Reviewer #3 (Public review):

Summary:

This work is a detailed and thorough analysis of the morphogenesis of the posterior signaling center (PSC), a hematopoietic niche in the Drosophila larva. Live imaging is performed from the stage of PSC determination until the appearance of a compact lymph gland and PSC in the stage 16 embryo. This analysis is combined with genetic studies that clarify the involvement of adjacent tissue, including the visceral mesoderm, alary muscle, and cardioblasts/dorsal vessel. Lastly, the Slit/Robo signaling system is clearly implicated in the normal formation of the PSC.

Strengths:

The data are clearly presented and well documented, and fully support the conclusions drawn from the different experiments.

The authors have addressed all of my previous comments, in particular concerning the role of epidermal cell rearrangements during dorsal closure as a possible force acting on the movement of PSC cells. The authors have clarified their definition of "collective migration" as it applies to the movement of PSC. The revised paper will make an important contribution to our understanding of the mechanisms driving morphogenesis.

Author response:

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

Reviewer #1 (Public review):

Summary:

The study by Nelson et al. is focused on formation of the Drosophila Posterior Signaling Center (PSC) which ultimately acts as a niche to support hematopoietic stem cells of the lymph gland (LG). Using a combination of genetics and live imaging, the authors show that PSC cells migrate as a tight collective and associate with multiple tissues during a trajectory that positions them at the posterior of the LG.

This is an important study that identifies Slit-Robo signaling as a regulator of PSC morphogenesis, and highlights the complex relationship of interacting cell types - PSC, visceral mesoderm (VM) and cardioblasts (CBs) - in coordinated development of these three tissues during organ development. However, one point requiring clarification is the idea that PSC cells exhibit a collective cell migration; it is not clear that the cells are migrating rather than being pushed to a more dorsal position through dorsal closure and/or other similar large scale embryo movement. This does not detract from the very interesting analysis of PSC morphogenesis as presented.

This Public Review by Reviewer #1 is identical to their original Public Review, thus we are unsure whether Reviewer #1 assessed the revised version of our manuscript, and whether they read our responses to their original Public Review. Below we summarize our original responses to the weaknesses listed for the first version of our manuscript.

Strengths:

• Using expression of Hid or Grim to ablate associated tissues, they find evidence that the VM and CB of the dorsal vessel affect PSC migration/morphology whereas the alary muscles do not. Slit is expressed by both VM and CBs, and therefore Slit-Robo signaling was investigated as PSCs express Robo.

• Using a combination of approaches, the authors convincingly demonstrate that Slit expression in the CBs and VM acts to support PSC positioning. A strength is the ability to knockdown slit levels in particular tissue types using the Gal4 system and RNAi.

• Although in the analysis of robo mutants, the PSC positioning phenotype is weaker in the individual mutants (robo1 and robo2) with only the double mutant (robo1,robo2) exhibiting a phenotype comparable to the slit RNAi. The authors make a reasonable argument that Slit-Robo signaling has an intrinsic effect, likely acting within PSCs, because PSCs show a phenotype even when CBs do not (Fig 4G).

• New insight into dorsal vessel formation by VM is presented in Fig 4A,B, as loss of the VM can affect dorsal vessel morphogenesis. This result additionally points to the VM as important.

Weaknesses:

• The authors are cautioned to temper the result that Slit-Robo signaling is intrinsic to PSC since loss of robo may affect other cell types (besides CBs and PSCs) to indirectly affect PSC migration/morphogenesis. In fact, in the robo2, robo1 mutant, the VM appears to be incorrectly positioned (Fig. 4G).

We maintain our conclusion, and, we point out that the Reviewer stated, “The authors make a reasonable argument that Slit-Robo signaling has an intrinsic effect, likely acting within PSCs”. We already added a statement to the Discussion reminding the reader of the possibility of secondary defects (“Finally, it is possible that PSC cells do not intrinsically require Robo activation, but rather CB-independent PSC mis-positioning in sli or robo mutants could be a secondary defect caused by compromised Slit-Robo signaling in some other tissue.”).

• If possible, the authors should use RNAi to knockdown Robo1 and Robo2 levels specifically in the PSCs if a Gal4 is available; might Antp.Gal4 (Fig 1K) be useful? Even if knockdown is achieved in PSCs+CBs, this would be a better/complementary experiment to support the approach outlined in Fig 4D.

As described in our first response, use of Antp-GAL4 with RNAi would be no better than a whole animal double Robo mutant.

• Movies are hard to interpret, as it seems unclear that the PSCs actively migrate rather than being pushed/moved indirectly due to association with VM and CBs/dorsal vessel.

Vm does not directly contact the PSC, so the Vm cannot be physically pushing the PSC. In their original review, Reviewer #3 expressed similar concerns (Weaknesses #1 and #2), and upon their review of our revised manuscript they determined we addressed these concerns.

Reviewer #2 (Public review):

The paper by Nelson KA, et al. explored the collective migration, coalescence and positioning of the posterior signaling center (PSC) cells in Drosophila embryo. With live imaging, the authors observed the dynamic progress of PSC migration. Throughout this process, visceral mesoderm (VM), alary muscles (Ams) and cardioblasts (CBs) are in proximity of PSC. Genetic ablation of these tissues reveals the requirement for VM and CBs, but not AMs in this process. Genetic manipulations further demonstrated that Slit-Robo signaling was critical during PSC migration and positioning. While the genetic mechanisms of positioning the PSC were explored in much detail, including using live imaging, the functional consequence of mispositioning or (partial) absence of PSC cells has not been addressed, but would much increase the relevance of their findings. A few additional issues need to be addressed as well in this otherwise well-done study.

Previous major points:

(1) The only readout in their experiments is the relative correctness of PSC positioning. Importantly, what is the functional consequence if PSC is not properly positioned? This would be particularly important with robo-sli manipulations, where the PSC is present but some cells are misplaced. What is the consequence? Are the LGs affected, like specification of their cell types, structure and function? To address this for at least the robo-slit requirement in the PSC, it may be important to manipulate them directly in the PSC with a split Gal4 system, using Antp and Odd promoters.

We state in our original response that exploring the functional consequences of PSC mis-positioning was outside the scope of this study. Given that the necessary cis-regulatory modules have not been identified at Antp or Odd, creating a split-GAL4 with ‘Antp and Odd promoters’ cannot be accomplished in a reasonable time frame, as we previously detailed in our original response.

(2) The densely, parallel aligned fibers in the lower part of Figure 1J seemed to be visceral mesoderm, but further up (dorsally) that may be epidermis. It is possible that the PSC migrate together with the epidermis? This should be addressed.

This was directly addressed by the additional data included in our revision. When epidermal closure is stalled, the PSC is able to migrate past the stalled leading edge, closer to the midline.

(3) Although the authors described the standards of assessing PSC positioning as "normal" or "abnormal", it is rather subtle at times and variable in the mutant or KD/OE examples. The criteria should be more clearly delineated and analyzed double-blind, also since this is the only readout. Further examples of abnormal positioning in supplementary figures would also help.

We addressed this comment in detail in our original response. Briefly, double-blinding was oftentimes not possible due to the obviousness of the genotype in the image. The criteria we outline for normal PSC positioning is as comprehensive as possible given the subtlety variability of mis-positioning phenotypes. Two of the authors independently analyzed the relatively large sets of samples and arrived at the same conclusions.

(4) Discussion is very lengthy and should shortened.

We shortened the Discussion in the revised version.

Comments on revised version:

Although the authors have responded to my concerns as they deemed suitable, these concerns still stand for the revised version.

Given our responses above and the lack of detail in this comment, we are unsure why the Reviewer is still concerned.

Reviewer #3 (Public review):

Summary:

This work is a detailed and thorough analysis of the morphogenesis of the posterior signaling center (PSC), a hematopoietic niche in the Drosophila larva. Live imaging is performed from the stage of PSC determination until the appearance of a compact lymph gland and PSC in the stage 16 embryo. This analysis is combined with genetic studies that clarify the involvement of adjacent tissue, including the visceral mesoderm, alary muscle, and cardioblasts/dorsal vessel. Lastly, the Slit/Robo signaling system is clearly implicated in the normal formation of the PSC.

Strengths:

The data are clearly presented and well documented, and fully support the conclusions drawn from the different experiments.

The authors have addressed all of my previous comments, in particular concerning the role of epidermal cell rearrangements during dorsal closure as a possible force acting on the movement of PSC cells. The authors have clarified their definition of "collective migration" as it applies to the movement of PSC. The revised paper will make an important contribution to our understanding of the mechanisms driving morphogenesis.

We are appreciative of the time spent by the Reviewer reading our responses and assessing the revision.

---------

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The study by Nelson et al. is focused on the formation of the Drosophila Posterior Signaling Center (PSC) which ultimately acts as a niche to support hematopoietic stem cells of the lymph gland (LG). Using a combination of genetics and live imaging, the authors show that PSC cells migrate as a tight collective and associate with multiple tissues during a trajectory that positions them at the posterior of the LG.

This is an important study that identifies Slit-Robo signaling as a regulator of PSC morphogenesis, and highlights the complex relationship of interacting cell types - PSC, visceral mesoderm (VM), and cardioblasts (CBs) - in the coordinated development of these three tissues during organ development. However, one point requiring clarification is the idea that PSC cells exhibit a collective cell migration; it is not clear that the cells are migrating rather than being pushed to a more dorsal position through dorsal closure and/or other similar large-scale embryo movement. This does not detract from the very interesting analysis of PSC morphogenesis as presented.

Since each referee asked for clarification concerning collective cell migration, we present a combined response further below, placed after the comments from Reviewer #3.

Strengths:

(1) Using the expression of Hid or Grim to ablate associated tissues, they find evidence that the VM and CB of the dorsal vessel affect PSC migration/morphology whereas the alary muscles do not. Slit is expressed by both VM and CBs, and therefore Slit-Robo signaling was investigated as PSCs express Robo.

(2) Using a combination of approaches, the authors convincingly demonstrate that Slit expression in the CBs and VM acts to support PSC positioning. A strength is the ability to knockdown slit levels in particular tissue types using the Gal4 system and RNAi.

(3) Although in the analysis of robo mutants, the PSC positioning phenotype is weaker in the individual mutants (robo1 and robo2) with only the double mutant (robo1,robo2) exhibiting a phenotype comparable to the slit RNAi. The authors make a reasonable argument that Slit-Robo signaling has an intrinsic effect, likely acting within PSCs because PSCs show a phenotype even when CBs do not (Figure 4G).

(4) New insight into dorsal vessel formation by VM is presented in Figure 4A, B, as loss of the VM can affect dorsal vessel morphogenesis. This result additionally points to the VM as important.

Weaknesses:

(1) The authors are cautioned to temper the result that Slit-Robo signaling is intrinsic to PSC since the loss of robo may affect other cell types (besides CBs and PSCs) to indirectly affect PSC migration/morphogenesis. In fact, in the robo2, robo1 mutant, the VM appears to be incorrectly positioned (Figure 4G).

We have reexamined our wording in the relevant Results section and, given that this referee agrees that we, “make a reasonable argument that Slit-Robo signaling has an intrinsic effect, likely acting within PSCs because PSCs show a phenotype even when CBs do not (Figure 4G)”, it was not clear how we might temper our conclusions more. Given that PSC cells express Robo1 and Robo2, and that the Vm does not contact the PSC, our ‘reasonable argument’ appears fair and parsimonious. Since we agree with the referee that a reader should be made as aware as possible of alternatives, we will add a comment to the Discussion, reminding the reader of the possibility of a secondary defect.

(2) If possible, the authors should use RNAi to knockdown Robo1 and Robo2 levels specifically in the PSCs if a Gal4 is available; might Antp.Gal4 (Fig 1K) be useful? Even if knockdown is achieved in PSCs+CBs, this would be a better/complementary experiment to support the approach outlined in Figure 4D.

While we agree that PSC-specific knockdown of Robo1 and Robo2 simultaneously would be ideal, this is not possible. First, the most-effective UAS-RNAi transgenes (that is, those in a Valium 20 backbone) are both integrated at the same chromosomal position; these cannot be simultaneously crossed with a GAL4 transgenic line to attempt double knock down. Additionally, as with all RNAi approaches that must rely on efficient knockdown over the rapid embryonic period, even having facile access to the above does not ensure the RNAi approach will cause as effective depletion as the genetic null condition that we use. Second, as the referee concedes, there is no embryonic PSC-specific GAL4. The proposed use of Antp-GAL4 would cause knockdown in many tissues (PSC, CB, Vm, epidermis and amnioserosa). This would lead to a reservation similar to that caused by our use of the straight genetic double mutant, as regards potential indirect requirement for Robo function.

(3) Movies are hard to interpret, as it seems unclear that the PSCs actively migrate rather than being pushed/moved indirectly due to association with VM and CBs/dorsal vessel.

First, the Vm does not directly contact the PSC, so it cannot be pushing the PSC dorsally. We will re-examine our text to be certain to make this clear. Second, in our analysis of bin mutants, which lack Vm, LGs and PSCs are able to reach the dorsal midline region in the absence of Vm. Finally, please see our response to Reviewer #3, point 2, for why we maintain that PSC cells are “migrating” even though some PSC cells are attached to CBs.

Reviewer #2 (Public Review):

The paper by Nelson KA, et al. explored the collective migration, coalescence, and positioning of the posterior signaling center (PSC) cells in Drosophila embryo. With live imaging, the authors observed the dynamic progress of PSC migration. Throughout this process, visceral mesoderm (VM), alary muscles (Ams), and cardioblasts (CBs) are in proximity to PSC. Genetic ablation of these tissues reveals the requirement for VM and CBs, but not AMs in this process. Genetic manipulations further demonstrated that Slit-Robo signaling was critical during PSC migration and positioning. While the genetic mechanisms of positioning the PSC were explored in much detail, including using live imaging, the functional consequence of mispositioning or (partial) absence of PSC cells has not been addressed, but would much increase the relevance of their findings. A few additional issues need to be addressed as well in this otherwise well-done study.

Major points:

(1) The only readout in their experiments is the relative correctness of PSC positioning. Importantly, what is the functional consequence if PSC is not properly positioned? This would be particularly important with robo-sli manipulations, where the PSC is present but some cells are misplaced. What is the consequence? Are the LGs affected, like the specification of their cell types, structure, and function? To address this for at least the robo-slit requirement in the PSC, it may be important to manipulate them directly in the PSC with a split Gal4 system, using Antp and Odd promoters.

We agree that the functional consequence of PSC mis-positioning is important and a relevant question to eventually address. However, virtually all markers and reagents used to assess the effect of the PSC on progenitor cells and their differentiated descendants are restricted to analyses carried out on the third larval instar - some three days after the experiments reported here. Most of the manipulated conditions in our work are no longer viable at this phase and, thus, addressing the functional consequences of a malformed PSC will require the field to develop new tools. 

As we noted in the Introduction, the consistency with which the wildtype PSC forms as a coalesced collective at the posterior of the LG strongly suggests importance of its specific positioning and shape, as has now been found for other niches (citations in manuscript). Additionally, in the Discussion we mention the existence of a gap junction-dependent calcium signaling network in the PSC that is important for progenitor maintenance. Without continuity of this network amongst all PSC cells (under conditions of PSC mis-positioning), we strongly anticipate that the balance of progenitors to differentiated hemocytes will be mis-managed, either constitutively, and / or under immune challenge conditions. 

Finally, to our knowledge, the tools do not exist to build a “split Gal4 system using Antp and Odd promoters”. The expression pattern observed using the genomic Antp-GAL4 line must be driven by endogenous enhancers–none of which have been defined by the field, and thus cannot be used in constructing second order drivers. Similarly, for odd skipped, in the embryo the extant Odd-GAL4 driver expresses only in the epidermis, with no expression in the embryonic LG. Thus, the cis regulatory element controlling Odd expression in the embryonic LG is unknown. In the future, the discovery of an embryonic PSC-specific driver will aid in addressing the specific functional consequences of PSC mis-positioning.

(2) The densely, parallel aligned fibers in the part of Figure 1J seemed to be visceral mesoderm, but further up (dorsally) that may be epidermis. It is possible that the PSC migrate together with the epidermis? This should be addressed.

See response to Reviewer #3.

(3) Although the authors described the standards of assessing PSC positioning as "normal" or "abnormal", it is rather subtle at times and variable in the mutant or KD/OE examples. The criteria should be more clearly delineated and analyzed double-blind, also since this is the only readout. Further examples of abnormal positioning in supplementary figures would also help.

We appreciate the Reviewer’s concern and acknowledge that the phenotypes we observed were indeed variable, and, at times subtle. As we show and discuss in the paper, our results revealed that the signaling requirements for proper PSC positioning are complex; this was favorably commented upon by Reviewer #1 (“...highlights the complex relationship of interacting cell types - PSC, visceral mesoderm (VM), and cardioblasts (CBs) - in the coordinated development of these three tissues during organ development.…”). We suspect the phenotypic variability is attributable to any number of biological differences such as heterogeneity of PSC cells and an accompanying difference in the timing of their competence to receive and respond to Slit-Robo signaling, the timing of release of Slit from CBs and Vm, number of cells in a given PSC, which PSC cells in the cluster respond to too little or too much signaling, and/or typical variability between organisms. Furthermore, PSC positioning analyses were conducted by two of the authors, who independently came to the same conclusions. For many of the manipulations double blinding was not possible since the genotype of the embryo was discernible due to the obvious phenotype of the manipulated tissue.

(4) The Discussion is very lengthy and should shortened.

We will re-examine the prose and emphasize more conciseness, while maintaining clarity for the reader.

Reviewer #3 (Public Review):

Summary:

This work is a detailed and thorough analysis of the morphogenesis of the posterior signaling center (PSC), a hematopoietic niche in the Drosophila larva. Live imaging is performed from the stage of PSC determination until the appearance of a compact lymph gland and PSC in the stage 16 embryo. This analysis is combined with genetic studies that clarify the involvement of adjacent tissue, including the visceral mesoderm, alary muscle, and cardioblasts/dorsal vessels. Lastly, the Slit/Robo signaling system is clearly implicated in the normal formation of the PSC.

Strengths:

The data are clearly presented, well documented, and fully support the conclusions drawn from the different experiments. The manuscript differs in character from the mainstay of "big data" papers (for example, no sets of single-cell RNAseq data of, for instance, PSC cells with more or less Slit input, are offered), but what it lacks in this regard, it makes up in carefully planned and executed visualizations and genetic manipulations.

Weaknesses:

A few suggestions concerning improvement of the way the story is told and contextualized.

(1) The minute cluster of PSC progenitors (5 or so cells per side) is embedded (as known before and shown nicely in this study) in other "migrating" cell pools, like the cardioblasts, pericardial cells, lymph gland progenitors, alary muscle progenitors. These all appear to move more or less synchronously. What should also be mentioned is another tissue, the dorsal epidermis, which also "moves" (better: stretches?) towards the dorsal midline during dorsal closure. Would it be reasonable to speculate (based on previously published data) that without the force of dorsal closure, operating in the epidermis, at least the lateral>medial component of the "migration" of the PSC (and neighboring tissues) would be missing? If dorsal closure is blocked, do essential components of PSC and lymph gland morphogenesis (except for the coming-together of the left and right halves) still occur? Are there any published data on this?

Each of the Reviewers is interested in our response to this very relevant question, and, thus, we will address the issue en bloc here. First, we will add a Supplementary Figure showing that LG and CBs are still able to progress medially towards the dorsal midline when dorsal closure stalls.  This rules out any major effect for the most prominent “large-scale embryo cell sheet movement” in positioning the PSC. Second, published work by Haack et. al. and Balaghi et. al. shows that CBs and leading edge epidermal cells are independently migratory, and we will add this context to the manuscript for the reader.

(2) Along similar lines: the process of PSC formation is characterized as "migration". To be fair: the authors bring up the possibility that some of the phenotypes they observe could be "passive"/secondary: "Thus, it became important to test whether all PSC phenotypes might be 'passive', explained by PSC attachment to a malforming dorsal vessel. Alternatively, the PSC defects could reflect a requirement for Robo activation directly in PSC cells." And the issue is resolved satisfactorily. But more generally, "cell migration" implies active displacement (by cytoskeletal forces) of cells relative to a substrate or to their neighbors (like for example migration of hemocytes). This to me doesn't seem really clearly to happen here for the dorsal mesodermal structures. Couldn't one rather characterize the assembly of PSC, lymph gland, pericardial cells, and dorsal vessel in terms of differential adhesion, on top of a more general adhesion of cells to each other and the epidermis, and then dorsal closure as a driving force for cell displacement? The authors should bring in the published literature to provide a background that does (or does not) justify the term "migration".

Before addressing this specifically, we remind readers of our response above that states the rationale ruling out large, embryo-scale movements, such as epidermal dorsal closure, in driving PSC positioning. So, how are PSC cells arriving at their reproducible position? This manuscript reports the first live-imaging of the PSC as it comes to be positioned in the embryo. We interpret these movies to suggest strongly that these cells are a ‘collective’ that migrates. Neither the data, nor we, are asserting that each PSC cell is ‘individually’ migrating to its final position. Rather, our data suggest that the PSC migrates as a collective. The most paradigmatic example of directed, collective cell migration, is of Drosophila ovarian border cells. That cell cluster is surrounded at all times by other cells (nurse cells, in that case), and for the collective to traverse through the tissue, the process requires constant remodeling of associations amongst the migrating cells in the collective (the border cells), as well as between cells in the collective and those outside of it (the nurse cells). In fact, the nurse cells are considered the substrate upon which border cells migrate. Note also that in collective border cell migration cells within the collective can switch neighbors, suggesting dynamic changes to cell associations and adhesions. 

In our analysis, the PSC cells exhibit qualities reminiscent of the border cells, and thus we infer that the PSC constitutes a migratory cell collective.  We also show in Figure 1H that PSC cells exhibit cellular extensions, and thus have a very active, intrinsic actin-based cytoskeleton. In fact, in Figure 1I, we point out that PSC cells shift position within the collective, which is not only a direct feature of migration, but also occurs within the border cell collective as that collective migrates. Additionally, the fact that the lateral-most PSC cells shift position in the collective while remaining a part of the collective–and they do this while executing net directional movement–makes a strong argument that the PSC is migratory, as no cell types other than PSCs are contacting the surfaces of those shifting PSC cells. Lastly, the Reviewer’s supposition that, rather than migration, dorsal mesoderm structures form via “differential adhesion, on top of a more general adhesion of cells to each other” is, actually, precisely an inherent aspect of collective cell migration as summarized above for the ovarian border collective.

In our resubmission we will adjust text citing the existing literature to better put into context the reasoning for why PSC formation based on our data is an example of collective cell migration.

(3) That brings up the mechanistic centerpiece of this story, the Slit/Robo system. First: I suggest adding more detailed data from the study by Morin-Poulard et al 2016, in the Introduction, since these authors had already implicated Slit-Robo in PSC function and offered a concrete molecular mechanism: "vascular cells produce Slit that activates Robo receptors in the PSC. Robo activation controls proliferation and clustering of PSC cells by regulating Myc, and small GTPase and DE-cadherin activity, respectively". As stated in the Discussion: the mechanism of Slit/Robo action on the PSC in the embryo is likely different, since DE-cadherin is not expressed in the embryonic PSC; however, it maybe not be THAT different: it could also act on adhesion between PSC cells themselves and their neighbors. What are other adhesion proteins that appear in the late lateral mesodermal structures?

Could DN-cadherin or Fasciclins be involved?

We agree with the Reviewer that Slit-Robo signaling likely acts in part on the PSC by affecting PSC cell adhesion to each other and/or to CBs (lines 428-435). As stated in the Discussion, we do not observe Fasciclin III expression in the PSC until late stages when the PSC has already been positioned, suggesting that Fasciclin III is not an active player in PSC formation. Assessing whether the PSC expresses any other of the suite of potential cell adhesion molecules such as DN-Cadherin or other Fasciclins, and then study their potential involvement in the Slit-Robo pathway in PSC cells, would be part of a follow-up study.  

Recommendations for the authors:

Reviewing Editor Comments:

The authors are encouraged to address several key issues and provide more explicit clarification when interpreting the behavior of the PSC cells as "migration." It is recommended that the authors engage with all reviewers' comments and refine the text based on the feedback they find valuable.

Reviewer #1 (Recommendations For The Authors):

Major concerns:

(1) Is it possible to assay robo1 and/or robo1 RNAi in a tissue-specific manner to further explore an intrinsic role in the PSC? Might the VM indirectly affect PSCs in a CB-independent manner? How does this affect the interpretation of results in Figure 4.

See also our response to Reviewer #1, Public review weaknesses #2.

Though we agree with the Reviewer that this is the better experiment to test for an intrinsic role for Robo in the PSC, this experiment is not possible at this time. As we noted in the manuscript, we do not yet have an embryonic PSC-specific GAL4, though we have been putting efforts towards identifying/developing such a tool. The Antp-GAL4 driver we used in this study will drive not only in both PSCs and CBs, but also in Vm, epidermis, and amnioserosa, as well as other tissues. The other available embryonic PSC drivers are not specific to the PSC and will drive expression in CBs and Vm, at minimum. This, combined with the reality that RNAi can be ineffective in embryonic tissues, resulted in our use of whole organism mutants to best address this question. 

We acknowledge that it is possible the Vm indirectly effects the PSC in a CB-independent manner in the double Robo mutant, and we added a statement to the Discussion reiterating this point. However, because the PSC expresses Robo1 and Robo2, we maintain that the simplest interpretation of the results in Figure 4 is that PSC cells require intrinsic Robo signaling. And, as we state in the manuscript, it is possible that Slit signals directly from Vm to Robo on the PSC.

(2) As this is the first study to be presenting PSC formation as involving collective cell migration, can the authors provide experimental evidence and rationale for this categorization?

We have added our rationale to the Results section in the revision.

See also our response to Reviewer #3, Public review weakness #2.

(3) The Slit staining presented in Fig 3 W', Z' should be quantified. Furthermore, what is the VM phenotype when Robo1 is overexpressed? Is there a VM-specific phenotype and could this indirect effect cause the PSC to misform/mismigrate?

We didn’t quantify Slit levels in the Vm-specific Robo overexpression condition because there was a visually striking difference compared to controls (increased intensity and specific localization to Vm membranes), and the manipulation resulted in a PSC phenotype. Thus, the evidence we show appears sufficient to strongly suggest that our genetic manipulation resulted in successful trapping of Slit on the Vm.

As to a Vm phenotype when Robo1 is overexpressed Vm-specifically: we know Vm is present, but we haven’t performed an in-depth phenotypic analysis. In the manuscript we show that this manipulation at least affects organization of PSC-adjacent CBs, which we go on to show is correlated with mis-positioned PSCs. Thus, the PSC phenotype in this condition is not solely due to a Vm-specific phenotype.

Minor concerns/suggestions:

(1) I might have missed it but where are the Movies referenced in the text? Are legends provided for the videos? It is important that this is included in the final version (or more clearly presented if I missed it).

We thank you the Reviewer for pointing this out; we now direct the reader to the movies at appropriate places within the text.

(2) In Figure 5, it might be helpful to add a third column to A in which the PSCs are pseudo-colored and thus highlighted because it is difficult to discern the white (not pink) PSCs...

We appreciate the suggestion and now include these panels as Figure 5A’’ in the revision.

(3) If I am following correctly, the lost PSC cells in Figure 5 don't move. Doesn't this suggest that what is critical is that the PSCs attach to the VM and/or CBs, and not necessarily that they are an actively migrating cell type? They "move" but might be passively carried.

See also the response to Reviewer #3, Public reviews weaknesses #2.

The Reviewer is correct that the PSC cells in Fig. 5 don’t move very much, but we interpret this differently from the Reviewer. After detachment of the cells in question they undergo dramatic shape changes, indicating active cytoskeletal remodeling, so the molecular machinery needed for migration appears to remain intact. Thus, we suggest that this observation actually emphasizes our finding that collectivity is needed for the migration. Given the consistency of PSC coalescence/collectivity and the intricate regulation that controls it, we believe it to be an integral part of PSC identity. When PSC cells become detached, they likely lose an aspect of their identity. In various manipulations we’ve noted instances of severely dispersed PSC cells expressing very low levels of identity markers Antp or Odd. Cells in such cases are likely compromised for their function, and this can include, for example, whether they can properly sense cues for migration.

Reviewer #2 (Recommendations For The Authors):

Minor points:

(1) The expression pattern of Antp-Gal4 > myrGFP in the whole embryo should be shown to better demonstrate the overlap with Odd. How does it compare with Antp-Gal4 > CD8::GFP?

We do not understand the question posed. We are not suggesting that Antp and Odd overlap in all cells, nor even many cells. It has been demonstrated by the field that co-expression among mesodermal cells, in the position where LG cells are specified, is a marker for the PSC. We have not thoroughly investigated all reporter lines for the GAL4 drivers used by the field.

(2) Does Tincdelta4-Gal4 not at all express in the PSC? This should be verified.

This question appears to refer to depletion of Slit by RNAi or cell killing driven by tinCΔ4-GAL4. TinCΔ4-GAL4 is expressed in CBs and in precisely 1 embryonic PSC cell. First, Slit isn’t expressed by any PSC cells to our eye, so any PSC mis-positioning observed upon tinCΔ4>Sli RNAi implicates CB involvement in PSC positioning. In designing tests for CB involvement, we were unable to identify any mutant known to lack CBs (or have fewer CBs) that didn’t also affect specification of the LG/PSC. The cell killing approach seemed best.  It is possible that, in this scenario, perhaps ablation of a single, key PSC cell could affect final positioning of the other PSCs, but we think that less likely than a role for CBs. We also retain our original conclusion due to the fact that we often find mis-positioned PSC cells adjacent to mis-positioned CBs, including in the panel representing the CB ablation experiment, Figure 2S.  

(3) Line 212: The data provide evidence that Vm is necessary, but clearly not sufficient, as CBs are also necessary.

We see how this wording was misleading and have adjusted the text accordingly.

(4) The CBs are not visible in Figure 3B.

We are unsure what the Reviewer is referring to, as we are certain that the signal between the blue outlines is indeed Slit expression in CBs.

Reviewer #3 (Recommendations For The Authors):

One minor mistake (I believe): in line 229 it should say "3C and 3D"

We have corrected this error.

eLife Assessment

This study unveils important data describing cell states of olfactory ensheathing cells, and how these cell states may relate to repair after spinal cord injury. The framework used for characterizing these cells is solid, although the study would be strengthened by additional quantification of immunohistochemical data and by complementing expression data with functional outcomes. This work will be of interest to stem cell biologists and spinal cord injury researchers.

Joint Public Review:

Summary

This manuscript explores the transcriptomic identities of olfactory ensheathing cells (OECs), glial cells that support life-long axonal growth in olfactory neurons, as they relate to spinal cord injury repair. The authors show that transplantation of cultured, immunopurified rodent OECs at a spinal cord injury site can promote injury-bridging axonal regrowth. They then characterize these OECs using single-cell RNA sequencing, identifying five subtypes and proposing functional roles that include regeneration, wound healing, and cell-cell communication. They identify one progenitor OEC subpopulation and also report several other functionally relevant findings, notably, that OEC marker genes contain mixtures of other glial cell type markers (such as for Schwann cells and astrocytes), and that these cultured OECs produce and secrete Reelin, a regrowth-promoting protein that has been disputed as a gene product of OECs.

Strengths

This manuscript offers an extensive, cell-level characterization of OECs, supporting their potential therapeutic value for spinal cord injury and suggesting potential underlying repair mechanisms. The authors use various approaches to validate their findings, providing interesting images that show the overlap between sprouting axons and transplanted OECs, and showing that OEC marker genes identified using single-cell RNA sequencing are present in vivo, in both olfactory bulb tissue and spinal cord after OEC transplantation.

Challenges

Despite the breadth of information presented, and although many of the suggestions in the initial review were addressed well, some points related to quantification and discussion of sex differences are not fully addressed in this revision.

(1) The request for quantification of OEC bridges is not fully addressed. We note that this revision includes the following statement (page 6): "We note, however, that such bridge formation is rare following a severe spinal cord injury in adult mammals." However, the title of the paper states that olfactory ensheathing cells promote neural repair and the abstract states that "OECs transplanted near the injury site modify the inhibitory glial scar and facilitate axon regeneration past the scar border and into the lesion." Statements such as these make it more crucial to include quantification of OEC bridges, because if single images are shown of remarkable, unusual bridges, but only one sentence acknowledges the low frequency of this occurrence, then this information taken together might present the wrong takeaway to readers.

Including some sort of quantification of bridging, whether it be the number of rats exhibiting bridges, the percentage area of OECs near a lesion site, or some other meaningful analysis, would add rigor and clarity to the manuscript.

(2) The additional discussion of sex differences in OEC bridging elaborates on the choice to study female rats, citing bladder challenges in male rats, but does not note salient clinical implications of this choice. Men account for ~80% of spinal cord injuries and likely also have worsened urinary tract issues, so it would be important to acknowledge this clinical fact and consider including males in future studies.

Author response:

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

Joint Public Review:

Summary

This manuscript explores the transcriptomic identities of olfactory ensheathing cells (OECs), glial cells that support life-long axonal growth in olfactory neurons, as they relate to spinal cord injury repair. The authors show that transplantation of cultured, immunopurified rodent OECs at a spinal cord injury site can promote injury-bridging axonal regrowth. They then characterize these OECs using single-cell RNA sequencing, identifying five subtypes and proposing functional roles that include regeneration, wound healing, and cell-cell communication. They identify one progenitor OEC subpopulation and also report several other functionally relevant findings, notably, that OEC marker genes contain mixtures of other glial cell type markers (such as for Schwann cells and astrocytes), and that these cultured OECs produce and secrete Reelin, a regrowth-promoting protein that has been disputed as a gene product of OECs.

This manuscript offers an extensive, cell-level characterization of OECs, supporting their potential therapeutic value for spinal cord injury and suggesting potential underlying repair mechanisms. The authors use various approaches to validate their findings, providing interesting images that show the overlap between sprouting axons and transplanted OECs, and showing that OEC marker genes identified using single-cell RNA sequencing are present in vivo, in both olfactory bulb tissue and spinal cord after OEC transplantation.

Despite the breadth of information presented, however, further quantification of results and explanation of experimental approaches would be needed to support some of the authors' claims. Additionally, a more thorough discussion is needed to contextualize their findings relative to previous work.

(1) a. Important quantification is lacking for the data presented. For example, multiple figures include immunohistochemistry or immunocytochemistry data (Figures 1, 5, 6), but they are presented without accompanying measures like fractions of cells labeled or comparisons against controls.

We would like to clarify that the immunohistochemistry or immunocytochemistry data presented are meant to be qualitative rather than quantitative. The main purpose of the images is to show the presence or absence of markers of OEC subtypes rather than how much is present. That being said, in the revision we now add quantitative estimates of cell fractions for OECs along with other major cell types in Supplemental Table 1 and each OEC subtype marker in Supplemental Table 2. 

b. As a result, for axons projecting via OEC bridges in Figure 1, it is unclear how common these bridges are in the presence or absence of OECs.

We note that the number of spinal cord transected rats with bridges of axons crossing the lesion core are extremely rare following a severe spinal cord injury in adult mammals. Our first example of axon bridging following a complete spinal cord transection followed by OEC transplants was reported in Thornton et al., (2018) and compared to an incomplete transection in a fibroblast-transplanted control in his Figure 4. That figure also appeared the cover of Experimental Neurology when the paper was published. Figure 1 in the current paper was from an independent experiment which replicated the previously observed rare bridge formation. We noted this in the revised manuscript.

Page 6: “We note, however, that such bridge formation is rare following a severe spinal cord injury in adult mammals.”

c. For Figure 6., it is unclear whether cells having an alternative OEC morphology coincide with progenitor OEC subtype marker genes to a statistically significant degree. (see top paragraph on page 11)

Franceschini & Barnett (1996) suggested that there were 2 distinct types of OECs that could be distinguished by their different morphology: one type resembling a Schwann cell and the other, an astrocyte. The purpose of Figure 6 is to determine if there is a link between our OEC subtypes based on scRNAseq with those previously described based on morphology alone (Franceschini and Barnett, 1996). There could be agreement between large, flat or small fusiform OECs morphological and their progenitor status, but it is not required that the two classification types would significantly overlap. Here we report the percentage of morphology-based cell subtypes that show expression of our OEC subtype markers to estimate the overlap between the two. Our results indicate the two types of OEC morphologies share a certain degree of overlap, a finding that indicates similarities as well as differences between the two classification methods.

In our results section we show that ~3/4ths of the Ki67-expressing OEC progenitor cells sampled were astrocyte-like, i.e., flat in shape and weakly Ngfr^p75-labeled. The remaining ~1/4th of the Ki67-labeled  OECs were fusiform in shape and expressed Ngfr^p75 strongly. We feel that this is important to include as it is the only previous report of OB-OEC subtypes. The statistics of these results were in our original manuscript on page 11 and we further revise the text as follows:

Page 12: “To determine if the proliferative OECs differ in appearance from adult OECs, and whether there is concordance between our OEC subtypes based on gene expression markers and previously described morphology-based OEC subtyping (Franceschini & Barnett, 1996), we analyzed OECs identified with the anti-Ki67 nuclear marker and anti-Ngfr^p75  (Figure 6g-h). Of the Ki67-positive OECs in our cultures, 24% ± 8% were strongly Ngfr^p75-positive and spindle-shaped, whereas 76% ± 8% were flat and weakly Ngfr^p75-labeled (n=4 cultures, p\= 0.023). Here we show that a large percentage (~3/4ths) of proliferative OECs are characterized by large, flat morphology and weak Ngfr^p75 expression resembling the previously described morphology-based astrocyte-like subtype. Our results indicate the two types of OEC classifications share a certain degree of overlap, indicating similarities but also differences between the two classification methods.”

d. Similar quantification is missing in other types of data such as Western blot images (Fig. 9) and OEC marker gene data (for which p-values are not reported; Table S2). 

Response on Western blots: The Western blot signals shown in Figure 9 are from experiments that were designed to be qualitative rather than quantitative, by addressing the question, “Can we detect Reelin signals or not? in the different samples.” Both Western blots show that Reln^+/+ mouse olfactory bulbs (d) or cortices (e) contain Reelin whereas Reln^-/-  samples do not and therefore provide positive and negative controls, respectively. The rat olfactory nerve layer (ONL, laminae I-II of olfactory bulb, d lane 1; e lane 3) contains mainly OECs wrapped around the axons of the olfactory sensory neurons that transmit olfactory signals into the olfactory bulb. To address your request for quantification, Dr. Khankan measured the density of the three isoforms of Reelin, 400 kD, 300 kD and 180 kD in Fig. 9e and normalized them against the GADPH control (37 kD). The graph below shows the normalized band density in arbitrary units on the Y-axis relative to the first 3 conditions, i.e., Reln^+/+ and Reln^-/- mouse cerebral cortices and rat  Reln^+/+ ONL. Because the conditioned medium was collected from tissue culture medium rather than cells or tissue, the GAPDH control was not present and therefore these data cannot be normalized in a similar analysis.  

Author response image 1.

Response for OEC marker gene data: We now add new full supplementary Table S1 (for major cell types) and Table S2 (for OEC subtypes) to report statistical p values and adjusted p values, as well as additional statistics information including percent cell expressing a subtype marker in a given subtype versus in other subtypes. 

e. The addition of quantitative measures and, where appropriate, statistical comparisons with p-values or other significance measures, would be important for supporting the authors' claims and more rigorously conveying the results.

As detailed in the above responses, we now add quantifications and statistics to support the claims and enhance the rigor of our analysis.

(2) a. Some aspects of the experimental design that are relevant to the interpretation of the results are not explained. For example, OECs appear to be collected from only female rats, but the potential implications of this factor are not discussed.

We added a short explanation in the Discussion and Methods section regarding why spinal cord injury studies are carried out on female rats.

Page 24, Discussion: “Due to the extensive urinary tract dysfunction in spinal cord transected rats, most studies prefer females as their short urethra facilitates daily manual bladder expression. Our study, therefore, was carried out only on adult female rats, so sex differences and the generalizability of our findings to adult male rats would require further investigation.”

Page 26, Methods: “Only females were used in order to match the sex of previous SCI studies conducted exclusively on female rats (Dixie, 2019; Khankan et al., 2016; Takeoka et al., 2011; Thornton et al., 2018). Following complete thoracic spinal cord transection, an adult rat is unable to urinate voluntarily and therefore urine must be manually “expressed” twice a day throughout the experiment. Females have a shorter urethra than males, and thus their bladders are easier to empty completely.”

b. Additionally, it is unclear from the manuscript to what degree immunopurified cells are OECs as opposed to other cell types. The antibody used to retain OECs, nerve growth factor receptor p75 (Ngfr-p75), can also be expressed by non-OEC olfactory bulb cell types including astrocytes [1-3]. The possible inclusion of Ngfr-p75-positive but non-OEC cell types in the OEC culture is not sufficiently addressed.

(a) Cragnolini, A.B. et al., Glia, (2009), doi: 10.1002/glia.20857.

(b) Vickland H. et al., Brain Res., (1991), doi: 10.1016/0006-8993(91)91659-O.

(c) Ung K. et al., Nat Commun., (2021), doi: 10.1038/s41467-021-25444-3.

Our OECs are dissected primarily from the olfactory nerve layer that is concentrated medially and ventrally around the olfactory bulb together with a small part of the glomerular layer (layer II). OECs are the only glia present in olfactory nerve layer. Thus, although it is possible that other cell types also express Ngfr-p75 as pointed out by the reviewer and in the references provided, our OEC dissection method severely limits the number of astrocytes that might be included in our cultures. We further provide additional evidence (see updated Figure 2d and the detailed responses to the next question) that our immunopanned OECs using our dissection method consistently express all classic OEC markers but do not consistently express the majority of classic markers for other glial cell types such as astrocytes or oligodendrocytes.

Such non-OEC cell types are also not distinguished in the analysis of single-cell RNA sequencing data (only microglia, fibroblasts, and OECs are identified; Figure 2). Thus, it is currently unclear whether results related to the OEC subtype may have been impacted by these experimental factors.

We need to clarify that when determining potential cell types in Figure 2, we compared our cell cluster marker genes against a broad array of cell types including astrocytes, oligodendrocytes and Schwann cells, but the gene overlap was only significant for microglia, fibroblasts, and OECs, which we labeled in new Figure 2d. We added more details in methods and results to clarify how we determined the cell types in Figure 2 (text added below). We did consider all the potential cell types that could have been present in our OEC cultures, including astrocytes. However, astrocyte or oligodendrocyte markers were not significantly enriched in the clusters, but markers for microglia, fibroblasts, and OECs were prominent in the cell clusters.

In the revised Figure 2d, we now illustrate that the OEC clusters not only express typical OEC markers, but also express a few but not all marker genes from other glial cells. We show the comparative data on markers for astrocytes, oligodendrocytes, and Schwann cells in Figure 2d in parallel with the marker genes for OECs, microglia, and fibroblasts. For each of the other glial cell types, there are some genes which overlap with OECs, and that is the reason why we identified OECs as hybrid glia.

Page 6, Results: “Based on previously reported cell type marker genes for fibroblasts and major glial cell types including OECs, astrocytes, oligodendrocytes, and microglia, we found elevated expression of OEC marker genes in clusters 2, 3 and 7, microglia marker genes in clusters 4, 6, and 7, and fibroblast marker genes in clusters 0, 1, and 5 (Figure 2d).”

Page 33, Methods: “Additional marker genes for fibroblasts and multiple glial cell types including astrocytes, oligodendrocytes, and microglia were also used to compare with those of the cell clusters.”

(3) The introduction, while well written, does not discuss studies showing no significant effect of OEC implantation after spinal cord injury. The discussion also fails to sufficiently acknowledge this variability in the efficacy of OEC implantation. This omission amplifies bias in the text, suggesting that OECs have significant effects that are not fully reflected in the literature. The introduction would need to be expanded to properly address the nuance suggested by the literature regarding the benefits of OECs after spinal cord injury. Additionally, in the discussion, relating the current study to previous work would help clarify how varying observations may relate to experimental or biological factors.

We appreciate the insightful comment and have now included information about the variability in OEC transplantation in previous studies in both the introduction and discussion sections. We discuss technical differences that lead to variability in the Introduction and how our findings could help interpret the variability in the Discussion.

Page 4-5: Text added to the Introduction: “The outcomes of OEC transplantation studies after spinal cord injury vary substantially in the literature due to many technical differences between their experimental designs. The source of OECs has a great impact on the outcome, with OB-OECs showing more promise than peripheral lamina propria-derived OECs, and purified, freshly-prepared OECs being required for optimal OEC survival. Other important variables include the severity of the injury (hemisection to complete spinal cord transection), the age of the spinal cord injured host (early postnatal versus adult), and OEC transplant strategies (delayed or acute transplantation, cell transplants with or without a matrix; Franssen et al., 2007). Franssen et al. (2007) evaluated studies that used only OECs as a transplant, and reported that 41 out of 56 studies showed positive effects, such as OEC stimulation of regeneration, positive interactions with the glial scar and remyelination of axons. More recent systematic reviews and meta-analyses on the effects of OEC transplantation following different spinal cord injury models reported that OECs significantly improved locomotor function (Watzlawick et al.2016; Nakjavan-Shahraki et al., 2018), but did not improve neuropathic pain (Nakjavan-Shahraki et al., 2018.)”

Pages 24-25: Discussion on OEC source variability  “Extensive differences between OEC preparations contribute to the large variation in results from OEC treatments following spinal cord injury. This scRNA-seq study focused entirely on OB-OECs, and the next step would be to carry out similar studies on the peripheral, lamina-propria-derived OECs to discern the differences between these OEC populations. Such comparative studies using scRNA-seq will help define the underlying mechanisms and help resolve the variability in results from OEC-based therapy. Detailed studies of the composition of different OEC transplant types will contribute to identifying the most reparative cell transplantation treatments.”

Reviewer #1 (Recommendations For The Authors):

This is an extremely well-written and impactful series of experiments from a renowned leader in the field. The experimental questions are timely, with similar therapeutic approaches being prepared for clinical trial. The results address a gap that has persisted in the field for several decades and one that has been considered by many scientists long before technology existed to find answers. This highlights the importance of these experiments and the results reported here. With these things in mind, there are only a few minor factors that I have, that should be addressed to strengthen the paper.

We truly appreciate the positive evaluations from the reviewer!

Primary concerns

(1) Quantification of results: The authors report on the data with broad brush strokes, missing the opportunity to quantify results and strengthen the interpretations. For instance, when describing gene expression, what proportion of cells analyzed were expressing these genes? How did this compare with detectable levels of protein? Can the author draw correlations between data sets collected that could offer even more insight into the identities of the cells studied? There is also a missed opportunity to evaluate how transplantation into injured neural tissue might alter gene expression of the phenotypes identified prior to transplantation.

We appreciate these insightful comments and have added quantitative information and other relevant discussions in the revision. We now add Suppl Tables 1 (for major cell types including OECs, fibroblast, and microglia) and 2 (for OEC subtypes) to indicate the proportion of cells expressing each marker gene in each given cell cluster/subtype in the column. “Percentage of cells expressing the gene in the subtype/cell type” versus the proportion of cells expression the given marker genes in other cell types in the column “Percentage of cells expressing the gene in the other subtypes/cell types.” In the new supplementary tables, we report statistical p values and adjusted p values after multiple testing correction to indicate statistical significance.

Regarding the comparison with protein levels, we carried out immunohistochemistry experiments to confirm the proteins corresponding to OEC subtype markers. Our findings show that proteins for the gene markers can be detected, and thereby supports our sc-seq findings. However, the immunofluorescence only provides a qualitative measure of protein levels in situ, so we cannot perform a correlation analysis. This is something we plan to  pursue in a follow-up study with measurable protein levels. We also discuss future directions to examine the genes and proteins in in vivo transplantation studies in the Discussion.

(2) Discussion and interpretation: Greater depth to interpretation and discussion of data and its impact on future work is needed. For example, on pages 20-21, the authors reflect briefly on why Reelin might be of interest (it could lead to Dab-1 expression), but why is that important? There are several instances like this where it would be useful for the authors to provide a little more insight into the potential impact of these data and interpretations.

We appreciate these valuable suggestions. We have revised our Results and Discussion sections to offer deeper insight and interpretation of the importance of the data, especially that for Reelin.

Page 17: Results: “In the canonical Reelin-signaling pathway, Reelin binds to the very-low-density lipoprotein receptor (Vldlr) and apolipoprotein E receptor 2 (ApoER2) and induces Src-mediated tyrosine phosphorylation of the intracellular adaptor protein Disabled-1 (Dab1). Both Reelin and Dab1 are highly expressed in embryos and contribute to correct neuronal positioning.”

Page 22-23, Discussion: “Reelin is a developmentally expressed protein detected in specific neurons, in addition to OECs and Schwann cells. The canonical Reelin-signaling pathway involves neuronal-secreted Reelin binding to Vldlr and ApoER2 receptors expressed on Dab1-labeled neurons. Following Reelin binding, Dab1 is phosphorylated by Src family kinases which initiates multiple downstream pathways. Very little is known, however, about Reelin secreted by glia. Panteri et al. (2006) reported that Schwann cells express low levels of Reelin in adults, and that it is upregulated following a peripheral nerve crush, as is reported above for many neurotrophic factors. Reelin loss in Schwann cells reduced the diameter of small myelinated axons but did not affect unmyelinated axons (Panteri et al., 2005). In the olfactory system, OECs ensheath the Dab1-labeled, unmyelinated axons of olfactory sensory neurons which are continuously generated and die throughout life. OEC transplantation following spinal cord injury would provide an exogenous source of Reelin that could phosphorylate Dab1-containing neurons or their axons. Dab1 is expressed at high levels in the axons of some projection neurons, such as the corticospinal pathway (Abadesco et al., 2014). Future experiments are needed to explore the function that glial-secreted Reelin may have on axonal regeneration.”

Minor concerns

(3) The authors reflect on the spontaneous glial bridge that develops in the repairing spinal cord of Zebrafish, but perhaps even more relevant is that this same phenomenon occurs in mammals as well if the spinal cord is injured during early development (opossum; Lane et al, EJN 2007). This should be considered and the statement that there is little regeneration in the mammalian spinal cord should be clarified.

We appreciate this insightful comment. We now add discussions of the axonal regeneration and bridging observed following severe spinal cord injury in young developing mouse and opossum spinal cords.

Page 23: “Adult mammals show little evidence of spontaneous axonal regeneration after a severe spinal cord injury in contrast to transected neonatal rats (Bregman, 1987; Bregman et al., 1993) and young postnatal opossums (Lane et al., 2007). In immature mammals, axons continue to project across or bridge the spinal cord transection site during development. Lower organisms such as fish, show even more evidence of regeneration following severe SCI. Mokalled et al. (2016) reported that glial secretion of Ctgfa/Ccn2 was both necessary and sufficient to stimulate a glial bridge for axon regeneration across the zebrafish transection site. Cells in the injury site that express Ctgf include ependymal cells, endothelial cells, and reactive astrocytes (Conrad et al., 2005; Mokalled et al., 2016; Schwab et al., 2001). Here we show that, although rare, Ctgf-positive OECs can contribute to glial bridge formation in adult rats. The most consistent finding among our severe SCI studies combined with OEC transplantation is the extent of remodeling of the injury site and axons growing into the inhibitory lesion site, together with OECs and astrocytes. The formation of a glial bridge across the injury was critical to the spontaneous axon generation seen in zebrafish (Mokalled et al., 2016) and likely contributed to the axon regeneration detected in our OEC transplanted, transected rats (Dixie, 2019; Khankan et al., 2016; Takeoka et al., 2011; Thornton et al., 2018).

Reviewer #2 (Recommendations For The Authors):

(1) The manuscript title and abstract must include the species and sex studied.

The title and abstract have been modified as suggested.

Page 1: “Olfactory ensheathing cells from adult female rats are hybrid glia that promote neural repair”

(2) OECs submitted for sequencing were like those about to be transplanted; however, the phenotype of the cells would likely change immediately and shift over time post-implantation. Please briefly address or discuss this point in the Discussion (or Results).

We have added this important discussion point.

Pages 23-24: Discussion: “We recognize that this study is a single snapshot of OEC gene expression derived from adult female rats before they are transplanted above and below the spinal cord transection site. We would expect the gene expression of transplanted OECs to change in each new environment, i.e. as they migrate into the injury site, integrate into the glial scar, and wrap around axons. Based on our past studies, OECs survived in an outbred Sprague-Dawley rat model for ~ 4 weeks (Khankan et al., 2016) and in an inbred Fischer 344 model for 5 months (Dixie, 2019). As spinal cord injury transplant procedures are further enhanced and OEC survival improves, these hybrid glial cells should be examined at multiple time points to better evaluate their proregenerative characteristics.”

(3) Page 12: Use of "monocytes" - the word "monocyte" implies a circulating, undifferentiated innate immune cell. This should not be used interchangeably with macrophage or microglia.

We agree and now refer to microglia or macrophages depending on the context. We did leave the term monocyte in Table 3 if these cells were found in a top 20 gene reported in the references.

(4) Page 12: "We now show that these unique monocytes reported between the bundles of olfactory axons surrounded by OECs (Smithson & Kawaja, 2010), are in fact, a distinct subtype of OECs."

Is it possible to conclude that these cells are a "distinct subtype of OECs?" Perhaps these cells are a hybrid between microglia/macrophages and OECs? This is speculative, so should be worded more carefully - especially in the Results section. Please clarify, dampen conclusions, and/or better justify the wording here.

We agree and have modified the entire paragraph to dampen and more carefully explain our conclusions. We also added an additional observation that strengthens the relationship between OECs and microglial/macrophages.  

Page 12, Results: Additional observation: “In fact, all top 20 genes in cluster 3 are expressed in microglia, macrophages, and/or monocytes (Suppl. Table 3).”

Page 13, Results: The statement referenced in your review was deleted and we wrote the following: “Smithson and Kawaja (2010) identified unique microglial/macrophages that immunolabeled with Iba-1 (Aif1) and Annexin A3 (Anxa3) in the olfactory nerve and outer nerve layer of the olfactory bulb. These authors proposed that Iba1-Anxa3 double-labeled cells were a distinct population of microglia/macrophages that protected the olfactory system against viral invasion into the cranial cavity. Based on our scRNA-seq data we offer an alternative interpretation that at least some of these Iba-1-Anxa3 cells may be a hybrid OEC-microglial cell type. Supporting this interpretation, there are a number of reports that suggest OECs frequently function as phagocytes (e.g., Khankan et al., 2016; Nazareth et al., 2020; Su et al. 2013).”

(5) Page 13: "Pseudotime trajectory analysis, a widely used approach to predict cell plasticity and lineages based on scRNA-seq data, suggests that there are potential transitions between specific OEC subclusters." This is interesting but is somewhat unclear. Please add one more sentence to aid the reader's understanding regarding how this analysis is performed.

Thank you for your valuable feedback. We have revised the text for clarity as follows:

Page 14, Results: “We performed pseudotime trajectory analysis using the Slingshot algorithm to infer lineage trajectories, cell plasticity and lineages by ordering cells in pseudotime based on their transcriptional progression reflected in scRNA-seq data. Transcriptional progression refers to the changes in gene expression profiles of cells as they undergo differentiation or transition through different states. The trajectory analysis results suggest that there are potential transitions between specific OEC subclusters.”

(6) The authors could discuss potential reasons for variability in OEC treatment results after spinal cord injury between studies and labs. How might sequencing results here inform the debate about whether OECs are helpful or not?

In response to the Public Review, we added discussions about the variability in OEC treatments between studies in both the Introduction and Discussion, and these comments are copied on pages 6-7 of this document. In the Discussion we included a statement about how the current findings may inform the debate on OECs.

(7) Discussion: please add a discussion of limitations and future directions that addresses the following points:

a) Please add one sentence on the lack of studying sex differences - only females were studied here.

b) There is no correlation or modulation of any target genes, so all results here are correlative.

c) Please add a brief paragraph with future directions for the field, including acknowledgment that the role of OECs in repair after SCI is not fully resolved and that future studies might consider targeting some of the specific pathways described herein.

d) Which pathways and OEC subpopulations likely best support repair, and how might these be reinforced or better maintained in the SCI environment? If not known, what are the next steps for identifying the most reparative OEC subtype?

Thank you for the valuable suggestions. We have added these to the discussion as detailed below.

Pages 23-25, Discussion:

“Limitations of these OEC scRNA-Seq studies”

“We recognize that this study is a single snapshot of OEC gene expression derived from adult female rats before they are transplanted above and below the spinal cord transection site. We would expect the gene expression of transplanted OECs to change in each new environment, i.e. as they migrate into the injury site, integrate into the glial scar, and wrap around axons. Based on our past studies, OECs survived in an outbred Sprague-Dawley rat model for ~ 4 weeks (Khankan et al., 2016) and in an inbred Fischer 344 model for 5 months (Dixie, 2019). As spinal cord injury transplant procedures are further enhanced and OEC survival improves, these hybrid glial cells should be examined at multiple time points to better evaluate their proregenerative characteristics.”

“Due to the extensive urinary tract dysfunction in spinal cord transected rats, most studies are conducted on females as their short urethra facilitates daily manual bladder expression. Our study was carried out only on adult female rats, so sex differences and the generalizability of our findings to adult male rats would require further investigation. We also did not modulate any of the genes or proteins in the identified OEC subtypes to test their causal and functional roles, thus our findings remain correlative in the current study. Future gene/protein modulation studies are necessary to understand the functional roles of the individual OEC subtypes in the context of their reparative functions to determine which pathways and subtypes are more critical and can be enhanced for neural repair. Our current findings build the foundation for these future studies to help resolve the role of OECs in spinal cord injury repair.” 

“Extensive differences between OEC preparations contribute to the large variation in results from OEC treatments following spinal cord injury. This scRNA-seq study focused entirely on OB-OECs, and the next step would be to carry out similar studies on the peripheral, lamina-propria-derived OECs to discern the differences between the two OEC populations. Such comparative studies using scRNA-seq will help define the underlying mechanisms and resolve the variability in results from OEC-based therapy. Detailed studies of the composition of different OEC transplant types will contribute to identifying the most reparative cell transplantation treatments.”

(8) Figure 6: What is the major point of this figure and its related immunocytochemistry? Please clarify.

Franceschini & Barnett (1996) suggested that there were 2 distinct types of OECs that could be distinguished by their different morphology: One type resembling a Schwann cell and the other, an astrocyte. The purpose of Figure 6 is to determine if there is a link between our scRNA-seq-based OEC subtypes with those previously described based on morphology alone (Franceschini and Barnett, 1996). In our results section we show that ~3/4ths of the OECs sampled that were Ki67+ progenitor cells and were astrocyte-like, i.e., flat in shape and weakly Ngfr^p75-labeled. The remainder were Schwann cell-like, fusiform in shape and strongly Ngfr^p75-labeled. Our results indicate the two types of OEC classifications share certain degrees of overlap, indicating similarities but also differences between the different classification methods.

(9) Figure 9, caption: "OEC whole cell lysates (WCL; lanes: 4, 6, and 8), and OEC conditioned medium (CM; lanes: 5 and 7)."  This statement is unclear - please clarify the result here.

We added clarification to the legend for Figure 9d. 

Page 50: (d) “Western blot confirms the expression of Reelin in rat olfactory nerve layer I and layer II (ONL; lane 1 of western blot). Reln^+/+ and Reln^-/- mouse olfactory bulbs were used as positive and negative controls, respectively (lanes: 2 and 3). Reelin that was synthesized by cultured OECs was found in whole cell lysates (WCL; lanes: 4, 6, and 8), whereas Reelin that was secreted by cultured OECs into tissue culture medium was measured in the OEC “conditioned medium” (CM; lanes: 5 and 7). GAPDH was the loading control for tissue homogenates (lanes 1-4, 6, 8).”

(10) Methods: A Cat. No. for all antibodies and key supplies should be included.

Response: All of the antibody information in the revised version is in Suppl. Table 4. Information for other key supplies is included in the extensive methods section.

(11) Methods: How was primary antibody specificity validated for less-used antibodies? Background staining can be a major issue after SCI; e.g., with the CTGF antibody used in Figure 5.

The spinal cord section shown in Figure 5 was compared to sections from the same SCI cohort that had been injected with control cells, i.e. skin fibroblasts. We have used the first two antibodies (anti-Glial fibrillary acidic protein and anti-Green fluorescent protein) for many years so only the CTGF was a “less-used antibody.” Our strategy for working with “less-used” or “newly-purchased” antibodies was as follows.

First, we studied the literature to find the best antibodies for neuronal tissue. Many of the images in Figure 7 were generated with antibodies purchased just for this study. Our goal was to characterize them on normal adult lamina propria and olfactory bulb tissues rather than in the injured spinal cord where background can be an issue. In the olfactory bulb we examined the olfactory nerve layer where OECs are concentrated and then examined the olfactory epithelium, lamina propria, and the deep layers of the olfactory bulb to find regions without immunolabel. As described above, we tested anti-CTGF antibodies in SCI sections implanted with skin fibroblasts controls when conducting experiments for CTGF in sections with OECs. New antibodies were tested at multiple concentrations and we tried different immunocytochemical techniques. Anti-CTFG is expressed by several different cell types, but expression is low in most of the areas above and below the injury site. Despite our success with many “newly-purchased” antibodies there were at least 4 of them that we were never able obtain specific labeling. 

(12) Will the data (especially the sequencing data) be shared publicly?

The data has been uploaded to and shared via the public data repository GEO. Data availability is stated on the title page of this manuscript.

eLife Assessment

This important paper provides solid evidence for an alternative conceptualization of the functional role of the place and grid cell network in the medial temporal lobe for memory as opposed to spatial processing or navigation. The theory is extensive, tightly integrating data on various spatial cell types. It accounts for many experimental results and generates strong predictions for future studies that will be of interest to researchers in this field. The impact of the work would be strengthened if future experiments reveal that grid cells do indeed encode specific nonspatial features.

Reviewer #1 (Public review):

Huber proposes a theory where the role of the medial temporal lobe (MTL) is memory, where properties of spatial cells in the MTL can be explained through memory function rather than spatial processing or navigation. Instantiating the theory through a computational model, the author shows that many empirical phenomena of spatial cells can be captured, and may be better accounted through a memory theory. It is an impressive computational account of MTL cells with a lot of theoretical reasoning and aims to tightly relate to various spatial cell data.

In general, the paper is well written, and has been greatly improved after revision for clarity and situating the model in the context of the literature. Below are a few responses to the author's rebuttal.

(2 & 3) In response to my previous review point 2 and 3, the author has now added "According to this model, hexagonally arranged grid cells should be the exception rather than the rule when considering more naturalistic environments." It is good to know that it captures data that show non-grid like responses in more complex and realistic environments. However, the model still focuses on explaining the spatial firing aspect of grid cells even though they are not supposed to be spatial. I noted in my previous review, "If it's not encoding a spatial attribute, it doesn't have to have a spatial field. For example, it could fire in the whole arena". The author notes inhibitory drive and habituation. Habituation happens, but then spatial cell responses are supposed (or assumed) to be still strong after many visits to that environment. More generally, I am more convinced that grid-like and spatial coding are a special case - both in navigation and memory. In a way I believe the author agrees, though the work here focuses on capturing spatial properties (which is understandable given the literature). In conclusion, though there may be theoretical disagreements, I find the points the author raises fair.

(4) The difference between mEC and lEC or PRC for encoding non-spatial vs spatial attributes is still not clear to me - though not crucial for the point of this paper.

(5) Thank you for providing a video - this makes it extremely clear how learning occurs.

Reviewer #3 (Public review):

The author presents a novel theory and computational model suggesting that grid cells do not encode space, but rather encode non-spatial attributes. Place cells in turn encode memories of where those specific attributes occurred. The theory accounts for many experimental results and generates useful predictions for future studies. The model's simplicity and potential explanatory power will interest others in the field. There are, however, a few weaknesses outlined below which undermine the theory.

Main criticisms:

(1) A crucial assumption of the model is that grid cells express grid-like firing patterns if and only if the content of experience is constant in space. It is difficult to imagine a real world example that satisfies this assumption. Odors and sounds are used as examples. While they are often more spatially diffuse than an object on the ground, odors and sounds have sources that are readily detectable and thus are not constant in space. Animals can easily navigate to a food source or to a vocalizing conspecific. This assumption is especially problematic because it predicts that all grid cells should become silent when their preferred non-spatial attribute (e.g. a specific odor) is missing. I'm not aware of any experimental data showing that grid cells become silent. On the contrary, grid cells are known to remain active across all contexts that have been tested, including across sleep/wake states. Unlike place cells, grid cells have never been shown to turn off. Since grid cells are active in all contexts, their preferred attribute must also be present in all contexts, and therefore they would not convey any information about the specific content of an experience. The author lists many attributes that could in theory be constant in a laboratory setting, but there is no data I'm aware of that shows this is true in practice. As it stands, this crucial assumption of the model remains mere speculation.

(2) The proposed novelty of this theory is that other models all assume that grid cells encode space. This is not quite true of models based on continuous attractor networks, the discussion of which is essentially absent. More specifically, attractor models focus on the importance of intrinsic dynamics within entorhinal cortex in generating the grid pattern. While this firing pattern is aligned to space during navigation and therefore can be used a representation of that space, the neural dynamics are preserved even during sleep. Similarly, it is because the grid pattern does not strictly encode physical space that grid-like signals are also observed in relation to other two-dimensional continuous variables.

(3) The use of border cells or boundary vector cells as the main (or only) source of spatial information in the hippocampus is not well supported by experimental data. Border cells in entorhinal cortex are not active in the center of an environment. Boundary-vector cells can fire farther away from the walls, but are not found in entorhinal cortex. They are located in the subiculum, a major output of the hippocampus. While the entorhinal-hippocampal circuit is a loop, the route from boundary-vector cells to place cells is much less clear than from grid cells. Moreover, both border cells and boundary-vector cells (which are conflated in this paper) comprise a small population of neurons compared to grid cells.

Minor comments:

(1) There is substantial theoretical and experimental work supporting the idea that grid cell modules instantiate continuous attractor networks, yet this class of models is largely ignored:

p. 7 "In contrast, most grid cell models (Bellmund et al., 2016; Bush et al., 2015; Castro & Aguiar, 2014; Hasselmo, 2009; Mhatre et al., 2012; Solstad et al., 2006; Sorscher et al., 2023; Stepanyuk, 2015; Widloski & Fiete, 2014) are domain specific models of spatial navigation"

The following references should be added:

McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E. I. & Moser, M.-B. Path integration and the neural basis of the 'cognitive map'. Nat. Rev. Neurosci. 7, 663-678 (2006).

Fuhs, M. C. & Touretzky, D. S. A spin glass model of path integration in rat medial entorhinal cortex. J. Neurosci. 26, 4266-4276 (2006).

Burak, Y. & Fiete, I. R. Accurate path integration in continuous attractor network models of grid cells. PLoS Comput. Biol. 5, e1000291 (2009).

Guanella, A., Kiper, D. & Verschure, P. A model of grid cells based on a twisted torus topology. Int. J. Neural Syst. 17, 231-240 (2007).

Couey, J. J. et al. Recurrent inhibitory circuitry as a mechanism for grid formation. Nat. Neurosci. 16, 318-324 (2013).

(Note: the Bellmund et al. (2016) citation is likely a mistake and was intended to be Bellmund et al. (2018).)

(2) The author claims in two places that this model is the first to explain that grid cell population activity lies on a torus. While it may be the first explicit computational account of why grid cell activity is mapped onto a torus, these claims should be moderated for clarity, for example by adding "but see McNaughton et al. (2006) and others".

Box 1. Results Uniquely Explained by this Memory Model - the population code of grid cells lies on a torus

p.11 "In addition, this simplifying assumption allows the model to capture the finding that the population of grid cells lies on a torus (Gardner et al., 2022), although I note that the model was developed before this result was known."

(3) Lateral entorhinal cortex is largely ignored in this memory model. It should be considered that the predominance of spatial representations reported in the literature is due to historical reasons. Namely, the discovery of hippocampal place cells spurred interest in looking upstream for the source of spatial information, which was later abundantly clear in medial entorhinal cortex. Lateral entorhinal cortex is relatively understudied, but is known to encode odors, objects, and time in a way that medial entorhinal cortex does not. It is therefore confusing to assume that these attributes are instead encoded by grid cells.

Author response:

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

Public Reviews

Reviewer #1 (Public Review): 

(1) Although the theory is based on memory, it also is based on spatially-selective cells.

Not all cells in the hippocampus fulfill the criteria of place/HD/border/grid cells, and place a role in memory. E.g., Tonegawa, Buszaki labs' work does not focus on only those cells, and there are certainly a lot of non-pure spatial cells in monkeys (Martinez-Trujillo) and humans (iEEG). Does the author mainly focus on saying that "spatial cells" are memory, but do not account for non-spatial memory cells? This seems to be an incomplete account of memory - which is fine, but the way the model is set up suggests that *all* memory is, place (what/where), and non-spatial attributes ("grid") - but cells that don't fulfil these criteria in MTL (Diehl et al., 2017, Neuron; non-grid cells; Schaeffer et al., 2022, ICML; Luo et al., 2024, bioRxiv) certainly contribute to memory, and even navigation. This is also related to the question of whether these cell definitions matter at all (Luo et al., 2024). The authors note "However, this memory conjunction view of the MTL must be reconciled with the rodent electrophysiology finding that most cells in MTL appear to have receptive fields related to some aspect of spatial navigation (Boccara et al., 2010; Grieves & Jeffery, 2017). The paucity of non-spatial cells in MTL could be explained if grid cells have been mischaracterized as spatial." Is the author mainly talking about rodent work?

There is a new section in the introduction that deals with these issues, titled ‘Why Model the Rodent Navigation Literature with a Memory Model?’ That section reads:

“Spatial navigation is inherently a memory problem – learning the spatial arrangement of a new enclosure requires memory for the conjunction of what and where. This has long been realized and in the introduction to ‘Hippocampus as a Cognitive Map’, O’Keefe and Nadel (1978) wrote “We shall argue that the hippocampus is the core of a neural memory system providing an objective spatial framework within which the items and events of an organism's experience are located and interrelated” (emphasis added). Furthermore, in the last chapter of their book, they extended cognitive map theory to human memory for non-spatial characteristics. However, in the decades since the development of cognitive map theory, the rodent spatial navigation and human memory literatures have progressed somewhat independently.

The ideas proposed in this model are an attempt to reunify these literatures by returning to the original claim that spatial navigation is inherently a memory problem. The goal of the current study is to explain the rodent spatial navigation literature using a memory model that has the potential to also explain the human memory literature. In contrast, most grid cell models (Bellmund et al., 2016; Bush et al., 2015; Castro & Aguiar, 2014; Hasselmo, 2009; Mhatre et al., 2012; Solstad et al., 2006; Sorscher et al., 2023; Stepanyuk, 2015; Widloski & Fiete, 2014) are domain specific models of spatial navigation and as such, they do not lend themselves to explanations of human memory. Thus, the reason to prefer this model is parsimony. Rather than needing to develop a theory of memory that is separate from a theory of spatial navigation, it might be possible to address both literatures with a unified account.

This study does not attempt to falsify other theories of grid cells. Instead, this model reaches a radically different interpretation regarding the function of grid cells; an interpretation that emerges from viewing spatial navigation as a memory problem. All other grid cell models assume that an entorhinal grid cell displaying a spatially arranged grid of firing fields serves the function of spatial coding (i.e., spatial grid cells exist to support a spatial metric). In contrast, the proposed memory model of grid cells assumes that the hexagonal tiling reflects the need to keep memories separate from each other to minimize confusion and confabulation – the grid pattern is the byproduct of pattern separation between memories rather than the basis of a spatial code. 

It is now understood that grid-like firing fields can occur for non-spatial twodimensional spaces. For instance, human entorhinal cortex exhibits grid-like responses to video morph trajectories in a two-dimensional bird neck-length versus bird leg-length space (Constantinescu et al., 2016). As a general theory of learning and memory, the proposed memory model of grid cells is easily extended to explain these results (e.g., relabeling the border cell inputs in the model as neck-length and leg-length inputs). However, there are other grid cell models that can explain both spatial grid cells as well as non-spatial grid-like responses (Mok & Love, 2019; Rodríguez-Domínguez & Caplan, 2019; Stachenfeld et al., 2017; Wei et al., 2015). Similar to this memory model of grid cells, these models are also positioned to explain both the rodent spatial navigation and human memory literatures. Nevertheless, there is a key difference between this model and other grid cell models that generalize to non-spatial representations. Specifically, these other models assume that grid cells exhibiting spatial receptive fields serve the function of identifying positions in the environment (i.e., their function is spatial). As such, these models do not explain why most of the input to rodent hippocampus appears to be spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). This memory model of grid cells provides an answer to the apparent paucity of nonspatial cell types in rodent MTL by proposing that grid cells with spatial receptive fields have been misclassified as spatial (they are what cells rather than where cells) and that place cells are fundamentally memory cells that conjoin what and where.”

(2) Related to the last point, how about non-grid multi-field mEC cells? In theory, these also should be the same; but the author only presents perfect-look grid cells. In empirical work, clearly, this is not the case, and many mEC cells are multi-field non-grid cells (Diehl et al., 2017). Does the model find these cells? Do they play a different role? As noted by the author "Because the non-spatial attributes are constant throughout the two-dimensional surface, this results in an array of discrete memory locations that are approximately hexagonal (as explained in the Model Methods, an "online" memory consolidation process employing pattern separation rapidly turns an approximately hexagonal array into one that is precisely hexagonal). " If they are indeed all precisely hexagonal, does that mean the model doesn't have non-grid spatial cells? 

Grid cells with irregular firing fields are now considered in the discussion with the following paragraphs

“According to this model, hexagonally arranged grid cells should be the exception rather than the rule when considering more naturalistic environments. In a more ecologically valid situation, such as with landmarks, varied sounds, food sources, threats, and interactions with conspecifics, there may still be remembered locations were events occurred or remembered properties can be found, but because the non-spatial properties are non-uniform in the environment, the arrangement of memory feedback will be irregular, reflecting the varied nature of the environment. This may explain the finding that even in a situation where there are regular hexagonal grid cells, there are often irregular non-grid cells that have a reliable multi-location firing field, but the arrangement of the firing fields is irregular (Diehl et al., 2017). For instance, even when navigating in an enclosure that has uniform properties as dictated by experimental procedures, they may be other properties that were not well-controlled (e.g., a view of exterior lighting in some locations but not others), and these uncontrolled properties may produce an irregular grid (i.e., because the uncontrolled properties are reliably associated with some locations but not others, hippocampal memory feedback triggers retrieval of those properties in the associations locations).

In this memory model, there are other situations in which an irregular but reliable multilocation grid may occur, even when everything is well controlled. In the reported simulations, when the hippocampal place cells were based on variation in X/Y (as defined by Border cells), nothing else changed as a function of location, and the model rapidly produced a precise hexagonal arrangement of hippocampal place cell memories. When head direction was included (i.e., real-world variation in X, Y, and head direction), the model still produced a hexagonal arrangement as per face-centered cubic packing of memories, but this precise arrangement was slower to emerge, with place cells continuing to shift their positions until the borders of the enclosure were sufficiently well learned from multiple viewpoints. If there is real-world variation in four or more dimensions, as is likely the case in a more ecologically valid situation, it will be even harder for place cell memories to settle on a precise regular lattice. Furthermore, in the case of four dimensions, mathematicians studying the “sphere packing problem” recently concluded that densest packing is irregular (Campos et al., 2023). This may explain why the multifield grid cells for freely flying bats have a systematic minimum distance between firing fields, but their arrangement is globally irregular (Ginosar et al., 2021). Assuming that the memories encoded by a bat include not just the three real-world dimensions of variation, but also head direction, the grid will likely be irregular even under optimal conditions of laboratory control.”

(3) Theoretical reasons for why the model is put together this way, and why grid cells must be coding a non-spatial attribute: Is this account more data-driven (fits the data so formulated this way), or is it theoretical - there is a reason why place, border, grid cells are formulated to be like this. For example, is it an efficient way to code these variables? It can be both, like how the BVC model makes theoretical sense that you can use boundaries to determine a specific location (and so place cell), but also works (creates realistic place cells). 

The motivation for this model is now articulated in the new section, quoted above, titled ‘Why Model the Rodent Navigation Literature with a Memory Model?’ Regarding the assumption that border cells provide a spatial metric, this assumption is made for the same reasons as in the BVC model. Regarding this, the text said: “These assumptions regarding border cells are based on the boundary vector cell (BVC) model of Barry et al. (2006). As in the BVC model, combinations of border cells encode where each memory occurred in the realworld X/Y plane.”. A new sentence is added to model methods, stating: “This assumption is made because border cells provide an efficient representation of Euclidean space (e.g., if the animal knows how far it is from different walls of the enclosure, this already available information can be used to calculate location).”

But in this case, the purpose of grid cell coding a non-spatial attribute, and having some kind of system where it doesn't fire at all locations seems a little arbitrary. If it's not encoding a spatial attribute, it doesn't have to have a spatial field. For example, it could fire in the whole arena - which some cells do (and don't pass the criteria of spatial cells as they are not spatially "selective" to another location, related to above).  

Some cells have a constant high firing rate, but they are the exception rather than the rule. More typically, cells habituate in the presence of ongoing excitatory drive and by doing so become sensitive to fluctuations in excitatory drive. Habituation is advantageous both in terms of metabolic cost and in terms of function (i.e., sensitivity to change). This is now explained in the following paragraph:

“In theory, a cell representing a non-spatial attribute found at all locations of an enclosure (aka, a grid cell in the context of this model), could fire constantly within the enclosure. However, in practice, cells habituate and rapidly reduce their firing rate by an order of magnitude when their preferred stimulus is presented without cessation (Abbott et al., 1997; Tsodyks & Markram, 1997). After habituation, the firing rate of the cell fluctuates with minor variation in the strength of the excitatory drive. In other words, habituation allows the cell to become sensitive to changes in the excitatory drive (Huber & O’Reilly, 2003). Thus, if there is stronger top-down memory feedback in some locations as compared to others, the cell will fire at a higher rate in those remembered locations rather than in all locations even though the attribute is found at all locations. In brief when faced with constant excitatory drive, the cell accommodates, and becomes sensitive to change in the magnitude of the excitatory drive. In the model simulation, this dynamic adaptation is captured by supposing that cells fire 5% of the time on-average across the simulation, regardless of their excitatory inputs.”

(4) Why are grid cells given such a large role for encoding non-spatial attributes? If anything, shouldn't it be lateral EC or perirhinal cortex? Of course, they both could, but there is less reason to think this, at least for rodent mEC.  

This is a good point and the following paragraph has been added to the introduction to explain that lateral EC is likely part of the explanation. But even when including lateral EC, it still appears that most of the input to hippocampus is spatial.

“One possible answer to the apparent lack of non-spatial cells in MTL is to highlight the role of the lateral entorhinal cortex (LEC) as the source of non-spatial what information for memory encoding (Deshmukh & Knierim, 2011). LEC can be contrasted with mEC, which appears to only provide where information (Boccara et al., 2010a; Diehl et al., 2017). Although it is generally true that LEC is involved in non-spatial processing, there is evidence that LEC provides some forms of spatial information (Knierim et al., 2014). The kind of non-spatial information provided by LEC appears to be in relation to objects (Connor & Knierim, 2017; Wilson et al., 2013). However, in a typical rodent spatial navigation study there are no objects within the enclosure. Thus, although the distinction between mEC and LEC is likely part of the explanation, it is still the case that rodent entorhinal input to hippocampus appears to heavily favor spatial information.”

(5) Clarification: why do place cells and grid cells differ in terms of stability in the model? Place cells are not stable initially but grid cells come out immediately. They seem directly connected so a bit unclear why; especially if place cell feedback leads to grid cell fields. There is an explanation in the text - based on grid cells coding the on-average memories, but these should be based on place cell inputs as well. So how is it that place fields are unstable then grid fields do not move at all? I wonder if a set of images or videos (gifs) showing the differences in spatial learning would be nice and clarify this point.  

In this revision, I provide a new video focused on learning of place cell memories that include head direction. This second video is in relation to the results reported in Figure 9. The short answer is that the grid fields for the non-spatial cell are based on the average across several view-dependent memories (i.e., across several place cells that have head direction sensitivity) and the average is reliable even if the place cells are unstable. The text of this explanation now reads:

“Why was the grid immediately apparent for the non-spatial attribute cell whereas the grid took considerable prior experience for the head direction cells? The answer relates to memory consolidation and the shifting nature of the hippocampal place cells. Head direction cells only produced a reliable grid once the hippocampal place cells (aka, memory cells) assumed stable locations. During the first few sessions, the hippocampal place cells were shifting their positions owing to pattern separation and consolidation. But once the place cells stabilized, they provided reliable top-down memory feedback to the head direction cells in some places but not others, thus producing a reliable grid arrangement to the firing maps of the head direction cells. In other words, for the head direction cells, the grid only appeared once the place cells stabilized. This slow stabilization of place fields is a known property (Bostock et al., 1991; Frank et al., 2004).

In the simulation, the place cells did not stabilize until a sufficient number of place cells were created (Figure 9C). Specifically, these additional memories were located immediately outside the enclosure, around all borders (Figure 9D). These “outside the box” memories served to constrain the interior place cells, locking them in position despite ongoing consolidation. This dynamic can be seen in a movie showing a representative simulation. The movie shows the positions of the head direction sensitive place cells during initial learning, and then during additional sessions of prior experience as the movie speeds up (see link in Figure 9 capture).

Why did the non-spatial grid cell (k) produce a grid immediately, before the place cells stabilized? As discussed in relation to Figure 8, the non-spatial grid cell is the projection through the 3D volume of real-world coordinates that includes X, Y, and head direction. Each grid field of a non-spatial grid cell reflects feedback from several place cells that each have a different head direction sensitivity (see for instance the allocentric pairs of memories illustrated in Figure 8C and 8D). Thus, each grid field is the average across several memories that entail different viewpoints and this averaging across memories provides stability even if the individual memories are not yet stable. This average of unstable memories produces a blurry sort of grid pattern without any prior experience.

A final piece of the puzzle relies on the same mechanism that caused the grid pattern to align with the borders as reported in the results of Figures 6 and 7. Specifically, there are some “sticky” locations with ongoing consolidation because the connection weights are bounded. Because weights cannot go below their minimum or above their maximum, it is slightly more difficult for consolidation to push or pull connection weights over the peak value or under the minimum value of the tuning curve. Thus, the place cells tend to linger in locations that correspond to the peak or trough of a border cell. There are multiple peak and trough locations but for the parameter values in this simulation, the grid pattern seen in Figure 9C shows the set of peak/trough locations that satisfy the desired spacing between memories. Thus, the average across memories shows a reliable grid field at these locations even though the memories are unstable.”

(6) Other predictions. Clearly, the model makes many interesting (and quite specific!) predictions. But does it make some known simple predictions? 

• More place cells at rewarded (or more visited) locations. Some empirical researchers seem to think this is not as obvious as it seems (e.g., Duvellle et al., 2019; JoN; Nyberg et al., 2021, Neuron Review).  

• Grid cell field moves toward reward (Butler et al., 2019; Boccera et al., 2019).  

• Grid cells deform in trapezoid (Krupic et al., 2015) and change in environments like mazes (Derikman et al., 2014).  

Thank you for these suggestions and I have added the following paragraph to the discussion:

“In terms of the animal’s internal state, all locations in the enclosure may be viewed as equally aversive and unrewarding, which is a memorable characteristic of the enclosure. Reward, or lack thereof, is arguably one of the most important nonspatial characteristics and application of this model to reward might explain the existence of goal-related activity in place cells (Hok et al., 2007; although see Duvelle et al., 2019), reflecting the need to remember rewarding locations for goal directed behavior. Furthermore, if place cell memories for a rewarding location activate entorhinal grid cells, this may explain the finding that grid cells remap in an enclosure with a rewarded location such that firing fields are attracted to that location (Boccara et al., 2019; Butler et al., 2019). Studies that introduce reward into the enclosure are an important first step in terms of examining what happens to grid cells when the animal is placed in a more varied environment.”

Regarding the changes in shape of the environment, this was discussed in the section of the paper that reads “As seen in Figure 12, because all but one of the place cells was exterior when the simulated animal was constrained to a narrow passage, the hippocampal place cell memories were no longer arranged in a hexagonal grid. This disruption of the grid array for narrow passages might explain the finding that the grid pattern (of grid cells) is disrupted in the thin corner of a trapezoid (Krupic et al., 2015) and disrupted when a previously open enclosure is converted to a hairpin maze by insertion of additional walls within the enclosure (Derdikman et al., 2009).” This particular section of the paper now appears in the Appendix and Figure 12 is now Appendix Figure 2.

Reviewer #2 (Public Review): 

The manuscript describes a new framework for thinking about the place and grid cell system in the hippocampus and entorhinal cortex in which these cells are fundamentally involved in supporting non-spatial information coding. If this framework were shown to be correct, it could have high impact because it would suggest a completely new way of thinking about the mammalian memory system in which this system is non-spatial. Although this idea is intriguing and thought-provoking, a very significant caveat is that the paper does not provide evidence that specifically supports its framework and rules out the alternate interpretations. Thus, although the work provides interesting new ideas, it leaves the reader with more questions than answers because it does not rule out any earlier ideas. 

Basically, the strongest claim in the paper, that grid cells are inherently non-spatial, cannot be specifically evaluated versus existing frameworks on the basis of the evidence that is shown here. If, for example, the author had provided behavioral experiments showing that human memory encoding/retrieval performance shifts in relation to the predictions of the model following changes in the environment, it would have been potentially exciting because it could potentially support the author's reconceptualization of this system. But in its current form, the paper merely shows that a new type of model is capable of explaining the existing findings. There is not adequate data or results to show that the new model is a significantly better fit to the data compared to earlier models, which limits the impact of the work. In fact, there are some key data points in which the earlier models seem to better fit the data.  

Overall, I would be more convinced that the findings from the paper are impactful if the author showed specific animal memory behavioral results that were only supported by their memory model but not by a purely spatial model. Perhaps the author could run new experiments to show that there are specific patterns of human or animal behavior that are only explained by their memory model and not by earlier models. But in its current form, I cannot rule out the existing frameworks and I believe some of the claims in this regard are overstated. 

As previously detailed in Box 1 and as explained in the text in several places, the model provides an explanation of several findings that remain unexplained by other theories (see “Results Uniquely Explained by the Memory Model”). But more generally this is a good point, and the initial draft failed to fully articulate why a researcher might choose this model to guide future empirical investigations. A new section in the introduction that deals with these issues, titled ‘Why Model the Rodent Navigation Literature with a Memory Model?’ That section reads:

“Spatial navigation is inherently a memory problem – learning the spatial arrangement of a new enclosure requires memory for the conjunction of what and where. This has long been realized and in the introduction to ‘Hippocampus as a Cognitive Map’, O’Keefe and Nadel (1978) wrote “We shall argue that the hippocampus is the core of a neural memory system providing an objective spatial framework within which the items and events of an organism's experience are located and interrelated” (emphasis added). Furthermore, in the last chapter of their book, they extended cognitive map theory to human memory for non-spatial characteristics. However, in the decades since the development of cognitive map theory, the rodent spatial navigation and human memory literatures have progressed somewhat independently.

The ideas proposed in this model are an attempt to reunify these literatures by returning to the original claim that spatial navigation is inherently a memory problem. The goal of the current study is to explain the rodent spatial navigation literature using a memory model that has the potential to also explain the human memory literature. In contrast, most grid cell models (Bellmund et al., 2016; Bush et al., 2015; Castro & Aguiar, 2014; Hasselmo, 2009; Mhatre et al., 2012; Solstad et al., 2006; Sorscher et al., 2023; Stepanyuk, 2015; Widloski & Fiete, 2014) are domain specific models of spatial navigation and as such, they do not lend themselves to explanations of human memory. Thus, the reason to prefer this model is parsimony. Rather than needing to develop a theory of memory that is separate from a theory of spatial navigation, it might be possible to address both literatures with a unified account.

This study does not attempt to falsify other theories of grid cells. Instead, this model reaches a radically different interpretation regarding the function of grid cells; an interpretation that emerges from viewing spatial navigation as a memory problem. All other grid cell models assume that an entorhinal grid cell displaying a spatially arranged grid of firing fields serves the function of spatial coding (i.e., spatial grid cells exist to support a spatial metric). In contrast, the proposed memory model of grid cells assumes that the hexagonal tiling reflects the need to keep memories separate from each other to minimize confusion and confabulation – the grid pattern is the byproduct of pattern separation between memories rather than the basis of a spatial code. 

It is now understood that grid-like firing fields can occur for non-spatial twodimensional spaces. For instance, human entorhinal cortex exhibits grid-like responses to video morph trajectories in a two-dimensional bird neck-length versus bird leg-length space (Constantinescu et al., 2016). As a general theory of learning and memory, the proposed memory model of grid cells is easily extended to explain these results (e.g., relabeling the border cell inputs in the model as neck-length and leg-length inputs). However, there are other grid cell models that can explain both spatial grid cells as well as non-spatial grid-like responses (Mok & Love, 2019; Rodríguez-Domínguez & Caplan, 2019; Stachenfeld et al., 2017; Wei et al., 2015). Similar to this memory model of grid cells, these models are also positioned to explain both the rodent spatial navigation and human memory literatures. Nevertheless, there is a key difference between this model and other grid cell models that generalize to non-spatial representations. Specifically, these other models assume that grid cells exhibiting spatial receptive fields serve the function of identifying positions in the environment (i.e., their function is spatial). As such, these models do not explain why most of the input to rodent hippocampus appears to be spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). This memory model of grid cells provides an answer to the apparent paucity of nonspatial cell types in rodent MTL by proposing that grid cells with spatial receptive fields have been misclassified as spatial (they are what cells rather than where cells) and that place cells are fundamentally memory cells that conjoin what and where.”

- The paper does not fully take into account all the findings regarding grid cells, some of which very clearly show spatial processing in this system. For example, findings on grid-bydirection cells (e.g., Sargolini et al. 2006) would seem to suggest that the entorhinal grid system is very specifically spatial and related to path integration. Why would grid-bydirection cells be present and intertwined with grid cells in the author's memory-related reconceptualization? It seems to me that the existence of grid-by-direction cells is strong evidence that at least part of this network is specifically spatial.

Head by direction grid cells were a key part of the reported results. These grid cells naturally arise in the model as the animal forms memories (aka, hippocampal place cells) that conjoin location (as defined by border cells), head direction at the time of memory formation, and one or more non-spatial properties found at that location. In this revision, I have attempted to better explain how including head direction in hippocampal memories naturally gives rise to these cell types. The introduction to the head direction module simulations now reads:

“According to this memory model of spatial navigation, place cells are the conjunction of location, as defined by border cells, and one or more properties that are remembered to exist at that location. Such memories could, for instance, allow an animal to remember the location of a food cache (Payne et al., 2021). The next set of simulations investigates behavior of the model when one of the to-be-remembered properties is head direction at the time when the memory was formed (e.g., the direction of a pathway leading to a food cache). Indicating that head direction is an important part of place cell representations, early work on place cells in mazes found strong sensitivity to head direction, such that the place field is found in one direction of travel but not the other (McNaughton et al., 1983; Muller et al., 1994). Place cells can exhibit a less extreme version of head direction sensitivity in open field recordings (Rubin et al., 2014), but the nature of the sensitivity is more complicated, depending on location of the animal relative to the place field center (Jercog et al., 2019).

It is possible that some place cell memories do not receive head direction input, as was the case for the simulations reported in Figures 6/7 – in those simulations, place cells were entirely insensitive to head direction, owing to a lack of input from head direction cells. However, removal of head direction input to hippocampus affects place cell responses (Calton et al., 2003) and grid cell responses (Winter et al., 2015), suggesting that head direction is a key component of the circuit. Furthermore, if place cells represent episodic memories, it seems natural that they should include head direction (i.e., viewpoint at the time of memory formation).

In the simulations reported next, head direction is simply another property that is conjoined in a hippocampal place cell memory. In this case, a head direction cell should become a head direction conjunctive grid cell (i.e., a grid cell, but only when the animal is heading in a particular direction), owing to memory feedback from the hexagonal array of hippocampal place cell memories. When including head direction, the real-world dimensions of variation are across three dimensions (X, Y, and head direction) rather than two, and consolidation will cause the place cells to arrange in a three-dimensional volume. The simulation reported below demonstrates that this situation provides a “grid module”.”

- I am also concerned that the paper does not do enough to address findings regarding how the elliptical shape of grid fields shifts when boundaries of an environment compress in one direction or change shape/angles (Lever et al., & Krupic et al). Those studies show compression in grid fields based on boundary position, and I don't see how the authors' model would explain these findings.  

This finding was covered in the original submission: “For instance, perhaps one egocentric/allocentric pair of mEC grid modules is based on head direction (viewpoint) in remembered positions relative to the enclosure borders whereas a different egocentric/allocentric pair is based on head direction in remembered positions relative to landmarks exterior to the enclosure. This might explain why a deformation of the enclosure (moving in one of the walls to form a rectangle rather than a square) caused some of the grid modules but not others to undergo a deformation of the grid pattern in response to the deformation of the enclosure wall (see also Barry et al., 2007). More specifically, if there is one set of non-orthogonal dimensions for enclosure borders and the movement of one wall is too modest as to cause avoid global remapping, this would deform the grid modules based the enclosure border cells. At the same time, if other grid modules are based on exterior properties (e.g., perhaps border cells in relation to the experimental room rather than the enclosure), then those grid modules would be unperturbed by moving the enclosure wall.”

I apologize for being unclear in describing how the model might explain this result. The paragraph has been rewritten and now reads:

“Consider the possibility that one mEC grid modules is based on head direction (viewpoint) in remembered positions relative to the enclosure borders (e.g., learning the properties of the enclosure, such as the metal surface) while a different grid module is based on head direction in remembered positions relative to landmarks exterior to the enclosure (e.g., learning the properties of the experimental room, such as the sound of electronics that the animal is subject to at all locations). This might explain why a deformation of the enclosure (moving one of the walls to form a rectangle rather than a square) caused some of the grid modules but not others to undergo a deformation of the grid pattern in response to the deformation of the enclosure wall (see also Barry et al., 2007). More specifically, suppose that the movement of one wall is modest and after moving the wall, the animal views the enclosure as being the same enclosure, albeit slightly modified (e.g., when a home is partially renovated, it is still considered the same home). In this case, the set of non-orthogonal dimensions associated with enclosure borders would still be associated with the now-changed borders and any memories in reference to this border-determined space would adjust their positions accordingly in real-world coordinates (i.e., the place cells would subtly shift their positions owing to this deformation of the borders, producing a corresponding deformation of the grid). At the same time, there may be other sets of memories that are in relation to dimensions exterior to the enclosure. Because these exterior properties are unchanged, any place cells and grid cells associated with the exterior-oriented memories would be unchanged by moving the enclosure wall.”

- Are findings regarding speed modulation of grid cells problematic for the paper's memory results? 

- A further issue is that the paper does not seem to adequately address developmental findings related to the timecourses of the emergence of different cell types. In their simulation, researchers demonstrate the immediate emergence of grid fields in a novel environment, while noting that the stabilization of place cell positions takes time. However, these simulation findings contradict previous empirical developmental studies (Langston et al., 2010). Those studies showed that head direction cells show the earliest development of spatial response, followed by the appearance of place cells at a similar developmental stage. In contrast, grid cells emerge later in this developmental sequence. The gradual improvement in spatial stability in firing patterns likely plays a crucial role in the developmental trajectory of grid cells. Contrary to the model simulation, grid cells emerge later than place cells and head direction cells, yet they also hold significance in spatial mapping. 

- The model simulations suggest that certain grid patterns are acquired more gradually than others. For instance, egocentric grid cells require the stabilization of place cell memories amidst ongoing consolidation, while allocentric grid cells tend to reflect average place field positions. However, these findings seemingly conflict with empirical studies, particularly those on the conjunctive representation of distance and direction in the earliest grid cells. Previous studies show no significant differences were found in grid cells and grid cells with directional correlates across these age groups, relative to adults (Wills et al., 2012). This indicates that the combined representation of distance and direction in single mEC cells is present from the earliest ages at which grid cells emerge. 

These are good points and they have been addressed in a new section of the introduction titled ‘The Scope of the Proposed Model’. That section reads:

“The reported simulations explain why most mEC cell types in the rodent literature appear to be spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). Assuming that rodents can form non-spatial memories, rodent hippocampus must receive non-spatial input from entorhinal cortex. These simulations suggest that characterization of the rodent mEC cortex as primarily spatial might be incorrect if most grid cells (except perhaps head direction conjunctive grid cells) have been mischaracterized as spatial. Other literatures with other species find non-spatial representations in MTL (Gulli et al., 2020; Quiroga et al., 2005; Wixted et al., 2014) and non-spatial hippocampal memory encoding has been found in rodents (Liu et al., 2012; McEchron & Disterhoft, 1999). The proposed memory model is compatible with these results – the ideas contained in this model could be applied to nonspatial memory representations. However, surveys of cell types in rodent entorhinal cortex seem to indicate that most cells are spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). How can the rodent hippocampus encode nonspatial memories if most of its input is spatial? The goal of the reported simulations is to explain the apparent paucity of non-spatial cells in rodent entorhinal cortex by proposing that grid cells have been misclassified as spatial (see also Luo et al., 2024).

Given the simplicity of the proposed model, there are important findings that the model cannot address -- it is not that the model makes the wrong predictions but rather that it makes no predictions. The role of running speed (Kraus et al., 2015) is one such variable for which the model makes no predictions. Similarly, because the model is a rate-coded model rather than a model of oscillating spiking neurons, it makes no predictions regarding theta oscillations (Buzsáki & Moser, 2013). The model is an account of learning and memory for an adult animal, and it makes no predictions regarding the developmental (Langston et al., 2010; Muessig et al., 2015; Wills et al., 2012) or evolutionary (Rodrıguez et al., 2002) time course of different cell types. This model contains several purely spatial representations such as border cells, head direction cells, and head direction conjunctive grid cells and it may be that these purely spatial cell types emerged first, followed by the evolution and/or development of non-spatial cell types. However, this does not invalidate the model. Instead, this is a model for an adult animal that has both episodic memory capabilities and spatial navigation capabilities, irrespective of the order in which these capabilities emerged.

This model has the potential to explain context effects in memory (Godden & Baddeley, 1975; Gulli et al., 2020; Howard et al., 2005). According to this model, different grid cells represent different non-spatial characteristics and place cells represent the combination of these “context” factors and location. In the simulation, just one grid cell is simulated but the same results would emerge when simulating hundreds of different non-spatial inputs provided that all of the simulated non-spatial inputs exist throughout the recording session. However, there is evidence that hippocampus can explicitly represent the passage of time (Eichenbaum, 2014), and time is assuredly an important factor in defining episodic memory (Bright et al., 2020). Thus, although the current model addresses unique combinations of what and where, it is left to future work to incorporate representations of when in the memory model.”

Reviewer #3 (Public Review): 

A crucial assumption of the model is that the content of experience must be constant in space. It's difficult to imagine a real-world example that satisfies this assumption. Odors and sounds are used as examples. While they are often more spatially diffuse than an objects on the ground, odors and sounds have sources that are readily detectable. Animals can easily navigate to a food source or to a vocalizing conspecific. This assumption is especially problematic because it predicts that all grid cells should become silent when their preferred non-spatial attribute (e.g. a specific odor) is missing. I'm not aware of any experimental data showing that grid cells become silent. On the contrary, grid cells are known to remain active across all contexts that have been tested, including across sleep/wake states. Unlike place cells, grid cells do not seem to turn off. Since grid cells are active in all contexts, their preferred attribute must also be present in all contexts, and therefore they would not convey any information about the specific content of an experience.  

These are good points and in this revision I have attempted to explain that there is a great deal of contextual similarity across all recording sessions. One paragraph in the discussion now reads

“In a typical rodent spatial navigation study, the non-spatial attributes are wellcontrolled, existing at all locations regardless of the enclosure used during testing (hence, a grid cell in one enclosure will be a grid cell in a different enclosure). Because labs adopt standard procedures, the surfaces, odors (e.g., from cleaning), external lighting, time of day, human handler, electronic apparatus, hunger/thirst state, etc. might be the same for all recording sessions. Additionally, the animal is not allowed to interact with other animals during recording and this isolation may be an unusual and highly salient property of all recording sessions. Notably, the animal is always attached to wires during recording. The internal state of the animal (fear, aloneness, the noise of electronics, etc.) is likely similar across all recording situations and attributes of this internal state are likely represented in the hippocampus and entorhinal input to hippocampus. According to this model, hippocampal place cells are “marking” all locations in the enclosure as places where these things tend to happen.”

The proposed novelty of this theory is that other models all assume that grid cells encode space. This isn't quite true of models based on continuous attractor networks, the discussion of which is notably absent. More specifically, these models focus on the importance of intrinsic dynamics within the entorhinal cortex in generating the grid pattern. While this firing pattern is aligned to space during navigation and therefore can be used as a representation of that space, the neural dynamics are preserved even during sleep. Similarly, it is because the grid pattern does not strictly encode physical space that gridlike signals are also observed in relation to other two-dimensional continuous variables. 

These models were briefly discussed in the general discussion section and in this revision they are further discussed in the introduction in a new section, titled ‘Why Model the Rodent Navigation Literature with a Memory Model?’ That section reads:

“Spatial navigation is inherently a memory problem – learning the spatial arrangement of a new enclosure requires memory for the conjunction of what and where. This has long been realized and in the introduction to ‘Hippocampus as a Cognitive Map’, O’Keefe and Nadel (1978) wrote “We shall argue that the hippocampus is the core of a neural memory system providing an objective spatial framework within which the items and events of an organism's experience are located and interrelated” (emphasis added). Furthermore, in the last chapter of their book, they extended cognitive map theory to human memory for non-spatial characteristics. However, in the decades since the development of cognitive map theory, the rodent spatial navigation and human memory literatures have progressed somewhat independently.

The ideas proposed in this model are an attempt to reunify these literatures by returning to the original claim that spatial navigation is inherently a memory problem. The goal of the current study is to explain the rodent spatial navigation literature using a memory model that has the potential to also explain the human memory literature. In contrast, most grid cell models (Bellmund et al., 2016; Bush et al., 2015; Castro & Aguiar, 2014; Hasselmo, 2009; Mhatre et al., 2012; Solstad et al., 2006; Sorscher et al., 2023; Stepanyuk, 2015; Widloski & Fiete, 2014) are domain specific models of spatial navigation and as such, they do not lend themselves to explanations of human memory. Thus, the reason to prefer this model is parsimony. Rather than needing to develop a theory of memory that is separate from a theory of spatial navigation, it might be possible to address both literatures with a unified account.

This study does not attempt to falsify other theories of grid cells. Instead, this model reaches a radically different interpretation regarding the function of grid cells; an interpretation that emerges from viewing spatial navigation as a memory problem. All other grid cell models assume that an entorhinal grid cell displaying a spatially arranged grid of firing fields serves the function of spatial coding (i.e., spatial grid cells exist to support a spatial metric). In contrast, the proposed memory model of grid cells assumes that the hexagonal tiling reflects the need to keep memories separate from each other to minimize confusion and confabulation – the grid pattern is the byproduct of pattern separation between memories rather than the basis of a spatial code. 

It is now understood that grid-like firing fields can occur for non-spatial two dimensional spaces. For instance, human entorhinal cortex exhibits grid-like responses to video morph trajectories in a two-dimensional bird neck-length versus bird leg-length space (Constantinescu et al., 2016). As a general theory of learning and memory, the proposed memory model of grid cells is easily extended to explain these results (e.g., relabeling the border cell inputs in the model as neck-length and leg-length inputs). However, there are other grid cell models that can explain both spatial grid cells as well as non-spatial grid-like responses (Mok & Love, 2019; Rodríguez-Domínguez & Caplan, 2019; Stachenfeld et al., 2017; Wei et al., 2015). Similar to this memory model of grid cells, these models are also positioned to explain both the rodent spatial navigation and human memory literatures. Nevertheless, there is a key difference between this model and other grid cell models that generalize to non-spatial representations. Specifically, these other models assume that grid cells exhibiting spatial receptive fields serve the function of identifying positions in the environment (i.e., their function is spatial). As such, these models do not explain why most of the input to rodent hippocampus appears to be spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). This memory model of grid cells provides an answer to the apparent paucity of nonspatial cell types in rodent MTL by proposing that grid cells with spatial receptive fields have been misclassified as spatial (they are what cells rather than where cells) and that place cells are fundamentally memory cells that conjoin what and where.”

The use of border cells or boundary vector cells as the main (or only) source of spatial information in the hippocampus is not well supported by experimental data. Border cells in the entorhinal cortex are not active in the center of an environment. Boundary-vector cells can fire farther away from the walls but are not found in the entorhinal cortex. They are located in the subiculum, a major output of the hippocampus. While the entorhinalhippocampal circuit is a loop, the route from boundary-vector cells to place cells is much less clear than from grid cells. Moreover, both border cells and boundary-vector cells (which are conflated in this paper) comprise a small population of neurons compared to grid cells.

AUTHOR RESPONSE: The model can be built without assuming between-border cells (early simulations with the model did not make this assumption). Regarding this issue, the text reads “Unlike the BVC model, the boundary cell representation is sparsely populated using a basis set of three cells for each of the three dimensions (i.e., 9 cells in total), such that for each of the three non-orthogonal orientations, one cell captures one border, another the opposite border, and the third cell captures positions between the opposing borders (Solstad et al., 2008). However, this is not a core assumption, and it is possible to configure the model with border cell configurations that contain two opponent border cells per dimension, without needing to assume that any cells prefer positions between the borders (with the current parameters, the model predicts there will be two border cells for each between-border cell). Similarly, it is possible to configure the model with more than 3 cells for each dimension (i.e., multiple cells representing positions between the borders).” The Solstad paper found a few cells that responded in positions between borders, but perhaps not as many as 1 out of 3 cells, such as this particular model simulation predicts. If the paucity of between-border cells is a crucial data point, the model can be reconfigured with opponent-border cells without any between border cells. The reason that 3 border cells were used rather than 2 opponent border cells was for simplicity. Because 3 head direction cells were used to capture the face-centered cubic packing of memories, the simulation also used 3 border cells per dimensions to allow a common linear sum metric when conjoining dimensions to form memories. If the border dimensions used 2 cells while head direction used 3 cells, a dimensional weighting scheme would be needed to allow this mixing of “apples and oranges” in terms of distances in the 3D space that includes head direction.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

Specific questions/clarifications:  

(1) Assumption of population-based vs single unit link to biological cells: At the start, the author assumes that each unit here can be associated with a population: "the simulated activation values can be thought of as proportional to the average firing rate of an ensemble of neurons with similar inputs and outputs (O'Reilly & Munakata, 2000)." But is a 'grid cell' found here a single cell or an average of many cells? Does this mean the model assumes many cells that have different fields that are averaged, which become a grid-like unit in the model? But in biology, these are single cells? Or does it mean a grid response is an average of the place cell inputs? 

I apologize for being unclear about this. The grid cells in the model are equivalent to real single cells except that the simulation uses a ratecoded cell rather than a spiking cell. The averaging that was mentioned in the paper is across identically behaving spiking cells rather than across cells with different grid field arrangements. To better explain this, I have added the following text:

“For instance, consider a set of several thousand spiking grid cells that are identical in terms of their firing fields. At any moment, some of these identically-behaving cells will produce an action potential while others do not (i.e., the cells are not perfectly synchronized), but a snapshot of their behavior can be extracted by calculating average firing rate across the ensemble. The simulated cells in the model represent this average firing rate of identically-behaving ensembles of spiking neurons.” 

This is a mathematical short-cut to avoid simulating many spiking neurons. Because this model was compared to real spike rate maps, this real-valued average firing rate is down-sampled to produce spikes by finding the locations that produced the top 5% of real-valued activation values across the simulation.

(2) It is not clear to me why they are circular border cells/basis sets.  

In the initial submission, there was a brief paragraph describing this assumption. In this revision, that paragraph has been expanded and modified for greater clarity. It now reads:

“Because head direction is necessarily a circular dimension, it was assumed that all dimensions are circular (a circular dimension is approximately linear for nearby locations). This assumption of circular dimensions was made to keep the model relatively simple, making it easier to combine dimensions and allowing application of the same processes for all dimensions. For instance, the model requires a weight normalization process to ensure that the pattern of weights for each dimension corresponds to a possible input value along that dimension. However, the normalization for a linear dimension is necessarily different than for a circular dimension. Because the neural tuning functions were assumed to be sine waves, normalization requires that the sum of squared weights add up to a constant value. For a linear dimension, this sum of squares rule only applies to the subset of cells that are relevant to a particular value along the dimension whereas for a circular dimension, this sum of squares rule is over the entire set of cells that represent the dimension (i.e., weight normalization is easier to implement with circular dimensions). Although all dimensions were assumed to be circular for reasons of mathematical convenience and parsimony, circular dimensions may relate to the finding that human observers have difficultly re-orienting themselves in a room depending on the degree of rotational symmetry of the room (Kelly et al., 2008). In addition, this simplifying assumption allows the model to capture the finding that the population of grid cells lies on a torus (Gardner et al., 2022), although I note that the model was developed before this result was known.”

(3) Why is it 3 components? I realise that the number doesn't matter too much, but I believe more is better, so is it just for simplicity? 

In this revision, additional text has been added to explain this assumption: “To keep the model simple, the same number of cells was assumed for all dimensions and all dimensions were assumed to be circular (head direction is necessarily circular and because one dimension needed to be circular, all dimensions were assumed to be circular). Three cells per dimensions was chosen because this provides a sparse population code of each dimension, with few border cells responding between borders, with few border cells responding between borders, while allowing three separate phases of grid cells within a grid cell module (in the model, a grid cell module arises from combination of a third dimension, such as head direction, with the real-world X/Y dimensions defined by border cells).”

As a reminder, the text explaining the sparse coding of border cells reads: “However, this is not a core assumption, and it is possible to configure the model with border cell configurations that contain two opponent border cells per dimension, without needing to assume that any cells prefer positions between the borders (with the current parameters, the model predicts there will be two border cells for each between-border cell). Similarly, it is possible to configure the model with more than 3 cells for each dimension (i.e., multiple cells representing positions between the borders).”

The model can work with just two opponent cells or with more than three cells per basis set. In different simulations, I have explored these possibilities. Three was chosen because it is a convenient way to highlight the face-centered cubic packing of memories that tends to occur (FCP produces 3 alternating layers of hexagonally arranged firing fields). Thus, each of the three head direction cells captures a different layer of the FCP arrangement. A more realistic simulation might combine 6 different head direction cells tiling the head direction dimension with opponent border cells (just 2 cells for each border dimensions). Such a combination would produce responses at borders, but no responses between borders and, at the same time, the head direction cells would still reveal the FCP arrangement. However, it is not easy to find the right parameters for such a mix-and-match simulation in which different dimensions have different numbers of tuning functions (e.g., some dimensions having 2 cells while others have 3 or 6 and some dimensions being linear while others are circular). When all of the dimensions are of the same type, the simple sum that arises from multiplying the input by the weight values gives rise to Euclidean distance (see Figure 3B). With a mix-and-match model of different dimension-types, it should be possible to adjust the sum to nevertheless produce a monotonic function with Euclidean distance although I leave this to future work. To keep things simple, I assumed that all dimensions are of the same type (circular, with 3 cells per dimension).  

(4) Confusion due to the border cells/box was unclear to me. "If the period of the circular border cells was the same as the width of the box, then a memory pushed outside the box on one side would appear on the opposite side of the box, in which case the partial grid field on one side should match up with its remainder on the other side. This would entail complete confusion between opposite sides of the box, and the representation of the box would be a torus (donut-shaped) rather than a flat two-dimensional surface. To reduce confusion ..." Is this confusion of the model? Of the animal?  

This would be confusion of the animal (e.g., a memory field overlapping with one border would also appear at the opposite border in the corresponding location). At one point in model development, I made the assumption that one side of the box wraps to the other side, and I asked Trygve Solstad to run some analyses of real data to see if cells actually wrap around in this manner. He did not find any evidence of this, and so I decided to include outsidethe-box representational area which, as it turned out, allowed the model to capture other behaviors as detailed in the paper.

This section of the paper now reads:

“The cosine tuning curves of the simulated border cells represent distance from the border on both sides of the border (i.e., firing rate increases as the animal approaches the border from either the inside or the outside of the enclosure). Experimental procedures do not allow the animal to experience locations immediately outside the enclosure, but these locations remain an important part of the hypothetic representation, particularly when considering the modification of memories through consolidation (i.e., a memory created inside the enclosure might be moved to a location outside the enclosure). This symmetry about the border cell’s preferred location is needed to maintain an unbiased representation, with a constant sum of squares for the border cell inputs (see methods section). Rather than using linear dimensions, all dimensions were assumed to be circular to keep the model relatively simple. This assumption was made because head direction is necessarily a circular dimension and by having all dimensions be circular, it is easy to combine dimensions in a consistent manner to produce multidimensional hippocampal place cell memories. Thus, the border cells define a torus (or more accurately a three-torus) of possible locations. This provides a hypothetical space of locations that could be represented.

In light of the assumption to represent border cells with a circular dimension, when a memory is pushed outside the East wall of the enclosure, it would necessarily be moved to the West wall of the enclosure if the period of the circular dimension was equal to the width of the enclosure. If this were true, then the partial grid field on one side of the enclosure would match up with its remainder on the other side. Such a situation would cause the animal to become completely confused regarding opposite sides of the enclosure (a location on the West wall would be indistinguishable from the corresponding location on the East wall). To reduce confusion between opposite sides of the enclosure, the width of the enclosure in which the animal navigated (Figure 5) was assumed to be half as wide as the full period of the border cells. In other words, although the space of possible representations was a three-torus, it was assumed that the real-world twodimensional enclosure encompassed a section of the torus (e.g., a square piece of tape stuck onto the surface of a donut). The torus is better thought of as “playing field” in which different sizes and shapes of enclosure can be represented (i.e., different sizes and shapes of tape placed on the donut). Furthermore, this assumption provides representational space that is outside the box without such locations wrapping around to the opposite side of the box.”

(5) Figure 3 - This result seems to be related to whether you use Euclidean or city-block distance. If you use Euclidean distances in two dimensions wouldn't this work out fine?  

Euclidean distance was the metric used in the analysis of the two-dimensional simulation, but this did not work out. To make this clear, I have changed the label on the x-axes to read “Euclidean distance” for both the two- and three-dimensional simulations. The two-dimensional simulation produced city block behavior rather than Euclidean behavior because memory retrieval is the sum of the two dimensions, as is standard in neural networks, rather than the Euclidian distance formula, which would require that memory retrieval be the square root of the sum of squares of the two dimensions. One way to address this problem with the two-dimensional simulation would be to use a specific Euclidean-mimicking activation function rather than a simple sum of dimensions. The very first model I developed used such an activation function as applied to opponent border cells with just two dimensions (so 4 cells in total – left/right and top/down). This produced Euclidean behavior, but the activation function was implausible and did not generalize to simulations that also included head direction. In contrast, with three non-orthogonal dimensions, the simple sum of dimensions is approximately Euclidean.

(6) Final sentence of the Discussion: "However, unlike the present model, these models still assume that entorhinal grid cells represent space rather than a non-spatial attribute." I am not sure if the authors of the cited papers will agree with this. They consider the spatial cases, but most argue they can treat non-spatial features as well. What the author might mean is that they assume non-spatial features are in some metric space that, in a way, is spatial. However, I am not sure if the author would argue that non-spatial features cannot be encoded metrically (e.g., Euclidean distance based on the similarity of odours). 

In this section, when referring to “entorhinal grid cells” I was specifically referring to traditional grid cells in a rodent spatial navigation experiment. I did not mean to imply that these other theories cannot explain nonspatial grid fields, such as in the two-dimensional bird space grid cells found with humans. The way in which the proposed memory model and these other models differ is in terms of what they assume regarding the function of grid cells that exhibit spatial grid fields. In this revision, I have changed this text to read:

“These models can capture some of the grid cell results presented in the current simulations, including extension to non-spatial grid-like responses (e.g., grid field that cover a two-dimensional neck/leg length bird space). Furthermore, these models may be able to explain memory phenomena similar to the model proposed in this study. However, unlike the proposed model, these models assume that the function of entorhinal grid cells that exhibit spatial X/Y grid fields during navigation is to represent space. In contrast, the memory model proposed in this study assume that the function of spatial X/Y grid cells is to represent a non-spatial attribute; the only reason they exhibit a spatial X/Y grid is because memories of that non-spatial attribute are arranged in a hexagonal grid owing to the uncluttered/unvarying nature of the enclosure. Thus, these model do not explain why most of the input to rodent hippocampus appears to be spatial (Boccara et al., 2010b; Diehl et al., 2017; Grieves & Jeffery, 2017) whereas the proposed model can explain this situation as reflecting the miss-classification of grid cells with a spatial arrangement as providing spatial input to hippocampus.”

(7) It would be interesting to see videos/gifs of the model learning, and an idea of how many steps of trials it takes (is it capturing real-time rodent cell firing whilst foraging, or is it more abstracted, taking more trials). 

The short answer is “yes”, the model is capturing real-time rodent cell firing while foraging. This is particularly true when simulating place cell memories in the absence of head direction information, as was shown in a video provided in the initial submission in relation to Figure 4. In this revision, I have provided a second video of learning when simulating place cell memories that include head direction. This second video is in relation to the results reported in Figure 9. This shows that even when learning a three-dimensional real-world space (X, Y, and head direction), the model rapidly produces an on-average hexagonal arrangement of place cells memories owing to the slight tendency of the place cell memories to linger in some locations as compared to others during consolidation. More specifically, they are more likely to linger in the locations that are the intersections of the peaks and/or troughs of the border cells and it is this tendency that supports the immediate appearance of grid cells. However, because the place cell memories are still shifting, head direction conjunctive grid cells are slower to emerge (the head direction conjunctive grid cells require stabilization of the place cells). The video then speeds up the learning process to so how place cells eventually stabilize after sufficient learning of the borders of the enclosure from different head/view directions.

(8) One question is whether all the results have to be presented in the main text. It was difficult to see which key predictions fit the data and do so better than a spatial/navigation account. 

Thank you for this suggestion. To make the paper more readable and easier for different readers with different interests to choose different aspects of the results to read, the second half of the results have been put in an appendix. More specifically, the second half of the results concerned place cells rather than grid cells. Thus, in this revision, the main text concerns grid cell results and the appendix concerns place cell results.

Reviewer #3 (Recommendations For The Authors):  

The title could usefully be shortened to focus on the main argument that observed firing patterns could be consistent with mapping memories instead of space. It's a stretch to argue that memory is the primary role when no such data is presented (i.e., there is no comparison of competing models). 

This is a good point (I do not present evidence that conclusively indicates the function of MTL). This original title was chosen to make clear how this account is a radical departure from other accounts of grid cells. The revised title highlights that: 1) a memory model can also explain rodent single cell recording data during navigation; and 2) grid cell may not be non-spatial. The revised title is: “A Memory Model of Rodent Spatial Navigation: Place Cells are Memories Arranged in a Grid and Grid Cells are Non-spatial”

When arguing that the main role of the hippocampus is memory, I strongly suggest engaging with the work of people like Howard Eichenbaum who spent the better part of their career arguing the same (e.g. DOI:10.1152/jn.00005.2017.)  

Thank you for pointing out this important oversight. Early in introduction, I now write: “The proposal that hippocampus represents the multimodal conjunctions that define an episode is not new (Marr et al., 1991; Sutherland & Rudy, 1989) and neither is the proposal that hippocampal memory supports spatial/navigation ability (Eichenbaum, 2017). This view of the hippocampus is consistent with “feature in place” results (O’Keefe & Krupic, 2021) in which hippocampal cells respond to the conjunction of a non-spatial attribute affixed to a specific location, rather than responding more generically to any instance of a non-spatial attribute. In other words, the what/where conjunction is unique. Furthermore, the uniqueness of the what/where conjunction may be the fundamental building block of spatial memory and navigation. In reviewing the hippocampal literature, Howard Eichenbaum (2017) concludes that ‘the hippocampal system is not dedicated to spatial cognition and navigation, but organizes experiences in memory, for which spatial mapping and navigation are both a metaphor for and a prominent application of relational memory organization.’”

With a focus on episodic memory, there should be a mention of the temporal component of memory. While it may rightfully be beyond the scope of this model, it's confusing to omit time completely from the discussion. 

This issue and several others are now addressed in a new section in the introduction titled ‘The Scope of the Proposed Model’. That section reads:

“The reported simulations explain why most mEC cell types in the rodent literature appear to be spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). Assuming that rodents can form non-spatial memories, rodent hippocampus must receive non-spatial input from entorhinal cortex. These simulations suggest that characterization of the rodent mEC cortex as primarily spatial might be incorrect if most grid cells (except perhaps head direction conjunctive grid cells) have been mischaracterized as spatial. Other literatures with other species find non-spatial representations in MTL (Gulli et al., 2020; Quiroga et al., 2005; Wixted et al., 2014) and non-spatial hippocampal memory encoding has been found in rodents (Liu et al., 2012; McEchron & Disterhoft, 1999). The proposed memory model is compatible with these results – the ideas contained in this model could be applied to nonspatial memory representations. However, surveys of cell types in rodent entorhinal cortex seem to indicate that most cells are spatial (Boccara et al., 2010; Diehl et al., 2017; Grieves & Jeffery, 2017). How can the rodent hippocampus encode nonspatial memories if most of its input is spatial? The goal of the reported simulations is to explain the apparent paucity of non-spatial cells in rodent entorhinal cortex by proposing that grid cells have been misclassified as spatial (see also Luo et al., 2024).

Given the simplicity of the proposed model, there are important findings that the model cannot address -- it is not that the model makes the wrong predictions but rather that it makes no predictions. The role of running speed (Kraus et al., 2015) is one such variable for which the model makes no predictions. Similarly, because the model is a rate-coded model rather than a model of oscillating spiking neurons, it makes no predictions regarding theta oscillations (Buzsáki & Moser, 2013). The model is an account of learning and memory for an adult animal, and it makes no predictions regarding the developmental (Langston et al., 2010; Muessig et al., 2015; Wills et al., 2012) or evolutionary (Rodrıguez et al., 2002) time course of different cell types. This model contains several purely spatial representations such as border cells, head direction cells, and head direction conjunctive grid cells and it may be that these purely spatial cell types emerged first, followed by the evolution and/or development of non-spatial cell types. However, this does not invalidate the model. Instead, this is a model for an adult animal that has both episodic memory capabilities and spatial navigation capabilities, irrespective of the order in which these capabilities emerged.

This model has the potential to explain context effects in memory (Godden & Baddeley, 1975; Gulli et al., 2020; Howard et al., 2005). According to this model, different grid cells represent different non-spatial characteristics and place cells represent the combination of these “context” factors and location. In the simulation, just one grid cell is simulated but the same results would emerge when simulating hundreds of different non-spatial inputs provided that all of the simulated non-spatial inputs exist throughout the recording session. However, there is evidence that hippocampus can explicitly represent the passage of time (Eichenbaum, 2014), and time is assuredly an important factor in defining episodic memory (Bright et al., 2020). Thus, although the current model addresses unique combinations of what and where, it is left to future work to incorporate representations of when in the memory model.”

I recommend explaining the motivation of the theory in more detail in the introduction. It reads as "what if it's like this?" It would be helpful to instead highlight the limitations of current theories and argue why this theory is either a better fit for the data or is logically simpler. 

This issue and several others are now addressed in the new section in the introduction titled ‘Why Model the Rodent Navigation Literature with a Memory Model?’, which I quoted above in response to the public reviews.

It's worth considering shortening the results section to include only those that most convincingly support the main claim. The manuscript is quite long and appears to lack focus at times. 

Thank you for this suggestion. To make the paper more readable and easier for different readers with different interests to choose different aspects of the results to read, the second half of the results have been put in an appendix. More specifically, the second half of the results concerned place cells rather than grid cells. Thus, in this revision, the main text concerns grid cell results and the appendix concerns place cell results.

The discussion of path dependence on the formation of the grid pattern is important but only briefly discussed. It may be useful to add simulations testing whether different paths (not random walks) produce distorted grid patterns. 

The short answer is that the path doesn’t affect things in general. The consolidation rule ensures equally spaced memories even if, for instance, one side of the enclosure is explored much more than the other side. As just one example, I have run simulations with a radial arm maze and even though the animal is constrained to only run on the maze arms. The memories still arrange hexagonally as memories become pushed outside the arms. Rather than adding additional simulations to study, I now briefly describe this in the model methods:

“Of note, the ability of the model to produce grid cell responses does not depend on this decision to simulate an animal taking a random walk – the same results emerge if the animal is more systematic in its path. All that matters for producing grid cell responses is that the animal visits all locations and that the animal takes on different head directions for the same location in the case of simulations that also include head direction as an input to hippocampal place cells.”

I struggle to understand in Figure 3 why retrieval strength ought to scale monotonically with Euclidean distance, and why that justifies a more complex model (three non-orthogonal dimensions). 

The introduction to this section now reads: “Animals can plan novel straight line paths to reach a known position and evidence suggests they do so by learning Euclidean representations of space (Cheng & Gallistel, 2014; Normand & Boesch, 2009; Wilkie, 1989). Thus, it was assumed that hippocampal place cells represent positions in Euclidean space (as opposed to non-Euclidean space, such a occurs with a city-block metric).”

p.17 "although the representational space is a torus (or more specifically a three-torus), it is assumed that the real-world two-dimensional surface is only a section of the torus (e.g., a square piece of tape stuck onto the surface of a donut)." I fail to understand how the realworld surface is only a part of the torus. In the existing theoretical and experimental work on toroidal topology of grid cell activity, the torus represents a very small fraction of the real world, and repeating activity on the toroidal manifold is a crucial feature of how it maps 2D space in a regular manner. Why then here do you want the torus to be larger than the realworld? 

This section has been rewritten to better explain these assumptions. The relevant paragraphs now read:

“The cosine tuning curves of the simulated border cells represent distance from the border on both sides of the border (i.e., firing rate increases as the animal approaches the border from either the inside or the outside of the enclosure). Experimental procedures do not allow the animal to experience locations immediately outside the enclosure, but these locations remain an important part of the hypothetic representation, particularly when considering the modification of memories through consolidation (i.e., a memory created inside the enclosure might be moved to a location outside the enclosure). This symmetry about the border cell’s preferred location is needed to maintain an unbiased representation, with a constant sum of squares for the border cell inputs (see methods section). Rather than using linear dimensions, all dimensions were assumed to be circular to keep the model relatively simple. This assumption was made because head direction is necessarily a circular dimension and by having all dimensions be circular, it is easy to combine dimensions in a consistent manner to produce multidimensional hippocampal place cell memories. Thus, the border cells define a torus (or more accurately a three-torus) of possible locations. This provides a hypothetical space of locations that could be represented.

In light of the assumption to represent border cells with a circular dimension, when a memory is pushed outside the East wall of the enclosure, it would necessarily be moved to the West wall of the enclosure if the period of the circular dimension was equal to the width of the enclosure. If this were true, then the partial grid field on one side of the enclosure would match up with its remainder on the other side. Such a situation would cause the animal to become completely confused regarding opposite sides of the enclosure (a location on the West wall would be indistinguishable from the corresponding location on the East wall). To reduce confusion between opposite sides of the enclosure, the width of the enclosure in which the animal navigated (Figure 5) was assumed to be half as wide as the full period of the border cells. In other words, although the space of possible representations was a three-torus, it was assumed that the real-world twodimensional enclosure encompassed a section of the torus (e.g., a square piece of tape stuck onto the surface of a donut). The torus is better thought of as “playing field” in which different sizes and shapes of enclosure can be represented (i.e., different sizes and shapes of tape placed on the donut). Furthermore, this assumption provides representational space that is outside the box without such locations wrapping around to the opposite side of the box.”

p.28 "More specifically, egocentric grid cells (e.g., head direction conjunctive grid cells) require stabilization of the place cell memories in the face of ongoing consolidation whereas allocentric grid cells reflect on-average place field positions." and p.32 "if place cells represent episodic memories, it seems natural that they should include head direction (an egocentric viewpoint)." But the head direction signal is not egocentric, it is allocentric. I'm unsure whether this is a typo or a potentially more serious conceptual misunderstanding. 

Any reference to egocentric has been removed in this revision. In the initial submission, when I used egocentric, I was referring to memories that depended on the head direction of the animal at the time of memory formation. I was using “egocentric” in relation to whether the memory was related to the animal’s personal bodily experience at the time of memory formation. But I concede that this is confusing since the ego/allo distinction is typically used to differentiate angular directions that are relative to the person (left/right) versus earth (East/West). Instead, throughout the manuscript I now refer to these as view-dependent memories since head direction would entail having a different view of the environment at the time of memory formation. I still refer to the stacking of multiple view-dependent memories on the same X/Y location as being the development of an allocentric representation however, since this can be thought of as one way to learn a cognitive map of the enclosure that is view independent.

p.37 "But if the border cells had changed their alignment with the new enclosure (e.g., if the E border dimension aligned with the North-South borders), then the place cells would have appeared to undergo global remapping as their positions rotated by 90 degrees and the grid pattern would have also rotated." But this would not be interpreted as global remapping by standard analyses of place and grid cell responses. A coherent rotation of firing patterns is not interpreted as remapping. 

This sentence now reads: “But if the border cells had changed their alignment with the new enclosure (e.g., if the E border dimension aligned with the North-South borders), then the place cells would remain in their same positions relative to the now-rotated borders (i.e., no remapping relative to the enclosure) and the corresponding grid cells would also retain their same alignment relative to the enclosure.”

p.37 "this is more accurately described as partial remapping (nearly all place fields were unaffected)." If nearly all place fields were unaffected, this should be interpreted as a stable map. Partial remapping is a mix of stability, rate remapping, and global remapping within a population of place cells. 

This sentence has been removed.

p.40 "The dependence of grid cell responses on memory may help explain why grid cells have been found for bats crawling on a two-dimensional surface (Yartsev et al., 2011), but three-dimensional grid cells have never been observed for flying bats." This is not true. Ginosar et al. (2021) observed 3D grid cells in flying bats.  

Thank you for highlighting this issue. In the initial submission I was using “grid cell” to mean a cell that produced a precise hexagonal grid, which is not the case for the 3D grid cells in bats. In this revision, I now discuss grid cell that produce irregular grid fields, writing:

“According to this model, hexagonally arranged grid cells should be the exception rather than the rule when considering more naturalistic environments. In a more ecologically valid situation, such as with landmarks, varied sounds, food sources, threats, and interactions with conspecifics, there may still be remembered locations were events occurred or remembered properties can be found, but because the non-spatial properties are non-uniform in the environment, the arrangement of memory feedback will be irregular, reflecting the varied nature of the environment. This may explain the finding that even in a situation where there are regular hexagonal grid cells, there are often irregular non-grid cells that have a reliable multi-location firing field, but the arrangement of the firing fields is irregular (Diehl et al., 2017). For instance, even when navigating in an enclosure that has uniform properties as dictated by experimental procedures, they may be other properties that were not well-controlled (e.g., a view of exterior lighting in some locations but not others), and these uncontrolled properties may produce an irregular grid (i.e., because the uncontrolled properties are reliably associated with some locations but not others, hippocampal memory feedback triggers retrieval of those properties in the associations locations).

In this memory model, there are other situations in which an irregular but reliable multi-location grid may occur, even when everything is well controlled. In the reported simulations, when the hippocampal place cells were based on variation in X/Y (as defined by Border cells), nothing else changed as a function of location, and the model rapidly produced a precise hexagonal arrangement of hippocampal place cell memories. When head direction was included (i.e., real-world variation in X, Y, and head direction), the model still produced a hexagonal arrangement as per face centered cubic packing of memories, but this precise arrangement was slower to emerge, with place cells continuing to shift their positions until the borders of the enclosure were sufficiently well learned from multiple viewpoints. If there is realworld variation in four or more dimensions, as is likely the case in a more ecologically valid situation, it will be even harder for place cell memories to settle on a precise regular lattice. Furthermore, in the case of four dimensions, mathematicians studying the “sphere packing problem” recently concluded that densest packing is irregular (Campos et al., 2023). This may explain why the multifield grid cells for freely flying bats have a systematic minimum distance between firing fields, but their arrangement is globally irregular (Ginosar et al., 2021). Assuming that the memories encoded by a bat include not just the three realworld dimensions of variation, but also head direction, the grid will likely be irregular even under optimal conditions of laboratory control.”

Multiple typos are found on page 25, end of paragraph 3: "More specifically, if there is one set of non-orthogonal dimensions for enclosure borders and the movement of one wall is too modest as to cause avoid global remapping, this would deform the grid modules based the enclosure border cells."

As detailed above in the response the public reviews, this paragraph has been rewritten.

eLife Assessment

The authors studied the development of mesentery borders in the rice coral Montipora, a new experimental system, to complement existing data from the sea anemone Nematostella. They make a solid case that in Montipora, there is a sequence of Hox-Gbx genes whose staggered expression in the unsegmented larva is suggestive of their role in subdividing the gastric cavity into repeated units bordered by mesenteries, as in the sea anemone Nematostella. Pharmacological experiments also point to the involvement of the BMP pathway in this process, but additional experiments validating this are necessary. This is a valuable contribution to the field of cnidarian evolution, suggesting that BMP- and "Hox-Gbx code"-dependent patterning of the directive axis was ancestral for Anthozoa.

Reviewer #1 (Public review):

Summary:

The manuscript of He et al. compares the roles of Hox/Gbx genes between the well-established anthozoan model, the burrowing sea anemone Nematostella, and the new scleractinian model Montipora. The authors show staggered expression of Anthox6a.1, Anthox8 and Gbx of the Montipora larva and argue that their BMP-dependent expression is responsible for the segmentation of the endomesoderm, just like they have previously demonstrated in Nematostella (despite some differences in the timing, formation of extra mesenteries, etc). The authors posit that Hox/Gbx-dependent segmentation of the endomesoderm represents an ancestral anthozoan trait. The study addresses a remarkably interesting question, but it has several important shortcomings, which the authors should try to rectify.

Strengths:

The authors introduce a new scleractinian model Montipora and present interesting data on the composition of its compact Hox cluster, its embryonic and larval development, metamorphosis, and segmentation. They also show staggered expression of Gbx, Anthox6a.1, and Anthox8, which is suggestive of their involvement in the partitioning of the gastrodermis of the polyp.

Weaknesses:

He et al. claim that Gbx and Hox genes are responsible for the segmentation of the directive axis in Montipora based on expression patterns of these genes before the onset of segmentation. In the absence of functional analyses, this claim (although likely correct) is not supported. Moreover, the authors do not show that staggered Gbx and Hox gene expression correlates with the position of the segment boundaries.

The authors use two inhibitors of BMP signaling and show that segmentation is lost in the treated animals. However, they do not provide controls, which would show that the effect of the treatment is specific to the loss of BMP function. Moreover, their transcriptomic analyses suggest that the whole BMP signaling system in Montipora is wired completely differently than in Nematostella, but they do not acknowledge and discuss this striking difference. If true, this is a very interesting result, but it requires thorough validation.

Reviewer #2 (Public review):

Building on their detailed dissection of the role of Hox-Gbx genes in endomesodermal segmentation in Nematostella, He and colleagues attempt to understand the evolutionary conservation of this process in anthozoans. In a move that should be congratulated, the authors perform this work in the coral M. capitata, a species that is not well established in the lab. The authors show convincing expression data using both RNAseq and in-situ hybridization and discover the conserved expression of Hox-Gbx genes preceding the segmentation of the enodmesoderm. The authors further attempt to understand whether BMP signalling is playing a role in this process and present data that certainly points to this being the case.

Strength:

The overall quality of the data is very high and the authors show very convincing expression data for the Hox-Gbx genes as well as putting forward a well-thought-out hypothesis for segment evolution.

Weakness:

There are a number of weaknesses in the paper which I believe can be easily addressed:

(1) The authors in many cases claim to have provided functional evidence for the role of Hox-Gbx genes in M. capitata. This is not, however, the case, and although the expression data along with their previous work in Nematostella make their claims very likely I still believe it is necessary to set a higher bar for claiming to understand function. In the abstract, for example, they claim: "These findings demonstrate the existence of a functionally conserved Hox-Gbx module....", something which is not substantiated by the data presented. At the end of the introduction, they say they "systematically interrogate the molecular functions of Hox-Gbx genes" (line 75) which again is not what is presented in the manuscript. Finally, on line 289-291 the authors state: "Taken together, our findings strongly suggest that the heterochronic deployment of a conserved Hox-Gbx module contributes to the divergent adult body plans observed between Edwardsiidae and other anthozoans." I would remove "Strongly" given the absence of functional data. There are also other examples where functional understanding is implied and I would suggest the authors tone this down throughout the manuscript.

(2) On Line 185, the authors state "To determine the function of the Hox-Gbx network in M.capitata segmentation..." when introducing their BMP experiments. I would reword this since they are looking at BMP signalling and do not look directly at Hox-Gbx function.

(3) Although the BMP inhibitor experiments are very interesting I think there is a lack of basic understanding of BMP signalling in this system. Where are the BMP components expressed and how would this match with the hypothesis derived from the data? The authors present some expression patterns in Figure S3 but do not discuss them. In addition, the authors do not show pSMAD staining etc, and do not validate that the inhibitors have an effect on this. I entirely understand the difficulties in doing such experiments in a system like this and would not suggest the authors should now do them but an acknowledgment of this in the discussion would be very welcome.

(4) In both lines 88 and 294 the authors talk about the mechanism of gastrulation. It is not clear to me how they infer this from the figure. If the authors could include some more high-resolution images that show this it would be very helpful and interesting.

(5) On line 169/170 the authors state that two Anthox6 paralogs, McAnthox6 and McAnthox6.1, were specifically expressed at the time of settlement. This is not what I see in the images. I see that McAnthox6 is expressed at 14 hpf more strongly than at the later time point. The authors should clarify this point.

(6) On lines 259-261 the authors state "How temporally and spatially coordinated gene expression can be achieved in this scenario remains an interesting and open question." This seems like a strange statement to include given that they have shown that there is no spatial and temporal collinearity in cnidarians. Surely it is not an open question to ask how it would work if there is none. I would simply remove this.

(7) The authors should cite the sources of information contained in Fig. S2 including how orthology was assigned.

Reviewer #3 (Public review):

Summary:

The authors analyze the expression of a series of genes from the Hox/Gbx family of transcription factors in the settling larva of the rice coral Montipora capitata. The first achievement of the work is developing a protocol for artificial induction of settlement in this species. In the synchronized settlers, the authors were able to follow the sequence of the subdivision of the body cavity to form individual cavities separated by mesenteries. This process has been previously studied in the starlet sea anemone, Nematostella vectensies, and this same group showed that there is a spatio-temporal sequence of expression of genes from the Hox/Gbx group, reminiscent of the sequence of Hox genes in bilaterians. The authors now repeat this analysis with orthologous genes in Montipora, and demonstrate a similar pattern. Finally, they manipulate the BMP pathway and demonstrate that in the absence of BMP signaling, the subdivision of the gastric cavity is abrogated.

Strengths:

The authors have developed a new experimental system for embryological work on cnidarians, where only a handful of systems are available. They identified orthologs of a number of homeobox genes and tested their expression. There is a detailed description of the sequence of the formation of the mesenteries, which differs from that of Namatostella, raising interesting questions about the evolution of mesentery number and the homology of mesenteries.

Weaknesses:

The in situ hybridization experiments describing the expression of the Hox/Gbx genes are not as clean and sharp as could be hoped for. This is evidently a limitation of the system. The discussion of the evolution of mesentery number does not really give new insights into the question (although just raising the discussion is interesting in its own right).

eLife Assessment

This manuscript develops a theoretical model of osmotic pressure adaptation in microbes by osmolyte production and wall synthesis. The prediction of a rapid increase in growth rate on osmotic shock is experimentally validated using fission yeast. By using phenomenological rules rather than detailed molecular mechanisms, the model can potentially apply to a wide range of microbes, providing important insights that would be of interest to the wider community studying the regulation of cell size and mechanics. However, because the core assumptions of the model have not been tested across a range of microbial organisms, the evidence for the universality of the model remains incomplete.

Reviewer #1 (Public review):

Summary:

A theoretical model for microbial osmoresponse was proposed. The model assumes simple phenomenological rules: (i) the change of free water volume in the cell due to osmotic imbalance based on pressure balance, (ii) Osmoregulation that assumes change of the proteome partitioning depending on the osmotic pressure that affects the osmolyte-producing protein production, (iii) The cell-wall synthesis regulation where the change of the turgor pressure to the cell-wall synthesis efficiency to go back to the target turgor pressure, (iv) Effect of Intracellular crowding assuming that the biochemical reactions slow down for more crowding and stops when the protein density (protein mass divided by free water volume) reaches a critical value. The parameter values were found in the literature or obtained by fitting to the experimental data. The authors compare the model behavior with various microorganismcs (E. coli, B. subtils, S. Cerevisiae, S. pombe), and successfully reproduced the overall trend (steady state behavior for many of them, dynamics for S. pombe). In addition, the model predicts non-trivial behavior such as the fast cell growth just after the hypoosmotic shock, which is consistent with experimental observation. The authors further make experimentally testable predictions regarding mutant behavior and transient dynamics.

Strength:

The theory assumes simple mechanistic dependence between core variables without going into specific molecular mechanisms of regulations. The simplicity allows the theory to apply to different organisms by adjusting the time scales with parameters, and the model successfully explains broad classes of observed behaviours. Mathematically, the model provides analytical expressions of the parameter dependences and an understanding of the dynamics through the phase space without being buried in the detail. This theory can serve as a base to discuss the universality and diversity of microbial osmoresponse.

Weakness:

The core part of this model is that everything is coupled with growth physiology, and, as far as I understand, the assumption (iv) (eq. 8) that imposes the global reaction rate dependence on crowding plays a crucial role. I would think this is a strong and interesting assumption. However, the abstract or discussion does not discuss the importance of this assumption. In addition, the paper does not discuss gene regulation explicitly, and some comparison with a molecular mechanism-oriented model may be beneficial to highlight the pros and cons of the current approach.

Reviewer #2 (Public review):

Summary:

In this study, Ye et al. have developed a theoretical model of osmotic pressure adaptation by osmolyte production and wall synthesis.

Strengths:

They validate their model predictions of a rapid increase in growth rate on osmotic shock experimentally using fission yeast. The study has several interesting insights which are of interest to the wider community of cell size and mechanics.

Weaknesses:

Multiple aspects of this manuscript require addressing, in terms of clarity and consistency with previous literature. The specifics are listed as major and minor comments.

Major comments:

(1) The motivation for the work is weak and needs more clarity.

(2) The link between sections is very frequently missing. The authors directly address the problem that they are trying to solve without any motivation in the results section.

(3) The parameters used in the models (symbols) need to be explained better to make the paper more readable.

(4) Throughout the paper, the authors keep switching between organisms that they are modelling. There needs to be some consistency in this aspect where they mention what organism they are trying to model, since some assumptions that they make may not be valid for both yeast as well as bacteria.

(5) The extent of universality of osmoregulation i.e the limitations are not very well highlighted.

(6) Line 198-200: It is not clear in the text what organisms the authors are writing about here. "Experiments suggested that the turgor pressure induce cell-wall synthesis, e.g., through mechanosensors on cell membrane [45, 46], by increasing the pore size of the peptidoglycan network [5], and by accelerating the moving velocity of the cell-wall synthesis machinery [31]". This however is untrue for bacteria as shown by the study (reference 22 is this paper:  E. Rojas, J. A. Theriot, and K. C. Huang, Response of escherichia coli growth rate to osmotic shock, Proceedings of the National Academy of Sciences 111, 7807 (2014).

(7) The time scale of reactions to hyperosmotic shocks does not agree with previous literature (reference 22). Therefore defining which organism you are looking at is important. Hence the statement " Because the timescale of the osmoresponse process, which is around hours (Figure 3B), is much longer than the timescale of the supergrowth phase, which is about 20 minutes, the turgor pressure at the growth rate peak can be well approximated by its immediate value after the shock." from line 447 does not seem to make sense. The authors need to address this.

eLife Assessment

This potentially important study describes the progressive transformation of olfactory information across five different brain regions in the olfactory pathway. While the dataset could be of broad interest to olfactory researchers, the analysis is incomplete and would benefit from a reconsideration of the data sampling window, a more uniform analysis framework, and greater clarity of presentation.

Reviewer #1 (Public review):

In this important study, the authors characterized the transformation of neural representations of olfactory stimuli from the primary sensory cortex to multisensory regions in the medial temporal lobe and investigated how they were affected by non-associative learning. The authors used high-density silicon probe recordings from five different cortical regions while familiar vs. novel odors were presented to a head-restrained mouse. This is a timely study because unlike other sensory systems (e.g., vision), the progressive transformation of olfactory information is still poorly understood. The authors report that both odor identity and experience are encoded by all of these five cortical areas but nonetheless some themes emerge. Single neuron tuning of odor identity is broad in the sensory cortices but becomes narrowly tuned in hippocampal regions. Furthermore, while experience affects neuronal response magnitudes in early sensory cortices, it changes the proportion of active neurons in hippocampal regions. Thus, this study is an important step forward in the ongoing quest to understand how olfactory information is progressively transformed along the olfactory pathway.

The study is well-executed. The direct comparison of neuronal representations from five different brain regions is impressive. Conclusions are based on single neuronal level as well as population level decoding analyses. Among all the reported results, one stands out for being remarkably robust. The authors show that the anterior olfactory nucleus (AON), which receives direct input from the olfactory bulb output neurons, was far superior at decoding odor identity as well as novelty compared to all the other brain regions. This is perhaps surprising because the other primary sensory region - the piriform cortex - has been thought to be the canonical site for representing odor identity. A vast majority of studies have focused on aPCx, but direct comparisons between odor coding in the AON and aPCx are rare. The experimental design of this current study allowed the authors to do so and the AON was found to convincingly outperform aPCx. Although this result goes against the canonical model, it is consistent with a few recent studies including one that predicted this outcome based on anatomical and functional comparisons between the AON-projecting tufted cells vs. the aPCx-projecting mitral cells in the olfactory bulb (Chae, Banerjee et. al. 2022). Future experiments are needed to probe the circuit mechanisms that generate this important difference between the two primary olfactory cortices as well as their potential causal roles in odor identification.

The authors were also interested in how familiarity vs. novelty affects neuronal representation across all these brain regions. One weakness of this study is that neuronal responses were not measured during the process of habituation. Neuronal responses were measured after four days of daily exposure to a few odors (familiar) and then some other novel odors were introduced. This creates a confound because the novel vs. familiar stimuli are different odorants and that itself can lead to drastic differences in evoked neural responses. Although the authors try to rule out this confound by doing a clever decoding and Euclidian distance analysis, an alternate more straightforward strategy would have been to measure neuronal activity for each odorant during the process of habituation.

Reviewer #2 (Public review):

Summary:

This manuscript investigates how olfactory representations are transformed along the cortico-hippocampal pathway in mice during a non-associative learning paradigm involving novel and familiar odors. By recording single-unit activity in several key brain regions (AON, aPCx, LEC, CA1, and SUB), the authors aim to elucidate how stimulus identity and experience are encoded and how these representations change across the pathway.

The study addresses an important question in sensory neuroscience regarding the interplay between sensory processing and signaling novelty/familiarity. It provides insights into how the brain processes and retains sensory experiences, suggesting that the earlier stations in the olfactory pathway, the AON aPCx, play a central role in detecting novelty and encoding odor, while areas deeper into the pathway (LEC, CA1 & Sub) are more sparse and encodes odor identity but not novelty/familiarity. However, there are several concerns related to methodology, data interpretation, and the strength of the conclusions drawn.

Strengths:

The authors combine the use of modern tools to obtain high-density recordings from large populations of neurons at different stages of the olfactory system (although mostly one region at a time) with elegant data analyses to study an important and interesting question.

Weaknesses:

(1) The first and biggest problem I have with this paper is that it is very confusing, and the results seem to be all over the place. In some parts, it seems like the AON and aPCx are more sensitive to novelty; in others, it seems the other way around. I find their metrics confusing and unconvincing. For example, the example cells in Figure 1C show an AON neuron with a very low spontaneous firing rate and a CA1 with a much higher firing rate, but the opposite is true in Figure 2A. So, what are we to make of Figure 2C that shows the difference in firing rates between novel vs. familiar odors measured as a difference in spikes/sec. This seems nearly meaningless. The authors could have used a difference in Z-scored responses to normalize different baseline activity levels. (This is just one example of a problem with the methodology.)

(2) There are a lot of high-level data analyses (e.g., decoding, analyzing decoding errors, calculating mutual information, calculating distances in state space, etc.) but very little neural data (except for Figure 2C, and see my comment above about how this is flawed). So, if responses to novel vs. familiar odors are different in the AON and aPCx, how are they different? Why is decoding accuracy better for novel odors in CA1 but better for familiar odors in SUB (Figure 3A)? The authors identify a small subset of neurons that have unusually high weights in the SVM analyses that contribute to decoding novelty, but they don't tell us which neurons these are and how they are responding differently to novel vs. familiar odors.

(3) The authors call AON and aPCx "primary sensory cortices" and LEC, CA1, and Sub "multisensory areas". This is a straw man argument. For example, we now know that PCx encodes multimodal signals (Poo et al. 2021, Federman et al., 2024; Kehl et al., 2024), and LEC receives direct OB inputs, which has traditionally been the criterion for being considered a "primary olfactory cortical area". So, this terminology is outdated and wrong, and although it suits the authors' needs here in drawing distinctions, it is simplistic and not helpful moving forward.

(4) Why not simply report z-scored firing rates for all neurons as a function of trial number? (e.g., Jacobson & Friedrich, 2018). Figure 2C is not sufficient. For example, in the Discussion, they say, "novel stimuli caused larger increases in firing rates than familiar stimuli" (L. 270), but what does this mean? Odors typically increase the firing in some neurons and suppress firing in others. Where does the delta come from? Is this because novel odors more strongly activate neurons that increase their firing or because familiar odors more strongly suppress neurons?

(5) Lines 122-124 - If cells in AON and aPCx responded the same way to novel and familiar odors, then we would say that they only encode for odor and not at all for experience. So, I don't understand why the authors say these areas code for a "mixed representation of chemical identity and experience." "On the other hand," if LEC, CA1, and SUB are odor selective and only encode novel odors, then these areas, not AON and aPCx, are the jointly encoding chemical identity and experience. Also, I do not understand why, here, they say that AON and PCx respond to both while LEC, CA1, and SUB were selective for novel stimuli, but the authors then go on to argue that novelty is encoded in the AON and PCx, but not in the LEC, CA1, and SUB.

(6) Lines 132-140 - As presented in the text and the figure, this section is poorly written and confusing. Their use of the word "shuffled" is a major source of this confusion, because this typically is the control that produces outcomes at the chance level. More importantly, they did the wrong analysis here. The better and, I think, the only way to do this analysis correctly is to train on some of the odors and test on an untrained odor (i.e., what Bernardi et al., 2021 called "cross-condition generalization performance"; CCGP).

Reviewer #3 (Public review):

In this manuscript, the authors investigate how odor-evoked neural activity is modulated by experience within the olfactory-hippocampal network. The authors perform extracellular recordings in the anterior olfactory nucleus (AON), the anterior piriform (aPCx) and lateral entorhinal cortex (LEC), the hippocampus (CA1), and the subiculum (SUB), in naïve mice and in mice repeatedly exposed to the same odorants. They determine the response properties of individual neurons and use population decoding analyses to assess the effect of experience on odor information coding across these regions.

The authors' findings show that odor identity is represented in all recorded areas, but that the response magnitude and selectivity of neurons are differentially modulated by experience across the olfactory-hippocampal pathway.

Overall, this work represents a valuable multi-region data set of odor-evoked neural activity. However, limitations in the interpretability of odor experience of the behavioral paradigm, and limitations in experimental design and analysis, restrict the conclusions that can be drawn from this study.

eLife Assessment

In this useful study, the authors use published scRNA-seq data to highlight the importance of mast cells (MCs) in TB granulomas, reporting a comparative assessment of chymase- and tryptase-expressing MCs in the lungs of tuberculosis-infected individuals and non-human primates, with MC-deficient mice showing reduced lung bacterial burden and pathology during infection. Whilst the findings are helpful, the evidence to support conclusions is inconsistent across models and thus incomplete. Specifically, the data supporting a role for MCs in coordinating cytokine responses to modulate pathology, susceptibility to tuberculosis, and dissemination during infection are weak.

Reviewer #1 (Public review):

Summary:

The study by Gupta et al. investigates the role of mast cells (MCs) in tuberculosis (TB) by examining their accumulation in the lungs of M. tuberculosis-infected individuals, non-human primates, and mice. The authors suggest that MCs expressing chymase and tryptase contribute to the pathology of TB and influence bacterial burden, with MC-deficient mice showing reduced lung bacterial load and pathology.

Strengths:

(1) The study addresses an important and novel topic, exploring the potential role of mast cells in TB pathology.

(2) It incorporates data from multiple models, including human, non-human primates, and mice, providing a broad perspective on MC involvement in TB.

(3) The finding that MC-deficient mice exhibit reduced lung bacterial burden is an interesting and potentially significant observation.

Weaknesses:

(1) The evidence is inconsistent across models, leading to divergent conclusions that weaken the overall impact of the study.

(2) Key claims, such as MC-mediated cytokine responses and conversion of MC subtypes in granulomas, are not well-supported by the data presented.

(3) Several figures are either contradictory or lack clarity, and important discrepancies, such as the differences between mouse and human data, are not adequately discussed.

(4) Certain data and conclusions require further clarification or supporting evidence to be fully convincing.

Reviewer #2 (Public review):

Summary:

The submitted manuscript aims to characterize the role of mast cells in TB granuloma. The manuscript reports heterogeneity in mast cell populations present within the granulomas of tuberculosis patients. With the help of previously published scRNAseq data, the authors identify transcriptional signatures associated with distinct subpopulations.

Strengths:

(1) The authors have carried out a sufficient literature review to establish the background and significance of their study.

(2) The manuscript utilizes a mast cell-deficient mouse model, which demonstrates improved lung pathology during Mtb infection, suggesting mast cells as a potential novel target for developing host-directed therapies (HDT) against tuberculosis.

Weaknesses:

(1) The manuscript requires significant improvement, particularly in the clarity of the experimental design, as well as in the interpretation and discussion of the results. Enhanced focus on these areas will provide better coherence and understanding for the readers.

(2) Throughout the manuscript, the authors have mislabelled the legends for WT B6 mice and mast cell-deficient mice. As a result, the discussion and claims made in relation to the data do not align with the corresponding graphs (Figure 1B, 3, 4, and S2). This discrepancy undermines the accuracy of the conclusions drawn from the results.

(3) The results discussed in the paper do not add a significant novel aspect to the field of tuberculosis, as the majority of the results discussed in Figure 1-2 are already known and are a re-validation of previous literature.

(4) The claims made in the manuscript are only partially supported by the presented data. Additional extensive experiments are necessary to strengthen the findings and enhance the overall scientific contribution of the work.

Author Response:

Reviewer #1 (Public Review):

Summary:

The study by Gupta et al. investigates the role of mast cells (MCs) in tuberculosis (TB) by examining their accumulation in the lungs of M. tuberculosis-infected individuals, non-human primates, and mice. The authors suggest that MCs expressing chymase and tryptase contribute to the pathology of TB and influence bacterial burden, with MC-deficient mice showing reduced lung bacterial load and pathology.

Strengths:

(1) The study addresses an important and novel topic, exploring the potential role of mast cells in TB pathology.

(2) It incorporates data from multiple models, including human, non-human primates, and mice, providing a broad perspective on MC involvement in TB.

(3) The finding that MC-deficient mice exhibit reduced lung bacterial burden is an interesting and potentially significant observation.

Weaknesses:

(1) The evidence is inconsistent across models, leading to divergent conclusions that weaken the overall impact of the study.

The strength of the study is the use of multiple models including mouse, non-human primate as well as human samples. The conclusions have now been refined to reflect the complexity of the disease and the use of multiple models.

(2) Key claims, such as MC-mediated cytokine responses and conversion of MC subtypes in granulomas, are not well-supported by the data presented.

To address the reviewer’s comments, we will carry out further experimentation to strengthen the link between MC subtypes and cytokine responses.

(3) Several figures are either contradictory or lack clarity, and important discrepancies, such as the differences between mouse and human data, are not adequately discussed.

We will further clarify the figures and streamline the discussions between the different models used in the study.

(4) Certain data and conclusions require further clarification or supporting evidence to be fully convincing.

We will either provide clarification or supporting evidence for some of the key conclusions in the paper.

Reviewer #2 (Public review):

Summary:

The submitted manuscript aims to characterize the role of mast cells in TB granuloma. The manuscript reports heterogeneity in mast cell populations present within the granulomas of tuberculosis patients. With the help of previously published scRNAseq data, the authors identify transcriptional signatures associated with distinct subpopulations.

Strengths:

(1) The authors have carried out a sufficient literature review to establish the background and significance of their study.

(2) The manuscript utilizes a mast cell-deficient mouse model, which demonstrates improved lung pathology during Mtb infection, suggesting mast cells as a potential novel target for developing host-directed therapies (HDT) against tuberculosis.

Weaknesses:

(1) The manuscript requires significant improvement, particularly in the clarity of the experimental design, as well as in the interpretation and discussion of the results. Enhanced focus on these areas will provide better coherence and understanding for the readers.

The strength of the study is the use of multiple models including mouse, non-human primate as well as human samples. The conclusions have now been refined to reflect the complexity of the disease and the use of multiple models.

(2) Throughout the manuscript, the authors have mislabelled the legends for WT B6 mice and mast cell-deficient mice. As a result, the discussion and claims made in relation to the data do not align with the corresponding graphs (Figure 1B, 3, 4, and S2). This discrepancy undermines the accuracy of the conclusions drawn from the results.

We apologize for the discrepancy which will be corrected in the revised manuscript

(3) The results discussed in the paper do not add a significant novel aspect to the field of tuberculosis, as the majority of the results discussed in Figure 1-2 are already known and are a re-validation of previous literature.

This is the first study which has used mouse, NHP and human TB samples from Mtb infection to characterize and validate the role of MC in TB. We believe the current study provides significant novel insights into the role of MC in TB.

(4) The claims made in the manuscript are only partially supported by the presented data. Additional extensive experiments are necessary to strengthen the findings and enhance the overall scientific contribution of the work.

We will either provide clarification or supporting evidence for some of the key conclusions in the paper.

eLife Assessment

This interesting study explores whether tumor cells can manipulate their Hydra hosts, and includes important findings on the consequences for the fitness of the host Hydra. The evidence supporting these findings is convincing. The work will be of broad interest to many fields including development biology, evolutionary biology and tumor biology.

Reviewer #1 (Public review):

Summary:

In this manuscript, BOUTRY et al examined a cnidarian Hydra model system where spontaneous tumors manifest in laboratory settings, and lineages featuring vertically transmitted neoplastic cells (via host budding) have been sustained for over 15 years. They observed that hydras harboring long-term transmissible tumors exhibit an unexpected augmentation in tentacle count. In addition, the presence of extra tentacles, enhancing the host's foraging efficiency, correlated with an elevated budding rate, thereby promoting tumor transmission vertically. This study provided the evidence that tumors, akin to parasitic entities, can also exert control over their hosts.

Strengths:

The manuscript is well-written, and the phenotype is intriguing.

Reviewer #2 (Public review):

Background and Summary: 

This study addresses the intriguing question of whether and how tumours can develop in the freshwater polyp hydra and how they influence the fitness of the animals. Hydra is notable for its significant morphogenetic plasticity and nearly unlimited capacity for regeneration. While its growth through asexual reproduction (budding) and the associated processes of pattern formation have been extensively studied at the cellular level, the occurrence of tumours was only recently described in two strains of Hydra oligactis (Domazet-Lošo et al, 2014). Here, tumour-like tissue bulges formed within the ectodermal epithelial layer and contained increased numbers of interstitial cell-like cells which exhibited female germline markers, but none specific for somatic derivatives of interstitial stem cells (e.g., nematocytes, neurons or glandular cells). It seems likely that the cellular basis of these malformations is a misregulation of oogenesis. In wild-type polyps, interstitial-cell-related germline precursors give rise to oocytes and nurse cells, which are subsequently phagocytosed by the growing egg cell. By comparison, in the mutant strains, this uptake is disturbed, but the homeostasis between germline cells and epithelial cells must remain functional enabling further growth pattern formation in hydra. Determining whether this differentiation arrest constitutes a neoplasm also remains a challenge. 

Clonal lines of both strains have been maintained in the laboratory for years and have also been used by Boutry and colleagues. They published two further papers on the ecological and evolutionary aspects of hydra tumour formation (Boutry et al 2022, 2023), which is also the focus of this manuscript. In their paper, the authors demonstrate an increase in the number of tentacles when "tumour tissue" was transplanted to intact gastric tissue of wildtype and mutant strains. While the impact on tentacle formation is relatively modest, small, it indicates a potential influence on the cross-talk between epithelial and interstitial cells in growth control (proportion regulation). The presented data are of interest, although the underlying molecular processes remain to be demonstrated. The authors offer a different interpretation. They conclude that this growth pattern (increased number of tentacles) is correlated with "reducing the burden on the host by (over-) compensating for the reproductive costs of tumours" and claim that "transmissible tumours in hydra have evolved strategies to manipulate the phenotype of their host". 

Strength <br /> The question of whether and how tumours can develop in simple systems, here the freshwater polyp hydra, is of general interest. The authors describe transplantation experiments by using mutant strains that indicate an influence of tumour-like malformation on pattern formation. The experiments also suggest an interaction between epithelial cells and germline cells during oogenesis, interfering with the homeostatic growth control between the cell lineages. 

Weaknesses <br /> Although it is stimulating to consider a fresh perspective from other disciplines (here, ecological and evolutionary aspects), it appears that this interpretation of the data (reducing the burden on the host by (over-) compensating for the reproductive costs of tumours) is somewhat beyond what can be reasonably inferred from the evidence presented. It is essential, particularly in the context of evolutionary biology, to conduct further analysis of the underlying cell biology of these intriguing mutant hydra strains. Such cellular analysis is a relatively straightforward approach that could provide a mechanistic understanding of the phenomenon described by the authors.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, BOUTRY et al examined a cnidarian Hydra model system where spontaneous tumors manifest in laboratory settings, and lineages featuring vertically transmitted neoplastic cells (via host budding) have been sustained for over 15 years. They observed that hydras harboring long-term transmissible tumors exhibit an unexpected augmentation in tentacle count. In addition, the presence of extra tentacles, enhancing the host's foraging efficiency, correlated with an elevated budding rate, thereby promoting tumor transmission vertically. This study provided evidence that tumors, akin to parasitic entities, can also exert control over their hosts.<br /> Strengths:

The manuscript is well-written, and the phenotype is intriguing.

Weaknesses:

The quality of this manuscript could be improved if more evidence were to be provided regarding the beneficial versus detrimental effects of the tumors.

We thank the reviewer for taking the time to examine our work carefully and for their highly relevant comments and precise suggestions. We have incorporated these suggestions, which greatly improved the clarity of our manuscript concerning the beneficial and detrimental effects of tumors. Specifically, we have added a new analysis and rephrased the results section, as well as the corresponding sentences in the discussion, to enhance clarity.

Additionally, regarding the impact of tumor size on the development of supernumerary tentacles, we have included as suggested a new analysis that was previously only available in the supplementary materials of the earlier version. This addresses the reviewer's question and significantly enhances the quality of our paper.

We have thanked the two referees in the Acknowledgements section of our article.

Reviewer #2 (Public Review):

Background and Summary:

This study addresses the intriguing question of whether and how tumors can develop in the freshwater polyp hydra and how they influence the fitness of the animals. Hydra is notable for its significant morphogenetic plasticity and nearly unlimited capacity for regeneration. While its growth through asexual reproduction (budding) and the associated processes of pattern formation have been extensively studied at the cellular level, the occurrence of tumors was only recently described in two strains of Hydra oligactis (Domazet-Lošo et al, 2014). In that research, an arrest in the differentiation of female germ cells led to an accumulation of germline cells that failed to develop into eggs. In hydra, fertile egg cells typically incorporate nurse cells, which originate from large interstitial stem cells (ISCs) restricted to the germline, through apoptosis. However, this increase in apoptosis activity is absent in "germline tumors," and germline ISCs instead form slowly growing patches that do not compromise tissue integrity. Despite the upregulation of certain genes associated with mammalian neoplasms (such as tpt1 and p23) in this tissue, determining whether this differentiation arrest and the resulting egg patches truly constitute neoplasms remains a challenge.

The authors have recently published two papers on the ecological and evolutionary aspects of hydra tumor formation (Boutry et al 2022, 2023), which is also the focus of this manuscript. They transplanted tissues derived from animals with germline tumors to wildtype animals and analyzed their growth patterns, specifically the number of tentacles in the host tissue. They observed that such tissues induced the growth of additional tentacles compared to tissues without germline tumors. The authors conclude that this growth pattern (increased number of tentacles) is correlated with "reducing the burden on the host by (over-)compensating for the reproductive costs of tumors" and claim that "transmissible tumors in hydra have evolved strategies to manipulate the phenotype of their host". While it might be stimulating to add a fresh view from other disciplines (here, ecological and evolutionary aspects), the authors completely ignore the current knowledge of the underlying cell biology of the processes they analyze.

Strengths:

The study focuses on intriguing questions. Whether and how tumors can develop in the freshwater polyp hydra, and how they influence the fitness of the animals?

Weaknesses:

Concept of germline tumors.

The conceptual foundation of their experiments on germline tumors was the study of Domazet-Lošo et al (2014) introducing the concept of germline tumors in hydra (see above). While this is an intriguing hypothesis, there has been little advancement in comprehending the molecular mechanisms underlying tumor formation in hydra beyond this initial investigation. Germline tumors in hydra do not fully meet the typical criteria for neoplasms observed in mammalian tissues. More importantly, a similar phenotype was already reported by the work of Paul Brien and described as "crise gametique" (Brien, 1966, Biologie de la reproduction animale - Blastogenèse, Gamétogenèse, Sexualisation, ed. Masson & Cie, Paris). This phenomenon of gametic crisis is unique to Hydra oligactis, a stenotherm, cold-adapted cosmopolitan species. In this species, gametogenesis severely impacts the vitality of the polyps, often leading to complete exhaustion and death (Tardent, 1974). Animals can only be rescued during the initial phase of the cold-induced sexual period (see also the research of Littlefield (1984, 1985, 1986, 1991). The observed arrest in differentiation arrest in germline tumors might represent an epigenetically established consequence of surviving gametogenesis. Regrettably, this important work was not mentioned by the authors or by Domazet-Lošo et al. (2014), highlighting a notable gap in the recognition of basic research in this area that might challenge the hydra tumor hypothesis.

"Super-nummary" tentacles in graft experiments.

The authors describe that after grafting tissue from animals with germline tumors to wild-type animals, the number of tentacles in the host tissue increased when the donor tissue had germline tumors. A maximum effect of four additional tentacles was found with donor strain H. oligactis robusta and three additional tentacles with donor strain H.oligactis St Petersburg. In general, H.oligactis wild-type host strains had fewer tentacles than H.oligactis St Petersburg strains. This is consistent with the results of Domazet-Lošo et al (2014) who showed that the number of tentacles increased in the strains with germline tumors. What conclusions can be drawn from these experiments? 

The authors might want to conclude that transmissible tumors in Hydra have developed strategies to manipulate the phenotype of their host. But there is no evidence for this, as essential controls are missing. It is known that the size of hydra polyps is proportion-regulated, i.e. the number of tentacles varies with the size and number of (epithelial) cells. Such controls are missing in the experiments. There is also a lack of controls from wild-type animals in gametogenesis: it is very likely that grafts with wild-type animals with egg spots of comparable size as the germline tumors (see above) will result in similar numbers of tentacles in host tissue.

We thank the reviewer for their thoughtful comments. While we appreciate the concerns raised, we maintain that the evidence provided by Domazet-Lošo et al. (2014, Nature Communications) supports the relevance of this model, including the suggested comparisons with the expression profiles of individuals undergoing induced sexual reproduction. Our study focuses primarily on the impact of these tumors on the host phenotype rather than their origin. Tumors are defined as accumulations of abnormally proliferating cells. This includes the definition provided by the referee, which describes “apoptosis activity as absent in 'germline tumors,' with germline ISCs forming slowly growing patches.” Compromise of tissue integrity is not a criterion for defining neoplasms, and many benign neoplasms do not meet this criterion. We are interested in continuing this discussion with the referee to better understand the expected evidence and agree that histological nomenclature could be improved. While further investigation into the cell biology of these tumors would be valuable, this is currently beyond the scope of our article but is being pursued in separate research.

We also appreciate the points raised regarding the definition of germline tumors and the reference to the pioneering work of Paul Brien. However, in that publication, the concept of gametic crisis in H. oligactis describes reproductive exhaustion leading to death, rather than abnormal cell proliferation indicative of a tumor-like phenotype. This distinction likely explains why this specific paper was not cited previously.

Our study builds on prior research using the same model (e.g., Domazet-Lošo et al. 2014; Boutry et al. 2023) and describes observations across different hydra strains from various locations worldwide (not just two), all conducted under stable warm temperatures that are not conducive to sexual development. These investigations reveal a phenomenon distinct from the senescence observed post-reproduction in H. oligactis. The phenotype we describe, characterized by an accumulation of cells in the ectoderm, aligns with studies referenced by the reviewer from leading groups in hydra research, known for their expertise in hydra cellular biology. We have relied on these studies after carefully reviewing their results and receiving training from these experts. Furthermore, our team is focused on eco-evolutionary topics and does not aim to specialize in cellular biology, as other teams are already dedicated to that field.

We also thank the reviewer for their comments on the relevance of our findings and the missing controls. However, we have noted that the reviewer may have misunderstood our experimental design and results.

Firstly, it appears that the reviewer based their critique mainly on the initial sentences of our Results section (illustrated in Figure 2), which outline the donor groups used in our study rather than presenting the results of the grafting experiments. This description alone is insufficient for drawing conclusions, which is why we conducted further analyses using these donor groups grafted onto different recipients. The maximum effects mentioned by the reviewer (+10 tentacles with St. Petersburg tumoral tissue and +8 tentacles with Robusta tumoral tissue, Results Section 2) represent only a part of our study. We encourage the reviewer to focus on the model analyses presented in Results Section 2, which directly relate to the grafting experiments and provide a more comprehensive evaluation of our results and conclusions. These analyses include comparisons between transmissible tumors and spontaneous tumors, offering deeper insights into their effects on tentacle development.

In our methods (as depicted in Figure 3), we explicitly compared different types of tumorous tissue from various donors, distinguishing between spontaneous and transmissible tumors. Although we avoid labeling spontaneous tumors as "controls" to prevent confusion with healthy tissue controls, they serve as controls to the “treatment” that involves transmissible tumors, and thus are appropriate comparisons for assessing the size effect suggested by the reviewer. Spontaneous and transmissible tumors share similar size and cellular characteristics but differ significantly in the number of tentacles their hosts possess. Furthermore, we refer the reviewer to a relevant study (Ngo et al. 2021) that found no increase in tentacle numbers with larger polyps of healthy tissue. This reference has been included in the revised discussion (line 309 to 312), which now also addresses the potential effect of body size with additional explanations.

Regarding the suggestion to include controls from animals undergoing gametogenesis, we did not find evidence in the literature indicating an increase in tentacle numbers during this process in hydra. If such studies exist, we kindly request the complete references so we can include them in our discussion. Additionally, as noted in Brien's work, Hydra oligactis undergoing gametogenesis are known to either die or experience significant degeneration afterward. Transplanting tissue from dead or dying (and reproducing) hydras poses technical challenges and raises questions about whether any observed effects result from incomplete gametogenesis, the onset of senescence, or both. While these questions are intriguing, they fall outside the scope of our article.

In conclusion, we appreciate the opportunity to address these points and reaffirm that our study offers valuable insights into the evolutionary dynamics of interactions between transmissible tumor tissues and host phenotypes in hydra. We remain open to further discussion and welcome any additional feedback to enhance the clarity and robustness of our manuscript.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) If the fitness of hydra is altered in those with spontaneous tumors is the increased number of tentacles associated with those with transmitted tumors able to rescue this phenotype?

We thank the reviewer for reformulating our results. Indeed, fitness can be restored and even improved in tumorous polyps harboring supernumerary tentacles. This phenomenon, which we referred to as compensation and over-compensation in Section 3 and Figure 4, was initially discussed only in the discussion section. To improve the clarity of our manuscript, we have now specified this in the Conclusion (lines 345 to 347 and some minor rewording in the same paragraph) in the Results section (lines 284 to 286).

(2) Does the size of the tumor predict the number of tentacles formed?

We agree that this would be a valuable complementary analysis. We have conducted an analysis considering the qualitative size of the tumors (based on visual categories) and the number of tentacles, which is now included in our paper (lines 160-161; lines 193 to 198; lines 253 to 259; lines 314 - 322).

(3) Considering the mentioned association of body size with tentacle numbers for hydra, is a change in size a phenotype associated with transmitted tumors, and is such a phenotype transmittable. 

All tumorous individuals, regardless of their tumor type, exhibit a swollen body. We have added a sentence in the introduction to clarify this point (line 62).

(4) Is there anything unique about the Rob population that would explain their mass mortality following transplantation? For instance, their resistance to spontaneous tumor formation? Similarly, is there a difference in transplantation success based on the type of tissue transplanted? The authors could address this point in the discussion.

It is a very old lineage described nearly 80 years ago. It is unknown whether natural populations of Robusta exist, and no reports of any male individuals have been documented. We have added a sentence in the Materials and Methods section to clarify this information (lines 98 to 102).

(5) What downsides are known about the transmittable tumors in hydra and how present are they in the grafted individuals? Are other physiological aspects such as mobility, regeneration, or sexual reproduction hindered?

Transmissible tumors have been associated with increased vulnerability to predation and alterations in life history traits, including a higher budding rate and decreased sexual reproduction. While we were unable to measure behavioral traits in this study of our grafted individuals, this is an intriguing avenue for further research. We have included this perspective in the discussion section as a concluding remark (lines 375 to 382). Thanks a lot for the suggestion of this conclusion.

(6) It is important to explore the mechanisms behind the phenotypic variation conferred by the types of tumors, whether of different lineage or transmissibility. For this purpose, RNA-Seq on the recipients seems like a good starting point.

Thanks for this suggestion, we've reworded the sentence about this perspective in our discussion to be more precise (line 320).

Boutry, Justine, Marie Buysse, Sophie Tissot, Chantal Cazevielle, Rodrigo Hamede, Antoine M. Dujon, Beata Ujvari, et al. 2023. « Spontaneously Occurring Tumors in Different Wild-Derived Strains of Hydra ». Scientific Reports 13 (1): 7449. https://doi.org/10.1038/s41598-023-34656-0.

Domazet-Lošo, Tomislav, Alexander Klimovich, Boris Anokhin, Friederike Anton-Erxleben, Mailin J. Hamm, Christina Lange, et Thomas C. G. Bosch. 2014. « Naturally occurring tumours in the basal metazoan {Hydra} ». Nat Commun 5 (1): 4222. https://doi.org/10.1038/ncomms5222.

Ngo, Kha Sach, Berta R-Almási, Zoltán Barta, et Jácint Tökölyi. 2021. « Experimental Manipulation of Body Size Alters Life History in Hydra ». Ecology Letters 24 (4): 728‑38. https://doi.org/10.1111/ele.13698.

eLife Assessment

This important study provides proof of principle that C. elegans models can be used to accelerate the discovery of candidate treatments for human Mendelian diseases by detailed high-throughput phenotyping of strains harboring mutations in orthologs of human disease genes. The data are compelling and support an approach that enables the potential rapid repurposing of FDA-approved drugs to treat rare diseases for which there are currently no effective treatments. The work will be of interest to all geneticists.

Reviewer #3 (Public review):

In this study, O'Brien et al. address the need for scalable and cost-effective approaches to finding lead compounds for the treatment of the growing number of Mendelian diseases. They used state-of-the-art phenotypic screening based on an established high-dimensional phenotypic analysis pipeline in the nematode C. elegans.

First, a panel of 25 C. elegans models was created by generating CRISPR/Cas9 knock-out lines for conserved human disease genes. These mutant strains underwent behavioral analysis using the group's published methodology. Clustering analysis revealed common features for genes likely operating in similar genetic pathways or biological functions. The study also presents results from a more focused examination of ciliopathy disease models.

Subsequently, the study focuses on the NALCN channel gene family, comparing the phenotypes of mutants of nca-1, unc-77, and unc-80. This initial characterization identifies three behavioral parameters that exhibit significant differences from the wild type and could serve as indicators for pharmacological modulation.

As a proof-of-concept, O'Brien et al. present a drug repurposing screen using an FDA-approved compound library, identifying two compounds capable of rescuing the behavioral phenotype in a model with UNC80 deficiency. The relatively short time and low cost associated with creating and phenotyping these strains suggest that high-throughput worm tracking could serve as a scalable approach for drug repurposing, addressing the multitude of Mendelian diseases. Interestingly, by measuring a wide range of behavioural parameters, this strategy also simultaneously reveals deleterious side effects of tested drugs that may confound the analysis.

Considering the wealth of data generated in this study regarding important human disease genes, it is regrettable that the data is not made accessible to researchers less versed in data analysis methods. This diminishes the study's utility. It would have a far greater impact if an accessible and user-friendly online interface were established to facilitate data querying and feature extraction for specific mutants. This would empower researchers to compare their findings with the extensive dataset created here.

Another technical limitation of the study is the use of single alleles. Large deletion alleles were generated by CRISPR/Cas9 gene editing. At first glance, this seems like a good idea because it limits the risk that background mutations, present in chemically-generated alleles, will affect behavioral parameters. However, these large deletions can also remove non-coding RNAs or other regulatory genetic elements, as found, for example, in introns. Therefore, it would be prudent to validate the behavioral effects by testing additional loss-of-function alleles produced through early stop codons or targeted deletion of key functional domains.

Comments on revisions:

In this final round of revisions, the authors have improved their manuscript and provide useful information about analysis procedures and code and updated figures.

Author response:

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

This important study provides proof of principle that C. elegans models can be used to accelerate the discovery of candidate treatments for human Mendelian diseases by detailed high-throughput phenotyping of strains harboring mutations in orthologs of human disease genes. The data are compelling and support an approach that enables the potential rapid repurposing of FDA-approved drugs to treat rare diseases for which there are currently no effective treatments. The authors should provide a clearer explanation of how the statistical analyses were performed, as well as a link to a GitHub repository to clarify how figures and tables in the manuscript were generated from the phenotypic data.

We have amended our description of the statistical analysis in the materials and methods section of the manuscript. We have also updated the GitHub repository link to a dedicated repository for this study, this contains all of the code needed to generated all the figures made from the phenotypic data provided. Additionally, we have updated the Zenodo repository to contain both the code and datasets within the same file.

We have also updated the GitHub repository link to a dedicated repository for this manuscript, that contains all of the code needed to generate all figures from the phenotypic data provided. Additionally, we have updated the Zenodo repository link to contain both the code and datasets within the same folder structure. 

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The authors have responded to previous review to improve the presentation of the work. The paper more than meets publication standards.

No response required.

Reviewer #2 (Recommendations for the authors):

The authors have addressed all of my questions and concerns. I'm happy to see this updated paper of record.

No response required.

Reviewer #3 (Recommendations for the authors):

Regarding the interactive heatmap

The html version and the panel in Figure 2C appear not to coincide visually. Maybe the features are ordered in a different way?

The html version of Figure 2C is for the entire feature set extract per strain and not the condensed Tierpsy256 set shown in the panel figure. We have now remade this figure to show this reduced feature set (aligning with what is shown in Figure 2C) and included both versions of the interactive heatmaps as static html files within the same repository.

Regarding data accessibility overall

More generally, the html file does not address my initial concern about the accessibility of the data to non-experts. Making the full dataset available was a necessary first step, but the hermetic nature of its format and the lack of a simple way to query the data remains an issue for me that limits the usefulness of this data to the broadest audience.

We agree, but unfortunately do not currently have the resources to build a public-facing database to facilitate this.

eLife Assessment

Saijilafu et al. describe that MLCK and MLCP bidirectionally regulate NMII phosphorylation ultimately impinging on axonal growth during regeneration in the central and peripheral nervous systems. However, the evidence is in most cases incomplete, since some key controls are missing, some major claims are too broad to be supported by data and contrasting data on how MLCK and MLCP regulates NMII activity is not fully addressed or discussed. In sum, this knowledge is potentially useful for the field due to the relevance of identifying mechanisms that regulate axonal regeneration.

Reviewer #1 (Public review):

This paper examines the role of MLCK (myosin light chain kinase) and MLCP (myosin light chain phosphatase) in axon regeneration. Using loss-of-function approaches based on small molecule inhibitors and siRNA knockdown, the authors explore axon regeneration in cell culture and in animal models from central and peripheral nervous systems. Their evidence shows that MLCK activity facilitates axon extension/regeneration, while MLCP prevents it.

Major concerns:

(1) In the title, authors indicate that the observed effects from loss-of-function of MLCK/MLCP take place via F-actin redistribution in the growth cone. However, there are no experiments showing a causal effect between changes in axon growth mediated by MLCK/MLCP and F-actin redistribution.

(2) The author combines MLCK inhibitors with Bleb (Figure 6), trying to verify if both pairs of inhibitors act on the same target/pathway. MLCK may regulate axon growth independent of NMII activity. However, this has very important implications for the understanding not only on how NMII works and affects axon extension, but also in trying to understand what MLCP is doing. One wonders if MLCP actions, which are opposite of MLCK, also independent of NMII activity? The authors, in the discussion section, try to find an explanation for this finding, but I consider it fails since the whole rationale of the manuscript is still around how MLCK and MLCP affect NMII phosphorylation.

What follows is a discussion of the merits and limitations of different claims of the manuscript in light of the evidence presented.

(1) Using western blot and immunohistochemical analyses, authors first show that MLCK expression is increased in DRG sensory neurons following peripheral axotomy, concomitant to an increase in MLC phosphorylation, suggesting a causal effect (Figure 1). The authors claim that it is common that axon growth-promoting genes are upregulated. It would have been interesting at this point to study in this scenario the regulation of MLCP.

(2) Using DRG cultures and sciatic nerve crush in the context of MLCK inhibition (ML-7) and down-regulation, authors conclude that MLCK activity is required for mammalian peripheral axon regeneration both in vitro and in vivo (Figure 2). In parallel, the authors show that these treatments affect as expected the phosphorylation levels of MLC.

The in vitro evidence is of standard methods and convincing. However, here, as well as in all other experiments using siRNAs, no Control siRNAs were used. Authors do show that the target protein is downregulated, and they can follow transfected cells with GFP. Still, it should be noted that the standard control for these experiments has not been done.

(3) The authors then examined the role of the phosphatase MLCP in axon growth during regeneration. The authors first use a known MLCP blocker, phorbol 12,13-dibutyrate (PDBu), to show that is able to increase the levels of p-MLC, with a concomitant increase in the extent of axon regrowth of DRG neurons, both in permissive as well as non-permissive substrates. The authors repeat the experiments using the knockdown of MYPT1, a key component of the MLC-phosphatase, and again can observe a growth-promoting effect (Figure 3).

The authors further show evidence for the growth-enhancing effect in vivo, in nerve crush experiments. The evidence in vivo deserves more evidence and experimental details (see comment 2). A key weakness of the data was mentioned previously: no control siARN was used.

(4) In the next set of experiments (presented in Figure 4) authors extend the previous observations in primary cultures from the CNS. For that, they use cortical and hippocampal cultures, and pharmacological and genetic loss-of-function using the above-mentioned strategies. The expected results were obtained in both CNS neurons: inhibition or knockdown of the kinase decreases axon growth, whereas inhibition or knockdown of the phosphatase increases growth. A main weakness in this set is that drugs were used from the beginning of the experiment, and hence, they would also affect axon specification. As pointed in Materials and Method (lines 143-145) authors counted as "axons" neurites longer than twice the diameter of the cell soma, and hence would not affect the variable measured. In any case, to be sure one is only affecting axon extension in these cells, the drugs should have been used after axon specification and maturation, which occurs at least after 5 DIV.

(5) In Figure 7, the authors a local cytoskeletal action of the drug, but the evidence provided does not differentiate between a localized action of the drugs and a localized cell activity.

References:

(1) Eun-Mi Hur 1, In Hong Yang, Deok-Ho Kim, Justin Byun, Saijilafu, Wen-Lin Xu, Philip R Nicovich, Raymond Cheong, Andre Levchenko, Nitish Thakor, Feng-Quan Zhou. 2011. Engineering neuronal growth cones to promote axon regeneration over inhibitory molecules. Proc Natl Acad Sci U S A. 2011 Mar 22;108(12):5057-62. doi: 10.1073/pnas.1011258108.

(2) Garrido-Casado M, Asensio-Juárez G, Talayero VC, Vicente-Manzanares M. 2024. Engines of change: Nonmuscle myosin II in mechanobiology. Curr Opin Cell Biol. 2024 Apr;87:102344. doi: 10.1016/j.ceb.2024.102344.

(3) Karen A Newell-Litwa 1, Rick Horwitz 2, Marcelo L Lamers. 2015. Non-muscle myosin II in disease: mechanisms and therapeutic opportunities. Dis Model Mech. 2015 Dec;8(12):1495-515. doi: 10.1242/dmm.022103.

Reviewer #2 (Public review):

Summary:

Saijilafu et al. demonstrate that MLCK/MLCP proteins promote axonal regeneration in both the central nervous system (CNS) and peripheral nervous system (PNS) using primary cultures of adult DRG neurons, hippocampal and cortical neurons, as well as in vivo experiments involving sciatic nerve injury, spinal cord injury, and optic nerve crush. The authors show that axon regrowth is possible across different contexts through genetic and pharmacological manipulation of these proteins. Additionally, they propose that MLCK/MLCP may regulate F-actin reorganization in the growth cone, which is significant as it suggests a novel strategy for promoting axonal regeneration.

Strengths:

This manuscript presents a wide range of experimental models to address its hypothesis and biological question. Notably, the use of multiple in vivo models significantly enhances the overall validity of the study.

Weaknesses:

-The authors previously published that blocking myosin II activity stimulates axonal growth and that MLCK activates myosin II. The present work shows that inhibiting MLCK blocks axonal regeneration while blocking MLCP (the protein that dephosphorylates myosin II) produces the opposite effect. Although this contradiction is discussed, no new evidence has been added to the manuscript to clarify this mechanism or address the remaining questions. Critical unresolved questions include: what happens to myosin II expression when both MLCK and MLCP are inhibited? If MLCK/MLCP are acting through an independent mechanism, what would that mechanism be?<br /> -In the discussion, the authors mention the existence of two myosin II isoforms with opposing effects on axonal growth. Still, there is no evidence in the manuscript to support this point.<br /> -It is also unclear how MLCK/MLCP acts on the actin cytoskeleton. The authors suggest that proteins such as ADF/cofilin, Arp 2/3, Eps8, Profilin, Myosin II, and Myosin V could regulate changes in F-actin dynamics. However, this study provides no experimental evidence to determine which proteins may be involved in the mechanism.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

This paper examines the role of MLCK (myosin light chain kinase) and MLCP (myosin light chain phosphatase) in axon regeneration. Using loss-of-function approaches based on small molecule inhibitors and siRNA knockdown, the authors explore axon regeneration in cell culture and in animal models. Their evidence shows that MLCK activity facilitates axon extension/regeneration, while MLCP prevents it.

Major concern:

A global inconsistency in the conclusions of the authors is evident when trying to understand the role of NMII in axon growth and to understand the present results in light of previous reports by the authors and many others on the role of NMII in axon extension. The discussion of the matter fails to acknowledge a vast literature on how NMII activity is regulated. The authors study enzymes responsible for the phosphorylation and dephosphorylation of NMII, referring to something that is strongly proven elsewhere, that phosphorylation activates NMII and dephosphorylation deactivates it. The authors mention their own previous evidence using inhibitors of NMII ATPase activity (blebbistatin, Bleb for short) and inhibitors of a kinase that phosphorylates NMII (ROCK), highlighting that Bleb increases axon growth. Since Bleb inhibits the ATPase activity of NMII, it follows that NMII is in itself an inhibitor of axon growth, and hence when NMII is inhibited, the inhibition on axon growth is relieved, and axonal growth takes place (REF1). It is known that NMII exists in an inactive folded state, and ser19 phosphorylation (by MLCK or ROCK) extends the protein, allowing NMII filament formation, ATPase activity, and force generation on actin filaments (REF2). From this, it is derived that if MLCK is inhibited, then there is no NMII phosphorylation, and hence no NMII activity, and, according to their previous work, this should promote axon growth. On the contrary, the authors show the opposite effect: in the lack of phospho-MLC, authors show axon growth inhibition.

We thank the Reviewer for taking time to review our manuscript, and we really appreciated the comments from the reviewer. We have tried our best to revise the manuscript to address all the comments raised by the Reviewer.

Reporting evidence challenging previous conclusions is common business in scientific endeavors, but the problem with the current manuscript is that it fails to point to and appropriately discuss this contradiction. Instead, the authors refer to the fact that MLCK and Bleb inhibit NMII in different steps of the activation process. While this is true, this explanation does not solve the contradiction. There are many options to accommodate the information, but it is not the purpose of this revision to provide them. Since the manuscript is focused solely on phosphorylation states of MLC and axon extension, the claims are simply at odds with the current literature, and this important finding, if true, is not properly discussed.

Thank you for reviewer's very good comments. As suggested by Reviewer, we discuss more detail it in our revised manuscripts (line 357-368; line 373-374).

What follows is a discussion of the merits and limitations of different claims of the manuscript in light of the evidence presented.

(1) Using western blot and immunohistochemical analyses, authors first show that MLCK expression is increased in DRG sensory neurons following peripheral axotomy, concomitant to an increase in MLC phosphorylation, suggesting a causal effect (Figure 1). The authors claim that it is common that axon growth-promoting genes are upregulated. It would have been interesting at this point to study in this scenario the regulation of MLCP, which is a main subject in this work, and expect its downregulation.

We thank the Reviewer for taking time to review our manuscript, and we really appreciated the positive comments from the Reviewer.

(2) Using DRG cultures and sciatic nerve crush in the context of MLCK inhibition and down-regulation, authors conclude that MLCK activity is required for mammalian peripheral axon regeneration both in vitro and in vivo (Figure 2).

The in vitro evidence is of standard methods and convincing. However, here, as well as in all other experiments using siRNAs, it is not clear what the control is about (the identity of the plasmids and sequences, if any).

We used the pCMV–EGFP–N3 as control, and the pCMV–EGFP–N3 plasmid was from Clontech, Inc. (line 114-115).   

Related to this, it is not helpful to show the same exact picture as a control example in Figures 2 and 3 (panels J and E, respectively). Either because they should not have received the same control treatment, or simply because it raises concern that there are no other control examples worth showing. In these images, it is not also clear where and how the crush site is determined in the GFP channel. This is of major importance since the axonal length is measured from the presumed crush site. Apart from providing further details in the text, the authors should include convincing images.

Thank you so much for your comments. We changed the control example in Figure 3J. For sciatic nerve regeneration experiments, the sciatic nerve was exposed at the sciatic notch by a small incision 2 days after the in vivo electroporation. The nerve was then crushed, and the crush site was marked with a 11-0 nylon epineural suture. After surgeries, the wound was closed, and the mice were allowed to recover. Three days after the sciatic nerve crush, the whole sciatic nerves from the perfused animals were dissected out and postfixed overnight in 4% PFA at 4°C. Before whole-mount flattening, it was confirmed that the place of epineural suture matched the injury site, and experiments were included in the analysis only when the crush site was clearly identifiable. Using whole mounted tissue, all identifiable EGFP-labeled axons in the sciatic nerve were manually traced from the crush site to the distal growth cone to measure the length of axon regeneration. (line 159-164).

(3) The authors then examined the role of the phosphatase MLCP in axon growth during regeneration. The authors first use a known MLCP blocker, phorbol 12,13-dibutyrate (PDBu), to show that is able to increase the levels of p-MLC, with a concomitant increase in the extent of axon regrowth of DRG neurons, both in permissive as well as non-permissive. The authors repeat the experiments using the knockdown of MYPT1, a key component of the MLC-phosphatase, and again can observe a growth-promoting effect (Figure 3).

The authors further show evidence for the growth-enhancing effect in vivo, in nerve crush experiments. The evidence in vivo deserves more evidence and experimental details (see comment 2). Some key weaknesses of the data were mentioned previously (unclear RNAi controls and duplication of shown images), but in this case, it is also not clear if there is a change only in the extent of growth, or also in the number of axons that are able to regenerate.

Thank you so much for your comments. We used same control as in vitro experiments (the pCMV– EGFP–N3 plasmid was from Clontech, Inc), and we also changed the control image in Figure 3J. For in vivo axon regeneration experiments, we measured the lengths of all identifiable EGFP-labelled axons in the sciatic nerve from the crush site to the distal axonal ends. The number of EGFP labeled regenerating axons were actually determined by the electroporation rate of EGFP, which is similar, but not identical, in different mice. Thus, our data only can show the differences in axon lengths among different experimental conditions. Such approach has been used in many of our previously published papers (e.g. Saijilafu et al. Nature Communications, 2011, Saijilafu et al. Nature Communications, 2013). (line 152-153).

(4) In the next set of experiments (presented in Figure 4) authors extend the previous observations in primary cultures from the CNS. For that, they use cortical and hippocampal cultures, and pharmacological and genetic loss-of-function using the above-mentioned strategies. The expected results were obtained in both CNS neurons: inhibition or knockdown of the kinase decreases axon growth, whereas inhibition or knockdown of the phosphatase increases growth. A main weakness in this set is that it is not indicated when (at what day in vitro, DIV) the treatments are performed. This is important to correctly interpret the results, since in the first days in vitro these neurons follow well-characterized stages of development, with characteristic cellular events with relevance to what is being evaluated. Importantly, this would be of value to understand whether the treatments affect axonal specification and/or axonal extension. Although these events are correlated, they imply a different set of molecular events.

The treatments were started from the initial of cell culture period, and this procedure may affect axon specification as the Reviewer point out. However, we mainly focused on axon length in our experiments, thus, for quantification of axon length, neurons with processes longer than twice the diameter of cell bodies were photographed, and the longest axon of each neuron was measured. We revised the manuscript as suggested by the reviewer (line 143-145).

The title of this section is misleading: line 241 "MLCK/MLCP activity regulated axon growth in the embryonic CNS"... the title (and the conclusion) implies that the experiments were performed in situ, looking at axons in the developing brain. The most accurate title and conclusion should mention that the evidence was collected in CNS primary cultures derived from embryos.

We have revised the manuscript as suggested by the reviewer (line 251).

(5) Performing nerve crush injury in CNS nerves (optic nerve and spinal cord), and the local application of PBDu, the author shows contrasting results (Figure 5). In the ON nerve, they can see axons extending beyond the lesion site due to PBDu. On the contrary, the authors fail to observe so in the corticospinal tract present in the spinal cord. The authors fail to discuss this matter in detail. Also, they accommodate the interpretation of the evidence in light of a process known as axon retraction, and its prevention by MLCP inhibition. Since the whole paper is on axon extension, and it is known that mechanistically axon retraction is not merely the opposite of axon extension, the claim needs far more evidence.

Thank you so much for your comments. Compared to optic nerve axons, corticospinal tract axons exhibit a reduced intrinsic axon growth capability. Consequently, we observed that PBDu stimulates optic nerve axon regeneration. However, unfortunately, we did not detect any enhancement in corticospinal tract axons beyond the injury site in SCI following the inhibition of myosin light chain phosphatase (MLCP) with PBDu.

In panel 5F and the supplementary data, the authors mention the occurrence of retraction bulbs, but the images are too small to support the claim, and it is not clear how these numbers were normalized to the number of axons labeled in each condition.

Thank you so much for your comments. In this study, we used a similar method from Ertürk et al. (2007) to quantify the retraction bulb. Both maximum width of the enlarged distal tip of the axon and the width of its immediately adjacent axon shaft was measured. Then, the ratio of these two widths was then calculated. An axonal tip was considered as a retraction bulb if its tip/shaft ratio exceeded 4. Averages number of retraction bulb were calculated from 3 sections in every mice for each group (n=5). (line 187-191).

[Ref] Ertürk A, Hellal F, Enes J, and Bradke F (2007). Disorganized microtubules underlie the formation of retraction bulbs and the failure of axonal regeneration. J. Neurosci 27, 9169–9180. [PubMed:17715353].

(6) The author combines MLCK and MLCP inhibitors with Bleb, trying to verify if both pairs of inhibitors act on the same target/pathway (Figure 6). The rationale is wrong for at least two reasons.<br /> a- Because both lines of evidence point to contrasting actions of NMII on axon growth, one approach could never "rescue" the other.

If MLCK regulates axon growth through the activation of Myosin, the inhibitory effect of ML-7 (an MLCK inhibitor) on axon growth might be influenced by Bleb, a NMII inhibitor. However, our findings reveal that the combination of Bleb and ML-7 does not alter the rate of axon outgrowth compared to ML-7 alone. This suggests that the roles of ML-7 and Bleb in axon growth are independent. It means MLCK may regulates axon growth independent of NMII activity.

b. Because the approaches target different steps on NMII activation, one could never "prevent" or rescue the other. For example, for Bleb to provide a phenotype, it should find any p-MLC, because it is only that form of MLC that is capable of inhibiting its ATPase site. In light of this, it is not surprising that Bleb is unable to exert any action in a situation where there is no p-MLC (ML-7, which by inhibiting the kinase drives the levels of p-MLC to zero, Figure 4A). Hence, the results are not possible to validate in the current general interpretation of the authors. (See 'major concern').

The reported mechanism of blebbistatin is not through competition with the ATP binding site of myosin. Instead, it selectively binds to the ATPase intermediate state associated with ADP and inorganic phosphate, which decelerates the phosphate release. Importantly, blebbistatin does not impede myosin's interaction with actin or the ATP-triggered disassociation of actomyosin. It rather inhibits the myosin head when it forms a product complex with a reduced affinity for actin. This indicates that blebbistatin functions by stabilizing a particular myosin intermediate state that is independent of the phosphorylation status of myosin light chain (MLC).

[Ref] Kovács M, Tóth J et al. Mechanism of blebbistatin inhibition of myosin II. J Biol Chem. 2004 Aug 20;279(34):35557-63. doi: 10.1074/jbc.M405319200.

(7) In Figure 7, the authors argue that the scheme of replating and using ML7 before or after replating is evidence for a local cytoskeletal action of the drug. However, an alternative simpler explanation is that the drug acts acutely on its target, and that, as such, does not "survive" the replating procedure. Hence, the conclusion raised by the evidence shown is not supported.

In our study, we meticulously assessed the neuronal survival rates across various experimental groups. The findings indicate no significant variation in survival rates among the groups. This suggests that the drug treatment exerts no discernible influence on cell viability but primarily modulates axonal elongation."

Author response image 1.

(8) In Figure 8, the authors show that the inhibitory treatments on MLCK and MLCP (ML7 and PRBu) alter the morphology of growth cones. However, it is not clear how this is correlated with axon growth. The authors also mention in various parts of the text that a local change in the growth cone is evidence for a local action/activity of the drug or enzyme. However, the local change<->local action is not a logical truth. It can well be that MLCK and MLCP activity trigger molecular events that ultimately have an effect elsewhere, and by looking at "elsewhere" one observes of course a local effect but is not because the direct action of MLCK or MLCP are localized. To prove true localized effects there are numerous efforts that can be made, starting from live imaging, fluorescent sensors, and compartmentalized cultures, just to mention a few.

About the relationship between growth cone size and its growth rate, the previous published literatures found that a fast-growing axon tended to have small growth cones (Mason C. et al. 1997). A recent study on Aplysia further supports this by noting that growth cones enlarge significantly when axonal elongation halts (Miller and Suter, 2018). Consistent with these findings, our data indicate that inhibiting MLCP with PDBu treatment leads to a reduction in growth cone size, which in turn promotes axon regeneration.

[Ref] Mason CA, Wang LC. Growth cone form is behavior-specific and, consequently, position-specific along the retinal axon pathway. J Neurosci. 1997; 13:1086–1100. [PubMed: 8994063]

[Ref] Miller KE, Suter DM. An Integrated Cytoskeletal Model of Neurite Outgrowth. Front Cell Neurosci. 2018 Nov 26;12:447. doi: 10.3389/fncel.2018.00447. eCollection 2018.

References:

(1) Eun-Mi Hur 1, In Hong Yang, Deok-Ho Kim, Justin Byun, Saijilafu, Wen-Lin Xu, Philip R Nicovich, Raymond Cheong, Andre Levchenko, Nitish Thakor, Feng-Quan Zhou. 2011. Engineering neuronal growth cones to promote axon regeneration over inhibitory molecules. Proc Natl Acad Sci U S A. 2011 Mar 22;108(12):5057-62. doi: 10.1073/pnas.1011258108.

(2) Garrido-Casado M, Asensio-Juárez G, Talayero VC, Vicente-Manzanares M. 2024. Engines of change: Nonmuscle myosin II in mechanobiology. Curr Opin Cell Biol. 2024 Apr;87:102344. doi: 10.1016/j.ceb.2024.102344.

(3) Karen A Newell-Litwa 1, Rick Horwitz 2, Marcelo L Lamers. 2015. Non-muscle myosin II in disease: mechanisms and therapeutic opportunities. Dis Model Mech. 2015 Dec;8(12):1495-515. doi: 10.1242/dmm.022103.

Reviewer #2 (Public review):

Summary:

Saijilafu et al. demonstrate that MLCK/MLCP proteins promote axonal regeneration in both the central nervous system (CNS) and peripheral nervous system (PNS) using primary cultures of adult DRG neurons, hippocampal and cortical neurons, as well as in vivo experiments involving sciatic nerve injury, spinal cord injury, and optic nerve crush. The authors show that axon regrowth is possible across different contexts through genetic and pharmacological manipulation of these proteins. Additionally, they propose that MLCK/MLCP may regulate F-actin reorganization in the growth cone, which is significant as it suggests a novel strategy for promoting axonal regeneration.

Strengths:

This manuscript presents a comprehensive array of experimental models, addressing the biological question in a broad manner. Particularly noteworthy is the use of multiple in vivo models, which significantly strengthens the overall validity of the study.

We thank the Reviewer for taking time to review our manuscript, and we really appreciated the positive comments from the Reviewer.

Weaknesses:

The following aspects apply:

(1) The manuscript initially references prior research by the authors suggesting that NMII inhibition enhances axonal growth and that MLCK activates NMII. However, the study introduces a contradiction by demonstrating that MLCK inhibition (via ML-7 or siMLCK) inhibits axonal growth. This inconsistency is not adequately addressed or discussed in the manuscript.

Thank you for reviewer's very good comments. As suggested by Reviewer, we discuss more detail it in our revised manuscripts (line 357-368; line373-374).

(2) While the study proposes that MLCK/MLCP regulates F-actin redistribution in the growth cone, the mechanism is not explored in depth. The only figure showing how pharmacological manipulation affects the growth cone suggests that not only F-actin but also the microtubule cytoskeleton might be affected, indicating that the mechanism may not be specific. A deeper exploration of this relationship in DRG neurons, in addition to cortical neurons, as shown in the study, would be beneficial.

Thank you for your insightful suggestion. However, our study primarily focuses on actin and myosin dynamics in the context of axonal elongation, as indicated by our direct observations in growing dorsal root ganglia (DRGs). Athamneh et al. (2017) elegantly demonstrated that the bulk movement of microtubules (MTs), rather than their assembly, predominantly drives MT advance during axonal elongation. Consequently, our manuscript concentrates on the actomyosin system, which is central to our findings. While the role of MTs in axonal growth is indeed significant and fascinating, the data we present is predominantly concerned with the actomyosin mechanism.

[Ref] Athamneh, A. I. M. et al. Neurite elongation is highly correlated with bulk forward translocation of microtubules. Scientific Reports 7, (2017).

(3) In the sciatic nerve injury experiments, it would be crucial to include additional controls that clearly demonstrate that siMYPT1 treatment increases MLCP in the L4-L5 ganglia. Additionally, although the manuscript mentions quantifying axons expressing EGFP, the Materials and Methods section only discusses siMYPT1 electroporation, which could lead to confusion.

Thank you for your suggestion. However, due to the unavailability of a suitable commercial MLCP antibody, we were unable to directly detect MLCP expression. Instead, we assessed the phosphorylation level of myosin light chain (MLC) as a proxy to indicate that siMYPT1 transfection effectively downregulates MLCP activity in L4/5 dorsal root ganglia (DRG). This approach was taken to ensure the integrity of our findings despite the limitations in antibody availability.

About the electroporation method section, we have now included detailed information about the control plasmid used in our experiments to ensure a clear understanding of our experimental setup and to validate our results. A 1 μl solution containing indicated siRNAs together with the plasmid encoding EGFP (pCMV–EGFP–N3) was then microinjected into the L4–L5 DRG….. (line 152-153).

(4) In some panels, it is difficult to differentiate the somas from the background (Figures 3, 4, 7). In conditions where images with shorter axonal lengths are represented, it is unclear whether this is due to fewer cells or reduced axonal growth (Figures 2, 4, 6).

In the original submission, there was some loss of image quality while converting the TIFF to PDF. We improved the quality of images in our revised manuscripts.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

There are a number of typos and language errors that should be thoroughly revised. For example, line 219: "It is well known that the opposite role of MLCK and MLCP to regulate the MLC phosphorylation status". The term "opposite role" is vague. Using "opposite roles" and specifying that they are in regulating MLC phosphorylation status clarifies the relationship between MLCK and MLCP. Also, the original phrase "to regulate" was not correctly integrated into the sentence. Rephrasing it to "in regulating" makes the role of MLCK and MLCP clearer.

We have revised the manuscript as suggested by the reviewer (line 229).

In the same line, there is a high number of panels that are not referred to in the text or references for panels that have another letter. Just to mention a few:

- line 199: "(Figure 1F, G)", → BUT figure 1 contains no G panel.

We have revised the manuscript as suggested by the reviewer (line 209).

- line 203: "The results showed that ML-7 administration led to a significant reduction in MLC phosphorylation levels (Figure 2A, B) and impaired axonal growth in sensory neurons (Figure 2C, D). → BUT panel C is related to A and B, and only D and E show impaired axonal growth.

We have revised the manuscript as suggested by the reviewer (line 214; line 215; line 217; line 219 ).

Reviewer #2 (Recommendations for the authors):

(1) Improving the quality of the images would significantly strengthen the results presented.

In the original submission, there was some loss of image quality while converting the TIFF to PDF. We improved the quality of images in our revised manuscripts.

(2) The representative images of controls do not always show the same number of cells or axonal growth (e.g., Figure 4).

We have changed some images as suggested by the reviewer.

(3) The text has citation errors when referring to the figure labels.

Upon thorough review, we have carefully examined our manuscript and have made the necessary corrections to address the identified errors. We appreciate the opportunity to enhance the quality of our work and believe that these revisions have significantly improved the clarity of our manuscript.

(4) What happens to MLCK levels when MLCP activity is inhibited in the optic nerve?

Upon analyzing our experimental data, we observed no significant alterations in the protein levels of MLCK when the activity of MLCP was inhibited. This finding suggests that the regulatory mechanisms governing MLCK expression may not be directly influenced by short-term MLCP inhibition. It is plausible that the duration of the inhibition period was insufficient to elicit a detectable change in MLCK expression levels.

(5) The text in line 266: "In contrast, local PBS administration at the injury site or intravitreal PDBu injection induced little axon regeneration beyond the injury site (Figure 5 A-C)." However, this is not reflected in the figure.

In our revised manuscript, we have provided a more precise description of our findings: In contrast, local PBS administration at the injury site or intravitreal PDBu injection did not significantly enhance axon regeneration beyond the injury site (Figure 5 A-C). This observation suggests that the only treatment employed in the injury site (the inhibition of MLCP activity within the growth cone) effective promote axonal growth. (line 276-279).

(6) Line 287: The phrase "Consistent with our previous study" requires a citation to support it.

We added the reference paper; Consistent with our previous study 1, the inhibition of myosin II activity with 25 μM blebbistatin markedly promoted axonal growth (Figure 6A, B). (line 298)

(7) Line 333: The paper cited by Yu P et al. (2012) does not mention MLCK or p-MLC, so it appears to be misquoted.

Thank you for comments. We rechecked this cited paper and confirmed that the author provided the western data C in the supplementary figure 1, it showed that Bleb did not alter the phosphorylation status of MLC.

eLife Assessment

This study provides important new insights into the contributions of local DNA features to the complex molecular mechanisms and dynamics of copy number variation (CNV) formation during adaptive evolution. While limited to a single CNV of interest, the study is well-designed and carefully controlled, presenting compelling evidence that supports the conclusions. This work will be of general interest to those studying genome architecture and evolution from yeast biologists to cancer researchers.

Reviewer #1 (Public review):

Summary:

The work by Chuong et al. provides important new insights into the contribution of different molecular mechanisms in the dynamics of CNV formation. It will be of interest to anyone curious about genome architecture and evolution from yeast biologists to cancer researchers studying genome rearrangements.

Strengths:

Their results are especially striking in that the "simplest" mechanism of GAP1 amplification (non-allelic homologous recombination between the flanking Ty-LTR elements) is not the most common route taken by the cells, emphasizing the importance of experimentally testing what might seem on the surface to be obvious outcome. One of the important developments of their work is the use of their neural network simulation-based inference (nnSBI) model to derive rates of amplicon formation and their fitness effects.

Weaknesses:

The nnSBI model that derives rates of amplicon formation and fitness is still opaque to this reviewer. All of the other criticisms made in the first review have been clarified/corrected in this much-improved version of the manuscript.

Reviewer #2 (Public review):

Summary:

This study examines how local DNA features around the amino acid permease gene GAP1 influence adaptation to glutamine limited conditions through changes in GAP1 Copy Number Variation (CNV). The study is well motivated by the observation of numerous CNVs documented in many organisms, but difficulty in distinguishing the mechanisms by which they are formed, and whether or how local genomic elements influence their formation. The main finding is convincing and is that a nearby Autonomous Replicating Sequence (ARS) influences the formation of GAP1 CNVs and this is consistent with a predominate mechanism of Origin Dependent Inverted Repeat Amplification (ODIRA). These results along with finding and characterizing other mechanisms of GAP1 CNV formation will be of general interest to those studying CNVs in natural systems, experimental evolution and in tumor evolution. While the results are limited to a single CNV of interest (GAP1), the carefully controlled experimental design and quantification of CNV formation will provide a useful guide to studying other CNVs and CNVs in other organisms.

Strengths:

The study was designed to examine the effects of two flanking genomic features next to GAP1 on CNV formation and adaptation during experimental evolution. This was accomplished by removing two Long Terminal Repeats (LTRs), removing a downstream ARS, and removing both LTRs and the ARS. Although there was some heterogeneity among replicates, later shown to include the size and breakpoints of the CNV and the presence of an unmarked CNV, both marker assisted tracking of CNV formation and modeling of CNV rate and fitness effects showed that deletion of the ARS caused a clear difference compared to the control and the LTR deletion.

The consequence of deletion of local features (LTR and ARS) was quantified by genome sequencing of adaptive clones to identify the CNV size, copy number and infer the mechanism of CNV formation. This greatly added value to the study as it showed that i) ODIRA was the most common mechanism but ODIRA is enhanced by a local ARS, ii) non-allelic homologous recombination (NAHR) is also used but depends on LTRs, and iii) de novo insertion of transposable elements mediate NAHR in strains with both ARS and LTR deletions. Together, these results show how local features influence the mechanism of CNV formation, but also how alternative mechanism can substitute when primary ones are unavailable.

Weaknesses:

The CNV mutation rate and effect on fitness are hard to disentangle. The frequency of the amplified GFP provides information about mutation rate differences as well as fitness differences. The data and analysis show that each evolved population has multiple GAP1 CNV lineages within it, with some being unmarked by GFP. Thus, estimates of CNV fitness are more of a composite view of all CNV amplifications increasing in frequency during adaptation. Another unknown but potential complication is whether the local (ARS, LTR) deletions influence GAP1 expression and thus the fitness gain of GAP1 CNVs. The neural network simulation based inference does a good job at estimating both mutation rates and fitness effects, while also accounting for unmarked CNVs. However, the model does not account for population heterogeneity of CNVs and their fitness effects. Despite these limitations of distinguishing mutation rate and fitness differences, the authors' conclusions are well supported in that the LTR and ARS deletions have a clear impact on the CNV mediated evolutionary outcome and the mechanism of CNV formation.

Reviewer #4 (Public review):

Summary:

Various 'simple' models are used to mechanistically explain the formation of genomic rearrangements, often based on local sequence elements. Here the authors show these models to be lacking for the well characterised GAP1 locus as, although predicted events are observed at reasonable frequency, mutating relevant local sequence elements has surprisingly little impact on the emergence of GAP1 CNV. Rather, a similar set of mechanisms occur (at in some cases somewhat lower frequency) using different genomic elements, the outcome being that that CNV frequency is largely independent of local genomic elements, although this does of course strongly influence the actual structure of the CNVs.

Strengths:

This is a very thorough study of a very complex system.

Weaknesses:

There are limitations as previous reviews have noted, but these are well justified in the revised text and rebuttal

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The work by Chuong et al. provides important new insights into the contribution of different molecular mechanisms in the dynamics of CNV formation. It will be of interest to anyone curious about genome architecture and evolution from yeast biologists to cancer researchers studying genome rearrangements.

Thank you for recognizing the broad significance of our study.

Strengths:

Their results are especially striking in that the "simplest" mechanism of GAP1 amplification-non-allelic homologous recombination between the flanking Ty-LTR elements is not the most common route taken by the cells, emphasizing the importance of experimentally testing what might seem on the surface to be obvious answers. One of the important developments of their work is the use of their neural network simulation-based inference (nnSBI) model to derive rates of amplicon formation and their fitness effects.

We agree with this assessment as the results of our study challenge our intuition that the simplest path to structural variation is the most likely and reveals the great diversity in mechanisms that can lead to large scale changes in the genome.

Weaknesses:

The manuscript reads as though two different people wrote two different sections of the manuscript - an experimental evolutionist and a computational scientist. If the goal is to reach both groups of readers, there needs to be more explanation of both types of work. I found the computational sections to be particularly dense but even the experimental sections need clearer explanations and more specific examples of the rearrangements found. I will point out these areas in the detailed remarks to the authors. While I have no reason to question their conclusions, I couldn't independently verify the results that ODIRA was the majority mechanism since the sequence of amplified clones was not made available during the review. I've encouraged the authors to include specific, detailed sequence information for both ODIRA events as well as the specific clones where GAP1 was amplified but the flanking gene GFP was not.

We have revised the manuscript to expand explanations of both the experimental and computational aspects of our study and to provide additional information for the reader. In doing so, we have edited the text to improve readability. We have made all raw data publicly available through the NCBI short read archive (SRA) and are hosting all sequence data for easy visualization in JBrowse using a public server.

Reviewer #2 (Public Review):

Summary:

This study examines how local DNA features around the amino acid permease gene GAP1 influence adaptation to glutamine-limited conditions through changes in GAP1 Copy Number Variation (CNV). The study is well motivated by the observation of numerous CNVs documented in many organisms, but difficulty in distinguishing the mechanisms by which they are formed, and whether or how local genomic elements influence their formation. The main finding is convincing and is that a nearby Autonomous Replicating Sequence (ARS) influences the formation of GAP1 CNVs and this is consistent with a predominate mechanism of Origin Dependent Inverted Repeat Amplification (ODIRA). These results along with finding and characterizing other mechanisms of GAP1 CNV formation will be of general interest to those studying CNVs in natural systems, experimental evolution, and in tumor evolution. While the results are limited to a single CNV of interest (GAP1), the carefully controlled experimental design and quantification of CNV formation will provide a useful guide to studying other CNVs and CNVs in other organisms.

Thank you for this positive assessment of our study.

Strengths:

The study was designed to examine the effects of two flanking genomic features next to GAP1 on CNV formation and adaptation during experimental evolution. This was accomplished by removing two Long Terminal Repeats (LTRs), removing a downstream ARS, and removing both LTRs and the ARS. Although there was some heterogeneity among replicates, later shown to include the size and breakpoints of the CNV and the presence of an unmarked CNV, both marker-assisted tracking of CNV formation and modeling of CNV rate and fitness effects showed that deletion of the ARS caused a clear difference compared to the control and the LTR deletion.

The consequence of deletion of local features (LTR and ARS) was quantified by genome sequencing of adaptive clones to identify the CNV size, copy number and infer the mechanism of CNV formation. This greatly added value to the study as it showed that i) ODIRA was the most common mechanism but ODIRA is enhanced by a local ARS, ii) non-allelic homologous recombination (NAHR) is also used but depends on LTRs, and iii) de novo insertion of transposable elements mediate NAHR in strains with both ARS and LTR deletions. Together, these results show how local features influence the mechanism of CNV formation, but also how alternative mechanisms can substitute when primary ones are unavailable.

We agree with this assessment.

Weaknesses:

The CNV mutation rate and its effect on fitness are hard to disentangle. The frequency of the amplified GFP provides information about mutation rate differences as well as fitness differences. The data and analysis show that each evolved population has multiple GAP1 CNV lineages within it, with some being unmarked by GFP. Thus, estimates of CNV fitness are more of a composite view of all CNV amplifications increasing in frequency during adaptation. Another unknown but potential complication is whether the local (ARS, LTR) deletions influence GAP1 expression and thus the fitness gain of GAP1 CNVs. The neural network simulation-based inference does a good job at estimating both mutation rates and fitness effects, while also accounting for unmarked CNVs. However, the model does not account for the population heterogeneity of CNVs and their fitness effects. Despite these limitations of distinguishing mutation rate and fitness differences, the authors' conclusions are well supported in that the LTR and ARS deletions have a clear impact on the CNV-mediated evolutionary outcome and the mechanism of CNV formation.

While it is true that the inferred mutation rate and fitness effect are negatively correlated, as in other studies (Gitschlag et al., 2023; Caspi et al., 2023; Avecilla et al., 2022), our modeling approach does generate an estimate of each parameter that is best explained by the data. By reporting the confidence intervals (i.e. the 95% HDI) we define the set of parameter values that are consistent with the data. It is true that our model doesn't explicitly account for population heterogeneity; rather, following Hegreness et al. (2006), we employ a single effective fitness effect and mutation rate for all GAP1 CNVs. It is interesting to consider whether the ARS and LTR affect GAP1 expression; however, we have no evidence that this is the case.

Reviewer #3 (Public Review):

Summary:

The authors represent an elegant and detailed investigation into the role of cis-elements, and therefore the underlying mechanisms, in gene dosage increase. Their most significant finding is that in their system copy number increase frequently occurs by what they call replication errors that result from the origin of replication firing.

The authors somewhat quantitatively determine the effect of the presence of a proximal origin of replication or LTR on the different CNV scenarios.

Strengths:

(1) A clever and elegant experimental design.

(2) A quantitative determination of the effect of a proximal origin of replication or LTR on the different CNV scenarios. Measuring directly the contribution of two competing elements.

(3) ODIRA can occur by firing of a distal ARS element.

(4) Re-insertion of Ty elements is interesting.

We agree that these are interesting and novel findings from our study.

Weaknesses:

(1) Overall, the research does not considerably advance the current knowledge. The research does not investigate what the maximum distance between ARS for ODIRA is to occur. This is an important point since ODIRA was previously described. A considerable contribution to the field would be to understand under what conditions ODIRA wins NAHR.

We agree that these are important questions and they are ones that we are pursuing in future studies.

(2) The title and some sentences in the abstract give a wrong impression of the generality and the novelty of the observations presented. Below are some examples of much earlier work that dealt with mechanisms of CNV and got different conclusions. The Lobachev lab (Cell 2006) published a different scenario years ago, with a very different mechanism (hair-pin capped breaks). The Argueso lab found something different (NAHR) (Genetics 2013).

In fact, the CUP1 system presents a good example of this point. The Houseley group showed a complex replication transcription-based mechanism (NAR 2022, cited), the Argueso group showed Ty-based amplification and the Resnick group showed aneuploidy-based amplification. While aneuploidy is a minor factor here the numerous works in Candida albicans, Cryptococcus neoformans, and Yeast suggest otherwise (Selmecki et al Science 2006, Yona et al PNAS 2013, Yang et al Microbiology Spectrum 2021).

As the reviewer points out there have been several important published studies investigating mechanisms by which structural variation is generated. It is important to note that we are explicitly looking at CNVs in the context of adaptive evolution and the role of genomic features that enable different mechanisms of CNV formation. To emphasize this point, we have changed the title of our manuscript to “Template switching during DNA replication is a prevalent source of adaptive gene amplification”. Aneuploidy is indeed a mechanism of adaptive gene amplification in our current and previously reported studies. We have expanded our discussion to place our study in the context of previous studies reporting mechanisms of gene amplification.

(3) The authors added a mathematical model to their experimental data. For me, it was very difficult to understand the contribution of the model to the research. I anticipated, for example, that the model would make predictions that would be tested experimentally. For example, " ARSΔ and ALLΔ are predicted to be almost eliminated by generation 116, as the average predicted WT proportion is 0.998 and 0.999" But to my understanding without testing the model.

In our previous publication (Avecilla et al. 2022, PLoS Biology) we experimentally validated the use of nnSBI to infer evolutionary parameters. In this study, we have extended our modeling framework to quantify differences between genotypes, which was not previously possible. Our results reveal that the local ARS has a key role in the overall supply rate of CNVs at this locus.

Recommendations for the authors:

We have addressed all public reviews and recommendations.

Reviewer #1 (Recommendations For The Authors):

Specific comments about the work are covered in the order of appearance in the text or Figures. I apologize in advance for the number of comments. They are made out of curiosity, enthusiasm for the research, and a desire to help highlight the most interesting aspects of this work.

We are grateful for the thoughtful comments that have helped us to significantly improve our manuscript.

(1) I would appreciate the inclusion of several references to the work on the ODIRA model.

a) Page 3 last paragraph: "(2) DNA replication-based mechanisms (Harel et al., 2015; Hastings, Lupski, et al., 2009; Malhotra & Sebat, 2012; Pös et al., 2021; Zhang, Gu, et al., 2009; Brewer et al., 2011)" (Addition of Brewer et al., 2011).

We have added all suggested references.

b) Page 4 top: (Brewer et al., 2011; Brewer et al., 2015; Martin et al., 2024). (Addition of Brewer et al., 2011).

We have added all suggested references.

c) Page 14 top: "Recent work has proposed that ODIRA CNVs are a major mechanism of CNVs in human genomes (Brewer et al., 2015; Martin et al., 2024; Brewer et al., 2024)." Brewer et al., 2024 focuses specifically on ODIRA and human CNVs. (Addition of Brewer et al., 2024).

We have added all suggested references.

(2) Page 6, third paragraph: I was surprised that a single inoculating strain was used to establish the replicate chemostats because of the possibility of non-independence of the resulting GAP1 CNVs. A nnSBI model was used to correct for this possibility later in the paper. It seems like it could have been avoided by a simple change in protocol to inoculate each chemostat with an independent inoculum. Was there a reason that the replicate chemostats were not conducted as independent events? Establishing the presence of 'founder' GAP1 CNVs without GFP seems rather secondary to the point of the paper (examining the CNVs that arise during evolution) and I would recommend it being moved to the supplement.

As is typical in microbial experimental evolution studies, we aimed to start with genetically identical homogenous populations and observe the emergence and selection of de novo variation. Therefore, we founded independent populations from a single inoculum. However, this study, and our prior work using lineage tracking barcodes, has clearly demonstrated that during the initial growth of the culture used for the inoculum CNVs are generated that contribute to the adaptation dynamics on all derived populations. This unanticipated result now suggests that the reviewer’s suggestion is a valid one - independent populations should be derived from independent inocula and this will be our standard practice in future studies.

We believe that our results, presented in Figure 2, establishing the presence of pre-existing GAP1 CNVs without the GFP are important as it highlights a limitation of the use of CNV reporters of gene copy number that was not previously known. However, we subsequently show that this class of variant - CNVs that are not detected by the reporter system - can be incorporated into our modeling framework enabling estimation of evolutionary parameters, which we believe is an important finding warranting inclusion in the main text.

(3) Page 7 first full paragraph: "Finally, we also observe a significant delay (ANOVA, p = 0.00833) in the generation at which the CNV frequency reaches equilibrium in ARS∆ (~generation 112) compared to WT (pairwise t-test, adjusted p = 0.05) . . .". Is the delay in reaching a plateau in Figure 1E just a consequence of the later appearance of CNVs or do the authors believe there are two separate events responsible for this delay? E.g. if the authors think that the delay in reaching a plateau is related to lower selection coefficients of the CNVs that do arise compared to the CNVs of other strains, then this should be explicitly discussed.

We believe that the delay in reaching equilibrium is a consequence of both a lower CNV formation and reduced selection coefficients. Lower values for the fitness coefficient and formation rate in ARS∆ explain both the delay in CNV appearance and CNV equilibrium as shown by the predicted dynamics (Figure S3B). We have added an explicit discussion of the effect of the ARS on CNV dynamics in paragraph 2 of the Discussion section paragraph 2 starting at line 456.

(4) Page 7: Incorporating pre-existing CNVs into an evolutionary model: The rationale for how you are able to discount the formation rate of GFP-free CNVs (C-) in your model isn't clear to me. How are you able to assume that these C- events don't form after timepoint 0? Why do you assume a starting population of C- events but not a starting population of C+ events?

We explored the possibility of modeling C- (amplifications of GAP1 without amplification of the reporter) during the evolution experiment. However, because the rate at which C- events occurs is slower than the rate at which C+ events occur (GAP1 amplifications with amplification of the reporter) we found that the effect was negligible. Importantly, the simple model is sufficient to describe the observed dynamics and thus we do not include these possible rare events.

(5) Figure 1:

(a) Panel B: Please put the tRNAs on the line diagrams of the four strains. I first interpreted ALLΔ as missing the tRNAs, too.

Thank you for this suggestion. We added tRNAs to all diagrams to provide additional detail about the structure of the GAP1 locus.

(b) Panels C, D, and E: the dark shade of the colored boxplots obscures the individual points. I recommend reducing the opacity of the box or choosing a lighter shade so that the individual points are visible on top of the box. Is the percent increase in CNVs per generation (Panel D) based on the slopes of the curves in panel B? By eye the slopes of ARS∆ and ALL∆ appear at least as steep as those of wild type and LTR∆.

Thank you for this suggestion. We have now made the individual points visible on top of the boxplots in Figures 1C, 1D, and 1E. The lines in Figure 1B show the median value across populations per time point whereas each point in Figure 1D is the slope from linear regression using values from individual populations (data from individual populations are shown in Figure 3C).

(6) Figure 2:

(a) Panel A: Please remind the readers what FSC-A is measuring and label the different groups of cells in each sample. Are we supposed to assume the upper scatter in generation 8 is the pre-existing CNV variants? Are the three species at generation 50 due to 1, 2, and 3 copies of GFP? Is the new species in generation 137 further amplification of the locus? And if so, how many copies does it represent? I find it fascinating that what I assume is the 2-copy CNV (presumably a direct oriented amplicon produced by NAHR) at 50 generations is lost (out-competed by a potential inverted triplication) at later times, but I didn't find any mention of this phenomenon in the text. What do the different mutant strains look like over the same time course? Please supply supplemental figures with the flow cytometry gating and vertically aligned histograms of the GFP signal so that the peaks are more easily compared. And provide this information for each of the altered strains in supplementary materials.

Thank you for these useful suggestions. We have added a gating legend to the figure to clearly indicate the copy-number for each subpopulation. We have edited the caption and main text to explain forward scatter (FSC-A). Raw flow cytometry plots are now provided as Supplementary figure 2 and distributions of cell-size normalized GFP signal are provided in Supplementary figure 3. Although our primary objective with Figure 2A was to show the persistence of the 1-copy GFP population the reviewer is correct that we did not highlight interesting aspects of the CNV dynamics. We have added additional text starting at line 251 to point out these features of the data.

(b) Panel B: It would help to label the different colored boxes inside cells in Figure 2B - it took me a while to identify the white box as an unrelated adaptive mutation elsewhere in the genome. The linear arrangement of these small colored blocks seems to indicate their structural arrangement. Is that the case? And are they inverted or direct amplicons? Perhaps the authors are being agnostic at this point but it would be better if each of the blocks were separate. If there are other mutations that can explain these GFP-non-amplified survivors, were they identified in your whole genome sequencing?

We have now included a complete legend for Figure 2B indicating that the white box reflects other beneficial mutations. We have separated this class of beneficial mutation from the GAP1 and reporter elements to reflect that they are not linked. We did not identify additional beneficial mutations but plan to pursue this question in a future project.

(c) Panel C: Are the two sets of lines mislabeled? One would expect the "reported" CNV proportions to be lower than the total CNV proportions, not the other way around. Maybe the labels "total CNVs" and "reported CNVs" are unclear to me and I am misunderstanding what "reported" refers to. Please clarify.

Thank you for identifying this mistake. The lines were mislabeled and have now been corrected in the revised version.

(7) Figure 3:

(a) A fuller discussion of panels A and B is needed. The results of panel A in particular seem like an excellent opportunity for connecting the computation to the biology. Can the authors speculate on why the ALL∆ strain has a higher CNV formation rate (𝛿c) than the ARS∆ strain? I would think that taking away one means of amplification would decrease CNV formation. Likewise, could the authors discuss why the selection coefficient (sc) for the LTR∆ strain would be the same as for the wild type? Overall, I would like to see more discussion about what these differences in formation rates and selection coefficients could mean for the types of amplicons arising in the chemostats. (In panel B I don't see the shaded area referred to in the figure legend.) A side-by-side comparison of the data in Panel A with the data shown in Supplemental Figure S3A would be instructive..

Thank you for raising these points. We have added substantial text to the manuscript to address these findings. Starting at line 456 we state:

“The lower CNV formation rate in the LTR∆ could be a closer approximation of ODIRA formation rates at this locus as ODIRA CNVs are the predominant CNV mechanism in the LTR∆ strain (Figure 4F). Furthermore, the low formation rates in the LTR∆ relative to WT might suggest that the presence of the flanking long terminal repeats may increase the rate of ODIRA formation through an otherwise unknown combinatorial effect of DNA replication across these flanking LTRs and template switching at the GAP1 locus. ARS∆ has the lowest CNV formation rate and it could be an approximation of the rates of NAHR between flanking LTRs and ODIRA at distal origins. We find that the ALL∆ has a higher CNV formation rate than the ARS∆, even though three elements are deleted instead of one. One explanation for this is that the deletion of the flanking LTRs in ALL∆ gives opportunity for novel transposon insertions and subsequent LTR NAHR. Indeed we find an enrichment of novel transposon-insertions in the ALL∆ (Figure 4F) and subsequent CNV formation through recombination of the Ty1-associated repeats (Figure 4H, ALL∆). Both events, transposon insertion followed by LTR NAHR, would have to occur quickly at a rate that explains our estimated CNV rate in ALL∆. While remarkable, increased transposon activity has been associated with nutrient stress (Curcio & Garfinkel, 1999; Lesage & Todeschini, 2005; Todeschini et al., 2005) and therefore feasible explanation for the CNV rate estimated in the ALL∆. Additionally, ARS∆ clones rely more on LTR NAHR to form CNVs (Figure 4F). The prevalence of ODIRA in ARS∆ and ALL∆ are similar. LTR NAHR usually occurs after double strand breaks at the long terminal repeats to give rise to CNVs (Argueso et al., 2008). Because we use haploid cells, such double strand break and homology-mediated repair would have to occur during S-phase after DNA replication with a sister chromatid repair template to form tandem duplications. Therefore the dependency on LTR NAHR to form CNVs and the spatial (breaks at LTR sequences) and temporal (S-phase) constraints could explain the lower formation rate in ARS∆.”

In addition, we added a discussion of the different selection coefficients estimated and how the simulated competitions help us understand the decreased selection coefficients in the architecture mutants. In newly added text starting at line 479 we state:

“The genomic elements have clear effects on the evolutionary dynamics in simulated competitive fitness experiments. The similar selection coefficients in WT and LTR∆ suggest that CNV clones formed in these background strains are similar. Indeed, the predominant CNV mechanism in both is ODIRA followed by LTR NAHR (Figure 4F). While LTR NAHR is abolished in the LTR∆, it seems that CNVs formed by ODIRA allow adaptation to glutamine-limitation similar to WT. The lower selection coefficients in ARS∆ and ALL∆ suggest that GAP1 CNVs formed in these strains have some cost. In a competition, they would get outcompeted by CNV alleles in the WT and LTR∆ background.”

(b) The data shown in panel C seems redundant to what is shown more clearly in Supplemental Figure S3B. It seems to me the more important comparison to make in panel C would be the overlay of the predicted data to the median proportion of cells obtained from the experimental data (Figure 1B). Also, overlays of the cultures from each strain could be added to S3A. It is difficult to see the variation within each strain when the data from all four strains are superimposed as they are in Figure 3C.

We agree and have edited Figure 3C to incorporate these suggestions and more clearly convey the intra- and interstrain variation.

(8) Figure 4:

(a) Panels A, B, and C are nice summaries and certainly helpful for understanding panel E, but it would be instructive to see some actual rearrangements of the ODIRA events, the NAHR, and the transposon-mediated rearrangements. It isn't clear to me what these last events look like. A figure that shows the specific architecture of example clones for each category would be helpful. I am also having a hard time reconciling ODIRA events with a copy number of 2. Are these rearrangements free isochromosomes with amplification to the telomere or are they secondary rearrangements like those described in Brewer et al., 2024? And what about the non-aneuploid rearrangement that includes the centromere? Is it a dicentric?

We have now added more detailed depictions of CNVs in Figure 4A and provide links to visualize the alignment files. We have added additional discussion starting at line 397 of the non-canonical ODIRA events and putative neochromosome amplicons with reference to Brewer et al 2024. Starting at line 397 we state:

“Surprisingly, we found CNVs with breakpoints consistent with ODIRA that contained only 2 copies of the amplified region, whereas ODIRA typically generates a triplication. In the absence of additional data, we cannot rule out inaccuracy in our read-depth estimates of copy numbers for these clones (ie. they have 3 copies). An alternate explanation is a secondary rearrangement of an original inverted triplication resulting in a duplication (Brewer et al., 2024); however, we did not detect evidence for secondary rearrangements in the sequencing data. A third alternate explanation is that a duplication was formed by hairpin capped double-strand break repair (Narayanan et al., 2006). Notably, we found 3 additional ODIRA clones that end in native telomeres, each of which had amplified 3 copies. In these clones the other breakpoint contains the centromere, indicating the entire right arm of chromosome XI was amplified 3 times via ODIRA, each generating supernumerary chromosomes. Thus,ODIRA can result in amplifications of large genomics regions from segmental amplifications to supernumerary chromosomes.”

(b) In Panel B the violin plots appear to indicate that there are two size categories for amplicons in the ARS∆ strain. Do clones from these different sub-populations share a common CNV architecture?

Thank you for making this point. (Please note that the violin plots are now Figure 4E) We added a short discussion and Supplementary Figure 14. In line 432, we state:

“In ARS∆, we find two CNV length groups (Figure 4E) that correspond with two different CNV mechanisms (Supplementary Figure 14). 100% of smaller CNVs (6-8kb) (Supplementary Figure 14) correspond with a mechanism of NAHR between LTRs flanking the GAP1 gene (Figure 4H, ARS∆, bottom left green points). Larger CNVs (8kb-200kb) (Supplementary Figure 14) correspond with other mechanisms that tend to produce larger CNVs, including ODIRA and NAHR between one local and one distal LTR element (Figure 4H).”

(c) Panels D and E: There is great information in these two panels but I find the color keys confusing. There doesn't seem to be any reason for the strain color key in panel E. I am assuming that the key should go with Panel D. Is there some way to indicate in Panel D which events are in which CNV category? It is cumbersome to find that information from Panel E. Perhaps the color-coding from Panel E could be applied to the row labels in Panel D. Being able to link amplicon to the mechanism of CNV formation is especially important for seeing which ODIRA events contain an origin.

Thank you for this suggestions. We now indicate the mechanism of CNV formation using a consistent color coding in panels G and H (previously panels D and E).

(d) Panel E: I don't understand the two axes in Panel E. If both axes are log scales, why is the origin 0 for the X-axis and 1 for the Y-axis? And why are the focal amplicons (most of which are recombination events between the two LTRs) scattered in both X and Y coordinates? Shouldn't they form a single point? The same for the recombinants with distal LTRs. Also, orange and red (ODIRA and complex CNVs, respectively) are very hard to distinguish. All of these data need to be presented in a spreadsheet identifying each clone's strain ID, chemostat number, GAP1 and GFP copy numbers, sequence across the junction, and their coordinates. The SRA project (PRJNA1016460) for the sequence data was not found in SRA. Will this data be available to easily look at read depth across chromosome XI for all of the sequenced strains - perhaps as .bam files?

Thank you for calling these issues with data visualization to our attention. Indeed, the focal amplifications do form around a single point. We originally had jittered the data to show each individual focal amplification but agree that this is confusing. We now overlay the individual points and have altered opacity to enable visualization of individual values. The suggested table of clone data is provided in Supplementary File 2 and the SRA project is now publicly available. Moreover, we are providing all alignment (.bam) files, split, and discordant read depth profiles for each CNV strain and their corresponding ancestor aligned to our custom reference genomes in a public jbrowse server at:

https://jbrowse.bio.nyu.edu/gresham/?data=data/ee_gap1_arch_muts for WT strains, https://jbrowse.bio.nyu.edu/gresham/LTRKO_clones for LTR∆ strains, https://jbrowse.bio.nyu.edu/gresham/ARSKO_clones for ARS∆ strains, https://jbrowse.bio.nyu.edu/gresham/ALLKO_clones for ALL∆ strains.

(e) Supplementary Table 1 and Supplementary Figure S2: Please indicate which rearrangements (of the 8 reported in Figure S2A) were identified in each of the clones described in the table. If each of the 8 amplicons is identified by a letter, then this information could be added as a column in the table. I am assuming that each of the eight rearrangements was found in more than one chemostat. Showing these data is crucial for establishing the possibility that they were preexisting at the time of chemostat inoculation. The other possibility is that the clones with amplified GAP1 but a single copy of GFP could have been created by a secondary rearrangement in the outgrowth of the clones that originally had amplified both genes to the same extent. What is the structure of these amplicons? Is there a common junction between GAP1 and GFP? I couldn't find these data in the paper. A suggestion for Supplemental Figure S2A - include a zoomed-in inset for the GAP1 GFP region for each of the 8 read-depth plots. It is hard to see the exact location of GFP and GAP1 across all 8 tracks without getting out a ruler. Were these sequences aligned to your custom reference genome or the reference genome without GFP? If they were aligned to the custom reference that includes the GFP reporter, the reader could visually confirm the absence of GFP amplification.

Thank you for these suggestions. We edited Supplementary Table 1 and Supplementary Figure 1A as requested. We now provide the precise CNV breakpoints in the GFP-GAP1 region (supplemental figure 1B) displaying both genome read depth and split read depth tracks. These sequences were aligned to the custom reference containing the GFP reporter, which is now clearer in the figure and caption text in line 1226.

The clones in this figure were sampled from the five different chemostats and we have clarified this in the edited table and text at line 210. We did not detect the same CNV allele in different chemostats and therefore we do not have evidence to support GAP1 amplification without the GFP reporter pre-existing at time of inoculation. We are not able to definitively distinguish whether the amplicons were pre-existing at the time of inoculation or occurred after as we do not have barcoded lineages. We isolated clones carrying this class of amplification from the 1-GFP-copy subfraction late in the experimental evolution (generation 165-182). Given that the alleles appear to differ between populations we think the most parsimonious explanation is that these amplifications occurred after chemostat inoculation but early in the evolution experiment. We explicitly state this in the text starting in line 219.

(9) Page 8-9: I am sorry to say that I can't evaluate the "HDI of posterior distributions". It is out of my competency range. So I am not sure what this analysis is adding to the paper. The same goes for the rest of the supplementary figures.

HDI is a measure of certainty in an estimate, similar to confidence interval. We state this in the text in line 276. With the editing of the text we hope the modeling and its supplementary figures are more clear now.

(10) Page 9 top: Deletion of the ARS appears to lower the fitness of the amplified GAP1 variants. Can the authors speculate on why the ARS deletion would reduce fitness? Did they consult published replication profiles to determine the size of the origin-free gap that could result from the deletion of this mid-S phase origin? Could it explain the delay in the appearance of GAP1 amplicons in the ARS-deletion strains and be responsible for their reduced selection coefficients? Did you examine the growth properties of the starting strain or any of the amplified GAP1 derivatives? Perhaps this consideration could contribute to the discussion. Could there be a bit fuller discussion on the interaction between CNV length differences as shown in Figure 4A and differences in selection coefficient as determined by the nnSBI?

Thank you for raising this point. We have now added text to our discussion of the reduced fitness in ARS∆ in relation to DNA replication starting on line 359:

“ARS1116 is a major origin (McGuffee et al., 2013) and ODIRA CNVs found around this origin corroborate its activity. GAP1 is highly transcribed in glutamine-limited chemostats (Airoldi et al., 2016). Head-on transcription-replication collisions at this locus may be contributing to the higher CNV formation rate in wild type and LTR∆. Elimination of the local ARS could result in less transcription-replication collisions and the slower CNV formation rates estimated. Once formed they get outcompeted by faster-forming CNVs and thus in theory are less fit than CNVs in other strain backgrounds. These simulated competitions further suggest that the ARS is a more important contributor to adaptive evolution mediated by GAP1 CNVs.”

We examined replication profiles in McGuffee et al. Mol Cell. 2013 but could not determine the size of the origin-free gap. ARS1116 and its neighboring ARSs, ARS1118 downstream and ARS1115 upstream are efficient firing origins (Supplement 1 of McGuffee et al. 2013) and therefore the gap is likely to be minimal. The dynamics of the distal firing ARS elements involved in creating ODIRA CNVs might explain the reduced fitness, but further experiments would be required to address this. Regarding growth properties, the growth rate at steady-state in the chemostat is the same as the dilution rate regardless of strain background. Because we had the same dilution rate for each chemostat, the ARS∆ populations would have the same replication rate as the other three strains even if there may be replication rate differences in bulk culture growth. Finally, we found no significant interaction between CNV length and selection coefficients and we state this in line 359.

(11) Page 10: WT competition simulations: It may help to explicitly state that the competition modeling approach was experimentally validated in Avecilla 2022 as opposed to just citing the paper. I found the results much more convincing after reading Avecilla 2022, but I imagine many readers may skip that.

We added a sentence to state that the nnSBI method was experimentally validated in Avecilla et 2022 at line 249.

Reviewer #2 (Recommendations For The Authors):

(1) Figure 2: says reported CNV proportions (dashed). This may be a typo since I think the GFP reported should be solid, not dashed. Also, (C) isn't bold.

Thank you for identifying these mistakes. We have corrected the figure’s caption in line 1157.

(2) "compared to 898/345 clones" Does this refer to transposition/clone? Seems more natural to compare clones with transpositions to a total number of clones. This could be clarified.

We rephrased the sentence (lines 519-520) to clarify that in their study Hays et al. 2023 found 898 novel Ty insertions across 345 nitrogen-evolved clones. As a result of this high rate of transposition, some clones are expected to have multiple Ty insertions.

(3) The methods state that Kan replaces the Nat cassette that was used to make the deletions. It should be made more clear whether Kan is present and where Kan is with respect to GFP and GAP1.

Thank you for pointing this out. To clarify we added the following sentence to the methods starting in line 567:

“The CNV reporter is 3.1 kb and located 1117 nucleotides upstream of the GAP1 coding sequence. It consists of, in the following order, an ACT1 promoter, mCitrine (GFP) coding sequence, ADH1 terminator, and kanamycin cassette under control of a TEF promoter and terminator.”

Additionally in line 571 we clarify the drug resistance of the genomic architecture ∆ strains that are kanamycin(+) and nourseothricin(-).

Reviewer #3 (Recommendations For The Authors):

(1) The major advancement of the manuscript is stated in the title "DNA replication errors are a major source of adaptive gene amplification" First, in my humble opinion the term replication errors is not quite right; the term template switching is more accurate. In that regard, recently a paper was published just on this topic (Martin et al Plos Genetics, 2024).

We have changed the title to “Template-switching during DNA replication is a prevalent source of adaptive gene amplification”. We cite Martin et al Plos Genetics 2024 throughout the main text in lines 93, 126, 159, 502, 555.

(2) I find the statement "We find that 49% of all GAP1 CNVs are mediated by the DNA replication-based mechanism Origin Dependent Inverted Repeat Amplification (ODIRA) regardless of background strain." Somewhat misleading, there were considerable differences between the strains. If I am not mistaken the range was 20-80%.

Thank you for pointing this out. Indeed, the range was 26-80% across the four strains. We updated this sentence in the abstract at line 40, and in the main text at line 141 to clearly state the range.

(3) In their attempt to fill the gap of knowledge regarding the fitness effect of the adaptive CNV the authors use a mathematical model. As an experimental biologist, I found the description lacking. It is hard for me to evaluate the contribution of the model to understanding the results and I think the authors could improve this part.

We have edited the text regarding the modeling and associated results and hope that it is now more clear. The mathematical model describes the experiment in a simplified manner. We use it to predict the outcomes of additional experiments without additional experimental work. For example, we used it to simulate a competition between two strains, predict the total proportion of GAP1 CNVs, and predict the relative genetic diversity.

(4) Experiments the authors may want to consider to increase the novelty of their work:

a) Place the GAP1 gene right in the middle of the two most distant ARS elements and test the mechanism of CNV.

Thank you for this proposed experiment. It is beyond the scope of this paper and will be pursued in future studies.

b) The finding of de-novo Ty element insertion is interesting. What happens if the overdose strain of Jef Boeke is used (Retrotransposon overdose and genome integrity, PNAS 2009) or in contrast, a reverse transcriptase deficient strain?

We agree. Our study has revealed a critical role for novel Ty insertion in mediating CNVs. The suggested experiments as well as using strains that lack Ty sequences will be very interesting to explore in followup studies.

c) The genomic analyses were based on single colony isolates. To my understanding, the CNV events are identified at least partly by split reads. Therefore, each event may have a "signature" that is unique and can be concluded from single reads and not necessarily from the assembled genome. If true, a distinction between the scenarios could be achieved if bulk cultures are sequenced with enough depth. Thus, a truly dynamic and quantitative determination of the different events, rate of appearance, and disappearance can be made.

Thank you for this suggestion, which is a good idea but not currently feasible for several reasons. First, although split reads are a powerful way to detect CNV breakpoints, we have found that even at high coverage (21-153X, median 78.5X), in clonal samples that are rare with only 3-30 split reads (median 14) detected. These observations are from a total of 23 breakpoints across 16 sequenced clones. Thus, when sequencing heterogeneous cultures, in which different CNVs only comprise a fraction of the population, our ability to detect single CNV alleles by split reads and quantify their frequency is limited. Given our observations, with a median of 14 split reads when sequencing to 78.5X genome-wide read coverage it is possible we may be able to detect an individual CNV allele once it makes up (14/78.5) 17% of the population. However, our previous study has shown that there are tens to hundreds of unique CNV alleles initially and thus this would only be feasible at very late timepoints. Second, recurrent CNVs may occur independently at the same exact location, such as LTR NAHR. Thus, unique signatures may not be obtained even if they are independent events. Third, it would be not appropriate to pursue this analysis with our current dataset, as we lack lineage tracking barcodes to validate the results.

eLife Assessment

This important work by Veneditto and colleagues developed a new modeling approach, called a mixture-of-agent hidden Markov model (MoA-HMM), in which choice behaviors are modeled as transitions between discrete states defined by different weighting of several reinforcement learning and decision strategies. The authors apply this approach to their previous data collected from rats performing the two-step task, and show that this method predicts fluctuations in neural and other behavioral data and provides better fits to the data than previous methods. The revision has greatly improved the manuscript, the evidence supporting the conclusions is convincing, and the method is of general interest to the field.

Reviewer #1 (Public review):

Summary:

Motivated by the existence of different behavioral strategies (e.g. model-based vs. model-free), and potentially different neural circuits that underlie them, Venditto et al. introduce a new approach for inferring which strategies animals are using from data. In particular, they extend the mixture of agents (MoA) framework to accommodate the possibility that the weighting among different strategies might change over time. These temporal dynamics are introduced via a hidden Markov model (HMM), i.e. with discrete state transitions. These state transition probabilities and initial state probabilities are fit simultaneously along with the MoA parameters, which include decay/learning rate and mixture weightings, using the EM algorithm. The authors test their model on data from Miller et al., 2017, 2022, arguing that this formulation leads to (1) better fits and (2) improved interpretability over their original model, which did not include the HMM portion. Lastly, they claim that certain aspects of OFC firing are modulated by internal state as identified by the MoA-HMM.

Strengths:

The paper is very well written and easy to follow, especially for one with a significant modeling component. Furthermore, the authors do an excellent job explaining and then disentangling many threads that are often knotted together in discussions of animal behavior and RL: model-free vs. model-based choice, outcome vs. choice-focused, exploration vs. exploitation, bias, perserveration. Each of these concepts is quantified by particular parameters of their models. Model recovery (Fig. 3) is convincing post-revision and licenses their fits to animal behavior later. While the specific claims made about behavior and neural activity are not especially surprising (e.g. the animals begin a session, in which rare vs. common transitions are not yet known, in a more exploratory mode), the MoA-HMM framework seems broadly applicable to other tasks in the field and useful for the purpose of quantification here. Overall, I believe this paper is certainly worthy of publication in a journal like eLife.

Weaknesses:

I am pleased with the authors' responses to my initial comments, and I thank them for their efforts. My main note of caution to readers is just that when it comes to applying this method to neural data, the benefits may be subtle. On one extreme, it may be possible to capture many of these effects simply by explicitly modeling time, although the authors do a good job showing that they can beat this benchmark in their case. On the other extreme, there may be multiple switches that cannot simply be a monotonic time effect, but these might be at a faster timescale than can be easily captured in this model (in Fig. 7Aii, for example, there is still lots of variance unexplained by the latent state). Quantitative justification will be required for using this model over simpler alternatives, but again, I commend the authors for providing that justification in this paper.

Reviewer #2 (Public review):

Summary:

This is an interesting and well-performed study that develops a new modeling approach (MoA-HMM) to simultaneously characterize reinforcement learning parameters of different RL agents, as well as latent behavioral states that differ in the relative contributions of those agents to the animal's choices. They performed this study in rats trained to perform the two-step task. While the major advance of the paper is developing and rigorously validating this novel technical approach, there are also a number of interesting conceptual advances. For instance, humans performing the two-step task are thought to exhibit a trade-off between model-free and model-based strategies. However, the MoA-HMM did not reveal such a trade-off in the rats, but instead suggested a trade-off between model-based exploratory vs. exploitative strategies. Additionally, the firing rates of neurons in the orbitofrontal cortex (OFC) reflected latent behavioral states estimated from the HMM, suggesting that (1) characterizing dynamic behavioral strategies might help elucidate neural dynamics supporting behavior, and (2) OFC might reflect the contributions of one or a subset of RL agents that are preferentially active or engaged in particular states identified by the HMM.

Strengths:

The claims were generally well-supported by the data. The model was validated rigorously, and was used to generate and test novel predictions about behavior and neural activity in OFC. The approach is likely to be generally useful for characterizing dynamic behavioral strategies.

Author response:

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

Reviewer #1 (Public Review):

The authors sometimes seem to equivocate on to what extent they view their model as a neural (as opposed to merely behavioral) description. For example, they introduce their paper by citing work that views heterogeneity in strategy as the result of "relatively independent, separable circuits that are conceptualized as supporting distinct strategies, each potentially competing for control." The HMM, of course, also relates to internal states of the animal. Therefore, the reader might come away with the impression that the MoA-HMM is literally trying to model dynamic, competing controllers in the brain (e.g. basal ganglia vs. frontal cortex), as opposed to giving a descriptive account of their emergent behavior. If the former is really the intended interpretation, the authors should say more about how they think the weighting/arbitration mechanism between alternative strategies is implemented, and how it can be modulated over time. If not, they should make this clearer.

The MoA-HMM is meant to be descriptive in identifying behaviorally distinct strategies. Our intention in connecting it with a “mixture-of-strategies” view of the brain is that the results of the MoA-HMM could be indicative of an underlying arbitration process, but not modeling that process per se, that can be used to test neural hypotheses driven by this idea. We’ve added additional clarification in the discussion to highlight this point.

Explicitly, we added the following sentence in the discussion: “For example, while the MoA-HMM itself is a descriptive model of behavior and is not explicitly modeling an underlying arbitration of controllers in the brain, the resulting behavioral states may be indicative of underlying neural processes and help identify times when different neural controllers are prevailing”

Second, while the authors demonstrate that model recovery recapitulates the weight dynamics and action values (Fig. 3), the actual parameters that are recovered are less precise (Fig. 3 Supplement 1). The authors should comment on how this might affect their later inferences from behavioral data. Furthermore, it would be better to quantify using the R^2 score between simulated and recovered, rather than the Pearson correlation (r), which doesn't enforce unity slope and zero intercept (i.e. the line that is plotted), and so will tend to exaggerate the strength of parameter recovery.

In the methods section, we noted that the interaction between parameters can cause the recovery of randomly drawn parameter sets to fail, as seen in Figure 3 Supplement 1. This is because there are parameter regimes (specifically when a softmax temperature is near zero) which causes choices to be random, and therefore other parameters no longer matter. To address this, we included a second supplemental figure, Figure 3 Supplement 2, where we recovered model parameters from data simulated solely from models inferred from the behavioral data. Recovery of these models is much more precise, which credits our later inferences from the behavioral data.

To make this point clearer, we changed the reference to Figure 3 Supplements 1 & 2 to: “(Figure 3 – figure supplement 1 for recovery of randomized parameters with noted limitations, and figure supplement 2 for recovery of models fit to real data)” We additionally added the following to the Figure 3 Supplement 1 caption: “Due to the interaction between different model parameters (e.g. a small 𝛽 weight will affect the recoverability of the agent’s learning rate 𝛼), a number of “failures” can be seen.”

Furthermore, we added an R^2 score that enforces unity slope and zero intercept alongside the Pearson correlation coefficient for more comprehensive metrics of recovery. The R^2 scores are plotted on both Figure 3 Supplements 1 & 2 as “R2”, and the following text was added in both captions: “"r" is the Pearson's correlation coefficient between the simulated and recovered parameters, and "R2" is the coefficient of determination, R2, calculating how well the simulated parameters predict the recovered parameters.”

Finally, the authors are very aware of the difficulties associated with long-timescale (minutes) correlations with neural activity, including both satiety and electrode drift, so they do attempt to control for this using a third-order polynomial as a time regressor as well as interaction terms (Fig. 7 Supplement 1). However, on net there does not appear to be any significant difference between the permutation-corrected CPDs computed for states 2 and 3 across all neurons (Fig. 7D). This stands in contrast to the claim that "the modulation of the reward effect can also be seen between states 2 and 3 - state 2, on average, sees a higher modulation to reward that lasts significantly longer than modulation in state 3," which might be true for the neuron in Fig. 7C, but is never quantified. Thus, while I am convinced state modulation exists for model-based (MBr) outcome value (Fig. 7A-B), I'm not convinced that these more gradual shifts can be isolated by the MoA-HMM model, which is important to keep in mind for anyone looking to apply this model to their own data.

We agree with the reviewers that our initial test of CPD significance was not sufficient to support the claims we made about state differences, especially for Figure 7D. To address this, we updated the significance test and indicators in Figure 7B,D to instead signify when there is a significant difference between state CPDs. This updated test supports a small, but significant difference in early post-outcome reward modulation between states 2 and 3.

We clarified and updated the significance test in the methods with the following text:

“A CPD (for a particular predictor in a particular state in a particular time bin) was considered significant if that CPD computed using the true dataset was greater than 95% of corresponding CPDs (same predictor, same state, same time bin) computed using these permuted sessions. For display, we subtract the average permuted session CPD from the true CPD in order to allow meaningful comparison to 0.

To test whether neural coding of a particular predictor in a particular time bin significantly differed according to HMM state, we used a similar test. For each CPD that was significant according to the above test, we computed the difference between that CPD and the CPD for the same predictor and time bin in the other HMM states. We compare this difference to the corresponding differences in the circularly permuted sessions (same predictor, time bin, and pair of HMM states). We consider this difference to be significant if the difference in the true dataset is greater than 95% of the CPD differences computed from the permuted sessions.”

We updated the significance indicators above the panels in Figure 7B,D (colored points) to refer to significant differences between states, with additional text to the left of each row of points to specify the tested state and which states it is significantly greater than. We updated the figure caption for both B and D to reflect these changes.

We also changed text in the results to focus on significant differences between states. Specifically, we replaced the sentence “Looking at the CPD of expected outcome value split by state (Figure 7B) reveals that the trend from the example neuron is consistent across the population of OFC units, where state 2 shows the greatest CPD.” with the sentence “Looking at the CPD of expected outcome value split by state (Figure 7B) reveals that the trend from the example neuron is consistent across the population of OFC units, where state 2 has a significantly greater CPD than states 1 and 3.”

We also replaced the sentence “Suggestively, the modulation of the reward effect can also be seen between states 2 and 3 – state 2, on average, sees a higher modulation to reward that lasts significantly longer than modulation in state 3.” with the sentence “Additionally, the modulation of the reward effect can also be seen between states 2 and 3 — immediately after outcome, we see a small but significantly higher modulation to reward during state 2 than during state 3.”

Reviewer #2 (Public Review):

There were a lot of typos and some figures were mis-referenced in the text and figure legends.

We apologize for the numerous typos and errors in the text and are grateful for the assistance in identifying many of them. We have taken another thorough pass through the manuscript to address those identified by the reviewer as well as fix additional errors. To reduce redundancy, we’ll address all typoand error-related suggestions from both reviewers here.

● We fixed all Figure 1 references. We additionally reversed the introduction order of the agents in Figure 1 and in the results section “Reinforcement learning in the rat two-step task”, where we introduce both model-free agents before both model-based agents. This is to make the model-based choice agent description (which references the model-free choice agent in the statement “That is, like MFc, this agent tends to repeat or switch choices regardless of reward”) come after introducing the model-free choice agent.

● We fixed all Figure 4 references.

● We fixed all Figure 6 references and fixed the panel references in the figure caption to match the figure labeling: Starting with panel B, the reference to (i) was removed, and the reference to (ii) was updated to C. The previous reference to C was updated to D.

● All line-numbered suggestions were addressed.

● The text “(move to supplement?)” was removed from the methods heading, and the mistaken reference to Q_MBr was fixed.

● We removed all “SR” acronyms from the statistics as it was an artifact from an earlier draft.

● We homogenized notation in Figure 2, replacing all “c” variable references with “y”, as well as homogenized notation of β

● We replaced many uses of the word “action” with the word “choice” for consistency throughout the manuscript.

● We addressed many additional minor errors

Reviewer #1 (Recommendations For The Authors):

(1) Could the authors comment on why the cross-validated accuracy continues to increase, albeit non-significantly, after four states, as opposed to decreasing (as I would naively expect would be the result due to overfitting)?

Due to the large amounts of trials and sessions obtained from each rat (often >100 sessions with >200 trials per session) and the limited number of training iterations (capped at 300 iterations), it is not guaranteed that the cross-validated accuracy would decrease over the range of states we included in Figure 4, especially given that the number of total parameters in the largest model shown (7-states, 95 parameters) is greatly less than the number of observations. Since we’re mainly interested in using this tool to identify interpretable, consistent structure across animals, we did not focus on interpreting the regime of larger models.

(2) It seems like the model was refit multiple times with different priors ("Estimation of Population Prior"), each derived from the previous step of fitting. I'm not very familiar with fitting these kinds of models. Is this standard practice? It gives off the feeling of double-dipping. It would be helpful if the authors could cite some relevant literature here or further justify their choices.

We adopted a “one-step” hierarchical approach, where we estimate the population prior a single time on (nearly) unconstrained model fits, and use it for a second, final round of model fits which were used for analysis. Since the prior is only estimated once, in practice there isn’t risk of converging on an overly constrained prior. This is a somewhat simplified approach motivated by analogy to the first step of EM fit in a hierarchical model, in which population- and subject-level parameters are iteratively re-estimated in terms of one another until convergence (Huys et al., 2012; Daw 2010). We have clarified this approach in the methods with citations by adding the following paragraph:

“Hierarchical modeling gives a better estimate of how model parameters can vary within a population by additionally inferring the population distribution over which individuals are likely drawn (Daw, 2011). This type of modeling, however, is notoriously difficult in HMMs; therefore, as a compromise, we adopt a “one-step” hierarchical model, where we estimate population parameters from “unconstrained” fits on the data, which are then used as a prior to regularize the final model fits. This approach is motivated by analogy to the first step of EM fit in a hierarchical model, in which population- and subject-level parameters are iteratively re-estimated in terms of one another until convergence (Daw, 2011; Huys et al., 2012). It is important to emphasize, since we aren’t inferring the population distributions directly, that we only estimate the population prior a single time on the “unconstrained” fits as follows.”

Reviewer #2 (Recommendations For The Authors):

Figure 3a.iii: Did the model capture the transition probabilities correctly as well?

We have updated Figure 3E to include additional panels (iii) and (iv) to show the recovered initial state probabilities and transition matrix.

For Figure 6, panel B makes it look like there is a larger influence of state on ITI rate after omission, in both the top and bottom plots. However, the violin plots in panel C show a different pattern, where state has a greater effect on ITIs following rewarded trials. Is it that the example in panel B is not representative of the population, or am I misinterpreting?

We thank the reviewer for catching this issue, as the colors were erroneously flipped in panel C. We have fixed this figure by ensuring that the colors appropriately matched the trial type (reward or omission). Additionally, we updated the colors in B and C that correspond to reward (previously gray, now blue) and omission (previously gold, now red) trials to match the color scheme used in Figure 1. We also inverted the corresponding line styles (reward changed to solid, omission changed to dashed) to match the convention used in Figure 7. To differentiate from the reward/omission color changed, we additionally changed the colors in Figure 6D and Figure 7 Supplement 1, where the color for “time” was changed from blue to gray, and the color for “state” was changed from red to gold.

For figure 4B right, I am confused. The legend says that this is the change in model performance relative to a model with one fewer state. But the y-axis says it's the change from the single-state model. Please clarify.

The plot is showing the increase in performance from the single-state model, while the significance tests were done between consecutive numbered states. We updated the significance indicators on the plot to more clearly identify that adjacent models are being compared (with the exception of the 2-state model, which is being compared to 0). We updated the Figure 4B caption text for the left panel to state: “Change in normalized, cross-validated likelihood when adding additional hidden states into the MoA-HMM, relative to the single-state model. Significant changes are computed with respect to models with one fewer states (e.g. 2-state vs 1-state, 3-state vs 2-state)”

eLife Assessment

This is an important study that combines an array of genetic, cell biological, and genomic techniques to elucidate the role of the transcription factor Hamlet in reproductive development. It provides compelling evidence that Hamlet is a master regulator of cell fate and differentiation to reveal transcriptional targets that mediate epithelial tissue fusion. While the genetic and genomic analyses are convincing, the images that report the phenotypes are difficult to interpret for non-experts. This, and other identified issues, should be addressed.

Reviewer #1 (Public review):

Summary:

Wang et al. identify Hamlet, a PR-containing transcription factor, as a master regulator of reproductive development in Drosophila. Specifically, the fusion between the gonad and genital disc is necessary for the development of continuous testes and seminal vesicle tissue essential for fertility. To do this, the authors generate novel Hamlet null mutants by CRISPR/Cas9 gene editing and characterize the morphological, physiological, and gene expression changes of the mutants using immunofluorescence, RNA-seq, cut-tag, and in-situ analysis. Thus, Hamlet is discovered to regulate a unique expression program, which includes Wnt2 and Tl, that is necessary for testis development and fertility.

Strengths:

This is a rigorous and comprehensive study that identifies the Hamlet-dependent gene expression program mediating reproductive development in Drosophila. The Hamlet transcription targets are further characterized by Gal4/UAS-RNAi confirming their role in reproductive development. Finally, the study points to a role for Wnt2 and Tl as well as other Hamlet transcriptionally regulated genes in epithelial tissue fusion.

Weaknesses:

The image resolution and presentation of figures is a major issue in this study. As a non-expert, it is nearly impossible to see the morphological changes as described in the results. Quantification of all cell biological phenotypes is also lacking therefore reducing the impact of this study to those familiar with tissue fusion events in Drosophila development.

Reviewer #2 (Public review):

Strengths:

Wang and colleagues successfully uncovered an important function of the Drosophila PRDM16/PRDM3 homolog Hamlet (Ham) - a PR domain-containing transcription factor with known roles in the nervous system in Drosophila. To do so, they generated and analyzed new mutants lacking the PR domain, and also employed diverse preexisting tools. In doing so, they made a fascinating discovery: They found that PR-domain containing isoforms of ham are crucial in the intriguing development of the fly genital tract. Wang and colleagues found three distinct roles of Ham: (1) specifying the position of the testis terminal epithelium within the testis, (2) allowing normal shaping and growth of the anlagen of the seminal vesicles and paragonia and (3) enabling the crucial epithelial fusion between the seminal vesicle and the testis terminal epithelium. The mutant blocks fusion even if the parts are positioned correctly. The last finding is especially important, as there are few models allowing one to dissect the molecular underpinnings of heterotypic epithelial fusion in development. Their data suggest that they found a master regulator of this collective cell behavior. Further, they identified some of the cell biological players downstream of Ham, like for example E-Cadherin and Crumbs. In a holistic approach, they performed RNAseq and intersected them with the CUT&TAG-method, to find a comprehensive list of downstream factors directly regulated by Ham. Their function in the fusion process was validated by a tissue-specific RNAi screen. Meticulously, Wang and colleagues performed multiplexed in situ hybridization and analyzed different mutants, to gain a first understanding of the most important downstream pathways they characterized, which are Wnt2 and Toll.

This study pioneers a completely new system. It is a model for exploring a process crucial in morphogenesis across animal species, yet not well understood. Wang and colleagues not only identified a crucial regulator of heterotypic epithelial fusion but took on the considerable effort of meticulously pinning down functionally important downstream effectors by using many state-of-the-art methods. This is especially impressive, as the dissection of pupal genital discs before epithelial fusion is a time-consuming and difficult task. This promising work will be the foundation future studies build on, to further elucidate how this epithelial fusion works, for example on a cell biological and biomechanical level.

Weaknesses:

The developing testis-genital disc system has many moving parts. Myotube migration was previously shown to be crucial for testis shape. This means, that there is the potential of non-tissue autonomous defects upon knockdown of genes in the genital disc or the terminal epithelium, affecting myotube behavior which in turn affects fusion, as myotubes might create the first "bridge" bringing the epithelia together. The authors clearly showed that their driver tools do not cause expression in myoblasts/myotubes, but this does not exclude non-tissue autonomous defects in their RNAi screen. Nevertheless, this is outside the scope of this work.

However, one point that could be addressed in this study: the RNAseq and CUT&TAG experiments would profit from adding principal component analyses, elucidating similarities and differences of the diverse biological and technical replicates.

Author response:

Thank you for reviewing our manuscript and providing constructive feedback. We are grateful that you recognize the importance of our work and find the evidences presented compelling. We will revise our manuscripts in accordance with reviewers’ recommendations. Below is our plan.

(1) As recommended by Reviewer 1, we will improve the image resolution and presentation in the figures, by adjusting dark colors into brighter ones, including single-channel images, and incorporating schematic illustrations to dipict morphological changes.

(2) Following the suggestions of reviewer 2, we will provide explanations and speculative insights into potential non-tissue autonomous effects.

(3) As suggested by reviewer 2, we will perform principal component analyses on our RNA-seq and Cut&Tag data. 

(2) Once we have addressed all the major and minor points raised by the reviewers, we will provide a detailed point-to-point response and submit the revised version of the manuscript.

eLife Assessment

This useful study investigates the role of frontotemporal regions in integrating linguistic and extra-linguistic information during communication, focusing on the inferior frontal gyrus and posterior middle temporal gyrus. It uses brain stimulation and electroencephalography to study speech-gesture integration. While the research question is interesting, the methods are insufficient for studying tightly-coupled brain regions over short timescales, leading to incomplete support for the claims due to conceptual and methodological limitations.

Reviewer #1 (Public review):

Summary:

The authors quantified information in gesture and speech, and investigated the neural processing of speech and gestures in pMTG and LIFG, depending on their informational content, in 8 different time-windows, and using three different methods (EEG, HD-tDCS and TMS). They found that there is a time-sensitive and staged progression of neural engagement that is correlated with the informational content of the signal (speech/gesture).

Strengths:

A strength of the paper is that the authors attempted to combine three different methods to investigate speech-gesture processing.

Weaknesses:

(1) One major issue is that there is a tight anatomical coupling between pMTG and LIFG. Stimulating one area could therefore also result in stimulation of the other area (see Silvanto and Pascual-Leone, 2008). I therefore think it is very difficult to tease apart the contribution of these areas to the speech-gesture integration process, especially considering that the authors stimulate these regions in time windows that are very close to each other in both time and space (and the disruption might last longer over time).

(2) Related to this point, it is unclear to me why the HD-TDCS/TMS is delivered in set time windows for each region. How did the authors determine this, and how do the results for TMS compare to their previous work from 2018 and 2023 (which describes a similar dataset+design)? How can they ensure they are only targeting their intended region since they are so anatomically close to each other?

(3) As the EEG signal is often not normally distributed, I was wondering whether the authors checked the assumptions for their Pearson correlations. The authors could perhaps better choose to model the different variables to see whether MI/entropy could predict the neural responses. How did they correct the many correlational analyses that they have performed?

(4) The authors use ROIs for their different analyses, but it is unclear why and on the basis of what these regions are defined. Why not consider all channels without making them part of an ROI, by using a method like the one described in my previous comment?

(5) The authors describe that they have divided their EEG data into a "lower half" and a "higher half" (lines 234-236), based on entropy scores. It is unclear why this is necessary, and I would suggest just using the entropy scores as a continuous measure.

Reviewer #2 (Public review):

Summary:

The study is an innovative and fundamental study that clarified important aspects of brain processes for integration of information from speech and iconic gesture (i.e., gesture that depicts action, movement, and shape), based on tDCS, TMS, and EEG experiments. They evaluated their speech and gesture stimuli in information-theoretic ways and calculated how informative speech is (i.e., entropy), how informative gesture is, and how much shared information speech and gesture encode. The tDCS and TMS studies found that the left IFG and pMTG, the two areas that were activated in fMRI studies on speech-gesture integration in the previous literature, are causally implicated in speech-gesture integration. The size of tDC and TMS effects are correlated with the entropy of the stimuli or mutual information, which indicates that the effects stem from the modulation of information decoding/integration processes. The EEG study showed that various ERP (event-related potential, e.g., N1-P2, N400, LPC) effects that have been observed in speech-gesture integration experiments in the previous literature, are modulated by the entropy of speech/gesture and mutual information. This makes it clear that these effects are related to information decoding processes. The authors propose a model of how the speech-gesture integration process unfolds in time, and how IFG and pMTG interact with each other in that process.

Strengths:

The key strength of this study is that the authors used information theoretic measures of their stimuli (i.e., entropy and mutual information between speech and gesture) in all of their analyses. This made it clear that the neuro-modulation (tDCS, TMS) affected information decoding/integration and ERP effects reflect information decoding/integration. This study used tDCS and TMS methods to demonstrate that left IFG and pMTG are causally involved in speech-gesture integration. The size of tDCS and TMS effects are correlated with information-theoretic measures of the stimuli, which indicate that the effects indeed stem from disruption/facilitation of the information decoding/integration process (rather than generic excitation/inhibition). The authors' results also showed a correlation between information-theoretic measures of stimuli with various ERP effects. This indicates that these ERP effects reflect the information decoding/integration process.

Weaknesses:

The "mutual information" cannot fully capture the interplay of the meaning of speech and gesture. The mutual information is calculated based on what information can be decoded from speech alone and what information can be decoded from gesture alone. However, when speech and gesture are combined, a novel meaning can emerge, which cannot be decoded from a single modality alone. When example, a person produces a gesture of writing something with a pen, while saying "He paid". The speech-gesture combination can be interpreted as "paying by signing a cheque". It is highly unlikely that this meaning is decoded when people hear speech only or see gestures only. The current study cannot address how such speech-gesture integration occurs in the brain, and what ERP effects may reflect such a process. Future studies can classify different types of speech-gesture integration and investigate neural processes that underlie each type. Another important topic for future studies is to investigate how the neural processes of speech-gesture integration change when the relative timing between the speech stimulus and the gesture stimulus changes.

Reviewer #3 (Public review):

In this useful study, Zhao et al. try to extend the evidence for their previously described two-step model of speech-gesture integration in the posterior Middle Temporal Gyrus (pMTG) and Inferior Frontal Gyrus (IFG). They repeat some of their previous experimental paradigms, but this time quantifying Information-Theoretical (IT) metrics of the stimuli in a stroop-like paradigm purported to engage speech-gesture integration. They then correlate these metrics with the disruption of what they claim to be an integration effect observable in reaction times during the tasks following brain stimulation, as well as documenting the ERP components in response to the variability in these metrics.

The integration of multiple methods, like tDCS, TMS, and ERPs to provide converging evidence renders the results solid. However, their interpretation of the results should be taken with care, as some critical confounds, like difficulty, were not accounted for, and the conceptual link between the IT metrics and what the authors claim they index is tenuous and in need of more evidence. In some cases, the difficulty making this link seems to arise from conceptual equivocation (e.g., their claims regarding 'graded' evidence), whilst in some others it might arise from the usage of unclear wording in the writing of the manuscript (e.g. the sentence 'quantitatively functional mental states defined by a specific parser unified by statistical regularities'). Having said that, the authors' aim is valuable, and addressing these issues would render the work a very useful approach to improve our understanding of integration during semantic processing, being of interest to scientists working in cognitive neuroscience and neuroimaging.

The main hurdle to achieving the aims set by the authors is the presence of the confound of difficulty in their IT metrics. Their measure of entropy, for example, being derived from the distribution of responses of the participants to the stimuli, will tend to be high for words or gestures with multiple competing candidate representations (this is what would presumptively give rise to the diversity of responses in high-entropy items). There is ample evidence implicating IFG and pMTG as key regions of the semantic control network, which is critical during difficult semantic processing when, for example, semantic processing must resolve competition between multiple candidate representations, or when there are increased selection pressures (Jackson et al., 2021). Thus, the authors' interpretation of Mutual Information (MI) as an index of integration is inextricably contaminated with difficulty arising from multiple candidate representations. This casts doubt on the claims of the role of pMTG and IFG as regions carrying out gesture-speech integration as the observed pattern of results could also be interpreted in terms of brain stimulation interrupting the semantic control network's ability to select the best candidate for a given context or respond to more demanding semantic processing.

In terms of conceptual equivocation, the use of the term 'graded' by the authors seems to be different from the usage commonly employed in the semantic cognition literature (e.g., the 'graded hub hypothesis', Rice et al., 2015). The idea of a graded hub in the controlled semantic cognition framework (i.e., the anterior temporal lobe) refers to a progressive degree of abstraction or heteromodal information as you progress through the anatomy of the region (i.e., along the dorsal-to-ventral axis). The authors, on the other hand, seem to refer to 'graded manner' in the context of a correlation of entropy or MI and the change in the difference between Reaction Times (RTs) of semantically congruent vs incongruent gesture-speech. The issue is that the discourse through parts of the introduction and discussion seems to conflate both interpretations, and the ideas in the main text do not correspond to the references they cite. This is not overall very convincing. What is it exactly the authors are arguing about the correlation between RTs and MI indexes? As stated above, their measure of entropy captures the spread of responses, which could also be a measure of item difficulty (more diverse responses imply fewer correct responses, a classic index of difficulty). Capturing the diversity of responses means that items with high entropy scores are also likely to have multiple candidate representations, leading to increased selection pressures. Regions like pMTG and IFG have been widely implicated in difficult semantic processing and increased selection pressures (Jackson et al., 2021). How is this MI correlation evidence of integration that proceeds in a 'graded manner'? The conceptual links between these concepts must be made clearer for the interpretation to be convincing.

Author response:

Responses to Editors:

We appreciate Reviewer 1’s first concern regarding the difficulty of disentangling the contributions of tightly-coupled brain regions to the speech-gesture integration process—particularly due to the close temporal and spatial proximity of the stimulation windows and the potential for prolonged disruption. We would like to provide clarification and evidence supporting the validity of our methodology.

Our previous study (Zhao et al., 2021, J. Neurosci) employed the same experimental protocol—using inhibitory double-pulse transcranial magnetic stimulation (TMS) over the inferior frontal gyrus (IFG) and posterior middle temporal gyrus (pMTG) in one of eight 40-ms time windows. The findings from that study demonstrated a time-window-selective disruption of the semantic congruency effect (i.e., reaction time costs driven by semantic conflict), with no significant modulation of the gender congruency effect (i.e., reaction time costs due to gender conflict). This result establishes that double-pulse TMS provides sufficient temporal precision to independently target the left IFG and pMTG within these 40-ms windows during gesture-speech integration. Importantly, by comparing the distinctively inhibited time windows for IFG and pMTG, we offered clear evidence of distinct engagement and temporal dynamics between these regions during different stages of gesture-speech semantic processing.

Furthermore, we reviewed prior studies utilizing double-pulse TMS on structurally and functionally connected brain regions to explore neural contributions across timescales as brief as 3–60 ms. These studies, which encompass areas from the tongue and lip areas of the primary motor cortex (M1) to high-level semantic regions such as the pMTG and ATL (Author response table 1), consistently demonstrate the methodological rigor and precision of double-pulse TMS in disentangling the neural dynamics of different regions within these short temporal windows.

Author response table 1.

Double-pulse TMS studies on brain regions over 3-60 ms time interval

Response to Reviewer #1:

(1) For concern on the difficulty of disentangling the contributions of tightly-coupled brain regions to the speech-gesture integration process:

We trust that the explanation provided above has clarified this issue.

(2) For concern on the rationale for delivering HD-tDCS/TMS in set time windows for each region, as well as how these time windows were determined and how the current results compare to our previous studies from 2018 and 2023:

The current study builds on a series of investigations that systematically examined the temporal and spatial dynamics of gesture-speech integration. In our earlier work (Zhao et al., 2018, J. Neurosci), we demonstrated that interrupting neural activity in the IFG or pMTG using TMS selectively disrupted the semantic congruency effect (reaction time costs due to semantic incongruence), without affecting the gender congruency effect (reaction time costs due to gender incongruence). These findings identified the IFG and pMTG as critical hubs for gesture-speech integration. This informed the brain regions selected for subsequent studies.

In Zhao et al. (2021, J. Neurosci), we employed a double-pulse TMS protocol, delivering stimulation within one of eight 40-ms time windows, to further examine the temporal involvement of the IFG and pMTG. The results revealed time-window-selective disruptions of the semantic congruency effect, confirming the dynamic and temporally staged roles of these regions during gesture-speech integration.

In Zhao et al. (2023, Frontiers in Psychology), we investigated the semantic predictive role of gestures relative to speech by comparing two experimental conditions: (1) gestures preceding speech by a fixed interval of 200 ms, and (2) gestures preceding speech at its semantic identification point. We observed time-window-selective disruptions of the semantic congruency effect in the IFG and pMTG only in the second condition, leading to the conclusion that gestures exert a semantic priming effect on co-occurring speech. These findings underscored the semantic advantage of gesture in facilitating speech integration, further refining our understanding of the temporal and functional interplay between these modalities.

The design of the current study—including the choice of brain regions and time windows—was directly informed by these prior findings. Experiment 1 (HD-tDCS) targeted the entire gesture-speech integration process in the IFG and pMTG to assess whether neural activity in these regions, previously identified as integration hubs, is modulated by changes in informativeness from both modalities (i.e., entropy) and their interactions (mutual information, MI). The results revealed a gradual inhibition of neural activity in both areas as MI increased, evidenced by a negative correlation between MI and the tDCS inhibition effect in both regions. Building on this, Experiments 2 and 3 employed double-pulse TMS and event-related potentials (ERPs) to further assess whether the engaged neural activity was both time-sensitive and staged. These experiments also evaluated the contributions of various sources of information, revealing correlations between information-theoretic metrics and time-locked brain activity, providing insights into the ‘gradual’ nature of gesture-speech integration.

We acknowledge that the rationale for the design of the current study was not fully articulated in the original manuscript. In the revised version, we will provide a more comprehensive and coherent explanation of the logic behind the three experiments, ensuring clear alignment with our previous findings.

(3) For concern about the use of Pearson correlation and the normality of EEG data.

We appreciate the reviewer’s thoughtful consideration. In Figure 5 of the manuscript, we have already included normal distribution curves that illustrate the relationships between the average ERP amplitudes within each ROI or elicited clusters and the three information models. Additionally, multiple comparisons were addressed using FDR correction, as outlined in the manuscript.

To further clarify the data, we will calculate the Shapiro-Wilk test, a widely accepted method for assessing bivariate normality, for both the MI/entropy and averaged ERP data. The corresponding p-values will be provided in the following-up point-to-point responses.

(4) For concern about the ROI selection, and the suggestion of using whole-brain electrodes to build models of different variables (MI/entropy) to predict neural responses:

For the EEG data, we conducted both a traditional region-of-interest (ROI) analysis, with ROIs defined based on a well-established work (Habets et al., 2011), and a cluster-based permutation approach, which utilizes data-driven permutations to enhance robustness and address multiple comparisons. The latter method complements the hypothesis-driven ROI analysis by offering an exploratory, unbiased perspective. Notably, the results from both approaches were consistent, reinforcing the reliability of our findings.

To make the methods more accessible to a broader audience, we will provide a clear description of the methods used and how they relate to each other in the revised manuscript.

Reference:

Habets, B., Kita, S., Shao, Z.S., Ozyurek, A., and Hagoort, P. (2011). The Role of Synchrony and Ambiguity in Speech-Gesture Integration during Comprehension. J Cognitive Neurosci 23, 1845-1854. 10.1162/jocn.2010.21462

(5) For concern about the median split of the data:

To identify ERP components or spatiotemporal clusters that demonstrated significant semantic differences, we split each model into higher and lower halves, focusing on indexing information changes reflected by entropy or mutual information (MI). To illustrate the gradual activation process, the identified components and clusters were further analyzed for correlations with each information matrix. Remarkably, consistent results were observed between the ERP components and clusters, providing robust evidence that semantic information conveyed through gestures and speech significantly influenced the amplitude of these components or clusters. Moreover, the semantic information was shown to be highly sensitive, varying in tandem with these amplitude changes.

We acknowledge that the rationale behind this approach may not have been sufficiently clear in the initial manuscript. In our revision, we will ensure a more detailed and precise explanation to enhance the clarity and coherence of this logical framework.

Response to Reviewer #2:

We greatly appreciate Reviewer2 ’s concern regarding whether "mutual information" adequately captures the interplay between the meanings of speech and gesture. We would like to clarify that the materials used in the present study involved gestures performed without actual objects, paired with verbs that precisely describe the corresponding actions. For example, a hammering gesture was paired with the verb “hammer”, and a cutting gesture was paired with the verb “cut”. In this design, all gestures conveyed redundant meaning relative to the co-occurring speech, creating significant overlap between the information derived from speech alone and that from gesture alone.

We understand the reviewer’s concern about cases where gestures and speech may provide complementary rather than redundant information. To address this, we have developed an alternative metric for quantifying information gains contributed by supplementary multisensory cues, which will be explored in a subsequent study. However, for the present study, we believe that the observed overlap in information serves as an indicator of the degree of multisensory convergence, a central focus of our investigation.

Regarding the reviewer’s concern about how the neural processes of speech-gesture integration may change with variations in the relative timing between speech and gesture stimuli, we would like to highlight findings from our previous study (Zhao, 2023, Frontiers in Psychology). In that study, we explored the semantic predictive role of gestures relative to speech under two conditions: (1) gestures preceding speech by a fixed interval of 200 ms, and (2) gestures preceding speech of its semantic identification point. Interestingly, only in the second condition did we observe time-window-selective disruptions of the semantic congruency effect in the IFG and pMTG. This led us to conclude that gestures play a semantic priming role for co-occurring speech. Building on this, we designed the present study with gestures preceding speech of its semantic identification point to reflect this semantic priming relationship. Additionally, ongoing research is exploring gesture and speech interactions in natural conversational settings to investigate whether the neural processes identified here are consistent across varying contexts.

To prevent any similar concerns from causing doubt among the audience and to ensure clarity regarding the follow-up study, we will provide a detailed discussion of the two issues in the revised manuscript.

Response to Reviewer #3:

The primary aim of this study is to investigate whether the degree of activity in the established integration hubs, IFG and pMTG, is influenced by the information provided by gesture-speech modalities and/or their interactions. While we provided evidence for the differential involvement of the IFG and pMTG by delineating their dynamic engagement across distinct time windows of gesture-speech integration and associating these patterns with unisensory information and their interaction, we acknowledge that the mechanisms underlying these dynamics remain open to interpretation. Specifically, whether the observed effects stem from difficulties in semantic control processes, as suggested by Reviewer 3, or from resolving information uncertainty, as quantified by entropy, falls outside the scope of the current study. Importantly, we view these two interpretations as complementary rather than mutually exclusive, as both may be contributing factors. Nonetheless, we agree that addressing this question is a compelling avenue for future research. In the revised manuscript, we will include an exploratory analysis to investigate whether the confounding difficulty, stemming from the number of lexical or semantic representations, is limited to high-entropy items. Additionally, we will address and discuss alternative interpretations.

Regarding the concern of conceptual equivocation, we would like to emphasize that this study represents the first attempt to focus on the relationship between information quantity and neural engagement. In our initial presentation, we inadvertently conflated the commonly used term "graded hub," which refers to anatomical distribution, with its usage in the present context. We sincerely apologize for this oversight and are grateful for the reviewer’s careful critique. In the revised manuscript, we will clearly articulate the study’s objectives, clarify the representations of entropy and mutual information, and accurately describe their association with neural engagement.

Reference

Teige, C., Mollo, G., Millman, R., Savill, N., Smallwood, J., Cornelissen, P. L., & Jefferies, E. (2018). Dynamic semantic cognition: Characterising coherent and controlled conceptual retrieval through time using magnetoencephalography and chronometric transcranial magnetic stimulation. Cortex, 103, 329-349.

Amemiya, T., Beck, B., Walsh, V., Gomi, H., & Haggard, P. (2017). Visual area V5/hMT+ contributes to perception of tactile motion direction: a TMS study. Scientific reports, 7(1), 40937.

Muessgens, D., Thirugnanasambandam, N., Shitara, H., Popa, T., & Hallett, M. (2016). Dissociable roles of preSMA in motor sequence chunking and hand switching—a TMS study. Journal of Neurophysiology, 116(6), 2637-2646.

Vernet, M., Brem, A. K., Farzan, F., & Pascual-Leone, A. (2015). Synchronous and opposite roles of the parietal and prefrontal cortices in bistable perception: a double-coil TMS–EEG study. Cortex, 64, 78-88.

Pitcher, D. (2014). Facial expression recognition takes longer in the posterior superior temporal sulcus than in the occipital face area. Journal of Neuroscience, 34(27), 9173-9177.

Bardi, L., Kanai, R., Mapelli, D., & Walsh, V. (2012). TMS of the FEF interferes with spatial conflict. Journal of cognitive neuroscience, 24(6), 1305-1313.

D’Ausilio, A., Bufalari, I., Salmas, P., & Fadiga, L. (2012). The role of the motor system in discriminating normal and degraded speech sounds. Cortex, 48(7), 882-887.

Pitcher, D., Duchaine, B., Walsh, V., & Kanwisher, N. (2010). TMS evidence for feedforward and feedback mechanisms of face and body perception. Journal of Vision, 10(7), 671-671.

Gagnon, G., Blanchet, S., Grondin, S., & Schneider, C. (2010). Paired-pulse transcranial magnetic stimulation over the dorsolateral prefrontal cortex interferes with episodic encoding and retrieval for both verbal and non-verbal materials. Brain Research, 1344, 148-158.

Kalla, R., Muggleton, N. G., Juan, C. H., Cowey, A., & Walsh, V. (2008). The timing of the involvement of the frontal eye fields and posterior parietal cortex in visual search. Neuroreport, 19(10), 1067-1071.

Pitcher, D., Garrido, L., Walsh, V., & Duchaine, B. C. (2008). Transcranial magnetic stimulation disrupts the perception and embodiment of facial expressions. Journal of Neuroscience, 28(36), 8929-8933.

eLife Assessment

This important study will allow for a better understanding of anthelmintic drug resistance in nematodes, a growing concern for public health. The authors provide a detailed analysis of the role of UBR-1 and its underlying mechanism in ivermectin resistance using convincing behavioural and genetic experiments with C. elegans. What is not yet clear is how much of this study applies to parasitic nematodes, in which resistance has naturally emerged in different populations. The study will be of relevance to colleagues working on the evolution of drug resistance and to parasitologists in general.

Reviewer #1 (Public review):

Summary:

The drug Ivermectin is used to effectively treat a variety of worm parasites in the world, however resistance to Ivermectin poses a rising challenge for this treatment strategy. In this study, the authors found that loss of the E3 ubiquitin ligase UBR-1 in the worm C. elegans results in resistance to Ivermectin. In particular, the authors found that ubr-1 mutants are resistant to the effects of Ivermectin on worm viability, body size, pharyngeal pumping, and locomotion. The authors previously showed that loss of UBR-1 disrupts homeostasis of the amino acid and neurotransmitter glutamate resulting in increased levels of glutamate in C. elegans. Here, the authors found that the sensitivity of ubr-1 mutants to Ivermectin can be restored if glutamate levels are reduced using a variety of different methods. Conversely, treating worms with exogenous glutamate to increase glutamate levels also results in resistance to Ivermectin supporting the idea that increased glutamate promotes resistance to Ivermectin. The authors found that the primary known targets of Ivermectin, glutamate-gated chloride channels (GluCls), are downregulated in ubr-1 mutants providing a plausible mechanism for why ubr-1 mutants are resistant to Ivermectin. Although it is clear that loss of GluCls can lead to resistance to Ivermectin, this study suggests that one potential mechanism to decrease GluCl expression is via disruption of glutamate homeostasis that leads to increased glutamate. This study suggests that if parasitic worms become resistant to Ivermectin due to increased glutamate, their sensitivity to Ivermectin could be restored by reducing glutamate levels using drugs such as Ceftriaxone in a combination drug treatment strategy.

Strengths:

(1) The use of multiple independent assays (i.e., viability, body size, pharyngeal pumping, locomotion, and serotonin-stimulated pharyngeal muscle activity) to monitor the effects of Ivermectin

(2) The use of multiple independent approaches (got-1, eat-4, ceftriaxone drug, exogenous glutamate treatment) to alter glutamate levels to support the conclusion that increased glutamate in ubr-1 mutants contributes to Ivermectin resistance.

Weaknesses:

(1) The primary target of Ivermectin is GluCls so it is not surprising that alteration of GluCl expression or function would lead to Ivermectin resistance.

(2) It remains to be seen what percent of Ivermectin-resistant parasites in the wild have disrupted glutamate homeostasis as opposed to mutations that more directly decrease GluCl expression or function.

Reviewer #2 (Public review):

Summary:

The authors provide a very thorough investigation of the role of UBR-1 in anthelmintic resistance using the non-parasitic nematode, C. elegans. Anthelmintic resistance to macrocyclic lactones is a major problem in veterinary medicine and likely just a matter of time until resistance emerges in human parasites too. Therefore, this study providing novel insight into the mechanisms of ivermectin resistance is particularly important and significant.

Strengths:

The authors use very diverse technologies (behavior, genetics, pharmacology, genetically encoded reporters) to dissect the role of UBR-1 in ivermectin resistance. Deploying such a comprehensive suite of tools and approaches provides exceptional insight into the mechanism of how UBR-1 functions in terms of ivermectin resistance.

Weaknesses:

I do not see any major weaknesses in this study. My only concern is whether the observations made by the authors would translate to any of the important parasitic helminths in which resistance has naturally emerged in the field. This is always a concern when leveraging a non-parasitic nematode to shed light on a potential mechanism of resistance of parasitic nematodes, and I understand that it is likely beyond the scope of this paper to test some of their results in parasitic nematodes.

Reviewer #3 (Public review):

Summary:

Li et al propose to better understand the mechanisms of drug resistance in nematode parasites by studying mutants of the model roundworm C. elegans that are resistant to the deworming drug ivermectin. They provide compelling evidence that loss-of-function mutations in the E3 ubiquitin ligase encoded by the UBR-1 gene make worms resistant to the effects of ivermectin (and related compounds) on viability, body size, pharyngeal pumping rate, and locomotion and that these mutant phenotypes are rescued by a UBR-1 transgene. They propose that the mechanism is resistance is indirect, via the effects of UBR-1 on glutamate production. They show mutations (vesicular glutamate transporter eat-4, glutamate synthase got-1) and drugs (glutamate, glutamate uptake enhancer ceftriaxone) affecting glutamate metabolism/transport modulate sensitivity to ivermectin in wild-type and ubr-1 mutants. The data are generally consistent with greater glutamate tone equating to ivermectin resistance. Finally, they show that manipulations that are expected to increase glutamate tone appear to reduce expression of the targets of ivermectin, the glutamate-gated chloride channels, which is known to increase resistance.

There is a need for genetic markers of ivermectin resistance in livestock parasites that can be used to better track resistance and to tailor drug treatment. The discovery of UBR-1 as a resistance gene in C. elegans will provide a candidate marker that can be followed up in parasites. The data suggest Ceftriaxone would be a candidate compound to reverse resistance.

Strengths:

The strength of the study is the thoroughness of the analysis and the quality of the data. There can be little doubt that ubr-1 mutations do indeed confer ivermectin resistance. The use of both rescue constructs and RNAi to validate mutant phenotypes is notable. Further, the variety of manipulations they use to affect glutamate metabolism/transport makes a compelling argument for some kind of role for glutamate in resistance.

Weaknesses:

The proposed mechanism of ubr-1 resistance i.e.: UBR-1 E3 ligase regulates glutamate tone which regulates ivermectin receptor expression, is broadly consistent with the data but somewhat difficult to reconcile with the specific functions of the genes regulating glutamatergic tone. Ceftriaxone and eat-4 mutants reduce extracellular/synaptic glutamate concentrations by sequestering available glutamate in neurons, suggesting that it is extracellular glutamate that is important. But then why does rescuing ubr-1 specifically in the pharyngeal muscle have such a strong effect on ivermectin sensitivity? Is glutamate leaking out of the pharyngeal muscle into the extracellular space/synapse? Is it possible that UBR-1 acts directly on the avr-15 subunit, both of which are expressed in the muscle, perhaps as part of a glutamate sensing/homeostasis mechanism?

The use of single ivermectin dose assays can be misleading. A response change at a single dose shows that the dose-response curve has shifted, but the response is not linear with dose, so the degree of that shift may be difficult to discern and may result from a change in slope but not EC50.

Similarly, in Figure 3C, the reader is meant to understand that eat-4 mutant is epistatic to ubr-1 because the double mutant has a wild-type response to ivermectin. But eat-4 alone is more sensitive, so (eyeballing it and interpolating) the shift in EC50 caused by the ubr-1 mutant in a wild type background appears to be the same as in an eat-4 background, so arguably you are seeing an additive effect, not epistasis. For the above reasons, it would be desirable to have results for rescuing constructs in a wild-type background, in addition to the mutant background.

The added value of the pumping data in Figure 5 (using calcium imaging) over the pump counts (from video) in Figure 1G, Figure 2E, F, K, & Figure 3D, H is not clearly explained. It may have to do with the use of "dissected" pharynxes, the nature/advantage of which is not sufficiently documented in the Methods/Results.

Author response:

We would like to express our sincere gratitude to both of you, and the reviewers, for the time and effort you have invested in reviewing our manuscript. We greatly appreciate the constructive feedback provided and are committed to addressing the suggested revisions.

In response to the public reviews, we would like to outline the following plan of action:

(1) Addressing Weaknesses in the Manuscript: We have carefully considered the comments regarding the weaknesses identified in the manuscript. Specifically, we will:

- Provide further clarification on the mechanism of IVM resistance in our study.

- Expand our discussion of the limitations and future directions of the research, addressing the concerns related to the potential translation of our findings to parasitic nematodes.

(2) Additional Experiments: We are currently conducting additional experiments to address the reviewers' suggestions, which include:

- Testing whether the overexpression of a relevant GluCl, such as AVR-15, can restore Ivermectin sensitivity in ubr-1 mutants.

- Examining the impact of Ceftriaxone treatment on the Ivermectin resistance in worms lacking key GluCls, such as avr-15, avr-14, and glc-1.

- Incorporating an analysis of major human parasitic nematodes in the phylogeny and discussing the conservation of relevant mechanisms across species.

- Double-checking the Dye filling (Dyf) phenotype in ubr-1 mutants, as suggested.

(3) Point-by-Point response: We will respond to both sets of comments (public reviews and editorial recommendations) in a comprehensive point-by-point manner in the revised manuscript.

(4) Timely Revisions: We aim to complete all revisions within a single round, ensuring that we address all comments thoroughly while maintaining the integrity of the data.

eLife Assessment

This important study describes a first-in-human trial of autologous p63+ stem cells in patients with idiopathic pulmonary fibrosis, a lethal condition for which effective treatments are lacking. The authors provide convincing evidence that P63+ progenitor cell therapy can be safely delivered in patients with ILD. However, given that this is a Phase 1 study, conclusions regarding efficacy should not yet be made.

Reviewer #1 (Public review):

Summary:

IPF is a disease lacking regressive therapies which has a poor prognosis, and so new therapies are needed. This ambitious phase 1 study builds on the authors' 2024 experience in Sci Tran Med with positive results with autologous transplantation of P63 progenitor cells in patients with COPD. The current study suggests that P63+ progenitor cell therapy is safe in patients with ILD. The authors attribute this to the acquisition of cells from a healthy upper lobe site, removed from the lung fibrosis. There are currently no cell-based therapies for ILD and in this regard the study is novel with important potential for clinical impact if validated in Phase 2 and 3 clinical trials.

Strengths:

This study addresses the need for an effective therapy for interstitial lung disease. It offers good evidence that the cells used for therapy are safe. In so doing it addresses a concern that some P63+ progenitor cells may be proinflammatory and harmful, as has been raised in the literature (articles which suggested some P63+ cells can promote honeycombing fibrosis; references 26 &35). The authors attribute the safety they observed (without proof) to the high HOPX expression of administered cells (a marker found in normal Type 1 AECs. The totality of the RNASeq suggests the cloned cells are not fibrogenic. They also offer exploratory data suggesting a relationship between clone roundness and PFT parameters (and a negative association between patient age and clone roundness).

Weaknesses:

The authors can conclude they can isolate, clone, expand, and administer P63+ progenitor cells safely; but with the small sample size and lack of a placebo group, no efficacy should be implied.

Specific points:

(1) The authors acknowledge most study weaknesses including the lack of a placebo group and the concurrent COVID-19 in half the subjects (the high-dose subjects). They indicate a phase 2 trial is underway to address these issues.

(2) The authors suggest an efficacy signal on pages 18 (improvement in 2 subjects' CT scans) and 21 (improvement in DLCO) but with such a small phase 1 study and such small increases in DLCO (+5.4%) the authors should refrain from this temptation (understandable as it is).

(3) Likewise most CT scans were unchanged and those that improved were in the mid-dose group (albeit DLCO improved in the 2 patients whose CT scans improved).

(4) The authors note an impressive 58m increase in 6MWTD in the high-dose group but again there is no placebo group, and the low-dose group has no net change in 6MWTD at 24 weeks.

(5) I also raise the question of the enrollment criteria in which 5 patients had essentially normal DLCO/VA values. In addition there is no discussion as to whether the transplanted stem cells are retained or exert benefit by a paracrine mechanism (which is the norm for cell-based therapies).

Reviewer #2 (Public review):

Summary:

This manuscript describes a first-in-human clinical trial of autologous stem cells to address IPF. The significance of this study is underscored by the limited efficacy of standard-of-care anti-fibrotic therapies and increasing knowledge of the role p63+ stem cells in lung regeneration in ARDS. While models of acute lung injury and p63+ stem cells have benefited from widespread and dynamic DAD and immune cell remodeling of damaged tissue, a key question in chronic lung disease is whether such cells could contribute to the remodeling of lung tissue that may be devoid of acute and dynamic injury. A second question is whether normal regions of the lung in an otherwise diseased organ can be identified as a source of "normal" p63+ stem cells, and how to assess these stem cells given recently identified p63+ stem cell variants emerging in chronic lung diseases including IPF. Lastly, questions of feasibility, safety, and efficacy need to be explored to set the foundation for autologous transplants to meet the huge need in chronic lung disease. The authors have addressed each of these questions to different extents in this initial study, which has yielded important if incomplete information for many of them.

Strengths:

As with a previous study from this group regarding autologous stem cell transplants for COPD (Ref. 24), they have shown that the stem cells they propagate do not form colonies in soft agar or cancers in these patients. While a full assessment of adverse events was confounded by a wave of Covid19 infections in the study participants, aside from brief fevers it appears these transplants are tolerated by these patients.

Weaknesses:

The source of stem cells for these autologous transplants is generally bronchoscopic biopsies/brushings from 5th-generation bronchi. Although stem cells have been cloned and characterized from nasal, tracheal, and distal airway biopsies, the systematic cloning and analysis of p63+ stem cells across the bronchial generations is less clear. For instance, p63+ stem cells from the nasal and tracheal mucosa appear committed to upper airway epithelia marked by 90% ciliated cells and 10% goblet cells (Kumar et al., 2011. Ref. 14). In contrast, p63+ stem cells from distal lung differentiate to epithelia replete with Club, AT2, and AT1 markers. The spectrum of p63+ stem cells in the normal bronchi of any generation is less studied. In the present study, cells are obtained by bronchoscopy from 3-5 generation bronchi and expanded by in vitro propagation. Single-cell RNAseq identifies three clusters they refer to as C1, C2, and C3, with the major C1 cluster said to have characteristics of airway basal cells and C2 possibly the same cells in states of proliferation. Perhaps the most immediate question raised by these data is the nature of the C1/C2 cells. Whereas they are clearly p63/Krt5+ cells as are other stem cells of the airways, do they display differentiation character of "upper airway" marked by ciliated/goblet cell differentiation or those of the lung marked by AT2 and AT1 fates? This could be readily determined by 3-D differentiation in so-called air-liquid interface cultures pioneered by cystic fibrosis investigators and should be done as it would directly address the validity of the sourcing protocol for autologous cells for these transplants. This would more clearly link the present study with a previous study from the same investigators (Shi et al., 2019, Ref. 9) whereby distal airway stem cells mitigated fibrosis in the murine bleomycin model. The authors should also provide methods by which the autologous cells are propagated in vitro as these could impact the quality and fate of the progenitor cells prior to transplantation.

The authors should also make a more concerted effort to compare Clusters 1, 2, and 3 with the variant stem cell identified in IPF (Wang et al., 2023, Ref. 27). While some of the markers are consistent with this variant stem cell population, others are not. A more detailed informatics analysis of normal stem cells of the airways and any variants reported could clarify whether the bronchial source of autologous stem cells is the best route to these transplants.

Other than these issues the authors should be commended for these first-in-human trials for this important condition.

Author response:

Reviewer #1 (Public review):

Summary:

IPF is a disease lacking regressive therapies which has a poor prognosis, and so new therapies are needed. This ambitious phase 1 study builds on the authors' 2024 experience in Sci Tran Med with positive results with autologous transplantation of P63 progenitor cells in patients with COPD. The current study suggests that P63+ progenitor cell therapy is safe in patients with ILD. The authors attribute this to the acquisition of cells from a healthy upper lobe site, removed from the lung fibrosis. There are currently no cell-based therapies for ILD and in this regard the study is novel with important potential for clinical impact if validated in Phase 2 and 3 clinical trials.

Strengths:

This study addresses the need for an effective therapy for interstitial lung disease. It offers good evidence that the cells used for therapy are safe. In so doing it addresses a concern that some P63+ progenitor cells may be proinflammatory and harmful, as has been raised in the literature (articles which suggested some P63+ cells can promote honeycombing fibrosis; references 26 &35). The authors attribute the safety they observed (without proof) to the high HOPX expression of administered cells (a marker found in normal Type 1 AECs. The totality of the RNASeq suggests the cloned cells are not fibrogenic. They also offer exploratory data suggesting a relationship between clone roundness and PFT parameters (and a negative association between patient age and clone roundness).

We thank the reviewer for the important comments.

Weaknesses:

The authors can conclude they can isolate, clone, expand, and administer P63+ progenitor cells safely; but with the small sample size and lack of a placebo group, no efficacy should be implied.

We thank the reviewer for the suggestion and agree that we should be more cautious to discuss the efficacy of current study.

Specific points:

(1) The authors acknowledge most study weaknesses including the lack of a placebo group and the concurrent COVID-19 in half the subjects (the high-dose subjects). They indicate a phase 2 trial is underway to address these issues.

N/A

(2) The authors suggest an efficacy signal on pages 18 (improvement in 2 subjects' CT scans) and 21 (improvement in DLCO) but with such a small phase 1 study and such small increases in DLCO (+5.4%) the authors should refrain from this temptation (understandable as it is).

We believe that exploring potential efficacy signal is also one important aim of this study in addition to safety evaluation. All these efficacy endpoint analyses had been planned in prior to the start of clinical trials (as registered in ClinicalTrial.gov) and the results anyhow need be analyzed and reported in the manuscript. And we will cautiously discuss the significance of the efficacy signal and avoid over-interpretation.

(3) Likewise most CT scans were unchanged and those that improved were in the mid-dose group (albeit DLCO improved in the 2 patients whose CT scans improved).

Yes, it is.

(4) The authors note an impressive 58m increase in 6MWTD in the high-dose group but again there is no placebo group, and the low-dose group has no net change in 6MWTD at 24 weeks.

Yes.

(5) I also raise the question of the enrollment criteria in which 5 patients had essentially normal DLCO/VA values. In addition there is no discussion as to whether the transplanted stem cells are retained or exert benefit by a paracrine mechanism (which is the norm for cell-based therapies).

Thank you for your detailed feedback.  The enrollment criteria are based on DLCO instead of DLCO/VA. And we would like to further discuss the possible benefit by paracrine mechanism in the revised manuscript.

Reviewer #2 (Public review):

Summary:

This manuscript describes a first-in-human clinical trial of autologous stem cells to address IPF. The significance of this study is underscored by the limited efficacy of standard-of-care anti-fibrotic therapies and increasing knowledge of the role p63+ stem cells in lung regeneration in ARDS. While models of acute lung injury and p63+ stem cells have benefited from widespread and dynamic DAD and immune cell remodeling of damaged tissue, a key question in chronic lung disease is whether such cells could contribute to the remodeling of lung tissue that may be devoid of acute and dynamic injury. A second question is whether normal regions of the lung in an otherwise diseased organ can be identified as a source of "normal" p63+ stem cells, and how to assess these stem cells given recently identified p63+ stem cell variants emerging in chronic lung diseases including IPF. Lastly, questions of feasibility, safety, and efficacy need to be explored to set the foundation for autologous transplants to meet the huge need in chronic lung disease. The authors have addressed each of these questions to different extents in this initial study, which has yielded important if incomplete information for many of them.

Strengths:

As with a previous study from this group regarding autologous stem cell transplants for COPD (Ref. 24), they have shown that the stem cells they propagate do not form colonies in soft agar or cancers in these patients. While a full assessment of adverse events was confounded by a wave of Covid19 infections in the study participants, aside from brief fevers it appears these transplants are tolerated by these patients.

We thank the reviewer for the important comments.

Weaknesses:

The source of stem cells for these autologous transplants is generally bronchoscopic biopsies/brushings from 5th-generation bronchi. Although stem cells have been cloned and characterized from nasal, tracheal, and distal airway biopsies, the systematic cloning and analysis of p63+ stem cells across the bronchial generations is less clear. For instance, p63+ stem cells from the nasal and tracheal mucosa appear committed to upper airway epithelia marked by 90% ciliated cells and 10% goblet cells (Kumar et al., 2011. Ref. 14). In contrast, p63+ stem cells from distal lung differentiate to epithelia replete with Club, AT2, and AT1 markers. The spectrum of p63+ stem cells in the normal bronchi of any generation is less studied. In the present study, cells are obtained by bronchoscopy from 3-5 generation bronchi and expanded by in vitro propagation. Single-cell RNAseq identifies three clusters they refer to as C1, C2, and C3, with the major C1 cluster said to have characteristics of airway basal cells and C2 possibly the same cells in states of proliferation. Perhaps the most immediate question raised by these data is the nature of the C1/C2 cells. Whereas they are clearly p63/Krt5+ cells as are other stem cells of the airways, do they display differentiation character of "upper airway" marked by ciliated/goblet cell differentiation or those of the lung marked by AT2 and AT1 fates? This could be readily determined by 3-D differentiation in so-called air-liquid interface cultures pioneered by cystic fibrosis investigators and should be done as it would directly address the validity of the sourcing protocol for autologous cells for these transplants. This would more clearly link the present study with a previous study from the same investigators (Shi et al., 2019, Ref. 9) whereby distal airway stem cells mitigated fibrosis in the murine bleomycin model. The authors should also provide methods by which the autologous cells are propagated in vitro as these could impact the quality and fate of the progenitor cells prior to transplantation.

We totally agree that the sub-population of the progenitor cells should be further analyzed. We would try this in the revised manuscript. And the methods to expand P63+ lung progenitor cells have been described in full details by Frank McKeon/Wa Xian group (Rao, et.al., STAR Protocols, 2020), which is adapted to pharmaceutical-grade technology patented by Regend Therapeutics, Ltd.

The authors should also make a more concerted effort to compare Clusters 1, 2, and 3 with the variant stem cell identified in IPF (Wang et al., 2023, Ref. 27). While some of the markers are consistent with this variant stem cell population, others are not. A more detailed informatics analysis of normal stem cells of the airways and any variants reported could clarify whether the bronchial source of autologous stem cells is the best route to these transplants. 

We thank for reviewer for the good suggestion and would like to make more detailed comparison in the revised manuscript.

Other than these issues the authors should be commended for these first-in-human trials for this important condition.

Thank you so much for the kind compliment.

eLife Assessment

The paper introduces DeepTX, a valuable deep-learning framework linking stochastic, mechanistic modelling with single-cell RNA sequencing data to investigate transcriptional burst kinetics on a genome-wide scale. This tool has been employed by the authors to evaluate transcriptional changes under DNA-damaging treatments, with observations that are of value to the systems biology and bioinformatics communities. The evidence supporting these findings is solid, though some concerns remain regarding specific technical details. This methodological advancement holds potential for application in diverse contexts, such as linking mechanistic models of signalling pathways to transcriptional data.

Joint Public Review:

In this work, the authors develop a new computational tool, DeepTX, for studying transcriptional bursting through the analysis of single-cell RNA sequencing (scRNA-seq) data using deep learning techniques. This tool aims to describe and predict the transcriptional bursting mechanism, including key model parameters and the steady-state distribution associated with the predicted parameters. By leveraging scRNA-seq data, DeepTX provides high-resolution transcriptional information at the single-cell level, despite the presence of noise that can cause gene expression variation. The authors apply DeepTX to DNA damage experiments, revealing distinct cellular responses based on transcriptional burst kinetics. Specifically, IdU treatment in mouse stem cells increases burst size, promoting differentiation, while 5FU affects burst frequency in human cancer cells, leading to apoptosis or, depending on the dose, to survival and potential drug resistance. These findings underscore the fundamental role of transcriptional burst regulation in cellular responses to DNA damage, including cell differentiation, apoptosis, and survival. Although the insights provided by this tool are mostly well supported by the authors' methods, certain aspects would benefit from further clarification.

The strengths of this paper lie in its methodological advancements and potential broad applicability. By employing the DeepTXSolver neural network, the authors efficiently approximate stationary distributions of mRNA counts through a mixture of negative binomial distributions, establishing a simple yet accurate mapping between the kinetic parameters of the mechanistic model and the resulting steady-state distributions. This innovative use of neural networks allows for efficient inference of kinetic parameters with DeepTXInferrer, reducing computational costs significantly for complex, multi-gene models. The approach advances parameter estimation for high-dimensional datasets, leveraging the power of deep learning to overcome the computational expense typically associated with stochastic mechanistic models. Beyond its current application to DNA damage responses, the tool can be adapted to explore transcriptional changes due to various biological factors, making it valuable to the systems biology, bioinformatics, and mechanistic modelling communities. Additionally, this work contributes to the integration of mechanistic modelling and -omics data, a vital area in achieving deeper insights into biological systems at the cellular and molecular levels.

This work also presents some weaknesses, particularly concerning specific technical aspects. The tool was validated using synthetic data, and while it can predict parameters and steady-state distributions that explain gene expression behaviour across many genes, it requires substantial data for training. The authors account for measurement noise in the parameter inference process, which is commendable, yet they do not specify the exact number of samples required to achieve reliable predictions. Moreover, the tool has limitations arising from assumptions made in its design, such as assuming that gene expression counts for the same cell type follow a consistent distribution. This assumption may not hold in cases where RNA measurement timing introduces variability in expression profiles.

The authors present a deep learning pipeline to predict the steady-state distribution, model parameters, and statistical measures solely from scRNA-seq data. Results across three datasets appear robust, indicating that the tool successfully identifies genes associated with expression variability and generates consistent distributions based on its parameters. However, it remains unclear whether these results are sufficient to fully characterise the transcriptional bursting parameter space. The parameters identified by the tool pertain only to the steady-state distribution of the observed data, without ensuring that this distribution specifically originates from transcriptional bursting dynamics.

A primary concern with the TXmodel is its reliance on four independent parameters to describe gene state-switching dynamics. Although this general model can capture specific cases, such as the refractory and telegraph models, accurately estimating the parameters of the refractory model using only steady-state distributions and typical cell counts proves challenging in the absence of time-dependent data.

The claim that the GO analysis pertains specifically to DNA damage response signal transduction and cell cycle G2/M phase transition is not fully accurate. In reality, the GO analysis yielded stronger p-values for pathways related to the mitotic cell cycle checkpoint signalling. As presented, the GO analysis serves more as a preliminary starting point for further bioinformatics investigation that could substantiate these conclusions. Additionally, while GSEA analysis was performed following the GO analysis, the involvement of the cardiac muscle cell differentiation pathway remains unclear, as it was not among the GO terms identified in the initial GO analysis.

As the advancement is primarily methodological, it lacks a comprehensive comparison with traditional methods that serve similar functions. Consequently, the overall evaluation of the method, including aspects such as inference accuracy, computational efficiency, and memory cost, remains unclear. The paper would benefit from being contextualised alongside other computational tools aimed at integrating mechanistic modelling with single-cell RNA sequencing data. Additional context regarding the advantages of deep learning methods, the challenges of analysing large, high-dimensional datasets, and the complexities of parameter estimation for intricate models would strengthen the work.

Author response:

Public Review:

In this work, the authors develop a new computational tool, DeepTX, for studying transcriptional bursting through the analysis of single-cell RNA sequencing (scRNA-seq) data using deep learning techniques. This tool aims to describe and predict the transcriptional bursting mechanism, including key model parameters and the steady-state distribution associated with the predicted parameters. By leveraging scRNA-seq data, DeepTX provides high-resolution transcriptional information at the single-cell level, despite the presence of noise that can cause gene expression variation. The authors apply DeepTX to DNA damage experiments, revealing distinct cellular responses based on transcriptional burst kinetics. Specifically, IdU treatment in mouse stem cells increases burst size, promoting differentiation, while 5FU affects burst frequency in human cancer cells, leading to apoptosis or, depending on the dose, to survival and potential drug resistance. These findings underscore the fundamental role of transcriptional burst regulation in cellular responses to DNA damage, including cell differentiation, apoptosis, and survival. Although the insights provided by this tool are mostly well supported by the authors' methods, certain aspects would benefit from further clarification.

The strengths of this paper lie in its methodological advancements and potential broad applicability. By employing the DeepTXSolver neural network, the authors efficiently approximate stationary distributions of mRNA count through a mixture of negative binomial distributions, establishing a simple yet accurate mapping between the kinetic parameters of the mechanistic model and the resulting steady-state distributions. This innovative use of neural networks allows for efficient inference of kinetic parameters with DeepTXInferrer, reducing computational costs significantly for complex, multi-gene models. The approach advances parameter estimation for high-dimensional datasets, leveraging the power of deep learning to overcome the computational expense typically associated with stochastic mechanistic models. Beyond its current application to DNA damage responses, the tool can be adapted to explore transcriptional changes due to various biological factors, making it valuable to the systems biology, bioinformatics, and mechanistic modelling communities. Additionally, this work contributes to the integration of mechanistic modelling and -omics data, a vital area in achieving deeper insights into biological systems at the cellular and molecular levels.

We thank the reviewers for their positive opinion on our manuscript. As reflected in our detailed responses to the reviewers’ comments, we will make significant changes to address their concerns comprehensively.

This work also presents some weaknesses, particularly concerning specific technical aspects. The tool was validated using synthetic data, and while it can predict parameters and steady-state distributions that explain gene expression behaviour across many genes, it requires substantial data for training. The authors account for measurement noise in the parameter inference process, which is commendable, yet they do not specify the exact number of samples required to achieve reliable predictions. Moreover, the tool has limitations arising from assumptions made in its design, such as assuming that gene expression counts for the same cell type follow a consistent distribution. This assumption may not hold in cases where RNA measurement timing introduces variability in expression profiles.

Thank you for your detailed and constructive feedback on our work. We will address the key concerns raised from the following points:

(1) Clarification on the required sample size: We tested the robustness of our inference method on simulated datasets by varying the number of single-cell samples. Our results indicated that the predictions of burst kinetics parameters become accurate when the number of cells reaches 500 (Supplementary Figure S3d, e). This sample size is smaller than the data typically obtained with current single-cell RNA sequencing (scRNA-seq) technologies, such as 10x Genomics and Smart-seq3 (Zheng GX et al., 2017; Hagemann-Jensen M et al., 2020). Therefore, we believed that our algorithm is well-suited for inferring burst kinetics from existing scRNA-seq datasets, where the sample size is sufficient for reliable predictions. We will clarify this point in the main text to make it easier for readers to use the tool.

(2) Assumption-related limitations: One of the fundamental assumptions in our study is that the expression counts of each gene are independently and identically distributed (i.i.d.) among cells, which is a commonly adopted assumption in many related works (Larsson AJM et al., 2019; Ochiai H et al., 2020; Luo S et al., 2023). However, we acknowledged the limitations of this assumption. The expression counts of the same gene in each cell may follow distinct distributions even from the same cell type, and dependencies between genes could exist in realistic biological processes. We recognized this and will deeply discuss these limitations from assumptions and prospect as an important direction for future research.

The authors present a deep learning pipeline to predict the steady-state distribution, model parameters, and statistical measures solely from scRNA-seq data. Results across three datasets appear robust, indicating that the tool successfully identifies genes associated with expression variability and generates consistent distributions based on its parameters. However, it remains unclear whether these results are sufficient to fully characterize the transcriptional bursting parameter space. The parameters identified by the tool pertain only to the steady-state distribution of the observed data, without ensuring that this distribution specifically originates from transcriptional bursting dynamics.

We appreciate your insightful comments and the opportunity to clarify our study’s contributions and limitations. Although we agree that assessing whether the results from these three realistic datasets can represent the characterize transcriptional burst parameter space is challenging, as it depends on data property and conditions in biology, we firmly believe that DeepTX has the capacity to characterize the full parameter space. This believes stems from the extensive parameters and samples we input during model training and inference across a sufficiently large parameter range (Method 1.3). Furthermore, the training of the model is both flexible and scalable, allowing for the expansion of the transcriptional burst parameter space as needed. We will clarify this in the text to enable readers to use DeepTX more flexibly.

On the other hand, we agree that parameter identification is based on the steady-state distribution of the observed data (static data), which loses information about the fine dynamic process of the burst kinetics. In principle, tracking the gene expression of living cells can provide the most complete information about real-time transcriptional dynamics across various timescales (Rodriguez J et al., 2019). However, it is typically limited to only a small number of genes and cells, which could not investigate general principles of transcriptional burst kinetics on a genome-wide scale. Therefore, leveraging the both steady-state distribution of scRNA-seq data and mathematical dynamic modelling to infer genome-wide transcriptional bursting dynamics represents a critical and emerging frontier in this field. For example, the statistical inference framework based on the Markovian telegraph model, as demonstrated in (Larsson AJM et al., 2019), offers a valuable paradigm for understanding underlying transcriptional bursting mechanisms. Building on this, our study considered a more generalized non-Mordovian model that better captures transcriptional kinetics by employing deep learning method under conditions such as DNA damage. This provided a powerful framework for comparative analyses of how DNA damage induces alterations in transcriptional bursting kinetics across the genome. We will highlight the limitations of current inference using steady-state distributions in the text and look ahead to future research directions for inference using time series data across the genome.

A primary concern with the TXmodel is its reliance on four independent parameters to describe gene state-switching dynamics. Although this general model can capture specific cases, such as the refractory and telegraph models, accurately estimating the parameters of the refractory model using only steady-state distributions and typical cell counts proves challenging in the absence of time-dependent data.

We thank you for highlighting this critical concern regarding the TXmodel's reliance on four independent parameters to describe gene state-switching dynamics. We acknowledge that estimating the parameters of the TXmodel using only steady-state distributions and typical single-cell RNA sequencing (scRNA-seq) data poses significant challenges, particularly in the absence of time-resolved measurements.

As described in the response of last point, while time-resolved data can provide richer information than static scRNA-seq data, it is currently limited to a small number of genes and cells, whereas static scRNA-seq data typically capture genome-wide expression. Our framework leverages deep learning methods to link mechanistic models with static scRNA-seq data, enabling the inference of genome-wide dynamic behaviors of genes. This provides a potential pathway for comparative analyses of transcriptional bursting kinetics across the entire genome.

Nonetheless, the refractory model and telegraphic model are important models for studying transcription bursts. We will discuss and compare them in terms of the accuracy of inferred parameters. Certainly, we agree that inferring the molecular mechanisms underlying transcriptional burst kinetics using time-resolved data remains a critical future direction. We will include a brief discussion on the role and importance of time-resolved data in addressing these challenges in the discussion section of the revised manuscript.

The claim that the GO analysis pertains specifically to DNA damage response signal transduction and cell cycle G2/M phase transition is not fully accurate. In reality, the GO analysis yielded stronger p-values for pathways related to the mitotic cell cycle checkpoint signalling. As presented, the GO analysis serves more as a preliminary starting point for further bioinformatics investigation that could substantiate these conclusions. Additionally, while GSEA analysis was performed following the GO analysis, the involvement of the cardiac muscle cell differentiation pathway remains unclear, as it was not among the GO terms identified in the initial GO analysis.

We thank the reviewer for this valuable feedback and for pointing out the need for clarification regarding the GO and GSEA analyses. We agree that the connection between the cardiac muscle cell differentiation pathway identified in the GSEA analysis and the GO terms from the initial analysis requires further clarification. This discrepancy arises because GSEA examines broader sets of pathways and may capture biological processes not highlighted by GO analysis due to differences in the statistical methods and pathway definitions used. We will revise the manuscript to address this point, explicitly discussing the distinct yet complementary nature of GO and GSEA analyses and providing a clearer interpretation of the results.

As the advancement is primarily methodological, it lacks a comprehensive comparison with traditional methods that serve similar functions. Consequently, the overall evaluation of the method, including aspects such as inference accuracy, computational efficiency, and memory cost, remains unclear. The paper would benefit from being contextualised alongside other computational tools aimed at integrating mechanistic modelling with single-cell RNA sequencing data. Additional context regarding the advantages of deep learning methods, the challenges of analysing large, high-dimensional datasets, and the complexities of parameter estimation for intricate models would strengthen the work.

We greatly appreciate your insightful feedback, which highlights important considerations for evaluating and contextualizing our methodological advancements. Below, we emphasize our advantages from both the modeling perspective and the inference perspective compared with previous model. As our work is rooted in a model-based approach to describe the transcriptional bursting process underlying gene expression, the classic telegraph model (Markovian) and non-Markovian models which are commonly employed are suitable for this purpose:

Classic telegraph model: The classic telegraph model allows for the derivation of approximate analytical solutions through numerical integration, enabling efficient parameter point estimation via maximum likelihood methods, e.g., as explored in (Larsson AJM et al., 2019). Although exact analytical solutions for the telegraph model are not available, certain moments of its distribution can be explicitly derived. This allows for an alternative approach to parameter inference using moment-based estimation methods, e.g., as explored in (Ochiai H et al., 2020). However, it is important to note that higher-order sample moments can be unstable, potentially leading to significant estimation bias.

Non-Markovian Models: For non-Markovian models, analytical or approximate analytical solutions remain elusive. Previous work has employed pseudo-likelihood approaches, leveraging statistical properties of the model’s solutions to estimate parameters, e.g., as explored in (Luo S et al., 2023). However, the method may suffer from low inference efficiency.

In our current work, we leverage deep learning to estimate parameters of TXmodel, which is non-Markovian model. First, we represent the model's solution as a mixture of negative binomial distributions, which is obtained by the deep learning method. Second, through integration with the deep learning architecture, the model parameters can be optimized using automatic differentiation, significantly improving inference efficiency. Furthermore, by employing a Bayesian framework, our method provides posterior distributions for the estimated dynamic parameters, offering a comprehensive characterization of uncertainty. Compared to traditional methods such as moment-based estimation or pseudo-likelihood approaches, we believe our approach not only achieves higher inference efficiency but also delivers posterior distributions for kinetics parameters, enhancing the interpretability and robustness of the results. We will present and emphasize the computational efficiency and memory cost of our methods the revised version.

Reference

Zheng, G.X., Terry, J.M., Belgrader, P., Ryvkin, P., Bent, Z.W., Wilson, R., Ziraldo, S.B., Wheeler, T.D., McDermott, G.P., Zhu, J., Gregory, M.T., Shuga, J., Montesclaros, L., Underwood, J.G., Masquelier, D.A., Nishimura, S.Y., Schnall-Levin, M., Wyatt, P.W., Hindson, C.M., Bharadwaj, R., Wong, A., Ness, K.D., Beppu, L.W., Deeg, H.J., McFarland, C., Loeb, K.R., Valente, W.J., Ericson, N.G., Stevens, E.A., Radich, J.P., Mikkelsen, T.S., Hindson, B.J., Bielas, J.H. 2017. Massively parallel digital transcriptional profiling of single cells. Nature Communications 8: 14049. DOI: https://dx.doi.org/10.1038/ncomms14049, PMID: 28091601

Hagemann-Jensen, M., Ziegenhain, C., Chen, P., Ramsköld, D., Hendriks, G.J., Larsson, A.J.M., Faridani, O.R., Sandberg, R. 2020. Single-cell RNA counting at allele and isoform resolution using Smart-seq3. Nat Biotechnol 38: 708-714. DOI: https://dx.doi.org/10.1038/s41587-020-0497-0, PMID: 32518404

Larsson, A.J.M., Johnsson, P., Hagemann-Jensen, M., Hartmanis, L., Faridani, O.R., Reinius, B., Segerstolpe, A., Rivera, C.M., Ren, B., Sandberg, R. 2019. Genomic encoding of transcriptional burst kinetics. Nature 565: 251-254. DOI: https://dx.doi.org/10.1038/s41586-018-0836-1, PMID: 30602787

Ochiai, H., Hayashi, T., Umeda, M., Yoshimura, M., Harada, A., Shimizu, Y., Nakano, K., Saitoh, N., Liu, Z., Yamamoto, T., Okamura, T., Ohkawa, Y., Kimura, H., Nikaido, I. 2020. Genome-wide kinetic properties of transcriptional bursting in mouse embryonic stem cells. Science Adavances 6: eaaz6699. DOI: https://dx.doi.org/10.1126/sciadv.aaz6699, PMID: 32596448

Luo, S., Wang, Z., Zhang, Z., Zhou, T., Zhang, J. 2023. Genome-wide inference reveals that feedback regulations constrain promoter-dependent transcriptional burst kinetics. Nucleic Acids Research 51: 68-83. DOI: https://dx.doi.org/10.1093/nar/gkac1204, PMID: 36583343

Rodriguez, J., Ren, G., Day, C.R., Zhao, K., Chow, C.C., Larson, D.R. 2019. Intrinsic dynamics of a human gene reveal the basis of expression heterogeneity. Cell 176: 213-226.e218. DOI: https://dx.doi.org/10.1016/j.cell.2018.11.026, PMID: 30554876

Luo, S., Zhang, Z., Wang, Z., Yang, X., Chen, X., Zhou, T., Zhang, J. 2023. Inferring transcriptional bursting kinetics from single-cell snapshot data using a generalized telegraph model. Royal Society Open Science 10: 221057. DOI: https://dx.doi.org/10.1098/rsos.221057, PMID: 37035293

eLife Assessment

This study addresses fundamental questions surrounding otitis media effusion in Down syndrome, identifying DYRK1A as a key gene involved in the condition. The findings are compelling, highlighting DYRK1A as a promising therapeutic target for addressing hearing loss in patients with Down syndrome.

Reviewer #1 (Public review):

Summary:

In this study, Hilda Tateossian et al. sought to identify the specific gene linked to hearing loss caused by otitis media effusion (OME) in individuals with Down syndrome (DS). They approached this by analyzing a series of mouse models of DS (referred to as the DpTyb lines), which include various duplications that encompass the regions of the mouse genome analogous to the human chromosome 21 (Hsa21). This allowed them to pinpoint genetic loci that may be associated with OME in DS. To control for external variables, such as genetic background and environmental influences, which could affect the development of chronic OME, all DpTyb mouse lines were maintained on a uniform C57BL/6J genetic background. The authors could show that chronic OME phenotypes were consistently reproducible across two research centers, the Francis Crick Institute and MRC Harwell Institute, supporting their conclusion while also reducing the likelihood that environmental factors could affect results.

The authors then focused on a significant locus on chromosome 16 in the Dp5Tyb mouse model that was strongly associated with OME. This locus contains only 12 genes, and it overlapped with the duplicated genomic regions in three additional mouse models (Dp1Tyb, Dp3Tyb, and Ts1Rhr), strengthening the link between this locus and OME. To identify the gene responsible within this critical interval, they conducted targeted crosses of Dp mouse lines (Dp1Tyb, Dp3Tyb, and Dp5Tyb) with gene knockout models. This strategy enabled them to normalize the copy number of specific genes within the progeny and assess the effect on OME. They found that reducing the gene dosage of Dyrk1a specifically restored a wild-type phenotype, implicating Dyrk1a as a key player in the development of OME in DS.

Given the broad biological roles of DYRK1A in various cellular pathways, the researchers also explored its effects on downstream proteins and pathways within the middle ear epithelium using immunohistochemistry and RT-qPCR. They uncovered several pathological mechanisms by which DYRK1A triplication could promote middle ear inflammation and increased vascular permeability. These mechanisms included the interaction between DYRK1A and TGFβ signaling, which affects proinflammatory cytokines IL-6 and IL-17, as well as elevated levels of VEGF in the middle ear that were accompanied by increased Hif1a expression.

At the morphological level, analyses by scanning electron microscopy further revealed a loss of cilia on the epithelial cells in the middle ears of 2-month-old Dp3Tyb and Dp5Tyb mutant mice, which likely contributes to the development of OME in DS.

Finally, to validate the relevance of their findings in humans, the researchers examined the expression of the 12 genes within the Dp5Tyb locus in samples from children with DS compared to unaffected parental controls, using qPCR. They found that among the 12 genes, DYRK1A showed the most significant fold increase in expression, further supporting its potential role in OME associated with DS.

Strengths:

(1) The manuscript is well-written and clearly presents both experimental design and results, together supporting the main conclusions.

(2) The experiments are carefully designed and executed, with data that convincingly support the identification of DYRK1A as a key gene involved in OME in DS. The use of gene knockouts to normalize Dyrk1a gene dosage within the Dp mouse lines was a thorough and successful strategy to strengthen and validate DYRK1A's causal inference in OME.

(3) The study goes beyond simple gene identification by exploring the downstream pathways and cellular effects of DYRK1A triplication. This mechanistic focus provides actionable insights into the potential molecular underpinnings of OME in DS.

(4) The study addresses a clinically important issue - OME in children with DS - and proposes DYRK1A as a practical therapeutic target. Based on data in mice and the high dose of DYRK1A in human clinical samples, the authors suggest that suppressing the activity of this gene by localized delivery of inhibitors to the middle ear cavity in DS patients can be a potential strategy for future treatment of OME.

Weaknesses:

No major weakness is identified.

The authors could discuss further the potential involvement of the other genes within the Dp5Tyb interval, and whether interactions among these genes could impact the disease or whether additional contributions to OME might be overlooked. Beyond DYRK1A expression, discussion of a more extensive analysis of the other genes within the locus in larger cohorts of individuals with DS and OME could add strength to the translational relevance of the findings.

Reviewer #2 (Public review):

This manuscript investigates the genetic basis of otitis media with effusion (OME) in children with Down syndrome (DS). Utilizing an impressive number of mouse models, the study identifies a significant locus on mouse chromosome 16 that contributes to the development of OME. Notably, the gene Dyrk1a is identified as a critical factor for OME in DS; Normalizing Dyrk1a dosage in Dp3Tyb mice restores the wild-type phenotype, highlighting its major contribution to OME in DS. The research also explores the downstream pathways affected by DYRK1A, revealing interactions with TGFβ signaling and the modulation of pro-inflammatory cytokines like IL-6 and IL-17, as well as increased VEGF levels linked to middle ear inflammation.

This work is novel in its comprehensive approach to linking specific genetic loci and genes to the development of OME in DS, and offers a refined genetic analysis, pinpointing Dyrk1a as a key gene. Additionally, the identification of some of the signaling pathways involved provides new insights into the pathophysiology of OME in DS. The findings have significant clinical implications, as they suggest that targeting Dyrk1a could be a potential therapeutic strategy for managing OME in children with DS. This could lead to improved treatment options that go beyond current surgical interventions, reducing the need for repeated tympanostomy tube placements and potentially mitigating the associated risks. Overall, this research enhances our understanding of the genetic factors underlying OME in DS, motivates future studies on the newly identified genetic loci, and opens avenues for future therapeutic developments.

Strengths:

(1) Robust methodology: The use of a comprehensive set of mouse models allows for precise localization of genetic loci associated with OME, an advancement over previous studies.

(2) Identification of Key Genes: The clear demonstration of Dyrk1a's role in OME provides a strong basis for further exploration of targeted therapies.

(3) Pathway Insights: The exploration of signaling pathways, including TGFβ and IL-6 interactions, enriches the discussion around the inflammatory mechanisms that contribute to OME in DS.

Weaknesses:

(1) Limited Human Data: While the mouse models are robust, the translation of findings to human populations could be further strengthened with comparative studies.

(2) Pathway Complexity: The study primarily focuses on Dyrk1a and its immediate inflammatory pathways, which may oversimplify the multifactorial nature of OME in DS; exploring additional genetic interactions, and further exploring the implications of the potential ciliogenesis role of DYRK1A in OME could provide a more complete view.

This study is a valuable contribution to the field of genetic research in Down syndrome, providing critical insights that could inform future therapeutic strategies for managing OME. The implications for treatment and understanding of DS phenotypes in mouse models are particularly noteworthy. The findings are well-supported and present clear avenues for further research.

Reviewer #3 (Public review):

Summary:

The authors used mouse models with nested duplications of genomic regions syntenic to human chromosome 21 to identify specific loci responsible for otitis media with effusion (OME) in people with Down syndrome. They identified two loci: one highly penetrant major locus containing the candidate gene Dyrk1a and one minor locus resulting in low penetrant OME. By normalizing the gene dosage of Dyrk1a, the authors showed it mitigated OME. Further investigation of the molecular mechanisms by which DYRK1A exerts its effect, unveiled interactions with TGFβ signaling, elevated proinflammatory cytokines (IL-6 and IL-17), and increased VEGF levels coupled with increased Hif1a activity in the middle ear.

Strengths:

(1) The manuscript is well-written and includes appropriate figures. I especially liked Figure 4, which provides an excellent graphical abstract for the genetic study.

(2) Using a panel of mouse models with nested duplications is an elegant, systematic approach to narrowing down the genetic loci linked to OME. This is a robust method for dissecting complex traits like those observed in Down syndrome.

(3) Identifying DYRK1A as a major genetic contributor to highly penetrant OME in DS could be extrapolated to individuals with isolated (nonsyndromic) OME, thus paving the way for broader exploration of its role in general OME susceptibility. This discovery also opens the door to developing genetic testing for individuals with recurrent or chronic OME, helping with diagnosis and personalized management.

(4) Identifying DYRK1A as a potential therapeutic target highlights the study's translational relevance and potential impact on treating OME in children with DS.

Weaknesses:

(1) While the mouse model findings are robust, the study lacks validation in humans. Collaborating with researchers studying OM in human cohorts to screen for DYRK1A variants and correlate these to human phenotypes could have significantly strengthened the study's translational relevance.

(2) More compelling evidence could have been provided by generating a DYRK1A overexpression knock-in mouse model in the ROSA26 locus. This approach would allow for the functional evaluation of the impact of the overexpression of this single gene. The authors could make the KI model inducible allowing for a more localized study of the gene in a subset of cells.

(3) The lack of histological findings in the cochlea does not rule out sensorineural hearing loss. The authors did not provide compelling evidence ruling out a sensorineural component. Given DYRK1A expression in various cochlear cell types (according to the gEAR resource), it is plausible that overexpression could cause dysfunction there too. Additional analysis of ABR waves, including amplitude and latency measurements, would help clarify whether the defect is exclusively middle ear-related.

(4) Although Dyrk1a is implicated as a critical gene, the study does not fully explore the potential contributions of the other 11 genes in the identified locus. These genes might also play roles in OME, whether independently or synergically.

(5) While TGFβ signaling and cytokine production are investigated, the study does not explore the full and broader pathway and network interactions. Using transcriptomics in these mice models could provide a deeper and more comprehensive understanding of the molecular mechanisms involved.

(6) The difference in wild-type phenotype restoration between double mutants: Dp3Tyb has the best rescue with no significant difference with wild type, versus Dp5Tyb failing to restore the wild-type phenotype needs further investigation. Understanding the factors accounting for these differences could identify additional modifiers within this locus.

(7) The authors stated, "We detected a one-third increase, as expected, of the number of cells positive for DYRK1A in Dp3Tyb mice (56.6%) compared to wild-type littermates (36.4%)". This measurement refers to the number of cells expressing DYRK1A rather than the actual level of DYRK1A protein expression within these cells. The number of expressing cells does not directly correlate with gene dosage, as it is likely the level of DYRK1A protein within individual cells that has a more significant impact on the phenotype. The authors should quantify the protein levels using Western blot, for example, to strengthen their findings. If the authors believe it is the number of expressing cells that is relevant, then they should provide a clear rationale for how this measure reflects gene dosage effects and its biological significance in this context.

eLife Assessment

This valuable study presents an interesting analysis of the role of the polyamine precursor putrescine in the pili-dependent surface motility of a laboratory strain of Escherichia coli. The genetic data convincingly shows this role, yet without putrescine measurements, the evidence remains incomplete. This work will be of interest to biomedical scientists studying uropathogenic bacteria, and those studying bacterial polyamine function.

Reviewer #1 (Public review):

Summary:

Ita Mehta and colleagues have investigated the role of putrescine in the pili-dependent surface motility of a laboratory strain of Escherichia coli. Enterobacteria, and particularly E. coli and Salmonella Typhimurium contain an enormous amount of putrescine and cadaverine compared to other bacteria. It has been estimated by Igarashi and colleagues that putrescine is present in E. coli at levels of at least 30 mM. Therefore, an investigation of the role of putrescine in E. coli is a welcome and important contribution to understanding polyamine function. The authors have used a comprehensive suite of E. coli gene deletion strains of putrescine biosynthetic, transport, and catabolic genes to understand the role of putrescine in pili-dependent surface motility.

Strengths:

Single gene deletions of arginine decarboxylase (speA) and agmatine ureohydrolase (speB), and a double gene deletion of the constitutive ornithine decarboxylase (speC) and the acid-inducible ornithine decarboxylase (speF), all of which are involved in putrescine biosynthesis, were found by the authors to be less efficient at pili-dependent surface motility. In addition, the putrescine transport genes plaP and potF are also required for efficient pili-dependent surface motility. Furthermore, the putrescine catabolic genes patA and puuA, when co-deleted, reduce pili-dependent surface motility. Transcriptomic analysis of the agmatine ureohydrolase (speB) gene deletion strain compared to the parental strain indicates a coordinated response to the speB gene deletion, including upregulation of ornithine biosynthetic genes and a downregulation of energy metabolic genes.

Weaknesses:

Because the cellular content of putrescine and other polyamines in the E. coli strains was not measured at any point in this study, and the gene deletions were not genetically complemented, it is not possible to definitively attribute physiological changes to the gene deletion strains specifically to changes in putrescine levels. Furthermore, the GT medium used for the mobility experiments contains trypsinated casein (tryptone), which may contain polyamines and most certainly contain arginine. There are two modes of putrescine biosynthesis in E. coli: one mode is the direct formation of putrescine from L-ornithine mediated by ornithine decarboxylase, and the other is the indirect pathway involving decarboxylation of arginine to form agmatine, followed by hydrolysis of agmatine to form putrescine and urea. In the absence of external arginine, putrescine is made by ornithine decarboxylase, however, in the presence of external arginine, ornithine biosynthesis is repressed and arginine decarboxylase becomes the primary biosynthetic route for putrescine biosynthesis. The GT medium used by the authors will tend to favor putrescine production from arginine. The speB gene deletion, which is used for the transcriptomic analyses, will even in the absence of external arginine, accumulate a very large amount of agmatine, greater than the level of putrescine. This will confound the interpretation of the effect of the speB gene deletion, because agmatine accumulation may be responsible for some of the effects, and the addition of external putrescine may repress agmatine accumulation. In the absence of polyamine level measurements, the relative levels of agmatine, the putrescine structural analog cadaverine, and the accumulation of decarboxylated S-adenosylmethionine, are not known. Changes to these metabolites could affect pili-dependent surface motility. Furthermore, it is not possible to conclude that the effects of gene deletions to biosynthetic, transport or catabolic genes on pili-dependent surface motility are due to changes in putrescine levels unless one takes it on faith that there must be changes to putrescine levels. Since E. coli contains such an enormous amount of putrescine, it is important to know how much putrescine must be depleted in order to exert a physiological effect.

The authors have tackled an important biomedical problem relevant to infections of the urogenital tract and also important for understanding the very unusual high level of putrescine in E. coli and related species. However, without confirmation of putrescine levels in their various strains, it would be difficult to unequivocally conclude that putrescine, or changes to its concentrations, are responsible for the physiological changes seen with the gene deletion strains.

Reviewer #2 (Public review):

Summary:

Mehta et al., in constructing E. coli strains unable to synthesize polyamines, noted that strains deficient in putrescine synthesis showed decreased movement on semisolid agar. They show that strains incapable of synthesizing putrescine have decreased expression of Type I pilin and, hence, decreased ability to perform pilin-dependent surface motility.

Strengths:

The authors characterize the specific polyamine pathways that are important for this phenomenon. RNAseq provides a detailed overview of gene expression in the strain lacking putrescine. The data suggest homeostatic control of polyamine synthesis and metabolic changes in response to putrescine.

Weaknesses:

In this version, the authors ignore phase variation of the pil operon promoter, which can be monitored via PCR. The gene expression data suggest that shifting to the pilin "off" state could help explain the phenotype.

Reviewer #3 (Public review):

Summary:

This study by Mehta et al. describes the mechanisms behind the observation that putrescine biosynthesis mutants in Escherichia coli strain W3110 are affected by surface motility. The manuscript shows that the surface motility phenotype is dependent on Type I fimbriae and that putrescine levels affect the expression level of fimbriae. The results further suggest that without putrescine, the metabolism of the cell is shifted towards the production of putrescine and away from energy metabolism.

There are two main aspects in the manuscript.

(1) The first observation is that a fimA mutant modified/decreased the motility phenotype. From this result, the authors conclude that type I fimbriae (or pili) are involved in the surface motility phenotype. Type I fimbriae are typically known to be involved in non-motile phenotypes, such as biofilm formation or adhesion. Type I fimbriae are also co-regulated with other surface structures that might impact motility. Thus, more controls are needed before concluding that the surface motility requires the type I fimbriae. For instance, the authors should have complemented the mutants and should have verified the flagella expression/motility in the fimA mutant.

(2) The second observation is that putrescine also impacts the surface motility phenotype and the expression of type I fimbriae. Although there is no genetic complementation, here the exogenous addition of putrescine to the speB mutant provides a chemical complementation method, which makes the data stronger.

In addition, testing the effect of putrescine on motility and type I fimbriae expression in additional strains of E. coli would strengthen the conclusion. This is especially important since the results are somewhat different from previous results obtained with a different strain of E. coli. The authors do note that experimental conditions are different, but testing their theory would make the conclusions stronger.

eLife Assessment

This valuable study combines experiments and modelling to advance our understanding of the nonlinear nature of homeostatic structural plasticity and its interaction with synaptic scaling. The methodology and findings are solid, although additional work is needed to better link models with experiments and support some of the conclusions drawn. This study will be of interest to theoretical and experimental neuroscientists working in homeostatic plasticity.

Reviewer #1 (Public review):

This manuscript investigates homeostatic structural plasticity and its interplay with synaptic scaling. It uses an integrated approach with models and experiments.<br /> First, electrophysiology and chronic imaging are used to investigate the influence of different levels of AMPA-receptor antagonist NBQX, which allows for gradual activity reduction. Low levels of NBQX lead to a decrease of activity and a homeostatic increase of synapse density, whereas high levels block neural activity and lead to a reduced number of synapses after 3 days. The authors conclude that there must be a non-linear dependency between neuronal activities and rewiring. As a mathematical model for this, a biphasic structural plasticity rule is used, which, for increasing neural activities, switches from net synapse removal to growth and back, yielding two stable states at zero activity and the homeostatic target.<br /> This rule is tested in various situations in silico, yet without attempting to reproduce the experiment. First, in network development, the biphasic rule generates a lot of unconnected silent neurons and a reasonable network structure only emerges when the neurons are additionally supported by a facilitating input current. For comparison, a linear and a simpler nonlinear homeostatic plasticity model, which had been ruled out by the experimental data, need no external drive. Second, the consequences of lasting, altered stimulation in a subgroup of neurons is explored. As expected by the design of the rule, a small increase and decrease in stimulation leads to a decrease and increase of synaptic connectivity, respectively, and stimulation silencing led to a complete disconnection of the sub-population with restoration of activity. Unlike in previous studies, an asymmetry of pre- and postsynaptic plasticity mechanisms cannot rescue this. Third, silencing only for a short time period and then overstimulating the network led to overly strong activity, which may, however, also hold without silencing. For a transiently silenced stimulation, recovery is possible, but only when there is enough recurrent excitation from the rest of the network.<br /> Following this, the second part of the manuscript explores whether synaptic scaling may adapt and up-regulate the recurrent excitation, such that activity in a normally silenced subpopulation can be restored. Indeed, fast enough synaptic scaling leads to a recovery of neuronal activity in simulations, but leads to highly synchronous activity. A systematic model analysis shows at which scaling and rewiring speeds the activity and connectivity for a silenced sub-population can be restored. In between, however, the authors analyze spine sizes and changes in their whole population AMPAR-blocking experiments that demonstrate synaptic scaling and that structural plasticity and scaling effects may be jointly regulated. This experimental "break" between a simulation and its systematic analysis makes the paper harder to read and seems unnecessary as the analyses from the experiments are not repeated for the model.

Overall, the combination of experiments and simulations is a promising approach to investigate network self-organization. Especially the gradual blocking of activity is very valuable to inform mathematical models and distinguish them from alternatives. However, it remains unclear whether the model would actually reproduce the experiment. When switching from one to the other, this entails a detour to the conceptual level which makes the narrative sometimes hard to follow.

In summary, this manuscript makes a valuable contribution to discern the mathematical shape of a homeostatic structural plasticity model and understanding the necessity of synaptic scaling in the same network. Both experimental and computational methods are solid and well described. Yet, both parts could be linked better in order to obtain conclusions with more impact and generality.

Reviewer #2 (Public review):

This manuscript by Lu et al addresses the understudied interplay between structural and functional changes underlying homeostatic plasticity. Using hippocampal organotypic slice cultures allowing chronic imaging of dendritic spines, the authors showed that a partial or complete inhibition of AMPA-type glutamate receptors differentially affects spine density, respectively leading to an increase or decrease. Based on that dataset, they built a model where activity-dependent synapse formation is regulated by a biphasic rule and tested it in stimulation- or deprivation-induced homeostatic plasticity. The model matches experimental data (from the authors and the literature) quite well, and provides a framework within which functional and structural changes coexist to regulate firing rate homeostasis.

While the correlation between changes in AMPAR numbers and in spine number/size has been well characterized during Hebbian plasticity, the situation is much less clear in homeostatic plasticity due to multiple studies yielding diverging results. This manuscript adds new experimental results to the existing data and presents a valuable effort to generate a model that can explain these divergences in a unifying framework.

The model and its successive implantation steps are well presented along a clear thread. However, the manuscript would benefit from clarifications at several key points (Hebbian vs homeostatic timeline).

First of all, it would have benefited from having an actual timeline of structural changes throughout the three days of AMPAR inhibition, especially as their experimental model allows it. This would have provided much-needed and otherwise entirely lacking information on spine dynamics (especially on transient spines) and on the respective timescale of the structural and functional changes, instead of modelling an entire timeline based solely on an experimental endpoint.

Additionally, the model would have been strengthened by an experimental dataset with homeostatic plasticity induced by higher activity (e.g. with bicuculline). To the best of my knowledge, there is currently no data on structural plasticity following scaling down, and it is also known that scaling up and down are mediated by different molecular pathways. The extension of the model from scaling up (in response to silencing) to scaling down (in response to increased activity) offers an interesting perspective, but its biological relevance is limited as there is no experimental data to support it.

Finally, the difference between weak and complete inhibition could have been more extensively characterized. The authors focus indeed on the effects of either condition on spine number, but only integrate synaptic weights following complete inhibition. This is a pity, as they show some intriguing data suggesting a differential effect on spine size by partial or complete AMPAR inhibition (although further work is required to support some of their interpretations). Since the model aims at correlating structural and functional homeostatic plasticity, the fact that it is only demonstrated for one of the two conditions tested severely undermines the claims of the authors in the discussion that the model tackles that question.

Author response:

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

Reviewer #1 (Recommendations For The Authors):

(1) Gap of knowledge:

From the introduction, I got the impression that the manuscript tries to answer the question of whether homeostatic structural plasticity is functionally redundant to synaptic scaling. However, the importance of this question needs to be worked out better. Also, I think it is hard to tackle this question with the shown experiments as one would have to block all other redundant mechanisms and see whether HSP functionally replaces them.

We appreciate the reviewer’s valuable feedback regarding the relationship between homeostatic structural plasticity (HSP) and synaptic scaling. The main objective of our study is indeed to investigate whether structural plasticity is homeostatically regulated, and if so, whether it acts as a redundant or heterogeneous mechanism in relation to synaptic scaling, which is widely recognized as a primary homeostatic process.

In our revised introduction, we have clarified this central question and its significance. Specifically, we explored why experimentally observed changes in spine density, a measure of structural plasticity, do not exhibit the same homeostatic characteristics as changes in spine head size, which reflects synaptic scaling, particularly under conditions of activity blockade.

We hypothesized two key points:

(1) Structural plasticity may not follow a monotonically activity-dependent rule as strictly as synaptic scaling.

(2) The observed changes in spine density may be influenced by the simultaneous modulation of spine size, suggesting that structural plasticity and synaptic scaling interact within the same biological system.

Both hypotheses were tested through a combination of experimental observations and systematic computer simulations. Our conclusions demonstrate that spine-number-based structural plasticity follows a biphasic activity-dependent rule. While it largely overlaps with synaptic scaling under typical conditions, it exhibits heterogeneity under extreme conditions, such as activity silencing. Furthermore, our simulations revealed that both mechanisms can compete and complement each other within neural networks.

We believe that these results offer a nuanced understanding of the interaction between structural plasticity and synaptic scaling, highlighting their redundancy under most conditions but also their heterogeneity under specific circumstances. Blocking all other redundant mechanisms, as suggested, would provide a more reductionist view, which may not capture the complexity and interplay of these processes in a physiological setting. Our approach reflects this complexity, providing insight into how these mechanisms operate together in a naturalistic context.

We have revised the introduction to better convey these points and emphasize the significance of this question for understanding the dynamics of homeostatic regulation in neural networks.

Similarly, the simulations do not really tackle redundancy as, e.g. network growth cannot be achieved by scaling alone.

We appreciate the reviewer’s comment regarding synaptic scaling's limitations in achieving network growth. We would like to clarify that we did not intend to suggest that structural plasticity and synaptic scaling are fully redundant. In fact, it is well established in the literature that structural plasticity plays a dominant role during development, particularly in network growth, which synaptic scaling alone cannot achieve.

The primary objective of our study was to investigate the interaction between structural plasticity and synaptic scaling under conditions of activity perturbation, rather than during network growth or development. To avoid any confusion regarding developmental processes, we chose to grow the network using only structural plasticity in our simulations. Synaptic scaling was then introduced (or not) during the phase of activity deprivation to specifically examine its role in regulating homeostasis under these conditions.

We have revised the corresponding sections of the manuscript to clarify this distinction, and we have ensured that the simulations reflect our focus on activity perturbation rather than network development. This distinction should help readers avoid conflating developmental processes with the specific goals of our study.

Instead, the section on "Integral feedback mechanisms" (L112-129) contains a much better description of the actual goals of the paper than is given in the introduction. Moreover, this section does not seem to include any new results (at least the Ca-dependent structural plasticity and synaptic scaling rules seem to be very common for me). I, therefore, suggest fusing this paragraph in the introduction to obtain a clearer and better understandable gap of knowledge, which is addressed by the paper.

We agree that the "Integral feedback control" section provides key information relevant to both the Introduction and Methodology. It outlines the theoretical framework and serves as a basis for the experimental design.

To better reflect this, we have revised the Introduction to include the gap in knowledge. However, we opted to retain the section in the Results, slightly modified, to set the context for the first experiment.

Along this line, as it seems a central point of the manuscript to distinguish the controller dependencies on Calcium, the different dependencies (working models) should be described in more detail. Also, the description of the inconsistencies of the previous results on HSP can be moved from the discussion (l419-l441) to the introduction.

We have revised the manuscript to place less emphasis on the controller models while retaining the core principles of control theory. The description of the HSP model has been moved to the Introduction, as suggested, while the detailed history remains in the Discussion to maintain the manuscript's consistency.

Systematic text revision: Regarding comment (1), we thank the reviewer for suggesting the text reorganization. We have adjusted several parts in the introduction, M&M section, and results section to increase clarity.

(2) Pharmacological Choice:

It should be discussed why NBQX is used to induce the homeostatic effect instead of TTX. As there are studies showing that it might block homeostatic rewiring (doi.org/10.1073/pnas.0501881102) as well as synaptic scaling (10.1523/JNEUROSCI.3753-08.2009), it seems unclear whether the observed effects are actually corresponding to those in other publications.

The rationale for using NBQX in our experiments, rather than TTX, is detailed in the public response. We selected NBQX based on specific experimental motivations relevant to our study’s objectives, while acknowledging the potential differences in effects compared to other studies.

Local text revision: We added one paragraph in the discussion section to explain the idea better.

(3) Model-Experiment Connection:

The paper combines simulations with experimental work, which is very good. However, in my opinion, the only connection between the two parts is that the experiments suggest a non-monotonic dependency between firing rate and synapse density (i.e. the biphasic dependency). The rest of the experimental results seem to be neglected in the modeling part. It is not even shown that the model reproduces the experiments. Instead, the model is tested in different situations and paradigms (blocking AMPARs in the whole culture vs network growth or silencing a sub-population). I think it would make the paper stronger and more consequential when a reproduction of the experiment by the model is demonstrated (with analogue analyses).

The experimental results serve three main purposes. First, as the reviewer noted, the spine analysis was conducted to inform the biphasic rule. Second, spine size analysis was performed to replicate published findings and confirm our modeling results, showing that activity deprivation leads to fewer synapses with larger sizes or higher weights. Third, the correlation analysis of spine density and size across dendritic segments suggested a hybrid combination of two types of plasticity across different neurons.

While we addressed these aspects in the Results and Discussion sections, the collective presentation in Fig. 2 may have caused some confusion. To improve clarity, we have now split the experimental results, presenting them alongside the relevant modeling data in Fig. 2, Fig. 8, and Fig. 9.

Also, there are a few more mismatches between the experiment and the model that you will want to discuss:

• The size-dependent homeostatic effect (l154ff, Fig2F) is not reflected by the used scaling model.

We revised Fig 8 and the corresponding text to explain how the scaling model reflects such an effect.

• The model assumes reduced Ca levels. Yet, the experimental protocol blocks AMPARs, which are to my knowledge not the primary source of Ca influx, but rather the NMDARs.

The model is based on neural activity, with calcium concentration serving as an internal integral signal of the firing rate, allowing for integral control. While calcium plays a critical role in homeostasis, we caution against drawing a strict correspondence between the model's calcium dynamics and the experimental protocol, as calcium can be sourced from multiple pathways in neurons beyond AMPARs, such as NMDARs, voltage gated calcium channels, and intracellular stores. Also, our recent work demonstrated that under baseline conditions, the majority of AMPARs are not Ca2+ permeable, i.e., GluA2-lacking (Kleidonas et al., 2023)

Improving the calcium dynamics, including secondary calcium release and calcium stores, is part of our future plan to refine the HSP model and address experimental findings that are not fully explained by the current model.

• The model further assumes silencing by input removal, whereas the recurrent connections stay intact. Wouldn't this rather correspond to a deafferentation experiment, where connections to another brain area are cut?

Thank you for pointing at this. The modeling section was not intended to directly replicate the tissue culture experiments but rather to provide insights into a broader range of scenarios, including pharmacological treatments, deafferentation, lesions, and even monocular deprivation.

Systematic text revision: Regarding comment (3), the goal of our modeling work was more than reproducing. To better serve the purposes of experimental results used in the present study, to inform, confirm, and inspire, we have systematically adjusted the layout of experimental and modeling results to link them better.

(4) Is the recurrent component too weak?

Your results show that HSP does not restore activity after silencing (deafferentation), whereas you discuss that earlier models did achieve this by active neighbors in a spatially organized network. However, the silenced neurons in your simulations also receive inputs through the "recurrent" connections from their neighbors (at least shortly after silencing). Therefore, given the recurrent input is strong enough, they should be able to recover in a similar way as the spatially organized ones. As a consequence, I obtained the impression that, in your model networks, activity is strongly driven by external stimulation and less by recurrent connections. I understand that this is important to achieve silencing through removing the Poisson stimulation. Yet, this fact may be responsible for the failure to restore activity such that presented effects are only applicable for networks that are strongly driven by external inputs, but not for strongly recurrent networks, which would severely limit the generality of the results. As a consequence, the paper would benefit from a systematic analysis of the trade-off between recurrent strength and input strength. Maybe, different constant negative currents could be injected in all neurons, such that HSP creates more recurrent synapses in the network.

We appreciate this insight. However, increasing recurrent input strength is beyond the scope of the current study, as it would fundamentally alter the predefined network dynamics of the Brunel network used. As noted in the manuscript, complete isolation or cell death is not always the outcome after input deprivation, lesion, or stroke, which cannot be fully explained by the Gaussian HSP rule alone. Butz and colleagues offered a solution using growth rules that maximized recurrent input, and we recognize the importance of their work.

That said, we approached the issue from a different angle, emphasizing the role of synaptic scaling in recurrence rather than relying solely on recurrent input strength. In biological networks, external inputs may vary, recurrency can be weak or strong, and synaptic scaling can dominate. Our model offers a complementary hypothesis, suggesting that these factors, in combination, contribute to the diverse and sometimes contradictory results found in the literature, rather than posing a strict constraint on network topology.

Local text revision: We emphasized these points in the Discussion section again.

(5) Missing conclusions / experimental predictions

As already described, the modelling work is not reproducing the presented or previous experimental data. Hence, the goal of modelling should be to derive a more general understanding and make experimental predictions. Yet, the conclusions in the discussion stay superficial and vague and there are no specific experimental predictions derived from the model results.

For example, the authors report that the recovery of activity in silenced cultures is observed in a previously spatially structured model but not in theirs -- at least with slow or no scaling. Yet it is left to the reader to think about whether the current model is an improvement to the previous one, how they could be experimentally distinguished, or to which experimental findings they relate or compare, which I would expect at this point. I would advise reworking the discussion and thoroughly working out which new insights the modelling part of the study has generated (not to be confused with the assumptions of the model aka the biphasic plasticity rule) and relating them to experimental pre- and postdiction.

We recognize the reviewer’s concern, which is closely related to comment (4). We have addressed these points by reorganizing the text to better clarify the purpose of our experimental work and its connection to the modeling results.

Specifically, we have reworked the discussion to highlight the new insights gained from the modeling, and how these can inform experimental predictions and interpretations. This includes distinguishing our model from previous ones and providing clearer connections to experimental findings.

Systematic text revision: Most of the comments on combining experiments and modeling results and on developing the story based on our expectations raised here are sincere and may also reflect the expectations and concerns of a broader readership, so we have accordingly adjusted the text in the Results and Discussion sections to make our points clear.

Suggestions for minor changes:

Fig 1I: Please check the graph and make it more self-explaining. For example, mark the "setpoint" activity (in my opinion, both curves should be at baseline there. In that case, however, I do not see the biphasic behavior anymore). Maybe the table and the graph can be aligned along the activity axis? Also: synaptic inhibition should be increased and not decreased, right?

Local text and figure revision: I guess the reviewer meant for Fig. 2I? We have improved the visualization to avoid confusion.

L74-81: I would reverse the order of associative and homeostatic plasticity in this paragraph.

Local text and figure revision: We have fine-tuned the order in the first and second paragraphs to match the readers' expectations.

L74-75: Provide references for such theories.

Local text and figure revision: fixed.

L84-86: Please provide a reference for the claim that negative feedback, redundancy, and heterogeneity contribute to robustness.

Local text and figure revision: fixed.

L 95-97: I think the heterogeneity aspect needs to be worked out a bit better. Do you mean that the described mechanisms contribute to firing rate homeostasis in a different mixture for each neuron (as shown assumed in the last figure)?

Local text and figure revision: The term heterogeneity is used in the manuscript for two major different settings: (1) heterogeneity in terms of control theory and (2) different combinations of HSP and SS rules. We have named the second condition as diversity to avoid confusion.

L 132: The question of linearity has not been posed so far. Also, I think "monotonous" would be a much better term than linear (as a test for linearity would require more than 2 datapoints).

Local text and figure revision: We agreed linear is not a good term. We replaced it with ‘monotonic’ throughout the manuscript.

Fig2 Bii: The data for 50um is clearly not Gaussian.

We did not imply that the 50 µM condition is Gaussian. Instead, we noted that the non-linearity observed in both the 200 nM and 50 µM data suggests a non-monotonic growth rule rather than a linear one. We applied the Gaussian rule because it has been extensively studied in previous simulations, allowing us to benchmark our findings against those results.

Fig2 D, E inset: The point at time 0 does not convey any information and could be left out.

The time zero data is included to demonstrate that the three groups have a similar baseline, ensuring that any observed differences are due to the treatment and not pre-existing biases in the grouping.

L 178: As the Gaussian rule drops below zero above the upper set-point again, it is rather tri-phasic than bi-phasic.

We intended to convey that inhibition results in either spine growth or deletion, reflecting a bi-phasic response rather than a true tri-phasic one.

Fig 6A: You may want to mark the eta variables in the curves.

Local text and figure revision: fixed.

Fig 6E: The curve of the S population extending to the next panel looks a bit messy.

We retained the curve extension to visually convey the impression of excessive network activity.

L272: It needs to be better described/motivated how protocol 1 and 2 are supposed to study the role of recurrent connection as well as what kind of biological situation this may be.

Local text and figure revision: The corresponding text has been adjusted to avoid confusion.

L 272: It is not clear how faster simulation leads to less recurrent connectivity, when the stimulation protocol and the rates stay the same and the algorithm compensates for the timestep properly. Maybe you rather want to say that you silence 10x longer and stimulate 10x longer?

Local text revision: The corresponding text has been adjusted to avoid confusion.

L. 302: "reactivate"?

Local text revision: fixed.

L 322f: I would suggest showing the connectivity matrix for a time-point with restored activity as well.

Local text and figure revision: fixed.

Fig 8A: The use of the morphological reconstructions is a bit misleading as the model uses point neuron.

Local text revision: Now after reorganization, it is in Fig.9. We kept the reconstruction figure for motivational purposes, suggesting how to understand the meaning of the combinations in more biologically realistic scenarios. The corresponding text has been adjusted to avoid confusion.

Fig 8E-F: the y axis should be in the same orientation as in panel D.

Local text and figure revision: Good idea and fixed in the new Fig. 9.

Fig. 8F: The results here look a little bit random. Maybe more runs with the same parameters would smooth out the contours or reveal a phase transition.

Local text and figure revision: Thank you for the suggestion. We conducted an additional ten random trials to average the traces and heatmaps, improving the clarity of the results now presented in Fig. 9.

L411: Note that there are earlier HSP models by Damasch and van Ooyen & van Pelt, that might be worth discussing here.

Local text revision: fixed.

L416 "beyond synaptic scaling" reference needed.

Local text revision: fixed.

L419: The biphasic rule was suggested by Butz already.

Local text revision: We adjusted the text to emphasize our contribution in suggesting/confirming the biphasic rule based on direct experimental observations.

L 419-44: Most of this is actually state-of-the art and may be better placed in the introduction to justify the use of NBQX as a competititve blocker.

Local text revision: We adjusted the text in the introduction and Discussion sections to cover the raised points.

L487: In my opinion, although scaling adapts the weights quickly, the information about deviating firing rate is still stored in the calcium signal such that it will also give rise to structural changes (although they may be small when the rate is low). Thus, I think that fast scaling does not abolish structural changes.

Local text revision: We adjusted the text to account for other factors that could lead to the same or opposite conclusions.

L502f: Sentence unclear. Do you mean Ca is an integrated (low-pass filtered) version of the firing rate?

Yes.

L504: What is the cumulative temporal effect of error in estimating firing rates?

We were referring to the potential instability in numeric simulations if the firing rate is not tracked by an integral signal (calcium concentration) but is instead estimated through average spike counts over time. In our model, calcium serves as a proxy for the firing rate to guide homeostatic structural plasticity. The intake and decay constants are set to minimize the accumulation of errors over time, making long-term error accumulation unlikely. In any case, this is not intended to be a precise measure of the firing rate but rather a smooth guide for homeostatic control.

Local text revision: We rewrote the section so as not to cause extra concerns.

L505: Which two rules are meant here? Ca- and firing rate based or HSP and scaling?

Local text revision: The two rules are the HSP rule and the HSS rule. We have adjusted the text to improve clarity.

L505ff: I did not really understand the control theoretic view here and Supp Fig 5 is not self-explaining enough to help. In my view, scaling is a proportional controller for the calcium level (the setpoint is defined for calcium and not firing rate). Also, all of the HSP rules do neither contain an integral nor a differential of the error and are thus nonlinear but proportional controllers in first approximation. If this part is supposed to stay in the manuscript, the supporting information should contain a more detailed mathematical explanation. Relevant previous work on homeostatic control by synaptic scaling and homeostatic rewiring, e.g. doi: 10.23919/ECC54610.2021.9655157 should be discussed

Local text revision: We have updated the last paragraph to increase clarity. The HSP and HSS rules are proportional and integral for neural activity, as neural firing rate homeostasis is the meaningful goal. However, it is also correct that the integral component is gone if we view calcium concentration as the goal or setpoint. This paper is discussed and cited in a paragraph above this one.

Reviewer #2 (Recommendations For The Authors):

I have some additional suggestions and questions for the authors, which I am presenting following the order of the figures.

Fig 1A: I'm a little bit puzzled by the timescales between Hebbian and homeostatic plasticity; a wealth of data suggests that Hebbian plasticity acts on a faster timescale than homeostatic plasticity, while Aii-Aiii implies the opposite. In lesion-induced degeneration, for instance, which is mentioned later by the authors, spine loss has been suggested to be Hebbian (LTD) while the subsequent recovery is homeostatic. Additionally, it will not be clear to the reader if the same stimulus could induce Hebbian and homeostatic plasticity, or why; the rest of the manuscript seems to imply that any stimulus could and would trigger homeostatic plasticity, which is not the case. Finally, there should be a mention somewhere that Hebbian structural plasticity also exists.

Local text and figure revision: We thank the reviewer for pointing out the time scale issue, which was not explicitly considered here and is now updated.

Fig. 2Bii: There is no significant difference at 200nm NBQX for sEPSC amplitude, contrary to what is stated in the text (line 136). Which one is it?

Local text revision: We thank the reviewer for pointing out the mistake. We have inspected the original statistical file and corrected the text.

Fig. 2F: The description of Fig. 2F in the text confused me for the longest time. I am still unsure why 200nm NBQX is described as leading to a general size increase when it follows the control line so closely, crosses 0 at the same point, and is even below the control line for the largest spine sizes. Similarly, 50um NBQX neatly overlaps with the control condition except for the smallest and largest spines, so the "shrinkage of middle-sized spines" doesn't seem different from the control condition. I also couldn't find any data supporting the statement that 50um NBQX increased only the size of "a small subset of large spines". Maybe the authors could clarify this section? I would also suggest adding statistics between the treatments at each spine size bin to support the claims, as they are central to the rest of the paper.

Importantly, there is no description of the normalization nor the quantification of the difference between days in the methods; I am assuming post-pre for the difference and (post-pre)/pre for the normalization, but this should be much more detailed in the methodology. I was happy to see the baseline raw spine sizes in Supplementary Fig. 1, and would also suggest adding the raw spine sizes after treatment for comparison.

Local text and figure revision: We have adjusted the text and figure to improve clarity.

Fig. 2G/S2A: a scale for the label sizes would be helpful. I would also like to have the same correlation for 50um NBQX treatment and the control condition (at least in the supplementary figures).

Local text and figure revision: We have adjusted the text and figure to improve clarity.

Fig. 2I: I might be missing something, but why is the activity line flat when there are changes in spine density and size?

Local text and figure revision: We have adjusted the text and figure to improve clarity.

Fig. 3C-D: they are referenced in the text as Fig. 1C-D (lines 188-194).

Local text revision: fixed.

Fig. 5: it is interesting that the biphasic model captures both spine loss and recovery, fitting well with lesion-induced degeneration and recovery. Does this mean that the model captures other types of plasticity, or does it suggest to the authors that both steps are homeostatic?

Indeed, the biphasic HSP rule captures two types of activity dependence. The pioneering work by Gallinaro and Rotter (2018) also demonstrated that the HSP rule, even in its monotonic/linear form, exhibits associative properties, which are typically associated with Hebbian plasticity.

Fig. 6A: This figure requires a more detailed legend - what are the various insets? Does the top right graph only have one curve because they are overlapping and the growth rules are the same for axons and dendrites?

Local text revision: fixed.

Fig. 6E: There is usually an overshoot when a stimulus is removed, in this case at the end of the silencing period (as shown in Fig. 1Aiii). Is there a reason why this is not recapitulated here? It shouldn't be as extreme as in the right panel so there should be no degeneration.

We agree that removing the stimulus would typically trigger an opposite homeostatic process. However, in this protocol, we aimed to emphasize the role of recurrency by presenting extreme cases to illustrate potential scenarios for the readers.

Local text revision: We revised this paragraph to walk the readers through the rationale better.

Fig. 6: the authors mention distance-dependent connectivity (line 268), but I couldn't find any data related to that statement. I was particularly curious about that aspect, so I would like to know what this statement is based on, especially as they touch again on the role of morphology in Fig. 8, and distance-dependent connectivity is more prominent in the discussion. On a similar note, would the authors have data from other layers of CA1 that would show similar or other rules? Please note that I am not asking to include these data in the present paper - I am just curious if these data exist (or if the experiments are considered).

Such an extensive dataset is included and thoroughly investigated in another study that has just been published in Lenz et al., 2023. We updated the reference in the revised text.

Fig. 7E top: the scalebar is missing.

Local text revision: fixed.

Fig. 8A: do the colors have meaning? If yes, please state them. Also indicate that the left two neurons are pyramidal cells from CA1 and the right neurons are granule cells from the dentate gyrus.

Local text revision: fixed.

Line 302: "reactive" should be "reactivate".

Local text revision: fixed.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this work, the authors investigate the functional difference between the most commonly expressed form of PTH, and a novel point mutation in PTH identified in a patient with chronic hypocalcemia and hyperphosphatemia. The value of this mutant form of PTH as a potential anabolic agent for bone is investigated alongside PTH(1-84), which is a previously used anabolic therapy. The authors have achieved the aims of the study. Their conclusion, however, that this suggests a "new path of therapeutic PTH analog development" seems unfounded; the benefit of this PTH variant is not clear, but the work is still interesting.

The work does not identify why the patient with this mutation has hypocalcemia and hyperphosphatemia; this was not the goal of the study, but the data are useful for helping to understand that.

Strengths:

The work is novel, as it describes the function of a novel, naturally occurring, variant of PTH in terms of its ability to dimerise, to lead to cAMP activation, to increase serum calcium, and its pharmacological action compared to normal PTH.

Weaknesses:

(1) The use of very young, 8-10 week old, mice as a model of postmenopausal osteoporosis is a major limitation of this study. At 8 weeks, the effect of ovariectomy leads to lack of new trabecular bone formation, rather than trabecular bone loss due to a defect in bone remodelling. Although the findings here provide a comparison between two forms of PTH, it is unlikely to be of direct relevance to the patient population. For example, the authors find an inhibitory effect of PTH on osteoclast surface, which is very unusual. Adding to this concern is that the authors have not described the regions used for histomorphometry, and from their figures (particularly the TRAP stain), it seems that the primary spongiosa (which is a region of growth) has been used for histomorphometry, rather than the secondary spongiosa (which more accurately reflects bone remodelling). Much further detail is needed to justify the use of this very young model, and a section on the limitations of this model is needed. Please provide that section in the revised manuscript.

Thank you for your crucial comment. We obtained 8-week-old female mice and stabilized them in our facility for 2 weeks. Then, we performed OVX using 10-week-old mice and determined the effects of dimeric ^R25CPTH(1-34) on bone after 8 weeks because of 4 weeks for recovery and 4 weeks for PTH or ^R25CPTH(1-34). Therefore, we sacrificed the mice at 18-week-old mice. We revised the method section on page 18, line 436-441 and page 18, line 442-448 as follows.

- ‘Eight-week-old C57BL/6N female mice were purchased from KOATECH (Gyeonggi-do, Republic of Korea), and stabilized mice for 2 weeks. All animal care and experimental procedures were conducted under the guidelines set by the Institutional Animal Care and Use Committees of Kyungpook National University (KNU-2021-0101). The mice were housed in a specific pathogen-free environment, with 4-5 mice per cage, under a 12-h light cycle at 22 ± 2°C. They were provided with standard rodent chow and water ad libitum.’

- ‘An ovariectomized (OVX) mouse model was established using 10-week-old C57BL/6N female mice. Following surgery, mice were divided into the following four groups (n = 6 mice/group) as follows: sham, OVX control group, OVX + PTH (1–34) treated group (40 µg/kg/day), and OVX + dimeric ^R25CPTH treated group (40-80 µg/kg/day). OVX mice were allowed to recover for 4 weeks after surgery. Afterward, PTH (1–34) or ^R25CPTH was injected subcutaneously 5 times a week for 4 weeks. Micro-computed tomography (μ-CT) and histological analyses were performed on 4 groups at 18 weeks of age.’

We also appreciate the reviewer's helpful comment on histology analysis. We agree with the reviewer’s comment that the primary spongiosa does not fully reflect bone remodeling. For histomorphometry analysis in young or male mice, we commonly use the secondary spongiosa, which more accurately reflects bone remodeling. However, in aged or OVX-induced osteoporosis mouse models, we use the primary and secondary spongiosa for histomorphometry analysis because of the barely detectable bone in the secondary spongiosa. In the TRAP staining, we observed an inhibitory effect of PTH on the osteoclast surface/bone surface, which was due to an increased bone surface in the PTH treatment group and less bone in the OVX-vehicle group. Serum CTX1 levels showed no significant difference between the OVX+vehicle and OVX+PTH(1-34) groups. We revised the Materials and Methods (page 21, line 502) and Discussion (page 14, line 330) sections as follows.

- ‘In the histomorphometry analysis for TRAP staining, we used the secondary and primary spongiosa for the trabecular ROI because of the barely detectable in the secondary spongiosa of OVX model.’

- ‘This study has several limitations. First, it is urgently necessary to determine whether dimeric ^R25CPTH is present in human patient serum. Second, TRAP staining showed an inhibitory effect of PTH treatment on the primary spongiosa area. However, the secondary spongiosa, which more accurately reflects bone remodeling (55), was not examined due to the barely detectable bone in this area in OVX-induced osteoporosis mouse models. Third, it is unclear whether similar bone phenotypes exist between human ^R25CPTH patients and dimeric ^R25CPTH-treated mice, particularly regarding low bone strength. Although the dimeric ^R25CPTH-treated group showed higher cortical BMD compared to WT-Sham or PTH groups, there was no difference in bone strength compared to the osteoporotic mouse model. Fourth, our study showed that PTH or ^R25CPTH treatment decreased circumferential length; it is uncertain if this phenotype is also present in PTH-treated or ^R25CPTH patients. Finally, we did not analyze the ^R25CPTH mutant mouse model, which would allow us to compare phenotypes that most closely resemble those of human patients.’

(2) It is also somewhat concerning that the age range is from 8-10 weeks, increasing the variability within the model. Did the age of mice differ between the groups analysed?

We utilized mice of the same age (10 weeks) across all experiments involving the surgically induced ovariectomy (OVX) model described as above.

(3) Methods are not sufficiently detailed. For example, the regions used for histomorphometry are not described, there is no information on micro-CT thresholds, no detail on the force used for mechanical testing. Please address this request.

Thank you for your comment. Let me address your points step by step.

(1) Thresholds for analysis were determined manually based on grayscale values for each experimental group as follows: trabecular bone: 3000; cortical bone: 5000 for all samples. We utilized an HA (calcium hydroxyapatite) phantom with HA content ranging from 0 to 1200 mg CaHA/cm³ to measure the grayscale values via µ-CT. These measurements were then used to generate a standard curve.

Author response image 1.

(2) Bone parameters and density were analyzed in the region between 0.3–1.755 mm (Voxel size: 9.7um, 150 slices) from the bottom of the growth plate. Analysis of bone structure was performed using adaptive thresholding in a CT Analyser.

Author response image 2.

(3) Three‐point bending test, the left femur of the mouse was immersed in 0.9 % NaCl solution, wrapped in gauze, and stored at −20°C until ready for a three-point bending test. In this test, we placed the mouse femurs positioned horizontally with the anterior surface facing upwards, centered on the supports, and the compressive force was applied vertically to the mid-shaft. The pressure sensor was positioned at a distance that allowed for the maximum allowable pressure (200N) without interfering with the test (20.0 mm for the femur). A miniature material testing machine (Instron, MA, USA) was used for this test. The crosshead speed was decreased to 1 mm/min until failure. During the test, force-displacement data were collected to determine the maximum load and slope of the bones.

(4)  As the reviewer’s suggestion, we revised the methods on page 20, line 477 and line 482-486 as follows.

- ‘Bone parameters and density were analyzed in the region between 0.3–1.755 mm (150 slices) from the bottom of the growth plate. Analysis of bone structure was performed using adaptive thresholding in a µ-CT Analyser. Thresholds for analysis were determined manually based on grayscale values for each experimental group: trabecular bone: 3000; cortical bone: 5000 for all samples.’

-  ‘The left femur of the mouse was immersed in 0.9 % NaCl solution, wrapped in gauze, and stored at −20°C until ready for a three-point bending test. In this test, we placed the mouse femurs horizontally with the anterior surface facing upwards, centered on the supports, and the compressive force was applied vertically to the mid-shaft. The pressure sensor was positioned at a distance that allowed maximum allowable pressure (1000N) without interfering with the test (20.0 mm for the femur). A miniature material testing machine (Instron, MA, U.S.A.) was used for this test. The crosshead speed was decreased to 1 mm/min until failure. During the test, force-displacement data were collected to determine the maximum load and slope of the bones.’

(4) There are three things unclear about the calvarial injection mouse model. Firstly, were the mice injected over the calvariae or with a standard subcutaneous injection (e.g. at the back of the neck)? If they were injected over the calvaria, why were both surfaces measured? Secondly, why was the dose of the R25C-PTH double that of PTH(1-34)? Thirdly, there is no justification for the use of "more intense coloration" as a marker of new bone; this requires calcein labelling to prove it new bone. It would be more reliable to measure and report the thickness of the calvaria. Please address these technical questions.

Thank you for your valuable feedback on the calvarial injection mouse model. Below are our responses to the specific points mentioned:

(1) Injection method and measurement sites: The injections were administered subcutaneously above the calvaria, rather than at the standard subcutaneous site such as the back of the neck. This approach was chosen to ensure direct delivery of the peptide to the target area, enhancing the localized effects on bone formation. Measurements were taken at two different parts of the calvaria to account for any variation in the spread and absorption of the administered substance following injection. By analyzing both surfaces, we aimed to provide a comprehensive assessment of the impact on calvarial bone thickness.

(2) Dose of ^R25CPTH compared to PTH(1-34): The dose of ^R25CPTH used in our study was determined based on molecular weight calculations. The molecular weight of the dimeric ^R25CPTH(1-34) is approximately twice that of the monomeric PTH(1-34). Therefore, to maintain a consistent molar concentration and ensure comparable biological effects, the dose of ^R25CPTH was adjusted accordingly.

(3) Use of "more intense coloration" as a marker of new bone: We acknowledge that calcein labeling would provide a more reliable and quantifiable way to identify new bone formation. The use of “more intense coloration” was intended as a qualitative indicator in this study, and we recognize the technical limitations of this approach.

(5) The presentation of mechanical testing data is not sufficient. Example curves should be shown, and data corrected for bone size needs to be shown. The difference in mechanical behaviour is interesting, but does it stem from a difference in the amount of bone, or two a difference in the quality of the bone? Please explain this matter better in the manuscript.

Thank you for your comment.

As a reviewer's comment, we provided example curves for the rat femur three-point bending test as shown below.

Author response image 3.

(1) The cortical bone area was decreased in the OVX-Vehicle and OVX-^R25CPTH(1-34) groups but not in the OVX-PTH(1-34) group compared to the Sham group. However, the total bone area was decreased in the PTH(1-34) and ^R25CPTH(1-34) treated groups, with no significant difference in the OVX-Vehicle group compared to the Sham group. Collectively, there was an increase in cortical thickness which resulted in a narrowing of the bone marrow space in OVX-^R25CPTH(1-34) groups. Accordingly, we revised Fig 5B with the addition of Tt.Ar and Ct.Ar.

(2) As the reviewer’s suggestion, we revised the results on page 10, line 220-228 s follows.

- ‘Quantitative micro-computed tomography (μ-CT) analysis of the femurs obtained from each group revealed that, as compared to OVX + vehicle controls, treatment with PTH(1–34) increased femoral trabecular bone volume fraction (Tb.BV/TV) by 121%, cortical bone volume fraction (Ct.BV/TV) by 128%, cortical thickness (Ct.Th) by 115%, cortical area (Ct.Ar) by 110%, and cortical area fraction (Ct.Ar/Tt.Ar) by 118% while decreased total tissue area (Tt.Ar) by 93% (Figure 5A and 5B). Treatment with dimeric ^R25CPTH(1-34) had similar effects on the femoral cortical bone parameters, as it increased Ct.BMD by 104%, Ct.BV/TV by 125%, Ct.Th by 107%, and Ct.Ar/Tt.Ar by 116%, while decreased Tt.Ar 86% (Figure 5). Considering the reduction of Tt.Ar and no change of Ct.Ar compared to the OVX+vehicle controls, the increase of Ct.Ar/Tt.Ar indicates a decrease in bone marrow space. The increase in cortical bone BMD was significant with dimeric ^R25CPTH(1-34) but not with PTH(1-34), whereas an increase in femoral trabecular bone was only observed with PTH(1-34).’

(6) The micro-CT analysis of the cortical bone in the OVX model is insufficient. Please indicate whether cross-sectional area has increased. Is there an increase in the size of the bones, or is the increase in cortical thickness due to a narrowing of the marrow space? This may help resolve the apparent contradiction between the cortical thickness data (where there is no difference between the two PTH formulations) and the mechanical testing data (where there is a difference). Please explain this matter better in the manuscript.

Thank you for your comment.

(1) The cortical bone area was decreased in the OVX-Vehicle and OVX-^R25CPTH(1-34) groups but not in the OVX-PTH(1-34) group compared to the Sham group. However, the total bone area was decreased in the PTH(1-34) and ^R25CPTH(1-34) treated groups, with no significant difference in the OVX-vehicle group compared to the Sham group. Taken together, there was an increase in cortical thickness due to a narrowing of the bone marrow space in OVX-^R25CPTH(1-34) groups. Therefore, we revised as above.

(2) As the reviewer’s suggestion, we revised the results on page 10, line 220-228 as follows.

- ‘Quantitative micro-computed tomography (μ-CT) analysis of the femurs obtained from each group revealed that, as compared to OVX + vehicle controls, treatment with PTH(1–34) increased femoral trabecular bone volume fraction (Tb.BV/TV) by 121%, cortical bone volume fraction (Ct.BV/TV) by 128%, cortical thickness (Ct.Th) by 115%, cortical area (Ct.Ar) by 110%, and cortical area fraction (Ct.Ar/Tt.Ar) by 118% while decreased total tissue area (Tt.Ar) by 93% (Figure 5A and 5B). Treatment with dimeric ^R25CPTH(1-34) had similar effects on the femoral cortical bone parameters, as it increased Ct.BMD by 104%, Ct.BV/TV by 125%, Ct.Th by 107%, and Ct.Ar/Tt.Ar by 116%, while decreased Tt.Ar 86% (Figure 5B). Considering the reduction of Tt.Ar and no change of Ct.Ar compared to the OVX+vehicle controls, the increase of Ct.Ar/Tt.Ar indicates a decrease in bone marrow space. The increase in cortical bone BMD was significant with dimeric ^R25CPTH(1-34) but not with PTH(1-34), whereas an increase in femoral trabecular bone was only observed with PTH(1-34).’

(7) The evidence that dimeric PTH has a different effect to monomeric PTH is very slim; I am not sure this is a real effect. Such differences take a long time to sort out (e.g. the field is still trying to determine whether teriparatide and abaloparatide are different). I think the authors need to look more carefully at their data - almost all effects are the same. Ultimately, the statement that dimeric PTH may be a more effective anabolic therapy than monomeric PTH are not supported by the data, and this should be removed. There is little to no difference found between normal PTH and the variant in their effects on calcium and phosphate homeostasis or on bone mass. However, the analysis has been somewhat cursory, with insufficient mechanical testing or cortical data presented. Many of the effects seem to be the same (e.g. cortical thickness, P1NP, ALP, vertebral BV/TV and MAR), but the way it is written it sounds like there is a difference. Please remove some of the unfounded claims that you have made in this manuscript.

Thank you for your insightful comments. We strongly agree with your conclusion that PTH and dimeric ^R25CPTH indeed exhibit similar activities. We have toned-down our statement, however, there are still some elements showing statistical significance that need to be clearly stated. Specifically, when we changed the statistical method from t-test to one-way ANOVA, the significance of bone formation markers were only observed in dimeric PTH treated samples, and we have revised the manuscript of Results section on page 9, line 206-212 as follows to reflect the change.

- ‘These analyses revealed that both PTH(1-34) and dimeric ^R25CPTH(1-34) significantly increased the width of the new bone area by approximately four-fold, as compared to the vehicle group (Figure 4B). These findings thus support a capacity of dimeric ^R25CPTH(1-34) to induce new bone formation in vivo, similar to PTH, despite molecular and structural changes.’

Although it is unclear whether ^R25CPTH circulate as dimeric form or mutant monomeric form, the absence of bone resorption associated with long-term PTH exposure in the patients suggests the potential for a bone anabolic drug without side effects. Also, continued observation of the recently reported young patient in Denmark is expected to clarify this effect further. However, we acknowledge that our current data alone are insufficient to claim that ^R25CPTH may be a more effective anabolic therapy than wild type PTH, and we have adjusted our tone accordingly.

(8) Statistical analysis used multiple t-tests. ANOVA would be more appropriate.

We agree with your suggestion. To compare the means among three or more groups, ANOVA is more appropriate than the t-test. Accordingly, we performed new statistical analyses using one-way and two-way ANOVA. One-way ANOVA was applied to figure 4, 5, and 6 (In previous, figure 5, 6, and 7), and two-way ANOVA was applied to Figure 3, considering both time and treatment variables. We revised some of the figures and descriptions to reflect the changes in significance.

Thank you for Reviewer #1’s thorough and thoughtful review. We greatly appreciate the suggestions and will incorporate them to enhance the quality of our paper.

Reviewer #2 (Public Review):

Summary:

The study conducted by Noh et al. investigated the effects of parathyroid hormone (PTH) and a dimeric PTH peptide on bone formation and serum biochemistry in ovariectomized mice as a model for postmenopausal osteoporosis. The authors claimed that the dimeric PTH peptide has pharmacological benefits over PTH in promoting bone formation, despite both molecules having similar effects on bone formation and serum Ca2+. However, after careful evaluation, I am not convinced that this manuscript adds a significant contribution to the literature on bone and mineral research.

Strengths:

Experiments are well performed, but strengths are limited to the methodology used to evaluate bone formation and serum biochemical analysis.

Weaknesses:

(1) Limited significance of this study:

• This study follows a previous study (not cited) reporting the effect of the dimeric R25CPTH(1-34) on bone regeneration in an osteoporotic dog (Beagle) model (Jeong-Oh Shin et al., eLife 13:RP93830, 2024). It's unclear why the authors tested the dimeric R25C-PTH peptide on a rodent animal model, which has limitations because the healing mechanism of human bone is more similar in dogs than in mice.

Thank you for your interest in our research. To address the paper by Shin et al. (2024, DOI:10.7554/eLife.93830.1), we would like to clarify that our research on dimeric ^R25CPTH(1-34) was conducted first. Initially, we confirmed dimerization under in vitro conditions and observed its effects in a mouse model. Recognizing the need for additional animal models, we collaborated with Shin et al.'s team. Due to delays during the submission process, our paper was submitted later, which seems to have led to this misunderstanding. However, Shin et al. (2024) cited our pre-print article on bioRxiv (Noh, M., Che, X., Jin, X., Lee, D. K., Kim, H. J., Park, D. R., ... & Lee, S. (2024). Dimeric R25CPTH (1-34) Activates the Parathyroid Hormone-1 Receptor in vitro and Stimulates Bone Formation in Osteoporotic Female Mice. bioRxiv, 2024-03.DOI: 10.1101/2024.03.13.584815). Both Shin et al., and our mouse work supports the action of dimeric R25CPTH(1-34) on regulating bone metabolism.

• The authors should clarify why they tested the effects of dimeric ^R25CPTH(1-34) and not dimeric ^R25CPTH(1-84)?

Thank you for your valid comments. Here are several reasons why we used the 1-34 fragment peptide in our experiment. Currently, PTH analog peptides for medical purposes include human parathyroid hormone fragment 1-34 (PTH(1-34)) and full-length recombinant human parathyroid hormone (rhPTH(1-84)). PTH(1-34) is used as a bone anabolic agent, while rhPTH(1-84) is used for PTH replacement therapy in hypoparathyroid patients with hypocalcemia. We aimed to compare the bone formation effects of R25CPTH with wild-type PTH, for which PTH(1-34) was deemed more appropriate. Additionally, previous studies have shown that both PTH(1-34) and PTH(1-84) possess equal ligand binding affinity for the PTH1 receptor. Key sites within the first 34 N-terminal amino acids of PTH are critical for high-affinity interactions and receptor activation. Alterations in the N-terminal sequence of PTH(1-84) significantly reduce receptor binding, while truncations at the C-terminal end do not affect receptor affinity. The peptide used in our experiment was synthetic, and if the length does not affect affinity to its receptor affinity, the shorter length of PTH(1-34) made its synthesis more reasonable. Consequently, we tested the effects of PTH(1-34) and dimeric R25CPTH(1-34) due to its known efficacy on bone anabolic effect and relevance in receptor interactions. However, we aim to conduct functional analysis of the dimeric R25CPTH(1-84) in further study.

• The study is descriptive with no mechanism.

We recognize that your concern is legitimate. While our study includes descriptive elements, it extends beyond mere observation. The R25CPTH research, which began with a case report, has evolved to utilize molecular techniques to better understand the unique physiological phenomena observed in patients. We have validated the peptide’s dimerization caused by mutations in vitro and assessed their effects in both in vitro cell line models and in vivo mouse models. Although we have not yet confirmed whether ^R25CPTH exists as a dimer or monomer in patient blood, we anticipate it may exist in dimeric form at least some fractions and are currently conducting mass spectrometry on patient blood samples to determine this. Therefore, this paper serves as the first report on this PTH mutant suggesting that it may form a homodimer. Importantly, we are actively investigating the molecular mechanisms and downstream signaling pathways that differentiate normal PTH from dimeric ^R25CPTH. This includes analyzing differences in proteome and transcriptome induced by PTH and dimeric ^R25CPTH and examining the direct molecular characteristics and structural changes responsible for these mutations. Through this comprehensive approach, we aim to provide a detailed mechanistic understanding of ^R25CPTH in the subsequent publication.

(2) Statistics are inadequately described or performed for the experimental design:

• The statistical analysis in Figure 5 needs to be written in a way that makes it clearer how statistics were done; t-test or one-way ANOVA?

Sorry for the inconvenience and thank you for your thorough review. Initially, we conducted the statistical analysis using a t-test. However, during the revision process, we performed a new statistical analysis using one-way ANOVA, as it is more appropriate for comparing the means among three or more groups. Despite this change, there were no differences in statistical significance, so the descriptions remained unchanged.

• Statistics in Figures 6 and 7 should be performed by one-way ANOVA to compare the mean values of one variable among three or more groups, and not t-test.

Thank you for your thorough review, and I apologize for any inconvenience. I agree with your suggestion that ANOVA is more appropriate than the t-test for comparing means among three or more groups. Accordingly, we performed new statistical analyses using one-way ANOVA. When we changed the statistical method from t-test to one-way ANOVA, the significance of bone formation markers, P1NP and ALP, appeared only in dimeric R25CPTH and not in wild-type PTH. We have reflected these findings in the text.

(3) Misleading and confused discussion:

• The first paragraph lacks clarity in the PTH nomenclature and the authors should provide a clear statement that the PTH mutant found in patients is likely a monomeric R25CPTH(1-84), considering that there has been no proof of a dimeric form.

Thank you for your insightful comments. I agree that there was some ambiguity in the nomenclature used in the first paragraph of the Discussion section. However, we do not believe that no proof of a dimeric form of the ^R25CPTH(1-84) mutant necessarily indicates that the PTH mutant in the blood is solely monomeric. Identifying the in vivo structure of ^R25CPTH(1-84) is one of the goals of our ongoing project. While the exact form of ^R25CPTH(1-84) in patients is still elusive, we are investigating the possibility that some fraction may exist as a dimer. On page 12, line 274-276, we have revised the content to address this issue and improve clarity as follows.

- ‘In this study, we show the introduction of a cysteine mutation at the 25th amino acid position of mature parathyroid hormone (^R25CPTH) facilitates the formation of homodimers comprised of the resulting dimeric R25CPTH peptide in vitro.’

• Moreover, the authors should discuss the study by White et al. (PNAS 2019), which shows that there are defective PTH1R signaling responses to monomeric R25CPTH(1-34). This results in faster ligand dissociation, rapid receptor recycling, a short cAMP time course, and a loss of calcium ion allosteric effect.

Sorry for the inconvenience and thank you for your thorough review. The authors were aware of the referenced paper and deeply apologize for its omission during the writing and editing process. Citing this paper will enhance the credibility of our findings. We have now included this citation and made the necessary adjustments to the manuscript of Discussion section on page 12, line 295-296 as follows.

- ‘We also observed that the potency of cAMP production in cells was lower for dimeric ^R25CPTH as compared to the monomeric ^R25CPTH, in accordance with a lower PTH1R-binding affinity. Previous reports indicated that a mutation at the 25th position of PTH results in the loss of calcium ion allosteric effects on monomeric ^R25CPTH, leading to faster ligand dissociation, rapid receptor recycling, and a shorter cAMP time course (50). Correspondingly, the weaker receptor affinity and reduced cAMP production observed in dimeric ^R25CPTH suggest a possibility that the formation of a disulfide bond at the 25th position significantly alters the function of PTH as a PTH1R ligand. These structural effects are not yet fully understood and need to be investigated further.’

• The authors should also clarify what they mean by "the dimeric form of R25CPTH can serve as a new peptide ...(lines 328-329)" The dimeric R25CPTH(1-34) induces similar bone anabolic effects and calcemic responses to PTH(1-34), so it is unclear what the new benefit of the dimeric PTH is.

We apologize for any confusion in our previous description. We concur that, as you mentioned, PTH and dimeric ^R25CPTH indeed exhibit similar activities. We have toned-down our statement, however, there are still some elements showing statistical significance that need to be clearly stated. Specifically, when we changed the statistical method from t-test to one-way ANOVA, the significance of bone formation markers was only observed in dimeric PTH treated samples, and we have revised the manuscript of Results section on page 9, line 206-212 as follows to reflect the change.

- ‘These analyses revealed that both PTH(1-34) and dimeric ^R25CPTH(1-34) significantly increased the width of the new bone area by approximately four-fold, as compared to the vehicle group (Figure 4B). These findings thus support a capacity of dimeric ^R25CPTH(1-34) to induce new bone formation in vivo, similar to PTH, despite molecular and structural changes.’

Although it is unclear whether ^R25CPTH circulate as dimeric form or mutant monomeric form, the absence of bone resorption associated with long-term PTH exposure in the patients suggests the potential for a bone anabolic drug without side effects. Also, continued observation of the recently reported young patient in Denmark is expected to clarify this effect further. However, we acknowledge that our current data alone are insufficient to claim that ^R25CPTH may be a more effective anabolic therapy than wild type PTH, and we have adjusted our tone accordingly.

Thank you for Reviewer #2’s comprehensive and considerate review. We are grateful for the ideas, and we have revised our manuscript accordingly them to improve our paper.

Reviewer #1 (Recommendations For The Authors):

(1) Figure 1D lacks molecular weight markers.

Thank you for your thorough review. We added protein molecular weight markers in the figure.

(2) The lack of change in plasma cAMP is very surprising, particularly given that there is no difference in the effect of the two forms of PTH on serum calcium or phosphate, or urinary phosphate. This data is somewhat of a distraction since no effort has been made to assess the difference in the effects of these PTH forms on kidney function. I suggest removing this data and spending time working on the origin of this difference.

Thank you for your insightful comments and valuable suggestions on our manuscript. We also could not precisely explain the discrepancy between the cell line and animal model experiments. However, since the results were consistently observed, we included them in the paper as they may be significant. We acknowledge that in the context of our current research, these data lack sufficient correlation with other findings. Therefore, we have removed the data about the lack of change in plasma cAMP by PTH injection (Figure 4. Effect of cAMP production by PTH injection in CD1 female mice) and revised the manuscript accordingly (Page 8, line 188-194; page 12, line 301-306; page 19, line 454-456). We are currently conducting further research with multiomics data analysis to elucidate potential differences in the sub-signaling pathways between PTH and dimeric R25CPTH, to identify the specific functions affected by these variations, and to understand the underlying mechanisms. The lack of changes in plasma cAMP levels in vivo will be addressed in a subsequent publication detailing our findings.

(3) Introduction, line 61. The authors state that "most" anti-resorptive therapies cannot stimulate new bone formation. I don't believe that ANY anti-resorptive therapies stimulate new bone formation! If there is one, this should be referenced.

Thank you for pointing out important aspects. Romosozumab, a humanized monoclonal anti-sclerostin antibody, has a dual effect by enhancing bone formation and inhibiting bone resorption. Sclerostin, a protein produced by osteocytes, plays a role in the regulation of bone metabolism. It promotes osteoclast differentiation, which is associated with bone resorption, and suppresses osteoblast activity, which is crucial for bone formation. By binding to sclerostin, Romosozumab prevents it from blocking the signaling pathways necessary for osteogenesis. Consequently, Romosozumab therapy not only regulates bone resorption but also affects new bone formation. We added the references to that information.

(4) The authors tend to include a lot of methods in the results section (e.g. describing the number of replicates, and details of histological analysis). This should be minimized.

Thank you for your thorough review, and sorry for the inconvenience. We have minimized the methodological details in the results section, ensuring that only essential information for understanding the findings and the procedures remain.

(5) Lines 302-305: If retaining the blood cAMP data, please provide references for the assertion that renal PTH receptors mediate this response.

PTH exerts its effects primarily through the PTH1 receptor (PTH1R), a G protein-coupled receptor present in various tissues, including bone and kidney (Chase et al., 1968, Chase et al., 1970). When activated by PTH, this receptor stimulates the production of cyclic AMP (cAMP), with the kidneys playing a significant role in this process (Maeda et al., 2013). In the initial manuscript, the importance of renal PTH receptors in mediating the blood cAMP response may have been overemphasized. We appreciate your feedback on this point, and we have provided references to support this assertion. However, by process following the former ‘Recommendations for the Authors’, we removed the data about the lack of change in plasma cAMP by PTH injection, the description of the renal PTH receptors mediate this response of blood cAMP also removed.

- Chase, Lewis R., and G. D. Aurbach. "Renal adenyl cyclase: anatomically separate sites for parathyroid hormone and vasopressin." Science 159.3814 (1968): 545-547.DOI:10.1126/science.159.3814.545

- Chase, Lewis R., and G. D. Aurbach. "The effect of parathyroid hormone on the concentration of adenosine 3', 5'-monophosphate in skeletal tissue in vitro." Journal of Biological Chemistry 245.7 (1970): 1520-1526.DOI:10.1016/S0021-9258(19)77126-9

- Maeda, Akira, et al. "Critical role of parathyroid hormone (PTH) receptor-1 phosphorylation in regulating acute responses to PTH." Proceedings of the National Academy of Sciences 110.15 (2013): 5864-5869.DOI: 10.1073/pnas.1301674110

(6) Eosin stains bone pink and haematoxylin stains cells purple. This has been incorrectly described in the manuscript.

Thank you for your thorough review, and I apologize for any confusion caused by the poor description. It appears that the terms were used interchangeably during the editing process. We have corrected the description in the manuscript and will ensure such mistakes do not occur again in the future.

(7) Sodium thiosulphate is a fixative for Von Kossa staining, not an agent that removes nonspecific binding.

Thank you for your careful review. However, there seems to be a misunderstanding of sodium formaldehyde as sodium thiosulfate. A 5% sodium thiosulfate solution is a critical in vitro diagnostic agent used in various staining kits. As a reducing agent, it effectively removes excess silver ions in staining kits based on silver impregnation techniques. In our experiment, sodium thiosulfate was specifically used to remove residual silver ions in Von Kossa staining. For more details, please refer to the following link: https://www.morphisto.de/en/shop/detail/d/Natriumthiosulfat_5//12825/.

Reviewer #2 (Recommendations For The Authors):

Moderate-to-Minor points:

• Line 73: it's either class B GPCR or secretin receptor family but not class B GPCR family.

Thank you for your thorough review, and I apologize for any confusion in our previous description. We corrected the description in the manuscript as class B GPCR.

• Line 79: correct "adenylate cyclase" to "transmembrane adenylate cyclases"

Thank you for your thorough review, and I apologize for any confusion in our previous description. We corrected the description in the manuscript as transmembrane adenylate cyclases.

• Line 89: should "hypothyroidism" be "hypoparathyroidism"?

Thank you for your thorough review, and I apologize for any confusion in our previous description. We corrected the description in the manuscript as hypoparathyroidism.

• Line 159: all agonists display higher binding affinities when their receptors are coupled to G proteins, so it's unclear why the higher affinity of the dimeric ^R25CPTH(1-34) for the RG state seems to be important for the authors.

Thank you for your insightful comments. First of all, comparing the binding affinities of the R0 (G protein-uncoupled) and RG (G protein-coupled) conformations of the receptor is inappropriate. This is because the form and size of the radio-label ligand bound to each conformation differ, which consequently affects their binding affinities and, in turn, influences the binding strength of target ligands such as PTH, monomeric ^R25CPTH, and dimeric ^R25CPTH. Therefore, it is preferable to compare how the binding strengths of test ligands differ for each conformation. Additionally, the fact that significant binding affinity is lost for R⁰ while remaining high for the RG conformation of PTH1R is important because typical PTH exhibits high binding affinity for R0, whereas PTHrP shows higher affinity for the RG conformation. This suggests that dimeric ^R25CPTH may possess distinct molecular characteristics and potentially induce different downstream signaling pathways compared to typical PTH.

• Line 169-170 and Fig. 2: According to the theory of receptor pharmacology established in the 60s' for native receptors (Arch. Int. Pharmacodyn. 127:459-478 (1960); Arch. Int. Pharmacodyn. 136:385-413 (1962)) and verified later in the 80-90's for recombinant GPCRs, the activity constant (Kact or EC50) value of hormone actions in various tissues or cells is equal to the dissociation constant (Kd) of the hormone when receptors are not overexpressed (EC50 = Kd). When receptors are overexpressed (presence of spare receptors), then EC50 < Kd. Assuming that after Cheng-Prussof correction for data in Fig. 2, IC50 < Ki = Kd, how do the authors explain that IC50 values for RG are about 1-Log lower than EC50s (i.e., EC50 > Kd)?

We appreciate your insightful comment and fully acknowledge the established theory of receptor pharmacology, which states that Kd equals EC50, and when the receptor is overexpressed, EC50 is less than Kd. After having read your comments, we have revisited this paper Okazaki et al, PNAS, 2008 to better understand the PTH interaction with PTH1R. While our data might appear to contradict this theory, we believe that a direct comparison between the IC50 of RG and the EC50 in Figure 2 may not be entirely appropriate for the following reasons. First, the IC50 was determined from membrane preparations of a receptor-overexpressing cell line (GP-2.3), whereas the EC50 was calculated based on the cAMP response in SaOS-2 cells. These different experimental conditions contribute to the observed discrepancies. Second, the peptides used in the competition assays differ. R⁰ utilized radiolabeled PTH(1-34), while RG employed M-PTH(1-15) with several amino acid substitutions and a shorter length. This further complicates a direct comparison between the EC50 and IC50 values in our study.

Thank you for all the reviewers’ thorough and thoughtful reviews. We greatly appreciate your suggestions and have addressed all the issues to enhance the quality of our paper.

eLife Assessment

The study is noteworthy for its effort to achieve a deeper understanding of PTH-1 Receptor signaling. The PTH-1 pathway underpins the control of calcium and phosphate metabolism throughout life in land-dwelling animals and it can be targeted to therapeutic benefit in patients with osteoporosis. We consider the significance of the findings in this paper to be valuable to the community of investigators working on PTH receptor signaling, with convincing evidence.

Reviewer #1 (Public review):

Summary:

In this work, the authors investigate the functional difference between the most commonly expressed form of PTH, and a novel point mutation in PTH identified in a patient with chronic hypocalcemia and hyperphosphatemia. The value of this mutant form of PTH as a potential anabolic agent for bone is investigated alongside PTH(1-84), which is a current anabolic therapy. The authors have achieved the aims of the study. Their conclusion that this suggests a "new path of therapeutic PTH analog development" seems unfounded; the benefit of this PTH variant is not clear, but the work is still interesting.

The work does not identify why the patient with this mutation has hypocalcemia and hyperphosphatemia; this was not the goal of the study, but the data is useful for helping to understand it.

Strengths:

The work is novel, as it describes the function of a novel, naturally occurring, variant of PTH in terms of its ability to dimerise, to lead to cAMP activation, to increase serum calcium, and its pharmacological action compared to normal PTH.

Weaknesses:

(1) The use of very young, 10 week old, mice as a model of postmenopausal osteoporosis remains a limitation of this study, but this is now quite clearly described as a limitation,, including justifying the use of the primary spongiosa as a measurement site.

(2) Methods have been clarified. It is still necessary to properly define the micro-CT threshold in mm HA/cc^3. I think it might be at about 200mg HA/cc^3 in this study.

(3) The apparent contradiction between the cortical thickness data (where there is no difference between the two PTH formulations) and the mechanical testing data (where there is a difference) remains unresolved. It is still not clear whether there is a material defect in the bone, which can be partially assessed by reporting the 3 point bending test, corrected for the diameters of the bone (i.e. as stress / strain curves).

(4) It is also puzzling that both dimeric and monomeric PTH lead to a reduction in total bone area (cross sectional area?). This would suggest a reduction in bone growth. This should be discussed in the work.

eLife Assessment

Aging reduces tissue regeneration capacity, posing challenges for an aging population. In this fundamental study, Reeves et al. show that by combining Wnt-mediated osteoprogenitor expansion (using a special bandage) with intermittent fasting, calvarial bone healing can be restored in aged animals. Intermitted fasting improves osteoprogenitor function by rescuing aging-related mitochondrial dysfunction, which can also be achieved by nicotinamide mononucleotide (NMN) supplementation or by modulating the gut microbiome. By employing rigorous histological, transcriptomic, and imaging analyses in a clinically relevant model, the authors provide compelling evidence supporting the conclusions. The therapeutic approach presented in this study shows promise for rejuvenating tissue repair, not only in bones but potentially across other tissues.

Reviewer #1 (Public review):

Aging reduces tissue regeneration capacity, posing challenges for an aging population. In this study, the authors investigate impaired bone healing in aging, focusing on calvarial bones, and introduce a two-part rejuvenation strategy. Aging depletes osteoprogenitor cells and reduces their function, which hinders bone repair. Simply increasing the number of these cells does not restore their regenerative capacity in aged mice, highlighting intrinsic cellular deficits. The authors' strategy combines Wnt-mediated osteoprogenitor expansion with intermittent fasting, which remarkably restores bone healing. Intermittent fasting enhances osteoprogenitor function by targeting NAD+ pathways and gut microbiota, addressing mitochondrial dysfunction - an essential factor in aging. This approach shows promise for rejuvenating tissue repair, not only in bones but potentially across other tissues.

This study is exciting, impressive, and novel. The data presented is robust and supports the findings well.

Reviewer #2 (Public review):

Reeves et al explore a model of bone healing in the context of aging. They show that intermittent fasting can improve bone healing, even in aged animals. Their study combines a 'bone bandage' which delivers a canonical Wnt signal with intermittent fasting and shows impacts on the CD90 progenitor cell population and the healing of a critical-sized defect in the calvarium. They also explore potential regulators of this process and identify mitochondrial dysfunction in the age-related decline of stem cells. In this context, by modulating NAD+ pathways or the gut microbiota, they can also enhance healing, hinting at an effect mediated by complex impacts on multiple pathways associated with cellular metabolism.

The study shows a remarkable finding: that age-related decreases in bone healing can be restored by intermittent fasting. There is ample evidence that intermittent fasting can delay aging, but here the authors provide evidence that in an already-aged animal, intermittent fasting can restore healing to levels seen in younger animals. This is an important finding as it may hint at the potential benefits of intermittent fasting in tissue repair.

Reviewer #3 (Public review):

Summary:

This study aims to address the significant challenge of age-related decline in bone healing by developing a dual therapeutic strategy that rejuvenates osteogenic function in aged calvarial bone tissue. Specifically, the authors investigate the efficacy of combining local Wnt3a-mediated osteoprogenitor stimulation with systemic intermittent fasting (IF) to restore bone repair capacity in aged mice. The highlights are:

(1) Novel Approach with Aged Models:<br /> This pioneering study is among the first to demonstrate the rejuvenation of osteoblasts in significantly aged animals through intermitted fasting, showcasing a new avenue for regenerative therapies.

(2) Rejuvenation Potential in Aged Tissues:<br /> The findings reveal that even aged tissues retain the capacity for rejuvenation, highlighting the potential for targeted interventions to restore youthful cellular function.

(3) Enhanced Vascular Health:<br /> The study also shows that vascular structure and function can be significantly improved in aged tissues, further supporting tissue regeneration and overall health.<br /> Through this innovative approach, the authors seek to overcome intrinsic cellular deficits and environmental changes within aged osteogenic compartments, ultimately achieving bone healing levels comparable to those seen in young mice.

Strengths:

The study is a strong example of translational research, employing robust methodologies across molecular, cellular, and tissue-level analyses. The authors leverage a clinically relevant, immunocompetent mouse model and apply advanced histological, transcriptomic, and functional assays to characterise age-related changes in bone structure and function. Major strengths include the use of single-cell RNA sequencing (scRNA-seq) to profile osteoprogenitor populations within the calvarial periosteum and suture mesenchyme, as well as quantitative assessments of mitochondrial health, vascular density, and osteogenic function. Another important point is the use of very old animals (up to 88 weeks, almost 2 years) modelling the human bone aging that usually starts >65 yo. This comprehensive approach enables the authors to identify critical age-related deficits in osteoprogenitor number, function, and microenvironment, thereby justifying the combined Wnt3a and IF intervention.

[Editors' note: The manuscript was evaluated positively by all three reviewers originally. In the revised manuscript, the authors included some new data following the reviewers' suggestions, while other comments were clarified in the response to the reviewers, and by revising the manuscript text. The new data further support the major conclusions of the paper.]

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Aging reduces tissue regeneration capacity, posing challenges for an aging population. In this study, the authors investigate impaired bone healing in aging, focusing on calvarial bones, and introduce a two-part rejuvenation strategy. Aging depletes osteoprogenitor cells and reduces their function, which hinders bone repair. Simply increasing the number of these cells does not restore their regenerative capacity in aged mice, highlighting intrinsic cellular deficits. The authors' strategy combines Wnt-mediated osteoprogenitor expansion with intermittent fasting, which remarkably restores bone healing. Intermittent fasting enhances osteoprogenitor function by targeting NAD+ pathways and gut microbiota, addressing mitochondrial dysfunction - an essential factor in aging. This approach shows promise for rejuvenating tissue repair, not only in bones but potentially across other tissues.

Strengths:

This study is exciting, impressive, and novel. The data presented is robust and supports the findings well.

Weaknesses:

As mentioned above the data is robust and supports the findings well. I have minor comments only.

We thank the reviewer for their enthusiastic and positive assessment of our study. We appreciate the recognition of the novelty and robustness of our data and findings. We have carefully considered the reviewer's comments and have revised the manuscript accordingly. We believe these revisions further strengthen the clarity and impact of our work.

Reviewer #2 (Public review):

Summary:

Reeves et al explore a model of bone healing in the context of aging. They show that intermittent fasting can improve bone healing, even in aged animals. Their study combines a 'bone bandage' which delivers a canonical Wnt signal with intermittent fasting and shows impacts on the CD90 progenitor cell population and the healing of a critical-sized defect in the calvarium. They also explore potential regulators of this process and identify mitochondrial dysfunction in the age-related decline of stem cells. In this context, by modulating NAD+ pathways or the gut microbiota, they can also enhance healing, hinting at an effect mediated by complex impacts on multiple pathways associated with cellular metabolism.

Strengths:

The study shows a remarkable finding: that age-related decreases in bone healing can be restored by intermittent fasting. There is ample evidence that intermittent fasting can delay aging, but here the authors provide evidence that in an already-aged animal, intermittent fasting can restore healing to levels seen in younger animals. This is an important finding as it may hint at the potential benefits of intermittent fasting in tissue repair.

Weaknesses:

The authors explore potential mechanisms by which the intermittent fasting protocol might impact bone healing. However, they do not identify a magic bullet here that controls this effect. Indeed, the fact that their results with intermittent fasting can be replicated by changing the gut microbiota or modulating fundamental pathways associated with NAD, suggests that there is no single mechanism that drives this effect, but rather an overall complex impact on metabolic processes, which may be very difficult to untangle.

We thank the reviewer for their positive assessment of our study and for highlighting the significant finding that intermittent fasting can restore age-related declines in bone healing. We appreciate the observation that our results suggest a complex interplay of metabolic processes rather than a single "magic bullet" mechanism. Indeed, the ability of gut microbiota modulation or NAD+ pathway targeting to replicate intermittent fasting's benefits underscores this complexity. While we recognize the challenges of disentangling these interconnected pathways, we believe our findings offer valuable insights into the multifaceted nature of intermittent fasting's impact on aged tissue repair. We hope this study serves as a foundation for future research aimed at identifying the individual contributions of these pathways and developing targeted therapeutic strategies.

Reviewer #3 (Public review):

Summary:

This study aims to address the significant challenge of age-related decline in bone healing by developing a dual therapeutic strategy that rejuvenates osteogenic function in aged calvarial bone tissue. Specifically, the authors investigate the efficacy of combining local Wnt3a-mediated osteoprogenitor stimulation with systemic intermittent fasting (IF) to restore bone repair capacity in aged mice. The highlights are:

(1) Novel Approach with Aged Models:

This pioneering study is among the first to demonstrate the rejuvenation of osteoblasts in significantly aged animals through intermitted fasting, showcasing a new avenue for regenerative therapies.

(2) Rejuvenation Potential in Aged Tissues:

The findings reveal that even aged tissues retain the capacity for rejuvenation, highlighting the potential for targeted interventions to restore youthful cellular function.

(3) Enhanced Vascular Health:

The study also shows that vascular structure and function can be significantly improved in aged tissues, further supporting tissue regeneration and overall health.<br /> Through this innovative approach, the authors seek to overcome intrinsic cellular deficits and environmental changes within aged osteogenic compartments, ultimately achieving bone healing levels comparable to those seen in young mice.

Strengths:

The study is a strong example of translational research, employing robust methodologies across molecular, cellular, and tissue-level analyses. The authors leverage a clinically relevant, immunocompetent mouse model and apply advanced histological, transcriptomic, and functional assays to characterise age-related changes in bone structure and function. Major strengths include the use of single-cell RNA sequencing (scRNA-seq) to profile osteoprogenitor populations within the calvarial periosteum and suture mesenchyme, as well as quantitative assessments of mitochondrial health, vascular density, and osteogenic function. Another important point is the use of very old animals (up to 88 weeks, almost 2 years) modelling the human bone aging that usually starts >65 yo. This comprehensive approach enables the authors to identify critical age-related deficits in osteoprogenitor number, function, and microenvironment, thereby justifying the combined Wnt3a and IF intervention.

Weaknesses:

One limitation is the use of female subjects only and the limited exploration of immune cell involvement in bone healing. Given the known role of the immune system in tissue repair, future studies including a deeper examination of immune cell dynamics within aged osteogenic compartments could provide further insights into the mechanisms of action of IF.

We thank the reviewer for their thorough summary and positive assessment of our study, particularly highlighting its translational nature, the robust methodologies employed, and the relevance of our aged animal model. We appreciate the insightful suggestion to include male subjects and to explore immune cell dynamics in future investigations.

We acknowledge the limitation of using only female mice in the current study and agree that future studies incorporating both sexes and investigating immune cell contributions within aged osteogenic compartments would offer valuable insights into the mechanisms underlying intermittent fasting and its impact on bone healing.

Our focus on female mice was informed by their distinct characteristics, including delayed healing and higher fracture risk (PMID: 37508423, PMID: 34434120). Importantly, female mice present a more challenging case for bone repair, making them a stringent test for evaluating the effectiveness of our rejuvenation approaches. Moreover, our research protocol, approved under animal license, adhered to ethical principles and the 3Rs, allowing us to reduce the number of animals required by focusing on a single sex.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) The authors should provide a justification for the use of female mice in this study. Additionally, the section on animal methods should be expanded to align with ARRIVE guidelines.

We thank the reviewer for their valuable feedback. In response to the comment regarding the use of female mice, we have included a justification in the updated manuscript. As noted, female mice were selected for this study due to their distinct characteristics, such as delayed healing and higher fracture risk (PMID: 37508423, PMID: 34434120), which provide a more challenging model for evaluating bone repair strategies. We believe this made our study a stringent test of the efficacy of the rejuvenation approaches being investigated.

Additionally, we have revised the animal methods section to ensure it aligns with the ARRIVE guidelines.

(2) Intermittent fasting can influence circadian rhythms in various ways. In the RNA-seq data, do the authors observe any changes related to circadian rhythm pathways?

The reviewer raises an important point regarding the influence of intermittent fasting (IF) on circadian rhythms. Our RNA-seq data revealed significant alterations in circadian rhythm pathways, particularly within the aged periosteal CD90+ cell population during IF. Specifically, the PAR bZip family transcription factors Dbp, Hlf, and Tef (q < 0.05) were significantly upregulated, consistent with their established roles as circadian rhythm regulators (PMID: 16814730, PMID: 31428688).

In suture CD90+ cells from the Aged + IF group, Dbp expression was significantly elevated compared to the Aged AL control group. Moreover, several other circadian-controlled genes, including Sirt1, Kat2b, Csnk1e, Ezh2, Fbxw11, and Ucp2 (p < 0.05), were also upregulated (Fig. 4b), suggesting enrichment of Clock/Per2/Arntl transcriptional targets, essential components of the circadian clock.

The observed upregulation of circadian rhythm effectors like Dbp, Hlf, and Tef further suggests a potential role for circadian transcription in CD90+ cell rejuvenation and bone repair in aged mice. While previous studies have primarily focused on the role of circadian rhythms in osteoblasts in vitro (PMID: 34579752, PMID: 30290183), our findings provide compelling evidence for their involvement in bone regeneration in vivo, providing compelling evidence for future investigation into this mechanism.

Chip-SEQ studies have shown D-box sites near promoters in Wnt/β-catenin components (e.g. Lrp6, Lrp5, Wnt8a, Fzd4) in pro-osteogenic transcription factor Zbtb16 (and see Fig 5), and in 11 of the 44 mouse collagen genes (PMID: 31428688). These components are known to regulate osteogenesis, and their proximity to circadian-controlled transcription factors suggests a possible overlap between circadian regulation and Wnt signaling in promoting bone repair.  Additionally, circadian rhythmicity, stem cell function, and Wnt signaling are interlinked (PMID: 29277155, PMID: 25414671). Food intake is a powerful regulator of the circadian rhythm in several organs (PMID: 11114885, PMID: 32363197), but little is known about the diet-circadian interaction in bone repair. The possibility that circadian transcription can be harnessed to target Aged stem cell function towards bone repair is a promising prospect.

We have incorporated this information in Figure 2 - figure supplement 3G-H, the results section as well as in the discussion.

Reviewer #2 (Recommendations for the authors):

(1) The authors refer to 'altered cellular mechanobiology', 'age-related changes in mechanobiology', etc. Here, they are using this terminology to refer to changes in F-actin intensity and nuclear shape. While I agree that these measures are indicators of a cellular response to mechanical cues, calling this 'changes in mechanobiology' doesn't sound quite correct to me. 'Mechanobiology' to me, is a field of study. Perhaps the authors should consider changing their terminology.

We appreciate the reviewer’s insightful comment on the terminology used in our manuscript. We agree that the term "mechanobiology" is a broad field of study and using it in the context of changes in F-actin intensity and nuclear shape may be misleading. We have revised the text to better reflect the specific cellular responses to mechanical cues, such as changes in the cytoskeleton and nuclear morphology, rather than referring to them as "altered mechanobiology." The updated terminology more accurately conveys the observed cellular alterations in response to mechanical forces. We have made these adjustments throughout the manuscript for clarity and precision.

(2) Three of the measures the authors use to highlight age-related changes (and rejuvenation) in their animal model are F-actin intensity, nuclear shape, and vascularisation. However, they never really explain what they believe these readouts mean practically/functionally. Indeed, it makes sense that less vascularisation would be associated with an aged phenotype and preclude healing, but this is only mentioned somewhat cursorily in the discussion. While vascularisation is discussed in the context of aging in the discussion, it is not discussed in the context of healing (which would seem relevant in the context of vascularisation being used as a readout in the healing models in response to Akk and IF treatment). Similarly, the changes in F-actin intensity and nuclear shape might suggest changes in the stiffness of the periosteum (as mentioned in the discussion), which could indeed be an indicator of an aged phenotype; however, their role in healing (in response to Akk and IF) are not clearly articulated.

We appreciate the reviewer’s insightful comments and have made revisions to clarify the implications of age-related changes in vascularization, F-actin intensity, and nuclear shape, as well as the functional significance of these observations in the context of healing and rejuvenation.

Vascularization:

Vascularization and modulation of blood flow are critical for calvarial bone repair, as supported by multiple studies (e.g., PMID: 38032405, PMID: 21156316, PMID: 25640220). Early in the calvarial repair process, blood vessels grow independently of osteoprogenitor cells, establishing a supportive environment that promotes osteoprogenitor migration and subsequent ossification (PMID: 38834586). Furthermore, angiogenic vessels from the periosteum at defect edges contribute to creating a specialized microenvironment essential for bone healing (PMID: 38834586, PMID: 38032405). Compromised vascularization significantly impairs the healing of critical-sized calvarial defects (PMID: 29702250).

Our data reveal a decline in periosteal vascularization with age, potentially compromising this microenvironment and impairing repair in aged animals. Importantly, our findings indicate that intermittent fasting (IF) reverses this phenotype by restoring periosteal vascularization. This rejuvenation of the vascular microenvironment aligns with improved bone repair outcomes in aged mice subjected to IF. We have revised the manuscript to emphasize the importance of vascularization in healing and to highlight the role of IF in restoring this critical aspect of the bone healing microenvironment.

F-actin intensity and nuclear shape:

Age-related changes in F-actin intensity and nuclear shape are associated with increased tissue stiffness, a hallmark of aging. Tissue stiffness has been shown to impair progenitor cell function and hinder repair in various systems, including neuroprogenitors (PMID: 31413369). Softening the extra cellular matrix in aged tissues has been demonstrated to partially restore progenitor function and improve repair outcomes, as seen in the case of neuroprogenitors (PMID: 31413369). In our study, IF reversed age-associated changes in F-actin expression and nuclear shape, restoring these parameters to a phenotype resembling that of younger animals. This suggests that IF mitigates the mechanical changes associated with aging, reducing tissue stiffness and rejuvenating the periosteum to facilitate improved bone healing, similar to the outcomes observed in younger models.

Following the reviewer’s advice, we have revised the text to clearly articulate the correlations and interpretations of our data regarding tissue mechanics and bone repair. Thank you for highlighting these critical aspects.

(3) In relation to my point 2) on nuclear shape, there are reports that aging is linked to changes in Lamin B1. Have the authors considered this? It might provide a clearer link between their data and the tissue-level phenotypes they observe.

Thank you for your comment regarding the potential link between aging and changes in Lamin B1. Following your suggestion, we performed Lamin B1 immunostaining on samples from Young, Adult, Aged, and Aged + IF groups. However, no significant differences in Lamin B1 levels were observed across these groups. These findings indicate that changes in Lamin B1 in osteoprogenitors are not apparent during aging, suggesting that Lamin B1 alterations in the context of aging may be tissue- and cell-type-specific.

The new data was added in Figure 1 - figure supplement 2i-j.

(4) In the data associated with Figure 2, the authors find that in the aged mice, MMP9 expression is increased, but MMP2 expression is decreased. They associate the decrease in MMP2 expression with decreased migration, but the canonical function of MMP9 should be similar to that of MMP2. Are there tissue-specific differences in the activity of MMP2/9 that could account for this?

Thank you for the thoughtful comment. While both MMP-2 and MMP-9 are involved in ECM remodeling and share some overlapping canonical functions, their roles are context-dependent and exhibit tissue-specific differences that could explain the observed changes in aged mice. MMP-2 has been shown to play a critical role in maintaining the structural and functional integrity of flat bones, such as those in the craniofacial skeleton, by supporting bone remodeling (PMID: 17400654, PMID: 17440987, PMID: 16959767). The decreased expression of MMP-2 in aged mice may impair these local processes, leading to reduced migratory capacity of osteoprogenitors and contributing to aging-related changes in flat bone structure and function.

In contrast, MMP-9 is more prominently involved in long bone remodeling, particularly at the growth plate where it regulates hypertrophic chondrocyte turnover, vascularization, and ossification during endochondral bone formation (PMID: 21611966, PMID: 9590175, PMID: 23782745, PMID: 16169742 ). Additionally, MMP-2 and MMP-9 differ in their regulation of specific ECM substrates and their interactions with bone-resident cells, which may further drive divergent outcomes in distinct bone types. For example, MMP-9’s role in osteoclastogenesis and its regulation of ECM proteins like type I collagen could be more critical in long bones, while MMP-2’s involvement in fine-tuning ECM microarchitecture may hold greater importance in flat bones.

The increased expression of MMP-9 in aged calvarial osteoprogenitors may reflect a compensatory mechanism in response to the reduced MMP-2 activity, possibly in response to increased ECM turnover demands. Further studies examining the precise molecular pathways driving these changes in osteoprogenitors will help clarify the underlying mechanisms and their contributions to age-related alterations in flat bone structure and function.

(5) In lines 391-2, the authors conclude that the data from Figure 4 shows that "during IF, CD90 cells, despite being aged, are more capable of ECM modulation and migration". The authors certainly present evidence that this is true, but the RNAseq showed that the enriched GO terms were predominantly associated with immune responses ('response to cytokine') and the proliferation phenotype seems very strong. Therefore, I would suggest that this overarching statement regarding the findings be less focussed on this one aspect of the finding, which doesn't look to be the dominant phenotype of the cellular response. And indeed, the authors move on from here to explore a mechanism associated with metabolism, not specifically with ECM remodelling.

We greatly appreciate the reviewer insight regarding the interpretation of our findings, particularly the conclusion drawn from Figure 4.

In response, we have revised the conclusion to more accurately reflect these findings.

The revised text in the conclusion now reads: " Together, these findings suggest that IF rejuvenates aged CD90+ cells, in part, by enhancing proliferation, immune response, ECM remodeling, Wnt/β-catenin pathway, and metabolism, including increased ATP levels and decreased AMPK levels.”

We hope that this adjustment better aligns with your suggestion and provides a more accurate summary of the key findings.

(6) Fasting blood glucose levels are often cited as an indicator of metabolic health. Did the authors look at this in their animals who underwent the IF protocol? Could this have had an impact on the healing response?

We thank the reviewer for this insightful comment. Throughout our study, we have withdrawn blood from the animals for various analyses that were not included in this manuscript in order to maintain focus on the osteoprogenitors.

Our analysis included the assessment of the metabolic health of the animals using fasting blood glucose levels and the area under the curve (AUC) of the intraperitoneal glucose tolerance test (IPGTT).

Fasting blood glucose levels reflect the animals' ability to maintain stable glucose levels after fasting, while the AUC from the IPGTT measures how efficiently glucose is cleared from the bloodstream following a glucose challenge. Typically, lower fasting blood glucose levels and reduced AUC indicate improved insulin sensitivity, better glucose metabolism, and enhanced metabolic control (PMID: 18812462, PMID: 19638507).

Our findings show that intermittent fasting (IF) significantly reduced both the fasting blood glucose levels and the AUC in the IPGTT. This indicates that IF enhances metabolic flexibility, likely through improved insulin sensitivity and better glucose homeostasis. By lowering fasting blood glucose, IF reduces the reliance on excessive gluconeogenesis during fasting, while a reduced AUC indicates more efficient postprandial glucose clearance, consistent with enhanced insulin action and reduced fluctuations in blood glucose levels. The new data has been incorporated in Figure 3 - figure supplement 1d-g.

Methods:

“Blood glucose level measurement

Fasting blood glucose levels were measured (Accu-Check tests strips) from 6h fasting mice by blood sampling the tail vein. For intraperitoneal glucose tolerance test (IPGTT), glucose was injected intraperitoneally (2 g/kg), and the blood glucose levels were measured after 15, 30, 60 and 120 minutes.”

Improved metabolic health through lower fasting glucose and reduced AUC can have profound implications for tissue repair (PMID: 32809434). Stable glucose levels ensure a consistent energy supply for key cellular processes, such as cell proliferation, migration, and differentiation, which are essential for regeneration. Enhanced insulin sensitivity supports nutrient delivery to cells and reduces inflammation, creating an environment conducive to tissue healing. Additionally, intermittent fasting's ability to optimize glucose metabolism and regulate insulin secretion may enhance the function of stem and progenitor cells, further improving the tissue repair process (PMID: 28843700). Together, these findings suggest a mechanistic link between improved metabolic health and the enhanced healing observed in animals subjected to intermittent fasting.

(7) In Supplementary Figure 10, the authors look at bone remodelling by assessing TRAP staining, as an indicator of osteoclast activity. I'm not sure if these data add all that much to the study. The authors have looked at bone formation at a tissue level using microCT. Here, they look at bone resorption at a cellular level with the TRAP assay. Overall, this probably suggests more bone remodelling, but the TRAP assay on its own at the cellular level could also be interpreted as an osteoporosis-like phenotype. This is clearly not the case because the authors show robust bone healing by microCT. In short, as an isolated measure of osteoclast activity at the cellular level without cellular-level assays of osteoblast activity, the interpretation of these data is not that clear. The microCT speaks far more of the phenotype and is, in my opinion, sufficient to make this point.

We thank the reviewer for their comments regarding the interpretation of the TRAP staining data and its context within the study. We appreciate the concern that, without direct assays of osteoblast activity, the TRAP assay could lead to ambiguity.

We have shown that intermittent fasting significantly increases the number and function of osteoprogenitor cells, the precursors to osteoblasts. While we acknowledge that these data do not directly measure osteoblast numbers or activity, they strongly suggest an increased capacity for osteoblast differentiation and bone formation. This aligns with the microCT findings of robust bone structure and healing.

After careful consideration and given that the microCT and histology findings  already provide robust and comprehensive evidence for bone structure and healing, we have decided to remove the TRAP staining data from the manuscript. We believe this change simplifies the manuscript and strengthens its focus on the most impactful data.

(8) In the discussion, the authors make a number of links between aging and IF. However, one of the exciting conclusions of this manuscript is that IF aids in healing in aged animals. In this context, IF has not impacted the aging process itself because the animals have not experienced an IF protocol across their lifespan, but rather only after injury. In this context, perhaps the authors should also be focussing their discussion on evidence of the short-term response to IF rather than its effects on aging, which are longer-term.

We appreciate the reviewer's comment and agree that emphasizing the short-term effects of intermittent fasting is crucial. Our study is the first to examine this protocol in Aged animals.

To address this, we have revised the discussion and highlighted how short-term IF enhances metabolic health, promotes osteoprogenitor functionality, and supports bone remodeling, as observed in our study.

Reviewer #3 (Recommendations for the authors):

(1) The authors should clarify details on intermittent fasting protocols, especially regarding caloric intake differences between fasting and non-fasting days, to aid reproducibility.

We appreciate the reviewer's suggestions and have incorporated them by clarifying the relevant details. The new data are presented in Figure 3 - figure supplement 1a-c.

Methods:

“Caloric intake calculation

To assess the caloric intake of mice, the food was weighted when made available to the mice (Win), and when removed (Wout). The daily consumed food was calculated based on the weight difference (Win - Wout), then converted to kcal (1 g = 3.02 kcal, LabDiet, 5053), and expressed as kcal/mouse/day for each cage (n cage ³ 3 with 1 to 5 mice/cage).”

(2) Did the authors evaluate the effect of their intermittent fasting protocol on fasting blood glucose levels?

Following the reviewer comment we included two measurements: 1) Fasting blood glucose, which reflects the ability to maintain glucose homeostasis during fasting, and 2) fasting blood glucose levels and the area under the curve (AUC) of the intraperitoneal glucose tolerance test (IPGTT), which measures glucose clearance efficiency after a glucose challenge. Lower values for both typically indicate improved insulin sensitivity, glucose metabolism, and metabolic control.

Our findings demonstrate that intermittent fasting significantly reduced both fasting blood glucose and IPGTT AUC, suggesting enhanced metabolic flexibility, likely through improved insulin sensitivity and glucose homeostasis. Lower fasting blood glucose with IF indicates reduced reliance on gluconeogenesis during fasting, while a reduced AUC suggests more efficient postprandial glucose clearance, consistent with enhanced insulin action and reduced blood glucose fluctuations. This new data is included in Figure 3 - figure supplement 1.

Generally, the improved metabolic environment supports tissue repair by ensuring adequate energy for cell proliferation and migration, reducing inflammation, and promoting the function of stem cells involved in tissue regeneration. Thus, this outcome of intermittent fasting may create a more favorable environment for tissue repair, potentially accelerating the healing of damaged tissues and improving overall regenerative capacity.

(3) In Figure 1E-F, the nuclei have an interesting shape and the authors quantified F-actin. Given the role of lamin B in nuclear integrity, an analysis of lamin B expression and its structural integrity in aged osteoprogenitors could provide valuable insights into cellular aging mechanisms and their potential reversal with intermittent fasting.

In response to the reviewer's comment, we performed Lamin B1 immunostaining on samples from Young, Adult, Aged, and Aged + IF groups. We observed no significant differences in Lamin B1 levels across these groups. This suggests that age-related changes in Lamin B1 are not evident in osteoprogenitors and may be tissue- or cell-type specific. The new data was added in Figure 1 - figure supplement 2i-j.

(4) The authors should explain, in the main text or the methods section, why are they only using females in this study.

We appreciate the reviewer's comment regarding the use of female mice. Female mice were chosen for this study due to their delayed healing and higher fracture risk (PMID: 37508423, PMID: 34434120), presenting a more challenging model for evaluating bone repair strategies and providing a stringent test of our rejuvenation approaches. This justification has been added to the revised manuscript. The animal methods section has also been updated to comply with ARRIVE guidelines.

(5) This story stands alone and has an incredible amount of data. However, for a follow-up study, I would like to suggest consideration of including a broader analysis of immune cell involvement within the osteogenic compartments to strengthen the mechanistic understanding of IF's impact.

We thank the reviewer for this insightful suggestion. We agree that investigating the role of immune cells within the osteogenic compartments could provide valuable mechanistic insights into how intermittent fasting influences tissue regeneration. Immune cells are key mediators of inflammation and repair, and their interactions with osteoprogenitors and other cells in the bone healing environment likely contribute to IF's effects.

While our study focuses on IF's impact on osteoprogenitor function and tissue repair, we acknowledge the importance of future research exploring immune cell involvement. Techniques like single-cell RNA sequencing or flow cytometry could characterize immune cell populations and their functional states within osteogenic niches, allowing for a deeper understanding of immune-skeletal interactions during IF-mediated bone healing. We appreciate the reviewer highlighting this promising avenue for future research.

Minor corrections to the text and figures:

(1) References formatting should be revised (eg. line 41).

The reference formatting was corrected.

(2) Line 144 - what do the authors mean by p2 in the references?

Thank you for your comment, we corrected the error and removed p2 from the reference.

Author response:

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

Reviewer #1 (Public review): 

Summary: 

Nitric oxide (NO) has been implicated as a neuromodulator in the retina. Specific types of amacrine cells (ACs) produce and release NO in a light-dependent manner. NO diffuses freely through the retina and can modulate intracellular levels of cGMP, or directly modify and modulate proteins via S-nitrosylation, leading to changes in gap-junction coupling, synaptic gain, and adaptation. Although these system-wide effects have been documented, it is not well understood how the physiological function of specific neuronal types is affected by NO. This study aims to address this gap in our knowledge. 

There are two major findings. 1) About a third of the retinal ganglion cells display cell-type specific adaptation to prolonged stimulus protocols. 2) Application of NO specifically affected Off-suppressed ganglion cells designated as G32 cells. The G32 cluster likely contains 3 ganglion cell types that are differentially affected. 

This is the first comprehensive analysis of the functional effects of NO on ganglion cells in the retina. The cell-type specificity of the effects is surprising and provides the field with valuable new information. 

Strengths: 

NO was expected to produce small effects, and considerable effort was expended in validating the system to ensure that changes in the state of the preparation would not confound any effects of NO. The authors used a sequential stimulus protocol to control for changes in the sensitivity of the retina during the extended recording periods. The approach potentially increases the sensitivity of the measurements and allows more subtle effects to be observed. 

Neural activity was measured by Ca-imaging. Responsive ganglion cells were grouped into 32 types using a clustering analysis. Initial control experiments demonstrated that the celltypes revealed by the analysis largely recapitulate those from their earlier landmark study using a similar approach. 

Application of NO to the retina modulated responses of a single cluster of cells, labeled G32, while having little effect on the remaining 31 clusters. In separate experiments, ganglion cell spiking activity was recorded on a multi-electrode array (MEA). Together the Ca-imaging and MEA recordings provide complementary approaches and demonstrate that NO modulates the temporal but not spatial properties of affected cell-types.

Weaknesses: 

The concentration of NO used in these experiments was ~0.25µM, which is 5- to 10-fold lower than the endogenous concentration previously measured in rodent retina. It is perhaps surprising that this relatively low NO concentration produced significant effects. However, the endogenous measurements were done in an eye-cup preparation, while the current experiments were performed in a bare (no choroid) preparation. Perhaps the resting NO level is lower in this preparation. It is also possible that the low concentration of NO promoted more selective effects.

Reviewer #2 (Public review): 

Neuromodulators are important for circuit function, but their roles in the retinal circuitry are poorly understood. This study by Gonschorek and colleagues aims to determine the modulatory effect of nitric oxide on the response properties of retinal ganglion cells. The authors used two photon calcium imaging and multi-electrode arrays to classify and compare cell responses before and after applying a NO donor DETA-NO. The authors found that DETA-NO selectively increases activity in a subset of contrast-suppressed RGC types. In addition, the authors found cell-type specific changes in light response in the absence of pharmacological manipulation in their calcium imaging paradigm. This study focuses on an important question and the results are interesting. The limitations of the method and data interpretation are adequately discussed in the revised manuscript. 

The authors have addressed my previous comments, included additional discussions on the limitations of the method, and provided a more careful interpretation of their data. 

Recommendations for the authors: 

Please correct the citation that reviewer #1 mentioned. In addition, a little more discussion of the NO concentration issue would be helpful. The low NO concentration is not a weakness in the data; it simply raises questions regarding the interpretation.

Thank you for these recommendations.

Regarding the citation error, we are not sure if Reviewer #1 refers to a citation   formatting error or incorrect placement. In any case, we modified the text: We  specified the extracted information regarding the NO concentrations and put the  applied concentration into that context (Lines 621-635). In addition, we made clear  that the citation of Guthrie (2014) refers to the dissertation, which can be easily  retrieved via Google Scholar. We also cited the mentioned ARVO abstract by   Guthrie and Mieler (2014). 

We hope that these modifications solve the above-mentioned issues. 

---

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

Reviewer #1 (Public Review):  

Summary: 

Nitric oxide (NO) has been implicated as a neuromodulator in the retina. Specific types of amacrine cells (ACs) produce and release NO in a light-dependent manner. NO diffuses freely through the retina and can modulate intracellular levels of cGMP, or directly modify and modulate proteins via S-nitrosylation, leading to changes in gap-junction coupling, synaptic gain, and adaptation. Although these system-wide effects have been documented, it is not well understood how the physiological function of specific neuronal types is affected by NO. This study aims to address this gap in our knowledge. 

Strengths: 

NO was expected to produce small effects, and considerable effort was expended in validating the system to ensure that any effects of NO would not be confounded by changes in the state of the preparation. The authors used a paired stimulus protocol to control for changes in the sensitivity of the retina during the extended recording periods. The approach potentially increases the sensitivity of the measurements and allows more subtle effects to be observed. 

Neural activity was initially measured by Ca-imaging. Responsive ganglion cells were grouped into 32 types using a clustering analysis. Initial control experiments demonstrated that the cell-types revealed here largely recapitulate those from their earlier landmark study using the same approach (Fig. 2). 

Application of NO to the retina strongly modulated responses of a single cluster of cells, labeled G32, while having little effect on the remaining 31 clusters. This result is evident in Fig. 3e. 

Separate experiments measured ganglion cell spiking activity on a multi-electrode array (MEA). Clustering analysis of the peri-stimulus spike-time histograms (PSTHs) obtained from the MEA data also revealed 32 clusters. The PSTHs for each cluster were aligned to the Ca-imaging data using a convolution approach. The higher temporal resolution of the MEA recordings indicated that NO increased the speed of sub-cluster 2 responses but had no effect on receptive field size. The physiological significance of the small change in kinetics remains unclear. 

We thank the reviewer for their detailed and constructive comments.

Weaknesses: 

The G32 cluster was further divided into three sub-types using Bayesian Information Criterion (BIC) based on the temporal properties of the Ca-responses. This sub-clustering result seems questionable due to the small difference in the BIC parameter between 2 and 3 clusters. Three sub-clusters of the G32 cluster were also revealed for the PSTH data, however, the BIC analysis was not applied to further validate this result. 

(1.1) We agree with the reviewer that this is an important point to be clarified. To this end, we repeated the analysis with n=2 clusters (see Author response image 1 below). In brief, we found that the overall interpretation did not change: Both clusters in the Ctrl1-dataset showed barely any type-specific adaptational effects, whereas under NO application, temporal contrast responses decreased (see Author response image 1 below). If requested, we would be happy to add this image to the supplementary material. 

Author response image 1.

In an additional analysis, we evaluated if n=2 or n=3 was the “better” choice for the number of clusters. In the new Supplementary Fig. S4, we compared the clusters with n=2 (top) and n=3 (bottom). For n=2, the two clusters are relatively strongly correlated for both visual stimuli, whereas for n=3, the clusters become more distinct, especially with respect to differences in the correlations for the two stimuli (Fig. S4b). For n=2, the low intra-cluster correlation (ICC) strongly suggests that cluster 2 contains multiple response types (ICC(C2) = 0.5 ± 0.48, mean ± s.d.; Fig. S4c). For n=3, the mean ICC values are high for all three clusters (ICC(C1) = 0.81 ± 0.16; ICC(C2) = 0.86 ± 0.07; ICC(C3) = 0.83 ± 0.1; mean ± s.d.). Together, this suggests that n=3 clusters captures the response diversity in G32 better than n=2 clusters. 

Finally, we performed a BIC analysis for the MEA dataset and found the optimal number of clusters to be also n=3 (see new Suppl. Fig. S5).

The alignment of sub-clusters 1, 2, and 3 identified in the Ca-imaging and the MEA recordings seemed questionable, because the temporal properties of clusters did not align well, nor did the effects of NO. 

(1.2) To address this important point, we analyzed the correlations between the control responses of the three clusters from the Ca²⁺-dataset with the ones from the MEA-dataset (see new Suppl. Fig. S7). To avoid confusion, we named the clusters in the MEA-dataset i,ii,iii (see Fig. 8). We found two of the three clusters to be highly correlated (Ca²⁺ clusters 2,3 and MEA clusters iii, ii), whereas one cluster was much less so (cluster 1 vs. cluster i), likely due to differences in response kinetics. In clusters i and ii NO application led to a release of suppression for temporal contrasts – similar to what we observed in the Ca²⁺ data (see also our new analysis of the MEA data in Suppl. Fig. S6, as discussed further below).

We agree that the cell types underlying the Ca²⁺ and MEA G32 clusters may not be the same – aligning functional types between those two methods is challenging due to several factors, mainly because while Ca²⁺ is a proxy for spiking activity, other Ca²⁺ sources as well as sub-threshold membrane potential changes affect the intracellular Ca²⁺, potentially in a cell type-specific way. We explain this now better in the text.

In any case, our main point was not to unambiguously align the cell types but to show that in both datasets, we find three subclusters of G₃₂, which are affected by NO in a differential manner, particularly their suppression to temporal contrasts.

The title of the paper indicates that nitric oxide modulates contrast suppression in a subset of mouse retinal ganglion cells, however, this result appears to be inferred from previous results showing that G32 is identified as a "suppressed-by-contrast" cell. The present study does not explicitly evaluate the amount of contrast-suppression in G32 cells. 

(1.3) The reviewer is correct in that we did not quantify contrast-suppression in G₃₂ in detail but focused on the responses to temporal contrast (chirp and moving bar) and its modulation by NO (Fig. 5). In this context, please note that G₃₂’s responses to the moving bar stimulus suggests that the cells are also suppressed by spatial contrast (i.e., an edge appearing in their RF). The functional RGC type G₃₂ (“Off suppressed 2”) was defined in an earlier study (Baden et al. 2016); it was assigned to the “Suppressed-by-Contrast” (SbC) category mainly because temporal contrast suppresses its responses. Already then, coverage analysis indicated that G₃₂ may indeed contain several RGC types – in line with our clustering analysis. It is still unclear if G₃₂ contains one (or more) of the SbC cells described by Jacoby & Schwartz (2018); in their recent study, Wienbar and Schwarz (2022) introduced the novel bursty-SbC RGC, which Goetz et al. (2022) speculated to potentially align with G₃₂.<br /> We now discuss the relationship between G₃₂ and the SbC RGCs defined in other studies in the revised manuscript.

In its current form, the work is likely to have limited impact, since the morphological and functional properties of the affected sub-cluster remain unknown. The finding that there can be cell-specific adaptation effects during experiments on in vitro retina is important new information for the field.

(1.4) Again, we thank the reviewer for the detailed and helpful feedback. We hope that the reviewer finds our revised manuscript improved.

Reviewer #1 (Recommendations For The Authors):  

Most of the calcium activity traces (dF/F) throughout the paper have neither vertical nor horizontal calibration bars. Presumably, most values are positive, but this is unclear as a zero level is not indicated anywhere. Without knowing where zero dF/F is, it is not possible to determine whether the NO increased the Ca-signal or blocked a decrease in the Ca-signal. 

Both ∆F/F and z-scoring, as we used here, are ways to normalize Ca²⁺ traces. We decided against using ∆F/F₀ because this typically assumes that F represents the cell’s Ca²⁺ resting level (F₀; without activity). However, in our measurements, the “resting” Ca²⁺ levels (i.e. before presenting a stimulus) may indeed reflect no spiking activity (e.g., in an ON RGC) but may also reflect baseline spiking activity (e.g., in an G₃₂, which has a baseline firing rate of ~10 Hz; see Fig. S6). Hence, we used z-scoring, which carries no assumption of resting Ca²⁺ level equal to no activity. In practice, we normalized all traces to the Ca²⁺ level prior to the light stimulus and defined this as zero (as described in the Methods).

We considered the reviewer’s suggestion of adding zero lines to every trace but felt that this would hamper the overall readability of the figures.

Regarding calibration bars: We made sure that horizontal bars (indicating time) are present in all figures. We decided to leave out vertical bars in Ca²⁺ responses, because as explained above, the traces are normalized (and unit-free), and within a figure all traces are scaled the same.

Points of clarification for the Methods: 

(1) The stimulus field was 800 x 600 µm. Presumably, both scan fields were contained within this region when scanning either Field 1 or Field 2 so that the adaptation level of the preparation at both locations was maintained? 

Yes, the stimulation field is always kept centered on the respective recording (scan) field and the adaptation level for each recording field was maintained.

(2) There appeared to be an indeterminate amount of time between the initial 10-minute adaptation period and Ctrl1, whereas there were no such gaps between subsequent scans. Is this likely to produce differences in adaptation state and thus represent a systematic error? 

At this time point, recording (scan) fields were selected to make sure that the cells in the field were uniformly labelled with the Ca²⁺ indicator and responsive to light stimuli. This typically happened already at the end of the light adaptation phase and/or right after. When selecting the fields, light stimuli were presented (to test responsiveness) and thereby the adaptation level was maintained independent of the duration of this procedure, minimizing systematic errors.

(3) Was the dense white noise stimulus applied during the wash-in period to maintain the adaptation state of the preparation prior to the subsequent scan? 

The dense noise was not applied throughout the wash-in period but at least 5-10min before the field was recorded with a drug (e.g., NO). 

Fig. 1d illustrates very nicely how the stimuli align with the responses. It would have been helpful to have this format continue throughout the paper but unfortunately, the vertical lines are dropped in Fig. 2a and then the stimulus waveform is omitted in Fig. 2e onwards. 

Thanks, good idea. We added the vertical lines and the stimulus waveform to the figures where they were missing to improve the readability. 

What was the rationale for selecting the concentration of the NO donor used? Is it likely to mimic natural levels? 

A DETA/NO concentration of 100 µM is commonly used in studies investigating NOinduced effects. DETA/NO has a half-life time (t_0.5) of 20 hours, which makes it more suitable for application in tissues (like our whole-mount preparation), because the donor can penetrate into the issue before releasing NO. In turn, this long t0.5 means that only a fraction of the bound NO is released per time unit.

Based on t_0.5 for DETA/NO and NO, one can roughly estimate the NO range as follows: t_0.5 of NO strongly depends on the tissue and is estimated in the second to minute range (Beckman & Koppenol, 1996). Assuming a t_0.5 for NO of 2 minutes, a freshly prepared 100 µM DETA/NO solution is expected to result within the first hour a NO concentration of approx. 0.25 µM (taking into account that 1 mole of DETA/NO releases 1.5 moles of NO molecules; see Ramamurthi & Lewis 1997).

In general, it is difficult to determine the physiological concentration of NO in the retina. Different measurements point at peaks of a few 100 nM (e.g., frog retina, ganglion cells: 0.25 µM, Kalamkarov et al. 2016; rodent inner retina, 0.1 to 0.4 µM, Micah et al. 2014). Hence, the NO concentrations we apply should be within the measured physiological range.

Fig. 3e: what are the diamond symbols? If these are the individual cells, it might be better to plot them on top of the box plots so all are visible. 

Indeed, the diamond symbols represent individual cells, yet outliers only. We decided not to plot all cells as a dot plot on top of the box plots since the readability will suffer as there are too many individual dots to show, e.g., n=251 for G₃₂ Ctrl and n=135 for G₃₂ DETA/NO.

Fig. 3: please explain more clearly the x-axis units in a-d and the y-axis units in e. 

To estimate potential response differences between the first and the second scan (i.e. either Ctrl 2 or NO), the traces were subtracted cell-pairwise (∆ Ctrl: Ctrl 2 – Ctrl 1; ∆ DETA/NO: NO – Ctrl 1). As all Ca²⁺ traces were normalized, they are unit-free. Therefore, the x-axes in Fig. 3a-d represent the mean differences of each cell per cell type, e.g., a value of zero would mean that the traces of Ctrl 1 and Ctrl 2 for a cell are identical. The y-axis in Fig. 3e is also unit-free, because technically, it is the same measure as Fig. 3a-d. But since it summarizes the control- and NO-data, we refer to this as “delta mean trace.” We tried to make this clearer in the revised manuscript and a detailed description can be found in the Methods.

Fig. 3: "...a substantial number of RGC types (34%) changed their responses to chirp and/or moving bar stimuli in the absence of any pharmacological perturbation in a highly reproducible manner...". How many of the cell types showed a significant difference? Two cell-types with p<0.001are highlighted with 3 asterisks. It would be helpful to indicate on this plot which of the other cells showed significant differences. 

Yes, this is a good idea. Thank you. We tried to add this information to the figure, but it became rather crowded. Therefore, we added a new Suppl. Fig. S3 (same style as Fig. 3) where we exclusively summarized the control-dataset. 

Fig. 7: To illustrate the transform from PSTH to Ca-imaging, why not use G32 data as an example?

Fair point. We modified the figure and added G₃₂ as an example.

It would be clearer if the cells were labeled consistently throughout the paper using their Baden cluster numbers rather than switching to the older nomenclature (JAM-B, local edge, alpha, etc), e.g. Fig. 7a,b. 

In the revised manuscript, we now changed the nomenclature to the Ca2+ Baden et al. (2016) terminology. We used the alternative cell type names here because where Fig. 7a is discussed in the manuscript, the cell type matching did not happen yet. But we agree that a consistent nomenclature is helpful.

The evidence supporting the sub-clustering of the G32 cells for the two recording methods could have been stronger. In Fig. 5, the BIC difference between 2 and 3 clusters is rather small. Is this result robust enough to justify 3 rather than 2 clusters? The BIC analysis should also be performed on the PSTH data-set to support the notion that the MEA G32 cluster also contains 3 rather than 2 sub-clusters. 

Regarding the sub-clustering of G₃₂ into n=2 or n=3 clusters for both datasets, please see our detailed reply #1.1 in our response to the public comments above.

The alignment of the three sub-clusters across the Ca-imaging and MEA data looked questionable. For example, the cluster 2 and cluster 3 traces in Fig. 5e,f look similar, with cluster 1 being more different. In Fig. 8c on the other hand, cluster 1 and 3 look similar with cluster 2 being more different. The pharmacological results also did not align well. For the Ca-imaging, NO appeared to have a large effect on cluster 1, a more modest effect on cluster 2 and less effect on cluster 3 (Fig. 5f). In comparison, the MEA results diverged, with NO producing the largest effect on cluster 2 and very modest if any effects on clusters 1 and 3 (Fig. 8c). Moreover, the temporal properties of cluster 1 and cluster 3 look very different between the Ca-imaging and MEA data. Without further comment, these differences raise concerns about the reliability of the clustering and the validity of comparisons made across the two sets of experiments. 

We agree that this is a critical point. Please see our reply #1.2 in our response to the public comments above.

Fig. 8: Transforming the PSTHs into Ca-traces is important to align the MEA recordings with the Ca-imaging data. It would also be very informative to see a more detailed overall presentation of the PSTH data since it provides a much higher temporal resolution of the responses. For example, illustrating the average PSTHs for the G32 cells under all the experimental conditions could be quite illuminating. 

To address this point, we added a new Supplementary Fig. S6, which shows the pseudo-Ca²⁺ traces for each cluster and condition next to the PSTHs. In addition, we quantified the cumulative firing rate for response features (time windows) where temporal suppression was observed in the Ca²⁺ data. This new analysis shows that during NO-application, we can see an increase in firing rate in all clusters. Nevertheless, the effect of NO on the PSTHs is admittedly small and it is better visible in the pseudo-Ca²⁺ transformed traces. One possible explanation for this difference may be that the overall firing rates are quite dynamic in G₃₂ such that a significant increase in “suppression” phases relative to the peak firing appears small.

Reviewer #2 (Public Review):  

Neuromodulators are important for circuit function, but their roles in the retinal circuitry are poorly understood. This study by Gonschorek and colleagues aims to determine the modulatory effect of nitric oxide on the response properties of retinal ganglion cells. The authors used two photon calcium imaging and multi-electrode arrays to classify and compare cell responses before and after applying a NO donor DETA-NO. The authors found that DETA-NO selectively increases activity in a subset of contrast-suppressed RGC types.

In addition, the authors found cell-type specific changes in light response in the absence of pharmacological manipulation in their calcium imaging paradigm. While this study focuses on an important question and the results are interesting, the following issues need further clarification for better interpretation of the data. 

We thank the reviewer for her/his detailed and constructive comments.

(1) Design of the calcium imaging experiments: the control-control pair has a different time course from the control-drug pair (Fig 1e). First, the control-control pair has a 10 minute interval while the control-drug pair has a 25 minute interval. Second, Control 1 Field 2 was imaged 10 min later than Control 1 Field 1 since the start of the calcium imaging paradigm. 

Given that the control dataset is used to control for time-dependent adaptational changes throughout the experiment, I wonder why the authors did not use the same absolute starting time of imaging and the same interval between the first and second round of imaging for both the control-control and the control-drug pairs. This can be readily done in one of the two ways: 1. In a set of experiment, add DETA/NO between "Control 1 Field 1 and "Control 2 Field 1" in Fig. 1e as the drug group; or 2. Omit DETA/NO in the Fig. 1e protocol as the control group to monitor the time course of adaptational changes. 

Thank you for raising this point. We hope that in the following we can clarify the reasoning behind our protocol and the analysis approach.

(2.1) Initially, we performed these experiments in different ways (also in the sequence suggested by the reviewer), before homing in on the paradigm illustrated in Fig. 1. We chose this paradigm for two reasons: First, we wanted to have for each retina both Ctrl1/Ctrl2 and Ctr1/NO data sets, to be sure that the time-dependent (adaptational) effects were not related to the general condition of an individual retina preparation. Second, we did not see obvious differences in time-dependent or NO-induced effects between paradigms. Therefore, while we cannot exclude that the absolute time between recordings can affect the observed changes, we do not think that such effects are substantial enough to change our conclusions.

In the revised manuscript, we now explicitly point at the different intervals. 

Related to the concern above, to determine NO-specific effect, the authors used the criterion that "the response changes observed for control (ΔR(Ctrl2−Ctrl1)) and NO (ΔR(NO−Ctrl1)) were significantly different". This criterion assumes that without DETA-NO, imaging data obtained at the time points of "Control 1 Field 2" and "DETA/NO Field 2" would give the same value of ΔR as ΔR(Ctrl2−Ctrl1) for all RGC types. It is not obvious to me why this should be the case, because of the unknown time-dependent trajectory of the adaptational change for each RGC type. For example, a RGC type could show stable response in the first 30 min and then change significantly in the following 30 min. DETA/NO may counteract this adaptational change, leading to the same ΔR as the control condition (false negative). Alternatively, DETA/NO may have no effect, but the nonlinear timedependent response drift can give false positive results. 

(2.2) Initially, we assumed that after adapting the retina to a certain light level, RGCs exhibit stable responses over time, such that when adding a pharmacological agent, we can identify drug-induced response changes (e.g., by calculating the response difference). To our surprise, we found that for some RGC types the responses changed between the first and the second recording (referred to as cell type-specific adaptational effects), which is why we devised the Ctrl1/Ctrl2 vs. Ctr2/NO analysis. 

The reviewer is correct in that we assume in our analysis that the adaptational- and NO-induced effects are independent and sum linearly. Further, we agree with the reviewer that there may be other possibilities, two of which are highlighted by the reviewer:

(a) Interaction: for instance, if NO compensates for the adaptational effect, we would not be able to measure this; or, if this compensation was partial, underestimate both effects. 

(b) More complex time-dependency: for example, if an RGC shows a pronounced adaptational effect with a longer delay (i.e. only after the second scan), or that a very transient NO effect has already disappeared when we perform the second scan. On the one hand, as we only can take snapshots of the RGC responses, we cannot exclude these possibilities. On the other hand, both effects (adaptational- and NO-dependent) were type-specific and reproducible between experiments (also with varying timing, see reply #2.1), which makes complex time dependencies less likely.

The revised manuscript now reflects these limitations of our recording paradigm and points out which effects can be detected, and which likely not.

I also wonder why washing-out, a standard protocol for pharmacological experiments, was not done for the calcium protocol since it was done in the MEA experiments. A reversible effect by washing in and out DETA/NO in the calcium protocol would provide a much stronger support that the observed NO modulation is due to NO and not to other adaptive changes. 

(2.3) We agree that a clear wash-out would strengthen our findings. Indeed, in the beginning of our experiments, we tried to wash-out the agent in the Ca²⁺ recordings, as we did in the MEA recordings. We soon stopped doing this in the Ca²⁺ experiments, because response quality decreased for the third scan of the same field, likely due to bleaching of fluorescent indicator and photopigment. This is why we typically restrict the total recording time of the same field of RGCs to about 30 min (~ two scans with all light stimuli). Moreover, our MEA data showed that DETA/NO can largely be washed-out, which supports that we observed NO-specific effects. Therefore, we decided against further attempts to establish the wash-out also in the Ca²⁺ experiments (e.g., shortening the recording time by presenting fewer light stimuli).

(2) Effects of Strychnine: In lines 215-219, " In the light-adapted retina, On-cone BCs boost light-Off responses in Off-cone BCs through cross-over inhibition (83, 84) and hence, strychnine affects Off-response components in RGCs - in line with our observations (Fig. S2)" However, Fig. S2 doesn't seem to show a difference in the Off-response components. Rather, the On response is enhanced with strychnine. In addition, suppressed-by-contrast cells are known to receive glycinergic inhibition from VGluT3 amacrine cells (Tien et al., 2016). However, the G32 cluster in Fig. S2 doesn't seem to show a change with strychnine. More explanation on these discrepancies will be helpful.

(2.4) We thank the reviewer for this comment. Regarding the first part, we agree that the figure does not support differences in the Off-response components. We therefore rephrased the corresponding text accordingly. Additionally, we now show all RGC types with n>3 cells per recording condition in the revised Suppl. Fig. S2 and added statistics.

Regarding the second part, there are several possible explanations for these discrepancies:

(a) The SbC (transient Off SbC) studied in Tien et al. (2016) likely corresponds to the RGC type G₂₈ (see Höfling et al. 2024). As mentioned above (see reply #1.2), it is unclear if G₃₂ corresponds to a previously described SbC, and if so, to which. Goetz et al. (2022) proposed that G₃₂ may align with the bursty-SbC (bSbC) type (their Supplemental Table 3), as described also by Wienbar and Schwartz (2022). An important feature of the bSbC type is that its contrast response function is mainly driven by intrinsic properties rather than synaptic input. If G₃₂ indeed included the bSbC, this may explain why strychnine does not interfere with the suppression of temporal contrast.

(b) In Tien et al. (2016), the authors genetically removed the VG3-ACs (see their Fig. 3) and show that this ablation reduces the inhibition of tSbC cells in a stimulus size-dependent manner. Specifically, larger light stimuli (600 µm) only show marginal effects on the IPSCs and inhibitory synaptic conductance (see their Figs. 3c,d and 3e,f, respectively). In our study, the full-field chirp had a size of 800 x 600 µm. Therefore – and assuming that G₃₂ indeed included tSbCs – our observation that strychnine did not affect temporal suppression in the full-field chirp responses would be in line with Tien et al. (2016).   

(3) This study uses DETA-NO as an NO donor for enhancing NO release. However, a previous study by Thompson et al., Br J Pharmacol. 2009 reported that DETA-NO can rapidly and reversible induce a cation current independent of NO release at the 100 uM used in the current study, which could potentially cause the observed effect in G32 cluster such as reduced contrast suppression and increased activity. This potential caveat should at least be discussed, and ideally excluded by showing the absence of DETA-NO effects in nNOS knockout mice, and/or by using another pharmacological reagent such as the NO donor SNAP or the nNOS inhibitor l-NAME. 

Thank you for pointing out this potential caveat. We certainly cannot exclude such side effects. However, we think that this explanation of our observations is unlikely, because Thompson et al. barely see effects at 100 µM DETA/NO; in fact, their data suggests that clear NO-independent effects on the cation-selective channel occur at much higher DETA/NO concentrations, such as 3 mM. 

In any case, in the revised manuscript, we refer to this paper in the Discussion. 

(4) Clarification of methods: In the Methods, lines 1119-1127, the authors describe the detrending, baseline subtraction, and averaging. Then, line 1129, " the mean activity r(t) was computed and then traces were normalized such that: max t(|r(t)|) = 1. How is the normalization done? Is it over the entire recording (control and wash in) for each ROI? Or is it normalized based on the mean trace under each imaging session (i.e. twice for each imaging field)? 

The normalization (z-scoring) was done for each ROI individually per stimulus and condition (Ctrl 1, Ctrl 2, DETA/NO). We normalized the traces, because the absolute Ca²⁺ signal depends on factors, such as “resting” state of the cell (e.g., silent vs. baseline spiking activity in the absence of a light stimulus) and its fluorescent dye concentration. This also means that absolute response amplitudes are difficult to interpret. Hence, we focused on analyzing relative changes per ROI and condition, which still allowed us to investigate adaptational and drug-induced effects. In the revised manuscript, we changed the corresponding paragraph for clarification.

As for the clustering of RGC types, I assume that each ROI's cluster identity remains unchanged through the comparison. If so, it may be helpful to emphasize this in the text.

Yes, this is correct. We identified G₃₂ RGCs based on their Ctrl1 responses and then compares these responses with those for Ctrl2 or NO. We now clarified this in the revised manuscript.

Reviewer #2 (Recommendations For The Authors):  

The manuscript would benefit from a discussion of how the findings in this study relate to known mechanisms of NO modulation and previously reported effects of NO manipulations on RGC activity. 

Thank you for the recommendation. We already refer to known mechanisms of NO within the retina in the Introduction. In the revised manuscript, we now added information to the Discussion.

In the abstract, "a paired-recording paradigm" could be misleading because paired recording generally refers to the simultaneous recording of two neurons. However, the paradigm in this study is essentially imaging experiments done at two time points. 

We agree with the reviewer. To avoid any confusion with paired electrophysiological recordings, we changed the term “paired-recording paradigm” to “sequential recording paradigm” and replaced the term “pair-/ed” with “sequentially recorded”.

eLife Assessment

This important study is the first comprehensive analysis of the modulatory effects of nitric oxide (NO) on the response properties of retinal ganglion cells (RGCs) in the mouse retina using two-photon calcium imaging and multi-electrode arrays (MEA). The results provide compelling evidence that a subset of suppressed-by-contrast RGCs are affected. These unexpected findings are likely of broad interest to visual neuroscientists.

Reviewer #1 (Public review):

Summary:

Nitric oxide (NO) has been implicated as a neuromodulator in the retina. Specific types of amacrine cells (ACs) produce and release NO in a light-dependent manner. NO diffuses freely through the retina and can modulate intracellular levels of cGMP, or directly modify and modulate proteins via S-nitrosylation, leading to changes in gap-junction coupling, synaptic gain, and adaptation. Although these system-wide effects have been documented, it is not well understood how the physiological function of specific neuronal types is affected by NO. This study aims to address this gap in our knowledge.

There are two major findings. 1) About a third of the retinal ganglion cells display cell-type specific adaptation to prolonged stimulus protocols. 2) Application of NO specifically affected Off-suppressed ganglion cells designated as G32 cells. The G32 cluster likely contains 3 ganglion cell types that are differentially affected.

This is the first comprehensive analysis of the functional effects of NO on ganglion cells in the retina. The cell-type specificity of the effects is surprising and provides the field with valuable new information.

Strengths:

NO was expected to produce small effects, and considerable effort was expended in validating the system to ensure that changes in the state of the preparation would not confound any effects of NO. The authors used a sequential stimulus protocol to control for changes in the sensitivity of the retina during the extended recording periods. The approach potentially increases the sensitivity of the measurements and allows more subtle effects to be observed.

Neural activity was measured by Ca-imaging. Responsive ganglion cells were grouped into 32 types using a clustering analysis. Initial control experiments demonstrated that the cell-types revealed by the analysis largely recapitulate those from their earlier landmark study using a similar approach.

Application of NO to the retina modulated responses of a single cluster of cells, labeled G32, while having little effect on the remaining 31 clusters. In separate experiments, ganglion cell spiking activity was recorded on a multi-electrode array (MEA). Together the Ca-imaging and MEA recordings provide complementary approaches and demonstrate that NO modulates the temporal but not spatial properties of affected cell-types.

Weaknesses:

The concentration of NO used in these experiments was ~0.25µM, which is 5- to 10-fold lower than the endogenous concentration previously measured in rodent retina. It is perhaps surprising that this relatively low NO concentration produced significant effects. However, the endogenous measurements were done in an eye-cup preparation, while the current experiments were performed in a bare (no choroid) preparation. Perhaps the resting NO level is lower in this preparation. It is also possible that the low concentration of NO promoted more selective effects.

Reviewer #2 (Public review):

Neuromodulators are important for circuit function, but their roles in the retinal circuitry are poorly understood. This study by Gonschorek and colleagues aims to determine the modulatory effect of nitric oxide on the response properties of retinal ganglion cells. The authors used two photon calcium imaging and multi-electrode arrays to classify and compare cell responses before and after applying a NO donor DETA-NO. The authors found that DETA-NO selectively increases activity in a subset of contrast-suppressed RGC types. In addition, the authors found cell-type specific changes in light response in the absence of pharmacological manipulation in their calcium imaging paradigm. This study focuses on an important question and the results are interesting. The limitations of the method and data interpretation are adequately discussed in the revised manuscript.

The authors have addressed my previous comments, included additional discussions on the limitations of the method, and provided a more careful interpretation of their data.

Author response:

Reviewer 1:

Summary:

This paper describes molecular dynamics simulations (MDS) of the dynamics of two T-cell receptors (TCRs) bound to the same major histocompatibility complex molecule loaded with the same peptide (pMHC). The two TCRs (A6 and B7) bind to the pMHC with similar affinity and kinetics, but employ different residue contacts. The main purpose of the study is to quantify via MDS the differences in the inter- and intra-molecular motions of these complexes, with a specific focus on what the authors describe as catch-bond behavior between the TCRs and pMHC, which could explain how T-cells can discriminate between different peptides in the presence of weak separating force.

Strengths:

The authors present extensive simulation data that indicates that, in both complexes, the number of high-occupancy interdomain contacts initially increases with applied load, which is generally consistent with the authors’ conclusion that both complexes exhibit catch-bond behavior, although to different extents. In this way, the paper somewhat expands our understanding of peptide discrimination by T-cells.

The reviewer makes thoughtful assessments of our manuscript. While our manuscript is meant to be a “short” contribution, our significant new finding is that even for TCRs targeting the same pMHC, having similar structures, and leading to similar functional outcomes in conventional assays, their response to applied load can be different. This supports out recent experimental work where TCRs targeting the same pMHC differed in their catch bond characteristics, and importantly, in their response to limiting copy numbers of pMHCs on the antigen-presenting cell (Akitsu et al., Sci. Adv., 2024; cited in our manuscript). Our present manuscript provides the physical basis where two similar TCRs respond to applied load differently. In the revised manuscript, we will make this point clearer.

Weaknesses:

While generally well supported by data, the conclusions would nevertheless benefit from a more concise presentation of information in the figures, as well as from suggesting experimentally testable predictions.

Following the reviewers’ suggestions, we will update figures and use Figure Supplements to make the main figures more concise and to simplify the overall presentation.

Regarding testable predictions, one prediction would be that B7 TCR will exhibit weaker catch bond behavior than A6. This is an important prediction because the two TCRs targeting the same pMHC have similar structures and are functionally similar in conventional assays. This prediction can be tested by single-molecule optical tweezers experiments. We also predict the A6 TCR may perform better when the number of pMHC molecules presented are limited, analogous to our recent experiments on different TCRs, Akitsu et al., Sci. Adv. (2024).

Another testable prediction for the conservation of the basic allostery mechanism is to test the Cβ FG-loop deletion mutant located at the hinge region of the β chain, yet its deletion severely impairs the catch bond formation. These predictions will be mentioned and discussed in the updated manuscript.

Reviewer 2:

In this work, Chang-Gonzalez and coworkers follow up on an earlier study on the force-dependence of peptide recognition by a T-cell receptor using all-atom molecular dynamics simulations. In this study, they compare the results of pulling on a TCR-pMHC complex between two different TCRs with the same peptide. A goal of the paper is to determine whether the newly studied B7 TCR has the same load-dependent behavior mechanism shown in the earlier study for A6 TCR. The primary result is that while the unloaded interaction strength is similar, A6 exhibits more force stabilization.

This is a detailed study, and establishing the difference between these two systems with and without applied force may establish them as a good reference setup for others who want to study mechanobiological processes if the data were made available, and could give additional molecular details for T-Cell-specialists. As written, the paper contains an overwhelming amount of details and it is difficult (for me) to ascertain which parts to focus on and which results point to the overall take-away messages they wish to convey.

As mentioned above and as the reviewer correctly pointed out, the condensed appearance of this manuscript arose largely because we intended it to be a Research Advances article as a short follow up study of our previous paper on A6 TCR published in eLife. Most of the analysis scripts for the A6 TCR study are already available on Github. We will additionally deposit sample structures and simulation scripts for the B7 TCR. Trajectory will be provided upon request given their large size.

Regarding the focus issue, it is in part due to the complex nature of the problem, which required simulations under different conditions and multi-faceted analyses. Concisely presenting the complex analyses also has been a challenge in our previous papers on TCR simulations (Hwang et al., PNAS 2020; Chang-Gonzalez et al., eLife, 2024 – both are cited in our manuscript). With updated figures and texts, we expect that the presentation will be a lot clearer. But even in the present form, the reviewer points out the main take-away message well: “The primary result is that while the unloaded interaction strength is similar, A6 exhibits more force stabilization.

Detailed comments:

(1) In Table 1 - are the values of the extension column the deviation from the average length at zero force (that is what I would term extension) or is it the distance between anchor points (which is what I would assume based on the large values. If the latter, I suggest changing the heading, and then also reporting the average extension with an asterisk indicating no extensional restraints were applied for B7-0, or just listing 0 load in the load column. Standard deviation in this value can also be reported. If it is an extension as I would define it, then I think B7-0 should indicate extension = 0+/- something.

The distance between anchor points could also be labeled in Figure 1A.

“Extension” is the distance between anchor points (blue spheres at the ends of the added strands in Fig. 1A). While its meaning should be clear in the section “Laddered extensions” in MD simulation protocol, at first glance it may lead to confusion. In a strict sense, use of “extension” for the distance is a misnomer, but we have used it in our previous two papers (Hwang et al., PNAS 2020; Chang-Gonzalez et al., eLife, 2024), so we prefer to keep it for consistency. Instead, in the caption of Table 1, we will explain its meaning, and also explicitly label it in Fig. 1A, as the reviewer suggested.

Please also note that the no-load case B7⁰ does not have a particular extension that yields zero load on average. It would in fact be very difficult to find such an extension (distance between two anchor points). To simulate the system without load, we separately built a TCR-pMHC complex without added linkers, and held the distal part of pMHC with weak harmonic restraints (explained in sections “Structure preparation” and “Systems without load”). In this way, no external force is applied to TCR as it moves relative to pMHC. We will clarify this when introducing B7⁰ in the Results section.

(2) As in the previous paper, the authors apply ”constant force” by scanning to find a particular bond distance at which a desired force is selected, rather than simply applying a constant force. I find this approach less desirable unless there is experimental evidence suggesting the pMHC and TCR were forced to be a particular distance apart when forces are applied. It is relatively trivial to apply constant forces, so in general, I would suggest this would have been a reasonable comparison. Line 243-245 speculates that there is a difference in catch bonding behavior that could be inferred because lower force occurs at larger extensions, but I do not believe this hypothesis can be fully justified and could be due to other differences in the complex.

There is indeed experimental evidence that the TCR-pMHC complex operates under constant separation. The spacing between a T-cell and an antigen-presenting cell is maintained by adhesion molecules such as the CD2CD58 pair, as explained in our paper on the A6 TCR, (Chang-Gonzalez et al., eLife, 2024; please see the bottom paragraph on page 4 of the paper). In in vitro single-molecule experiments, pulling to a fixed separation and holding is also commonly done. Detailed comparison between constant extension vs. constant force simulations is definitely a subject of our future study. We will clarify these points when explaining about the constant extension (or separation).

Regarding line 243–245, we agree with the reviewer that without further tests, lower forces at larger extensions per se cannot be an indicator that B7 forms a weaker catch bond. But with additional insight, it does have an indirect relevance. In addition to fewer TCR-pMHC contacts (Fig. 1C of our manuscript), the intra-TCR contacts are also reduced compared to those of A6 (Fig. 1D vs. Chang-Gonzalez et al., eLife, 2024, Fig. 8A,B, first column; reproduced in the figure in our response to reviewer 3 below). This shows that the B7 TCR forms a looser complex with pMHC compared to A6. With its higher compliance, the B7 TCR-pMHC complex needs to be under a greater extension than A6 to apply comparable levels of force, and it would be more difficult to achieve load-induced stabilization of the TCR-pMHC interface, hence a weaker catch bond. We will add this point when explaining the weaker catch bond behavior of B7.

(3) On a related note, the authors do not refer to or consider other works using MD to study force-stabilized interactions (e.g. for catch bonding systems), e.g. these cases where constant force is applied and enhanced sampling techniques are used to assess the impact of that applied force: https://www.cell.com/biophysj/fulltext/S0006-3495(23)00341-7, https://www.biorxiv.org/content/10.1101/2024.10.10.617580v1. I was also surprised not to see this paper on catch bonding in pMHC-TCR referred to, which also includes some MD simulations: https://www.nature.com/articles/s41467-023-38267-1

We thank the reviewer for bringing the three papers to our attention, which are:

(1) Languin-Cattoën, Sterpone, and Stirnemann, Biophys. J. 122:2744 (2023): About bacterial adhesion protein FimH.

(2) Peña Ccoa, et al., bioRxiv (2024): About actin binding protein vinculin.

(3) Choi et al., Nat. Comm. 14:2616 (2023): About a mathematical model of the TCR catch bond.

Catch bond mechanisms of FimH and vinculin are different from that of TCR in that FimH and vinculin have relatively well-defined weak- and strong-binding states where there are corresponding crystal structures. Availability of the end-state structures enable using simulation approaches such as enhanced sampling of individual states and studying the transition between the two states. In contrast, TCR does not have any structurally well-defined weakor strong-binding states, which requires a different approach. As demonstrated in our current manuscript as well as in our previous two papers (Hwang et al., PNAS 2020; Chang-Gonzalez et al., eLife, 2024), our microsecond-long simulations of the complex under realistic pN-level loads and a combination of analysis methods are effective for elucidating the catch bond mechanism of TCR. In the revised manuscript, we will cite the two papers, to compare the TCR catch bond mechanism with those of FimH and vinculin, which will offer a broader perspective.

The third paper (Choi, 2023) proposes a mathematical model to analyze extensive sets of data, and also perform new experiments and additional simulations. Of note, their model assumptions are based mainly on the steered MD (SMD) simulation in their previous paper (Wu, et al., Mol. Cell. 73:1015, 2019). In their model, formation of a catch bond (called catch-slip bond in Choi’s paper) requires partial unfolding of MHC and tilting of the TCR-pMHC interface. While further studies are needed to find whether those changes are indeed required, even so, the question remains regarding how the complex in the fully folded state can bear load and enter such a state in the first place. Our current and previous simulation studies suggest a mechanism by which ligand- and load-dependent responses occur as the first obligatory step of catch bond formation, after which partial unfolding and/or extensive conformational transitions may occur, as described in our recent paper (Akitsu et al., Sci. Adv., 2024). In the revised manuscript, we will cite Wu’s paper and briefly explain the above.

(4) The authors should make at least the input files for their system available in a public place (github, zenodo) so that the systems are a more useful reference system as mentioned above. The authors do not have a data availability statement, which I believe is required.

As mentioned above, we will make sample input files and coordinates available on Github. Data availability statement will be added.

Reviewer 3:

Summary:

The paper by Chang-Gonzalez et al. is a molecular dynamics (MD) simulation study of the dynamic recognition (load-induced catch bond) by the T cell receptor (TCR) of the complex of peptide antigen (p) and the major histocompatibility complex (pMHC) protein. The methods and simulation protocols are essentially identical to those employed in a previous study by the same group (Chang-Gonzalez et al., eLife 2024). In the current manuscript, the authors compare the binding of the same pMHC to two different TCRs, B7 and A6 which was investigated in the previous paper. While the binding is more stable for both TCRs under load (of about 10-15 pN) than in the absence of load, the main difference is that, with the current MD sampling, B7 shows a smaller amount of stable contacts with the pMHC than A6.

Strengths:

The topic is interesting because of the (potential) relevance of mechanosensing in biological processes including cellular immunology.

Weaknesses:

The study is incomplete because the claims are based on a single 1000-ns simulation at each value of the load and thus some of the results might be marred by insufficient sampling, i.e., statistical error. After the first 600 ns, the higher load of B7high than B7low is due mainly to the simulation segment from about 900 ns to 1000 ns (Figure 1D). Thus, the difference in the average value of the load is within their standard deviation (9 +/- 4 pN for B7low and 14.5 +/- 7.2 for B7high, Table 1). Even more strikingly, Figure 3E shows a lack of convergence in the time series of the distance between the V-module and pMHC, particularly for B70 (left panel, yellow) and B7low (right panel, orange). More and longer simulations are required to obtain a statistically relevant sampling of the relative position and orientation of the V-module and pMHC.

The reviewer uses data points during the last 100 ns to raise an issue with sampling. But since we are using realistic pN range forces, force fluctuates more slowly. In fact, in our simulation of B7^high, while the force peaks near 35 pN at 500 ns (Fig. 1D of our manuscript; reproduced as panels C and D below), the contact heat map shows no noticeable changes around 500 ns (Fig. 2C of our manuscript). Thus, a wider time window must be considered rather than focusing on instantaneous force.

We believe the reviewer’s concern about sampling arose also due to a lack of clear explanation. Author response image 1 below contains panels from our earlier eLife paper on the A6 TCR. Panels A and B are from Fig. 8 of the A6 paper, and panels C and D are from Fig. 1D of our present manuscript. The high-load simulations in both cases (outlined circles) fluctuate widely in force so that one might argue that sampling was insufficient. However, unless one is interested in finding the precise value of force for a given extension, sampling in our simulations was reasonable enough to distinguish between high- and low-force behaviors. To support this, we show panel E below, which is from Appendix 3–Fig. 1 of our A6 paper. Added to this panel are the average forces and standard deviations of B7^low and B7^high from Table 1 of our manuscript (red squares). Please note that all of the data were measured after 500 ns. Except for Y8A^low and dFG^low of A6 (explained below), all of the data points lie on nearly a straight line.

Author response image 1.

Thermodynamically, the force and position of the restraint (blue spheres in Fig. 1A of our manuscript) form a pair of generalized force and the corresponding spatial variable in equilibrium at temperature 300 K, which is akin to the pressure P and volume V of an ideal gas. If V is fixed, P fluctuates. Denoting the average and std of pressure as ⟨P⟩ and ∆P, respectively, Burgess showed that ∆P/⟨P⟩ is a constant (Eq. 5 of Burgess, Phys. Lett. A, 44:37; 1973). In the case of the TCRαβ-pMHC system, although individual atoms are not ideal gases, since their motion leads to the fluctuation in force on the restraints, the situation is analogous to the case where pressure arises from individual ideal gas molecules hitting the confining wall as the restraint. Thus, the near-linear behavior in panel E above is a consequence of the system being many-bodied and at constant temperature. The linearity is also an indirect indicator that sampling of force was reasonable. The fact that A6 and B7 data show a common linear profile further demonstrates the consistency in our force measurement. That said, the B7 data points (red in panel E) are elevated slightly above nearby A6 data points. This is consistent with B7 forming an overall weaker complex, both at the TCR-pMHC interface (panels A vs. C) and within intra-TCR interfaces (panels B vs. D), which can be seen by the wider ranges of color bars in panels A and B for A6 compared to panels C and D for B7.

About the two outliers of A6, Y8A^low is for an antagonist peptide and dFG^low is the Cβ FG-loop deletion mutant. Interestingly, both cases had reduced numbers of contacts with pMHC, which likely caused a wider conformational motion, hence greater fluctuation in force.

A similar argument applies to Fig. 3E of our manuscript. If precise values of the V-module to pMHC distance were needed, longer or duplicate simulations would be necessary, however, Fig. 3E as it currently stands clearly shows that B7^high maintains more stable interface compared to B7^low, which is consistent with all other measures we used, such as Fig. 3B (Hamming distance), Fig. 3C (buried surface area), and Fig. 4A–E (Vα-Vβ motion and CDR3 distance). They are also consistent with our simulations of A6.

Thus, rather than relying on peculiarities of individual trajectories, we analyze data in multiple ways and draw conclusions based on features that are consistent across different simulations. Please also note that reviewer 1 mentioned that our conclusions are “generally well supported by data.”

We will update our manuscript to concisely explain the above and also will add Panel E above as a supplement of Fig. 1.

It is not clear why ”a 10 A distance restraint between alphaT218 and betaA259 was applied” (section MD simulation protocol, page 9).

αT218 and βA259 are the residues attached to a leucine-zipper handle in in vitro optical trap experiments (Das, et al., PNAS 2015). In T cells, those residues also connect to transmembrane helices. Author response image 2 is a model of N15 TCR used in experiments in Das’ paper, constructed based on PDB 1NFD. Blue spheres represent Cα atoms corresponding to αT218 and βA259 of B7 TCR. Their distance is 6.7 ˚A. The 10-˚A distance restraint in simulation was applied to mimic the presence of the leucine zipper that prevents excessive separation of the added strands. The distance restraint is a flat-bottom harmonic potential which is activated only when the distance between the two atoms exceeds 10 ˚A, which we did not clarify in our original manuscript. The same restraint was used in our previous studies on JM22 and A6 TCRs.

We will add the figure as a supplement of Fig. 1, cite Das’ paper, and also update description of the distance restraint in the MD simulation protocol section.

Author response image 2.

eLife Assessment

This useful study reports detailed molecular dynamics simulations of T-cell receptors in complex with a peptide/MHC complex, for a better understanding of the mechanism of T-cell activation. The key observation was that tensile force applied in the direction of separation between TCR/pMHC appears to strengthen the interface, which is consistent with the catch bond scenario, although the effect is less apparent than that studied in the earlier work despite many similarities. The analyses are systematic and thus generally solid, although the level of evidence could be considered incomplete due to limited sampling based on a single trajectory for each load.

Reviewer #1 (Public review):

Summary:

This paper describes molecular dynamics simulations (MDS) of the dynamics of two T-cell receptors (TCRs) bound to the same major histocompatibility complex molecule loaded with the same peptide (pMHC). The two TCRs (A6 and B7) bind to the pMHC with similar affinity and kinetics, but employ different residue contacts. The main purpose of the study is to quantify via MDS the differences in the inter- and intra-molecular motions of these complexes, with a specific focus on what the authors describe as catch-bond behavior between the TCRs and pMHC, which could explain how T-cells can discriminate between different peptides in the presence of weak separating force.

Strengths:

The authors present extensive simulation data that indicates that, in both complexes, the number of high-occupancy inter-domain contacts initially increases with applied load, which is generally consistent with the authors' conclusion that both complexes exhibit catch-bond behavior, although to different extents. In this way, the paper somewhat expands our understanding of peptide discrimination by T-cells.

Weaknesses:

While generally well supported by data, the conclusions would nevertheless benefit from a more concise presentation of information in the figures, as well as from suggesting experimentally testable predictions.

Reviewer #2 (Public review):

In this work, Chang-Gonzalez and coworkers follow up on an earlier study on the force-dependence of peptide recognition by a T-cell receptor using all-atom molecular dynamics simulations. In this study, they compare the results of pulling on a TCR-pMHC complex between two different TCRs with the same peptide. A goal of the paper is to determine whether the newly studied B7 TCR has the same load-dependent behavior mechanism shown in the earlier study for A6 TCR. The primary result is that while the unloaded interaction strength is similar, A6 exhibits more force stabilization.

This is a detailed study, and establishing the difference between these two systems with and without applied force may establish them as a good reference setup for others who want to study mechanobiological processes if the data were made available, and could give additional molecular details for T-Cell-specialists. As written, the paper contains an overwhelming amount of details and it is difficult (for me) to ascertain which parts to focus on and which results point to the overall take-away messages they wish to convey.

Detailed comments:

(1) In Table 1 - are the values of the extension column the deviation from the average length at zero force (that is what I would term extension) or is it the distance between anchor points (which is what I would assume based on the large values. If the latter, I suggest changing the heading, and then also reporting the average extension with an asterisk indicating no extensional restraints were applied for B7-0, or just listing 0 load in the load column. Standard deviation in this value can also be reported. If it is an extension as I would define it, then I think B7-0 should indicate extension = 0+/- something. The distance between anchor points could also be labeled in Figure 1A.

(2) As in the previous paper, the authors apply "constant force" by scanning to find a particular bond distance at which a desired force is selected, rather than simply applying a constant force. I find this approach less desirable unless there is experimental evidence suggesting the pMHC and TCR were forced to be a particular distance apart when forces are applied. It is relatively trivial to apply constant forces, so in general, I would suggest this would have been a reasonable comparison. Line 243-245 speculates that there is a difference in catch bonding behavior that could be inferred because lower force occurs at larger extensions, but I do not believe this hypothesis can be fully justified and could be due to other differences in the complex.

(3) On a related note, the authors do not refer to or consider other works using MD to study force-stabilized interactions (e.g. for catch bonding systems), e.g. these cases where constant force is applied and enhanced sampling techniques are used to assess the impact of that applied force: https://www.cell.com/biophysj/fulltext/S0006-3495(23)00341-7, https://www.biorxiv.org/content/10.1101/2024.10.10.617580v1. I was also surprised not to see this paper on catch bonding in pMHC-TCR referred to, which also includes some MD simulations: https://www.nature.com/articles/s41467-023-38267-1

(4) The authors should make at least the input files for their system available in a public place (github, zenodo) so that the systems are a more useful reference system as mentioned above. The authors do not have a data availability statement, which I believe is required.

Reviewer #3 (Public review):

Summary:

The paper by Chang-Gonzalez et al. is a molecular dynamics (MD) simulation study of the dynamic recognition (load-induced catch bond) by the T cell receptor (TCR) of the complex of peptide antigen (p) and the major histocompatibility complex (pMHC) protein. The methods and simulation protocols are essentially identical to those employed in a previous study by the same group (Chang-Gonzalez et al., eLife 2024). In the current manuscript, the authors compare the binding of the same pMHC to two different TCRs, B7 and A6 which was investigated in the previous paper. While the binding is more stable for both TCRs under load (of about 10-15 pN) than in the absence of load, the main difference is that, with the current MD sampling, B7 shows a smaller amount of stable contacts with the pMHC than A6.

Strengths:

The topic is interesting because of the (potential) relevance of mechanosensing in biological processes including cellular immunology.

Weaknesses:

The study is incomplete because the claims are based on a single 1000-ns simulation at each value of the load and thus some of the results might be marred by insufficient sampling, i.e., statistical error. After the first 600 ns, the higher load of B7high than B7low is due mainly to the simulation segment from about 900 ns to 1000 ns (Figure 1D). Thus, the difference in the average value of the load is within their standard deviation (9 +/- 4 pN for B7low and 14.5 +/- 7.2 for B7high, Table 1). Even more strikingly, Figure 3E shows a lack of convergence in the time series of the distance between the V-module and pMHC, particularly for B70 (left panel, yellow) and B7low (right panel, orange). More and longer simulations are required to obtain a statistically relevant sampling of the relative position and orientation of the V-module and pMHC.

It is not clear why "a 10 A distance restraint between alphaT218 and betaA259 was applied" (section MD simulation protocol, page 9).

Author response:

Reviewer #1 (Public review):

The significance of the target molecule and mechanisms may help in understanding the molecular mechanisms of metformin.

We greatly appreciate the reviewer’s insightful comment regarding the significance of the target molecule and its mechanisms in understanding the molecular actions of metformin. ATP5I is responsible for the dimerization of the F₁F₀-ATPase(1-3). Hence, we propose conducting BN-PAGE followed by a western blot using the β-subunit of the F1 domain of F1F0-ATP synthase to investigate whether metformin affects its dimerization. This will provide a more direct evidence of the on target action of metformin on ATP5I. Due to the high abundance of F₁F₀-ATP synthase in cells and the slow ability of metformin to enter mitochondria, we plan to perform long-term treatments (3 and 6 days) with high concentrations of metformin (10 mM) to enhance the likelihood of detecting subtle yet biologically relevant shifts in the monomer and dimer populations. Prolonged exposure is expected to reveal the cumulative effects of metformin on F₁F₀-ATP synthase dimers/monomers ratio. We do not expect that metformin will totally mimic the cumulative effect of the dimerization as in ATP5I KO cells but we think it will be important to report to what extent this ratio is affected.

Reviewer #2 (Public review):

(1) The interpretation of the cellular co-localization of the biotin-biguanide conjugate with TOMM20 (Figure 1-D) as mitochondrial "accumulation" of the conjugate is overstated because it cannot exclude binding of the conjugate to the mitochondrial membrane. It would have been more convincing if additional incubations with the biotin-biguanide conjugate in combination with metformin had shown that metformin is competitive with the biotin-conjugate.

We appreciate the reviewer’s insightful comment and agree that the resolution provided by fluorescence microscopy makes it challenging to pinpoint the specific mitochondrial compartment where the biotin-biguanide conjugate localizes, even with additional markers such as TOMM20 antibodies for the inner mitochondrial membrane. While it remains a possibility that the conjugate binds to the mitochondrial surface, another plausible explanation is that the biotin moiety may facilitate entry into mitochondria through a biotin-specific transporter, adding further mechanistic intricacies. Furthermore, while a competition assay with metformin might help investigate interactions with mitochondrial targets and transporters (OCT family), it would not compete for biotin-mediated transport. Thus, while we acknowledge the reviewer’s suggestion, we believe such an experiment may not provide conclusive evidence regarding the conjugate’s mitochondrial localization or mechanism of entry. Instead, we will revise the manuscript to more accurately describe the findings as "mitochondrial association" rather than "mitochondrial accumulation," ensuring that our interpretation remains consistent with the resolution and limitations of the data presented.

(2) The manuscript reports the identification of 69 proteins by mass spectrometry of the pull-down assay of which 31 proteins were eluted by metformin. However, no Mass Spectrometry data is presented of the peptides identified. The methodology does not state the minimum number of peptides (1, 2?) that were used for the identification of the 31/69 proteins.

Concerning the mass spectrometry results, our intention was to provide a comprehensive table summarizing these findings in a separate data sheet, as part of the data availability section. To address the reviewer’s comment and ensure full transparency, we will include this table as supplementary material in the revised manuscript. Additionally, we will update the methodology section to explicitly state these criteria and ensure clarity regarding the identification process.

(3) The validation of ATP5I was based on the use of recombinant protein (which was 90% pure) for the SPR and the use of a single antibody to ATP5I. The validity of the immunoblotting rests on the assumption that there is no "non-specific" immunoactivity in the relevant mol wt range. Information on the validation of the antibody would be helpful.

Regarding the recombinant protein used for SPR, its purity was evaluated using a Coomassie-stained gel. For the antibody used in immunoblotting, its specificity was validated through knockout cell lines, ensuring minimal concerns about non-specific immunoactivity within the relevant molecular weight range. Unfortunately, the KO data comes in the paper after the first immunoblots are presented. In the revised manuscript, we will clearly outline these validation steps in the methods section and additional manufacturer documentation for the antibody we used.

(4) Knock-out of ATP5I markedly compromised the NAD/NADH ratio (Fig.3A) and cell proliferation (Figure 3D). These effects may be associated with decreased mitochondrial membrane potential which could explain the low efficacy of metformin (and most of the data in Figures 3-5). This possibility should be discussed. Effects of [metformin] on the NAD/NADH ratio in control cells and ATP5I-KO would have been helpful because the metformin data on cell growth is normalized as fold change relative to control, whereas the NAD/NADH ratio would represent a direct absolute measurement enabling comparison of the absolute effect in control cells with ATP5I KO.

The mitochondrial membrane potential depends on a functional electron transport chain which drives proton pumping from the matrix to the intermembrane space. Metformin can decrease the mitochondrial membrane potential and this usually explained as a consequence of complex I inhibition(4). It has been published the metformin requires this membrane potential to accumulate in mitochondria so the actions of metformin are self-limiting due to this requirement. The reviewer is right that ATP5I KO cells could be resistant to metformin because they may have a lower membrane potential. We do not believe this to be the case because the response to phenformin, another biguanide that can enter mitochondria through the membrane without the need of the OCT transporters(5), is also affected in ATP5IKO cells. Of note, compensatory mechanisms such as enhanced glycolysis, as observed in ATP5I-KO cells (elevated ECAR and increased sensitivity to 2-D-deoxyglucose), and the ATPase activity of F₁F₀-ATP synthase could potentially help maintain membrane potential suggesting that this might not be an issue in the ATP5I KO cells. We will discuss these possibilities in the revised manuscript.

Nevertheless, to experimentally address this point, we propose measuring mitochondrial membrane potential using tetramethylrhodamine methyl ester (TMRE) and ATP levels using luciferase-based assays (CellTiter-Glo) in ATP5I-KO cells.

Regarding the NAD+/NADH in both control and KO cells may not be very helpful because this ratio can be corrected by LDH which is induced as part of the glycolytic adaptation that occurs after inhibition of respiration. Since our KO cells have been propagated already for several passages, the extent of this adaptation is likely different from metformin-treated cells. As we mentioned in answering Reviewer 1, we will provide a more direct measurement of metformin acting on ATP5I: the levels of F1F0-ATPase dimers and monomers.

(5) Figure-6 CRISPR/Cas9 KO at 16mM metformin in comparison with 70nM rotenone and 2 micromolar oligomycin (in serum-containing medium). The rationale for the use of such a high concentration of metformin has not been explained. In liver cells metformin concentrations above 1mM cause severe ATP depletion, whereas therapeutic (micromolar) concentrations have minimal effects on cellular ATP status. The 16mM concentration is ~2 orders of magnitude higher than therapeutic concentrations and likely linked to compromised energy status. The stronger inhibition of cell proliferation by 16mM metformin compared with rotenone or oligomycin raises the issue of whether the changes in gene expression may be linked to the greater inhibition of mitochondrial metabolism. Validation of the cellular ATP status and NAD/NADH with metformin as compared with the two inhibitors could help the interpretation of this data.

To address the reviewer’s final comment, we would like to clarify the rationale behind our experimental approach. NALM-6 cells are very glycolytic, have low respiration rates, and weak dependence on ATP5I (DepMap score: -0.47)(6). The concentration of 16 mM metformin was chosen based on the IC50 for this cell line. This approach aligns with our focus on the anticancer mechanism of action rather than the antidiabetic effects of metformin. Both ATP status and NAD+/NADH ratios will depend on the extent of the compensatory glycolysis. On the other hand, our genetic screening evaluates cell proliferation as an integration of all metabolic activities required for the process. This unbiased screening revealed a common pathway affected by metformin and oligomycin different that the pathway affected by rotenone, which is consistent with the finding that metformin acts of the F₁F₀ATPase.

Reviewer #3 (Public review):

(1) Most of the data are based on measurements of the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) measured by the Seahorse analyser in control and ATP5l KO cells. However, these measurements are conducted by a single injection of a biguanide, followed over time and presented as fold change. By doing so, the individual information on the effect of metformin and derivate on control and KO cells are lost. In addition, the usual measurement of OCR is coupled with certain inhibitors and uncouplers, such as oligomycin, FCCP, and Antimycin A/rotenone, to understand the contribution of individual complexes to respiration. Since biguanides and ATP5l KO affect protein levels of components of complex I and IV, it would be informative to measure their individual contributions/effects in the Seahorse. To further strengthen the data, it would be helpful to obtain measurements of actual ATP levels in these cells, as this would explain the activation of AMPK.

We appreciate the reviewer’s observations regarding the Seahorse measurements and acknowledge the potential limitations of presenting the data as fold change. Due to experimental challenges in maintaining KP-4 and ATP5I-KO cells with sufficient nutrients, caused by their rapid glucose uptake and subsequent lactate production, it was more practical to present the Seahorse results in this format. Using inhibitors at each time point during the Seahorse experiment was not feasible, as the delay between inhibitor injections and the corresponding changes in oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) would introduce variability and complicate the interpretation of dynamic responses. Nevertheless, we recognize the importance of understanding the contributions of specific respiratory complexes to OCR and ECAR. To address this, we will include a representative figure showcasing a typical Seahorse analysis, highlighting ATP turnover and proton leak after oligomycin addition, maximal respiration with FCCP, and disruption with rotenone and antimycin A. While these experiments are inherently complex due to the metabolic demands of ATP5I-KO cells, this approach will provide a clearer breakdown of mitochondrial activity. Furthermore, as mentioned in our response to Reviewer 2, we will measure ATP levels using a luciferase-based assay (CellTiter-Glo) in both control and ATP5I-KO cells to better explain AMPK activation. This will provide additional context to strengthen the interpretation of mitochondrial function and metabolic compensation mechanisms in these cells.

(2) The authors report on alterations in mitochondrial morphology upon ATP5l KO, which is measured by subjective quantifications of filamentous versus puncta structures. Fiji offers great tools to quantify the mitochondrial network unbiasedly and with more accuracy using deconvolution and skeletonization of the mitochondria, providing the opportunity to measure length, shape, and number quantitatively. This will help to understand better, whether mitochondria are really fragmented upon ATP5l KO and rescued by its re-introduction.

Concerning the analysis of mitochondrial morphology, we acknowledge the potential benefits of using Fiji and additional plugins such as MiNA for more accurate and unbiased quantification. Indeed, this approach could provide stronger evidence for mitochondrial fragmentation upon ATP5I-KO and its potential rescue by ATP5I reintroduction. We will consider integrating this methodology into our analysis to enhance the precision and robustness of our findings.

(3) Finally, the authors report in the last part of the paper a genetic CRISPR/Cas9 KO screen in NALM-6 cells cultured with high amounts of metformin to identify potential new mediators of metformin action. It is difficult to connect that to the rest of the paper because a) different concentrations of metformin are used and b) the metabolic effects on energy consumption are not defined. They argue about the molecular function of the obtained hits based on literature and on a comparison of the pattern of genetic alterations based on treatments with known inhibitors such as oligomycin and rotenone. However, a direct connection is not provided, thus the interpretation at the end of the results that "the OMA1-DEL1-HRI pathway mediates the antiproliferative activity of both biguanides and the F1ATPase inhibitor oligomycin" while increasing glycolysis, needs to be toned down. This is an interesting observation, but no causality is provided. In general, this part stands alone and needs to be better connected to the rest of the paper.

NALM-6 are very glycolytic, have low respiration rates, and weak dependence on ATP5I(6), forcing us to use higher concentrations of metformin to inhibit their growth. Recent results show that metformin targets PEN2 in the cytosol to increase AMPK activity, controlling both the glucose lowering and the life span extension abilities of metformin 7. This work raises the question whether the antiproliferative and anticancer effects of metformin are due to a mitochondrial activity or are controlled by this new pathway of AMPK activation. Hence, the genetic screening was performed to unbiasedly find how metformin works. The results provide compelling evidence for mitochondria and in particular the ATP synthase as potential targets of metformin and a foundation for future studies. We will revise the text and abstract to better reflect the exploratory nature of this finding and ensure clarity.

(1) Paumard, P. et al. Two ATP synthases can be linked through subunits i in the inner mitochondrial membrane of Saccharomyces cerevisiae. Biochemistry 41, 10390-10396 (2002). https://doi.org/10.1021/bi025923g

(2) Paumard, P. et al. The ATP synthase is involved in generating mitochondrial cristae morphology. EMBO J 21, 221-230 (2002). https://doi.org/10.1093/emboj/21.3.221

(3) Habersetzer, J. et al. ATP synthase oligomerization: from the enzyme models to the mitochondrial morphology. Int J Biochem Cell Biol 45, 99-105 (2013). https://doi.org/10.1016/j.biocel.2012.05.017

(4) Xian, H. et al. Metformin inhibition of mitochondrial ATP and DNA synthesis abrogates NLRP3 inflammasome activation and pulmonary inflammation. Immunity 54, 1463-1477 e1411 (2021). https://doi.org/10.1016/j.immuni.2021.05.004

(5) Hawley, S. A. et al. Use of cells expressing gamma subunit variants to identify diverse mechanisms of AMPK activation. Cell metabolism 11, 554-565 (2010). https://doi.org/10.1016/j.cmet.2010.04.001

(6) Hlozkova, K. et al. Metabolic profile of leukemia cells influences treatment efficacy of L-asparaginase. BMC Cancer 20, 526 (2020). https://doi.org/10.1186/s12885-020-07020-y

(7) Ma, T. et al. Low-dose metformin targets the lysosomal AMPK pathway through PEN2. Nature 603, 159-165 (2022). https://doi.org/10.1038/s41586-022-04431-8

eLife Assessment

This valuable manuscript describes ATP5I, a subunit of F1Fo-ATP synthase, as a key target of medicinal biguanides, however, it provides incomplete evidence of a direct interaction between ATP5I and metformin. The knockout of ATP5I in pancreatic cancer cells mimics biguanide treatment, inducing a metabolic switch from OXPHOS to glycolysis due to a compromised expression of the Complex I protein NDUFB8. This results in a markedly decreased NAD/NADH ratio and decreased cell proliferation. These findings point out ATP5I as a promising mitochondrial target for cancer therapies and contribute to our understanding of metformin's mechanism of action since many of its molecular mechanisms remain poorly understood.

Reviewer #1 (Public review):

Summary:

In the manuscript entitled 'The Role of ATP Synthase Subunit e (ATP5I) in 1 Mediating the Metabolic and Antiproliferative 2 Effects of Biguanides', Lefrancois G et al. identifies ATP5I, a subunit of F1Fo-ATP synthase, as a key target of medicinal biguanides. ATP5I stabilizes F1Fo-ATP synthase dimers, essential for cristae morphology, but its role in cancer metabolism is understudied. The research shows ATP5I interacts with a biguanide analogue, and its knockout in pancreatic cancer cells mimics biguanide treatment effects, including altered mitochondria, reduced OXPHOS, and increased glycolysis. ATP5I knockout cells resist biguanide-induced antiproliferative effects, but reintroducing ATP5I restores the effects of metformin and phenformin. These findings highlight ATP5I as a promising mitochondrial target for cancer therapies. The manuscript is well written.

Strengths:

Demonstrated the experiments in systematic and well-accepted methods.

Weaknesses:

The significance of the target molecule and mechanisms may help in understanding the molecular mechanisms of metformin.

Reviewer #2 (Public review):

Summary:

The mechanism(s) by which the therapeutic drug metformin lowers blood glucose in type 2 diabetes and inhibits cell proliferation at higher concentrations remain contentious. Inhibition of complex 1 of the mitochondrial respiratory chain with consequent changes in cellular metabolites which favour allosteric activation of phosphofructokinase-1, allosteric inhibition of fructose bisphosphatase-1 and cAMP signalling and activation of AMPK which phosphorylates transcription factors are candidate mechanisms. The current manuscript proposes the e-subunit of ATP-synthase as a putative binding protein of biguanides and demonstrates that it regulates the expressivity of the Complex 1 protein NDUFB8.

Strengths:

(1) The metformin conjugate and metformin show comparable efficacy on inhibition of cell proliferation in the millimolar range.

(2) Demonstration of compromised expression of the Complex I protein NDUFB8 by the ATP5I knockout and its reversal by ATP5I expression is an important strength of the study. This shows that the decreased "sensitivity" to metformin in the ATP5I knock-out cells could be due to various proteins.

(3) Demonstration of converse effects of ATP5I KO and re-expression ATP5I on the NAD/NADH ratio.

Weaknesses:

(1) The interpretation of the cellular co-localization of the biotin-biguanide conjugate with TOMM20 (Figure 1-D) as mitochondrial "accumulation" of the conjugate is overstated because it cannot exclude binding of the conjugate to the mitochondrial membrane. It would have been more convincing if additional incubations with the biotin-biguanide conjugate in combination with metformin had shown that metformin is competitive with the biotin-conjugate.

(2) The manuscript reports the identification of 69 proteins by mass spectrometry of the pull-down assay of which 31 proteins were eluted by metformin. However, no Mass Spectrometry data is presented of the peptides identified. The methodology does not state the minimum number of peptides (1, 2?) that were used for the identification of the 31/69 proteins.

(3) The validation of ATP5I was based on the use of recombinant protein (which was 90% pure) for the SPR and the use of a single antibody to ATP5I. The validity of the immunoblotting rests on the assumption that there is no "non-specific" immunoactivity in the relevant mol wt range. Information on the validation of the antibody would be helpful.

(4) Knock-out of ATP5I markedly compromised the NAD/NADH ratio (Fig.3A) and cell proliferation (Figure 3D). These effects may be associated with decreased mitochondrial membrane potential which could explain the low efficacy of metformin (and most of the data in Figures 3-5). This possibility should be discussed. Effects of [metformin] on the NAD/NADH ratio in control cells and ATP5I-KO would have been helpful because the metformin data on cell growth is normalized as fold change relative to control, whereas the NAD/NADH ratio would represent a direct absolute measurement enabling comparison of the absolute effect in control cells with ATP5I KO.

(5) Figure-6 CRISPR/Cas9 KO at 16mM metformin in comparison with 70nM rotenone and 2 micromolar oligomycin (in serum-containing medium). The rationale for the use of such a high concentration of metformin has not been explained. In liver cells metformin concentrations above 1mM cause severe ATP depletion, whereas therapeutic (micromolar) concentrations have minimal effects on cellular ATP status. The 16mM concentration is ~2 orders of magnitude higher than therapeutic concentrations and likely linked to compromised energy status. The stronger inhibition of cell proliferation by 16mM metformin compared with rotenone or oligomycin raises the issue of whether the changes in gene expression may be linked to the greater inhibition of mitochondrial metabolism. Validation of the cellular ATP status and NAD/NADH with metformin as compared with the two inhibitors could help the interpretation of this data.

Reviewer #3 (Public review):

Most of the data are based on measurements of the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) measured by the Seahorse analyser in control and ATP5l KO cells. However, these measurements are conducted by a single injection of a biguanide, followed over time and presented as fold change. By doing so, the individual information on the effect of metformin and derivate on control and KO cells are lost. In addition, the usual measurement of OCR is coupled with certain inhibitors and uncouplers, such as oligomycin, FCCP, and Antimycin A/rotenone, to understand the contribution of individual complexes to respiration. Since biguanides and ATP5l KO affect protein levels of components of complex I and IV, it would be informative to measure their individual contributions/effects in the Seahorse. To further strengthen the data, it would be helpful to obtain measurements of actual ATP levels in these cells, as this would explain the activation of AMPK.

The authors report on alterations in mitochondrial morphology upon ATP5l KO, which is measured by subjective quantifications of filamentous versus puncta structures. Fiji offers great tools to quantify the mitochondrial network unbiasedly and with more accuracy using deconvolution and skeletonization of the mitochondria, providing the opportunity to measure length, shape, and number quantitatively. This will help to understand better, whether mitochondria are really fragmented upon ATP5l KO and rescued by its re-introduction.

Finally, the authors report in the last part of the paper a genetic CRISPR/Cas9 KO screen in NALM-6 cells cultured with high amounts of metformin to identify potential new mediators of metformin action. It is difficult to connect that to the rest of the paper because a) different concentrations of metformin are used and b) the metabolic effects on energy consumption are not defined. They argue about the molecular function of the obtained hits based on literature and on a comparison of the pattern of genetic alterations based on treatments with known inhibitors such as oligomycin and rotenone. However, a direct connection is not provided, thus the interpretation at the end of the results that "the OMA1-DEL1-HRI pathway mediates the antiproliferative activity of both biguanides and the F1ATPase inhibitor oligomycin" while increasing glycolysis, needs to be toned down. This is an interesting observation, but no causality is provided. In general, this part stands alone and needs to be better connected to the rest of the paper.

eLife Assessment

This valuable study utilizes a newly developed approach to culture T gondii bradyzoites in myotubes, and then takes advantage of the antiparasitic compound collection known as the Pathogen Box, to find compounds that target both tachyzoite and bradyzoite forms of the parasite. A set of compounds yielding patterns consistent with targeting the mitochondrial bc1 complex was explored further, with solid evidence for changes in ATP production in bradyzoites to support the conclusions about the importance of this complex. The paper will be interesting for parasitologists studying drug discovery of apicomplexan parasites.

Reviewer #1 (Public review):

Summary:

The authors' goal was to advance the understanding of metabolic flux in the bradyzoite cyst form of the parasite T. gondii, since this is a major form of transmission of this ubiquitous parasite, but very little is understood about cyst metabolism and growth.

Nonetheless, this is an important advance in understanding and targeting bradyzoite growth.

Strengths:

The study used a newly developed technique for growing T. gondii cystic parasites in a human muscle-cell myotube format, which enables culturing and analysis of cysts. This enabled the screening of a set of anti-parasitic compounds to identify those that inhibit growth in both vegetative (tachyzoite) forms and bradyzoites (cysts). Three of these compounds were used for comparative Metabolomic profiling to demonstrate differences in metabolism between the two cellular forms.

One of the compounds yielded a pattern consistent with targeting the mitochondrial bc1 complex and suggests a role for this complex in metabolism in the bradyzoite form, an important advance in understanding this life stage.

Weaknesses:

Studies such as these provide important insights into the overall metabolic differences between different life stages, and they also underscore the challenge of interpreting individual patterns caused by metabolic inhibitors due to the systemic level of some of the targets, so that some observed effects are indirect consequences of the inhibitor action. While the authors make a compelling argument for focusing on the role of the bc1 complex, there are some inconsistencies in the patterns that underscore the complexity of metabolic systems.

Reviewer #2 (Public review):

Summary:

A particular challenge in treating infections caused by the parasite Toxoplasma gondii is to target (and ultimately clear) the tissue cysts that persist for the lifetime of an infected individual. The study by Maus and colleagues leverages the development of a powerful in vitro culture system for the cyst-forming bradyzoite stage of Toxoplasma parasites to screen a compound library for candidate inhibitors of parasite proliferation and survival. They identify numerous inhibitors capable of inhibiting both the disease-causing tachyzoite and the cyst-forming bradyzoite stages of the parasite. To characterize the potential targets of some of these inhibitors, they undertake metabolomic analyses. The metabolic signatures from these analyses lead them to identify one compound (MMV1028806) that interferes with aspects of parasite mitochondrial metabolism. The authors claim that MV1028806 targets the bc1 complex of the mitochondrial electron transport chain of the parasite, although the evidence for this is indirect and speculative. Nevertheless, the study presents an exciting approach for identifying and characterizing much-needed inhibitors for targeting tissue cysts in these parasites.

Strengths:

The study presents convincing proof-of-principle evidence that the myotube-based in vitro culture system for T. gondii bradyzoites can be used to screen compound libraries, enabling the identification of compounds that target the proliferation and/or survival of this stage of the parasite. The study also utilizes metabolomic approaches to characterize metabolic 'signatures' that provide clues to the potential targets of candidate inhibitors, although falls short of identifying the actual targets.

Weaknesses:

(1) The authors claim to have identified a compound in their screen (MMV1028806) that targets the bc1 complex of the mitochondrial electron transport chain (ETC). The evidence they present for this claim is indirect (metabolomic signatures and changes in mitochondrial membrane potential) and could be explained by the compound targeting other components of the ETC or affecting mitochondrial biology or metabolism in other ways. In order to make the conclusion that MMV1028806 targets the bc1 complex, the authors should test specifically whether MMV1028806 inhibits bc1-complex activity (i.e. in a direct enzymatic assay for bc1 complex activity). Testing the activity of MMV1028806 against other mitochondrial dehydrogenases (e.g. dihydroorotate dehydrogenase) that feed electrons into the ETC might also provide valuable insights. The experiments the authors perform also do not directly measure whether MMV1028806 impairs ETC activity, and the authors could also test whether this compound inhibits mitochondrial O2 consumption (as would be expected for a bc1 inhibitor).

(2) The authors claim that compounds targeting bradyzoites have greater lipophilicity than other compounds in the library (and imply that these compounds also have greater gastrointestinal absorbability and permeability across the blood-brain barrier). While it is an attractive idea that lipophilicity influences drug targeting against bradyzoites, the effect seems pretty small and is complicated by the fact that the comparison is being made to compounds that are not active against parasites. If the authors are correct in their assertion that lipophilicity is a major determinant of bradyzoicidal compounds compared to compounds that target tachyzoites alone, you would expect that compounds that target tachyzoites alone would have lower lipophilicity than those that target bradyzoites. It would therefore make more sense to (statistically) compare the bradyzoicidal and dual-acting compounds to those that are only active in tachyzoites (visually the differences seem small in Figure S2B). This hypothesis would be better tested through a structure-activity relationship study of select compounds (which is beyond the scope of the study). Overall, the evidence the authors present that high lipophilicity is a determinant of bradyzoite targeting is not very convincing, and the authors should present their conclusions in a more cautious manner.

(3) Page 11 and Figure 7. The authors claim that their data indicate that ATP is produced by the mitochondria of bradyzoites "independently of exogenous glucose and HDQ-target enzymes." The authors cite their previous study (Christiansen et al, 2022) as evidence that HDQ can enter bradyzoites, since HDQ causes a decrease in mitochondrial membrane potential. Membrane potential is linked to the synthesis of ATP via oxidative phosphorylation. If HDQ is really causing a depletion of membrane potential, is it surprising that the authors observe no decrease in ATP levels in these parasites? Testing the importance of HDQ-target enzymes using genetic approaches (e.g. gene knockout approaches) would provide better insights than the ATP measurements presented in the manuscript, although would require considerable extra work that may be beyond the scope of the study. Given that the authors' assay can't distinguish between ATP synthesized in the mitochondrion vs glycolysis, they may wish to interpret their data with greater caution.

Reviewer #3 (Public review):

Summary:

The authors describe an exciting 400-drug screening using a MMV pathogen box to select compounds that effectively affect the medically important Toxoplasma parasite bradyzoite stage. This work utilises a bradyzoites culture technique that was published recently by the same group. They focused on compounds that affected directly the mitochondria electron transport chain (mETC) bc1-complex and compared them with other bc1 inhibitors described in the literature such as atovaquone and HDQs. They further provide metabolomics analysis of inhibited parasites which serves to provide support for the target and to characterise the outcome of the different inhibitors.

Strengths:

This work is important as, until now, there are no effective drugs that clear cysts during T. gondii infection. So, the discovery of new inhibitors that are effective against this parasite stage in culture and thus have the potential to battle chronic infection is needed. The further metabolic characterization provides indirect target validation and highlights different metabolic outcomes for different inhibitors. The latter forms the basis for new studies in the field to understand the mode of inhibition and mechanism of bc1-complex function in detail.

The authors focused on the function of one compound, MMV1028806, that is demonstrated to have a similar metabolic outcome to burvaquone. Furthermore, the authors evaluated the importance of ATP production in tachyzoite and bradyzoites stages and under atovaquone/HDQs drugs.

Weaknesses:

Although the authors did experiments to identify the metabolomic profile of the compounds and suggested bc-1 complex as the main target of MMV1028806, they did not provide experimental validation for that.

Author response:

We thank the reviewers for taking the time to read and critically assess our manuscript.

We agree with the main points and they will be addressed in both writing and in additional experiments in a revised version of the paper.

The shared and major point of criticism are non-conclusive metabolomic data that indicate the bc1-complex in T. gondii as a MMV1028806 target tachyzoites and bradyzoites. Regarding the former, our conclusion was mainly based on both metabolite abundance changes that are observed after treatment with one bona-fide bc1-complex inhibitor atovaquone and also steady-state stable isotope incorporation patterns. While it is true that secondary effects of metabolic inhibition occur and are often dominant, isotope labelling equilibria take more time to establish and may reflect more accurately blocked metabolic reactions i.e. the primary target.

Regardless, we will follow the excellent suggestions to functionally assay particular mitochondrial electron transfer reactions to corroborate or revise our conclusions regarding the primary MMV1028806 target.

For more details please refer the full author responses that will accompany the revised manuscript.

eLife Assessment

The study provides valuable insights into the mechanisms underlying neurovascular coupling; the authors present solid evidence demonstrating that layer II/III pyramidal neurons can induce vasoconstriction under conditions of intense optogenetic stimulation. They identify three distinct signalling pathways responsible for this effect, involving direct action on smooth muscle cells, as well as indirect modulation via interneurons or astrocytes. This work will be of interest to the broader neuroscience community and has potential implications for understanding pathological microcirculation in the brain, particularly in conditions characterized by strong excitatory neuronal activation. There are however questions that should be clarified, especially the conflict between three identified parallel pathways and the observed complete inhibition of the constriction by blockage of the NPY Y1 receptors.

Reviewer #1 (Public review):

Neuronal activity spatiotemporal fine-tuning of cerebral blood flow balances metabolic demands of changing neuronal activity with blood supply. Several 'feed-forward' mechanisms have been described that contribute to activity-dependent vasodilation as well as vasoconstriction leading to a reduction in perfusion. Involved messengers are ionic (K+), gaseous (NO), peptides (e.g., NPY, VIP), and other messengers (PGE2, GABA, glutamate, norepinephrine) that target endothelial cells, smooth muscle cells, or pericytes. Contributions of the respective signaling pathways likely vary across brain regions or even within specific brain regions (e.g., across the cortex) and are likely influenced by the brain's physiological state (resting, active, sleeping) or pathological departures from normal physiology.

The manuscript "Elevated pyramidal cell firing orchestrates arteriolar vasoconstriction through COX-2-derived prostaglandin E2 signaling" by B. Le Gac, et al. investigates mechanisms leading to activity-dependent arteriole constriction. Here, mainly working in brain slices from mice expressing channelrhodopsin 2 (ChR2) in all excitatory neurons (Emx1-Cre; Ai32 mice), the authors show that strong optogenetic stimulation of cortical pyramidal neurons leads to constriction that is mediated through the cyclooxygenase-2 / prostaglandin E2 / EP1 and EP3 receptor pathway with contribution of NPY-releasing interneurons and astrocytes releasing 20-HETE. Specifically, using a patch clamp, the authors show that 10-s optogenetic stimulation at 10 and 20 Hz leads to vasoconstriction (Figure 1), in line with a stimulation frequency-dependent increase in somatic calcium (Figure 2). The vascular effects were abolished in the presence of TTX and significantly reduced in the presence of glutamate receptor antagonists (Figure 3). The authors further show with RT-PCR on RNA isolated from patched cells that ~50% of analyzed cells express COX-1 or -2 and other enzymes required to produce PGE2 or PGF2a (Figure 4). Further, blockade of COX-1 and -2 (indomethacin), or COX-2 (NS-398) abolishes constriction. In animals with chronic cranial windows that were anesthetized with ketamine and medetomidine, 10-s long optogenetic stimulation at 10 Hz leads to considerable constriction, which is reduced in the presence of indomethacin. Blockade of EP1 and EP3 receptors leads to a significant reduction of the constriction in slices (Figure 5). Finally, the authors show that blockade of 20-HETE synthesis caused moderate and NPY Y1 receptor blockade a complete reduction of constriction.

The mechanistic analysis of neurovascular coupling mechanisms as exemplified here will guide further in-vivo studies and has important implications for human neuroimaging in health and disease. Most of the data in this manuscript uses brain slices as an experimental model which contrasts with neurovascular imaging studies performed in awake (headfixed) animals. However, the slice preparation allows for patch clamp as well as easy drug application and removal. Further, the authors discuss their results in view of differences between brain slices and in vivo observations experiments, including the absence of vascular tone as well as blood perfusion required for metabolite (e.g., PGE2) removal, and the presence of network effects in the intact brain. The manuscript and figures present the data clearly; regarding the presented mechanism, the data supports the authors' conclusions. Some of the data was generated in vivo in head-fixed animals under anesthesia; in this regard, the authors should revise the introduction and discussion to include the important distinction between studies performed in slices, or in acute or chronic in-vivo preparations under anesthesia (reduced network activity and reduced or blockade of neuromodulation, or in awake animals (virtually undisturbed network and neuromodulatory activity). Further, while discussed to some extent, the authors could improve their manuscript by more clearly stating if they expect the described mechanism to contribute to CBF regulation under 'resting state conditions' (i.e., in the absence of any stimulus), during short or sustained (e.g., visual, tactile) stimulation, or if this mechanism is mainly relevant under pathological conditions; especially in the context of the optogenetic stimulation paradigm being used (10-s long stimulation of many pyramidal neurons at moderate-high frequencies) and the fact that constriction leading to undersupply in response to strongly increased neuronal activity seems counterintuitive?

Reviewer #2 (Public review):

The present study by Le Gac et al. investigates the vasoconstriction of cerebral arteries during neurovascular coupling. It proposes that pyramidal neurons firing at high frequency lead to prostaglandin E2 (PGE2) release and activation of arteriolar EP1 and EP3 receptors, causing smooth muscle cell contraction. The authors further claim that interneurons and astrocytes also contribute to vasoconstriction via neuropeptide Y (NPY) and 20-hydroxyeicosatetraenoic acid (20-HETE) release, respectively. The study mainly uses brain slices and pharmacological tools in combination with Emx1-Cre; Ai32 transgenic mice expressing the H134R variant of channelrhodopsin-2 (ChR2) in the cortical glutamatergic neurons for precise photoactivation. Stimulation with 470 nm light using 10-second trains of 5-ms pulses at frequencies from 1-20 Hz revealed small constrictions at 10 Hz and robust constrictions at 20 Hz, which were abolished by TTX and partially inhibited by a cocktail of glutamate receptor antagonists. Inhibition of cyclooxygenase-1 (COX-1) or -2 (COX-2) by indomethacin blocked the constriction both ex vivo (slices) and in vivo (pial artery), and inhibition of EP1 and EP3 showed the same effect ex vivo. Single-cell RT-PCR from patched neurons confirmed the presence of the PGE2 synthesis pathway.

While the data are convincing, the overall experimental setting presents some limitations. How is the activation protocol comparable to physiological firing frequency? The delay (minutes) between the stimulation and the constriction appears contradictory to the proposed pathway, which would be expected to occur rapidly. The experiments are conducted in the absence of vascular "tone," which further questions the significance of the findings. Some of the targets investigated are expressed by multiple cell types, which makes the interpretation difficult; for example, cyclooxygenases are also expressed by endothelial cells. Finally, how is the complete inhibition of the constriction by the NPY Y1 receptor antagonist BIBP3226 consistent with a direct effect of PGE2 and 20-HETE in arterioles? Overall, the manuscript is well-written with clear data, but the interpretation and physiological relevance have some limitations. However, vasoconstriction is a rather understudied phenomenon in neurovascular coupling, and the present findings may be of significance in the context of pathological brain hypoperfusion.

eLife Assessment

This valuable study combines experiments and theory to investigate the putative role of spontaneous correlated activity in establishing aligned topographic maps of neural activity in higher-order sensory areas, and will be of interest to researchers studying multisensory integration and brain development. However, the evidence presented is incomplete, as there are notable disconnects between the experimental data and the modeling setup, and there are methodological details that are either unclear or missing, limiting the strength of the claims.

Reviewer #1 (Public review):

Dwulet et al. combined experimental and modeling approaches to investigate how correlated spontaneous activity in the mouse's primary visual (V1) and primary somatosensory (S1) areas drives the development of multisensory integration in area RL. Notably, they focused on early developmental stages, before sensory experience occurs. Consistent with previous experimental findings, the authors first demonstrated that spontaneous activity becomes more sparse across development in all three areas, as measured by event amplitude, event duration, and participation ratio. Using a linear mixed model analysis to compare the maturation of this spontaneous activity, they found evidence that S1 matured the fastest. The authors then presented experimental evidence suggesting that these spontaneous events were moderately correlated both spatially and temporally.

They hypothesized that activity-dependent mechanisms use these correlations to establish connectivity across these regions. To test this hypothesis, the authors modeled a feedforward network with connections from S1 to RL and from V1 to RL, where the strength of connections depended on a Hebbian term for potentiation and a heterosynaptic term for depression. By investigating different levels of V1-S1 correlations, they found that moderate levels of correlation led to the significant development of topographically organized connectivity while maintaining a mix of bimodal and unimodal cells in RL. Additionally, when simulating a network with a more mature S1, they observed that topographical maps improved not only between S1 and RL but also between V1 and RL. Finally, the authors use linear regression to suggest that the mixture of bimodal and unimodal cells in RL is optimal for encoding the maximum amount of information from both V1 and S1.

However, there are significant gaps between the experimental data and the modeling setup, which weaken the paper's conclusions. Additionally, some key details are omitted, making it difficult to fully assess their analysis and interpret some of their figures.

(1) Some of the statistical measures and techniques in Figure 1 could benefit from clearer definitions. While the thresholds for activation (peak with at least 5% dF/F0) and events (20% of recorded cells activated simultaneously) are provided, event duration and participation rate are not clearly defined. Based on this definition of event alone, it is unclear why the minimum participation rate in Figure 1F is not 20%. Additionally, the conclusion that S1 matures earlier than RL and V1 could be strengthened by including a direct comparison between S1 and RL, as the current analysis only compares these areas to V1.

(2) The wide-field experiments in Figure 2 could be expanded to support the feedforward modeling assumptions. Currently, the spatial and temporal correlations presented leave open the possibility that these spontaneous events are traveling waves propagating from V1 to RL to S1 (or vice versa). This scenario would suggest a different connectivity scheme for the model. Clarifying this point with additional data analysis, specifically including temporal correlations involving RL, could provide stronger support for the model's assumptions.

(3) The functional correlation map in Figure 2D appears contradictory to the authors' modeling assumption that inputs are correlated spatially in V1 and S1. While V1 seed points align topographically with RL, this organization breaks down when extended into S1. In contrast, and in support of the modeling assumption, Figure 2E shows clearer topography across all three regions. A discussion of this discrepancy would be helpful, as it's a key conclusion of the figure. Additionally, it is unclear when this data was collected during development. Clarifying the developmental stage and analyzing how this map changes over time could strengthen the results.

(4) The modeling of spontaneous events with fixed amplitude and duration seems inconsistent with the experimental data in Figure 1, which shows variability in these parameters. This is particularly confusing in Figure 4, where S1 maturation is modeled as a stronger topographical alignment with RL, but the experimental data defines maturation based on amplitude, duration, and event rates. Justifying these modeling choices or adapting the model to reflect experimental variability would create a better connection between the theory and data.

(5) Several important details of the mathematical model are missing or unclear, partly due to typos. The Results section mentions the general framework of the input correlation matrix (e.g., "S1 and V1 neurons were driven by a combination of events, independent and shared in each V1 and S1" and "each independent event activated a randomly chosen, contiguous set of neurons"), but the specifics are not fully explained. Additionally, the caption of Figure 5 refers to a non-linear transfer function (a sigmoid), but these details are not provided in the Methods section, which instead suggests a linear model was used. A careful review of the main text and Methods section would help ensure that all the necessary details are included and that the story is both complete and accurate.

(6) While Figure 5 supports the paper's conclusion that a mixture of unimodal and bimodal neurons in RL optimizes information encoding, the authors missed an opportunity to strengthen the connection between the model and experimental data. Specifically, they could apply this reconstruction method to the experimental data and examine how RL's ability to reconstruct V1/S1 activity changes across development. Their model predicts that this performance would improve over time, and if this trend is observed in the experimental data, it would provide strong validation that these feedforward connections are developing in line with the model's predictions.

Reviewer #2 (Public review):

The authors aim to investigate the role of spontaneous activity in shaping the development of multisensory integration in the brain, specifically focusing on the connections between primary visual and somatosensory sensory areas (V1 and S1) and a higher-order cortical area rostro-lateral to V1 (RL). They seek to understand how spontaneous activity guides the formation of aligned topographic maps and the emergence of bimodal neurons in RL.

First, the authors found that spontaneous activity in all three areas sparsifies over time, but S1 exhibits more mature patterns earlier than V1 and RL. They claimed that correlated activity among neighboring regions of these areas during development carries topographic information. These data were used to implement a computational model that employed Hebbian rules of synaptic plasticity. The model indicated that correlated spontaneous activity can generate topographic connectivity between S1/V1 and RL and bimodal neurons in RL. The model suggested that the more mature spontaneous activity in S1 can guide map alignment between V1 and RL. In addition, the model also suggested that a mixture of bimodal and unimodal neurons in RL is optimal for decoding information from V1 and S1.

While the data presented in the manuscript is promising and provides preliminary insights into the role of spontaneous activity in multisensory integration, it would be beneficial to strengthen the experimental foundation regarding the correlation between V1, S1, and RL. Incorporating more rigorous spatio-temporal analyses of spontaneous activity could enhance the robustness of these findings.

Here are some important concerns:

(1) The analysis of how spatial topography influences activity correlations in Figure 2 has several issues.<br /> 1a. While squares in V1 and S1 covered a small area of these sensory areas, the correlated territories in RL covered the entire area of RL. The topographic map in V1 continues caudally, so where is the rest of the map in RL? Something similar applies to the relationship between S1 and RL.<br /> 1b. It is essential to know how areas were drawn. High precision is required.<br /> 1c. It is not clear if correlated activity means different events in sync or large events that cover 2 or all 3 cortical areas of interest. The figure points to the second option, which contradicts the size of events at these stages, mainly in the oldest mice analyzed here.<br /> 1d. It is fundamental to know in detail and provide examples of how the detection of events was performed. For instance, could the dispersion of light from an event in V1 close to RL cause the detection of activity in RL?

(2) For the correlations among V1, S1, and RL, it is crucial to have a consistent method to delineate the borders of cortical areas. The authors mention in one sentence that areas were drawn according to a reference map. More details are needed to convince the reader that the borders are accurate, especially because their shape and position change with age.

(3) The results from the model seem to be based on the initial bias in connectivity between neighboring cells from the different areas. Then, it seems straightforward that implementing correlated activity with Hebbian and synaptic depression rules will force the strengthening of connections between spatially close cells. Despite this apparent predisposition of the model towards a defined outcome, the flaws in the experimental data used prevent a rigorous interpretation of the computational model.

(4) In the Introduction, the authors nicely and briefly explain the role of primary and higher-order sensory cortices in information processing. They also explain how spontaneous activity during development helps to build these circuits by refining connections or establishing hierarchies. They continue explaining the relevance of aligning different topographic maps to allow multisensory integration. Then they provide some examples of sites of multisensory integration. This provides a general context for the data presented in the Results section; however, and importantly, there is no specific introduction of why they are interested in RL and its interaction with V1 and S1. The authors should introduce the RL area and explain why it is an interesting site for multisensory processing.

(5) The results shown in Figure 1 corroborate published data from Golshani et al, Rochefort et al, Murakami et al. While the reproduction of data is more than welcome, the authors should specify which part of the data is completely new and acknowledge clearly the rest as corroboration of previous data. The sentence "As described in previous experiments ..." partially acknowledges this fact but is not clear enough. In addition, the transition between this part of the manuscript and the next data is not smooth. Data seems to be used to feed the model so perhaps the organization of the manuscript leaves room for improvement.

Reviewer #3 (Public review):

Summary:

The study by Dwulet et al. explores how the development of spontaneous neural activity in primary sensory cortices influences the co-alignment of multiple sensory modalities in higher-order brain areas (HOAs). To address this question, they focus on connectivity between the primary visual (V1) and somatosensory (S1) cortices and an associative cortical area (RL) in mice. The authors combine experimental (wide-field and two-photon calcium imaging) and computational approaches to show that spontaneous activity matures at a different pace across these brain regions. Their data indicate that S1 develops more rapidly than V1, which is possibly beneficial for RL's integration of visual and somatosensory inputs through correlated spontaneous activity. Using a computational model, they demonstrate that a moderate correlation between V1 and S1 activity can optimally guide the formation of bimodal neurons in RL, which are crucial for maximizing the decodability of multisensory stimuli. This finding highlights the role of correlated spontaneous activity in primary sensory cortices in establishing co-aligned topographic multimodal sensory representations in downstream circuits.

Strengths:

The manuscript is well written and it provides strong enough evidence to support the main claim of the authors. The insights on the role of correlated activity on instructing co-aligned multisensory maps in HOAs are not trivial and are an important advancement for the field.

Weaknesses:

In the opinion of this reviewer, the study has no major weaknesses. A drawback of the work is that none of the predictions of the computational modeling have been corroborated through mechanistic experimental manipulations of early brain activity.

Author response:

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

The reviewers suggest a number of experiments and re-analyses to strengthen their claims and enhance the impact of the study. While a number of these are longer term, below is a summary of experiments and analyses recommended by the reviewers that can be accomplished in the shorter term:

(1) Clarification of statistical approaches, quantification, data presentation and description of cerebellar anatomical nomenclature (e.gs. detailed statistical methods for the GEO dataset analysis, FDR correction, quantification in Figs 2-4)

The revised manuscript will provide detailed statistical methods including FDR  correction for GEO dataset analyses and quantification. Please see specific responses to GEO dataset analyses below.

(2) Improved quality of images for select immunostains and in situ hybridization

The revised manuscript will address the quality of the images as indicated by the reviewers.

(3) Include a control group of hGFAP-Cre mice with loxP sites but without Sufu deletion to assess the effects of Cre-induced double-strand breaks on phosphorylated H2AX-DSB signaling.

The breeding scheme we used to generate homozygous SUFU conditional mutants will not generate pups carrying only hGFAP-Cre. Thus, we are unable to compare expression of gH2AX expression in littermates that do not carry loxP sites. The reviewer is correct in pointing out the possibility of Cre recombinase activity inducing double-strand breaks on its own. However, it is likely that any hGFAP-Cre induced double-strand breaks does not sufficiently cause the phenotypes we observed in homozygous mutants (Sufu-cKO) mice because the cerebellum of mice carry heterozygous SUFU mutations (hGFAP-Cre;Sufu-fl/+) do not display the profound cerebellar phenotypes observed in Sufu-cKO mice. We cannot rule out, however, any undetectable abnormalities that could be present which may require further analyses.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

SUFU modulates Sonic hedgehog (SHH) signaling and is frequently mutated in the B-subtype of SHH-driven medulloblastoma. The B-subtype occurs mostly in infants, is often metastatic, and lacks specific treatment. Yabut et al. found that Fgf5 was highly expressed in the B-subtype of SHH-driven medulloblastoma by examining a published microarray expression dataset. They then investigated how Fgf5 functions in the cerebellum of mice that have embryonic Sufu loss of function. This loss was induced using the hGFAP-cre transgene, which is expressed in multiple cell types in the developing cerebellum, including granule neuron precursors (GNPs) derived from the rhombic lip. By measuring the area of Pax6+ cells in the external granule cell layer (EGL) of Sufu-cKO mice at postnatal day 0, they find Pax6+ cells occupy a larger area in the posterior lobe adjacent to the secondary fissure, which is poorly defined. They show that Fgf5 RNA and phosphoErk1/2 immunostaining are also higher in the same disrupted region. Some of the phosphoErk1/2+ cells are proliferative in the Sufu-cKO. Western blot analysis of Gli proteins that modulate SHH signaling found reduced expression and absence of Gli1 activity in the region of cerebellar dysgenesis in Sufu-cKO mice. This suggests the GNP expansion in this region is independent of SHH signaling. Amazingly, intraventricular injection of the FGFR1-2 antagonist AZD4547 from P0-4 and examined histologically at P7 found the treatment restored cytoarchitecture in the cerebella of Sufu-cKO mice. This is further supported by NeuN immunostaining in the internal granule cell layer, which labels mature, non-diving neurons, and KI67 immunostaining, indicating dividing cells, and primarily found in the EGL. The mice were treated beginning at a timepoint when cerebellar cytoarchitecture was shown to be disrupted and it is indistinguishable from control following treatment. Figure 3 presents the most convincing and exciting data in this manuscript.

Sufu-cKO do not readily develop cerebellar tumors. The authors detected phosphorylated H2AX immunostaining, which labels double-strand breaks, in some cells in the EGL in regions of cerebellar dysgenesis in the Sufu-cKO, as was cleaved Caspase 3, a marker of apoptosis. P53, downstream of the double-strand break pathway, the protein was reduced in Sufu-cKO cerebellum. Genetically removing p53 from the Sufu-cKO cerebellum resulted in cerebellar tumors in 2-month old mice. The Sufu;p53-dKO cerebella at P0 lacked clear foliation, and the secondary fissure, even more so than the Sufu-cKO. Fgf5 RNA and signaling (pERK1/2) were also expressed ectopically.

The conclusions of the paper are largely supported by the data, but some data analysis need to be clarified and extended.

(1) The rationale for examining Fgf5 in medulloblastoma is not sufficiently convincing. The authors previously reported that Fgf15 was upregulated in neocortical progenitors of mice with conditional loss of Sufu (PMID: 32737167). In Figure 1, the authors report FGF5 expression is higher in SHH-type medulloblastoma, especially the beta and gamma subtypes mostly found in infants. These data were derived from a genome-wide dataset and are shown without correction for multiple testing, including other Fgfs. Showing the expression of other Fgfs with FDR correction would better substantiate their choice or moving this figure to later in the manuscript as support for their mouse investigations would be more convincing.

To assess FGF5 (ENSG00000138675) expression in MB tissues, we used Geo2R (Barrett et al., 2013) to analyze published human MB subtype expression arrays from accession no. GSE85217 (Cavalli et al., 2017). GEO2R is an interactive web tool that compares expression levels of genes of interest (GOI) between sample groups in the GEO series using original submitter-supplied processed data tables. We entered the GOI Ensembl ID and organized data sets according to age and MB subgroup or MB^SHH subtype classifications. GEO2R results presented gene expression levels as a table ordered by FDR-adjusted (Benjamini & Hochberg) p-values, with significance level cut-off at 0.05, processed by GEO2R’s built-in limma statistical test. Resulting data were subsequently exported into Prism (GraphPad). We generated scatter plots presenting FGF5 expression levels across all MB subgroups (Figure 1A) and MB^SHH subtypes (Figure 1D). We performed additional statistical analyses to compare FGF5 expression levels between MB subgroups and MB^SHH subtypes and graphed these data as violin plots (Figure 1B, 1C, and 1E). For these analyses, we used one-way ANOVA with Holm-Sidak’s multiple comparisons test, single pooled variance. P value ≤0.05 was considered statistically significant. Graphs display the mean ± standard error of the mean (SEM).

Author response image 1.

Comparative expression of FGF ligands, FGF5, FGF10, FGF12, and FGF19, across all MB subgroups. FGF12 expression is not significantly different, while FGF5, FGF10, and FGF19, show distinct upregulation in MB<sup>SHH subgroup (MB^WNT n=70, MB^SHH n=224, MB^GR3 n=143, MB^GR4 n=326).

Expression of the 21 known FGF ligands were also analyzed. Many FGFs did not exhibit differential expression levels in MB^SHH compared to other MB subgroups, such as with FGF12 in Figure 1. FGF5, FGF10, and FGF19 (the human orthologue of mouse FGF15) all showed specific upregulation in MB^SHH compared to other MB subgroups (Author response image 1), supporting our previous observations that FGF15 is a downstream target of SHH signaling (Yabut et al., 2020), as the reviewer pointed out. However, further stratification of MB^SHH patient data revealed that only FGF5 specifically showed upregulation in infants with MBSHH (MB^SHHβ and MB^SHHγ Author response image 2) indicating a more prominent role for FGF5 in the developing cerebellum and driver of MB^SHH tumorigenesis in this dynamic environment.

Author response image 2.

Comparative expression of FGF5, FGF10, and FGF19 in different MB^SHH subtypes. FGF5 specifically show mRNA relative levels above 6 in 81% of MB^SHH infant patient tumors (n=80 MB^SHHα and MB^SHHγ tumors) unlike 35% of MB^SHHα (n=65) or 0% of MB^SHHδ  (n=75) tumors.

(2) The Sufu-cKO cerebellum lacks a clear anchor point at the secondary fissure and foliation is disrupted in the central and posterior lobes. It would be helpful for the authors to review Sudarov & Joyner (PMID: 18053187) for nomenclature specific to the developing cerebellum.

The reviewers are correct that the cerebellar foliation is severely disrupted in central and posterior lobes, as per Sudarov and Joyner (Neural Development 2007). This nomenclature may be referred to describe the regions referred in this manuscript.

(3) The metrics used to quantify cerebellar perimeter and immunostaining are not sufficiently described. It is unclear whether the individual points in the bar graph represent a single section from independent mice, or multiple sections from the same mice. For example, in Figures 2B-D. This also applies to Figure 3C-D.

All quantification were performed from 2-3 20 um cerebellar sections of 3-6 independent mice per genotype analyzed. Individual points in the bar graphs represent the average cell number (quantified from 2-3 sections) from each mice. Figure 2B show data points from n=4 mice per genotype. Figure 2C show data from n=3 mice per genotype. Figure 2D show data from n=6 mice per genotype.  Figure 3C-D show data from n=3 mice per genotype.

(4) The data on Fgf5 RNA expression presented in Figure 2E are not sufficiently convincing. The perimeter and cytoarchitecture of the cerebellum are difficult to see and the higher magnification shown in 2F should be indicated in 2E.

The lack of foliation in Sufu-cKO cerebellum is clear particularly when visualizing the perimeter via DAPI labeling (Figure 2E). The expression area of FGF5 is also visibly larger, given that all images in Figure 2E are presented in the same scale (scale bars = 500 um). 

(5) The data presented in Figure 3 are not sufficiently convincing. The number of cells double positive for pErk and KI67 (Figure 3B) are difficult to see and appear to be few, suggesting the quantification may be unreliable.

We used KI67+ expression to provide a molecular marker of regions to be quantified in both WT and Sufu-cKO sections. Quantification of labeled cells were performed in images obtained by confocal microscopy, enabling imaging of 1-2 um optical slices since Ki67 or pERK expression might not localize within the same cellular compartments. We relied on continuous DAPI nuclear staining to distinguish individual cells in each optical slice and the colocalization of of Ki67 and pERK. All quantification were performed from 2-3 20 um cerebellar sections of 3-6 independent mice per genotype analyzed. Individual points in the bar graphs represent the average cell number (quantified from 2-3 sections) from each mice.

(6) The data presented in Figure 4F-J would be more convincing with quantification. The Sufu;p53-dKO appears to have a thickened EGL across the entire vermis perimeter, and very little foliation, relative to control and single cKO cerebella. This is a more widespread effect than the more localized foliation disruption in the Sufu-cKO. 

We agree with the reviewers that quantification of these phenotypes provide a solid measure of the defects. The phenotypes of Sufu:p53-dKO cerebellum are so profound requiring  in-depth characterization that will be the focus of future studies.

(7) Figure 5 does not convincingly summarize the results. Blue and purple cells in sagittal cartoon are not defined. Which cells express Fgf5 (or other Fgfs) has not been determined. The yellow cells are not defined in relation to the initial cartoon on the left.

The revised manuscript will address this confusion by clearly labeling the cells and their roles in the schematic diagram.

Reviewer #2 (Public Review):

Summary:

Mutations in SUFU are implicated in SHH medulloblastoma (MB). SUFU modulates Shh signaling in a context-dependent manner, making its role in MB pathology complex and not fully understood. This study reports that elevated FGF5 levels are associated with a specific subtype of SHH MB, particularly in pediatric cases. The authors demonstrate that Sufu deletion in a mouse model leads to abnormal proliferation of granule cell precursors (GCPs) at the secondary fissure (region B), correlating with increased Fgf5 expression. Notably, pharmacological inhibition of FGFR restores normal cerebellar development in Sufu mutant mice.

Strengths:

The identification of increased FGF5 in subsets of MB is novel and a key strength of the paper.

Weaknesses:

The study appears incomplete despite the potential significance of these findings. The current paper does not fully establish the causal relationship between Fgf5 and abnormal cerebellar development, nor does it clarify its connection to SUFU-related MB. Some conclusions seem overstated, and the central question of whether FGFR inhibition can prevent tumor formation remains untested.

Reviewer #3 (Public Review):

Summary:

The interaction between FGF signaling and SHH-mediated GNP expansion in MB, particularly in the context of Sufu LoF, has just begun to be understood. The manuscript by Yabut et al. establishes a connection between ectopic FGF5 expression and GNP over-expansion in a late-stage embryonic Sufu LoF model. The data provided links region-specific interaction between aberrant FGF5 signaling with the SHH subtype of medulloblastoma. New data from Yabut et al. suggest that ectopic FGF5 expression correlates with GNP expansion near the secondary fissure in Sufu LoF cerebella. Furthermore, pharmacological blockade of FGF signaling inhibits GNP proliferation. Interestingly, the data indicate that the timing of conditional Sufu deletion (E13.5 using the hGFAP-Cre line) results in different outcomes compared to later deletion (using Math1-cre line, Jiwani et al., 2020). This study provides significant insights into the molecular mechanisms driving GNP expansion in SHH subgroup MB, particularly in the context of Sufu LoF. It highlights the potential of targeting FGF5 signaling as a therapeutic strategy. Additionally, the research offers a model for better understanding MB subtypes and developing targeted treatments.

Strengths:

One notable strength of this study is the extraction and analysis of ectopic FGF5 expression from a subset of MB patient tumor samples. This translational aspect of the study enhances its relevance to human disease. By correlating findings from mouse models with patient data, the authors strengthen the validity of their conclusions and highlight the potential clinical implications of targeting FGF5 in MB therapy.

The data convincingly show that FGFR signaling activation drives GNP proliferation in Sufu, conditional knockout models. This finding is supported by robust experimental evidence, including pharmacological blockade of FGF signaling, which effectively inhibits GNP proliferation. The clear demonstration of a functional link between FGFR signaling and GNP expansion underscores the potential of FGFR as a therapeutic target in SHH subgroup medulloblastoma.

Previous studies have demonstrated the inhibitory effect of FGF2 on tumor cell proliferation in certain MB types, such as the ptc mutant (Fogarty et al., 2006)(Yaguchi et al., 2009). Findings in this manuscript provide additional support suggesting multiple roles for FGF signaling in cerebellar patterning and development.

Weaknesses:

In the GEO dataset analysis, where FGF5 expression is extracted, the reporting of the P-value lacks detail on the statistical methods used, such as whether an ANOVA or t-test was employed. Providing comprehensive statistical methodologies is crucial for assessing the rigor and reproducibility of the results. The absence of this information raises concerns about the robustness of the statistical analysis.

The revised manuscript will include the following detailed explanation of the statistical analyses of the GEO dataset:

For the analysis of expression values of FGF5 (ENSG00000138675), we obtained these values using Geo2R (Barrett et al., 2013), which directly analyze published human MB subtype expression arrays from accession no. GSE85217 (Cavalli et al., 2017). GEO2R is an interactive web tool that compares expression levels of genes of interest (GOI) between sample groups in the GEO series using original submitter-supplied processed data tables. We simply entered the GOI Ensembl ID and organized data sets according to age and MB subgroup or MBSHH subtype classifications. GEO2R results presented gene expression levels as a table ordered by FDR-adjusted (Benjamini & Hochberg) p-values, with significance level cut-off at 0.05, processed by GEO2R’s built-in limma statistical test. Resulting data were subsequently exported into Prism (GraphPad). We generated scatter plots presenting FGF5 expression levels across all MB subgroups (Figure 1A) and MBSHH subtypes (Figure 1D). We performed additional statistical analyses to compare FGF5 expression levels between MB subgroups and MBSHH subtypes and graphed these data as violin plots (Figure 1B, 1C, and 1E). For these analyses, we used one-way ANOVA with Holm-Sidak’s multiple comparisons test, single pooled variance. P value ≤0.05 was considered statistically significant. Graphs display the mean ± standard error of the mean (SEM). Sample sizes were:

Author response table 1.

Another concern is related to the controls used in the study. Cre recombinase induces double-strand DNA breaks within the loxP sites, and the control mice did not carry the Cre transgene (as stated in the Method section), while Sufu-cKO mice did. This discrepancy necessitates an additional control group to evaluate the effects of Cre-induced double-strand breaks on phosphorylated H2AX-DSB signaling. Including this control would strengthen the validity of the findings by ensuring that observed effects are not artifacts of Cre recombinase activity.

The breeding scheme we used to generate homozygous SUFU conditional mutants will not generate pups carrying only hGFAP-Cre. Thus, we are unable to compare expression of gH2AX expression in littermates that do not carry loxP sites. The reviewer is correct in pointing out the possibility of Cre recombinase activity inducing double-strand breaks on its own. However, it is likely that any hGFAP-Cre induced double-strand breaks does not sufficiently cause the phenotypes we observed in homozygous mutants (Sufu-cKO) mice because the cerebellum of mice carry heterozygous SUFU mutations (hGFAP-Cre;Sufu-fl/+) do not display the profound cerebellar phenotypes observed in Sufu-cKO mice. We cannot rule out, however, any undetectable abnormalities that could be present which may require further analyses.

Although the use of the hGFAP-Cre line allows genetic access to the late embryonic stage, this also targets multiple celltypes, including both GNPs and cerebellar glial cells. However, the authors focus primarily on GNPs without fully addressing the potential contributions of neuron-glial interaction. This oversight could limit the understanding of the broader cellular context in which FGF signaling influences tumor development. 

The reviewer is correct in that hGFAP-Cre also targets other cell types, such as cerebellar glial cells, which are generated when Cre-expression has begun. It is possible that cerebellar glial cell development is also compromised in Sufu-cKO mice and may disrupt neuron-glial interaction, due to or independently of FGF signaling. In-depth studies are required to interrogate how loss of SUFU specifically affect development of cerebellar glial cells and influence their cellular interactions in the developing cerebellum.

Recommendations for the authors:

Editorial Comments:

The reviewers suggest a number of steps to improve the manuscript that include additional experiments and a deeper analyses and re-evaluation of existing data. Short of significant new experiments, there appears to be number of straightforward analyses that can improve the study:

(1) Reanalyses of statistical and quantitative approaches used (e.gs FDR correction, cerebellar deficits, GEO analyses.

The revised manuscript will include detailed information on the statistical and quantitative approaches as addressed in our response to the reviewer’s comments.

(2) More clear presentation of qualitative labeling approaches (immunohistochemistry and in situ hybridization).

A detailed description of the protocols used will be included in  the methods section for labeling methods in the revised manuscript.

Reviewer #1 (Recommendations For The Authors):

AZD4547 treatment of the dKO mice would provide more convincing evidence that FGF-targeted treatments could curtail tumor growth in these mice or refute the suggestion that FGF-targeted treatment could prevent tumor growth.

We agree that performing AZD4547 treatment on Sufu-dKO mice will strengthen these studies. However, we are unable to address since these mice are now unavailable. We hope that future studies will address these.

Atoh1 is referred to as Math1 (older nomenclature) and should be corrected.

The revised manuscript will include this change in nomenclature.

Check verb tense throughout the manuscript.

We will edit the manuscript further to check verb tenses prior to submission of the revised manuscript.

Reviewer #2 (Recommendations For The Authors):

Specific Comments:

(1) The identification of increased FGF5 in subsets of MB is novel and a key strength of the paper. However, the causal relationship between FGF5 and MB remains unestablished. Based on the current data, FGF5 can only be considered a biomarker for stratifying MB.

We agree with the reviewer that our studies do not provide direct evidence that FGF5 cause MB. Future investigation focusing on determining if FGF5 inhibition leads to phenotypic rescue will strongly establish the relationship between FGF5 and MB. The reviewer is also correct that our studies reveal that FGF5 acts as a potential biomarker, as we mentioned in the Discussion section.

(2) The upregulation of Fgf5 in Sufu-deficient cerebella is crucial to this study, yet the presented data are unconvincing to support this conclusion. In comparing Fgf5 expression between WT and Sufu mutants (Figures 2E, F and 4I), the cerebellar sections differ significantly, with mutant sections seemingly from a more lateral position. The authors should provide images of mutant sections from more comparable positions to accurately assess the effect of Sufu deficiency on Fgf5 expression. Additionally, the signals in Figure 2F resemble non-specific backgrounds rather than specific RNAscope signals.

The WT and mutant sections analyzed were carefully selected from comparable levels. The abnormal foliation in Sufu-cKO make the mutant sections look like they are from the lateral cerebellum.

Figure 2F (enlarged regions) point to punctate RNAScope signals which is characteristic of this labeling method (see RBFOX3 or GFAP labeling in DAPI-labeled cells in the mouse brain at https://acdbio.com/science/applications/research-areas/neuroscience). The higher number of punctate signals in some, but not all, DAPI-labeled cells in Figure 2F indicate that the FGF5 RNAScope signal is specific.

(3) Jiwani et al. (2020) reported that Fgf8 also expressed in region B of the EGL, is upregulated in Sufu-deficient cerebella and is necessary and sufficient for Sufu mutant GCP proliferation. The current study does not distinguish whether the FGFR inhibitor AZD4547 blocks Fgf5 and Fgf8 function in restoring cerebellar histology in Sufu mutants.

AZD4547 potently inhibits FGFR1, FGFR2, and FGFR3 autophosphorylation (Gavine et al., Cancer Research, 2012). FGF8 is reported to bind to these receptors (Ornitz and Itoh, 2015). Thus, the reviewer is correct that the studies will not distinguish between FGF5 or FGF8 activity. Further investigation on FGF8 expression and the effects of its inhibition in the Sufu-cKO neonatal cerebellum will determine whether tumorigenic processes are driven by either FGF5 or FGF8. Nevertheless, we postulate that FGF5 is exerting a greater effect in activating FGF signaling in the developing cerebellum given that it is highly expressed along the external granule layers of the developing cerebellum (Author response image 3).

Author response image 3.

Expression of FGF5 and FGF8 in the P4 mouse cerebellum (Allen Brain Atlas, https://developingmouse.brain-map.org )

(4) The authors should show whether AZD4547 treatment restores normal Fgf5 expression. Importantly, they need to test whether AZD4547 rescues the proliferation defect observed in Sufu;p53 double mutants.

We agree that performing AZD4547 treatment on Sufu-dKO mice will strengthen these studies. However, we are unable to address since these mice are now unavailable. We hope that future studies will address these.

(5) Jiwani et al. (2020) showed that deleting Sufu with Atoh1-Cre promotes Gli3R and suppresses Gli2 levels, leading to increased cell proliferation and delayed cell cycle exit in the central lobe. The findings of the current study (Supplementary Figure 1) seem to differ from this previous report, yet both studies conclude that Sufu-KO disrupts differentiation. The authors should provide an explanation for this discrepancy.

Our results align completely with the findings by Jiwani et al. (2020). Both studies showed reduced levels of Gli3R, showing nearly 50% reduction, when Sufu is deleted (see Figure 4A-4D in Jiwani et al., 2020).

(6) The hGFAP-Cre mouse line is used to delete Sufu from the cerebellum, but it is not commonly used for GCP-specific deletion. The authors need to provide a reference or more details on the temporal and spatial activity of the Cre line, as the cited paper describes its generation but offers little information on its activity in the developing cerebellum.

We appreciate the reviewer’s reminder to include the reference for the Schuller et al. 2008 paper. This study characterized the hGFAP-Cre temporal and spatial expression in the developing cerebellum, including granule cell precursors. We will include this reference in the revised manuscript.

(7) Based on the provided data, it is difficult to determine which cell types express Fgf5. Given that hGFAP-cre may delete Sufu in other cerebellar cell types, the authors should demonstrate that Fgf5 is expressed in granule cells or granule cell precursors.

Future studies will focus on further characterization of the role of FGF5 in cerebellar development, including the identity cells expressing FGF5. The reviewer is correct in that hGFAP-Cre also targets other cell types and that Sufu deletion in these cells induced ectopic FGF5 expression.

(8) The provided data show an increase in pERK+ cells in GCPs at the secondary fissure. This increase may simply reflect an accumulation of GCPs. It is unconvincing that there is an increase in pERK due to the loss of Sufu.

The reviewer is correct that the increase in GCPs will also result increase the number of pERK+ cells. To control for this, our quantification reflects the number of cells per unit area where Ki67+ cells. With these parameters, we found that there is an increased density of pERK+ cells in a given Ki67+ region. All quantification were performed from 2-3 20 um cerebellar sections of 3-6 independent mice per genotype analyzed. Individual points in the bar graphs represent the average cell number (quantified from 2-3 sections) from each mice.

(9) No data are provided on MB formation in Sufu-cKO; p53- mutants, and it is unknown whether FGFR inhibitors block tumor formation.

We agree that performing AZD4547 treatment on Sufu-dKO mice will strengthen these studies. However, we are unable to address since these mice are now unavailable. We hope that future studies will address these.

(10) The authors frequently mention "preneoplastic lesions" of GCPs in Sufu mutant mice. What evidence supports this claim?

Preneoplastic lesions are defined as cells carrying genetic and phenotypic alterations that show higher risk of malignancy (such as MB) but lack the capacity to grow autonomously in the absence of a secondary factor (Feo et al., 2011). In Sufu-cKO mice, we see abnormally proliferating and behaving granule precursor cells that do not grow autonomously, in the absence of a p53 LOF. The combined deletion of Sufu and p53 transforms these cells to become neoplastic.

(11) Fgf5 is normally expressed in region B. What is its potential function? Does AZD4547 affect normal development? 

Future studies will focus on further characterization of the role of FGF5 in cerebellar development, including the identity cells expressing FGF5. Regarding AZD4547, we did not observe any obvious difference between AZD4547-treated and vehicle-treated cerebelli. These indicate that AZD4547 inhibition of FGFRs under physiologic conditions does not significantly disrupt normal cerebellar development.

(12) Figure 3G: It is unclear which specimens were treated with AZD4547. The authors mention treatment in line 281 but contradict themselves in the figure legend.

We thank the reviewer for pointing out this typo. Cerebellar tissues shown in Figure 3G were all treated with AZD4547. The figure legend will be corrected in the revised manuscript.

(13) Figure 4J: The higher magnification images of the pERK/Ki67 staining appear identical in the control and Sufu;p53-dKO. The authors need to correct the mistake.

We thank the reviewer for pointing this out. We will correct this figure in the revised manuscript.

Minor Comments:

(1) Whenever possible, images comparing WT and mutants should be presented at the same scale within a figure. For example, readers might easily conclude that mutant brains are smaller than controls in Figure 4G.

Unfortunately, because the cerebellum of Sufu;p53-dKO mice are significantly bigger, we are unable to show the whole cerebellum in the same scale in Figure 4G. We wanted to emphasize the significant and abnormal cerebellar growth in this figure.

(2) The figure legend for Supplementary Figure 2 is missing.

Thank you for pointing this out. We will add a figure legend in this Supplementary data in the revised manuscript.

(3) The authors state that the expansion of Pax6+ GNPs in the newborn Sufu-cKO cerebellum (Figure 2) occurs in similar anatomical subregions where infantile MB tumors typically arise (Tan et al., 2018). The cited paper describes more abundant SHH MB in the cerebellar hemisphere. The authors need to elaborate on their statement to clarify this point.

The reviewer is correct in that Tan et al., 2018 observed tumors arising from the cerebellar hemisphere. More specifically, these tumors arise in the posterior/ventral regions of the cerebellar hemispheres (Figure 2 in Tan et al., 2018). Similarly, Sufu-cKO mice have more severe defects in the posterior/ventral regions of the cerebellar hemisphere (Figures 2A and 3F) and therefore corroborate the findings by Tan et al., that abnormal SHH signaling in these regions results in increased sensitivity to MB formation.

Reviewer #3 (Recommendations For The Authors):

Figure1 [Upregulated FGF5 expression in MBS-HH tumors]

- Statistical analysis from the Geo expression dataset does not provide enough detail. At least, the authors should mention whether they have made any adjustments from the default settings and how they extracted/plotted the FGF5 expression (Figure 1BCE).

For the analysis of expression values of FGF5 (ENSG00000138675), we obtained these values using Geo2R (Barrett et al., 2013), which directly analyze published human MB subtype expression arrays from accession no. GSE85217 (Cavalli et al., 2017). GEO2R is an interactive web tool that compares expression levels of genes of interest (GOI) between sample groups in the GEO series using original submitter-supplied processed data tables. We simply entered the GOI Ensembl ID and organized data sets according to age and MB subgroup or MBSHH subtype classifications. GEO2R results presented gene expression levels as a table ordered by FDR-adjusted (Benjamini & Hochberg) p-values, with significance level cut-off at 0.05, processed by GEO2R’s built-in limma statistical test. Resulting data were subsequently exported into Prism (GraphPad). We generated scatter plots presenting FGF5 expression levels across all MB subgroups (Figure 1A) and MB^SHH subtypes (Figure 1D). We performed additional statistical analyses to compare FGF5 expression levels between MB subgroups and MB^SHH subtypes and graphed these data as violin plots (Figure 1B, 1C, and 1E). For these analyses, we used one-way ANOVA with Holm-Sidak’s multiple comparisons test, single pooled variance. P value ≤0.05 was considered statistically significant. Graphs display the mean ± standard error of the mean (SEM). See Author response table 1 for sample sizes.

Figure 3 [Ectopic activation of FGF signaling in the EGL of P0 Sufu-cKO cerebellum]

- Gil1-lz mice reference wrong. Correct Bai CB, et al. 2002

- Generation of Sufu-cKO;Gli1-LacZ triple transgenic mice not described 

- Veh vs. treated not labelled (Figure 3F)

We will address these minor text changes in the revised manuscript. A more detailed description of the generation of Sufu-cKO;Gli1-LacZ triple transgenic will also be included in the Methods section.

Figure 5 [Proposed model]

- In the text, Figure 5 is mistaken for Figure 8. 

We will address these minor text changes in the revised manuscript.

eLife Assessment

This study provides valuable new insight into the role of Fgf signalling in SUFU mutation-linked cerebellar tumors and indicates novel therapeutic interventions via inhibition of Fgf signalling. The potential impact of this work is therefore very high and it is supported by solid evidence. However, due to current limitations in the full identification of the cell types secreting FGF5, and issues with robustness of evaluation of genetically engineered animals, the validation of some interpretations awaits future experiments.

Reviewer #1 (Public review):

Summary:

SUFU modulates Sonic hedgehog (SHH) signaling and is frequently mutated in the B-subtype of SHH driven medulloblastoma. The B-subtype occurs mostly in infants, is often metastatic, and lacks specific treatment. Yabut et al. found Fgf5 was highly expressed in the B-subtype of SHH driven medulloblastoma by examining a published microarray expression dataset. They then investigated how Fgf5 functions in the cerebellum of mice that have embryonic Sufu loss of function. This loss was induced using the hGFAP-cre transgene, which is expressed multiple cell types in the developing cerebellum, including granule neuron precursors (GNPs) derived from the rhombic lip. By measuring the area of Pax6+ cells in the external granule cell layer (EGL) of Sufu-cKO mice at postnatal day 0, they find Pax6+ cells occupy a larger area in the posterior lobe adjacent to the secondary fissure, which is poorly defined. They show that Fgf5 RNA and phosphoErk1/2 immunostaining are also higher in the same disrupted region. Some of the phosphoErk1/2+ cells are proliferative in the Sufu-cKO. Western blot analysis of Gli proteins that modulate SHH signaling found reduced expression and absence of Gli1 activity in the region of cerebellar dysgenesis in Sufu-cKO mice. This suggests the GNP expansion in this region is independent of SHH signaling. Amazingly, intraventricular injection of the FGFR1-2 antagonist AZD4547 from P0-4 and examined histologoically at P7 found the treatment restored cytoarchitecture in the cerebella of Sufu-cKO mice. This is further supported by NeuN immunostaining in the internal granule cell layer, which labels mature, non-diving neurons, and KI67 immunostaining, indicating dividing cells, and primarily found in the EGL. The mice were treated beginning at a timepoint when cerebellar cytoarchitecture was shown to be disrupted and it is indistinguishable from control following treatment. Fig.3 presents the most convincing and exciting data in this manuscript.

Sufu-cKO do not readily develop cerebellar tumors. The authors detected phosphorylated H2AX immunostaining, which labels double strand breaks, was in some cells in the EGL in regions of cerebellar dysgenesis in the Sufu-cKO, as was cleaved Caspase 3, a marker of apoptosis. P53, downstream of the double strand break pathway, protein was reduced in Sufu-cKO cerebellum. Genetically removing p53 from the Sufu-cKO cerebellum resulted in cerebellar tumors in 2 mo mice. The Sufu;p53-dKO cerebella at P0 lacked clear foliation, and the secondary fissure, even more so than the Sufu-cKO. Fgf5 RNA and signaling (pERK1/2) were also expressed ectopically.

In the revised manuscript, additional details have been added to clarify statistical analyses and to state limitations of the reported results in the absence of further experimental analyses.

Reviewer #2 (Public review):

Summary:

Mutations in SUFU are implicated in SHH medulloblastoma (MB). SUFU modulates Shh signaling in a context-dependent manner, making its role in MB pathology complex and not fully understood. This study reports that elevated FGF5 levels are associated with a specific subtype of SHH MB, particularly in pediatric cases. The authors demonstrate that Sufu deletion in a mouse model leads to abnormal proliferation of granule cell precursors (GCPs) at the secondary fissure (region B), correlating with increased Fgf5 expression. Notably, pharmacological inhibition of FGFR restores normal cerebellar development in Sufu mutant mice.

Strengths:

The identification of increased FGF5 in subsets of MB is novel and a key strength of the paper.

Weaknesses:

The study appears incomplete despite the potential significance of these findings. The current paper does not fully establish the causal relationship between Fgf5 and abnormal cerebellar development, nor does it clarify its connection to SUFU-related MB. Some conclusions seem overstated, and the central question of whether FGFR inhibition can prevent tumor formation remains untested.

Comments on revisions:

In this revised version, many of the concerns and comments raised by this and other reviewers remain unaddressed and require attention in future studies. The manuscript does not demonstrate significant improvement.

Specific Comments:

(1) In the figure provided by the authors, FGF5 appears to be highly expressed beneath the GCPs. Regarding our comment and Reviewer 1's Comment 7, it is essential to identify the cell types secreting FGF5 and clarify whether it functions in a paracrine or autocrine manner. This should be incorporated into the working model illustrated in Figure 5.<br /> (2) Contrary to the authors' claim that their results align completely with Jiwani et al. (2020), there is a discrepancy in the data. Jiwani et al. reported an increase in Gli2 levels in the Sufu mutant, whereas the current study shows the opposite result. This inconsistency needs to be addressed.

Reviewer #3 (Public review):

Summary:

The interaction between FGF signaling and SHH-mediated GNP expansion in MB, particularly in the context of Sufu LoF, has just begun to be understood. The manuscript by Yabut et al. establishes a connection between ectopic FGF5 expression and GNP over-expansion in a late stage embryonic Sufu LoF model. The data provided links region-specific interaction between aberrant FGF5 signaling with SHH subtype of medulloblastoma. New data from Yabut et al. suggest that ectopic FGF5 expression correlates with GNP expansion near the secondary fissure in Sufu LoF cerebella. Furthermore, pharmacological blockade of FGF signaling inhibits GNP proliferation. Interestingly, the data indicate that the timing of conditional Sufu deletion (E13.5 using the hGFAP-Cre line) results in different outcomes compared to later deletion (using Math1-cre line, Jiwani et al., 2020). This study provides significant insights into the molecular mechanisms driving GNP expansion in SHH subgroup MB, particularly in the context of Sufu LoF. It highlights the potential of targeting FGF5 signaling as a therapeutic strategy. Additionally, the research offers a model for better understanding MB subtypes and developing targeted treatments.

Strengths:

One notable strength of this study is the extraction and analysis of ectopic FGF5 expression from a subset of MB patient tumor samples. This translational aspect of the study enhances its relevance to human disease. By correlating findings from mouse models with patient data, the authors strengthen the validity of their conclusions and highlight the potential clinical implications of targeting FGF5 in MB therapy.

The data convincingly show that FGFR signaling activation drives GNP proliferation in Sufu conditional knockout models. This finding is supported by robust experimental evidence, including pharmacological blockade of FGF signaling, which effectively inhibits GNP proliferation. The clear demonstration of a functional link between FGFR signaling and GNP expansion underscores the potential of FGFR as a therapeutic target in SHH subgroup medulloblastoma.

Previous studies have demonstrated the inhibitory effect of FGF2 on tumor cell proliferation in certain MB types, such as the ptc mutant (Fogarty et al., 2006)(Yaguchi et al., 2009). Findings in this manuscript provide additional support suggesting multiple roles for FGF signaling in cerebellar patterning and development.

Weaknesses:

In the GEO dataset analysis, where FGF5 expression is extracted, the reporting of the P-value lacks detail on the statistical methods used, such as whether an ANOVA or t-test was employed. Providing comprehensive statistical methodologies is crucial for assessing the rigor and reproducibility of the results. The absence of this information raises concerns about the robustness of the statistical analysis.

Another concern is related to the controls used in the study. Cre recombinase induces double-strand DNA breaks within the loxP sites, and the control mice did not carry the Cre transgene (as stated in the Method section), while Sufu-cKO mice did. This discrepancy necessitates an additional control group to evaluate the effects of Cre-induced double-strand breaks on phosphorylated H2AX-DSB signaling. Including this control would strengthen the validity of the findings by ensuring that observed effects are not artifacts of Cre recombinase activity.

Although the use of the hGFAP-Cre line allows genetic access to late embryonic stage, this also targets multiple cell types, including both GNPs and cerebellar glial cells. However, the authors focus primarily on GNPs without fully addressing the potential contributions of neuron-glial interaction. This oversight could limit the understanding of the broader cellular context in which FGF signaling influences tumor development.

- Statistical analysis from the Geo expression dataset:<br /> The reviewer is satisfied with the revisions provided by the author and considers Figure 1 substantially improved.

- Generation of Sufu-cKO;Gli1-LacZ triple transgenic mice not described:<br /> >The reviewer finds that the supplementary Figure 1 revisions provided by the author do not fully address the concerns raised, and the issue remains unresolved.

- Request control group to evaluate the effects of Cre induced double-strand breaks on phosphorylated H2AX-DSB signaling:<br /> >Despite the revisions, control group (hGFAPCre;Sufu-fl/+) highlighted in the author response does not present in the revision, leaving this issue unresolved.

- hGFAP-Cre line also targets multiple celltypes, including both GNPs and cerebellar glial cells:<br /> >The author acknowledges the limitations of the study, and the reviewer concurs, noting that it enhances the contextual understanding of the findings.

eLife Assessment

Briola and co-authors determined the structure of the human CTF18 clamp loader bound to PCNA to high resolution, analyzed the structure, and tested a new mechanism involving a human-specific Ctf18 beta-hairpin docking onto Rfc5, which represents a valuable contribution. The data are solid and complement data recently published by others.

Reviewer #1 (Public review):

Summary:

The authors report the structure of the human CTF18-RFC complex bound to PCNA. Similar structures (and more) have been reported by the O'Donnell and Li labs. This study should add to our understanding of CTF18-RFC in DNA replication and clamp loaders in general. However, there are numerous major issues that I recommend the authors fix.

Strengths:

The structures reported are strong and useful for comparison with other clamp loader structures that have been reported lately.

Weaknesses:

The structures don't show how CTF18-RFC opens or loads PCNA. There are recent structures from other groups that do examine these steps in more detail, although this does not really dampen this reviewer's enthusiasm. It does mean that the authors should spend their time investigating aspects of CTF18-RFC function that were overlooked or not explored in detail in the competing papers. The paper poorly describes the interactions of CTF18-RFC with PCNA and the ATPase active sites, which are the main interest points. The nomenclature choices made by the authors make the manuscript very difficult to read.

Reviewer #2 (Public review):

Summary

Briola and co-authors have performed a structural analysis of the human CTF18 clamp loader bound to PCNA. The authors purified the complexes and formed a complex in solution. They used cryo-EM to determine the structure to high resolution. The complex assumed an auto-inhibited conformation, where DNA binding is blocked, which is of regulatory importance and suggests that additional factors could be required to support PCNA loading on DNA. The authors carefully analysed the structure and compared it to RFC and related structures.

Strength & Weakness

Their overall analysis is of high quality, and they identified, among other things, a human-specific beta-hairpin in Ctf18 that flexibly tethers Ctf18 to Rfc2-5. Indeed, deletion of the beta-hairpin resulted in reduced complex stability and a reduction in a primer extension assay with Pol ε. This is potentially very interesting, although some more work is needed on the quantification. Moreover, the authors argue that the Ctf18 ATP-binding domain assumes a more flexible organisation, but their visual representation could be improved.

The data are discussed accurately and relevantly, which provides an important framework for rationalising the results.

All in all, this is a high-quality manuscript that identifies a key intermediate in CTF18-dependent clamp loading.

Reviewer #3 (Public review):

Summary:

CTF18-RFC is an alternative eukaryotic PCNA sliding clamp loader that is thought to specialize in loading PCNA on the leading strand. Eukaryotic clamp loaders (RFC complexes) have an interchangeable large subunit that is responsible for their specialized functions. The authors show that the CTF18 large subunit has several features responsible for its weaker PCNA loading activity and that the resulting weakened stability of the complex is compensated by a novel beta hairpin backside hook. The authors show this hook is required for the optimal stability and activity of the complex.

Relevance:

The structural findings are important for understanding RFC enzymology and novel ways that the widespread class of AAA ATPases can be adapted to specialized functions. A better understanding of CTF18-RFC function will also provide clarity into aspects of DNA replication, cohesion establishment, and the DNA damage response.

Strengths:

The cryo-EM structures are of high quality enabling accurate modelling of the complex and providing a strong basis for analyzing differences and similarities with other RFC complexes.

Weaknesses:

The manuscript would have benefitted from more detailed biochemical analysis to tease apart the differences with the canonical RFC complex.

I'm not aware of using Mg depletion to trap active states of AAA ATPases. Perhaps the authors could provide a reference to successful examples of this and explain why they chose not to use the more standard practice in the field of using ATP analogues to increase the lifespan of reaction intermediates.

Overall appraisal:

Overall the work presented here is solid and important. The data is sufficient to support the stated conclusions and so I do not suggest any additional experiments.

eLife Assessment

This manuscript describes an important study of the giant virus Jyvaskylavirus. The characterisation presented is solid, although, in the current form, it is not clear to what extent these findings change our perception of how giant viruses, especially those isolated from a cold environment, function. The work will be of interest to virologists working on giant viruses as well as those working with other members of the PRD1/Adenoviridae lineage.

Reviewer #1 (Public review):

This study presents Jyvaskylavirus, a new member of the Marseilleviridae family, infecting Acanthamoeba castellanii. The study provides a detailed and comprehensive genomic and structural analysis of Jyvaskylavirus. The authors identified ORF142 as the capsid penton protein and additional structural proteins that comprise the virion. Using a combination of imaging techniques the authors provide new insights into the giant virus architecture and lifecycle. The study could be improved by providing atomic coordinates and refinement statistics, comparisons with available giant virus structures could be expanded, and the novelty in terms of the first isolated example of a giant virus from Finland could be expounded upon.

The study contributes new structural and genomic diversity to the Marseilleviridae family, hinting at a broader distribution and ecological significance of giant viruses than previously thought.

Reviewer #2 (Public review):

Summary:

This paper describes the molecular characterisation of a new isolate of the giant virus Jyvaskylavirus, a member of the Marseilleviridae family infecting Acanthamoeba castellanii. The isolate comes from a boreal environment in Finland, showcasing that giant viruses can thrive in this ecological niche. The authors came up with a non-trivial isolation procedure that can be applied to characterise other members of the family and will be beneficial for the virology field. The genome shows typical Marseilleviridae features and phylogenetically belongs to their clade B. The structural characterisation was performed on the level of isolated virion morphology by negative stain EM, virions associated with cells either during the attachment or release by helium microscopy, the visualisation of the virus assembly inside cells using stained thin sections, and lastly on the protein secondary structure level by reconstructing ~6 A icosahedral map of the massive virion using cryoEM. The cryoEM density combined with gene product structure prediction enabled the identification and functional assessment of various virion proteins.

Strengths:

The detailed description of the virus isolation protocol is the largest strength of the paper and this reviewer believes it can be modified for isolating various viruses infecting small eukaryotes. The cryoEM map allows us to understand how exceptionally large virions of these viruses are stabilised by minor capsid proteins and nicely demonstrates the integration of medium-resolution cryoEM with protein structure prediction in deciphering virion protein function. The visualisation of ongoing virus assembly inside virus factories brings interesting hypotheses about the process that; however, needs to be verified in the next studies.

Weaknesses:

The conclusions from helium microscopy images are overinterpreted, as the native membrane structure cannot be preserved in a fixed and dehydrated sample. In the image, there are many other parts of the curved membrane and a lot of virions, to me it seems the specific position of the highlighted virion could arise by a random chance. The claim that the cells were imaged in the near-original state by this method should be therefore omitted. Also, no mass spectrometry data are presented that would supplement and confirm the identity of virion proteins which predicted models were fitted into the cryoEM density. For a general virology reader outside of the giant virus field, the results presented in the current state might not have enough influence and the section should be rewritten to better showcase the novelty of findings.

eLife Assessment

This important study describes a novel flow-responsive gene and its role in regulating the inflammation-associated transcription factor IRF5. While the in vivo experiments are solid, the in vitro data is inadequate since embryonic fibroblasts are used throughout despite the work aiming to investigate mechanisms of endothelial cell activation in atherosclerosis.

Reviewer #1 (Public review):

Summary:

The authors report the role of a novel gene Aff3ir-ORF2 in flow-induced atherosclerosis. They show that the gene is anti-inflammatory in nature. It inhibits the IRF5-mediated athero-progression by inhibiting the causal factor (IRF5). Furthermore, the authors show a significant connection between shear stress and Aff3ir-ORF2 and its connection to IRF5 mediated athero-progression in different established mice models which further validates the ex vivo findings.

Strengths:

(1) An adequate number of replicates were used for this study.<br /> (2) Both in vitro and in vivo validation was done.<br /> (3) The figures are well presented.<br /> (4) In vivo causality is checked with cleverly designed experiments.

Weaknesses:

(1) Inflammatory proteins must be measured with standard methods e.g ELISA as mRNA level and protein level does not always correlate.

(2) RNA seq analysis has to be done very carefully. How does the euclidean distance correlate with the differential expression of genes. Do they represent the neighborhood? If they do how does this correlation affect the conclusion of the paper?

(3) The volcano plot does not indicate the q value of the shown genes. It is advisable to calculate the q value for each of the genes which represents the FDR probability of the identified genes.

(4) GO enrichment was done against the Global gene set or a local geneset? The authors should provide more detailed information about the analysis.

(5) If the analysis was performed against a global gene set. How does that connect with this specific atherosclerotic microenvironment?

(6) What was the basal expression of genes and how did the DGE (differential gene expression) values differ?

(7) How was IRF5 picked from GO analysis? was it within the 20 most significant genes?

(8) Microscopic studies should be done more carefully? There seems to be a global expression present on the vascular wall for Aff3ir-ORF2 and the expression seems to be similar to AFF3 in Figure 1.

Reviewer #2 (Public review):

Summary:

The authors recently uncovered a novel nested gene, Aff3ir, and this work sets out to study its function in endothelial cells further. Based on differences in expression correlating with areas of altered shear stress, they investigate a role for the isoform Aff3ir-ORF2 in endothelial activation and development of atherosclerosis downstream of disturbed shear stress. Using a knockout mouse model and in vivo overexpression experiments, they demonstrate a strong potential for Aff3ir-ORF2 to alleviate atherosclerosis. They find that Aff3ir-ORF2 interacts with the pro-inflammatory transcription factor IRF5 and retains it in the cytoplasm, hence preventing upregulation of inflammation-associated genes. The data expands our knowledge of IRF5 regulation which could be relevant to researchers studying various inflammatory diseases as well as adding to our understanding of atherosclerosis development.

Strengths:

The in vivo data is solid using immunofluorescence staining to assess AFF3ir-ORF2 expression, a knockout mouse model, overexpression and knockdown studies, and rescue experiments in combination with two atherosclerotic models to demonstrate that Aff3ir-ORF2 can lessen atherosclerotic plaque formation in ApoE-/- mice.

Weaknesses:

While the in vivo data is generally convincing, a few data panels have issues and will need addressing. Also, the knockout mouse model will need to be described, since the paper referred to in the manuscript does not actually report any knockout mouse model. Hence it is unclear how Aff3ir-ORF2 is targeted, but Figure S2B shows that targeting is partial, since about 30% expression remains at the RNA level in MEFs isolated from the knockout mice.

While the effect on atherosclerosis is clear, the conclusion that this is the result of reduced endothelial cell activation is not supported by the data. The mouse model is described as a global knockout and the shRNA knockdowns (Figure 5) and overexpression data in Figure 2 are not cell type-specific. Only the overexpression construct in Figure 6 uses an ICAM-2 promoter construct, which drives expression in endothelial cells, though leaky expression of this promoter has been reported in the literature. Therefore, other cell types such as smooth muscle cells or macrophages could be responsible for the effects observed.

The weakest part of the manuscript is the in vitro experiments. While they are solidly executed, all experiments are performed in MEFs, and results are interpreted as being equivalent to endothelial cell responses. There is also an RNA-seq experiment performed on MEFs from the Aff3ir-ORF2 knockout and control mice, but the data is not disclosed other than showing some non-identifiable expression differences. The data is used to hypothesise on a role for IRF5 in the effects observed with Aff3ir-ORF2 knockout.

Overall, the paper succeeds in demonstrating a link between Aff3ir-ORF2 and atherosclerosis, but the cell types involved and mechanisms remain unclear. The study also shows a functional interaction between Aff3ir-ORF2 and IRF5 in embryonic fibroblasts, but any relevance of this mechanism for atherosclerosis or any cell types involved in the development of this disease remains largely speculative.

Reviewer #3 (Public review):

This study is to demonstrate the role of Aff3ir-ORF2 in the atheroprone flow-induced EC dysfunction and ensuing atherosclerosis in mouse models. Overall, the data quality and comprehensiveness are convincing. In silico, in vitro, and in vivo experiments and several atherosclerosis were well executed. To strengthen further, the authors can address human EC relevance.

Major comments:

(1) The tissue source in Figures 1A and 1B should be clarified, the whole aortic segments or intima? If aortic segment was used, the authors should repeat the experiments using intima, due to the focus of the current study on the endothelium.

(2) Why were MEFs used exclusively in the in vitro experiments? Can the authors repeat some of the critical experiments in mouse or human ECs?

(3) The authors should explain why AFF3ir-ORF2 overexpression did not affect the basal level expression of ICAM-1, VCAM-1, IL-1b, and IL-6 under ST conditions (Figure 2A-C).

(4) Please include data from sham controls, i.e., right carotid artery in Figure 2E.

(5) Given that the merit of the study lies in the effect of different flow patterns, the legion areas in AA and TA (Figure 3B, 3C) should be separately compared.

(6) For confirmatory purposes for the variations of IRF5 and IRF8, can the authors mine available RNA-seq or even scRNA-seq data on human or mouse atherosclerosis? This approach is important and could complement the current results that are lacking EC data.

(7) With the efficacy of using AAV-ICAM2-AFF3ir-ORF2 in atherosclerosis reduction (Figure 6), the authors are encouraged to use lung ECs isolated from the AFF3ir-ORF2-/-mice to recapitulate its regulation of IRF5.

eLife Assessment

This important study demonstrates that screening by artificial intelligence can identify relevant novel compounds for interacting with KATP channels. The experimental work is compelling. The broader significance of this work relates to the possibility that KATP channel mutations linked to congenital hyperinsulinism may be effectively rescued to the cell surface with a drug, which could normalize insulin secretion or enhance the effectiveness of existing KATP channel activators such as diazoxide.

Reviewer #1 (Public review):

Summary:

Multiple compounds that inhibit ATP-sensitive potassium (KATP) channels also chaperone channels to the surface membrane. The authors used an artificial intelligence (AI)-based virtual screening (AtomNet) to identify novel compounds that exhibit chaperoning effects on trafficking-deficient disease-causing mutant channels. One compound, which they named Aekatperone, acts as a low affinity, reversible inhibitor and effective chaperone. A cryoEM structure of KATP bound to Aekatperone showed that the molecule binds at the canonical inhibitory site.

Strengths and weaknesses:

The details of the AI screening itself are inevitably opaque but appear to differ from classical virtual screening in not involving any physical docking of test compounds into the target site. The authors mention criteria that were used to limit the number of compounds so that those with high similarity to known binders and 'sequence identity' (does this mean structural identity) were excluded. The identified molecules contain sulfonylurea-like moieties. How different are they from other sulfonylure4as?

The experimental work confirming that Aekatperone acts to traffic mutant KATP channels to the surface and acts as a low affinity, reversible, inhibitor is comprehensive and clear, with very convincing cell biological and patch-clamp data, as is the cryoEM structural analysis, for which the group are leading experts. In addition to the three positive chaperone-effective molecules, the authors identified a large number of compounds that are predicted binders but apparently have no chaperoning effect. Did any of them have an inhibitory action on channels? If so, does this give clues to separating chaperoning from inhibitory effects?

The authors suggest that the novel compound may be a promising therapeutic for the treatment of congenital hyperinsulinism due to trafficking defective KATP mutations. Because they are low-affinity, reversible, inhibitors. This is a very interesting concept, and perhaps a pulsed dosing regimen would allow trafficking without constant channel inhibition (which otherwise defeats the therapeutic purpose), although it is unclear whether the new compound will offer advantages over earlier low-affinity sulfonylurea inhibitor chaperones. These include tolbutamide which has very similar affinity and effect to Aekatperone. As the authors point out this (as well as other sulfonylureas) is currently out of favor because of potential adverse cardiovascular effects, but again, it is unclear why Aekatperone should not have the same concerns.

Reviewer #2 (Public review):

Summary:

In their study 'AI-Based Discovery and CryoEM Structural Elucidation of a KATP Channel Pharmacochaperone', ElSheikh and colleagues undertake a computational screening approach to identify candidate drugs that may bind to an identified binding pocket in the SUR1 subunit of KATP channels. Other KATP channel inhibitors such as glibenclamide have been previously shown to bind in this pocket, and in addition to inhibition of KATP channel function, these inhibitors can very effectively rescue cell surface expression of trafficking deficient KATP mutations that cause excessive insulin secretion (Congenital Hyperinsulinism). However, a challenge for their utility for the treatment of hyperinsulinism has been that they are powerful inhibitors of the channels that are rescued to the channel surface. In contrast, successful therapeutic pharmacochaperones (eg. CFTR chaperones) permit the function of the channels rescued to the cell membrane. Thus, a key criterion for the authors' approach, in this case, was to identify relatively low-affinity compounds that target the glibenclamide binding site (and be washed off) - these could potentially rescue KATP surface expression but also permit KATP function.

Strengths:

The main findings of the manuscript include:

(1) Computational screening of a large virtual compound library, followed by functional screening of cell surface expression, which identified several potential candidate pharmacochaperones that target the glibenclamide binding site.

(2) Prioritization and functional characterization of Aekatperone as a low-affinity KATP inhibitor which can be readily 'washed off' in patch clamp and cell-based efflux assays. Thus the drug clearly rescues cell surface expression but can be manipulated experimentally to permit the function of rescued channels.

(3) Determination of the binding site and dynamics of this candidate drug by cryo-EM, and functional validation of several residues involved in drug sensitivity using mutagenesis and patch clamp.

The experiments are well-conceived and executed, and the study is clearly described. The results of the experiments are very straightforward and clearly support the conclusions drawn by the authors. I found the study to provide important new information about the KATP chaperone effects of certain drugs, with interesting considerations in terms of ion channel biology and human disease.

Weaknesses:

I don't have any major criticisms of the study as described, but I had some remaining questions that could be addressed in a revision.

(1) The chaperones can effectively rescue KATP trafficking mutants, but clearly not as strongly as the higher affinity inhibitor glibenclamide. Is this relationship between inhibitory potency, and efficacy of trafficking an intrinsic challenge of the approach? I suspect that it may be an intractable problem in the sense that the inhibitor-bound conformation that underlies the chaperone effect cannot be uncoupled from the inhibited gating state. But this might not be true (many partial agonist drugs with low efficacy can be strongly potent, for example). In this case, the approach is really to find a 'happy medium' of a drug that is a weak enough inhibitor to be washed away, but still strong enough to exert some satisfactory chaperone effect. Could some additional clarity be added in the discussion on whether the chaperone and gating effects can be 'uncoupled'?

(2) Based on the western blots in Figure 2B, the rescue of cell surface expression appears to require a higher concentration of AKP compared to the concentration-response of channel inhibition (~9 microM in Figure 3, perhaps even more potent in patch clamp in Figure 2C). Could the authors clarify/quantify the concentration response for trafficking rescue?

(3) A future challenge in the application of pharmacochaperones of this type in hyperinsulinism may be the manipulation of chaperone concentration in order to permit function. In experiments, it is straightforward to wash off the chaperone, but this would not be the case in an organism. I wondered if the authors had attempted to rescue channel function with diazoxide in the presence of AKP, rather than after washing off (ie. is AKP inhibition insurmountable, or can it be overcome by sufficient diazoxide).

(4) Do the authors have any information about the turnover time of KATP after the wash-off of the chaperone (how stable are the rescued channels at the cell surface)? This is a difficult question to probe when glibenclamide is used as a chaperone, but may be much simpler to address with a lower affinity chaperone like AKP.

eLife Assessment

In this important manuscript, the authors investigate the phospho-regulation of the C. elegans kinesin-2 motor protein OSM-3, revealing that the kinase, NEKL-3, phosphorylates a serine/threonine patch at the hinge region of the motor to mediate autoinhibition until it reaches the ciliary middle segment. The findings are supported by robust genetic data, in vivo imaging, and motility assays with wild-type and mutant motors, although the methods section lacks detailed protocols for NEKL-3 assays and in silico analyses. Overall, the study provides a solid contribution to understanding the regulation of OSM-3 kinesin activity.

Reviewer #1 (Public review):

Summary:

This manuscript is a focused investigation of the phosphor-regulation of a C. elegans kinesin-2 motor protein, OSM-3. In C-elegans sensory ciliary, kinesin-2 motor proteins Kinesin-II complex and OSM-3 homodimer transport IFT trains anterogradely to the ciliary tip. Kinesin-II carries OSM-3 as an inactive passenger from the ciliary base to the middle segment, where kinesin-II dissociates from IFT trains and OSM-3 gets activated and transports IFT trains to the distal segment. Therefore, activation/inactivation of OSM-3 plays an essential role in its ciliary function.

Strengths:

In this study, using mass spectrometry, the authors have shown that the NEKL-3 kinase phosphorylates a serine/threonine patch at the hinge region between coiled coils 1 and 2 of an OSM-3 dimer, referred to as the elbow region in ubiquitous kinesin-1. Phosphomimic mutants of these sites inhibit OSM-3 motility both in vitro and in vivo, suggesting that this phosphorylation is critical for the autoinhibition of the motor. Conversely, phospho-dead mutants of these sites hyperactivate OSM-3 motility in vitro and affect the localization of OSM3 in C. elegans. The authors also showed that Alanine to Tyrosine mutation of one of the phosphorylation rescues OS-3 function in live worms.

Weaknesses:

Collectively, this study presents evidence for the physiological role of OSM-3 elbow phosphorylation in its autoregulation, which affects ciliary localization and function of this motor. Overall, the work is well performed, and the results mostly support the conclusions of this manuscript. However, the work will benefit from additional experiments to further support conclusions and rule out alternative explanations, filling some logical gaps with new experimental evidence and in-text clarifications, and improving writing.

Reviewer #2 (Public review):

Summary:

The regulation of kinesin is fundamental to cellular morphogenesis. Previously, it has been shown that OSM-3, a kinesin required for intraflagellar transport (IFT), is regulated by autoinhibition. However, it remains totally elusive how the autoinhibition of OSM-3 is released. In this study, the authors have shown that NEKL-3 phosphorylates OSM-3 and releases its autoinhibition.

The authors found NEKL-3 directly phosphorylates OSM-3 (although the method is not described clearly) (Figure 1). The phophorylated residue is the "elbow" of OSM-3. The authors introduced phospho-dead (PD) and phospho-mimic (PM) mutations by genome editing and found that the OSM-3(PD) protein does not form cilia, and instead, accumulates to the axonal tips. The phenotype is similar to another constitutive active mutant of OSM-3, OSM-3(G444A) (Imanishi et al., 2006; Xie et al., 2024). osm-3(PM) has shorter cilia, which resembles with loss of function mutants of osm-3 (Figure 3). The authors did structural prediction and showed that G444E and PD mutations change the conformation of OSM-3 protein (Figure 3). In the single-molecule assays G444E and PD mutations exhibited increased landing rate (Figure 4). By unbiased genetic screening, the authors identified a suppressor mutant of osm-3(PD), in which A489T occurs. The result confirms the importance of this residue. Based on these results, the authors suggest that NEKL-3 induces phosphorylation of the elbow domain and inactivates OSM-3 motor when the motor is synthesized in the cell body. This regulation is essential for proper cilia formation.

Strengths:

The finding is interesting and gives new insight into how the IFT motor is regulated.

Weaknesses:

The methods section has not presented sufficient information to reproduce this study.

eLife Assessment

This important study provides a framework for applying single-cell transcriptome data and network analysis from genetically diverse mouse cells to identify novel driver genes underlying the role of genetic loci associated with bone mineral density. The evidence supporting the identification of the driver genes and the conclusion of the paper is convincing. Overall, this approach may be broadly applicable and of interest to researchers investigating the genetics of complex diseases.

Reviewer #1 (Public review):

In this manuscript, Dillard and colleagues integrate cross-species genomic data with a systems approach to identify potential driver genes underlying human GWAS loci and establish the cell type(s) within which these genes act and potentially drive disease. Specifically, they utilize a large single-cell RNA-seq (scRNA-seq) dataset from an osteogenic cell culture model - bone marrow-derived stromal cells cultured under osteogenic conditions (BMSC-OBs) - from a genetically diverse outbred mouse population called the Diversity Outbred (DO) stock to discover network driver genes that likely underlie human bone mineral density (BMD) GWAS loci. The DO mice segregate over 40M single nucleotide variants, many of which affect gene expression levels, therefore making this an ideal population for systems genetic and co-expression analyses. The current study builds on previously published work from the same group that used co-expression analysis to identify co-expressed "modules" of genes that were enriched for BMD GWAS associations. In this study, the authors utilize a much larger scRNA-seq dataset from 80 DO BMSC-OBs, infer co-expression-based and Bayesian networks for each identified mesenchymal cell type, focused on networks with dynamic expression trajectories that are most likely driving differentiation of BMSC-OBs, and then prioritized genes ("differentiation driver genes" or DDGs) in these osteogenic differentiation networks that had known expression or splicing QTLs (eQTL/sQTLs) in any GTEx tissue that colocalized with human BMD GWAS loci. The systems analysis is impressive, the experimental methods are described in detail, and the experiments appear to be carefully done. The computational analysis of the single-cell data is comprehensive and thorough, and the evidence presented in support of the identified DDGs, including Tpx2 and Fgfrl1, is for the most part convincing. Some limitations in the data resources and methods hamper enthusiasm somewhat and are discussed below. Overall, while this study will no doubt be valuable to the BMD community, the cross-species data integration and analytical framework may be more valuable and generally applicable to the study of other diseases, especially for diseases with robust human GWAS data but for which robust human genomic data in relevant cell types is lacking.

Specific strengths of the study include the large scRNA-seq dataset on BMSC-OBs from 80 DO mice, the clustering analysis to identify specific cell types and sub-types, the comparison of cell type frequencies across the DO mice, and the CELLECT analysis to prioritize cell clusters that are enriched for BMD heritability (Figure 1). The network analysis pipeline outlined in Figure 2 is also a strength, as is the pseudotime trajectory analysis (results in Figure 3). One weakness involves the focus on genes that were previously identified as having an eQTL or sQTL in any GTEx tissue. The authors rightly point out that the GTEx database does not contain data for bone tissue, but the reason that eQTLs can be shared across many tissues - this assumption is valid for many cis-eQTLs, but it could also exclude many genes as potential DDGs with effects that are specific to bone/osteoblasts. Indeed, the authors show that important BMD driver genes have cell-type-specific eQTLs. Furthermore, the mesenchymal cell type-specific co-expression analysis by iterative WGCNA identified an average of 76 co-expression modules per cell cluster (range 26-153). Based on the limited number of genes that are detected as expressed in a given cell due to sparse per-cell read depth (400-6200 reads/cell) and dropouts, it's hard to believe that as many as 153 co-expression modules could be distinguished within any cell cluster. I would suspect some degree of model overfitting here and would expect that many/most of these identified modules have very few gene members, but the methods list a minimum module size of 20 genes. How do the numbers of modules identified in this study compare to other published scRNA-seq studies that use iterative WGCNA?

In the section "Identification of differentiation driver genes (DDGs)", the authors identified 408 significant DDGs and found that 49 (12%) were reported by the International Mouse Knockout [sic] Consortium (IMPC) as having a significant effect on whole-body BMD when knocked out in mice. Is this enrichment significant? E.g., what is the background percentage of IMPC gene knockouts that show an effect on whole-body BMD? Similarly, they found that 21 of the 408 DDGs were genes that have BMD GWAS associations that colocalize with GTEx eQTLs/sQTLs. Given that there are > 1,000 BMD GWAS associations, is this enrichment (21/408) significant? Recommend performing a hypergeometric test to provide statistical context to the reported overlaps here.

Reviewer #2 (Public review):

Summary:

In this manuscript, Farber and colleagues have performed single-cell RNAseq analysis on bone marrow-derived stem cells from DO Mice. By performing network analysis, they look for driver genes that are associated with bone mineral density GWAS associations. They identify two genes as potential candidates to showcase the utility of this approach.

Strengths:

The study is very thorough and the approach is innovative and exciting. The manuscript contains some interesting data relating to how cell differentiation is occurring and the effects of genetics on this process. The section looking for genes with eQTLs that differ across the differentiation trajectory (Figure 4) was particularly exciting.

Weaknesses:

The manuscript is in parts hard to read due to the use of acronyms and there are some questions about data analysis that need to be addressed.

eLife Assessment

The present study described GEARBOCS, an adeno-associated virus tool for in vivo gene editing in astrocytes, which is both timely and of importance for glial biologists, who often are troubled by efficient gene targeting in astrocytes. Overall, the finding is valuable, and the strength of the evidence is solid. Presumably, there will be great potential associated with GEARBOCS applications in the future.

Reviewer #1 (Public review):

Summary:

The manuscript by Bindu et al. created an AAV-based tool (GEARAOCS) to perform in vivo genome editing of mouse astrocytes. The authors engineered a versatile AAV vector that allows for gene deletion through NHNJ, site-specific knock-in by HDR, and gene trap. By utilizing this tool, the authors deleted Sparcl1 virally in subsets of astrocytes and showed that thalamocortical synapses in cortical layer IV are indeed reduced during a critical period of ocular dominance plasticity and in adulthood, whereas there is no change in excitatory synapse number in cortical layer II/III. Furthermore, the authors made a VAMP2 gene-trap AAV vector and showed that astrocyte-derived VAMP2 is required for the maintenance of both excitatory and inhibitory synapses.

Strengths:

This AAV-based tool is versatile for astrocytic gene manipulation in vivo. The work is innovative and exciting, given the paucity of tools available to probe astrocytes in vivo.

Weaknesses:

Several important considerations need to be made for the validation and usage of this tool, including:

Major points:

(1) Efficiency and specificity of spCas9-sgRNA mediated gene knockout in astrocytes. In Figure 3, the authors utilized Sparcl1 gene deletion as the proof-of-principle experiment. The readout for Sparcl1 KO efficiency is solely the immunoreactivity using an antibody raised against Sparcl1. As the method is based on NHEJ, the indels can be diverse and can occur in one allele or two. For the tool and proof-of-principle experiment, it will be important to know the percentage of editing near the PAM site, as well as the actual sequences of indels. This can be done by single-cell PCR of edited astrocytes, similar to the published work (Ye... Chen, Nature Biotechnology 2019).

(2) Along the same line, the authors showed that GEARBOCS TagIn of Sparcl1 resulted in 12.49% efficiency based on the immunohistochemistry of mCherry tag. It is understandable that the knock-in efficiency is much reduced as compared to gene knockout. However, it remains unclear if those 12.49% knock-in cells represent sequence-correct ones, as spCas9-mediated HDR is also an error-prone process, and it may accidentally alter nucleotides near the PAM site without causing the frameshift. The author will need to consider the related evidence or make comments in the discussion.

(3) What are the efficiencies of Sparcl1 GEARBOCS GeneTrap (Figure 3V) and Vamp2 GeneTrap and HA TagIn (Figure 5)?

Minor points:

(1) Figure 3H-J. The authors only showed the representative images of Sparcl1 KO. Please consider including the control (without gRNA), given that there are still many Sparcl1+ signals in Figure 3I (likely because of its expression in other cell types?).

(2) In figure 3Q-T, it appears that some Cas9-EGFP+ astrocytes (Q) do not express Sparcl1 (R). Is Sparcl1 expressed in subsets of astrocytes? Does Cas9-EGFP or Sparcl1-TagIn alter Sparcl1 endogenous expression?

(3) On Page 8, for the explanation of the design of the GEARBOCS construct, the authors have made a self-citation (#43). That was a BioRxiv paper that is being reviewed currently.

(4) For Figures 4 and 6, the graphs seem to be made in R with the x-axis labeled as "Condition". The y-axis labels are too small to read properly, especially in print. It would be better to make the graphs clearer like Figure 2 and Figure 3.

(5) On Page 13, "Figures 3V-Y" were referred to. However, there are no Figures 3W, X, and Y.

(6) There are a few typos in the manuscript, including line 900 "immunofluorescence microscopy images of a Cas9-EGFP-positive astrocytes (green)".

Reviewer #2 (Public review):

Summary:

The present study described GEARBOCS, an adeno-associated virus tool for in vivo gene editing in astrocytes. This tool is timely and important for glial biologists who often are troubled by efficient gene targeting in astrocytes. Overall the significance of the finding is valuable, and the strength of the evidence is solid. Presumably, there will be great potential associated with GEARBOCS applications in the future.

Strengths:

As efficient tools for targeting non-neuronal cells in the brains are rather limited for astrocytes and microglia, GEARBOCS adds to the small pool of currently available tools and will provide new options for glial biologists studying these tools. As the study revealed, GEARBOCS are capable of knockout and knockin manipulations for genes of interest, also ascribed with reporter tracking and gene-trap strategy. The promising multi-functional tool will advance our understanding of astrocytes and help to further elucidate the mechanism of neuron-glia interaction.

Weaknesses:

Even though the tool seems promising and powerful. the authors failed to provide more evidence on the robustness and specificity of GEARBOCS. Also, the advantages of GEARBOCS over some of the traditional methods were not clearly stated. Some of these concerns are described below.

Reviewer #3 (Public review):

Summary:

Sivadasan Bindu et al. developed a CRISPR/Cas9-based gene-editing strategy using a single AAV vector, named GEARBOCS (Gene Editing in AstRocytes Based On CRISPR/Cas9 System), which enables precise genome manipulation in astrocytes. This tool was shown to effectively perform knockout, tagging, and reporter knock-in gene modifications. The utility of GEARBOCS was demonstrated in two cases: establishing astrocytes as essential for the synaptogenic factor Sparcl1 in thalamocortical synapse maintenance, and revealing that cortical astrocytes express the Vamp2 protein, which is vital for maintaining synapse numbers.

Strengths:

Astrocytes play a crucial role in brain development and function, but studying them in vivo has been challenging due to limited molecular tools for manipulation. Sivadasan Bindu et al. developed a valuable system called GEARBOCS for effective astrocyte infection via retro-orbital injection.

Weaknesses:

The manuscript provides data only from the cerebral cortex and results from P42. Additional data from other brain regions and various time points (e.g., P0-15) are needed. Results from local injection experiments would also enhance the utility of this tool for the broader glial research community.

eLife Assessment

This is an important study that describes the development of optical biosensors for various Rab GTPases and explores the contributions of Rab10 and Rab4 to structural and functional plasticity at hippocampal synapses during glutamate uncaging. Most of the evidence supporting the conclusions of the paper is solid, while the evidence supporting the finding that Rab10 activation during structural LTP is sustained is incomplete due to the characterization of the relevant sensor.

Reviewer #1 (Public review):

Summary:

Wang et al. created a series of specific FLIM-FRET sensors to measure the activity of different Rab proteins in small cellular compartments. They apply the new sensors to monitor Rab activity in dendritic spines during induction of LTP. They find sustained (30 min) inactivation of Rab10 and transient (5 min) activation of Rab4 after glutamate uncaging in zero Mg. NMDAR function and CaMKII activation are required for these effects. Knockdown of Rab4 reduced spine volume change while knockdown of Rab10 boosted it and enhanced functional LTP (in KO mice). To test Rab effects on AMPA receptor exocytosis, the authors performed FRAP of fluorescently labeled GluA1 subunits in the plasma membrane. Within 2-3 min, new AMPARs appear on the surface via exocytosis. This process is accelerated by Rab10 knock-down and slowed by Rab4 knock-down. The authors conclude that CaMKII promotes AMPAR exocytosis by i) activating Rab4, the exocytosis driver and ii) inhibiting Rab10, possibly involved in AMPAR degradation.

Strengths:

The work is a technical tour de force, adding fundamental insights to our understanding of the crucial functions of different Rab proteins in promoting/preventing synaptic plasticity. The complexity of compartmentalized Ras signaling is poorly understood and this study makes substantial inroads. The new sensors are thoroughly characterized, seem to work very well, and will be quite useful for the neuroscience community and beyond (e.g. cancer research). The use of FLIM for read-out is compelling for precise activity measurements in rapidly expanding compartments (i.e., spines during LTP).

Weaknesses:

The interpretation of the FRAP experiments (Figure 5, Ext. Data Figure 13) is not straightforward as spine volume and surface area greatly expand during uncaging. I appreciate the correction for the added spine membrane shown in Extended Data Figure 14i, but shouldn't this be a correction factor (multiplication) derived from the volume increase instead of a subtraction?

Also, experiments were not conducted or analyzed blind, risking bias in the selection/exclusion of experiments for analysis. This reduces my confidence in the results.

Reviewer #2 (Public review):

Summary:

Wang et al. developed a set of optical sensors to monitor Rab protein activity. Their investigation into Rab activity in dendritic spines during structural long-term plasticity (sLTP) revealed sustained Rab10 inactivation (>30min) and transient Rab4 activation (~5 min). Through pharmacological and genetic manipulation to constitutively activate or inhibit Rab proteins, they found that Rab10 negatively regulates sLTP and AMPA receptor insertion, while Rab4 positively influences sLTP but only in the transient phase. The optical sensors provide new tools for studying Rab activity in cells and neurobiology. However, a full understanding of the timing of Rab activity will require a detailed characterization of sensor kinetics.

Strengths:

(1) Introduction of a series of novel sensors that can address numerous questions in Rab biology.

(2) Multiple methods to manipulate Rab proteins to reveal the roles of Rab10 and rab4 in LTP.

(3) Discovery of Rab4 activation and Rab10 inhibition with different kinetics during sLTP, correlating with their functional roles in the transient (Rab4) and both transient and sustained (Rab10) phases of sLTP.

Weaknesses:

(1) Lack of characterization of sensor kinetics, making it difficult to determine if the observed Rab kinetics during sLTP were due to sensor behavior or actual Rab activity.

(2) It is crucial to assess whether the overexpression of Rab proteins as reporters, affects Rab activity and cellular structure and physiology (e.g. spine number and size).

(3) The paper does not explain the apparently different results between NMDA receptor activation and glutamate uncaging. NMDA receptor activation increased Rab10 activity, while glutamate uncaging decreased it. NMDA receptor activation resulted in sustained Rab4 activation, whereas glutamate uncaging caused only brief activation of about 5 minutes. A potential explanation, ideally supported by data, is needed.

(4) There is a discrepancy between spine phenotype and sLTP potential with Rab10 perturbation. Rab10 perturbation affected spine density but not size, suggesting a role in spinogenesis rather than sLTP. However, glutamate uncaging affected sLTP, and spinogenesis was not examined. Explaining the discrepancy between spine size and sLTP potential is necessary. Exploring spinogenesis with glutamate uncaging would strengthen these results. Additionally, Figure 4j shows no change in synaptic transmission with Rab10 knockout, despite an increase in spine density. An explanation, ideally supported by data, is needed for the unchanged fEPSP slope despite an increase in spine density.

(5) Spine volume was imaged using acceptor fluorophores (mCherry, or mCherry/Venus) at 920nm, where the two-photon cross-section of mCherry is minimal. 920nm was also used to excite the donor fluorophore, hence the spine volume measurement based on total red channel fluorescence is the sum of minimal mCherry fluorescence from direct 920nm excitation, bleed-through from the green channel, and FRET. This confounded measurement requires correction and clarification.

Reviewer #3 (Public review):

Summary:

This study examines the roles of Rab10 and Rab4 proteins in structural long-term potentiation (sLTP) and AMPA receptor (AMPAR) trafficking in hippocampal dendritic spines using various different methods and organotypic slice cultures as the biological model.

The paper shows that Rab10 inactivation enhances AMPAR insertion and dendritic spine head volume increase during sLTP, while Rab4 supports the initial stages of these processes. The key contribution of this study is identifying Rab10 inactivation as a previously unknown facilitator of AMPAR insertion and spine growth, acting as a brake on sLTP when active. Rab4 and Rab10 seem to be playing opposing roles, suggesting a somewhat coordinated mechanism that precisely controls synaptic potentiation, with Rab4 facilitating early changes and Rab10 restricting the extent and timing of synaptic strengthening.

Strengths:

The study combines multiple techniques such as FRET/FLIM imaging, pharmacology, genetic manipulations, and electrophysiology to dissect the roles of Rab10 and Rab4 in sLTP. The authors developed highly sensitive FRET/FLIM-based sensors to monitor Rab protein activity in single dendritic spines. This allowed them to study the spatiotemporal dynamics of Rab10 and Rab4 activity during glutamate uncaging-induced sLTP. They also developed various controls to ensure the specificity of their observations. For example, they used a false acceptor sensor to verify the specificity of the Rab10 sensor response.

This study reveals previously unknown roles for Rab10 and Rab4 in synaptic plasticity, showing their opposing functions in regulating AMPAR trafficking and spine structural plasticity during LTP.

Weaknesses:

In sLTP, the initial volume of stimulated spines is an important determinant of induced plasticity. To address changes in initial volume and those induced by uncaging, the authors present Extended Data Figure 2. In my view, the methods of fitting, sample selection, or both may pose significant limitations for interpreting the overall results. While the initial spine size distribution for Rab10 experiments spans ~0.1-0.4 fL (with an unusually large single spine at the upper end), Rab4 spine distribution spans a broader range of ~0.1-0.9 fL. If the authors applied initial size-matched data selection or used polynomials rather than linear fitting, panels a, b, e, f, and g might display a different pattern. In that case, clustering analysis based on initial size may be necessary to enable a fair comparison between groups not only for this figure but also for main Figures 2 and 3.

Another limitation is the absence of in vivo validation, as the experiments were performed in organotypic hippocampal slices, which may not fully replicate the complexity of synaptic plasticity in an intact brain, where excitatory and inhibitory processes occur concurrently. High concentrations of MNI-glutamate (4 mM in this study) are known to block GABAergic responses due to its antagonistic effect on GABA-A receptors, thereby precluding the study of inhibitory network activity or connectivity [1], which is already known to be altered in organotypic slice cultures.

[1] https://www.frontiersin.org/journals/neural-circuits/articles/10.3389/neuro.04.002.2009/full

eLife Assessment

This study provides a valuable set of analyses and theoretical derivations to understand the mechanisms used by recurrent neural networks (RNNs) to perform context-dependent accumulation of evidence. The novelty of some of the findings needs clarification, and additional details need to be provided for some of the analyses. However, the results regarding the dimensionality and neural dynamical signatures of RNNs are solid and provide new avenues to study the mechanisms underlying context-dependent computations.

Reviewer #1 (Public review):

Summary:

This paper investigates how recurrent neural networks (RNNs) can perform context-dependent decision-making (CDM). The authors use low-rank RNN modeling and focus on a CDM task where subjects are presented with sequences of auditory pulses that vary in location and frequency, and they must determine either the prevalent location or frequency based on an external context signal. In particular, the authors focus on the problem of differentiating between two distinct selection mechanisms: input modulation, which involves altering the stimulus input representation, and selection vector modulation, which involves altering the "selection vector" of the dynamical system.

First, the authors show that rank-one networks can only implement input modulation and that higher-rank networks are required for selection vector modulation. Then, the authors use pathway-based information flow analysis to understand how information is routed to the accumulator based on context. This analysis allows the authors to introduce a novel definition of selection vector modulation that explicitly links it to changes in the effective coupling along specific pathways within the network.

The study further generates testable predictions for differentiating selection vector modulation from input modulation based on neural dynamics. In particular, the authors find that:<br /> (1) A larger proportion of selection vector modulation is expected in networks with high-dimensional connectivity.<br /> (2) Single-neuron response kernels exhibiting specific profiles (peaking between stimulus onset and choice onset) are indicative of neural dynamics in extra dimensions, supporting the presence of selection vector modulation.<br /> (3) The percentage of explained variance (PEV) of extra dynamical modes extracted from response kernels at the population level can serve as an index to quantify the amount of selection vector modulation.

Strengths:

The paper is clear and well-written, and it draws bridges between two recent important approaches in the study of CDM: circuit-level descriptions of low-rank RNNs, and differentiation across alternative mechanisms in terms of neural dynamics. The most interesting aspect of the study involves establishing a link between selection vector modulation, network dimensionality, and dimensionality of neural dynamics. The high correlation between the networks' mechanisms and their dimensionality (Figure 7d) is surprising since differentiating between selection mechanisms is generally a difficult task, and the strength of this result is further corroborated by its consistency across multiple RNN hyperparameters (Figure 7-Figure Supplement 1 and Figure 7-figure supplement 2). Interestingly, the correlation between the selection mechanism and the dimensionality of neural dynamics is also high (Figure 7g), potentially providing a promising future avenue for the study of neural recordings in this task.

Weaknesses:

The first part of the manuscript is not particularly novel, and it would be beneficial to clearly state which aspects of the analyses and derivations are different from previous literature. For example, the derivation that rank-1 RNNs cannot implement selection vector modulation is already present in the Extended Discussion of Pagan et al., 2022 (Equations 42-43). Similarly, it would be helpful to more clearly explain how the proposed pathway-based information flow analysis differs from the circuit diagram of latent dynamics in Dubreuil et al., 2022.

With regard to the results linking selection vector modulation and dimensionality, more work is required to understand the generality of these results, and how practical it would be to apply this type of analysis to neural recordings. For example, it is possible to build a network that uses input modulation and to greatly increase the dimensionality of the network simply by adding additional dimensions that do not directly contribute to the computation. Similarly, neural responses might have additional high-dimensional activity unrelated to the task. My understanding is that the currently proposed method would classify such networks incorrectly, and it is reasonable to imagine that the dimensionality of activity in high-order brain regions will be strongly dependent on activity that does not relate to this task.

Finally, a number of aspects of the analysis are not clear. The most important element to clarify is how the authors quantify the "proportion of selection vector modulation" in vanilla RNNs (Figures 7d and 7g). I could not find information about this in the Methods, yet this is a critical element of the study results. In Mante et al., 2013 and in Pagan et al., 2022 this was done by analyzing the RNN linearized dynamics around fixed points: is this the approach used also in this study? Also, how are the authors producing the trial-averaged analyses shown in Figures 2f and 3f? The methods used to produce this type of plot differ in Mante et al., 2013 and Pagan et al., 2022, and it is necessary for the authors to explain how this was computed in this case.

I am also confused by a number of analyses done to verify mathematical derivations, which seem to suggest that the results are close to identical, but not exactly identical. For example, in the histogram in Figure 6b, or the histogram in Figure 7-figure supplement 3d: what is the source of the small variability leading to some of the indices being less than 1?

Reviewer #2 (Public review):

This manuscript examines network mechanisms that allow networks of neurons to perform context-dependent decision-making.

In a recent study, Pagan and colleagues identified two distinct mechanisms by which recurrent neural networks can perform such computations. They termed these two mechanisms input-modulation and selection-vector modulation. Pagan and colleagues demonstrated that recurrent neural networks can be trained to implement combinations of these two mechanisms, and related this range of computational strategies with inter-individual variability in rats performing the same task. What type of structure in the recurrent connectivity favors one or the other mechanism however remained an open question.

The present manuscript addresses this specific question by using a class of mechanistically interpretable recurrent neural networks, low-rank RNNs.

The manuscript starts by demonstrating that unit-rank RNNs can only implement the input-modulation mechanism, but not the selection-vector modulation. The authors then build rank three networks that implement selection-vector modulation and show how the two mechanisms can be combined. Finally, they relate the amount of selection-vector modulation with the effective rank, ie the dimensionality of activity, of a trained full-rank RNN.

Strengths:

(1) The manuscript is written in a straightforward manner.<br /> (2) The analytic approach adopted in the manuscript is impressive.<br /> (3) Very clear identification of the mechanisms leading to the two types of context-dependent modulation.<br /> (4) Altogether this manuscript reports remarkable insights into a very timely question.

Weaknesses:

- The introduction could have been written in a more accessible manner for any non-expert readers.

eLife Assessment

This valuable study introduces a novel method for controlling generalization and interference in neural networks undergoing continual learning. The authors provide solid evidence that their parsimonious method performs better than online gradient descent in several continual learning situations while providing biologically plausible links to three-factor learning rules. However, empirical validation is limited to linear networks, which raises questions about the generality of the results in non-linear networks. While the work is interesting to theoretical and experimental neuroscientists, improving the article presentation by clearly defining terminology before using it and providing more details on the setup of the simulation experiments would be vital to make the article more accessible.

Reviewer #1 (Public review):

Summary:

This paper advances a new understanding of plasticity in artificial neural networks. It shows that weight changes can be decomposed into two components: the first governs the magnitude (or gain) of responses in a particular layer; the second governs the relationship of those responses to the input to that layer. Then, it shows that separate control of these two factors via a surprise-based metaplasticity can avoid catastrophic forgetting as well as induce successful generalization in different conditions, through a series of simulation experiments in linear networks. The authors argue that separate control of the two factors may be at work in the brain and may underlie the ability of humans and other animals to perform successful sequential learning. The paper is hampered by confusing terminology and the precise setup of some of the simulations is unclear. The paper also focuses exclusively on the linear case, which limits confidence in the generality of the results. The paper would also benefit from the inclusion of specific predictions for neural data that would confirm the idea that the separate control of these two factors underlies successful continual learning in the brain.

Strengths:

(1) The theoretical framework developed by the paper is interesting, and could have wide applicability for both training networks and for understanding plasticity.

(2) The simulations convincingly show benefits to the coordinated eligibility model of plasticity advanced by the authors.

Weaknesses:

(1) The simulation results are limited to simple tasks in linear networks, it would be interesting to see how the intuitions developed in the linear case extend to nonlinear networks.

(2) The terminology is somewhat confusing and this can make the paper difficult to follow in some places.

(3) The details of some of the simulations are lacking.

Reviewer #2 (Public review):

Summary:

Scott and Frank propose a new method for controlling generalization and interference in neural networks that undergo continual learning. Their method called coordinated eligibility models (CEM), relies on the factorization of synaptic updates into input-driven and output-driving factors. They subsequently employ the fact that it is sufficient to orthogonalize any one of these two factors across different data points to nullify the interference during learning. They exemplify this on a number of toy tasks while comparing their result to vanilla gradient.

Strengths:

The specific mechanism proposed here is novel (while, as authors acknowledge, there is a large number of other mechanisms for the selective recruitment of synapses for the prevention of catastrophic forgetting). Furthermore, it is simple, elegant, and to a large extent biologically plausible, potentially pointing to specific and testable aspects of learning dynamics.

Weaknesses:

(1) Scope and toy nature of experiments: the model was only applied to very simple problems tailored specifically to demonstrate the strengths of the CEM method. Furthermore, single hyperparameter setting is presented for every scenario which leaves it questionable how general the numerical results are. The selection of input, output dimensionality and data set size also seems to be underexplored. Will a larger curriculum, smaller or larger dimension, compromise any of the CEM ingredients? Restriction to linear models seems arbitrary (it should be a no-time test to add non-linearity within a pytorch framework that authors used), and applicability for any non-synthetic problem is not obvious.

It is also unclear to what extent of domain knowledge is needed for surprise signals to be successfully generated. Can the authors make a stronger case about novel curriculum entries being easily recognizable by cosine distance, either in the brain or in machine learning? Can they alternatively demonstrate their method on a less toy benchmark (e.g. permuted MNIST from Kirkpatrick et al 2017 that they cite)?

Another limitation is that unlike smoother models of plasticity budgets (e.g. Kirkpatrick et al 17, Zenke et al 17), here eligibility seems to be lost forever, once surprise is applied. What happens to the model if more data from a previously visited task becomes available? Will the system be able to continue learning within the right context and how does CEM perform compared to other catastrophic-forgetting-prevention strategies?

(2) The clarity and organization must be improved. Specifically, the balance between verbal descriptions, equations, figures, and their captions needs to be improved. For example - two full-size equations are dedicated to the application of linear regression (around lines 183 and 236) while by far less obvious math such as settings for fig 7, including 'feature loadings', 'demands', etc., is presented in a hardly readable mixture figure and main text. Similarly, the surprise mechanism which is a key ingredient for the model is presented in a very non-straightforward fashion, scattered between the main text, figure, and methods. The figure legends are poorly informative in many cases as well (see minor comments for examples).

Reviewer #3 (Public review):

Summary:

This paper describes a modification of gradient descent learning, and shows in several simulations that this modification allows online learning of linear regression problems where naive gradient descent fails. The modification starts from the observation that the rank-1 weight update of online gradient learning can be written as the outer product Δw ∝ g xᵀ of a vector g and the input x. Modifying this update rule, by projecting g or x to some subspaces, i.e. Δw ∝ Pg (Qx)ᵀ, allows for preventing the typical catastrophic forgetting behavior of online gradient descent, as confirmed in the simulations. The projection matrices P and Q are updated with a "surprise"-modulation rule.

Strengths:

I find it interesting to explore the benefits of alternatives to naive online gradient learning for continual learning.

Weaknesses:

The novelty and advancement in our theoretical understanding of plasticity in neural systems are unclear. I appreciate gaining insights from simple mathematical arguments and simulations with toy models, but for this paper, I do not yet clearly see what I learned: on the mathematical/ML/simulation side it is unclear how it relates to the continual learning literature, on the neuroscience/surprise side I see only a number of papers cited but not any clear connection to data or novel insights.

More specifically:

(1) It is unclear what exactly the "coordinated eligibility theory" is. Is any update rule that satisfies Equation 4 included in the coordinated eligibility theory? If yes, what is the point: any update rule can be written in this way, including standard online gradient descent. If no, what is it? It is not Equation 5 it seems, because this is called "one of the simplest coordinated eligibility models".

(2) There is a lot of work on continual learning which is not discussed, e.g. "Orthogonal Gradient Descent for Continual Learning" (Farajtabar et al. 2019), "Continual learning in low-rank orthogonal subspaces" (Chaudhry et al. 2020), or "Keep Moving: identifying task-relevant subspaces to maximise plasticity for newly learned tasks" (Anthes et al. 2024), to name just a few. What is the novelty of this work relative to these existing works? Is the novelty in the specific projection operator? If yes, what are the benefits of this projection operator in theory and simulations? How would, for example, the approach of Farajtabar et al. 2019 perform on the tasks in Figures 3-7?

(3) There is also work on using surprise signals for multitask learning in models of biological neural networks, e.g. "Fast adaptation to rule switching using neuronal surprise" (Barry et al. 2023).

(4) What is the motivation for the projection to the unit sphere in Equation 5?

(5) What is the motivation for the surprise definition? For example, why cos(x⋅μ) = cos(|x||μ|cos(θ)) = cos(cos(θ))? (Assuming x and μ have unit length and θ is the angle between x and μ).

eLife Assessment

This study presents an important contribution to the understanding of neural speech tracking, demonstrating how minimal background noise can enhance the neural tracking of the amplitude-onset envelope. The evidence supporting the claims of the author is solid, through a well-designed series of EEG experiments. This work will be of interest to auditory scientists, particularly those investigating biological markers of speech processing.

Reviewer #1 (Public review):

This paper presents a comprehensive study of how neural tracking of speech is affected by background noise. Using five EEG experiments and Temporal response function (TRF), it investigates how minimal background noise can enhance speech tracking even when speech intelligibility remains very high. The results suggest that this enhancement is not attention-driven but could be explained by stochastic resonance. These findings generalize across different background noise types and listening conditions, offering insights into speech processing in real-world environments.

I find this paper well-written, the experiments and results are clearly described. However, I have a few comments that may be useful to address.

(1) The behavioral accuracy and EEG results for clear speech in Experiment 4 differ from those of Experiments 1-3. Could the author provide insights into the potential reasons for this discrepancy? Might it be due to linguistic/ acoustic differences between the passages used in experiments? If so, what was the rationale behind using different passages across different experiments?

(2) Regarding peak amplitude extraction, why were the exact peak amplitudes and latencies of the TRFs for each subject not extracted, and instead, an amplitude average within a 20 ms time window based on the group-averaged TRFs used? Did the latencies significantly differ across different SNR conditions?

(3) How is neural tracking quantified in the current study? Does improved neural tracking correlate with EEG prediction accuracy or individual peak amplitudes? Given the differing trends between N1 and P2 peaks in babble and speech-matched noise in experiment 3, how is it that babble results in greater envelope tracking compared to speech-matched noise?

(4) The paper discusses how speech envelope-onset tracking varies with different background noises. Does the author expect similar trends for speech envelope tracking as well? Additionally, could you explain why envelope onsets were prioritized over envelope tracking in this analysis?

Reviewer #2 (Public review):

The author investigates the role of background noise on EEG-assessed speech tracking in a series of five experiments. In the first experiment, the influence of different degrees of background noise is investigated and enhanced speech tracking for minimal noise levels is found. The following four experiments explore different potential influences on this effect, such as attentional allocation, different noise types, and presentation mode.

The step-wise exploration of potential contributors to the effect of enhanced speech tracking for minimal background noise is compelling. The motivation and reasoning for the different studies are clear and logical and therefore easy to follow. The results are discussed in a concise and clear way. While I specifically like the conciseness, one inevitable consequence is that not all results are equally discussed in depth.

Based on the results of the five experiments, the author concludes that the enhancement of speech tracking for minimal background noise is likely due to stochastic resonance. Given broad conceptualizations of stochastic resonance as a noise benefit this is a reasonable conclusion.

This study will likely impact the field as it provides compelling support questioning the relationship between speech tracking and speech processing.

eLife Assessment

This important study identifies neurotrophin signaling as a molecular mechanism underlying previous findings of structural plasticity in central dopaminergic neurons of the adult fly brain. The authors present solid evidence for neurotrophin signaling in shaping the structure and synapses of certain dopaminergic circuits. The work suggests an intriguing potential link between neurotrophin signaling and experience-induced structural plasticity but further research will be necessary to establish this connection definitively.