeostasis-- both basic science and clincal expertise (more than 300 peer reviewed publications on these topics)

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the authors have studied the role of iron deficiency in the host response to Plasmodium infection using a transgenic mouse model that carries a mutation in the transferrin receptor. They show that restricted cellular iron acquisition attenuated P. chabaudi infection- induced splenic and hepatic immune responses which in turn mitigated the immunopathology, even though the peak parasitemia was significantly high in the mutant mice. Interestingly, the course of parasite infection doesn't seem to be affected in the mutant mice compared to the wildtype mice despite the induction of poor immune responses. The authors show that the decreased cellular iron uptake broadly impact both innate and adaptive components of the immune system. Conversely, free iron supplementation restored the immune cell functions.

• The study is well performed, and the manuscript is well written. However, the authors should show how conserved the role of cellular iron is across other rodent malaria parasite species at least with P. yoelii or P. berghei blood stage infection models. This question becomes critical to address in order to understand broad relevance to human malaria infections where both the host and parasites are genetically diverse.
• Since, restricted cellular iron uptake mitigates the immunopathology, the authors should explore whether this could also relieve the cerebral malaria condition that is caused by the hyper inflammation in the brain. They should use the P. berghei ANKA parasite strain which causes t cerebral malaria in mice. I think would increase impact of the paper.

Minor comments:

• Line 222: repeating word, "iron iron-supplemented...."
• Figure 3C, S4C & S5F: Why Mann-Whitney test is performed in these particular graphs, whereas rest of the two groups comparison were done using Welch's test? The authors should clearly mention this in the methods section.
• Have authors explored whether gamma-delta T cell responses are affected in the mutant mouse strain compared to wildtype mice as they are one of the early responders and the key cytokine producing cells against the Plasmodium blood stage infection.

Significance

Overall, the study provides novel insights into the role of iron in the immune response to Plasmodium blood stage infection using a rodent malaria model and the interplay of infection, immunity and the development of pathology. As such it is an important 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 #1

Evidence, reproducibility and clarity

In this paper by Wideman et al, the authors seek to determine the role of cellular iron homeostasis in the pathogenesis of murine malaria.

The authors to attempt to disentangle the effects of anemia from that of cellular iron deficiency. The authors elegantly make use of a murine model of a rare human mutation in the transferrin receptor. This mutation leads to decreased receptor internalization and decreased cellular iron, but otherwise healthy mice. Using this model, the authors use a P. chabaudi infection model and show an increase in pathogen burden and a decrease in pathology. They show in some detail that the immune response to P. chabaudi infection is blunted, both T and B-cell responses are attenuated in the TfRY20H/Y20H model, and the block in proliferation can be rescued by exogenous iron supplementation. They also show that decreased cellular iron attenuates liver pathology through potentially multiple mechanisms.

Minor comments:

• The peak of parasitemia is relatively low (approx..3%) compared to other published studies (e.g. PMID: 22100995, 16714546, 31110285) where the peak in C57BL/6 mice reached 25 - 40%. Can the authors account for this low parasitemia?
• Figure 1K - At homeostasis, serum iron is low in TfR mice however increases to significantly higher than the WT mice at 8 days post infection. Do the authors have an explanation on why these dramatic changes in serum iron are seen?
• Figure S3 - Is it surprising that no effects on splenic neutrophils are seen? Were neutrophils quantified at any other point? These would also be expected to have a role in both the control of malaria infection and on any pathology

Changes to the text

• Fig S1EandF - Please add to the figure legend that these were measured at homeostasis
• Figure 3 - In the legend, H and I are the wrong way around.
• Figure 4 - please add the units of concentration of FeSO4 to all panels
• Line 246 - The authors state: "there was some evidence of decreased malaria-induced hepatomegaly" however there is no significant difference between WT and TfR mice and both show significant hepatomegaly. I feel that this line should be reworded.

Significance

This work is one of the first to attempt to define the requirements for cellular iron in malaria infection. This is a difficult topic, as infection and associated inflammation and the red blood cell destruction caused by malaria all have complex effects on iron within the body. This study fits well with previous observations showing that anemia can be protective as it both prevents parasite growth and limit immunopathology. This work advances the field by demonstrating a cell intrinsic role for iron in malaria infection. There is a broad possible audience for this work, including malaria researchers, immunologists and people interested in the role or iron, both at a cellular level and systemically.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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 comments and insights, we feel the manuscript is now greatly improved. Please find below our answers to the reviewer’s queries

Reviewer #1 (Evidence, reproducibility and clarity):

The manuscript by Niccoli et al. describes the identification of a novel modifier of C9orf72-derived toxicity based on the manipulation of the brain metabolic pathways. The premise for this work is supported by strong literature describing the aberrant glucose metabolism in FTD, AD and other degenerative disorders. The idea tested here is whether increasing the import of pyruvate produced in glia into neurons. They test three different types of importers and find that one of them, Bumpel, the orthologue of human SLC5A12, suppresses toxicity and reduces the accumulation of arginine-containing repeats, GP and PR. The authors investigate several potential mechanisms mediating this reduction of toxic DPRs, but do not find strong evidence linking pyruvate import and increase autophagy or mitochondria metabolism.

Overall, this is an interesting discovery based on a candidate approach that shows the power of Drosophila to efficiently identify novel mediators of neurodegeneration. The article is well written, although more detailed explanations of some experiments would be helpful. The weaknesses of the manuscript are the lack of a clear mechanism mediating the protective activity of pyruvate, the incomplete experiments lacking relevant controls, and the presentation of western blots.

Specific comments:

1. The reduced levels of DPRs require that the expression of C9 mRNA or the GR and PR constructs is examined by qPCR. In figure 3E, GP is not even detectable_

We agree with the reviewer, ideally we would have measured the RNA by qPCR. However, the C9 repeats and the DPR constructs are highly repetitive, it is therefore impossible to do a qPCR for them. The upstream and downstream sequence is identical for the C9 and the bumpel constructs, there isn’t, to our knowledge any unique sequence we can use to measure levels of expression in the presence of bumpel.

We did run a GFP control (Fig 2D) and did not see any difference and we have now carried out a qPCR for Gal4-GeneSwitch (Fig S3) to show that the levels of the driver do not change.

1. I wonder if there are constructs available to silence Bumpel or overexpress the human orthologues of bumpel. These would be nice controls for the effects observed with the Bumpel overexpression

This would be an extremely interesting experiment, however bumpel is normally only expressed in glia, therefore we can’t down-regulated it in glia whilst upregulating 36R in neurons, as we are limited to one driver (since everything is driven by the Gal4/UAS system). Expression of C9 in glia does not have a clear phenotype (our observation), so we can’t drive both in glia. We tried over-expressing the human homologue SLC5A12 , but it did not rescue the C9 phenotype (data not shown), possibly because it requires (like other human SLC5A type transporters) PDZK1 as extra co-factor (Srivastava S. et al, 2019), and this is not present in flies.

1. The argument about bumpel modulating autophagy downstream of Atg1 is not supported by the experimental data

We now have imaging data showing that bumpel modulates the formation of lysosomes, downstream of Atg1 (Fig 5). We also show that bumpel and Atg1 can act synergistically, leading to a much stronger rescue of C9 expression (See Fig 5I.), which also suggests that the two are acting at different points in the same pathway. We also show that bumpel rescues the downregulation of TFEB targets (Fig 5J)

1. Western blots throughout show no control lanes and in several occasions are created with cutout bands. The standard for this type of experiments should be more stringent, with entire gels showing all experimental conditions, which requires consistent methods and results vs selecting the best bands from different gels.

We apologise if this was mis-understood, the lanes shows are all from the same blot, where other samples were run too, and it would be confusing for the reader to include them. We have re-run samples where we had remaining sample from our quantifications, so that the lanes are now contiguous and we provide original blot images in the supplemental information for those we could not re-run. The control for all experiments are the C9 expressing line without bumpel, and this is always present, if the reviewer means we are missing -RU controls, these do not produce any DPRs so are not included in western blot or ELISA quantifications as the signal is not above back-ground.

1. For figures 2B and 5C, please, show representative WBs

These are ELISA quantifications, not western blots, we choose to run these when possible, as they are more quantitative.

1. Figure 5D describes the survival curve as significantly rescued. Statistical tests can indicate differences, but that is in no way convincing. The test may show the curves are different, but the abeta Atg1 flies also seem to start falling early, so an argument could be made in both directions, as a suppressor or an enhancer.

We agree the rescue is not strong enough, we have now removed this lifespan.

1. It is unclear why several results are placed in the supplemental materials. In general, all this material seems highly relevant and related to what is shown in the main figures

We are happy to include them in the main manuscript if this would help the reader, and we have now placed all mitochondrial data in Fig 4.

Minor comments:

Please, define several abbreviations throughout

We apologise for this over-sight, we have now does this.

A couple of sections could be improved by carefully sequencing human vs Drosophila background to advance the argument rather than going in circles. There is also a section on mitophagy in between two sections related to autophagy that could be sequenced better.

We have re-structured the sections, we think this has improved the flow.

There is a sentence at the end of page 6 that seems misplaced

We apologise for the over-sight, and we have removed this

Reviewer #1 (Significance):

Overall, this is an interesting discovery based on a candidate approach that shows the power of Drosophila to efficiently identify novel mediators of neurodegeneration. The article is well written, although more detailed explanations of some experiments would be helpful. The weaknesses of the manuscript are the lack of a clear mechanism mediating the protective activity of pyruvate, the incomplete experiments lacking relevant controls, and the presentation of western blots.

We thank the reviewer for the helpful comments, we have added some details in the methods section, we apologise for not having made it clear that the westerns were all derived from the same blot (we have now placed the originals in the supplemental materials). Regarding mechanism, we now show that bumpel over-expression increases clearance of late stage autolysosomes, possibly by increasing transcription of TFEB target lysosomal genes.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary:<br /> Project investigates the role in dementias of glial glucose uptake, conversion to lactate and shuttling via transporters to neurons to produce pyruvate to fuel TCA cycle production of ATG. The experiments are conducted in Drosophila melanogaster, which have become a powerful model system for understanding neurodegeneration mechanisms associated with ALS/FTD associated C9orf72 pathology. Bumple misexpression is shown to rescue early death phenotype in flies expressing a C9orf72 expansion and flies expressing arginine containing di-peptide repeat proteins. The report describes novel insight into the function of bumpel, demonstrating that this conserved orthologue of human SLC14A functions as a sodium exchange transporter for monocarboxylates pyruvate and lactate. These findings conclude that increased neuronal pyruvate, but not its metabolites, rescues C9orf72 associated pathology.<br /> The authors next set out to describe the mechanism by which increase pyruvate rescues survival in C9orf72 expressing flies. Levels of autolysosomes were increased in C9orf72 expressing flies, and stimulation of autophagy by overexpression of atg1 shown to decrease levels of DPRs (though not to same extent as bumple expression). Expression of bumple in C9orf72 flies led to a modest increase in LC3-II, indicating increased autophagy. Co-overexpression of bumple and atg1 did not have an additive effect, suggesting bumple activates autophagy downstream or independent of atg1 activity. Finally the author extend their findings to amyloid models, suggest a common protective mechanism for elevating neuronal pyruvate levels in neurodegenerative disease.

Major comments

Prior data suggests that bumpel is expressed in glia (for example Yildirim et al 2022). In their study the authors do not present any data to demonstrate that the transporter is normally expressed in neurons in flies. This calls into questions the physiological relevance of their findings, that neuronal upregulation of bumpel is protective against C9orf72 associated pathology in neurons, from which it is reasonable for a reader to conclude that bumpel may be a neuronal target for therapeutic intervention. However, the report well demonstrates that regardless of whether the transporter in native to neurons, the increase in monocarboxylates it facilitates is projective against C9orf72 pathology and thus the overall conclusion of the project is supported by experimental evidence. The point of upregulation of a natively expressed gene versus misexpression of a glial enriched transporter should be considered in a bit more detail in the discussion text. The authors may consider speculating the identify of members of the sodium coupled monocarboxylate transporters that are enriched in neurons. Are any of the bumple human orthologues expressed in neurons?_

We thank the reviewer for this comment and suggestion. The reviewer correctly points out that we do not show whether there is a defect in pyruvate import in C9 expressing flies. We could not identify a validated sodium coupled pyruvate transporter in flies with a strong neuronal expression, we have added a comment in the discussion about this. There are a number of human homologues, some, such as SLC5A8, are expressed in neurons, thus providing a possible therapeutic target. We have added a sentence to this regard in the discussion.

[_OPTIONAL] cDNA overexpression of neuron specific sodium coupled monocarboxylate transporters in C9orf72 fly models would strengthen the conclusion their physiological relevance for ALS/FTD. Fly lines for these are not available in repositories, but could be generated and tested at reasonable cost (<£700, ~3 month duration).

This would be an ideal experiment, however, we could not find a neuronal sodium coupled transporter which is known to import monocarboxylates. There are a number of sodium coupled neuronal transporters, but they are mostly homologous to SLC5A6, which is a glucose coupled transporter. Going forward, we will screen a number of transporters to identify if there are any which import pyruvate.

The role of bumple expression in survival (Figure 1) could be a technical artifact due to dilution of Gal4 between C9orf72 and bumple-ORF transgenes. No expression control is shown (for example GFP, LacZ etc). This theory is unlikely as no improvement in survival was seen for the SLC14A class of transporters which have a matching site directed transgene insertion. For clarity this point relating to controls should be commented on in the text.

The reviewer is correct, there could be a dilution of the Gal4. We don’t like using GFP as a control as we have often seen a worsening when expressing other highly stable proteins at high levels. We have generated an “empty” flyORF line (generated by injecting the empty plasmid into the identical attP site), and used it as a control to check for dilution effects, bumpel still rescued relative to this control, we now include this is the supplementary (Fig S1B).

Reduced Mito-GFP levels are used to support a role for bumple in increasing mitophagy. As mito-GFP is a marker for mitochondria but not specifically mitophagy, an alternative explanation for decreased levels could be reduced mitochondria biogenesis. The text should be amended to clarify this point.<br /> The role of Pink1 RNAi in modifying mitophagy is a bit overstated. Whilst Pink1 is involved in stress associated mitophagy, its role in basal mitochondria turnover is less well defined. Text should be adapted.

We have added qualifying statements regarding the possibility of reduced mitochondrial biogenesis, and the fact that Pink1’s role in basal mitophagy is not very clear. The use of the mitophagy inducer drug, Kaempferol, however, suggests that mitophagy is unlikely to be a cause of the DRP reduction.

Minor comments

Introduction well describes current state of C9orf72 fly models. Introduction would benefit from a few comparable lines for AD models. The first paragraph of reports may also be better placed in the introduction._

We thank the reviewer for the suggestion, and have added a more in depth introduction to Aß and have moved the first paragraph of the results section to the introduction

Figure 1 presents survival for three SLC16A transporters and bumple. The C9 control curve appears to be consistent between charts, likely indicating the same control used across experiments, rather than independent controls for each chart. The authors should considered showing either all SLC16A and bumple data on a single chart, or clarify in the figure legend that a common control dataset is used. GFP control is used in later experiments (Figure 2).

We have now indicated that the SLC16A transporters were run together in the figure legend.

Choice of amyloid model needs a line of explanation, particularly with regard to extra/intracellular deposition of amyloid in this model.

We have now added a few sentences describing this when the model is introduced

Fruit Fly Injection method section needs a bit more detail to describe site of injection (head, body etc). This is not clear in the result section either.

We have now added this, the injection was done in the abdomen.

How were bumple orthologues identified? What degree of conservation (sequence homology etc?)

The bumpel orthologues are those identified as most similar by flybase. We have now added the degree of conservation in the text

The speculative mechanism for C9 pathology modification involves interaction of neurons and glia, monocarboxylate transporters and changes in autophagy activity. For clarity a diagram showing the model may be a helpful addition.

We have now added a diagram explaining how we think the rescue is achieved

Typos:<br /> Figure 1 Legend - "p values of ona way ANOVA "

We apologise for the error, and have now corrected it

Figure S2 Legend - Atg1 RNAi genotypes from S2 legend are mentioned erroneously

We apologise for the error, and have now corrected it

Repetition of text in results: "Bumpel, together with its paralogues kumpel and rumpel, is expressed in glia in flies, where it is thought to promote transport of substrates across the brain (31)."

We apologise and have rectified this

"Modulation of Atg1 when bumpel was co-overexpressed, however, did not affect GP<br /> levels (Fig 4E, F)" - Should be refering to Fig 4D, E)

We apologise and have rectified this

Reviewer #2 (Significance):

The study will be of broadly of interest to researcher working in the fields of neurodegeneration and metabolism, providing evidence for a protective role of elevated pyruvate in neuron that provide new understand relating to pathology in C9orf72 associated motor neuron disease and frontotemporal dementia.

Strengths:<br /> The study presents novel data to demonstrate that overexpression of fly monocarboxylate transporter bumple rescues an early death phenotype associate with ALS/FTD gene C9orf72. Any novel therapeutic strategies of ALS are of interest to the field, and the strategy demonstrated here may be readily translated to human cell culture systems for proof of principle translational studies to a more physiologically relevant system. This study further demonstrates the utility of invertebrate models to generate novel understanding of C9orf72 pathology.

Limitations:<br /> The study speculates that there is a link between pyruvate levels and increased autophagy, however the mechanisms by which this occurs is not defined in present study. This is a limitation of the experiment, though opens up an interesting question for future studies._

We thank the reviewer for their comments, and we have now added experiments characterising the role of bumpel in autophagy, particularly showing its rescue of a late autolysosomal block.

Reviewer expertise: The reviewer researches ALS and dementia associated neurodegeneration, utilising Drosophila, rodent and stem cell derived model systems.

Reviewer #3 (Evidence, reproducibility and clarity):

This is an interesting manuscript in which the authors provide evidence that elevated neuronal expression of the pyruvate transporter bumpel can partially rescue shortened lifespan in fly models of frontotemporal dementia and Alzheimer's disease. In addition, elevated neuronal bumpel expression can reduce accumulation of arginine containing FTD-linked dipeptide repeat proteins. Some evidence is presented that elevated neuronal bumpel expression may activate autophagy. These findings are novel and may have implications for therapeutic interventions based on pyruvate import/metabolism to treat neurodegenerative disorders. However, I have several concerns as follows:

Major Comments:

1. The authors provide no explanation as to why they targeted bumpel overexpression in neurons. Endogenous bumpel appears to be predominately expressed in glia cells so why not target these cells instead?

We wanted to increase pyruvate import in neurons, so we over-expressed a number of pyruvate transporter that were available in the fly ORF stock centre (so that they would all be inserted into the same site and therefore directly comparable), we were mainly interested in cell autonomous effects of importing glycolytic metabolites. Over-expressing bumpel in glia would be indeed an extremely interesting experiment, unfortunately we do not have the ability to express C9 in neurons while over-expressing bumpel in glia as we only have one over-expression system that works. We are working towards generating a new C9 model so we can then use the Gal 4 system to over-express bumpel in glia, but this is currently not available yet. Over-expression of C9 in glia is not toxic and not a good model of disease.

1. Data is shown that overexpressed bumpel can suppress GR and PR dipeptide repeat toxicity when these peptides are translated using an ATG start codon (Fig 2D,E). Does bumpel mediated neuroprotection also correlate with a reduction in DPR levels driven with an ATG start codon?

This would be a very interesting question, unfortunately, whist the Isaacs lab kindly made available the GR antibody for the initial ELISA experiment, we no longer have that antibody available and we do not have a working PR antibody. GR and PR westerns are not possible to carry out as the proteins are too positively charged to run. We do show that bumpel can down-regulate Aß from a UAS promoter, so its effect is not specific to RAN translation.

1. The authors provide some evidence suggesting that overexpression of bumpel increases autophagy in the fly brain. However, knockdown of Atg1 while co-expressing bumpel (Fig 4E) did not result in increased GP protein levels. In addition, Atg1 knockdown did not attenuate the protective effects of bumpel overexpression (Fig 4I), suggesting that bumpel is working through a pathway independent of autophagy to promote DPR clearance and protection against toxic peptide accumulation. The authors need to modify the interpretation of their data and temper their claim that autophagy contributes to bumpel-mediated protective effects in the CNS.

We apologise the data was not strong enough. We have now added evidence that bumpel acts downstream of Atg1, on late stage autolysosomal clearance. We also show that bumpel and Atg1 can act synergistically to improve the C9 phenotype when over-expressed, this is now described in Fig 5.

1. Although the authors present evidence that increased bumpel expression can activate autophagy, the data is not convincing that the neuroprotective effects associated with bumpel are mediated through autophagy. Pyruvate, in some circumstances, can non-enzymatically scavenge hydrogen peroxide or in other cases trigger oxidative stress resistance through hormetic ROS signaling. The authors should consider these alternative possibilities.

These are indeed possibilities, we have added a sentence to that effect in the discussion, we have now also showed that bumpel is affecting late clearance of autolysosomes, and is leading to an increase in TFEB targets.

1. The authors rely on overexpressing bumpel to attenuate C9 toxicity in flies. They should perform the opposite experiment and knockdown bumpel to demonstrate that reduced bumpel expression results in potentiation of C9 and amyloid beta neurotoxicity. In addition, then should show that knockdown of bumpel expression has some effect on autophagy.

This would be a very interesting experiment, unfortunately bumpel is expressed only in a few glia subtypes in a wild type fly, and we can’t downregulate it in glia while over-expressing toxic proteins in neurons, because of limitations of our expression system, both genes need to be over-expressed in the same cell type. We have tried downregulating bumpel in neurons, and don’t get an effect on phenotype, and no effect on DPR levels, but bumpel expression in neurons is extremely low. Moreover, bumpel has 2 paralogs, rumpel and kumpel,(also only present in glia) and all three need to be knocked out for phenotypes to become visible in glia (Yildirim et al, 2022). These experiments would be interesting but outside out scope.

We are in the process of generating new C9 models to be able to do these experiments, but these are currently outside the scope of this work.

Minor Comments:

1. Neuronal overexpression of bumpel appears to shorten lifespan of wild type flies (Fig 2A). It is possible that neuronal import of pyruvate may drive mitochondrial oxidative phosphorylation and ROS formation. The authors should comment on this possibility in the discussion._

This is a very good point, we have added a point to that effect.

1. In Fig 3 the authors used a mixture of sodium pyruvate and ethyl pyruvate to demonstrate the import properties of bumpel. The rationale for using ethyl pyruvate is unclear as this membrane-permeable metabolite can by-pass any transporters.

The ethyl pyruvate was only used in the injection of flies, not for the FRET experiments looking at the import properties of bumpel. Since we were not over-expressing bumpel, we needed the pyruvate to by-pass the requirement for a transporter. We were showing that delivery of pyruvate by another methods (other than by a transporter) was able to phenocopy the over-expression of bumpel, thus showing the effect is mediated by pyruvate entrance into the cell.

1. In the introduction several acronyms are used (i.e. GRN, MAPT, TREM2) that are not defined.

We apologise and have now rectified this.

Reviewer #3 (Significance):

To my knowledge, this is the first study to identify that bumpel can permit the import of pyruvate and lactate into neurons when ectopically expressed in the fly brain. The fact that increased neuronal pyruvate import can partially protect against toxic peptide accumulation is unexpected and quite novel. Although some evidence is presented that bumpel can trigger autophagy, it is not clear if autophagy is mediating bumpel neuroprotective effects. Alternative mechanisms related to pyruvate effects on ROS and oxidative stress resistance should be considered.

We thank the reviewer for their comments, and have added clarifying statements regarding the potential role of ROS.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This is an interesting manuscript in which the authors provide evidence that elevated neuronal expression of the pyruvate transporter bumpel can partially rescue shortened lifespan in fly models of frontotemporal dementia and Alzheimer's disease. In addition, elevated neuronal bumpel expression can reduce accumulation of arginine containing FTD-linked dipeptide repeat proteins. Some evidence is presented that elevated neuronal bumpel expression may activate autophagy. These findings are novel and may have implications for therapeutic interventions based on pyruvate import/metabolism to treat neurodegenerative disorders. However, I have several concerns as follows:

Major Comments:

1. The authors provide no explanation as to why they targeted bumpel overexpression in neurons. Endogenous bumpel appears to be predominately expressed in glia cells so why not target these cells instead?
2. Data is shown that overexpressed bumpel can suppress GR and PR dipeptide repeat toxicity when these peptides are translated using an ATG start codon (Fig 2D,E). Does bumpel mediated neuroprotection also correlate with a reduction in DPR levels driven with an ATG start codon?
3. The authors provide some evidence suggesting that overexpression of bumpel increases autophagy in the fly brain. However, knockdown of Atg1 while co-expressing bumpel (Fig 4E) did not result in increased GP protein levels. In addition, Atg1 knockdown did not attenuate the protective effects of bumpel overexpression (Fig 4I), suggesting that bumpel is working through a pathway independent of autophagy to promote DPR clearance and protection against toxic peptide accumulation. The authors need to modify the interpretation of their data and temper their claim that autophagy contributes to bumpel-mediated protective effects in the CNS.
4. Although the authors present evidence that increased bumpel expression can activate autophagy, the data is not convincing that the neuroprotective effects associated with bumpel are mediated through autophagy. Pyruvate, in some circumstances, can non-enzymatically scavenge hydrogen peroxide or in other cases trigger oxidative stress resistance through hormetic ROS signaling. The authors should consider these alternative possibilities.
5. The authors rely on overexpressing bumpel to attenuate C9 toxicity in flies. They should perform the opposite experiment and knockdown bumpel to demonstrate that reduced bumpel expression results in potentiation of C9 and amyloid beta neurotoxicity. In addition, then should show that knockdown of bumpel expression has some effect on autophagy.

Minor Comments:

1. Neuronal overexpression of bumpel appears to shorten lifespan of wild type flies (Fig 2A). It is possible that neuronal import of pyruvate may drive mitochondrial oxidative phosphorylation and ROS formation. The authors should comment on this possibility in the discussion.
2. In Fig 3 the authors used a mixture of sodium pyruvate and ethyl pyruvate to demonstrate the import properties of bumpel. The rationale for using ethyl pyruvate is unclear as this membrane-permeable metabolite can by-pass any transporters.
3. In the introduction several acronyms are used (i.e. GRN, MAPT, TREM2) that are not defined.

Significance

To my knowledge, this is the first study to identify that bumpel can permit the import of pyruvate and lactate into neurons when ectopically expressed in the fly brain. The fact that increased neuronal pyruvate import can partially protect against toxic peptide accumulation is unexpected and quite novel. Although some evidence is presented that bumpel can trigger autophagy, it is not clear if autophagy is mediating bumpel neuroprotective effects. Alternative mechanisms related to pyruvate effects on ROS and oxidative stress resistance should be considered.

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:

Project investigates the role in dementias of glial glucose uptake, conversion to lactate and shuttling via transporters to neurons to produce pyruvate to fuel TCA cycle production of ATG. The experiments are conducted in Drosophila melanogaster, which have become a powerful model system for understanding neurodegeneration mechanisms associated with ALS/FTD associated C9orf72 pathology. Bumple misexpression is shown to rescue early death phenotype in flies expressing a C9orf72 expansion and flies expressing arginine containing di-peptide repeat proteins. The report describes novel insight into the function of bumpel, demonstrating that this conserved orthologue of human SLC14A functions as a sodium exchange transporter for monocarboxylates pyruvate and lactate. These findings conclude that increased neuronal pyruvate, but not its metabolites, rescues C9orf72 associated pathology.

The authors next set out to describe the mechanism by which increase pyruvate rescues survival in C9orf72 expressing flies. Levels of autolysosomes were increased in C9orf72 expressing flies, and stimulation of autophagy by overexpression of atg1 shown to decrease levels of DPRs (though not to same extent as bumple expression). Expression of bumple in C9orf72 flies led to a modest increase in LC3-II, indicating increased autophagy. Co-overexpression of bumple and atg1 did not have an additive effect, suggesting bumple activates autophagy downstream or independent of atg1 activity. Finally the author extend their findings to amyloid models, suggest a common protective mechanism for elevating neuronal pyruvate levels in neurodegenerative disease.

Major comments

Prior data suggests that bumpel is expressed in glia (for example Yildirim et al 2022). In their study the authors do not present any data to demonstrate that the transporter is normally expressed in neurons in flies. This calls into questions the physiological relevance of their findings, that neuronal upregulation of bumpel is protective against C9orf72 associated pathology in neurons, from which it is reasonable for a reader to conclude that bumpel may be a neuronal target for therapeutic intervention. However, the report well demonstrates that regardless of whether the transporter in native to neurons, the increase in monocarboxylates it facilitates is projective against C9orf72 pathology and thus the overall conclusion of the project is supported by experimental evidence. The point of upregulation of a natively expressed gene versus misexpression of a glial enriched transporter should be considered in a bit more detail in the discussion text. The authors may consider speculating the identify of members of the sodium coupled monocarboxylate transporters that are enriched in neurons. Are any of the bumple human orthologues expressed in neurons?<br /> [OPTIONAL] cDNA overexpression of neuron specific sodium coupled monocarboxylate transporters in C9orf72 fly models would strengthen the conclusion their physiological relevance for ALS/FTD. Fly lines for these are not available in repositories, but could be generated and tested at reasonable cost (<£700, ~3 month duration).<br /> The role of bumple expression in survival (Figure 1) could be a technical artifact due to dilution of Gal4 between C9orf72 and bumple-ORF transgenes. No expression control is shown (for example GFP, LacZ etc). This theory is unlikely as no improvement in survival was seen for the SLC14A class of transporters which have a matching site directed transgene insertion. For clarity this point relating to controls should be commented on in the text.<br /> Reduced Mito-GFP levels are used to support a role for bumple in increasing mitophagy. As mito-GFP is a marker for mitochondria but not specifically mitophagy, an alternative explanation for decreased levels could be reduced mitochondria biogenesis. The text should be amended to clarify this point.<br /> The role of Pink1 RNAi in modifying mitophagy is a bit overstated. Whilst Pink1 is involved in stress associated mitophagy, its role in basal mitochondria turnover is less well defined. Text should be adapted.

Minor comments

Introduction well describes current state of C9orf72 fly models. Introduction would benefit from a few comparable lines for AD models. The first paragraph of reports may also be better placed in the introduction.

Figure 1 presents survival for three SLC16A transporters and bumple. The C9 control curve appears to be consistent between charts, likely indicating the same control used across experiments, rather than independent controls for each chart. The authors should considered showing either all SLC16A and bumple data on a single chart, or clarify in the figure legend that a common control dataset is used. GFP control is used in later experiments (Figure 2).

Choice of amyloid model needs a line of explanation, particularly with regard to extra/intracellular deposition of amyloid in this model.

Fruit Fly Injection method section needs a bit more detail to describe site of injection (head, body etc). This is not clear in the result section either.

How were bumple orthologues identified? What degree of conservation (sequence homology etc?)

The speculative mechanism for C9 pathology modification involves interaction of neurons and glia, monocarboxylate transporters and changes in autophagy activity. For clarity a diagram showing the model may be a helpful addition.

Typos:

Figure 1 Legend - "p values of ona way ANOVA "

Figure S2 Legend - Atg1 RNAi genotypes from S2 legend are mentioned erroneously

Repetition of text in results: "Bumpel, together with its paralogues kumpel and rumpel, is expressed in glia in flies, where it is thought to promote transport of substrates across the brain (31)."

"Modulation of Atg1 when bumpel was co-overexpressed, however, did not affect GP<br /> levels (Fig 4E, F)" - Should be refering to Fig 4D, E)

Significance

The study will be of broadly of interest to researcher working in the fields of neurodegeneration and metabolism, providing evidence for a protective role of elevated pyruvate in neuron that provide new understand relating to pathology in C9orf72 associated motor neuron disease and frontotemporal dementia.

Strengths:

The study presents novel data to demonstrate that overexpression of fly monocarboxylate transporter bumple rescues an early death phenotype associate with ALS/FTD gene C9orf72. Any novel therapeutic strategies of ALS are of interest to the field, and the strategy demonstrated here may be readily translated to human cell culture systems for proof of principle translational studies to a more physiologically relevant system. This study further demonstrates the utility of invertebrate models to generate novel understanding of C9orf72 pathology.

Limitations:

The study speculates that there is a link between pyruvate levels and increased autophagy, however the mechanisms by which this occurs is not defined in present study. This is a limitation of the experiment, though opens up an interesting question for future studies.

Reviewer expertise: The reviewer researches ALS and dementia associated neurodegeneration, utilising Drosophila, rodent and stem cell derived model systems.

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 Niccoli et al. describes the identification of a novel modifier of C9orf72-derived toxicity based on the manipulation of the brain metabolic pathways. The premise for this work is supported by strong literature describing the aberrant glucose metabolism in FTD, AD and other degenerative disorders. The idea tested here is whether increasing the import of pyruvate produced in glia into neurons. They test three different types of importers and find that one of them, Bumpel, the orthologue of human SLC5A12, suppresses toxicity and reduces the accumulation of arginine-containing repeats, GP and PR. The authors investigate several potential mechanisms mediating this reduction of toxic DPRs, but do not find strong evidence linking pyruvate import and increase autophagy or mitochondria metabolism.

Overall, this is an interesting discovery based on a candidate approach that shows the power of Drosophila to efficiently identify novel mediators of neurodegeneration. The article is well written, although more detailed explanations of some experiments would be helpful. The weaknesses of the manuscript are the lack of a clear mechanism mediating the protective activity of pyruvate, the incomplete experiments lacking relevant controls, and the presentation of western blots.

Specific comments:

1. The reduced levels of DPRs require that the expression of C9 mRNA or the GR and PR constructs is examined by qPCR. In figure 3E, GP is not even detectable
2. I wonder if there are constructs available to silence Bumpel or overexpress the human orthologues of bumpel. These would be nice controls for the effects observed with the Bumpel overexpression
3. The argument about bumpel modulating autophagy downstream of Atg1 is not supported by the experimental data
4. Western blots throughout show no control lanes and in several occasions are created with cutout bands. The standard for this type of experiments should be more stringent, with entire gels showing all experimental conditions, which requires consistent methods and results vs selecting the best bands from different gels.
5. For figures 2B and 5C, please, show representative WBs
6. Figure 5D describes the survival curve as significantly rescued. Statistical tests can indicate differences, but that is in no way convincing. The test may show the curves are different, but the abeta Atg1 flies also seem to start falling early, so an argument could be made in both directions, as a suppressor or an enhancer.
7. It is unclear why several results are placed in the supplemental materials. In general, all this material seems highly relevant and related to what is shown in the main figures

Minor comments:

Please, define several abbreviations throughout

A couple of sections could be improved by carefully sequencing human vs Drosophila background to advance the argument rather than going in circles. There is also a section on mitophagy in between two sections related to autophagy that could be sequenced better.

There is a sentence at the end of page 6 that seems misplaced

Significance

Overall, this is an interesting discovery based on a candidate approach that shows the power of Drosophila to efficiently identify novel mediators of neurodegeneration. The article is well written, although more detailed explanations of some experiments would be helpful. The weaknesses of the manuscript are the lack of a clear mechanism mediating the protective activity of pyruvate, the incomplete experiments lacking relevant controls, and the presentation of western blots.

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 the reviewers for their insightful comments. We believe that the changes that have been suggested will add greatly to this paper, and we will endeavor to incorporate as many of these suggestions as we can.

Reviewer #1

This is an interesting study, which presents yet another mechanism involved in the regulation of tumour associated paraneoplastic syndromes, such as muscle wasting. It suggest the intriguing possibility of using a hight fat diet and modulating mitochondrial metabolism as a means of alleviating cachectic muscle wasting. However, as it stands, these aspects of the study remains rather preliminary. This is particularly the case regarding the role of dietary interventions in the model and understanding of the type of metabolic reprogramming in wasting muscles, which lack direct experimental evidence. If the authors were able to further develop this aspects of the study with robust experimental work, it will make it a very valuable and impactful report.

1- All the mitochondrial phenotypes presented should be compared in the two different tumour models (Gal4/UAS and the QF/QUAS driven), which are indistinctively used throughout the study.

We will ensure that mitochondrial size and TMRE staining are performed in the two different tumour models so that they can be compared.

2- The mitochondrial phenotype of wasting muscles is only evident towards the late stages of tumourigenesis (7 day old larvae). Mitochondria of 5 day old tumour bearing animals is indistinct from the control ones. Given that 5 days is the oldest wild type larvae available, the authors need to assess the mitochondrial size and function in muscles form developmentally delayed, no-tumour bearing larvae to discard a trivial contribution of failed metamorphosis in such phenotype.

We will examine mitochondrial size and TMRE in pmhGal4 > torsoRNAi animals (which undergo delayed metamorphosis) compared with control animals.

4- TMRE staining presented in Figure 1 is not convincing. If available, a biochemical and/or more quantitative method to address mitochondrial function should be used.

We will perform ATP synthesis and O2 consumption assays to provide a biochemical method to accompany the TMRE assays.

5- Related to the point above. The extent of the mitochondrial phenotype following genetic manipulations in the tumour or muscle is not consistently analysed. In some cases, mitochondrial size and activity is assessed but in multiple cases, only mitochondrial size is measured. Mitochondrial activity should be assessed in all cases also.

We will assess mitochondrial activity in a time course of RasV12DlgRNAi vs w1118, as well as tumor-bearing animals treated with nicotinamide, QF-QUAS RasV12scribRNAi, MHC> foxoRNAi, and RasV12DlgRNAi > Impl2RNAi.

6- Are mitochondrial fusion proteins such as Marf upregulated in muscles undergoing wasting in Rasv12dlg RNAi animals?

Regulation of neither Opa1 nor Marf are altered in our proteomics study.

7- Is overexpression of mitochondrial fusion proteins alone sufficient to induce muscle wasting?

No, overexpression of Marf was not sufficient to induce muscle wasting, however overexpression of Marf caused worsened muscle wasting in tumour-bearing animals. We will include this data in our revised manuscript.

8- Is there a change in the expression of ATP5A in the muscles of bearing animals RasV12dlgRNAi, which has dysfunctional mitochondria compared to the control?

There is no change in ATP5A expression in our proteomics study.

9- Regarding measures of insulin signaling activity in muscle (Figure 2): the data provide on FOXO staining is not very convincing. Improved staining and robust and more quantitative measure of insulin signaling activity, such as western blot analysis of pAkt should be provided. Apart from the nucleus, there is an overall increase in FOXO expression in the muscle cells of RasV12dlgRNAi compared to the control. In control animals, there is no signal of FOXO. How do you explain this?

We have attempted western blots of pAkt in tumour-bearing muscle previously and found that tumour metastases caused unreliable results, making immuno-staining a more reliable option. However, pAkt antibody staining also does not work well in the muscles. The control image we displayed was an extreme example, so we will choose more representative images that show more consistent FOXO staining.

12- In S3 J-L, Since MHC expression is also dependent upon muscle health and integrity, it would be better to use another, and more universal, readout for protein translation/synthesis. For example, labelling the tissue with Puromycin or staining for translation initiation factors.

We will perform O-propargyl-puromycin (OPP) staining for a w1118 vs RasV12DlgRNAi time course to provide another translation readout to accompany the MHC staining.

13- How does lipid/high fat diet restore muscle wasting? What happens to the tumours of high fat and Nicotinamide feed animals? In all cases, the impact on tumour size upon genetic manipulations of the muscle should be shown.

We will measure tumour size in tumour-bearing animals on both nicotinamide and high-fat diets, as well as QF-QUAS RasV12scribRNAi MHC> foxoRNAi, marfRNAi and whdRNAi animals. Impl2RNAi in tumour-bearing animals has been shown already (Lodge et al., 2021).

14- Does NAM feeding or High-fat diet restore whd transcript levels??

We will perform qPCR to examine whd transcript levels in tumour-bearing animals on nicotinamide diets as well as high-fat diets.

15- Do these feeding regimes restore insulin signaling in RasV12dlgRNAi animals?

We have demonstrated that for RasV12dlgRNAi animals fed a nicotinamide diet, FOXO levels are decreased (Fig 5D). We will do the same experiment for tumour-bearing animals fed a high fat diet.

17- Related to the point above, DAPI and phalloidin should be included when showing lipid staining to understand better the cellular structures present in the field of view along with the lipid droplets.

DAPI and phalloidin staining is not compatible with lipid staining, as they require the use of PBST (detergent) which breaks down extracellular lipids. We will include more representative, raw images in which the details of the muscle can be seen.

Minor comments<br /> 1. The order of panels in the figures and the main text should be the same for better readability.

We will revisit the figures to ensure readability is improved.

1. Figure S3 G-H: The image looks out of focus. Is Atg8 expression high near to the nucleus?

Atg8a expression is highest near the nucleus, and is decreased in RasV12dlgRNAi > Impl2RNAi animals. We will provide more representative images to make this clearer.

Reviewer #2

This manuscript proposes and interesting new mechanism how tumours non-autonomously induce muscle mass loss (cachexia) in a genetic Drosophila model. These effects can be modified by diet. Hence results are interesting for both basic and more clinically interested audience.<br /> The weak point of the paper is the limited quantification of mitochondria sizes/morphologies, which is an important point that asks for significant improvement of either the imaging conditions or the image analysis.

1. The authors provide evidence that eye or imaginal disc tumours induce larger mitochondria in muscles. The authors try to quantify mitochondrial sizes using an automated analysis. This is a tricky task from their light microscopy images that appear to be limited in resolution. By looking at the Suppl. Figure 1, I wonder how relevant an increase of a "large" mitochondria fraction from 7 to 12 % is in the tumour larvae, considering that a significant fraction of the mitochondria are currently not counted, as they are too large to be investigated (white colours in S1F, G). Can the authors increase resolution to resolve these large clumps that likely consist of individual mitochondria to reliably segment all of them, and not only a sub fraction. It would be useful to display the size profiles of all mitochondria in various conditions and not only of a very selected subset of "large" mitochondria. This comment applies to all figures in which mitochondria size was quantified and hence is critical for the entire manuscript.

We will utilise a newly developed segmentation and centroid tracking-based analysis pipeline based in MATLAB, that may be able to separate the large clumps of mitochondria, to ensure that as many mitochondria can be quantified as possible. We will also provide size profiles of all mitochondria sizes from all conditions in which we performed mitochondria size analysis.

1. Comparing MitoTracker to TMRE is a valid approach to estimate mitochondria activity/health. The images shown in 1H,I are overview images that seem to show large regional differences in the muscles of unclear origin. High resolution images of representative regions as shown for the ATP5A stains would be more convincing as these can resolve individual mitochondria to hopefully see damaged ones next to normal ones. Would "active" mitochondria not be expected to be the ones that oxidise a lot of fatty acid break down products?

We will take representative zoomed in images for 1H & I to better demonstrate mitochondria morphology.

1. The authors find that co-overexpressing FOXO in muscles results in a more severe muscle degeneration phenotype in tumour bearing animals than tumour alone. However, it seems the important control of FOXO overexpression in an otherwise wildtype animal is missing. In order to judge if the muscles really detach in these genotypes, instead of shrink and finally rupture, high resolution images of muscle attachment sites would be needed.

We will assess if MHCGal4 > UAS dFOXO causes loss of muscle integrity. In addition, in both wildtype and tumour-bearing animals, we will overexpress FOXO in the muscles and stain for muscle attachment proteins such as tiggrin to determine if the phenotype seen is caused by a mislocalisation of proteins at attachment sites.

1. The strongly reduced lipid droplets in the tumour bearing animals is interesting. To better normalise for the reduced size of the muscles, a counter staining for muscle and a following normalisation would make the statement stronger and thus better support the conclusion.

As mentioned above we will provide more representative images to help visualize muscle structures in LipidTOX experiments. In addition, we will normalize the amount of lipid droplets detected to a set area, as opposed to just measuring total lipid droplets.

Reviewer #3

The strength of the study is the use of suitable in vivo model systems, combined with genetic manipulations to study the mechanisms behind cancer cachexia. The weak points of the study is the lack of functional assays such as quantitative measurements of oxidative phosphorylation and metabolites.

1, Throughout the manuscript the authors use TMRE staining to evaluate mitochondrial function. To me it is not clear what function they are actually referring to. I assume they mean respiration/respiratory chain function, as this generates the proton motif force measured, but neither oxygen consumption nor aerobic ATP synthesis is ever mentioned or measured. Especially considering that the authors suggest that an increased flux through beta oxidation, which is a mitochondrial function, results in muscle wasting, the authors might want to consider measuring respiration with different substrates, using either a seahorse or Oroboros or equivalent.

We do not have the necessary equipment or resources to perform Seahorse or Oroboros experiments. Therefore, we will perform O2 consumption and ATP synthesis assays for RasV12dlgRNAi and QF-QUAS RasV12scribRNAi vs w1118, RasV12dlgRNAi > Impl2RNAi, QF-QUAS RasV12scribRNAi > marfRNAi, whdRNAi, and tumour-bearing animals fed high fat diets to provide more insights into mitochondria function.

3, It is difficult to understand that it is even possible for beta oxidation to exceed the capacity of the OXPHOS system. In that case one would have excess of acetyl CoA and NADH, inevitably inhibiting further beta oxidation and the TCA cycle due to lack of NAD, as well as numerous regulatory mechanisms. Additionally, one would expect increased ketone body production. The authors might want to clarify how the excess redox potential, due to increased beta oxidation is utilised.

We will perform acetyl-CoA and NAD/NADH assays in RasV12dlgRNAi and QF-QUAS RasV12scribRNAi vs w1118 to determine if beta-oxidation is occurring in excess. In addition, we will clarify in the text that we hypothesize that increased beta-oxidation is utilizing the muscle’s resources to the point that there is none left to continue energy production.

Minor:

Line 223 "Together, this data suggests that FOXO lies upstream of beta-oxidation, and mitochondria function lies downstream of beta-oxidation".<br /> I would suggest to rephrase. Of course beta-oxidation and the TCA takes place inside mitochondria, so what mitochondrial functions do the authors refer to?

As mentioned earlier, we will perform O2 consumption and ATP synthesis assays to strengthen this claim. In addition, we will rephrase this sentence to avoid confusion.

Line 238 "Overall, this data suggests that the depletion of muscle lipid stores via beta oxidation affects mitochondrial function and is negatively correlated with muscle health in cachectic flies, mice and patients" - The mechanism is not fully clear to me as other energy sources are still available to the fly. The authors might want to expand here.

We will clarify that there may be other energy sources available that were not investigated in this paper.

Line 93 : "To test whether this increase in mitochondrial size could lead to compromised mitochondrial function, we performed live staining with tetramethylrhodamine ethyl ester (TMRE), a compound used to measure the membrane potential of mitochondria." - I am not sure that size on its own correlates with mitochondrial function, but rather the energetic and metabolic state of the cell. Increased biogenesis is a common response to dysfunction, but often reflected in increased mass.

We will clarify the that the increase in size may be a reflection of increased metabolic need of the muscle.

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

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

Reviewer 1:

3- In all cases, the age of experimental animals must be clearly indicated in figures and/or figure legends.

We have already put the ages of the experimental animals in the bottom of the figure legends.

11- Does insulin signaling influence Lipid metabolism in muscle?

We demonstrate in the manuscript that FoxoRNAi in the muscle of tumour-bearing animals reduces whd transcript levels (Fig 4C), and Impl2RNAi in the tumour restores muscle lipid droplet levels (Fig 3G-I).

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

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

Reviewer 1:

10- The phenotype of increased fatty acid oxidation in wasting muscles is inferred as per the proteomic signature but not directly demonstrated. TCA metabolite tracing using 13C-Palmitate should be used to demonstrate this, which is a central point of the manuscript.

The examination of 13C-palmitate would require metabolomic approaches, for which we do not have the necessary equipment and is beyond our timeframe. Thus, we will aim to examine changes in mitochondria metabolism through other measures mentioned above.

16- The lipid phenotype in cachectic fly muscles is not consistent with that reported in humans and shown by the authors in their xenograft model. While loss of lipid droplets is observed in the fly muscle cells, there is increase in the lipid content within the mouse muscle and only extramyocellular lipid is decreased. The relevance of the extracellular lipid is unclear.

We hypothesize that this is due to a transport of lipids from extracellular lipid droplets to mitochondria for utilization, as has been suggested previously (Rambold et al., 2015). Examining in detail if this is the case in our models is beyond the scope of this paper.

Reviewer 3:

2, The authors suggest that an increase in beta oxidation exceeds mitochondrial function (?), which in turn induces a change in mitochondrial morphology, further contributing to the muscle wasting. The authors may want to demonstrate that there is indeed excess beta oxidation, by measuring a toxic accumulation of different lengths of acylcarnitines. For instance, it is well known that patients with beta oxidation defects accumulate toxic intermediates of beta oxidation that can ultimately affect mitochondrial function.<br /> The manuscript would be much improved if oxygen consumption is measured and combined with analysis of acylcarnitines.

The examination of acylcarnitines would require lipidomic approaches, and is beyond our timeframe for these revisions. To try to address the need for investigations if beta-oxidation is in excess, we will perform oxygen consumption assays as mentioned and alter the manuscript to de-emphasize excess beta-oxidation.

4, Unfortunately the supplementary information is in a format I can't open, thus I can't evaluate the method for identifying large mitochondria and other results in these files. This makes part of the reviewing process difficult.

N/A

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 study the mechanisms behind cancer cachexia, using drosophila cancer models. They find that muscle wasting in cachexia is mediated via two different mechanisms: either via insulin signalling and FOXO activation or beta oxidation via mitochondrial fusion.<br /> It is well known that many cancers can induce a catabolic state, compatible with a decrease in insulin signalling and one of the mechanisms proposed. Additionally, the authors suggest that an imbalance between mitochondrial capacity and beta oxidation flux leads to muscle wasting.

Major comments:

1. Throughout the manuscript the authors use TMRE staining to evaluate mitochondrial function. To me it is not clear what function they are actually referring to. I assume they mean respiration/respiratory chain function, as this generates the proton motif force measured, but neither oxygen consumption nor aerobic ATP synthesis is ever mentioned or measured. Especially considering that the authors suggest that an increased flux through beta oxidation, which is a mitochondrial function, results in muscle wasting, the authors might want to consider measuring respiration with different substrates, using either a seahorse or Oroboros or equivalent.
2. The authors suggest that an increase in beta oxidation exceeds mitochondrial function (?), which in turn induces a change in mitochondrial morphology, further contributing to the muscle wasting. The authors may want to demonstrate that there is indeed excess beta oxidation, by measuring a toxic accumulation of different lengths of acylcarnitines. For instance, it is well known that patients with beta oxidation defects accumulate toxic intermediates of beta oxidation that can ultimately affect mitochondrial function.<br /> The manuscript would be much improved if oxygen consumption is measured and combined with analysis of acylcarnitines.
3. It is difficult to understand that it is even possible for beta oxidation to exceed the capacity of the OXPHOS system. In that case one would have excess of acetyl CoA and NADH, inevitably inhibiting further beta oxidation and the TCA cycle due to lack of NAD, as well as numerous regulatory mechanisms. Additionally, one would expect increased ketone body production. The authors might want to clarify how the excess redox potential, due to increased beta oxidation is utilised.
4. Unfortunately the supplementary information is in a format I can't open, thus I can't evaluate the method for identifying large mitochondria and other results in these files. This makes part of the reviewing process difficult.

Minor:

Line 223 "Together, this data suggests that FOXO lies upstream of beta-oxidation, and mitochondria function lies downstream of beta-oxidation".<br /> I would suggest to rephrase. Of course beta-oxidation and the TCA takes place inside mitochondria, so what mitochondrial functions do the authors refer to?

Line 238 "Overall, this data suggests that the depletion of muscle lipid stores via beta oxidation affects mitochondrial function and is negatively correlated with muscle health in cachectic flies, mice and patients" - The mechanism is not fully clear to me as other energy sources are still available to the fly. The authors might want to expand here.

Line 93 : "To test whether this increase in mitochondrial size could lead to compromised mitochondrial function, we performed live staining with tetramethylrhodamine ethyl ester (TMRE), a compound used to measure the membrane potential of mitochondria." - I am not sure that size on its own correlates with mitochondrial function, but rather the energetic and metabolic state of the cell. Increased biogenesis is a common response to dysfunction, but often reflected in increased mass.

Significance

General assessment: The strength of the study is the use of suitable in vivo modelsystems, combined with genetic manipulations to study the mechanisms behind cancer cachexia. The weak points of the study is the lack of functional assays such as quantitative measurements of oxidative phosphorylation and metabolites.

Advance: The main advance of this study is attributed to mechanistic insights behind cancer cachexia and the role of mitochondria in more conditions as opposed to the its involvement in inherited mitochondria disease.

Audience: This report should be of interest to a broad audience since it's studying a condition connected to cancer and cancer metabolism.

Reviewers field of expertise: Mitochondrial disease/dysfunction, in vivo modelling, molecular biology, bioenergetics and metabolism

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

Chen and colleagues are using the Drosophila larval muscles model to investigate how a tumour can non-autonomously induce muscle mass loss, a known phenomenon called cancer cachexia. They report that tumours change muscle mitochondria morphologies, specifically their size and their chemistry. These changes correlate with increase in fat metabolism and a depletion of fat and glycogen reserves. Regarding the molecular mechanism, the authors propose that tumour cells secrete IGF binding protein that reduces the level of insulin and thus insulin signalling in muscle. They test this hypothesis by reducing FOXO activity, a negative regulator of insulin signalling, or mitochondrial fusion in muscles of tumour carrying larvae, which indeed appears to result in muscle improvements. These insights from Drosophila muscles suggest that tumour-caused reduced insulin signalling in muscles can be responsible for tumour induced muscle loss. A similar mechanism may apply to mammals and hence these findings are of clinical interest.

Major comments

1. The authors provide evidence that eye or imaginal disc tumours induce larger mitochondria in muscles. The authors try to quantify mitochondrial sizes using an automated analysis. This is a tricky task from their light microscopy images that appear to be limited in resolution. By looking at the Suppl. Figure 1, I wonder how relevant an increase of a "large" mitochondria fraction from 7 to 12 % is in the tumour larvae, considering that a significant fraction of the mitochondria are currently not counted, as they are too large to be investigated (white colours in S1F, G). Can the authors increase resolution to resolve these large clumps that likely consist of individual mitochondria to reliably segment all of them, and not only a sub fraction. It would be useful to display the size profiles of all mitochondria in various conditions and not only of a very selected subset of "large" mitochondria.<br /> This comment applies to all figures in which mitochondria size was quantified and hence is critical for the entire manuscript.
2. Comparing MitoTracker to TMRE is a valid approach to estimate mitochondria activity/health. The images shown in 1H,I are overview images that seem to show large regional differences in the muscles of unclear origin. High resolution images of representative regions as shown for the ATP5A stains would be more convincing as these can resolve individual mitochondria to hopefully see damaged ones next to normal ones. Would "active" mitochondria not be expected to be the ones that oxidise a lot of fatty acid break down products?
3. The authors find that co-overexpressing FOXO in muscles results in a more severe muscle degeneration phenotype in tumour bearing animals than tumour alone. However, it seems the important control of FOXO overexpression in an otherwise wildtype animal is missing. In order to judge if the muscles really detach in these genotypes, instead of shrink and finally rupture, high resolution images of muscle attachment sites would be needed.
4. The strongly reduced lipid droplets in the tumour bearing animals is interesting. To better normalise for the reduced size of the muscles, a counter staining for muscle and a following normalisation would make the statement stronger and thus better support the conclusion.

Significance

This manuscript proposes and interesting new mechanism how tumours non-autonomously induce muscle mass loss (cachexia) in a genetic Drosophila model. These effects can be modified by diet. Hence results are interesting for both basic and more clinically interested audience.<br /> The weak point of the paper is the limited quantification of mitochondria sizes/morphologies, which is an important point that asks for significant improvement of either the imaging conditions or the image analysis.

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

Larvae bearing RasV12; dlgRNAi eye tumours recapitulate aspects of cachexia, such as muscle wasting. In this manuscript, the authors use their previously characterized RasV12; dlgRNAi larval model of cancer cachexia to show that tumour induced cachectic muscle wasting is associated with excessive mitochondrial fusion, resulting in the formation of enlarged dysfunctional mitochondria in wasted muscle cells. Muscle specific blockade of mitochondrial fusion prevents muscle wasting and restores mitochondrial potential in tumour bearing animals. The authors also link increased mitochondrial size to decreased insulin signaling (increased foxo) caused by the tumour induced pro-cachexia factor and insulin inhibitor Impl2. Consistently, downregulation of ImpL2 from the tumour decreases foxo levels in muscle and reduces mitochondrial size. Finally, the authors show that wasting muscles in flies show decrease lipid droplets and a molecular and proteomic signature indicative of increased fatty acid oxidation. Muscle wasting, loss of lipids and mitochondrial integrity can be restored upon inhibition of Impl2 in the tumour, downregulation of the mitochondrial lipid transporter CPT1 or feeding animals with a high fat diet.

Major comments

1. All the mitochondrial phenotypes presented should be compared in the two different tumour models (Gal4/UAS and the QF/QUAS driven), which are indistinctively used throughout the study.
2. The mitochondrial phenotype of wasting muscles is only evident towards the late stages of tumourigenesis (7 day old larvae). Mitochondria of 5 day old tumour bearing animals is indistinct from the control ones. Given that 5 days is the oldest wild type larvae available, the authors need to assess the mitochondrial size and function in muscles form developmentally delayed, no-tumour bearing larvae to discard a trivial contribution of failed metamorphosis in such phenotype.
3. In all cases, the age of experimental animals must be clearly indicated in figures and/or figure legends.
4. TMRE staining presented in Figure 1 is not convincing. If available, a biochemical and/or more quantitative method to address mitochondrial function should be used.
5. Related to the point above. The extent of the mitochondrial phenotype following genetic manipulations in the tumour or muscle is not consistently analysed. In some cases, mitochondrial size and activity is assessed but in multiple cases, only mitochondrial size is measured. Mitochondrial activity should be assessed in all cases also.
6. Are mitochondrial fusion proteins such as Marf upregulated in muscles undergoing wasting in Rasv12dlg RNAi animals?
7. Is overexpression of mitochondrial fusion proteins alone sufficient to induce muscle wasting?
8. Is there a change in the expression of ATP5A in the muscles of bearing animals RasV12dlgRNAi, which has dysfunctional mitochondria compared to the control?
9. Regarding measures of insulin signaling activity in muscle (Figure 2): the data provide on FOXO staining is not very convincing. Improved staining and robust and more quantitative measure of insulin signaling activity, such as western blot analysis of pAkt should be provided. Apart from the nucleus, there is an overall increase in FOXO expression in the muscle cells of RasV12dlgRNAi compared to the control. In control animals, there is no signal of FOXO. How do you explain this?
10. The phenotype of increased fatty acid oxidation in wasting muscles is inferred as per the proteomic signature but not directly demonstrated. TCA metabolite tracing using 13C-Palmitate should be used to demonstrate this, which is a central point of the manuscript.
11. Does insulin signaling influence Lipid metabolism in muscle?
12. In S3 J-L, Since MHC expression is also dependent upon muscle health and integrity, it would be better to use another, and more universal, readout for protein translation/synthesis. For example, labelling the tissue with Puromycin or staining for translation initiation factors.
13. How does lipid/high fat diet restore muscle wasting? What happens to the tumours of high fat and Nicotinamide feed animals? In all cases, the impact on tumour size upon genetic manipulations of the muscle should be shown.
14. Does NAM feeding or High-fat diet restore whd transcript levels??
15. Do these feeding regimes restore insulin signaling in RasV12dlgRNAi animals?
16. The lipid phenotype in cachectic fly muscles is not consistent with that reported in humans and shown by the authors in their xenograft model. While loss of lipid droplets is observed in the fly muscle cells, there is increase in the lipid content within the mouse muscle and only extramyocellular lipid is decreased. The relevance of the extracellular lipid is unclear.
17. Related to the point above, DAPI and phalloidin should be included when showing lipid staining to understand better the cellular structures present in the field of view along with the lipid droplets.

Minor comments

1. The order of panels in the figures and the main text should be the same for better readability.
2. Figure S3 G-H: The image looks out of focus. Is Atg8 expression high near to the nucleus?

Significance

This is an interesting study, which presents yet another mechanism involved in the regulation of tumour associated paraneoplastic syndromes, such as muscle wasting. It suggest the intriguing possibility of using a hight fat diet and modulating mitochondrial metabolism as a means of alleviating cachectic muscle wasting. However, as it stands, these aspects of the study remains rather preliminary. This is particularly the case regarding the role of dietary interventions in the model and understanding of the type of metabolic reprogramming in wasting muscles, which lack direct experimental evidence. If the authors were able to further develop this aspects of the study with robust experimental work, it will make it a very valuable and impactful report.

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 the reviewers for their comments and suggestions, which were very helpful to improve our manuscript. The revised manuscript notably includes the following improvements:

• To evaluate the relevance of identified candidate targets genes, we integrated an additional screening step in our method, corresponding to the analysis of RNAseq datasets specific of blood or brain cells. RNAseq data from irradiated hematopoietic stem cells or splenic cells were analyzed and included in the new Table S19, and RNAseq data from zika virus-infected neural progenitors were analyzed and included in the new Table S28. In addition, we also verified that the expression of a subset of blood related genes was decreased in the bone marrow cells of p53Δ31/Δ31 mice, known to exhibit increased p53 activity and to phenocopy dyskeratosis congenita (new Figure S8).
• Luciferase data were expanded to show that, for promoters exhibiting a significant p53-mediated repression in luciferase assays, the p53-dependent regulation was abrogated after mutation of the putative DREAM binding site (new Figures 2e and 2i).
• We found putative DREAM binding sites for 151 targets, and the predicted binding sites were precisely mapped relative to the position of ChIP peaks of DREAM subunits (E2F4 and LIN9) and to transcription start sites of target genes. These additional analyses, shown in the new Figures 3a and 3b, further suggest the reliability of our predicted binding sites. Notably, hypergeometric tests of the distribution of DREAM binding sites relative to E2F4/LIN9 ChIP peaks reveal a significant >1300-fold enrichment of these sites at ChIP peaks.
• We now present a detailed comparison of our results with those reported in other studies, notably the predicted E2F and CHR sites from the Target gene regulation database (new Figure S11), or the list of candidate DREAM targets suggested from Lin37 KO cells (new Figure S10 and new Table S35). This also leads us to discuss the different types of DREAM binding sites (bipartite sites (e.g. CDE/CHR or E2F/CLE) vs sites composed of a single E2F or a single CHR motif).
• We integrated updates of the Human phenotype ontology website to include the latest lists of genes related to blood or brain ontology terms in our analysis. In the previous version of the manuscript we had analyzed a total of 811 genes downregulated ≥ 1.5 fold upon bone marrow cell differentiation. Our revised manuscript now includes the analysis of 883 genes.
• Several improvements were made to present our results more clearly and with more details : 1) additional evidence that the differentiation of Hoxa9ER cells correlates with p53 activation is now provided in the new Figure S1; 2) the precise values for gene expression after bone marrow cell differentiation, as well as p53 regulation scores from the Target gene regulation databases are included in the new Tables S1, S5, S8, S11, S14, S20 and S23; 3) A Venn-like diagram was included to summarize the different steps of our approach in the new Figure 3c, with detailed lists of genes selected at each step in new Tables S17 and S26; 4) for genes associated with blood or brain genetic disorders, bibliographic references describing gene mutations and clinical traits were included in a new Table S36; 5) Figure 4a and Table S37 were improved to include evidence that increased BRD8 in glioblastoma cells leads to a decreased expression of several genes transactivated by p53.

Reviewer #1 (Evidence, reproducibility and clarity):

Summary<br /> In this paper the authors describe a data driven approach to identify and prioritise p53-DREAM targets whose repression might contribute to abnormal haematopoiesis and brain abnormalities observed in p53-CTD deleted mice. The premise is that in these mice, (where they have previously demonstrated p53 to be hyperactive in at least a subset of tissues), that the p53-p21-E2F/DREAM axis is at least in part responsible for observed phenotypes due to the repression of E2F and CDE/CHE element containing genes. Their approach to home in on relevant genes is based on transcriptomic gene ontology analysis of genes repressed in these disease settings where they primarily use publicly available data from HOXA9-ER regulated model of HSC expansion wherein they observe increases on p53-p21 expression upon differentiation where they demonstrate that p53-p21 DREAM target genes are suppressed as we would expect in this scenario where p53-p21 is activating withdrawal from cell cycle. They then spend a lot of effort analysing this datasets combining "gene-ontology", "disease phenotype" and "meta-ChIP-seq" analysis of public data to support the observation that mutations of genes suppressed in this manner are disproportionately linked to heritable haematopoetic and brain disorders. While these results are interesting in terms of framing a hypothesis about how mutations in p53-p21-DREAM regulated targets contribute to such conditions, they are to be expected given the now very well described impact of p53-p21 on both E2F4/DREAM targets.

We agree with the referee that the impact of p53-p21 on both E2F4/DREAM targets is well described. However, discussions with many scientists or clinicians specialized in bone marrow failure syndromes or microcephaly diseases led us to realize that most were not familiarized with the p53-DREAM pathway, so that a study that would bridge the gap between DREAM experts and bone marrow or microcephaly specialists would be particularly useful. In addition, we thought that strategies that would rely on disease-based ontology terms were likely to identify new targets, compared to previous studies that considered cell cycle regulation instead of disease phenotypes. Consistent with this, many genes we identified as candidate DREAM targets were not reported in previous studies. In addition, as detailed below, our positional frequency matrices led to identify DREAM binding sites that had not been predicted by previous approaches.

The natural progression of this work would be to go on to show this occurs in relevant cells or tissues derived from the p53-CTD mice as well as look at modulating target genes to understand underlying mechanisms and consequences.<br /> Rather than this, they focus on validating that a sub-set of these targets are indeed suppressed by specific p53 activation by MDM2 inhibitor Nutlin-3A in MEFs by qPCR and that mutation of predicted CDE CHR elements in luciferase constructs leads to increase luciferase activity. While these findings support their predictions, the results are entirely expected based on what is known about such targets and demonstrating that this occurs in MEFs does not closely relate to haematopoietic and brain cells they suggest this regulation is important. In fact, in the discussion, the authors comment on the importance of cell type context specificity in terms of discordance between predictions of TF binding sites and public datasets.

We agree that additional data from relevant cells or tissues were required to strengthen our conclusions. In the revised manuscript, we evaluated the relevance of candidate target genes related to blood ontology terms by integrating an additional screening step in our method, corresponding to the analysis of RNAseq datasets specific of blood cells. We analyzed dataset GSE171697, with RNAseq data from hematopoietic stem cells of unirradiated p53 KO, or unirradiated or irradiated WT mice, as well as dataset GSE204924, with RNAseq data from splenic cells of irradiated p53Δ24/- or p53+/- mice. The latter dataset appeared interesting because p53Δ24 is a mouse model prone to bone marrow failure and the spleen is a hematopoietic organ in mice. The analysis of these datasets is included in the new Table S19. In the datasets,increased p53 activity correlated with the downregulation of most of the 269 candidate DREAM targets. However, 56 genes which appeared upregulated in cells with increased p53 activity were considered poor candidate p53-DREAM targets and removed from further analyses, leading to a list of 213 genes that appeared as better candidate p53-DREAM targets related to blood abnormalities. Furthermore, we also verified that the expression of a subset of blood-related candidate genes was decreased in the bone marrow cells of p53Δ31/Δ31 mice (prone to bone marrow failure) compared to bone marrow cells from WT mice. This result is presented in the new Figure S8.

As for genes related to brain development, we discussed in the previous version of the manuscript that most genes mutated in syndromes of microcephaly or cerebellar hypoplasia are involved in ubiquitous cellular functions (chromosome condensation, mitotic spindle activity, tRNA splicing…), which suggested that our analysis of transcriptomic changes associated with bone marrow cell differentiation might also be used to identify brain specific targets. However, we agree with the referee that confirmation of these brain specific targets in a more relevant cellular context was preferable. In the revised manuscript, we included the analysis of datasets GSE78711 and GSE80434, containing RNAseq data from human cortical neural progenitors infected by the Zika virus (ZIKV) or mock-infected, because ZIKV was shown to cause p53 activation in cortical neural progenitors and microcephaly. This analysis is detailed in the new supplementary Table S28. In both datasets, increased p53 activity correlated with the downregulation of most of the 226 candidate DREAM targets. Sixty-four genes which appeared more expressed in ZIKV-infected cells were considered poor candidate p53-DREAM targets and removed from further analyses, leading to a list of 162 candidate p53-DREAM targets related to brain abnormalities. We think this significantly increases the relevance of our analysis of brain-specific targets.

Finally, they try and contextualise effects in glioblastoma data by correlating target gene expression with levels of BRD8 since it has recently been shown to attenuate p53 function in glioblastoma and show that some of the brain disease associated genes are expressed at higher levels in BRD8 high patient samples. It seems strange here that they do not also look at expression of p21 or other p53 targets that would help ascertain if p53 activity is indeed suppressed. Moreover, much more elegant methods for predicting transcription factor activity could be applied to this data.

We agree with the referee. Indeed, when we had performed the analysis of glioblastoma cells, we first verified that increased BRD8 levels correlated with decreased p21 levels in these cells. However, we had not included this verification in the previous version of the manuscript. In this revision, we improved the Figure 4 (and Table S37) reporting the analysis of glioblastoma cells to address this point. In Figure 4a, we now show the variations in mRNA levels between BRD8Low and BRD8High tumors, for BRD8 itself, as well as 5 genes well-known to be transactivated by p53 (p21, MDM2, BAX, GADD45A and PLK3) and the 77 p53-DREAM targets associated with microcephaly or cerebellar hypoplasia. The data clearly show that tumors with high BRD8 exhibit a decrease in the expression of p53 transactivated targets, and an increase in p53-DREAM repressed targets.

Major Comments<br /> The major result of this paper as it stands is the prioritisation of candidate genes in the p53-DREAM pathway involved in these conditions, and their refined approach used to identify and prioritise these genes and is such more of a starting point for further investigation. They fall short of demonstrating the relevance of their predictions physiologically in tissues from the mice and do not demonstrate functional importance of regulation of targets they put forward. Given that these genes will be co-ordinately regulated, without a mechanistic experiment in physiologically relevant model it is impossible to infer causality. For example, depleting individual targets in the HOXA9 model and evaluating impact on survival, proliferation and differentiation may be a (relatively) simple way to explore this, perhaps comparing to effects of p53 activating agents such as Nutlin-3A. Of note the authors (Jaber 2016 PMID: 27033104) and several other groups had (Fischer 2014 PMID: 25486564 McDade 2014 PMID: 24823795) previously demonstrated the link between p53-p21 and suppression of DNA-repair/Damage related genes (as is also observed here in particular FA-related genes that they discuss briefly here. I would have thought that this would be an obvious starting point for some mechanistic experiments and in fact I note this has been demonstrated before (Li et al 2018 PMID: 29307578)

The starting point of our study is not the prioritization of DREAM target genes, but rather the detailed phenotyping of p53Δ31/Δ31 mice that we performed in previous publications (Simeonova et al. Cell Rep 2013, Toufektchan et al. Nat. Commun. 2016), in which we mentioned phenotypical traits typical of dyskeratosis congenita and Fanconi anemia, including notably bone marrow failure and cerebellar hypoplasia.

We understand that depleting individual targets in the Hoxa9 system and evaluating impact on survival, proliferation and differentiation might seem appropriate to explore their potential causality. However, our previous work on Fanc genes leads us to think that this might not be informative. Regarding this, we now clearly discuss in the revised version of the manuscript : “Finding a functionally relevant [DREAM binding site] for Fanca, mutated in 60% of patients with Fanconi anemia [59,60], may help to understand how a germline increase in p53 activity can cause defects in DNA repair. Importantly however, we previously showed that p53Δ31/Δ31 cells exhibited defects in DNA interstrand cross-link repair, a typical property of Fanconi anemia cells, that correlated with a subtle but significant decrease in expression for several genes of the Fanconi anemia DNA repair pathway, rather than the complete repression of a single gene in this pathway [25]. Thus, the Fanconi-like phenotype of p53Δ31/Δ31 cells most likely results from a decreased expression of not only Fanca, but also of additional p53-DREAM targets mutated in Fanconi anemia such as Fancb, Fancd2, Fanci, Brip1, Rad51, Palb2, Ube2t or Xrcc2, for which functional or putative [DREAM binding sites] were also found with our systematic approach.” We further discuss in the manuscript how this may also apply to telomere-, ribosome-, of microcephaly-related genes.

The analysis of brain specific targets and the link to BRD8 sits largely as an aside and the analysis of patient data from glioblastomas is underdeveloped as noted above.

As we previously mentioned, the revised manuscript includes the analysis of RNAseq datasets from human cortical neural progenitors infected by the Zika virus (ZIKV) or mock-infected, which significantly increases the relevance of our analysis of brain-specific targets. Furthermore, we improved Figure 4 to present more clearly the impact of BRD8 levels on the expression of genes transactivated by p53 or repressed by p53-DREAM.

The computational methods applied are robust, albeit predominantly coorelative, in terms of identifying regulation of potential causative target genes, validated across human and mouse cell lines, and this indicates a role of these genes in the relevant conditions. However, further validation through application in a bulk or single cell RNAseq patient cohort, or at least an in vivo model would strengthen these conclusions and complement the work presented here which is based on in vitro mouse and human cells. This is pertinent as this study improves upon previously published approaches by focusing on "clinically relevant target genes". Additionally, this would exhibit the potential applications of the findings presented.

We thank the referee for this comment. As mentioned above, in the revised manuscript we analyzed RNAseq data from hematopoietic stem cells of unirradiated WT or p53 KO mice, or irradiated WT mice, and from splenic cells of irradiated p53D24/- or p53+/- mice, and quantified the expression of a subset of blood-related candidate genes in the bone marrow cells of p53Δ31/Δ31 mice (prone to bone marrow failure) and WT mice (new Figure S8 and Table S19). For genes related to brain development, we included the analysis of RNAseq data from human cortical neural progenitors infected by the Zika virus (ZIKV) or mock-infected (Table S28). These RNAseq analyses were added as an additional screening criterion in our approach, which significantly increased the relevance of the target genes identified.

In terms of statistical analysis, the hypergeometric test should be applied to assess significant enrichment of genes for example with CDE/CHR regions within the previously identified lists.

In the revised manuscript, we precisely mapped the DREAM binding sites in 50 bp windows within regions bound by E2F4 and/or LIN9, an analysis included in new Figure 3a. We then compared the distribution of DREAM binding sites at the level of ChIP peaks compared to their distribution over the entire genome and found a > 1300-fold enrichment of these sites at ChIP peaks. This significant enrichment (f=3 10-239 in a hypergeometric test) is most likely underestimated because mouse-human DNA sequence conservations were not determined for putative DBS over the full genome. These new analyses clearly reinforce our previous conclusions.

Minor Comments<br /> References are required for the genes listed which play a role in the diseases of interest.

In the revised manuscript, references are provided for genes which play a role in the diseases of interest. Due to the large number of added references, these were included in a new supplementary table, Table S36.

This paper would benefit from the inclusion of summary schematics and tables throughout (rather than relying only on somewhat unwieldy heatmaps which show little other than all these genes are co-ordinately regulated), this could include summaries of the methods applied, gene or CDE/CHR inclusion criteria, and Venn diagrams indicating the subsets of final genes identified through this approach.

We thank the referee for this suggestion. In the revised manuscript we provide a Venn-like diagram of the different steps of our approach (new Figure 3c), as well as tables listing the genes retained after each step of the selection (new Tables S17 and S26) and these additions improve the clarity of our manuscript.

Reviewer #1 (Significance):

In its current form this is a very limited study that would require significant additional work to move conclusions beyond correlation and hypothesis generation.<br /> Overall, while limited largely to target prioritisation, this research nicely exemplifies how genes affected by the p53-DREAM pathway can be robustly identified, providing a potential resource for individuals working on this pathway or on abnormal haematopoiesis and brain abnormalities. These results are complementary to work previously published by Fischer et al, which has been referenced throughout the analysis (highlighting Target Gene Regulation Database p53 and DREAM target genes) and discussion.

This paper will be of interest to researchers of blood/neurological diseases who can assess if these genes are dysregulated in their datasets, or those investigating the p53-DREAM pathway. This work represents a useful resource detailing genes affected by this pathway in these disease settings, however researchers of the p53-DREAM pathway may find this paper useful when planning an approach to identify and prioritise genes of interest.

We thank the reviewer for considering that our study represents a useful resource for researchers working on the p53-DREAM pathway, abnormal haematopoiesis and brain abnormalities, because it was exactly the purpose of our work. As mentioned above, we think that a study bridging the gap between DREAM experts and bone marrow or microcephaly specialists should be particularly useful.

We also agree with the referee that our approach could be used to identify DREAM targets relevant to other disease settings, and we now mentioned this clearly in the revised manuscript.

While our results are complementary to work previously published by Fischer et al and included in the Target gene regulation database, in the revised manuscript we discuss the novelty of our results in more details, notably by performing additional analyses. For example, our method identified bipartite DREAM binding sites for 151 candidate DREAM targets (of which 56 genes were not previously mentioned by Fischer et al.) and we now provide a detailed mapping (using 50 bp windows) of the bipartite DREAM binding sites we identified relative to ChIP peaks for DREAM subunits, then performed a similar mapping of the E2F and CHR sites included in the Target gene regulation database. Our predicted DREAM binding sites coincided with ChIP peaks more frequently (Figure 3a) than the predicted E2F or CHR from the Target gene regulation database (Figure S11), which further indicates the usefulness of our study as a resource.

Reviewer #2 (Evidence, reproducibility and clarity):

The authors used various systems including Hoxa9-indubible BMCs, human and mouse cells, WT and p53 knockout MEF, glioblastoma cells to screen p53-DREAM targets and observed distinct finding for each system. Since different cell types have various p53 activation and p53 target genes expression, the authors might want to select proper cell type(s) to screen p53-DREAM target genes and design experiments to confirm that these genes are really p53-DREAM target genes.

We agree that additional data from relevant cells or tissues were required to strengthen our conclusions. As mentioned in response to referee #1, in the revised manuscript we evaluated the relevance of candidate target genes related to blood ontology terms by integrating an additional screening step in our method, corresponding to the analysis of RNAseq dataset GSE171697, with data from hematopoietic stem cells of unirradiated or irradiated WT mice and unirradiated p53 KO mice , as well as RNAseq dataset GSE204924, with data from splenic cells of irradiated p53D24/- or p53+/- mice. As for genes related to brain development, we included the analysis of RNAseq datasets GSE78711 and GSE80434 for validation, two datasets from human cortical neural progenitors infected by the Zika virus or mock-infected. Together, the 4 datasets provide evidence for a p53-dependent downregulation in blood- and brain- relevant settings (new Tables S19 and S28).

Importantly, in the revision we also compared our list of 151 genes appearing as the best p53-DREAM candidates with the results of Magès et al., who analyzed, in murine cells with a CRISPR-mediated KO of Lin37 (a subunit of DREAM), the transcriptomic changes that follow a reintroduction of Lin37. This comparison is detailed in the discussion section, with the new Figure S10 and Table S35. We mention: “Our list of 151 genes overlaps only partially with the list of candidate DREAM targets obtained with this approach, with 51/151 genes reported to be downregulated in Lin37-rescued cells [17]. To better evaluate the reasons for this partial overlap, we extracted the RNAseq data from Lin37 KO and Lin37-rescued cells and focused on the 151 genes in our list. For the 51 genes that Mages et al. reported as downregulated in Lin37-rescued cells, an average downregulation of 14.8-fold was observed (Figure S10, Table S35). Furthermore, when each gene was tested individually, a downregulation was observed in all cases, statistically significant for 47 genes, and with a P value between 0.05 and 0.08 for the remnant 4 genes (Table S35). By contrast, for the 100 genes not previously reported to be downregulated in Lin37-rescued cells, an average downregulation of 4.7-fold was observed (Figure S10, Table S35), and each gene appeared downregulated, but this downregulation was statistically significant for only 35/100 genes, and P values between 0.05 and 0.08 were found for 23/100 other genes (Table S35). These comparisons suggest that, for the additional 100 genes, a more subtle decrease in expression, together with experimental variations, might have prevented the report of their DREAM-mediated regulation in Lin37-rescued cells.”

This comparison provides additional evidence that the 151 candidate target genes we identified are bona fide DREAM targets.

Specific comments:<br /> The authors need to describe and define HSC and Diff in Figure 1.

This has been corrected in the revised manuscript. “HSC” was replaced by “Hematopoietic Stem / Progenitor cells (+OHT)” and “Diff” was replaced by “Differentiated cells (5 days – OHT).

Are Figure 1B and 1D list genes p53 targets in bone marrow cells?

In the revised manuscript, we now analyzed RNAseq data to address this point. The question refers to lists of telomere-related genes (Figure 1b in both versions of the manuscript) and Fanconi-related genes (Figure 1d in the previous version, now Figure S2a), but could also apply to other lists of genes related to blood ontology terms (Figures S3-S5 in the revised manuscript). As mentioned in response to referee #1, in the revised manuscript we integrated an additional screening step in our method, corresponding to the analysis of RNAseq datasets specific of blood cells. We analyzed dataset GSE171697, with RNAseq data from hematopoietic stem cells of unirradiated WT or p53 KO mice, or irradiated WT mice, as well as dataset GSE204924, with RNAseq data from splenic cells of irradiated p53D24/- or p53+/- mice. The latter dataset appeared interesting because p53D24 is a mouse model prone to bone marrow failure and the spleen is a hematopoietic organ in mice. Furthermore, we also verified that the expression of a subset of blood-related candidate genes was decreased in the bone marrow cells of p53Δ31/Δ31 mice (prone to bone marrow failure) compared to bone marrow cells from WT mice, a result presented in the new Figure S8.

Where is the detailed information for mouse and human cells in Figure 1 and Figure 2?

In the first draft of the manuscript, supplementary tables provided precise values for ChIP binding. In the revised manuscript, we also provide the precise values for gene expression after bone marrow cell differentiation, as well as p53 regulation scores from the Target gene regulation databases. This additional information is included in the new Tables S1, S5, S8, S11, S14, S20 and S23.

Are Figure 3B list genes also p53 target genes in other cell types such as bone marrow cells and glioblastoma?

For genes in the Figure 3B of the previous version of the manuscript (now Figure 2B in the revised version), we now provide evidence that the blood-related genes are less expressed in the bone marrow cells of p53Δ31/Δ31 mice (mice with increased p53 activity and prone to bone marrow failure) compared to bone marrow cells from WT mice. This result is presented in the new Figure S8. For the brain-related genes of the same Figure, evidence of their p53-mediated regulation is provided by the RNAseq datasets GSE78711 and GSE80434, from human cortical neural progenitors infected by the Zika virus or mock-infected (analyzed in the new Table S28). Evidence of that a decreased p53 activity in glioblastomas correlates with increased expression of the brain-related genes of the same Figure is provided in supplementary Table S37.

Does BRD8high has high p53 and p21?

We now clearly show, in both Figure 4a and Table S37, that glioblastoma cells with high BRD8 exhibit a decreased expression of CDKN1A/p21 and other genes known to be transactivated by p53 (BAX, GADD45A, MDM2, PLK3), consistent with the fact that BRD8 attenuates p53 activity.

Are genes listed in Figure 4B all p53 target genes? can some validation be done?

For genes in Figure 4B, in the revision we focused on the genes that appeared more relevant, i.e. the 77 genes mutated in diseases with microcephaly or cerebellar hypoplasia. All the genes in Figure 4B are repressed in neural progenitors upon infection by the Zika virus, a virus known to cause p53 activation in those cells. This is reported in the new Table S28.

Reviewer #2 (Significance):

This is a potentially interesting study. The major limitation is the absence of validation from the screening. This study would definitely benefit the research community as long as some of the key findings are validated.

We thank the referee for this comment. We hope the new evidence in this revision provide the validation requested by the referee.

Reviewer #3 (Evidence, reproducibility and clarity):

In their work submitted to Review Commons, Rakotopare et al. aim to identify p53-DREAM target genes associated with blood or brain abnormalities. To this end, they utilize published data generated with a cellular model that results in cell-cycle exit and differentiation of murine bone marrow progenitor cells upon inducible expression of Hoxa9. By analyzing this gene expression data set published by Muntean et al., they find that multiple of the 3631 genes which are downregulated more than 1.5-fold in differentiated BMCs are also mutated in several disorders connected to proliferation and differentiation defects during hematopoiesis and brain development. By screening ChIP-seq data sets available at ChIP-Atlas, they find that the promoters of many of these genes are bound by DREAM complex components, and most of them were identified as genes indirectly repressed by p53 before (Fischer et al. 2016, targetgenereg.org). They then use a computational approach to identify putative CDE/CHR DREAM-binding sites in the promoters of 372 genes associated with blood/brain abnormalities which are downregulated in differentiated BMCs and bound by DREAM components. Out of the 173 candidate genes, they select twelve to analyze whether mutation of the putative DREAM binding sites results in increased activity of the promoters in luciferase reporter assays. The authors conclude that their findings suggest a general role for the p53-DREAM pathway in regulating hematopoiesis and brain development.<br /> While the study supports a large body of publications proving that repression of cell cycle genes by the DREAM complex is crucial for cell cycle arrest and exit, it is noted that none of the main conclusions here are unexpected or particularly exciting. All the analyses are based on data sets that compare gene expression in highly proliferative cells with cells that underwent terminal cell cycle exit. Thus, a large portion of the genes that are downregulated in differentiated BMCs are cell cycle genes and well-established targets of DREAM and E2F:RB complexes. Furthermore, it is not surprising that some of these pro-proliferative genes are mutated in diseases connected to proliferation defects like anemias or microcephaly.

We agree with the referee that the DREAM complex is well known to regulate cell cycle genes – in fact, this is what we mention in the first sentence of our introduction in both versions of our manuscript. However, as we already pointed out in response to Referee #1, many scientists or clinicians specialized in bone marrow failure syndromes or microcephaly diseases are not familiarized with the p53-DREAM pathway, and we think our study will be particularly useful to them. Furthermore, our strategy relying on disease-based ontology terms rather than cell cycle regulation led to identify many DREAM targets that were not reported in previous studies, and our positional frequency matrices led to identify DREAM binding sites not predicted by previous approaches. As discussed below, our revised manuscript provides a more detailed comparison of our findings with those from previous studies.

Additionally, I am not very enthusiastic about this manuscript because of several major concerns:

1. The authors draw conclusions about the p53-DREAM pathway based on data that was generated in a cellular differentiation model without convincingly showing that p53 plays a central role in gene repression in this experimental setup.<br /> (A) Rakotopare et al. define p53-DREAM target genes based on RNA expression data from proliferating precursor cells and non-proliferating, differentiated BMCs (Muntean et al., 2010). This paper has not studied whether p53 gets activated in the particular experimental setup during Hox9a-induced BMC differentiation. On page 4 of their manuscript, the authors state: "Consistent with the fact that BMC differentiation strongly correlates with p53 activation..." without citing any literature or explaining why this is supposed to be a fact. Furthermore, they imply that cell cycle gene repression in this model system depends on p53 because mRNA expression of the p53 targets p21 and Mdm2 was found to be increased in the differentiated cells (Fig. 1A, 5-fold and 2-fold, respectively). However, defining a large set of "p53-DREAM target genes" based on the moderate increase in mRNA levels of two genes that are known to be activated by p53 without showing any evidence that p53 is even involved in this effect during BMC differentiation is not appropriate.

We agree that Muntean et al. did not study whether p53 gets activated when BMCs differentiate in the Hox9a-ER system. We previously mentioned: “We observed that p53 activation correlated with cell differentiation in this system, because genes known to be transactivated by p53 (e.g. Cdkn1a, Mdm2) were induced, whereas genes repressed by p53 (e.g. Rtel1, Fancd2) were downregulated after tamoxifen withdrawal (Figure 1a)”. We had provided examples for 2 genes transactivated and 2 genes repressed, but clearly mentioned that they were given as examples. In the revised manuscript, we provide additional evidence with a new supplementary Figure that includes changes in expression for 15 additional genes known to be transactivated by p53, and 5 additional genes known to be repressed by p53 (Figure S1). In total, we now correlate HSC differentiation with p53 activation based on the expression of 24 well-known p53-regulated genes, which we hope is more convincing.

In addition, we changed our phrasing and mention “Consistent with the notion that BMC differentiation strongly correlates with p53 activation in this system, 72 of these 76 genes have negative score(s) in the Target gene regulation (TGR) database”.

(B) Interestingly, p53 is among the genes that get repressed on mRNA level in differentiated BMCs (Fig. 1B; Trp53), and the authors also identify the DREAM components E2F4 and LIN9 as bound to the p53 promoter by screening ChIP-Atlas data (Fig. 1C). Given that p53 has never been described as a DREAM target, I find this rather surprising and it makes me wonder whether appropriate parameters were selected for analyzing the ChIP data, particularly since the authors do not provide binding data for sets of non-cell cycle genes as a negative control.

We retrieved ChIP data from the ChIP Atlas database without any specific parameters, thus in a completely unbiased manner. Importantly however, for reasons detailed in the manuscript, we clearly mentioned that total ChIP scores <979/4000 were considered too low to reflect significant DREAM binding. The ChIP score for Trp53 was 630, which rapidly led us to eliminate this gene from our screen.

This ChIP score criterion was already mentioned in the previous version of our manuscript, but we think the addition of a Venn-like diagram (Figure 3c) and summary tables (S17 and S26) in the revised manuscript will probably make it easier to understand.

(C) Finally, the authors utilize the targetgenereg.org database to show that many of the genes they describe as p53-repressed were already identified as p53 targets. This database (Fischer et al. 2016) was created by performing a meta-analysis integrating a plethora of RNA-seq and ChIP-seq datasets with the aim to identify whether a particular gene gets up- or downregulated by p53, shows cell-cycle-dependent expression, is a DREAM/MuvB or E2F:RB target, etc. For example, 57 datasets analyzing p53-dependent RNA expression in human and 15 datasets generated with mouse cells were included, and a positive or negative score shows in how many of these experiments the gene was found to be up (positive score) or downregulated (negative score). Combining a large number of datasets in such a study is very helpful to get an idea if a gene is indeed generally regulated by a transcription factor, or if it just showed up in a few experiments - either as a false positive or because the regulation depends on a particular biological setting. The authors find most of the genes they identify as repressed in differentiated BMCs also as downregulated by p53 in targetgenereg.org, however, it remains unclear what parameters they used to define a gene as p53-repressed. For example, in the caption of Fig. 1C, they state: "According to the Target gene regulation database, 72/76 genes are downregulated upon mouse and/or human p53 activation." The four exemptions are SLX1B (human score: 0, mouse score : na), PML (+41, +9), RAD50 (0, na), and TNKS2 (+17, +4). However, there are several other genes that do not appear to be generally repressed by p53, e.g. HMBOX1 (+4, -2); UPF1 (+1, -2), SMG6 (+18, -2), CTC1 (-5, +11), etc. Thus, without providing details regarding the parameters they use to define p53-target genes, such statements are rather misleading. An easy way to solve this problem would be to show the p53 scores in the tables together with the E2F4/LIN9 ChIP data.

All the genes mentioned as downregulated by p53 had a negative TGR score in human and/or mouse cells. In the revised manuscript, we mention clearly what a negative TGR score means, by stating: “Consistent with the notion that BMC differentiation strongly correlates with p53 activation in this system, 72 of these 76 genes have negative p53 expression score(s) in the Target gene regulation (TGR) database [23], which indicates that they were downregulated upon p53 activation in most experiments carried out in mouse and/or human cells (Figure 1b, Table S1).” We agree with the referee that adding precise TGR scores is informative. In the revised manuscript, we provide the TGR scores for all the genes analyzed, as part of the new supplementary Tables S1, S5, S8, S11, S14, S20 and S23, together with their expression levels in undifferentiated or differentiated cells (as requested by Referee #2). The ChIP data are provided in separate tables (Tables S2, S3, S6, S7, S9, S10, S12, S13, S15, S16, S21, S22, S24 and S25).

1. The authors define a large set of genes containing "CDE-CHR" promoter elements and thereby ignore how these elements are defined and what properties they have.<br /> (A) At the beginning of the introduction, the authors state: "The DREAM complex typically represses the transcription of genes whose promoter contain a bipartite CDE/CHR binding site, with a cell cycle-dependent element (CDE) bound by E2F4 or E2F5, and a cell cycle gene homology region (CHR) bound by LIN54, the DNA binding subunit of MuvB (Zwicker et al., 1995; Müller and Engeland, 2010)."<br /> This statement is incorrect. The authors ignore that the CDE/CHR tandem site is just one of four promoter elements that have been shown to recruit DREAM for the transcriptional repression of several hundred genes. It has been studied in detail that DREAM can bind to the following promoter sites:<br /> (I) CHR elements - bound by DREAM via LIN54; also bound by the activator MuvB complexes B-MYB-MuvB and FOXM1-MuvB which results in maximum gene expression in G2/M<br /> (II) CDE-CHR tandem elements - like (I) but binding of DREAM can be stabilized via E2F4/DP interacting with a truncated E2F binding site. Since CDE elements do not represent functional E2F sites, E2F:RB complexes do not bind.<br /> (III) E2F binding sites - bound by DREAM via E2F4/DP; also bound by E2F:RB complexes and activator E2Fs which results in maximum gene expression in G1/S<br /> (IV) E2F-CLE tandem elements - like (III) but binding of DREAM can be stabilized via LIN54 interacting with a non-canonical CHR-like element. Since CLE elements do not represent functional CHR sites, B-MYB-MuvB and FOXM1-MuvB do not bind.<br /> Thus, these promoter sites have different functions and can be clearly distinguished from each other based on their properties - a fact that is completely ignored by the authors. Since the authors do not differentiate between G1/S and G2/M expressed genes and (CDE)-CHR and E2F-(CLE) sites, they identify CDE-CHR elements in G1/S genes that are functional E2F-(CLE) sites. A good example of this is the Rad51ap1 gene (and also the Rad51 gene that the Toledo lab described before as a CDE-CHR gene (Jaber et al. 2016)): these genes get expressed in G1/S and the promoters contain highly conserved E2F sites (parts of which the authors define as CDEs), and CLEs (which the authors define as CHRs). Furthermore, E2F:RB complexes bind to the promoters. Again: even though (CDE)-CHR and E2F-(CLE) sites both bind DREAM, they are otherwise functionally different in their ability to recruit non-DREAM complexes.

We agree that in the previous version of our manuscript we should have presented in more details the different types of DREAM binding sites and have corrected this in the revised manuscript. We now mention in the introduction that “The DREAM complex was initially reported to repress the transcription of genes whose promoter sequences contain a bipartite binding motif called CDE/CHR [19,20] (or E2F/CHR [21]), with a GC-rich cell cycle dependent element (CDE) that may be bound by E2F4 or E2F5, and an AT-rich cell cycle gene homology region (CHR) that may be bound by LIN54, the DNA-binding subunit of MuvB [19,20]. Later studies indicated that DREAM may also bind promoters with a single E2F binding site, a single CHR element, or a bipartite E2F/CHR-like element (CLE), and concluded that E2F and CHR elements are required for the regulation of G1/S and G2/M cell cycle genes, respectively [14,22].”

We hope that the referee will agree with this complete yet concise way of presenting DREAM binding sites. Importantly, we agree that CDE/CHR and E2F/CLE are sites bound by different non-DREAM complexes, but both sites are bound by DREAM, so it makes perfect sense to use them together to define positional frequency matrices for DREAM binding predictions. We would also like to point out that terms used to define DREAM binding sites may vary in the literature. For example, to our knowledge Müller et al. were the first to propose a clear distinction between “CDE/CHR” and “E2F/CLE” sites (Müller et al. (2017) Oncotarget 8, 97737-97748), yet Müller recently co-authored a review in which these two distinct terms were not used, but were replaced by a single, apparently more generic term of “E2F/CHR” (Fischer et al., (2022) Trends Biochem. Sci. 47, 1009-1022). In the revised manuscript we now clearly mention that we designed our positional frequency matrices to search for “bipartite DREAM binding sites”, i.e. sites that might be referred to as CDE/CHR, E2F/CLE or E2F/CHR sites in various publications.

(B) The authors identified putative CDE-CHR in the promoters of genes by building two position weight matrices (PWMs) based on 10 or 22 "validated CDE-CHR elements". However, since they include several genes that are clearly expressed in G1/S and contain E2F-(CLE) sites (e.g. Mybl2/B-myb, Rad51, Fanca, Fen1), it is not surprising that they identify a lot of putative CDE-CHR sites in genes that do not contain such elements.

As discussed above, both CDE/CHR and E2F/CLE are bipartite DREAM binding sites, and we now clearly state that we used bipartite DREAM binding sites to generate our positional frequency matrices and predict DREAM binding.

(C) Finally, in the discussion, the authors state: "A recent update (2.0) of the Target gene regulation database of p53 and cell cycle genes (www.targetgenereg.org) was recently reported to include putative DREAM binding sites for human genes (Fischer et al., 2022). However, this update only suggests potential E2F or CHR binding sites independently, a feature of little help to identify CDE/CHR elements. For example, targetgenereg 2.0 suggests several potential E2F sites, but no CHR site close to the transcription start site of FANCD2, despite the fact that we previously identified a functionally CDE/CHR element near the transcription start site of this gene (Jaber et al., 2016)." This statement highlights again that the authors don't seem to be aware of what specific properties distinct DREAM binding sites have, and that analyzing promoters for CHR and E2F sites separately generates much more meaningful results than the approach they chose. Also, the FANCD2 promoter binds DREAM as well as E2F:RB complexes and contains a highly conserved E2F binding site - which Jaber et al. mutated together with a potential downstream CLE element and named it "CDE/CHR".

In the revised manuscript, we provide a more detailed comparison between the bipartite DREAM binding sites predicted with our positional frequency matrices for 151 genes and the separate E2F and CHR predicted sites reported in the Target gene regulation database for the same set of genes. We now mention: “The Target gene regulation (TGR) database of p53 and cell-cycle genes was reported to include putative DREAM binding sites for human genes, based on separate genome-wide searches for 7 bp-long E2F or 5 bp-long CHR motifs [23]. We analyzed the predictions of the TGR database for the 151 genes for which we had found putative bipartite DBS. A total of 342 E2F binding sites were reported at the promoters of these genes, but only 64 CHR motifs. The similarities between the predicted E2F or CHR sites from the TGR database and our predicted bipartite DBS appeared rather limited: only 14/342 E2F sites overlapped at least partially with the GC-rich motif of our bipartite DBS, while 27/64 CHR motifs from the TGR database exhibited a partial overlap with the AT-rich motif. Importantly, most E2F and CHR sites from the TGR database mapped close to E2F4 and LIN9 ChIP peaks, but only 16% of E2Fs (54/342), and 33% of CHRs (21/64) mapped precisely at the level of these peaks (Figure S11), compared to 55% (83/151) of our bipartite DBS (Figure 3a). Thus, at least for genes with bipartite DREAM binding sites, our method relying on PFM22 appeared to provide more reliable predictions of DREAM binding than the E2F and CHR sites reported separately in the TGR database. Importantly however, predictions of the TGR database may include genes regulated by a single E2F or a single CHR that would most likely remain undetected with PFM22, suggesting that both approaches provide complementary results.”

1. The experimental approach chosen to validate CDE-CHR elements in a set of twelve promoters by luciferase reporter assays is not adequate.<br /> (A) Since the authors introduce point mutations in putative CDE and CHR elements in parallel, it is impossible to identify functional CDE elements. As explained above, a functional CDE is not required for binding of MuvB complexes and gene repression, and mutating the CHR alone would already lead to a loss of DREAM binding and to de-repression of a promoter. Thus, without mutating both sites of CDE-CHR elements separately, it is impossible to provide evidence that a putative CDE is functional.<br /> (B) As the putative CDE-CHR elements identified by the authors with a computational approach can overlap with functional E2F-(CLE) elements, the authors inactivate such sites by introducing mutations which leads to loss of DREAM binding and upregulation of the promoters, however, because of the problems described above, this experimental approach in the best case identifies DREAM binding sites, but does not differentiate between (CDE)-CHR and E2F-(CLE) elements.

Yes, we agree with this comment. As discussed above, our goal was to identify DREAM-binding sites, not to differentiate between CDE/CHR and E2F/CLE elements. In other words, we wanted to identify genes regulated by p53 and DREAM, but not distinguish between genes regulated by p53, DREAM and E2F/Rb versus those regulated by p53, DREAM and BMyb-MuvB or FoxM1-MuvB.

(C) The authors analyze the activities of wild-type and mutant promoters in proliferating NIH3T3 cells. Since the mutated promoters showed increased activity (about 2-3 fold), which would be expected when binding of DREAM gets abolished, they conclude: "...these experiments indicated that we could identify functional CDE/CHRs for 12/12 tested genes." In addition to the problems described above, a slight upregulation of promoter activities caused by the introduction of multiple point mutations close to the TSS is not sufficient to verify these elements. The increase in activity could occur independent of DREAM-binding by unrelated mechanisms. The authors should at least analyze the activities of the promoters with and without induction of p53. A loss of p53-dependent repression of the mutated promoters would prove that the elements are essential for p53-dependent repression. Furthermore, there are several experimental approaches to analyze whether DREAM binds to the putative promoter element and whether the introduced mutations disrupt binding (ChIP, DNA affinity purification, etc.).

In the revised manuscript, we show that the promoters of 7 of the tested genes, when cloned in luciferase reporter plasmids and transfected into NIH3T3 cells, exhibited a significant (> 1.4 fold) repression upon p53 activation by cell treatment with Nutlin, the Mdm2 antagonist. For these promoters, we showed that the p53-dependent repression was abrogated by mutating the identified DREAM binding site, which provided direct evidence that our positional frequency matrices can identify functionally relevant DREAM binding sites essential for p53-mediated repression. These experiments were added in Figures 2e and 2i.

Furthermore, as previously mentioned in response to referee #1, in the revised manuscript we precisely mapped the predicted DREAM binding sites for 151 genes in 50 bp windows within regions bound by E2F4 and/or LIN9, an analysis included in new Figure 3a. The distribution of these peaks clearly indicates that most predicted DREAM binding sites map precisely within a 50 bp-window encompassing the ChIP peaks, which represents an enrichment of at least a 1300-fold compared to the rest of the genome. This mapping strongly suggests that our predicted DREAM binding sites are functionally relevant.

Importantly, as shown in the new Figure S11, we carried out a similar mapping of the predicted E2F and CHR sites reported in the Target gene regulation (TGR) database and found that our predicted DREAM binding sites co-mapped with E2F4/LIN9 ChIP peaks more frequently than the E2F and CHR sites of the TGR database, which supports the conclusion that our positional frequency matrices bring new and improved predictions for DREAM binding.

1. Taken together, while over-simplifying mechanisms of cell cycle gene regulation, the authors largely ignore recent findings and publications regarding gene regulation by p53, E2F:RB, and DREAM/MuvB complexes:<br /> (A) Publications that show how DREAM binds to (CDE)-CHR sites and that experimentally defined a consensus motif for CHR elements (e.g. PMID: 27465258, PMID: 25106871).<br /> (B) Publications that identify p53-DREAM target genes by activating p53 in cells with or without functional DREAM complex (e.g. PMID: 31667499, PMID: 31400114).<br /> (C) Identification and comparison of (CDE)-CHR and E2F-(CLE) DREAM binding sites that have distinct functions in the activation of cell-cycle expression in G1/S and G2/M (e.g. PMID: 29228647, PMID: 25106871).<br /> These findings have been summarized in several review articles (e.g. PMID: 29125603, PMID: 28799433, PMID: 35835684). All of them describe the mechanisms I have mentioned above in detail, and since Rakotopare et al. cite one of the papers (Engeland 2018), I wonder even more why they did not design their experiments based on current knowledge.

The points (A) and (C) of this comment were largely discussed in our response to points 2 and 3 of the same referee. Briefly, in the revised manuscript we clearly mention CDE/CHR, E2F/CLE and E2F/CHR sites, as well as the functional differences between E2F and CHR sites with regards to cell cycle regulation, but all these sites were considered together in our positional frequency matrices because our goal was to identify genes regulated by p53 and DREAM, not to distinguish between genes regulated by p53, DREAM and E2F/Rb versus those regulated by p53, DREAM and BMyb-MuvB or FoxM1-MuvB.

Regarding point (B) of this comment, in the revised manuscript we performed a detailed comparison of our results with those of Mages et al. who analyzed, in murine cells with a CRISPR-mediated KO of Lin37 (a subunit of DREAM), the transcriptomic changes that follow a reintroduction of Lin37 (Mages et al. (2017) elife 6, e26876). This comparison is detailed in the discussion section, with New Figure S10 and Table S35. As mentioned in response to referee #2, this comparison is perfectly consistent with DREAM regulating the 151 genes for which we identified DREAM binding sites.

Minor concerns:

1. The authors state: "Importantly however, the relative importance of the p53-p21-DREAM pathway (called below p53-DREAM) remains controversial, because multiple mechanisms were proposed to account for p53-mediated gene repression (Peuget and Selivanova, 2021)." Even though Peuget & Selivanova do not agree that genes get repressed in response to p53 activation exclusively by the p21-DREAM pathway, they do not question that this mechanism is essential for the p53-dependent repression of a core set of cell cycle genes. Since I am also not aware of any publications that challenge the importance of the p53-p21-DREAM pathway, I do not agree with this statement.

As the referee pointed out, in the first version of the manuscript we wrote that “the relative importance of the p53-p21-DREAM pathway (called below p53-DREAM) remains controversial, because multiple mechanisms were proposed to account for p53-mediated gene repression (Peuget and Selivanova, 2021)”. The term “relative” was crucial in this sentence, because we wanted to say that the relative proportion of genes regulated by DREAM remained controversial. It seems to us that the title of the review by Peuget & Selivanova (“p53-dependent repression: DREAM or reality?”) emphasizes this controversy. Nevertheless, in the revised manuscript, we now mention : “The relative importance of this pathway remains to be fully appreciated, because multiple mechanisms were proposed to account for p53-mediated gene repression [18]”. We hope the referee will find this phrasing more acceptable.

1. Some parts of the manuscript are tiring to read - for example, pages 6, 7, and 8 which contain long listings and numbers of genes that are downregulated in differentiated BMC, found to be mutated in various disorders, bind DREAM components, were identified as downregulated by p53, etc. The authors may consider combining central parts of these data in a table that they show in the main manuscript which would make it easier to digest the information and at the same time significantly shorten the manuscript.

We apologize if some parts of the article were tiring to read. We hope that the addition of Tables S17 and S26, as well as the Venn-like diagram in Figure 3c, will improve the reading of the manuscript.

1. The supplementary tables (S1-S26) are combined in one Excel file with multiple tabs. The authors should label the tabs accordingly to make it easier for the reader to find a particular table.

We labelled the Excel tabs in the revised manuscript, as suggested.

1. At the end of page 6, the authors show that 17 genes found to be downregulated in differentiated BMCs are mutated in multiple bone marrow disorders, however, since they don't include references, it remains unclear where these mutations were originally described.

In the revised manuscript, we included a supplementary table (Table S36) with appropriate references for blood and/or brain related phenotypes for the 106 genes associated with blood or brain abnormalities.

1. On page 9, the authors state: "As a prerequisite to luciferase assays, we first verified that the expression of these genes, as well as their p53-mediated repression, can be observedin mouse embryonic fibroblasts (MEFs), because luciferase assays rely on transfections into MEFs (Figure 3b)." The authors don't explain why luciferase assays rely on transfections into MEFs and based on the caption of Fig. 3C, the luciferase assays were not performed in MEFs, but in NIH3T3 cells: "WT or mutant luciferase reporter plasmids were transfected into NIH3T3 cells..."

According to the American Type Culture Collection (ATCC), the NIH3T3 cell line is a mouse embryonic fibroblastic (MEF) cell line, which explains why we had tested the expressions of candidate target genes in MEFs. However, as we now clearly mention in the manuscript, this cell line exhibits an attenuated p53 pathway, which improves cell survival after transfection but leads to decreased p53-mediated repression. These points are now clearly mentioned in the text and in a new supplemental Figure (Figure S9).

Reviewer #3 (Significance):

While the study supports a large body of publications proving that repression of cell cycle genes by the DREAM complex is crucial for cell cycle arrest and exit, it is noted that none of the main conclusions here are unexpected or particularly exciting. All the analyses are based on data sets that compare gene expression in highly proliferative cells with cells that underwent terminal cell cycle exit. Thus, a large portion of the genes that are downregulated in differentiated BMCs are cell cycle genes and well-established targets of DREAM and E2F:RB complexes. Furthermore, it is not surprising that some of these pro-proliferative genes are mutated in diseases connected to proliferation defects like anemias or microcephaly.

Again, we agree with the referee that the DREAM complex is well known to regulate cell cycle genes, but many scientists or clinicians specialized in bone marrow failure syndromes or microcephaly diseases are not familiarized with the p53-DREAM pathway, and we think our study will be particularly useful to them. As for DREAM specialists, our strategy relying on disease-based ontology terms rather than cell cycle regulation led to identify many DREAM targets that were not reported in previous studies, and our positional frequency matrices led to identify DREAM binding sites not predicted by previous approaches. We hope that, by considering all these points together, the referee will acknowledge that our study provides a valuable resource for different types of readerships.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In their work submitted to Review Commons, Rakotopare et al. aim to identify p53-DREAM target genes associated with blood or brain abnormalities. To this end, they utilize published data generated with a cellular model that results in cell-cycle exit and differentiation of murine bone marrow progenitor cells upon inducible expression of Hoxa9. By analyzing this gene expression data set published by Muntean et al., they find that multiple of the 3631 genes which are downregulated more than 1.5-fold in differentiated BMCs are also mutated in several disorders connected to proliferation and differentiation defects during hematopoiesis and brain development. By screening ChIP-seq data sets available at ChIP-Atlas, they find that the promoters of many of these genes are bound by DREAM complex components, and most of them were identified as genes indirectly repressed by p53 before (Fischer et al. 2016, targetgenereg.org). They then use a computational approach to identify putative CDE/CHR DREAM-binding sites in the promoters of 372 genes associated with blood/brain abnormalities which are downregulated in differentiated BMCs and bound by DREAM components. Out of the 173 candidate genes, they select twelve to analyze whether mutation of the putative DREAM binding sites results in increased activity of the promoters in luciferase reporter assays. The authors conclude that their findings suggest a general role for the p53-DREAM pathway in regulating hematopoiesis and brain development.

While the study supports a large body of publications proving that repression of cell cycle genes by the DREAM complex is crucial for cell cycle arrest and exit, it is noted that none of the main conclusions here are unexpected or particularly exciting. All the analyses are based on data sets that compare gene expression in highly proliferative cells with cells that underwent terminal cell cycle exit. Thus, a large portion of the genes that are downregulated in differentiated BMCs are cell cycle genes and well-established targets of DREAM and E2F:RB complexes. Furthermore, it is not surprising that some of these pro-proliferative genes are mutated in diseases connected to proliferation defects like anemias or microcephaly.

Additionally, I am not very enthusiastic about this manuscript because of several major concerns:

1. The authors draw conclusions about the p53-DREAM pathway based on data that was generated in a cellular differentiation model without convincingly showing that p53 plays a central role in gene repression in this experimental setup.

(A) Rakotopare et al. define p53-DREAM target genes based on RNA expression data from proliferating precursor cells and non-proliferating, differentiated BMCs (Muntean et al., 2010). This paper has not studied whether p53 gets activated in the particular experimental setup during Hox9a-induced BMC differentiation. On page 4 of their manuscript, the authors state: "Consistent with the fact that BMC differentiation strongly correlates with p53 activation..." without citing any literature or explaining why this is supposed to be a fact. Furthermore, they imply that cell cycle gene repression in this model system depends on p53 because mRNA expression of the p53 targets p21 and Mdm2 was found to be increased in the differentiated cells (Fig. 1A, 5-fold and 2-fold, respectively). However, defining a large set of "p53-DREAM target genes" based on the moderate increase in mRNA levels of two genes that are known to be activated by p53 without showing any evidence that p53 is even involved in this effect during BMC differentiation is not appropriate.

(B) Interestingly, p53 is among the genes that get repressed on mRNA level in differentiated BMCs (Fig. 1B; Trp53), and the authors also identify the DREAM components E2F4 and LIN9 as bound to the p53 promoter by screening ChIP-Atlas data (Fig. 1C). Given that p53 has never been described as a DREAM target, I find this rather surprising and it makes me wonder whether appropriate parameters were selected for analyzing the ChIP data, particularly since the authors do not provide binding data for sets of non-cell cycle genes as a negative control.

(C) Finally, the authors utilize the targetgenereg.org database to show that many of the genes they describe as p53-repressed were already identified as p53 targets. This database (Fischer et al. 2016) was created by performing a meta-analysis integrating a plethora of RNA-seq and ChIP-seq datasets with the aim to identify whether a particular gene gets up- or downregulated by p53, shows cell-cycle-dependent expression, is a DREAM/MuvB or E2F:RB target, etc. For example, 57 datasets analyzing p53-dependent RNA expression in human and 15 datasets generated with mouse cells were included, and a positive or negative score shows in how many of these experiments the gene was found to be up (positive score) or downregulated (negative score). Combining a large number of datasets in such a study is very helpful to get an idea if a gene is indeed generally regulated by a transcription factor, or if it just showed up in a few experiments - either as a false positive or because the regulation depends on a particular biological setting. The authors find most of the genes they identify as repressed in differentiated BMCs also as downregulated by p53 in targetgenereg.org, however, it remains unclear what parameters they used to define a gene as p53-repressed. For example, in the caption of Fig. 1C, they state: "According to the Target gene regulation database, 72/76 genes are downregulated upon mouse and/or human p53 activation." The four exemptions are SLX1B (human score: 0, mouse score : na), PML (+41, +9), RAD50 (0, na), and TNKS2 (+17, +4). However, there are several other genes that do not appear to be generally repressed by p53, e.g. HMBOX1 (+4, -2); UPF1 (+1, -2), SMG6 (+18, -2), CTC1 (-5, +11), etc. Thus, without providing details regarding the parameters they use to define p53-target genes, such statements are rather misleading. An easy way to solve this problem would be to show the p53 scores in the tables together with the E2F4/LIN9 ChIP data.<br /> 2. The authors define a large set of genes containing "CDE-CHR" promoter elements and thereby ignore how these elements are defined and what properties they have.

(A) At the beginning of the introduction, the authors state: "The DREAM complex typically represses the transcription of genes whose promoter contain a bipartite CDE/CHR binding site, with a cell cycle-dependent element (CDE) bound by E2F4 or E2F5, and a cell cycle gene homology region (CHR) bound by LIN54, the DNA binding subunit of MuvB (Zwicker et al., 1995; Müller and Engeland, 2010)."

This statement is incorrect. The authors ignore that the CDE/CHR tandem site is just one of four promoter elements that have been shown to recruit DREAM for the transcriptional repression of several hundred genes. It has been studied in detail that DREAM can bind to the following promoter sites:

(I) CHR elements - bound by DREAM via LIN54; also bound by the activator MuvB complexes B-MYB-MuvB and FOXM1-MuvB which results in maximum gene expression in G2/M

(II) CDE-CHR tandem elements - like (I) but binding of DREAM can be stabilized via E2F4/DP interacting with a truncated E2F binding site. Since CDE elements do not represent functional E2F sites, E2F:RB complexes do not bind.

(III) E2F binding sites - bound by DREAM via E2F4/DP; also bound by E2F:RB complexes and activator E2Fs which results in maximum gene expression in G1/S

(IV) E2F-CLE tandem elements - like (III) but binding of DREAM can be stabilized via LIN54 interacting with a non-canonical CHR-like element. Since CLE elements do not represent functional CHR sites, B-MYB-MuvB and FOXM1-MuvB do not bind.

Thus, these promoter sites have different functions and can be clearly distinguished from each other based on their properties - a fact that is completely ignored by the authors. Since the authors do not differentiate between G1/S and G2/M expressed genes and (CDE)-CHR and E2F-(CLE) sites, they identify CDE-CHR elements in G1/S genes that are functional E2F-(CLE) sites. A good example of this is the Rad51ap1 gene (and also the Rad51 gene that the Toledo lab described before as a CDE-CHR gene (Jaber et al. 2016)): these genes get expressed in G1/S and the promoters contain highly conserved E2F sites (parts of which the authors define as CDEs), and CLEs (which the authors define as CHRs). Furthermore, E2F:RB complexes bind to the promoters. Again: even though (CDE)-CHR and E2F-(CLE) sites both bind DREAM, they are otherwise functionally different in their ability to recruit non-DREAM complexes.

(B) The authors identified putative CDE-CHR in the promoters of genes by building two position weight matrices (PWMs) based on 10 or 22 "validated CDE-CHR elements". However, since they include several genes that are clearly expressed in G1/S and contain E2F-(CLE) sites (e.g. Mybl2/B-myb, Rad51, Fanca, Fen1), it is not surprising that they identify a lot of putative CDE-CHR sites in genes that do not contain such elements.

(C) Finally, in the discussion, the authors state: "A recent update (2.0) of the Target gene regulation database of p53 and cell cycle genes (www.targetgenereg.org) was recently reported to include putative DREAM binding sites for human genes (Fischer et al., 2022). However, this update only suggests potential E2F or CHR binding sites independently, a feature of little help to identify CDE/CHR elements. For example, targetgenereg 2.0 suggests several potential E2F sites, but no CHR site close to the transcription start site of FANCD2, despite the fact that we previously identified a functionally CDE/CHR element near the transcription start site of this gene (Jaber et al., 2016)." This statement highlights again that the authors don't seem to be aware of what specific properties distinct DREAM binding sites have, and that analyzing promoters for CHR and E2F sites separately generates much more meaningful results than the approach they chose. Also, the FANCD2 promoter binds DREAM as well as E2F:RB complexes and contains a highly conserved E2F binding site - which Jaber et al. mutated together with a potential downstream CLE element and named it "CDE/CHR".<br /> 3. The experimental approach chosen to validate CDE-CHR elements in a set of twelve promoters by luciferase reporter assays is not adequate.

(A) Since the authors introduce point mutations in putative CDE and CHR elements in parallel, it is impossible to identify functional CDE elements. As explained above, a functional CDE is not required for binding of MuvB complexes and gene repression, and mutating the CHR alone would already lead to a loss of DREAM binding and to de-repression of a promoter. Thus, without mutating both sites of CDE-CHR elements separately, it is impossible to provide evidence that a putative CDE is functional.

(B) As the putative CDE-CHR elements identified by the authors with a computational approach can overlap with functional E2F-(CLE) elements, the authors inactivate such sites by introducing mutations which leads to loss of DREAM binding and upregulation of the promoters, however, because of the problems described above, this experimental approach in the best case identifies DREAM binding sites, but does not differentiate between (CDE)-CHR and E2F-(CLE) elements.

(C) The authors analyze the activities of wild-type and mutant promoters in proliferating NIH3T3 cells. Since the mutated promoters showed increased activity (about 2-3 fold), which would be expected when binding of DREAM gets abolished, they conclude: "...these experiments indicated that we could identify functional CDE/CHRs for 12/12 tested genes." In addition to the problems described above, a slight upregulation of promoter activities caused by the introduction of multiple point mutations close to the TSS is not sufficient to verify these elements. The increase in activity could occur independent of DREAM-binding by unrelated mechanisms. The authors should at least analyze the activities of the promoters with and without induction of p53. A loss of p53-dependent repression of the mutated promoters would prove that the elements are essential for p53-dependent repression. Furthermore, there are several experimental approaches to analyze whether DREAM binds to the putative promoter element and whether the introduced mutations disrupt binding (ChIP, DNA affinity purification, etc.).<br /> 4. Taken together, while over-simplifying mechanisms of cell cycle gene regulation, the authors largely ignore recent findings and publications regarding gene regulation by p53, E2F:RB, and DREAM/MuvB complexes:

(A) Publications that show how DREAM binds to (CDE)-CHR sites and that experimentally defined a consensus motif for CHR elements (e.g. PMID: 27465258, PMID: 25106871).

(B) Publications that identify p53-DREAM target genes by activating p53 in cells with or without functional DREAM complex (e.g. PMID: 31667499, PMID: 31400114).

(C) Identification and comparison of (CDE)-CHR and E2F-(CLE) DREAM binding sites that have distinct functions in the activation of cell-cycle expression in G1/S and G2/M (e.g. PMID: 29228647, PMID: 25106871).

These findings have been summarized in several review articles (e.g. PMID: 29125603, PMID: 28799433, PMID: 35835684). All of them describe the mechanisms I have mentioned above in detail, and since Rakotopare et al. cite one of the papers (Engeland 2018), I wonder even more why they did not design their experiments based on current knowledge.

Minor concerns:

1. The authors state: "Importantly however, the relative importance of the p53-p21-DREAM pathway (called below p53-DREAM) remains controversial, because multiple mechanisms were proposed to account for p53-mediated gene repression (Peuget and Selivanova, 2021)." Even though Peuget & Selivanova do not agree that genes get repressed in response to p53 activation exclusively by the p21-DREAM pathway, they do not question that this mechanism is essential for the p53-dependent repression of a core set of cell cycle genes. Since I am also not aware of any publications that challenge the importance of the p53-p21-DREAM pathway, I do not agree with this statement.
2. Some parts of the manuscript are tiring to read - for example, pages 6, 7, and 8 which contain long listings and numbers of genes that are downregulated in differentiated BMC, found to be mutated in various disorders, bind DREAM components, were identified as downregulated by p53, etc. The authors may consider combining central parts of these data in a table that they show in the main manuscript which would make it easier to digest the information and at the same time significantly shorten the manuscript.
3. The supplementary tables (S1-S26) are combined in one Excel file with multiple tabs. The authors should label the tabs accordingly to make it easier for the reader to find a particular table.
4. At the end of page 6, the authors show that 17 genes found to be downregulated in differentiated BMCs are mutated in multiple bone marrow disorders, however, since they don't include references, it remains unclear where these mutations were originally described.
5. On page 9, the authors state: "As a prerequisite to luciferase assays, we first verified that the expression of these genes, as well as their p53-mediated repression, can be observed<br /> in mouse embryonic fibroblasts (MEFs), because luciferase assays rely on transfections into MEFs (Figure 3b)." The authors don't explain why luciferase assays rely on transfections into MEFs and based on the caption of Fig. 3C, the luciferase assays were not performed in MEFs, but in NIH3T3 cells: "WT or mutant luciferase reporter plasmids were transfected into NIH3T3 cells..."

Significance

While the study supports a large body of publications proving that repression of cell cycle genes by the DREAM complex is crucial for cell cycle arrest and exit, it is noted that none of the main conclusions here are unexpected or particularly exciting. All the analyses are based on data sets that compare gene expression in highly proliferative cells with cells that underwent terminal cell cycle exit. Thus, a large portion of the genes that are downregulated in differentiated BMCs are cell cycle genes and well-established targets of DREAM and E2F:RB complexes. Furthermore, it is not surprising that some of these pro-proliferative genes are mutated in diseases connected to proliferation defects like anemias or microcephaly.

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 used various systems including Hoxa9-indubible BMCs, human and mouse cells, WT and p53 knockout MEF, glioblastoma cells to screen p53-DREAM targets and observed distinct finding for each system. Since different cell types have various p53 activation and p53 target genes expression, the authors might want to select proper cell type(s) to screen p53-DREAM target genes and design experiments to confirm that these genes are really p53-DREAM target genes.

Specific comments:

The authors need to describe and define HSC and Diff in Figure 1.

Are Figure 1B and 1D list genes p53 targets in bone marrow cells?

Where is the detailed information for mouse and human cells in Figure 1 and Figure 2?

Are Figure 3B list genes also p53 target genes in other cell types such as bone marrow cells and glioblastoma?

Does BRD8high has high p53 and p21?

Are genes listed in Figure 4B all p53 target genes? can some validation be done?

Significance

This is a potentially interesting study. The major limitation is the absence of validation from the screening. This study would definitely benefit the research community as long as some of the key findings are validated.

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 paper the authors describe a data driven approach to identify and prioritise p53-DREAM targets whose repression might contribute to abnormal haematopoiesis and brain abnormalities observed in p53-CTD deleted mice. The premise is that in these mice, (where they have previously demonstrated p53 to be hyperactive in at least a subset of tissues), that the p53-p21-E2F/DREAM axis is at least in part responsible for observed phenotypes due to the repression of E2F and CDE/CHE element containing genes. Their approach to home in on relevant genes is based on transcriptomic gene ontology analysis of genes repressed in these disease settings where they primarily use publicly available data from HOXA9-ER regulated model of HSC expansion wherein they observe increases on p53-p21 expression upon differentiation where they demonstrate that p53-p21 DREAM target genes are suppressed as we would expect in this scenario where p53-p21 is activating withdrawal from cell cycle. They then spend a lot of effort analysing this datasets combining "gene-ontology", "disease phenotype" and "meta-ChIP-seq" analysis of public data to support the observation that mutations of genes suppressed in this manner are disproportionately linked to heritable haematopoetic and brain disorders. While these results are interesting in terms of framing a hypothesis about how mutations in p53-p21-DREAM regulated targets contribute to such conditions, they are to be expected given the now very well described impact of p53-p21 on both E2F4/DREAM targets. The natural progression of this work would be to go on to show this occurs in relevant cells or tissues derived from the p53-CTD mice as well as look at modulating target genes to understand underlying mechanisms and consequences.<br /> Rather than this, they focus on validating that a sub-set of these targets are indeed suppressed by specific p53 activation by MDM2 inhibitor Nutlin-3A in MEFs by qPCR and that mutation of predicted CDE CHR elements in luciferase constructs leads to increase luciferase activity. While these findings support their predictions, the results are entirely expected based on what is known about such targets and demonstrating that this occurs in MEFs does not closely relate to haematopoietic and brain cells they suggest this regulation is important. In fact, in the discussion, the authors comment on the importance of cell type context specificity in terms of discordance between predictions of TF binding sites and public datasets.<br /> Finally, they try and contextualise effects in glioblastoma data by correlating target gene expression with levels of BRD8 since it has recently been shown to attenuate p53 function in glioblastoma and show that some of the brain disease associated genes are expressed at higher levels in BRD8 high patient samples. It seems strange here that they do not also look at expression of p21 or other p53 targets that would help ascertain if p53 activity is indeed suppressed. Moreover, much more elegant methods for predicting transcription factor activity could be applied to this data.

Major Comments

The major result of this paper as it stands is the prioritisation of candidate genes in the p53-DREAM pathway involved in these conditions, and their refined approach used to identify and prioritise these genes and is such more of a starting point for further investigation. They fall short of demonstrating the relevance of their predictions physiologically in tissues from the mice and do not demonstrate functional importance of regulation of targets they put forward. Given that these genes will be co-ordinately regulated, without a mechanistic experiment in physiologically relevant model it is impossible to infer causality. For example, depleting individual targets in the HOXA9 model and evaluating impact on survival, proliferation and differentiation may be a (relatively) simple way to explore this, perhaps comparing to effects of p53 activating agents such as Nutlin-3A. Of note the authors (Jaber 2016 PMID: 27033104) and several other groups had (Fischer 2014 PMID: 25486564 McDade 2014 PMID: 24823795) previously demonstrated the link between p53-p21 and suppression of DNA-repair/Damage related genes (as is also observed here in particular FA-related genes that they discuss briefly here. I would have thought that this would be an obvious starting point for some mechanistic experiments and in fact I note this has been demonstrated before (Li et al 2018 PMID: 29307578)<br /> The analysis of brain specific targets and the link to BRD8 sits largely as an aside and the analysis of patient data from glioblastomas is underdeveloped as noted above.<br /> The computational methods applied are robust, albeit predominantly coorelative, in terms of identifying regulation of potential causative target genes, validated across human and mouse cell lines, and this indicates a role of these genes in the relevant conditions. However, further validation through application in a bulk or single cell RNAseq patient cohort, or at least an in vivo model would strengthen these conclusions and complement the work presented here which is based on in vitro mouse and human cells. This is pertinent as this study improves upon previously published approaches by focusing on "clinically relevant target genes". Additionally, this would exhibit the potential applications of the findings presented.<br /> In terms of statistical analysis, the hypergeometric test should be applied to assess significant enrichment of genes for example with CDE/CHR regions within the previously identified lists.

Minor Comments

References are required for the genes listed which play a role in the diseases of interest. This paper would benefit from the inclusion of summary schematics and tables throughout (rather than relying only on somewhat unwieldy heatmaps which show little other than all these genes are co-ordinately regulated), this could include summaries of the methods applied, gene or CDE/CHR inclusion criteria, and Venn diagrams indicating the subsets of final genes identified through this approach.

Significance

In its current form this is a very limited study that would require significant additional work to move conclusions beyond correlation and hypothesis generation.

Overall, while limited largely to target prioritisation, this research nicely exemplifies how genes affected by the p53-DREAM pathway can be robustly identified, providing a potential resource for individuals working on this pathway or on abnormal haematopoiesis and brain abnormalities. These results are complementary to work previously published by Fischer et al, which has been referenced throughout the analysis (highlighting Target Gene Regulation Database p53 and DREAM target genes) and discussion.

This paper will be of interest to researchers of blood/neurological diseases who can assess if these genes are dysregulated in their datasets, or those investigating the p53-DREAM pathway. This work represents a useful resource detailing genes affected by this pathway in these disease settings, however researchers of the p53-DREAM pathway may find this paper useful when planning an approach to identify and prioritise genes of interest.

My expertise is in the field of transcription factor and p53 family biology in cancer and disease. Our group utilises functional genomics and computational approaches to harness this information to identify causal regulators of downstream effects or indeed novel ways to exploit p53 family

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 referees for their interest, comments and advice on our manuscprit. The reply to the reviewers follows the revision plan proposed Review Commons.

1. Description of the planned revisions

R#1 major comments :

R#1 raised three major points relative to quantitative data. For the two first, we have optimized our quantification methods, as explained in the next section, which clearly improved the results, but some data still have to be re-analyzed for figures 5,6,7.

For the third one, here is the comment and our proposed additional experiments to answer it :

c) Quantifications of lateral fraction Col IV in mosaic experiments do not support decreased lateral secretion in Rab8 OE (3G) or Dys- (S5C), which are central tenets of the study. “

We have endeavoured to detect such differences in Dys mutants and Rab8 OE and do not see any possible improvement in the quantification method and therefore propose, instead, additional experiments.

With respect to Rab8 OE, we suspect that this gain of function is not sufficiently effective under the specific conditions of the experimental setup described in Figure 3, as its effect appears to be more subtle than that of Rab10 OE in Figure 2. We therefore propose to repeat this experiment on a sensitised background in which Rab10 function is partially affected. Unpublished data indicate that this downregulation of Rab10 is not sufficient to induce significant differences in this experimental setup. However, based on the genetic interactions described in the other figures, an additive/synergistic effect between rab8 OE and Rab10 KD can be expected, which would allow to confirm the involvement of Rab8 in basal secretion.

With regard to Dys mutant, we considered another possible explanation for this observation, namely that DAPC could affect the secretion of other BM proteins but not that of collagen IV. It should be noted that Perlecan and LamininA, which are also found in BM fibrils, are ligands for Dystroglycan, which is not the case for collagen IV. Unfortunately, there is no existing transgene with a UAS promoter and a tagged version of these proteins that would allow this hypothesis to be tested within a reasonable timeframe using the same method as that described in Fig.3. Therefore, we propose an alternative approach that would determine whether the secretion of endogenous laminin and/or perlecan is affected by Dystroglycan overexpression (i.e. secreted more laterally) and test whether this effect is Dys-dependent. It should be noted that this hypothesis would be fully consistent with all our data and in particular with that shown in Fig. 5.

R#2 major comments

From the data presented in Figure S1B, the authors state that the basement membrane mislocalization observed in Rab8/10KD has no major impact on polarity maintenance. They based this statement only on the localization of the apical marker aPKC. Although the aPKC data are convincing, it would be more compelling if the authors observe the distribution of other polarity proteins such as Dlg, E-Cadherin, and armadillo to better assess if the overall epithelial polarity is maintained in this condition.

We will complete fig S1 and perform ECad and Dlg staining to provide a better description of apical-basal polarity in the different Rab knock-down conditions.

R#3 comments

Results Figure 4

The authors suggest basal Rab10 expression domain near the Golgi exit point. Can the authors use a Trans-Golgi marker in order to confirm this statement other than the references stated?

Such a staining will be included in the final version with for instance golgin245 staining.

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

R#1 major comments :

a) Rab8 KD does not significantly increase apical fraction of Collagen IV with respect to control (Fig. 1H). The image in 1C clearly shows that Col IV is present apically, something that has been shown by others and that never occurs in the wild type. Failure of the quantifying method to detect a difference can only mean the quantifying method is not adequate. A 10% average in the control when it's clear that no Col IV at all is found apically in the wild type suggests that the authors are quantifying background signal that they should not be acquiring, and, if acquired, they should be subtracting Rab8/Rab10 double knock down is said to show a synergistic effect, when an additive effect would be more consistent with alternative routes. Other problematic deductions drawn from apical fraction quantifications are found in Fig. 5J (Dys- enhancing Rab8 KD but not Rab10 KD) and Fig. 7D (Exo70- enhancing Rab10 KD but not Rab8 KD)

We agree that this quantification was not optimal. We improved it by quantifying a narrower and more precise region for each domain. The new results are shown in Figure 1H. This improvement reduces the apical signal in the control from 10% to 6% and allows us to detect a significant increase between the control and Rab8 KD, thus resolving the problem raised. After verification, we did not subtract the background because there was no electronic background in our images (i.e. black is really black and equal to zero). Thus, the remaining signal is the true cytoplasmic GFP signal and it may not be appropriate to subtract it. Other data (fig 5J and 7D, now named fig 5L and 7H) were also re-analyzed with no major change.

b) Similar to apical fraction, measurements of planar polarization (trailing/lateral ratio) show average ratios near 1 for Dg, Rab10 and Dys, which is striking given that the localization of these proteins is so clearly polarized. Ratios lower than 1, which are reported for many individual cells in these graphs, should mean reversed polarity. In light of this, I would not be too confident on the effects reported in 5O-Q. In fact, on two occasions, the authors obtain significant differences in these planar polarization measurements that they themselves disregard: Fig. 6J (Rab10 in Exo70-) and Fig. 7I (Dys in Rab8 KD).

We agree that this quantification could be improved. Our initial quantification of the planar polarised proteins, Rab10 and Dys, found at the trailing edge, was confounded by their lateral spread. We have now reported with only the front half of the lateral side. By doing this in Figure 5, we increase the ratio in the control conditions, with almost no points below the value of 1, while the conditions in which polarity is visually affected are unchanged and still close to 1. We did not have time to re-analyze all the data (Figures 6 and 7), but will do so in the final revised version.

R#1 minor comments :

All the minor points raised by R#1 have been addressed by changes in the main text and/or the figures with the exception of the following :

• Fig. S1B does not seem to make a significant point in the context of this study.

Although we understand this comment, we followed suggestion of R#2 who asked in its major comments for more details with other cell polarity markers. These data are not yet included but will be generated for the fully revised version.

• I suggest drawing a summary scheme to aid readers better assess interpretations alternative to the ones given in the text.

While we will be happy to provide such a scheme in the final version, we prefer to wait for the results of the proposed complementary experiments to be as accurate as possible.

R#2 major comments

In the text for Figure 1G-H (page 4), the authors stated that the basal secretion was not restored in Rab8, 10, and 11 triple KD, in our opinion, it is unclear how the authors came to this strong conclusion from the presented data. It would be good if the authors explicitly explain how they come to this conclusion. Is it only based on the weak Coll-IV-GFP signal in the Rab8, 10, and 11 triple KD data compare to the control? If so, the authors should statistically quantify the difference with the control. In Figure 1H, no statistical analysis is provided between the control and triple KD conditions.

We agree that it was not entirely appropriate to give such conclusions on the basis of the quantifications available. A new graph showing basal fluorescence intensity (new Figure 1H) (and not just the ratio of apical to apical plus basal as in Figure 1I) has been added to better support the text. A relevant statistical comparison has been added to Figure 1H (old Figure 1I). We apologize for this oversight.

R#2 minor comments :

We took in account all these comments and changed accordingly the text and the figures

R#3 comments :

Results figure 1

The authors use RNAi lines to arrive at their conclusions, however, the extent of inhibition of gene expression achieved by the RNAi, has not been justified. Also observations from only one RNAi stock may not be completely conclusive:

i) Efficiency of RNAi has not been tested or shown. No supporting data. Rab10-RNAi stock is 26289 BDSC which is in Valium10, which is a weak RNAi line and needs a Dicer.

ii) Can same observations be made using classic alleles or generate somatic clones on follicular epithelial cells?<br />

R#3 raised several questions regarding the efficiency of RNAi, the use of different lines and/or the use of classical mutants as an alternative method.

For Rab10, we tested three different lines with similar results as shown now in Figure S1A-B. These data are also consistent with those obtained by overexpression of a dominant-negative form of Rab10 (Lerner et al, 2013). Unfortunately, Rab10 is located extremely close to the X chromosome centromere and is even more proximal than the FRT transgenes. It is therefore impossible to generate somatic mutant clones.

Regarding Rab8, it is already published that Rab8 RNAi, expression of a dominant-negative form of Rab8 and Rab8 mutant cells obtained by somatic clones give similar defects (Devergne et al, 2017). The text has been modified to better illustrate the available data validating our approach.

In addition, mutant clones would not allow analysis of genetic interactions in complex genetic contexts such as double and triple KDs. Similarly, the choice of the Rab10 line was motivated by the ease of obtaining the appropriate genetic combination according to their genomic location.

iii) Intensity of Collagen IV in the basement membrane in Rab11 knock-down mutants seems to be significantly low as compared to the Rab8 and Rab10 knock downs in supplementary Fig 1B. Are the authors very sure that Rab11 has no functions in basement membrane basal organization?

Good catch! Indeed, Rab11 RNAi significantly reduces basal secretion as now shown on fig 1H. Rab11 has pleiotropic functions in epithelial cells notably for their polarity (Choubey and Roy, 2017, Fletcher et al, 2012…) and, accordingly aPKC is partially disrupted in Rab11 RNAi conditions (Fig S1). Thus, the reason for such a decrease is unclear and could be an indirect consequence of an overall abnormal epithelial structure. Thus, we now report this observation but have not taken its interpretation too far.

iv) Authors need to show where and how fluorescence intensities have been measured.

Magenta rectangles with dashed lines on figure 1A illustrate the ROIs used for this analysis and more details have been added in the ‘experimental procedures’ section.

Results Figures 2 and 3:

Texts and figures have been modified as suggested.

Results Figure 4 :

The authors suggest a UAS-Rab10-RFP transgene show same results as endogenous Rab10-YFP as compared to spatial expression pattern. This is worrisome as expression of full length functional gene tagged with a fluorophore may be an overexpression. A control experiment would be helpful in suggesting/comparing with the Rab10 OE phenotype and that will be more convincing.

We are not sure that we fully understand the reviewer's comment. However, we initially compared endogenous Rab10 and UAS-RAB10 at 25°C, a temperature at which the latter has no visible impact on BM structure (Cerqueira-Campos et al, 2020). Furthermore, even when higher expression was induced (by increasing the temperature and therefore Gal4 activity) and this had an impact on BM structure, this did not change the subcellular localization of Rab10, i.e. it was still planarly polarized, as shown in Fig 5S. The text has been modified to emphasize this point.

Result Figure 5

The authors may provide a Rab10 expression profile in DAPC null or KD mutants which would make their claims more comprehensive.

Data showing Rab10 localization in Dys mutant cells were already shown on Figure S5A-B. Of notice, we also tried similar experiments using RAB10 knock-in line. However, for unexpected reason, having one copy of the chromosome with YFP insertion in Rab10 strongly enhanced DAPC mutant phenotypes in terms of F-actin orientation and follicle elongation ((as described in Cerqueira-Campos et al, 2020). We therefore considered these data as inappropriate.

General comment

In general some immunostainings should be carried out if not in all at least in some experiments with some cell domain specific markers, more specifically PCP markers such as Flamingo/Vangl and basolateral markers such as Lgl/Dlg. This makes the positions specific claims of the authors more valid in the eyes of the reader.

We agree that this may help the reader but the pcp markers mentioned are not expressed in this tissue. However, the tissue planar orientation is now systematically indicated and consistent in all figures. We did not generally perform immunostaining for lateral markers but routinely included F-actin staining to detect cellular cortex. Our quantifications or cortical segmentations were based on the cell outline provided by this stain. On the basis of this staining, the outline of the cells was added on certain figures to facilitate understanding of the images.

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

R#3 comments :

Result Figure 6 :

Exo 70 is a versatile molecule and Rho kinases such as Cdc42 can direct Polarised exocytosis through interaction of Rab effectors with Exo 70. Have the authors considered this?

We agree that it is an interesting prospect, but we consider it as beyond the scope of this article.

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 the present work the authors have elucidated a novel mechanistic model of basement membrane morphogenesis using Drosophila ovarian follicle cells as a model. The authors have employed extensive quantification approaches to justify the spatio-temporal expression of the molecules under study such as Collagen IV, Rab8, Rab10, DAPC, etc. The authors suggest distinct exit domains of BM protein Collagen IV facilitated by Rab8 and Rab10 via distinct routes as they interact with each other. The authors further show that DAPC plays an essential role in Rab10 mediated baso-lateral fibrillar BM synthesis whereas Rab8 functions are more Exocyst (Exo-70 dependent).

Major comments:

Result 1:

The authors use RNAi lines to arrive at their conclusions, however, the extent of inhibition of gene expression achieved by the RNAi, has not been justified. Also observations from only one RNAi stock may not be completely conclusive:

1. Efficiency of RNAi has not been tested or shown. No supporting data.<br /> Rab10-RNAi stock is 26289 BDSC which is in Valium10, which is a weak RNAi line and needs a Dicer.
2. Can same observations be made using classic alleles or generate somatic clones on follicular epithelial cells?
3. Intensity of Collagen IV in the basement membrane in Rab11 knock-down mutants seems to be significantly low as compared to the Rab8 and Rab10 knock downs in supplementary Fig 1B. Are the authors very sure that Rab11 has no functions in basement membrane basal organization?
4. Authors need to show where and how fluorescence intensities have been measured.

Result 2:

Confusing diagram. The authors should clarify whether the BM fibrils indicate lateral or planar BM components which they show to be more prominently expressed in Rab10 over-expression mutants.

A short note or an accompanying explanatory diagram on the source of the BM fibrils in the cellular context should make things less confusing.

FF calculation is an ingenious way of trying to look into functions.

The term Opposite effects/functions may be reconsidered as Rab8 and Rab10 compete with each other to deposit Collagen at spatially distinct domains. Opposite functions may give an impression that Rab8 actually represses Rab10 activity or vice versa, which may not be the case here.

Result 3:

Why was anti-GFP Ab detected with Cy3-Cy5 secondary Ab. GFP itself is green so why detect it with a Red secondary? Logic? How clone Collagen GFP and ECM collagen GFP was differentiated? Please justify

A panel with a dotted line joining the peripheral or lateral Collagen as shown in panels D' E' of Fig 3 would support the cartoon provided and link the cartoon to the actual microscopic images.

Result 4:

The authors suggest a UAS-Rab10-RFP transgene show same results as endogenous Rab10-YFP as compared to spatial expression pattern. This is worrisome as expression of full length functional gene tagged with a fluorophore may be an overexpression. A control experiment would be helpful in suggesting/comparing with the Rab10 OE phenotype and that will be more convincing.

When the authors mention back of cells, where do the authors exactly mean? A cartoon of "the back of follicle cells", wrt the entire ovarian follicle would be helpful.

The authors suggest basal Rab10 expression domain near the Golgi exit point. Can the authors use a Trans-Golgi marker in order to confirm this statement other than the references stated?

Result 5:

The authors may provide a Rab10 expression profile in DAPC null or KD mutants which would make their claims more comprehensive.

Result 6:

Exo 70 is a versatile molecule and Rho kinases such as Cdc42 can direct Polarised exocytosis through interaction of Rab effectors with Exo 70. Have the authors considered this?

In general some immunostainings should be carried out if not in all at least in some experiments with some cell domain specific markers, more specifically PCP markers such as Flamingo/Vangl and basolateral markers such as Lgl/Dlg. This makes the positions specific claims of the authors more valid in the eyes of the reader.

Significance

The findings impinge on a critical cellular process of Rab protein interactions in the genesis of the basement membrane which is of potential interest.

This falls under basic research. Since Rab molecules have emerged as molecules governing membrane morphogenesis, Cell and Molecular Biologists as well as a wide audience including clinicians will be interested on this.

Our group works on the roles of Rab11 in membrane morphogenesis in Drosophila model and we are now trying to venture with Rab11 in mammalian wound healing.

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 manuscript, Dennis et al. identify different secretory routes and cell exit sites involved in basement membrane secretion and diversification in epithelial cells. Using the follicular epithelium of the Drosophila ovary as their model system coupled with genetics, imaging, and image analysis approaches, they show that two previously identified RabGTPases, Rab8 and Rab10, work in parallel routes for basement membrane secretion. These two small GTPases work in a partially redundant manner, where Rab8 promotes basal secretion leading to a homogenous basement membrane, while Rab10 promotes lateral and planer-polarized secretion, leading to the formation of fibrils. The authors also show that Rab10 and the dystrophin-associated protein act together to regulate lateral secretion, and dystrophin (Dys) is necessary for dystroglycan (Dg) to recruit Rab10. Furthermore, DAPC is shown to be essential for fibril formation and is sufficient to reorient Collagen IV to the Rab10-dependent secretory route. Dys was also shown to interact directly with exocyst subunit Exo70. Using overexpression and loss of function approaches the authors claim that Exo70 limits the planer polarization of Dys, and as a result, Rab10, hence limiting basement membrane fibril formation. Finally, the authors state that the Exocyst (Exo70) is also required for the Rab8-dependent basement membrane route. Overall, the data described in this manuscript are convincing and the authors' claims are supported by the presented data. We have mainly minor comments and only a few major comments that need to be addressed.

Major Comments:

• In the text for Figure 1G-H (page 4), the authors stated that the basal secretion was not restored in Rab8, 10, and 11 triple KD, in our opinion, it is unclear how the authors came to this strong conclusion from the presented data. It would be good if the authors explicitly explain how they come to this conclusion. Is it only based on the weak Coll-IV-GFP signal in the Rab8, 10, and 11 triple KD data compare to the control? If so, the authors should statistically quantify the difference with the control. In Figure 1H, no statistical analysis is provided between the control and triple KD conditions.
• From the data presented in Figure S1B, the authors state that the basement membrane mislocalization observed in Rab8/10KD has no major impact on polarity maintenance. They based this statement only on the localization of the apical marker aPKC. Although the aPKC data are convincing, it would be more compelling if the authors observe the distribution of other polarity proteins such as Dlg, E-Cadherin, and armadillo to better assess if the overall epithelial polarity is maintained in this condition.

Minor Comments:

General comments:

• In the text describing their data, we recommend that the authors clearly indicate which panel(s) they are referring to.
• The authors should also be consistent with the diction throughout the manuscript when referring to the cortical domain or region of the cell (back/rear/trailing edge/leading edge).
• Several references are missing in the manuscript.

The following specific comments are in order of appearance in the manuscript.

Introduction Section:

The following statements in the introduction should be supported by specific references:

• "BM is critical for tissue development, homeostasis and regeneration, as exemplified in humans by its implication in many congenital and chronic disorders."
• "BM is assembled from core components conserved throughout evolution: type IV collagen (Col IV), the heparan sulfate proteoglycan perlecan, and the glycoproteins laminin and nidogen."
• "During development, the dynamic interplay between cells and BM participates in sculpting organs and maintaining their shape."
• "BM protein secretion shows some specificities, mainly because of the large size of the protein complexes (e.g., procollagen) that must transit from the endoplasmic reticulum to the cell surface". This statement could be supported with references including specific Drosophila references. Additionally, the authors need to clarify what they mean by "some specifies".

Results section:

• In the text describing Fig. 2 (page 5), the authors describe two different basement membrane types: fibrils and homogenous. Moreover, the manuscript focuses on the role of Rab8 and Rab10 in the formation of these two structures. Thus, the authors must better describe the two different types of basement membrane structures and their known roles. This will be helpful for the readers to analyze the presented data, especially for those that are not familiar with the system. In Figure 2A, the authors describe stage 3 basement membrane as uniform BM, do they mean homogenous?
• In the text describing the data for Fig. 3 (page 6), the authors should clearly explain the reason to use anti-GFP antibodies in a non-permeabilized condition (i.e., to detect specifically the extracellular secretion of BM proteins). This will help the readers to interpret the data presented.
• On page 9, the authors stated that the precise localization of Dg in follicle cells is unknown. This statement is incorrect. It has been shown, using a Dg antibody, that Dg localizes at a high level at the basal side of the follicle cells and at a lower level at the apical side (Deng et al, 2003 and Denef et al. 2008).

Discussion Section:

• The following statement is not clear: "Thus, three different Rab proteins are targeted towards the three distinct domains of epithelial cells defined by apical basal polarity, and at least of them is also planar polarized". The authors should rephrase and describe specifically which Rabs they are talking about.
• This statement is vague: "These three Rab GTPases have been jointly involved in different processes (Knödler et al, 2010; Sato et al, 2014; Vogel et al, 2015; Eguchi et al, 2018; Häsler et al, 2020)". The authors could also mention the processes in which Rab8, 10, and 11 are involved.
• The following statements need to be supported by references. "Therefore, more investigations are required to define exactly how the DAPC allows the formation of BM fibrils. Nonetheless, given the importance of the DAPC and BM proteins in muscular dystrophies, our results will pave the way to determine whether a similar function is present also in muscle cells. Interestingly, the extracellular matrix is different between the myotendinous junction and the interjunctional sarcolemmal basement membrane and may provide another developmental context where several routes targeted to different subcellular domains may be implicated".

Experimental Procedure Section:

• In the dissection and immunostaining section (p14), there is a typo: it should be for "20 min" instead of "2for 0 min"
• For the GST pulldown experiments, the authors mention that they use a standard protocol to produce S35 Exo 70 and the GST pulldown experiments. The authors should provide references.

Figure and Figure Legend:

• General comment: The orientation of the images showing the rotation and leading and trailing edges need to be consistent in the different figures (e.g., In Figures 3 and 7, the leading edge is oriented to the top while in Figures 4, S4, 5, 6, the leading edge is oriented to the bottom). This will help the readers to analyze the data.
• In Figure 1 C-G the scale bars are missing and should be added as Fig. 1B.
• Figure S1A: The data presented in Figure S1A is convincing. However, a control panel should be added showing the absence of apical Coll IV for comparison. This information will help with the interpretation of the data.
• In Figure 3 legend: it should be "immunostained" for GFP instead of stain for f-actin and GFP.
• In Figure 4, some scale bars are missing.
• In Figure 4 legend: it should be "(A, E)" after (i.e 0.8 µm above the basal surface) instead of "(C, G)"
• In Figure 5A-E, the authors show quantification of the fibril fraction for Dys-, Rab10 OE, and Rab10OE+Dys, Rab8KD, and Rab8KD+Dys-, and images of the collagen fibril for all the conditions except Dys-, it will be informative that the authors present a representative image of the Coll IV fibril in Dys- condition for comparison. The above comment also applies to Figure 5F-J, and it will be also informative to have a representative image of Dys- condition.
• In Figure 5 legend (p23), it should be "plane" and not "plan".
• Overall, the legend for Fig. S5 is not clear and we recommend the authors to clearly described the different panels. (e.g., it should be "(D)" instead of "(H-J)")
• In Figure 6, some scale bars are missing.

Significance

Despite the important roles of the basement membrane for mechanical support, tissue and organ development, and function, the mechanisms that control the polarized deposition of basement membrane proteins are largely unknown. The contribution of Rab 8 and Rab 10 in the polarized deposition of the basement membrane was previously shown. However, by identifying two competitive secretory routes for the basal secretion of the basement membrane proteins that required these two different RabGTPases, controlled by the DAPC and the exocyst complexes, the authors make a novel contribution to our understanding of the mechanism that leads to the polarized secretion of basement membrane proteins (in that case Collagen IV). Since the basement membrane has critical roles in tissue and organ morphogenesis and functions, and its misregulation has been associated with developmental defects and pathological conditions, this research sheds light on the mechanisms important in these morphogenetic processes and will give insights into their deregulations in pathological conditions.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript by Dennis et al. reports a study of the polarized secretion of basement membrane Collagen IV in the Drosophila (fruit fly) follicular epithelium. Using genetic manipulations and confocal imaging, the authors show that Rab-GTPases Rab8 and Rab10, both known to be required for proper basal secretion of Collagen IV (work by the labs of Sally Horne-Badovinac and Trudi Schupbach, respectively), mediate two alternative secretion routes: Rab8 mediates basal-most secretion of soluble Collagen IV that is incorporated homogenously into the basement membrane, whereas Rab10 mediates basal-lateral secretion of Collagen IV that produces insoluble fibers. The authors additionally study the relation between Rab10 and Dystroglycan/Dystrophin (Dystrophin-associated protein complex, DAPC), which they previously showed to be essential for fibril formation (Cerqueira-Campos et al., 2020). They show here that Dystrophin and Rab10 colocalize at the basal trailing side of follicle cells and that overexpressed Dystroglycan can recruit Rab10 to the plasma membrane; however, they also show that Dystrophin mutants fail to display an effect on Rab10 localization, leaving the significance of the proposed Rab10-DAPC interaction unresolved. Finally, the authors present convincing evidence that the exocyst complex opposes fibril formation, and suggestive but comparatively weaker results pointing that this opposition is due to two independent separate exocyst roles: an inhibitory interaction exocyst-Dystrophin (Dystrophin being required for fibril formation), and a positive role in the alternative Rab8 non-fibril route.

Major comment:

• There are several instances throughout the study in which the authors seem to have problems quantifying results. This affects some assertions central to the message of the paper that are not supported by the quantifications presented. It also casts doubts on accessory points deduced from quantitative differences (or lack of difference) that do not seem fully reliable. I would urge the authors to reevaluate their quantification methods.

a) Rab8 KD does not significantly increase apical fraction of Collagen IV with respect to control (Fig. 1H). The image in 1C clearly shows that Col IV is present apically, something that has been shown by others and that never occurs in the wild type. Failure of the quantifying method to detect a difference can only mean the quantifying method is not adequate. A 10% average in the control when it's clear that no Col IV at all is found apically in the wild type suggests that the authors are quantifying background signal that they should not be acquiring, and, if acquired, they should be subtracting. Rab8/Rab10 double knock down is said to show a synergistic effect, when an additive effect would be more consistent with alternative routes. Other problematic deductions drawn from apical fraction quantifications are found in Fig. 5J (Dys- enhancing Rab8 KD but not Rab10 KD) and Fig. 7D (Exo70- enhancing Rab10 KD but not Rab8 KD).

b) Similar to apical fraction, measurements of planar polarization (trailing/lateral ratio) show average ratios near 1 for Dg, Rab10 and Dys, which is striking given that the localization of these proteins is so clearly polarized. Ratios lower than 1, which are reported for many individual cells in these graphs, should mean reversed polarity. In light of this, I would not be too confident on the effects reported in 5O-Q. In fact, on two occasions, the authors obtain significant differences in these planar polarization measurements that they themselves disregard: Fig. 6J (Rab10 in Exo70-) and Fig. 7I (Dys in Rab8 KD).

c) Quantifications of lateral fraction Col IV in mosaic experiments do not support decreased lateral secretion in Rab8 OE (3G) or Dys- (S5C), which are central tenets of the study.

Minor comments:

• It is stated that Rab10 and Dys associate with tubular endosomes, but no data here support identification as endosomes of these tubular structures, to my understanding.

• The authors call sup-basal the cell region immediately apical to the most basal. Is there sufficient reason to not call this lateral? If a new term is needed, shouldn't it be supra-basal?

• In Fig. S1A and B, Col IV is labeled as green but represented in cyan.

• Fig. S1A should present a wild type control.

• Fig. S1B does not seem to make a significant point in the context of this study.

• Fig. 3C'-E' label suggests a gradient made from multiple images, but it looks like just two images and two colors.

• Graphs in Fig. 3H-J, S5D and 7B are not legible.

• It is not clear where Y2H results in Fig 6A come from.

• I suggest drawing a summary scheme to aid readers better assess interpretations alternative to the ones given in the text.

Significance

This study reports important new information on the secretion of Collagen IV by polarized cells of the Drosophila follicular epithelium. It complements previous studies on the roles of Rab8, Rab10 and Dystroglycan/Dystrophin, additionally uncovering a role for the exocyst complex. Addressing some issues with quantitative imaging should increase confidence in its most critical conclusions.

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):

1) It is interesting MxDnaK1 seems to prefer cytosolic proteins while Mx-DnaK2 prefers inner membrane proteins. The domain-swapping experiments seem to suggest that the NBD is important for this difference. How NBD is important is not addressed. Is it due to ATP hydrolysis, NBD-SBD interaction, or co-chaperone interactions?

Answer: Thanks for your comments. We speculate that the co-chaperone interaction might be the key factor contributing to substrate differences. According to the working principle of Hsp70, its functional diversity is largely determined by substrate differences. Co-chaperones, such as JDPs, play a crucial role in this process as they possess the ability to bind substrates and facilitate their targeted delivery. Therefore, much of the functional diversity of the HSP70s is driven by a diverse class of JDPs 1,2. We found that NBD played important roles in cochaperone recognition of MxDnaKs. Additionally, it is generally accepted that the efficiency of ATP hydrolysis does not significantly impact the substrate recognition of Hsp70. Furthermore, if the NBD-SBD interaction is crucial, the substitution of either the NBD or SBDβ domain might result in similar cell phenotypes, as both alterations disrupt the original NBD-SBDβ interaction. We believe the DnaK proteins and their cochaperones both determine the substrate spectrums. We made corresponding modifications in the revised manuscript. (Page22; Line 488-494 in the marked-up manuscript)

2) About the interactome analysis, since apyrase was added to remove ATP, it's surprising multiple Hsp40s were found in their analysis. Hsp70-Hsp40 interaction is known to require ATP. This may suggest some of the proteins found in their interactome analysis are artifacts. The authors should perform negative controls for their interactome analysis, such as using a control antibody for their CO-IP and analyze any non-specific binding to their resin.

In addition, since JDPs were pull-down, is it possible some of the substrates identified are actually substrates for JDPs, not binding directly to DnaKs?

Answer: This is an interesting question. As you correctly noted, the interaction between Hsp70 and Hsp40 requires ATP. In our experiment, we used apyrase to remove ATP in order to promote tight binding of substrate by DnaK. This methodology was initially described by Calloni, G. et al in 20123, and the authors also identified the co-chaperone protein DnaJ, but with a concentration higher than 77% of the interactors. In our opinions, the incomplete removal of ATP could be the underlying cause of this phenomenon.

We apologize for the undetailed description in Methods. Actually, we implemented negative controls for each MxDnaK in order to eliminate the potential non-specific interactions with Protein A/G beads or antibodies. Specifically, we conducted a CO-IP experiment without the presence of antibodies to assess any non-specific binding to the Protein A/G beads. To further investigate non-specific binding to the antibodies of MxDnaK2 and MxDnaK1, we utilized the mxdnak2-deleted mutant (strain YL2216) and the MxDnaK1 swapping strain with the MxDnaK2 SBDα (strain YL2204), respectively. As the SBDα of MxDnaK1 was employed as antigen to generate antibodies, and YL2204 can’t be recognized by anti-MxDnaK1 (Figure S5). We believe these controls allowed us to evaluate and exclude the non-specific interactions in our CO-IP. We have improved our description in methods. (Page 27; Line 596-607)

While one of the main functions of JDPs is to interact with unfolded substrates and facilitate their delivery to Hsp70, there may still be substrates that do not directly bind to Hsp70. It’s thus possible that some of the substrates identified only bind to JDPs. We made corresponding modifications in the revised manuscript. (Page 14; Line 290-292)

3) For Figure S7, the pull-down assay used His6-tagged JDPs. Ni resin is known to bind Hsp70s non-specifically. It's not surprising DnaK showed up in all the pull-down lanes, especially considering how much DnaK was over-expressed. For some pull-down lanes, the amount of DnaK is much more than that of JDPs, further indicating artifact. The author should include negative controls such as JDPs without His6-tag or any irrelevant protein with His6 tag.

Answer: Thanks for your suggestion. As you and another reviewer pointed out, there were some flaws in the experimental design of the pulldown assay. These include the non-specific binding of Hsp70 proteins to nickel resin, the absence of a negative control without a tag, and the inappropriate selection of the MBP tag. Thus, we employed the nLuc assay as an alternative to the pulldown experiment to validate the interaction between DnaK and JDP (Figure S9). While our manuscript employed nLuc to confirm protein dimerization, it is worth noting that nLuc assay was originally devised for investigating protein interactions 4.

4) For the proposed dimer formation in Fig. 4C, there are multiple bands above the monomer bands. What are these forms? It seems the majority of the Cys residues that could form disulfide bonds are in the NBD of MxDnaK2 since constructs with MxDnaK2-NBD form some sort of high-MW bands above the monomer. Does MxDnaK1-NBD also contain Cys at the analogous positions? The fact that MxDnaK1 didn't show disulfide-bonded bands doesn't mean it doesn't form dimer. It depends on where the Cys residues are.

It's nice the authors did Fig. 4D. However, the authors should include a positive control to show how strong the signal is for a true interaction before interpreting their results.

Answer: Thank you very much for your comments. In at least three independent experiments, we consistently observed two unidentified bands within the molecular weight range of 70-100 kDa during the purification process of His6-MxDnaK2. These bands appeared to be intermediate in size between the monomeric and dimeric forms of His6-MxDnaK2, and disappeared upon DTT treatment. the unidentified band compositions have been confirmed by LC/MS. The upper band included MxDnaK2 (65.3 kDa) and anti-FlhDC factor of E. coli (WP_001300634.1, 27 kDa). In the lower band, we detected the presence of MxDnaK2 and the 50S ribosomal protein L28 of E. coli (WP_000091955.1, 9 kDa). Based on these findings, we conclude that these two additional bands are the result of the interaction between His6-MxDnaK2 and these two E. coli proteins. The related explanations have been added in the legend of Figure 5. (Page 42; Line 938-942)

We analyzed the presence of Cys in MxDnaK1 and MxDnaK2. The NBD region of MxDnaK2 contains two Cys, located at positions 15 and 319. MxDnaK1-NBD contain a Cys at position of 316, which is the analogous position of 319-Cys of MxDnaK2. The analogous position of 15-Cys of MxDnaK2 is a Val in MxDnaK1, which might be an important factor contributing to the inability of MxDnaK1 to form oligomers.

Thanks for your suggestion to add the positive control. We re-performed the nLuc assays including a positive control(αSyn). According to the working principle of the nLuc assay, the amount of fluorescent substrate is limited. Therefore, even for proteins that interact with each other, the fluorescence value gradually decreases and reaches a plateau, similar to the negative control. This gradual decline in fluorescence is a significant indicator of protein interaction. In Figure 4D (Figure 5D in the revision version), we only presented the results of the first 20 minutes of detection. The complete two-hour detection results have been added in the supplementary figure (Figure S14).

5) line 48: "human HSC70 and HSP70 are 85% identical, and the phenotypes of their knockout mutants are different, which is consistent with their largely nonoverlapping substrates" The authors completely ignored that the promoters of HSC70 and HSP70 are very different.

Answer: This is our carelessness. Yes, HSC70 and HSP70 exhibit distinct expression patterns, which play important roles in their functional diversity. We modified the sentence in the new version (Page 5; Line 58)

6) Line 69: "The two PRK00290 proteins, not the other Myxococcus Hsp70s, could alternatively compensate the functions of EcDnaK (DnaK of E. coli) for growth." Please add references for this statement.

Answer: Added, thanks.

7) line 191: What's the mechanism for DnaK's role in oxidative stress? Is the disulfide bond formation in Fig. 4 related? Does disulfide-bond change the activity of DnaK?

Answer: Thanks for your pertinent comments. Honestly, we have no idea about the mechanism for MxDnaK2's role in oxidative stress. In our previous studies, we determined that the deletion of mxdnaK2 resulted in a longer lag phase after H2O2 treatment. Here, our aim was to investigate the impact of region swapping on the cellular function of MxDnaK2. In other bacteria, the critical role that DnaK plays in resistance to oxidative stress stems from the pleotropic functions of this chaperone. By shortening the dwelling time that proteins spend in the unfolded state, the DnaK/DnaJ chaperone system minimizes the risk of metal-catalyzed carbonylation of the side chains of proline, lysine, arginine, and threonine residues, but none of them linked to the dimerization characteristic of DnaK 5-7.

8) Fig. S9 seems redundant.

Answer: Deleted, thanks.

9) line 263, "but the NBD exchange was almost equal to the deletion of the gene with respect to phenotypes." But, the mutant has >50% activity in Fig. 3F.

Answer: We apologize for the confusing description. The “phenotypes” here indicates “cell phenotypes”. What we really tried to say with this sentence is that the NBD swapping strain of either MxDnaK1 or MxDnaK2 presented identical cell phenotypes with the gene-deleted strain. As we have already provided a detailed description of this result earlier, now we consider this sentence to be redundant and have therefore deleted it. (Page 17; Line 355-356)

10) line 221-226: the logic is not quite clear.

Answer: We apologize for the confusing description. In M. xanthus DK1622, MxDnaK1 is essential for cell survival, and an insertion of a second copy of mxdnaK1 in the genome is required for deletion of the in-situ gene. Thus, To verify whether the NBD region is required for the essentiality of MxDnaK1, we performed the region swapping of the in situ MxDnaK1 gene in the att::mxdnaK1 mutant (a DK1622 mutant containing a second copy of mxdnaK1 at attB site), and successfully obtained the MxDnaK1 mutant swapped with the MxDnaK2 NBD region. The experiment indicated that the NBD of MxDnaK1 is essential for the cellular functions of the chaperone. We have added the information and modified the sentences in the manuscript. (Page 15; Line 308-319)

Minor concerns:

Please check spelling. There are some typos such as "HPPES" in the Methods.

Answer: Corrected. Many thanks.

My areas of expertise are protein biochemistry, genetics, and structural biology on heat shock proteins.

Reviewer #2 (Evidence, reproducibility and clarity):

Major comments:

The manuscript is very nice and interesting, although some of the authors' conclusions are perhaps not well supported by their data. For example:

1) In the pulldown experiments the lack of interaction between 2747-MxDnaK2, 3015-MxDnaK2 and 1145-MxDnaK1 should be shown in order to support the conclusion made in line 197-198,

Answer: This is our carelessness. As you and another reviewer pointed out, there are some flaws in the experimental design of the pulldown assay. These include the non-specific binding of Hsp70 proteins to nickel resin, the absence of a negative control without a tag, and the inappropriate selection of the MBP tag. Thus, we employed the nLuc assay as an alternative to the pulldown experiment to validate the interaction between DnaK and JDP (including 2747-MxDnaK2, 3015-MxDnaK2 and 1145-MxDnaK1 interaction) (Figure S9). While our manuscript employed nLuc to confirm protein dimerization, it is worth noting that nLuc assay was originally devised for investigating protein interactions 4.

2) The only evidence that the NBD of MxDnaK1 is essential for bacterial growth is that this mutation couldn´t be obtained in M. xanthus. However, it could be purified in E. coli. Could the authors do some experiments with the M. xanthus strain without the chromosomal MxDnaK1 and then introduce a plasmid with the mutated gene?

Answer: We apologize for the confusing description. Actually, we determined the NBD is essential not only from the mutation couldn’t be obtained. In M. xanthus DK1622, MxDnaK1 is essential for cell survival, and in-situ deletion of the gene could be obtained after an insertion of a second copy of mxdnaK1 in the genome at the attB site. To verify whether the NBD region is required for the essentiality of MxDnaK1, we performed the region swapping of the in situ MxDnaK1 gene in the att::_mxdnaK_1 mutant (a DK1622 mutant containing a second copy of _mxdnaK_1), and successfully obtained the MxDnaK1 mutant swapped with the MxDnaK2 NBD region. The experiment indicated that the NBD of MxDnaK1 is essential for the cellular functions of the chaperone. We have added the information and modified the sentences in the manuscript. (Page 15; Line 308-319)

3) All the experiments with purified proteins were done with MxDnaKs bearing His-tags. It doesn't say explicitly its position, but as they employed a pET28A it is likely that the tag is at the N-terminus, which is close to the linker region. As this tag might interfere, it should be removed for the experiments, or at least a control done with the tag removed.

Answer: We apologize for the lack of detailed description. As you pointed out, the His-tags are located at the N-terminus of DnaKs. The full lengths of MxDnaK1 and MxDnaK2 are 638 and 607 amino acids. The linker regions are located at amino acid positions 381-386 for MxDnaK1 and 387-392 for MxDnaK2. Therefore, we believe that the His-tag is not close to the linker regions. We have included the information in new manuscript. (Page 24; Line 544-546)

The purified His6-DnaK proteins were employed for holdase activity assays and in vitro dimerization assays. Several previous studies have utilized the same holdase activity assay method with His-tagged DnaK 8,9. We suggested that the His-tag did not interfere with the holdase activity of DnaK. To exclude the influence of His-tag on oligomerization, we conducted a control with the tag removed in the in vitro dimerization assay and the result show no difference (Figure S13).

4) The authors state that MxDnaK dimerized in vitro with the NBD, and to disrupt the dimer they used 100 mM DTT, which is a very high concentration. As the protein has the His-tag, it should be removed to corroborate that it is not interfering with the dimerization.

Answer: Thanks for your suggestion. As mentioned above, to exclude the influence of the His-tag on oligomerization, we conducted a control with the tag removed in the in vitro dimerization assay and the result show no difference (Figure S13).

5) Why were the pulldown experiments done with MBP-MxDnaKs? Can you show a negative control between the MBP and the JDPs to rule out this interaction? It will be more suitable to do the pulldown assays with the purified MxDnaK´s without the His-tags (and the His-tags JDP that were employed).

Answer: Thanks for your suggestion. As mentioned above, there are some flaws in the experimental design of the pulldown assay. Thus, we employed the nLuc assay as an alternative to the pulldown experiment to validate the interaction between MxDnaKs and JDPs (Figure S9).

Minor comments:

• E. coli´s DnaK is only essential in heat shock conditions and for lambda phage cycle. If MxDnaK1 is similar to this Hsp70, why the substitution of its NBD for the NBD MxDnaK2 would be lethal for bacterial growth?

Answer: Thanks for the comments. As you correctly point out, DnaK is nonessential in E. coli. But in some other bacteria, DnaK also plays an essential role in cell growth for different reasons 10-12. In our previous studies, we determined that MxDnaK1 is essential in M. xanthus DK1622, and the MxDnaK2 is nonessential. In this study, we performed region swapping and found that only the NBD of MxDnaK1 was unreplaceable. In our opinions, the result indicated that NBD play important roles in the functional diversity between MxDnaK1 and MxDnaK2.

• I think that the writing should be revised and in the supporting information the captions of the figures should include more information.

Answer: Thanks a lot for the suggestion. We revised the manuscript and added more information in the legends of supplementary figures.

Reviewer #2 (Significance):

-General assessment: This is a nice piece of work which would benefit from revision to address the comments above. The authors showed the roles and differences between two DnaK in the same organism. They track these differences to the subdomains of the MxDnaK´s and co-chaperones. It will be interesting for future works to explore more deeply the co-chaperones and their interactions.

-Advance: I think that this manuscript fills a gap regarding the role of DnaK duplicated in bacterial strains. -Audience: I would say that the audience is broad and includes scientists interested in protein folding and chaperones, as well as myxobacteria.

1. Rosenzweig, R., Nillegoda, N. B., Mayer, M. P. & Bukau, B. The Hsp70 chaperone network. Nat Rev Mol Cell Biol 20, 665-680, doi:10.1038/s41580-019-0133-3 (2019).
2. Kampinga, H. H. & Craig, E. A. The HSP70 chaperone machinery: J proteins as drivers of functional specificity. Nat Rev Mol Cell Biol 11, 579-592, doi:10.1038/nrm2941 (2010).
3. Calloni, G. et al. DnaK functions as a central hub in the E. coli chaperone network. Cell Rep 1, 251-264, doi:10.1016/j.celrep.2011.12.007 (2012).
4. Dixon, A. S. et al. NanoLuc Complementation Reporter Optimized for Accurate Measurement of Protein Interactions in Cells. ACS Chem Biol 11, 400-408, doi:10.1021/acschembio.5b00753 (2016).
5. Fredriksson, A., Ballesteros, M., Dukan, S. & Nystrom, T. Defense against protein carbonylation by DnaK/DnaJ and proteases of the heat shock regulon. J Bacteriol 187, 4207-4213, doi:10.1128/JB.187.12.4207-4213.2005 (2005).
6. Santra, M., Dill, K. A. & de Graff, A. M. R. How Do Chaperones Protect a Cell's Proteins from Oxidative Damage? Cell Syst 6, 743-751 e743, doi:10.1016/j.cels.2018.05.001 (2018).
7. Fredriksson, A., Ballesteros, M., Dukan, S. & Nystrom, T. Induction of the heat shock regulon in response to increased mistranslation requires oxidative modification of the malformed proteins. Mol Microbiol 59, 350-359, doi:10.1111/j.1365-2958.2005.04947.x (2006).
8. Chang, L., Thompson, A. D., Ung, P., Carlson, H. A. & Gestwicki, J. E. Mutagenesis reveals the complex relationships between ATPase rate and the chaperone activities of Escherichia coli heat shock protein 70 (Hsp70/DnaK). J Biol Chem 285, 21282-21291, doi:10.1074/jbc.M110.124149 (2010).
9. Thompson, A. D., Bernard, S. M., Skiniotis, G. & Gestwicki, J. E. Visualization and functional analysis of the oligomeric states of Escherichia coli heat shock protein 70 (Hsp70/DnaK). Cell Stress Chaperones 17, 313-327, doi:10.1007/s12192-011-0307-1 (2012).
10. Shonhai, A., Boshoff, A. & Blatch, G. L. The structural and functional diversity of Hsp70 proteins from Plasmodium falciparum. Protein Sci 16, 1803-1818, doi:10.1110/ps.072918107 (2007).
11. Vermeersch, L. et al. On the duration of the microbial lag phase. Curr Genet 65, 721-727, doi:10.1007/s00294-019-00938-2 (2019).
12. Burkholder, W. F. et al. Mutations in the C-terminal fragment of DnaK affecting peptide binding. Proc Natl Acad Sci U S A 93, 10632-10637, doi:10.1073/pnas.93.20.10632 (1996).

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 manuscript describes interesting studies of two paralogues of the E. coli Hsp70, DnaK, from of M. xanthus: MxDnaK1 and MxDnaK2. MxDnaK1 is similar to E. coli DnaK in terms of heat shock response, subcellular localization, etc. while MxDnaK2 is involved with membrane proteins and does not participate in the heat shock response. The interactome of the Mx DnaK´s are larger than that of E. coli DnaK, and their subcellular localization is also different. Regarding the differences between M. xanthus DnaK´s, MxDnaK2 prefers proteins with a higher hydrophobicity score, consistent with its role associated with membrane proteins. The phenotype of diverse mutants with domain swapping showed that the substitution of the NBD of MxDnaK2 for the NBD of MxDnaK1 led to similar phenotypes as the deletion of MxDnaK2 in terms of sporulation and S motility. Consistently, the interactomes of these variants were reduced in number of substrates in comparison with the wild type enzymes. No obvious effect was observed when the SBD´s subdomains were swept. Both MxDnaK interact with JDPs and NEF cochaperones. However, MxDnaK2 interacts only with one of the NEFs, and it depends on the NBD, and has one specific JDP, whichdepends on the beta-subdomain of the SBD (no information provided regarding NBD). MxDnaK1 interacts with both NEFs and has two specific JDPs, which also seems to depend on the beta subdomain of the SBD. Finally, a phylogenetic analysis reveals that the duplication of the dnak gene in Mx is correlated with the complexity of the proteome.

Major comments:

• The manuscript is very nice and interesting, although some of the authors' conclusions are perhaps not well supported by their data. For example: 1) In the pulldown experiments the lack of interaction between 2747-MxDnaK2, 3015-MxDnaK2 and 1145-MxDnaK1 should be shown in order to support the conclusion made in line 197-198, 2) The only evidence that the NBD of MxDnaK1 is essential for bacterial growth is that this mutation couldn´t be obtained in M. xanthus. However, it could be purified in E. coli. Could the authors do some experiments with the M. xanthus strain without the chromosomal MxDnaK1 and then introduce a plasmid with the mutated gene?
• All the experiments with purified proteins were done with MxDnaKs bearing His-tags. It doesn't say explicitly its position, but as they employed a pET28A it is likely that the tag is at the N-terminus, which is close to the linker region. As this tag might interfere, it should be removed for the experiments, or at least a control done with the tag removed.
• The authors state that MxDnaK dimerized in vitro with the NBD, and to disrupt the dimer they used 100 mM DTT, which is a very high concentration. As the protein has the His-tag, it should be removed to corroborate that it is not interfering with the dimerization.
• Why were the pulldown experiments done with MBP-MxDnaKs? Can you show a negative control between the MBP and the JDPs to rule out this interaction? It will be more suitable to do the pulldown assays with the purified MxDnaK´s without the His-tags (and the His-tags JDP that were employed).

Minor comments:

• E. coli´s DnaK is only essential in heat shock conditions and for lambda phage cycle. If MxDnaK1 is similar to this Hsp70, why the substitution of its NBD for the NBD MxDnaK2 would be lethal for bacterial growth?
• I think that the writing should be revised and in the supporting information the captions of the figures should include more information.

Significance

General assessment: This is a nice piece of work which would benefit from revision to address the comments above. The authors showed the roles and differences between two DnaK in the same organism. They track these differences to the subdomains of the MxDnaK´s and co-chaperones. It will be interesting for future works to explore more deeply the co-chaperones and their interactions.

Advance: I think that this manuscript fills a gap regarding the role of DnaK duplicated in bacterial strains.

Audience: I would say that the audience is broad and includes scientists interested in protein folding and chaperones, as well as myxobacteria.

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 study, Pan et al. characterized two Hsp70 DnaKs from Myxococcus xanthus DK1622. Through determining interactomes, the authors defined the differences and similarities between these two DnaKs in interacting with co-chaperones and substrates. Using domain-swapping, the authors analyzed the domain requirements for their functions. Lastly, their bioinformatics analyses seem to suggest the presence of these two DnaKs (i.e., DnaK duplication) is due to the increase of proteomic complexity. Overall, the results are interesting although not surprising. As the authors pointed out, many organisms have multiple Hsp70s with different but overlapping functions. Although multiple experimental approaches were used, the manuscript is generally descriptive without revealing any major mechanistic insights.

1. It is interesting MxDnaK1 seems to prefer cytosolic proteins while Mx-DnaK2 prefers inner membrane proteins. The domain-swapping experiments seem to suggest that the NBD is important for this difference. How NBD is important is not addressed. Is it due to ATP hydrolysis, NBD-SBD interaction, or co-chaperone interactions?
2. About the interactome analysis, since apyrase was added to remove ATP, it's surprising multiple Hsp40s were found in their analysis. Hsp70-Hsp40 interaction is known to require ATP. This may suggest some of the proteins found in their interactome analysis are artifacts. The authors should perform negative controls for their interactome analysis, such as using a control antibody for their CO-IP and analyze any non-specific binding to their resin.<br /> In addition, since JDPs were pull-down, is it possible some of the substrates identified are actually substrates for JDPs, not binding directly to DnaKs?
3. For Figure S7, the pull-down assay used His6-tagged JDPs. Ni resin is known to bind Hsp70s non-specifically. It's not surprising DnaK showed up in all the pull-down lanes, especially considering how much DnaK was over-expressed. For some pull-down lanes, the amount of DnaK is much more than that of JDPs, further indicating artifact. The author should include negative controls such as JDPs without His6-tag or any irrelevant protein with His6 tag.
4. For the proposed dimer formation in Fig. 4C, there are multiple bands above the monomer bands. What are these forms? It seems the majority of the Cys residues that could form disulfide bonds are in the NBD of MxDnaK2 since constructs with MxDnaK2-NBD form some sort of high-MW bands above the monomer. Does MxDnaK1-NBD also contain Cys at the analogous positions? The fact that MxDnaK1 didn't show disulfide-bonded bands doesn't mean it doesn't form dimer. It depends on where the Cys residues are.<br /> It's nice the authors did Fig. 4D. However, the authors should include a positive control to show how strong the signal is for a true interaction before interpreting their results.
5. line 48: "human HSC70 and HSP70 are 85% identical, and the phenotypes of their knockout mutants are different, which is consistent with their largely nonoverlapping substrates." The authors completely ignored that the promoters of HSC70 and HSP70 are very different.
6. Line 69: "The two PRK00290 proteins, not the other Myxococcus Hsp70s, could alternatively compensate the functions of EcDnaK (DnaK of E. coli) for growth." Please add references for this statement.
7. line 191: What's the mechanism for DnaK's role in oxidative stress? Is the disulfide bond formation in Fig. 4 related? Does disulfide-bond change the activity of DnaK?
8. Fig. S9 seems redundant.
9. line 263, "but the NBD exchange was almost equal to the deletion of the gene with respect to phenotypes." But, the mutant has >50% activity in Fig. 3F.
10. line 221-226: the logic is not quite clear.

Minor concerns:

Please check spelling. There are some typos such as "HPPES" in the Methods.

Significance

In this study, Pan et al. characterized two Hsp70 DnaKs from Myxococcus xanthus DK1622. Through determining interactomes, the authors defined the differences and similarities between these two DnaKs in interacting with co-chaperones and substrates. Using domain-swapping, the authors analyzed the domain requirements for their functions. Lastly, their bioinformatics analyses seem to suggest the presence of these two DnaKs (i.e., DnaK duplication) is due to the increase of proteomic complexity. Overall, the results are interesting although not surprising. As the authors pointed out, many organisms have multiple Hsp70s with different but overlapping functions. Although multiple experimental approaches were used, the manuscript is generally descriptive without revealing any major mechanistic insights.

My areas of expertise are protein biochemistry, genetics, and structural biology on heat shock proteins.

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

Learn more at Review Commons

---

Reply to the reviewers

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

Summary: Kellner and Berlin present their research findings pertaining to the effect of GRIN2B variants that modify NMDA receptor function and pharmacology. While these mutations were published previously, the new manuscript provides a more thorough investigation into the effects that these variants pose when incorporated into heteromeric complexes with either wildtype GluN2B or GluN2A - NMDA receptors containing only a single mutated GluN2B subunits is more relevant to the disease cases because the associated patients are heterozygous for the variant. The authors achieved selective expression of receptor heteromeric complexes by utilising an established trafficking control system. The authors found that while a single variant subunit in the receptor complex is largely dominant in its effect on reducing glutamate potency of the NMDA receptor, it 's effect on receptor pharmacology varied. Unlike diheteromeric receptors containing mutated subunits, polyamine spermine potentiated GluN1/2B (but not GluN1/2A/2B) receptors that contained a single mutated GluN2B. In contrast, the neurosteroid, pregnenolone-sulfate (PS), was effective at potentiating the NMDA receptor currents (to varying degrees) regardless of the subunit composition. The potentiation of NMDA receptor currents by PS was also observed in neurons overexpressing the variants.

The techniques used in this study were appropriate to address the objectives and the overall effects are large, and generally convincing. I like the way the results are presented, although have a few (easily addressable) comments.

We thank the reviewer for the positive remarks on our manuscript.

Major comments:

#1 When incrementally adding drugs (e.g. traces in figures 5 and 6), it doesn't always appear like the response has plateaued before changing the solutions/drugs. Therefore, I am curious to what extent the effects observed are underestimated.

The reviewer is correct to note that some responses do not necessarily reach a plateau, despite our efforts reach steady-state (as shown in most figures, e.g., Figs. 1-4, 6b, etc.), in particular when applying pregnenolone-sulfate (PS) (Fig. 5a, all traces in middle and bottom rows). However, in several instances, this was unobtainable due the very slow effect of the neurosteroid (its mode of action is from within the membrane) and the very large size of the cell (~1 mm). For these reasons, these experiments mandated excessively long exposures (~minutes) of oocytes to glutamate and PS (see scale bar- 20 secs) to try to reach steady-state, however this also caused deterioration to some cells (which did not return to baseline- and were therefore discarded). Thus, we eventually converged on settings whereby we did not expose oocytes to more than 4 minutes of the drug. Nevertheless, to try to estimate the extent of the underestimation (if any), we fitted the currents (standard mono-exponential fit, as previously reported1–3 (Suppl. 5a). We found that our application times of PS were, on average, three time the response’s time constants (tau) (Suppl. 5b), and we found a very weak relationship (R2 = 0.09) between the response to PS and time of its application (Suppl. 5c). These are now explicitly mentioned in the text (line #203), and in the legend of Suppl. 5. These thereby suggest that the reaction reached approximately 95% (1 - 1/e^3) of the steady-state value, and we are therefore confident that we have very small, if any, underestimation the extent of PS potentiation.

2 Also, in relation to figure 6, to what extent does agonist application cause desensitization here? Looking at traces in Figure 6b it appears that there is some desensitization and it isn’t clear to what extent this persists during the solution changes.

Agonist desensitization of NMDARs-currents is a well-known phenomenon, but it is very well established that it is not always observed in cells, including neurons (e.g., 4–7). In general, we did not observe very frequent desensitization’s (we provide a larger variety of traces of desensitizing and non-desensitizing currents (Fig. 6b Suppl. 7e and Suppl. 8a). Nevertheless, we explicitly note that in neurons, currents that didn’t reach steady-state after application of 100 mM NMDA were excluded from analysis (Methods - Patch clamping of cultured neurons, line #474), and in most cases desensitization was minor (or absent) following application of 100 mM NMDA and 100 mM PS (Fig. 6b).

3 Could the authors conduct/show the controls where NMDA alone (for 50-60s), or NMDA followed by PE-S (without ifenprodil).

These recordings are now shown in Fig. 6b and Suppl. 8a, (as opposed to Suppl. 7e).

#4 Finally, figure 5 shows the effect of the neurosteroid (and ifenprodil) on NMDA-evoked currents in neurons overexpressing the GluN2B variants in neurons. However, there currents probably reflect a mixture of extrasynaptic and synaptic receptors. To what extent are synaptic NMDA receptors affected by the variants?

To show the extent of the effect of the variants over synaptic receptors, we recorded miniature NMDA-dependent EPSCs; mEPSCNMDA), as described in our previous report8. We find that the varinats completely eliminate the appearance of mEPSCs (Suppl. 7a, b). Change in minis’ frequency is not the result of a presynaptic change or a change in synapse number9, as we have shown that AMPAR-mEPSC frequency was unaffected by the variants (i.e., synapse number and probability of presynaptic release are unchanged by the variants).

  To further address this, we also explored the relative synaptic vs. extrasynaptic distribution of the variants by using the established MK-801-protocol (to block all synaptic receptors during spontaneous activity, leaving extrasynaptic receptors unblocked)10,11. In neurons overexpressing the GluN2B-*wt* subunit, we obtained an extrasynaptic fraction of 38%, highly consistent with previous reports12,13. Overexpression of the variants, however, yielded a significantly and higher fraction (~50%) of the remaining current, supposedly suggesting more variant receptors at extrasynaptic loci (__Suppl. 8b, c__). However, due to the experimental settings we have chosen, the results from this experiment represent quite the inverse when involving extreme LoF variants. Firstly, 100 mM NMDA does not saturate variant receptors (whether pure, mixed di- or tri-heteromers, see __Table 1__). Secondly, normal neurotransmission does not open synaptic receptors containing mutant GluN2B-subunits, attested by the complete absence of mEPSCs (see __Suppl. 7a, b and __8,9). Thus, during the 10 minutes exposure to MK-801, only (mostly) purely *wt* receptors are blocked by spontaneous synaptic activity, and thus the second bout of 100 mM NMDA solely exposes the remaining *wt*-receptors. An increase in the number reflects more *wt*-receptors at the extrasynapse than the synapse. Thus, the observed increase in the fraction of extrasynaptic receptors in neurons overexpressing the variants, implies that the number of *wt*-receptors is necessarily decreased from the synapse and increases at the extrasynapse. We deem this to ensue due to the incorporation of the variants at the synapse. This increase cannot be explained by an overall increase in membrane expression of *wt*-receptors in neurons overexpressing the variants, as these cells show a strong reduction in Imax  (see __Fig. 6c and Suppl. 7e__). This is now detailed in the text (lines #270-290).

Minor comments:

5 Looking at the fits in the graph of Figure 2b it appears that the slope on the concentration response curves is less steep for the mixed 2B-diheteromeric NMDA receptors. How much are the Hill coefficients changing and can this be interpreted to provide more mechanistic insight? Wouldn't it make sense to include the Hill coefficients in Table 1?

We agree with the reviewer’s observation. Actually, the mixed di-heteromers have a similar Hill coefficient (nH) as the purely di-heteromeric GluN2Bwt receptors (see Table 1), and these show the typical near nH ~1 (e.g., 14–16). The only diverging groups are the purely di-heteromeric variant-containing channels (G689C/S only containing receptors; nH~2). Although these may suggest positive cooperation between the subunits, we are less inclined to infer insights from the latter owing to the fact that we limited our examination to 10 mM glutamate (we limit exposure of oocytes to 10 mM glutamate due to artifacts arising past this concentration, as discussed in Kellner et al.8: Fig. 2—figure supplement 1). (this description is now mentioned in page lines #149, 318, 319).

6 The authors illustrate the changes in potency by the shift in the concentration response curves, but is there any change in efficacy? A simple way to illustrate this would be also present a simple graph showing the maximum current amplitudes (i.e. to 10 mM glutamate) for each of the receptor complexes.

We now provide these data in (Suppl. 2a, b). We would like to note however that the expression pattern of the tailed-receptors (i.e., subunits with carboxy-termini tagged with C1/C2 tails, see Fig. 1a) are less expressive in general when compared with the native subunits (Suppl. 2c). This description is detailed in lines# 162-166.

#7 The authors characterize the 'apparent' affinity (or potency) of the receptor using concentration-response curves, but numerous points in the manuscript refer to changes in affinity. None of the experiments shown directly measure affinity (which would require ligand-binding assays) and so the use of the word affinity is inaccurate/misleading. I suggest the authors replace the instances of the word 'affinity' with 'potency'.

We apologize for the confusion surrounding our use of the term affinity. In fact, we do initially define this term in introduction (page #4): “apparent glutamate affinity (EC50)” to differentiate from affinity (KD). Regardless, and to avoid confusion, we replaced all terms, as suggested by reviewer to potency.

#8 In the third line of the abstract, the authors wrote, 'for which there are no treatments' in relation to GRINopathies. My understanding is that there are symptomatic treatments but that there are no disease-modifying treatments.

Indeed, all current treatments are supportive, rather than provide a bona fide cure or disease-modifying. These are now better defined in the abstract.

#9 The authors have interchangeably used the terms NMDAR or GluNRs throughout the manuscript. I suggest sticking to one of these terms. I would suggest NMDARs since this is less likely to be misread as a a specific NMDA receptor subunit.

Agreed and corrected throughout manuscript.

#10 Typos: 1) Results paragraph 2 sentence one: 'We thereby produced GluN2B-wt, GluN2B-G689C and GluN2B-G689S subunits tagged with C1 or C2, co-expressed these along with the GluN1a-wt subunits in...') Results paragraph 2: '...but these were mainly noticeable when oocytes are were exposed to high (saturating) glutamate concentrations...'
3) Last sentence in the second to last paragraph of the results section entitled 'Mixed di-and tri-heteromeric channels...': 'This , PS may serve to rescue...'
4) Last sentence in last paragraph of the results section entitled 'Mixed di-and tri-heteromeric channels...': 'Despite the latter, we found no evidence for any direct effect of three different physiologically relevant concentrations of the drug on di- or tri-heteromeric receptors'

All typos corrected.

#11 Figures 1e, 2b, 3b: it would be helpful to add a legend to the graph so that the curves can be interpreted without having to read through the figure legend.

Corrected.

#12 The bar graphs in Figure 6 show individual data points but those in figures 4b and 5b don't. Can the authors please add the data points to these graph.

Individual data points have been added.

#13 It would be helpful to reviewers that future manuscripts by the authors include page numbers and line numbers.

Included.

**Referees cross-commenting**

#14 Reviewers 2 and 3 highlight an important issue concerning figure 6 and the extent to which the overexpressed variants subunits can compete and assemble with endogenous NMDA receptors (unlike the system where the surface expression of specific receptor complexes is controlled). Indeed in the recent paper by the same authors, the two variants differed in their surface expression (in HEK cells), with G689C expressing particularly poorly. With reference to the second minor comment of Reviewer 1, the maximum current amplitudes would of course need to be normalized to cell surface expression of the receptor to gain any insight into efficacy.

We provide maximal current amplitudes (Imax) as a proxy for expression level as typically done (e.g.,8,17). These are now shown in Suppl. 2a, b (and see our response to comment #6, above). We would like to emphasize that we find it challenging to gain insights about efficacy of the variants in neuronal synapses, as we purposefully express non-C1/C2 tagged subunits in neurons (as we covet assembly of the variants with endogenous subunits). Moreover, the C1/C2-tagged subunits (whether wt or variants) are less expressive compared to their non-tagged NMDAR-counterparts. For instance, tagged GluN2B-wt subunits express at ~50% compared to non-tagged GluN2B wt subunits (Suppl. 2c). Thus, we find that efficacy of the C1/C2 tagged-subunits is less relatable to the non-tagged subunits (which are used in neurons and likely more relevant to the disease).

Despite the latter, we deem that we have specifically addressed this issue by measuring miniature EPSCs (mEPSCs) (see our reply to comment #4, Suppl. 7a, b). Briefly, even though the non-tagged G689C expresses at ~40% compared to other subunits (in oocytes and mammalian cells8), in neurons it engenders a robust (and highly significant) negative effect over synaptic currents (mEPSCs), as strong as the G689S-variant which expresses much more robustly (non-tagged G689S expresses to same extent as wt subunits). This demonstrates that the reduced efficacy the tagged subunits is less relatable to the non-tagged subunits and, importantly, it does not hinder the variants’ ability to incorporate within the synapse and affect function (i.e., exert a dominant negative effect). Here, we extend these observations towards the major postnatal channel subtype, namely tri-heteromers (2A/2B*), and therefore demonstrate that the robust dominant negative effect of G689C and G689S variants is likely due to their ability to incorporate within the predominant receptor subtype at the synapse (Suppl. 8).

Reviewer #1 (Significance (Required)):

This study emphasizes the complex pattern of effects that variants can have on glutamate receptor function and pharmacology, especially considering the context of receptor subunit composition. The authors have followed up their previous findings on the same mutants (Kellner et al, 2021, Elife), but used a trafficking control system here to characterize properties of mutated receptor complexes that are most likely to exist in neurons. The authors show that the defective currents mediated by NMDA receptors containing a loss-of-function GluN2B variant can be enhanced by neurosteroids (and in the case of GluN1/2B receptors, polyamines also). Development and approval of neurosteroids for the clinic would be required for the findings to translate to a therapy for patients. Readers should also be aware that neurosteroids act on other receptors too (e.g. GABA receptors), which could complicate the outcome. The expertise of the reviewer is in glutamate receptors and synaptic transmission.

We agree with the reviewer’s comment pertaining to challenges in translating PS to the clinic. Indeed, we explicitly mentioned its inhibitory effect on GABAA receptors (see line #366-367 and reference 18), as well as note its potential negative effect over GluN2C/D-containing receptors (line #365 and reference 19). We further describe alternative neurosteroids and means to bypass the limitations of PS, for instance by use of 24(S)-hydroxycholesterol6,18 or synthetic analogues (SGE-201, SGE-301)6. Lastly, we also propose a novel therapeutic approach, for which we did not find any mentions in the literature with regard to GRINopathies, consisting of the use of the FDA-approved Efavirenz (anti-retroviral compound20) to promote activity of cytochrome P450 46A1 (CYP46A1) to increase amounts of 24-S in the brain (discussion, lines #370-383).

---

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

#1 The objective of this paper is to assess whether a single mutated subunit of GRIN can affect the function of various forms of NMDA receptors. In particular, this study investigates the functional consequences of a GRIN variant when assembled within tri-heteromers, containing 2 GluN1, 1 GluN2A and 1 GluN2B subunits, the major postnatal receptor type. For this purpose, the authors artificially forced the subunits to associate in predefined complexes, using chimeras of GRIN subunits fused to GABAb receptor retention control sites at the endoplasmic reticulum. This trick allows to control the stoichiometry of the channels at the membrane and thus to focus on the function of a single type of NMDA receptor.The take home message of the paper is that a single GluN2B‐variant, whether assembled with a GluN2B‐wt subunit to form mixed di‐heteromer or with a GluN2A‐wt‐subunit (tri‐heteromer), strongly impairs the receptor functioning, as reported by a decrease of the apparent glutamate affinity of the receptor.

Altogether, this is a straightforward study of great interest for the GRIN community.

We greatly appreciate the reviewer’s comment about the relevance of our work towards the GRIN-community.

2 However, the way the background and purpose of the study (title and abstract) are presented is a bit confusing for non-specialists and could be easily improved. Technical information, which is crucial to validate the conclusions drawn from data analysis, should be added to the article. Some additional experiments are suggested to consolidate the work. Finally, additional discussion points are strongly encouraged.

We apologize for not making the paper more accessible to a broader readership. We did so for the sake of brevity. Nevertheless, we have re-written major parts of the manuscript to address this issue and retitled the report: “Rescuing Tri-Heteromeric NMDA Receptor Function: The potential of Pregnenolone-Sulfate in Loss-of-Function GRIN2B Mutations”.

Specific comments

Abstract / Title:

3 This work shows that a single GRIN variant impairs the function of various forms of NMDA receptors. Several sentences in the title and the abstract are confusing for a non-specialized audience. "Two extreme Loss‐of‐Function GRIN2B‐mutations are detrimental to triheteromeric NMDAR‐function, but rescued by pregnanolone‐sulfate." "Here, we have systematically examined how two de novo GRIN2B variants (G689C and G689S) affect the function of di‐ and tri‐heteromers." The number of variants tested is not of capital importance in the title, especially because one could believe that both are tested at the same time; similarly, when variants are named in the abstract, the fact that only 1 variant is studied at a time should be clarified (G689C OR G689S). Indeed, the problem is obvious to those familiar with GRIN disorders, but if this paper is to be published in a journal reaching a large audience rather than a specialized audience, the title of the paper should be modified.

As noted in our reply to comment #1 of this reviewer, we apologize for not making the paper more accessible and have therefore changed the title and re-written major parts of the manuscript to address this issue. We would like to note that we appreciate the reviewer’s comment and intent to increase the readership of our manuscript.

#4 "We find that the inclusion of a single GluN2B‐variant within mixed di‐ or tri‐heteromeric channels is sufficient to prompt a strong reduction in the receptors' glutamate affinity, but these reductions are not as drastic as in purely di‐heterometric receptors containing two copies of the variants. This observation is supported by the ability of a GluN2B‐selective potentiator (spermine) to potentiate mixed diheteromeric channels." Please, clarify the link between the two sentences. How do spermin potentiation of mixed diheteromeric channels supports the observation that the inclusion of a single GluN2B‐variant has less effect than the inclusion of two variants?

Our intention was to highlight that mixed di-heteromeric channels (2B/2B*) are less “damaged” (this is the link) than purely di-heteromeric channels (2B*/2B*).Explicitly mixed di-heteromers show less reduction in glutamate potency AND are also spermine-responsive, whereas purely mutant di-heteromers (2B*/2B*) show reduced potency, BUT do not respond to spermine at all. We have rephrased the sentences in our current manuscript to be clearer:

For instance: The positive responses of mixed di-heteromers, compared to the null effect over pure di-heteromers, is likely the result of the restored pH-sensitivity of mixed di-heteromers (Suppl. 3). This was surprising as the minimal and essential rules of engagement for potentiation by spermine are not well established, particularly in the case of tri-heteromers21,22 (see discussion, lines #341-353).

Methods

#5 All this study is based on the use of a unique ER‐retention technique to limit expression of a desired receptor‐population at the membrane of cells. According to the ER system retention of GABAb receptor, used in this study, while C1/C1-fused subunits are retained in the ER, C2/C2 reach the cell surface and the association of C1/C2 in the ER enables cell-surface targeting of the heterodimer. However, GB2 does not contain any retention signal and can reach the cell surface in the absence of GB1, as a functionally inactive homo-dimer (doi: 10.1042/BJ20041435). If there is an experimental trick that prevents the addressing of C2/C2 to the cell membrane, it should be specified and explained. This is critically important for understanding which receptor populations the data are derived from: receptors containing C1/C2 fused subunits only as stated by the authors, or C1/C2 and C2/C2 fused subunits?

We base our experiments on two seminal reports—23,24—that have developed this unique method (which we also refer to in the text, lines# 112-116). Briefly, the method employs the binding motifs of GABAB1 (GB1) and GB2 subunits and ER-retention motifs (these are now better detailed in Methods section, line # 448). Previous reports explicitly demonstrate that C1/C1- OR C2/C2-containing receptors do not reach the plasma (or very minimally) and we have reproduced these data with our variants (C1/C1: Suppl. 1a-d; C2/C2: Fig. 1a-c).

Figures #6 NMDA-receptor current amplitude should be normalized by the membrane expression of the receptors. A preliminary experiment should measure the effective cell surface expression of each of the subunits in the different transfection conditions.

To address the effective cell surface expression, we employed Imax as a proxy for functional expression (e.g.,8,17). These are now shown in Suppl. 2a, b (and see our response to reviewer 1, comments #6 and 14). Expectedly, we find significantly reduced efficacy by the varinats compared to wt-receptors, and the purely mutant di-heteromeric receptors exhibit the weakest efficacy. We have also addressed this issue by measuring miniature EPSCs (mEPSCs) (see our reply to reviewer 1, comment #4,). We find the variants to abolish mEPSCNMDA frequency (Suppl. 7a, b). This shows that the variants’ reduced efficacy translates to elimination of synaptic activity (dominant negative effect) (also seen in Suppl. 8).

#7 Fig.1a

The scheme should include C2-C2 complexes and mention whether these complexes are expressed at the cell surface (see previous and following comments).

As noted in our reply to comment #5 of this reviewer (above), C2/C2-containing receptors do not reach the plasma membrane (Fig. 1a-c). To avoid confusion, we have now added this scheme to the cartoon presented in Fig. 1a and have provided a more detailed description of the method and clones produced in the Methods section (line # 448).

#8 Fig.1b and c

Current from cells transfected with GluN2B‐wt‐C1 and GluN2B‐wt‐C2 should be compared to current expressed in cells expressing untagged receptor subunits: GluN2B‐wt Current from cells transfected with GluN2B‐wt‐C1 alone should be shown as well (although expected to be retained in the ER) (as performed for GluN2A‐wt‐C1 GluN2B‐wt‐C1 in suppl Fig. 1a)

Current comparisons of oocytes expressing tagged GluN2B‐wt‐C1 and GluN2B‐wt‐C2 and non-tagged GluN2B‐wt are now demonstrated in Suppl. 2c. The results indicate that the “tags” (C1 and C2) affect the expression of the subunits. We have also added a sample trace of current from a cell expressing the GluN2B‐wt‐C1 alone (Fig. 1b).

9 How could you explain the null current from cells transfected with GluN2B‐wt‐C2 alone (Fig.1b middle, and 1c)? since GB2 does not contain any retention signal and can reach the cell surface in the absence of GB1, GluN2B‐wt‐C2 is supposed to reach the cell surface. This is a very important point to clarify (I am probably missing a technical detail) because if the sub-unit tagged with C2 does reach the cell surface, then all the results and conclusions drawn from the C1-C2 conditions are wrong and could be attributed to a mix of complexes containing either C1-C2 or C2-C2.

We now realize that the reviewer was missing a crucial technical detail, namely how the clones are designed. Briefly, all clones have ER retention motifs and cannot reach plasma membrane unless they necessarily assemble as C1/C223,24. Also, please see our replies to comments #5, 7 to this reviewer (and Methods section, line # 448).

My following comments are based on the assumption that only receptors containing C1-C2 tagged subunits reach the membrane (as assumed by the authors and suggested in Figure 1b middle), but explanations should absolutely be provided to convince the reader. Fig. 4a and 5a (see our above replies to comments #5, 7 and 9; and references 23,24).

#10 Please, keep the current scale constant between all current illustrations within the same figure (4a and 5a). Indeed, not only the Spermin- or SP- induced potentiation is an important data (which is presently quantified on the histograms of fig. 4b and 5b) but also knowing whether the amount of current recorded in cells expressing one mutant subunit in presence of SP (for example GluN2A‐wt‐C1 GluN2B‐G689S‐C2) is comparable to the current recorded in wt receptor-expressing cells (GluN2A‐wt‐C1 GluN2B‐wt‐C2) in absence of SP would be an excellent added value for the paper. A special figure could quantify this rescue effect of SP, measuring and comparing the mean currents recorded in these conditions (one current illustration is not sufficient given variations between similar samples). By the way 5mM glutamate might be an excessive concentration. At 1mM, the expected synaptic concentration of glutamate following action potential, according to figures 3 and Suppl1 the response of the mutated receptor is much lower than that of the WT which is already almost maximal. In these conditions, SP-induced potentiation by a factor of two of GluN2A‐wt‐C1 GluN2B‐G689S‐C2 current could be equivalent to control currents recorded in GluN2A‐wt‐C1 GluN2B‐wt‐C2 cells.

We have rescaled all current amplitudes in Figs. 4 and 5 to be identical in size for easier comparison.

We have added all current amplitudes to try to examine the rescue effect of the two drugs in cell overexpressing a specific channel subtype, as requested (Suppl. 4). We find that; indeed, the potentiated currents of the mutant receptors reach (or even surpass) the basal Imax (i.e., current before potentiation) of the non-mutated receptors (Suppl. 4, dashed statistics bar).

In neurons, we address this in two ways. First, we show that the total NMDA-current is reduced by expression of the variants, and this current is “normalized” by PS (Fig. 6a-c). Similar reductions in Imax (by the variants) are shown in Suppl. 7e (to provide more examples). Secondly, neurotransmission (i.e., 1 mM glutamate25,26) is not sufficient for activating mutant receptors, certainly not pre-di-heteromers (see Table1, EC50 and Suppl. 7a, b- no mEPSCs)27–29. Therefore, 5 mM was required. Together, these strongly suggest that PS may normalize the currents of different receptors that respond to PS (under physiological settings and not 1- or 5mM NMDA). As suggested by the reviewer, there are many subtypes, and some may be activated by ambient glutamate (as suggested by application of PS onto neurons without opening the receptors by NMDA; see Suppl. 7c, d).

#11 Fig. 6

Figure 6 is not convincing because cultured hippocampal neurons do express endogenous NMDA receptors. To what extent the recording currents are affected by endogenous, non-mutated GluN2B subunits? Western Blots showing an extinction of endogenous subunits expression when transfected tagged subunits are competitively expressed would be required.

We have previously shown that the two variants incorporate very efficiently within synapses, causing a very robust elimination of synaptic currents (by measuring miniature NMDA-dependent EPSCs; minis) [see Fig. 8 in Kellner et al. eLIFE, 202127, and see review by Sabo et al.9 ). Change in minis’ frequency can be interpreted as either a presynaptic change or a change in synapse number, however we observed that AMPAR-mEPSC frequency was unaffected by these variants. These imply that synapse number and probability of release are unchanged by the variants. As the experiments are performed in wild-type neurons, (which express wild-type GluN2A and -2B), the dramatic effects we observed on minis suggests a dominant-negative effect of these disease-associated GluN2B variants. These are consistent with our observations that mutant subunits can co-assemble with wild-type GluN2B and/or GluN2A subunits. We have now reproduced this experiment (in fact, we employ this strategy prior each experiment to ensure expression of the variants) (Suppl. 7a, b). This thereby shows that there are no available wt-receptors at the synapse.

As there are various pools of NMDARs at synaptic and extrasynaptic sites, we did not think that a western blot would sufficiently differentiate between the latter, and thereby would not provide insight about extinction of wt-receptors (which could be simply pushed to other sites compared to synapse). Moreover, the intracellular pool of receptors is much larger than the amount of NMDARs that can be detected at the membrane (e.g., 30,31), and therefore electrophysiology seemed to be a better means to monitor membrane receptors only:

Thus, to examine the distribution of the variants between synaptic- and extrasynaptic loci, we employed a standard procedure consisting of the use of the activity-dependent blocker MK-801 (Methods). Briefly, neurons were persistently bathed in TTX during which they were probed for Imax using 100 mM NMDA (to refrain from activating other GluRs), followed by application of MK-801 for 10 minutes to exclusively blocks synaptic receptors (that open following action-potential independent miniature neurotransmission). This thereby spares all extrasynaptic receptors from being blocked by MK-801, which are subsequently revealed by a second application of 100 mM NMDA (Suppl. 8a, inset)12. In neurons overexpressing the GluN2B-wt subunit, we obtained an extrasynaptic fraction of 38%, highly consistent with previous reports12,13. Overexpression of the variants, however, yielded a significantly and higher fraction (~50%) of remaining current (Suppl. 8b, c), but instead of reflecting a larger pool of extrasynaptic receptors, the experiment represents quite the inverse when involving LoF variants. Firstly, 100 mM NMDA does not saturate variant receptors (whether pure, mixed di- or tri-heteromers, see Table 1). Secondly, normal neurotransmission does not open synaptic receptors containing mutant GluN2B-subunits, attested by the complete absence of mEPSCs (see Suppl. 7). Thus, during the 10 minutes exposure to MK-801, only wt receptors are blocked by spontaneous synaptic activity, and thus the second bout of 100 mM NMDA solely exposes the remaining wt-receptors at the extrasynapse. Thus, the observed increase in the fraction of extrasynaptic receptors, in neurons overexpressing the variants, implies that the number of wt-receptors is necessarily decreased from the synapse and increases at the extrasynapse, most likely due to the incorporation of the variants at the synapse. This increase cannot be explained by an overall increase in membrane expression of wt-receptors in neurons overexpressing the variants, as these cells show, yet again, a strong reduction in Imax as seen above (see Fig. 6c and Suppl. 7e) (lLines #270-291). These thereby suggest that purely wt-receptors are not necessarily eliminated from the membrane (extinct), rather pushed outside of the synapse.

12 Fig.6b “PE-S” on the graph should be replaced by “PS”

Typo corrected.

Discussion #13 The authors are surprised by the fact (Fig.2) that 1 variant reduces the apparent glutamate affinity of the receptor, but not as much as 2 variants, despite the fact that "NMDARs opening requires all four subunits to be liganded (i.e., occupied by a ligand) which implies that the least affine subunit should have dominated the final affinity of the receptor". I agree that the difference is noticeable, however the glutamate affinity for receptors containing 1 variant is much closer to that of receptors containing 2 variants than that of wild-type receptors. Hence, the results obtained do not seem so surprising and could result, as rightly explained by the authors from a possible cooperativity between the subunits.

We agree with the reviewer that glutamate potency of receptors containing 1 variant subunit is much closer to that of receptors containing 2 variant subunits. However, we maintain our surprise because we expected it to equal (not just close) to the potency of the least affine subunit (the limiting factor). This is based on the notion that all four subunits need to be liganded for channel opening4,32–34. We gently raise the possibility of potential cooperativity (Table 1, see Hill-coefficient and 33,35,36), as well as mention that this may also stem from the variants’ lower proton sensitivity (Suppl. 3), which has also been shown to promote motions of the ATD (amino terminal domain) and increase open probability (positive cooperativity)36. Nevertheless, we are very careful with interpreting the Hill coefficient , as we limited exposure of oocytes to 10 mM glutamate due to artifacts arising past this concentration (see Kellner et al.8: Fig. 2—figure supplement 1). This description is now mentioned in page lines #149, 318, 319. Thus, even the slightest underestimation of the maximal reposnse would surely affect the slope.

#14 On the other hand, the data in Figure 6 are much more difficult to interpret and reconcile with the nature of the expressed receptor subunits (which this time is not controlled) nor their association within the same receptor. However, this aspect, which is essential to the understanding of the consequences of 1 variant on neuronal signalling, is not discussed: Whatever the stoichiometry of the complexes in the heterozygous disease, the mutated and wild type GluN2B subunits coexist in the same cell: Either within the same di-heteromeric complexes GluN2B-wt + GluN2B-mutant, or in separate complexes but nevertheless expressed in the same cell, in di heteromeric (GluN2B-wt + GluN2B-wt and GluN2B-mutant + GluN2B-mutant); or tri-heteromeric (GluN2A-wt + GluN2B-wt and GluN2A-wt + GluN2B-mutant) complexes. Assuming that half of the complexes remain wild-type, e.g. (GluN2A-wt + GluN2B-wt and GluN2A-wt + GluN2B-mutant) we would expect (Fig. 6) a small decrease in NMDA current (carried only by the half that expresses the mutated subunit, and whose function is not zero but only decreased by about 20% in response to 5 mM Glutamate, Fig. 3b). The same reasoning applies to the di-heteromeric conditions (GluN2B-wt + GluN2B-wt; GluN2B-mutant + GluN2B-mutant), here again the decrease observed Fig. 6b is difficult to reconcile with the responses measured Fig. 2b.

In other words, how to explain a 50% decrease of the currents, instead of the 10% expected by the previous reasoning. In this experiment we do not know which subunits are expressed, their proportions, nor how they are associated in functional complexes, which makes the interpretation of the data impossible. The only explanation, far-fetched, for 50 % decrease would be that the complexes were to contain all (or the vast majority) 1 wild-type subunits associated with 1 variant, then a homogeneous 50% reduction in current could be expected. But this extreme condition could only be possible in the case of di-heteromers, which is unlikely the case in Fig.6 as GluN2A currents are measured in presence of Ifenprodil. To conclude

1) the comparison of the currents in transfected and non-transfected neurons does not make sense in figure 6b which is not convincing because we do not know the nature of the currents actually measured. A comparison in controlled condition would make more sense (as I suggested in the criticism of figures 4, 5).

2) The reality of the combinations of expression and association between subunits within different complexes expressed in the same cell must be considered and taken into account in the interpretation of the data. Undoubtedly, the means of restoring the NMDA current will be different depending on the presence of mutated subunits in all functional channels or not.

Indeed, neurons express a variety of different combinations of channel stoichiometry, including following transfection with the variants. We do find find that the effect on whole-cell current is indeed ~50% (Fig. 6b, c), thereby safe to assume that 50% remain “wt”, but we do not know how they distribute between synaptic and extrasynaptic loci. Our results however argue against 50% remaining receptors at the synapse. First, mEPSCNMDA disappear (Suppl. 7a, b and see reply to comment #11 of this reviewer), but wt-receptors are still at the membrane, and they seem to be moving out of synapse (Suppl. 8). Thus, we can only state with higher certainty that the variant subunits are very efficient in incorporating within mixed or pure receptors, especially at the synapse.

  We also consider that the reduction in the whole-cell current observed in __Fig. 6b, c__ is not due to the remaining 50% GluN2B-*wt*-containing receptors, rather likely due to other variants, notably GluN2A, which are more prominent at postnatal stages37, such as in our case. In support, we see a large remaining current after saturating ifenprodil application (__Suppl. 7 e, f__)38. Thus, the variants incorporate within all 2A/2B membrane receptors, at the synapse and outside it (i.e., extrasynaptic) (see __Suppl. 8, c__).

**Referees cross-commenting**

The referees' comments are highly relevant. In particular, referee 3's comment 1 seems very interesting because it may help to better understand the discrepancy in the results observed in neurons, i.e. a 50% decrease in the currents induced by the expression of the mutant and wild type subunits in the same cells, whereas theoretically one would expect a 10% decrease of this current (cf. referee 2's 2nd comment in the discussion section). This comment 1 of referee 3 indeed stresses the fact that the control (non-transfected neurons) to which the heterozygous condition is compared is not the correct control, which should rather be neurons transfected with wild type receptor subunits. More generally, this comment underlines the importance of monitoring the effective membrane expression of the different subunits in each of the experimental conditions in order to be able to compare conditions and draw conclusions.

We initially did not perform this control as the literature paints a clear picture whereby expression of the GluN2B-subunit (without adding excess of the GluN1 subunit) does not instigate a robust increase in surface expression of NMDARs (and thus current remains the ~same) 4,39–43, and see our reply to comment #14 (above), and reviewer 3 comment 1 (below). Nevertheless, we have now performed this test by overexpressing GluN2B-wt. In support of previous reports, we do not find any statistical difference in current size between non-transfected neurons and neurons solely overexpressing the GluN2B-wt subunit (Fig. 6a, b). Furthermore, application of PS onto naïve or GluN2Bwt expressing neurons yields identical currents (Imax) and potentiation (Fig. 6c, d). These argue that we did not obtain “overexpression”.

We suggest that the 50% reduction in current size between neurons expressing the mutant and wt expressing neurons stems from the integration of mutant subunits and their dominant negative effect. Evidence for this incorporation is provided by the very strong reduction in synaptic currents (suppl 7a, b), and the supposedly higher abundance of wt-containing receptors in extrasynaptic regions (see reviewer 1 comment 4 and suppl 8). This is

Reviewer #2 (Significance required):

The novelty of the study, is to evaluate the consequences of a single mutated subunit within NMDA receptors affected by GRIN variant, to mimic the heterozygous condition of GRIN encephalopathies, this is of potential value for the field and the interest could also be extended to other genetic diseases (at least the experimental way to study the functioning of only one desired stoichiometric configuration). The strength of this paper is precisely to isolate technically and to study the functioning of a desired stoichiometric configuration only. The main limitation of the paper is the interpretation that the authors make of their data in a physio-pathological context. This work could be of interest for general audience, providing the title and summary are slightly modified. My area of expertise could not be closer to the topic of the article: Glutamate receptors; GRIN; molecular tinkering, cell culture, electrophysiology, receptor stoichiometry...

We thank the reviewer for noting the value in our work and its potential contribution and interest to the field and other diseases. Per reviewer’s suggestion, we have modified the title and text to suit a larger audience.

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

This paper is a follow up of an earlier paper published by the group (Kellner et al., eLife 2021), which aimed at characterizing the functional properties of two de novo GluN2B mutations in patients suffering from severe pediatric diseases, GluN2B-G689C and -G689S. NMDA receptors (NMDARs) are tetramers composed of two GluN1 and two GluN2 subunits. A single receptor can incorporate either two identical GluN2 subunits (di-heteromers) or two different GluN2 subunits (tri-heteromers), leading to a large diversity of NMDAR subtypes. The main NMDAR subtypes in the adult forebrain are GluN1/GluN2A and GluN1/GluN2B di-heteromers, as well as GluN1/GluN2A/GluN2B tri-heteromers. While the exact proportions of these three subtypes are still contentious, there are evidence that in the adult N1/2A/2B tri-heteromers form the major population of synaptic NMDARs in the adult forebrain. In addition, patients bearing pathogenic mutations are often heterozygous for the mutation, giving rise to mixed NMDARs incorporating one mutated and one intact GluN2 subunit. In their previous paper, Kellner et al. had shown that purely di-heteromeric GluN1/GluN2B-G689C and -G689S mutants display a drastic (> 1,000-fold) decrease of glutamate sensitivity and a decrease of surface expression. In the current paper, the authors characterize the effects of the -G689C and -G689S mutations on N1/2A/2B tri-heteromeric receptors, as well as on mixed di-heteromeric GluN1/GluN2B receptors containing one copy of the wild-type GluN2B subunit and one copy of the mutated GluN2B subunit. They show that one copy of the mutant subunit, either within mixed diheteromeric or tri-heteromeric receptors, is sufficient to decrease drastically glutamate sensitivity, although the shift in glutamate EC50 is not as strong as in pure di-heteromeric receptors (≈ 500-fold). They furthermore explore strategies to counteract the hypofunction induced by these mutations by testing the effect of positive allosteric modulators (PAMs). They show that spermine, a GluN2B-specific PAM, can potentiate the activity of mixed diheteromeric N1/2B but not N1/2A/2B tri-heteromers. However pregnenolone sulfate (a 2A/2B-specific PAM) can potentiate both the activity of mixed diheteromeric and tri-heteromeric NMDAR populations, either in oocytes or cultured neurons.I have very few major comments to make. The experiments are straightforward and the adequate controls have been made. Here are my two only major comments:

We thank the reviewer for the very detailed overview of our work and for appreciating our controls and methods.

#1 About the experiment on cultured neurons. The authors compare the currents of cultured neurons transfected with GluN2B-G689C and -G689S to non transfected neurons. The adequate control is rather neurons transfected with the wild-type GluN2B subunit to even out any phenomenon linked to transfection of the neuron. Given the overexpression that can occur after transfection, the effect of the mutations on the size of NMDAR currents might be even stronger than what the authors show. However in that case PS might not completely rescue mutant NMDAR currents to wild-type levels.

We initially did not perform this control as the literature paints a clear picture whereby expression of the GluN2B-subunit (without adding excess of the GluN1 subunit) does not instigate a robust increase in surface expression of NMDARs (and thus current remains the ~same) 4,39–43, and see our reply to comment #14 (above), and reviewer 3 comment 1 (below). Nevertheless, we have now performed this test by overexpressing GluN2B-wt. In support of previous reports, we do not find any statistical difference in current size between non-transfected neurons and neurons solely overexpressing the GluN2B-wt subunit (Fig. 6a, b). Furthermore, application of PS onto naïve or GluN2Bwt expressing neurons yields identical currents (Imax) and potentiation (Fig. 6c, d). These argue that we did not obtain “overexpression”. Thus, our results and interpretations hold true, and are therefore not underestimation of the effects of PS in neurons.

2 How come high concentrations of glutamate (>100µM) produce additional current on wt GluN1/GluN2B (with retention signals) compared to 100 µM glutamate, which is supposed to be saturating? It does not seem to stem from an osmotic effect since 10 mM glutamate does not produce any current on uninjected oocytes. Knowing that this "artefactual" effect might also occur in the mutant receptors, how do you take this effect into account when calculating the glutamate EC50s of the mutants? Given the drastic shift in EC50 produced by the mutant, taking into account this artefact is not going to change the conclusion, but the actual EC50s will be affected.

GluN1/GluN2B-wt receptors (with or without retention signals) are indeed saturated at 100 mM glutamate. However, excessively large concentrations of glutamate (>100 mM) may yield artefacts even in non-injected oocytes (in 10 mM, this occurs in ~20% of the cells, see Kellner et al 20218—Fig. 2 and Suppl. 1c, d) as well as in GluN2B-wt injected oocytes (supplementary Table 1 in 44). This is not due to osmolarity, as rightly mentioned by the reviewer (and below), rather possibly by endogenous glutamate receptors and transporters that do not readily contribute to current amplitude (these are extremely small currents), but can cause deterioration of the cell (and enhance ‘leak’) when activated for prolonged times by very large concentrations (e.g.,45). In fact, we explicitly report these to highlight potential artefacts, as these are often overlooked in the field. Regardless, most reports do no go past 100 mM glutamate, not even when describing GRIN2 mutations since most mutations do not cause such drastic shifts in potency as we observed (to the best of our knowledge only one report describes such an extreme LoF mutation for a GluN2A variant46). Of note, these effects are not seen when glycine is applied at high concentrations (supporting lack of effect by osmolarity)47. Thus, we refrained from testing concentrations past 10 mM, aware that it may yield a slight underestimation of glutamate potency (and perhaps the reason for the larger Hill coefficient, nH; see our reply to reviewer #1, comment #5). Importantly, despite the potential underestimation of the EC50, it does not change our conclusions as all groups are measured side-by-side (thus, the same underestimation equally applies to all other groups as well). We now mention this more in detail in the methods under the section – “Two Electrode Voltage Clamp recordings in Xenopus Laevis oocytes”.

Minor comments:

3 In the first paragraph of the "Results" section, when describing the design of the constructs used to force a heteromeric stoichiometry in recombinant systems, the authors do as if they had designed the constructs themselves "Briefly, we tagged...are retained in the ER (Fig. 1a)". Please rewrite this paragraph to show that you used constructs that had been previously designed by another group (Hansen et al., 2014).

We apologize. We did not mean to express that we have developed the method and indeed refer readers to the seminal works of those who did (Stroebel et al., 2014 and Hansen et al. 2014, lines #109-116). We did not go into details for the sake of brevity. We have rewritten this part to give proper acknowledgement to the method’s developers (also see Methods, line# 448).

4 I do not see any evidence of "positive cooperativity" between subunits in ref. 32. Ref. 32, to the best of my knowledge, states that in N1/2A/2B tri-heteromers, the 2A subunit sets the biophysical properties of the tri-heteromer. But there is no account of mixed di-heteromers. In addition, the cooperativity between the glutamate and glycine binding sites is negative.

The reviewer is correct, and we apologize for the mis-citation. Indeed, the cooperativity between glutamate- and glycine-binding is typically reported as negative48,49, and our intention was to highlight the strong cooperativity (whether positive or negative) observed between NMDAR-subunits and meant to cite the works of: 33,35,50 (lines . We have now rephrased the sentence: The divergence from this scenario suggests that the slight amelioration in potency could stem from positive cooperativity between the subunits50 (but see Hill coefficients in Table 1). Indeed, mixed receptors show restored proton sensitivity (Suppl. 3), which has been suggested to be coupled to other receptor features, notably increase in open probability.

5 Interpretation of spermine action within the Results section: it is striking indeed to observe that the mutations in the context of a mixed di-heteromer still allow spermine potentiation, while they abolish this potentiation in pure di-heteromers. As rightly said in the discussion, the regain of spermine potentiation in the mixed compared to the pure diheteromers is likely due to a more favorable transduction of spermine signaling to the pore, likely via a higher pH sensitivity of mixed di-heteromers compared to di-heteromers. I would thus avoid the terms of "one single intact interface" for the mixed di-heteromer, since both spermine binding sites are likely intact in this NMDAR configuration. How is pH sensitivity affected in the mixed di-heteromers?

We have performed a detailed pH dose-responses for the various channel types (Suppl 3). We find that GluN2B mixed di-heteromers exhibit similar IC­50 as pure GluN2B-wt di-heteromers, thus explaining their ability to undergo potentiation by spermine via alleviation of proton inhibition. We therefore further suggest that mixed di-heteromers’ have higher pH-sensitivity compared to pure mutant di-heteromers and this mat also contributes to their higher spermine sensitivity. Lastly, we observed that all GluN2A-wt-containing tri-heteromeric receptors were non-responsive to spermine (Fig. 4a). In fact, under our experimental conditions tri-heteromers underwent slight inhibition by spermine, regardless the identity of the GluN2B subunit (whether wt or variant) (Fig. 4b). Thus, as the tri-heteromers used here exhibit identical pH-sensitivity as 2B-di-heteromers, the only diverging aspect is the missing interface between the GluN1a and GluN2B subunits, demonstrating that potentiation by spermine requires at least one GluN2B-subunit with an intact proton sensitivity, and mandates two intact interfaces between GluN1-wt and GluN2B-wt subunits (Table 1)21.

6 In the methods section, the oocyte recording solution (likely Ringer and not Barth) does not contain any potassium. This is probably a typo. Could you correct the composition of your Ringer?

Corrected. We record NMDARs currents by use of a Barth solution containing (in mM): 100 NaCl, 0.3 BaCl2, 5 HEPES, pH 7.3 (adjusted with KOH, at ~2.5 mM) (as in 4,51).

7 There are several typos, especially in the Discussion.

We have corrected the typos throughout the publication.

**Referees cross-commenting**

I overall agree with the comments of reviewers 1 and 2. In particular, I agree that it is pointless to compare the absolute currents of non transfected neurons vs mutant-transfected neurons without an idea of receptor cell-surface expression.

We have performed this experiment (Fig. 6) and please see our reply to this reviewer’s comment #1.

I would like however to give some precisions about some comments of Reviewer 2. About the ER retention technique to express tri-heteromers: I didn't know that the C2 signal could be addressed to the membrane on its own. The lack of leak current stemming from C1-C1 or C2-C2 combinations has been demonstrated in the paper establishing the technique (Hansen et al, 2014), as well as in another paper that developed an analog technique based on GABAB retention signals (Stroebel et al., J Neurosci 2014). So it is fair to consider that the authors were not surprised by the lack of current when co-expressing two GluN2B subunits carrying the C2 signal.

We thank you for this addition and support for our observations.

About the comparison about absolute currents wt vs mutants, +/- spermine (Fig. 4a and 5a). I agree with reviewer 2 that being able to compare absolute currents of wt without spermine to mutant + spermine would be very interesting to see if spermine can actually rescue mutant hypofunction. However, to the defense of the authors, comparing absolute current values of recordings from Xenopus oocytes is meaningless. Indeed the variability of currents for the same construct and same day of experiment is too high (there can be up to a ten-fold difference between the lowest and the highest current of oocytes expressing the same construct the same experimental day). A way to investigate this aspect would be to estimate the open probability of the different constructs with or without spermine via the inhibition kinetics of an open channel blocker (e.g. MK801) and measure surface expression by Western blot, but I am not sure these experiments are worth it for the spermine experiment.

We agree with this reviewer about current size. It is quite variable among cells and would therefore introduce an additional variable and variability: the expression of these modified (C1/C2-tagged) subunits is dually affected by the mutation itself (Kellner et al. 2021) and by the introduction of the tagging (which really hampers there trafficking to membrane, Suppl. 2c); with unknown contribution of each variable. We thereby do not think these provide an added value to our conclusions, yet to grant reviewers’ no 2 request we have added __Suppl. 4 __which shows the rescue effect of the different drugs.

Reviewer #3 (Significance (Required)):

This paper is not of high significance since most of the characterization of the 2B-G689C and -G689S de novo mutants found in patients has already been published (Kellner et al., eLife 2021). However, this paper is worth publishing since it brings new data on the effect of the mutations on tri-heteromeric and mixed di-heteromeric NMDAR populations, which are likely the most abundant NMDAR populations in the patient's brain at adult stage. Tri-heteromeric and mixed NMDAR populations have often been overlooked when studying pathogenic NMDAR mutations due to the difficulty to express them specifically in recombinant systems. This paper (in addition to other papers in the field, see for instance Elmasri et al., Brain Sci. 2022; Li et al., Hum. Mutat. 2019) shows that the effect of the mutations on the receptor biophysical and pharmacological properties (but also on trafficking) differ whether the receptor contains one or two copies of the mutant subunit. This paper is of interest to scientists interested in NMDA receptor structure-function and pharmacology, as well as clinicians interested in GRINopathies (pathologies linked to NMDAR mutations).

I, the reviewer, am an expert in NMDAR structure-function and pharmacology. I believe I have sufficient expertise to evaluate the entirety of the paper.

We thank the reviewer for appreciating and acknowledging the merits of our work for publication.

References:

1. Berlin, S. et al. Gαi and Gβγ Jointly Regulate the Conformations of a Gβγ Effector, the Neuronal G Protein-activated K+ Channel (GIRK). J. Biol. Chem. 285, 6179–6185 (2010).
2. Kahanovitch, U., Berlin, S. & Dascal, N. Collision coupling in the GABAB receptor–G protein–GIRK signaling cascade. FEBS Lett. 591, 2816–2825 (2017).
3. Berlin, S. et al. A Collision Coupling Model Governs the Activation of Neuronal GIRK1/2 Channels by Muscarinic-2 Receptors. Front. Pharmacol. 11, (2020).
4. Berlin, S. et al. A family of photoswitchable NMDA receptors. eLife 5, e12040 (2016).
5. Reyes-Guzman, E. A., Vega-Castro, N., Reyes-Montaño, E. A. & Recio-Pinto, E. Antagonistic action on NMDA/GluN2B mediated currents of two peptides that were conantokin-G structure-based designed. BMC Neurosci. 18, 44 (2017).
6. Paul, S. M. et al. The Major Brain Cholesterol Metabolite 24(S)-Hydroxycholesterol Is a Potent Allosteric Modulator of N-Methyl-D-Aspartate Receptors. J. Neurosci. 33, 17290–17300 (2013).
7. Yakovlev, A. V., Kurmasheva, E. D., Ishchenko, Y., Giniatullin, R. & Sitdikova, G. F. Age-Dependent, Subunit Specific Action of Hydrogen Sulfide on GluN1/2A and GluN1/2B NMDA Receptors. Front. Cell. Neurosci. 11, 375 (2017).
8. Kellner, S. et al. Two de novo GluN2B mutations affect multiple NMDAR-functions and instigate severe pediatric encephalopathy. eLife 10, e67555 (2021).
9. Sabo, S. L., Lahr, J. M., Offer, M., Weekes, A. L. & Sceniak, M. P. GRIN2B-related neurodevelopmental disorder: current understanding of pathophysiological mechanisms. Front. Synaptic Neurosci. 14, (2023).
10. Martel, M.-A. et al. The Subtype of GluN2 C-terminal Domain Determines the Response to Excitotoxic Insults. Neuron 74, 543–556 (2012).
11. Papouin, T. et al. Synaptic and Extrasynaptic NMDA Receptors Are Gated by Different Endogenous Coagonists. Cell 150, 633–646 (2012).
12. Harris, A. Z. & Pettit, D. L. Extrasynaptic and synaptic NMDA receptors form stable and uniform pools in rat hippocampal slices. J. Physiol. 584, 509–519 (2007).
13. Moldavski, A., Behr, J., Bading, H. & Bengtson, C. P. A novel method using ambient glutamate for the electrophysiological quantification of extrasynaptic NMDA receptor function in acute brain slices. J. Physiol. 598, 633–650 (2020).
14. Curras, M. C. & Dingledine, R. Selectivity of amino acid transmitters acting at N-methyl-D-aspartate and amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptors. Mol. Pharmacol. 41, 520–526 (1992).
15. Laube, B., Hirai, H., Sturgess, M., Betz, H. & Kuhse, J. Molecular Determinants of Agonist Discrimination by NMDA Receptor Subunits: Analysis of the Glutamate Binding Site on the NR2B Subunit. Neuron 18, 493–503 (1997).
16. Esmenjaud, J. et al. An inter‐dimer allosteric switch controls NMDA receptor activity. EMBO J. 38, (2019).
17. Liu, S. et al. A Rare Variant Identified Within the GluN2B C-Terminus in a Patient with Autism Affects NMDA Receptor Surface Expression and Spine Density. J. Neurosci. 37, 4093–4102 (2017).
18. Geoffroy, C., Paoletti, P. & Mony, L. Positive allosteric modulation of NMDA receptors: mechanisms, physiological impact and therapeutic potential. J. Physiol. 600, 233–259 (2022).
19. Malayev, A., Gibbs, T. T. & Farb, D. H. Inhibition of the NMDA response by pregnenolone sulphate reveals subtype selective modulation of NMDA receptors by sulphated steroids. Br. J. Pharmacol. 135, 901–909 (2002).
20. Petrov, A. M. et al. CYP46A1 Activation by Efavirenz Leads to Behavioral Improvement without Significant Changes in Amyloid Plaque Load in the Brain of 5XFAD Mice. Neurotherapeutics 16, 710–724 (2019).
21. Mony, L., Zhu, S., Carvalho, S. & Paoletti, P. Molecular basis of positive allosteric modulation of GluN2B NMDA receptors by polyamines. EMBO J. 30, 3134–3146 (2011).
22. Stroebel, D., Casado, M. & Paoletti, P. Triheteromeric NMDA receptors: from structure to synaptic physiology. Curr. Opin. Physiol. 2, 1–12 (2018).
23. Hansen, K. B., Ogden, K. K., Yuan, H. & Traynelis, S. F. Distinct Functional and Pharmacological Properties of Triheteromeric GluN1/GluN2A/GluN2B NMDA Receptors. Neuron 81, 1084–1096 (2014).
24. Stroebel, D., Carvalho, S., Grand, T., Zhu, S. & Paoletti, P. Controlling NMDA Receptor Subunit Composition Using Ectopic Retention Signals. J. Neurosci. 34, 16630–16636 (2014).
25. Clements, J. D., Lester, R. A. J., Tong, G., Jahr, C. E. & Westbrook, G. L. The Time Course of Glutamate in the Synaptic Cleft. Science 258, 1498–1501 (1992).
26. Budisantoso, T. et al. Evaluation of glutamate concentration transient in the synaptic cleft of the rat calyx of Held: Glutamate concentration in synapse. J. Physiol. 591, 219–239 (2013).
27. Kellner, S. et al. Two de novo GluN2B mutations affect multiple NMDAR-functions and instigate severe pediatric encephalopathy. eLife 10, e67555 (2021).
28. McAllister, A. K. & Stevens, C. F. Nonsaturation of AMPA and NMDA receptors at hippocampal synapses. Proc. Natl. Acad. Sci. 97, 6173–6178 (2000).
29. Ishikawa, T., Sahara, Y. & Takahashi, T. A Single Packet of Transmitter Does Not Saturate Postsynaptic Glutamate Receptors. Neuron 34, 613–621 (2002).
30. Washbourne, P., Liu, X.-B., Jones, E. G. & McAllister, A. K. Cycling of NMDA Receptors during Trafficking in Neurons before Synapse Formation. J. Neurosci. 24, 8253–8264 (2004).
31. Yan, Y.-G. et al. Clustering of surface NMDA receptors is mainly mediated by the C-terminus of GluN2A in cultured rat hippocampal neurons. Neurosci. Bull. 30, 655–666 (2014).
32. Kussius, C. L. & Popescu, G. K. Kinetic basis of partial agonism at NMDA receptors. Nat. Neurosci. 12, 1114–1120 (2009).
33. Sun, W., Hansen, K. B. & Jahr, C. E. Allosteric interactions between NMDA receptor subunits shape the developmental shift in channel properties. Neuron 94, 58-64.e3 (2017).
34. Benveniste, M. & Mayer, M. L. Kinetic analysis of antagonist action at N-methyl-D-aspartic acid receptors. Two binding sites each for glutamate and glycine. Biophys. J. 59, 560–573 (1991).
35. Lü, W., Du, J., Goehring, A. & Gouaux, E. Cryo-EM structures of the triheteromeric NMDA receptor and its allosteric modulation. Science 355, eaal3729 (2017).
36. Vyklicky, V., Stanley, C., Habrian, C. & Isacoff, E. Y. Conformational rearrangement of the NMDA receptor amino-terminal domain during activation and allosteric modulation. Nat. Commun. 12, 2694 (2021).
37. Stroebel, D., Casado, M. & Paoletti, P. Triheteromeric NMDA receptors: from structure to synaptic physiology. Curr. Opin. Physiol. 2, 1–12 (2018).
38. Borza, I. & Domany, G. NR2B Selective NMDA Antagonists: The Evolution of the Ifenprodil-Type Pharmacophore. Curr. Top. Med. Chem. 6, 687–695 (2006).
39. Tang, Y. P. et al. Genetic enhancement of learning and memory in mice. Nature 401, 63–69 (1999).
40. Gonda, S. et al. GluN2B but Not GluN2A for Basal Dendritic Growth of Cortical Pyramidal Neurons. Front. Neuroanat. 14, (2020).
41. Sceniak, M. P. et al. A GluN2B mutation identified in Autism prevents NMDA receptor trafficking and interferes with dendrite growth. J. Cell Sci. jcs.232892 (2019) doi:10.1242/jcs.232892.
42. Philpot, B. D. et al. Effect of transgenic overexpression of NR2B on NMDA receptor function and synaptic plasticity in visual cortex. Neuropharmacology 41, 762–770 (2001).
43. Barria, A. & Malinow, R. Subunit-Specific NMDA Receptor Trafficking to Synapses. Neuron 35, 345–353 (2002).
44. Platzer, K. et al. GRIN2B encephalopathy: novel findings on phenotype, variant clustering, functional consequences and treatment aspects. J. Med. Genet. 54, 460–470 (2017).
45. Green, T., Rogers, C. A., Contractor, A. & Heinemann, S. F. NMDA Receptors Formed by NR1 in Xenopus laevis Oocytes Do Not Contain the Endogenous Subunit XenU1. Mol. Pharmacol. 61, 326–333 (2002).
46. Swanger, S. A. et al. Mechanistic Insight into NMDA Receptor Dysregulation by Rare Variants in the GluN2A and GluN2B Agonist Binding Domains. Am. J. Hum. Genet. 99, 1261–1280 (2016).
47. Madry, C., Betz, H., Geiger, J. R. P. & Laube, B. Supralinear potentiation of NR1/NR3A excitatory glycine receptors by Zn2+ and NR1 antagonist. Proc. Natl. Acad. Sci. 105, 12563–12568 (2008).
48. Regalado, M. P., Villarroel, A. & Lerma, J. Intersubunit Cooperativity in the NMDA Receptor. Neuron 32, 1085–1096 (2001).
49. Durham, R. J. et al. Conformational spread and dynamics in allostery of NMDA receptors. Proc. Natl. Acad. Sci. 117, 3839–3847 (2020).
50. Vyklicky, V., Stanley, C., Habrian, C. & Isacoff, E. Y. Conformational rearrangement of the NMDA receptor amino-terminal domain during activation and allosteric modulation. Nat. Commun. 12, 2694 (2021).
51. Kellner, S. et al. Two de novo GluN2B mutations affect multiple NMDAR-functions and instigate severe pediatric encephalopathy. eLife 10, e67555 (2021).

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This paper is a follow up of an earlier paper published by the group (Kellner et al., eLife 2021), which aimed at characterizing the functional properties of two de novo GluN2B mutations in patients suffering from severe pediatric diseases, GluN2B-G689C and -G689S. NMDA receptors (NMDARs) are tetramers composed of two GluN1 and two GluN2 subunits. A single receptor can incorporate either two identical GluN2 subunits (di-heteromers) or two different GluN2 subunits (tri-heteromers), leading to a large diversity of NMDAR subtypes. The main NMDAR subtypes in the adult forebrain are GluN1/GluN2A and GluN1/GluN2B di-heteromers, as well as GluN1/GluN2A/GluN2B tri-heteromers. While the exact proportions of these three subtypes are still contentious, there are evidence that in the adult N1/2A/2B tri-heteromers form the major population of synaptic NMDARs in the adult forebrain. In addition, patients bearing pathogenic mutations are often heterozygous for the mutation, giving rise to mixed NMDARs incorporating one mutated and one intact GluN2 subunit. In their previous paper, Kellner et al. had shown that purely di-heteromeric GluN1/GluN2B-G689C and -G689S mutants display a drastic (> 1,000-fold) decrease of glutamate sensitivity and a decrease of surface expression. In the current paper, the authors characterize the effects of the -G689C and -G689S mutations on N1/2A/2B tri-heteromeric receptors, as well as on mixed di-heteromeric GluN1/GluN2B receptors containing one copy of the wild-type GluN2B subunit and one copy of the mutated GluN2B subunit. They show that one copy of the mutant subunit, either within mixed diheteromeric or tri-heteromeric receptors, is sufficient to decrease drastically glutamate sensitivity, although the shift in glutamate EC50 is not as strong as in pure di-heteromeric receptors (≈ 500-fold). They furthermore explore strategies to counteract the hypofunction induced by these mutations by testing the effect of positive allosteric modulators (PAMs). They show that spermine, a GluN2B-specific PAM, can potentiate the activity of mixed diheteromeric N1/2B but not N1/2A/2B tri-heteromers. However pregnenolone sulfate (a 2A/2B-specific PAM) can potentiate both the activity of mixed diheteromeric and tri-heteromeric NMDAR populations, either in oocytes or cultured neurons.

I have very few major comments to make. The experiments are straightforward and the adequate controls have been made. Here are my two only major comments:

1. About the experiment on cultured neurons. The authors compare the currents of cultured neurons transfected with GluN2B-G689C and -G689S to non transfected neurons. The adequate control is rather neurons transfected with the wild-type GluN2B subunit to even out any phenomenon linked to transfection of the neuron. Given the overexpression that can occur after transfection, the effect of the mutations on the size of NMDAR currents might be even stronger than what the authors show. However in that case PS might not completely rescue mutant NMDAR currents to wild-type levels.
2. How come high concentrations of glutamate (>100µM) produce additional current on wt GluN1/GluN2B (with retention signals) compared to 100 µM glutamate, which is supposed to be saturating? It does not seem to stem from an osmotic effect since 10 mM glutamate does not produce any current on uninjected oocytes. Knowing that this "artefactual" effect might also occur in the mutant receptors, how do you take this effect into account when calculating the glutamate EC50s of the mutants? Given the drastic shift in EC50 produced by the mutant, taking into account this artefact is not going to change the conclusion, but the actual EC50s will be affected.

Minor comments:

1. In the first paragraph of the "Results" section, when describing the design of the constructs used to force a heteromeric stoichiometry in recombinant systems, the authors do as if they had designed the constructs themselves "Briefly, we tagged...are retained in the ER (Fig. 1a)". Please rewrite this paragraph to show that you used constructs that had been previously designed by another group (Hansen et al., 2014).
2. I do not see any evidence of "positive cooperativity" between subunits in ref. 32. Ref. 32, to the best of my knowledge, states that in N1/2A/2B tri-heteromers, the 2A subunit sets the biophysical properties of the tri-heteromer. But there is no account of mixed di-heteromers. In addition, the cooperativity between the glutamate and glycine binding sites is negative.
3. Interpretation of spermine action within the Results section: it is striking indeed to observe that the mutations in the context of a mixed di-heteromer still allow spermine potentiation, while they abolish this potentiation in pure di-heteromers. As rightly said in the discussion, the regain of spermine potentiation in the mixed compared to the pure diheteromers is likely due to a more favorable transduction of spermine signaling to the pore, likely via a higher pH sensitivity of mixed di-heteromers compared to di-heteromers. I would thus avoid the terms of "one single intact interface" for the mixed di-heteromer, since both spermine binding sites are likely intact in this NMDAR configuration. How is pH sensitivity affected in the mixed di-heteromers?
4. In the methods section, the oocyte recording solution (likely Ringer and not Barth) does not contain any potassium. This is probably a typo. Could you correct the composition of your Ringer?
5. There are several typos, especially in the Discussion.

Referees cross-commenting

I overall agree with the comments of reviewers 1 and 2. In particular, I agree that it is pointless to compare the absolute currents of non transfected neurons vs mutant-transfected neurons without an idea of receptor cell-surface expression.

I would like however to give some precisions about some comments of Reviewer 2. About the ER retention technique to express tri-heteromers: I didn't know that the C2 signal could be addressed to the membrane on its own. The lack of leak current stemming from C1-C1 or C2-C2 combinations has been demonstrated in the paper establishing the technique (Hansen et al, 2014), as well as in another paper that developed an analog technique based on GABAB retention signals (Stroebel et al., J Neurosci 2014). So it is fair to consider that the authors were not surprised by the lack of current when co-expressing two GluN2B subunits carrying the C2 signal. About the comparison about absolute currents wt vs mutants, +/- spermine (Fig. 4a and 5a). I agree with reviewer 2 that being able to compare absolute currents of wt without spermine to mutant + spermine would be very interesting to see if spermine can actually rescue mutant hypofunction. However, to the defense of the authors, comparing absolute current values of recordings from Xenopus oocytes is meaningless. Indeed the variability of currents for the same construct and same day of experiment is too high (there can be up to a ten-fold difference between the lowest and the highest current of oocytes expressing the same construct the same experimental day). A way to investigate this aspect would be to estimate the open probability of the different constructs with or without spermine via the inhibition kinetics of an open channel blocker (e.g. MK801) and measure surface expression by Western blot, but I am not sure these experiments are worth it for the spermine experiment.

Significance

This paper is not of high significance since most of the characterization of the 2B-G689C and -G689S de novo mutants found in patients has already been published (Kellner et al., eLife 2021). However, this paper is worth publishing since it brings new data on the effect of the mutations on tri-heteromeric and mixed di-heteromeric NMDAR populations, which are likely the most abundant NMDAR populations in the patient's brain at adult stage. Tri-heteromeric and mixed NMDAR populations have often been overlooked when studying pathogenic NMDAR mutations due to the difficulty to express them specifically in recombinant systems. This paper (in addition to other papers in the field, see for instance Elmasri et al., Brain Sci. 2022; Li et al., Hum. Mutat. 2019) shows that the effect of the mutations on the receptor biophysical and pharmacological properties (but also on trafficking) differ whether the receptor contains one or two copies of the mutant subunit. This paper is of interest to scientists interested in NMDA receptor structure-function and pharmacology, as well as clinicians interested in GRINopathies (pathologies linked to NMDAR mutations). I, the reviewer, am an expert in NMDAR structure-function and pharmacology. I believe I have sufficient expertise to evaluate the entirety of the paper.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The objective of this paper is to assess whether a single mutated subunit of GRIN can affect the function of various forms of NMDA receptors. In particular, this study investigates the functional consequences of a GRIN variant when assembled within tri-heteromers, containing 2 GluN1, 1 GluN2A and 1 GluN2B subunits, the major postnatal receptor type. For this purpose, the authors artificially forced the subunits to associate in predefined complexes, using chimeras of GRIN subunits fused to GABAb receptor retention control sites at the endoplasmic reticulum. This trick allows to control the stoichiometry of the channels at the membrane and thus to focus on the function of a single type of NMDA receptor. The take home message of the paper is that a single GluN2B‐variant, whether assembled with a GluN2B‐wt subunit to form mixed di‐heteromer or with a GluN2A‐wt‐subunit (tri‐heteromer), strongly impairs the receptor functioning, as reported by a decrease of the apparent glutamate affinity of the receptor.

Altogether, this is a straightforward study of great interest for the GRIN community. However, the way the background and purpose of the study (title and abstract) are presented is a bit confusing for non-specialists and could be easily improved. Technical information, which is crucial to validate the conclusions drawn from data analysis, should be added to the article. Some additional experiments are suggested to consolidate the work. Finally, additional discussion points are strongly encouraged.

Specific comments

Abstract / Title:

This work shows that a single GRIN variant impairs the function of various forms of NMDA receptors. Several sentences in the title and the abstract are confusing for a non-specialized audience. "Two extreme Loss‐of‐Function GRIN2B‐mutations are detrimental to triheteromeric NMDAR‐function, but rescued by pregnanolone‐sulfate." "Here, we have systematically examined how two de novo GRIN2B variants (G689C and G689S) affect the function of di‐ and tri‐heteromers." The number of variants tested is not of capital importance in the title, especially because one could believe that both are tested at the same time; similarly, when variants are named in the abstract, the fact that only 1 variant is studied at a time should be clarified (G689C OR G689S). Indeed, the problem is obvious to those familiar with GRIN disorders, but if this paper is to be published in a journal reaching a large audience rather than a specialized audience, the title of the paper should be modified.

Introduction:

"We find that the inclusion of a single GluN2B‐variant within mixed di‐ or tri‐heteromeric channels is sufficient to prompt a strong reduction in the receptors' glutamate affinity, but these reductions are not as drastic as in purely di‐heterometric receptors containing two copies of the variants. This observation is supported by the ability of a GluN2B‐selective potentiator (spermine) to potentiate mixed diheteromeric channels." Please, clarify the link between the two sentences. How do spermin potentiation of mixed diheteromeric channels supports the observation that the inclusion of a single GluN2B‐variant has less effect than the inclusion of two variants?

Methods

All this study is based on the use of a unique ER‐retention technique to limit expression of a desired receptor‐population at the membrane of cells. According to the ER system retention of GABAb receptor, used in this study, while C1/C1-fused subunits are retained in the ER, C2/C2 reach the cell surface and the association of C1/C2 in the ER enables cell-surface targeting of the heterodimer. However, GB2 does not contain any retention signal and can reach the cell surface in the absence of GB1, as a functionally inactive homo-dimer (doi: 10.1042/BJ20041435). If there is an experimental trick that prevents the addressing of C2/C2 to the cell membrane, it should be specified and explained. This is critically important for understanding which receptor populations the data are derived from: receptors containing C1/C2 fused subunits only as stated by the authors, or C1/C2 and C2/C2 fused subunits?

Figures

NMDA-receptor current amplitude should be normalized by the membrane expression of the receptors. A preliminary experiment should measure the effective cell surface expression of each of the subunits in the different transfection conditions.

Fig.1a - The scheme should include C2-C2 complexes and mention whether these complexes are expressed at the cell surface (see previous and following comments).

Fig.1b and c - Current from cells transfected with GluN2B‐wt‐C1 and GluN2B‐wt‐C2 should be compared to current expressed in cells expressing untagged receptor subunits: GluN2B‐wt - Current from cells transfected with GluN2B‐wt‐C1 alone should be shown as well (although expected to be retained in the ER) (as performed for GluN2A‐wt‐C1 GluN2B‐wt‐C1 in suppl Fig. 1a) - How could you explain the null current from cells transfected with GluN2B‐wt‐C2 alone (Fig.1b middle, and 1c)? since GB2 does not contain any retention signal and can reach the cell surface in the absence of GB1, GluN2B‐wt‐C2 is supposed to reach the cell surface. This is a very important point to clarify (I am probably missing a technical detail) because if the sub-unit tagged with C2 does reach the cell surface, then all the results and conclusions drawn from the C1-C2 conditions are wrong and could be attributed to a mix of complexes containing either C1-C2 or C2-C2. My following comments are based on the assumption that only receptors containing C1-C2 tagged subunits reach the membrane (as assumed by the authors and suggested in Figure 1b middle), but explanations should absolutely be provided to convince the reader.

Fig. 4a and 5a - Please, keep the current scale constant between all current illustrations within the same figure (4a and 5a). Indeed, not only the Spermin- or SP- induced potentiation is an important data (which is presently quantified on the histograms of fig. 4b and 5b) but also knowing whether the amount of current recorded in cells expressing one mutant subunit in presence of SP (for example GluN2A‐wt‐C1 GluN2B‐G689S‐C2) is comparable to the current recorded in wt receptor-expressing cells (GluN2A‐wt‐C1 GluN2B‐wt‐C2) in absence of SP would be an excellent added value for the paper. A special figure could quantify this rescue effect of SP, measuring and comparing the mean currents recorded in these conditions (one current illustration is not sufficient given variations between similar samples). By the way 5mM glutamate might be an excessive concentration. At 1mM, the expected synaptic concentration of glutamate following action potential, according to figures 3 and Suppl1 the response of the mutated receptor is much lower than that of the WT which is already almost maximal. In these conditions, SP-induced potentiation by a factor of two of GluN2A‐wt‐C1 GluN2B‐G689S‐C2 current could be equivalent to control currents recorded in GluN2A‐wt‐C1 GluN2B‐wt‐C2 cells.

Fig. 6 - Figure 6 is not convincing because cultured hippocampal neurons do express endogenous NMDA receptors. To what extent the recording currents are affected by endogenous, non-mutated GluN2B subunits? Western Blots showing an extinction of endogenous subunits expression when transfected tagged subunits are competitively expressed would be required. - Fig.6b "PE-S" on the graph should be replaced by "PS"

Discussion

The authors are surprised by the fact (Fig.2) that 1 variant reduces the apparent glutamate affinity of the receptor, but not as much as 2 variants, despite the fact that "NMDARs opening requires all four subunits to be liganded (i.e., occupied by a ligand) which implies that the least affine subunit should have dominated the final affinity of the receptor". I agree that the difference is noticeable, however the glutamate affinity for receptors containing 1 variant is much closer to that of receptors containing 2 variants than that of wild-type receptors. Hence, the results obtained do not seem so surprising and could result, as rightly explained by the authors from a possible cooperativity between the subunits.

On the other hand, the data in Figure 6 are much more difficult to interpret and reconcile with the nature of the expressed receptor subunits (which this time is not controlled) nor their association within the same receptor. However, this aspect, which is essential to the understanding of the consequences of 1 variant on neuronal signalling, is not discussed: Whatever the stoichiometry of the complexes in the heterozygous disease, the mutated and wild type GluN2B subunits coexist in the same cell: Either within the same di-heteromeric complexes GluN2B-wt + GluN2B-mutant, or in separate complexes but nevertheless expressed in the same cell, in di heteromeric (GluN2B-wt + GluN2B-wt and GluN2B-mutant + GluN2B-mutant); or tri-heteromeric (GluN2A-wt + GluN2B-wt and GluN2A-wt + GluN2B-mutant) complexes. Assuming that half of the complexes remain wild-type, e.g. (GluN2A-wt + GluN2B-wt and GluN2A-wt + GluN2B-mutant) we would expect (Fig. 6) a small decrease in NMDA current (carried only by the half that expresses the mutated subunit, and whose function is not zero but only decreased by about 20% in response to 5mM Glutamate, Fig. 3b). The same reasoning applies to the di-heteromeric conditions (GluN2B-wt + GluN2B-wt; GluN2B-mutant + GluN2B-mutant), here again the decrease observed Fig. 6b is difficult to reconcile with the responses measured Fig. 2b. In other words, how to explain a 50% decrease of the currents, instead of the 10% expected by the previous reasoning. In this experiment we do not know which subunits are expressed, their proportions, nor how they are associated in functional complexes, which makes the interpretation of the data impossible. The only explanation, far-fetched, for 50 % decrease would be that the complexes were to contain all (or the vast majority) 1 wild-type subunits associated with 1 variant, then a homogeneous 50% reduction in current could be expected. But this extreme condition could only be possible in the case of di-heteromers, which is unlikely the case in Fig.6 as GluN2A currents are measured in presence of Ifenprodil. To conclude 1) the comparison of the currents in transfected and non-transfected neurons does not make sense in figure 6b which is not convincing because we do not know the nature of the currents actually measured. A comparison in controlled condition would make more sense (as I suggested in the criticism of figures 4, 5). 2) The reality of the combinations of expression and association between subunits within different complexes expressed in the same cell must be considered and taken into account in the interpretation of the data. Undoubtedly, the means of restoring the NMDA current will be different depending on the presence of mutated subunits in all functional channels or not.

Referees cross-commenting

The referees' comments are highly relevant. In particular, referee 3's comment 1 seems very interesting because it may help to better understand the discrepancy in the results observed in neurons, i.e. a 50% decrease in the currents induced by the expression of the mutant and wild type subunits in the same cells, whereas theoretically one would expect a 10% decrease of this current (cf. referee 2's 2nd comment in the discussion section). This comment 1 of referee 3 indeed stresses the fact that the control (non-transfected neurons) to which the heterozygous condition is compared is not the correct control, which should rather be neurons transfected with wild type receptor subunits. More generally, this comment underlines the importance of monitoring the effective membrane expression of the different subunits in each of the experimental conditions in order to be able to compare conditions and draw conclusions.

Significance

The novelty of the study, is to evaluate the consequences of a single mutated subunit within NMDA receptors affected by GRIN variant, to mimic the heterozygous condition of GRIN encephalopathies, this is of potential value for the field and the interest could also be extended to other genetic diseases (at least the experimental way to study the functioning of only one desired stoichiometric configuration).

The strength of this paper is precisely to isolate technically and to study the functioning of a desired stoichiometric configuration only. The main limitation of the paper is the interpretation that the authors make of their data in a physio-pathological context

This work could be of interest for general audience, providing the title and summary are slightly modified My area of expertise could not be closer to the topic of the article: Glutamate receptors; GRIN; molecular tinkering, cell culture, electrophysiology, receptor stoichiometry...

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:

Kelner and Berlin present their research findings pertaining to the effect of GRIN2B variants that modify NMDA receptor function and pharmacology. While these mutations were published previously, the new manuscript provides a more thorough investigation into the effects that these variants pose when incorporated into heteromeric complexes with either wildtype GluN2B or GluN2A - NMDA receptors containing only a single mutated GluN2B subunits is more relevant to the disease cases because the associated patients are heterozygous for the variant. The authors achieved selective expression of receptor heteromeric complexes by utilising an established trafficking control system. The authors found that while a single variant subunit in the receptor complex is largely dominant in its effect on reducing glutamate potency of the NMDA receptor, it 's effect on receptor pharmacology varied. Unlike diheteromeric receptors containing mutated subunits, polyamine spermine potentiated GluN1/2B (but not GluN1/2A/2B) receptors that contained a single mutated GluN2B. In contrast, the neurosteroid, pregnenolone-sulfate (PS), was effective at potentiating the NMDA receptor currents (to varying degrees) regardless of the subunit composition. The potentiation of NMDA receptor currents by PS was also observed in neurons overexpressing the variants.

The techniques used in this study were appropriate to address the objectives and the overall effects are large, and generally convincing. I like the way the results are presented, although have a few (easily addressable) comments.

Major comments:

• When incrementally adding drugs (e.g. traces in figures 5 and 6), it doesn't always appear like the response has plateaued before changing the solutions/drugs. Therefore, I am curious to what extent the effects observed are underestimated.
• Also, in relation to figure 6, to what extent does agonist application cause desensitization here? Looking at traces in Figure 6b it appears that there is some desensitization and it isn't clear to what extent this persists during the solution changes. Could the authors conduct/show the controls where NMDA alone (for 50-60s), or NMDA followed by PE-S (without ifenprodil).
• Finally, figure 5 shows the effect of the neurosteroid (and ifenprodil) on NMDA-evoked currents in neurons overexpressing the GluN2B variants in neurons. However, there currents probably reflect a mixture of extrasynaptic and synaptic receptors. To what extent are synaptic NMDA receptors affected by the variants?

Minor comments:

• Looking at the fits in the graph of Figure 2b it appears that the slope on the concentration response curves is less steep for the mixed 2B-diheteromeric NMDA receptors. How much are the Hill coefficients changing and can this be interpreted to provide more mechanistic insight? Wouldn't it make sense to include the Hill coefficients in Table 1?
• The authors illustrate the changes in potency by the shift in the concentration response curves, but is there any change in efficacy? A simple way to illustrate this would be also present a simple graph showing the maximum current amplitudes (i.e. to 10 mM glutamate) for each of the receptor complexes.
• The authors characterize the 'apparent' affinity (or potency) of the receptor using concentration-response curves, but numerous points in the manuscript refer to changes in affinity. None of the experiments shown directly measure affinity (which would require ligand-binding assays) and so the use of the word affinity is inaccurate/misleading. I suggest the authors replace the instances of the word 'affinity' with 'potency'.
• In the third line of the abstract, the authors wrote, 'for which there are no treatments' in relation to GRINopathies. My understanding is that there are symptomatic treatments but that there are no disease-modifying treatments.
• The authors have interchangeably used the terms NMDAR or GluNRs throughout the manuscript. I suggest sticking to one of these terms. I would suggest NMDARs since this is less likely to be misread as a a specific NMDA receptor subunit..
• Typos: 
1) Results paragraph 2 sentence one: 'We thereby produced GluN2B-wt, GluN2B-G689C and GluN2B-G689S subunits tagged with C1 or C2, co-expressed these along with the GluN1a-wt subunits in...'
2) Results paragraph 2: '...but these were mainly noticeable when oocytes are were exposed to high (saturating) glutamate concentrations...'
3) Last sentence in the second to last paragraph of the results section entitled 'Mixed di-and tri-heteromeric channels...': 'This , PS may serve to rescue...'
4) Last sentence in last paragraph of the results section entitled 'Mixed di-and tri-heteromeric channels...': 'Despite the latter, we found no evidence for any direct effect of three different physiologically relevant concentrations of the drug on di- or tri-heteromeric receptors'
• Figures 1e, 2b, 3b: it would be helpful to add a legend to the graph so that the curves can be interpreted without having to read through the figure legend.
• The bar graphs in Figure 6 show individual data points but those in figures 4b and 5b don't. Can the authors please add the data points to these graph.
• It would be helpful to reviewers that future manuscripts by the authors include page numbers and line numbers

Referees cross-commenting

Reviewers 2 and 3 highlight an important issue concerning figure 6 and the extent to which the overexpressed variants subunits can compete and assemble with endogenous NMDA receptors (unlike the system where the surface expression of specific receptor complexes is controlled). Indeed in the recent paper by the same authors, the two variants differed in their surface expression (in HEK cells), with G689C expressing particularly poorly. With reference to the second minor comment of Reviewer 1, the maximum current amplitudes would of course need to be normalized to cell surface expression of the receptor to gain any insight into efficacy.

Significance

This study emphasizes the complex pattern of effects that variants can have on glutamate receptor function and pharmacology, especially considering the context of receptor subunit composition. The authors have followed up their previous findings on the same mutants (Kellner et al, 2021, Elife), but used a trafficking control system here to characterize properties of mutated receptor complexes that are most likely to exist in neurons. The authors show that the defective currents mediated by NMDA receptors containing a loss-of-function GluN2B variant can be enhanced by neurosteroids (and in the case of GluN1/2B receptors, polyamines also). Development and approval of neurosteroids for the clinic would be required for the findings to translate to a therapy for patients. Readers should also be aware that neurosteroids act on other receptors too (e.g. GABA receptors), which could complicate the outcome. The expertise of the reviewer is in glutamate receptors and synaptic transmission.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Major:

- The statement (line 149'Together, our data suggest that systemic ecdysone levels are unlikely to be involved in modulating tumour-induced muscle detachment or to mediate the role of fatbody Insulin signalling in regulating muscle detachment.') is derived from an experiment with sterol free diet (in which 20HE is genetically addressed) and a pleiotropic experiment (PG>RasG12V). In neither paper nor the current manuscript, 20HE levels have been directly addressed.

Therefore, this statement needs further experimental support and discussion. Ecdysone is a critical hormone during development and especially growth-related effects central to this study. The authors should consider doing pharmacology or augment their claims here with genetic manipulation experiments of 20HE related genes in larvae (Leopold, Rewitz, Rideout, Drummond-Barbosa, Schuldiner labs) and adult animals using genetics, pharmacology or direct assessment of 20HE levels (RIPA, Edgar and Reiff labs).

The main point we were trying to convey is that we do not think global ecdysone levels plays a role in modulating fatbody insulin or tgfb signalling, which in turn affects muscle detachment. We are not claiming that edysone levels is not changing in control vs. tumour bearing animals. In fact, we predict that 20HE levels will be different in tumour bearing vs. control animals (as tumour bearing animals undergo developmental delay), but this is not the main point of our conclusions. We believe that our conclusions are supported by the experiment demonstrating global ecdysone alterations (via feeding sterol-free food) did not affect how fatbody Akt activation altered tgfb signalling and enhanced muscle integrity (Figure S1). Therefore, we don’t think measuring 20HE helps to support our conclusions. Pharmacological inhibition via feeding ecdysone inhibitors effectively demonstrate a similar point to feeding sterol-free food which we have already performed. We are happy to try direct manipulation of 20HE related genes (eip75B-RNAi) in the fatbody to see if this affects muscle detachment or pAkt and pMad levels in tumour bearing animals.

- In Fig.7 the authors used a sog-LacZ stock to show transcriptional activation in fatbody cells. This stock is based on P-element insertion in the according regulatory regions and supposed to express lacZ with an nls. I can clearly see lacZ in nuclei in Fig. 7H, whereas this is very hard to see in nuclei in Fig7i in the tumour model. In addition, lacZ is known for its high stability and not the best option. As this finding is vital for central claims of this study, it should be complemented by either qPCR for sog on fat body cells or using another readout by converting one of the two Mimic lines (BL42189/44958) into GFP sensors for sog.

We will add a counterstain to these images. We will also perform qPCR in the fatbody of control and cachectic animals to assess whether Sog transcription is altered. We agree converting one of the Mimic lines to a GFP sensor would be a good option, but this experiment would require getting new fly lines into Australia, which takes at least 2 months because of quarantine laws. We don’t believe this experiment would change the general conclusions of the paper, therefore would prefer not to do this experiment.

- I have similar problems with Fig.7B-F, as phosphorylated Mad should be translocated to the nucleus. In 7F the authors measure pMad over Dapi, which is the right way but it is hard to see pMad in the nucleaus apart from Fig7B, wheras in D and E, where the authors measure higher levels, I cannot identify clear pMad in nuclei. These images either need to show the Dapi channel or more representative images should be chosen like in Fig.4 with arrows pointing to measured nuclei. Fig.7C something went wrong with the compression of this image.

We will show more representative examples and fix Fig 7C.

- The proper function of RNAi stocks targeting genes like sog, mad, etc. is vital for this study as these lines are used throughout the study. Functional evidence of specific knockdown efficiency should be provided or references given in which these stocks were shown to provide functional knockdown on transcript or protein level.

We agree with the reviewer that this is an important point. We will demonstrate the knockdown of sog and mad (and other RNAis) used in the study by either referring to published data or demonstrate knockdown ourselves.

- Fig.S7 discusses appearance of gbb/Bmp7 and Sog/CHRD in human patients. The analysis the authors performed shows a correlation between both factors, but is hampered by the fact that datasets for peripheral tissues of cachexia patients are unavailable. The authors may consider sorting these after tumor entities in which cachexia occurs frequently vs. low occurrence and then check for both genes.

We will try this analysis.

Fig.5 M-P pMAd is not indicated in the Panels only the legend.

We will fix this error.

- Please follow FlyBase nomenclature, e.g. dlg1 for discs large 1 and unify in the whole manuscript and figure for all genes.

We will fix this error.

- For endogenous fusion proteins like Viking-GFP (e.g. vkg::GFP) choose a format to clearly decipher them from transcriptional readout stocks like sog-lacZ.

We will fix this error.

- The quantifications in most figures are quite small with tiny lettering and XY axis are difficult to read in letter/A4 size.

We will enlarge font size.

Minor:

1. Adjust in-figure caption alignments

2. Line 104: add comma RasV12, dlgRNAi

3. Line 114: replace little  not significant (n.s.)

4. Line 334: 'sogRNAi overexpression' to my knowledge, RNAi are expressed, not overexpressed.

5. Line 454: italicize r4>

6. Fig S4E: remove frame

7. Figures 6: It would be better to number and explain the pathway presented in the figure in text and fig legend.

8. Just a personal preference. Lettering of images in images is commonly done horizontally, here it appears like a mix between vertical and horizontal.

We will fix these minor errors.

Reviewer #2

Major comment

Their genetic experiments clearly showed that the reduction of insulin signaling activity in the fatbody induces upregulation of TGF-β signaling and Collagen accumulation. Then, how does TGF-β signaling induce Collagen accumulation?

From the experiments we have carried out, we do not have insights into how TGF-B signalling induce Collagen accumulation.

They showed that Rab10 knockdown and SPARC overexpression reduced the accumulation of fatbody ECM. Are Rab10 and SPARC expression regulated by TGF-β signaling?

We can address this point by assessing if Rab10 and SPARC expression is altered in cachectic fatbody.

Minor comments

Line 90: "Disc Large (Dlg) RNAi in the eye" must be "Discs Large (Dlg1) RNAi in the eye imaginal discs".

we will fix this error.

Figures 1D and 1L are from the same image. Also, Figures 1C and 1M are from the same image. Are both of them necessary to be shown in the different panels?

The duplication of 1C and 1M, was an error, we thank the reviewer for picking this up. We will fix this error. We will use different images for 1D and 1L.

Why are the staining patterns of anti-pAkt shown in Figures 1L and 1U so different? pAkt is not detected in the nuclei in Fig. 1L but its nuclear signal is clear in Fig. 1U.

We will show more representative images of these staining.

Figure 1: Images of counter staining for nuclei like DAPI should be also included for all these fatbody images.

We will show counter staining for DAPI.

Line 101: "Tumour specific ImpL2 inhibition was sufficient to reduce fatbody pAkt levels." Is this correct? ImpL2 inhibition in tumors should elevate the pAKT level in fatbody.

This was a mistake, we will fix this error.

Figure S1~S4: These figures and their legends do not correspond to each other. We thank the reviewer in picking up this error, there was an error in inserting the images into the text. S2 and S3 were swapped.

We will fix this error.

Line 189: The pAkt level in the muscle of tumour-bearing animals should be examined to confirm the activity of the insulin signaling is downregulated.

We will include this data.

Line 189: If the authors conclude that muscle insulin signaling predominantly regulates translation and atrophy, OPP assay for the muscle cells should be examined in the same experimental settings.

We will carry out OPP assay upon Akt overexpression in the muscle.

Line 247: The expression level of Rab10 and SPARC should be examined in the fatbody of tumour-bearing animals to see whether Rab10 is upregulated and SPARC is downregulated.

Line 247: If Rab10 upregulation and SPARC downregulation are the causes of the accumulation of ECM proteins in the fatbody of tumour-bearing animals, how the overexpressed Collagen proteins can be secreted from the fatbody cells?

We are not sure, but the overexpression of Collagen proteins is at an extremely high level, therefore, it is possible that some of it can be processed and secreted despite Rab10 upregulation and SPARC downregulation. We have carried out an experiment to overexpress Collagen proteins in the muscle, in this case, this manipulation did not rescue. This indicates that processing of Collagen in the fatbody is important, however, we do not know how the processing is regulated.

Line 347: Sog is a secreted BMP antagonist. Thus, it can be expected that the Sog overexpression downregulates TGF-β signaling in fatbody and muscle tissues. If the rescued phenotypes with Sog overexpression can be explained by this logic, pMad level should be examined in these experiments.

We have shown this data in Figure R-T. We will refer back to this data in Line 347.

Reviewer #3

Major comments:

- Are the key conclusions convincing?

Most of the conclusions are convincing. It is not clear however whether the ECM accumulation in the fat body of tumor animals is fibrotic and whether it is extracellular or in the cell cortex.

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

-The authors state in line 71 'This deposition of disorganized ECM leads to fibrotic ECM

accumulation.' The authors haven't really provided evidence for the ECM being fibrotic. The authors could either rephrase this or provide additional experimental evidence of fibrosis in the fat body.

We will tone down the claim that the ECM accumulation is fibrotic.

- Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

-The authors state in line 147" Finally, in tumor-bearing animals fed a sterol-free diet, that underwent a prolonged 3rd instar stage due to reduced ecdysone levels (Parkin and Burnet, 1986), we activated insulin signalling in the fatbody via Akt overexpression (QRasV12, scribRNAi). We found that this manipulation caused a significant decrease in pMad levels in the fatbody and a rescue of muscle detachment (Figure S1 D-I), similar to animals fed a standard diet (Figure 1 O-Q, Figure 2 F-H)." Since it's not already known what the extent of muscle integrity defect there is in tumors with additional sterol free diet, it would be important to show a non-tumor control for comparison in FigS1F. This would also then make it clear to what extent the defect is rescued by Akt overexpression.

We will include a non-tumour control for Fig S1F.

-The authors state in line 158 'Upon the knockdown of Impl2, we found that tumor gbb was not significantly altered (Figure S3A).' Even though this shows an indication that Gbb levels are not reduced, the n number is too low to state that it is non-significant. The authors should increase the n number here.

N=3 is generally enough to see a difference, we will include data done in parallel which shows Impl2 RNAi is sufficient to induce a reduction in Impl2 RNA levels. This will demonstrate that n=3 is sufficient to demonstrate a reduction in transcript levels if there is a reduction.

-The authors state in line 171 'Conversely, knockdown of gbb alone or knockdown of gbb together with ImpL2 significantly rescued the Nidogen overaccumulation defects observed at the plasma membrane of fatbody from tumor-bearing animals, while ImpL2RNAi alone did not (Figure S2 Q-U).' This is a somewhat misleading representation, since again no non-tumor control was used, so the extent of the rescue by gbb knowdown is not obvious. In FigS2P Nidogen levels in the tumor seem ~100% higher than in control. But in FigS2U, in which no control was included, the tumor+gbb knowdown seems ~ 20% lower than tumor. So it is probably a more moderate rescue, but that's only possible to assess by including a non-tumor control in FigS2U. Also the images in FigS2Q-T don't seem representative since they appear to show a much bigger difference in fluorescence intensity than ~20%. Please show more representative images.

We will include a non-tumour control for S2Q-T and show more representative pictures.

-The authors state in line 174 'Finally, co-knockdown of gbb and ImpL2 in the tumor significantly rescued the reduction in OPP and Nidogen levels observed in the muscles of tumor-bearing animals (Figure S3 B-I).'

Again, the single knockdowns and the non-tumor control are not shown in FigS3E and I and should be included for comparison and to see the contribution of each knockdown and to be able to judge the extent of the rescue.

We will include the single knockdowns and a wildtype control

-Regarding Fig3O: Is there a significant tumor muscle attachment defect here? In this graph the tumor only looks about 10% lower than the WT (rather than 40% in Fig2E). The other issue is the extremely low n number for WT. I would recommend increasing the n number for WT here and to indicate in the graph whether the tumor is significantly different to WT (or non-significant, in which case RabRNAi wouldn't actually 'rescue' the defect). In the present form, this graph is not very convincing.

We will increase the n number for WT for this experiment. The reduction in muscle detachment is 10% rather than 40% here is because this experiment was done at day 6, which we will indicate in the figure legend. The 40% reduction in Fig2E is because these samples were processed at day7. Rab10RNAi experiment was carried out at day 6, because by day7, the Rab10RNAi rescue is so good, most of the tumour bearing animals have pupated, thus the experiment could only be carried out at day6.

- Regarding Fig3W: A non-tumor control would be important to include to be able to judge the extent of muscle attachment defects and the extent of the rescue for UAS-Sparc. This will allow to assess the severity of muscle integrity defect in this particular experiment (since it appears to vary in different experiments e.g. muscle defect in tumor 40% in Fig2E and ~10% in Fig3O) and to assess the extent of rescue for the various genotypes.

We will include a non-tumour control for 3W.

-The authors show an accumulation of ECM in the fat body of tumors. It is not clear, whether this ECM accumulates intracellularly near the cell surface or extracellularly. The authors should assess this, maybe by doing electron microscopy.

We do not have an EM facility that can accommodate this experiment, thus doing EM is not an option for us. However, we can address whether the accumulation of ECM is intracellular or extracellular by performing an experiment, where we try perform antibody staining against Viking-GFP without permeabilizing the cells. If Viking is detected without permeabilization, it would indicate the accumulations are extracellular. This approach has been previously used to address this question in Zang et al., elife, 2015.

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

-These suggested experiments should be quite straightforward since they are mostly just repeating previous experiments with the appropriate controls and n numbers. I would think that they can be done within a few months. The electron microscopy should not take more than a few weeks and not be costly.

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

-The details on how old animals used in each experiment were, are not easy to find and not written very clearly. They should be included in the each figure legend rather than summarising those details in the methods.

We will add the number of days in the figure legend.

-Also, in line 788 in the methods, several stocks are indicated as coming from particular labs (e.g. UAS-FOXO (Kieran Harvey), UAS-GFP (Kieran Harvey), UAS-lacZRNAi (Kieran Harvey), UAS-RasV12 (Helena Richardson), UAS-cg25C;UAS-Vkg (Brian Stramer)).

However, it is not clear whether these labs actually made these stocks and if so whether it has already been described in their papers how the lines were made. If the lines are unpublished, the detailed information should be given on how the lines were made. Or if the lines are published, the authors should provide the reference.

We will fix these references.

- Are the experiments adequately replicated and statistical analysis adequate?

In general, the n number is rather low in several experiments, especially n of 3 for many controls. And as I mentioned before, rescues of tumor phenotypes are often shown without including a non-tumor control, making it hard to judge the extent of the rescue. Sometimes this information can be found in other figures, but the reader should not have to search for it. And also the severity of the phenotype can vary from experiment to experiment.

We will include a non-tumour control when appropriate to address this.

Minor comments:

- Specific experimental issues that are easily addressable.

- Are prior studies referenced appropriately?

Yes, as far as I can tell.

- Are the text and figures clear and accurate?

-In the literature, people usually call it 'fat body' rather than 'fatbody'.

We will fix this error.

-The authors state in line 265 "Vkg accumulated in the membranes of fatbody where p60 was overexpressed using r4-GAL4 (Figure 5 A-C)."

This must be a typo. I think it is shown in Fig5E-G. Unless it's labelled wrongly in the figure and B, C and D show p60 rather than TorDN.

We will fix this error.

-The authors state in line 188 'This manipulation significantly rescued muscle integrity (Figure S4 A-C) and muscle atrophy (Figure S4 D-F), without affecting muscle ECM levels (Figure S4 G-H).' According to the graph in FigS4H this does actually 'affect muscle ECM levels' significantly, as in that it reduced Nidogen levels further. The authors could rephrase this.

We will reword this statement.

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:

Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

This paper uses a Drosophila tumor model induced by the expression of RasV12+Scrib-IR or RasV12+Dlg-IR in the eye imaginal disc to understand how inter-organ communication affects cachexia in the fat body and muscle. The tumor has previously been shown to secrete the factors ImpL2 and Gbb which decreases insulin signalling and increases TGF-beta signalling in the fat body, respectively, and results in fat body and muscle defects. Here they dissect the role of insulin and TGF-beta signalling in the fat body in regulating muscle integrity further. They show that these two pathways converge via Sog in the fat body of tumor-bearing animals and result in aberrant ECM accumulation in the fat body which hinders ECM secretion. This then results in the muscle receiving less fat body-derived ECM which causes muscle attachment defects. Interestingly, these muscle defects can be ameliorated by activating insulin signalling or inhibiting TGF-beta signalling or even by increasing ECM secretion in the fat body. The authors also provide some evidence that the insulin and TGF-beta signalling pathways can converge in non-tumor settings.

Major comments:

• Are the key conclusions convincing?

Most of the conclusions are convincing. It is not clear however whether the ECM accumulation in the fat body of tumor animals is fibrotic and whether it is extracellular or in the cell cortex.<br /> - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?<br /> - The authors state in line 71 'This deposition of disorganized ECM leads to fibrotic ECM<br /> accumulation.' The authors haven't really provided evidence for the ECM being fibrotic. The authors could either rephrase this or provide additional experimental evidence of fibrosis in the fat body.<br /> - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.<br /> - The authors state in line 147" Finally, in tumor-bearing animals fed a sterol-free diet, that underwent a prolonged 3rd instar stage due to reduced ecdysone levels (Parkin and Burnet, 1986), we activated insulin signalling in the fatbody via Akt overexpression (QRasV12, scribRNAi). We found that this manipulation caused a significant decrease in pMad levels in the fatbody and a rescue of muscle detachment (Figure S1 D-I), similar to animals fed a standard diet (Figure 1 O-Q, Figure 2 F-H)." Since it's not already known what the extent of muscle integrity defect there is in tumors with additional sterol free diet, it would be important to show a non-tumor control for comparison in FigS1F. This would also then make it clear to what extent the defect is rescued by Akt overexpression.<br /> - The authors state in line 158 'Upon the knockdown of Impl2, we found that tumor gbb was not significantly altered (Figure S3A).' Even though this shows an indication that Gbb levels are not reduced, the n number is too low to state that it is non-significant. The authors should increase the n number here.<br /> - The authors state in line 171 'Conversely, knockdown of gbb alone or knockdown of gbb together with ImpL2 significantly rescued the Nidogen overaccumulation defects observed at the plasma membrane of fatbody from tumor-bearing animals, while ImpL2RNAi alone did not (Figure S2 Q-U).' This is a somewhat misleading representation, since again no non-tumor control was used, so the extent of the rescue by gbb knowdown is not obvious. In FigS2P Nidogen levels in the tumor seem ~100% higher than in control. But in FigS2U, in which no control was included, the tumor+gbb knowdown seems ~ 20% lower than tumor. So it is probably a more moderate rescue, but that's only possible to assess by including a non-tumor control in FigS2U. Also the images in FigS2Q-T don't seem representative since they appear to show a much bigger difference in fluorescence intensity than ~20%. Please show more representative images.<br /> - The authors state in line 174 'Finally, co-knockdown of gbb and ImpL2 in the tumor significantly rescued the reduction in OPP and Nidogen levels observed in the muscles of tumor-bearing animals (Figure S3 B-I).'<br /> Again, the single knockdowns and the non-tumor control are not shown here in Fig3E and I and should be included for comparison and to see the contribution of each knockdown and to be able to judge the extent of the rescue.<br /> - Regarding Fig3O: Is there a significant tumor muscle attachment defect here? In this graph the tumor only looks about 10% lower than the WT (rather than 40% in Fig2E). The other issue is the extremely low n number for WT. I would recommend increasing the n number for WT here and to indicate in the graph whether the tumor is significantly different to WT (or non-significant, in which case RabRNAi wouldn't actually 'rescue' the defect). In the present form, this graph is not very convincing.<br /> - Regarding Fig3W: A non-tumor control would be important to include to be able to judge the extent of muscle attachment defects and the extent of the rescue for UAS-Sparc. This will allow to assess the severity of muscle integrity defect in this particular experiment (since it appears to vary in different experiments e.g. muscle defect in tumor 40% in Fig2E and ~10% in Fig3O) and to assess the extent of rescue for the various genotypes.<br /> - The authors show an accumulation of ECM in the fat body of tumors. It is not clear, whether this ECM accumulates intracellularly near the cell surface or extracellularly. The authors should assess this, maybe by doing electron microscopy.<br /> - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.<br /> - These suggested experiments should be quite straightforward since they are mostly just repeating previous experiments with the appropriate controls and n numbers. I would think that they can be done within a few months. The electron microscopy should not take more than a few weeks and not be costly.<br /> - Are the data and the methods presented in such a way that they can be reproduced?<br /> - The details on how old animals used in each experiment were, are not easy to find and not written very clearly. They should be included in the each figure legend rather than summarising those details in the methods.<br /> - Also, in line 788 in the methods, several stocks are indicated as coming from particular labs (e.g. UAS-FOXO (Kieran Harvey), UAS-GFP (Kieran Harvey), UAS-lacZRNAi (Kieran Harvey), UAS-RasV12 (Helena Richardson), UAS-cg25C;UAS-Vkg (Brian Stramer)).<br /> However, it is not clear whether these labs actually made these stocks and if so whether it has already been described in their papers how the lines were made. If the lines are unpublished, the detailed information should be given on how the lines were made. Or if the lines are published, the authors should provide the reference.<br /> - Are the experiments adequately replicated and statistical analysis adequate?<br /> In general, the n number is rather low in several experiments, especially n of 3 for many controls. And as I mentioned before, rescues of tumor phenotypes are often shown without including a non-tumor control, making it hard to judge the extent of the rescue. Sometimes this information can be found in other figures, but the reader should not have to search for it. And also the severity of the phenotype can vary from experiment to experiment.

Minor comments:

Specific experimental issues that are easily addressable.

• Are prior studies referenced appropriately?

Yes, as far as I can tell.<br /> - Are the text and figures clear and accurate?<br /> - In the literature, people usually call it 'fat body' rather than 'fatbody'.<br /> - The authors state in line 265 "Vkg accumulated in the membranes of fatbody where p60 was overexpressed using r4-GAL4 (Figure 5 A-C)."<br /> This must be a typo. I think it is shown in Fig5E-G. Unless it's labelled wrongly in the figure and B, C and D show p60 rather than TorDN.<br /> - The authors state in line 188 'This manipulation significantly rescued muscle integrity (Figure S4 A-C) and muscle atrophy (Figure S4 D-F), without affecting muscle ECM levels (Figure S4 G-H).' According to the graph in FigS4H this does actually 'affect muscle ECM levels' significantly, as in that it reduced Nidogen levels further. The authors could rephrase this.<br /> - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

Significance

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

The field of inter-organ communication in cancer is a very interesting and trending research field. Several labs including this one have provided new insights into how the tumor, the fat body and the muscle communicate and affect each other and how this can cause cachexia. Previous work from the Chen lab already showed that the tumor secretes the factors ImpL2 and Gbb which decreases insulin signalling and increases TGF-beta signalling in the fat body, respectively and results in fat body and muscle defects. Here they dissect this role of insulin and TGF-beta signalling in the fat body in regulating muscle integrity during cachexia further. They show that these two pathways converge via Sog in the fat body of tumor-bearing animals and result in aberrant ECM accumulation in the fat body which hinders ECM secretion. As a result of this, the muscle receives less fat body-derived ECM and displays muscle attachment defects. Interestingly, the authors show that these muscle defects can be ameliorated by activating insulin signalling or inhibiting TGF-beta signalling or even by increasing ECM secretion in the fat body. This has potentially important implications for the clinic since it suggests that targeting ECM secretion or ECM remodeling in the fat tissue could be a promising treatment for cachexia.<br /> Moreover, the authors also provide some evidence that the insulin and TGF-beta signalling pathways can converge in tumor and non-tumor settings. This might also reveal new drug targets to treat cachexia.<br /> - Place the work in the context of the existing literature (provide references, where appropriate).

The Chen lab showed previously that MMP1 secreted from the tumor induces ECM disruption in the fat body as well as muscle, ultimately causing fat body remodeling and muscle wasting (Lodge et al. 2021). They showed that this is via TGF-beta activation in the fat body. Another contributing factor is tumor-secreted Impl2 which decreases Insulin signalling in the fat body and tumor. However, it remained unknown, how ECM accumulation in the fat body might cause muscle wasting. In this paper, the authors look into this.<br /> - State what audience might be interested in and influenced by the reported findings.

This paper would be of interest for scientists and clinicians interested in inter-organ communication in cancer, particularly in the context of cachexia.<br /> - Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

My expertise lies in the field of Drosophila fat body and ECM, and to some extent tumors but less so signalling pathways.

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 paper, the authors show how the interaction of two signaling pathways, insulin/PI3K and TGF-β signaling, in the fatbody plays an important role in cachectic muscle detachment in tumor-bearing animals. The Drosophila tumor models and the genetic experimental tools are sophisticated and the conclusion is well supported by the data from these genetic experiments. They found that the TGF-β signaling activation (phosphorylation of Mad) is negatively regulated by insulin/PI3K signaling in the fatbody. They also identified the functional involvements of two molecules secreted from tumour tissues, Impl2 (a negative regulator of insulin signaling) and Gbb (one of the TGF-β ligands), in protein synthesis and ECM accumulation in the fatbody, respectively. They also showed that the cachectic fatbody traps ECM proteins and prevents ECM secretion to the muscle, causing muscle degradation. Finally, they identified a secreted BMP antagonist, Sog, as an important player in this process. They found that Sog is reduced in the hemolymph of tumour-bearing animals and that Sog expression is regulated by insulin signaling. Furthermore, Sog overexpression in the tumours, fatbody, and muscle rescues cachectic muscle detachment.

Major comment

Their genetic experiments clearly showed that the reduction of insulin signaling activity in the fatbody induces upregulation of TGF-β signaling and Collagen accumulation. Then, how does TGF-β signaling induce Collagen accumulation? They showed that Rab10 knockdown and SPARC overexpression reduced the accumulation of fatbody ECM. Are Rab10 and SPARC expression regulated by TGF-β signaling?

Minor comments

1. Line 90: "Disc Large (Dlg) RNAi in the eye" must be "Discs Large (Dlg1) RNAi in the eye imaginal discs".
2. Figures 1D and 1L are from the same image. Also, Figures 1C and 1M are from the same image. Are both of them necessary to be shown in the different panels?
3. Why are the staining patterns of anti-pAkt shown in Figures 1L and 1U so different? pAkt is not detected in the nuclei in Fig. 1L but its nuclear signal is clear in Fig. 1U.
4. Figure 1: Images of counter staining for nuclei like DAPI should be also included for all these fatbody images.
5. Line 101: "Tumour specific ImpL2 inhibition was sufficient to reduce fatbody pAkt levels." Is this correct? ImpL2 inhibition in tumors should elevate the pAKT level in fatbody.
6. Figure S1~S4: These figures and their legends do not correspond to each other.
7. Line 189: The pAkt level in the muscle of tumour-bearing animals should be examined to confirm the activity of the insulin signaling is downregulated.
8. Line 189: If the authors conclude that muscle insulin signaling predominantly regulates translation and atrophy, OPP assay for the muscle cells should be examined in the same experimental settings.
9. Line 247: The expression level of Rab10 and SPARC should be examined in the fatbody of tumour-bearing animals to see whether Rab10 is upregulated and SPARC is downregulated.
10. Line 247: If Rab10 upregulation and SPARC downregulation are the causes of the accumulation of ECM proteins in the fatbody of tumour-bearing animals, how the overexpressed Collagen proteins can be secreted from the fatbody cells?
11. Line 347: Sog is a secreted BMP antagonist. Thus, it can be expected that the Sog overexpression downregulates TGF-β signaling in fatbody and muscle tissues. If the rescued phenotypes with Sog overexpression can be explained by this logic, pMad level should be examined in these experiments.

Significance

I found these results from their genetic experiments described here very interesting and of high quality. Although the mechanism by which the TGF-β signaling induces ECM accumulation in fatbody is not clear, this study represents several important advances to understand the key processes in tumor-induced muscle degradation. These data will attract broad audiences not only from cancer biology but also from the research fields including interorgan interactions, systemic signaling in homeostasis, and developmental 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

In this manuscript, Bakopolous et al. investigated on the function of Insulin and TGF beta signaling in the converging regulation of sog (BMP antagonist) and how it controls ECM remodeling. Therefore, the authors used a Drosophila model of cachexia established in Lodge et al., 2021. The authors have shown that the tumors increase Impl2 and Gbb in the fatbody leading to the inhibition of insulin signaling and activation of TGF-β signaling respectively. This lead to the accumulation of ECM proteins that contributes to muscle ECM deficit and muscle detachment. These findings are a major advance in the field of cachexia and of broad interest. The authors demonstrate that state-of-the-art genetics in flies allows acquisition of genetically precise data along with important and complex discoveries on signaling pathways with relevance not only for basic, but for biomedical research as well.

The manuscript is concise and very well written. The experiments overt a clear logical order and are comprehensively described. The authors provide exhaustive data to support their novel claims of broad interest to the scientific community. Please find below some minor recommendations and experiments that could shed further light on some aspects of this manuscript.

Major:

• The statement (line 149'Together, our data suggest that systemic ecdysone levels are unlikely to be involved in modulating tumour-induced muscle detachment or to mediate the role of fatbody Insulin signalling in regulating muscle detachment.') is derived from an experiment with sterol free diet (in which 20HE is genetically addressed) and a pleiotropic experiment (PG>RasG12V). In neither paper nor the current manuscript, 20HE levels have been directly addressed.<br /> Therefore, this statement needs further experimental support and discussion. Ecdysone is a critical hormone during development and especially growth-related effects central to this study. The authors should consider doing pharmacology or augment their claims here with genetic manipulation experiments of 20HE related genes in larvae (Leopold, Rewitz, Rideout, Drummond-Barbosa, Schuldiner labs) and adult animals using genetics, pharmacology or direct assessment of 20HE levels (RIPA, Edgar and Reiff labs).
• In Fig.7 the authors used a sog-LacZ stock to show transcriptional activation in fatbody cells. This stock is based on P-element insertion in the according regulatory regions and supposed to express lacZ with an nls. I can clearly see lacZ in nuclei in Fig. 7H, whereas this is very hard to see in nuclei in Fig7i in the tumour model. In addition, lacZ is known for its high stability and not the best option. As this finding is vital for central claims of this study, it should be complemented by either qPCR for sog on fat body cells or using another readout by converting one of the two Mimic lines (BL42189/44958) into GFP sensors for sog.
• I have similar problems with Fig.7B-F, as phosphorylated Mad should be translocated to the nucleus. In 7F the authors measure pMad over Dapi, which is the right way but it is hard to see pMad in the nucleaus apart from Fig7B, wheras in D and E, where the authors measure higher levels, I cannot identify clear pMad in nuclei. These images either need to show the Dapi channel or more representative images should be chosen like in Fig.4 with arrows pointing to measured nuclei. Fig.7C something went wrong with the compression of this image.
• The proper function of RNAi stocks targeting genes like sog, mad, etc. is vital for this study as these lines are used throughout the study. Functional evidence of specific knockdown efficiency should be provided or references given in which these stocks were shown to provide functional knockdown on transcript or protein level.
• Fig.S7 discusses appearance of gbb/Bmp7 and Sog/CHRD in human patients. The analysis the authors performed shows a correlation between both factors, but is hampered by the fact that datasets for peripheral tissues of cachexia patients are unavailable. The authors may consider sorting these after tumor entities in which cachexia occurs frequently vs. low occurrence and then check for both genes.
• Fig.5 M-P pMAd is not indicated in the Panels only the legend.
• Please follow FlyBase nomenclature, e.g. dlg1 for discs large 1 and unify in the whole manuscript and figure for all genes.
• For endogenous fusion proteins like Viking-GFP (e.g. vkg::GFP) choose a format to clearly decipher them from transcriptional readout stocks like sog-lacZ.
• The quantifications in most figures are quite small with tiny lettering and XY axis are difficult to read in letter/A4 size.

Minor:

1. Adjust in-figure caption alignments
2. Line 104: add comma RasV12, dlgRNAi
3. Line 114: replace little  not significant (n.s.)
4. Line 334: 'sogRNAi overexpression' to my knowledge, RNAi are expressed, not overexpressed.
5. Line 454: italicize r4>
6. Fig S4E: remove frame
7. Figures 6: It would be better to number and explain the pathway presented in the figure in text and fig legend.
8. Just a personal preference. Lettering of images in images is commonly done horizontally, here it appears like a mix between vertical and horizontal.

Significance

In this manuscript, Bakopolous et al. investigated on the function of Insulin and TGF beta signaling in the converging regulation of sog (BMP antagonist) and how it controls ECM remodeling. Therefore, the authors used a Drosophila model of cachexia established in Lodge et al., 2021. The authors have shown that the tumors increase Impl2 and Gbb in the fatbody leading to the inhibition of insulin signaling and activation of TGF-β signaling respectively. This lead to the accumulation of ECM proteins that contributes to muscle ECM deficit and muscle detachment. These findings are a major advance in the field of cachexia and of broad interest. The authors demonstrate that state-of-the-art genetics in flies allows acquisition of genetically precise data along with important and complex discoveries on signaling pathways with relevance not only for basic, but for biomedical research as well.

The manuscript is concise and very well written. The experiments overt a clear logical order and are comprehensively described. The authors provide exhaustive data to support their novel claims of broad interest to the scientific community

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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 #4

Evidence, reproducibility and clarity

Summary

The authors in this manuscript create in vitro degron models of DNMT1 as tools to investigate the roles and functions of DNA methylation in molecular and cellular processes. Degron models can directly target the tagged protein of interest leading to its degradation. When it comes to DNMT1, this system can bypass the use DNMT inhibitors, like DAC and GSK3685032 that can have secondary cytotoxic effects. More specifically, the authors create DNMT1 degron tagged models of two cell lines (DLD-1 and RPE1), as well as a DNMT1 degron tagged model of a DNMT3BKO DLD-1 cell line. These systems allowed the authors to investigate the passive demethylation occurring over consecutive cell divisions, and particularly the role of DNMT1 and DNMT3B and their cooperativity in maintaining DNA methylation levels and how this differs among different genomic regions. The authors characterise the cell fitness of the models they established when DNMT1 is degraded, and methylation levels are lost, and observe a reduction of fitness due to G1 arrest. Finally, the authors show that the loss of DNA methylation observed in these cells leads to reduced levels of heterochromatin (H3K9me3) as well as changes in chromatin and nuclear compartmentalization. Overall, the authors, show an appealing in vitro model that can directly target DNMT1, allowing for more delicate experiments that address the impact of DNA methylation levels in somatic cells, to de-convolute their exact roles from other epigenetic marks and cellular processes that are often correlated with.

Major comments

• The auxin degron system relies on the ectopic expression of OsTir1, which is described in materials and methods under 'Plasmids and Cell line generation'. However, OsTir1 expression is never addressed during the manuscript. Quantification of OsTir1 expression levels across the different cell lines is very important in order to more comprehensively characterise this system. This is especially when considering one of the key points of the authors is to establish these new in vitro models as a new tool to study DNA methylation dynamics in the field.
• The degron system requires an endogenous tag of the protein of interest. Specifically in this work, a tag including the mNGreen and the AID sequence are incorporated at the N-terminus of DNMT1. It is unlikely that there is major interference of the tag to protein function as the tagged cells for DLD-1 and RPE1 are both viable and demonstrate high methylation levels. However, the authors do not consider or discuss that the tag might interfere with the function of the protein at all. It would be useful if the authors compared the tagged cell lines (untreated) with wildtype controls for their methylation levels and/or DNMT1 expression and/or DNMT1 localisation with imaging. These experiments would better substantiate the use of untreated cells as 'wildtype' equivalents and contribute to the better characterisation of these systems as in vitro models.

Furthermore, DNMT1 can have different transcripts that begin from different sites. Do the authors consider whether the tag is included in all/most isoforms of DNMT1, or if there are any expressed without it? - The authors observe that DNMT1 is important for maintaining methylation levels as well as proper cell proliferation. They also observe that DNMT1 depletion does not lead to complete lethality as previously observed (Rhee et al., 2000 Nature, Chen et al., 2007 Nature Genetics). They hypothesise that this might be due to non-specific toxic effects (from CRE) and suggest that the degron system is better suited to bypass such toxicity effects. While this might be true and degron systems do provide a direct and acute protein depletion without non-specific toxicity, the authors do not discuss the implications p53 activity might have on the lack of lethality they observe. Omitting the role of p53 in hypomethylation models and drawing conclusions about toxicity effects between different systems can be misleading and should be corrected. Specifically, it has been shown that hypomethylation triggers p53 dependent apoptosis (Jackson-Grusby et al., 2001 Nature Genetics). The authors do acknowledge the difference in p53 activity when comparing between DLD-1 and RPE-1 DNMT1 depleted cells. The reduced proliferation of RPE-1 cells would suggest that irrespective to the degron system, viability depends on tolerance of each cell line to hypomethylation (whether this is p53 dependent or not). DLD-1 cells seem to have a single nucleotide variant in p53 (p.Ser241Phe (c.722C>T)) (Liu et al., 2006 PNAS), that could potentially explain their viability upon hypomethylation, although further work is required to conclusively suggest such interaction. Furthermore, DNA methylation levels and chromatin organisation of RPE-1 NADNMT1 cells are not characterised in the manuscript and is unclear why. - Figure 1D, 1E: The authors provide a Western blot of DNMT (1/3A/3B) across the established cell lines. While some effects like the degradation of DNMT1 based on the degron system or the KO of DNMT3B are convincing (and work well to validate the cell lines), the observation about upregulation of DNMT3B when DNMT1 is degraded, or levels of DNMT1 after wash out, are not as convincing when only showing one blot. This is especially when considering that the DNMTs might have cell cycle expression differences. Additional replicas of the western blot and quantification of bands across replicas, or qPCR to show upregulation of DNMT transcripts, or imaging (like figure S1E), would help make the claim of DNMT3B upregulation and DNMT1 recovery more convincing. - The authors show that during wash out (after stopping the IAA treatment), DNMT1 levels can recover slightly and show the methylation levels of specific sites (figure 2B). However, the authors do not make any characterisation of the global levels of methylation levels and their potential recovery (?) after wash out. This could be either done by imaging (like in figure 1F and 1G) or dot blot (like figure S2A) or mass-spec.

The authors note that recovery of DNMT1 after wash out is to a lesser extent in the NADNMT1/DNMT3B-/- background. The authors do not speculate why would this be. Past reports of degron tagged proteins show that after treatment endogenous protein levels can recover. Does this hint towards a viability issue of the line due to excessive hypomethylation? While difficult to prove it would be useful to speculate why this effect occurs. - The authors employ DNAme arrays to assess the DNA methylation loss after degradation of DNMT1 and study where in the genome this occurs. Specifically, the authors look on differentially methylated probes between treated/non treated samples and demonstrate their abundance over different genomic regions (figure 2E and S2 H, I, J, K). However, this way of visualising the data is a bit difficult to interpret as differences can be small. Furthermore, number of probes across the genome is not uniformly distributed, so it would be useful to include these numbers. It would be helpful if authors can provide genome browser snapshots with methylation levels and accompanying histone marks (from available data, Rokavec et al., 2017?) like done in figure 4F, S4B and S5C to show representative regions that showcase their observations. Coverage of the EPIC array will mean that these tracks will not have high coverage and thus gaps, and ideally one would need whole genome bisulfite data, however hopefully some snapshots can demonstrate locus specific changes better.

Considering the function of DNMT1 in remethylating the DNA after replication, one would assume that methylation is lost equally across the genome as a simplistic model. Of course, there are many reasons like secondary functions of DNMT1, DNMT3A/3B and TET activity etc that could alter this and provide biases over regions of the genome. The authors discuss this and note most probes show such loss (106,647 of 178,529). It would be useful for the authors to better describe where the rest of the probes (that do not lose the expected methylation, annotated as 'late') are located and speculate what mechanisms might be involved. This is partly addressed in figures S2H and J, but it is not immediately clear what distinguishes late regions from early. Genome tracks with methylation levels and histone tracks as mentioned above could provide examples of regions.

The authors briefly discuss the role of DNMT1 and DNMT3B in methylating specific regions and their cooperativity as well as the underexplored de novo activity of DNMT1. Based on their findings, can the authors draw any new mechanistic conclusions/observations about the activity of DNMT1 and/or DNMT3B and how it is directed? Are there any sequence signatures or histone mark profiles that could explain the hypomethylation or remethylation (after wash out) of specific loci? - The authors observe that 70% of DMPs display an increased methylation in the DNMT3BKO cell line compares to NADNMT1. The authors speculate that this is due to an 'uncontrolled activity' of DNMT1 in the absence of DNMT3B. The increased levels observed could be a clonal effect when generating the KO line. While including additional clonal lines can be a significant amount of work, the authors should acknowledge the effects of clonality in their findings when comparing between the cell lines used (that do not relate to the IAA treatments). - In figures 3D and S3D, the authors compare the viability between IAA treated cells as well as DAC and GSK3685032 and observe increased toxicity/lethality in the case of DAC and GSK3685032. It would be helpful for the authors to discuss the dosage and concentration they used for each drug and why. In order to compare the viability of cells between treatment of different drugs, one would expect dosages that lead to equivalent extents of hypomethylation. - The authors show in figure 3 that the cell lines used have major cell cycle defects, with pronounce G1 arrest, when treated with IAA. Then the authors proceed to perform HiC in treated and untreated sample in figure 4. Can cell cycle differences be cofounding in chromatin compartments and thus affect the data observed in HiC? - For figure 4F and G the authors note a global reduction of H3K9me3 levels after treatment. It would be helpful if the authors include assessment of global levels of H3K9me3 (for e.g. by WB) or ChIP qPCR on loci of interest or specify the use of spike-in in methods, as alterations in global levels of a mark can lead to skewed normalisation/quantifications between samples. Alternatively, comparing the peaks/domains of a mark (and whether they are conserved across cell lines) but not directly compare levels can provide a safer interpretation of the data. - For figures 4F and S5C different days of treatment are provided, with HiC and H3K9me3 being done after 10d of IAA and CpG methylation after 4d of IAA. It is not explained why this discrepancy in days of treatment has occurred, which can be misleading as 10d treated cells should have lower methylation levels from 4d treated cells.

Minor comments:

• Typo in introduction: germiline
• Introduction has some sentences that might need rewording. For example: 'Somatic DNAme domains are erased right after fertilization to establish a totipotent germiline epigenotype, deposited de novo during early development and undergo massive re-shaping during differentiation, lineage specification, and in response to external cues; then, they are maintained and inherited through cell divisions'. It would be good if this is broken into smaller sections as it is hard to follow.
• Introduction does not include the degron technologies and their advancement in the last couple of years. Considering the main point of the paper is to establish an in vitro tool to study DNA methylation based on degrons, it would be helpful to include some information about the technology in the introduction.
• Introduction does not include HiC technologies and the different compartments (A/B, and further subcategories) that the genome can be divided in by them. As the authors then proceed to use HiC data and perform such genome compartmentalisation, it would be helpful if this is addressed briefly at the introduction.
• The authors do not mention the DNMT3BKO strategy they employed. Specifically, the exact strategy should be listed under 'Plasmids and Cell line generation'. A genotyping PCR at supplementary (like figure S1B) could be added. A schematic like Supplementary Figure S1A would also be helpful, but not necessary.
• The duration and concentration of DAC and GSK368503 are not always indicated in figure legends.
• Figure 1C. Homozygous intensity of GFP is much more heterogeneous than the heterozygous levels. It would be interesting if authors could speculate why this is.
• Figure S1D, S1E: Quantification of imaging experiments is shown, however there is no representative images of the staining performed. Incorporate an example image of each staining would be helpful to accompany the quantifications.
• Typo: 106,647 ("early") of 178,529 probes
• Figure 2D: DNA methylation levels in somatic cell lines usually have a bimodal distribution with highly and lowly methylated regions, thus the representation of the data with a boxplot can be misleading.
• Figure 3E: The no. of colonies after IAA removal (from figure 3D) is not included, as suggested from the text.
• Figure S3E: Aneuploidy will be dependent on number of cell divisions so it would be helpful if authors specified how long after treatment the experiment was performed.
• Figure S4B typo: On top track blue compartment is annotated as DLD1-H, while I think it should be DLD1-B2/B3?
• It would be helpful if the authors include an example image of how the segmentation and quantifications for figure 4A and 4B-C were performed as a supplementary figure, demonstrating the area they consider as periphery.
• Figure 3B-C have no error bars and figure legend mentions N>15643 cells per condition. It would be helpful if the number of cells per condition is included in the legend and error bars are included in the figure.
• The authors note that there must be a cooperative effect of DNMT1 and DNMT3B in maintaining DNA methylation and that they observe a strong additive effect in cell survival in double DNMT1/3B depleted cells. These observations have already been observed in the past in HCT116 cells, so it would be useful to cite these papers along with their observations. For e.g. Rhee et al., 2002 Nature, Cai et al., 2017 Genome Research
• A degron tagged DNMT1 in HCT116 cells has already been shown at Onoda et al 2022 bioRxiv that would be good to reference. While the authors in this preprint do not perform any characterisation of methylation levels of the tagged line as in this work, it provides a similar in vitro model that is helpful to include.
• The effects of extensive hypomethylation due to the lack of DNMT activity and its effect in 3D genome integrity has also been shown in the best and would be helpful to mention. For e.g. Du et al., 2021 Cell Reports

Significance

The authors in this work generate and characterise an untransformed (DLD-1) and cancer (RPE-1) cell line model of DNMT1 with a degron tag, as well as DNMT3BKO line of DLD-1 with the degron tagged DNMT1. These in vitro degron models allow for acute deletion of DNMT1 and induced hypomethylation and can be valuable tools to study the effect of DNA methylation in other epigenetic marks and cellular processes. The authors demonstrate the role of DNMT1 and DNMT3B and their cooperativity in maintaining DNA methylation levels in these cells, as previously demonstrated in similar somatic cell models. They also characterise the fitness of these cell lines after DNMT1 degradation and note their viability over DAC and GSK3685032 treatments that can have secondary cytotoxic effects. However, the viability of the cells and the reasons of observed lethality in some systems is underexplored, with the extent of hypomethylation in each system not specified. Finally, the authors show that DNMT1 and DNMT3B impact heterochromatin and the loss of DNA methylation leads to changes in chromatin compartmentalization (with HiC), which have been observed before. While the DNA methylation levels and chromatin organisation of DLD-1 cells was investigated, the authors do not provide any characterisation of these in RPE-1 cells. Furthermore, it appears that RPE-1 cells show more pronounced cell cycle defects and reduced viability hinting towards p53 dependent apoptosis due to loss of methylation, something which is not extensively explored. These observations suggest that the viability of the DLD-1 cells is 'DLD-1 specific'/p53 dependent and not due to the degron system overall. Nevertheless, these in vitro tools will be highly valuable in the epigenetics and specifically DNA methylation fields and their more comprehensive characterisation and will be of high significance.

My field of expertise lies within DNA methylation mechanisms and have limited expertise in HiC experiments.

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 Scelfo et al. describes the establishment of an auxin-inducible degron system for depleting DNMT1 in human cancer and immortalized cell lines. Using this system, the authors show that lack of DNMT1 leads to a profound passive loss in 5mC that is enhanced in the context of DNMT3B knock-out. The decrease in 5mC is further associated with cell-cycle arrest in G1. In addition, they demonstrate through microscopy that the peripheral distribution of heterochromatin in the nucleus depends on DNA methylation. By running Hi-C analysis, the authors further show that specific chromatin domain interactions depend on DNMT1 levels while others depend on DNMT3B. Finally, restoring DNMT1 levels through auxin wash out, although with different kinetics, partially alleviates the above-mentioned effects of DNMT1 loss.

This is a well-designed study and in general the data support the conclusions that are drawn by the authors. I have one concern though regarding the description of the genomic distribution of DMPs (Page 11 of the text and Fig. S2H and S2I). Indeed, the authors indicate that "DNAme at active promoters was unaffected" but Fig. S2H and S2I show that, if I understood correctly how data are represented, 2.5 % (S2H) to 5 % (S2I) of DMPs are falling within the active promoter category. I agree that these are under-represented when compared to the % of CpGs falling in this category in the Epic array but one cannot say that no DNAme change is targeting these regions. A similar concern applies to the description of Fig. S2J,K. In addition, I found the authors could describe and justify in a more detailed way their choice of the two cell lines used in the present study (DLD1 and RPE cells). Also, it would be useful to have information on the cell cycle duration in these cells in order to be able to fully interpret the impact of DNMT1 loss on cell cycle in the time frame used here. Finally, information regarding the cell lines used to acquire data is missing in a number of figure legends. This is the case for instance for Fig. 1B,D,E,F,G, Fig. 3A and Fig. S3E in which it is not specified whether the authors used RPE or DLD1 cells.

Significance

The novel system described here will certainly be of interest to researcher in the field of DNA methylation and chromatin organization. This article presents convincing and original data showing that DNMT1 levels can be reversibly down-regulated through the auxin-inducible degron thus providing opportunities to study the effects of DNA methylation loss on chromatin organization without the drawbacks usually observed in long-term KO experiments or treatment with toxic DNMT inhibitors. Example of such data obtained with the degron system are convincingly showing that peripheral heterochromatin relies on DNA methylation by DNMT1 and that interaction of heterochromatic domains also depend on DNMT1 activity. Another original finding is that the spatial organization of different silent chromatin domains can also depend on DNMT3B activity, independently of DNMT1. The fact that DNMT1 levels can be restored after wash out of auxin medium is probably one of the most interesting aspects of this study since it allows to run assays that are not possible in the context of DNMT KO experiments. Using this strategy, the authors demonstrated that, concomitant to an increase in DNA methylation, heterochromatin relocates to the periphery of the nucleus and that DLD1-BA compartmentalization is restored. In this respect, the authors observed that compartmentalization of B4 is rapid and near complete at a time when DNA methylation recovery is still partial, suggesting that DNMT1 could have catalytic-independent roles in this process. Although this is possible, another explanation could be that a partial re-methylation of DNA is sufficient for recovering homotypic interactions.

Regarding Hi-C data, similar results obtained with DNMT1/DNMT3B DKO and 5-aza-deoxycytidine-treated HCT116 cells were already described in a previous study from the authors (Spracklin et al. Nat Struct Mol Biol, 2023). However, differences in the reorganization of chromatin contacts after auxin treatment of DLD1 cells compared to 5-aza-dC or DKO HCT116 cells can be evidenced and are possibly linked to a difference in the organization of heterochromatin between DLD1 cells and HCT116, highlighting the usefulness of running these analyses in different cell types.

Although not really crucial in the context of the present study, information on the transcriptomic changes induced by DNMT1 loss could add some insights into the cellular state induced by auxin treatment. Indeed, cells are arrested in G1 and peripheral heterochromatin seems to undergo spatial rearrangement. This is reminiscent of senescence-associated processes and a loss of DNA methylation during replicative aging has already been documented. Especially, knock-down of DNMT1 is known to trigger premature senescence entry (Cruickshanks et al., Nat Cell Biol, 2013). Hence, a further characterization of the G1-arrested cells upon auxin treatment would clearly add some value to the manuscript.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this paper, Scelfo, A. et al. investigated the mechanisms underlying the cooperative maintenance of DNA methylation by DNMT1 and DNMT3B. Using a rapid degradation of DNMT1 by the auxin-inducible Degron system, which allows the assessment of reversible and time-dependent effects of DNMT1 loss with low cytotoxicity, the authors revealed a cooperative activity between DNMT1 and DNMT3B to maintain DNA methylation. Furthermore, they showed that gradual loss of DNA methylation is accompanied by progressive and reversible changes in heterochromatin abundance, compartmentalization, and peripheral localization. Collectively, this study provides a new cellular model to investigate the fundamental and biological role of DNA methylation and the molecular mechanism underlying its establishment and maintenance.

Important comments:

1. Are there any data showing no change in DNA methylation level of WT DLD-1 and DNMT NA DLD-1 cells?
2. In Fig. S2A and S2D, differences in global DNA methylation between NA-DNMT1-IAA-Day4 and DAC-treated cannot be determined from these images alone because the blot intensities appear similar. It is better to present the results in a more quantitative manner with appropriate statistical analysis.
3. In Figure 2D, it is better to present the results in a more quantitative manner with appropriate statistical analysis.
4. In Figures 3A and S3B, the authors show that DNMT1 depletion decreases the percentage of cells in S phase and increases the percentage of cells in G1 phase and sub-G1 phase in NA-DNMT1/DNMT3B KO. The authors postulate that the decrease in cell proliferation after DNMT1 depletion is due to activation of p53. Data demonstrating activation of the p53 pathway are needed.

Minor comments:

1. Regarding IAA-induced DNMT1 degradation, the authors should provide complete DNMT1 blots to show that no additional isoforms are present.
2. In Fig. S1C. Molecular weight was not labeled in the immunoblot.
3. In Fig. S2B. NeonGreen-IAA 10days images appear to be exposed for different lengths of time.
4. In Fig. 4B-C, S1E, S2E. It is better to present the results in a more quantitative manner with appropriate statistical analysis.
5. Contrary to Fig. S2A, the PCA analysis does not seem to show any difference between NA-DNMT1-IAA Day2 and Day4.

Significance

Collectively, this study provides a new cellular model to investigate the fundamental and biological role of DNA methylation and the molecular mechanism underlying its establishment and maintenance.

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, Scelfo et al. describe how different DNMTs cooperate to maintain DNA methylation and the impact of decreased DNA methylation on chromatin structure. Toward this, they established an inducible degradation system for DNMT1 in untransformed and cancer cell models. The experiments revealed that DNMT1 and DNMT3B are required to maintain and control DNA methylation patterns throughout the genome. The authors also demonstrate that heterochromatic regions are highly susceptible to DNA demethylation, with loss of their localization to the nuclear periphery and disappearance of their compartmentalization patterns. Together, this work will allow better temporal resolution analysis of DNA methylation abnormalities and will be useful for clarifying the role of DNA methylation and its regulatory mechanism.

Major comments:

1. In Figure 2G, the authors report increased DNA methylation at selected loci in the absence of DNMT3B and suggest a compensatory role for DNMT1 in de novo methylation, as this increased DNA methylation is lost upon DNMT1 depletion. However, how do the authors rule out the possibility that this methylation is catalyzed by DNMT3A and maintained by DNMT1?
2. In Figure 3, the authors demonstrate that DNMT1 depletion leads to cell cycle arrest at G1. Since p53-proficient RPE-1 cells showed faster G1 arrest (Figure S3B), the authors suggest that DNMT1 depletion activates the cell cycle checkpoint. The authors might want to check p53, p16, and p21 levels in line with their suggestion.
3. Figure 4F and 4G demonstrate a global reduction of H3K9me3 levels upon DNMT1 depletion. Is a similar effect seen with H3K27me3?

Significance

Understanding how DNMTs regulate chromatin structure and cell fitness is critical to understand better how DNA methylation impact cell fate and function.

Strength: This work established an inducible DNMT1-degradation system with reversibility, temporal control, and low toxicity.

Weakness: this work is merely descriptive and preliminary to understand the observed phenotypes clearly.

Advance: Although the presented experiments are accurate and well-designed, it has already been reported that DNMT1 and DNMT3A/3B cooperate to maintain DNA methylation patterns and that reduced DNA methylation leads to the disruption of H3K9me3/HP1-enriched heterochromatin structure. In addition, the molecular understanding underlying these phonotypes remains unexplored. In its current form, the contributions of this study to the field will be limited.

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

Rios-Szwed and co-authors show that the depletion of FAM111A results in faster replication speed, longer intra-origin distances, and less chromatin-bound RPA even without induction of replication stress in U2OS cells. Induction of replication stress in FAM111A-depleted cells results in blunted response with less DNA damage, decreased checkpoint activation and resistance to the replication-stress inducing agent, HU. They show that cells without FAM111A display lower levels of single stranded DNA after treatment.

In the second part, the authors show that FAM111A and FAM111B form a complex, although the similarities and differences of their functions are not explored in detail. From the little data shown, it looks like they might be working together in controlling amount of ssDNA. They find that both proteins are expected to have two conserved UBL domains, with one of them overlapping with ssDNA binding domain. Finally, the authors use overexpression of WT and mutant proteins to show that expression of WT and patient-derived mutant has increased level of DNA damage, increased levels of ssDNA, with and without DNA damage, and that the peptidase domain is necessary for the phenotypes.

The data from the first two figures are consistent with FAM111A being involved in regulation of single stranded DNA formation during normal replication and during replication stress. Unfortunately, the work gives no indication of the mechanism of such regulation. I am not convinced that the function has much to do with controlling origin activation (see below). The data from the last two figures is also descriptive. Until the substrates of FAM111A are identified, there will be no understanding of its true function and the data will continue to be descriptive.

Specific points:

Figure 1: The siFAM11A-2 has a stronger phenotype in growth assay but has very little change in levels of cells in G1. No complementation of the phenotype is given.

1D- there is no total RPA so it is unclear if there is no change in pRPA in relation to total RPA. Small differences will be missed without DNA damage and it would be helpful to use more sensitive assays to identify the reduction in ssDNA under unperturbed conditions.

1E- what does the data look like if the lengths of IdU are plotted? This would be a measure of speed of the ongoing forks. Generally, this would be better than the CldU measurement.

1F- the Inter-CldU distance increase could be secondary (indirect effect) of the increased replication speed

1G- It looks like there are many more data points in the siFAM11A-1 and many fewer in the siFAM111A-2. The increase in the MCM quantified in H is bigger with si2 even though the G1 distribution has less change than with si1. Consequently, these data are incolclusive.

1I- no plot is shown for si2 but it is quantified. It would be informative to see the plots for easy comparison.

Figure 2: This is the most interesting part of the paper and generally is well done. As mentioned above, I believe that the phenotype the authors see in Figure 1 is the same phenotype as seen here- less production of ssDNA but it is hard to see this under unperturbed conditions, thus more data should be gathered to test that.

Figure 3: shows novel findings but it is unclear how it relates to the rest of the paper except that it suggests that the paralogs may work together in the pathway that has been explored in Figure 1 and 2. The authors perform computational and predictive analysis that identifies two UBL domains in the FAM111A/B paralogs. The FAM111A UBL2 domain is known to bind ssDNA. The authors might test if the domain can also bind ssDNA in FAM111B and if FAM111B has similar ability to promote ssDNA formation

Figure 4: The human mutations provide some insight as to the requirement for functional peptidase activity for the function of the protein. The work would also be strengthened if a ssDNA binding mutant was made and tested given the authors interest in defining the UBL domains.

Not sure why they use a term "ssDNA exposure"? It implies a removal of something that was covering it which they certainly do not show. I would use ssDNA levels, maybe ssDNA production, formation?

Other points:

As QIBC is used throughout the paper, it would be nice to have a brief explanation of the technique when it is first introduced.

The authors write that the function of FAM111A in promoting ssDNA formation is "distinct from overcoming protein-DNA complexes ahead of the replisome by Top1 or PARP1". It is not clear to this reader how they have determined that they are not the result of the same mechanism as the phenotypes seem very related. I would clarify this point.

Since the authors are including patient mutations, more introduction to the diseases would be useful.

Referee cross-commenting

I have no further comments.

Significance

The findings add to the growing literature on the FAM111 proteins and will be of interest to scientists who are studying them and those interested in replication and replication stress response.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript by Rios-Szwed et al have investigated the role of FAM111A in DNA replication. Previous studies had identified that FAM111A suppresses DNA replication via an interaction with RFC and that hyperactive mutants induce apoptosis. Now, Rios-Szwed et al discovered that FAM111A knockdown affects inter origin distance without checkpoint induction. In particular, the firing of dormant origins when dNTPs are limiting is supressed and less ssDNA is produced. Although FAM111B is a strong interactor of FAM111A, no additive effect on DNA replication was detected when both proteins were depleted. On the other hand, overexpression or hyperactive mutants promote more gammaH2AX and ssDNA even in the presence of a caspase inhibitor, suggesting that the protease functions in ssDNA production prior to apoptosis.

Major comments:

Dormant origins are frequently inhibited by phosphatases - is there any evidence that phosphatases are the target of FAM111A. In this context I would suggest to blot for Treslin, as it is one of the first factors being recruited in a kinase dependent manner to the MCM2-7 complex.

Minor comments:

Abstract: Unclear why too much FAM111A causes cell death

Introduction: the R569H point mutant needs to be better introduced - e.g. explain where the mutation is localised or what it affects e.g. it is localised in the predicted peptidase domain

Figure 1A and 1D - are all the lanes shown originating from the same gel - if not please repeat.

Page 3 - I am not sure that in FAM111A depleted cells the DNA synthesis rate is reduced. Could it be, that just fewer cells are in S-phase.

Page 3 - It is stated: "In contrast, the inter-fork distance was slightly increased in FAM111A depleted cells (Fig. S1E)", however, the data but the data do not fully support this statement.

Figure 4C - the quantification of the last lane looks wrong. Is the average or the median? Please find information in the figure and methods section.

Question: If both FAM111A and FAM111B are overexpressed - is this better tolerated?

Is there a homologue in other species?

Referee cross-commenting

I agree with the other reviewers that the study has a descriptive nature. I guess this could be acceptable dependent on the journal choice.

Significance

In general, I really like the study as it establishes how initiation of DNA replication is affected by inhibition and activation of FAM111A. The work is done well and deserves to be seen in a good journal.

The study helps the field to move forward and will allow a more targeted search for specific protease targets. In this way it will help clinicians and also researchers.

My expertise is in initiation of DNA replication.

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

Rios-Szwed and colleagues investigate functions of FAM111A, a protease that Dr. Alabert has previously shown to localize at nascent DNA and promote PCNA loading. In this manuscript, the authors first describe that FAM111A facilitates efficient activation of replication origins by using DNA combing experiments and by analyzing chromatin loading of DNA replication proteins. Next, they show that FAM111A KO cells show reduced levels of ssDNA exposure after replication stress. Then the authors move on to show that the major FAM111A interactor is FAM111B, which they show to localize at nascent DNA and is epistatic to FAM111A in promoting DNA replication as well as RPA loading after replication stress. Finally, the authors show that unregulated FAM111A activity, either by overexpression of WT FAM111A or disease-associated mutants, causes extensive exposure of ssDNA.

Major comments

1. Fig. S1G: Actual inter-origin distances (distance between replication tracks in which a CldU track is flanked by IdU tracks on both sides) should be plotted to estimate the changes in origin firing frequencies. The results should be presented as inter-origin distances, not ratios between UCN-01-treated and untreated. The revised experiment should be included in the main figures as this is central to the conclusion, and statistics should be included.
2. The claim "FAM111A ... promotes DNA replication initiation of active and dormant origins" (page 4, line 4) is not fully supported by experiments. Does FAM111A localize at replication origins? Without direct evidence of FAM111A being present at replication origins, it remains possible that the changes in origin activity is secondary to the loss of FAM111A function at forks or something else.
3. Fig. S1G: If FAM111A's function to promote activation of dormant origins in response to UCN-01 is unrelated to the function of FAM111A at forks, it is expected to be independent of the PIP motif. Is it the case?
4. Fig. 2B: Increased survival after HU treatment might be secondary to reduced S-phase populations in FAM111A-depleted cells (Fig. 1C) as HU would affect only S-phase cells.
5. Fig. 2B-I: Similarly, the blunted response to replication stress in FAM111A depleted cells could be simply explained by reduced number of forks per cell as indicated by increased inter-fork distance (Fig. 1F). Similarly, the authors' group has previously reported reduced PCNA levels on chromatin (Alabert et al, 2014), suggesting that there are reduced number of active forks per nucleus.
6. Fig. 2H "FAM111A depletion reduced ssDNA exposure upon HU treatment (Fig. 2H, 2I)": The figure in Fig. 2H does not appear to be treated with FAM111A RNAi. If this is FAM111A RNAi cells, siControl cells need to be shown as a comparison.
7. Fig. 3B,C: The interaction between FAM111A and FAM111B needs to be validated by coimmunoprecipitation-WB of endogenous proteins.
8. Fig. 4A-C: Induction of DNA damage and apoptosis by FAM111A WT and disease mutants (including T338A that the authors claim unstudied) has been reported by Hoffman et al. and therefore not novel.
9. Fig. 4E: The increase in ssDNA intensities is mild and might not be biologically significant.
10. Fig. 4G: Cell cycle status needs to be assessed by FACS after treatment with each drug. Bleomycin might induce G1/S arrest if G1/S checkpoint is intact.
11. ssDNA exposure after FAM111A OE might not be because FAM111A has a function in promoting ssDNA exposure, but could be simply explained by replication fork stalling, for example, due to degradation of essential proteins as proposed before (Hoffman et al, 2020).
12. Page 8, line 17, "Altogether, these data revealed that unrestrained FAM111A peptidase activity leads to ssDNA exposure upstream of apoptosis.": Just because the caspase inhibitor did not block the ssDNA exposure, it does not mean ssDNA exposure is upstream of apoptosis - it could be happening in parallel and might be unrelated. A similar unsupported conclusion "ssDNA exposure is upstream of apoptosis" appears in other places: page 8, line 30; page 9, line 22.
13. Whether protease activity is necessary for the FAM111A function in regulation of origin activation and in ssDNA exposure is not addressed. Can the phenotypes of FAM111A KO cells be rescued by FAM111A WT but not an active site mutant?
14. Similarly, the authors need to test whether the PIP motif of FAM111A is required for the function of FAM111A at forks, such as promoting ssDNA exposure.

Minor comments

1. Page 2, Line 8, "FAM111A catalytic activity has not been shown in vitro": Protease activity of FAM111A has been shown using recombinant proteins in vitro by Hoffman et al, 2020.
2. Page 7, line 26, "T338A is a previously unstudied GCLEB patient mutation.": The T338A mutant was studied by Hoffman et al. and shown to have hyperactivity in vitro and to cause DNA damage when overexpressed in cells.

Referee cross-commenting

I feel that this study has problems even as a descriptive study. As I mentioned in my review, there are alternative explanations for their observations that the authors have not ruled out. If the authors remove all unsupported claims, then there is not much to conclude from this study. I am not saying their conclusions are wrong - I think this study is just premature.

Significance

This study could be of interest to the audience in DNA replication/DNA repair field and could be unveiling a new function of FAM111A in DNA replication. However, in the current form, this study appears to be a collection of loosely connected observations of FAM111A-manipulated cells without a clear message of what FAM111A does at replication forks and origins. Each observation appears to be loosely tied together with a keyword of ssDNA exposure, but how FAM111A regulates or changes ssDNA exposure is not addressed. The described phenotypes are potentially interesting, but for each observation there is an alternative explanation that could affect authors' interpretation. As outlined in my comments, lack of mechanism, lack of clear conclusion, and misinterpretation of some of the data led to this less enthusiastic review.

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 manuscript entitled "A population instrinsic timer controls Hox gene expression and cell dispersion during progenitor addition to the body axis", Busby and colleagues investigate the topic of "cell type identity" in the context of body axis elongation in chick embryos. To this end, they performed heterochronic grafts from HH8 stage embryos to HH4 stage embryos and compared these to HH4 homochronic grafts. They found that HH8 grafts ingressed but were then delayed at a stage they termed cell dispersion. By scRNAseq this new cell state was characterized further. While HH8 cells adjusted their expression pattern to their surroundings, Hox gene expression was maintained as in the host developmental stage. Hox gene expression and collinearity of expression changes were also maintained when HH4 cells were grafted into HH8 embryos or cells were cultured ex vivo. Finally, the authors found differences in migration properties between HH4 and HH8 cells, when cultured ex vivo, with HH4 cells migrating faster than HH8.

This constitutes an elegant work to describe the existence of a "cell-intrinsic timer" that regulates cell identity and progressive body axis extension. Experiments and analysis have been performed adequately and conclusions have been drawn appropriately.

There are a rather minor comments I would suggest for further analysis, discussion and potentially experiments to further support this paper:

• A major finding is that grafted cells keep their Hox expression pattern, independent of whether it is from HH4 to HH8 or vice versa. Moreover, grafted HH8 cells pause at the cell dispersion stage and do not mix, unless grafted in very low cell populations. The authors conclude that Hox gene expression seems to be cell intrinsically regulated. However, for pausing of cells after ingression, I wonder if it is rather the difference to the neighbors than a cell-intrinsic effect that prevents the cells from dispersing. One possibility is that differences in adhesion could account for this, since sorting of cell populations based on differential expression of adhesion molecules has been observed in various model systems. This possibility is excluded here, since adhesion-related genes were not differentially expressed in their expression data. However, I would not exclude this possibility at this stage for the following reasons: 1. The authors detect different migration speeds for HH4 and HH8 cell clusters with HH4 cells migrating faster. Differential migration rate could indeed hint at differential adhesion and mechanical properties of the cells. 2. Hox genes have been shown to be upstream of and modulate adhesion molecules, which might be an interesting link. 3. So far, the authors have only analyzed expression of adhesion molecules at mRNA levels. However, the functional components are the adhesion proteins themselves. It might therefore be useful to stain embryos for some "obvious" candidate adhesion molecules, such as cadherins. If no further experiments are performed, then this should at least be discussed.
• The authors describe a new, intermediate stage, namely cell dispersion, in which HH8 MSP pause when grafted into HH4 embryos. They perform scRNAseq and GO term analyses to analyze these cells in more detail. The also perform gene set enrichment analysis. However, I am still wondering about the exact identity of these cells. What are they? What markers do they express? Do they upregulate certain signalling pathways? Etc. I would for instance be interested if there are differences in FGF or Wnt levels/ activities. It would be useful if the authors could analyze their scRNAseq data further in this regard.
• At several points in the manuscript, expression levels and patterns of HH4 and HH8 grafts are compared to each other. It does not become clear what the differences and similarities to the non-grafted cells of the same clusters are. Does grafting itself change the expression patterns?
• The authors found differences in cell cycle stages of HH4 and HH8 grafts. A more detailed discussion of this aspect would be useful rather than just excluding any cell cycle-related genes from the comparisons. Why could there be this difference? What effect could this have? Etc.

Optional: Other experiments that could increase the relevance of the work:

• As discussed by the authors, they specifically compare HH4 stage to HH8, which represents primitive streak stage and 4-somite stage, respectively. It would therefore be interesting to perform grafts from HH8 to later stages, such as HH10, or vice versa, when the process of somitogenesis is more similar. This could reveal if their findings are specific for pre to post node development or more general. However, this might be outside of the scope of this study.

Significance

General assessment: This study provides a systematic analysis of the interaction of embryonic cell clusters from different developmental stages. To this end, "classical" developmental biology techniques, i.e. grafting (complicated techniques that probably less and less people can perform nowadays), is combined with more modern ways of analysis, i.e. scRNAseq. This allows the authors to dissect the differential behaviour of hetero- and homochonic grafts. In the longer term this data can provide the basis for further in-depth mechanistic analyses, some of which could be added here already. The involvement of Hox genes in the control of developmental time is interesting and should be placed into context of our current knowledge. Here, Hox genes are rather used as readout of developmental time rather than active players.

Advance: How developmental time is maintained during embryonic development is a long-standing question in the field. This study provides conceptual advance in this question by describing a cell-intrinsic timer.

Audience: This study is relevant for developmental biologists in general, since it describes how developmental time can be kept by a cell-intrinsic timer, at least in early stages of somite formation in chick embryos.

My expertise: developmental biology, somitogenesis

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the authors investigate the importance of intrinsic and extrinsic factors in the timing of progenitor addition to the elongating primary body axis. During development, progenitor populations have to combine their self-renewal with the gradual contribution to the full length of the body axis. The mechanisms underlying the population dynamics that ensure the formation of a proportioned body plan remain poorly understood. By combining heterochronic (HH8 to HH4) and homochronic grafting (HH4 to HH4) of somitic progenitors with next generation sequencing and imaging, the authors observe that the older HH8 tissue shows intrinsic delays in migration and does not disperse within the surrounding mesodermal tissue after ingression through the primitive streak. This behavior correlates with intrinsic and tissue-specific differences in the expression of Hox genes but not with differences in the expression of cell adhesion/migration genes.

Overall, this study provides new data exploring how progenitors control their contribution to the body axis. By combining classic embryology techniques with single-cell sequencing, the authors describe novel cell states that might help understand the progenitor population dynamics. There are however a number of further analyses and experiments that should be performed to support the main claims of the manuscript.

Major comments:

1. The authors claim that grafted HH8 cells are paused after the ingression stage and before the dispersion stage. The grafted cells ingress through the primitive streak and then remain as a distinct cluster of cells that does not disperse throughout the mesoderm. This is in contrast with other observations where overexpression of late hox genes delays the cells at the point of ingression. The authors should better demonstrate that their grafts are actually ingressing and then stopping once in the mesoderm compartment. Figure 3B' shows grafted HH8 cells (GFP positive) present in the mesoderm (ME) compartment 3 h after grafting. It is surprising that a cluster of cells can ingress through the primitive streak in a short period of time and then remain paused. It would be helpful to have the equivalent figure right after grafting to assess the differences in the location of the HH8 GFP+ cells and potentially observe them while ingressing.
2. The authors describe a novel transcription state, namely clusters 6, 12 and 8 in Figure 2B, populated by HH8 cells 3 h after grafting. It is surprising that the UMAP looks very different between 0 h and 3 h in the HH8-HH4 grafts (Figure 2E and F). The authors should clarify where the HH4 (GFP negative) cells are present in Figure F, I. In the current figures, it looks as if both HH8 and HH4 cells changed completely their transcription profile in only 3 h and populated the central clusters (6, 12, 8). The authors claim that these central clusters are present in normal development and that cells rapidly transit through them. However, it is not clear whether this state happens before or after HH4. For example, the cells may be moving from right to left in UMAP_1 according to time (HH4 in the right, HH8 in the left and a central transient cluster). This would mean that in Figure 2F HH8 grafted cells are regressing to an earlier development state and not a new one. Including RNA velocity analysis could help clarify how the cells are changing their expression profiles.
3. Related to the previous point, the striking changes between 0 h and 3 h in the HH8-HH4 grafts (Figure 2E and F) may suggest an effect of the grafting procedure on the transcription profile of the cells. The authors should demonstrate that the grafting of cells does not have a huge impact on the transcriptome and that these changes are specific to the previously undescribed delayed state of HH8 cells. For this, they should include scRNA data of HH4-HH4 3 h. If grafting does not have a significant effect on the transcriptome, they should see GFP positive and negative cells in HH4-HH4 3 h remain intermixed.
4. The authors observe that when doing smaller heterochronic grafts, cells can disperse throughout the mesoderm. Nevertheless, the Hox gene expression does not change depending on the size of the grafts. This is in sharp contrast with their observations and claims done for big heterochronic grafts. The result is interesting as it demonstrates that the expression of Hox genes, but not dispersion, is cell intrinsic. However, the uncoupling of hox gene expression and cell dispersion requires further investigation. The authors should repeat the heterochronic grafting of Figure 1 using smaller grafts and check the contribution of grafted cells to the somites. If cells can readily disperse without delay, they might be able to contribute to all somites as observed with homochronic grafts. Similarly, the authors should repeat the explant spreading assay using smaller HH8 grafts and quantify whether differences in the migratory dynamics are observed. The authors already discuss the possibility that other factors apart from Hox expression might affect dispersion. Nevertheless, they should assess the importance of graft size in their experimental system. If smaller grafts maintain the expression profile but have a different capacity to contribute to the body axis, the initial observations might have been influenced by extrinsic factors of the graft (size, cell-to-cell contact, ECM...) and not by cell intrinsic properties (gene expression). This would change the conclusion of the work.

Minor comments:

1. It would be informative to have a better time-resolved description of the heterochronic graft behavior in Figure 1. For homochronic grafts, several timepoints are provided allowing the visualization of cells travelling through the body axis. For heterochronic grafts, by contrast, only an early and final timepoint are provided.
2. In Figure 4, the authors show that explants of HH4 and HH8 embryos have different migratory dynamics, with HH4 cells migrating faster than HH8. In Figure 4E, HH8 explants seem not to change their area for about 15 h and then start spreading. This indicates that there is a great delay in migration compared to HH4 explants. However, once they start spreading, it seems that the area starts to increase exponentially in a similar manner to what is observed for HH4 at earlier time points. It would be interesting to monitor the HH8 for a longer time to see the behaviors at later time points. HH8 explants may be just delayed, and once they start fully spreading, the speed may not be so different from the one of HH4 explants.
3. The authors conclude in Supp. Table 2 that HH8 and HH4 do not have different expressions of adhesion-related genes upon grafting. This observation is very important to understand the potential mechanism behind the different dispersion behaviors, and thus it should be included in the main figure.

Significance

The authors combine classic embryology with single-cell RNA-seq and imaging techniques to explore progenitor population dynamics during addition to the body axis. They conclude that the delayed contribution of older cells to axis formation correlates with the intrinsic expression of posterior hox genes. While the idea of intrinsic regulation of hox genes during axial specification is not conceptually new, the authors use modern techniques to describe with finer detail the progenitor population states. For this reason, this manuscript will be of interest to researchers in the development field who want to better understand the hox control and influence during axial elongation.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript describes the differential behavior of the epiblast region of chicken embryos containing the progenitor cells for the medial half of the somites (MSP) at HH4 (building the first 4-5 somites) and HH8 (building more caudal somites). Their approach combines grafting experiments with imaging, single cell and whole mount expression analyses of the grafts. The basic experiment involves the comparison of HH4 to HH4 homochronic grafts with HH8 to HH4 heterochronic grafts. They show that homochronic grafts undergo a dispersion stage after ingression through the primitive streak before they contribute to somites. The same region of HH8 embryos, when grafted into the MSP region of HH4 embryos, however, fail to undergo this dispersion and do not contribute to the first 4-5 somites. They also show that Hox gene expression follows the patterns observed in the grafted tissue, failing to acquire the expression profiles of the receiving host. The authors conclude that the MSP cells contain an internal timer involved in the regulation of their changing behavior as development proceeds.

The general findings reported in this manuscript are novel and can provide insights to further our understanding of the differences between formation of the first 4-5 somites and more caudal somites. However, I think that ADDITIONAL EXPERIMENTS are important to properly evaluate the data and the conclusions of this manuscript.

1. A control HH8 to HH8 homochronic graft to check the behavior of the grafted cells: do they disperse in their natural environment after ingression through the primitive streak or they are also paused as a distinctive cell cluster?
2. The reverse heterochronic grafting experiment, namely HH4 cells into HH8. Do HH4 cells maintain their dispersal behavior at the ectopic position, or they behave differently?

While the authors assessed the intrinsic properties of HH4 and HH8 tissue by incubating it on fibronectin, this experiment does not properly reproduce the environment of the embryonic region receiving the graft, which might be different at HH4 and HH8. The experiments I am suggesting take this variable into consideration and will therefore help assessing the possible involvement of the host tissue in the behavior of the grafts.

In addition to those experiments, a MORE EXTENSIVE ANALYSES of the already reported experiments could also improve the manuscript.

1. When the cells staying in the MSP region after homochronic HH4 grafts reach later stages (e.g. approaching HH8), do they keep dispersing as at earlier stages after ingression through the primitive streak or they remain as a distinct cluster? And does Hox gene expression within those grafts follow the same activation profile observed in the host cells as development proceeds?
2. In the experiment reported on fig. 5I, HH4 MSP grafted into HH8 embryos fail to activate Hox genes like Hoxa2, even after 6 hours of incubation. When these grafted embryos develop even further (for the period of time required for a HH4 embryo to reach the HH8 stage), do they activate Hoxa2 or Hoxa3 or they remain negative for these genes?
3. The differential GO terms between HH4 and HH8 tissue in cluster 6 include chromatin organization, DNA methylation and C5-methylation of cytosine. This suggests that epigenetic changes might be involved in the behavioral differences between the MSP of the two stages, which can affect many different processes involved in cell activity, including the activation of Hox genes.

SOME COMMENTS ON DATA INTERPRETATION.

1. It is clear that Hox gene expression in the grafts matches the profile of the donor tissue, indicating the existence of a Hox "timer". However, in my opinion, the authors place too much emphasis on the possible meaning of these observations in what concerns the differential behavior of the grafted cells. If they want to focus on Hox genes they should include some experiment testing their involvement in cell dispersal, either by misexpression or downregulation of specific genes (although there is plenty of information arguing against this possibility, maybe with the exception of that of Iimura and Pourquie, 2006; in this regard, the authors' own data already indicate that dispersion is independent of Hox gene expression).
2. The authors disregard the involvement of differential patterns of cell adhesion molecules as the origin of the differential behavior of HH4 and HH8 grafts in the HH4 context. However, in their data on supplementary fig. 2A there are several genes differentially expressed between the HH8 and HH4 cells (e.g. Ptk7, Spon1 or Nfasc) that could indeed play a role in the differential interaction between cells from the two embryonic stages. It might be interesting to perform HCR experiments with some of these factors to see if they are differentially expressed at the two embryonic stages. Also, although it might be somewhat far reaching, if differential expression is observed by HCR, it might be interesting to experimentally manipulate expression of the relevant gene (s) (misexpression or down-regulation, depending on the stage) to evaluate its/their potential functional relevance.

Minor points

1. The authors write that the HH8 specific clusters are 0, 4 and 6. However, I think that it is #7 and not #6 the one belonging to this group. I guess that this is typo, but becomes confusing, as a large part of the analysis of the single cell data is centered on cluster #6.
2. In the introduction the authors state that the first 4-5 somites do not develop ganglia, citing Lim et al 1987. I think that the way this is written is imprecise, as it sort of implies that more caudal somites develop ganglia (which would mean that the dorsal root ganglia are somite derivatives). However, somites at any level do not develop ganglia; the anterior half of their sclerotomes are permissive to migration of the neural crest that will eventually build the ganglia, something that seems not to happen in the more anterior somites.

A side note

Different alternative transcripts have been reported at least for Ptk7 and Nfasc. This might be relevant considering that another of the prominent differential GO terms identified in supplementary fig 2C is related to RNA splicing. Would different alternative transcripts for some of these genes be specifically associated with the cells from one of the embryonic stages?

Significance

It has been known for many decades that the first 4-5 somites of amniotes are different to the rest of the somites in several ways, from the structures they generate to the way they are generated or the gene regulatory networks controlling their morphogenesis. Much is known about how the posterior somites are generated and the mechanisms of their differentiation. Conversely, relatively little is known about the same processes in the most anterior somites. The work described in this manuscript shows that the progenitor cells from the epiblast that will contribute to the 4-5 first somites already behave different than those generating more caudal somites. Also, they show that progenitors generating more caudal somites are unable to contribute to the rostral somites. These two sets of observations show that the differences in the rostral and caudal somites are already present in their progenitors and that those features are quite stable within the cells, at least when they are kept as a group.

So far, the single cell analyses shown in this manuscript failed to provide clear hints to explain the different behavior of the two sets of progenitors. However, they represent an important resource to further explore this important biological question. The authors focus on Hox genes as potential regulators of the differential behavior of the HH4 and HH8 MSPs, I guess that prompted by the report by Iimura and Pourquie (2006) indicating the involvement of Hox genes in the migratory properties of the somite progenitors. However, there is plenty of information, mostly genetic studies in mice, indicating that Hox genes might have very little influence in the differential behavior of rostral and caudal somites. In this regard, expression does not mean causation. I think that this manuscript is interesting, most particularly for developmental biologists involved in understanding the mechanisms governing the basic layout of the vertebrate body plan.

Research in my laboratory also explores this type of biological questions, although using more genetic approaches and in a different model system, namely the mouse. I therefore consider myself in a position that allows a knowledgeable evaluation of this manuscript.

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

We thank the four reviewers for their generally positive feedback on the manuscript. Below, we provide a point-by-point response to each reviewer.

We are performing new FCS and gradient measurements as suggested by the reviewers. We are confident we can have these completed within three months (accounting for the summer break).

---

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

*This manuscript reports a very thorough and careful study of the mobility of Bicoid in the early embryo, explored with single-point fluorescence correlation spectroscopy. Although previous groups have looked into this question in the past, the work presented here is novel and interesting because of the different Bicoid mutants and constructs the authors have examined, in particular with the goal of understanding the role of the protein DNA-binding homeodomain. The authors convincingly show that there is a significant increase in Bicoid dynamics from the anterior to the posterior region of the embryo, and that the homeodomain plays an important role in regulating the protein's dynamics. Their experiments are very well designed and carefully analyzed. The authors also modelled gradient formation to see whether this change in dynamics might play a role in setting the shape of the gradient. I am not sure I fully agree with their conclusion that it does, as mentioned in my comment below. However, it is an interesting discussion to have, and I think this paper makes a significant advance in our understanding of Bicoid's behavior in the early embryo. *

We thank the Reviewer for their positive comments and their suggestions for improving the manuscript. We will resolve the concerns raised by the reviewer with clarity in the revision. We will also add additional comment in the Discussion regarding the interpretation of our results.

*Major comments: *

• 1) Gradient profile quantification: Some of the conclusions made by the authors rely on the comparison between their model of gradient formation (as captured in the equations in lines 232 and 233) and the Bcd intensity profile measured in the embryos. Since the differences in gradient shape predicted by the different models are very small (see Fig. 3B, which is on a log scale and therefore emphasize small differences, and Fig. 3C), it is very important to understand how reliable the experimental concentration profiles are.*

This is a fair comment. It is worth noting that the key differences between the 1- and 2-component models are only apparent at large distances (and hence low concentrations) from the source.

We performed the quantification of the gradients in a manner similar to the Gregor lab, whereby the midsagittal plane is analysed. We used 488nm illumination (rather than 2-photon, as the Gregor lab does) so our measurements are likely noisier. However, we are not investigating the variability in the gradient here, but the mean extent. We currently correct background with a uniform subtraction, but we appreciate that is not the optimal method.

In the revised manuscript, we will repeat the above experiments using a 2-photon microscope. Further, we will image lines expressing His::mcherry without eGFP under the same imaging conditions to more accurately estimate the background signal. While we expect this to improve the data quality, we do not envisage significant change to the observed profiles based on prior experience.

At the moment, I do not find the evidence that [Bcd] concentration profile is more consistent with a 2-component diffusion model than a 1-component model very strong. A few comments related to this: * * 1a. Line 249, it is mentioned that: "observations ... incompatible with the SDD model". Which observations exactly are incompatible with the SDD model?

The key points are in the preceding paragraph. We will improve the model presentation in the Results and also include further contextualisation in the Discussion.

1b. In Fig. 3D, only the prediction of the 2-component model is shown. What would the simple 1-component diffusion model look like? Is it really incompatible with the data?

We agree with this comment and will provide the 1-component fit to the gradient profiles. We expect it to fit well for the anterior half of the embryo but fail at larger distances (as has been previously shown).

Regarding the FCS data, we also show one and two component fits. We will show the alternative fits – a 2 particle fit is clearly an improvement (see also related response to reviewer 2).

1c. Line 243: "The increased fraction in the fast form ... consistent with experimental observation of Bcd in the most posterior" (Mir et al.)". I am not sure how this is significant, since the simple model also predicts there will be Bcd in the posterior - the only difference is how much is there (as shown in Fig. 3C), and it's a very small difference.

The absolute differences are not large between the two models, but due to the observed clustering (Mir et al. 2018), even small differences can have very large effects. In the revision we will provide estimates of the actual concentration differences.

We are performing new experiments with the Fritzsche lab at Oxford to estimate if there is clustering of Bcd. We will also repeat our FCS experiments to validate our key conclusion of AP differences in diffusion of Bcd. These should be completed by the end of the summer.

1d. Since the difference between models is in the posterior region where Bcd concentration is very low, when comparing the models to the data the question of background subtraction is essential. How was the subtracted background (mentioned line 612) estimated?

See above response to the first comment.

1e. Along the same line, were the detectors on the Zeiss LSM analog or photon counting detectors, and how confident can we be that signal is exactly proportional to concentration?

We used PMTs and did not directly do photon counting. But the intensity is still proportional to the concentration. It is possible to estimate the absolute concentration value, e.g., Zhang et al., 2021 (https://doi.org/10.1016/j.bpj.2021.06.035). However, our main conclusions – especially regarding the spatially varying Bcd dynamics – are not dependent on this.

1f. Can the gradients created by the two Bcd mutants (FIg. 4B) be quantified as well, and are they any different from the original Bcd gradient?

We agree this would be useful. We will provide the gradient quantifications of the bcd mutants in the revision.

1e. What is the pink line in Figure 5C (I am assuming the green one is the same as in Fig. 3D)? It could be better to not use normalization here, or normalize everything respective to the eGFP::Bcd data to make comparison in relative concentrations in the posterior for different constructs more evident (also maybe different colors for the three different data sets would help clarity).

This is a fair comment, and we will create graphs with new data for better visualisation.

1f. Discussion, lines 402-403: Does the detailed shape of the Bcd in the posterior region matter at all, since the posterior is not a region where Bicoid is active, as far as we know? Could a varying Bcd dynamics have other consequences that would be more biologically relevant?

Bcd is now known to act at 70% EL (Singh et al., Cell Reports 2022). So, the gradient is relevant for a large extent of the embryo length, though it is not known if there is any effect in the most posterior region.

2) Model for gradient formation (lines 231-238): * * 2a. Whether the molecules of Bcd can change from their fast to slow form is never questioned. How do we know (or why might we suspect) they do exchange?

This is a good point. Within the nucleus, and based on our mutant data, we suspect the fast/slow forms correspond to unbound/bound DNA states.

In the cytoplasm, the dynamics are less clear. Bcd can bind to cytoskeletal elements (Cai et al., PLoS One 2017) as well as to Caudal mRNA. Therefore, it seems reasonable to have different effective dynamic modes – yet, how such switching occurs remains unclear.

Ultimately, our model approximates multiple dynamic modes that are integrated to drive Bcd motion. Including switching between states is a reasonable assumption based on what is known about cytoskeletal and protein dynamics, but we do not have a specific mechanism.

It is challenging to estimate a specific kon / koff rate, as the dynamic changes also depend on the diffusion – which itself is changing. For now, we believe our level of abstraction is appropriate given what is known about the system. It will be very interesting to explore the specific interactions underlying such behaviour in the future, but that is beyond this current manuscript.

2b. The values used in the model for alpha, beta_0 and rho_0 should be mentioned. Maybe having a table with all the parameters in the method section, or even in the supplementary section, would help. The exact values of alpha and beta matter, because if they are large (fast exchange) a single exponential gradient is to be expected, if they are 0 (no exchange) a double exponential gradient is to be expected, with intermediate behavior in between. Which case are we in here?

We agree and will add a more complete table in the revision.

3) Discussion about anomalous diffusion (lines 386-388): The 2-component model used by the authors to interpret their FCS data seems very well justified here (excellent fits with very small residuals). I agree with the authors' conclusion that "the dynamics of Bcd within the nucleus are more complicated than a simple model of bound versus unbound Bcd", but I don't see how that can lead to a diagnostic of anomalous diffusion instead. Maybe it is just a matter of exactly explaining what is meant by anomalous diffusion here (since this term is often used to mean different things). A more likely scenario I think, is that there are more than just two Bcd components in the system.

This is a good point, and we can’t easily differentiate two/multi- component fits from anomalous diffusion ones. This is a known problem. But we have recently shown in a collaboration with the Laurent Heliot lab (Furlan et al, Biophys J 2019), that anomalous diffusion is a good stable indicator of changes, even if it might not be the right model. We use anomalous diffusion as it stably predicts changes. We do not claim, however, that diffusion is anomalous. We will improve the discussion of these points in the revised manuscript.

4) Line 440 and after: What is the evidence that the transition between the two forms might vary non-linearly with Bcd concentration? How would that help adapt to different embryo sizes? It would be good to be more explicit here instead of just referring to another paper.

We will improve this discussion. The central point is that the action of Bicoid is unlikely to simply depend linearly on concentration as in that case the ratio of fast to slow forms would be constant across the embryo. Related to the above comment, it is important to emphasise that we are using a phenomenological model, not one based on a specific mechanism.

5) Since an important aspect of this work is the study of different Bcd constructs in vivo, it is important that these constructs are very clearly described, so the section on the generation of the fly lines (Methods) should be expanded. In particular: * * 5a. It seems that the eGFP:: NLS control used here was different from that first described in Ref. 64 (and used for FCS experiments in Ref. 30 and 36)? If so, what NLS sequence was used here, and precisely what type of eGFP was used (in particular, was the A206K mutation that prevents dimerization present in the eGFP used)? If it is the same construct as in Ref. 64, it should be mentioned explicitly. * * 5b. Were the mutant N51A and R54A lines gifts as well, or have they been described before? If so, previous publications should be referenced. If not, how the plasmid was introduced in the embryo should be briefly explained.

We agree and will expand on the fly lines in the revision.

6) Concentration calibration measurements (Methods Fig. 2, line 568 and on). It is well known that background noise is going to interfere with the measurement of N when the signal becomes equivalent to the background noise (Koppel 197, Phys Rev A 10:1938-1945, and for a recent discussion of this effect for morphogens in fly embryos: Zhang et al., 2021, Biophysical Journal 120,4230-4241). It is almost certain that in the low signal regions of the embryo (e.g. posterior cytoplasm) this is affecting the reported concentration, and should be at least acknowledged.

We agree with the reviewer. We will provide the SBR. We will also correct the N values based on the method followed in Zhang et al., 2021, Biophysical Journal 120,4230-4241.

*7) Reference 3 is mis-characterized in two different ways in the manuscript: * * 7a. Line 50: The conclusion in Ref. 3 was not that the gradient was due to a diffusive process, on the contrary Gregor et al. argued that Bcd was too slow to form such a long-range gradient by diffusion. Studies that do present data consistent with a morphogen gradient formation mechanism driven by diffusion are reference 5, reference 30, Zhou et al., Curr. Biol. 2012;22(8):668-75 and Müller et al., Science 336 (2012) 721-724. *

Gregor et al., do not argue against a diffusion process – indeed, they utilise a SDD model in their paper. However, they do extensively discuss how the predicted dynamics from the SDD model are not compatible with gradient formation as observed after n.c. 13. This problem was resolved to some degree by FCS measurements of Bcd (e.g., Dostatni lab, Development 2011) and the use of a Bcd tandem reporter which showed that production and degradation change during n.c. 14 (Durrieu et al., MSB 2018). We will improve the framing of these results in the revision.

7b. The diffusion coefficient estimated from FRAP measurements and reported in Ref. 3 (D = 0.4 micron^2/s) is mentioned a couple of times in the manuscript (line 66, line 395, line 411). However, this number is simply incorrect. When fast components (such as the ones clearly detected here by FCS) are present, they diffuse out of the photobleached area during the photobleaching step. If that is not corrected for during the analysis (and it wasn't in Ref. 3), then the recovery time measured is just equal to the photobleaching time, and has nothing to do with either the fast or slow fraction of the studied molecule - it has no other meaning than to give a lower bound on the value of the actual effective diffusion coefficient of the molecule. This effect (called the halo effect) is well known in the FRAP community (see e.g. Weiss 2004, Traffic 5:662-671), it has been experimental demonstrated to occur for Bcd-eGFP in the conditions used in Ref. 3 (Reference 30), and the actual diffusion coefficient that should have been extracted from the data presented in Ref. 3 has been recalculated by another group to be instead D = 0.9 micron^2/s (Castle et al., 2011, Cell. Mol. Bioeng. 4:116-121). It would therefore be better to report the corrected value from Castle et al. to help the field converge towards an accurate description of Bcd mobility.

We fully agree and will use the improved FRAP estimated value for Bcd.

*Minor comments and suggestions: *

• 8) Figure 1: From panel A, it seems that what is called "Anterior" and "Posterior" is about 150 micron away from the embryo mid-section, i.e. about 100 micron from either the anterior pole or the posterior pole (so not the tip of the embryo, but somewhere in the anterior half or posterior half). Maybe this should be made clear in the text. *

We have made changes in Figure 1A to indicate the region within which the FCS measurements are carried out. We have added the relevant details in the legend of figure 1 lines 137-138.

*9) Fig. 2A; It might be good to put this graph on a log scale, so that cytoplasmic values are seen more clearly. Also, what about reporting on nuclear to cytoplasmic ratios? *

We will rework on this graph and make necessary changes.

*10) Fig. 2: It could be interesting to plot D_effective as a function of the measured concentration of Bicoid in different locations, since the (interesting) suggestion is made several time that [Bcd] could the a determinant of the protein mobility. *

Our work provides an indication that Bcd concentration is connected to the diffusion. We did this by measuring at two locations. To extend this to a rigorous model would require substantial new measurement along the whole length of the embryo. While interesting, this represents a very large investment of time and lies beyond the current manuscript.

*11) Figure 3B&C: Is the curve for 2-component diffusion (without concentration dependence) for steady-state missing? *

We will clarify in the revision.

*12) Lines 78 and 471: What do the authors mean by "new reagents"? The word reagent evokes a chemical reaction, but there are none here. Do the authors mean new constructs? or new mutants? *

We have changed lines 78 and 479 from “new reagents” to new Bcd mutant eGFP lines”.

*13) Lines 57-59: Another good reference for FCS measurements performed to study the dynamics of a morphogen (in this case Dpp) is Zhou et al., Curr. Biol. 2012;22(8):668-75 *

We added this reference in no.70.

*14) Lines 109-111: A word must be missing. Precisely determined what? *

Precisely measure within cytoplasm, and nuclear compartments and also during interphase stages. We have changed to “precisely measure in the cytoplasmic and nuclear regions during the interphase stages of nuclear cycles (n.c.)12-14.” in line no.111-112.

*15) Line 278: The increase in the slow mode is expected. Maybe explicitly mention why. *

In line 286, we have added “due to the loss of Bcd binding to the DNA”.

*16) Line 282: "with the fast component increasing", maybe replace with "with the diffusion coefficient of the fast component increasing" or "with the fraction of the fast component increasing". *

We have changed line 289 “with the diffusion component of fast component increasing towards the posterior”.

*17) Line 517: Is there a reason why the dorsal surface is always placed in the coverslip? *

We have added these details in line 528-529 in Methods.

*18) Line 524 and on: FCS measurements: What was the duration of each individual FCS measurement? It is great that the exact number of measurements are reported in the supplementary! *

Thank you for the complement. Typically, cytoplasmic measurements are 60secs and nuclear measurements are 20-40s. We have added this in line no.528-529. We also added a column to indicate the duration of each of the measurements in the supplementary tables.

*19) An Airy unit of 120 um seems large in combination with an objective with a NA of 1.2, is there a reason for that? What was the radius of the resulting detection volume? *

Olympus microscopes have a 3x magnification stage in their confocals. This leads to the change in the Airy unit. Otherwise, it would be 40 mm.

*20) Thank you for detailing the reasons behind the choice of excitation power, an important and often omitted details. Where in the excitation path were the values of the laser power measured (before or after the objective?)? *

Thank you for the complement. The laser power is measured before the objective. We removed the objective and measured the laser power in the objective path.

*21) Line 585: "since the brightness of eGFP::Bcd..." do the authors mean the molecular brightness of a single eGFP::Bcd molecule, or the total fluorescence signal? *

It is the total fluorescence signal. We have edited line no.592.

*22) It would be good for reference to mention the approximate value of the molecular brightness recorded for these eGFP constructs at the laser power used. *

We will measure and tabulate in the revised manuscript.

*23) Reference 766: The year (and maybe other things) is missing. *

We have corrected this reference.

24) Figure 2 (Methods): The concentrations shown on the figure should be in nM not uM. * * Thanks for noticing – we have changed.

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

MAJOR POINTS

• 1) FCS measurements and fits *
• a) Please state the duration of each individual FCS measurement. *

In the cytoplasm, the measurements were carried out for 60 secs and in nuclei it is between 20-40s. We could not measure for 60s in the nuclei as the nuclear position fluctuates from its initial position. We will add another column to indicate the duration of FCS measurements in the supplementary tables.

b) The authors acknowledge potential issues with fluorophore photophysics and use different lag time ranges for the calibration dye Atto-488 (0.001 ms in Method Fig. 2) and eGFP (0.1 ms in the main figures). Given the strong influence of different parameters on data interpretation and conclusions, Method Fig. 2 should be repeated with purified eGFP. This is particularly relevant for the noisy FCS measurements in posterior regions.

Performing the experiment with purified eGFP will be a volume calibration. We routinely performed this before each imaging session, and that should be fluorophore independent. As noted by Reviewer 1, it is also important to be clear about background correction. We will provide brightness data for eGFP and background values in the revised manuscript. We can then use this to estimate the corrected concentrations.

We use 0.1 ms to start, as at that point any contribution from the photo-physics should have decayed (0.1 ms is about 3-5 times the day rate of the photophysical process, Sun et al., Analytical Chem 2015).

c) Please explain why no data is shown for "AN" around 0.1 ms lag time in Fig. 1B in contrast to all other figures.

We will add the data for AN from 0.01 in the revised figures.

d) Please state what the estimated diffusion coefficients with one-component model fits are. Please also explain why the fits in Fig. S1E do not reach a value of 1 and why they plateau higher than the experimental data at long lag times. Please constrain the fits to G=1 at 0.1 ms tau and G=0 at 1 s tau to make a fair comparison.

The experimental ACF curves reach 0 at long lag times as would be expected. The one-component fits, however, don’t describe the data well and as a result they do not reach 1 and 0 at short and long lag times, respectively. The fitting is done using a mean-squared estimation of the best approximation of the particular model function to the data. Fixing the parameters can be done, but it will further reduce fit accuracy and deviations will be larger. We will perform this analysis and tabulate the one component fits in supplementary 1 with necessary corrections.

e) Please assess the validity of all multi-component fits by comparing the relative quality of the models to the number of estimated parameters using the Akaike information criterion or similar approaches.

We will provide the values denoting the quality of the fits in the revision. We will provide the 3D 1 particle fit, the 3D 1 particle fit with triplet, the 3D 2 particle fit and the 3D 2 particle fit with triple and will provide appropriate measures of fit quality.

f) Please also present the Bcd-GFP fits with 0.001 ms that are mentioned in line 590, and present the results for the data that did not give comparable tau_D1 and tau_D2 values mentioned in line 593.

We will provide all the curves from 0.001ms in the supplementary. We did not provide these details as we have followed the methods from Abu Arish et al., 2010. As our cytoplasmic and nuclear TauD values match with Abu Arish et al., 2010 and Porcher et al., 2010, we thought the excess data would be redundant.

3) Bicoid gradient and modeling * a) Little et al. 2011 observed that the Bcd gradient decreases around n.c. 13. Can the authors of the present work observe a similar concentration decrease using FCS? This is important to i) validate the FCS concentration measurements, and ii) to resolve the controversy regarding "previous claims based on imaging the Bcd profile within nuclei, which predicted decrease in Bcd diffusion in later stages".*

This is a good point regarding conclusions from the previous literature. The Little et al. paper inferred that diffusion had to decrease from fitting to the gradient profiles. However, subsequent analysis from our lab (Durrieu et al., MSB 2018 [which uses a different method involving a tandem reporter for Bicoid] and this manuscript) strongly suggest that Bicoid remains dynamic, at least through n.c. 13 and early n.c. 14. One way to test this is to use SPIM-FCS, where longer time courses can be taken (though with slower time resolution in the FCS). We have performed preliminary experiments with SPIM-FCS and we will revisit this data to see if we can find evidence for changes in the diffusion.

We will also extend the Discussion to make the results clearer in terms of previous models and literature.

b) Please explain why the experimental Bcd-GFP gradient data does not reach a value of 1 (e.g. in Fig. 3D) despite normalization. Please also explain why the fits become flatter in Fig. 5B compared to the steep fit in Fig. 3D.

Both lines were measured under identical conditions. Therefore, we normalised to the maximum value of both experiments. We will redo, normalising to each individual experiment. Regarding Fig. 5C, the Bcd::eGFP curve is identical to Fig. 3D. The flatter curve is the line with eGFP tagged to a NLS alone.

c) For modeling, please take into account observations that the Bcd source is graded with a wide distribution (30-40% EL, see Spirov et al. 2009, Little et al. 2011, Cai et al. 2017 etc.). The extent of the source used in the present work (x_s=20 um, line 620) is at least five times too small.

Care must be taken in defining the source extent. The most careful measurements are reported in Little et al., PLoS Biology 2011 who performed single molecule FISH. They conclude “We demonstrate that all but a few mRNA particles are confined to the anterior 20% of the egg”. Further, the peak in the particle density is around 20-30um from the anterior (Figure 3, Little et al., PLoS Biology 2011), with the vast majority of counts being with 10% of the anterior pole. Further, Durrieu et al. MSB 2018, showed using a Bcd tandem reporter that there was unlikely to be an extended gradient of bcd mRNA (maximum extent of around 50um). Here, we used a simple source domain, which was arguable a little narrow, but not significantly so. We will increase the value in the revision, but the claim that there is an extended bcd mRNA gradient (Spirov et al., Development 2009) has not been substantiated by later experiments.

• d) Please discuss in the paper how well the simulations in Fig. 3B agree with the experimental data.*

We will provide these details in the revision.

• e) Please provide a precise estimate for the statement "Even with an effective diffusion coefficient of 7 μm2s-1, few molecules would be expected at the posterior given the estimated Bcd lifetime (30-50 minutes)" to turn this into a quantitative argument. How many molecules are expected to reach posterior in which model, and how does it compare to experimental observations?*

This can be estimated based on the root-mean-square distance for diffusive processes. We will provide this in the revision.

• f) The sentence "we find that a model of Bcd dynamics that explicitly incorporates fast and slow forms of Bcd (rather than a single "effective" dynamic mode) is consistent with a range of observations that are otherwise incompatible with the standard SDD model" needs to be toned down and corrected since a simple SDD appears to be sufficient to account for the observed gradients. If the authors disagree, please specifically point out in the paragraph around line 249 what observations exactly are incompatible with a standard SDD model.*

This is similar to the point raised by Reviewer 1. While the standard SDD model can explain the overall gradient shape, it is not compatible with the observed time scales and Bcd puncta tracked in the posterior pole. We will improve the Discussion around this point to make the distinctions between the models clearer.

• 5) Data presentation *
• a) In line 27 and 122 it would be better to rephrase the wording "find/found" and give credit to previous papers that first made these observations. *

We will edit in the revision.

• b) For the statement "This suggests that the dynamics of the fast fraction were not captured by previous FRAP measurements", please explain why this should not be the case even though the fast fraction is shown to be larger than the slow fraction in the current work.*

We will edit in the revision.

• c) Similarly, the sentence "The dynamics of the slower mode correspond closely to measured Bcd dynamics from FRAP" likely needs to be corrected since it neglects the contribution of the faster mode, which is fluorescent as well and should also contribute to the dynamics from FRAP.*

This is similar to the point raised by Reviewer 1 and we will edit in the revision.

d) In the absence of further evidence (see above), the sentences "We establish that such spatially varying differences in the Bcd dynamics are sufficient to explain how Bcd can have a steep exponential gradient in the anterior half of the embryo and yet still have an observable fraction of Bcd near the posterior pole" and "These results explain how a long- ranged gradient can form while retaining a steep profile through much of its range" in the abstract need to be toned down.

We are not sure here what needs to be toned down. Our results show that there are (at least) two dynamic forms of Bcd and, combined, they are capable of forming a long-ranged gradient while also ensuring the gradient remains steep in the anterior (because the diffusion coefficient itself varies across the embryo). We will go through these statements and make sure the meaning is clear.

e) The authors state that "However, we show that eGFP::Bcd in its fastest form can move quickly (~18 μm2s-1), and the fraction of eGFP::Bcd in this form increases at lower concentrations", but this has not been directly shown. Please tone down this statement or directly test the prediction that Bcd has a higher fraction of the fast form in earlier nuclear cycles when Bcd concentration is smaller.

This is a good suggestion, and we will test whether early nuclear cycles of the anterior domain show faster dynamics.

*MINOR POINTS * * 1) Introduction * * a) Please explain explicitly what exactly the contention in Bcd, Nodal and Wingless dynamics is in the cited references. *

We will add in the revision. b) In line 95, it would be better to state that this is a variation of the SDD model rather than "a new model". * We changed from “a new model” to “an improved version of SDD model” in the current version of the manuscript. 2) Methods * * a) The authors state that "The same software was also used to calculate the cross-correlation function", but I couldn't find any cross-correlation analyses. Please clarify. *

It is line 538. There is no cross correlation. We changed this to the autocorrelation function.

b) Please correct the "uM" typo to "nM" in the legend of Method Fig. 2A.

We have changed this in the current version.

• c) In the sentence "Further, since the brightness eGFP:Bcd in the anterior and posterior cytoplasm is lower compared to the nuclei", "brightness" probably needs to be changed to "concentration" since the molecular brightness is unlikely to change. *

We edited the line no.591.

• d) Please explain the background-correction method mentioned in line 612. Please also state at what temperature the experiments were performed.*

We will add a better background correction in the revision. Currently, it is the non-embryo background as background noise. The measurements are carried out at 25oC.

*3) Results * * a) Please provide labels for anterior, posterior, dorsal and ventral in Fig. 1A. * * b) Please explain the colors in Fig. 5C. * * c) Please explain the dashed lines in Fig. 3C. * We have edited Figure 1A and Figure 5C. We will edit Figure 3C in further revision.

*OPTIONAL * * 1) If possible, it would be helpful to mention whether the transgenic animals have any abnormal phenotypes or whether they can rescue the bcd mutant. * We will update in the revision.

*2) To validate the concentration measurements, it would be ideal if the authors could determine the Bcd concentration gradient using FCS along the anterior-posterior axis. This would also address whether there are further unexpected changes in diffusivity in medial regions and along the anterior-posterior axis that would have to be considered for modeling. * To measure the Bcd concentration using FCS along the whole axis would be a very challenging undertaking. To get the data for the two positions analysed already represents a significant amount of work. We have done SPIM-FCS measurements, and we will be repeating our FCS measurements in the Fritzsche lab at Oxford. Combined, we believe this provides sufficient corroboration of our results.

*3) Local photoconversion experiments, e.g. in Bcd-Dendra2 embryos if available, would provide compelling support for the relevance of the measurements in the current work. * This is a nice idea, but this would represent a substantial project in its own right and lies beyond the current work.

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

*In my estimation the experimental work is rigorous and the results fully support the conclusions of the authors. I was surprised, however, that the HD-only form localizes via very different and simpler dynamics than does full-length Bcd, but nevertheless forms at least a qualitatively similar gradient. That leads to the question as to whether the existence of the fast and slow forms and their different ratios in different parts of the embryo actually are physiologically relevant. I don't see a straightforward way to test this experimentally, because the mutations that effect Bcd gradient formation also affect essential functions of the protein that if abrogated produce severe downstream effects on embryonic development and lethality. However I would like to see this point at least addressed in the discussion. The data and the methods are presented in such a manner that they can be reproduced, and the number of replicates and statistical analysis is overall robust. * We thank the Reviewer for the positive and constructive review. They, like both previous reviewers, raise the issue of the model and how it fits with the data. As outlined above, we will improve this part of the data presentation and also the Discussion to make sure the main results are clear.

We agree that the underlying importance of the different dynamic forms of Bicoid – and why they change across the embryo – remains unknown. We believe that our careful characterisation of such behaviour is important nonetheless, as it reveals that: (1) morphogen dynamics are more complicated than typically modelled, and this may be just as relevant for ligands moving through extracellular space; and (2) dynamics can vary in space/time, providing an additional possible mechanism of control for regulating morphogen gradient profiles.

Of course, we would like to explore potential physiological relevance. Further exploration of the homeodomain and its role in regulating dynamics is a potential route, but that belongs in future work.

*Minor comments: *

• The presentation of the graphical data measuring Bcd levels along the a-p axis (Fig 1C, 1D, 4C-F and others) needs to be improved, because the grey lines that represent ACF curves are essentially invisible. This is partly because there is usually extensive overlap between the grey lines and other lines. This may be solved by using a more vivid colour than grey for the ACF curves, or perhaps the ACF lines could be made thicker but with some transparency so that overlapping data can be seen. In any event this aspect of the presentation needs to be improved. * We have made the ACF lines thicker to distinguish from the model fit.

*In Figs 2D and 2I measurements of statistical significance between the proportion of protein in fast and slow modes need to be added. * We will add in the revision.

*Relevant to line 174 and Fig 2, NLS should be defined when first used, the source of the NLS should be given (is it from Bcd?) and the rationale for looking at eGFP::NLS should be made explicit. *

We have added details on how the eGFP::NLS is generated in the methods.

*In Fig 3D the dashed lines need to be defined. I assume these are experimental error bars but this is not stated. *

We now state this in the legends.

*On lines 344-5, shouldn't this conclusion concern the HD rather than the NLS? * Yes, thanks for pointing it out it is related to only NLS not NLSHD. We removed this statement from line 351.

*On line 432, CAP is not an acronym, the correct term is 5' 'cap' or 'cap structure'. Also Cho et al. PMID 15882623 should be added to the references here. * We changed the corresponding section and added the references.

*On lines 446, 456, 469, and throughout: replace 'blastocyst' with 'blastoderm'. The former term is generally used for embryos that undergo full cellular divisions and cleavage in early embryogenesis, not for syncytial embryos such as Drosophila. * We have changed blastocyst to blastoderm throughout the manuscript.

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

Major comments: The averaged autocorrelation curves were fitted to models of diffusion with one and two components. The one-component model was insufficient to reproduce the data and the two-component model seems to fit the data. Have the authors tested models with more than two components? Could it be possible to distinguish more Bcd populations?

While it is possible to fit with further components, it rarely provides useful further insight. In particular, the error in measuring three tau_D’s is typically very large. In addition, the improvement in the fit will be marginal, and thus the extra components cannot be justified statistically. Of course, we cannot exclude a third (or more) possible dynamic modes, but within the resolution of our FCS measurements two components with triplets are in general the maximum that can be accommodated without overfitting. We will provide evidence for this claim in the supplement of the revised manuscript.

In Figure 2E, the same concentration of eGFP::NLS is estimated to exist in the cytoplasm and nucleus. Since the NLS should target eGFP to the nucleus, what is the explanation for this observation? Is it possible that the method used to estimate the concentration of molecules is underestimating the concentration in the nucleus or the opposite in the cytoplasm?

This is a good observation. There are two possible explanations. First, the regular division cycles “reset” the nuclear levels. Therefore, differences may not be so large. Second, FCS measurements of concentration can be noisy, as they depend on the very short time scales in the measurement. We will double check our measurements and clarify this in our revision.

*In the simulation of the SDD model (Figure 3B), simulations at 10 min, 25 min and 120 min are shown. Assuming that 120 min corresponds to early nc14, are simulations at earlier timepoints corresponding to nc12 and nc13 indistinguishable from the profile at 120 min? This demonstration would further support the option to merge the data from all nuclear cycles. *

This is a good point. Here, we were primarily focused on showing the time evolution of the model, rather than directly mapping onto experiment. We will clarify in the revision.

*The results obtained with the BcdN51A mutant show an increase in diffusion speed, while retaining similar proportions of fast and slow populations. In the slow fraction, a new population is found. Assuming that the BcdN51A molecules cannot bind specifically to DNA due to the mutation, what would this newly found population correspond to? Could the authors explore the possibility of nonspecific binding to DNA? The article would also win by discussing more on this aspect or other options. *

This is an interesting question. Dslow for anterior nuclei of N51A mutants increases (Dslow from ~0.2um2/s to ~1.5 um2/s), and the proportion is similar to the slow fraction of WT Bcd in the anterior nuclei (F=50%). The Dslow values of bcdWT suggest that 0.2um2/s is a result of DNA binding. For bcdN51A, Dslow of 1.5 um2/s is suggestive of nonspecific interaction of bcdN51A to the DNA. Such a nonspecific interaction is also noticed in the case of NLS::eGFP, where we see a significant amount (Dslow~ 1-1.5 um2/s , F=20%) of slow form in the anterior nuclei, likely due to non-specific interaction with the DNA.

It is worth noting that the inactive homeodomain of transcription factor sex comb reduced (scr) also interacts non-specifically with DNA at high concentration (Vukojevic et al., PNAS 2010). Non-specific interaction of eGFP fluorophore is also noted to be higher in the nuclei of AT-1 cells that suggest “obstacle-free accessible space” is low in the nuclei (Wachsmuth et al., JMB 2000). Therefore, though we do not understand the specific mechanism, our results for N51 mutants are aligned with previous observations of intra-nuclei dynamics.

The experimental rational behind the BcdMM reporter needs to be better explained as it is not clear. It was previously shown that the N51A mutation disturbs zygotic hb activation and Caudal gradient formation (see Figure 3 in Niessing et al., 2000). Since N51A already causes a strong phenotype by disturbing hb expression and Cad gradient formation, what is the reasoning being adding extra mutations to this background? Since the mutations in the PEST domain and YIRPYL motif are involved in cad translational repression, it would be more interesting to add them to the R54A mutation and further study the repression of cad? It would also shed light on the unexpected no difference or even decrease in diffusion in the cytoplasm of the R54A mutant which should increase if indeed the cad mRNA binding is being repressed.

Our rationale was to remove more elements of Bcd to see if there was some degree of redundancy – at least in terms of the dynamics.

The Bicoid homeodomain N51A mutation is physiologically known to cause de-repression of caudal and inhibit hunchback expression. Mechanistically, nuclear Bcd activates hb transcription. However, in the cytoplasm Bcd interacts with other proteins and forms a complex to de-repress caudal. Bcd binds to caudal mRNA through its HD at one end of the complex. However, in the other end, other proteins in the complex are bound to the 5’cap region caudal mRNA. Our rationale for generating the MM mutation was that the N51A mutation may not be sufficient for Bcd to be released from the protein complex. Therefore, additional mutations to N51A may release Bcd from interactions with either DNA or with other proteins through PEST domain and YIRPYL motif.

*Have the authors confirmed that their BcdR54A indeed inhibits cad translation? *

We have not tested the eGFP:bcdR54A to inhibit cad translation. We will add the data in the revision.

*How many embryos of BcdMM were analysed? The authors should also provide a table with all the values in SI as they have done for all the other reporters. *

We will add this data with the revision.

*The claims with eGFP::NLSBcdHD need to be supported by data from multiple embryos. Even if multiple ACF curves are obtained from one embryo, analysing only one embryo is not sufficient. This would clarify the fact that this reporter seems to be able to reproduce the mobility of Bcd in the nucleus. *

We agree and we are arranging to collect more data. This should be completed by the end of the summer.

*According to the methods, all reporters were expressed in a bcd null background, made with the bcd1 allele. This allele is also known as bcd085 and according to Driever and Nusslein-Volhard, 1988 (PMID: 3383244), this allele only causes an intermediate phenotype. This indicates that a truncated version of the protein probably still exists on the embryo. Do the conclusions obtained here still hold if a truncated version of the Bcd protein exists in addition to their reporters? *

We used the bcdE1 mutant, a null mutant of bcd. This was used by Gregor et al., Cell 2007 in their generation of the original Bcd::eGFP. We have also recently generated a more complete bcdKO mutation (Huang et al., eLife 2017). Our embryos do not have a clear phenotype that we can relate to the specific bcd- background used. Nonetheless, we agree it is an important point to be clear about the genetic background and we will clarify in the revised manuscript.

Minor comments: * * In line 45: "Morphogens are signalling molecules", the authors should consider removing the word "signalling" since not all morphogens are, especially the one being studied, Bicoid. * * In lines 80-81 (and also throughout the text): "We measure the Bcd dynamics at multiple locations along the embryo AP-axis", should be more accurate and changed to anterior and posterior of the embryo. Using "multiple locations along the AP axis" is ambiguous and not exact for what was done.

Yes, this is a fair comment. We have edited these sections in the current manuscript.

*Throughout the article, the authors refer multiple times to "modes for/of Bcd transport". Since they or others have not proven that Bcd is being transported, which would involve at least another factor, the authors should replace transport by movement, diffusion or a similar word with which they are comfortable. *

We have changed transport to movement wherever relevant in the text.

*Suggestion: The authors claim that the Bcd gradient is exponential up to 60% of embryo length. Would this information allow a more precise calculation of the gradient decay length in the exponential region than the 80-100µm stated on line 202? *

This is an interesting point, but our results suggest that the idea of the decay length is not so applicable in the posterior region. There, the Bcd dynamics are generally quicker, thereby increasing l. Of course, we cannot discount possible spatial variation in degradation. However, in previous work, our Bcd tandem reporter (which is sensitive to changes in degradation) did not reveal spatial variation in degradation.

In lines 258-259, the sentence "Further, Bcd binds to caudal mRNA, repressing its expression in the cytoplasm" should be improved to clarify the role of Bcd in caudal mRNA translation repression and references should be added. This should also be corrected in the following paragraph.

We will add the necessary corrections in the revision.

*In line 262, "mutations" should be singular since it corresponds to only one amino acid mutation. *

We have corrected this.

*Figure 4J needs to be corrected as the fractions of the slow and fast populations do not correspond to what is shown in Table 3. For example, Fslow fraction of AC is ~45% in the figure while it is 36% in Table 3. The problem occurs in all fractions. *

We are sorry there is a mislabelling in the corresponding figure. AN is in the place of AC. We have edited figure 4J and removed the mislabelling.

*In the discussion, in lines 379-380, "Given the changing fractions of the fast and slow populations in space, the interactions between the populations are likely non-linear". What is the reasoning for non-linearity and not interchangeability? *

If the interactions between the two populations were linear, then the fraction in each form would be constant across the embryo. Some degree of nonlinearity is required in order to have spatially varying relative populations.

*In line 432 caudal should be italicized. *

We have edited this.

*In the discussion, the authors conclude that "In the nucleus, the two populations can be largely (though not completely) explained by Bcd binding to DNA". The discussion would win by explaining all the possible options. * We will add the necessary changes in the discussion. This is also related to above reviewer 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 #4

Evidence, reproducibility and clarity

Summary:

Athilingam et al. are interested in understanding how the Bicoid (Bcd) morphogen gradient is formed in the early stages of Drosophila embryonic development. Using fluorescence correlation spectroscopy (FCS), the authors quantify the dynamics of Bcd in nuclei and cytoplasm at anterior and posterior regions of the embryo. First, they characterize the dynamics of eGFP::Bcd in space and time. Analysing FCS autocorrelation curves at the anterior and posterior regions of the embryo during the interphases of nuclear cycles 12 to 14, they detect differences in Bcd diffusion and are able to infer and distinguish two Bcd populations with slow and fast diffusions. Moreover, these dynamics do not vary between nuclear cycles. Using the different diffusions of the slow and fast populations, they find a model capable of explaining the formation of the gradient at larger distances and the existence of the Bcd molecules in the posterior, compatible with the SDD model widely accepted in the community. Lastly, given that Bcd has multiple roles binding to DNA and also to RNA through its homeodomain, mutations affecting DNA/RNA and RNA only binding are used. Using a mutant without the ability to bind to DNA, their determine that the slower diffusion in the nuclei are due to DNA binding. They further confirm this by fusing the homeodomain of Bcd to eGFP::NLS.

Major comments:

The averaged autocorrelation curves were fitted to models of diffusion with one and two components. The one-component model was insufficient to reproduce the data and the two-component model seems to fit the data. Have the authors tested models with more than two components? Could it be possible to distinguish more Bcd populations? In Figure 2E, the same concentration of eGFP::NLS is estimated to exist in the cytoplasm and nucleus. Since the NLS should target eGFP to the nucleus, what is the explanation for this observation? Is it possible that the method used to estimate the concentration of molecules is underestimating the concentration in the nucleus or the opposite in the cytoplasm? In the simulation of the SDD model (Figure 3B), simulations at 10 min, 25 min and 120 min are shown. Assuming that 120 min corresponds to early nc14, are simulations at earlier timepoints corresponding to nc12 and nc13 indistinguishable from the profile at 120 min? This demonstration would further support the option to merge the data from all nuclear cycles. The results obtained with the BcdN51A mutant show an increase in diffusion speed, while retaining similar proportions of fast and slow populations. In the slow fraction, a new population is found. Assuming that the BcdN51A molecules cannot bind specifically to DNA due to the mutation, what would this newly found population correspond to? Could the authors explore the possibility of nonspecific binding to DNA? The article would also win by discussing more on this aspect or other options. The experimental rational behind the BcdMM reporter needs to be better explained as it is not clear. It was previously shown that the N51A mutation disturbs zygotic hb activation and Caudal gradient formation (see Figure 3 in Niessing et al., 2000). Since N51A already causes a strong phenotype by disturbing hb expression and Cad gradient formation, what is the reasoning being adding extra mutations to this background? Since the mutations in the PEST domain and YIRPYL motif are involved in cad translational repression, it would be more interesting to add them to the R54A mutation and further study the repression of cad? It would also shed light on the unexpected no difference or even decrease in diffusion in the cytoplasm of the R54A mutant which should increase if indeed the cad mRNA binding is being repressed. Have the authors confirmed that their BcdR54A indeed inhibits cad translation? How many embryos of BcdMM were analysed? The authors should also provide a table with all the values in SI as they have done for all the other reporters. The claims with eGFP::NLSBcdHD need to be supported by data from multiple embryos. Even if multiple ACF curves are obtained from one embryo, analysing only one embryo is not sufficient. This would clarify the fact that this reporter seems to be able to reproduce the mobility of Bcd in the nucleus. According to the methods, all reporters were expressed in a bcd null background, made with the bcd1 allele. This allele is also known as bcd085 and according to Driever and Nusslein-Volhard, 1988 (PMID: 3383244), this allele only causes an intermediate phenotype. This indicates that a truncated version of the protein probably still exists on the embryo. Do the conclusions obtained here still hold if a truncated version of the Bcd protein exists in addition to their reporters?

Minor comments:

In line 45: "Morphogens are signalling molecules", the authors should consider removing the word "signalling" since not all morphogens are, especially the one being studied, Bicoid. In lines 80-81 (and also throughout the text): "We measure the Bcd dynamics at multiple locations along the embryo AP-axis", should be more accurate and changed to anterior and posterior of the embryo. Using "multiple locations along the AP axis" is ambiguous and not exact for what was done. Throughout the article, the authors refer multiple times to "modes for/of Bcd transport". Since they or others have not proven that Bcd is being transported, which would involve at least another factor, the authors should replace transport by movement, diffusion or a similar word with which they are comfortable. Suggestion: The authors claim that the Bcd gradient is exponential up to 60% of embryo length. Would this information allow a more precise calculation of the gradient decay length in the exponential region than the 80-100µm stated on line 202? In lines 258-259, the sentence "Further, Bcd binds to caudal mRNA, repressing its expression in the cytoplasm" should be improved to clarify the role of Bcd in caudal mRNA translation repression and references should be added. This should also be corrected in the following paragraph. In line 262, "mutations" should be singular since it corresponds to only one amino acid mutation. Figure 4J needs to be corrected as the fractions of the slow and fast populations do not correspond to what is shown in Table 3. For example, Fslow fraction of AC is ~45% in the figure while it is 36% in Table 3. The problem occurs in all fractions. In the discussion, in lines 379-380, "Given the changing fractions of the fast and slow populations in space, the interactions between the populations are likely non-linear". What is the reasoning for non-linearity and not interchangeability? In line 432 caudal should be italicized. In the discussion, the authors conclude that "In the nucleus, the two populations can be largely (though not completely) explained by Bcd binding to DNA". The discussion would win by explaining all the possible options.

Significance

The results presented in this article advance the knowledge of the field by adding data and quantifications of Bcd mobility at four locations: anterior nucleus, posterior nucleus, anterior cytoplasm and posterior cytoplasm. Until now, FCS studies have focused mostly on measuring the dynamics of Bcd in nuclei at the anterior (Abu-Arish et al. 2010; Porcher et al. 2010) of the embryo. The results are also consistent with what was previously found for eGFP:Bcd in the anterior nucleus. Still, this is not surprising as they use the same reporter as the previous studies.

This article will be interesting to an audience comprising biologists and biophysicists interested in protein diffusion.

As a biologist, I do not have sufficient expertise to completely evaluate if the modelling is performed flawlessly. However, in my understanding, the FCS analysis is crucial for the results and their conclusions, hence the comment on the certainty of the existence of only two Bcd populations, though these populations being previously described with FCS. Comments from a physicist with experience in analysing FCS data are thus necessary.

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 presents experiments that combine a number of techniques to produce an extremely detailed view of Bcd gradient formation in the early Drosophila embryo. The authors provide both empirical and theoretical evidence that support a conclusion that Bcd has both slow and fast moving forms, that the proportion of fast vs slow Bcd is higher at the posterior of the embryo, and that a theoretical model involving both fast and slow moving components fits empirical observations substantially better than a simple diffusion model. Next, they extend the work by investigating what functional motifs of Bcd are required for gradient formation. They demonstrate that a missense mutation N51A in the homeodomain (HD) that affects both DNA binding and cad regulation has major effects on Bcd gradient formation, while the R54A mutation that only affects cad regulation does not. They further show that a protein composed of only an NLS and the Bcd HD can form a gradient that is similar to full-length Bcd, although the two-component dynamics are not recapitulated by the HD-only form.

In my estimation the experimental work is rigorous and the results fully support the conclusions of the authors. I was surprised, however, that the HD-only form localizes via very different and simpler dynamics than does full-length Bcd, but nevertheless forms at least a qualitatively similar gradient. That leads to the question as to whether the existence of the fast and slow forms and their different ratios in different parts of the embryo actually are physiologically relevant. I don't see a straightforward way to test this experimentally, because the mutations that effect Bcd gradient formation also affect essential functions of the protein that if abrogated produce severe downstream effects on embryonic development and lethality. However I would like to see this point at least addressed in the discussion. The data and the methods are presented in such a manner that they can be reproduced, and the number of replicates and statistical analysis is overall robust.

Minor comments:

The presentation of the graphical data measuring Bcd levels along the a-p axis (Fig 1C, 1D, 4C-F and others) needs to be improved, because the grey lines that represent ACF curves are essentially invisible. This is partly because there is usually extensive overlap between the grey lines and other lines. This may be solved by using a more vivid colour than grey for the ACF curves, or perhaps the ACF lines could be made thicker but with some transparency so that overlapping data can be seen. In any event this aspect of the presentation needs to be improved.

In Figs 2D and 2I measurements of statistical significance between the proportion of protein in fast and slow modes need to be added.

Relevant to line 174 and Fig 2, NLS should be defined when first used, the source of the NLS should be given (is it from Bcd?) and the rationale for looking at eGFP::NLS should be made explicit.

In Fig 3D the dashed lines need to be defined. I assume these are experimental error bars but this is not stated.

On lines 344-5, shouldn't this conclusion concern the HD rather than the NLS?

On line 432, CAP is not an acronym, the correct term is 5' 'cap' or 'cap structure'. Also Cho et al. PMID 15882623 should be added to the references here.

On lines 446, 456, 469, and throughout: replace 'blastocyst' with 'blastoderm'. The former term is generally used for embryos that undergo full cellular divisions and cleavage in early embryogenesis, not for syncytial embryos such as Drosophila.

Significance

General assessment: The main strength of the paper is that it extends our knowledge about the dynamics of Bcd gradient formation, and by so doing it advances our understanding of the physical parameters underlying morphological gradients and patterning. The authors are admirably open about questions that were unanswered by the study, which include identifying molecular interactions that affect Bcd dynamics in the cytoplasm, whether Bcd diffusion is dependent on its concentration, and whether other morphogens form gradients in a similar manner. Answering any of these questions would involve substantial experimental work and it is appropriate to leave these questions for subsequent manuscripts.

Advance: A number of high-profile papers were published on this topic in the late 2000s and early 2010s that reached difficult conclusions, so the nature of the mechanisms underlying Bcd gradient formation have remained controversial and somewhat opaque. While this paper does not provide a definitive answer to all relevant open questions, it nevertheless represents a significant advance toward their resolution.

Audience: This paper will be of interest both to developmental biologists interested in gradients and pattern formation, and to biophysicists interested in physical parameters affect molecular movements. It is fundamental research that does not have an obvious clinical or translational component.

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, Athilingam et al. investigated Bicoid (Bcd) protein dynamics along the anterior-posterior axis of Drosophila embryos. They performed Fluorescence Correlation Spectroscopy (FCS) experiments to analyze Bcd protein diffusion and found that Bcd had a 1.7-fold higher mobility in the posterior compared to the anterior. The authors also generated Bcd homeodomain mutations and analyzed their impact on gradient formation. They found that the BcdN51A mutation exhibited altered nuclear dynamics with faster diffusion, while cytoplasmic dynamics remained unchanged. BcdR54A embryos showed dynamics more similar to the wild-type Bcd, with a minor decrease in slow diffusion in the posterior region. Interestingly, the Bcd homeodomain alone, when fused to eGFP-NLS, was able to replicate the observed Bcd protein dynamics, particularly in the slower diffusive mode. The paper provides evidence that the Bcd homeodomain has a significant influence on protein dynamics and suggests a complex interplay of interactions between Bcd, nuclear DNA, and cytoplasmic RNA that regulate Bcd diffusion and function.

Overall, the data and methods presented in the paper are clearly described. The experiments are adequately replicated, and the statistical analysis appears sufficient. Prior studies are referenced appropriately, and the claims and conclusions are mostly supported by the data. However, the following points should be addressed to corroborate the conclusions:

Major points

1. FCS measurements and fits
   • a) Please state the duration of each individual FCS measurement.
   • b) The authors acknowledge potential issues with fluorophore photophysics and use different lag time ranges for the calibration dye Atto-488 (0.001 ms in Method Fig. 2) and eGFP (0.1 ms in the main figures). Given the strong influence of different parameters on data interpretation and conclusions, Method Fig. 2 should be repeated with purified eGFP. This is particularly relevant for the noisy FCS measurements in posterior regions.
   • c) Please explain why no data is shown for "AN" around 0.1 ms lag time in Fig. 1B in contrast to all other figures.
   • d) Please state what the estimated diffusion coefficients with one-component model fits are. Please also explain why the fits in Fig. S1E do not reach a value of 1 and why they plateau higher than the experimental data at long lag times. Please constrain the fits to G=1 at 0.1 ms tau and G=0 at 1 s tau to make a fair comparison.
   • e) Please assess the validity of all multi-component fits by comparing the relative quality of the models to the number of estimated parameters using the Akaike information criterion or similar approaches.
   • f) Please also present the Bcd-GFP fits with 0.001 ms that are mentioned in line 590, and present the results for the data that did not give comparable tau_D1 and tau_D2 values mentioned in line 593.
2. Bicoid gradient and modeling
   • a) Little et al. 2011 observed that the Bcd gradient decreases around n.c. 13. Can the authors of the present work observe a similar concentration decrease using FCS? This is important to i) validate the FCS concentration measurements, and ii) to resolve the controversy regarding "previous claims based on imaging the Bcd profile within nuclei, which predicted decrease in Bcd diffusion in later stages".
   • b) Please explain why the experimental Bcd-GFP gradient data does not reach a value of 1 (e.g. in Fig. 3D) despite normalization. Please also explain why the fits become flatter in Fig. 5B compared to the steep fit in Fig. 3D.
   • c) For modeling, please take into account observations that the Bcd source is graded with a wide distribution (30-40% EL, see Spirov et al. 2009, Little et al. 2011, Cai et al. 2017 etc.). The extent of the source used in the present work (x_s=20 um, line 620) is at least five times too small.
   • d) Please discuss in the paper how well the simulations in Fig. 3B agree with the experimental data.
   • e) Please provide a precise estimate for the statement "Even with an effective diffusion coefficient of 7 μm2s-1, few molecules would be expected at the posterior given the estimated Bcd lifetime (30-50 minutes)" to turn this into a quantitative argument. How many molecules are expected to reach posterior in which model, and how does it compare to experimental observations?
   • f) The sentence "we find that a model of Bcd dynamics that explicitly incorporates fast and slow forms of Bcd (rather than a single "effective" dynamic mode) is consistent with a range of observations that are otherwise incompatible with the standard SDD model" needs to be toned down and corrected since a simple SDD appears to be sufficient to account for the observed gradients. If the authors disagree, please specifically point out in the paragraph around line 249 what observations exactly are incompatible with a standard SDD model.
3. Data presentation
   • a) In line 27 and 122 it would be better to rephrase the wording "find/found" and give credit to previous papers that first made these observations.
   • b) For the statement "This suggests that the dynamics of the fast fraction were not captured by previous FRAP measurements", please explain why this should not be the case even though the fast fraction is shown to be larger than the slow fraction in the current work.
   • c) Similarly, the sentence "The dynamics of the slower mode correspond closely to measured Bcd dynamics from FRAP" likely needs to be corrected since it neglects the contribution of the faster mode, which is fluorescent as well and should also contribute to the dynamics from FRAP.
   • d) In the absence of further evidence (see above), the sentences "We establish that such spatially varying differences in the Bcd dynamics are sufficient to explain how Bcd can have a steep exponential gradient in the anterior half of the embryo and yet still have an observable fraction of Bcd near the posterior pole" and "These results explain how a long- ranged gradient can form while retaining a steep profile through much of its range" in the abstract need to be toned down.
   • e) The authors state that "However, we show that eGFP::Bcd in its fastest form can move quickly (~18 μm2s-1), and the fraction of eGFP::Bcd in this form increases at lower concentrations", but this has not been directly shown. Please tone down this statement or directly test the prediction that Bcd has a higher fraction of the fast form in earlier nuclear cycles when Bcd concentration is smaller.

Minor points

1. Introduction
   • a) Please explain explicitly what exactly the contention in Bcd, Nodal and Wingless dynamics is in the cited references.
   • b) In line 95, it would be better to state that this is a variation of the SDD model rather than "a new model".
2. Methods
   • a) The authors state that "The same software was also used to calculate the cross-correlation function", but I couldn't find any cross-correlation analyses. Please clarify.
   • b) Please correct the "uM" typo to "nM" in the legend of Method Fig. 2A.
   • c) In the sentence "Further, since the brightness eGFP:Bcd in the anterior and posterior cytoplasm is lower compared to the nuclei", "brightness" probably needs to be changed to "concentration" since the molecular brightness is unlikely to change.
   • d) Please explain the background-correction method mentioned in line 612. Please also state at what temperature the experiments were performed.
3. Results
   • a) Please provide labels for anterior, posterior, dorsal and ventral in Fig. 1A.
   • b) Please explain the colors in Fig. 5C.
   • c) Please explain the dashed lines in Fig. 3C.

Optional

1. If possible, it would be helpful to mention whether the transgenic animals have any abnormal phenotypes or whether they can rescue the bcd mutant.
2. To validate the concentration measurements, it would be ideal if the authors could determine the Bcd concentration gradient using FCS along the anterior-posterior axis. This would also address whether there are further unexpected changes in diffusivity in medial regions and along the anterior-posterior axis that would have to be considered for modeling.
3. Local photoconversion experiments, e.g. in Bcd-Dendra2 embryos if available, would provide compelling support for the relevance of the measurements in the current work.

Significance

The paper investigates the diffusion dynamics of Bicoid (Bcd), a transcription factor crucial for establishing the anterior-posterior axis during Drosophila embryogenesis. The authors utilize Fluorescence Correlation Spectroscopy (FCS) and various Bcd mutants fused with eGFP to understand the role of Bcd's homeodomain and other domains in its nuclear and cytoplasmic diffusion dynamics. They performed elegant experiments using insightful transgenic lines, showcasing well-designed and well-executed methodology. The paper is very nice to read, with a clear and engaging writing style and excellent presentation of the data, making it easy to follow and understand their findings.

However, there are a few limitations to consider. First, the paper does not provide evidence for the switching between slow and fast populations central for their modeling, leaving an important aspect of the dynamics unexplained. Second, there are doubts regarding the accuracy of the model used to fit the FCS data (see detailed comments in the section "Evidence, reproducibility and clarity"), underscored by the statement "While the increase in the slow mode was expected, the reason for the change in the fast mode is less clear". Third, the relevance of a potentially higher Bcd mobility for gradient formation remains unclear. For example, the fit in Fig. 3D deviates substantially from the data around 0 um embryo length, which is likely even larger than the error expected from a "simple diffusion" fit at the posterior end of the embryo. In addition, in Fig. 3C the lines for the different models appear to be indistinguishable given the noisy measurements, calling the relevance of the findings into question.

Overall, the paper extends our knowledge by providing new insights into the role of the Bcd homeodomain in determining Bcd gradient formation. The paper highlights that homeodomain interactions in the cytoplasm and nuclei are significant contributors to determining Bcd dynamics. Additionally, the paper suggests that additional components within Bcd itself or other proteins in the cytoplasm affect Bcd dynamics at different Bcd concentrations. The paper will be of interest to a broad audience in developmental biology, molecular biology, and biophysics. Researchers studying transcription factor dynamics, morphogen gradients, and Drosophila embryogenesis will find this study particularly valuable. While the study is primarily focused on basic research, the insights gained on Bcd diffusion dynamics and the role of the homeodomain may contribute to a broader understanding of transcription factor regulation and function in other systems.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This manuscript reports a very thorough and careful study of the mobility of Bicoid in the early embryo, explored with single-point fluorescence correlation spectroscopy. Although previous groups have looked into this question in the past, the work presented here is novel and interesting because of the different Bicoid mutants and constructs the authors have examined, in particular with the goal of understanding the role of the protein DNA-binding homeodomain. The authors convincingly show that there is a significant increase in Bicoid dynamics from the anterior to the posterior region of the embryo, and that the homeodomain plays an important role in regulating the protein's dynamics. Their experiments are very well designed and carefully analyzed. The authors also modelled gradient formation to see whether this change in dynamics might play a role in setting the shape of the gradient. I am not sure I fully agree with their conclusion that it does, as mentioned in my comment below. However, it is an interesting discussion to have, and I think this paper makes a significant advance in our understanding of Bicoid's behavior in the early embryo.

Major comments:

1. Gradient profile quantification: Some of the conclusions made by the authors rely on the comparison between their model of gradient formation (as captured in the equations in lines 232 and 233) and the Bcd intensity profile measured in the embryos. Since the differences in gradient shape predicted by the different models are very small (see Fig. 3B, which is on a log scale and therefore emphasize small differences, and Fig. 3C), it is very important to understand how reliable the experimental concentration profiles are. At the moment, I do not find the evidence that [Bcd] concentration profile is more consistent with a 2-component diffusion model than a 1-component model very strong. A few comments related to this:
   • 1a. Line 249, it is mentioned that: "observations ... incompatible with the SDD model". Which observations exactly are incompatible with the SDD model?
   • 1b. In Fig. 3D, only the prediction of the 2-component model is shown. What would the simple 1-component diffusion model look like? Is it really incompatible with the data?
   • 1c. Line 243: "The increased fraction in the fast form ... consistent with experimental observation of Bcd in the most posterior" (Mir et al.)". I am not sure how this is significant, since the simple model also predicts there will be Bcd in the posterior - the only difference is how much is there (as shown in Fig. 3C), and it's a very small difference.
   • 1d. Since the difference between models is in the posterior region where Bcd concentration is very low, when comparing the models to the data the question of background subtraction is essential. How was the subtracted background (mentioned line 612) estimated?
   • 1e. Along the same line, were the detectors on the Zeiss LSM analog or photon counting detectors, and how confident can we be that signal is exactly proportional to concentration?
   • 1f. Can the gradients created by the two Bcd mutants (FIg. 4B) be quantified as well, and are they any different from the original Bcd gradient?
   • 1e. What is the pink line in Figure 5C (I am assuming the green one is the same as in Fig. 3D)? It could be better to not use normalization here, or normalize everything respective to the eFP::Bcd data to make comparison in relative concentrations in the posterior for different constructs more evident (also maybe different colors for the three different data sets would help clarity).
   • 1f. Discussion, lines 402-403: Does the detailed shape of the Bcd in the posterior region matter at all, since the posterior is not a region where Bicoid is active, as far as we know? Could a varying Bcd dynamics have other consequences that would be more biologically relevant?
2. Model for gradient formation (lines 231-238):
   • 2a. Whether the molecules of Bcd can change from their fast to slow form is never questioned. How do we know (or why might we suspect) they do exchange?
   • 2b. The values used in the model for alpha, beta_0 and rho_0 should be mentioned. Maybe having a table with all the parameters in the method section, or even in the supplementary section, would help. The exact values of alpha and beta matter, because if they are large (fast exchange) a single exponential gradient is to be expected, if they are 0 (no exchange) a double exponential gradient is to be expected, with intermediate behavior in between. Which case are we in here?
3. Discussion about anomalous diffusion (lines 386-388): The 2-component model used by the authors to interpret their FCS data seems very well justified here (excellent fits with very small residuals). I agree with the authors' conclusion that "the dynamics of Bcd within the nucleus are more complicated than a simple model of bound versus unbound Bcd", but I don't see how that can lead to a diagnostic of anomalous diffusion instead. Maybe it is just a matter of exactly explaining what is meant by anomalous diffusion here (since this term is often used to mean different things). A more likely scenario I think, is that there are more than just two Bcd components in the system.
4. Line 440 and after: What is the evidence that the transition between the two forms might vary non-linearly with Bcd concentration? How would that help adapt to different embryo sizes? It would be good to be more explicit here instead of just referring to another paper.
5. Since an important aspect of this work is the study of different Bcd constructs in vivo, it is important that these constructs are very clearly described, so the section on the generation of the fly lines (Methods) should be expanded. In particular:
   • 5a. It seems that the eGFP:: NLS control used here was different from that first described in Ref. 64 (and used for FCS experiments in Ref. 30 and 36)? If so, what NLS sequence was used here, and precisely what type of eGFP was used (in particular, was the A206K mutation that prevents dimerization present in the eGFP used)? If it is the same construct as in Ref. 64, it should be mentioned explicitly.
   • 5b. Were the mutant N51A and R54A lines gifts as well, or have they been described before? If so, previous publications should be referenced. If not, how the plasmid was introduced in the embryo should be briefly explained.
6. Concentration calibration measurements (Methods Fig. 2, line 568 and on). It is well known that background noise is going to interfere with the measurement of N when the signal becomes equivalent to the background noise (Koppel 197, Phys Rev A 10:1938-1945, and for a recent discussion of this effect for morphogens in fly embryos: Zhang et al., 2021, Biophysical Journal 120,4230-4241). It is almost certain that in the low signal regions of the embryo (e.g. posterior cytoplasm) this is affecting the reported concentration, and should be at least acknowledged.
7. Reference 3 is mis-characterized in two different ways in the manuscript:
   • 7a. Line 50: The conclusion in Ref. 3 was not that the gradient was due to a diffusive process, on the contrary Gregor et al. argued that Bcd was too slow to form such a long-range gradient by diffusion. Studies that do present data consistent with a morphogen gradient formation mechanism driven by diffusion are reference 5, reference 30, Zhou et al., Curr. Biol. 2012;22(8):668-75 and Müller et al., Science 336 (2012) 721-724.
   • 7b. The diffusion coefficient estimated from FRAP measurements and reported in Ref. 3 (D = 0.4 micron^2/s) is mentioned a couple of times in the manuscript (line 66, line 395, line 411). However, this number is simply incorrect. When fast components (such as the ones clearly detected here by FCS) are present, they diffuse out of the photobleached area during the photobleaching step. If that is not corrected for during the analysis (and it wasn't in Ref. 3), then the recovery time measured is just equal to the photobleaching time, and has nothing to do with either the fast or slow fraction of the studied molecule - it has no other meaning than to give a lower bound on the value of the actual effective diffusion coefficient of the molecule. This effect (called the halo effect) is well known in the FRAP community (see e.g. Weiss 2004, Traffic 5:662-671), it has been experimental demonstrated to occur for Bcd-eGFP in the conditions used in Ref. 3 (Reference 30), and the actual diffusion coefficient that should have been extracted from the data presented in Ref. 3 has been recalculated by another group to be instead D = 0.9 micron^2/s (Castle et al., 2011, Cell. Mol. Bioeng. 4:116-121). It would therefore be better to report the corrected value from Castle et al. to help the field converge towards an accurate description of Bcd mobility.

Minor comments and suggestions:

1. Figure 1: From panel A, it seems that what is called "Anterior" and "Posterior" is about 150 micron away from the embryo mid-section, i.e. about 100 micron from either the anterior pole or the posterior pole (so not the tip of the embryo, but somewhere in the anterior half or posterior half). Maybe this should be made clear in the text.
2. Fig. 2A; It might be good to put this graph on a log scale, so that cytoplasmic values are seen more clearly. Also, what about reporting on nuclear to cytoplasmic ratios?
3. Fig. 2: It could be interesting to plot D_effective as a function of the measured concentration of Bicoid in different locations, since the (interesting) suggestion is made several time that [Bcd] could the a determinant of the protein mobility.
4. Figure 3B&C: Is the curve for 2-component diffusion (without concentration dependence) for steady-state missing?
5. Lines 78 and 471: What do the authors mean by "new reagents"? The word reagent evokes a chemical reaction, but there are none here. Do the authors mean new constructs? or new mutants?
6. Lines 57-59: Another good reference for FCS measurements performed to study the dynamics of a morphogen (in this case Dpp) is Zhou et al., Curr. Biol. 2012;22(8):668-75
7. Lines 109-111: A word must be missing. Precisely determined what?
8. Line 278: The increase in the slow mode is expected. Maybe explicitly mention why.
9. Line 282: "with the fast component increasing", maybe replace with "with the diffusion coefficient of the fast component increasing" or "with the fraction of the fast component increasing".
10. Line 517: Is there a reason why the dorsal surface is always placed in the coverslip?
11. Line 524 and on: FCS measurements: What was the duration of each individual FCS measurement? It is great that the exact number of measurements are reported in the supplementary!
12. An Airy unit of 120 um seems large in combination with an objective with a NA of 1.2, is there a reason for that? What was the radius of the resulting detection volume?
13. Thank you for detailing the reasons behind the choice of excitation power, an important and often omitted details. Where in the excitation path were the values of the laser power measured (before or after the objective?)?
14. Line 585: "since the brightness of eGFP::Bcd..." do the authors mean the molecular brightness of a single eGFP::Bcd molecule, or the total fluorescence signal?
15. It would be good for reference to mention the approximate value of the molecular brightness recorded for these eGFP constructs at the laser power used.
16. Reference 766: The year (and maybe other things) is missing.
17. Figure 2 (Methods): The concentrations shown on the figure should be in nM not uM.

Significance

Strengths:

Very careful and systematic study of Bcd's dynamics in the early embryo Use of several mutant and truncated forms of Bcd to pinpoint the importance of the DNA binding domain in setting this dynamics Uncovers a previously unknown change in Bcd dynamics from the anterior to the posterior of the embryo Modelling of the Bcd concentration gradient shape taking into account the measured dynamics

Limitations:

The quantitative comparison between modelled and measured gradient could be improved. The discussion of the biological implications of the work is limited

Advance:

Uncovers a previously unknown change in Bcd dynamics from the anterior to the posterior of the embryo. This raises very interesting questions about molecular mechanisms involving Bicoid. Other studies (cited in this manuscript) reported on Bcd dynamics, but the present study represents a very welcome expansion of these earlier studies, by looking at spatial dependence and by examining several Bcd constructs.

Audience:

Somewhat specialized, as this work should firstly be of interest to scientists studying morphogen gradients. However, it is also a beautiful example dynamical studies in vivo, so it will also be of interest to experimental physicists studying protein motions in vivo (a rather large 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

This manuscript characterizes the ULD mouse model as a new platform for pre-clinical TB vaccine testing. Using the current tuberculosis (TB) vaccine, BCG, the manuscripts shows that three distinct parameters of protective immunity can be assessed in this model: 1) reduction of bacterial burden (which is shown to be more durable in this model than in the conventional model); 2) prevention of dissemination to the contralateral lung; and 3) prevention of detectable infection. The last parameter of protection is notable because vaccines have not been previously shown to be capable of preventing Mycobacterium tuberculosis (Mtb) infection in the mouse model, and in fact, it has been widely believed that mice lack the immune effector mechanisms necessary to prevent detectable infection. We show here that this is not true. When mice are challenged with a physiologic infectious dose of Mtb, vaccine-induced immunity can indeed prevent detectable infection. Thus, we believe this physiologic dose challenge model, provides potential for an improved platform for preclinical vaccine testing, as it allows for measurement of protective parameters that could not previously be assessed and may provide a window to assess meaningful differences between vaccine candidates. We were happy that both reviewers recognized the significance of this work, noting that the study “offers a new avenue for evaluation of TB vaccines, especially vaccines to prevent establishment of long-term infection” and “is clinically useful to test new TB vaccine candidates by improving the conventional TB vaccine model.”

We thank the reviewers for their time and excellent comments. We have addressed the Reviewer’s comments as outlined below. Most could be fully addressed by minor modifications to the manuscript’s texts or figures. Reviewer 2 requested additional studies to further assess the model’s durability at timepoints later than day 125 post-infection. In response to this comment, we have modified the manuscript to soften our conclusions about the model’s durability. However, we do not believe that performing additional experiments (which would take up to a year to perform) to further examine BCG’s durability in this model is necessary to support the manuscript’s conclusions. Changes made to the manuscript in response to the reviewers’ comments are underlined.

This report provides evidence that the ULD infection model of M. tuberculosis provides the capacity to detect 3 types of protection by BCG: 1) durable reductions in Mtb burden, 2) inhibition of Mtb dissemination, 3) prevention of detectable bacteria. In general, this is well-reasoned, well-written, transparently presented and easy to follow. A few points.

Major issues:

In Figure 2A, 2B, 3A - the authors have a highly skewed distribution with a bimodal distribution between detected bacterial counts and zero bacterial counts. This type of distribution does not lend itself to a t-test. What is the mean of two exclusive categories? For these graphs, the authors should consider plotting the median and interquartile range, then applying a non-parametric test. I suspect the p-values will be as significant, if not more, and there will be some comparisons where the median of the BCG group will be 0 CFU.

We completely agree that a t-test would not be appropriate for data with bimodal distribution, such as if the mice with detected bacterial counts and those with zero bacterial counts were both included in this analysis. We apologize that we did not sufficiently explain that only mice with detectable bacterial counts were included in our analysis of BCG’s ability to reduce bacterial burdens; those with zero bacterial counts were excluded. We recognize that doing this may underestimate the ability of a vaccine to reduce bacterial burdens as a handful of mice that might have had detectable bacterial burdens in the absence of vaccination are not included in the analysis. However, including mice all mice with undetectable bacterial burdens would be confounded by the fact that some mice in the ULD model are never infected at all, at least measurably, even in the unvaccinated group. We decided that the best way to disentangle these issues would be to analyze reduction of bacterial burdens and proportion of mice with zero detectable CFUs separately. Thus, for the former, only those with detectable CFUs are considered, and separately, we compare the proportion of mice with mice with undetectable bacterial burdens in the vaccinated mice compared to the unvaccinated controls (Figure 5). For these reasons, we believe the t-test is an appropriate test for this analysis. However, in response to this comment we have made changes to the text, figure legend, and Methods, to clearly state how and why the analysis was done this way. We thank the reviewer for these comments, as we believe the original manuscript was not sufficiently clear in this respect, and it is very important to convey how the analysis is being performed.

In 3B, instead of presenting a model of the data, can the authors present the raw data across all experiments as datapoints with violin plots, or some other form of data visualization? They could present the fixed effect model as a supplementary figure. The fixed effect model works better for plotting proportions in 4A.

We thank the reviewer for this comment. In Figure 3B of the revised manuscript, the raw data across all experiments is now shown as datapoints with violin plots.

Minor points:

Abstract, line 35. The authors state that BCG does A, B, C in a small percentage of mice. It is not clear whether this means all of A, ,B and C happen in a small percentage. Or rather whether the small percentage refers to C alone. Perhaps this can be rewritten for clarity.

We thank the reviewer for this comment. The small percentage was meant to refer to C alone, but we agree this was not clear as it was written. In the revised manuscript, this sentence in the abstract is written as follows, “We show that BCG confers a reduction in lung bacterial burdens that is more durable than that observed after conventional dose challenge, curbs Mtb dissemination to the contralateral lung, and, in a small percentage of mice, prevents detectable infection.”

Line 123. 12/20 vs 10/20. Hard to say it appeared to prevent based on these numbers. Perhaps may prevent?

In the revised manuscript we have changed “appeared to prevent” to “may prevent”, as suggested.

Line 128. The authors use the word colonized here but don't use this term elsewhere. What is the difference between colonized and infected, and why is colonized used only here?

In the revised manuscript we have changed “colonized” to “infected”, as suggested.

The power calculations could be a supplementary table if space is tight.

We are amenable to moving the power calculations to a supplementary table if this is the preference of the editor.

Reviewer #2

Manuscript "Assessing vaccine-mediated protection in an ULD mycobacterium tuberculosis murine model" is an interesting, well-documented and comprehensive study to develop new TB murine model used to assess new TB vaccine candidates. Although TB vaccine are urgently needed, many TB vaccine candidates remain in the development pipeline mainly because the conventional vaccine evaluation strategy is hindered by the lack of reliable animal models that mimic human TB pathogenic cycle. This manuscript used the ULD mouse infection model to resembles human Mtb infection to test the ability of the ULD model as a TB vaccine testing strategy by assessing the BCG vaccination in I. durability in lung bacterial burden, II. the capacity to protect Mtb dissemination to other organs, and III. levels to protect Mtb infection. Overall, this study is quite extensive and potential interest as the result can be readily used for clinical settings.

Major This study started based upon one of the biggest problems of conventional TB infection model, in which the protection efficacy can be misinterpreted as CFU burden dissipates at later time points due to relatively high burden of initial infection load. To propose that vaccine efficacy test outcomes could be better in ULD murine model compared to that of conventional TB infection model as initial infection burden in ULD model is pathogenetically similar to human infection case. This reviewer is concerned about the authors' interpretation of the results as authors monitored all experimental outcomes at maximum day 125 when lung CFU was ~ 50 fold lower than conventional TB model. Because authors didn't monitor longer period of time, it is not clear if the ULD murine model is optimal to prevent lung CFU dissipation or dissemination to other lung lobes or organs. Authors need to provide additional evidence if the ULD model results are still positive to support authors' hypothesis or the BCG vaccine efficacy in the ULD model was attributed simply to yet lower bacterial burden.

We thank the reviewer for these comments. While we agree that it would be interesting to see if the protective effects of BCG immunization were durable even beyond 125 days post-infection, we don’t believe that defining the durability further is necessary to complete the study or to support our conclusions. In many ways, we believe we have been quite comprehensive and rigorous in this study, examining over 1,000 mice at timepoints ranging from 14 to 125 days post-infection. We believe we have conclusively shown that BCG’s reduction of lung bacterial burdens is more durable in the ULD model than with a conventional dose challenge (50-100 CFU); while the difference is maintained out to days 90-125 in the former, it wanes in the latter. Similarly, BCG’s ability to prevent dissemination to the contralateral lung, a parameter that cannot be assessed in the conventional dose model, is also durable to days 90-125. Finally, because we used a large number of mice, we showed for the first time that BCG can prevent detectable infection in mice challenged with physiologic Mtb dose (pIn response to the reviewer’s comments, we have softened our statements regarding showing that BCG confers durable reductions in lung bacterial burdens in the ULD model. Now, throughout the abstract and manuscript, we say that we show that BCG confers a reduction in lung bacterial burdens that is more durable than observed with conventional dose challenge.

Minor Fig. 1 - 10 out of 20 mouse in BCG vaccinated condition didn't show bacterial burden in the lung at any time. It is not clear that this even is attributed to the failed infection or BCG vaccination mediated protection.

We agree. In this same experiment, 7 out of 20 of the unvaccinated control mice also didn’t show bacterial burden in the lung. One of the features of the ULD model is that we use such a low dose that we intentionally leave some of the mice uninfected, even in the unvaccinated controls. We believe that it is necessary to do this to achieve many of the advantages of the model (e.g., assess dissemination to the contralateral lung and prevention of detectable infection), however, an inherent challenge of the model is that in a single experiment you cannot discern whether an individual mouse with no detectable lung bacteria had infection prevented or whether it would never have been infected in the first place. In the manuscript, we do not claim the difference observed in Figure 1 (7/10 vs. 10/20 with zero CFU) is meaningful. We state in lines 123-125 that “we also observed that 7/20 of the unimmunized mice and 10/20 of the BCG-immunized mice had no detectable infection in either lung (Figure 2A), a difference that was not statistically significant in this single experiment (p=0.53).” We go on to show (Figure 5), that if results from several experiments are pooled, the difference becomes highly significant (p

As shown in Fig. 1 and 2A, at 42 day post infection, conventional TB infection model reached 5 X 106 CFU in an unimmunized condition and 5 X 105 CFU in a BCG vaccinated condition. In this ULD model, the CFU was 1 X 105 CFU in an unimmunized condition and 1 X 104 CFU in an BCG vaccinated model even at 63 day post infection. If we directly compare the CFU between ULD and conventional TB infection model, the difference was ~ 50 folds. Authors may need to show the bacterial CFU burden is still plateaued and stable bacterial dissemination even after a longer period of infection.

We have shown that differences in lung bacterial burdens and bacterial dissemination are durable as long as we’ve looked, which is to days 90-125. As discussed above, we believe this is sufficient to support our conclusions and the goals of the study.

Fig. 2 - The conventional TB model may be included as a negative control.

We show results of BCG efficacy in the conventional TB model in Figure 1. Because conventional dose and ULD infections are different doses, they cannot be in the same infection chamber at the same time and therefore they need to be shown as separate experiments. Nevertheless, the results shown in Figure 1 are highly reproducible, as shown by us and by several other groups (as referenced).

Fig. 5A - Why total challenged mouse number gets increased ?

We presume the reviewer is asking why there are more mice challenged at later timepoints than at early timepoints. Our early experiments suggested that there might be relatively more vaccinated mice with undetectable infection at late timepoints than at earlier ones. As a result, we assessed more experiments at late timepoints than at earlier ones.

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

Manuscript "Assessing vaccine-mediated protection in an ULD mycobacterium tuberculosis murine model" is an interesting, well-documented and comprehensive study to develop new TB murine model used to assess new TB vaccine candidates.

Although TB vaccine are urgently needed, many TB vaccine candidates remain in the development pipeline mainly because the conventional vaccine evaluation strategy is hindered by the lack of reliable animal models that mimic human TB pathogenic cycle. This manuscript used the ULD mouse infection model to resembles human Mtb infection to test the ability of the ULD model as a TB vaccine testing strategy by assessing the BCG vaccination in I. durability in lung bacterial burden, II. the capacity to protect Mtb dissemination to other organs, and III. levels to protect Mtb infection. Overall, this study is quite extensive and potential interest as the result can be readily used for clinical settings.

Major

This study started based upon one of the biggest problems of conventional TB infection model, in which the protection efficacy can be misinterpreted as CFU burden dissipates at later time points due to relatively high burden of initial infection load. To propose that vaccine efficacy test outcomes could be better in ULD murine model compared to that of conventional TB infection model as initial infection burden in ULD model is pathogenetically similar to human infection case. This reviewer is concerned about the authors' interpretation of the results as authors monitored all experimental outcomes at maximum day 125 when lung CFU was ~ 50 fold lower than conventional TB model. Because authors didn't monitor longer period of time, it is not clear if the ULD murine model is optimal to prevent lung CFU dissipation or dissemination to other lung lobes or organs. Authors need to provide additional evidence if the ULD model results are still positive to support authors' hypothesis or the BCG vaccine efficacy in the ULD model was attributed simply to yet lower bacterial burden.

Minor

Fig. 1 - 10 out of 20 mouse in BCG vaccinated condition didn't show bacterial burden in the lung at any time. It is not clear that this even is attributed to the failed infection or BCG vaccination mediated protection. As shown in Fig. 1 and 2A, at 42 day post infection, conventional TB infection model reached 5 X 106 CFU in an unimmunized condition and 5 X 105 CFU in a BCG vaccinated condition. In this ULD model, the CFU was 1 X 105 CFU in an unimmunized condition and 1 X 104 CFU in an BCG vaccinated model even at 63 day post infection. If we directly compare the CFU between ULD and conventional TB infection model, the difference was ~ 50 folds. Authors may need to show the bacterial CFU burden is still plateaued and stable bacterial dissemination even after a longer period of infection.

Fig. 2 - The conventional TB model may be included as a negative control.

Fig. 5A - Why total challenged mouse number gets increased ?

Significance

This study is clinically useful to test new TB vaccine candidates by improving the conventional TB vaccine 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 #1

Evidence, reproducibility and clarity

This report provides evidence that the ULD infection model of M. tuberculosis provides the capacity to detect 3 types of protection by BCG: 1) durable reductions in Mtb burden, 2) inhibition of Mtb dissemination, 3) prevention of detectable bacteria. In general, this is well-reasoned, well-written, transparently presented and easy to follow. A few points.

Major issues:

In Figure 2A, 2B, 3A - the authors have a highly skewed distribution with a bimodal distribution between detected bacterial counts and zero bacterial counts. This type of distribution does not lend itself to a t-test. What is the mean of two exclusive categories? For these graphs, the authors should consider plotting the median and interquartile range, then applying a non-parametric test. I suspect the p-values will be as significant, if not more, and there will be some comparisons where the median of the BCG group will be 0 CFU.

In 3B, instead of presenting a model of the data, can the authors present the raw data across all experiments as datapoints with violin plots, or some other form of data visualization? They could present the fixed effect model as a supplementary figure. The fixed effect model works better for plotting proportions in 4A.

Minor points:

Abstract, line 35. The authors state that BCG does A, B, C in a small percentage of mice. It is not clear whether this means all of A, ,B and C happen in a small percentage. Or rather whether the small percentage refers to C alone. Perhaps this can be rewritten for clarity.

Line 123. 12/20 vs 10/20. Hard to say it appeared to prevent based on these numbers. Perhaps may prevent?

Line 128. The authors use the word colonized here but don't use this term elsewhere. What is the difference between colonized and infected, and why is colonized used only here?

The power calculations could be a supplementary table if space is tight.

Significance

This is a strong paper and the data are compelling.<br /> The significance is that this offers a new avenue for evaluation of TB vaccines, especially vaccines to prevent establishment of long-term infection.

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: Major comments: The key point of the manuscript is to provide resources for the plant community. The motivation for selecting these specific promoters, how they were obtained and cloned, what they are in detail and how they will be made publically available is all clearly described. The infection experiments presented in it are an added bonus and a proof of concept of the applicability of the system.

Thank you very much.

Minor comments: The promotor sequences will probably be included in the AddGene submission, however, it might be helpful to also deposit the promoter sequences at e.g. GenBank.

Indeed, we have sent all sequence files to AddGene and they will be available for download there. We will look into transferring them to GenBank as well. We have not done this before, but are generally always supportive of maintaining data in open repositories.

Line 133: "There are few exceptions to this rule...". It would probably helpful to list/mark these exceptions in Table 1

We agree. We have now marked them in the table, and included the sentence “There are a few exceptions to this rule (marked with a * in the ‘Bases’ column in table 2), where we used a defined stretch of DNA that has previously been described to complement a mutant” in lines 135-137.

Line 138: "A overhangs". In the GreenGate system, A-modules (promoters) are flanked by A- (5') and B- (3') overhangs (applies to line 144, too). Also, the B-overhang listed here (TTGT) is the reverse complement, which might be confusing for readers.

A very good point. We have modified these lines to “standard four base pair GreenGate promoter module overhangs (5´-ACCT and TTGT-3´) were added via primers during amplification of the promoter sequences (see Supplementary Table 1 for a list of primer sequences. Note that TTGT is the complementary sequence of the A-to-B-module overhang, as this is added via the reverse primer)” in lines 141-144.

Line 149 ff.: How many lines have been established per promoter tested? Did they all yield a similar expression pattern?

This is indeed a very important point which was somehow lost along the way during manuscript preparations, after being moved around between results and methods section. We have put it back in in lines 162-165 as “We recovered several independent transgenic lines for the PEP1 and 2, PEPR1 and 2, as well as BIK1 and RBOHD reporters. Out of those, a minimum of three (RBOHD) and up to seven (PEPR2) independent lines showed fluorescence, and out of those, all individual lines for each reporter showed the same expression patterns.”

Line 163: As someone not being familiar with microscoping Arabidopsis roots, I'm wondering how the authors can be sure that the tissue in question is the vasculature. Is this obvious for experts in the field?

Of course, we can’t give a totally objective answer here, but we believe that by including the transmitted light image next to the fluorescence image, it is indeed visible that the fluorescence is limited to the center of the root, not the complete circumference. At the same time, it is important to note that all images are stereomicroscopic images, not confocal images. Thus, it is indeed not possible to, e.g., conclude if pericycle cells are included or excluded in the region with expression. So, while it is, we believe, safe to assume that it is vascular cells, we can’t determine which cell types in the vascular cylinder are expressing the reporters. This would require confocal imaging, which would increase the resolution, but at the expense of a good overview, which we think is more valuable for such a proof-of-principle.

Discussion: Is there by any chance prior (cell-resolution) knowledge about the expression behaviour of any of the investigated promoters? E. g. by in-situ hybridizations? If so, do the expression patterns match?

No, the expression of these reporters in direct response to fungal infection have so far only been studied by transcriptomics.

Presentation and quality of the images need be improved. Scale bars are missing in all confocal images. In Figure 3 and 4, the name of genes examined can be labeled on the image, which will make it easier for readers. In addition, key information such as the inoculum and sampling time point after fungal inoculation should be described in the legend or the main text.

We have added the scale bars and gene names into the images. We agree that the gene names make it easier for the reader. Further, we have added the inoculum and sampling time to the legend.

More importantly, a "mock" inoculation or "before fungal inoculation" should be performed to reveal the expression changes of the marker genes after fungal inoculation.

This is information was provided in the text and via the supplemental figures, but I assume we didn’t make it clear that these results and images were indeed specific control/mock experiments, and not some ‘general’ expression analysis. We have now tried to make this clearer, specifically in lines 192-194.

Lines 172-174, the pictures are too small to see these details. The same for BIK1 (line 187).

We have split up figure 3 into two separate figures (figures 3 and 4), to allow for them to be displayed larger, so that more details can be observed. Of course, it would also be helpful to do some confocal microscopy on specific regions of interest of these stereomicroscopic images to obtain high-resolution images of these regions, but, unfortunately, we did not reach this point in this project, before our team was disbanded, and we therefore only have the overview images to get a general idea of the responsiveness of the different reporters.

Line 174-176, which results are these referring to? The same for line 200-203.

We assume that this was not clear because we previously failed to make it clear that the control supplementary figures are from experimental controls/mock. We have reworded both paragraphs to, hopefully, explain it a bit better, and included the supplementary figure number that refers to. It’s now in lines 212-215 and 237-242.

This study provides a valuable collection of vectors/constructs for investigation of transcriptional dynamics of plant immunity genes and should attract broad interest of the plant immunity field.

Thank you very much.

The current study by Calabria et al., entitled "pGG-PIP: A GreenGate (GG) entry vector collection with Plant Immune system Promoters (PIP)," reported the development of a set of GreenGate-compatible entry plasmids that contain promoter sequences of a series of immunity-related genes. This tool enables live-cell observation of immune responses at a cellular resolution. Being compatible with many other GreenGate tools, it opens up a door toward simultaneous visualization of different but overlapping immune pathways and ultimately describes the 4D dynamics of plant immunity. It is more than expected that these constructs will be used by a wide range of researchers and contribute to the ultimate understanding of plant innate immunity.

Thank you very much.

It is exciting that the authors observed the marker expression by a fluorescent stereomicroscope. This allows for non-destructive observation of response over time, keeping the system gnotobiotic. However, it was partly disappointing that the author did not take full advantage of this. It would have been much nicer if the authors observed the infection process over time, such that one could tell when and where the response starts, and whether local and systemic reactions occur simultaneously or instead require local-to-systemic signal transduction. They indeed seem to have done such time-course observation (line 378) however did not provide the results. I am curious to know what the authors could have found from those experiments. It would also be a strong appealing point of this method and is therefore highly encouraged

We absolutely agree that this temporal data would be valuable and interesting. So far, we always imaged the colonization sites in the root tips from the first day when they become visible, until the day when the entire root was colonized/dying. However, we only recorded the infection sites directly, and did not image the entire plants, and local as well as systemic responses. This is, of course, something that we would have liked to do, and planned to do in the future, but, so far, we have not gotten to that point. We also attempted to use the images of the infection sites that we have recorded over time to obtain information about disease progression, e.g., colonization speed of the fungus, but this data is not (yet) at a point, where we feel confident that we have enough information to draw solid conclusions. So, while we absolutely agree that this kind of whole-plant imaging with both, high spatial and temporal resolution, must be the aim, at this point, unfortunately, we simply are not at that place yet.

Immune responses are not always induction of expression but sometimes reduction. Some genes up-regulated in the first phase will also be down-regulated afterward in order to go back to the initial non-responding state. During such down-regulation, the expression of a fluorescence marker gene might not accurately reflect the real expression levels, because the translated proteins might stay longer even while its transcription is suppressed. To address this point, it is suggested that the authors observe the marker lines in the presence of a translation inhibitor, such as cycloheximide, and quantitatively analyze the dynamics of protein degradation when no new protein is synthesized.

This is indeed an excellent point. Unfortunately, we have to first say that due to funding issues we are currently unable to do this experiment. However, we did include two things in the revised manuscript: First, we have put in a note that this is indeed a caveat of the system that must be acknowledged (lines 334-337). Second, we have included some information from a different study, which at least addresses this point to some degree. We have imaged the transcriptional response of the WRKY11 transcription factor in response to colonization by Fo5176, and in this case, we not only see a local upregulation next to the colonization site, but we see a complete switch in expression pattern. As part of this switch, WRKY11 expression, which was expressed in all root tissues and cells in uninfected control experiments, switches expression off in all tissues and cells except the vascular cells close to the infection site. So here, we indeed have a downregulation of the reporter. In these experiments, signal from the fluorescent WRKY11 reporter disappears from the cells within a day. As we imaged once per day, we can, unfortunately not get more specific than this one-day window. The day before colonization of the tip, signal is seen in all tissues, one day later, if/when the vasculature if colonized in the tip, there is no weak/residual fluorescence left in the cells of the outer tissues. So we can at least state that we would probably also detect downregulation of expression, despite the protein lifetime. Importantly, all our imaging is done on a regular stereomicroscope, and thus, camera sensitivity is moderate. I could imagine that we may be able to detect some residual fluorescence with ultra-sensitive cameras at a spinning disc, or a sensitive detector at a laser-scanning microscope, but we have not tested this. We have added this information in lines 337-347. I apologize that we can’t add more information than this.

It is remarkable that the authors managed to clone 75 promoter sequences. However, whether all promoters work as expected was not clearly assessed in the present study. Did the authors only transform plants with PEP1, PEP2, PEPR1, and PEPR2 marker constructs? How would they know that the other promoters also work appropriately? In terms of providing these constructs to the research community, it is needed to disclose to which extent the expression has been validated in planta and which promoter has not been assessed.

This is indeed important information. We have not used the promoters in mutant complementation assays, and have added this caveat in lines 348-350.

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 current study by Calabria et al., entitled "pGG-PIP: A GreenGate (GG) entry vector collection with Plant Immune system Promoters (PIP)," reported the development of a set of GreenGate-compatible entry plasmids that contain promoter sequences of a series of immunity-related genes. This tool enables live-cell observation of immune responses at a cellular resolution. Being compatible with many other GreenGate tools, it opens up a door toward simultaneous visualization of different but overlapping immune pathways and ultimately describes the 4D dynamics of plant immunity. It is more than expected that these constructs will be used by a wide range of researchers and contribute to the ultimate understanding of plant innate immunity.

It is exciting that the authors observed the marker expression by a fluorescent stereomicroscope. This allows for non-destructive observation of response over time, keeping the system gnotobiotic. However, it was partly disappointing that the author did not take full advantage of this. It would have been much nicer if the authors observed the infection process over time, such that one could tell when and where the response starts, and whether local and systemic reactions occur simultaneously or instead require local-to-systemic signal transduction. They indeed seem to have done such time-course observation (line 378) however did not provide the results. I am curious to know what the authors could have found from those experiments. It would also be a strong appealing point of this method and is therefore highly encouraged.

Immune responses are not always induction of expression but sometimes reduction. Some genes up-regulated in the first phase will also be down-regulated afterward in order to go back to the initial non-responding state. During such down-regulation, the expression of a fluorescence marker gene might not accurately reflect the real expression levels, because the translated proteins might stay longer even while its transcription is suppressed. To address this point, it is suggested that the authors observe the marker lines in the presence of a translation inhibitor, such as cycloheximide, and quantitatively analyze the dynamics of protein degradation when no new protein is synthesized.

It is remarkable that the authors managed to clone 75 promoter sequences. However, whether all promoters work as expected was not clearly assessed in the present study. Did the authors only transform plants with PEP1, PEP2, PEPR1, and PEPR2 marker constructs? How would they know that the other promoters also work appropriately? In terms of providing these constructs to the research community, it is needed to disclose to which extent the expression has been validated in planta and which promoter has not been assessed.

Referee cross-commenting

I agree with reviewer #1 that the authors need to disclose how many independent lines were established and assessed for each construct.

I also agree with reviewer #2 that the figure and data presentation needs to be improved.

Significance

Overall, the current study already provides a widely useful set of tools for plant researchers, and some additional work would further increase its strength and value.

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 study provides a useful toolkit of reporter/marker constructs for investigating the gene expression of many immune-associated genes. The authors further used this toolkit to examine the expression pattern of several immune elicitor/receptor/downstream component genes after the inoculation of a fungal vascular pathogen Fusarium oxysporum. The study provides valuable tools for plant immunity study. I have some comments regarding the experiment design and data presentation as shown below.

Presentation and quality of the images need be improved. Scale bars are missing in all confocal images. In Figure 3 and 4, the name of genes examined can be labeled on the image, which will make it easier for readers. In addition, key information such as the inoculum and sampling time point after fungal inoculation should be described in the legend or the main text. More importantly, a "mock" inoculation or "before fungal inoculation" should be performed to reveal the expression changes of the marker genes after fungal inoculation.

Lines 172-174, the pictures are too small to see these details. The same for BIK1 (line 187). Line 174-176, which results are these referring to? The same for line 200-203.

Significance

This study provides a valuable collection of vectors/constructs for investigation of transcriptional dynamics of plant immunity genes and should attract broad interest of the plant immunity 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

Summary:

In their manuscript, Calabria et al. primarily present a collection of 75 plant (Arabidopsis thaliana) promoters cloned by them into the GreenGate system. These promoters represent different pathways of the plant immune system. Exemplarily they used this compilation to check the involvement of several components of the PLANT ELICITOR PEPTIDE (PEP)-pathway in the response of A. thaliana roots to infection with Fusarium oxysporum strain Fo5176 via transcriptional reporters.

Major comments:

The key point of the manuscript is to provide resources for the plant community. The motivation for selecting these specific promoters, how they were obtained and cloned, what they are in detail and how they will be made publically available is all clearly described. The infection experiments presented in it are an added bonus and a proof of concept of the applicability of the system.

Minor comments:

The promotor sequences will probably be included in the AddGene submission, however, it might be helpful to also deposit the promoter sequences at e.g. GenBank.

Line 133: "There are few exceptions to this rule...". It would probably helpful to list/mark these exceptions in Table 1.

Line 138: "A overhangs". In the GreenGate system, A-modules (promoters) are flanked by A- (5') and B- (3') overhangs (applies to line 144, too). Also, the B-overhang listed here (TTGT) is the reverse complement, which might be confusing for readers.

Line 149 ff.: How many lines have been established per promoter tested? Did they all yield a similar expression pattern?

Line 163: As someone not being familiar with microscoping Arabidopsis roots, I'm wondering how the authors can be sure that the tissue in question is the vasculature. Is this obvious for experts in the field?

Discussion: Is there by any chance prior (cell-resolution) knowledge about the expression behaviour of any of the investigated promoters? E. g. by in-situ hybridizations? If so, do the expression patterns match?

Significance

As stated above, this manuscript primarily describes a technical resource useful for the plant science community.

It is GreenGate-based and therefore easily compatible with other GreenGate-based resource collections. Its primary focus is in the area of plant immune research.

The key audience is plant immunologists. However, also researchers requiring e.g. tissue-specific and/or pathogen-inducible expression might find it helpful.

My own field of expertise is plant transformation and cloning systems, thus I went over the part dealing with the proof-of-principle only as a non-expert.

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

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: Yeh et al., present novel findings that Bitesize (Btsz) a Synaptotagmin-like protein, helps organize actin during the syncytial blastoderm stages of Drosophila embryo development. Depleting Btsz leads to phenotypes in the syncytial cycles that mimic Drosophila mutants where actin or membrane trafficking is disrupted. Perhaps most interestingly, the authors show that a non-Moesin Binding Domain-containing isoform of Btsz is important for cytoskeletal regulation during syncytial cycles. The authors generated a BtszB-C terminal recombinant protein and showed using imaging and biochemistry that this conserved segment of Btsz (which is present in all isoforms) can bind to and bundle F-actin in vitro. Lastly, the authors show that Btsz localizes to the apical region of pseudocleavage furrows and cell interfaces at gastrulation, which is consistent with previous literature regarding its role in regulating adherens junctions.

Major Concern: Both the imaging data, image analysis and biochemistry, are compelling. The findings regarding expression of alternative isoforms of Btsz are interesting within this developmental context. However, the final model is very simple and may benefit from the addition of experiments or at least further attention in the discussion. For example, it is not entirely clear what the division of labor may be between isoforms that do or do not bind Moesin. Do these isoforms work to accomplish a single function; or do they perform unique functions? Are the isoforms subject to similar or different regulation? At a minimum, the authors thoughts should be included in the Discussion, and a more integrated model presented. Relatedly, the authors mention the possibility that membrane trafficking may be impacted but end abruptly there. Additional experiments would obviously increase impact. If no experiments are added, the existing text should nonetheless be edited to include a more complete Discussion of the results.

Specific Concerns:

1. While the authors claim there is an actin defect, that defect is not readily revealed by a change in actin levels. Is the change perhaps in actin stability or in mechanical properties of the actin filaments (for example, if filaments can assemble but not be bundled or appropriately tethered to the plasma membrane in the mutant)? Have the authors tried either FRAP or laser cutting of furrows in mutant embryos?
2. While prior publications mention the role for Btsz in building adherens junctions, it would still be useful to see an analysis of junction phenotypes in the hands of these authors. Also, where do junction components concentrate and what do they do, if known, in syncytial embryos? It would be helpful to include this information in the text.
3. Does imaging of golgi or endosome markers reveal any differences in membrane compartments in Btsz mutant embryos? Even negative results would be interesting.
4. In describing the Myosin network phenotype during cellularization, it is not clear what is meant by the statement that the network has "constricted" over the positions where nuclei were lost. That sounds like an active process. It seems equally possible that the Myosin is just coating the membrane that now fills the gaps where nuclei should be.
5. Some aspects of Btsz gene expression are discussed and equated with a small number of previously described genes for cellularization. Are those genes only expressed during cellularization or beyond? It appears that Btsz is expressed beyond cellularization. Do those genes also have complex splicing patterns/multiple isoforms?
6. Could the authors comment on why they chose to describe the syncytial phenotypes in Cycle 12 but not other syncytial cycles?

Significance

For strengths and limitations, see above.

Advance: The authors advance the field of regarding Synaptotagmin-like proteins (Slps) by studying alternative isoforms of the proteins which lack a Moesin-binding domain (MBD). They find a novel function for Btsz isoforms that do not contain an MBD and show that a variant of these isoforms can directly bind to and bundle F-actin to regulate actin during syncytial nuclear divisions. Since the domain(s) they tested are conserved in all isoforms, this likely means that the actin binding function of Btsz could be conserved for most Slps, including Btsz isoforms which contain MBD.

Audience: This work is of interest to cell and developmental biologists who study the regulation of actin cytoskeleton. The work as presented also has some relevance to those who study adherens junctions, membrane trafficking, and Synaptotagmin-like proteins. More broadly, this work may be of interest to those who study alternatively-spliced proteins in the context of 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 manuscript by Yeh and colleagues examines the function of the Slp-family protein Bitesize in Drosophila. Btsz has been previously studied in the fly embryo and appreciated to regulate F-actin and adhesion-related properties in the formation of the epithelium, but in this manuscript Yeh et al. look at an earlier point of development (division in the early embryonic syncytium) and also examine the role of bitesize transcripts that lack the moesin-binding domain (MBD). The authors disrupt Btsz function and observe defects in actin-dependent structures, such as the apical actin cap and the pseudo-cleavage furrows, which leads to defect in the pseudo-cleavage divisions. They also perform actin-bundling assays with a btsz fragment that does not contain the MBD and see that btsz can still bundle actin, implicating the C2 domains in this function. The work appears well-done, with only one major area of concern (the imaging and analysis of actin caps, below). The manuscript is well-written, with a nice Introduction, and the Results are appropriately described and interpreted. The quantitation appears appropriate, and n number and the statistical tests used by the authors are consistently stated throughout the manuscript. A few more detailed comments are below:

1. It does not appear that the actin caps are being measured and imaged. None of the usual internal structures of the caps are apparent, and instead it appears that what is presented are the apical margins of the pseudocleavage furrows (or the very edges of the caps).
2. Along these lines, the argument that caps are smaller does not make much sense, since it appears that the "caps" are being measured late once the furrows have formed. Since these dimensions are set by the number of nuclei in the embryo, as long as the caps are growing large enough to get collisions between adjacent nuclei/caps, how can the caps be smaller unless there are fewer nuclei? These changes could also be secondary consequences of differences in nuclear distributions around the embryo periphery. For these reasons, and because of the close packing of nuclei together, usually cap growth rates are plotted in periods prior to cap collisions.
3. Sorry if this was missed, but are the cycles at which measurements are made listed in each appropriate figure? I saw "cycle 12" listed in one figure legend, but not others.
4. How do the authors know that the nuclear density defects in the CRISPR allele are due to the same mechanism? Could be through same mechanism, but could also be due to defect in nuclear anchoring, cortical portioning, etc...
5. The schematics and illustrations are nicely done.

Minor notes:

• a) Should there be actin in the top row of 1A?

Significance

This manuscript by Yeh and colleagues examines the function of the Slp-family protein Bitesize in Drosophila. Btsz has been previously studied in the fly embryo and appreciated to regulate F-actin and adhesion-related properties in the formation of the epithelium, but in this manuscript Yeh et al. look at an earlier point of development (division in the early embryonic syncytium) and also examine the role of bitesize transcripts that lack the moesin-binding domain (MBD). The authors disrupt Btsz function and observe defects in actin-dependent structures, such as the apical actin cap and the pseudo-cleavage furrows, which leads to defect in the pseudo-cleavage divisions. They also perform actin-bundling assays with a btsz fragment that does not contain the MBD and see that btsz can still bundle actin, implicating the C2 domains in this function. The work should be of interest to a developmental community and those workers interested in Slp-family function.