Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public Review):

Summary:

In this study, Osiurak and colleagues investigate the neurocognitive basis of technical reasoning. They use multiple tasks from two neuroimaging studies and overlap analysis to show that the area PF is central for reasoning, and plays an essential role in tool-use and non-tool-use physical problem-solving, as well as both conditions of mentalizing task. They also demonstrate the specificity of the technical reasoning and find that the area PF is not involved in the fluid-cognition task or the mentalizing network (INT+PHYS vs. PHYS-only). This work suggests an understanding of the neurocognitive basis of technical reasoning that supports advanced technologies.

Strengths:

-The topic this study focuses on is intriguing and can help us understand the neurocognitive processes involved in technical reasoning and advanced technologies.

-The researchers obtained fMRI data from multiple tasks. The data is rich and encompasses the mechanical problem-solving task, psychotechnical task, fluid-cognition task, and mentalizing task.

-The article is well written.

We sincerely thank Reviewer 1 for their positive and very helpful comments, which helped us improve the MS. Thank you.

Weaknesses:

- Limitations of the overlap analysis method: there are multiple reasons why two tasks might activate the same brain regions. For instance, the two tasks might share cognitive mechanisms, the activated regions of the two tasks might be adjacent but not overlapping at finer resolutions, or the tasks might recruit the same regions for different cognition functions.

Thus, although overlap analysis can provide valuable information, it also has limitations.

Further analyses that capture the common cognitive components of activation across different

tasks are warranted, such as correlating the activation across different tasks within subjects for a region of interest (i.e. the PF).

We thank Reviewer 1 for this comment. We added new analyses to address the two alternative interpretations stressed here by Reviewer 1, namely, the same-region-but-differentfonction interpretation and the adjacency interpretation. The new analyses ruled out both alternative interpretations, thereby reinforcing our interpretation.

“The conjunction analysis reported was subject to at least two key limitations that needed to be overcome to assure a correct interpretation of our findings. The first was that the tasks could recruit the same regions for different cognition functions (same-region-but-different-function interpretation). The second was that the activated regions of the different tasks could be adjacent but did not overlap at finer resolutions (adjacency interpretation). We tested the same-region-but-different-function interpretation by conducting additional ROI analyses, which consisted of correlating the specific activation of the left area PF (i.e., difference in terms of mean Blood-Oxygen Level Dependent [BOLD] parameter estimates between the experimental condition minus the control condition) in the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. This analysis did not include the mechanical problem-solving task because the sample of participants was not the same for this task. As shown in Fig. 5, we found significant correlations between all the tasks that were hypothesized as recruiting technical reasoning, i.e., the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .05). By contrast, no significant correlation was obtained between these three tasks and the fluid-cognition task (all p > .15). This finding invalidates the same-region-but-different-function interpretation by revealing a coherent pattern in the activation of the left area PF in situations in which participants were supposed to reason technically. We examined the adjacency interpretation by analysing the specific locations of individual peak activations within the left area PF ROI for the mechanical problemsolving task, the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. These peaks, which corresponded to the maximum value of activation obtained for each participant within the left area PF ROI, are reported in Fig. 6. As can be seen, the peaks of the fluid-cognition task were located more anteriorly, in the left area PFt (Parietal Ft) and the postcentral cortex, compared to the peaks of the other four tasks, which were more posterior, in the left area PF. Statistical analyses based on the y coordinates of the individual activation peaks confirmed this description (Fig. 6). Indeed, the y coordinates of the peaks of the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task were posterior to the y coordinates of the peaks of the fluid-cognition task (all p < .05), whereas no significant differences were reported between the four tasks (all p > .05). These findings speak against the adjacency interpretation by revealing that participants recruited the same part of the left area PF to perform tasks involving technical reasoning.” (p. 11-13)

Control tasks may be inadequate: the tasks may involve other factors, such as motor/ actionrelated information. For the psychotechnical task, fluid-cognition task, and mentalizing task, the experiment tasks need not only care about technical-cognition information but also motor-related information, whereas the control tasks do not need to consider motor-related information (mainly visual shape information). Additionally, there may be no difference in motor-related information between the conditions of the fluid-cognition task. Therefore, the regions of interest may be sensitive to motor-related information, affecting the research conclusion.

We thank Reviewer 1 for this comment. We added a specific section in the discussion that addresses this limitation.

“The second limitation concerns the alternative interpretation that the left area PF is not central to technical reasoning but to the storage of sensorimotor programs about the prototypical manipulation of common tools. Here we show that the left area PF is recruited even in situations in which participants do not have to process common manipulable tools. For instance, some items of the psychotechnical task consisted of pictures of tractor, boat, pulley, or cannon. The fact that we found a common activation of the left area PF in such tasks as well as in the mechanical problem-solving task, in which participants could nevertheless simulate the motor actions of manipulating novel tools, indicates that this brain area is not central to tool manipulation but to physical understanding. That being said, some may suggest that viewing a boat or a cannon is enough to incite the simulation of motor actions, so our tasks were not equipped to distinguish between the manipulation-based approach and the reasoning-based approach. We have already shown that the left area PF is more involved in tasks that focus on the mechanical dimension of the tool-use action (e.g., the mechanical interaction between a tool and an object) than its motor dimension (i.e., the interaction between the tool and the effector [e.g., 24, 40]). Nevertheless, we recognize that future research is still needed to test the predictions derived from these two approaches.” (p. 18-19)

-Negative results require further validation: the cognitive results for the fluid-cognition task in the study may need more refinement. For instance, when performing ROI analysis, are there any differences between the conditions? Bayesian statistics might also be helpful to account for the negative results.

We agree that our negative results required further validation. We conducted the ROI analyses suggested by Reviewer 1, which confirmed the initial whole-brain analyses.

“Region of interest (ROI) results. We conducted additional analyses to test the robustness of our findings. One of our results was that we did not report any specific activation of the left area PF in the fluid-cognition task contrary to the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. However, this negative result needed exploration at the ROI level. Therefore, we created a spherical ROI of the left area PF with a radius of 12 mm in the MNI standard space (–59; –31; 40). This ROI was literature-defined to ensure the independence of its selection (40). ROI results are shown in Fig. 4. The analyses confirmed the results obtained with the whole-brain analyses by indicating a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .35).” (p. 10-11)

Reviewer #1 (Recommendations For The Authors):

(1) I may not fully grasp some of the arguments. In the abstract, what does the term "intermediate-level" mean, and why is it an intermediate-level state? In the sentence "the existence of a specific cognitive module in the human brain dedicated to materiality", I cannot see a clear link between technical cognition and the word "materiality".

We used the term materiality to refer to a potential human trait that allows us to shape the physical world according to our ends, by using, making tools and transmiting them to others. This is a reference to Allen et al. (2020; PNAS): “We hope this empirical domain and modeling framework can provide the foundations for future research on this quintessentially human trait: using, making, and reasoning about tools and more generally shaping the physical world to our ends” (p. 29309). Scientists (including archaeologists, economists, psychologists, neuroscientists) interested in human materiality have tended to focus on how we manipulate things according to our thought (motor cognition) or how we conceptualize our behaviour to transmit it to others (language, social cognition). However, little has been said on the intermediate level, that is, technical cognition. We added the term “technical cognition” here, which should help to make the connection more quickly.

“Yet, little has been said about the intermediate-level cognitive processes that are directly involved in mastering this materiality, that is, technical cognition.” (p. 2)

(2) The introduction could provide more details on why the issue of "generalizability and specificity" is important to address, to clarify the significance of the research question.

We followed this comment and added a sentence to explain why it is important to address this research question. Again, we thank Reviewer 1 for their helpful comments.

“Here we focus on two key aspects of the technical-reasoning hypothesis that remain to be addressed: Generalizability and specificity. If technical reasoning is a specific form of reasoning oriented towards the physical world, then it should be implicated in all (the generalizability question) and only (the specificity question) the situations in which we need to think about the physical properties of our world.” (p. 5)

Reviewer #2 (Public Review):

Summary:

The goal of this project was to test the hypothesis that a common neuroanatomic substrate in the left inferior parietal lobule (area PF) underlies reasoning about the physical properties of actions and objects. Four functional MRI (fMRI) experiments were created to test this hypothesis. Group contrast maps were then obtained for each task, and overlap among the tasks was computed at the voxel level. The principal finding is that the left PF exhibited differentially greater BOLD response in tasks requiring participants to reason about the physical properties of actions and objects (referred to as technical reasoning). In contrast, there was no differential BOLD response in the left PF when participants engaged in fMRI variant of the Raven's progressive matrices to assess fluid cognition.

Strengths:

This is a well-written manuscript that builds from extensive prior work from this group mapping the brain areas and cognitive mechanisms underlying object manipulation, technical reasoning, and problem-solving. Major strengths of this manuscript include the use of control conditions to demonstrate there are differentially greater BOLD responses in area PF over and above the baseline condition of each task. Another strength is the demonstration that area PF is not responsive in tasks assessing fluid cognition - e.g., it may just be that PF responds to a greater extent in a harder condition relative to an easy condition of a task. The analysis of data from Task 3 rules out this alternative interpretation. The methods and analysis are sufficiently written for others to replicate the study, and the materials and code for data analysis are publicly available.

We sincerely thank Reviewer 2 for their precious comments, which helped us improve the MS. 

Weaknesses:

The first weakness is that the conclusions of the manuscript rely on there being overlap among group-level contrast maps presented in Figure 2. The problem with this conclusion is that different participants engaged in different tasks. Never is an analysis performed to demonstrate that the PF region identified in e.g., participant 1 in Task 2 is the same PF region identified in Participant 1 in Task 4.

We added new analyses that demonstrated that “the PF region identified in e.g., participant 1 in Task 2 is the same PF region identified in Participant 1 in Task 4”. We thank Reviewer 2 for this comment, because these new analyses reinforced our interpretation.

“The conjunction analysis reported was subject to at least two key limitations that needed to be overcome to assure a correct interpretation of our findings. The first was that the tasks could recruit the same regions for different cognition functions (same-region-but-different-function interpretation). The second was that the activated regions of the different tasks could be adjacent but did not overlap at finer resolutions (adjacency interpretation). We tested the same-region-but-different-function interpretation by conducting additional ROI analyses, which consisted of correlating the specific activation of the left area PF (i.e., difference in terms of mean Blood-Oxygen Level Dependent [BOLD] parameter estimates between the experimental condition minus the control condition) in the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. This analysis did not include the mechanical problem-solving task because the sample of participants was not the same for this task. As shown in Fig. 5, we found significant correlations between all the tasks that were hypothesized as recruiting technical reasoning, i.e., the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .05). By contrast, no significant correlation was obtained between these three tasks and the fluid-cognition task (all p > .15). This finding invalidates the same-region-but-different-function interpretation by revealing a coherent pattern in the activation of the left area PF in situations in which participants were supposed to reason technically. We examined the adjacency interpretation by analysing the specific locations of individual peak activations within the left area PF ROI for the mechanical problemsolving task, the psychotechnical task, the fluid-cognition task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. These peaks, which corresponded to the maximum value of activation obtained for each participant within the left area PF ROI, are reported in Fig. 6. As can be seen, the peaks of the fluid-cognition task were located more anteriorly, in the left area PFt (Parietal Ft) and the postcentral cortex, compared to the peaks of the other four tasks, which were more posterior, in the left area PF. Statistical analyses based on the y coordinates of the individual activation peaks confirmed this description (Fig. 6). Indeed, the y coordinates of the peaks of the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task were posterior to the y coordinates of the peaks of the fluid-cognition task (all p < .05), whereas no significant differences were reported between the four tasks (all p > .05). These findings speak against the adjacency interpretation by revealing that participants recruited the same part of the left area PF to perform tasks involving technical reasoning.” (p. 11-13)

A second weakness is that there is a variance in accuracy between tasks that are not addressed. It is clear from the plots in the supplemental materials that some participants score below chance (~ 50%). This means that half (or more) of the fMRI trials of some participants are incorrect. The methods section does not mention how inaccurate trials were handled. Moreover, if 50% is chance, it suggests that some participants did not understand task instructions and were systematically selecting the incorrect item.

It is true that the experimental conditions were more difficult than the control conditions, with some participants who performed at or below 50% in the experimental conditions. We added a section in the MS to stress this aspect. To examine whether this potential difficulty effect biased our interpretation, we conducted new ROI analyses by removing all the participants who performed at or below the chance level. These analyses revealed the same results as when no participant was excluded, suggesting that this did not bias our interpretation.

“As mentioned above, the experimental conditions of all the tasks were more difficult than their control conditions. As a result, the specific activation of the left area PF documented above could simply reflect that this area responds to a greater extent in a harder condition relative to an easy condition of a task. This interpretation is nevertheless ruled out by the results obtained with the fluid-cognition task. We did not report a specific activation of the left area PF in this task while its experimental condition was more difficult than its control condition. To test more directly this effect of difficulty, we conducted new ROI analyses by removing all the participants who performed at or below 50% (Fig. S2). These new analyses replicated the initial analyses by showing a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .48). In sum, the ROI analyses corroborated the wholebrain analyses and ruled out the potential effect of difficulty.” (p. 11)

A third weakness is related to the fluid cognition task. In the fMRI task developed here, the participant must press a left or right button to select between 2 rows of 3 stimuli while only one of the 3 stimuli is the correct target. This means that within a 10-second window, the participant must identify the pattern in the 3x3 grid and then separately discriminate among 6 possible shapes to find the matching stimulus. This is a hard task that is qualitatively different from the other tasks in terms of the content being manipulated and the time constraints.

We acknowledge that the fluid-cognition task involved a design that differed from the other tasks. However, this was also true for the other tasks, as the design also differed between the mechanical problem-solving task, the psychotechnical task, and the mentalizing task. Nevertheless, despite these distinctions, we found a consistent activation of the left area PF in these tasks with different designs including in the psychotechnical task, which seemed as difficult as the fluid-cognition task.

“Region of interest (ROI) results. We conducted additional analyses to test the robustness of our findings. One of our results was that we did not report any specific activation of the left area PF in the fluid-cognition task contrary to the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task. However, this negative result needed exploration at the ROI level. Therefore, we created a spherical ROI of the left area PF with a radius of 12 mm in the MNI standard space (–59; –31; 40). This ROI was literature-defined to ensure the independence of its selection (40). ROI results are shown in Fig. 4. The analyses confirmed the results obtained with the whole-brain analyses by indicating a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .35).” (p. 10-11)

In sum, this is an interesting study that tests a neuro-cognitive model whereby the left PF forms a key node in a network of brain regions supporting technical reasoning for tool and non-tool-based tasks. Localizing area PF at the level of single participants and managing variance in accuracy is critically important before testing the proposed hypotheses.

We thank Reviewer 2 for this positive evaluation and their suggestions. As detailed in our response, our revision took into consideration both the localization of the left area PF at the level of single participants and the variance in accuracy. 

Reviewer #2 (Recommendations For The Authors):

Did the fMRI data undergo high-pass temporal filtering prior to modeling the effects of interest? Participants engaged in a long (17-24 minutes) run of fMRI data collection. Highpass filtering of the data is critically important when managing temporal autocorrelation in the fMRI response (e.g., see Shinn et al., 2023, Functional brain networks reflect spatial and temporal autocorrelation. Nature Neuroscience).

Yes. We added this information.

“Regressors of non-interest resulting from 3D head motion estimation (x, y, z translation and three axes of rotation) and a set of cosine regressors for high-pass filtering were added to the design matrix.” (p. 25-26)

Including scales in Figure 2 would help the reader interpret the magnitude of the BOLD effects.

We added this information in Figure 3 (Figure 2 in the initial version of the MS).

It was difficult to inspect the small thumbnail images of the task stimuli in Figure 1. Higher resolution versions of those stimuli would help facilitate understanding of the task design and trial structure.

We changed both Figure 1 and Figure S1.

Reviewer #3 (Public Review):

Summary:

This manuscript reports two neuroimaging experiments assessing commonalities and differences in activation loci across mechanical problem-solving, technical reasoning, fluid cognition, and "mentalizing" tasks. Each task includes a control task. Conjunction analyses are performed to identify regions in common across tasks. As Area PF (a part of the supramarginal gyrus of the inferior parietal lobe) is involved across 3 of the 4 tasks, the investigators claim that it is the hub of technical cognition.

Strengths:

The aim of finding commonalities and differences across related problem-solving tasks is a useful and interesting one.

The experimental tasks themselves appear relatively well-thought-out, aside from the concern that they are differentially difficult.

The imaging pipeline appears appropriate.

We thank Reviewer 3 for their constructive comments, which helped us improve the MS.

Weaknesses:

(1) Methodological

As indicated in the supplementary tables and figures, the experimental tasks employed differ markedly in 1) difficulty and 2) experimental trial time. Response latencies are not reported (but are of additional concern given the variance in difficulty). There is concern that at least some of the differences in activation patterns across tasks are the result of these fundamental differences in how hard various brain regions have to work to solve the tasks and/or how much of the trial epoch is actually consumed by "on-task" behavior. These difficulty issues should be controlled for by 1) separating correct and incorrect trials, and 2) for correct trials, entering response latency as a regressor in the Generalized Linear Models, 3) entering trial duration in the GLMs.

We thank Reviewer 3 for this comment. It is true that the experimental conditions were more difficult than the control conditions, with some participants who performed at or below 50% in the experimental conditions. We added a section in the MS to stress this aspect. We could not conduct new analyses by separating correct and incorrect trials because, for each task, participants had to respond only on the last item of the block. Therefore, we did not record a response for each event. Nevertheless, we could examine whether this potential difficulty effect biased our interpretation, by conducting new ROI analyses in which we removed all the participants who performed at or below the chance level. These analyses revealed the same results as when no participant was excluded, suggesting that this did not bias our interpretation. 

“As mentioned above, the experimental conditions of all the tasks were more difficult than their control conditions. As a result, the specific activation of the left area PF documented above could simply reflect that this area responds to a greater extent in a harder condition relative to an easy condition of a task. This interpretation is nevertheless ruled out by the results obtained with the fluid-cognition task. We did not report a specific activation of the left area PF in this task while its experimental condition was more difficult than its control condition. To test more directly this effect of difficulty, we conducted new ROI analyses by removing all the participants who performed at or below 50% (Fig. S2). These new analyses replicated the initial analyses by showing a greater activation of the left area PF in the mechanical problem-solving task, the psychotechnical task, and the PHYS-Only and INT+PHYS conditions of the mentalizing task (all p < .001), but not in the fluid-cognition task (p \= .48). In sum, the ROI analyses corroborated the wholebrain analyses and ruled out the potential effect of difficulty.” (p. 11)

A related concern is that the control tasks also differ markedly in the degree to which they were easier and faster than their corresponding experimental task. Thus, some of the control tasks seem to control much better for difficulty and time on task than others. For example, the control task for the psychotechnical task simply requires the indication of which array contains a simple square shape (i.e., it is much easier than the psychotechnical task), whereas the control task for mechanical problem-solving requires mentally fitting a shape into a design, much like solving a jigsaw puzzle (i.e., it is only slightly easier than the experimental task).

It is true that some control conditions could be easier than other ones. These differences reinforced the common activation found in the left area PF in the tasks hypothesized as involving technical reasoning, because this activation survived irrespective of the differences in terms of experimental design. For us, the rationale is the same as for a meta-analysis, in which we try to find what is common to a great variety of tasks. The only detrimental consequence we identified here is that this difference explained why we did not report a specific activation of the left area PF in the fluid-cognition task, as if the left area PF was more responsive when the task was difficult. This possibility assumes that the experimental condition of the fluid-cognition task is much more difficult than its control condition compared to what can be seen in the other tasks. As Reviewer 2 stressed in Point 1, this interpretation is unlikely, because the differences between the experimental and control conditions were similar to the fluid-cognition task in the mechanical problem-solving and psychotechnical tasks. In addition, again, the new ROI analyses in which we removed all the participants who performed at or below the chance level in expetimental conditions reproduced our initital results.

(2) Theoretical 

The investigators seem to overlook prior research that does not support their perspective and their writing seems to lack scientific objectivity in places. At times they over-reach in the claims that can be made based on the present data. Some claims need to be revised/softened.

As this comment is also mentioned below, please find our response to it below.

Reviewer #3 (Recommendations For The Authors):

(1) Because of the high level of detail, Figures 1 and S2 (particularly the mentalizing task and mechanical problem-solving task, and their controls) are very hard to parse, even when examined relatively closely. It is suggested that these figures be broken down into separate panels for Experiment 1 and Experiment 2 to facilitate understanding.

We changed both Figure 1 and Figure S1.

(2) The behavioral data (including response latencies) should be reported in the main results section of the paper and not in a supplement.

The behavioural data are now reported in the main results. We did not report response latencies because participants were not prompted to respond as quickly as possible.

“Behavioural results. All the behavioural results are given in Fig. 2. As shown, scores were higher in the experimental conditions than for the control conditions for all the tasks (all p < .05). In other words, the experimental conditions were more difficult than the control conditions. This difference in terms of difficulty can also be illustrated by the fact that some participants performed at or below the chance level in the experimental conditions whereas none did so in the control conditions.” (p. 8)

(3) The investigators seem to overlook prior research that does not support their perspective and their writing seems to lack scientific objectivity in places. At times they over-reach in the claims that can be made based on the present data. For example, claims that need to be revised/softened include:

Abstract: "Area PF... can work along with social-cognitive skills to resolve day-to-day interactions that combine social and physical constraints". This statement is overly speculative.

This statement is based on the fact that we reported a combined activation of the technical-reasoning network and the mentalizing network in the INT+PHYS condition of the mentalizing task. This suggests that both networks need to work together for solving a day-today problem in which both the physical constraints of the situation and the intention of the individual must be integrated. Our findings replicated previous ones with a similar task (e.g., Brunet et al. 2000; Völlm et al., 2006), in which the authors gave an interpretation similar to ours in considering that this task requires understanding physical and social causes. Perhaps that the reference to the results of the mentalizing task was not explicit enough. We added “dayto-day” before “problem” in the part of the discussion in which we discuss this possibility to make this aspect clearer.

“In broad terms, the results of the mentalizing task indicate that causal reasoning has distinct forms and that it recruits distinct networks of the human brain (Social domain: Mentalizing; Physical domain: Technical reasoning), which can nevertheless interact together to solve day-to-day problems in which several domains are involved, such as in the INT+PHYS condition of the mentalizing task.” (p. 16)

Introduction: "The manipulation-based approach... remains silent on the more general cognitive mechanisms...that must also encompass the use of unfamiliar or novel tools". This statement seems to be based on an overly selective literature review. There are a number of studies in which the relationship between a novel and familiar tool selection/use has been explored (e.g., Buchman & Randerath, 2017; Mizelle & Wheaton, 2010; Silveri & Ciccarelli, 2009; Stoll, Finkel et al., 2022; Foerster, 2023; Foerster, Borghi, & Goslin, 2020; Seidel, Rijntjes et al., 2023).

We thank Reviewer 3 for this comment. Even if we accept the idea that we possess specific sensorimotor programs about tool manipulation, it remains that these programs cannot explain how an individual decides to bend a wire to make a hook or to pour water in a recipient to retrieve a target. As a matter of fact, such behaviour has been reported in nonhuman animals, such as crows (Weir et al., 2002, Nature) or orangutans (Mendes et al., 2007, Biology Letters). In these studies, the question is whether these nonhuman animals understand the physical causes or not, but the question of sensorimotor programs is never addressed (to our knowledge). This is also true in developmental studies on tool use (e.g., Beck et al., 2011, Cognition; Cutting et al., 2011, Journal of Experimental Child Psychology). This is what we meant here, that is, the manipulation-based approach is not equipped to explain how people solve physical problems by using or making tools – or any object – or by building constructions or producing technical innovations. However, we agree that some papers have been interested in exploring the link between common and novel tool use and have suggested that both could recruit common sensorimotor programs. It is noteworthy that these studies do not test the predictions from the manipulation-based approach versus the reasoning-based approach, so both interpretations are generally viable as stressed by Seidel et al. (2023), one of the papers recommended by Reviewer 3.

“Apparently, the presentation of a graspable object that is recognizable as a tool is sufficient to provoke SMG activation, whether one tends to see the function of SMG to be either “technical reasoning” (Osiurak and Badets 2016; Reynaud et al. 2016; Lesourd et al. 2018; Reynaud et al. 2019) or “manipulation knowledge” (Sakreida et al. 2016; Buxbaum 2017; Garcea et al. 2019b).” (Seidel et al., 2023; p. 9)

Regardless, as suggested by Reviewer 3, these papers deserve to be cited and this part needed to be rewritten to insist on the “making, construction, and innovation” dimension more than on the “unfamiliar and novel tool use” dimension to avoid any ambiguity.

“This manipulation-based approach has provided interesting insights (12–16) and even elegant attempts to explain how these sensorimotor programs could support the use of both unfamiliar or novel tools (17–20), but remains silent on the more general cognitive mechanisms behind human technology that include the use of common and unfamiliar or novel tools but must also encompass tool making, construction behaviour, technical innovations, and transmission of technical content.” (p. 3)

Introduction: "Here we focus on two important questions... to promote the technicalreasoning hypothesis as a comprehensive cognitive framework..."(italics added). This and other similar statements should be rewritten as testable scientific hypotheses rather than implying that the point of the research is to promote the investigators' preferred view.

We agree that our phrasing could seem inappropriate here. What we meant here is that the technical-reasoning hypothesis could become an interesting framework for the study of the cognitive bases of human technology only if we are able to verify some of its key facets. As suggested, we rewrote this part. We also rewrote the abstract and the first paragraph of the discussion.

“Here we focus on two key aspects of the technical-reasoning hypothesis that remain to be addressed: Generalizability and specificity. If technical reasoning is a specific form of reasoning oriented towards the physical world, then it should be implicated in all (the generalizability question) and only (the specificity question) the situations in which we need to think about the physical properties of our world.” (p. 5)

Introduction: The Goldenberg and Hagmann paper cited actually shows that familiar tool use may be based either on retrieval from semantic memory or by inferring function from structure (mechanical problem solving); in other words, the investigators saw a role for both kinds of information, and the relationship between mechanical problem solving and familiar tool use was actually relatively weak. This requires correction.

We disagree with Reviewer 3 on this point. The whole sentence is as follows:

“This silence has been initially broken by a series of studies initiated by Goldenberg and Hagmann (9), which has documented a behavioural link in left brain-damaged patients between common tool use and the ability to solve mechanical problems by using and even sometimes making novel tools (e.g., extracting a target out from a box by bending a wire to create a hook) (9, 17).” (p. 3-4)

We did not mention the interpretations given by Goldenberg and Hagmann about the link with the pantomime task, but only focused on the link they reported between common tool use and novel tool use. This is factual. In addition, we also disagree that the link between common tool use and novel tool use was weak.

“The hypothesis put forward in the introduction predicts that knowledge about prototypical tool use assessed by pantomime of tool use and the ability to infer function from structure assessed by novel tool selection can both contribute to the use of familiar tools. Indeed results of both tests correlated signicantly with the use of familiar tools pantomime of tool use: r \= 0.77, novel tool selection: r \= 0.62; both P < 0.001), but there was also a signicant correlation between the two tests r \= 0.64, P < 0.001).” (Goldenberg & Hagmann, 1998; p. 585)

As can be seen in this quote, they reported a significant correlation between novel tool selection and the use of familiar tools. It is also noteworthy that the novel tool selection test and the pantomime test correlated together. Georg Goldenberg told one of the authors (F. Osiurak; personal communication) that this result incited him to revise its idea that pantomime could assess “semantic knowledge”, which explains why he did not use it again as a measure of semantic knowledge. Instead, he preferred to use a classical semantic matching task in his 2009 Brain paper with Josef Spatt, in which they found a clearer dissociation between semantic knowledge and common/novel tool use not only at the behavioral level but also at the cerebral level.

Introduction: Please expand and clarify this sentence "However, this involvement seems to be task-dependent, contrary to the systematic involvement of left are PF. The IFG and LOTC activations observed in prior studies are of interest as well. Were they indeed all taskdependent in these studies?

We agree that this sentence is confusing. We meant that, in the studies reported just above in the paragraph, these regions were not systematically reported contrary to the left area PF. As we think that this information was not crucial for the logic of the paper, we preferred to remove it. 

Introduction: If implicit mechanical knowledge is acquired through interactions with objects, how is that implicit knowledge conveyed to pass on the material culture to others?

We thank Reviewer 3 for this comment. Although mechanical knowledge is implicit, it can be indirectly transmitted to other individuals, as shown in two papers we published in Nature Human Behaviour (Osiurak et al., 2021) and Science Advances (Osiurak et al., 2022). Actually, verbal teaching is not the only way to transmit information. There are many other ways of transmitting information such as gestural teaching (e.g., pointing the important aspects of a task to make them salient to the learner), observation without teaching (i.e., when we observe someone unbeknown to them) or reverse engineering (i.e., scrutinizing an artifact made by someone else). We have shown that even in reverse-engineering conditions, participants can benefit from what previous participants have done to increase their understanding of a physical system. In other words, all these forms of transmission allow the learners to understand new physical relationships without waiting that these relationships randomly occur in the environment. There is a wide literature on social learning, which describes very well how knowledge can be transmitted without using explicit communication. In fact, it is very likely that such forms of transmission were already present in our ancestors, allowing them to start accumulating knowledge without using symbolic language. We did not add this information in the MS because we think that this was a little bit beyond the scope of the MS. Nevetheless, we cited relevant literature on the topic to help the reader find it if interested in the topic.

“Yet, recent accounts have proposed that non-social cognitive skills such as causal understanding or technical reasoning might have played a crucial role in cumulative technological culture (6, 29, 66). Support for these accounts comes from micro-society experiments, which have demonstrated that the improvement of technology over generations is accompanied by an increase in its understanding (67, 68), or that learners’ technical-reasoning skills are a good predictor of cumulative performance in such micro-societies (33, 69).” (p. 19)

What distinguishes this implicit mechanical knowledge from stored knowledge about object manipulation? Are these two conceptualizations really demonstrably (testably) different?

We agree that it is complex to distinguish between these two hypotheses as suggested by Seidel et al. (2023) cited above (see Reviewer 3 Point 8). We have conducted several studies to test the opposite predictions derived from each hypothesis. The main distinction concerns the understanding of physical materials and forces, which is central to the technical-reasoning hypothesis but not to the manipulation-based approach. Indeed, sensorimotor programs about tool manipulation are not assumed to contain information about physical materials and forces. In the present study, the understanding of physical materials and forces was needed in the four tasks hypothesized as requiring technical reasoning, i.e., the mechanical problem-solving task, the psychotechnical task and the PHYS-Only and INT+PHYS conditions of the mentalizing task. We can illustrate this aspect with items of each of these tasks. Figure 1A is of the mechanical problem-solving task. 

As explained in the MS, participants had memorized the five possible tools before the scanner session. Thus, for 4 seconds, they had to imagine which of these tools could be used to extract the target out from the box. We did so to incit them to reason about mechanical solutions based on the physical properties of the problem. Then, they had 3 seconds to select the tool with the appropriate shape, here the right one. In this case, the motor action remains the same (i.e., pulling). Another illustration can be given, with the psychotechnical task (Figure 1B).

In this task, the participant had to reason as to whether the boat-tractor connection was better in the left picture or in the right picture. This needs to reason about physical forces, but there is no need to recruit sensorimotor programs about tool manipulation. Finally, a last example can be given with the PHYS-Only condition of the mentalizing task (but the logic is the same for the INT+PHYS condition except that the character’s intentions must also be taken into consideration) Figure 1D).

Here the participant must reason about which picture shows what is physically possible. In this task, there is no need to recruit sensorimotor programs about tool manipulation. In sum, what is common between these three tasks is the requirement to reason about physical materials and forces. We do not ignore that motor actions could be simulated in the mechanical problemsolving task, but no motor action needed to be simulated in the other three tasks. Therefore, what was common between all these tasks was the potential involvement of technical reasoning but not of sensorimotor programs about tool manipulation. Of course, an alternative is to consider that motor actions are always needed in all the situations, including situations where no “manipulable tool” is presented, such as a tractor and a boat, a pulley, or a cannon. We cannot rule out this alternative, which is nevertheless, for us, prejudicial because it implies that it becomes difficult to test the manipulation-based approach as motor actions would be everywhere. We voluntarily decided not to introduce a debate between the reasoning-based approach and the manipulation-based approach and preferred a more positive writing by stressing the insights from the present study. Note that we stressed the merits of the manipulation-based approach in the introduction because we sincerely think that this approach has provided interesting insights. However, we voluntarily did not discuss the debate between the two approaches. Given Reviewer 3’s comment (see also Reviewer 1 Point 2), we understand and agree that some words must be nevertheless said to discuss how the manipulation-based approach could interpret our results, thus stressing the potential limitations of our interpretations. Therefore, we added a specific section in the discussion in which we discussed this aspect in more details.

“The second limitation concerns the alternative interpretation that the left area PF is not central to technical reasoning but to the storage of sensorimotor programs about the prototypical manipulation of common tools. Here we show that the left area PF is recruited even in situations in which participants do not have to process common manipulable tools. For instance, some items of the psychotechnical task consisted of pictures of tractor, boat, pulley, or cannon. The fact that we found a common activation of the left area PF in such tasks as well as in the mechanical problem-solving task, in which participants could nevertheless simulate the motor actions of manipulating novel tools, indicates that this brain area is not central to tool manipulation but to physical understanding. That being said, some may suggest that viewing a boat or a cannon is enough to incite the simulation of motor actions, so our tasks were not equipped to distinguish between the manipulation-based approach and the reasoning-based approach. We have already shown that the left area PF is more involved in tasks that focus on the mechanical dimension of the tool-use action (e.g., the mechanical interaction between a tool and an object) than its motor dimension (i.e., the interaction between the tool and the effector [e.g., 24, 40]). Nevertheless, we recognize that future research is still needed to test the predictions derived from these two approaches.” (p. 18-19)

Introduction and throughout: The framing of left Area PF as a special area for technical reasoning is overly reductionistic from a functional neuroanatomic perspective in that it ignores a large relevant literature showing that the region is involved with many other tasks that seem not to require anything like technical cognition. Indeed, entering the coordinates - 56, -29, 36 (reported as the peak coordinates in common across the studied tasks) in Neurosynth reveals that 59 imaging studies report activations within 3 mm of those coordinates; few are action-related (a brief review indicated studies of verbal creativity, texture processing, reading, somatosensory processing, stress reactions, attentional selection etc). Please acknowledge the difficulty of claiming that a large brain region should be labeled the brain's technical reasoning area when it seems to also participate in so much else. The left IPL (including area PF) is densely connected to the ventral premotor cortex, and this network is activated in language and calculation tasks as well as tool use tasks (e.g., Matsumoto, Nair, et al., 2012). What other constructs might be able to unite this disparate literature, and are any of these alternative constructs ruled out by the present data? Lacking this objective discussion, the manuscript does read as a promotion of the investigators' preferred viewpoint.

We thank Reviewer 3 for this comment. As stressed in the initial version of the MS, we did not write that the left area PF is sufficient but central to the network that allows us to reason about the physical world. Regardless, we agree that an objective discussion was needed on this aspect to help the reader not misunderstand our purpose. We added a section in this aspect as suggested. 

“Before concluding, we would like to point out two potential limitations of the present study. The first limitation concerns the fact that the literature has documented the recruitment of the left area PF in many neuroimaging experiments in which there was no need to reason about physical events (e.g., language tasks). This can be easily illustrated by entering the left area PF coordinates in the Neurosynth database.

This finding could be enough to refute the idea that this brain area is specific to technical reasoning. Although this limitation deserves to be recognized, it is also true for many other findings. For instance, sensory or motor brain regions such as the precentral or the postcentral cortex have been found activated in many non-motor tasks, the visual word form area in non-language tasks, or the Heschl’s gyrus in nonmusical tasks. This remains a major challenge for scientists, the question being how to solve these inconsistencies that can result from statistical errors or stress that considerable effort is needed to understand the very functional nature of these brain areas. Thus, understanding that the left area PF is central to physical understanding can be viewed as a first essential step before discovering its fundamental function, as suggested by the functional polyhedral approach (56).” (p. 18)

Discussion: The discussion of a small cluster in the IFG (pars opercularis) that nearly survived statistical correction is noteworthy in light of the above point. This further underscores the importance of discussing networks and not just single brain regions (such as area PF) when examining complex processes. The investigators note, "a plausible hypothesis is that the left IFG integrates the multiple constraints posed by the physical situation to set the ground for a correct reasoning process, such as it could be involved in syntactic language processing". In fact, the hypothesis that the IFG and SMG are together related to resolving competition has been previously proposed, as has the more specific hypothesis that the SMG buffers actions and that the context-appropriate action is then selected by the IFG (e.g., Buxbaum & Randerath, 2018). The parallels with the way the SMG is engaged with competing lexical or phonological alternatives (e.g., Peramunage, Blumstein et al., 2011) have also been previously noted.

We added the Buxbaum and Randerath (2018)’s reference in this section.

“The functional role of the left IFG in the context of tool use has been previously discussed (24) and a plausible hypothesis is that the left IFG integrates the multiple constraints posed by the physical situation to set the ground for a correct reasoning process, such as it could be involved in syntactic language processing (for a somewhat similar view, see [51]).” (p. 16-17)

Introduction and Discussion: Please clarify how the technical reasoning network overlaps with or is distinct from the tool-use network reported by many previous investigators.

We added a couple of sentences in the discussion to clarify this point.

“It should be clear here that we do not advocate the localizationist position simply stating that activation in the left area PF is the necessary and sufficient condition for technical reasoning. We rather defend the view according to which it requires a network of interacting brain areas, one of them – and of major importance – being the left area PF. This allows the engagement of different configurations of cerebral areas in different technical-reasoning tasks, but with a central process acting as a stable component: The left area PF. Thus, when people intend to use physical tools, it can work in concert with brain regions specific to object manipulation and motor control, thereby forming another network, the tool-use network. It can also interact with brain regions specific to intentional gestures to form a “social-learning” network that allows people to enhance their understanding about the physical aspects of a technical task (e.g., the making of a tool) through communicative gestures such as pointing gestures (42). The major challenge for future research is to specify the nature of the cognitive process supported by the left area PF and that might be involved in the broad understanding of the physical world.” (p. 14)

Discussion: All of the experimental tasks require a response from a difficult choice in an array, and all of the tasks except for the fluid cognition task are likely to require prediction or simulation of a motion trajectory-whether an embodied or disembodied trajectory is unclear. The Discussion does mention the related (but distinct) idea of an "intuitive physics engine", a "kind of simulator", Please clarify how this study can rule out these alternative interpretations of the data. If the study cannot rule out these alternatives, the claims of the study (and the paper title which labels PF as a technical cognition area) should be scaled back considerably. 

We thank Reviewer 3 for this comment. The authors of the papers on intuitive physics engine or associative learning do not suggest that these processes are embodied. As discussed above, we clarified our perspective on the role of the left area PF and hope that these modifications help the reader better understand it. We warmly thank Reviewer 3 for their comments, which considerably helped us improve the MS.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Hüppe and colleagues had already developed an apparatus and an analytical approach to capture swimming activity rhythms in krill. In a previous manuscript they explained the system, and here they employ it to show a circadian clock, supplemented by exogenous light, produces an activity pattern consistent with "twilight" diel vertical migration (DVM; a peak at sunset, a midnight sink, and a peak in the latter half of the night).

They used light:dark (LD) followed by dark:dark (DD) photoperiods at two times of the year to confirm the circadian clock, coupled with DD experiments at four times of year to show rhythmicity occurs throughout the year along with DVM in the wild population. The individual activity data show variability in the rhythmic response, which is expected. However, their results showed rhythmicity was sustained in DD throughout the year, although the amplitude decayed quickly. The interpretation of a weak clock is reasonable, and they provide a convincing justification for the adaptive nature of such a clock in a species that has a wide distributional range and experiences various photic environments. These data also show that exogenous light increases the activity response and can explain the morning activity bouts, with the circadian clock explaining the evening and late-night bouts. This acknowledgement that vertical migration can be driven by multiple proximate mechanisms is important.

The work is rigorously done, and the interpretations are sound. I see no major weaknesses in the manuscript. Because a considerable amount of processing is required to extract and interpret the rhythmic signals (see Methods and previous AMAZE paper), it is informative to have the individual activity plots of krill as a gut check on the group data.

The manuscript will be useful to the field as it provides an elegant example of looking for biological rhythms in a marine planktonic organism and disentangling the exogenous response from the endogenous one. Furthermore, as high latitude environments change, understanding how important organisms like krill have the potential to respond will become increasingly important. This work provides a solid behavioral dataset to complement the earlier molecular data suggestive of a circadian clock in this species.

We appreciate the positive evaluation of our work by Reviewer 1, acknowledging our approach to record locomotor activity in krill and the importance of the findings in assessing krill’s potential to respond to environmental change in their habitat.

Reviewer #2 (Public review):

Summary:

This manuscript provides experimental evidence on circadian behavioural cycles in Antarctic krill. The krill were obtained directly from krill fishing vessels and the experiments were carried out on board using an advanced incubation device capable of recording activity levels over a number of days. A number of different experiments were carried out where krill were first exposed to simulated light:dark (L:D) regimes for some days followed by continuous darkness (DD). These were carried out on krill collected during late autumn and late summer. A further set of experiments was performed on krill across three different seasons (summer, autumn, winter), where incubations were all DD conditions. Activity was measured as the frequency by which an infrared beam close to the top of the incubation tube was broken over unit time. Results showed that patterns of increased and decreased activity that appeared synchronised to the LD cycle persisted during the DD period. This was interpreted as evidence of the operation of an internal (endogenous) clock. The amplitude of the behavioural cycles decreased with time in DD, which further suggests that this clock is relatively weak. The authors argued that the existence of a weak endogenous clock is an adaptation to life at high latitudes since allowing the clock to be modulated by external (exogenous) factors is an advantage when there is a high degree of seasonality. This hypothesis is further supported by seasonal DD experiments which showed that the periodicity of high and low activity levels differed between seasons.

Strengths

Although there has been a lot of field observations of various circadian type behaviour in Antarctic krill, relatively few experimental studies have been published considering this behaviour in terms of circadian patterns of activity. Krill are not a model organism and obtaining them and incubating them in suitable conditions are both difficult undertakings. Furthermore, there is a need to consider what their natural circadian rhythms are without the overinfluence of laboratory-induced artefacts. For this reason alone, the setup of the present study is ideal to consider this aspect of krill biology. Furthermore, the equipment developed for measuring levels of activity is well-designed and likely to minimise artefacts.

We would like to thank Reviewer 2 for their positive assessment of our approach to study the influence of the circadian clock on krill behavior. We are delighted, that Reviewer 2 found our mechanistic approach in understanding daily behavioral patterns of Antarctic krill using the AMAZE set-up convincing, and that the challenging circumstances of working with a polar, non-model species are acknowledged.

Weaknesses

I have little criticism of the rationale for carrying out this work, nor of the experimental design. Nevertheless, the manuscript would benefit from a clearer explanation of the experimental design, particularly aimed at readers not familiar with research into circadian rhythms. Furthermore, I have a more fundamental question about the relationship between levels of activity and DVM on which I will expand below. Finally, it was unclear how the observational results made here related to the molecular aspects considered in the Discussion.

(1) Explanation of experimental design - I acknowledge that the format of this particular journal insists that the Results are the first section that follows the Introduction. This nevertheless presents a problem for the reader since many of the concepts and terms that would generally be in the Methods are yet to be explained to the reader. Hence, right from the start of the Results section, the reader is thrown into the detail of what happened during the LD-DD experiments without being fully aware of why this type of experiment was carried out in the first place. Even after reading the Methods, further explanation would have been helpful. Circadian cycle type research of this sort often entrains organisms to certain light cycles and then takes the light away to see if the cycle continues in complete darkness, but this critical piece of knowledge does not come until much later (e.g. lines 369-372) leaving the reader guessing until this point why the authors took the approach they did. I would suggest the following (1) that more effort is made in the Introduction to explain the exact LD/DD protocols adopted (2) that a schematic figure is placed early on in the manuscript where the protocol is explained including some logical flow charts of e.g. if behavioural cycle continues in DD then internal clock exists versus if cycle does not continue in DD, the exogenous cues dominate - followed by - major decrease in cyclic amplitude = weak clock versus minor decrease = strong clock and so on

We want to thank Reviewer 2 for pointing out that the experimental design and its rationale are not becoming clear early in the manuscript, especially for people outside the field of chronobiology. We added a new figure (now Fig. 1), illustrating the basic principle of chronobiological study design and how we adopted it. We also extended the description at the beginning of the Results section to clarify the rationale behind the experimental design.

(2) Activity vs kinesis - in this study, we are shown data that (i) krill have a circadian cycle - incubation experiments; (ii) that krill swarms display DVM in this region - echosounder data (although see my later point). My question here is regarding the relationship between what is being measured by the incubation experiments and the in situ swarm behaviour observations. The incubation experiments are essentially measuring the propensity of krill to swim upwards since it logs the number of times an individual (or group) break a beam towards the top of the incubation tube. I argue that krill may be still highly active in the rest of the tube but just do not swim close to the surface, so this approach may not be a good measure of "activity". Otherwise, I suggest a more correct term of what is being measured is the level of "upward kinesis". As the authors themselves note, krill are negatively buoyant and must always be active to remain pelagic. What changes over the day-night cycle is whether they decide to expend that activity on swimming upwards, downwards or remaining at the same depth. Explaining the pattern as upward kinesis then also explains by swarms move upwards during the night. Just being more active at night may not necessarily result in them swimming upwards.

We believe there is a slight misunderstanding in how what we call “activity” is measured. The experimental columns are equipped with five detector modules, evenly distributed over the height of the column. In our analysis we count all beam breaks caused by upward movement, i.e. every time a detector module is triggered after a detector module at a lower position has been triggered, and not only when the top detector module is triggered. In this way, we record upward swimming movements throughout the column, and not only when the krill swims all the way to the top of the column. This still means that what we are measuring is swimming activity, caused by upward swimming. We use this measure, to deliberately separate increased swimming activity, from baseline activity (i.e. swimming, which solely compensates for negative buoyancy) and inactivity (i.e. passive sinking).

Higher activity is thus at first interpreted as an increase in swimming activity, which in the field may result in upwards-directed swimming but also could mean a horizontal increase in activity, for example, representing increased foraging and feeding activity. This would explain the daily activity pattern observed under LD cycles (now Fig. 3), which shows a general increase in activity during the dark phase. This nighttime increase could be used for both upward directed migration during sunset and horizontal directed swimming for feeding and foraging throughout the night.

We added the following sentence to the description of the activity metric in the Methods section to clarify this point (lines 465-469):

“To accomplish this, we organized the raw beam break data from all five detector modules in each experimental column in chronological order. We selected only those beam break detections that occurred after a detection in the detector module positioned lower on the column. Like this, we consider upward swimming movements throughout the full height of the column.”

(3) Molecular relevance - Although I am interested in molecular clock aspects behind these circadian rhythms, it was not made clear how the results of the present study allow any further insight into this. In lines 282 to 284, the findings of the study by Biscontin et al (2017) are discussed with regard to how TIM protein is degraded by light via the clock photreceptor CRYTOCHROME 1. This element of the Discussion would be a lot more relevant if the results of the present study were considered in terms of whether they supported or refuted this or any other molecular clock model. As it stands, this paragraph is purely background knowledge and a candidate for deletion in the interest of shortening the Discussion.

We agree that this part is not directly related to the data presented in the manuscript. We, therefore, omitted this part in the revised version of the manuscript to keep the discussion concise and focused on the results.

Other aspects

(i) 'Bimodal swimming' was used in the Abstract and later in the text without the term being fully explained. I could interpret it to mean a number of things so some explanation is required before the term is introduced.

We thank the Reviewer for pointing this out. We provided an explanation for the term “bimodal” in the Results section, where the two clock driven activity bouts are described first, by extending the sentence in lines 161-164, which now reads:

“This suggests that the circadian clock drives a distinct bimodal activity pattern with two activity peaks in one day, i.e. the evening and late-night activity bouts, while. In contrast, the morning activity bout is triggered by the onset of illumination in the experimental set-up.”.

(ii) Midnight sinking - I was struck by Figure 2b with regards to the dip in activity after the initial ascent, as well as the rise in activity predawn. Cushing (1951) Biol Rev 26: 158-192 describes the different phases of a DVM common to a number of marine organisms observed in situ where there is a period of midnight sinking following the initial dusk ascent and a dawn rise prior to dawn descent. Tarling et al (2002) observe midnight sinking pattern in Calanus finmarchicus and consider whether it is a response to feeding satiation or predation avoidance (i.e. exogenous factors). Evidence from the present study indicates that midnight sinking (and potential dawn rise) behaviour could alternatively be under endogenous control to a greater or lesser degree. This is something that should certainly be mentioned in the Discussion, possibly in place of the molecular discussion element mentioned above - possibly adding to the paragraph Lines 303-319.

We would like to thank the Reviewer for pointing this out and agree that adding the idea of an endogenous control of midnight sinking would be interesting to the discussion. We added the following section to the Discussion (lines 335-343):

“Interestingly, the decrease in clock-controlled swimming activity during the early night, right after the evening activity bout, may further facilitate a phenomenon called “midnight sinking”, which describes the sinking of animals to intermediate depths after the evening ascent, followed by a second rise to the surface before the morning descend. This behavior has been observed in a number of zooplankton species, including calanoid copepods (see 69, 70 and references therein) and krill (71). While previous studies suggested several exogenous factors, such as satiation or predator presence, as drivers of the midnight sink (69, 70), our study suggests that this pattern may be partly under endogenous control.”

(iii) Lines 200-207 - I struggled to follow this argument regarding Piccolin et al identifying a 12 h rhythm whereas the present study indicates a ~24 h rhythm. Is one contradicting the other - please make this clear.

In our study, we found that the circadian clock drives a bimodal pattern of swimming activity in krill, meaning it controls two bouts of activity in a 24-hour cycle. Piccolin et al. (2020) identified a swimming activity pattern of ~12 h (i.e. two peaks in 24 h) at the group level, which aligns with our findings at the individual level. We revised the Section in the discussion for more clarity, which now reads:

“Data from Piccolin et al. (20) showed a strong damping of the amplitude and indication of a remarkably short (~12 h) free running period (FRP) of vertical swimming behavior of a group of krill under constant darkness (20). The short period found in Piccolin et al. (20) complements is in line with our findings of a bimodal activity pattern the pattern of swimming activity under DD conditions on the individual level found in the present study, suggesting that the ~12 h rhythm in group swimming behavior in Piccolin et al. (20) could have resulted from a bimodal activity pattern at the individual level, as found in our study.” (lines 212-219).  

(iv) Although I agree that the hydroacoustic data should be included and is generally supportive of the results, I think that two further aspects should be made clear for context (a) whether there was any groundtruthing that the acoustic marks were indeed krill and not potentially some other group know to perform DVM such as myctophids (b) how representative were these patterns - I have a sense that they were heavily selected to show only ones with prominent DVM as opposed to other parts of the dataset where such a pattern was less clear - I am aware of a lot of krill research where DVM is not such a clear pattern and it is disingenuous to provide these patterns as the definitive way in which krill behaves. I ask this be made clear to the reader (note also that there is a suggestion of midnight sinking in Fig 5b on 28/2).

To clarify the mentioned points concerning the hydroacoustic data:

a) As mentioned in the Methods section, only hydroacoustic data during active fishing was included in the analysis. E. superba occurs in large monospecific aggregations, and the fishery actively targets E. superba and monitors their catch and the proportion of non-target species continuously with cameras. Krill fishery bycatch rates are very low (0.1–0.3%, Krafft et al. 2022), and fishing operations would stop if non-target species were caught in significant proportions at any time. Therefore, and supported by our own observations when we conducted the experiments, we argue that it is a valid assumption that E. superba predominantly causes the backscattering signal shown in Figure 5 (now Fig. 6).

b) We are aware of the fact that DVM patterns of Antarctic krill are highly variable and that normal DVM patterns do not need to be the rule (e.g. see our cited study on the plasticity of krill DVM by Bahlburg et al. 2023). The visualized data were not selected for their DVM pattern but represent the period directly preceding the sampling for behavioral experiments in four seasons (experiment 2), including the day of sampling. These periods were chosen to assess the DVM behavior of krill swarms in the field in the days before and during the sampling for behavioral experiments.

To improve understanding, we modified the description in the Results, Discussion, and Methods sections, as well as the caption of Figure 5 (now Fig. 6), which now read:

“To investigate whether krill swarms exhibited daily behavioral patterns in swimming behavior in the field before they were sampled for seasonal experiments, hydroacoustic data were recorded from the fishing vessel, continuously over a three-day period prior to sampling for the seasonal experiments described above…” (lines 191-194).

“Furthermore, hydroacoustic recordings demonstrate that most krill swarms sampled exhibited synchronized DVM in the field in the days directly before sampling for behavioral experiments, indicating that in this region, krill remain behaviorally synchronized across a wide range of photoperiods.” (lines 397-400).

“Hydroacoustic data were collected using a hull-mounted SIMRAD ES80 echosounder (Kongsberg Maritime AS) aboard the Antarctic Endurance, covering three days before the sampling for each of the seasonal behavioral experiments of experiment 2” (lines 512-515).

“We only included data during active fishing periods and the vessel is specifically targeting E. superba, which occurs in large monospecific aggregations. Further, krill fishery bycatch rates are very low (0.1-0.3%, 84), which makes it highly probable that the recorded signal represents krill swarms.” (lines 523-526).

“Hydroacoustic recordings showing the vertical distribution of krill swarms in the upper water column (<220 m) below the vessel, visualized by the mean volume backscattering signal (200 kHz), on the three days prior to krill sampling for experiments…” (lines 802-804).

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

As noted in the public review, this is a logical and well-written manuscript. I have very few comments to consider addressing.

The Results lead with a paragraph outlining the experimental approach. This is good, but you use the term "experiments" to refer to both the two sets, and the two or four subsets of experiments. Perhaps consider the subset experiments as "treatments"? I understood what you meant, but it took a few read-throughs to be sure I got it.

We thank the reviewer for pointing this out and changed the nomenclature of the experiments throughout the manuscript. We now refer to the two sets of experiments as experiment 1 and 2, to the subsets of experiment 1 as “short day treatment” and “long day treatment”, and to the subsets of experiment 2 as summer treatment, late summer treatment, autumn treatment, and winter treatment. We also believe that the new Figure 1 is now helping to follow the experimental design more efficiently.

Ln 140: "...off and decrease at lights-on."

We adjusted the sentence accordingly.

Ln 244: Can you define "extreme photic conditions"? I get what you mean, but to be clear to the reader this would help.

We adjusted the sentence, which now reads:

“This could confer a significant adaptive advantage to species inhabiting environments characterized by extreme photic conditions (53, 54, 60), such as phases of polar night or midnight sun as well as rapid changes in daylength, or species that rely on precise photoperiodic time measurement for accurate seasonal adaptation.” (lines 258-261).

Figures: Consider adding an LSP for groups in Fig 1. Also, it would be useful to have LSP period estimates for each individual tested. This could be a separate table, or it could be added to the individual activity plots. Should S3 and S4 be reversed?

We thank the reviewer for their suggestion and added an LSP as figure 1d (now Fig. 2d) to statistically support the group activity shown in Figure 1c (now Fig. 2c) as suggested. We added the individual animals' LSP period estimates to supplementary figures S2, S7, S8, S9, and S10. We also reversed Figures S3 and S4 to match the appearance in the main text. 

Fig 5: are the light regime bars for b and c correct? They look similar, but there are only 15 days apart, so perhaps they are correct as is.

We double checked the light regime bars in Fig. 5b and c (now 6b and c) and they are correct as is.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

This manuscript by Kaya et al. studies the effect of food consumption on hippocampal sharp wave ripples (SWRs) in mice. The authors use multiple foods and forms of food delivery to show that the frequency and power of SWRs increase following food intake, and that this effect depends on the caloric content of food. The authors also studied the effects of the administration of various food-intake-related hormones on SWRs during sleep, demonstrating that ghrelin negatively affects SWR rate and power, but not GLP1, insulin, or leptin. Finally, the authors use fiber photometry to show that GABAergic neurons in the lateral hypothalamus, increase activity during a SWR event.

Strengths:

The experiments in this study seem to be well performed, and the data are well presented, visually. The data support the main conclusions of the manuscript that food intake enhances hippocampal SWRs. Taken together, this study is likely to be impactful to the study of the impact of feeding on sleep behavior, as well as the phenomena of hippocampal SWRs in metabolism.

Weaknesses:

Details of experiments are missing in the text and figure legends. Additionally, the writing of the manuscript could be improved.

We thank the reviewer for their favorable assessment of the work and its potential impact. We have added all requested details in the text and figure legends and revised the wording of the manuscript to improve its clarity.

Reviewer #2 (Public review):

Summary:

Kaya et al uncover an intriguing relationship between hippocampal sharp wave-ripple production and peripheral hormone exposure, food intake, and lateral hypothalamic function. These findings significantly expand our understanding of hippocampal function beyond mnemonic processes and point a direction for promising future research.

Strengths:

Some of the relationships observed in this paper are highly significant. In particular, the inverse relationship between GLP1/Leptin and Insulin/Ghrelin are particularly compelling as this aligns well with opposing hormone functions on satiety.

Weaknesses:

I would be curious if there were any measurable behavioral differences that occur with different hormone manipulations.

We thank the reviewer for their favorable assessment of the work and its contribution to our understanding of non-mnemonic hippocampal function. Whether there are behavioral differences that occur following administration of the different hormones is a great question, yet unfortunately our study design did not include fine behavioral monitoring to the degree that would allow answering it. While some previous studies have partially addressed the behavioral consequences of the delivery of these hormones (and we reference these studies in our Discussion), how these changes may interact with the hippocampal and hypothalamic effects we observe is a very interesting next step.

Reviewer #3 (Public review):

Summary:

The manuscript by Kaya et al. explores the effects of feeding on sharp wave-ripples (SWRs) in the hippocampus, which could reveal a better understanding of how metabolism is regulated by neural processes. Expanding on prior work that showed that SWRs trigger a decrease in peripheral glucose levels, the authors further tested the relationship between SWRs and meal consumption by recording LFPs from the dorsal CA1 region of the hippocampus before and after meal consumption. They found an increase in SWR magnitude during sleep after food intake, in both food restricted and ad libitum fed conditions. Using fiber photometry to detect GABAergic neuron activity in the lateral hypothalamus, they found increased activity locked to the onset of SWRs. They conclude that the animal's satiety state modulates the amplitude and rate of SWRs, and that SWRs modulate downstream circuits involved in regulating feeding. These experiments provide an important step forward in understanding how metabolism is regulated in the brain. However, currently, the paper lacks sufficient analyses to control for factors related to sleep quality and duration; adding these analyses would further support the claim that food intake itself, as opposed to sleep quality, is primarily responsible for changes in SWR activity. Adding this, along with some minor clarifications and edits, would lead to a compelling case for SWRs being modulated by a satiety state. The study will likely be of great interest in the field of learning and memory while carrying broader implications for understanding brain-body physiology.

Strengths:

The paper makes an innovative foray into the emerging field of brain-body research, asking how sharp wave-ripples are affected by metabolism and hunger. The authors use a variety of advanced techniques including LFP recordings and fiber photometry to answer this question. Additionally, they perform comprehensive and logical follow-up experiments to the initial food-restricted paradigm to account for deeper sleep following meal times and the difference between consumption of calories versus the experience of eating. These experiments lay the groundwork for future studies in this field, as the authors pose several follow-up questions regarding the role of metabolic hormones and downstream brain regions.

We thank the reviewer for their appreciation and constructive review of the work.

Weaknesses:

Major comments:

(1) The authors conclude that food intake regulates SWR power during sleep beyond the effect of food intake on sleep quality. Specifically, they made an attempt to control for the confounding effect of delta power on SWRs through a mediation analysis. However, a similar analysis is not presented for SWR rate. Moreover, this does not seem to be a sufficient control. One alternative way to address this confound would be to subsample the sleep data from the ad lib and food restricted conditions (or high calorie and low calorie, etc), to match the delta power in each condition. When periods of similar mean delta power (i.e. similar sleep quality) are matched between datasets, the authors can then determine if a significant effect on SWR amplitude and rate remains in the subsampled data.

This is an important point that we believe we addressed in a few complementary ways. First, the mediation analysis we implemented measures the magnitude and significance of the contribution of food on SWR power after accounting for the effects of delta power, showing a highly significant food-SWR contribution. While the objective of subsampling is similar, mediation is a more statistically robust approach as it models the relationship between food, SWR power, and delta power in a way that explicitly accounts for the interdependence of these variables. Further, subsampling introduces the risk of losing statistical power by reducing the sample size, due to exclusion of data that might contain relevant and valuable information. Mediation analysis, on the other hand, uses the full dataset and retains statistical power while modeling the relationships between variables more holistically. However, as we were not satisfied with a purely analytical approach to test this issue, we carried out a new set of experiments in ad-libitum fed mice, where there is no concern of food restriction impairing sleep quality in the presleep session. In these conditions food amount also significantly correlated with, and showed significant mediation of, the SWR power change. Finally, we acknowledge and discuss this point in the Discussion, highlighting that given the known relationship between cortical delta and SWRs, it is challenging to fully disentangle these signals. 

(2) Relatedly, are the animals spending the same amount of time sleeping in the ad lib vs. food restricted conditions? The amount of time spent sleeping could affect the probability of entering certain stages of sleep and thus affect SWR properties. A recent paper (Giri et al., Nature, 2024) demonstrated that sleep deprivation can alter the magnitude and frequency of SWRs. Could the authors quantify sleep quantity and control for the amount of time spent sleeping by subsampling the data, similar to the suggestion above?

Following the reviewer’s comment, we have quantified and compared the amount of time spent in NREM sleep in the Pre and Post session pairs in which the animals were food restricted, with 0-1.5 g of chow given between the sleep sessions. We found that there was no significant difference in the amount of time spent in NREM sleep in the Pre and Post sessions. We have added this result to the Results section of the manuscript and as a new Supplementary Fig. 1. 

Additionally, we have added details to the Methods section that were missing in the original submission that are relevant to this point. Specifically, within the sleep sessions, the ongoing sleep states were scored using the AccuSleep toolbox (https://github.com/zekebarger/AccuSleep) using the EEG and EMG signals. NREM periods were detected based on high EEG delta power and low EMG power, REM periods were detected based on high EEG theta power and low EMG power, and Wake periods were detected based on high EMG power. Importantly, only NREM periods were included for subsequent SWR detection, quantification and analyses (in particular, reported SWR rates reflect the number of SWRs per second of NREM sleep). 

(3) Plot 5I only reports significance but does not clearly show the underlying quantification of LH GABAergic activity. Upon reading the methods for how this analysis was conducted, it would be informative to see a plot of the pre-SWR and post-SWR integral values used for the paired t-test whose p-values are currently shown. For example, these values could be displayed as individual points overlaid on a pair of boxand-whisker plots of the pre- and post-distribution within the session (perhaps for one example session per mouse with the p-value reported, to supplement a plot of the distribution of p-values across sessions and mice). If these data are non-normal, the authors should also use a non-parametric statistical test.

We have generated the summary plots the reviewer requested and have now included them in Supplementary Fig. 2. 

Minor comments:

(4) A brief explanation (perhaps in the discussion) of what each change in SWR property (magnitude, rate, duration) could indicate in the context of the hypothesis may be helpful in bridging the fields of metabolism and memory. For example, by describing the hypothesized mechanistic consequence of each change, could the authors speculate on why ripple rate may not increase in all the instances where ripple power increases after feeding? Why do the authors speculate that ripple duration does not increase, given that prior work (Fernandez-Ruiz et al. 2019) has shown that prolonged ripples support enhanced memory?

This is an interesting point and we have added a section to the Discussion to discuss it (pg. 17, last paragraph)

(5) The authors suggest that "SWRs could modulate peripheral metabolism" as a future implication of their work. However, the lack of clear effects from GLP-1, leptin and insulin complicates this interpretation. It might be informative for readers if the authors expanded their discussion of what specific role they speculate that SWRs could play in regulating metabolism, given these negative results.

We have added a section to the Discussion proposing potential reasons for this point (pg. 16, last paragraph)

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors):

Major Comments:

(1) The experiments involve very precise windows of time for sleeping and eating that seem impossible to control. For example, the authors state that for the experiments in Figure 1, there was a 2-h sleep period, followed by a 1-h feeding period, followed by another 2-h sleep period. Without sleep deprivation procedures or other environmental manipulations, how can these periods be so well-defined? Even during the inactive period, mice typically don't sleep for 2-h bouts at once, and the addition of food would not likely lead to an exact 1-h period of wakefulness in the middle. The validity of these experimental times would be more believable if the authors provided much more data on these sessions. For example, the authors could provide a table or visual display of data for the actual timing of the pre-sleep, eating, and post-sleep phases with exact time measurements and/or visual display of sleep versus wakefulness.

This is an important point, which we were not clear enough about in the original submission. While the durations of the Pre-sleep, Wake and Post-sleep sessions were indeed 2 h, 1 h and 2 h respectively, the animals did not actually sleep during the entirety of the sleep sessions. Importantly, we performed sleep state scoring on all sessions, and only analyzed identified NREM sleep for all SWR analyses. Following the reviewer’s comment (and that of Reviewer 1), we have quantified and compared the amount of time spent in NREM sleep in the Pre and Post session pairs in which the animals were food restricted and 0-1.5 g of chow were given between the sleep sessions. We found that there was no significant difference in the amount of time spent in NREM sleep in the Pre and Post sessions. We have added this result to the Results section of the manuscript and as a new Supplementary Fig. 1. 

Additionally, we have added details to the Methods section that were missing in the original submission that are relevant to this point. Specifically, within the sleep sessions, the ongoing sleep states were scored using the AccuSleep toolbox (https://github.com/zekebarger/AccuSleep) using the EEG and EMG signals. NREM periods were detected based on high EEG delta power and low EMG power, REM periods were detected based on high EEG theta power and low EMG power, and Wake periods were detected based on high EMG power. Importantly, only NREM periods were included for subsequent SWR detection, quantification and analyses (in particular, reported SWR rates reflect the number of SWRs per second of NREM sleep). 

(2) I may have missed this (although I tried searching in the text and figure legend), but the authors did not state the difference between green versus red bar colors in Figure 1 C-E. For Figures 1 F-J, do the individual dots represent both the test (fed) animals and control animals, or just the test animals?

We thank the reviewer for the opportunity to clarify these points. Red bars in Fig. 1C-E represent the SWR changes observed following delivery of equal or more than 0.5 g of chow, while the green bars represent the changes observed following delivery of less than 0.5 g. Fig. 1F-J includes both the experimental and control animals- the control animals appearing as having received 0 food amount. This information has now been added to the figure legend.

(3) For the jello experiments in Figure 3, was there only 1 trial per animal? Previous studies show that animals learn the caloric value of jello after subsequent trials, so whether or not multiple trials took place in each animal is important for interpretation of the results.

In Figure 3, the datapoints within each panel represent different animals and this information has now been added to the figure legend. Nevertheless, the animals were previously habituated to all foods, including regular jello, sugar-free jello and chocolate. While we consider it unlikely that this prior experience was sufficient to underlie the differential effects on SWRs, we cannot fully rule out the possibility that it provided some ability to predict the caloric value and consequences of the different foods. We have added details to the acknowledgement of this point in the Discussion (pg. 17, second paragraph).

(4) The experiments in Figure 5 are informative but don't relate to the experiments in the rest of the study. It is difficult to interpret their meaning given that these experiments take place over seconds while the other experiments take place over hours. Some attempt should be made to bridge these experiments over the timescales relevant for the behaviors studied in Figures 1-4.

We have now further acknowledged and discussed the point that our investigation is limited to the timescale of seconds around SWRs, and thus identified a potential communication channel, but whether and how this communication changes across hours following feeding remains for future studies (pg. 18, second paragraph).

(5) Figure 5B should depict the x-axis in seconds, not an arbitrary set of times from a recording.

We have replaced these with a time scale bar.

Minor Comments:

(6) The writing of the manuscript can be improved in many places:

Sometimes the writing could be more precise. For example, the Abstract states: "hippocampal sharp wave ripples (SWRs)... have been shown to influence peripheral glucose metabolism." Could this be written in a more informative way, rather than just staying "has been shown to influence?" A few more words would provide a lot more information. Similarly, at the end of the Introduction: "we set out to test the hypothesis that SWRs are modulated following meal times as part of the systems-level response to changing metabolic needs." This is not a strong hypothesis... could it be written to boldly state how the SWRs will be modulated (increase or decrease) and provide more assertive information?

The writing can be grandiose at times. Phrases such as "life is a continuous journey" or "the hypothalamus is a master regulator of homeostasis" are a bit sophomoric and too colloquial.

Finally, a representative recording should be referred to as just that-a "representative recording," as opposed to a "snippet," which is also colloquial. This word is used in the figure legends to Figures 1 and 5, and misspelled as "sinpper" in Figure 1

We have reworded all these sentences and phrases to make them clearer, more concrete and more formal.

(7) The methods state that the study used both male and female mice. Were they used in equal numbers across experiments?

Only one female was used in the final dataset, and we have corrected the wording accordingly.

Reviewer #2 (Recommendations for the authors):

Great paper!

Thanks!

Reviewer #3 (Recommendations for the authors):

Below are some minor requests for clarification, including in figures:

(1) Fig. 5H y-axis should say "normalized dF/F."

Done

(2) Fig. 1B is missing a y-axis label. It may be clearer to display separate y-axis scale bars for each component (SWR envelope, ripple-filtered amplitude, etc).

Done

(3) Please include labels for brain areas and methodological components in Fig. 5A.

Done

(4) Should Fig. 5B have the same y-axis or scale bars as 1B?

We have edited the figure labels and legends to be visually similar

(5) In Fig. 5J, is the y-axis a count of sessions?

Yes, we have added that to the y-axis label

(6) Could the authors please clarify whether the sugar-free jello was sweetened with an artificial sweetener? If so, this is a robust control for the rewarding nature of the two jellos, so a quick clarification would highlight this strength of the experiment.

We thank the reviewer for this great point. Indeed, the sugar free jello contained artificial sweeteners (Aspartame and Acesulfame Potassium). We have added this information to the Results and Methods.

(7) It appears in Fig. 5 that there may be a reliable dip in activity **at** the time of SWR onset, followed by the increase afterward, as shown in the example FP trace and the individual ripple-triggered traces. Is this indeed the case, and does this dip fall significantly below baseline? This characterization would be interesting, but I acknowledge is not necessarily crucial to the study to include.

This would indeed be an interesting finding, but upon examination and statistical testing, we found that this is not the case. We believe this may appear as such due to the normalization of the traces.

(8) The authors mention a reduction in ripple rate following insulin under food restriction as the only significant effect for insulin, GLP-1, and leptin, yet there was also a significant increase (at p<0.05) in ripple duration for GLP-1 in the ab lib condition. Is this not considered noteworthy?

This is a fair point and we have reworded the description of this result to simply state that there were no robust, consistent, dose-dependent effects of GLP-1, leptin and insulin on SWR attributes.

Author response:

The following is the authors’ response to the previous reviews

Reviewer #1 (Public review):

This study presents evidence that a special group of place cells, those tuned to fast-gamma oscillations, play a key role in theta sequence development. How theta sequences are formed and developed during experience is an important question, because these sequences have been implicated in several cognitive functions of place cells, including memory-guided spatial navigation. The revised version of this paper has been significantly improved. Major concerns in the previous round of review on technical and conceptual aspects of the relationship between gamma oscillations and theta sequences are addressed. The main conclusion is supported by the data presented.

Reviewer #2 (Public review):

The authors have conducted new analysis to address the issues I and the other reviewers raised in our original revision. As a result, the revised manuscript has been substantially improved.

We thank the two reviewers for their positive comments.

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

There are, however, still a few remaining issues that need further clarification.

- Despite the authors explanation and comparison with Kitanishi et al., 2015, Neuron, I still find that the reduced number of significantly gamma phase-locked cells is at odds with most previous reports (e.g., Csicvari et al., 2003; Colgin et al., 2009; Belluscio et al., 2012; Schomburg et al., 2014; Cabral et al., 2014; Fernandez-Ruiz et al., 2017; Lopes dos Santos et al., 2018). There can be several issues to explain this difference, like the choice of LFP reference channel. The authors should at least acknowledge this difference in the text.

We thank the reviewer for this suggestion.  We discussed the potential reasons causing the different proportion of gamma phase locked cells in the Discussion (lines367-380).

- The new Figure R2 is very useful and should be included in the manuscript. It would be even better to expand the frequency range to higher frequencies to show where the maximum peak is. Still, the potential contribution of spike leakage should be acknowledged. While I agree that it will not account for all fast gamma spike modulation, it is certainly a contributing factor. A further evidence of this is that the coupling strength seems to keep increasing towards supra gamma frequency range in Fig R2. This is to be expected given that the authors have used the local LFP from the same tetrode where cells were recorded, which is never a good practice.

We thank the reviewer for this suggestion. Now the Fig R2 has been moved to the manuscript as a part of Figure 2-figure supplement 2 (lines133-135). In terms of the contribution of spike leakage by using the local LFP, we also detected FG-cells by using LFP from a different tetrode, i.e. the central one of the bundle that located in the cell body layer, and found approximate proportion of FG-cells which phase locked to ~75Hz (Fig R3, now the Figure 2-figure supplement 2C-F). Thus, we think using the local LFP would not affect the main conclusion and we decide to keep the original results. We also acknowledged the potential contribution of spike leakage in the Discussion (lines 372-377).

- From the authors answer I understand that recordings were almost exclusively conducted from the deep CA1 pyramidal layer. This would preclude any meaningful interpretation of the deep/ superficial differences in the distribution of FG and NFG cells. This is not a crucial point for the paper but needs to be acknowledged.

We thank the reviewer for this suggestion.  We acknowledged the meaningful interpretation of the deep/ superficial differences in the distribution of FG- and NFG-cells in the Discussion (lines 380-386).

- I am afraid that the authors interpreted my comment about authorship in the opposite way that I intended. I meant that the usual practice is that the last author of the manuscript is the person who has been the main intellectual driver of the work, not the most senior one necessarily. I guess that is Dr. Zheng not Dr. Ming. However, I leave this decision to the discretion of the authors.

We thank the reviewer for this rigorous consideration.  Dr. Ming and Dr. Zheng were both the main intellectual drivers of this work.  Therefore, we decide to keep the current authors in the manuscript.

Author response:

The following is the authors’ response to the original reviews

Public reviews:

Reviewer #1 (Public review):

Summary:

This very interesting manuscript proposes a general mechanism for how activating signaling proteins respond to species-specific signals arising from a variety of stresses. In brief, the authors propose that the activating signal alters the structure by a universal allosteric mechanism.

Strengths:

The unitary mechanism proposed is appealing and testable. They propose that the allosteric module consists of crossed alpha-helical linkers with similar architecture and that their attached regulatory domains connect to phosphatases or other molecules through coiled-coli domains, such that the signal is transduced via rigidifying the alpha helices, permitting downstream enzymatic activity. The authors present genetic and structural prediction data in favor of the model for the system they are studying, and stronger structural data in other systems.

Weaknesses:

The evidence is indirect - targeted mutations, structural predictions, and biochemical data. Therefore, these important generalizable conclusions are not buttressed by impeccable data, which would require doing actual structures in B. subtilis, confirming experiments in other organisms, and possibly co-evolutionary coupling. In the absence of such data, it is not possible to rule out variant models.

We thank the reviewer for their feedback. A challenge of studying flexible proteins is that it is often not possible to directly obtain high resolution structural data. For the case of B. subtilis RsbU, the independent experimental approaches we applied (including two unbiased genetic screens, targeted mutagenesis, SAXS, enzymology, and structure prediction, which includes evolutionary coupling) converged upon a model for activation, which we feel is well supported. Frustratingly, our attempts at determining high resolution experimental structures have been unsuccessful, which we think is due to the flexibility of the proteins revealed by our SAXS experiments. For example, we collected X-ray diffraction data from crystals of a fragment of B. subtilis RsbU containing the N-terminal domain and linker in which the linker was almost entirely disordered in the maps. We agree that doing experiments in other organisms would be valuable next steps to test the hypothesis that this coiled-coil based transduction mechanism is conserved across species, and will modify the text to differentiate this more speculative section of the manuscript.

We have modified the abstract to read:

“This coiled-coil linker transduction mechanism additionally suggests a resolution to the mystery of how shared sensory domains control serine/threonine phosphatases, diguanylate cyclases and histidine kinases.”

We have modified the results to read:

"These predictions suggest a testable hypothesis that RsbP is controlled through an activation mechanism similar to that of RsbU (Fig. 5A)”

“From this analysis, we speculate that linker-mediated phosphatase domain dimerization is an evolutionarily conserved, adaptable mechanism to control PPM phosphatase activity.”

Based on this critique (and the critiques of the other reviewers), we plan to do energetic analysis of the predicted coiled coils from the enzymes we analyzed from other species and to incorporate this into the manuscript.

We have modified the results to read:

Consistent with a model in which the stability of the linker plays a conserved regulatory role, the AlphaFold2 models for many of the predicted structures have unfavorable polar residues buried in the coiled-coil interface (positions a and d, for which non-polar residues are most favorable) (Figure 5 – figure supplement 2).”

Finally, in the manuscript, we have highlighted that this mechanism is not the only mechanism for activation of other proteins with effector domains connected to linkers, but rather one of many mechanisms (Fig 5G). The reviewer additionally made helpful suggestions about the text in detailed comments that we will incorporate as appropriate.

Reviewer #2 (Public review):

Summary:

While bacteria have the ability to induce genes in response to specific stresses, they also use the General Stress Response (GSR) to deal with growth conditions that presumably include a larger range of stresses (for instance, stationary phase growth). The activation of GSR-specific sigma factors is frequently at the heart of the induction of a GSR. Given the range of stresses that can lead to GSR induction, the regulatory inputs are frequently complex. In B. subtilis, the stressosome, a multi-protein complex, contains a set of proteins that, upon appropriate stresses, initiate partner switching cascades that free the sigma B sigma factor from an anti-sigma. The focus here is on the mode of activation of RsbU, a serine/threonine phosphatase of the PPM family, leading to sigB activation. RbsT, a component of the degradosome interacts with RsbU upon stress, activating the phosphatase activity. Once active, RsbU dephosphorylates its target (RsbV, an anti-antisigma), which in turn binds the anti-sigma. The conclusion is that flexible linker domains upstream of the phosphatase domain are the target for activation, via binding of proteins to the N-terminal domain, resulting in a crossed-linker dimeric structure. The authors then use the information on RsbU to suggest that parallel approaches are used to activate PPM phosphatases for the GSR response in other bacteria. (Biology vs. Mechanism, evolution?)

Strengths and Weaknesses:

Many of these have to do with clarifying what was done and why. This includes the presentation and content of the figures.

One issue relates to the background and context. A bit more information on the stresses that release RsbT would be useful here. The authors might also consider a figure showing the major conclusions and parallels for SpoIIE activation and possibly other partner switches that are discussed, introducing the switch change more clearly to set the stage for the work here (and the generalization). There are a lot of players to keep track of.

We plan to carefully review the manuscript to improve the clarity of presentation and background. In particular, we thank the reviewer for pointing out the missing information about the release of RsbT from the stressosome. We will incorporate this information into the introduction and provide an additional figure.

We have added the following text to the introduction:

“RsbT is sequestered in a megadalton stress sensing complex called the stressosome, and is released to bind RsbU in response to specific stress signals including ethanol, heat, acid, salt, and blue light”

We have added a new figure panel (2C) that shows the model for how Q94L, M166V, and RsbT binding induce conformational change of the PPM domain to recruit metal cofactor and activate RsbU (analogous, but slightly different from the mechanism for SpoIIE).

The reviewer additionally provided detailed helpful comments that we will incorporate in the text and figures.

Reviewer #3 (Public review):

Summary:

The authors present a study building on their previous work on activation of the general stress response phosphatase, RsbU, from Bacillus subtilis. Using computed structural models of the RsbU dimer the authors map previously identified activating mutations onto the structure and suggest further protein variants to test the role of the predicted linker helix and the interaction with RsbT on the activation of the phosphatase activity.

Using in vivo and in vitro activity assays, the authors demonstrate that linker variants can constitutively activate RsbU and increase the affinity of the protein for RsbT, thus showing a link between the structure of the linker region and RsbT binding.

Small angle X-ray scattering experiments on RsbU variants alone, and in complex with RsbT show structural changes consistent with a decreased flexibility of the RsbU protein, which is hypothesised to indicate a disorder-order transition in the linker when RsbT binds. This interpretation of the data is consistent with the biochemical data presented by the authors.

Further computed structure models are presented for other protein phosphates from different bacterial species and the authors propose a model for phosphatase activation by partner binding. They compare this to the activation mechanisms proposed for histidine kinase two-component systems and GGDEF proteins and suggest the individual domains could be swapped to give a toolkit of modular parts for bacterial signalling.

Strengths:

The key mutagenesis data is presented with two lines of evidence to demonstrate RsbU activation - in vivo sigma-b activation assays utilising a beta-galactosidase reporter and in vitro activity assays against the RsbV protein, which is the downstream target of RsbU. These data support the hypothesis for RsbT binding to the RsbU linker region as well as the dimerisation domain to activate the RsbU activity.

Weaknesses:

Small angle scattering curves are difficult to unambiguously interpret, but the authors present reasonable interpretations that fit with the biochemical data presented. These interpretations should be considered as good models for future testing with other methods - hydrogen/deuterium exchange mass spectrometry, would be a good additional method to use, as exchange rates in the linker region would be affected significantly by the disorder/order transition on RsbT binding.

We agree with the reviewer that the SAXS data has inherent ambiguity due to the nature of the measurement. However, SAXS is one of the best techniques to directly assess conformational flexibility. Our scattering data for RsbU have multiple signatures of flexibility supporting a high confidence conclusion. While the scattering data support a reduction in flexibility for the RsbT/RsbU complex, we agree that a high resolution structure would be valuable. However the combination of the scattering data with our biochemical and genetic data supports the validity of the AlphaFold predicted model. We thank the reviewer for the suggestion of future hydrogen/deuterium exchange experiments that would be complementary, but which we feel are beyond the scope of this work.

The interpretation of the computed structure models should be toned down with the addition of a few caveats related to the bias in the models returned by AlphaFold2. For the full-length models of RsbU and other phosphatase proteins, the relationship of the domains to each other is likely to be the least reliable part of the models - this is apparent from the PAE plots shown in Supplementary Figure 8. Furthermore, the authors should show models coloured by pLDDT scores in an additional supplementary figure to help the reader interpret the confidence level of the predicted structures.

We thank the reviewer for suggestions on how to clarify the discussion of AlphaFold models. We will decrease the emphasis on the computed models in the text and will add figures with the models colored by the pLDDT scores to aid in the interpretation.

We have modified the text of the Abstract: “This coiled-coil linker transduction mechanism additionally suggests a resolution to the mystery of how shared sensory domains control serine/threonine phosphatases, diguanylate cyclases and histidine kinases.”

We have modified the text of the Results: “These predictions suggest a testable hypothesis that RsbP is controlled through an activation mechanism similar to that of RsbU (Fig. 5A).”

“From this analysis, we speculate that linker-mediated phosphatase domain dimerization is an evolutionarily conserved, adaptable mechanism to control PPM phosphatase activity”

We have also added Figure 1 – figure supplement 2 with the AlphaFold2 models colored by the pLDDT scores.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Baral and colleagues investigate the regulatory mechanisms of the General Stress Response (GSR) in Bacillus subtilis, focusing on the phosphatase RsbU and its regulation by the protein RsbT. The GSR is a critical adaptive mechanism that allows bacteria to survive under various stress conditions by reshaping their physiology through a broad transcriptional response. RsbU, a key player in the GSR, facilitates the activation of the transcription factor SigB by dephosphorylating RsbV. This activation is mediated through a partner-switching mechanism involving RsbT. Baral and colleagues use a combination of genetic screening, structural predictions via AlphaFold2, and biophysical techniques such as SAXS and MALS to present a model for how RsbT regulates RsbU. Key findings include the identification of specific amino acid substitutions that enhance RsbU activity, the role of the α-helical linker in RsbU dimerization and activation, and the potential broader conservation of these mechanisms across bacterial species. However, as described below, additional work is required to solidify the results.

Major Points

(1) The manuscript is misnamed--it dissects a single step of the signal-transduction pathway regulating the general stress response. Instead, it is rather seeking a generalizable mechanism for kinase -phosphatase interactions across stresses.

We have edited the title to “A General Mechanism for Initiating the General Stress Response in Bacteria” to reflect that that this study addresses the initiating event of the general stress response.

(2) The genetic screen likely has limitations in detecting all possible variants that could affect RsbU activity. The readout is specific to σ^B activation, and the focus on specific amino acid substitutions may overlook other significant regions or mechanisms involved in the regulation of RsbU, particularly those involving RsbV and RsbT.

Our screens were specifically designed to identify features of RsbU that contribute to regulation. Importantly, RsbU does not have any known targets other than RsbV and the downstream σ^B response but agree that substitutions in either RsbV or RsbT could influence RsbU activation. In principle our suppressor screen with RsbU^Y28I could have identified RsbT variants (rsbT was mutagenized in this screen), but we did not identify any such variants in the screen. We conducted a separate screen (published elsewhere) that specifically addressed how RsbU recognizes RsbV.

(3) The authors largely focus on the biochemical and structural aspects of RsbU regulation. There is limited discussion on the broader functional implications of these findings in the context of bacterial physiology and stress response. Incorporating more in vivo studies to show how these mechanisms impact bacterial survival and adaptation would provide a more comprehensive understanding.

We appreciate this comment, but did not conduct additional studies of survival and adaptation because the phenotypes of σ^B deletion in B. subtilis under laboratory conditions are relatively mild and therefore difficult to assay. Future studies to address this in other systems could be highly informative.

(4) The results primarily support the model of linker-mediated dimerization and rigidity. However, other potential regulatory mechanisms or interacting partners might also play significant roles in RsbU activation. A more thorough exploration of these possibilities would strengthen the study's conclusions.

One of the major advantages of RsbU as a model for initiation of the general stress response is that the system is discreet with all evidence pointing to there being a single primary input (RsbT) and output (dephosphorylation of RsbV). While there are other possible variations on the system (for example RsbU may be directly activated by manganese stress), we focused on this system precisely because of its simplicity.

(5) While the study presents evidence for the conservation of the described mechanism across different species, this assumption is based on structural predictions and limited experimental data. Broader experimental validation across diverse bacterial species would be necessary to substantiate this claim. Coevolution coupling along with conservation/evolutionary studies could be considered.

We have altered the language in the paper to emphasize where we are making inferences from predictions that are therefore more speculative. We agree that a more detailed analysis of the evolutionary coupling would likely be fruitful. We note that these couplings are the major driving force of AlphaFold predictions, suggesting that these couplings contributed to the models that we analyzed.

(6) The reliance on AlphaFold2 for structural predictions introduces potential biases and uncertainties inherent in computational models. Experimental validation of these models through additional techniques such as cryo-EM or X-ray crystallography would strengthen the conclusions.

We agree with this point, which is why we performed extensive analysis and validation of the models for RsbU using SAXS, genetics, and biochemistry. The proposed techniques are made more challenging by flexibility and heterogeneity, which we detected in our experiments. Our attempts thus far at experimental structure determination are consistent with this being a major technical hurdle.

(7) SAXS data provide low-resolution structural information, and the interpretation of flexibility versus rigidification might be overemphasized in its interpretation. This part of the study was difficult to interpret. Improving readability by breaking down the text into sections with clear headings for each figure panel and clarifying descriptions of the panels and methods would help. Complementary high-resolution techniques could provide a more definitive view of the linker's conformational changes.

We have modified the presentation of the figures to clarify the SAXS analysis. The fact that the SAXS analysis suggests flexibility rather than a discrete inactive conformation means that high-resolution techniques may not be appropriate for this system.

(8) The study primarily focuses on the model where RsbT binding rigidifies the RsbU linker. Alternative hypotheses, such as subtle conformational adjustments without complete rigidification, are not extensively explored or ruled out.

Our analysis of the SAXS data strongly suggests that a subtle conformational change could not account for the scattering data that we obtained. We have modified the text to clarify this point.

“Indicative of significant deviation between the RsbU structure in solution to the AlphaFold2 model, the scattering intensity profile (I(q) vs. q) was a poor fit (χ² 12.53) to a profile calculated from the AlphaFold2 model of an RsbU dimer using FoXS (Schneidman-Duhovny et al. 2016; Schneidman-Duhovny et al. 2013) (Fig. 4A). We therefore assessed the SAXS data for the RsbU dimer for features that report on flexibility (Kikhney & Svergun 2015). First, the scattering intensity data lacked distinct features caused by the multi-domain structure of RsbU from the AlphaFold2 model (Fig.4A).”

(9) Future studies should aim to validate the AlphaFold2 predictions with high-resolution structural techniques. This would provide definitive evidence for the proposed conformational states of RsbU with and without RsbT.

The fact that the SAXS analysis suggests flexibility rather than a discrete inactive conformation means that high-resolution techniques may not be appropriate for this system.

(10) Investigating the RsbU-RsbT interaction in vivo using techniques like FRET, co-immunoprecipitation, or live-cell imaging would provide a more comprehensive understanding of their functional dynamics in a cellular context.

We appreciate the reviewer’s suggestions for future experiments.

(11) Exploring and testing alternative models of RsbU activation, such as partial rigidification or different modes of conformational change, would strengthen the conclusions.

While our data strongly support that a flexible-to-rigid transition controls RsbU activation, we agree that it is possible that other mechanisms of linker modification could control other phosphatases and we discuss this at some length in the discussion.

(12) The figure legends are quite dense and could benefit from some streamlining.

We have edited the figure legends for clarity and length.

Reviewer #2 (Recommendations for the authors):

(1) Activation assays (Figures 1, 3, S2) are presented here as blue or white spots (reflecting a reporter activity). While off and on these are fairly clear, it is more difficult to compare the degree of activity (for instance that rsbU^Q94L is more active than M166V). It would also be good to clearly present in the text the logic of asking if the mutant is RsbT independent or not (and the interpretation of that). Quantitative assays of these would be very useful.

We chose not to perform quantitative-LacZ assays here because of several complications to interpreting these results that we encountered in our previously published study (Ho and Bradshaw, 2021). However, the level of blue pigmentation shown in Figure 1B for RsbU Q94L and RsbU M166V is qualitatively different, making the comparison possible. Most importantly, we observed cell density dependent changes in LacZ activity in the absence of rsbT for rsbU^M166V expressing cells, meaning that comparisons between strains would be difficult. Additionally, we found that it was important to make a chromosomal replacement of rsbU to see the full effect of the M166V substitution. However, we were not able to construct a similar rsbU^Q94L strain, likely because the high level σ^B activity is lethal (we were able to construct this strain when σ^B was deleted but only obtained strains with additional loss-of-function mutations in RsbU when σ^B was present.

We have modified the text to explain the logic of identifying RsbT independent variants: “We previously conducted a genetic screen (Ho & Bradshaw 2021) to identify features of RsbU that are important for phosphatase regulation by isolating gain-of-function variants that are active in the absence of RsbT.”

(2) Explain Figure S8 graphs: as much as Alphafold is now in use, the authors should provide some further explanation of what is shown here. Blue (low error) is good, presumably. What are the A, B, C, and D sections showing? Different parts of a given letter region (and between them)? What is the x-axis? Is the top-ranked model used in every case in the text? How different are these models? The Methods section could be used for some of this (but doesn't in its current form). This also becomes important for the models generated later in the paper (Figure S7), which look rather different here.

We have modified figure S8 to include additional labels and have added structures with the pLDDT scores shown. We have additionally modified the figure legends and methods to provide the requested information.

(3) Figure 1C, D, Figure S2: amino acid ends of linker domains could be shown (text discusses 83-97 the linker as a two-turn coiled coil; Q94 is pretty close to the end of this coiled-coil? Figure S2 is even less clear - addresses of other amino acids would help, and or an added sequence showing the full linker and coiled-coil region). Some explanation for positions for readers to focus on for full coiled-coil would be useful in the legend of Figure S2. How strong a coiled-coil prediction is there for this region?

We have added the sequence of the coiled-coil regions to the figures with numbering. For these analyses we used the Socket2 program, which analyzes a PDB file to identify coiled-coil regions and thus does not provide a confidence score. However, inspection of the sequence and the confidence scores of the AlphaFold2 models indicates that the coiled-coil regions are not ideal, consistent with this being a regulatory feature.

Is it clear that the fully inactive proteins are still properly folded and soluble?

In the case of RsbU, our biophysical analysis indicates that the inactive form of the protein is soluble. While phosphatase activity is substantially reduced, our unpublished comparison of single- and multiple-turnover reactions in the absence of RsbT indicates that nearly all of the enzyme is active.

Finally, are there other positions that would also be expected, from this model, to stabilize the coiled-coil and thus bypass the requirement for RsbT? If so, it would be good to test these. Is it the burial of amino acid at position 94 that is important, or the ability to form crossed helices?

Because of how short the predicted coiled-coil region is, we did not identify any obvious positions that would likely have the same effect as Q94 substitution. We considered making helix-breaking mutations, which would be predicted to block RsbU activation, but favored analysis of the wildtype protein because of limitations in interpreting the effects of loss-of-function mutations.

(4) Figure 2A, RsbT binding to RsbU: It was not entirely clear to this reviewer why one would expect the RsbT binding, not needed for activation, to be increased by the mutation that stabilizes the crossed alpha helices. The change is impressive but doesn't the lack of a need for RsbT suggest that this mutation bypasses the normal mechanism? (Is dimerization enuf? Or other protein cross helices?).

We have modified the text to clarify this point: “One prediction of our hypothesis that RsbT stabilizes the crossed alpha helices of the RsbU dimer, is that RsbT should bind more tightly to rsbU^Q94L than to RsbU because the coiled-coil conformation that RsbT binds would be more energetically favorable.” Another way of putting this is that if the Q94L substitution activates RsbU through an on-pathway mechanism, RsbT must bind more tightly.

(5) Figure 3A, Figure S3: Please label the yellow (interface) residues in RsbU and RsbT in Fig. S3 and the green (suppressor) spheres in Figure 3A.

We have added labels to the figures as suggested.

If RbsT interacts with the N-terminal dimerization domain and linker, why were residues 174 and 178 (from PPM domain) shown to be implicated in binding?

The fact that residues in the switch region suppress a mutation that decreases RsbT binding suggests that this region is part of an allosteric network that links RsbT binding, the linker, and dimerization of the phosphatase domains. For example, any substitution that promotes a conformation of the phosphatase domain that is more favorable for dimerization would also promote RsbT binding. However, the precise details of how each mutation fits into this network is not clear and we have therefore chosen to not specify a particular model to avoid over interpreting our data.

Are these marked in Figure S3?

We have added labels to make this clear.

Are these part of a dimerization interface in the C-terminal domain? Are any/all of these RsbU mutants suppressed by Q94L, as one might predict (apparently Y28I is since Q94L was again identified)?

We chose to focus on Y28I because it was the best studied previously, but we would predict that Q94L would suppress other RsbT binding mutations.

(6) Line 191-192: Is it surprising that no suppressors were isolated in RsbT?

We didn’t have a preconception of whether or not it would be possible to identify similar suppressors in RsbT. Explanations for why we did not identify such suppressors could include that RsbT may be destabilized more easily by substitution, that RsbT is more constrained because it has other interaction partners, or that the particular substitutions that would suppress Y28I are less common by the PCR mutagenesis strategy we used.

(7) Figure 3: Would the same mutants arise if the screen had been done in the absence of RsbT? Was RsbT-dependent tested for the rsbU alleles?

Our prediction is that we would not have identified any of these mutations except for Q94L in the absence of rsbT. We tested a few of the alleles and found them all to be rsbT dependent, but did not systematically test all of the alleles and therefore did not include this analysis in the manuscript.

Given the findings earlier in the paper for Q94L, suggesting that this stabilizes the coiled-coil and shows some activity in the absence of RsbT, it seems that the interpretation of other mutants in this region (and Q94L itself) as evidence that RsbT contacts the linker directly and that contact is necessary for activation may be an overinterpretation. If these are in fact RsbT independent, they support the importance of the linker (do they further stabilize coiled-coil formation?), rather than the role of RsbT here. Are G92 and T89 on the outside of the coiled-coil? If Q94 is buried, is it qualitatively different from these others?

G92 and T89 are predicted to be exposed. The fact that these mutations are near Q94 is part of the reason that we focused on R91 and the predicted contact with D92 of RsbT as another approach to validate the predicted interface.

(8) Figure 3C addresses the issue of direct interaction of RsbT with the RsbU linker to some extent, given that RsbU R91E doesn't appear to have a lot of activity without RsbT. It would be helped by telling the reader what the R91 contact is initially.

We have modified the text to clarify this point: “To test the model that RsbT activates RsbU by directly interacting with the linker to dimerize the RsbU phosphatase domains, we introduced a charge swap at position R91 that would abolish a predicted salt-bridge with RsbT D92 (Fig. 3C).”

(9) Figure 4 and the discussion of it in the text is not likely to be easily understandable for many readers. Aside from providing a bit more explanation of what these analyses are showing, it would be useful to start the whole section (or maybe even much earlier in the paper) with the information found on lines 261-264, that other studies show that the N-terminus dimerizes stably on its own (and is it known that the C-terminus does not?). Then the discussion of the alternative models early in this section would be clearer.

We have updated the introduction to emphasize this point “RsbU has an N-terminal four-helix bundle domain that dimerizes RsbU and is also the binding site for RsbT, which activates RsbU as a phosphatase (Fig. 1C,D) (Delumeau et al. 2004).”

We have also added clarification to the model presented at the beginning of this section: “A second possibility is that inactive RsbU is dimerized by the N-terminal domains but that the linkers of inactive RsbU are flexible and that the phosphatase domains only interact with each other when RsbT orders the linkers into a crossing conformation.”

Is the dimerization of the N-terminal domains previously determined similar/the same as what is seen in the AlphaFold models used here (or the AlphaFold dimerization derived primarily from that data?).

Yes, the dimerization in the AlphaFold models matches closely to the published structure.

(10) Discussion and Figure 5: The final part of this work predicts AlphaFold models for a set of other phosphatases involved in initiating GSR across bacterial species, and suggests that linked-mediated phosphatase dimerization is the critical factor to activate the phosphatase. Clearly, this is the most speculative but interesting aspect of the paper. A number of possible questions are suggested by some of this:

a. Do any of the activating mutants In RsbU and RsbP in the PPM domain (that apparently improve dimerization and thus activation) do a similar job in the other modeled proteins?

This is an interesting question, but unfortunately most of these proteins have not been biochemically characterized. We highlight examples of RsbP and E. coli RssB for which similar activating mutations have been characterized.

b. The legend (Figure 5G) suggests that all of the linker combinations will be coiled-coils, but that they will undergo different types of activating (and dimerizing?) transitions. Is that in fact what is being proposed here?

Yes, this is our working hypothesis.

c. If there is no dimerization (as noted, only weak dimerization has been reported for E. coli RssB), does that generalize the model to there are linkers and their structures are important? At the least, would the folding up of the E. coli RssB linker with antiadaptor binding be considered another mode of signal transduction or rather some sort of storage form?

Interestingly, the P. aeruginosa RssB constitutively dimerizes, suggesting the E. coli is the outlier.

d. Would the "toolkit" model, in which different changes occur in the linker regions, suggest that the interacting proteins are going to be critical for the type of linker changes that will be important? Or something about the nature of the linkers themselves?

This is an interesting question that we cannot yet answer. We have chosen to focus on the possible flexibility of this mechanism and anticipate that a variety of mechanisms will be used.

e. Given the extensive comparison to E. coli RssB, the authors might consider a figure to clarify the relative domain architecture, sequences that are akin to switch regions, and others important to the discussion here.

We tried to highlight this in Figure 5C including coloring the regions similar to the switch regions.

Reviewer #3 (Recommendations for the authors):

Given the caveats noted above related to the reliability of computed structure models, I would recommend the authors make the following additions/modifications to their manuscript:

(1) The authors should show alpha fold models coloured by pLDDT scores in an additional supplementary figure to help the reader interpret the confidence level of the predicted structures.

We have added these models to figure 1 – figure supplement 2.

(2) Because of the points mentioned above the authors should tone down the generalisation relating to the activation mechanism of this family of phosphatases presented in the discussion.

We have modified the paper throughout to emphasize where we are speculating.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1:

Summary:

Kimura et al performed a saturation mutagenesis study of CDKN2A to assess functionality of all possible missense variants and compare them to previously identified pathogenic variants. They also compared their assay result with those from in silico predictors.

Strengths:

CDKN2A is an important gene that modulate cell cycle and apoptosis; therefore it is critical to accurately assess functionality of missense variants. Overall, the paper reads well and touches upon major discoveries in a logical manner.

Weaknesses:

The paper lacks proper details for experiments and basic data, leaving the results less convincing. Analyses are superficial and does not provide variant-level resolution. Many of which were addressed during the revision process.

Comments on revisions:

The manuscript was improved during the revision process.

We thank the reviewer for their comments. We are grateful for the opportunity to provide additional information and data to clarify our approach and study results.

Reviewer #2:

Summary:

This study describes a deep mutational scan across CDKN2A using suppression of cell proliferation in pancreatic adenocarcinoma cells as a readout for CDKN2A function. The results are also compared to in silico variant predictors currently utilized by the current diagnostic frameworks to gauge these predictors' performance. The authors also functionally classify CDKN2A somatic mutations in cancers across different tissues.

Review:

The goal of this paper was to perform functional classification of missense mutations in CDKN2A in order to generate a resource to aid in clinical interpretation of CDKN2A genetic variants identified in clinical sequencing. In our initial review, we concluded that this paper was difficult to review because there was a lack of primary data and experimental detail. The authors have significantly improved the clarity, methodological detail and data exposition in this revision, facilitating a fuller scientific review. Based on the data provided we do not think the functional characterization of CDKN2A variants is robust or complete enough to meet the stated goal of aiding clinical variant interpretation. We think the underlying assay could be used for this purpose but different experimental design choices and more replication would be required for these data to be useful. Alternatively, the authors could also focus on novel CDKN2A variants as there seems to be potential gain of function mutations that are simply lumped into "neutral" that may have important biological implications.

Major concerns:

Low experimental concordance. The p-value scatter plot (Figure 2 Figure Supplement 3A) across 560 variants shows low collinearity indicating poor replicability. These data should be shown in log2fold changes, but even after model fitting with the gamma GLM still show low concordance which casts strong doubt on the function scores.

Concordance among non-significant p-values is generally low because most of the signal comes from random variability across repeats. If the observed log2 fold change between the repeats is entirely due to noise, one would expect two repeated p-values to behave like independent random uniforms. True concordance is typically more evident in significant p-values because they reflect consistent effects above random noise. Functionally deleterious variants are called when their associated p-value is significant. To confirm this statement, a scatter plot with the log2 normalized fold change was added in Figure 2 Supplement 3C. We see low concordance between repeats in the log2 normalized fold changes centered around 0, corresponding to log log2 normalized changes mainly due to noise. The concordance increases as the variants become significant. One can notice that the correlation coefficient between duplicate assay results was almost identical between the model-based p-values and log2normalized fold change (Figure 2-figure supplement 3A and 3C, Appendix 1-table 4, and Appendix 1-table 6). Also, importantly, no variant was functionally deleterious in one replicate and functionally neutral in another, implying a perfect concordance in calls if we exclude variants that were called indeterminate in one of the two repeats. Finally, of variants with discordant classifications, only 6/560 repeats (1.1%) were functionally deleterious (significant p-value) in one replicate and of indeterminate function in another. We have updated the text as follows:

“Of variants with discordant classifications, 6 (1.1%) were functionally deleterious in one replicate and of indeterminate function in another. While 102 variants (18.2%) were functionally neutral in one replicate and of indeterminate function in another. Importantly, no variant that was functionally deleterious in one replicate and functionally neutral in another (Appendix 1 -table 4). Furthermore, the correlation coefficient between duplicate assay results was similar using the gamma GLM and log2 normalized fold change (Figure 2-figure supplement 3A and 3C).”

The more detailed methods provided indicate that the growth suppression experiment is done in 156 pools with each pool consisting of the 20 variants corresponding to one of the 156 aa positions in CKDN2A. There are several serious problems with this design.

Batch effects in each of the pools preventing comparison across different residues. We think this is a serious design flaw and not standard for how these deep mutational scans are done. The standard would be to combine all 156 pools in a single experiment. Given the sequencing strategy of dividing up CDKN2A into 3 segments, the 156 pools could easily have been collapsed into 3 (1 to 53, 54 to 110, 111 to 156). This would significantly minimize variation in handling between variants at each residue and would be more manageable for performance of further replicates of the screen for reproducibility purposes. The huge variation in confluency time 16-40 days for each pool suggest that this batch effect is a strong source of variation in the experiment.

While there is variation in time to confluency between different amino acid residues, we do not anticipate this batch effect to significantly affect variant classifications in our study. For example, our results were generally consistent with previous classifications. All synonymous variants (one per residue) and benchmark benign variants assayed were classified as functionally neutral. Furthermore, of benchmark pathogenic variants assayed, none were classified as functionally neutral. 84% were classified as functionally deleterious and 16 percent were classified as indeterminate function.

Lack of experimental/biological replication: The functional assay was only performed once on all 156 CDKN2A residues and was repeated for only 28 out of 156 residues, with only ~80% concordance in functional classification between the first and second screens. This is not sufficiently robust for variant interpretation. Why was the experiment not performed more than once for most aa sites?

In our study we determined functional classifications for all CDKN2A missense variants while assessing variability with replicates across 28 residues. Of these variants, only 6 (1.1%) were functionally deleterious in one replicate and of indeterminate function in another. Furthermore, no variant was functionally deleterious in one replicate and functionally neutral in another (Appendix 1 -table 4).  As noted above, we provided additional context in the manuscript.

For the screen, the methods section states that PANC-1 cells were infected at MOI=1 while the standard is an MOI of 0.3-0.5 to minimize multiple variants integrating into a single cell. At an MOI =1 under a Poisson process which captures viral integration, ~25% of cells would have more than 1 lentiviral integrant. So in 25% of the cells the effect of a variant would be confounded by one or more other variants adding noise to the assay.

As noted previously, we are not able to differentiate effects due to multiple viral integrations per cells. However, we do not anticipate multiple viral integrations to significantly affect variant classifications in our study as our results are consistent with previous classifications, as described above.

While the authors provide more explanation of the gamma GLM, we strongly advise that the heatmap and replicate correlations be shown with the log2 fold changes rather than the fit output of the p-values.

Thank you for the suggestion. As noted, we provide additional explanation in the manuscript about why we classified variants using a gamma GLM. Using a gamma GLM, classification thresholds were determined using the change in representation of 20 non-functional barcodes in a pool of PANC-1 cells stably expressing CDKN2A after a period of in vitro proliferation. Our variant classifications were therefore not based on assay outputs for previously reported – benchmark – pathogenic or begin variants to determine thresholds. We strongly prefer using p-values and classifications using the gamma GLM in the manuscript. However, comparison of assay outputs using a gamma GLM and log2 fold change are included in the manuscript. Read counts, log2 fold change, and classifications based on log2 fold change are presented in the manuscript, for all variants. Readers who wish to use these data may do so and we refer them to the manuscript text, Appendix 1 -table 4, Appendix 1 -table 6, and Figure 2 -figure supplement 2.

In this study, the authors only classify variants into the categories "neutral", "indeterminate", or "deleterious" but they do not address CDKN2A gain-of-function variants that may lead to decreased proliferation. For example, there is no discussion on variants at residue 104, whose proliferation values mostly consist of higher magnitude negative log2fold change values. These variants are defined as neutral but from the one replicate of the experiment performed, they appear to be potential gain-of-function variants.

We have added a comment to the discussion to highlight that we did not identify potential gain-of-function variants. Specifically:

“We classified CDKN2A missense variants using a gamma GLM, as either functionally deleterious, indeterminate functional or functionally neutral. However, we did not classify variants that may have gain-of-function effects, resulting in decreased representation in the cell pool. Future studies are necessary to determine the prevalence and significance of CDKN2A gain-of-function variants.”

Minor concerns:

The differentiation between variants of "neutral" and "indeterminate" function seems unnecessary and it seems like there are too many variants that fall into the "indeterminate" category. The authors seem to have set numerical thresholds for CDKN2A function using benchmark variants of known function. While the benchmark variants are important as a frame of reference for the "dynamic range" of the assay, their function scores should not necessarily be used to define hard cutoffs of whether a variant's function score can be interpreted.

We did not utilize benchmark variants to define thresholds for functional classifications using a gamma GLM. This is one of the strengths of using a gamma GLM model for classification. As explained in our manuscript, classification thresholds were determined using the change in representation of 20 non-functional barcodes in a pool of PANC-1 cells stably expressing CDKN2A after a period of in vitro proliferation. Our variant classifications were therefore not based on assay outputs for previously reported – benchmark – pathogenic or begin variants. While not required when using a gamma GLM, we included indeterminate classifications, which are not uncommon.

Figure 2 supplement 2 - on the x-axis, should "intermediate" be "indeterminate"?

This, and a similar typographical error in Figure 2 -figure supplement 3, has been corrected.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #2 (Public Review):

This study elucidates the toxic effects of the lipid aldehyde trans-2-hexadecenal (t-2-hex). The authors show convincingly that t-2-hex induces a strong transcriptional response, leads to proteotoxic stress and causes the accumulation of mitochondrial precursor proteins in the cytosol.

The data shown are of high quality and well-controlled. The genetic screen for mutants that are hyper-and hypo-sensitive to t-2-hex is elegant and interesting, even if the mechanistic insights from the screen are rather limited. Moreover, the authors show evidence that t-2-hex affects subunits of the TOM complex. However, they do not formally demonstrate that the lipidation of a TOM subunit is responsible for the toxic effect of t-2-hex. A t-2-hex-resistant TOM mutant was not identified. Nevertheless, this is an interesting and inspiring study of high quality. The connection of proteostasis, mitochondrial biogenesis and sphingolipid metabolism is exciting and will certainly lead to many follow-up studies.

Reviewer #3 (Public Review):

Summary:

The authors investigate the effect of high concentrations of the lipid aldehyde trans-2-hexadecenal (t-2-hex) in a yeast deletion strain lacking the detoxification enzyme. Transcriptomic analyses as global read out reveal that a large range of cellular functions across all compartments are affected (transcriptomic changes affect 1/3 of all genes). The authors provide additional analyses, from which they built a model that mitochondrial protein import caused by modification of Tom40 is blocked.

Our initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant as an experimental approach is only a starting experiment and was aimed to identify as many determinants of t-2-hex toxicity as possible as stated in the manuscript. From this, we developed multiple independent approaches in wild-type (and mutant) cells at low t-2-hex concentrations, demonstrating that proteostasis and mitochondrial protein trafficking are physiologically important targets of the pro-apoptotic lipid. Specifically, proteostasis-specific PACE reporters are robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10 mM t-2-hex in detoxification competent wild type cells (new Figure 3F).

We do not propose Tom40 as the lipid's primary target, while we show that several subunits of the TOM (and TIM) complex are directly targeted by low t-2-hex concentrations in vitro (Figure 8B), and Tom20 and Tom70 are important for lipid toxicity (Figure 8D) and mitochondrial protein trafficking in vivo (Suppl. Figure 2).

Strengths:

Global analyses (transcriptomic and functional genomics approach) to obtain an overview of changes upon yeast treatment with high doses of t-2-hex.

Weaknesses:

The use of high concentrations of t-2-hex in combination with a deletion of the detoxifying enzyme Hfd1 limits the possibility to identify physiological relevant changes. From the hundreds of identified targets the authors focus on mitochondrial proteins, which are not clearly comprehensible from the data.

The initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant is a starting experiment and was aimed to identify as many determinants of t-2-hex toxicity as possible as stated in the manuscript. As stated (page 4), genes up-regulated (>2 log2FC) by t-2-hex were selected and subjected to GO category enrichment analysis (Supplemental Table 1). We found that “Mitochondrial organization” was the most numerous GO group activated by t-2-hex.  Among the strongly t-2-hex induced genes encoding mitochondrial proteins, CIS1 represented the most inducible gene with a known mitochondrial function. Cis1 is the central protein of the MitoCPR pathway, which is specifically induced upon and protects from mitochondrial protein import stress. We further show that proteostasis and mitochondrial protein trafficking are physiologically important targets at low t-2-hex doses in several independent experimental approaches: proteostasis-specific PACE reporters are robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10mM t-2-hex in detoxification competent wild type cells (new Figure 3F); mitochondrial pre-protein accumulation is induced by 10mM t-2-hex in wild type cells (Figure 5G); several subunits of the TOM and TIM complexes are lipidated by low (10mM) t-2-hex doses in wild type cell extracts (Figure 8B), mitochondrial import assays with mt-GFP in intact yeast wild type cells reveal that t-2-hex significantly inhibits import at low (5mM) t-2-hex concentrations (new Suppl. Figure 1). 5-10mM t-2-hex applied here is considerably lower than the published data in human cells with ³ 25mM on intact cells or cell extracts (Jarugumilli et al. 2018).

The main claim of the manuscript that t-2-hex targets the TOM complex and inhibits mitochondrial protein import is not supported by experimental data as import was not experimentally investigated. The observed accumulation of precursor proteins could have many other reasons (e.g. dissipation of membrane potential, defects in mitochondrial presequence proteases, defects in cytosolic chaperones, modification of mitochondrial precursors by t-2-hex rendering them aggregation prone and thus non-import competent). However, none of these alternative explanations have been experimentally addressed or discussed in the manuscript.

We have now performed additional experiments, alternative to the pre-protein quantifications, showing that t-2-hex specifically inhibits mitochondrial protein import. We investigated the effect of t-2-hex on mitochondrial protein import using flow cytometric GFP assays in live yeast cells. Specifically, we compared the expression and maturation of GFP targeted either to the cytosol or the mitochondrial matrix and show that low doses of t-2-hex (≥5 μM) significantly inhibited mt-GFP activity compared to cytosolic GFP in wild-type cells (new Supplemental Figure 1B). In contrast, this inhibition was not observed with the saturated derivative, t-2-hex-H2. Flow cytometric rhodamine123 assays revealed that t-2-hex did not alter ΔΨm within the concentration range that efficiently inhibits mt-GFP activity (new Supplemental Figure 1C). Alternative t-2-hex effects such as the direct modification of mitochondrial pre-proteins or cytosolic chaperones, potentially making the precursors prone to aggregation, are less likely, as the mitochondrial and cytosolic GFP used in these import studies differ only by the small, cysteine-free PreSu9 pre-peptide. This information is now included in the Results and Discussion sections.

Furthermore, many of the results have been reported before (interaction of Tom22 and Tom70 with Hfd1) or observed before (TOM40 as target of t-2-hex in human cells).

The interaction of Tom22 or Tom70 with Hfd1 has been only reported in high throughput pull-down studies in yeast (Opalinski et al., 2018 and Burri et al., 2006), and no functional connection between Hfd1 lipid detoxification and TOM has been investigated. Here we corroborate these high throughput results by targeted pull-down experiments, which strengthens the new finding that Hfd1 functionally interacts with the TOM complex. Tom40 has been found to be lipidated by high t-2-hex concentrations in human cell extracts in high throughput in vitro proteomic studies (Jarugumilli et al., 2018), but no functional connection between human TOM and t-2-hex has been investigated so far. Here we corroborate these high throughput results by targeted experiments, which strengthens the new findings that t-2-hex and TOM interact functionally.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

Congratulations on this exciting study. Even if some of the mechanistic details will have to be addressed in further studies (which of the modified sites are physiologically relevant; which sites are modified in vivo without external addition of t-2-hex) this study is inspiring and opens a new direction of mitochondrial research. I therefore fully support publication of this nice study in its current form.

Reviewer #3 (Recommendations For The Authors):

Two of the reviewers pointed out that the observation of precursors in whole cell extract is not sufficient to draw conclusions on mitochondrial protein import rates. The authors did not provide any new experiments but argued that a recent publication (Weidberg and Amon, 2018) had used the same readout for this conclusion. Why this manuscript was accepted with this statement is not known to this reviewer, but it does not change the fact, that the conclusion is not valid. Many alternative explanations are possible (see public review) and the claim that the import competence of the TOM complex is affected upon t-2-hex treatment is not appropriate.

We have now performed new experiments addressing the inhibition of mitochondrial protein import by t-2-hex as an alternative to our precursor accumulation assays. We compared the induced expression of cytosolic and mitochondrial GFP by flow cytometry as a quantitative mitochondrial import assay (Sirk et al., Cytometry A. 2003 Nov; 56(1) 15-22). Low doses of t-2-hex (≥5 μM) significantly inhibited mt-GFP activity as compared to cytosolic GFP in wild-type cells (new Supplemental Figure 1B). This inhibition of mitochondrial GFP is independent of mitochondrial membrane potential perturbation (new Supplemental Figure 1C) and alternative t-2-hex effects, such as the direct modification of the mtGFP precursor or cytosolic chaperones are less likely, as the mitochondrial and cytosolic GFP used in these import studies differ only by the small, cysteine-free PreSu9 pre-peptide.

The first sentence of the abstract states that t-2-hex „induces mitochondrial dysfunction in a conserved manner from yeast to human". I find two issues with this statement: 1) if the mechanism is known what is the question addressed in the present manuscript and 2) the second sentence of the results fully contradicts the above sentence „In human cells, t-2-hex causes mitochondrial dysfunction by directly stimulating Bax-oligomerisation at the outer mitochondrial membrane. In yeast, however, t-2-hex efficiently interferes with mitochondrial function and cell growth in a Bax independent manner."

We agree that the first sentence was misleading, this has been fixed now in the revised version.

The first reviewer requested a repetition of key experiments with lower concentrations and the authors provided additional in vitro data, however, for this, 10 uM is still very high. To gain valuable and physiological relevant data the initial transcriptomic analysis should be repeated with a low amount and in a wild-type yeast background.

Published t-2-hex chemoproteomic experiments on human cell extracts were performed at higher concentrations (>25mM) and human Bax is hardly lipidated by 10mM t-2-hex (Jarugumilli et al., 2018), therefore the in vitro lipidation data provided in our study should be considered a low t-2-hex dose. The initial transcriptomic study with high doses of t-2-hex in a detoxifying mutant is a starting experiment and was aimed at identifying as many determinants of t-2-hex toxicity as possible. Building on this, we further show that proteostasis and mitochondrial protein trafficking, the relevant cellular functions for our study, are physiologically important targets at low t-2-hex doses in several independent experimental approaches: proteostasis-specific gene expression is robustly induced in a detoxification mutant by 5mM t-2-hex (Figure 3D,E) and significantly induced by 10mM t-2-hex in detoxification competent wild type cells (new Figure 3F); mitochondrial pre-protein accumulation is induced by 10mM t-2-hex in wild type cells (Figure 5G); several subunits of the TOM and TIM complexes are lipidated by low (10mM) t-2-hex doses in vitro in wild type extracts (Figure 8B), mitochondrial import assays with mt-GFP in intact yeast wild type cells reveal that t-2-hex significantly inhibits import at low (5mM) t-2-hex concentrations (new Suppl. Figure 1).

As already stated above there are many alternative explanations for the observed accumulation of precursor proteins, e.g. the decreased proteasome activity could be cause and not consequence. Also, the modification of precursors directly upon translation in the cytosol could likely impact on their further transport and result in direct aggregation in the cytosol.

As mentioned above, we have now corroborated the t-2-hex specific mitochondrial protein import defect by alternative in vivo experiments, which are not dependent on the accumulation of mitochondrial precursors. We have tested now the possibility that decreased proteasome activity could indirectly inhibit mitochondrial import. This is not the case because a rpn4 mutant with impaired proteasomal activity induces normal mtGFP levels (new Suppl. Figure 1D). We cannot exclude that the modification of precursors by t-2-hex upon translation might additionally impact on the transport of some mitochondrial pre-proteins. However, mitochondrial and cytosolic GFP used in the import studies only differ in the small cysteine-free PreSu9 pre-peptide making it very unlikely that precursor lipidation is secondarily responsible for the observed import defect.

Many of the comments after first reviewing the manuscript were not addressed experimentally although many of the suggested experiments are easy to perform. I can only encourage the authors to provide more experimental support and controls, as the claims are currently not sufficiently supported.

In the two revisions of our manuscript, we have included several control experiments to better link the pro-apoptotic lipid t-2-hex with mitochondrial import stress. These include: in vitro lipidation of TOM/TIM subunits by low t-2-hex concentrations, t-2-hex tolerance and recovery of mitochondrial protein import in specific tom mutants, inhibition of mitochondrial protein import (pre-protein and mtGFP assays) by low t-2-hex doses independently on mitochondrial membrane potential and proteasome activity, and induction of proteostasis specific gene expression by low t-2-hex doses.

Author response:

Reviewer #1 (Public review):

Summary:

The authors have developed self-amplifying RNAs (saRNAs) encoding additional genes to suppress dsRNA-related inflammatory responses and cytokine release. Their results demonstrate that saRNA constructs encoding anti-inflammatory genes effectively reduce cytotoxicity and cytokine production, enhancing the potential of saRNAs. This work is significant for advancing saRNA therapeutics by mitigating unintended immune activation.

Strengths:

This study successfully demonstrates the concept of enhancing saRNA applications by encoding immune-suppressive genes. A key challenge for saRNA-based therapeutics, particularly for non-vaccine applications, is the innate immune response triggered by dsRNA recognition. By leveraging viral protein properties to suppress immunity, the authors provide a novel strategy to overcome this limitation. The study presents a well-designed approach with potential implications for improving saRNA stability and minimizing inflammatory side effects.

We thank Reviewer #1 for their thorough review and for recognizing both the significance of our work and the potential of our strategy to expand saRNA applications beyond vaccines.

Weaknesses:

(1) Impact on Cellular Translation:

The authors demonstrate that modified saRNAs with additional components enhance transgene expression by inhibiting dsRNA-sensing pathways. However, it is unclear whether these modifications influence global cellular translation beyond the expression of GFP and mScarlet-3 (which are encoded by the saRNA itself). Conducting a polysome profiling analysis or a puromycin labeling assay would clarify whether the modified saRNAs alter overall translation efficiency. This additional data would strengthen the conclusions regarding the specificity of dsRNA-sensing inhibition.

We thank the reviewer for this helpful insight and suggestion. We aim to conduct a puromycin labelling assay to clarify the effect of the various saRNA constructs on translation efficiency.

(2) Stability and Replication Efficiency of Long saRNA Constructs:

The saRNA constructs used in this study exceed 16 kb, making them more fragile and challenging to handle. Assessing their mRNA integrity and quality would be crucial to ensure their robustness.

Furthermore, the replicative capacity of the designed saRNAs should be confirmed. Since Figure 4 shows lower inflammatory cytokine production when encoding srIkBα and srIkBα-Smad7-SOCS1, it is important to determine whether this effect is due to reduced immune activation or impaired replication. Providing data on replication efficiency and expression levels of the encoded anti-inflammatory proteins would help rule out the possibility that reduced cytokine production is a consequence of lower replication.

This is another very helpful comment. We will conduct an analysis of saRNA integrity and quality by denaturing gel electrophoresis. To examine replicative capacity of the saRNA constructs, we aim to conduct RT-qPCR experiments.

(3) Comparative Data with Native saRNA:

Including native saRNA controls in Figures 5-7 would allow for a clearer assessment of the impact of additional genes on cytokine production. This comparison would help distinguish the effect of the encoded suppressor proteins from other potential factors.

Thank you for your suggestion. We will implement this change in the next version of the manuscript.

(4) In vivo Validation and Safety Considerations:

Have the authors considered evaluating the in vivo potential of these saRNA constructs? Conducting animal studies would provide stronger evidence for their therapeutic applicability. If in vivo experiments have not been performed, discussing potential challenges - such as saRNA persistence, biodistribution, and possible secondary effects-would be valuable.

(5) Immune Response to Viral Proteins:

Since the inhibitors of dsRNA-sensing proteins (E3, NSs, and L*) are viral proteins, they would be expected to induce an immune response. Analyzing these effects in vivo would add insight into the applicability of this approach.

We recognize the importance of in vivo studies and immune cell responses and plan to incorporate in vivo imaging in future studies to investigate these interactions, as well as examining delivery of various cargoes via saRNA to determine potential therapeutic benefits in different animal models of inflammatory pain, but such studies are beyond the scope of this current investigation. As suggested by the reviewer, we will incorporate a section on potential challenges of in vivo saRNA work in the revised manuscript.

(6) Streamlining the Discussion Section:

The discussion is quite lengthy. To improve readability, some content - such as the rationale for gene selection-could be moved to the Results section. Additionally, the descriptions of Figure 3 should be consolidated into a single section under a broader heading for improved coherence.

Thank you for your suggestions, we will make these changes in the next revision.

Reviewer #2 (Public review):

Summary:

Lim et al. have developed a self-amplifying RNA (saRNA) design that incorporates immunomodulatory viral proteins, and show that the novel design results in enhanced protein expression in vitro in mouse primary fibroblast-like synoviocytes. They test constructs including saRNA with the vaccinia virus E3 protein and another with E3, Toscana virus NS protein and Theiler's virus L protein (E3 + NS + L), and another with srIκBα-Smad7-SOCS1. They have also tested whether ML336, an antiviral, enables control of transgene expression.

Strengths:

The experiments are generally well-designed and offer mechanistic insight into the RNA-sensing pathways that confer enhanced saRNA expression. The experiments are carried out over a long timescale, which shows the enhance effect of the saRNA E3 design compared to the control. Furthermore, the inhibitors are shown to maintain the cell number, and reduce basal activation factor-⍺ levels.

We thank Reviewer #2 for their detailed assessment and recognition of the mechanistic insights provided by our study.

Weaknesses:

One limitation of this manuscript is that the RNA is not well characterized; some of the constructs are quite long and the RNA integrity has not been analyzed. Furthermore, for constructs with multiple proteins, it's imperative to confirm the expression of each protein to confirm that any therapeutic effect is from the effector protein (e.g. E3, NS, L). The ML336 was only tested at one concentration; it is standard in the field to do a dose-response curve. These experiments were all done in vitro in mouse cells, thus limiting the conclusion we can make about mechanisms in a human system.

We agree that these are weaknesses of our work. We plan to address some of these weaknesses by performing a dose response curve for ML336, examining saRNA integrity through denaturing gel electrophoresis, and will also aim to provide additional evidence for effects of effector proteins through RT-qPCR. We are also looking into testing these constructs in patient-derived FLS.

Author response:

Thanks for the positive review of our manuscript and for appreciating our work.

We align in many ways with the reviewers comments.. Our initial finding concerning the slight shift of f_free in a/b neurons after conditioning is interesting but we agree it would certainly deserve a follow-up to substantiate its link with memory formation. We also agree that an analysis in distribution rather than through an averaged signal might be more sensitive.

We however have to cope with the fact that extending our investigation would require manpower resources that are no longer available. Therefore we appreciate the suggestion made by the 3 reviewers to restrain the claim and hence change the title to "In vivo NAD(P)H autofluorescence lifetime imaging reveals metabolic heterogeneity within the Drosophila mushroom body.". We find it matches better with the scope of this study which is mostly to showcase the potential of NAD(P)H FLIM to quantify variations in metabolism in Drosophila brain rather than firmly testing a specific hypothesis linked to memory formation. In this respect, we do provide quantitative results showing metabolic profile variations between brain tissues such as the somata and calyx regions but also between different Kenyon cells subtypes. We would then present the shifts of f_free induced by conditioning as a curio that might entice future work, as advised by Reviewer #2.

Altogether, in the revised version we will change the title to restrain the claim, move two supplementary figures as main figures to better focus on and describe the registration process. We will also correct the figure panels pointed by the reviewers and add individual samples to our boxplots. We will also slightly compress the introduction and expand the discussion on potential applications. Finally, we will evaluate if statistical tests based on distributions may be more sensitive to observe a significant shift in FLIM signal in the a/b KCs after conditioning, to strengthen our last observation if confirmed.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review): Summary:

The authors demonstrate that two human preproprotein human mutations in the BMP4 gene cause a defect in proprotein cleavage and BMP4 mature ligand formation, leading to hypomorphic phenotypes in mouse knock-in alleles and in Xenopus embryo assays.

Strengths:

They provide compelling biochemical and in vivo analyses supporting their conclusions, showing the reduced processing of the proprotein and concomitant reduced mature BMP4 ligand protein from impressively mouse embryonic lysates. They perform excellent analysis of the embryo and post-natal phenotypes demonstrating the hypomorphic nature of these alleles. Interesting phenotypic differences between the S91C and E93G mutants are shown with excellent hypotheses for the differences. Their results support that BMP4 heterodimers act predominantly throughout embryogenesis whereas BMP4 homodimers play essential roles at later developmental stages.

Weaknesses:

(1) A control of BMP7 alone in the Xenopus assays seems important to excludeBMP7 homodimer activity in these assays.

We and other have shown that BMP7 homodimers have weak or no activity while BMP4/7 heterodimers single at a much higher level than either BMP4 or BMP7 homodimers in Xenopus ectodermal and mesodermal cells. We have expanded the description of these published findings in the results section (lines 182-187). We have also added representative examples of experiments in which BMP4 and BMP7 alone controls are included (new Fig. S2). Since the level of activity of BMP7 + BMP4 variants is equivalent to that of BMP7 + WT BMP4, this cannot be accounted for by BMP7 homodimers.

(2) The Discussion could be strengthened by more in-depth explanations of how BMP4 homodimer versus heterodimer signaling is supported by the results, so that readers do not have to think it all through themselves. Similarly, a discussion of why the S91C mutant has a stronger phenotype than E93G early in the Discussion would be helpful or least mention that it will be addressed later.

We have revised the discussion as suggested by the reviewer. Please see responses to recommendations 2-4 below.

Reviewer #1 (Recommendations for the authors):

(1) A control of BMP7 injection alone seems missing when comparing the BMP4/7 variants. BMP4 in the embryo assays presented in Fig 1. Is it not possible that the activity observed is BMP7 homodimers, perhaps due to inhibited heterodimer formation by the BMP4 variant?

Multiple published studies have shown that BMP7 homodimers have weak or no activity in Xenopus ectodermal and mesodermal cells, and that ½ dose of RNA encoding BMP4 and BMP7 together signals at a higher level than does a full dose of RNA encoding either BMP4 or BMP7 alone. We have expanded our description of these published findings (lines 182-187), have included additional details about RNA doses that were injected (line 156, 175, 182) and have added representative examples of experiments in which BMP4 and BMP7 controls were included in a new Figure (Fig. S2).

(2) In reading the Discussion, I was continually thinking of the stronger phenotype of the S91C mutant compared to the E93G one, although both are discussed together throughout most of the Discussion. Only at the end of the Discussion is the stronger phenotype of S91C discussed with a compelling explanation for the stronger phenotype, not related to the phosphorylation site function. I wonder if it would be better placed earlier in Discussion or at least mentioned the difference in phenotypes that will be discussed later.

We have moved the possible explanation of differences between Bmp4^S91C and Bmp4^E93G mutants to immediately follow the introductory paragraph of the results section.

(3) Along these same lines, why is it that the E93G exhibits rather normal cleavage at E10.5? Might the mechanisms of cleavage vary in different contexts with phosphorylation-dependent cleavage not functioning at early stages of development? I believe the hypothesis is that it is cleaved due to heterodimerization with BMP7. More discussion of this excellent hypothesis should be provided with clear statements, rather than inferences, if I'm understanding this correctly. For example, I had to read 3 times the first sentence of the last paragraph on p.14 before I understood it. Better to break that sentence down and the one that follows it, so it is easier to understand.

We have rewritten and expanded the paragraphs describing phenotypic and biochemical evidence for defective homodimer but not heterodimer signaling as suggested (lines 343-375). We have also more explicitly stated the possibility that normal cleavage of BMP4^E93G in embryonic lystates may be due to a predominance of BMP4/7 heterodimers in early embryonic stages or spatiotemporal differences in phosphorylation-dependent cleavage of BMP4 homodimers (lines 369-372)

(4) Similarly the last paragraph of the Discussion mentions that the authors provide evidence of BMP4 homodimer signaling. I agree with the authors, but I had to think through the evidence myself. Better if the authors clearly explain the evidence that points to this, as this is a very good point of

See response to point 3, above. Thank you for these useful suggestions.

(5) Last sentence, first paragraph on p.11 should be qualified for the E93G mutant to E13.5, since it was normal at E10.5 regarding Figure 4 results.

Thank you for pointing this out. It has been corrected.

(6) Skip the PC acronym, since it is only repeated once in the text and hard to remember almost 10 pages later when it is used again.

We have corrected this.

(7) In the Discussion, a typo in "a single intramolecular disulfide bond that stabilizes the dimer", should be 'intermolecular'.

Thank you for catching our switch in the use of inter- and intramolecular. We have corrected this (lines 334-335).

(8) At times the E93G mutant is referred to having early lethality, often in conjunction with S91C, while other times it is referred to as late lethality. Considering that the homozygotes die postnatally after weaning, most would consider it late lethality. In contrast S91C is indeed an early lethal.

We have changed the wording in the introduction to state that “mice carrying Bmp4^S91C or Bmp4^E93G knock in mutations show embryonic or enhanced postnatal lethality, respectively,… (lines 141-143)” and have removed the word “early” from the title.

Reviewer #2 (Public review): Summary:

Kim et al. report that two disease mutations in proBMP4, Ser91Cys and Glu93Gly, which disrupt the Ser91 FAM20C phosphorylation site, block the activation of proBMP4 homodimers. Consequently, analysis of DMZ explants from Xenopus embryos expressing the proBMP4 S91C or E93G mutants showed reduced pSmad1 and tbxt1 expression. The block in BMP4 activity caused by the mutations could be overcome by co-expression of BMP7, suggesting that the missense mutations selectively affect the activity of BMP4 homodimers but not BMP4/7 heterodimers. The expert amphibian tissue transplant studies were extended to in vivo studies in Bmp4S91C/+ and Bmp4E93G/+ mice, demonstrating the impact of these mutations on embryonic development, particularly in female mice, in line with patient studies. Finally, studies in MEFs revealed that the mutations did not affect proBMP4 glycosylation or ER-to-Golgi transport but appeared to inhibit the furin-dependent cleavage of proBMP4 to BMP4. Based on these findings and AI (AlphaFold) modeling of proBMP4, the authors speculate that pSer91 influences access of furin to its cleavage site at Arg289AlaLysArg292.

Strengths:

The Xenopus and mouse studies are valuable and elegantly describe the impact of the S91C and E93G disease mutations on BMP signaling and embryonic development.

Weaknesses:

The interpretation of how the mutations may disturb the furin-mediated cleavage of proBMP4 is underdeveloped and does not consider all of their data. Understanding how pS91 influences the furin-dependent cleavage at Arg292 seems to be the crux of this work and thus warrants more consideration. Specifically:

(1) Figure S1 may be significantly more informative than implied. The authors report that BMP4S91D activates pSmad1 only incrementally better than S91C and much less than WT BMP4. However, Fig. S1B does not support the conclusion on page 7 (numbering beginning with title page); "these findings suggest that phosphorylation of S91 is required to generate fully active BMP4 homodimers". The authors rightly note that the S91C change likely has manifold effects beyond inhibiting furin cleavage. The E93G change may also affect proBMP4 beyond disturbing FAM20C phosphorylation. Additional mutation analyses would strengthen the work.

The major goal of generating and comparing the activity of the S91D mutant with S91C was to control for phosphorylation independent defects cause by the deleterious introduction of a cysteine residue, which might cause aberrant disulfide bonding. We opted to introduce S91D since “phosphomimics” can sometimes approximate the phosphorylated state. S91D has significantly higher activity than S91C (p<0.01) and has a less significant loss of activity (p<0.05) than does S91C (<p<0.0001) relative to wild type BMP4 (Fig. S1), consistent with deleterious effects of the cysteine residue and supporting a possible explanation for the more severe phenotype of S91C vs E93G mice. We have rewritten this section to clarify our interpretation (lines 165-174)and have changed our statement that our activity data “suggest the importance of phosphorylation” to a statement that they are consistent with this possibility (lines 179-180). We do not believe that further mutational analysis using activity assays in Xenopus would shed light on how or whether phosphorylation affects proteolytic activation of BMP4.

(2) These findings in Figure S1 are potentially significant because they may inform how proBMP4 is protected from cleavage during transit through the TGN and entry into peripheral cellular compartments. Intriguing modeling studies in Figure 6 suggest that pSer91 is proximal to the furin cleavage site. Based on their presentation, pSer91 may contact Arg289, the critical P4 residue at the furin site. If so, might that suggest how pS91 may prevent furin cleavage, thus explaining why the S91D mutation inhibits processing as presented, and possibly how proBMP4 processing is delayed until transit to distal compartments (perhaps activated by a change in the endosomal microenvironment or a Ser91 phosphatase)? Have the authors considered or ruled out these possibilities? In addition to additional mutation analyses of the FAM20C site, moving the discussion of this model to an "Ideas and Speculation" subsection may be warranted.

The model shown in Fig. 6B proposes the possibility that phosphorylation unmasks (rather than preventing) the furin cleavage motif due to the proximity of Ser91 to the cleavage site (lines 399-402). If S91D truly mimicked phosphorylation, we would predict it would facilitate processing rather than inhibiting it. We do not have data comparing cleavage of S91D relative to wild type BMP4 and have not generated knock in S91D mice to test this idea. While the reviewers questions are intriguing, they cannot be answered by mutational analysis of the FAM20C site and are beyond the scope of the current studies that sought to understand the impact of human pS91C and pE93G mutations and cell biological implications. We have moved the models to an “Ideas and Speculation” subsection as suggested (lines 377-414) since these models are meant to provoke further thought rather than provide definitive answers based on our data.

(3) The lack of an in vitro protease assay to test the effect of the S91 mutations on furin cleavage is problematic.

Although we routinely perform in vitro cleavage assays with recombinant furin, we don’t believe they would be informative on how S91 phosphorylation or mutation of this residue impacts cleavage since in vitro synthesized substrate used in these assays is neither dimerized not post-translationally modified, and cleavage would be tested in isolation from the endogenous trafficking environment that we propose influences cleavage.

Reviewer #2 (Recommendations for the authors):

(1) The impact of BMPS91A should be determined and paired with the S91D phosphomimic data to reveal if it causes proBMP4 to be cleaved prematurely and disturbs pSmad1 expression. Data for S93G should also be included.

Our major goal in comparing the activity of S91D with S91C was to control for phosphorylation independent defects cause by the deleterious introduction of a cysteine residue in S91C, which might cause aberrant disulfide bonding. We opted to introduce S91D since “phosphomimics” can sometimes approximate the phosphorylated state. We note that S91D has significantly higher activity than S91C, consistent with deleterious effects of the cysteine residue and supporting a possible explanation for the more severe phenotype of S91C vs E93G mice. We have revised the wording of this section to clarify this. Our models predict that S91D would be cleaved more efficiently than S91C or S91A, if it really mimics the endogenous phosphorylated state, rather than being cleaved prematurely. Our biochemical analysis compares cleavage of endogenous BMP4 in wild type and mutant MEFs. Generation of S91D, S91A or S93G mutant mice to compare cleavage is beyond the scope of the current work.

(2) Is the distance between pS91 and Arg289 close enough to form a hydrogen bond? If so, might this interaction influence furin access?

AI modeling does not provide high probability prediction of structures surrounding the furin motif (see Fig. S7) and thus we cannot comment on whether or not these residues are close enough to form a hydrogen bond. We have revised the wording of the discussion to state “This simple model building indicates the possibility of direct contact between pSer91 and Arg289, and that phosphorylation is required for furin to access the cleavage site, although we note that predictions surrounding the furin motif represent low probability conformations (Fig. S7) (lines 399-402).”

(3) The genotypes in Figure 2 are labeled awkwardly. Consider labeling the headers for the three subsections of panels (A-F, G-L, and M-O) differently.

We have revised Fig. 2 to clarify that the three subsections of panels are distinct, and to emphasize that the middle subsection represents views of the right and left side of the same embryo.

(4) The tables should be reformatted. As is, the labeling is frequently cut off, and the numbers of expected and observed progeny should both be stated to aid the reader.

We thank the reviewer for noting the formatting errors in the tables, which we have corrected. We have also changed the tables so that normal or abnormal mendelian distributions are reported as numbers of observed/expected progeny rather than numbers/percent observed progeny.

Reviewer #3 (Public review):

Summary:

The authors describe important new biochemical elements in the synthesis of a class of critical developmental signaling molecules, BMP4. They also present a highly detailed description of developmental anomalies in mice bearing known human mutations at these specific elements.

Strengths:

Exceptionally detailed descriptions of pathologies occurring in mutant mice. Novel findings regarding the interaction of propeptide phosphorylation and convertase cleavage, both of which will move the field forward. Provocative hypothesis regarding furin access to cleavage sites, supported by Alphafold predictions.

Weaknesses:

Figure 6A presents two testable models for pre-release access of furin to cleavage sites since physical separation of enzyme from substrate only occurs in one model; could immunocytochemistry resolve?

Available reagents are not sensitive enough to detect endogenous furin and BMP4 with high resolution. Because PC/substrate interactions are transient, whereas the bulk of furin and BMP4 is distributed throughout the secretory pathway, it is not possible to co-immunolocalize furin and BMP4 in vivo at present. Studies using more advanced cell biological techniques such along with tagged proteins may enable us to test these hypotheses in the future.

Reviewer #3 (Recommendations for the authors):

This interesting paper presents new data on an important family of developmental signaling molecules, BMPs. Mutations at FAM20C consensus sites within BMP prodomains are known to cause birth defects. The authors have here explored differential effects of human mutations on hetero- and homodimer activity and maturation, issues that may well arise during human development. In addition to demonstrating the profound effect of these mutations on development in Xenopus and mice, the authors also show differential processing of BMP4 precursors bearing these mutations in MEF cells prepared from mutant embryos. Finally, they show that FAM20C plays a role in BMP4 prodomain processing with quite differing outcomes in homo- vs heterodimers, which they suggest is due to structural differences impacting furin access. While this latter idea remains speculative due to the lack of crystal structures (models are based on Alphafold) it is a highly promising line of work.

The data are beautifully presented and will be of clear interest to all developmental biologists. Certain cell biology results may also extrapolate to other phosphorylated precursor molecules undergoing the interesting (and as yet unexplained) phenomenon of convertase cleavage immediately before secretion, for example, FGF23. I have only a few minor comments regarding the presentation, which is remarkably clear.

(1) The introduction of BMP7 in the Abstract is abrupt. It should be described as a preferred dimerization partner for BMP4.

Thank you for noting this. We have revised the first sentence of the abstract to better introduce BMP7(lines 49-50).

(2) In Figure 1A, what is the small light green box?

This is a small fragment released from the prodomain by the second cleavage. We have clarified this in the introduction (lines 112-114) and in the legend to Figure 1 (lines 758-759).

(3) In the Discussion it might be relevant to mention that FAM20C propeptide is not cleaved by convertases but by S1P (Chen 2021).

We have added this information to clarify (lines 394-396).

(4) Figure 3, define VSD; Figure 5, Endo H removes sugars only from immature (nonsialylated) sugars, not from all chains as implied. More importantly, EndoH and PNGase remove N-linked sugars, yet Results refer only to O-linked glycosylation.

Thank you for noting these oversights. We have defined VSD in Figure 3. We have also revised the headers for Fig. 5 and for the relevant subsection of the results to include N-linked glycosylation and note in the results that EndoH removes only immature N-linked carbohydrates (lines 301-304).

(5) Figure 5- for clarity, I suggest it be broken up into two larger panels labeled "Embryos" and "MEFs"

Thank you for this suggestion, we have subdivided the Figure into two panels.

(6) Figure 6A presents two testable models for pre-release access of furin to cleavage sites since the physical separation of the enzyme from substrate only occurs in one model; could confocal immunocytochemistry resolve?

Available reagents are not sensitive enough to detect endogenous furin and BMP4 with high resolution and PC/substrate interactions are transient whereas the bulk of both furin and BMP4 is in transit through the secretory pathway. For these reasons it is not possible to co-immunolocalize furin and BMP4 in vivo. Future studies using advanced cell biological techniques may enable us to test these hypotheses in the future.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

This manuscript uses the eye lens as a model to investigate basic mechanisms in the Fgf signaling pathway. Understanding Fgf signaling is of broad importance to biologists as it is involved in the regulation of various developmental processes in different tissues/organs and is often misregulated in disease states. The Fgf pathway has been studied in embryonic lens development, namely with regards to its involvement in controlling events such as tissue invagination, vesicle formation, epithelium proliferation and cellular differentiation, thus making the lens a good system to uncover the mechanistic basis of how the modulation of this pathway drives specific outcomes. Previous work has suggested that proteins, other than the ones currently known (e.g., the adaptor protein Frs2), are likely involved in Fgfr signaling. The present study focuses on the role of Shp2 and Shc1 proteins in the recruitment of Grb2 in the events downstream of Fgfr activation.

Strengths:

The findings reveal that the juxtamembrane region of the Fgf receptor is necessary for proper control of downstream events such as facilitating key changes in transcription and cytoskeleton during tissue morphogenesis. The authors conditionally deleted all four Fgfrs in the mouse lens that resulted in molecular and morphological lens defects, most importantly, preventing the upregulation of the lens induction markers Sox2 and Foxe3 and the apical localization of F-actin, thus demonstrating the importance of Fgfrs in early lens development, i.e. during lens induction. They also examined the impact of deleting Fgfr1 and 2, on the following stage, i.e. lens vesicle development, which could be rescued by expressing constitutively active KrasG12D. By using specific mutations (e.g. Fgfr1ΔFrs lacking the Frs2 binding domain and Fgfr2LR harboring mutations that prevent binding of Frs2), it is demonstrated that the Frs2 binding site on Fgfr is necessary for specific events such as morphogenesis of lens vesicle. Further, by studying Shp2 mutations and deletions, the authors present a case for Shp2 protein to function in a context-specific manner in the role of an adaptor protein and a phosphatase enzyme. Finally, the key surprising finding from this study is that downstream of Fgfr signaling, Shc1 is an important alternative pathway - in addition to Shp2 - involved in the recruitment of Grb2 and in the subsequent activation of Ras. The methodologies, namely, mouse genetics and state-of-the-art cell/molecular/biochemical assays are appropriately used to collect the data, which are soundly interpreted to reach these important conclusions. Overall, these findings reveal the flexibility of the Fgf signaling pathway and it downstream mediators in regulating cellular events. This work is expected to be of broad interest to molecular and developmental biologists.

Weaknesses:

A weakness that needs to be discussed is that Le-Cre depends on Pax6 activation, and hence its use in specific gene deletion will not allow evaluation of the requirement of Fgfrs in the expression of Pax6 itself. But since this is the earliest Cre available for deletion in the lens, mentioning this in the discussion would make the readers aware of this issue.

Reviewer #2 (Public review):

Summary

I have reviewed the revised manuscript submitted by Wang et al., which is entitled "Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development". In this paper, the authors first examined lens phenotypes in mice with Le-Cre-mediated knockdown (KD) of all four FGFR (FGFR1-4), and found that pERK signals, Jag1 and foxe3 expression are absent or drastically reduced, indicating that FGF signaling is essential for lens induction. Next, the authors examined lens phenotypes of FGFR1/2-KD mice and found that lens fiber differentiation is compromised and that proliferative activity and cell survival are also compromised in lens epithelium. Interestingly, Kras activation rescues defects in lens growth and lens fiber differentiation in FGFR1/2-KD mice, indicating that Ras activation is a key step for lens development, downstream of FGF signaling. Next, the authors examined the role of Frs2, Shp2 and Grb2 in FGF signaling for lens development. They confirmed that lens fiber differentiation is compromised in FGFR1/3-KD mice combined with Frs2-dysfunctional FGFR2 mutants, which is similar to lens phenotypes of Grb2-KD mice. However, lens defects are milder in mice with Shp2YF/YF and Shp2CS mutant alleles, indicating that involvement of Shp2 is limited for the Grb2 recruitment for lens fiber differentiation. Lastly, the authors showed new evidence on the possibility that another adapter protein, Shc1, promotes Grb2 recruitment independent of Frs2/Shp2-mediated Grb2 recruitment.

Strength

Overall, the manuscript provides valuable data on how FGFR activation leads to Ras activation through the adapter platform of Frs2/Shp2/Grb2, which advances our understanding on complex modification of FGF signaling pathway. The authors applied a genetic approach using mice, whose methods and results are valid to support the conclusion. The discussion also well summarizes the significance of their findings.

Weakness

The authors found that the new adaptor protein Shc1 is involved in Grb2 recruitments in response to FGF receptor activation. However, the main data on Shc1 are only histological sections and statistical evaluation of lens size. In the revised manuscript, the authors did not answer my major concern that cellular-level data are missing, which is not fully enough to support their main conclusion on the involvement of Shc1 in Grb2 recruitment of FGF signaling for lens development. Since the title of this manuscript is that Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development, it is important to provide the cellular-level evidence on Shc1.

Reviewer #3 (Public review):

Summary:

The manuscript entitled "Shc1 cooperates with Frs2 and Shp2 to recruit Grb2 in FGF-induced lens development" by Wang et al., investigates the molecular mechanism used by FGFR signaling to support lens development. The lens has long been known to depend on FGFR-signaling for proper development. Previous investigations have demonstrated the FGFR signaling is required for embryonic lens cell survival and for lens fiber cell differentiation. The requirement of FGFR signaling for lens induction has remained more controversial as deletion of both Fgfr1 and Fgfr2 during lens placode formation does not prevent the induction of definitive lens markers such as FOXE3 or αA-crystallin. Here the authors have used the Le-Cre driver to delete all four FGFR genes from the developing lens placode demonstrating a definitive failure of lens induction in the absence of FGFR-signaling. The authors focused on FGFR1 and FGFR2, the two primary FGFRs present during early lens development and demonstrated that lens development could be significantly rescued in lenses lacking both FGFR1 and FGFR2 by expressing a constitutively active allele of KRAS. They also showed that the removal of pro-apoptotic genes Bax and Bak could also lead to a substantial rescue of lens development in lenses lacking both FGFR1 and FGFR2. In both cases, the lens rescue included both increased lens size and the expression of genes characteristic of lens cells.

Significantly the authors concentrated on the juxtamembrane domain, a portion of the FGFRs associated with FRS2. Previous investigations have demonstrated the importance of FRS2 activation for mediating a sustained level of ERK activation. FRS2 is known to associate both with GRB2 and SHP2 to activate RAS. The authors utilized a mutant allele of Fgfr1, lacking the entire juxtamembrane domain (Fgfr1ΔFrs) and an allele of Fgfr2 containing two-point mutations essential for Frs2 binding (Fgfr2LR). When combining three floxed alleles and leaving only one functional allele (Fgfr1ΔFrs or Fgfr2LR) the authors got strikingly different phenotypes. When only the Fgfr1ΔFrs allele was retained, the lens phenotype matched that of deleting both Fgfr1 and Fgfr2. However, when only the Fgfr2LR allele was retained the phenotype was significantly milder, primarily affecting lens fiber cell differentiation, suggesting that something other than FRS2 might be interacting with the juxtamembrane domain to support FGFR signaling in the lens. The authors also deleted Grb2 in the lens and showed that the phenotype was similar to that of the lenses only retaining the Fgfr2LR allele, resulting a failure of lens fiber cell differentiation and decreased lens cell survival. However, mutating the major tyrosine phosphorylation site of GRB2 did not affect lens development. The authors additionally investigated the role of SHP2 in lens development by either deleting SHP2 or by making mutations in the SHP2 catalytic domain. The deletion of the SHP2 phosphatase activity did not affect lens development as severely as total loss of SHP2 protein, suggesting a function for SHP2 outside of its catalytic activity. Although the loss of Shc1 alone has only a slight effect on lens size and pERK activation in the lens, the authors showed that the loss of Shc1 exacerbated the lens phenotype in lenses lacking both Frs2 and Shp2. The authors suggest that SHC1 binds to the FGFR juxtamembrane domain allowing for the recruitment of GRB2 in independently of FRS2.

Strengths:

(1) The authors used a variety of genetic tools to carefully dissect the essential signals downstream of FGFR signaling during lens development.

(2) The authors made a convincing case that something other than FRS2 binding mediates FGFR signaling in the juxtamembrane domain.

(3) The authors demonstrated that despite the requirement of both the adaptor function and phosphatase activity of SHP2 are required for embryonic survival, neither of these activities is absolutely required for lens development.

(4) The authors provide more information as to why FGFR loss has a phenotype much more severe than the loss of FRS2 alone during lens development.

(5) The authors followed up their work analyzing various signaling molecules in the context of lens development with biochemical analyses of FGF-induced phosphorylation in murine embryonic fibroblasts (MEFs).

(6) In general, this manuscript represents a Herculean effort to dissect FGFR signaling in vivo with biochemical backing with cell culture experiments in vitro.

Weaknesses:

(1) The authors demonstrate that the loss of FGFR1 and FGFR2 can be compensated by a constitutive active KRAS allele in the lens and suggest that FGFRs largely support lens development only by driving ERK activation. However, the authors also saw that lens development was substantially rescued by preventing apoptosis through the deletion of BAK and BAX. To my knowledge, the deletion of BAK and BAX should not independently activate ERK. The authors do not show whether ERK activation is restored in the BAK/BAX deficient lenses. Do the authors suggest the FGFR3 and/or FGFR4 provide sufficient RAS and ERK activation for lens development when apoptosis is suppressed? Alternatively, is it the survival function of FGFR-signaling as much as a direct effect on lens differentiation?

(2) Do the authors suggest that GRB2 is required for RAS activation and ultimately ERK activation? If so, do the authors suggest that ERK activation is not required for FGFR-signaling to mediate lens induction? This would follow considering that the GRB2 deficient lenses lack a problem with lens induction.

(3) The increase in p-Shc is only slightly higher in the Cre FGFR1f/f FGFR2r/LR than in the FGFR1f/Δfrs FGFR2f/f. Can the authors provide quantification?

(4) The authors have not shown directly that Shc1 binds to the juxtamembrane region of either Fgfr1 or Fgfr2.

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

In the revised manuscript, the authors have responded to my recommendations to revise the original manuscript, except for three suggestions below.

(1) The original recommendation: Results (page 6, line 8): The authors mentioned "we observed .... expression of Foxe3 in ...mutant lens cells (Figure 1E, arrows). However, Foxe3-expressing lens cells are a very small population in Figure 1E. It is important to state the decreased number of Foxe3-expressing lens cells in FGFR1/2 mutants. In addition, I would like to request the authors to show histograms indicating sample size and statistical analysis for marker expression: Foxe3 (Figure 1E), Prox1 and aA-crystallin (Fig. 1F), cyclin D1 and TUNEL (Fig. 1G) and pmTOR and pS6 (Supplementary figure 1B).

Author's response: We added a statement indicating that the number of Foxe3-expressing cells is reduced in FGFR1/2 mutants, which is now quantified in Fig. 1H. Quantifications for Cyclin D1 and TUNEL are now shown in Fig. 1I and J, respectively. However, we chose not to quantify Prox1, αA-crystallin, pmTOR, and pS6, as the FGFR1/2 mutants showed no staining for these markers.<br /> My recommendation: Although the authors have responded to revise the quantification of Foxe3-expressing cells, Cyclin D1 and TUNEL, they did not conduct statistical analysis of Prox1, αA-crystallin, pmTOR, and pS6, because of absence of these marker signals. I understand that no signal makes statistical analysis no meaningful. However, it is still important to indicate how many the authors repeated experiments to confirm the same result. Please indicate the number of biological replicates or independent experiments in the figure legends, for example "Biological replicates, n=3" or "Three independent experiments show similar results". As for pS6 labeling, there seems to be a weak signal in Supplementary Figure 1B, so please show statistical analysis to indicate its histogram.

We have added the number of biological replicates for Prox1 and αA staining in the legend of Fig.1. The review is correct that there is weak staining of pS6, and also pmTOR. The quantification of pS6 and pmTOR staining are now shown in Supplementary Fig. 1C and D.

(2) The original recommendation: Results (page 6, line 19- page 7, line 6): The authors showed that inducible expression of constitutive active Kras, KrasG12D, using Le-Cre, recovered lens size to the half level of wild-type control. However, in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D, pERK was detected in the most posterior edge of the lens fiber core, whereas pERK was detected in the broader area of the lens in control. Furthermore, pMEK was detected in the whole lens of mice with Le-Cre; FGFR1/2f/f; and LSL-KrasG12D, whereas pMEK was detected only in the lens epithelial cells at the equator. So, the spatial profile of pERK and pMEK expression was different from those of wild-type, although the authors observed that Prox1 and Crystallin expression are normally induced in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D. I wonder whether the lens normally develops in mice with Le-Cre; LSL-KrasG12D? Is the lens growth enhanced in mice with Le-Cre; LSL-KrasG12D? Please add the panels of mice with Le-Cre; LSL-KrasG12D in Figure 2B and 2C. In addition, I wonder whether apoptosis is suppressed in the lens of mice with Le-Cre; FGFR1/2f/f; LSL-KrasG12D?

Authors' response: Response: As we previously reported (Developmental Biology 355, 2011, 12-20), Le-Cre; LSL-KrasG12D did not lead to enhanced lens growth. While we agree that including images of Le-Cre; LSL-KrasG12D as controls in Fig. 2B and C and evaluating apoptosis in Le-Cre; FGFR1/2f/f; LSL-KrasG12D mutants would be appropriate, we regretfully no longer have these animals available to conduct these experiments.

My recommendation: I would like to suggest the authors conduct these experiments again, because the recovery of lens formation by Bax/Bak KD in Fgfr1/2 KD mice (Fig. 2F) suggests that KrasG12D activates the AKT-mediated cell survival pathway as well as that MEK/MAPK pathway downstream of FGF signaling pathway. Regarding the availability of mouse strains, in general, it is necessary to keep animal strains available for sincere response to reviewers' suggestions. Please clarify why these strains are now not available and justify the reason in the response to reviewers' recommendations.

We acknowledge the reviewer's suggested experiments. However, our research utilized multiple mouse strains that are costly to maintain, a challenge that was exacerbated during and after the COVID-19 pandemic. Unfortunately, we no longer have access to the specific mouse strains required to conduct these additional studies.

(3) The original recommendation: Figures 7E, and 7F: The authors showed that lens morphology and lens size evaluation in genetic combinations: control, Frs2/Shc1 KD, Frs2/Shp2 KD, and Frs2/Shp2/Shc1 KD. However, I would like to request the authors to show more detailed data in these genetic combinations, for example, pERK, foxe3, Maf, Prox1, Jag1, p57, cyclin D3, g-crystallin, and TUNEL.

Authors' response: Unfortunately, we no longer have these mutant mice to perform these detailed staining.

My recommendation: As I mentioned in the statement on weakness above, it is important to provide the cellular-level evidence to support the main conclusion on the involvement of Shc1 in Grb2 recruitment of FGF signaling for lens development, because this is the main novel finding in this manuscript. Regarding the availability of mouse strains, it is generally necessary to keep animal strains available for sincere response to reviewers' suggestions. Please clarify why these strains are now not available and justify the reason in the response to the reviewers' suggestions.

We regret that we did not anticipate these experiments suggested by the reviewer. Unfortunately, we are unable to perform these studies as we no longer maintain the required mouse strains in our colony.

Reviewer #3 (Recommendations for the authors):

The changes made by the authors improved the manuscript. I have no further suggestions.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

Chang and colleagues used tetrode recordings in behaving rats to study how learning an audiovisual discrimination task shapes multisensory interactions in the auditory cortex. They found that a significant fraction of neurons in the auditory cortex responded to visual (crossmodal) and audiovisual stimuli. Both auditory-responsive and visually-responsive neurons preferentially responded to the cue signaling the contralateral choice in the two-alternative forced choice task. Importantly, multisensory interactions were similarly specific for the congruent audiovisual pairing for the contralateral side.

Strengths:

The experiments were conducted in a rigorous manner. Particularly thorough are the comparisons across cohorts of rats trained in a control task, in a unisensory auditory discrimination task, and the multisensory task, while also varying the recording hemisphere and behavioral state (engaged vs. anesthesia). The resulting contrasts strengthen the authors' findings and rule out important alternative explanations. Through the comparisons, they show that the enhancements of multisensory responses in the auditory cortex are specific to the paired audiovisual stimulus and specific to contralateral choices in correct trials and thus dependent on learned associations in a task-engaged state.

We thank Reviewer #1 for the thorough review and valuable feedback.

Weaknesses:

The main result is that multisensory interactions are specific for contralateral paired audiovisual stimuli, which is consistent across experiments and interpretable as a learned task-dependent effect. However, the alternative interpretation of behavioral signals is crucial to rule out, which would also be specific to contralateral, correct trials in trained animals. Although the authors focus on the first 150 ms after cue onset, some of the temporal profiles of activity suggest that choice-related activity could confound some of the results.

We thank the reviewer for raising this important point regarding the potential influence of choice-related activity on our results. In our experimental setup, it is challenging to completely disentangle the effects of behavioral choice from multisensory interaction. However, we conducted relevant analyses to examine the influence of choice-related components on multisensory interaction.

First, we analyzed neural responses during incorrect trials and found a significant reduction in multisensory enhancement for the A^10k-V^vt pairing (Fig. 4). In contrast, for the A^3k-V^hz pairing, there was no strong multisensory interaction during either correct (right direction) or incorrect (left direction) choices. This finding suggests that the observed multisensory interactions are strongly associated with specific cue combinations during correct task performance.

Second, we conducted experiments with unisensory training, in which animals were trained separately on auditory and visual discriminations without explicit multisensory associations. The results demonstrated that unisensory training did not lead to the development of selective multisensory enhancement or congruent auditory-visual preferences, as observed in the multisensory training group. This indicates that the observed multisensory interactions in the auditory cortex are specific to multisensory training and cannot be attributed solely to behavioral signals or choice-related effects.

Finally, we specifically focused on the early 0-150 ms time window after cue onset in our main analyses to minimize contributions from motor-related or decision-related activity, which typically emerge later. This time window allowed us to capture early sensory processing while reducing potential confounds.

Together, these findings strongly suggest that the observed choice-dependent multisensory enhancement is a learned, task-dependent phenomenon that is specific to multisensory training.

The auditory stimuli appear to be encoded by short transient activity (in line with much of what we know about the auditory system), likely with onset latencies (not reported) of 15-30 ms. Stimulus identity can be decoded (Figure 2j) apparently with an onset latency around 50-75 ms (only the difference between A and AV groups is reported) and can be decoded near perfectly for an extended time window, without a dip in decoding performance that is observed in the mean activity Figure 2e. The dynamics of the response of the example neurons presented in Figures 2c and d and the average in 2e therefore do not entirely match the population decoding profile in 2j. Population decoding uses the population activity distribution, rather than the mean, so this is not inherently problematic. It suggests however that the stimulus identity can be decoded from later (choice-related?) activity. The dynamics of the population decoding accuracy are in line with the dynamics one could expect based on choice-related activity. Also the results in Figures S2e,f suggest differences between the two learned stimuli can be in the late phase of the response, not in the early phase.

We appreciate the reviewer’s detailed observations and questions regarding the dynamics of auditory responses and decoding profiles in our study. In our experiment, primary auditory cortex (A1) neurons exhibited short response latencies that meet the established criteria for auditory responses in A1, consistent with findings from many other studies conducted in both anesthetized and task-engaged animals. While the major responses typically occurred during the early period (0-150ms) after cue onset (see population response in Fig. 2e), individual neuronal responses in the whole population were generally dynamic, as illustrated in Figures 2c, 2d, and 3a–c. As the reviewer correctly noted, population decoding leverages the distribution of activity across neurons rather than the mean activity, which explains why the dynamics of population decoding accuracy align well with choice-related activity. This also accounts for the extended decoding window observed in Figure 2j, which does not entirely match the early population response profiles in Figure 2e.

To address the reviewer’s suggestion that differences between the two learned stimuli might arise in the late phase of the response, we conducted a cue selectivity analysis during the 151–300 ms period after cue onset. The results, shown below, indicate that neurons maintained cue selectivity in this late phase for each modality (Supplementary Fig. 5), though the selectivity was lower than in the early phase. However, interpreting this late-phase activity remains challenging. Since A^3k, V^hz, and A^3k-V^hz were associated with the right choice, and A^10k, V^vt, and A^10k-V^vt with the left choice, it is difficult to disentangle whether the responses reflect choice, sensory features, or a combination of both.

To further investigate, we examined multisensory interactions during the late phase, controlling for choice effects by calculating unisensory and multisensory responses within the same choice context. Our analysis revealed no evident multisensory enhancement for any auditory-visual pairing, nor significant differences between pairings—unlike the robust effects observed in the early phase (Supplementary Fig. 5). We hypothesize that early responses are predominantly sensory-driven and exhibit strong multisensory integration, whereas late responses likely reflect task-related, choice-related, or combined sensory-choice activity, where sensory-driven multisensory enhancement is less prominent. As the focus of this manuscript is on multisensory integration and cue selectivity, we prioritized a detailed analysis of the early phase, where these effects are most prominent. However, the complexity of interpreting late-phase activity remains a challenge and warrants further investigation. We cited Supplementary Fig. 5 in revised manuscript as the following:

“This resulted in a significantly higher mean MSI for the A^10k-V^vt pairing compared to the A^3k-V^hz pairing (0.047 ± 0.124 vs. 0.003 ± 0.096; paired t-test, p < 0.001). Among audiovisual neurons, this biasing is even more pronounced (enhanced vs. inhibited: 62 vs. 2 in A^10k-V^vt pairing, 6 vs. 13 in A^3k-V^hz pairing; mean MSI: 0.119±0.105 in A^10k-V^vt pairing vs. 0.020±0.083 A^3k-V^hz pairing, paired t-test, p<0.00001) (Fig. 3f). Unlike the early period (0-150ms after cue onset), no significant differences in multisensory integration were observed during the late period (151-300ms after cue onset) (Supplementary Fig. 5).”

First, it would help to have the same time axis across panels 2,c,d,e,j,k. Second, a careful temporal dissociation of when the central result of multisensory enhancements occurs in time would discriminate better early sensory processing-related effects versus later decision-related modulations.

Thank you for this valuable feedback. Regarding the first point, we used a shorter time axis in Fig. 2j-k to highlight how the presence of visual cues accelerates the decoding process. This visualization choice was intended to emphasize the early differences in processing speed. For the second point, we have carefully analyzed multisensory integration across different temporal windows. The results presented in the Supplementary Fig. 5 (also see above) already address the late phase, where our data show no evidence of multisensory enhancement for any auditory-visual pairings. This distinction helps clarify that the observed multisensory effects are primarily related to early sensory processing rather than later decision-related modulations. We hope this addresses the concerns raised and appreciate the opportunity to clarify these points.

In the abstract, the authors mention "a unique integration model", "selective multisensory enhancement for specific auditory-visual pairings", and "using this distinct integrative mechanisms". I would strongly recommend that the authors try to phrase their results more concretely, which I believe would benefit many readers, i.e. selective how (which neurons) and specific for which pairings?

We appreciate the reviewer’s suggestion to clarify our phrasing for better accessibility. To address this, we have revised the relevant sentence in the abstract as follows:

"This model employed selective multisensory enhancement for the auditory-visual pairing guiding the contralateral choice, which correlated with improved multisensory discrimination."

Reviewer #2 (Public review):

Summary

In this study, rats were trained to discriminate auditory frequency and visual form/orientation for both unisensory and coherently presented AV stimuli. Recordings were made in the auditory cortex during behaviour and compared to those obtained in various control animals/conditions. The central finding is that AC neurons preferentially represent the contralateral-conditioned stimulus - for the main animal cohort this was a 10k tone and a vertically oriented bar. Over 1/3rd of neurons in AC were either AV/V/A+V and while a variety of multisensory neurons were recorded, the dominant response was excitation by the correctly oriented visual stimulus (interestingly this preference was absent in the visual-only neurons). Animals performing a simple version of the task in which responses were contingent on the presence of a stimulus rather than its identity showed a smaller proportion of AV stimuli and did not exhibit a preference for contralateral conditioned stimuli. The contralateral conditioned dominance was substantially less under anesthesia in the trained animals and was present in a cohort of animals trained with the reverse left/right contingency. Population decoding showed that visual cues did not increase the performance of the decoder but accelerated the rate at which it saturated. Rats trained on auditory and then visual stimuli (rather than simultaneously with A/V/AV) showed many fewer integrative neurons.

Strengths

There is a lot that I like about this paper - the study is well-powered with multiple groups (free choice, reversed contingency, unisensory trained, anesthesia) which provides a lot of strength to their conclusions and there are many interesting details within the paper itself. Surprisingly few studies have attempted to address whether multisensory responses in the unisensory cortex contribute to behaviour - and the main one that attempted to address this question (Lemus et al., 2010, uncited by this study) showed that while present in AC, somatosensory responses did not appear to contribute to perception. The present manuscript suggests otherwise and critically does so in the context of a task in which animals exhibit a multisensory advantage (this was lacking in Lemus et al.,). The behaviour is robust, with AV stimuli eliciting superior performance to either auditory or visual unisensory stimuli (visual were slightly worse than auditory but both were well above chance).

We thank the reviewer for their positive evaluation of our study.

Weaknesses

I have a number of points that in my opinion require clarification and I have suggestions for ways in which the paper could be strengthened. In addition to these points, I admit to being slightly baffled by the response latencies; while I am not an expert in the rat, usually in the early sensory cortex auditory responses are significantly faster than visual ones (mirroring the relative first spike latencies of A1 and V1 and the different transduction mechanisms in the cochlea and retina). Yet here, the latencies look identical - if I draw a line down the pdf on the population level responses the peak of the visual and auditory is indistinguishable. This makes me wonder whether these are not sensory responses - yet, they look sensory (very tightly stimulus-locked). Are these latencies a consequence of this being AuD and not A1, or ... ? Have the authors performed movement-triggered analysis to illustrate that these responses are not related to movement out of the central port, or is it possible that both sounds and visual stimuli elicit characteristic whisking movements? Lastly, has the latency of the signals been measured (i.e. you generate and play them out synchronously, but is it possible that there is a delay on the audio channel introduced by the amp, which in turn makes it appear as if the neural signals are synchronous? If the latter were the case I wouldn't see it as a problem as many studies use a temporal offset in order to give the best chance of aligning signals in the brain, but this is such an obvious difference from what we would expect in other species that it requires some sort of explanation.

Thank you for your insightful comments. I appreciate the opportunity to clarify these points and strengthen our manuscript. Below, I address your concerns in detail:

We agree that auditory responses are typically faster than visual responses due to the distinct transduction mechanisms. However, in our experiment, we intentionally designed the stimulus setup to elicit auditory and visual responses within a similar time window to maximize the potential for multisensory integration. Specifically, we used pure tone sounds with a 15 ms ramp and visual stimuli generated by an LED array, which produce faster responses compared to mostly used light bars shown on a screen (see Supplementary Fig. 2a). The long ramp of the auditory stimulus slightly delayed auditory response onset, while the LED-generated bar (compared to the bar shown on the screen) elicited visual responses more quickly. This alignment likely facilitated the observed overlap in response latencies.

Neurons’ strong spontaneous activity in freely moving animals complicates the measurement of first spike latencies. Despite that, we still can infer the latency from robust cue-evoked responses. Supplementary Fig. 2b illustrates responses from an exemplar neuron (the same neuron as shown in Fig. 2c), where the auditory response begins 9 ms earlier than the visual response. Given the 28 ms auditory response latency observed here using 15 ms-ramp auditory stimulus, this value is consistent with prior studies in the primary auditory cortex usually using 5 ms ramp pure tones, where latencies typically range from 7 to 28 ms. Across the population (n=559), auditory responses consistently reached 0.5 of the mean Z-scored response 15 ms earlier than visual responses (Supplementary Fig. 2c). The use of Gaussian smoothing in PSTHs supports the reliability of using the 0.5 threshold as an onset latency marker. We cited Supplementary Fig. 2 in the revised manuscript within the Results section (also see the following):

“This suggests multisensory discrimination training enhances visual representation in the auditory cortex. To optimize the alignment of auditory and visual responses and reveal the greatest potential for multisensory integration, we used long-ramp pure tone auditory stimuli and quick LED-array-elicited visual stimuli (Supplementary Fig. 2). While auditory responses were still slightly earlier than visual responses, the temporal alignment was sufficient to support robust integration.”

We measured the time at which rats left the central port and confirmed that these times occur significantly later than the neuronal responses analyzed (see Fig. 1c-d). While we acknowledge the potential influence of movements such as whiskering, facial movements, head direction changes, or body movements on neuronal responses, precise monitoring of these behaviors in freely moving animals remains a technical challenge. However, given the tightly stimulus-locked nature of the neuronal responses observed, we believe they primarily reflect sensory processing rather than movement-related activity.

To ensure accurate synchronization of auditory and visual stimuli, we verified the latencies of our signals. The auditory and visual stimuli were generated and played out synchronously with no intentional delay introduced. The auditory amplifier used in our setup introduces minimal latency, and any such delay would have been accounted for during calibration. Importantly, even if a small delay existed, it would not undermine our findings, as many studies intentionally use temporal offsets to facilitate alignment of neural signals. Nonetheless, the temporal overlap observed here is primarily a result of our experimental design aimed at promoting multisensory integration.

We hope these clarifications address your concerns and highlight the robustness of our findings.

Reaction times were faster in the AV condition - it would be of interest to know whether this acceleration is sufficient to violate a race model, given the arbitrary pairing of these stimuli. This would give some insight into whether the animals are really integrating the sensory information. It would also be good to clarify whether the reaction time is the time taken to leave the center port or respond at the peripheral one.

We appreciate your request for clarification. In our analysis, reaction time (RT) is defined as the time taken for the animal to leave the center port after cue onset. This measure was chosen because it reflects the initial decision-making process and the integration of sensory information leading to action initiation. The time taken to respond at the peripheral port, commonly referred to as movement time, was not included in our RT measure. However, movement time data is available in our dataset, and we are open to further analysis if deemed necessary.

To determine whether the observed acceleration in RTs in the audiovisual (AV) condition reflects true multisensory integration rather than statistical facilitation, we tested for violations of the race model inequality (Miller, 1982). This approach establishes a bound for the probability of a response occurring within a given time interval under the assumption that the auditory (A) and visual (V) modalities operate independently. Specifically, we calculated cumulative distribution functions (CDFs) for the RTs in the A, V, and AV conditions (please see Author response image 1). In some rats, the AV_RTs exceeded the race model prediction at multiple time points, suggesting that the observed acceleration is not merely due to statistical facilitation but reflects true multisensory integration. Examples of these violations are shown in Panels a-b of the following figure. However, in other rats, the AV_RTs did not exceed the race model prediction, as illustrated in Author response image 1c-d.

This variability may be attributed to task-specific factors in our experimental design. For instance, the rats were not under time pressure to respond immediately after cue onset, as the task emphasized accuracy over speed. This lack of urgency may have influenced their behavioral responses and movement patterns. The race model is typically applied to assess multisensory integration in tasks where rapid responses are critical, often under conditions that incentivize speed (e.g., time-restricted tasks). In our study, the absence of strict temporal constraints may have reduced the likelihood of observing consistent violations of the race model. Furthermore, In our multisensory discrimination task, animals should discriminate multiple cues and make a behavioral choice have introduced additional variability in the degree of integration observed across individual animals. Additionally, factors such as a decline in thirst levels and physical performance as the task progressed may have significantly contributed to the variability in our results. These considerations are important for contextualizing the race model findings and interpreting the data within the framework of our experimental design.

Author response image 1.

Reaction time cumulative distribution functions (CDFs) and race model evaluation. (a) CDFs of reaction times (RTs) for auditory (blue), visual (green), and audiovisual stimuli (red) during the multisensory discrimination task. The summed CDF of the auditory and visual conditions (dashed purple, CDF_Miller) represents the race model prediction under independent sensory processing. The dashed yellow line represents the CDF of reaction times predicted by the race model. According to the race model inequality, the CDF for audiovisual stimuli (CDF_AV) should always lie below or to the right of the sum of CDF_A and CDF_V. In this example, the inequality is violated at nearly t = 200 ms, where CDF_AV is above CDF_Miller. (b) Data from another animal, showing similar results. (c, d) CDFs of reaction times for two other animals. In these cases, the CDFs follow the race model inequality, with CDF_AV consistently lying below or to the right of CDF_A + CDF_V.

The manuscript is very vague about the origin or responses - are these in AuD, A1, AuV... ? Some attempts to separate out responses if possible by laminar depth and certainly by field are necessary. It is known from other species that multisensory responses are more numerous, and show greater behavioural modulation in non-primary areas (e.g. Atilgan et al., 2018).

Thank you for highlighting the importance of specifying the origin of the recorded responses. In the manuscript, we have detailed the implantation process in both the Methods and Results sections, indicating that the tetrode array was targeted to the primary auditory cortex. Using a micromanipulator (RWD, Shenzhen, China), the tetrode array was precisely positioned at stereotaxic coordinates 3.5–5.5 mm posterior to bregma and 6.4 mm lateral to the midline, and advanced to a depth of approximately 2–2.8 mm from the brain surface, corresponding to the primary auditory cortex. Although our recordings were aimed at A1, it is likely that some neurons from AuD and/or AuV were also included due to the anatomical proximity.

In fact, in our unpublished data collected from AuD, we observed that over 50% of neurons responded to or were modulated by visual cues, consistent with findings from many other studies. This suggests that visual representations are more pronounced in AuD compared to A1. However, as noted in the manuscript, our primary focus was on A1, where we observed relatively fewer visual or audiovisual modulations in untrained rats.

Regarding laminar depth, we regret that we were unable to determine the specific laminar layers of the recorded neurons in this study, a limitation primarily due to the constraints of our recording setup.

Reviewer #3 (Public review):

Summary:

The manuscript by Chang et al. aims to investigate how the behavioral relevance of auditory and visual stimuli influences the way in which the primary auditory cortex encodes auditory, visual, and audiovisual information. The main result is that behavioral training induces an increase in the encoding of auditory and visual information and in multisensory enhancement that is mainly related to the choice located contralaterally with respect to the recorded hemisphere.

Strengths:

The manuscript reports the results of an elegant and well-planned experiment meant to investigate if the auditory cortex encodes visual information and how learning shapes visual responsiveness in the auditory cortex. Analyses are typically well done and properly address the questions raised.

We sincerely thank the reviewer for their thoughtful and positive evaluation of our study.

Weaknesses:

Major

(1) The authors apparently primarily focus their analyses of sensory-evoked responses in approximately the first 100 ms following stimulus onset. Even if I could not find an indication of which precise temporal range the authors used for analysis in the manuscript, this is the range where sensory-evoked responses are shown to occur in the manuscript figures. While this is a reasonable range for auditory evoked responses, the same cannot be said for visual responses, which commonly peak around 100-120 ms, in V1. In fact, the latency and overall shape of visual responses are quite different from typical visual responses, that are commonly shown to display a delay of up to 100 ms with respect to auditory responses. All traces that the authors show, instead, display visual responses strikingly overlapping with auditory ones, which is not in line with what one would expect based on our physiological understanding of cortical visually-evoked responses. Similarly, the fact that the onset of decoding accuracy (Figure 2j) anticipates during multisensory compared to auditory-only trials is hard to reconcile with the fact that visual responses have a later onset latency compared to auditory ones. The authors thus need to provide unequivocal evidence that the results they observe are truly visual in origin. This is especially important in view of the ever-growing literature showing that sensory cortices encode signals representing spontaneous motor actions, but also other forms of non-sensory information that can be taken prima facie to be of sensory origin. This is a problem that only now we realize has affected a lot of early literature, especially - but not only - in the field of multisensory processing. It is thus imperative that the authors provide evidence supporting the true visual nature of the activity reported during auditory and multisensory conditions, in both trained, free-choice, and anesthetized conditions. This could for example be achieved causally (e.g. via optogenetics) to provide the strongest evidence about the visual nature of the reported results, but it's up to the authors to identify a viable solution. This also applies to the enhancement of matched stimuli, that could potentially be explained in terms of spontaneous motor activity and/or pre-motor influences. In the absence of this evidence, I would discourage the author from drawing any conclusion about the visual nature of the observed activity in the auditory cortex.

We thank the reviewers for highlighting the critical issue of validating the sensory origin of the reported responses, particularly regarding the timing of visual responses and the potential confound of motor-related activity.

We analyzed neural responses within the first 150 ms following cue onset, as stated in the manuscript. This temporal window encompasses the peak of visual responses. The responses to visual stimuli occur predominantly within the first 100 ms after cue onset, preceding the initiation of body movements in behavioral tasks. This temporal dissociation aligns with previous studies, which demonstrate that motor-related activity in sensory cortices generally emerges later and is often associated with auditory rather than visual stimuli

We acknowledge that auditory responses are typically faster than visual responses due to distinct transduction mechanisms. However, in our experiment, we intentionally designed the stimulus setup to elicit auditory and visual responses within a similar time window to maximize the potential for multisensory integration. Specifically, we used pure tone sounds with a 15 ms ramp and visual stimuli generated by an LED array, which produce faster responses compared to commonly used light bars shown on a screen. The long ramp of the auditory stimulus slightly delayed auditory response onset, while the LED-generated bar elicited visual responses more quickly (Supplementary Fig. 2). This alignment facilitated the observed overlap in response latencies. As we measured in neurons with robust visual response, first spike latencies is approximately 40 ms, as exemplified by a neuron with a low spontaneous firing rate and a strong, stimulus-evoked response (Supplementary Fig. 4). Across the population (n = 559 neurons), auditory responses reached 0.5 of the mean Z-scored response 15 ms earlier than visual responses on average (Supplementary Fig. 2). We cited Supplementary Fig. 4 in the Results section as follows:

“Regarding the visual modality, 41% (80/196) of visually-responsive neurons showed a significant visual preference (Fig. 2f). The visual responses observed within the 0–150 ms window after cue onset were consistent and unlikely to result from visually evoked movement-related activity. This conclusion is supported by the early timing of the response (Fig. 2e) and exemplified by a neuron with a low spontaneous firing rate and a robust, stimulus-evoked response (Supplementary Fig. 4).”

We acknowledge the growing body of literature suggesting that sensory cortices can encode signals related to motor actions or non-sensory factors. To address this concern, we emphasize that visual responses were present not only during behavioral tasks but also in anesthetized conditions, where motor-related signals are absent. Additionally, movement-evoked responses tend to be stereotyped and non-discriminative. In contrast, the visual responses observed in our study were highly consistent and selective to visual cue properties, further supporting their sensory origin.

In summary, the combination of anesthetized and behavioral recordings, the temporal profile of responses, and their discriminative nature strongly support the sensory (visual) origin of the observed activity within the early response period. While the current study provides strong temporal and experimental evidence for the sensory origin of the visual responses, we agree that causal approaches, such as optogenetic silencing of visual input, could provide even stronger validation. Future work will explore these methods to further dissect the visual contributions to auditory cortical activity.

(2) The finding that AC neurons in trained mice preferentially respond - and enhance - auditory and visual responses pertaining to the contralateral choice is interesting, but the study does not show evidence for the functional relevance of this phenomenon. As has become more and more evident over the past few years (see e.g. the literature on mouse PPC), correlated neural activity is not an indication of functional role. Therefore, in the absence of causal evidence, the functional role of the reported AC correlates should not be overstated by the authors. My opinion is that, starting from the title, the authors need to much more carefully discuss the implications of their findings.

We fully agree that correlational data alone cannot establish causality. In light of your suggestion, we will revise the manuscript to more carefully discuss the implications of our findings, acknowledging that the preferred responses observed in AC neurons, particularly in relation to the contralateral choice, are correlational. We have updated several sentences in the manuscript to avoid overstating the functional relevance of these observations. Below are the revisions we have made:

Abstract section

"Importantly, many audiovisual neurons in the AC exhibited experience-dependent associations between their visual and auditory preferences, displaying a unique integration model. This model employed selective multisensory enhancement for the auditory-visual pairing guiding the contralateral choice, which correlated with improved multisensory discrimination."

(Page 8, fourth paragraph in Results Section)

"This aligns with findings that neurons in the AC and medial prefrontal cortex selectively preferred the tone associated with the behavioral choice contralateral to the recorded cortices during sound discrimination tasks, potentially reflecting the formation of sound-to-action associations. However, this preference represents a neural correlate, and further work is required to establish its causal link to behavioral choices."

(rewrite 3rd paragraph in Discussion Section)

"Consistent with prior research(10,31), most AC neurons exhibited a selective preference for cues associated with contralateral choices, regardless of the sensory modality. This suggests that AC neurons may contribute to linking sensory inputs with decision-making, although their causal role remains to be examined. "

"These results indicate that multisensory training could drive the formation of specialized neural circuits within the auditory cortex, facilitating integrated processing of related auditory and visual information. However, further causal studies are required to confirm this hypothesis and to determine whether the auditory cortex is the primary site of these circuit modifications."

MINOR:

(1) The manuscript is lacking what pertains to the revised interpretation of most studies about audiovisual interactions in primary sensory cortices following the recent studies revealing that most of what was considered to be crossmodal actually reflects motor aspects. In particular, recent evidence suggests that sensory-induced spontaneous motor responses may have a surprisingly fast latency (within 40 ms; Clayton et al. 2024). Such responses might also underlie the contralaterally-tuned responses observed by the authors if one assumes that mice learn a stereotypical response that is primed by the upcoming goal-directed, learned response. Given that a full exploration of this issue would require high-speed tracking of orofacial and body motions, the authors should at least revise the discussion and the possible interpretation of their results not just on the basis of the literature, but after carefully revising the literature in view of the most recent findings, that challenge earlier interpretations of experimental results.

Thank you for pointing out this important consideration. We have revised the discussion (paragraph 8-9) as follows:

“There is ongoing debate about whether cross-sensory responses in sensory cortices predominantly reflect sensory inputs or are influenced by behavioral factors, such as cue-induced body movements. A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(48). Several studies have demonstrated sensory neurons can encode signals associated with whisking(49), running(50), pupil dilation (510 and other movements(52). In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset. This early timing suggests that the observed responses likely reflect direct sensory inputs, rather than being modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(53).

A recent study by Clayton et al. (2024) demonstrated that sensory stimuli can evoke rapid motor responses, such as facial twitches, within 50 ms, mediated by subcortical pathways and modulated by descending corticofugal input(56). These motor responses provide a sensitive behavioral index of auditory processing. Although Clayton et al. did not observe visually evoked facial movements, it is plausible that visually driven motor activity occurs more frequently in freely moving animals compared to head-fixed conditions. In goal-directed tasks, such rapid motor responses might contribute to the contralaterally tuned responses observed in our study, potentially reflecting preparatory motor behaviors associated with learned responses. Consequently, some of the audiovisual integration observed in the auditory cortex may represent a combination of multisensory processing and preparatory motor activity. Comprehensive investigation of these motor influences would require high-speed tracking of orofacial and body movements. Therefore, our findings should be interpreted with this consideration in mind. Future studies should aim to systematically monitor and control eye, orofacial, and body movements to disentangle sensory-driven responses from motor-related contributions, enhancing our understanding of motor planning’s role in multisensory integration.”

(2) The methods section is a bit lacking in details. For instance, information about the temporal window of analysis for sensory-evoked responses is lacking. Another example: for the spike sorting procedure, limited details are given about inclusion/exclusion criteria. This makes it hard to navigate the manuscript and fully understand the experimental paradigm. I would recommend critically revising and expanding the methods section.

Thank you for raising this point. We clarified the temporal window by including additional details in the methods section, even though this information was already mentioned in the results section. Specifically, we now state:

(Neural recordings and Analysis in methods section)

“...These neural signals, along with trace signals representing the stimuli and session performance information, were transmitted to a PC for online observation and data storage. Neural responses were analyzed within a 0-150ms temporal window after cue onset, as this period was identified as containing the main cue-evoked responses for most neurons. This time window was selected based on the consistent and robust neural activity observed during this period.”

We appreciate your concern regarding spike sorting procedure. To address this, we have expanded the methods section to provide more detailed information about the quality of our single-unit recordings. we have added detailed information in the text, as shown below (Analysis of electrophysiological data in methods section):

“Initially, the recorded raw neural signals were band-pass filtered in the range of 300-6000 Hz to eliminate field potentials. A threshold criterion, set at no less than three times the standard deviation (SD) above the background noise, was applied to automatically identify spike peaks. The detected spike waveforms were then subjected to clustering using template-matching and built-in principal component analysis tool in a three-dimensional feature space. Manual curation was conducted to refine the sorting process. Each putative single unit was evaluated based on its waveform and firing patterns over time. Waveforms with inter-spike intervals of less than 2.0 ms were excluded from further analysis. Spike trains corresponding to an individual unit were aligned to the onset of the stimulus and grouped based on different cue and choice conditions. Units were included in further analysis only if their presence was stable throughout the session, and their mean firing rate exceeded 2 Hz. The reliability of auditory and visual responses for each unit was assessed, with well-isolated units typically showing the highest response reliability.”

Reviewer #1 (Recommendations for the authors):

(1) Some of the ordering of content in the introduction could be improved. E.g. line 49 reflects statements about the importance of sensory experience, which is the topic of the subsequent paragraph. In the discussion, line 436, there is a discussion of the same findings as line 442. These two paragraphs in general appear to discuss similar content. Similarly, the paragraph starting at line 424 and at line 451 both discuss the plasticity of multisensory responses through audiovisual experience, as well as the paragraph starting at line 475 (but now audiovisual pairing is dubbed semantic). In the discussion of how congruency/experience shapes multisensory interactions, the authors should relate their findings to those of Meijer et al. 2017 and Garner and Keller 2022 (visual cortex) about enhanced and suppressed responses and their potential role (as well as other literature such as Banks et al. 2011 in AC).

We thank the reviewer for their detailed observations and valuable recommendations to improve the manuscript's organization. Below, we address each point:

We deleted the sentence, "Sensory experience has been shown to shape cross-modal presentations in sensory cortices" (Line 49), as the subsequent paragraph discusses sensory experience in detail.

To avoid repetition, we removed the sentence, "This suggests that multisensory training enhances AC's ability to process visual information" (Lines 442–443).

Regarding the paragraph starting at Line 475, we believe its current form is appropriate, as it focuses on the influence of semantic congruence on multisensory integration, which differs from the topics discussed in the other paragraphs.

We have cited the three papers suggested by the reviewer in the appropriate sections of the manuscript.

(Paragraph 6 in discussion section)

“…A study conducted on the gustatory cortex of alert rats has shown that cross-modal associative learning was linked to a dramatic increase in the prevalence of neurons responding to nongustatory stimuli (24). Moreover, in the primary visual cortex, experience-dependent interactions can arise from learned sequential associations between auditory and visual stimuli, mediated by corticocortical connections rather than simultaneous audiovisual presentations (26).”

(Paragraph 2 in discussion section)

“...Meijer et al. reported that congruent audiovisual stimuli evoke balanced enhancement and suppression in V1, while incongruent stimuli predominantly lead to suppression(6), mirroring our findings in AC, where multisensory integration was dependent on stimulus feature…”

(Paragraph 2 in introduction section)

“...Anatomical investigations reveal reciprocal nerve projections between auditory and visual cortices(4,11-15), highlighting the interconnected nature of these sensory systems. Moreover, two-photon calcium imaging in awake mice has shown that audiovisual encoding in the primary visual cortex depends on the temporal congruency of stimuli, with temporally congruent audiovisual stimuli eliciting balanced enhancement and suppression, whereas incongruent stimuli predominantly result in suppression(6).”

(2) The finding of purely visually responsive neurons in the auditory cortex that moreover discriminate the stimuli is surprising given previous results (Iurilli et al. 2012, Morrill and Hasenstaub 2018 (only L6), Oude Lohuis et al. 2024, Atilgan et al. 2018, Chou et al. 2020). Reporting the latency of this response is interesting information about the potential pathways by which this information could reach the auditory system. Furthermore, spike isolation quality and histological verification are described in little detail. It is crucial for statements about the auditory, visual, or audiovisual response of individual neurons to substantiate the confidence level about the quality of single-unit recordings and where they were recorded. Do the authors have data to support that visual and audiovisual responses were not restricted to posteromedial tetrodes or clusters with poor quality? A discussion of finding V-responsive units in AC with respect to literature is warranted. Furthermore, the finding that also in visual trials behaviorally relevant information about the visual cue (with a bias for the contralateral choice cue) is sent to the AC is pivotal in the interpretation of the results, which as far as I note not really considered that much.

We appreciate the reviewer’s thoughtful comments and have addressed them as follows:

Discussion of finding choice-related V-responsive units in AC with respect to literature and potential pathways

3rd paragraph in the Discussion section

“Consistent with prior research(10,31), most AC neurons exhibited a selective preference for cues associated with contralateral choices, regardless of the sensory modality. This suggests that AC neurons may contribute to linking sensory inputs with decision-making, although their causal role remains to be examined. Associative learning may drive the formation of new connections between sensory and motor areas of the brain, such as cortico-cortical pathways(35). Notably, this cue-preference biasing was absent in the free-choice group. A similar bias was also reported in a previous study, where auditory discrimination learning selectively potentiated corticostriatal synapses from neurons representing either high or low frequencies associated with contralateral choices(32)…”

6th paragraph in the Discussion section

“Our results extend prior finding(4,47), showing that visual input not only reaches the AC but can also drive discriminative responses, particularly during task engagement. This task-specific plasticity enhances cross-modal integration, as demonstrated in other sensory systems. For example, calcium imaging studies in mice showed that a subset of multimodal neurons in visual cortex develops enhanced auditory responses to the paired auditory stimulus following coincident auditory–visual experience(25)…”

8th paragraph in the Discussion section

“…In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset, suggesting that visual information reaches the AC through rapid pathways. Potential candidates include direct or fast cross-modal inputs, such as pulvinar-mediated pathways(8) or corticocortical connections(5，54), rather than slower associative mechanisms. This early timing indicates that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55).”

Response Latency

Regarding the latency of visually driven responses, we have included this information in our response to the second reviewer’s first weakness (please see the above). Briefly, we analyzed neural responses within a 0-150ms temporal window after cue onset, as this period captures the most consistent and robust cue-evoked responses across neurons.

Purely Visually Responsive Neurons in A1

We agree that the finding of visually responsive neurons in the auditory cortex may initially seem surprising. However, these neurons might not have been sensitive to target auditory cues in our task but could still respond to other sound types. Cortical neurons are known to exhibit significant plasticity during the cue discrimination tasks, as well as during passive sensory exposure. Thus, the presence of visually responsive neurons is not inconsistent with prior findings but highlights task-specific sensory tuning. We confirm that responses were not restricted to posteromedial tetrodes or low-quality clusters (see an example of a robust visually responsive neuron in supplementary Fig. 4). Histological analysis verified electrode placements across the auditory cortex.

For spike sorting, we have added detailed information in the text, as shown below:

“Initially, the recorded raw neural signals were band-pass filtered in the range of 300-6000 Hz to eliminate field potentials. A threshold criterion, set at no less than three times the standard deviation (SD) above the background noise, was applied to automatically identify spike peaks. The detected spike waveforms were then subjected to clustering using template-matching and built-in principal component analysis tool in a three-dimensional feature space. Manual curation was conducted to refine the sorting process. Each putative single unit was evaluated based on its waveform and firing patterns over time. Waveforms with inter-spike intervals of less than 2.0 ms were excluded from further analysis. Spike trains corresponding to an individual unit were aligned to the onset of the stimulus and grouped based on different cue and choice conditions. Units were included in further analysis only if their presence was stable throughout the session, and their mean firing rate exceeded 2 Hz. The reliability of auditory and visual responses for each unit was assessed, with well-isolated units typically showing the highest response reliability.”

(3) In the abstract it seems that in "Additionally, AC neurons..." the connective word 'additionally' is misleading as it is mainly a rephrasing of the previous statement.

Replaced "Additionally" with "Furthermore" to better signal elaboration and continuity.

(4) The experiments included multisensory conflict trials - incongruent audiovisual stimuli. What was the behavior for these trials given multiple interesting studies on the neural correlates of sensory dominance (Song et al. 2017, Coen et al. 2023, Oude Lohuis et al. 2024).

We appreciate your feedback and have addressed it by including a new figure (supplemental Fig. 8) that illustrates choice selection during incongruent audiovisual stimuli. Panel (a) shows that rats displayed confusion when exposed to mismatched stimuli, resulting in choice patterns that differed from those observed in panel (b), where consistent audiovisual stimuli were presented. To provide clarity and integrate this new figure effectively into the manuscript, we updated the results section as follows:

“...Rats received water rewards with a 50% chance in either port when an unmatched multisensory cue was triggered. Behavioral analysis revealed that Rats displayed notable confusion in response to unmatched multisensory cues, as evidenced by their inconsistent choice patterns (supplementary Fig. 8).”

(5) Line 47: The AC does not 'perceive' sound frequency, individual brain regions are not thought to perceive.

e appreciate the reviewer’s observation and have revised the sentence to ensure scientific accuracy. The updated sentence in the second paragraph of the Introduction now reads:

“Even irrelevant visual cues can affect sound discrimination in AC¹⁰.”

(6) Line 59-63: The three questions are not completely clear to me. Both what they mean exactly and how they are different. E.g. Line 60: without specification, it is hard to understand which 'strategies' are meant by the "same or different strategies"? And Line 61: What is meant by the quotation marks for match and mismatch? I assume this is referring to learned congruency and incongruency, which appears almost the same question as number 3 (how learning affects the cortical representation).

We have revised the three questions for improved clarity and distinction as follows:<br /> “This limits our understanding of multisensory integration in sensory cortices, particularly regarding: (1) Do neurons in sensory cortices adopt consistent integration strategies across different audiovisual pairings, or do these strategies vary depending on the pairing? (2) How does multisensory perceptual learning reshape cortical representations of audiovisual objects? (3) How does the congruence between auditory and visual features—whether they "match" or "mismatch" based on learned associations—impact neural integration?”

(7) Is the data in Figures 1c and d only hits?

Only correct trials are included. We add this information in the figure legend. Please see Fig. 1 legend. Also, please see below

“c Cumulative frequency distribution of reaction time (time from cue onset to leaving the central port) for one representative rat in auditory, visual and multisensory trials (correct only). d Comparison of average reaction times across rats in auditory, visual, and multisensory trials (correct only).”

(8) Figure S1b: Preferred frequency is binned in non-equidistant bins, neither linear nor logarithmic. It is unclear what the reason is.

The edges of the bins for the preferred frequency were determined based on a 0.5-octave increment, starting from the smallest boundary of 8 kHz. Specifically, the bin edges were calculated as follows:

8×2^0.5=11.3 kHz;

8×2¹=16 kHz;

8×2^1.5=22.6 kHz;

8×2²=32 kHz;

This approach reflects the common practice of using changes in octaves to define differences between pure tone frequencies, as it aligns with the logarithmic perception of sound frequency in auditory neuroscience.

(9) Figure S1d: why are the responses all most neurons very strongly correlated given the frequency tuning of A1 neurons? Further, the mean normalized response presented in Figure S2e does seem to indicate a stronger response for 10kHz tones than 3kHz, in conflict with the data from anesthetized rats presented in Figure S2e.

There is no discrepancy in the data. In Figure S1d, we compared neuronal responses to 10 kHz and 3 kHz tones, demonstrating that most neurons responded well to both frequencies. This panel does not aim to illustrate frequency selectivity but rather the overall responsiveness of neurons to these tones. For detailed information on sound selectivity, readers can refer to Figures S3a-b, which show that while more neurons preferred 10 kHz tones, the proportion is lower than in neurons recorded during the multisensory discrimination task. This distinction explains the observed differences and aligns with the results presented.

(10) Line 79: For clarity, it can be added that the multisensory trials presented are congruent trials (jointly indicated rewarded port), and perhaps that incongruent trials are discussed later in the paper.

We believe additional clarification is unnecessary, as the designations "A^3kV^hz" and "A^10kV^vt" clearly indicate the specific combinations of auditory and visual cues presented during congruent trials. Additionally, the discussion of incongruent trials is provided later in the manuscript, as noted by the reviewer.

(11) Line 111: the description leaves unclear that the 35% reflects the combination of units responsive to visual only and responsive to auditory or visual.

The information is clearly presented in Figure 2b, which shows the proportions of neurons responding to auditory-only (A), visual-only (V), both auditory and visual (A, V), and audiovisual-only (VA) stimuli in a pie chart. Readers can refer to this figure for a detailed breakdown of the neuronal response categories.

(12) Figure 2h: consider a colormap with diverging palette and equal positive and negative maximum (e.g. -0.6 to 0.6) and perhaps reiterate in the color bar legend which stimulus is preferred for which selectivity index.

We appreciate the suggestion; however, we believe that the current colormap effectively conveys the data and the intended interpretation. The existing color bar legend already provides clear information about the selectivity index, and the stimulus preference is adequately explained in the figure caption. As such, further adjustments are not necessary.

(13) Line 160: "a ratio of 60:20 for V^vt 160 preferred vs. V^hz preferred neurons." Is this supposed to add up to 100, or is this a ratio of 3:1?

We rewrite the sentence. Please see below:

“Similar to the auditory selectivity observed, a greater proportion of neurons favored the visual stimulus (V^vt) associated with the contralateral choice, with a 3:1 ratio of V^vt-preferred to V^hz-preferred neurons.”

(14) The statement in Figure 2g and line 166/167 could be supported by a statistical test (chi-square?).

Thank you for the suggestion. However, we believe that a statistical test is not required in this case, as the patterns observed are clearly represented in Figure 2g. The qualitative differences between the groups are evident and sufficiently supported by the data.

(15) Line 168, it is unclear in what sense 'dominant' is meant. Is audition perceived as a dominant sensory modality in a behavioral sense (e.g. Song et al. 2017), or are auditory signals the dominant sensory signal locally in the auditory cortex?

Thank you for the clarification. To address your question, by "dominant," we are referring to the fact that auditory inputs are the most prominent and influential among the sensory signals feeding into the auditory cortex. This reflects the local dominance of auditory signals within the auditory cortex, rather than a behavioral dominance of auditory perception. We have revised the sentence as follows:

“We propose that the auditory input, which dominates within the auditory cortex, acts as a 'teaching signal' that shapes visual processing through the selective reinforcement of specific visual pathways during associative learning.”

(16) Line 180: "we discriminated between auditory, visual, and multisensory cues." This phrasing indicated that the SVMs were trained to discriminate sensory modalities (as is done later in the manuscript), rather than what was done: discriminate stimuli within different categories of trials.

Thank you for your comment. We have revised the sentence for clarity. Please see the updated version below:

“Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates stimulus identity within the same modality (e.g., A^3k vs. A^10k for auditory stimuli, V^hz vs. V^vt for visual stimuli, A^3kV^hz vs. A^10kV^vt for multisensory stimuli).”

(17) Line 185: "a deeply accurate incorporation of visual processing in the auditory cortex." the phrasing is a bit excessive for a binary classification performance.

Thank you for pointing this out. We have revised the sentence to better reflect the findings without overstating them:

“Interestingly, AC neurons could discriminate between two visual targets with around 80% accuracy (Fig. 2j), demonstrating a meaningful incorporation of visual information into auditory cortical processing.”

(18) Figure 3, title. An article is missing (a,an/the).

Done. Please see below:

“Fig. 3 Auditory and visual integration in the multisensory discrimination task”

(19) Line 209, typo pvalue: p<-0.00001.

Done (p<0.00001).

(20) Line 209, the pattern is not weaker. The pattern is the same, but more weakly expressed.

Thank you for your valuable feedback. We appreciate your clarification and agree that our phrasing could be improved for accuracy. The observed pattern under anesthesia is indeed the same but less strongly expressed compared to the task engagement. We have revised the sentence to better reflect this distinction:

“A similar pattern, albeit less strongly expressed, was observed under anesthesia (Supplementary Fig. 3c-3f), suggesting that multisensory perceptual learning may induce plastic changes in AC.”

(21) Line 211: choice-free group → free-choice group.

Done.

(22) Line 261: wrong → incorrect (to maintain consistent terminology).

Done.

(23) Line 265: why 'likely'? Are incorrect choices on the A^3k-V^hz trials not by definition contralateral and vice versa? Or are there other ways to have incorrect trials?

We deleted the word of ‘likely’. Please see below:

“…, correct choices here correspond to ipsilateral behavioral selection, while incorrect choices correspond to contralateral behavioral selection.”

(24) Typo legend Fig 3a-c (tasks → task). (only one task performed).

Done.

(25) Line 400: typo: Like → like.

Done.

(26) Line 405: What is meant by a cohesive visual stimulus? Congruent? Rephrase.

Done. Please see the below:

“…layer 2/3 neurons of the primary visual cortex(7), and a congruent visual stimulus can enhance sound representation…”

(27) Line 412: Very general statement and obviously true: depending on the task, different sensory elements need to be combined to guide adaptive behavior.

We really appreciate the reviewer and used this sentence (see second paragraph in discussion section).

(28) Line 428: within → between (?).

Done.

(29) Figure 3L is not referenced in the main text. By going through the figures and legends my understanding is that this shows that most neurons have a multisensory response that lies between 2 z-scores of the predicted response in the case of 83% of the sum of the auditory and the visual response. However, how was the 0.83 found? Empirically? Figure S3 shows a neuron that does follow a 100% summation. Perhaps the authors could quantitatively support their estimate of 83% of the A + V sum, by varying the fraction of the sum (80%, 90%, 100% etc.) and showing the distribution of the preferred fraction of the sum across neurons, or by showing the percentage of neurons that fall within 2 z-scores for each of the fractions of the sum.

Thank you for your detailed feedback and suggestions regarding Figure 3L and the 83% multiplier.

(1) Referencing Figure 3L:

Figure 3L is referenced in the text. To enhance clarity, we have revised the text to explicitly highlight its relevance:

“Specifically, as illustrated in Fig. 3k, the observed multisensory response approximated 83% of the sum of the auditory and visual responses in most cases, as quantified in Fig. 3L.”

(2) Determination of the 0.83 Multiplier:

The 0.83 multiplier was determined empirically by comparing observed audiovisual responses with the predicted additive responses (i.e., the sum of auditory and visual responses). For each neuron, we calculated the auditory, visual, and audiovisual responses. We then compared the observed audiovisual response with scaled sums of auditory and visual responses (Fig. 3k), expressed as fractions of the additive prediction (e.g., 0.8, 0.83, 0.9, etc.). We found that when the scaling factor was 0.83, the population-wide difference between predicted and observed multisensory responses, expressed as z-scores, was minimized. Specifically, at this value, the mean z-score across the population was approximately zero (-0.0001±1.617), indicating the smallest deviation between predicted and observed responses.

(30) Figure 5e: how come the diagonal has 0.5 decoding accuracy within a category? Shouldn't this be high within-category accuracy? If these conditions were untested and it is an issue of the image display it would be informative to test the cross-validated performance within the category as well as a benchmark to compare the across-category performance to. Aside, it is unclear which conventions from Figure 2 are meant by the statement that conventions were the same.

The diagonal values (~0.5 decoding accuracy) within each category reflect chance-level performance. This occurs because the decoder was trained and tested on the same category conditions in a cross-validated manner, and within-category stimulus discrimination was not the primary focus of our analysis. Specifically, the stimuli within a category shared overlapping features, leading to reduced discriminability for the decoder when distinguishing between them. Our primary objective was to assess cross-category performance rather than within-category accuracy, which may explain the observed pattern in the diagonal values.

Regarding the reference to Figure 2, we appreciate the reviewer pointing out the ambiguity. To avoid any confusion, we have removed the sentence referencing "conventions from Figure 2" in the legend for Figure 5e, as it does not contribute meaningfully to the understanding of the results.

(31) Line 473: "movement evoked response", what is meant by this?

Thank the reviewer for highlighting this point. To clarify, by "movement-evoked response," we are referring to neural activity that is driven by the animal's movements, rather than by sensory inputs. This type of response is typically stereotyped, meaning that it has a consistent, repetitive pattern associated with specific movements, such as whisking, running, or other body or facial movements.

In our study, we propose that the visually-evoked responses observed within the 150 ms time window after cue onset primarily reflect sensory inputs from the visual stimulus rather than movement-related activity. This interpretation is supported by the response timing: visual-evoked activity occurs within 100 ms of the light flash onset, a timeframe too rapid to be attributed to body or orofacial movements. Additionally, unlike stereotyped movement-evoked responses, the visual responses we observed are discriminative, varying based on specific visual features—a hallmark of sensory processing rather than motor-driven activity.

We have revised the manuscript as follows (eighth paragraph in discussion section):

“There is ongoing debate about whether cross-sensory responses in sensory cortices predominantly reflect sensory inputs or are influenced by behavioral factors, such as cue-induced body movements. A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(49). Several studies have demonstrated sensory neurons can encode signals associated with whisking(50), running(51), pupil dilation(52) and other movements(53). In our study, the responses to visual stimuli in the auditory cortex occurred primarily within a 100 ms window following cue onset. suggests that visual information reaches the AC through rapid pathways. Potential candidates include direct or fast cross-modal inputs, such as pulvinar-mediated pathways(8) or corticocortical connections(5，54), rather than slower associative mechanisms. This early timing suggests that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55). ”

(32) Line 638-642: It is stated that a two-tailed permutation test is done. The cue selectivity can be significantly positive and negative, relative to a shuffle distribution. This is excellent. But then it is stated that if the observed ROC value exceeds the top 5% of the distribution it is deemed significant, which corresponds to a one-tailed test. How were significantly negative ROC values detected with p<0.05?

Thank you for pointing this out. We confirm that a two-tailed permutation test was indeed used to evaluate cue selectivity. In this approach, significance is determined by comparing the observed ROC value to both tails of the shuffle distribution. Specifically, if the observed ROC value exceeds the top 2.5% or falls below the bottom 2.5% of the distribution, it is considered significant at p< 0.05. This two-tailed test ensures that both significantly positive and significantly negative cue selectivity values are identified.

To clarify this in the manuscript, we have revised the text as follows:

“This generated a distribution of values from which we calculated the probability of our observed result. If the observed ROC value exceeds the top 2.5% of the distribution or falls below the bottom 2.5%, it was deemed significant (i.e., p < 0.05).”

(33) Line 472: the cited paper (reference 52) actually claims that motor-related activity in the visual cortex has an onset before 100ms and thus does not support your claim that the time window precludes any confound of behaviorally mediated activity. Furthermore, that study and reference 47 show that sensory stimuli could be discriminated based on the cue-evoked body movements and are discriminative. A stronger counterargument would be that both studies show very fast auditory-evoked body movements, but only later visually-evoked body movements.

We appreciate the reviewer’s comments. As Lohuis et al. (reference 55) demonstrated, activity in the visual cortex (V1) can reflect distinct visual, auditory, and motor-related responses, with the latter often dissociable in timing. In their findings, visually-evoked movement-related activity arises substantially later than the sensory visual response, generally beginning around 200 ms post-stimulus onset. In contrast, auditory-evoked activity in A1 occurs relatively early.

We have revised the manuscript as follows (eighth paragraph in discussion section):

“A recent study shows that sound-clip evoked activity in visual cortex have a behavioral rather than sensory origin and is related to stereotyped movements(49). ...This early timing suggests that the observed responses were less likely modulated by visually-evoked body or orofacial movements, which typically occur with a delay relative to sensory cue onset(55). ”

(34) The training order (multisensory cue first) is important to briefly mention in the main text.

We appreciate the reviewer’s suggestion and have added this information to the main text. The revised text now reads:

“The training proceeded in two stages. In the first stage, which typically lasted 3-5 weeks, rats were trained to discriminate between two audiovisual cues. In the second stage, an additional four unisensory cues were introduced, training the rats to discriminate a total of six cues.”

(35) Line 542: As I understand the multisensory rats were trained using the multisensory cue first, so different from the training procedure in the unisensory task rats where auditory trials were learned first.

Thank you for pointing this out. You are correct that, in the unisensory task, rats were first trained to discriminate auditory cues, followed by visual cues. To improve clarity and avoid any confusion, we have removed the sentence "Similar to the multisensory discrimination task" from the revised text.

(36) Line 546: Can you note on how the rats were motivated to choose both ports, or whether they did so spontaneously?

Thank you for your insightful comment. The rats' port choice was spontaneous in this task, as there was no explicit motivation required for choosing between the ports. We have clarified this point in the text to address your concern. The revised sentence now reads:

“They received a water reward at either port following the onset of the cue, and their port choice was spontaneous.”

(37) It is important to mention in the main text that the population decoding is actually pseudopopulation decoding. The interpretation is sufficiently important for interpreting the results.

Thank you for this valuable suggestion. We have revised the text to specify "pseudo-population" instead of "population" to clarify the nature of our decoding analysis. The revised text now reads:

“Our multichannel recordings enabled us to decode sensory information from a pseudo-population of AC neurons on a single-trial basis. Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates between stimuli.”

(38) The term modality selectivity for the description of the multisensory interaction is somewhat confusing. Modality selectivity suggests different responses to the visual or auditory trials. The authors could consider a different terminology emphasizing the multisensory interaction effect.

Thank you for your insightful comment. We have replaced " modality selectivity " with " multisensory interactive index " (MSI). This term more accurately conveys a tendency for neurons to favor multisensory stimuli over individual sensory modalities (visual or auditory alone).

(39) In Figures 3 e and g the color code is different from adjacent panels b and c and is to be deciphered from the legend. Consider changing the color coding, or highlight to the reader that the coloring in Figures 3b and c is different from the color code in panels 3 e and g.

We appreciate the reviewer’s observation. However, we believe that a change in the color coding is not necessary. Figures 3e and 3g differentiate symbols by both shape and color, ensuring accessibility and clarity. This is clearly explained in the figure legend to guide readers effectively.

(40) Figure S2b: was significance tested here?

Yes, we did it.

(41) Figure S2d: test used?

Yes, test used.

(42) Line 676: "as appropriate", was a normality test performed prior to statistical test selection?

In our analysis, we assessed normality before choosing between parametric (paired t-test) and non-parametric (Wilcoxon signed-rank test) methods. We used the Shapiro-Wilk test to evaluate the normality of the data distributions. When data met the assumption of normality, we applied the paired t-test; otherwise, we used the Wilcoxon signed-rank test.

Thank you for pointing this out. We confirm that a normality test was performed prior to the selection of the statistical test. Specifically, we used the Shapiro-Wilk test to assess whether the data distributions met the assumption of normality. Based on this assessment, we applied the paired t-test for normally distributed data and the Wilcoxon signed-rank test for non-normal data.

To ensure clarity, we update the "Statistical Analysis" section of the manuscript with the following revised text:

“For behavioral data, such as mean reaction time differences between unisensory and multisensory trials, cue selectivity and mean modality selectivity across different auditory-visual conditions, comparisons were performed using either the paired t-test or the Wilcoxon signed-rank test. The Shapiro-Wilk test was conducted to assess normality, with the paired t-test used for normally distributed data and the Wilcoxon signed-rank test for non-normal data.”

(43) Line 679: incorrect, most data is actually represented as mean +- SEM.

Thank you for pointing this out. In the Results section, we report data as mean ± SD for descriptive statistics, while in the figures, the error bars typically represent the standard error of the mean (SEM) to visually indicate variability around the mean. We have specified in each figure legend whether the error bars represent SD or SEM.

Reviewer #2 (Recommendations for the authors):

(1) Line 182 - here it sounds like you mean your classifier was trained to decode the modality of the stimulus, when I think what you mean is that you decoded the stimulus contingencies using A/V/AV cues?

Thank you for pointing out this potential misunderstanding. We would like to clarify that the classifier was trained to decode the stimulus identity (e.g., A^3k vs. A^10k for auditory stimuli, V^hz vs. V^vt for visual stimuli, and A^3kV^hz vs. A^10kV^vt for multisensory stimuli) rather than the modality of the stimulus. The goal of the analysis was to determine how well the pseudo-population of AC neurons could distinguish between individual stimuli within the same modality. We have revised the relevant text in the revised manuscript to ensure this distinction is clear. Please see the following:

“Our multichannel recordings enabled us to decode sensory information from a pseudo-population of AC neurons on a single-trial basis. Using cross-validated support vector machine (SVM) classifiers, we examined how this pseudo-population discriminates stimulus identity (e.g.,  A^3k vs. A^10k for auditory stimuli, V^hz vs. V^vt for visual stimuli,  A^3kV^hz vs. A^10kV^vt for multisensory stimuli).”

(2) Lines 256 - here the authors look to see whether incorrect trials diminish audiovisual integration. I would probably seek to turn the causal direction around and ask are AV neurons critical for behaviour - nevertheless, since this is only correlational the causal direction cannot be unpicked. However, the finding that contralateral responses per se do not result in enhancement is a key control. Showing that multisensory enhancement is less on error trials is a good first step to linking neural activity and perception, but I wonder if the authors could take this further however by seeking to decode choice probabilities as well as stimulus features in an attempt to get a little closer to addressing the question of whether the animals are using these responses for behaviour.

Thank you for your comment and for highlighting the importance of understanding whether audiovisual (AV) neurons are critical for behavior. As you noted, the causal relationship between AV neural activity and behavioral outcomes cannot be directly determined in our current study due to its correlational nature. We agree that this is an important topic for future exploration. In our study, we examined how incorrect trials influence multisensory enhancement. Our findings show that multisensory enhancement is less pronounced during error trials, providing an initial link between neural activity and behavioral performance. To address your suggestion, we conducted an additional analysis comparing auditory and multisensory selectivity between correct and incorrect choice trials. As shown in Supplementary Fig. 7, both auditory and multisensory selectivity were significantly lower during incorrect trials. This result highlights the potential role of these neural responses in decision-making, suggesting they may extend beyond sensory processing to influence choice selection. We have cited this figure in the Results section as follows: ( the paragraph regarding Impact of incorrect choices on audiovisual integration):

“Overall, these findings suggest that the multisensory perception reflected by behavioral choices (correct vs. incorrect) might be shaped by the underlying integration strength. Furthermore, our analysis revealed that incorrect choices were associated with a decline in cue selectivity, as shown in Supplementary Fig. 7.”

We acknowledge your suggestion to decode choice probabilities alongside stimulus features as a more direct approach to exploring whether animals actively use these neural responses for behavior. Unfortunately, in the current study, the low number of incorrect trials limited our ability to perform such analyses reliably. Nonetheless, we are committed to pursuing this direction in subsequent work. We plan to use techniques such as optogenetics in future studies to causally test the role of AV neurons in driving behavior.

(3) Figure 5E - the purple and red are indistinguishable - could you make one a solid line and keep one dashed?

We thank the reviewer for pointing out that the purple and red lines in Figure 5E were difficult to distinguish. To address this concern, we modified the figure by making two lines solid and changing the color of one square, as suggested. These adjustments enhance visual clarity and improve the distinction between them.

(4) The unisensory control training is a really nice addition. I'm interested to know whether behaviourally these animals experienced an advantage for audiovisual stimuli in the testing phase? This is important information to include as if they don't it is one step closer to linking audiovisual responses in AC to improved behavioural performance (and if they do, we must be suitably cautious in interpretation!).

Thank you for raising this important point. To address this, we have plotted the behavioral results for each animal (see Author response image 2). The data indicate that performance with multisensory cues is slightly better than with the corresponding unisensory cues. However, given the small sample size (n=3) and the considerable variation in behavioral performance across individuals, we remain cautious about drawing definitive conclusions on this matter. We recognize the need for further investigation to establish a robust link between audiovisual responses in the auditory cortex and improved behavioral performance. In future studies, we plan to include a larger number of animals and more thoroughly explore this relationship to provide a comprehensive understanding.

Author response image 2.

(5) Line 339 - I don't think you can say this leads to binding with your current behaviour or neural responses. I would agree there is a memory trace established and a preferential linking in AC neurons.

We thank the reviewer for raising this important point. In the revised manuscript, we have clarified that our data suggest the formation of a memory trace and preferential linking in AC neurons. The text has been updated to emphasize this distinction. Please see the revised section below (first paragraph in Discussion section).

“Interestingly, a subset of auditory neurons not only developed visual responses but also exhibited congruence between auditory and visual selectivity. These findings suggest that multisensory perceptual training establishes a memory trace of the trained audiovisual experiences within the AC and enhances the preferential linking of auditory and visual inputs. Sensory cortices, like AC, may act as a vital bridge for communicating sensory information across different modalities.”

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In their manuscript, Kong Fang et al describe a robust pipeline for the isolation of small extracellular vesicles through a combination of size exclusion chromatography and miniaturized density gradient separation. Subsequently, they prove that the method is reproducible and suitable for small-volume operations while at the same time not compromising the quality of vesicles.

Strengths:

The paper narrates a robust method for purifying high-quality sEVs from small amounts of blood plasma. They also demonstrate that through this approach, they can derive sEVs without compromising the protein composition, integrity of the vesicles, or contamination with other proteins or lipids.

Weaknesses:

The paper is a nice summary of how to enrich sEVs from blood samples. Although well performed and substantiated with data, the paper primarily deals with method development and optimisation.

We agree with the reviewer's assessment that this paper primarily focuses on the development and optimization of a method. Using this robust technique for isolating small extracellular vesicles (sEVs) from small blood volumes, our future research will investigate sEVs isolated from clinical samples, with a particular focus on their role in various diseases.

Reviewer #2 (Public Review):

Summary:

In this work, the authors manage to optimize a simple and rapid protocol using SEC followed by DGCU to isolate sEVs with adequate purity and yield from small volumes of plasma. Isolated fractions containing sEVs using SEC, DGCU, SEC-DGCU, and DGCU-SEC are compared in terms of their yield, purity surface protein profile, and RNA content. Although the combined use of these methodologies has already been evaluated in previous works, the authors manage to adapt them for the use of small volumes of plasma, which allows working in 1.5 mL tubes and reducing the centrifugation time to 2 hours.

The authors finally find that although both the SEC-DGCU and DGCU-SEC combinations achieve isolates with high purity, the SEC-DGCU combination results in higher yields.

This work provides an interesting tool for the rapid obtention of sEVs with sufficient yield and purity for detailed characterization which could be very useful in research and clinical therapy.

Strengths:

- The work is well-written and organized.

- The authors clearly state the problem they want to address, that is, optimizing a method that allows sEV to be isolated from small volumes of plasma.

- Although these methodologies have been tested in previous works, the authors manage to isolate sEVs of high purity and good performance through a simple and fast methodology.

- The characteristics of all isolated fractions are exhaustively analyzed through various state-of-the-art methodologies.

- They present a good interpretation of the results obtained through the methodologies used.

Weaknesses:

- Lack of references that support some of the results obtained.

- Although this work focuses on comparing different techniques and their combinations to find an optimal option, the authors do not use any statistical method that reliably shows the differences between these techniques, except when repeatability is measured.

We appreciate the reviewer's insightful comments and will incorporate the suggested missing references. We acknowledge that we did not perform statistical analyses when comparing the differences among the three methods. Nevertheless, the superiority of the SEC-DGUC method is evident from observations based on several independent characterization methods, including Cryo-EM, TEM, western blot, and total RNA quantification.

Firstly, repeated Cryo-EM observations consistently confirm that the SEC-alone method shows severe lipoprotein contamination while the SEC-DGUC method drastically reduces such lipoprotein contamination. In comparing the SEC-DGUC and DGUC-SEC methods, multiple independent characterization methods showed that the SEC-DGUC method yields significantly greater quantity of sEVs: 1) The western blot experiment showed much higher signal intensity for all four tested sEV markers (CD9, CD63, CD81, and TSG101), with estimated concentrations approximately 2.1, 2.1, 4.7, and 4.2 times higher than the DGUC-SEC method. 2) The total RNA analysis showed that SEC-DGUC-1 contained more than 4 times the total amount of RNA compared to DGUC-SEC-PF. 3) Establishing the normalization baseline, particle size distributions in SEC-DGUC-1 and DGUC-SEC-PF measured by TEM were found to be similar, suggesting comparable purity and distribution of the captured sEVs. For comparison purposes, within each independent characterization method, the same plasma source and total plasma volume were used, while across different methods, different plasma sources were used. These independent characterization methods have consistently demonstrated the superiority of the SEC-DGUC method over the DGUC-SEC or SEC-alone methods.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

In my opinion, this work is elegantly designed and supported by data, which would motivate more studies related to blood-derived microvesicles in the context of infectious and systemic diseases. Overall, the manuscript is well-written and explained in sufficient detail. I only have minor comments.

(1) Recruitment of volunteers for blood/plasma collection: there is a need for a statement that this was in accordance with ethical and biosafety regulations of the Institute/Clinic.

We added two sentences at the beginning of the Blood Collection section (under Materials and methods): “All procedures involving peripheral blood specimens were approved by the Singapore National Health Group Domain Specific Review Board (the central ethics committee) and were mutually recognized by the Nanyang Technological University Institutional Review Board (IRB#2018/00671). All blood specimens were de-identified prior to their use in the experiments.”

(2) Since this is a method development and validation article, it would be good to include an image of the iodixanol gradient with the high-density sEV zone, after centrifugation.

We have incorporated an image after centrifugation in Supplementary Figure 3.

(3) Although several sEV markers are shown in Figure 7A, flotillin is missing in this figure which was part of Figure 6B. Does flotillin show a different pattern? Flotillin is a DRM-associated marker, and hence may behave differently, would be interesting to add any insights.

We appreciate the reviewer’s careful observation. In Figure 6B, Flotillin was used to confirm the presence of sEVs in different density zones. However, for the purpose of comparing the yield between the SEC-DGUC and DGUC-SEC methods, as shown in Figure 7A, Flotillin was not included in the western blot analysis. No obvious pattern changes were observed in other sEV markers tested in both Figures 6B and 7A.   

(4) Methods section of LC/MS analysis- which protein database was used for protein identification?

We added the following sentence at the end of the LC/MS analysis section: “The protein database used for protein identification was Uniprot Human.”

Reviewer #2 (Recommendations For The Authors):

In line 43 some references are needed.

We added this reference: EL Andaloussi, S., Mäger, I., Breakefield, X. et al. Extracellular vesicles: biology and emerging therapeutic opportunities. Nat Rev Drug Discov 12, 347–357 (2013). https://doi.org/10.1038/nrd3978

In line 107, please avoid using short forms such as "it's".

We have revised that to “it is.”

In line 153: "...separates low-density particles from those of high density, but a considerable amount of..." the word "but" should not be in the sentence.

We have removed “but” in this sentence.

In line 181 the authors establish that "Notably, SEC-PF exhibited a high level of ApoB and low expression of sEV markers." Is there any explanation for this?

SEC-PF represents the eluate from the SEC step, collected before the DGUC step. This fraction contains a mixture of lipoproteins and sEVs. Due to the overwhelming abundance of lipoproteins compared to sEVs, the western blot predictably shows a high level of ApoB with minimal expression of sEV markers. This highlights that SEC alone effectively reduces plasma protein content but does not efficiently remove lipoproteins. Figure 6C further illustrates this point, as cryo-EM images of SEC-PF reveal the presence of sEVs, which are vastly outnumbered by lipoproteins.

In line 198, the sentence "Theoretically, the DGUC-SEC protocol should also effectively isolate sEVs from plasma" need to be supported by references.

See for instance:

- Holcar M, Ferdin J, Sitar S, Tušek-Žnidarič M, Dolžan V, Plemenitaš A, Žagar E, Lenassi M. 2020. Enrichment of plasma extracellular vesicles for reliable quantification of their size and concentration for biomarker discovery. Sci Rep 10:21346. doi:10.1038/s41598-020-78422-y.

- Jia Y, Yu L, Ma T, Xu W, Qian H, Sun Y, Shi H. 2022. Small extracellular vesicles isolation and separation: Current techniques, pending questions and clinical applications. Theranostics 12:6548-6575. doi:10.7150/thno.74305

- Vergauwen G, Dhondt B, Van Deun J, De Smedt E, Berx G, Timmerman E, Gevaert K, Miinalainen I, Cocquyt V, Braems G, Van den Broecke R, Denys H, De Wever O, Hendrix A. 2017. Confounding factors of ultrafiltration and protein analysis in extracellular vesicle research. Sci Rep 7:2704. doi:10.1038/s41598-017-02599-y

We have added this reference: Holcar M, Ferdin J, Sitar S, Tušek-Žnidarič M, Dolžan V, Plemenitaš A, Žagar E, Lenassi M. 2020. Enrichment of plasma extracellular vesicles for reliable quantification of their size and concentration for biomarker discovery. Sci Rep 10:21346. https://doi.org/10.1038/s41598-020-78422-y.  

In line 309 the authors establish that "NTA measured size distributions displayed well-overlapped histograms of particles". It is possible for the authors to analyze this overlapping using some statistical test as a chi-squared test?

We have conducted a statistical analysis of the histogram similarities using the Jensen-Shannon Divergence (JSD) method. This is reflected in the manuscript under the results section, “Repeatability and reliability of the SEC-DGUC protocol”, where we state: “We then compared size distributions for each plasma fraction using Jensen-Shannon Divergence (JSD). The JSD values, which are well below 0.1 (Figure 10B), indicate a consistent population of isolated particles, as further supported by Supplementary Figure 8.” Additionally, we included JSD values in the legend of Figure 10B: “JSD values for SEC-DGUC-1 to 4 are 0.015, 0.006, 0.001, and 0.002, indicating strong similarities among the histograms.” These additions demonstrate the robustness and repeatability of the SEC-DGUC protocol.

In line 360, "lasts ~ 16 hours or more." This statement needs a reference that supports this time.

We have added this reference: Vergauwen, G. et al. Robust sequential biophysical fractionation of blood plasma to study variations in the biomolecular landscape of systemically circulating extracellular vesicles across clinical conditions. J Extracell Vesicles 10, e12122 (2021).

In line 399, the reference format is different from the previously used format.

This is corrected. We thank the reviewer for this careful examination.

Line 466: This sentence is not quite clear. It can be understood that for every 0.5 mL of plasma, 2 mL of particle fraction are obtained and that for 6 mL of plasma, this method will give a total volume of 24 mL. However, it is not clear what is meant by the fact that it has been concentrated to 6 mL. While one can assume that those final 6 mL concentrates come from the initial 24 mL, perhaps the way this sentence was worded was not appropriate. I would recommend rewriting it for a simpler interpretation of how this method was performed.

We have changed the sentence to: “For the DGUC experiment using the 12 ml tube, 24 ml of PFs were obtained from 6 ml of plasma and subsequently concentrated to 6 ml. The 6 ml of concentrated PFs were then transferred to a Beckman Coulter ultra-clear centrifuge tube (344059, Beckman Coulter, USA) for further processing.”

Line 519: The authors established a second dilution to avoid absorbance values above 1.2. Is there any justification for this value, taking into account that the Lambert-Beer law presents more precision in the absorbance range of 0.2 to 0.8?

We have added this reference: https://diagnostic.serumwerk.com/wp-content/uploads/2021/05/V05-Serumwerk.pdf

Line 519-520: "Also included were water and 0.25 M sucrose as blanks". Perhaps authors could consider rephrasing this sentence.

We have changed the sentence to: “The absorbance measurements were made against water and 0.25 M sucrose blanks.”

In line 520, the sentence must say "each sample was made by triplicate".

We have changed the sentence to: “Each sample was prepared by triplicate to reduce error.” We thank the reviewer for this suggestion.

Line 673: The phrase "0.1% formic acid in 100% ACN" would be better, in my opinion, if it said "0.1% formic acid in ACN".

Yes, these two expressions have the same meaning. However, to ensure clarity, we have updated the description to “0.1% formic acid in ACN.”. We thank reviewer for this suggestion.

Supplementary Figure 1: in the Figure caption there is an error in the numbering: at the end, where it is written (E), it should be (F). Please, correct this.

We have made the necessary correction and sincerely appreciate the reviewer’s attentiveness.

Supplementary Figure 5: Some sEVs are hard to visualize due to poor image resolution. Is there any possibility for the authors to enhance these images?

We thank the reviewer for this valuable comment. To improve the visual clarity of the images, we have opted to display four sub-figures instead of nine.

Author response:

We appreciate the effort the reviewers have put into evaluating our work, and will take the opportunity to revise and improve our submission. In response to the reviewer's comments, we will carefully revisit our manuscript to address the concerns they have raised. Specifically, we will ensure that our revised version is coherent with our annotations and public databases, clarify any discrepancy between the investigated proteins and gene models, and re-examine our discussion of the evolutionary implications in light of their suggestions. We are confident that these revisions will strengthen our work and provide a clearer understanding of our research findings.

Author response:

We sincerely thank all three reviewers for their time, comments, and valuable suggestions, which will help improve our manuscript. Below, we provide preliminary remarks addressing some of the key issues that have been raised.

Reviewer 1:

We agree with the reviewer on the challenge of accurately mapping reads to multigene families. We carefully considered this issue and addressed it by evaluating the performance of multiple aligners using simulated RNA-seq reads. Our results indicate that kallisto performs particularly well in this context, outperforming widely used aligners such as Bowtie2 and STAR. This is likely due to kallisto’s expectation-maximization (EM) algorithm (described in the Materials and Methods section), which employs a probabilistic model to assign reads from similar transcripts. Previous studies have demonstrated the effectiveness of this approach in quantifying highly repetitive sequences, such as transposons (doi.org/10.1093/bioinformatics/btv422). In the revised manuscript, we are considering the inclusion of a supplementary figure to further support the selection of the mapping algorithm.

Reviewer 2:

We believe that obtaining experimental evidence on the influence of multiple multigene families would represent a significant advancement in the field. However, we would like to emphasize that this is a short communication centered on a specific and biologically relevant observation within a single multigene family. The manuscript is not intended to comprehensively address all aspects of the experiment but rather to highlight what we consider an important biological phenomenon with potential functional implications.

The influence of phenotypic heterogeneity and its possible advantages under environmental pressures has been previously proposed for Trypanosoma cruzi, related trypanosomatids, and other biological systems, ranging from bacteria to tumors (Seco-Hidalgo 2015, doi: 10.1098/rsob.150190 and Luzak 2021, doi: 10.1146/annurev-micro-040821-012953, for a comprehensive review on this topic). While the reviewer is correct in noting that our model does not demonstrate a functional role for TcS heterogeneity, the experimental approaches required to address this question in a large multigene family are highly complex and beyond the scope of this study. However, we acknowledge the importance of clarifying that the proposed functional implications remain speculative, so we will revise the manuscript accordingly.

As the reviewer suggests, in the revised version of the manuscript, we will include additional analyses on the characteristics of frequently expressed TcS genes to identify common features that may explain their expression patterns.

We appreciate the reviewer’s comments and suggestions regarding the clarity of methodological choices and the explanation of key concepts. Accordingly, we will refine the description of our methodology and ensure that our figures are more intuitive and self-explanatory.

Reviewer 3:

We recognize the limitations imposed by gene dropout in our data, as highlighted by the reviewer. In the manuscript, we have aimed to be transparent about this issue and discussed its impact in two separate sections (lines 110–121 and 175–181). To enhance clarity, we will revise these paragraphs to provide a more comprehensive discussion of this limitation. Unfortunately, gene dropout is an inherent limitation of 10x genomics data. Trypanosomatids are not an exception in this regard, and the general metrics of the single-cell RNA-seq data in other reports are equivalent to those obtained in our experiment.

Despite this important limitation, we believe that our comparative analyses (the contrast between TcS and ribosomal protein expression) provide valuable insights into a biological phenomenon with potential functional relevance for the parasite. Furthermore, we are actively working on generating single-cell RNA-seq data using alternative methodologies that improve gene dropout rates. We anticipate that these future studies will help clarify the extent of the phenomenon described in this work.

Author response:

Reviewer #1 (Public review):

Summary:

Liu et al., present glmSMA, a network-regularized linear model that integrates single-cell RNA-seq data with spatial transcriptomics, enabling high-resolution mapping of cellular locations across diverse datasets. Its dual regularization framework (L1 for sparsity and generalized L2 via a graph Laplacian for spatial smoothness) demonstrates robust performance of their model and offers novel tools for spatial biology, despite some gaps in fully addressing spatial communication.

Overall, the manuscript is commendable for its comprehensive benchmarking across different spatial omics platforms and its novel application of regularized linear models for cell mapping. I think this manuscript can be improved by addressing method assumptions, expanding the discussion on feature dependence and cell type-specific biases, and clarifying the mechanism of spatial communication.

The conclusions of this paper are mostly well supported by data, but some aspects of model development and performance evaluation need to be clarified and extended.

We thank the reviewer for their thoughtful comments. We will clarify the model assumptions and the feature selection process to make it more understandable. To clarify, the performance of glmSMA does not depend on cell type. For some rare cell types, the small number of cells can lead to a drop in performance. To better illustrate our results and reduce cell type-specific biases, we will shuffle and randomly sample the cell types.

(1) What were the assumptions made behind the model? One of them could be the linear relationship between cellular gene expression and spatial location. In complex biological tissues, non-linear relationships could be present, and this would also vary across organ systems and species. Similarly, with regularization parameters, they can be tuned to balance sparsity and smoothness adequately but may not hold uniformly across different tissue types or data quality levels. The model also seems to assume independent errors with normal distribution and linear additive effects - a simplification that may overlook overdispersion or heteroscedasticity commonly observed in RNA-seq data.

Thank you for this comment. We acknowledge that the non-linear relationships can be present in complex tissues and may not be fully captured by a linear model. 

Our choice of a linear model was guided by an investigation of the relationship in the current datasets, which include intestinal villus, mouse brain, and fly embryo.

There is a linear correlation between expression distance and physical distance [Nitzan et al]. Within a given anatomical structure, cells in closer proximity exhibit more similar expression patterns. In tissues where non-linear relationships are more prevalent—such as the human PDAC sample—our mapping results remain robust. We acknowledge that we have not yet tested our algorithm in highly heterogeneous regions like the liver, and we plan to include such analyses in future work if necessary. Regarding the regularization parameters, we agree that the balance between sparsity and smoothness is sensitive to tissue-specific variation and data quality. In our current implementation, we explored a range of values to find robust defaults.

(2) The performance of glmSMA is likely sensitive to the number and quality of features used. With too few features, the model may struggle to anchor cells correctly due to insufficient discriminatory power, whereas too many features could lead to overfitting unless appropriately regularized. The manuscript briefly acknowledges this issue, but further systematic evaluation of how varying feature numbers affect mapping accuracy would strengthen the claims, particularly in settings where marker gene availability is limited. A simple way to show some of this would be testing on multiple spatial omics (imaging-based) platforms with varying panel sizes and organ systems. Related to this, based on the figures, it also seems like the performance varies by cell type. What are the factors that contribute to this? Variability in expression levels, RNA quantity/quality? Biases in the panel? Personally, I am also curious how this model can be used similarly/differently if we have a FISH-based, high-plex reference atlas. Additional explanation around these points would be helpful for the readers.

Thank you for this thoughtful comment. The performance of our method is indeed sensitive to the number and quality of selected features. To optimize feature selection, we employed multiple strategies, including Moran’s I statistic, identification of highly variable genes, and the Seurat pipeline to detect anchor genes linking the spatial transcriptomics data with the reference atlas. The number of selected markers depends on the quality of the data. For high-quality datasets, fewer than 100 markers are typically sufficient for accurate prediction. To address this more clearly, we will revise the manuscript to include detailed descriptions of our feature selection process and demonstrate how varying the number of selected features impacts performance.

We evaluated our method across diverse tissue types and platforms—including Slide-seq, 10x Visium, and Virtual-FISH—which represent both sequencing-based and imaging-based spatial transcriptomics technologies. Our model consistently achieved strong performance across these settings. It's worth noting that the performance of other methods, such as CellTrek [Wei et al] and novoSpaRc [Nitzan et al], also depends heavily on feature selection. In particular, performance degrades substantially when fewer features are used.

We do not believe that the observed performance is directly influenced by cell type composition. Major cell types are typically well-defined, and rare cell types comprise only a small fraction of the dataset. For these rare populations, a single misclassification can disproportionately impact metrics like KL divergence due to small sample size. However, this does not necessarily indicate a systematic cell type–specific bias in the mapping. To mitigate this issue, we will implement shuffling and sampling procedures to reduce potential bias introduced by rare cell types.

(3) Application 3 (spatial communication) in the graphical abstract appears relatively underdeveloped. While it is clear that the model infers spatial proximities, further explanation of how these mappings translate into insights into cell-cell communication networks would enhance the biological relevance of the findings.

Thank you for this valuable feedback. We agree that further elaboration on the connection between spatial proximity and cell–cell communication would enhance the biological interpretation of our results. While our current model focuses on inferring spatial relationships, we may provide some cell-cell communications in the future.

(4) What is the final resolution of the model outputs? I am assuming this is dictated by the granularity of the reference atlas and the imposed sparsity via the L1 norm, but if there are clear examples that would be good. In figures (or maybe in practice too), cells seem to be assigned to small, contiguous patches rather than pinpoint single-cell locations, which is a pragmatic compromise given the inherent limitations of current spatial transcriptomics technologies. Clarification on the precise spatial scale (e.g., pixel or micrometer resolution) and any post-mapping refinement steps would be beneficial for the users to make informed decisions on the right bioinformatic tools to use.

Thank you for the comment. For each cell, our algorithm generates a probability vector that indicates its likely spatial assignment along with coordinate information. We will include the resolution and the number of cells assigned to each spot in future versions. In our framework, each cell is mapped to one or more spatial locations with associated probabilities. Depending on the amount of regularization through L1 and L2 norms, a cell may be localized to a small patch or distributed over a broader domain. For the 10x Visium data, we applied a repelling algorithm to enhance visualization [Wei et al]. If a cell’s original location is already occupied, it is reassigned to a nearby neighborhood to avoid overlap. The users can also see the entire regularization path by varying the penalty terms. 

Nitzan M, Karaiskos N, Friedman N, Rajewsky N. Gene expression cartography. Nature. 2019;576(7785):132-137. doi:10.1038/s41586-019-1773-3

Wei, R. et al. (2022) ‘Spatial charting of single-cell transcriptomes in tissues’, Nature Biotechnology, 40(8), pp. 1190–1199. doi:10.1038/s41587-022-01233-1. 

Reviewer #2 (Public review):

Summary:

The author proposes a novel method for mapping single-cell data to specific locations with higher resolution than several existing tools.

Thank you for recognizing our contribution. Our goal was to develop a method that achieves higher spatial resolution in mapping single-cell data compared to existing tools. We are encouraged by the results and will continue to refine the approach to improve accuracy and generalizability across platforms and tissue types.

Strengths:

The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus.

Thank you for this comment. We believe that evaluating our method across diverse tissue types—such as the mouse cortex, human PDAC, and intestinal villus—demonstrates its robustness and broad applicability. We plan to continue expanding these evaluations to additional tissue contexts and species to further validate the method’s generalizability.

Weakness:

(1) Although the researchers claim that glmSMA seamlessly accommodates both sequencing-based and image-based spatial transcriptomics (ST) data, their testing primarily focused on sequencing-based ST data, such as Visium and Slide-seq. To demonstrate its versatility for spatial analysis, the authors should extend their evaluation to imaging-based spatial data.

Thank you for the comment. We have tested our algorithm on the virtual FISH dataset from the fly embryo, which serves as an example of image-based spatial omics data. However, such datasets often contain a limited number of available genes. To address this, we will conduct additional testing on image-based data if needed. The Allen Brain Atlas provides high-quality ISH data, and we can select specific brain regions from this resource to further evaluate our algorithm if necessary [Lein et al]. Currently, we plan to focus more on the 10x Visium platform, as it supports whole-transcriptome profiling and offers a wide range of tissue samples for analysis.

(2) The definition of "ground truth" for spatial distribution is unclear. A more detailed explanation is needed on how the "ground truth" was established for each spatial dataset and how it was utilized for comparison with the predicted distribution generated by various spatial mapping tools.

Thank you for the comment. To clarify how ground truth is defined across different tissues, we provide the following details. Direct ground truth for cell locations is often unavailable in scRNA-seq data due to experimental constraints. To address this, we adopted alternative strategies for estimating ground truth in each dataset:

- 10x Visium Data: We used the cell type distribution derived from spatial transcriptomics (ST) data as a proxy for ground truth. We then computed the KL divergence between this distribution and our model's predictions for performance assessment.

- Slide-seq Data: We validated predictions by comparing the expression of marker genes between the reconstructed and original spatial data.

- Fly Embryo Data: We used predicted cell locations from novoSpaRc as a reference for evaluating our algorithm.

These strategies allowed us to evaluate model performance even in the absence of direct cell location data. In addition, we can apply multiple evaluation strategies within a single dataset.

(3) In the analysis of spatial mapping results using intestinal villus tissue, only Figure 3d supports their findings. The researchers should consider adding supplemental figures illustrating the spatial distribution of single cells in comparison to the ground truth distribution to enhance the clarity and robustness of their investigation.

Thank you for the comment. We will include additional details for this dataset in the supplementary figures. As the intestinal villus is a relatively simple tissue, most existing algorithms performed well on it. For this reason, we did not initially provide extensive details in the main text.

(4) The spatial mapping tests were conducted on various tissues, including the mouse cortex, human PDAC, and intestinal villus. However, the original anatomical regions are not displayed, making it difficult to directly compare them with the predicted mapping results. Providing ground truth distributions for each tested tissue would enhance clarity and facilitate interpretation. For instance, in Figure 2a and Supplementary Figures 1 and 2, only the predicted mapping results are shown without the corresponding original spatial distribution of regions in the mouse cortex. Additionally, in Figure 3c, four anatomical regions are displayed, but it is unclear whether the figure represents the original spatial regions or those predicted by glmSMA. The authors are encouraged to clarify this by incorporating ground truth distributions for each tissue.

Thank you for the comment. To improve visualization, we will include anatomical structures alongside the mapping results in the next version, wherever such structures are available (e.g., mouse brain cortex, human PDAC sample, etc.). Regions will be color-coded to enhance clarity and make the spatial organization easier to interpret.

(5) The cell assignment results from the mouse hippocampus (Supplementary Figure 6) lack a corresponding ground truth distribution for comparison. DG and CA cells were evaluated solely based on the gene expression of specific marker genes. Additional analyses are needed to further validate the robustness of glmSMA's mapping performance on Slide-seq data from the mouse hippocampus.

Thank you for the comment. The ground truth for DG and CA cells was not available. To better evaluate the model's performance, we will compute the KL divergence between the original and predicted cell type distributions, following the same approach used for the 10x Visium dataset.

(6) The tested spatial datasets primarily consist of highly structured tissues with well-defined anatomical regions, such as the brain and intestinal villus. Anatomical regions are not distinctly separated, such as liver tissue. Further evaluation of such tissues would help determine the method's broader applicability.

Thank you for the comment. We have already tested our algorithm on the fly embryo, where anatomical structures are not well defined or clearly separated. If needed, we can further apply glmSMA to more complex tissues such as the liver. To clarify the role of anatomical structures in our model: glmSMA does not require anatomical information as input. Instead, it leverages a distance matrix between cells to apply L2 norm regularization. Despite the absence of anatomical information, the model still demonstrates strong performance. We will include results to illustrate its effectiveness without anatomical input. Additionally, we plan to evaluate the model on tissues where anatomical regions are not clearly delineated.

Lein, E., Hawrylycz, M., Ao, N. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168–176 (2007). https://doi.org/10.1038/nature05453

Reviewer #3 (Public review):

Summary:

The authors aim to develop glmSMA, a network-regularized linear model that accurately infers spatial gene expression patterns by integrating single-cell RNA sequencing data with spatial transcriptomics reference atlases. Their goal is to reconstruct the spatial organization of individual cells within tissues, overcoming the limitations of existing methods that either lack spatial resolution or sensitivity.

Strengths:

(1) Comprehensive Benchmarking:

Compared against CellTrek and Novosparc, glmSMA consistently achieved lower Kullback-Leibler divergence (KL divergence) scores, indicating better cell assignment accuracy.

Outperformed CellTrek in mouse cortex mapping (90% accuracy vs. CellTrek's 60%) and provided more spatially coherent distributions.

(2) Experimental Validation with Multiple Real-World Datasets:

The study used multiple biological systems (mouse brain, Drosophila embryo, human PDAC, intestinal villus) to demonstrate generalizability.

Validation through correlation analyses, Pearson's coefficient, and KL divergence support the accuracy of glmSMA's predictions.

We thank reviewer #3 for their positive feedback and thoughtful recommendations.

Weaknesses:

(1) The accuracy of glmSMA depends on the selection of marker genes, which might be limited by current FISH-based reference atlases.

We agree that the accuracy of glmSMA is influenced by the selection of marker genes, and that current FISH-based reference atlases may offer a limited gene set. To address this, we incorporate multiple feature selection strategies, including highly variable genes and spatially informative genes (e.g., via Moran’s I), to optimize performance within the available gene space. As more comprehensive reference atlases become available, we expect the model’s accuracy to improve further.

(2) glmSMA operates under the assumption that cells with similar gene expression profiles are likely to be physically close to each other in space which not be true under various heterogeneous environments.

While this assumption effectively captures spatial continuity in many cases, we acknowledge that it may not hold across all biological contexts. To address this, we plan to refine our regularization strategy and evaluate the model's performance in heterogeneous tissue regions.

Author response:

Public Reviews

Reviewer #1 (Public review):

Summary:

Kwon et al present a very well-conducted and well-written sieve analysis of rotavirus infections in a passive surveillance network in the US, considering how relative vaccine efficacy changes with genetic distance from the vaccine strains including the whole genome. The results are compelling, supported by a number of sensitivity analyses, and the manuscript is generally easy to follow.

Strengths:

(1) The underlying study base, a surveillance network across multiple sites in the US.

(2) The use of a test-negative design, which is well established for rotavirus, to estimate vaccine efficacy.

(3) The use of genetic distance to measure differences between infecting and vaccine strains, and the innovative use of k-means clustering to make results more interpretable.

(4) The secondary and sensitivity analyses that provide additional context and support for the primary findings.

Weaknesses:

(1) As identified by the authors, there is a limited sample size for the analysis of RV1 (monovalent rotavirus vaccine).

(2) Sieve analyses were originally designed for randomized trials, in which setting their key assumptions are more likely to be met. There is little discussion in this paper of how those assumptions might be violated and what effect that might have on the results. The authors have access to some important confounders, but I believe some more discussion on potential biases in this observational study is warranted.

We appreciate the reviewer’s positive comments and the opportunity to discuss the application of sieve analysis in observational vaccine effectiveness studies, contrasting it with its traditional use in clinical trials assessing vaccine efficacy. We fully acknowledge the reviewer's point that sieve analysis was originally developed for, and is most frequently employed in, randomized controlled trials (RCTs).

Sieve analysis, as defined by Gilbert et al. (2001), has the following core assumptions: (A1) uniform susceptibility to infection for all participants except for vaccine-induced strain-specific effects; (A2) equal exposure (for each strain s = 1,…,K ) distribution between vaccine groups; and (A3), constant strain prevalence. RCTs ensure these through randomization. However, our observational design is vulnerable to violating these assumptions, especially A1 and A3. To address A1 and A3, we adjusted for age (in years), sample collection year, and clinical setting (i.e., outpatient, inpatient, ED), aiming to account for both individual-level and temporal variations.

A2 is particularly challenging in observational settings. We found that study site was correlated with both vaccination status (main predictor) and the strain distribution, potentially violating A2. However, adjusting for study site reversed the expected association. Upon further reflection, we realized that the site-specific differences in strain distributions likely reflect the population-level effect of vaccination, which we believe outweighs the potential confounding by study site as an independent cause of both individual-level vaccination status and strain distributions irrespective of vaccination. Thus, adjusting for site would have obscured this genuine population-level effect, and therefore we elected not to do so. We will include further discussion of this point in the revised manuscript.

Our study demonstrates the unique capacity of sieve analysis to disentangle individual- and population-level effects on vaccine effectiveness in observational settings. We will expand on these considerations, including the potential biases inherent to observational studies and the rationale for our analytical choices, within the discussion section of the revised manuscript.

Reviewer #2 (Public review):

Summary:

This study introduces a new metric for assessing the efficacy of rotavirus vaccines through the genetic distance clustering of strains. The authors analyzed variations in vaccine protection using whole genome sequencing.

Strengths:

Evaluating vaccine efficacy using whole genome sequencing can enhance our understanding of how pathogen evolution influences disease transmission and control.

Weaknesses:

While the study proposed a new method for evaluating vaccine efficacy using genetic information, its weaknesses arise from the insufficient evidence that analyses based on whole genome sequencing are more reliable than those that rely solely on VP7 and VP4 genotypes.

Though most cases received the RV5 vaccine (n=119 compared to n=30 for RV1), Figure 2 and the primary focus of the paper concentrate on RV1, as the authors identified a stronger association with genetic distance.

Additionally, it is unclear whether the difference between the two groups (j=0 versus j=1) is statistically significant for the analysis based on genetic distance to the RV1 strain, as well as for that based on minimum genetic distance to any of the RV5 vaccine strains. In both cases, the confidence intervals show substantial overlap

The authors do not seem to have used a criterion for model selection based on the number of clusters; therefore, k=2 may not represent the optimal number of clusters, particularly in relation to the genetic distance associated with the RV5 vaccine (Figure 1B), which does not appear to show a bimodal distribution.

Finally, outcomes for RV1 are highly associated with both homotypic and heterotypic antibody responses (Supplemental Figure 10), which have already been shown to impact vaccine effectiveness (The Pediatric Infectious Disease Journal 40(12):p 1135-1143, 2021, doi:10.1097/INF.0000000000003286). Given this strong association, the benefit of using genetic distance is unclear, as the GxPx genotype serves as a good proxy for genetic similarity. 

We sincerely appreciate reviewer's careful consideration of our manuscript and their constructive suggestions for improvement.

Regarding the comparison of whole-genome sequencing with traditional VP7/VP4 genotyping, we concur that a more explicit comparison would strengthen our findings. To this end, we plan to incorporate the direct comparison of genetic distance (GD) and genotype-specific vaccine effectiveness (VE) analyses into the main text. Additionally, we will conduct an analysis of VE based on homotypic, partially heterotypic, and fully heterotypic genotype groupings. This will provide a clearer demonstration of the potential added value of GD in refining VE estimates, particularly for future applications. Given the potential for reassortment among the rotavirus gene segments, our analysis highlights that relying solely on the VP7/VP4 genotype can at times be misleading. 

Regarding k-means clustering, we wish to clarify that the selection of k=2 was not arbitrary. It was determined using the elbow method on the total within-sum-of-squares (using the fviz_nbclust function in the factoextra R package, with n=5000 bootstrapping). While we acknowledge that other methods, such as silhouette and gap statistics, may yield different optimal cluster numbers, we prioritized maximizing group sample sizes. We will explicitly state this model selection criterion within the methods section of the revised manuscript.

We acknowledge the reviewer’s concern regarding the overlapping confidence intervals and the statistical significance of the differences between the VE for the j=0 and j=1 groups. One way to address this would be to modify our analysis. Instead of two separate logistic regression models (controls vs j=0 cases, and controls vs j=1 cases), we could employ a multinomial logistic regression model with three categories: controls (reference), j=0 cases, and j=1 cases, then conduct Wald test to directly compare the regression slopes for the j=0 and j=1 cases against controls. We intend to explore this approach in the revised manuscript, which will provide a more rigorous assessment of differences in VE by accounting for the relationship between groups within a single model.

Reviewer #3 (Public review):

Overall, this is an outstanding paper. It presents a novel approach to estimating rotavirus vaccine efficacy; is clearly written and presented; and has implications for this vaccine specifically as well as type-specific vaccine evaluation more generally. The analytical framework is a creative and there is rigorous use of data and statistical approaches. It has long been argued that rotavirus immunity/vaccine performance operates beyond the scale of G/P genotyping. This paper is the first to demonstrate that convincingly, using data on all 11 viral genes and whole genome sequence analysis. I have only minor comments that I recommend should be addressed.

We sincerely thank the reviewer for their highly positive assessment of our manuscript. We will carefully address their minor comments and incorporate their recommendations in the revised manuscript, which we believe will further enhance the clarity and impact of our study.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

The paper by Fournier et al. investigates the sensitivity of neural circuits to changes in intrinsic and synaptic conductances. The authors use models of the stomatogastric ganglion (STG) to compare how perturbations to intrinsic and synaptic parameters impact network robustness. Their main finding is that changes to intrinsic conductances tend to have a larger impact on network function than changes to synaptic conductances, suggesting that intrinsic parameters are more critical for maintaining circuit function.

The paper is well-written and the results are compelling, but I have several concerns that need to be addressed to strengthen the manuscript. Specifically, I have two main concerns:

(1) It is not clear from the paper what the mechanism is that leads to the importance of intrinsic parameters over synaptic parameters.

(2) It is not clear how general the result is, both within the framework of the STG network and its function, and across other functions and networks. This is crucial, as the title of the paper appears very general.

I believe these two elements are missing in the current manuscript, and addressing them would significantly strengthen the conclusions. Without a clear understanding of the mechanism, it is difficult to determine whether the results are merely anecdotal or if they depend on specific details such as how the network is trained, the particular function being studied, or the circuit itself. Additionally, understanding how general the findings are is vital, especially since the authors claim in the title that "Circuit function is more robust to changes in synaptic than intrinsic conductances," which suggests a broad applicability.

I do not wish to discourage the authors from their interesting result, but the more we understand the mechanism and the generality of the findings, the more insightful the result will be for the neuroscience community.

Major comments

(1) Mechanism

While the authors did a nice job of describing their results, they did not provide any mechanism for why synaptic parameters are more resilient to changes than intrinsic parameters. For example, from Figure 5, it seems that there is mainly a shift in the sensitivity curves. What is the source of this shift? Can something be changed in the network, the training, or the function to control it? This is just one possible way to investigate the mechanism, which is lacking in the paper.

(2) Generality of the results within the framework of the STG circuit

(a) The authors did show that their results extend to multiple networks with different parameters (the 100 networks). However, I am still concerned about the generality of the results with respect to the way the models were trained. Could it be that something in the training procedure makes the synaptic parameters more robust than intrinsic parameters? For example, the fact that duty cycle error is weighted as it is in the cost function (large beta) could potentially affect the parameters that are more important for yielding low error on the duty cycle.

(b) Related to (a), I can think of a training scheme that could potentially improve the resilience of the network to perturbations in the intrinsic parameters rather than the synaptic parameters. For example, in machine learning, methods like dropout can be used to make the network find solutions that are robust to changes in parameters. Thus, in principle, the results could change if the training procedure for fitting the models were different, or by using a different optimization algorithm. It would be helpful to at least mention this limitation in the discussion.

(3) Generality of the function

The authors test their hypothesis based on the specific function of the STG. It would be valuable to see if their results generalize to other functions as well. For example, the authors could generate non-oscillatory activity in the STG circuit, or choose a different, artificial function, maybe with different duty cycles or network cycles. It could be that this is beyond the scope of this paper, but it would be very interesting to characterize which functions are more resilient to changes in synapses, rather than intrinsic parameters. In other words, the authors might consider testing their hypothesis on at least another 'function' and also discussing the generality of their results to other functions in the discussion.

(4) Generality of the circuit

The authors have studied the STG for many years and are pioneers in their approach, demonstrating that there is redundancy even in this simple circuit. This approach is insightful, but it is important to show that similar conclusions also hold for more general network architectures, and if not, why. In other words, it is not clear if their claim generalizes to other network architectures, particularly larger networks. For example, one might expect that the number of parameters (synaptic vs intrinsic) might play a role in how resilient the function is with respect to changes in the two sets of parameters. In larger models, the number of synaptic parameters grows as the square of the number of neurons, while the number of intrinsic parameters increases only linearly with the number of neurons. Could that affect the authors' conclusions when we examine larger models?

In addition, how do the authors' conclusions depend on the "complexity" of the non-linear equations governing the intrinsic parameters? Would the same conclusions hold if the intrinsic parameters only consisted of fewer intrinsic parameters or simplified ion channels? All of these are interesting questions that the authors should at least address in the discussion.

We thank Reviewer #1 for their valuable input. We agree with the reviewer that generality of the results may have been overstated. To address this we changed the title of the manuscript to make it more specific to rhythmic circuits and we included a sentence to this effect in the discussion. 

(1) We were more interested in knowing which set of conductances is more robust in a population of models, rather than a mechanism. If such a mechanism exists it will be the subject of a different study.

(2) (a) It is impossible to explore the whole parameter space of these models. Our method to find circuits will leave subsets of circuits out of the study. Our sole goal in constructing the model database was that the activities were similar but the conductances were different.  (b) Of course one could devise a cost function targeting circuits that are more or less robust to changes in one parameter. Whether those exist is a different matter. This is not what we intended to do.

(3) For this we would need a different circuit that produces non-oscillatory activity. A normal pyloric rhythm circuit always produces oscillatory activity unless it is “crashed"either by temperature or perturbations, but even in this case because we don’t have a proper “control” activity (circuits crash in different ways) we would not be able to utilize the same approach.

We think it is a valuable idea to perform a similar study in another small circuit with nonoscillatory (or rhythmic) activities. 

(4) We did not explore the issue of how our results generalize to larger networks as it would be pure speculation. It could be potentially interesting to do a similar sensitivity analysis with a large network trained to perform a simple task. Our understanding is that many large trained networks are extremely sensitive to perturbations in synaptic weights, at the same time that the intrinsic properties of neurons in ANN are typically oversimplified and identical across units. 

Reviewer #2 (Public review):

Summary:

This manuscript presents an important exploration of how intrinsic and synaptic conductances affect the robustness of neural circuits. This is a well-deserved question, and overall, the manuscript is written well and has a logical progression.

The focus on intrinsic plasticity as a potentially overlooked factor in network dynamics is valuable. However, while the stomatogastric ganglion (STG) serves as a well-characterized and valuable model for studying network dynamics, its simplified structure and specific dynamics limit the generalizability of these findings to more complex systems, such as mammalian cortical microcircuits.

Strengths:

Clean and simple model. Simulations are carefully carried out and parameter space is searched exhaustively.

Weaknesses:

(1) Scope and Generalizability:

The study's emphasis on intrinsic conductance is timely, but with its minimalistic and unique dynamics, the STG model poses challenges when attempting to generalize findings to other neural systems. This raises questions regarding the applicability of the results to more complex circuits, especially those found in mammalian brains and those where the dynamics are not necessarily oscillating. This is even more so (as the authors mention) because synaptic conductances in this study are inhibitory, and changes to their synaptic conductances are limited (as the driving force for the current is relatively low).

(2) Challenges in Comparison:

A significant challenge in the study is the comparison method used to evaluate the robustness of intrinsic versus synaptic perturbations. Perturbations to intrinsic conductances often drastically affect individual neurons' dynamics, as seen in Figure 1, where such changes result in single spikes or even the absence of spikes instead of the expected bursting behavior. This affects the input to downstream neurons, leading to circuit breakdowns. For a fair comparison, it would be essential to constrain the intrinsic perturbations so that each neuron remains within a particular functional range (e.g., maintaining a set number of spikes). This could be done by setting minimal behavioral criteria for neurons and testing how different perturbation limits impact circuit function.

(3) Comparative Metrics for Perturbation:

Another notable issue lies in the evaluation metrics for intrinsic and synaptic perturbations. Synaptic perturbations are straightforward to quantify in terms of conductance, but intrinsic perturbations involve more complexity, as changes in maximal conductance result in variable, nonlinear effects depending on the gating states of ion channels. Furthermore, synaptic perturbations focus on individual conductances, while intrinsic perturbations involve multiple conductance changes simultaneously. To improve fairness in comparison, the authors could, for example, adjust the x-axis to reflect actual changes in conductance or scale the data post hoc based on the real impact of each perturbation on conductance. For example, in Figure 6, the scale of the panels of the intrinsic (e.g., g_na-bar) is x500 larger than the synaptic conductance (a row below), but the maximal conductance for sodium hits maybe for a brief moment during every spike and than most of the time it is close to null. Moreover, changing the sodium conductance over the range of 0-250 for such a nonlinear current is, in many ways, unthinkable, did you ever measure two neurons with such a difference in the sodium conductance? So, how can we tell that the ranges of the perturbations make a meaningful comparison?

We thank Reviewer #2 for their comments. We agree with both reviewers about scope and generalizability. We changed the title of the manuscript and included a sentence in the discussion to address this. 

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Line 63: Tau_b is tau in Fig 1B? What is the 'network period' tau_n? Both are defined in the methods, but it would be good to clarify here and also in the figure.

This was fixed. Tau_b is the  bursting period and we indicated it in the figure. Network period means the period of the network activity. This was rewritten.  

(2) Line 74: "maximal conductances g_i." What is i? I can imagine what you meant, but it would be good to clarify the notation.

There are multiple different currents. Letter ‘i' is an index over the different types. It now reads as follows,

"The activity of the network depends on the values of the maximal conductances g ̄ i, where i is an index corresponding to the different current types (Na,CaS,CaT,Kd,KCa,A,H,Leak IMI)"

(3) Line 78: "conductances are changed by a random amount." How much is the "random amount"? In percentages? 

We fixed this sentence. This is how it reads now, 

"The blue trace in Figure 1C corresponds to the activity of the same model when each  of the intrinsic conductances is changed by a random amount within a range between 0  (completely removing the conductance) and twice its starting value, 2×gi, or equivalently, an increment of 100%."

Similarly, in Line 87: "by a similar percent." Can you provide Figures 1E-F in percentages? Are the percentages the same?

The phrase "by a similar percent.” Is misleading and unimportant. Thank you, we removed it. 

(4) Line 113: Why did you add I_MI? Is it important for the results or for the conclusions?

I_MI was added because the current is known to be there and it is not more or less important for the results or conclusions than any other current. 

(5) Line 117: "We used a genetic algorithm to generate a database." Confusing. I guess you meant that you used genetic algorithms to optimize the cost function.

Thank you for this comment. We fixed this sentence, see below. 

“We used a genetic algorithm to optimize the cost function, and in this way generated a database of N = 100 models with different values of maximal conductances (Holland 88)."

(6) Line 136: "The models in the database were constrained to produce solutions whose features were similar to the experimental measurements." Why are there differences in the features? Is this an optimization issue? I thought you wanted to claim that there are degenerate solutions, that is, solutions where the parameters are different, but the output is identical. Please clarify.

The concept of degenerate solutions does not imply that the solutions are mathematically identical. In biology this means that they provide very similar functions, but do so with different underlying parameters (in this case, maximal conductances). The activity of the pyloric network is slightly different across animals, and it also changes over time within the same individual. Variation across models reflects individual variation in the biological circuit, and it is strength of our modeling approach. The function of the circuits are equally good because they produce biologically realistic patterns, although the details of the activity patterns show differences. 

(7) Line 139: "distributed (p > 0.05)." What test did you use? N? Similarly, at Lines 218, 241, 239, etc. Please be more rigorous when reporting statistical tests.

Thank you. We now specify the test we utilized every time we report a p value. 

(8) Line 143: "In this case, it is not possible to identify clusters, suggesting that there are no underlying relationships between the features in the model database." The 2D plot is misleading, as the features are in 11 dimensions. Claims should be about the 11D space, not projections onto 2D. In fact, I don't think you can rule out correlations between the features based on the 2D plots. For example, shouldn't there be correlations between the on and off phases and the burst durations?

Thank you. These sentences were confusing and were removed. We added the following sentence to the end of that paragraph.

"Because the feature vectors are similar, their t-SNE projections do not form groups or clusters."

(9) Related to this, I don't understand this sentence: "Even though the conductances are broadly distributed over many-fold ranges, the output of the circuits results in tight yet uncorrelated distributions.”

This sentence is confusing and was removed. 

(10) Line 158: Repetition of Line 152: Figure 3 shows the currentscapes of each cell in two model networks.

We removed the second instance of the repeated sentences. 

(11) Line 160: "yet the activity of the networks is similar." Well, they are similar, but not identical. I can also say that the current scapes are 'similar'. This should be better quantified and not left as a qualitative description.

While this is an interesting point it will not change the results and conclusions of the present study. The network models are different since the values of their maximal conductances are distributed over wide ranges.  

(12) Line 218: midpoint parameter? Is that b - the sharpness? Please be consistent. Regarding the mechanism (see above) - any ideas what leads to this shift in the sensitivity curves between the two types of parameters?

Yes, we made a mistake. ‘b’ is the midpoint parameter. This was fixed in the text, thank you.

(13) Figure 6 illustrates why synaptic parameters are more robust, but it is not quantified. Why not provide a quantitative measure for this claim? For example, calculate the colored area within the white square for each pair, for each cell, and for each model. Show that these measures can predict improved robustness for one model over another and for synaptic vs. intrinsic parameters.

The ratio of areas of the colored and non-colored regions in the whole hyperboxes (for intrinsic and synaptic conductances) is the number reported in the y-axis of the sensitivity curves when we include all conductances (and not just a pair). 

We computed the ratios of the colored/noncolored areas in all panels in figure 6 and now report these quantities as follows, 

"We computed the proportions of areas of the white boxes that correspond to pyloric activity. These values for the intrinsic conductances panels are PD = 0.58, LP = 0.50, PY = 0.49, and the proportions for the synaptic conductances panels are PDPY = 0.62, P DLP = 0.87, and LPPD = 0.94. The occupied areas for synaptic conductances are larger than in the intrinsic conductances panels, consistent with our finding that the circuits’ activities are more robust to changes in synaptic conductances versus changes in intrinsic conductances."

"As before, we computed the proportion of areas of pyloric activity within the white boxes: PD = 0.61, LP = 0.55, PY = 0.52, and the proportions for the synaptic conductances panels are PDPY = 0.88, PDLP = 0.87, and LPP D = 0.83. These results provide an intuition of the complexities of GP . Not only are these regions hard-to-impossible to characterize in one circuit, but they are also different across circuits.” 

(14) Does the sign of the synaptic weights affect the conclusions?

We did not explore this issue because all chemical synapses in this network are inhibitory.

(15) Line 492: typo: deltai.

We fixed this.

Reviewer #2 (Recommendations for the authors):

(1) Line 301 - you can also add Williams and Fletcher 2019 Neuron.

We added the reference. Thank you. 

(2) Line 316 - this is a strange comment as these exact regions that were shown intrinsic plasticity (e.g., Losonczy, Attila, Judit K. Makara, and Jeffrey C. Magee. "Compartmentalized dendritic plasticity and input feature storage in neurons." Nature 452.7186 (2008): 436-441).

We did not understand this comment. 

(3) I found only one citation for the work of Turrigiano, the most relevant of which is only mentioned in the Method section. This is odd, as her work directly relates how synaptic conductance perturbation results in changes in intrinsic conductance.

We included more references to the work of Turrigiano to provide more context. 

"Desai, Niraj S., Lana C. Rutherford, and Gina G. Turrigiano. "Plasticity in the intrinsic excitability of cortical pyramidal neurons." Nature neuroscience 2, no. 6 (1999): 515-520.” "Desai, Niraj S., Sacha B. Nelson, and Gina G. Turrigiano. "Activity-dependent regulation of excitability in rat visual cortical neurons." Neurocomputing 26 (1999): 101-106.”

(4) Line 329 - The list of citations is very limited regarding studies of ext/int balance which started really way before 2009. Please give some of the credit to the classics.

We included the following additional references.

Van Vreeswijk, Carl, and Haim Sompolinsky. "Chaos in neuronal networks with balanced excitatory and inhibitory activity." Science 274, no. 5293 (1996): 1724-1726.

Rubin, Ran, L. F. Abbott, and Haim Sompolinsky. "Balanced excitation and inhibition are required for high-capacity, noise-robust neuronal selectivity." Proceedings of the National Academy of Sciences 114, no. 44 (2017): E9366-E9375.

Wang, Xiao-Jing. "Macroscopic gradients of synaptic excitation and inhibition in the neocortex." Nature reviews neuroscience 21, no. 3 (2020): 169-178.

Lo, Chung-Chuan, Cheng-Te Wang, and Xiao-Jing Wang. "Speed-accuracy tradeoff by a control signal with balanced excitation and inhibition." Journal of Neurophysiology 114, no. 1 (2015): 650-661.

(5) In Figure 1B, why does it say 'OFF' when the neuron is spiking?

The label indicates the interval of time elapsed between the first spike in the PD neuron (taken as a reference), and the last spike in the burst (PD off). 

Summary of changes to figures:

Figure 1:

Fixed labels indicating bursting period and burst duration.

Figure 5:

Added labels in panels C and D specifying the symbol corresponding to the sigmoidal parameter.

Additional changes

We changed the title of the manuscript as follows:

"Rhythmic circuit function is more robust to changes in  synaptic than intrinsic conductances." We included the following sentence at the end of the Discussion Section. 

"We believe our results will hold for other rhythmic circuits and will be relevant for similar studies in other circuits with more complex functions.”

We realized we made a mistake with the units for maximal conductances. They were incorrectly expressed in nS (nano Siemens) in the figure labels, and correctly expressed in micro Siemens in the methods section. This was fixed and now conductances are expressed in micro Siemens consistently in the manuscript.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

This study takes a detailed approach to understanding the effect of menopausal hormone therapy (MHT) in the brain aging of females. Neuroimaging data from the UK Biobank is used to explore brain aging and shows an unexpected effect of current MHT use and poorer brain health outcomes relative to never users. There is considerable debate about the benefits of MHT and estrogens in particular for brain health, and this analysis illustrates that the effects are certainly not straightforward and require greater consideration.

Strengths:

(1) The detailed approach to obtaining important information about MHT use from primary care records. Prior studies have suggested that factors such as estrogen/progestin type, route of administration, duration, and timing of use relative to menopause onset can contribute to whether MHT benefits brain health.

(2) Consideration of type of menopause (spontaneous, or surgical) in the analysis, as well as sensitivity diagnoses to rule out the effect being driven by those with clinical conditions.

(3) The incorporation of the brain age estimate along with hippocampal volume to address brain health.

(4) The complex data are also well explained and interpretations are reasonable.

(5) Limitations of the UK Biobank data are acknowledged

We thank the reviewer for their time and the positive evaluation of our manuscript.

Weaknesses:

(1) Lifestyle factors are listed and the authors acknowledge group differences (at least between current users and never users of MHT). I was not able to find these analyses showing these differences.

We highlighted and tested for group differences in lifestyle scores, and the results are shown in Table 1-3, column p-value. As highlighted in the method section (page 9): “The lifestyle score was calculated using a published formula (69), and included data on sleep, physical activity, nutrition, smoking, and alcohol consumption (see supplementary Note 3, Table S2)”. In line with reviewer 1 suggestion to the authors, we now included an additional table testing for group differences in the specific lifestyle factors constituting the lifestyle score in the supplementary materials (Table S2). Please find a more detailed response below (Recommendations for the authors, Response to Comment 1).

(2) The distribution of women who were not menopausal was unequal across groups, and while the authors acknowledge this, one wonders to what extent this explains the observed findings.

We agree with the reviewer that the unequal distribution of women across groups can influence the observed findings. We have made minor edits to highlight this important topic more explicitly in the discussion:

Discussion (page 21): “Current MHT users were significantly younger than past- and never-users, and around 67 % were menopausal relative to over 80% in the past- and never-user groups. The unequal distribution of age and menopausal status across groups may have influenced the observed findings. For instance, a larger proportion of the current users might be in the perimenopausal phase, which is often associated with debilitating neurological and vasomotor symptoms (1). MHT is commonly prescribed to minimize such symptoms. Although MHT initiation during perimenopause has been associated with improved memory and hippocampal function, as well as lower AD risk later in life (15), the need for MHT might in itself be an indicator of neurological changes (71); here potentially reflected in higher BAG and lower hippocampal volumes. After the transition to menopause, symptoms might subside and some perimenopausal brain changes might revert or stabilize in the postmenopausal phase 5. Although the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, our results might be capturing subtle disturbances during perimenopause that later stabilize. This could explain why the largely postmenopausal groups of past MHT users and never-users present with lower GM and WM BAG than the current user group. Considering the critical window hypothesis emphasizing perimenopause as a key phase for MHT action (29,43), future longitudinal studies are crucial to clarify the interplay between neurological changes and MHT use across the menopause transition.”

Discussion (page 25): “In addition, previous studies highlight that UK Biobank participants are considered healthier than the general population based on several lifestyle and health-related factors (89, 90). This healthy volunteer bias increases with age, likely resulting in a disproportionate number of healthier older adults. Together with the imbalance in age distributions across groups, this might explain the less apparent brain aging in the older MHT user groups. We have previously highlighted that age is negatively associated with the number of APOE ε4 carriers in the UK Biobank (21), which is indicative of survivor bias.”

(3) While the interpretations are reasonable, and relevant theories (healthy cell & critical window) are mentioned, the discussion is missing a more zoomed-out perspective of the findings. While I appreciate wanting to limit speculation, the reader is left having to synthesize a lot of complex details on their own. A particularly difficult finding to reconcile is under what conditions these women benefit from MHT and when do they not (and why that may be).

We thank the reviewer for this comment. As the presented data is cross-sectional and does not enable causal inference, we have refrained from a more zoomed-out interpretation of the results to avoid undue speculations. However, where applicable, we have discussed our findings in a broader context such as the effects of MHT use on the brain across the menopausal transition (discussion page 21) and the effects of MHT use on the brain in the presence and absence of bilateral oophorectomy and/or hysterectomy (discussion page 25).

To best inform the reader about the scope of our paper, we would like to highlight the following sentences in our discussion (page 24):

“The current work represents the most comprehensive study of detailed MHT data, APOE ε4 genotype, and several brain measures in a large population-based cohort to date. Overall, our findings do not unequivocally support general neuroprotective effects of MHT, nor do they indicate severe adverse effects of MHT use on the female brain. The results suggest subtle yet complex relationships between MHT’s and brain health, highlighting the necessity for a personalized approach to MHT use. Importantly, our analyses provide a broad view of population-based associations and are not designed to guide individual-level decisions regarding the benefits versus risks of MHT use.”

And the conclusion (page 25): “In conclusion, our findings suggest that associations between MHT use and female brain health might vary depending on duration of use and past surgical history. Although the effect sizes were generally modest, future longitudinal studies and RCTs, particularly focused on the perimenopausal transition window, are warranted to fully understand how MHT use influences female brain health. Importantly, considering risks and benefits, decisions regarding MHT use should be made within the clinical context unique to each individual.”

Reviewer #1 (Recommendations for the authors):

Can the authors provide:

(1) More information about which aspects of lifestyle factors were different between the groups, and how these factors may have contributed to the observed findings (if possible, without burying this information in the supplemental)?

We thank the reviewer for this suggestion. We now added a table comparing lifestyle factors contained in the lifestyle score by MHT user status using t-tests (continuous variables) or χ2 tests (see Table S2). The results are referred to in the main manuscript result section under “Sample characteristics”, and the table (Table S2) is provided in the supplements not to overburden the main text, in line with input from reviewer 3.

We updated the main text to refer to Table S2 and updated the supplementary Note 3 (page 2-3) to include the results of the comparison of the lifestyle factors contained in the lifestyle score by MHT user status.

Methods, page 9:“The lifestyle score was calculated using a published formula (69), and included data on sleep, physical activity, nutrition, smoking, and alcohol consumption (see supplementary Note 3, Table S2).”

Results, page 13: “Sample demographics including lifestyle score, stratified by MHT user group, surgical history among MHT users, and estrogen only MHT or combined MHT use, are summarized in Table 1, 2 and 3, respectively. MHT user group differences for each lifestyle factor contained in the lifestyle score are shown in Table S2.”

“Note 3| Lifestyle Score

The lifestyle score was calculated based on sleep duration, time spent watching television, current and past smoking status, alcohol consumption frequency, physical activity level (number of days per week of moderate/vigorous activity for at least 10 minutes), intake of fruits and vegetables, and intake of oily fish, beef, lamb/mutton, pork and processed meat (for details see (10)). Each unhealthy lifestyle factor was scored with 1 point (e.g., smoking), and participants points were summed to generate an unweighted score (from 0-9): the higher the lifestyle score, the unhealthier the participant’s lifestyle.

A comparison of the lifestyle factors contained in the lifestyle score by MHT user status is presented in Table S2. In summary, we found that current MHT were more often smokers than never-users, had a higher alcohol intake than never- and past MHT users, reported the lowest fruit and vegetable intake relative to never-users and past MHT users, and stated lower moderate activity levels relative to past MHT users. Past MHT users reported higher alcohol intake than never-users, spend more time watching TV relative to never- and current-users, consumed more beef, pork, lamb/mutton, and processed meat than never-users, and reported lower vigorous activity levels relative to never-users. However, oily fish intake and fruit and vegetable intake was higher among past MHT users relative to never-and current-users. Self-reported sleep duration did not differ between MHT user groups.”

(2) A greater description of the 2 main theories of MHT effects on the brain (healthy cell vs critical window). Can the authors also provide a more thorough explanation for how the findings fit with these theories.

We thank the reviewer for this comment. We have described our findings in the context of the critical window hypothesis (discussion, page 21, paragraph 2), the healthy cell bias hypothesis (discussion, page 22, paragraph 3), and healthy user bias hypothesis (discussion, page 22, paragraph 4). We refrained from a more thorough explanation to avoid undue speculations.

(3) Reflect more on what the findings may indicate as to who benefits from MHT, and why. There are some references that the authors may want to add, particularly related to recent findings from premenopausal bilateral oophortectomies that also speak to when (and for whom) MHT use might benefit.

We thank the reviewer for this feedback. We have included additional references in the revised manuscript as follows:

Discussion, page 23: “It is also possible that the timing between MHT use and surgery is more tightly controlled and therefore more beneficial for brain aging (43). For instance, studies suggest that MHT may mitigate the potential long-term adverse effects of bilateral oophorectomy before natural menopause on bone mineral density as well as cardiovascular, cognitive and mental health (79-81). In addition, a 2024 UK Biobank study found that ever used MHT was associated with decreased odds of Alzheimer’s disease in women with bilateral oophorectomy (82).”  

(79) Blumel JE, Arteaga E, Vallejo MS, et al. Association of bilateral oophorectomy and menopause hormone therapy with mild cognitive impairment: the REDLINC X study. Climacteric 2022;25:195-202.

(80) Kaunitz AM, Kapoor E, Faubion S. Treatment of Women After Bilateral Salpingo-oophorectomy Performed Prior to Natural Menopause. JAMA 2021;326:1429-1430.

(81) Stuursma A, Lanjouw L, Idema DL, de Bock GH, Mourits MJE. Surgical Menopause and Bilateral Oophorectomy: Effect of Estrogen-Progesterone and Testosterone Replacement Therapy on Psychological Well-being and Sexual Functioning; A Systematic Literature Review. J Sex Med 2022;19:1778-1789.

(82) Calvo N, McFall GP, Ramana S, et al. Associated risk and resilience factors of Alzheimer's disease in women with early bilateral oophorectomy: Data from the UK Biobank. J Alzheimers Dis 2024;102:119-128.

Reviewer #2 (Public review):

Summary:

In this observational study, Barth et al. investigated the association between menopausal hormone therapy and brain health in middle- to older-aged women from the UK Biobank. The study evaluated detailed MHT data (never, current, or past user), duration of mHT use (age first/last used), history of hysterectomy with or without bilateral oophorectomy, APOEE4 genotype, and brain characteristics in a large, population-based sample. The researchers found that current mHT use (compared to never-users), but not past use, was associated with a modest increase in gray and white matter brain age gap (GM and WM BAG) and a decrease in hippocampal volumes. No significant association was found between the age of mHT initiation and brain measures among mHT users. Longer duration of use and older age at last MHT use post-menopause were associated with higher GM and WM BAG, larger WMH volumes, and smaller hippocampal volumes. In a sub-sample, after adjusting for multiple comparisons, no significant associations were found between detailed mHT variables (formulations, route of administration, dosage) and brain measures. The association between mHT variables and brain measures was not influenced by APOEE4 allele carrier status. Women with a history of hysterectomy with or without bilateral oophorectomy had lower GM BAG compared to those without such a history. Overall, these observational data suggest that the association between mHT use and brain health in women may vary depending on the duration of use and surgical history.

Strengths:

(1) The study has several strengths, including a large, population-based sample of women in the UK, and comprehensive details of demographic variables such as menopausal status, history of oophorectomy/hysterectomy, genetic risk factors for Alzheimer's disease (APOE ε4 status), age at mHT initiation, age at last use, duration of mHT, and brain imaging data (hippocampus and WMH volume).

(2) In a sub-sample, the study accessed detailed mHT prescription data (formulations, route of administration, dosage, duration), allowing the researchers to study how these variables were associated with brain health outcomes. This level of detail is generally missing in observational studies investigating the association of mHT use with brain health.

We thank the reviewer for their time and the positive evaluation of our manuscript.

Weaknesses:

(1) While the study has many strengths, it also has some weaknesses. As highlighted in an editorial by Kantarci & Manson (2023), women with symptoms such as subjective cognitive problems, sleep disturbances, and elevated vasomotor symptoms combined with sleep disturbances tend to seek mHT more frequently than those without these symptoms. The authors of this study have also indicated that the need of mHT use which might be associated with these symptoms may be indicators of preexisting neurological changes, potentially reflecting worse brain health scores, including higher BAG and lower hippocampal volume and/or higher WMH. However, among current users, how many of these women have these symptoms could not be reported in the study. Women with these vasomotor symptoms who are using mHT are more likely to stay longer in the healthcare system compared with those without these symptoms and no MHT use history. The authors noted that the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, limiting the study's ability to understand how these variables influence outcomes.

We thank the reviewer for the succint synopsis of the limitations highlighted in discussion, page 21. We have now added the mentioned reference, 2023 editoral by Kantarci & Manson, to the discussion as well (see reference 71).

Discussion (page 21): “Current MHT users were significantly younger than past- and never-users, and around 67 % were menopausal relative to over 80% in the past- and never-user groups. The unequal distribution of age and menopausal status across groups may have influenced the observed findings. For instance, a larger proportion of the current users might be in the perimenopausal phase, which is often associated with debilitating neurological and vasomotor symptoms (1). MHT is commonly prescribed to minimize such symptoms. Although MHT initiation during perimenopause has been associated with improved memory and hippocampal function, as well as lower AD risk later in life (15), the need for MHT might in itself be an indicator of neurological changes (71); here potentially reflected in higher BAG and lower hippocampal volumes. After the transition to menopause, symptoms might subside and some perimenopausal brain changes might revert or stabilize in the postmenopausal phase 5. Although the UK Biobank lacks detailed information on menopausal symptoms and perimenopausal staging, our results might be capturing subtle disturbances during perimenopause that later stabilize. This could explain why the largely postmenopausal groups of past MHT users and never-users present with lower GM and WM BAG than the current user group. Considering the critical window hypothesis emphasizing perimenopause as a key phase for MHT action (29,43), future longitudinal studies are crucial to clarify the interplay between neurological changes and MHT use across the menopause transition.”

(2)  Earlier observational studies have reported conflicting results regarding the association between mHT use and the risk of dementia and brain health. Contrary to some observational studies, three randomized trials (WHI, KEEPS, ELITE) (Espeland et al 2013, Gleason et al 2015; Henderson et al 2016) demonstrated neither beneficial nor harmful effects of mHT (with varying doses and formulations) when initiated closer to menopause (<5 years). While strong efforts were made to run proper statistical analyses to investigate the association between mHT use and brain health, these results reflect mainly associations, but not causal relationships as also stated by the authors.

We thank the reviewer for pointing that out.

(3)  Furthermore, observational studies have intrinsic limitations, such as a lack of control over switching mHT doses and formulations, a lack of laboratory measures to confirm mHT use, and reliance on self-reported data, which may not always be reliable. The authors caution that these findings should not guide individual-level decisions regarding the benefits versus risks of mHT use. However, the study raises new questions that should be addressed by randomized clinical trials to investigate the varying effects of MHT on brain health and dementia risk.

We thank the reviewer for making our efforts in providing proper disclaimers in the discussion visible.

Reviewer #2 (Recommendations for the authors):

(1) The study could benefit from extending these findings by adding plasma biomarkers of AD and PET imaging markers to further study the association of mHT variables with brain health.

We agree with the reviewer that such markers would be beneficial for elucidating the association between MHT variables and brain health. Unfortunately, these markers are not readily available in the UK Biobank.

(2) The study's reliance on a predominantly white cohort limits the generalizability of the findings to more diverse populations. This homogeneity may not capture the full spectrum of responses to MHT across different ethnic and genetic backgrounds.

We fully agree with the reviewers statement and state this limitation in the discussion (page 25) as follows:

“In addition to these inherent biases in aging cohorts, the ethnic background of the sample is homogeneous (> 96% white), further reducing the generalizability of the results.”

(3) The study may benefit by editing the following information in the introduction: "In summary, WHIMS, HERS, and KEEPS mainly relied on orally administered CEE in older-aged or recently postmenopausal females." KEEPS used two routes and formulations (transdermal estradiol and oCEE, both with micronized progesterone).

We thank the reviewer for catching this oversight. We removed the sentence to avoid ambiguities and revised the sentence specifically refering to the KEEPS study as follows:

Introduction, page 3: “In contrast, administering oral CEE or transdermal estradiol plus micronized progesterone in recently postmenopausal females did not alter cognition in the Kronos Early Estrogen Prevention Study (KEEPS) (28).”

(4) The study may benefit by editing the following statement in the introduction: "oral CEE use in combination with MPA seems to increase the risk for AD regardless of timing": I would suggest revising this statement, which is based on review article 29. The statement of the adverse effect of oCEE regardless of the time of start contradicts earlier randomized clinical findings. I think it is important to make a distinction between the outcomes of randomized control trials and observational studies. The WMIHS (Shumaker et al., 2003) (randomized control trial) reported that there was an increased risk of dementia for women (who were more than 10 years from the onset of menopause when the therapy was initiated) in oCEE + MPA compared to placebo. Two other long-duration randomized trials tested the effect of oral oestrogen and progesterone treatment on cognitive function in women who started treatment shortly after menopause (within 3 or 6 years) did not find evidence that treatment benefits or harms cognitive function compared with placebo (Gleason et al., 2015; Henderson et al., 2016). A short-term (4 months) randomized trial (Maki et al 2007 (Maki et al., 2007) (mentioned in ref 29) reported a potential negative effect of CEE/MPA on verbal memory in women who started HT shortly after menopause (within 3 years). The study did not investigate the risk of dementia, and the duration of use of HT was short-term.

We thank the reviewer for this detailed input. After checking the provided references, we rephrased the sentence as follows:

Introduction, page 4:“Although emerging evidence supports this hypothesis (30, 31), oral CEE use in combination with MPA has been found to increase the risk for memory decline regardless of timing (26, 29, 32).”

We believe this formulation is more in line with the evidence provided by Shumaker et al. 2003, Maki et al. 2007 and the other references provided in the review paper by Maki and colleagues (mentioned in ref. 29). The reviewer further refers to Gleason et al. 2015 and Henderson et al. 2016, however both RCTs use micronized progesterone, not MPA, thereby not supporting the statement.

(26) Shumaker SA, Legault C, Rapp SR, et al. Estrogen plus progestin and the incidence of dementia and mild cognitive impairment in postmenopausal women: the Women's Health Initiative Memory Study: a randomized controlled trial. JAMA 2003;289:2651-2662.

(29) Maki PM. Critical window hypothesis of hormone therapy and cognition: a scientific update on clinical studies. Menopause 2013;20:695-709.

(32) Maki PM, Gast MJ, Vieweg AJ, Burriss SW, Yaffe K. Hormone therapy in menopausal women with cognitive complaints: a randomized, double-blind trial. Neurology 2007;69:1322-1330.

Reviewer #3 (Public review):

In this study Barth et al. present results of detailed analyses of the relationships between menopausal hormone therapy (MHT), APOE ε4 genotype, and measures of anatomical brain age in women in the UK Biobank. While past studies have investigated the links between some of these variables (including works by the authors themselves), this new study adds more detailed MHT variables, surgical status, and additional brain aging measures. The UK biobank sample is large, but it is a population cohort and many of the MHT measures are self-reported (as the authors point out). However, the authors present a solid analysis of the available information which shows associations between MHT user status, length of MHT use, as well as surgical status with brain age. However, as the authors themselves state, the results do not unequivocally support the neuroprotective or adverse effect of MHT on the brain. I think this work strengthens the case for the need of better-designed longitudinal studies investigating the effect of MHT on the brain in the peri/post-menopausal stage.

Strengths:

(1) The authors addressed the statistical analyses rigorously. For example, multiple testing corrections, outlier removal, and sensitivity analysis were performed carefully. Ample background information is provided in the introduction allowing even individuals not familiar with the field to understand the motivation behind the work. The discussion section also does a great job of addressing open questions and limitations. Very detailed results of all statistical tests are provided either in the main text or in the supplementary information.

We thank the reviewer for their time and the positive evaluation of our manuscript.

Weaknesses:

(1) For me, the biggest weakness was the presentation of the results. As many variables are involved and past studies have investigated several of these questions, it would have helped to better clarify the analysis and questions that are addressed by this study in particular and what sets this work apart from past studies. The information is present in the manuscript but better organization might have helped. For example, a figure depicting the key questions near the beginning of the manuscript would have been very helpful for me. The Tables also contain a lot of information but I wonder if there might be a way to capture the most relevant information more succinctly (either in Table format or in a figure) for the main text.

We thank the reviewer for this comment. We do agree that with the large number of analyses it can be hard to keep an overview. We now added a Figure summarizing the main and sensitity analyses by sample.

(2) Another concern I had was the linear models investigating the effects of these MHT variables on the brain age gap. The authors have included "age" as one of the parameters in this analysis. I wonder if adding a quadratic age factor age2 in the model might have improved the fit since many brain phenotypes tend to show quadratic brain age effects in the 40 to 80-year age range.

We thank the reviewer for this suggestion. We have rerun the main analysis in the whole sample (model 1) with age squared as an additional covariate, and compared the gray matter brain age gap model fits using the corrected Akaike Information Criterion (AIC). All models with age squared had a better model fit than models without age squared (see Author response table 1). Hence, in the revised manuscript, we added a sensitivity analysis rerunning the model 1 with age squared to account for potential non-linear effect. The results were largely consistent. The manuscript was revised as follows to reflect the added analysis:

Sensitivity analysis (Methods, Page 11): “To test whether the results were influenced by the inclusion of participants with ICD-10 diagnosis or by non-linear effects of age, the main analyses (models 1-2) were re-run excluding the sub-sample with diagnosed brain disorders (see supplementary Note 2) or adding age(2) as additional covariate, respectively.”

Sensitivity analysis (Results, Page 20): “The results were consistent after removing participants with ICD-10 diagnoses known to impact the brain (see Table S9 for model 1 analyses and Table S10 for model 2 analyses), after additionally adjusting for age(2) (see Table S11), and after removing extreme values (see Table S12 for model 1 analyses).”

Author response table 1.

Gray matter brain age gap model selection based on corrected Akaike Information Criterion (AICc)

Abbreviations and explanations of parameters: MHT = menopausal hormone therapy, K = number of estimated parameters for each model, AICc = the information criterion requested for each model, ΔAICc = the appropriate delta AIC component depending on the information criteria selectedModelLik = the relative likelihood of the model given the data, AICcWT = Akaike weights to indicate the level of support in favor of any given model being the most parsimonious among the candidate model sets, LL = log-likelihood of each model.

Reviewer #3 (Recommendations for the authors):

(1) Please note typo in Figures 2 and 3 legend "GM WM".

We thank the reviewer for catching this typo and we changed it to BAG GM and BAG WM for all Figures for consistency.

Author response:

The following is the authors’ response to the previous reviews

Reply to the comments of the second referee

We sincerely appreciate the positive evaluation and the useful suggestions on our manuscript.

(1) The authors identified key metabolites affecting responses to perturbations in two ways: (i) by fixing a metabolite's value and (ii) by performing a sensitivity analysis. It would be helpful for the modeling community to understand better the differences and similarities in the obtained results. Do both methods identify substrate-level regulators? Is freezing a metabolite's dynamics dramatically changing the metabolic response (and if yes, which ones are so different in the two cases)? Does the scope of the network affect these differences and similarities? 

Thank you for these suggestions. We compared the Sobolʼ total sensitivity index with the absolute values of the change in the response coefficient (Figure S6 in the revised manuscript). There is no clear relationship between the two quantities. The Sobolʼ sensitivity analysis quantifies how a perturbation on the concentration of a metabolite X contributes to the overall dynamics. On the other hand, the analysis in which metabolitesʼ concentrations are fixed measures how strongly metabolite X helps propagate the perturbations on the other metabolites throughout the metabolic network. In other words, in the Sobolʼ analysis, we evaluate the outcome when the perturbation is applied directly to metabolite X, whereas in the fixing-metabolites analysis, we consider perturbations applied to other metabolites and assess how X influences those perturbations. We believe this conceptual difference explains why the two quantities do not correlate. We suspect that this lack of correlation is independent of the networkʼs scope, because each method evaluates a different aspect of the system.  We would say that both methods identify the effect of the metabolite dynamics on the overall dynamics whatever the form is, i.e. the methods do not distinguish the perturbation on the metabolite affecting the overall dynamics by whether the stoichiometric (reactant) way or, the substrate-level regulations. Thus, identifying the substrate-level regulation by utilizing the methods would be challenging. 

(2) Regarding the issues the authors encountered when performing the sensitivity analysis, they can be approached in two ways. First, the authors can check the methods for computing conserved moieties nicely explained by Sauro's group (doi:10.1093/bioinformatics/bti800) and compute them for large-scale networks (but beware of metabolites that belong to several conserved pools). Otherwise, the conserved pools of metabolites can be considered as variables in the sensitivity analysis-grouping multiple parameters is a common approach in sensitivity analysis. 

Thank you for this helpful suggestion. Following the method described in the reference, we have computed the Sobolʼ sensitivity index of NADH, NADPH, and Q8H2 (with their counterparts algebraically solved and treated as dependent variables). We have updated Figure S5 accordingly.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review): 

Summary:

The authors examine the role of the medial prefrontal cortex (mPFC) in cognitive control, i.e. the ability to use task-relevant information and ignore irrelevant information, in the rat. According to the central-computation hypothesis, cognitive control in the brain is centralized in the mPFC and according to the local hypothesis, cognitive control is performed in task-related local neural circuits. Using the place avoidance task which involves cognitive control, it is predicted that if mPFC lesions affect learning, this would support the central computation hypothesis whereas no effect of lesions would rather support the local hypothesis. The authors thus examine the effect of mPFC lesions in learning and retention of the place avoidance task. They also look at functional interconnectivity within a large network of areas that could be activated during the task by using cytochrome oxidase, a metabolic marker. In addition, electrophysiological unit recordings of CA1 hippocampal cells are made in a subset of (lesioned or intact) animals to evaluate overdispersion, a firing property that reflects cognitive control in the hippocampus. The results indicate that mPFC lesions do not impair place avoidance learning and retention (though flexibility is altered during conflict training), do not affect cognitive control seen in hippocampal place cell activity (alternation of frame-specific firing), a measure of location-specific firing variability, in pretraining. It nevertheless has some effect on functional interconnections. The results overall support the local hypothesis. 

Strengths:

Straightforward hypothesis: clarification of the involvement of the mPFC in the brain is expected and achieved. Appropriate use of fully mastered methods (behavioral task, electrophysiological recordings, measure of metabolic marker cytochrome oxidase) and rigorous analysis of the data. The conclusion is strongly supported by the data. 

Weaknesses:

No notable weaknesses in the conception, making of the study, and data analysis. The introduction does not mention important aspects of the work, i.e. cytochrome oxidase measure and electrophysiological recordings. The study is actually richer than expected from the introduction. 

The revised Introduction now includes:

“We used cytochrome oxidase, a metabolic marker of baseline neuronal activity, to confirm the mPFC lesions were effective and that there are non-local network consequences despite the local lesion. We first evaluated cytochrome oxidase activity in regions known to be associated with performance in the active place avoidance task, or regions with known connectivity to the mPFC. We then evaluated covariance of activity amongst the regions in an effort to detect network consequences of the lesion.”

Reviewer #2 (Public review): 

Park et al. set out to test two competing hypotheses about the role of the medial prefrontal cortex (PFC) in cognitive control, the ability to use task-relevant cues and ignore taskirrelevant cues to guide behavior. The "central computation" hypothesis assumes that cognitive control relies on computations performed by the PFC, which then interacts with other brain regions to accomplish the task. Alternatively, the "local computation" hypothesis suggests that computations necessary for cognitive control are carried out by other brain regions that have been shown to be essential for cognitive control tasks, such as the dorsal hippocampus and the thalamus. If the central computation hypothesis is correct, PFC lesions should disrupt cognitive control. Alternatively, if the local computation hypothesis is correct, cognitive control would be spared after PFC lesions. The task used to assess cognitive control is the active place avoidance task in which rats must avoid a section of a rotating arena using the stationary room cues and ignoring the local olfactory cues on the rotating platform. Performance on this task has previously been shown to be disrupted by hippocampal lesions and hippocampal ensembles dynamically represent the room and arena depending on the animal's proximity to the shock zone. They found no group (lesion vs. sham) differences in the three behavioral parameters tested: distance traveled, latency to enter the shock zone, and number of shock zone entries for both the standard task and the "conflict" task in which the shock zone was rotated by 180 degrees. The only significant difference was the savings index; the lesion group entered the new shock zone more often than the sham group during the first 5 minutes of the second conflict session. This deficit was interpreted as a cognitive flexibility deficit rather than a cognitive control failure. Next, the authors compared cytochrome oxidase activity between sham and lesion groups in 14 brain regions and found that only the amygdala showed significant elevation in the lesion vs. sham group. Pairwise correlation analysis revealed a striking difference between groups, with many correlations between regions lost in the lesion group (between reuniens and hippocampus, reuniens and amygdala and a correlation between dorsal CA1 and central amygdala that appeared in the lesion group and were absent in the sham group. Finally, the authors assessed dorsal hippocampal representations of the spatial frame (arena vs. room) and found no differences between lesion and sham groups. The only difference in hippocampal activity was reduced overdispersion in the lesion group compared to the sham group on the pretraining session only and this difference disappeared after the task began. Collectively, the authors interpret their findings as supporting the local computation hypothesis; computations necessary for cognitive control occur in brain regions other than the PFC. 

Strengths:

(1) The data were collected in a rigorous way with experimental blinding and appropriate statistical analyses. 

(2) Multiple approaches were used to assess differences between lesion and sham groups, including behavior, metabolic activity in multiple brain regions, and hippocampal singleunit recording. 

Weaknesses:

(1) Only male rats were used with no justification provided for excluding females from the sample.

This is a weakness we acknowledge. The experiments were performed at a time when we did not have female rats in the lab.

(2) The conceptual framework used to interpret the findings was to present two competing hypotheses with mutually exclusive predictions about the impact of PFC lesions on cognitive control. The authors then use mainly null findings as evidence in support of the local computation hypothesis. They acknowledge that some people may question the notion that the active place avoidance task indeed requires cognitive control, but then call the argument "circular" because PFC has to be involved in cognitive control. This assertion does not address the possibility that the active place avoidance task simply does not require cognitive control. 

We beg to differ that the possibility was not addressed. Prior to making the assertion, the manuscript describes the evidence that the active place avoidance task requires cognitive control. The evidence is multifold, and includes task design, behavior, and electrophysiology; we argue that this is more evidence than has been provided for other tasks that are asserted to require cognitive control. Specifically line 417 states:

“We have previously demonstrated cognitive control in the active place avoidance task variant we used (Fig. 1) because the rats must ignore local rotating place cues to avoid the stationary shock zone. Even when the arena does not rotate, rats distinctly learn to avoid the location of shock according to distal visual room cues and local olfactory arena cues, such that the distinct place memories can be independently manipulated using probe trials [49, 50]. When the arena rotates as in the present studies, neural manipulations that impair the place avoidance are no longer impairing when the irrelevant arena cues are hidden by shallow water [14, 15, 51, 52]. Furthermore, persistent hippocampal neural circuit changes caused by active place avoidance training are not detected when shallow water hides the irrelevant arena cues to reduce the cognitive control demand [10, 31, 33]. While these findings unequivocally demonstrate the salience of relevant stationary room cues to use for avoiding shock and irrelevant arena cues to ignore during active place avoidance, the most compelling evidence of cognitive control comes from recording hippocampal ensemble discharge. Hippocampal ensemble discharge purposefully represents current position using stationary room information when the subject is close to the stationary shock zone and alternatively represents rotating arena information when the mouse is far from the stationary shock zone [Fig. 4; 10].”

Line 436, however, acknowledges a fact that will always be true: no matter what anyone opines - until there are universally agreed upon objective criteria, it is logically possible that active place avoidance does not require cognitive control. The revision states: Despite this evidence from task design, behavioral observations, and direct electrophysiological representational switching as required to directly demonstrate cognitive control, one might still argue that it is logically possible that the active place avoidance task does not require cognitive control and this is why the mPFC lesion did not impair place avoidance of the initial shock zone. We consider such reasoning to be unproductive because it presumes that only tasks that require an intact mPFC can be cognitive control tasks. We nonetheless acknowledge that for some, we have not provided sufficient evidence that the active place avoidance requires cognitive control.

“We assert the evidence is compelling, and together these findings require rejecting the central-computation hypothesis that the mPFC is essential for the neural computations that are necessary for all cognitive control tasks.”

(3) The authors did not link the CO activity with the behavioral parameters even though the CO imaging was done on a subset of the animals that ran the behavioral task nor did they make any attempt to interpret these findings in light of the two competing hypotheses posed in the introduction. Moreover, the discussion lacks any mechanistic interpretations of the findings. For example, there are no attempts to explain why amygdala activity and its correlation with dCA1 activity might be higher in the PFC lesioned group. 

The CO study was performed to assess the effects of the lesion, as stated on line 262 “Cytochrome oxidase (CO), a sensitive metabolic marker for neuronal function [27], was used to evaluate whether lesion effects were restricted to the mPFC.” Furthermore, as a matter of fact, line 411 states “Thus, CO imaging and electrophysiological evidence identify changes in the brain beyond the directly damaged mPFC area. In particular, the dorsal hippocampus loses the inhibitory input from mPFC [45, 46] and loses the metabolic correlation with the nucleus reuniens, which is thought to be a relay between the mPFC and the dorsal hippocampus [47, 48].”

These CO measures assess baseline metabolic function and so it would be inappropriate to correlate them with the measures of behavior. Because the lesion and control groups do not differ on most measures of behavior, a relationship to CO measures is not expected. Importantly, even if there were differences in correlations between CO activity and behavioral measures, what could they mean? The study was designed to distinguish between two hypotheses, not to determine what CO differences could mean for behavior. As such, it is not at all clear how metabolic consequences of the lesion relate to the two hypotheses being evaluated, and so we consider it inappropriate to speculate. We did examine, and now include, the correlation between lesion size and conflict behavior. The Fig. 1 legend states “Savings was not related to lesion size r = 0.009, p = 0.98. *p < 0.05.”

(4) Publishing null results is important to avoid wasting animals, time, and money. This study's results will have a significant impact on how the field views the role of the PFC in cognitive control. Whether or not some people reject the notion that the active place avoidance task measures cognitive control, the findings are solid and can serve as a starting point for generating hypotheses about how brain networks change when deprived of PFC input. 

We thank the reviewer for the acknowledgement.

Reviewer #3 (Public review): 

Summary:

This study by Park and colleagues investigated how the medial prefrontal cortex (mPFC) influences behavior and hippocampal place cell activity during a two-frame active place avoidance task in rats. Rats learned to avoid the location of mild shock within a rotating arena, with the shock zone being defined relative to distal cues in the room. Permanent chemical lesions of the mPFC did not impair the ability to avoid the shock zone by using distal cues and ignoring proximal cues in the arena. In parallel, hippocampal place cells alternated between two spatial tuning patterns, one anchored to the distal cues and the other to the proximal cues, and this alteration was not affected by the mPFC lesion. Based on these findings, the authors argue that the mPFC is not essential for differentiating between task-relevant and irrelevant information. 

Strengths:

This study was built on substantial work by the Fenton lab that validated their two-frame active place avoidance task and provided sound theoretical and analytical foundations. Additionally, the effectiveness of mPFC lesions was validated by several measures, enabling the authors to base their argument on the lack of lesion effects on behavior and place cell dynamics. 

Weaknesses:

The authors define cognitive control as "the ability to judiciously use task-relevant information while ignoring salient concurrent information that is currently irrelevant for the task." (Lines 77-78). This definition is much simpler than the one by Miller and Cohen: "the ability to orchestrate thought and action in accordance with internal goals (Ref. 1)" and by Robbins: "processes necessary for optimal scheduling of complex sequence of behaviour." (Dalley et al., 2004, PMID: 15555683). Differentiating between task-relevant and irrelevant information is required in various behavioral tasks, such as differential learning, reversal learning, and set-shifting tasks. Previous rodent behavioral studies have shown that the integrity of the mPFC is necessary for set-shifting but not for differential or reversal learning (e.g., Enomoto et al., 2011, PMID: 21146155; Cho et al., 2015, PMID: 25754826). In the present task design, the initial training is a form of differential learning between proximal and distal cues, and the conflict training is akin to reversal learning. Therefore, the lack of lesion effects is somewhat expected. It would be interesting to test whether mPFC lesions impair set-shifting in their paradigm (e.g., the shock zone initially defined by distal cues and later by proximal cues). If the mPFC lesions do not impair this ability and associated hippocampal place dynamics, it will provide strong support for the authors' local computation hypothesis.

Thank you for these comments. In addressing them we have provided a significant revision to the manuscript’s Introduction. While authors like those cited by the reviewer have defined cognitive control, those definitions are difficult to test rigorously, as it is almost a matter of opinion whether a subject is displaying “the ability to orchestrate thought and action in accordance with internal goals" or whether they are using "processes necessary for optimal scheduling of complex sequence of behaviour." What would such definitions of cognitive control predict about neuronal activity? We have deliberately used a simple, operational definition of cognitive control because it is physiologically testable. In the revision, starting at line 93, we have provided an excerpt from Miller and Cohen (2001) with discussion. The importance of that work is that it provides explicit neuronal criteria and a means to operationally define cognitive control. As stated on Line 118 “Accordingly, cognitive control would be at work when there is sustained neuronal network representations of task-relevant information that suppresses or gates representations of salient task-irrelevant information in accord with purposeful judicious behavior.”

We used a R+A- task variant in which there is a stationary room-frame shock zone and task irrelevant arena-frame information. A strict correspondence to shift-shifting task design cannot be accomplished with active place avoidance because an A+R- task that requires avoiding an arena-frame shock zone in the absence of a room-frame shock zone can be accomplished trivially if the subject chooses to not move when it is in a place with no shock. However, the R+A+ task variant is readily learned, in which there is both a room-frame and an arena-frame shock zone (see cited work below). This task variant requires the subject to judiciously shift between avoiding the room-frame shock zone using stationary room information and avoiding the arena-frame shock zone using rotating arena information. This R+A+ task variant might meet the reviewer’s criteria for cognitive control. We have recorded hippocampal and entorhinal ensemble activity during the R+A+ task variant and it is very similar to the activity during the R+A- task we used. Nonetheless, future work will investigate the efect of mPFC lesion on the R+A+ task variant.

Cited work:

Fenton AA, Wesierska M, Kaminsky Y, Bures J (1998), Both here and there: simultaneous expression of autonomous spatial memories in rats. Proc Natl Acad Sci U S A 95:11493-11498. Kelemen E, Fenton AA (2010), Dynamic grouping of hippocampal neural activity during cognitive control of two spatial frames. PLoS Biol 8:e1000403.

Burghardt NS, Park EH, Hen R, Fenton AA (2012), Adult-born hippocampal neurons promote cognitive flexibility in mice. Hippocampus 22:1795-1808.

Park EH, Keeley S, Savin C, Ranck JB, Jr., Fenton AA (2019), How the Internally Organized Direction Sense Is Used to Navigate. Neuron 101:1-9.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors): 

(1) Incorporate the cytochrome oxidase and hippocampal recordings (rationale and hypothesis) in the introduction, explaining how these aspects are relevant to the general question. 

We have done this as requested. See lines 159-173 of the revised introduction.

(2) Figure 1C. On Day 4-5 (conflict training) in which the shock zone was relocated 180 deg from the initial location, the behavioral tracks did not show any presence of the rat in this sector (in particular for the lesion example). Figure 4 nevertheless indicates that entrances have been made (which was expected since rats have to know that the shock zone was relocated).

Thanks for pointing this out. The tracks are from the end of the sessions. The labels have been changed to specify which trials the tracks are from.

(3) Figure 1C. The caption is huge as it contains the statistical analyses details. I would prefer to have these details in the text and keep the caption at a "reasonable" length. At the end of the caption (l. 190-191), it would be less confusing the keep the numbering of the training days: replace D1T1 with D2T1 and D2T9 with D3T9).

The statistical details have been relocated to the main text and the numbering updated, as suggested, thank you.

(4) It was not inconsiderable to show that mPFC lesion had some effects in the present task if it were only to validate the effectiveness of the lesion. This brain area has been shown to be important for planning, cognitive flexibility, etc. Indeed the authors found that the saving index was greater in sham than in mPFC rats (overdispersion in hippocampal firing was also reduced in pretraining) and interpreted this result as impaired flexibility. Would an alternative explanation be a memory deficit? I nevertheless expected that impaired flexibility in mPFC rats would be expressed in conflict trials in the form of more entrances in the zone that was initially not associated with shock (at least in the first trials of Day 4). But it appears to not be the case.

A memory deficit is unlikely to explain the difference between the groups on the first trial of Day 5. Memory in the lesion rats was tested multiple times, specifically at the start of each trial (time to first entrance), including on the 24-h retention test, and no deficits were observed. Performance on Day 9 trial 1 is worse in the lesion group than in the controls, but it is not parsimonious to attribute this to a simple memory deficit since 24-h memory was good and similar between lesion and control rats on days 3 and 4, and memory on Day 5 was equally poor in both the lesion and control rats, as measured by time to first entrance.  

(5) Material and methods. The injected volume of ibotenic acid should be mentioned. 

The volume 0.2 µl was added. See line 531.

(6) The rationale for doing the conflict training session should be indicated somewhere. 

The rationale was provided. See lines 204-208.

Reviewer #2 (Recommendations for the authors): 

(1) Line 132: The text states that all sham rats improved and only 6/10 lesion rats improved is followed by a t-test, which tests the difference between means; it does not compare proportions. Also, what criterion was used to determine if an improvement was seen or not? 

The statistical comparison is provided (now lines 230: test of proportions z = 2.3, p = 0.03). Improvement was simply numerically fewer entrances.

(2) Line 138: This is a very long and confusing sentence. Consider revising for clarity. 

The sentence (now line 234) was revised.

(3) Figure 1B only includes data from 3 animals. Most published studies show the whole dataset by presenting the largest and smallest lesions. 

Supplemental Figure S2 was added with all the lesions depicted and quantified.

(4) Figure 1C suggestion to make the schematic shock zone line up with the shock zone shown for the tracking data. 

Graphically, it looks better as drawn as it uses to perspective to depict a three-dimensional structure.

(5) Methods: Clarify if the shock zone location was the same across all rats. 

Line 570 states that the shock zone was the same for all rats.

(6) Line 158: "Behavioral tracks" is not clear. Suggest more precise wording.

Reworded to “Tracked room-frame positions” (now line 249)

(7) Line 166: "effect of trial" - should this be the main effect of trial?; "interaction" - should this be "group x trial" interaction? 

Reworded (now line 181).

(8) Line 167: "or their interaction" is awkward in the context of the sentence. 

Reworded (now line 182).

(9) Line 182: Avoid talking about "trends" as if they are almost significant unless the authors suspect that they did not have sufficient statistical power to detect differences. In that case, a power analysis should be provided. 

Removed.

(10) Line 190: "left:...right..." is hard to follow, especially with acronyms like D1T1. Consider revising for clarity. 

Revised (now lines 246-248).

(11) Line 195: "effectiveness of the PFC to impair" is unnecessarily verbose. 

Reworded (now lines 255-257).

(12) Savings results: There is a lot of variability in the lesion group. It would be interesting to know if the extent of the lesion correlates with savings.

Savings was not related to lesion. See line 259.

(13) Line 300: The thalamic recording results are not reported in the results section (other than appearing in the table). Moreover, there is no detail about which thalamic nucleus these recordings are from.

Lines 411 and 614 provides these details.  

(14) Line 312: "no longer impair" contains a grammatical error. 

Corrected (now line 422)

(15) Line 325: "was not impairing" contains a grammatical error. 

Corrected (now line 437).

(16) Line 327: The sentence ending with "...opinion of others" seems unnecessarily confrontational. 

Previous reviewers at other journals have maintained this position, we therefore included such a strong statement in our initial submission. However, we now revised this statement to avoid appearing confrontational.

(17) Line 329: Sentence is awkward. Consider revising. 

Revised (now line 443).

(18) Line 384: The authors should disclose if there was an objective metric for determining the adequacy of the lesion. 

The lesion assessment and quantification is better explained in the Methods under “Cytochrome oxidase activity and Nissl staining,” (lines 708-714).

(19) Line 385: The authors should clarify how they got from 15 rats (Line 376) to 10. 

This information is provided in the methods.

(20) Line 390: It is not clear why skin irritation in the cage mate would prevent the rat from being tested. 

This has been explained in the Methods under “Behavioral analysis followed by cytochrome oxidase activity” (lines 515-518).

(21) Methods section: The authors should describe how the tracking data were acquired. Overhead camera? Tracker based on luminance or body position? What software program was used? What was the sampling rate? 

This is now better explained in the Methods under “Active place avoidance task) (lines 538551).

(22) Methods section: Include how fast the arena was rotating and other details about the task such as where rats were placed during the ITI. 

Better explained in the Methods under “Active place avoidance task”.

(23) Line 439: The recording system used (hardware & software) should be stated. 

This is now included in the Methods (line 538).

(24) Line 435: Though overdispersion calculation is described thoroughly, there is nothing in the paper that tells me what overdispersion means. 

What the measure means is now described in the Methods under “Electrophysiology data analysis” (lines 646-650).

(25) Line 561: The test used to assess effect sizes should be stated. 

Effect sizes corresponding to the statistical tests are provided.

Reviewer #3 (Recommendations for the authors): 

(1) At the end of the conflict training, rats with mPFC lesions learned to avoid the new shock zone (Figure 1F, Block 16), but their place cells did not show room-preferring activity near the shock zone (Figure 4B). This observation questions whether spatial frame-specific representation is relevant for active avoidance. Can the authors clarify this point?

This is a dynamic behavior and the hippocampal dynamics match, changing with a dynamic that is a few seconds, as we have shown in several published papers. The lack of a preference averaged over 20 minutes when the rats are avoiding both the current and former shock zones during the conflict session is pretty much what would be expected from such a coarse measurement. The important measure is the spatially-resolved measure of room versus arena preference. Figure 4B shows that in the lesion rats there is less of a frame preference during conflict, generally (consistent with poorer flexibility). However, Figure 4D quantifies the frame preference near and far from the shock zone and accordingly, there is no difference between the groups.

(2) Related to the point above, the author might consider including panels in Figures 4C and D to show the neural activity during the pretraining and conflict training retention period. I assume p(room) will be comparable between the Near and Far segment in both sessions, but the p(room) may be higher in the Conflict training session than the Pretraining session. This would show that the mPFC lesion impairs suppressing the place cell activity encoding the old shock location. 

Thanks for the suggestion. While we don’t think we can draw any strong conclusions from this analysis we are fine to show it. The issue is that during conflict, the rats have two perfectly reasonable representations of where there was shock, the initial location that was turned off to make the conflict, and the most recent conflict location of shock. Importantly, these recordings are during conflict retention after we turned off the shock for the retention recording (for the second time in the rat’s experience). Turning off the shock allows us to exactly match the physical conditions of pretraining, initial retention and conflict retention, which was the experimental design’s goal. However, the experiential history of the rats prior to initial retention and conflict retention cannot match, because during initial retention the rats had never experienced a changed shock zone whereas, by conflict retention, they had experienced multiple changes. Importantly, we have previously shown that mouse hippocampal ensembles represent both initial and conflict shock locations, as the animals consider their options during conflict trials (see Dvorak et al 2018, PLoS Biol 16:e2003354). Consequently, we cannot make any strong predictions about whether or not hippocampal activity during conflict retention should be room-frame preferring selectively in the vicinity of the current shock zone. As I am sure the reviewer appreciates from their own introspection, mental representations are mercifully not obliged to dictate behavior. In fact, that is what is interesting and controversial about cognitive control – it is a dynamic internal process and the innovation of our work lies in demonstrating that one cannot only rely on behavior to assess this process. Nonetheless, we did this analysis and now present it in the revised Fig. 4. During pretraining both lesion and sham groups express no particular spatially-modulated preference for either the room or the arena frame, as expected. During initial training both groups express a room-frame preference in the vicinity of the shock zone, as we initially reported. By inspection, during conflict, the sham rats express a preference for room-frame activity in the vicinity of the most recent shock zone location; this preference is weaker than what is expressed during initial retention. The lesion rats do not show this preference. These impressions are quantified in revised Fig. 4D; the comparisons within the conflict retention sessions did not reach statistical significance. We leave it to the reader to interpret what that means. Thanks for the nudge.

(3) The significant group difference in place cell overdispersion during the pretraining phase (Figure 3C) is interesting, but some readers would appreciate additional sentences on its functional implication. Does it mean the spatial tuning of place cells was disrupted by the mPFC lesion?

Only the reliability of spatial firing was altered, not the spatial tuning.

(4) Although the method section described how to calculate overdispersion and SFEP, some concise, intuitive descriptions of these measures in the result section would help readers understand these results.

Overdispersion is better explained. See lines 646-650.

(5) I recommend adding a figure of the task performance of the rats used in the electrophysiological recording experiment and a table summarizing the number of cells recorded per animal. 

We have included Table S2 with the cell counts and a summary of the performance for each of the rat in the electrophysiological recording experiment.

(6) Readers would appreciate additional information on task apparatus, such as the size, appearance, and rotating speed of the arena, as well as stationary cues available in the room. 

This is now provided in the Methods under “Active place avoidance task”.

(7) Lines 425-416: "On the fourth day of the behavioral training, the rats had a single trial with the shock on to test retention of the training." Shouldn't it be "shock off"? 

No the shock was on to prevent extinction learning and to increase the challenge for conflict learning.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1:

Major Concerns/Public Review

Comment 1: There is a mild disconnect between behavioral readout (reflexive pain) and neural circuits of interest (emotional). Considering that this circuit is likely engaged in the aversiveness of pain, it would have been interesting to see how carrageenan and/or AIE impacted non-reflexive pain measures. Perhaps this would reveal a potentiated or dysregulated phenotype that matches the neurophysiological changes reported. However, this critique does not take away from the value of the paper or its conclusions.

We agree that including measures of non-reflexive pain would enhance future studies and potentially reveal a phenotype that is closely related to the observed changes in neurophysiology.

Minor Concerns/Recommendations

Comment 1: There are a few minor grammatical errors in the text, mostly in the captions. A close read should be able to identify these errors.

We have fixed what grammatical errors we found.

Reviewer #2:

Major Concerns/Public Review

No major concerns.

Minor Concerns/Recommendations

Comment 1: If pain sensitivity was assessed at 3 time points post carrageenan administration, why were these data averaged? Were there no differences between the time points? The data from the 3 time points should be presented, either in a figure, table, or supplementary materials.

We averaged the pain sensitivity data across the 3 time points following carrageenan administration because we were trying to present this data in a more concise manner. Pain sensitivity did change over time following carrageenan administration. We have now included the unaveraged data in figure 2 (panels D, F, H, and J).

Comment 2: For the optically-evoked EPSCs and IPSCs, were the peak amplitudes the max responses that could be obtained? If not, how were levels of ChR2 expression or light intensity controlled for?

The peak amplitudes for EPSCs and IPSCs were half the maximal response that could be evoked by optical stimulation. The AMPA and NMDA currents were maximal responses as prior literature indicated some PVINs have small NMDA currents, and we wanted to ensure these currents would be detected reliably. We updated our methods section to include this information in the voltage clamp recordings section.

Comment 3: In the example traces for the aEPSC experiment, the figure legend states that the "+" symbol indicates an asynchronous event. However, there are several "|" or "-" symbols in the figure. Perhaps this is an issue with the resolution of the figure and those are supposed to be "+"s.

We have increased the resolution of the figures to ensure that the markings of the asynchronous events display properly. We apologize for not noticing that these symbols were not displayed correctly in the original figures included in the manuscript.

Comment 4: For the von Frey and the Hargreaves test, were animals acclimated to the apparatus in the days leading up to the first test, or was the 5-minute pre-test the only acclimation that was done? This information needs to be provided. If the latter, there is concern that the animals did not fully acclimate to the apparatus and handling prior to testing, which should be taken into consideration in the interpretation of the behavioral analyses.

The rats underwent handling once a day for three days prior to the first von Frey and Hargreaves tests. On the day prior to the first test, rats were acclimated to the von Frey and Hargreaves apparatuses. The acclimation period consisted of a 15-min exposure to the von Frey apparatus and a 30-min exposure to the Hargreaves apparatus for each animal. This information has been added to the revised methods section under the assessment of mechanical and thermal sensitivity heading.

Reviewer #3:

Major Concerns/Public Review

Comment 1: There is incomplete evidence supporting some of the conclusions drawn in this manuscript. The authors claim that the changes in feedforward inhibition onto pyramidal cells are due to the changes in parvalbumin interneurons, but evidence is not provided to support that idea. PV cells do not spontaneously fire action potentials spontaneously in slices (nor do they receive high levels of BLA activity while at rest in slices). It is possible that spontaneous GABA release from PV cells is increased after AIE but the authors did not report sIPSC frequency. Second, the authors did not determine that PV cells mediate the feedforward BLA op-IPSCs and changes following AIE (this would require manipulation to reduce/block PV-IN activity). This limitation in results and interpretation is important because prior work shows BLA-PFC feedforward IPSCs can be driven by somatostatin cells. Cholecystokinin cells are also abundant basket cells in PFC and have been recently shown to mediate feedforward inhibition from the thalamus and ventral hippocampus, so it's also possible that CCK cells are involved in the effects observed here.

The hypothesis that adolescent alcohol exposure could change spontaneous GABA release from PVINs is an interesting one that merits future exploration. Unfortunately, as the focus of this manuscript was on circuit-specific alterations in synaptic function, this experiment is somewhat outside the scope of the paper as sIPSCs and mIPSCs are not circuit specific measures of GABA activity and would not reflect spontaneous release from only GABA interneurons receiving input from the BLA. Despite this, a future study investigating spontaneous GABA release from PVINs in the PrL would be a valuable complement to the present study.

While we did not directly manipulate PVINs to demonstrate that decreased oIPSC amplitude at PrL^PAG neurons following AIE is due solely to changes in PVINs, it is notable that both the intrinsic excitability of PVINs and the BLA-driven E/I balance at PVINs were reduced following AIE. These changes would be consistent with decreased PVIN output onto PrL^PAG neurons. However, we agree that this does not preclude the possibility that changes in SST or CCK interneurons contribute to the observed decrease in BLA-driven inhibition at PrL^PAG neurons following AIE. As such, we have altered the wording in the discussion to indicate that reduced BLA-driven feedforward inhibition of PrL^PAG neurons may be related, at least in part, to the observed changes in PVINs.

Comment 2: The authors conclude that the changes in this circuit likely mediate long-lasting hyperalgesia, but this is not addressed experimentally. In some ways, the focused nature of the study is a benefit in this regard, as there is extensive prior literature linking this circuit with pain behaviors in alternative models (e.g., SNI), but it should be noted that these studies have not assessed hyperalgesia stemming from prior alcohol exposure. While the current studies do not include a causative behavioral manipulation, the strength of the association between BLA-PL-PAG function and hyperalgesia could be bolstered by current data if there were relationships detected between electrophysiological properties and hyperalgesia. Have the authors assessed this? In addition, this study is limited by not addressing the specificity of synaptic adaptations to the BLA-PL-PAG circuit. For instance, PL neurons send reciprocal projections to BLA and send direct projections to the locus coeruleus (which the authors note is an important downstream node of the PAG for regulating pain).

We have not assessed correlations between the electrophysiological properties and hyperalgesia. We feel that future studies using DREADDs to perform cell-type and circuit-specific manipulations can better address the involvement of this circuitry in long-lasting hyperalgesia following AIE. With respect to the circuit specificity of the observed changes, we have previously evaluated the effects of AIE on pyramidal neurons projecting from the PrL to the BLA (PrL^BLA). We found that following AIE exposure there was no change in the intrinsic excitability of these neurons. In addition, the amplitude and frequency of sEPSCs and sIPSCs onto PrL^BLA neurons was unchanged. While these results did not assess whether the BLA-PrL-BLA circuit undergoes synaptic adaptations similar to those observed in the BLA-PrL-vlPAG circuit, it is notable that the intrinsic excitability of PrL^BLA neurons was unchanged following AIE exposure. This indicates that the effects of AIE on the intrinsic excitability of pyramidal neurons in the PrL may be circuit specific. We agree that it would be interesting to study the effect of AIE on PrL neurons that project to the locus coeruleus, however due to the well-defined role of the BLA-PrL-vlPAG circuit in pain we chose to evaluate this circuit first.

Comment 3: I have some concerns about methodology. First, 5-ms is a long light pulse for optogenetics and might induce action-potential independent release. Does TTX alone block op-EPSCs under these conditions? Second, PV cells express a high degree of calcium-permeable AMPA receptors, which display inward rectification at positive holding potentials due to blockade from intracellular polyamines. Typically, this is controlled/promoted by including spermine in the internal solution, but I do not believe the authors did that. Nonetheless, the relatively low A/N ratios for this cell type suggest that CP-AMPA receptors were not sampled with the +40/+40 design of this experiment, raising concerns that the majority of AMPA receptors in these cells were not sampled during this experiment. Finally, it should be noted that asEPSC frequency can also reflect changes in a number of functional/detectable synapses. This measurement is also fairly susceptible to differences in inter-animal differences in ChR2 expression. There are other techniques for assessing presynaptic release probability (e.g., PPR, MK-801 sensitivity) that would improve the interpretation of these studies if that is intended to be a point of emphasis.

When we included TTX but not 4-AP we did not observe any optically evoked responses, so we don’t believe that the 5-ms pulse induced action-potential independent release in these experiments. With respect to the second point, we did not include spermine in the internal solution for the AMPA/NMDA recordings in PVINs, and it is possible that endogenous polyamines interfered with recording CP-AMPA receptors in the +40/+40 design. To address this concern, we recalculated the AMPA/NMDA ratio for PVINs using data from an optically evoked AMPA current that was collected while holding the cell at -70 mV. This data was collected at the end of the +40/+40 recording protocol as we were interested in assessing whether there would be any difference in the ratio of the +40/-70 AMPA current across treatment conditions. As there were no observed difference in the +40/-70 AMPA current ratio across treatment groups, we had originally used the +40 AMPA current for calculating the AMPA/NMDA ratio for PVINs to make the methods for calculating this ratio uniform for both PVINs and PrL^PAG neurons. The methods, results, and Fig. 10 have been updated to reflect the recalculated AMPA/NMDA ratio for PVINs. Notably, only the significance of the AIE x carrageenan interaction was altered by the change in the way the AMPA/NMDA ratio was calculated. Originally, this interaction displayed a trend toward significance (p = 0.0501), however when the recalculated AMPA/NMDA ratio was analyzed this interaction term became significant (p = 0.0131). We have also added the +40/-70 AMPA ratio to figure 10 as it might be of interest.

Finally, the point regarding aEPSC frequency reflecting not only release probability but also the number of functional/detectable synapses is an important consideration. For this manuscript, we intentionally selected aEPSC frequency for this reason. As the BLA to PrL projection continues to mature during adolescence, the number of BLA contacts onto GABA neurons in the PrL increases. Thus, we thought that it was possible that AIE would alter the number of detectable BLA inputs onto PVINs. We acknowledge that as this measure is sensitive to differences in ChR2 expression between animals/slices it can be difficult to interpret. We also agree that in the future it would be beneficial to include either PPR or MK-801 sensitivity to improve interpretability.

Comment 4: In a few places in the manuscript, results following voluntary drinking experiments (especially Salling et al. and Sicher et al.) are discussed without clear distinction from prior work in vapor models of dependence.

We have altered the manuscript to specifically note where voluntary drinking was used rather than vapor models.

Comment 5: Discussion (lines 416-420). The authors describe some differing results with the literature and mention that the maximum current injection might be a factor. To me, this does not seem like the most important factor and potentially undercuts the relevance of the findings. Are the cells undergoing a depolarization block? Did the authors observe any changes in the rheobase or AP threshold? On the other hand, a more likely difference between this and previous work is that the proportion of PAG-projecting cells is relatively low, so previous work in L5 likely sampled many types of pyramidal cells that project to other areas. This is a key example where additional studies by the current group assessing a distinct or parallel set of pyramidal cells would aid in the interpretation of these results and help to place them within the existing literature. Along these lines, PAG-projecting neurons are Type A cells with significant hyperpolarization sag. Previous studies showed that adolescent binge drinking stunts the development of HCN channel function and ensuing hyperpolarization sag. Have the authors observed this in PAG-projecting cells? Another interesting membrane property worth exploring with the existing data set is the afterhyperpolarization / SK channel function.

In discussing the maximum current injection as a factor in differing results on intrinsic excitability, we were principally considering how the additional data points increase the power of the analysis and thus the likelihood of detecting an effect. In focusing on this, however, we ignored other relevant and interesting factors that we should also have discussed. Additional analyses examining HCN and SK channel function have now been added to the manuscript and incorporated into the results section under the heading Adolescent Intermittent Ethanol Exposure and Carrageenan Enhanced the Intrinsic Excitability of Prelimbic Neurons Projecting to the Ventrolateral Periaqueductal Gray. We have also modified the third paragraph in the discussion to add additional context. Additional information on the biophysical properties of the neurons has been added to Figure 4.

Minor Concerns/Recommendations

Comment 1: Subheadings are vague. "Analysis of..." Should be rephrased to use active voice to describe key findings.

The subheadings have been rephrased to describe key findings.

Comment 2: Consider altering or consolidating the figure layout for clarity. For instance, it would be helpful for aEPSCs to be near the AMPA and NMDA experiments. The feedforward IPSCs could also be with the PV-IN recordings. This would be helpful in developing a cohesive picture of key findings. To that end, a working model or graphical abstract would be helpful.

It doesn’t appear that this journal allows graphical abstracts, but we have added a model that summarizes the principal findings in the discussion.

Comment 3: There are a lot of statistics punctuating the text in the Results. It can be hard to parse at times.

We considered moving the statistics to tables, but this became unwieldy.

Comment 4: The Discussion is quite long (10 paragraphs). Suggest consolidating to 3-4 most salient points.

We appreciate this comment and have made some edits to the discussion, albeit without consolidating it to only 3-4 points.

Author response:

eLife Assessment

The authors provide a useful summary of ten years of Brain Initiative funding including the historical development, the specific funding mechanisms, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed. The evidence for impact is incomplete due to the omission of a comparison group of funded grants.

In this combined version, we include a comparison group of non-BRAIN Initiative R01s derived from the parent notice of funding opportunity from FY2014-2022. We performed a bibliometric analysis of the publications, citations, RCR and budget productivity measure of the non-BRAIN parent R01. 

Public Reviews:

Reviewer #1 (Public review):

Summary:

This is a convincing description of approximately ten years of funding from the NIH BRAIN initiative. It is of particular value at this moment in history, given the cataclysmic changes in the US government structure and function occurring in early 2025.

Strengths:

The paper contains a fair bit of documentation so that the curious reader can actually parse what this BRAIN program funded.

Weaknesses:

There are too many acronyms, and the manuscript reads as if it were an internal NIH document, where the audience knows all of the NIH nomenclature and program details. It is not particularly friendly to the outside, lay reader.

In this version, we have attempted to minimize acronyms and explain NIH nomenclature and program details to make it more accessible to readers not familiar with NIH terminology.

Reviewer #2 (Public review):

Summary:

The authors provide an important summary of ten years of Brain Initiative funding including a description of the historical development of the initiative, the specific funding mechanisms utilized, and examples of grants funded and work produced. The authors also conduct analyses of the impact on overall funding in Systems and Computational Neuroscience, the raw and field normalized bibliographic impact of the work, the social media impact of the funded work, and the popularity of some tools developed.

Strengths:

This is a useful perspective on an important funding initiative over a ten-year period. It is clearly written and the illustrations and analyses are mostly useful for understanding the impact of the initiative.

Weaknesses:

The major limitation is that the bibliographic analysis does not provide a comparison group of funded grants. Because work that successfully competes for funding is likely to be more impactful than all work in a given area, the normalization of citations to field medians may reflect this "grant review" effect, rather than anything special about the Brain Initiative. Hopefully, this speculation is incorrect (I would guess that it is), but it would be helpful to try to demonstrate this more directly by including a funded comparison group.

In this version, we have provided a comparison group of parent R01s that are not funded through the BRAIN Initiative from FY2014-2022 in Figure 3. We include publication metrics and budget efficiency measures for this comparison group.  

There are also minor inconsistencies in the numbering of the figures that need to be cleared up.

We have updated the figure numbers.

Author response:

eLife Assessment 

The manuscript presents some useful accounts of experiences funding team projects within the BRAIN Initiative. These would be more appropriate to add to the companion manuscript since the present manuscript contains some overlapping analyses and does not stand well on its own. Therefore the evidence supporting the conclusions is incomplete. 

We appreciate the feedback on merging both manuscripts into one and have followed the advice in this version. 

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

In this useful narrative, the authors attempt to capture their experience of the success of team projects for the scientific community.  

Strengths: 

The authors are able to draw on a wealth of real-life experience reviewing, funding, and administering large team projects, and assessing how well they achieve their goals. 

Weaknesses: 

The utility of the RCR as a measure is questionable. I am not sure if this really makes the case for the success of these projects. The conclusions do not depend on Figure 1. 

We respectfully disagree about the utility of the RCR, particularly because it is metric that is normalized by both year and topical area. We have added a more detailed description of how the RCR is calculated on page 6-7. Please note that figure 1 is aimed to highlight the funding opportunities, investments and number of awards associated with small lab (exploratory) versus team (elaborated, mature) research rather than a description of publication metrics.  

Reviewer #2 (Public review): 

Summary: 

The authors review the history of the team projects within the Brain initiative and analyze their success in progression to additional rounds of funding and their bibliographic impact. 

Strengths: 

The history of the team projects and the fact that many had renewed funding and produced impactful papers is well documented. 

Weaknesses: 

The core bibliographic and funding impact results have largely been reported in the companion manuscript and so represent "double dipping" I presume the slight disagreement in the number of grants (by one) represents a single grant that was not deemed to address systems/computational neuroscience. The single figure is relatively uninformative. The domains of study are sufficiently large and overlapping that there seems to be little information gained from the graphic and the Sankey plot could be simply summarized by rates of competing success. 

While we sincerely appreciate the feedback, we chose to retain these plots on domains and models to provide a sense of the broad spectrum of research topics contained in our TeamBCP awards. Further details on the awards can be derived from the award links provided in the text. Additionally, we retained the Sankey plots because these are a visual depiction of how awards transition from one mechanism to another, evolve in their funding sources, and advance in their research trajectories. The plot is an example of our continuity analysis which is only reported in the text and not visually shown for the remaining BCP programs.

Author response:

We thank the reviewers for their detailed and constructive comments on our manuscript entitled “Activity-Dependent Changes in Ion Channel Voltage-Dependence Influence the Activity Patterns Targeted by Neurons.” We appreciate the time and effort the reviewers invested in critiquing our work and are grateful for the opportunity to clarify and improve our manuscript.

As noted by the reviewers, the main message of the manuscript is that the intrinsic properties and activity characteristics of targeted bursters depend on the timescale of half-(in)activation alterations in the homeostatic mechanism. However, the concerns of the reviewers reveal that the manuscript is organized in ways that detract from this message. Below we respond to the points the reviewers raise and close by outlining the changes that we will make to the manuscript as a result. Our goal will be to streamline the message of the paper while addressing the concerns of the reviewers.

Response to Reviewer #1:

Point 1: We interpret the reviewer’s question about “mechanism” to be: why do half-(in)activation alterations redirect degenerate bursters to different parameter regions? (A separate aspect of “mechanism,” namely how these alterations might be biologically implemented, is already addressed in the paper.)

We speculate that Figure 3 illustrates this process. As conductance densities slowly evolve, rapid half-(in)activation changes cause the sensor variable (α) to jump abruptly as it searches for a voltage-dependence configuration that meets calcium targets (Figure 3A). The channel densities are slightly altered and this process continues again. Slowing the half-(in)activations alterations reduces these abrupt fluctuations (Figure 3B). Making the alterations infinitely slow effectively removes half-(in)activation changes altogether, leaving the system reliant solely on slower alterations in maximal conductances (Figure 3C). Because each timescale of half-(in)activation produces a different channel repertoire at each time step, the neuron follows distinct trajectories through the space of activity characteristics and intrinsic properties over the long term.

Point 2: We appreciate the reviewer’s skepticism regarding our statistical approach with the “Group of 5” and “Group of 20.” These groups arose from historical aspects of our analysis and this analysis does not directly advance the main point—that changes in the timescale of channel voltage-dependence alterations impact the properties of bursters to which the homeostatic mechanism converges. Therefore, we plan to remove the references to the Group of 5 and focus on how the Group of 20 responds to variations in the timescale of voltage-dependent alterations.

Point 3: Our paper claims that the half-(in)activation mechanism is subordinate to the maximal conductance mechanism. We agree with the reviewer that making this claim requires more care. The simulations we run are controls in the spirit described below.

The reviewer notes that in our simulations, half-(in)activations are already near the range required for bursting, which forces maximal conductances to undergo larger changes and thus appear more critical. We however note that the opposite can also occur: if half-(in)activation values were already positioned in ranges required for bursting, an arrangement of small maximal conductances may potentially produce bursting. The latter might give the impression that maximal conductance alterations and half-(in)activation alterations are equally important. The simulations we ran are simply suggested this wasn’t true for these models.

Points 4 - 6: In Point 4, the reviewer highlights model choices (e.g., constraints on maximal conductance and half-(in)activation, use of the L8 norm) are not clearly justified. In Point 5, the reviewer suggests that the paper provides excessive detail about other model choices. Point 6 appears to reiterate concerns about insufficient justification for some modeling decisions.

Our intent was to acknowledge every caveat, which led us to include long section on Model Assumptions in the Discussion. However, as Point 5 notes, this makes the Discussion cumbersome. The Discussion should focus on remarks regarding the impact that timescale of half-(in)activation alterations have on the family of bursters targeted by the homeostatic mechanism. Consequently, we will relocate the extended discussion of model assumptions from the Discussion to the Methods section. This section already touches on how the constraints on half-(in)activation alterations compare to earlier versions of the model (noted in Point 6) and will be expanded to further explain our choice of the L8 norm (Point 4).

Response to Reviewer #2:

Weakness 1: The reviewer notes that the writing is “rather confusing.” This likely arises from the fact that we did not consistently emphasize the core message: the timescale of half-(in)activation alterations influences the intrinsic properties and activity characteristics of bursters targeted by the homeostatic mechanism. We will address this by reorganizing the manuscript to make that focus clearer, and we outline these planned revisions at the end of these responses.

The reviewer specifically points out that the state-of-the-art is not clearly articulated. We will reorganize the Introduction to highlight this. Briefly, work on activity-dependent homeostasis has historically focused on changes in channel density. This is supported by experiment and has been modelled theoretically. In comparison, changes in channel voltage-dependence, while documented, are less explored due to the challenges of measuring them. In this work, we attempt to study the impact that alterations in channel voltage-dependence have on activity-dependent homeostasis. To do this, we extend existing computational models of activity-dependent homeostasis—models that have hitherto only altered channel density—by incorporating a mechanism that also adjusts channel voltage-dependence.

Weakness 2: The Discussion highlights two potential implications of our findings—one for neuronal development and another for activity recovery following perturbations. However, they were outlined after the Model Assumptions section which, as Reviewer 1 points out, is quite detailed and cumbersome.

Another aspect that may contribute to the challenge in interpreting our results may be our conceptual approach to neuronal excitability, which relies on a computational model of activity-dependent homeostasis that abstracts much of the underlying biochemistry. Our message is general: the timescale of half-(in)activation alterations influences the intrinsic properties and activity characteristics of bursters targeted by a homeostatic mechanism. As such, the implications are general. Their value lies in circumscribing a conceptual framework from which experimentalists may devise and test new hypotheses. We do not aim to predict or explain any specific phenomenon in this work. To address this concern however, we will expand our discussion of how these findings may guide experimental considerations, particularly regarding neuronal development and activity recovery during perturbations, to better illustrate the practical utility of our results.

Response to Reviewer #3:

Point 1: This reviewer suggests that our core message—namely, that the timescale of half-(in)activation alterations affects the intrinsic properties and activity patterns targeted by a homeostatic mechanism—should also apply during perturbations. We plan to address this by extending our analysis on the Group of 20 models. We will perturb activity by increasing extracellular potassium concentration and change the timescale of half-(in)activation alterations during the perturbation. This should underscore how the neuron’s stabilized activity pattern depends on this timescale, reinforcing our central message.

Point 2: In this part of the Discussion, we noted that multiple half-activation shifts collectively shape the neuron’s global properties, and that averaging might obscure these effects. However, in light of the reviewers’ comments, we recognize that this observation alone does not directly advance the paper’s main message. To make it relevant, we would need to (1) identify correlations between intrinsic parameters (i.e., half-(in)activation and maximal conductance) and the resulting activity patterns, and (2) examine how these correlations shift under different timescales half-(in)activation alterations. Since we have not performed that analysis, we will revise this part of the Discussion to clarify its connection to the paper’s principal focus by noting that a deeper exploration of this notion using correlations will be the topic of future work.

Conclusion: We outline updates we will make to the paper here.

Introduction: In response to Reviewer 2, we will provide a clearer explanation of the state-of-the-art in activity-dependent homeostasis and highlight our specific contribution. We will emphasize that our conclusions, while generic, are relevant in experimental contexts.

Results: We will reorganize this section to underscore the main point: the timescale of half-(in)activation alterations affects the intrinsic properties and activity characteristics of bursters in the homeostatic mechanism. Figures 1 will remain as it is. It shows assembly from random initial conditions and explain that for these simulations we must always consider the half-(in)activation mechanism with a mechanism that alters maximal conductances as the half-(in)activation alterations alone cannot form bursters. Figure 2 will remain as is, but we will remove any discussion of the “Group of 5,” addressing Reviewer 1’s feedback. What is presently Figure 4 will then follow, illustrating how timescale differences shape the properties of 20 degenerate solutions. We then present Figure 3 to address Reviewer 1’s critique on mechanism. Here we will explain how different timescales of half-(in)activation alteration cause the homeostatic mechanism to update channel properties differently, leading to distinct trajectories through the space of intrinsic properties and activity characteristics (as described in the response of Point 1 of Reviewer 1’s feedback). Finally, following Point 1 of Reviewer 3, we will add a new figure highlighting the role of half-(in)activation timescale during perturbation.

Discussion: To streamline the Discussion, the “Model Assumptions” section will be moved to Methods. In line with Point 2 of Reviewer 3, we will clarify how the concept of "small half-(in)activation shifts lead to global changes in neuronal properties" aligns with our core message. Additionally, following Reviewer 2’s comments, we will expand our discussion of implications by including how experimentalists might use our findings to inform studies on perturbations and development.

Methods: We will expand “Model Assumptions” to explain in more detail why we chose the L8 norm.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Overall, the conclusions of the paper are mostly supported by the data but may be overstated in some cases, and some details are also missing or not easily recognizable within the figures. The provision of additional information and analyses would be valuable to the reader and may even benefit the authors' interpretation of the data.

We thank the reviewer for the thoughtful and constructive feedback. We are pleased that the reviewer found the overall conclusions of our paper to be well supported by the data, and we appreciate the suggestions for improving figure clarity and interpretive accuracy. Below we address each point raised:

The conclusion that DREADD expression gradually decreases after 1.5-2 years is only based on a select few of the subjects assessed; in Figure 2, it appears that only 3 hM4Di cases and 2 hM3Dq cases are assessed after the 2-year timepoint. The observed decline appears consistent within the hM4Di cases, but not for the hM3Dq cases (see Figure 2C: the AAV2.1-hSyn-hM3Dq-IRES-AcGFP line is increasing after 2 years.)

We agree that our interpretation should be stated more cautiously, given the limited number of cases assessed beyond the two-year timepoint. In the revised manuscript, we will clarify in both the Results and Discussion that the observed decline is based on a subset of animals. We will also state that while a consistent decline was observed in hM4Di-expressing monkeys, the trajectory for hM3Dq expression was more variable—with at least one case showing increased in signal beyond two years.

Given that individual differences may affect expression levels, it would be helpful to see additional labels on the graphs (or in the legends) indicating which subject and which region are being represented for each line and/or data point in Figure 1C, 2B, 2C, 5A, and 5B. Alternatively, for Figures 5A and B, an accompanying table listing this information would be sufficient.

We thank the reviewer for these helpful suggestions. In response, we will revise the relevant figures as noted in the “Recommendations for the authors”, including simplifying visual encodings and improving labeling. We will also provide a supplementary table listing the animal ID and brain regions for each data point shown in the graphs.

While the authors comment on several factors that may influence peak expression levels, including serotype, promoter, titer, tag, and DREADD type, they do not comment on the volume of injection. The range in volume used per region in this study is between 2 and 54 microliters, with larger volumes typically (but not always) being used for cortical regions like the OFC and dlPFC, and smaller volumes for subcortical regions like the amygdala and putamen. This may weaken the claim that there is no significant relationship between peak expression level and brain region, as volume may be considered a confounding variable. Additionally, because of the possibility that larger volumes of viral vectors may be more likely to induce an immune response, which the authors suggest as a potential influence on transgene expression, not including volume as a factor of interest seems to be an oversight.

We thank the reviewer for raising this important issue. We agree that injection volume is a potentially confounding variable. In response, we will conduct an exploratory analysis including volume as an additional factor. We will also expand the Discussion to highlight the need for future systematic evaluation of injection volume, especially in relation to immune responses or transduction efficiency in different brain regions.

The authors conclude that vectors encoding co-expressed protein tags (such as HA) led to reduced peak expression levels, relative to vectors with an IRES-GFP sequence or with no such element at all. While interesting, this finding does not necessarily seem relevant for the efficacy of long-term expression and function, given that the authors show in Figures 1 and 2 that peak expression (as indicated by a change in binding potential relative to non-displaced radioligand, or ΔBPND) appears to taper off in all or most of the constructs assessed. The authors should take care to point out that the decline in peak expression should not be confused with the decline in longitudinal expression, as this is not clear in the discussion; i.e. the subheading, "Factors influencing DREADD expression," might be better written as, "Factors influencing peak DREADD expression," and subsequent wording in this section should specify that these particular data concern peak expression only.

We appreciate this important clarification. In response, we will revise the title to “Factors influencing peak DREADD expression levels”, and we will specify that our analysis focused on peak ΔBP_ND values around 60 days post-injection. We will also explicitly distinguish these findings from the later-stage changes in expression seen in the longitudinal PET data in both the Results and Discussion sections.

Reviewer #2 (Public review):

Weaknesses

This study is a meta-analysis of several experiments performed in one lab. The good side is that it combined a large amount of data that might not have been published individually; the downside is that all things were not planned and equated, creating a lot of unexplained variances in the data. This was yet judiciously used by the authors, but one might think that planned and organized multicentric experiments would provide more information and help test more parameters, including some related to inter-individual variability, and particular genetic constructs.

We thank the reviewer for bringing this important point to our attention. We fully agree that the retrospective nature of our dataset, compiled from multiple studies conducted within a single laboratory, introduces variability due to differences in constructs, injection sites, and timelines. While this reflects the real-world constraints of long-term NHP research, we acknowledge the need for more standardized approaches. We will add a statement in the revised Discussion emphasizing that future multicenter and harmonized studies would be valuable for systematically examining specific parameters and inter-individual variability.

Reviewer #3 (Public review):

Minor weaknesses are related to a few instances of suboptimal phrasing, and some room for improvement in time course visualization and quantification. These would be easily addressed in a revision.

These findings will undoubtedly have a very significant impact on the rapidly growing but still highly challenging field of primate chemogenetic manipulations. As such, the work represents an invaluable resource for the community.

We thank the reviewer for the positive assessment of our manuscript and for the constructive suggestions noted in the “Recommendations for the authors”. In response, we will carefully review and revise the manuscript to improve visualization and quantification.

Author response:

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

We thank you for the time you took to review our work and for your feedback! We have made only minor changes in this submission and primarily wanted to respond to the concerns raised by reviewer 1.

Reviewer #1 (Public review): 

Summary: 

Fluorescence imaging has become an increasingly popular technique for monitoring neuronal activity and neurotransmitter concentrations in the living brain. However, factors such as brain motion and changes in blood flow and oxygenation can introduce significant artifacts, particularly when activitydependent signals are small. Yogesh et al. quantified these effects using GFP, an activity-independent marker, under two-photon and wide-field imaging conditions in awake behaving mice. They report significant GFP responses across various brain regions, layers, and behavioral contexts, with magnitudes comparable to those of commonly used activity sensors. These data highlight the need for robust control strategies and careful interpretation of fluorescence functional imaging data. 

Strengths: 

The effect of hemodynamic occlusion in two-photon imaging has been previously demonstrated in sparsely labeled neurons in V1 of anesthetized animals (see Shen and Kara et al., Nature Methods, 2012). The present study builds on these findings by imaging a substantially larger population of neurons in awake, behaving mice across multiple cortical regions, layers, and stimulus conditions. The experiments are extensive, the statistical analyses are rigorous, and the results convincingly demonstrate significant GFP responses that must be accounted for in functional imaging experiments. 

In the revised version, the authors have provided further methodological details that were lacking in the previous version, expanded discussions regarding alternative explanations of these GFP responses as well as potential mitigation strategies. They also added a quantification of brain motion (Fig. S5) and the fraction of responsive neurons when conducting the same experiment using GCaMP6f (Fig. 3D-3F), among other additional information. 

Weaknesses: 

(1) The authors have now included a detailed methodology for blood vessel area quantification, where they detect blood vessels as dark holes in GFP images and measure vessel area by counting pixels below a given intensity threshold (line 437-443). However, this approach has a critical caveat: any unspecific decrease in image fluorescence will increase the number of pixels below the threshold, leading to an apparent increase in blood vessel area, even when the actual vessel size remains unchanged. As a result, this method inherently introduces a positive correlation between fluorescence decrease and vessel dilation, regardless of whether such a relationship truly exists. 

To address this issue, I recommend labelling blood vessels with an independent marker, such as a red fluorescence dye injected into the bloodstream. This approach would allow vessel dilation to be assessed independently of GFP fluorescence -- dilation would cause opposite fluorescence changes in the green and red channels (i.e., a decrease in green due to hemodynamic occlusion and an increase in red due to the expanding vessel area). In my opinion, only when such ani-correlation is observed can one reliably infer a relationship between GFP signal changes and blood vessel dynamics. 

Because this relationship is central to the author's conclusion regarding the nature of the observed GFP signals, including this experiment would greatly strengthen the paper's conclusion. 

This is correct – a more convincing demonstration that blood vessels dilate or constrict anticorrelated with apparent GFP fluorescence would be a separate blood vessel marker. However, we don’t think this experiment is worth doing, as it is also not conclusive in the sense the reviewer may have in mind. The anticorrelation does not mean that occlusion drives all of the observed effect. Our main argument is instead that there is no other potential source than hemodynamic occlusion with sufficient strength that we can think of. The experiment one would want to do is block hemodynamic changes and demonstrate that the occlusion explains all of the observed changes. 

(2) Regarding mitigation strategy, the authors advocate repeating key functional imaging experiments using GFP, and state that their aim here is to provide a control for their 2012 study (Keller et al., Neuron). Given this goal, I find it important to discuss how these new findings impact the interpretation of their 2012 results, particularly given the large GFP responses observed. 

We are happy to discuss how the conclusions of our own work are influenced by this (see more details below), but the important response of the field should probably be to revisit the conclusions of a variety of papers published in the last two decades. This goes far beyond what we can do here. 

For example, Keller et al. (2012) concluded that visuomotor mismatch strongly drives V1 activity (Fig. 3A in that study). However, in the present study, mismatch fails to produce any hemodynamic/GFP response (Fig. 3A, 3B, rightmost bar), and the corresponding calcium response is also the weakest among the three tested conditions (Fig. 3D). How do these findings affect their 2012 conclusions? 

The average calcium response of L2/3 neurons to visuomotor mismatch is probably roughly similar to the average calcium response at locomotion onset (both are on the order of 1% to 5%, depending on indicator, dataset, etc.). In the Keller et al. (2012) paper, locomotion onset was about 1.5% and mismatch about 3% (see Figure 3A in that paper). What we quantify in Figure 3 of the paper here is the fraction of responsive neurons. Thus, mismatch drives strong responses in a small subset of neurons (approx. 10%), while locomotion drives a combination of a weak responses in a large fraction of the neurons (roughly 70%) and also large responses in a subset of neurons. A strong signal in a subset of neurons is what one would expect from a neuronal response, a weak signal from many neurons would be indicative of a contaminating signal. This all appears consistent. 

Regarding influencing the conclusions of earlier work, the movement related signals described in the Keller et al. (2012) paper are probably overestimated, but are also apparent in electrophysiological recordings (Saleem et al., 2013). Thus, the locomotion responses reported in the Keller et al. (2012) paper are likely too high, but locomotion related responses in V1 are very likely real. The only conclusion we draw in the Keller et al. 2012 paper on the strength of the locomotion related responses is that they are smaller than mismatch responses (this conclusion is unaffected by hemodynamic contamination). In addition, the primary findings of the Keller et al. (2012) paper are all related to mismatch, and these conclusions are unaffected. 

Similarly, the present study shows that GFP reveals twice as many responsive neurons as GCaMP during locomotion (Fig. 3A vs. Fig. 3D, "running"). Does this mean that their 2012 conclusions regarding locomotion-induced calcium activity need reconsideration? Given that more neurons responded with GFP than with GCaMP, the authors should clarify whether they still consider GCaMP a reliable tool for measuring brain activity during locomotion. 

Comparisons of the fraction of significantly responsive neurons between GFP and GCaMP are not straightforward to interpret. One needs to factor in the difference in signal to noise between the two sensors. (Please note, we added the GCaMP responses here upon request of the reviewers). Note, there is nothing inherently wrong with the data, and comparisons within dataset are easily made (e.g. more grating responsive neurons than running responsive neurons in GCaMP, and vice versa with GFP). The comparison across datasets is not as straightforward as we define “responsive neurons” using a statistical test that compares response to baseline activity for each neuron. GFP labelled neurons are very bright and occlusion can easily be detected. Baseline fluorescence in GCaMP recordings is much lower and often close to or below the noise floor of the data (i.e. we only see the cells when they are active). Thus occlusion in GCaMP recordings is preferentially visible for cells that have high baseline fluorescence. Thus, in the GCaMP data we are likely underestimating the fraction of responsive neurons. 

Regarding whether GCaMP (or any other fluorescence indicator used in vivo) is a reliable tool, we are not sure we understand. Whenever possible, fluorescence-sensor based measurements should be corrected for hemodynamic contamination – to quantify locomotion related signals this will be more difficult than e.g. for mismatch, but that does not mean it is not reliable. 

(3) More generally, the author should discuss how functional imaging data should be interpreted going forward, given the large GFP responses reported here. Even when key experiments are repeated using GFP, it is not entirely clear how one could reliably estimate underlying neuronal activity from the observed GFP and GCaMP responses. 

We are not sure we have a good answer to this question. The strategy for addressing this problem will depend on the specifics of the experiment, and the claims. Take the case of mismatch. Here we have strong calcium responses and no evidence of GFP responses. We would argue that this is reasonable evidence that the majority of the mismatch driven GCaMP signal is likely neuronal. For locomotion onsets, both GFP and GCaMP signals go in the same direction on average. Then one could use a response amplitude distribution comparison to conservatively exclude all neurons with a GCaMP amplitude lower than e.g. the 99th percentile of the GFP response. Etc. But we don’t think there is an easy generalizable fix for this problem.  

For example, consider the results in Fig. 3A vs. 3D: how should one assess the relative strength of neuronal activity elicited by running, grating, or visuomotor mismatch? Does mismatch produce the strongest neuronal activity, since it is least affected by the hemodynamic/GFP confounds (Fig. 3A)? Or does mismatch actually produce the weakest neuronal activity, given that both its hemodynamic and calcium responses are the smallest? 

See above, the reviewer may be confounding “response strength” with “fraction of responsive neurons” here. Regarding the relationship between neuronal activity and hemodynamics, it is very likely not just the average activity of all neurons, but a specific subset that drives blood vessel constriction and dilation. This would of course be a very interesting question to answer for the interpretation of hemodynamic based measurements of brain activity, like fMRI, but goes beyond the aim of the current paper.  

In my opinion, such uncertainty makes it difficult to robustly interpret functional imaging results. Simply repeating experiments with GFP does not fully resolve this issue, as it does not provide a clear framework for quantifying the underlying neuronal activity. Does this suggest a need for a better mitigation strategy? What could these strategies be? 

If the reviewer has a good idea - we would be all ears. We don’t have a better idea currently.  

In my opinion, addressing these questions is critical not only for the authors' own work but also for the broader field to ensure a robust and reliable interpretation of functional imaging data. 

We agree, having a solution to this problem would be important – we just don’t have one.  

(4) The authors now discuss various alternative sources of the observed GFP signals. However, I feel that they often appear to dismiss these possibilities too quickly, rather than appreciating their true potential impacts (see below). 

For example, the authors argue that brain movement cannot explain their data, as movement should only result in a decrease in observed fluorescence. However, while this might hold for x-y motion, movement in the axial (z) direction can easily lead to both fluorescence increase and decrease. Neurons are not always precisely located at the focal plane -- some are slightly above or below. Axial movement in a given direction will bring some cells into focus while moving others out of focus, leading to fluorescence changes in both directions, exactly as observed in the data (see Fig. S2). 

The reviewer is correct that z-motion can result in an increase of apparent fluorescence (just like x-y motion can as well). On average however, just like with x-y motion, z-motion will always result in a decrease. This assumes that the user selecting regions of interest (the outlines of cells used to quantify fluorescence), will select these such that the distribution of cells selected centers on the zplane of the image. Thus, the distribution of z-location of the cell relative to the imaging plane will be some Gaussian like distribution centered on the z-plane of the image (with half the cell above the zplane and half below). Because the peak of the distribution is located on the z-plane at rest, any zmovement, up or down, will move away from the peak of the distribution (i.e. most cells will decrease in fluorescence). This is the same argument as for why x-y motion always results in decreases (assuming the user selects regions of interest centered on the location of the cells at rest).  

Furthermore, the authors state that they discard data with 'visible' z-motion. However, subtle axial movements that escape visual detection could still cause fluorescence fluctuations on the order of a few percent, comparable to the reported signal amplitudes. 

Correct, but as explained above, z-motion will always result in average decreases of average fluorescence as explained above.  

Finally, the authors state that "brain movement kinematics are different in shape than the GFP responses we observe". However, this appears to contradict what they show in Fig. 2A. Specifically, the first example neuron exhibits fast GFP transients locked to running onset, with rapid kinematics closely matching the movement speed signals in Fig. S5A. These fast transients are incompatible with slower blood vessel area signals (Fig. 4), suggesting that alternative sources could contribute significantly. 

We meant population average responses here. We have clarified this. Some of the signals we observed do indeed look like they could be driven by movement artifacts (whole brain motion, or probably more likely blood vessel dilation driven tissue distortion). We show this neuron to illustrate that this can also happen. However, to illustrate that this is a rare event we also show the entire distribution of peak amplitudes and the position in the distribution this neuron is from.  

In sum, the possibility that alternative signal sources could significantly contribute should be taken seriously and more thoroughly discussed. 

All possible sources (we could think of) are explicitly discussed (in roughly equal proportion). Nevertheless, the reviewer is correct that our focus here is almost exclusively on the what we think is the primary source of the problem. Given that – in my experience – this is also the one least frequently considered, I think the emphasis on – what we think is – the primary contributor is warranted.  

(5) The authors added a quantification of brain movement (Fig. S5) and claim that they "only find detectable brain motion during locomotion onsets and not the other stimuli." However, Fig. S5 presents brain 'velocity' rather than 'displacement'. A constant (non-zero) velocity in Fig. S5 B-D indicates that the brain continues to move over time, potentially leading to significant displacement from its initial position across all conditions. While displacement in the x-y plane are corrected, similar displacement in the z direction likely occurs concurrently and cannot be easily accounted for. To assess this possibility, the authors should present absolute displacement relative to pre-stimulus frames, as displacement -- not velocity -- determines the size of movement-related fluorescence changes. 

We use brain velocity here as a natural measure when using frame times as time bins. The problem with using a signed displacement is that if different running onsets move the brain in opposing directions, this can average out to zero. To counteract this, one can take the absolute displacement in a response window away from the position in a baseline time window. If this is done with time bins that correspond to frame times, this just becomes displacement per frame, i.e. velocity. Using absolute changes in displacement (i.e. velocity) is more sensitive than signed displacement. The responses for signed displacement are shown below (Author response image 1), but given that we are averaging signed quantities here, the average is not interpretable. 

Author response image 1.

Average signed brain displacement. 

Regarding a constant drift, the reviewer might be misled by the fact that the baseline brain velocity is roughly 1 pixel per frame. The registration algorithm works in integer number of pixels only. 1 pixel per frame corresponds roughly to the noise floor of the registration algorithm. Registrations are done independently for each frame. As a consequence, the registration oscillates between a shift of 17 and 18 pixels – frame by frame – if the actual shift is somewhere between 17 and 18 pixels. This “jitter” results in a baseline brain velocity of about 1 pixel per frame. 

(6) In line 132-133, the authors draw an analogy between the effect of hemodynamic occlusion and liquid crystal display (LCD) function. However, there are fundamental differences between the two. LCDs modulate light transmission by rotating the polarization of light, which then passes through a crossed polarizer. In contrast, hemodynamic occlusion alters light transmission by changing the number and absorbance properties of hemoglobin. Additionally, LCDs do not involve 'emission' light - backillumination travels through the liquid crystal layer only once, whereas hemodynamic occlusion affects both incoming excitation light and the emitted fluorescence. Given these fundamental differences, the LCD analogy may not be entirely appropriate. 

The mechanism of occlusion is, as the reviewer correctly points out, different for an LCD. In both cases however, there is a variable occluder between a light source and an observer. The fact that with hemodynamic occlusion the light passes through the occluder twice (excitation and emission) does not appear to hamper the analogy to us. We have rephrased to highlight the time varying occlusion part. 

Reviewer #2 (Public review):

-  Approach 

In this study, Yogesh et al. aimed at characterizing hemodynamic occlusion in two photon imaging, where its effects on signal fluctuations are underappreciated compared to that in wide field imaging and fiber photometry. The authors used activity-independent GFP fluorescence, GCaMP and GRAB sensors for various neuromodulators in two-photon and widefield imaging during a visuomotor context to evaluate the extent of hemodynamic occlusion in V1 and ACC. They found that the GFP responses were comparable in amplitude to smaller GCaMP responses, though exhibiting context-, cortical region-, and depth-specific effects. After quantifying blood vessel diameter change and surrounding GFP responses, they argued that GFP responses were highly correlated with changes in local blood vessel size. Furthermore, when imaging with GRAB sensors for different neuromodulators, they found that sensors with lower dynamic ranges such as GRAB-DA1m, GRAB-5HT1.0, and GRAB-NE1m exhibited responses most likely masked by the hemodynamic occlusion, while a sensor with larger SNR, GRAB-ACh3.0, showed much more distinguishable responses from blood vessel change. They thoroughly investigate other factors that could contribute to these signals and demonstrate hemodynamic occlusion is the primary cause. 

-  Impact of revision 

This is an important update to the initial submission, adding much supplemental imaging and population data that provide greater detail to the analyses and increase the confidence in the authors conclusions. 

Specifically, inclusion of the supplemental figures 1 and 2 showing GFP expression across multiple regions and the fluorescence changes of thousands of individual neurons provides a clearer picture of how these effects are distributed across the population. Characterization of brain motion across stimulation conditions in supplemental figure 5 provides strong evidence that the fluorescence changes observed in many of the conditions are unlikely to be primarily due to brain motion associated imaging artifacts. The role of vascular area on fluorescence is further supported by addition of new analyses on vasoconstriction leading to increased fluorescence in Figures 4C1-4, complementing the prior analyses of vasodilation. 

The expansion of the discussion on other factors that could lead to these changes is thorough and welcome. The arguments against pH playing a factor in fluorescence changes of GFP, due to insensitivity to changes in the expected pH range are reasonable, as are the other discussed potential factors. 

With respect to the author's responses to prior critique, we agree that activity dependent hemodynamic occlusion is best investigated under awake conditions. Measurement of these dynamics under anesthesia could lead to an underestimation of their effects. Isoflurane anesthesia causes significant vasodilation and a large reduction in fluorescence intensity in non-functional mutant GRABs. This could saturate or occlude activity dependent effects. 

- Strengths 

This work is of broad interest to two photon imaging users and GRAB developers and users. It thoroughly quantifies the hemodynamic driven GFP response and compares it to previously published GCaMP data in a similar context, and illustrates the contribution of hemodynamic occlusion to GFP and GRAB responses by characterizing the local blood vessel diameter and fluorescence change. These findings provide important considerations for the imaging community and a sobering look at the utility of these sensors for cortical imaging. 

Importantly, they draw clear distinctions between the temporal dynamics and amplitude of hemodynamic artifacts across cortical regions and layers. Moreover, they show context dependent (Dark versus during visual stimuli) effects on locomotion and optogenetic light-triggered hemodynamic signals. 

The authors suggest that signal to noise ratio of an indicator likely affects the ability to separate hemodynamic response from the underlying fluorescence signal. With a new analysis (Supplemental Figure 4) They show that the relative degree of background fluorescence does not affect the size of the artifact. 

Most of the first generation neuromodulator GRAB sensors showed relatively small responses, comparable to blood vessel changes in two photon imaging, which emphasizes a need for improved the dynamic range and response magnitude for future sensors and encourages the sensor users to consider removing hemodynamic artifacts when analyzing GRAB imaging data. 

- Weaknesses 

The largest weakness of the paper remains that, while they convincingly quantify hemodynamic artifacts across a range of conditions, they provide limited means of correcting for them. However they now discuss the relative utility of some hemodynamic correction methods (e.g. from Ocana-Santero et al., 2024). 

The paper attributes the source of 'hemodynamic occlusion' primarily to blood vessel dilation, but leaves unanswered how much may be due to shifts in blood oxygenation. Figure 4 directly addresses the question of how much of the signal can be attributed to occlusion by measuring the blood vessel dilation, and has been improved by now showing positive fluorescence effects with vasoconstriction. They now also discuss the potential impact of oxygenation. 

Along these lines, the authors carefully quantified the correlation between local blood vessel diameter and GFP response (or neuropil fluorescence vs blood vessel fluorescence with GRAB sensors). We are left to wonder to what extent does this effect depend on proximity to the vessels? Do GFP/ GRAB responses decorrelate from blood vessel activity in neurons further from vessels (refer to Figure 5A and B in Neyhart et al., Cell Reports 2024)? The authors argue that the primary impact of occlusion is from blood vessels above the plane of imaging, but without a vascular reconstruction, their evidence for this is anecdotal. 

The choice of ACC as the frontal region provides a substantial contrast in location, brain movement, and vascular architecture as compared to V1. As the authors note, ACC is close to the superior sagittal sinus and thus is the region where the largest vascular effects are likely to occur. A less medial portion of M2 may have been a more appropriate comparison. The authors now include example imaging fields for ACC and interesting out-of-plane vascular examples in the supplementary figures that help assess these impacts. 

-Overall Assessment 

This paper is an important contribution to our understanding of how hemodynamic artifacts may corrupt GRAB and calcium imaging, even in two-photon imaging modes. While it would be wonderful if the authors were able to demonstrate a reliable way to correct for hemodynamic occlusion which did not rely on doing the experiments over with a non-functional sensor or fluorescent protein, the careful measurement and reporting of the effects here is, by itself, a substantial contribution to the field of neural activity imaging. It's results are of importance to anyone conducting two-photon or widefield imaging with calcium and GRAB sensors and deserves the attention of the broader neuroscience and invivo imaging community. 

We agree with this assessment.

Reviewer #3 (Public review):

Summary:

In this study, the authors aimed to investigate if hemodynamic occlusion contributes to fluorescent signals measured with two-photon microscopy. For this, they image the activity-independent fluorophore GFP in 2 different cortical areas, at different cortical depths and in different behavioral conditions. They compare the evoked fluorescent signals with those obtained with calcium sensors and neuromodulator sensors and evaluate their relationship to vessel diameter as a readout of blood flow.

They find that GFP fluorescence transients are comparable to GCaMP6f stimuli-evoked signals in amplitude, although they are generally smaller. Yet, they are significant even at the single neuronal level. They show that GFP fluorescence transients resemble those measured with the dopamine sensor GRABDA1m and the serotonin sensor GRAB-5HT1.0 in amplitude an nature, suggesting that signals with these sensors are dominated by hemodynamic occlusion. Moreover, the authors perform similar experiments with wide-field microscopy which reveals the similarity between the two methods in generating the hemodynamic signals. Together the evidence presented calls for the development and use of high dynamic range sensors to avoid measuring signals that have another origin from the one intended to measure. In the meantime, the evidence highlights the need to control for those artifacts such as with the parallel use of activity independent fluorophores.

Strengths:

- Comprehensive study comparing different cortical regions in diverse behavioral settings in controlled conditions.

- Comparison to the state-of-the-art, i.e. what has been demonstrated with wide-field microscopy.

- Comparison to diverse activity-dependent sensors, including the widely used GCaMP.

Comments on revisions:

The authors have addressed my concerns well. I have no further comments.

We agree with this assessment.  

The following is the authors’ response to the original reviews

The major changes to the manuscript are:

(1) Re-wrote the discussion, going over all possible sources of the signals we describe.

(2) We added a quantification of brain motion as Figure S5.

(3) We added an example of blood vessel contraction as Figure 4C.

(4) We added data on the fraction of responsive neurons when measured with GCaMP as Figures 3D-3F.

(5) We added example imaging sites from all imaged regions as Figure S1.

(6) We added GFP response heatmaps of all neurons as Figure S2.

(7) We add a quantification of the relationship between GFP response amplitude and expression level Figure S4.

A detailed point-by-point response to all reviewer concerns is provided below.

Public Reviews:

Reviewer #1 (Public Review):

Fluorescence imaging has become an increasingly popular technique for monitoring neuronal activity and neurotransmitter concentrations in the living brain. However, factors such as brain motion and changes in blood flow and oxygenation can introduce significant artifacts, particularly when activity-dependent signals are small. Yogesh et al. quantified these effects using GFP, an activity-independent marker, under two-photon and wide-field imaging conditions in awake behaving mice. They report significant GFP responses across various brain regions, layers, and behavioral contexts, with magnitudes comparable to those of commonly used activity sensors. These data highlight the need for robust control strategies and careful interpretation of fluorescence functional imaging data.

Strengths:

The effect of hemodynamic occlusion in two-photon imaging has been previously demonstrated in sparsely labeled neurons in V1 of anesthetized animals (see Shen and Kara et al., Nature Methods, 2012). The present study builds on these findings by imaging a substantially larger population of neurons in awake, behaving mice across multiple cortical regions, layers, and stimulus conditions. The experiments are extensive, the statistical analyses are rigorous, and the results convincingly demonstrate significant GFP responses that must be accounted for in functional imaging experiments. However, whether these GFP responses are driven by hemodynamic occlusion remains less clear, given the complexities associated with awake imaging and GFP's properties (see below).

Weaknesses:

(1) The authors primarily attribute the observed GFP responses to hemodynamic occlusion. While this explanation is plausible, other factors may also contribute to the observed signals. These include uncompensated brain movement (e.g., axial-direction movements), leakage of visual stimulation light into the microscope, and GFP's sensitivity to changes in intracellular pH (see e.g., Kneen and Verkman, 1998, Biophysical Journal). Although the correlation between GFP signals and blood vessel diameters supports a hemodynamic contribution, it does not rule out significant contributions from these (or other) factors. Consequently, whether GFP fluorescence can reliably quantify hemodynamic occlusion in two-photon microscopy remains uncertain.

We concur; our data do not conclusively prove that the effect is only driven by hemodynamic occlusion. We have attempted to make this clearer in the text throughout the manuscript. In particular we have restructured the discussion to focus on this point. Regarding the specific alternatives the reviewer mentions here:

a) Uncompensated brain motion. While this can certainly contribute, we think the effect is negligible in our interpretation for the following reasons. First, just to point out the obvious, as with all two-photon data we acquire in the lab, we only keep data with no visible z-motion (axial). Second, and more importantly, uncompensated brain motion results in a net decrease of fluorescence. As regions of interest (ROI) are selected to be centered on neurons (as opposed to be randomly selected, or next to, or above or below), movement will – on average – result in a decrease in fluorescence, as neurons are moved out of the ROIs. In the early days of awake two-photon imaging (when preps were still less stable) – we used this movement onset decrease in fluorescence as a sign that running onsets were selected correctly (i.e. with low variance). See e.g. the dip in the running onset trace at time zero in figure 3A of (Keller et al., 2012). Third, we find no evidence for any brain motion in the case of visual stimulation, while the GFP responses during locomotion and visual stimulation are of similar magnitude. We have added a quantification of brain motion (Figure S5) and a discussion of this point to the manuscript.

b) Leakage of stimulation light. First, all light sources in the experimental room (the projector used for the mouse VR, the optogenetic stimulation light, as well as the computer monitors used to operate the microscope) are synchronized to the turnaround times of the resonant scanner of the two-photon microscope. Thus, light sources in the room are turned off for each line scan of the resonant scanner and turned on in the turnaround period. With a 12kHz scanner this results in a light cycle of 24 kHz (see Leinweber et al., 2014 for details). While the system is not perfect, we can occasionally get detectable light leak responses at the image edges (in the resonant axis as a result of the exponential off kinetics of many LEDs & lasers), these are typically 2 orders of magnitude smaller than what one would get without synchronizing, and far smaller than a single digit percentage change in GFP responses, and only detectable at the image edges. Second, while in visual cortex, dark running onsets are different from running onsets with the VR turned on (Figures 5A and B), they are indistinguishable in ACC (Figure 5C). Thus, stimulation light artefacts we can rule out.

c) GFP’s sensitivity to changes in pH. Activity results in a decrease in neuronal intracellular pH (https://pubmed.ncbi.nlm.nih.gov/14506304/, https://pubmed.ncbi.nlm.nih.gov/24312004/) – decreasing pH decreases GFP fluorescence (https://pubmed.ncbi.nlm.nih.gov/9512054/).

To reiterate, we don’t think hemodynamic occlusion is the only possible source to the effects we observe, but we do think it is most likely the largest.

(2) Regardless of the underlying mechanisms driving the GFP responses, these activity-independent signals must be accounted for in functional imaging experiments. However, the present manuscript does not explore potential strategies to mitigate these effects. Exploring and demonstrating even partial mitigation strategies could have significant implications for the field.

We concur – however, in brief, we think the only viable mitigation strategy (we are capable of), is to repeat functional imaging with GFP imaging. To unpack this: There have been numerous efforts to mitigate these hemodynamic effects using isosbestic illumination. When we started to use such strategies in the lab for widefield imaging, we thought we would calibrate the isosbestic correction using GFP recordings. The idea was that if performed correctly, an isosbestic response should look like a GFP response. Try as we may, we could not get the isosbestic responses to look like a GFP response. We suspect this is a result of the fact that none of the light sources we used were perfectly match to the isosbestic wavelength the GCaMP variants we used (not for a lack of trying, but neither lasers nor LEDs were available for purchase with exact wavelength matches). Complicating this was then also the fact that the similarity (or dissimilarity) between isosbestic and GFP responses was a function of brain region. Importantly however, just because we could not successfully apply isosbestic corrections, of course does not mean it cannot be done. Hence for the widefield experiments we then resorted to mitigating the problem by repeating the key experiments using GFP imaging (see e.g. (Heindorf and Keller, 2024)). Note, others have also argued that the best way to correct for hemodynamic artefacts is a GFP recording based correction (Valley et al., 2019). A second strategy we tried was using a second fluorophore (i.e. a red marker) in tandem with a GCaMP sensor. The problem here is that the absorption of the two differs markedly by blood and once again a correction of the GCaMP signal using the red channel was questionable at best. Thus, we think the only viable mitigation strategy we have found is GFP recordings and testing whether the postulated effects seen with calcium indicators are also present in GFP responses. This work is our attempt at a post-hoc mitigation of the problem of our own previous two-photon imaging studies.

(3) Several methodology details are missing from the Methods section. These include: (a) signal extraction methods for two-photon imaging data (b) neuropil subtraction methods (whether they are performed and, if so, how) (c) methods used to prevent visual stimulation light from being detected by the two-photon imaging system (d) methods to measure blood vessel diameter/area in each frame. The authors should provide more details in their revision.

Please excuse, this was an oversight. All details have been added to the methods.

Reviewer #2 (Public Review):

In this study, Yogesh et al. aimed at characterizing hemodynamic occlusion in two photon imaging, where its effects on signal fluctuations are underappreciated compared to that in wide field imaging and fiber photometry. The authors used activity-independent GFP fluorescence, GCaMP and GRAB sensors for various neuromodulators in two-photon and widefield imaging during a visuomotor context to evaluate the extent of hemodynamic occlusion in V1 and ACC. They found that the GFP responses were comparable in amplitude to smaller GCaMP responses, though exhibiting context-, cortical region-, and depth-specific effects. After quantifying blood vessel diameter change and surrounding GFP responses, they argued that GFP responses were highly correlated with changes in local blood vessel size. Furthermore, when imaging with GRAB sensors for different neuromodulators, they found that sensors with lower dynamic ranges such as GRAB-DA1m, GRAB5HT1.0, and GRAB-NE1m exhibited responses most likely masked by the hemodynamic occlusion, while a sensor with larger SNR, GRAB-ACh3.0, showed much more distinguishable responses from blood vessel change.

Strengths

This work is of broad interest to two photon imaging users and GRAB developers and users. It thoroughly quantifies the hemodynamic driven GFP response and compares it to previously published GCaMP data in a similar context, and illustrates the contribution of hemodynamic occlusion to GFP and GRAB responses by characterizing the local blood vessel diameter and fluorescence change. These findings provide important considerations for the imaging community and a sobering look at the utility of these sensors for cortical imaging.

Importantly, they draw clear distinctions between the temporal dynamics and amplitude of hemodynamic artifacts across cortical regions and layers. Moreover, they show context dependent (Dark versus during visual stimuli) effects on locomotion and optogenetic light-triggered hemodynamic signals.

Most of the first generation neuromodulator GRAB sensors showed relatively small responses, comparable to blood vessel changes in two photon imaging, which emphasizes a need for improved the dynamic range and response magnitude for future sensors and encourages the sensor users to consider removing hemodynamic artifacts when analyzing GRAB imaging data.

Weaknesses

(1) The largest weakness of the paper is that, while they convincingly quantify hemodynamic artifacts across a range of conditions, they do not quantify any methods of correcting for them. The utility of the paper could have been greatly enhanced had they tested hemodynamic correction methods (e.g. from Ocana-Santero et al., 2024) and applied them to their datasets. This would serve both to verify their findings-proving that hemodynamic correction removes the hemodynamic signal-and to act as a guide to the field for how to address the problem they highlight.

See also our response to reviewer 1 comment 2.

In the Ocana-Santero et al., 2024 paper they also first use GFP recordings to identify the problem. The mitigation strategy they then propose, and use, is to image a second fluorophore that emits at a different wavelength concurrently with the functional indicator. The authors then simply subtract (we think – the paper states “divisive”, but the data shown are more consistent with “subtractive” correction) the two signals to correct for hemodynamics. However, the paper does not demonstrate that the hemodynamic signals in the red channel match those in the green channel. The evidence presented that this works is at best anecdotal. In our hands this does not work (meaning the red channel does not match GFP recordings), we suspect this is a combination of crosstalk from the simultaneously recorded functional channel and the fact that hemodynamic absorption is strongly wavelength specific, or something we are doing wrong. Either way, we cannot contribute to this in the form of mitigation strategy.

Given that the GFP responses are a function of brain area and cortical depth – it is not a stretch to postulate that they also depend on genetic cell type labelled. Thus, any GFP calibration used for correction will need to be repeated for each cell type and brain area. Once experiments are repeated using GFP (the strategy we advocate for – we don’t think there is a simpler way to do this), the “correction” is just a subtraction (or a visual comparison).

(2) The paper attributes the source of 'hemodynamic occlusion' primarily to blood vessel dilation, but leaves unanswered how much may be due to shifts in blood oxygenation. Figure 4 directly addresses the question of how much of the signal can be attributed to occlusion by measuring the blood vessel dilation, but notably fails to reproduce any of the positive transients associated with locomotion in Figure 2. Thus, an investigation into or at least a discussion of what other factors (movement? Hb oxygenation?) may drive these distinct signals would be helpful.

See also our response to reviewer 1 comment 1.

We have added to Figure 4 an example of a positive transient. At running onset, superficial blood vessels in cortex tend to constrict and hence result in positive transients.

We now also mention changes in blood oxygenation as a potential source of hemodynamic occlusion. And just to be clear, blood oxygenation (or flow) changes in absence of any fluorophore, do not lead to a two-photon signal. Just in case the reviewer was concerned about intrinsic signals – these are not detectable in two photon imaging.

(3) Along these lines, the authors carefully quantified the correlation between local blood vessel diameter and GFP response (or neuropil fluorescence vs blood vessel fluorescence with GRAB sensors). To what extent does this effect depend on proximity to the vessels? Do GFP/ GRAB responses decorrelate from blood vessel activity in neurons further from vessels (refer to Figure 5A and B in Neyhart et al., Cell Reports 2024)?

We indeed thought about quantifying this, but to do this properly would require having a 3d reconstruction of the blood vessel plexus above (with respect to the optical axis) the neuron of interest, as well as some knowledge of how each vessel dilates as a function of stimulus. The prime effect is likely from blood vessels that are in the 45 degrees illumination cone above the neuron (Author response image 2). Lateral proximity to a blood vessel is likely only of secondary relevance. Thus, performing such a measurement is impractical and of little benefit for others.

Author response image 2.

A schematic representation of the cone of illumination.

While imaging a neuron (the spot on the imaging plane at the focus of the cone of illumination), the relevant blood vessels that primarily contribute to hemodynamic occlusion are those in the cone of illumination between the neuron and the objective lens. Blood vessels visible in the imaging plane (indicated by gray arrows), do not directly contribute to hemodynamic occlusion. Any distance dependence of hemodynamic occlusion in the observed response of a neuron to these blood vessels in the imaging plane is at best incidental.

(4) Raw traces are shown in Figure 2 but we are never presented with the unaveraged data for locomotion of stimulus presentation times, which limits the reader's ability to independently assess variability in the data. Inclusion of heatmaps comparing event aligned GFP to GCaMP6f may be of value to the reader.

We fear we are not sure what the reviewer means by “the unaveraged data for locomotion of stimulus presentation times”. We suspect this should read “locomotion or stimulus…”. We have added heat maps of the responses of all neurons of the data shown in Figure 1 – as Figure S2.

(5) More detailed analysis of differences between the kinds of dynamics observed in GFP vs GCaMP6f expressing neurons could aid in identifying artifacts in otherwise clean data. The example neurons in Figure 2A hint at this as each display unique waveforms and the question of whether certain properties of their dynamics can reveal the hemodynamic rather than indicator driven nature of the signal is left open. Eg. do the decay rate and rise times differ significantly from GCaMP6f signals?

The most informative distinction we have found is differences in peak responses (Figure 2B). Decay and rise time measurements critically depend on the identification of “events”. As a function of how selective one is with what one calls an event (e.g. easy in example 1 of Figure 2 – but more difficult in examples 2 and 3), one gets very different estimates of rise and decay times. Due to the fact that peak amplitudes are lower in GFP responses – rise and decay times will be either slower or noisier (depending on where the threshold for event detection is set).

(6) The authors suggest that signal to noise ratio of an indicator likely affects the ability to separate hemodynamic response from the underlying fluorescence signal. Does the degree of background fluorescence affect the size of the artifact? If there was variation in background and overall expression level in the data this could potentially be used to answer this question. Could lower (or higher!) expression levels increase the effects of hemodynamic occlusion?

There may be a misunderstanding (i.e. we might be misunderstanding the reviewer’s argument here). Our statement from the manuscript that the signal to noise ratio of an indicator matters is based on the simple consideration that hemodynamic occlusion is in the range of 0 to 2 % ΔF/F. The larger the dynamic range of the indicator, the less of a problem 2% ΔF/F are. Imagine an indicator with average responses in the 100’s of % ΔF/F - then this would be a non-problem. For indicators with a dynamic range less than 1%, a 2% artifact is a problem.

Regarding “background” fluorescence, we are not sure what is meant here. In case the reviewer means fluorescence that comes from indicator molecules in processes (as opposed to soma) that are typically ignored (or classified as neuropil) – we are not sure how this would help. The occlusion effects are identical for both somatic and axonal or dendritic GFP (the source of the GFP fluorescence is not relevant for the occlusion effect). In case the reviewer means “baseline” fluorescence – above a noise threshold ΔF/F₀ should be constant independent of F₀ (i.e. baseline fluorescence). This also holds in the data, see Figure S4. We might be stating the trivial - the normalization of fluorescence activity as ΔF/F₀ has the effect that the “occluder" effect is constant for all values of all F₀.

(7) The choice of the phrase 'hemodynamic occlusion' may cause some confusion as the authors address both positive and negative responses in the GFP expressing neurons, and there may be additional contributions from changes in blood oxygenation state.

Regarding the potential confusion with regards to terminology, occlusion can decrease or increase.

Only under the (incorrect) assumption that occlusion is zero at baseline would this be confusing – no? If the reviewer has a suggestion for a different term, we’d be open to changing it.

Regarding blood oxygenation – this is absolutely correct, we did not explicitly point this out in the previous version of the manuscript. Occlusion changes are driven by a combination of changes to volume and “opacity” of the blood. Oxygenation changes would be in the second category. We have clarified this in the manuscript.

(8) The choice of ACC as the frontal region provides a substantial contrast in location, brain movement, and vascular architecture as compared to V1. As the authors note, ACC is close to the superior sagittal sinus and thus is the region where the largest vascular effects are likely to occur. The reader is left to wonder how much of the ROI may or may not have included vasculature in the ACC vs V1 recordings as the only images of the recording sites provided are for V1. We are left unable to conclude whether the differences observed between these regions are due to the presence of visible vasculature, capillary blood flow or differences in neurovasculature coupling between regions. A less medial portion of M2 may have been a more appropriate comparison. At least, inclusion of more example imaging fields for ACC in the supplementary figures would be of value.

Both the choice of V1 and ACC were simply driven by previous experiments we had already done in these areas with calcium indicators. And we agree, the relevant axis is likely distance from midline, not AP – i.e. RSC and ACC are likely more similar, and V1 and lateral M2 more similar. We have made this point explicitly in the manuscript and have added sample fields of view as Figure S1.

(9) In Figure 3, How do the proportions of responsive GFP neurons compare to GCaMP6f neurons?

We have added the data for GCaMP responses.

(10) How is variance explained calculated in Figure 4? Is this from a linear model and R^2 value? Is this variance estimate for separate predictors by using single variable models? The methods should describe the construction of the model including the design matrix and how the model was fit and if and how cross validation was run.

This is simply a linear model (i.e. R^2) – we have added this to the methods.

(11) Cortical depth is coarsely defined as L2/3 or L5, without numerical ranges in depth from pia.

Layer 2/3 imaging was done at a depth of 100-250 μm from pia, and the same for layer 5 was 400-600 μm. This has been added to the methods.

Overall Assessment:

This paper is an important contribution to our understanding of how hemodynamic artifacts may corrupt GRAB and calcium imaging, even in two-photon imaging modes. Certain useful control experiments, such as intrinsic optical imaging in the same paradigms, were not reported, nor were any hemodynamic correction methods investigated. Thus, this limits both mechanistic conclusions and the overall utility with respect to immediate applications by end users. Nevertheless, the paper is of significant importance to anyone conducting two-photon or widefield imaging with calcium and GRAB sensors and deserves the attention of the broader neuroscience and in-vivo imaging community.

Reviewer #3 (Public review):

In this study, the authors aimed to investigate if hemodynamic occlusion contributes to fluorescent signals measured with two-photon microscopy. For this, they image the activity-independent fluorophore GFP in 2 different cortical areas, at different cortical depths and in different behavioral conditions. They compare the evoked fluorescent signals with those obtained with calcium sensors and neuromodulator sensors and evaluate their relationship to vessel diameter as a readout of blood flow.

They find that GFP fluorescence transients are comparable to GCaMP6f stimuli-evoked signals in amplitude, although they are generally smaller. Yet, they are significant even at the single neuronal level. They show that GFP fluorescence transients resemble those measured with the dopamine sensor GRABDA1m and the serotonin sensor GRAB-5HT1.0 in amplitude an nature, suggesting that signals with these sensors are dominated by hemodynamic occlusion. Moreover, the authors perform similar experiments with wide-field microscopy which reveals the similarity between the two methods in generating the hemodynamic signals. Together the evidence presented calls for the development and use of high dynamic range sensors to avoid measuring signals that have another origin from the one intended to measure. In the meantime, the evidence highlights the need to control for those artifacts such as with the parallel use of activity independent fluorophores.

Strengths:

- Comprehensive study comparing different cortical regions in diverse behavioral settings in controlled conditions.

- Comparison to the state-of-the-art, i.e. what has been demonstrated with wide-field microscopy.

- Comparison to diverse activity-dependent sensors, including the widely used GCaMP.

Weaknesses:

(1) The kinetics of GCaMP is stereotypic. An analysis/comment on if and how the kinetics of the signals could be used to distinguish the hemodynamic occlusion artefacts from calcium signals would be useful.

We might be misunderstanding what the reviewer means by “the kinetics of GCaMP are stereotypic”. The kinetics are clearly stereotypic if one has isolated single action potential responses in a genetically identified cell type. But data recorded in vivo looks very different, see e.g. example traces in figure 1g of (Keller et al., 2012). And these are selected example traces, the average GCaMP trace looks perhaps more like the three example traces shown in Figure 2 (this is not surprising if the GCaMP signals one records in vivo are a superposition of calcium responses and hemodynamic occlusion). All quantification of kinetics relies on identifying “events”. We cannot identify events in any meaningful way for most of the data (see e.g. examples 2 and 3 in Figure 2). The one feature we can reliably identify as differing between GCaMP and GFP responses is peak response amplitude (as quantified in Figure 2).

(2) Is it possible that motion is affecting the signals in a certain degree? This issue is not made clear.

See also our response to reviewer 1 comment 1. In brief, we have added a quantification of motion artefacts as Figure S5, and argue that motion artefacts could only account for locomotion onset responses (there is no detectable brain motion to visual responses) and would predict a decrease in fluorescence (not an increase).

(3) The causal relationship with blood flow remains open. Hemodynamic occlusion seems a good candidate causing changes in GFP fluorescence, but this remains to be well addressed in further research.

We agree – we have made this clearer in the manuscript.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Figure 2A shows three neurons with convincing GFP responses, with amplitudes often exceeding 100%. However, after seeing these data, I actually feel less convinced that these responses are related to hemodynamic occlusion. Blood vessel diameter changes by at most a few percent during behavior -- how could such small changes lead to >100% changes in GFP fluorescence?

My guess is that these responses might instead be related to motion artifacts, particularly given the strong correlation between these responses and running speed (Figure 2A). One possible way to test this is by examining a pixelwise map of fluorescence changes (dF/F) during running vs. baseline. If hemodynamic effects are involved, one would likely see a shadow of the involved blood vessels in this map. Conversely, if motion artifacts are the primary factor, the map of dF/F should resemble the spatial gradients of the mean fluorescence image. Examining pixelwise maps of dF/F will likely provide insights regarding the nature of the GFP signals.

The underlying assumption (“blood vessel diameter changes by at most a few percent”) might be incorrect here. (Note also, relevant is likely the cross section, not diameter.) See Figure 4A1 and B1 for quantification of example blood vessel area changes - both example vessels change area by approximately 50%. Also note, example 1 in Figure 2 is an extreme example. The example was chosen to highlight that effects can be large. To try to illustrate that this is not typical however, we also show the distribution of all neurons in Figure 2B and mark all three example cells – example 1 is at the very tail of the distribution.

Regarding the analysis suggested, we have added examples of this for running onset to the manuscript (Figure S7). We have examples in which a blood vessel shadow is clearly visible. More typical however, is a general increase in fluorescence (on running onset) that we think is caused by blood vessels closer to the surface of the brain.

(2) Figure 3A shows strong GFP responses during running, while visuomotor mismatch elicit virtually no GFP-responsive neurons. This finding is puzzling, as visuomotor mismatch has been shown by the same group to activate L2/3 neurons more strongly than running (see Figure 3A, Keller et al., 2012, Neuron). Stronger neuronal activation should, in theory, result in more pronounced hemodynamic effects, and therefore, a higher proportion of GFP-responsive neurons. The absence of GFP responses during visuomotor mismatch raises questions about whether GFP signals are directly linked to hemodynamic occlusion.

An alternative explanation is that the strong GFP responses observed during running could instead be driven by motion artifacts, e.g., those associated with the increased head or body movements during running onsets. Such artifacts could explain the observed GFP responses, rather than hemodynamic occlusion.

This might be a misunderstanding. Mismatch responses are primarily observed in mismatch neurons. These are superficial L2/3 neurons (possibly the population that in higher mammals is L2 neurons). The fact that mismatch responses are primarily observed in this superficial population is likely the reason they were discovered using two-photon calcium imaging (which tends to have a bias towards superficial neurons as the image quality is best there), and seen in much fewer neurons when using electrophysiological techniques (Saleem et al., 2013) that are biased to deeper neurons. In response to Reviewer #2, we have now also added a quantification of the fraction of neurons responsive to these stimuli when using GCaMP (Figure 3D-F). The fraction of neurons responsive to visuomotor mismatch is smaller than those responsive on locomotion or to visual stimuli.

Thus, based on “average” responses across all cortical cell types (our L2/3 recordings here are as unbiased across all of L2/3 as possible) the response profiles (strong running onset and visual responses, and weak MM responses) are probably what one would expect in first approximation also in the blood vessel response profile. Complicating this is of course the fact that it is likely some cell type specific activity that contributes most to blood flow changes, not simply average neuronal activity.

See response to public review comment 1 for a discussion of alternative sources, including motion artefacts.

(3) Given the potential confound associated with brain motion, the authors might consider quantifying hemodynamic occlusion effects under more controlled conditions, such as in anesthetized animals, where brain movement is minimal. They could use drifting grating stimuli, which are known to produce wellcharacterized blood vessel and hemodynamic responses in V1. The effects of hemodynamic occlusion can then be quantified by imaging the fluorescence of an activity-independent marker. For maximal robustness, GFP should ideally be avoided, due to its known sensitivity to pH changes, as noted in the public review.

Brain motion is negligible to visual stimuli in the awake mouse as well (Figure S5). This is likely the better control than anesthetized recordings – anesthesia has strong effects on blood pressure, heart rate, breathing, etc. all of which would introduce more confounds.

(4) Regardless of the precise mechanism driving the observed GFP response, these activity-independent signals must be accounted for in functional imaging experiments. This applies not only to experiments using small dynamic range sensors but also to those employing 'high dynamic range' sensors like GCaMP6, which, according to the authors, exhibit responses only ~2-fold greater than those of GFP.

In this context, the extensive GFP imaging data are highly valuable, as they could serve as a benchmark for evaluating the effectiveness of correction methods. Ideally, effective correction methods should produce minimal responses when applied to GFP imaging data. With these data at hand, I strongly encourage the authors to explore potential correction methods, as such methods could have far-reaching impact on the field.

As discussed above, we have tested a number of such correction approaches for both widefield and two-photon imaging and could never recover a response profile that resembles the GFP response. The “correction method” we have come to favor, is repeating experiments using GFP (i.e. what we have done here).

(5) Several correction approaches could be considered: for instance, the strong correlation between GFP responses and blood vessel diameter (as shown in Figure 4) could potentially be leveraged to predict and compensate for the activity-independent signals. Alternatively, expressing an activity-independent marker alongside the activity sensor in orthogonal spectral channels could enable simultaneous monitoring and correction of activity-independent signals. Finally, computational procedure to remove common fluctuations, measured from background or 'neuropil' regions (see, e.g., Kerlin et al., 2010, Neuron; Giovannucci et al., 2019, eLife), may help reduce the contamination in cellular ROIs. The authors could try some or all of these methods, and benchmark their effectiveness by assessing, e.g., the number of GFP responsive neurons after correction.

Over the years we have tried many of these approaches. A correction using a second fluorophore of a different color likely fails because blood absorption is strongly wavelength dependent, making it challenging to calibrate the correction factor. Neuropil “correction” on GCaMP data, even with the best implementations, is just a common mode subtraction. The signal in the neuropil – as the name implies is just an average of many axons and dendrites in the vicinity – most of these processes are from nearby neurons making a neuropil response simply an average response of the neurons in some neighborhood. Adding the problem of hemodynamic responses (which on small scales will also influence nearby neurons and neuropil similarly) makes disentangling the two effects impossible (i.e. neuropil subtraction makes the problem worse, not better). However, just because we fail in implementing all of these methods, does not necessarily mean the method is faulty. Hence we have chosen not to comment on any such method, and simply provide the only mitigation strategy that works in our hands – record GFP responses.

(6) Given the potential usefulness of the GFP imaging data, I encourage the authors to share these data in a public repository to facilitate the development of correction methods.

Certainly – all of our data are always published. In the early years of the lab on an FMI repository here https://data.fmi.ch/ - more recently now on Zenodo.

(7) As noted in the public review, several methodology details are missing. Most importantly, I could not find the description in the Methods section explaining how fluorescence signals from individual neurons were extracted from two-photon imaging data. The existing section on 'Extraction of neuronal activity' appears to cover only the wide-field analysis, with details about two-photon analysis seemingly absent.

Please excuse the omission – this has all been added to the methods. In brief, to answer your questions:

Were regions of interest (ROIs) for individual cells identified manually or automatically?

We use a mixture of manual and automatic methods for our two-photon data. Based on a median filtered (spatially) version of the mean fluorescence image, we used a threshold based selection of ROIs. This was then visually inspected and manually corrected where necessary such that ROIs were at least 250 pixels and only labelled clearly identifiable neurons.

Was fluorescence within each ROI calculated by averaging signals across pixels, or were signal de-mixing algorithms (e.g., PCA, ICA, or NMF) applied?

We use the average fluorescence across pixels without any de-mixing algorithms here and in all our two-photon experiments. De-mixing algorithms can introduce a variety of artefacts.

Additionally, did the authors account for and correct the contribution of surrounding neuropil?

No neuropil correction was applied. It would also be difficult to see how this would help. If the model of hemodynamic occlusion is correct, one would expect occlusion effects to change on the length scale of blood vessels (i.e. tens to hundreds of microns). Thus, the effect of occlusion on neuropil and cells should be the similar. Neuropil “correction” is always based on the idea of removing signals that are common to both neuropil and somata, thereby complicating the interpretation of the resulting signal even further.

Without these methodological details, it is difficult to accurately interpret the two-photon signals reported in the manuscript.

(8) The rationale for using the average fluorescence of a ROI within the blood vessel as a proxy for blood vessel diameter is not entirely clear to me. The authors should provide a clearer justification for this approach in their revision.

Consider a ROI placed within a blood vessel at the focus of the illumination cone (Author response image 3). Given the axial point-spread-function of two-photon imaging is in the range of 0.5 μm laterally and 3 μm axially (indicated by the bicone), emitted photons from the fluorescent tissue outside of the blood vessel but within the two-photon volume will contribute to change in fluorescence in the ROI. A change in the blood vessel volume, say an increase on dilation, would decrease the amount of emission photons reaching the objective by, one, pushing more of the fluorescent tissue outside of the two-photon volume, and two, by presenting greater hemodynamic occlusion to the photons emitted by the fluorescent tissue immediately below the vessel. Conversely, on vasoconstriction there are more emission photons at the objective.

In line with this argument, as shown in Figure 4A1-A2, B1-B2 and C1-C2, we do find that the change in fluorescence of blood vessel ROI varies inversely with the area of the blood vessel. Of course, change in blood vessel ROI fluorescence is only a proxy for vessel size. Extracting blood vessel boundaries from individual two-photon frames was noisy and proved unreliable in the absence of specific dyes to label the vessel walls. We thus resorted to using blood vessel ROI fluorescence as a proxy for hemodynamic occlusion, and tested how much of the variance in GFP responses is explained by the change in blood vessel ROI response.

We have added an explanation to the manuscript, as suggested.

Author response image 3.

Average response of ROIs placed within blood vessels co-vary with hemodynamic occlusion.

(9) I find that the Shen et al., 2012, Nature Methods paper has gone quite far to demonstrate the effect of hemodynamic occlusion in two photon imaging. Therefore, I suggest the authors describe and cite this work not only in the discussion but also in the introduction, where they can highlight the key questions left unanswered by that study and explain how their manuscript aims to address them.

We have added the reference and point to the work in the introduction as suggested.

Reviewer #3 (Recommendations for the authors):

I appreciate very much that the study is presented in a very clear manner.

A few comments that could clarify it even further:

(1) Fig. 1: make clear on legend if it is an average of full FOVs.

The traces shown are the average over ROIs (neurons) – we have clarified in the figure legend as suggested.

(2) Give a more complete definition of hemodynamic occlusion to understand the hypothesis in the relationship between blood vessel dilation and GFP fluorescence (116-119). Maybe, move the phrase from conclusion "Since blood absorbs light, hemodynamic occlusion can affect fluorescence intensity measurements" (219-220).

Very good point – we expanded on the definition in the introduction.

(3) For clarity, mention in the main text the method used to assess how a parameter explains the variance (126-129).

Is implemented.

(4) Discuss the possible relationship of the signals to neuronal activity.

We have added this to the discussion.

(5) Discuss if the measurements could provide any functional insights, whether they could be used to learn something about the brain.

We have added this to the discussion.

Author response:

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

As to the exceptionally minor issue, namely, correction for multiple statistical tests (minor because the data and the error are presented in the text). We have now conducted one-way ANOVA to back the data displayed in Fig 4A., and Supp. Figs 19 and 21. In each case ANOVA revealed a highly significant difference among means: Dunnett’s post hoc test was then used to test each result against SBW25, with the multiple comparisons corrected for in the analysis.

This resulted in changes to the description of the statistical analysis in the following captions:

To Figure 4.

Where we previously referred to paired t-tests we now state:  ANOVA revealed a highly significant difference among means [F_7,16 = 8.19, p < 0.001] with Dunnett’s post-hoc test adjusted for multiple comparisons showing that five genotypes (*) differ significantly (p < 0.05) from SBW25.

To Supplementary Figure 19.

Where we previously referred to paired t-tests we now state: ANOVA revealed a highly significant difference among means [F_7,16 = 16.74, p < 0.001] with Dunnett’s post-hoc test adjusted for multiple comparisons showing that three genotypes (*) differ significantly (p < 0.05) from SBW25.

To Supplementary Figure 21.

Where we previously referred to paired t-tests we now state:  ANOVA revealed a highly significant difference among means [F_7,89 = 9.97, p < 0.0001] with Dunnett’s post-hoc test adjusted for multiple comparisons showing that SBW25 ∆mreB and SBW25 ∆PFLU4921-4925 are significantly different (*) from SBW25 (p < 0.05).

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

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

The authors performed experimental evolution of MreB mutants that have a slow-growing round phenotype and studied the subsequent evolutionary trajectory using analysis tools from molecular biology. It was remarkable and interesting that they found that the original phenotype was not restored (most common in these studies) but that the round phenotype was maintained. 

Strengths: 

The finding that the round phenotype was maintained during evolution rather than that the original phenotype, rod-shaped cells, was recovered is interesting. The paper extensively investigates what happens during adaptation with various different techniques. Also, the extensive discussion of the findings at the end of the paper is well thought through and insighXul. 

Weaknesses: 

I find there are three general weaknesses: 

(1) Although the paper states in the abstract that it emphasizes "new knowledge to be gained" it remains unclear what this concretely is. On page 4 they state 3 three research questions, these could be more extensively discussed in the abstract. Also, these questions read more like genetics questions while the paper is a lot about cell biological findings. 

Thank you for drawing attention to the unnecessary and gratuitous nature of the last sentence of the Abstract. We are in agreement. It has been modified, and we have taken  advantage of additional word space to draw attention to the importance of the two competing (testable) hypotheses laid out in the Discussion. 

As to new knowledge, please see the Results and particularly the Discussion. But beyond this, and as recognised by others, there is real value for cell biology in seeing how (and whether) selection can compensate for effects that are deleterious to fitness. The results will very often depart from those delivered from, for example, suppressor analyses, or bottom up engineering. 

In the work recounted in our paper, we chose to focus – by way of proof-of principle – on the most commonly observed mutations, namely, those within pbp1A.  But beyond this gene, we detected mutations  in other components of the cell shape / division machinery whose connections are not yet understood and which are the focus of on-going investigation.  

As to the three questions posed at the end of the Introduction, the first concerns whether selection can compensate for deleterious effects of deleting mreB (a question that pertains to evolutionary aspects); the second seeks understanding of genetic factors; the third aims to shed light on the genotype-to-phenotype map (which is where the cell biology comes into play).  Given space restrictions, we cannot see how we could usefully expand, let alone discuss, the three questions raised at the end of the Introduction in restrictive space available in the Abstract.   

(2) It is not clear to me from the text what we already know about the restoration of MreB loss from suppressors studies (in the literature). Are there suppressor screens in the literature and which part of the findings is consistent with suppressor screens and which parts are new knowledge?  

As stated in the Introduction, a previous study with B. subtilis (which harbours three MreB isoforms and where the isoform named “MreB” is essential for growth under normal conditions), suppressors of MreB lethality were found to occur in ponA, a class A penicillin binding protein (Kawai et al., 2009). This led to recognition that MreB plays a role in recruiting Pbp1A to the lateral cell wall. On the other hand, Patel et al. (2020) have shown that deletion of classA PBPs leads to an up-regulation of rod complex activity. Although there is a connection between rod complex and class A PBPs, a further study has shown that the two systems work semi-autonomously (Cho et al., 2016). 

Our work confirms a connection between MreB and Pbp1A, and has shed new light on how this interaction is established by means of natural selection, which targets the integrity of cell wall. Indeed, the Rod complex and class A PBPs have complementary activities in the building of the cell wall with each of the two systems able to compensate for the other in order to maintain cell wall integrity. Please see the major part of the Discussion. In terms of specifics, the connection between mreB and pbp1A (shown by Kawai et al (2009)) is indirect because it is based on extragenic transposon insertions. In our study, the genetic connection is mechanistically demonstrated.  In addition, we capture that the evolutionary dynamics is rapid and we finally enriched understanding of the genotype-to-phenotype map.

(3) The clarity of the figures, captions, and data quantification need to be improved.  

Modifications have been implemented. Please see responses to specific queries listed below.

Reviewer #2 (Public Review): 

Yulo et al. show that deletion of MreB causes reduced fitness in P. fluorescens SBW25 and that this reduction in fitness may be primarily caused by alterations in cell volume. To understand the effect of cell volume on proliferation, they performed an evolution experiment through which they predominantly obtained mutations in pbp1A that decreased cell volume and increased viability. Furthermore, they provide evidence to propose that the pbp1A mutants may have decreased PG cross-linking which might have helped in restoring the fitness by rectifying the disorganised PG synthesis caused by the absence of MreB. Overall this is an interesting study. 

Queries: 

Do the small cells of mreB null background indeed have no DNA? It is not apparent from the DAPI images presented in Supplementary Figure 17. A more detailed analysis will help to support this claim. 

It is entirely possible that small cells have no DNA, because if cell division is aberrant then division can occur prior to DNA segregation resulting in cells with no DNA. It is clear from microscopic observation that both small and large cells do not divide. It is, however, true, that we are unable to state – given our measures of DNA content – that small cells have no DNA. We have made this clear on page 13, paragraph 2.

What happens to viability and cell morphology when pbp1A is removed in the mreB null background? If it is actually a decrease in pbp1A activity that leads to the rescue, then pbp1A- mreB- cells should have better viability, reduced cell volume and organised PG synthesis. Especially as the PG cross-linking is almost at the same level as the T362 or D484 mutant.  

Please see fitness data in Supp. Fig. 13. Fitness of ∆mreB ∆pbp1A is no different to that caused by a point mutation. Cells remain round.  

What is the status of PG cross-linking in ΔmreB Δpflu4921-4925 (Line 7)? 

This was not analysed as the focus of this experiment was PBPs. A priori, there is no obvious reason to suspect that ∆4921-25 (which lacks oprD) would be affected in PBP activity.

What is the morphology of the cells in Line 2 and Line 5? It may be interesting to see if PG cross-linking and cell wall synthesis is also altered in the cells from these lines. 

The focus of investigation was restricted to L1, L4 and L7. Indeed, it would be interesting to look at the mutants harbouring mutations in :sZ, but this is beyond scope of the present investigation (but is on-going). The morphology of L2 and L5 are shown in Supp. Fig. 9.

The data presented in 4B should be quantified with appropriate input controls. 

Band intensity has now been quantified (see new Supp. Fig .20). The controls are SBW25, SBW25∆pbp1A, SBW25 ∆mreB and SBW25 ∆mreBpbp1A as explained in the paper.

What are the statistical analyses used in 4A and what is the significance value? 

Our oversight. These were reported in Supp. Fig. 19, but should also have been presented in Fig. 4A. Data are means of three biological replicates. The statistical tests are comparisons between each mutant and SBW25, and assessed by paired t-tests.  

A more rigorous statistical analysis indicating the number of replicates should be done throughout. 

We have checked and made additions where necessary and where previously lacking. In particular, details are provided in Fig. 1E, Fig. 4A and Fig. 4B. For Fig. 4C we have produced quantitative measures of heterogeneity in new cell wall insertion. These are reported in Supp. Fig. 21 (and referred to in the text and figure caption) and show that patterns of cell wall insertion in ∆mreB are highly heterogeneous.

Reviewer #3 (Public Review): 

This paper addresses an understudied problem in microbiology: the evolution of bacterial cell shape. Bacterial cells can take a range of forms, among the most common being rods and spheres. The consensus view is that rods are the ancestral form and spheres the derived form. The molecular machinery governing these different shapes is fairly well understood but the evolutionary drivers responsible for the transition between rods and spheres are not. Enter Yulo et al.'s work. The authors start by noting that deletion of a highly conserved gene called MreB in the Gram-negative bacterium Pseudomonas fluorescens reduces fitness but does not kill the cell (as happens in other species like E. coli and B. subtilis) and causes cells to become spherical rather than their normal rod shape. They then ask whether evolution for 1000 generations restores the rod shape of these cells when propagated in a rich, benign medium. 

The answer is no. The evolved lineages recovered fitness by the end of the experiment, growing just as well as the unevolved rod-shaped ancestor, but remained spherical. The authors provide an impressively detailed investigation of the genetic and molecular changes that evolved. Their leading results are: 

(1) The loss of fitness associated with MreB deletion causes high variation in cell volume among sibling cells after cell division. 

(2) Fitness recovery is largely driven by a single, loss-of-function point mutation that evolves within the first ~250 generations that reduces the variability in cell volume among siblings. 

(3) The main route to restoring fitness and reducing variability involves loss of function mutations causing a reduction of TPase and peptidoglycan cross-linking, leading to a disorganized cell wall architecture characteristic of spherical cells. 

The inferences made in this paper are on the whole well supported by the data. The authors provide a uniquely comprehensive account of how a key genetic change leads to gains in fitness and the spectrum of phenotypes that are impacted and provide insight into the molecular mechanisms underlying models of cell shape. 

Suggested improvements and clarifications include: 

(1) A schematic of the molecular interactions governing cell wall formation could be useful in the introduction to help orient readers less familiar with the current state of knowledge and key molecular players. 

We understand that this would be desirable, but there are numerous recent reviews with detailed schematics that we think the interested reader would be better consulting. These are referenced in the text.

(2) More detail on the bioinformatics approaches to assembling genomes and identifying the key compensatory mutations are needed, particularly in the methods section. This whole subject remains something of an art, with many different tools used. Specifying these tools, and the parameter settings used, will improve transparency and reproducibility, should it be needed. 

We overlooked providing this detail, which has now been corrected by provision of more information in the Materials and Methods. In short we used Breseq, the clonal option, with default parameters. Additional analyses were conducted using Genieous. The BreSeq output files are provided https://doi.org/10.17617/3.CU5SX1 (which include all read data).

(3) Corrections for multiple comparisons should be used and reported whenever more than one construct or strain is compared to the common ancestor, as in Supplementary Figure 19A (relative PG density of different constructs versus the SBW25 ancestor). 

The data presented in Supp Fig 19A (and Fig 4A) do not involve multiple comparisons. In each instance the comparison is between SBW25 and each of the different mutants. A paired t-test is thus appropriate.

(4) The authors refrain from making strong claims about the nature of selection on cell shape, perhaps because their main interest is the molecular mechanisms responsible. However, I think more can be said on the evolutionary side, along two lines. First, they have good evidence that cell volume is a trait under strong stabilizing selection, with cells of intermediate volume having the highest fitness. This is notable because there are rather few examples of stabilizing selection where the underlying mechanisms responsible are so well characterized. Second, this paper succeeds in providing an explanation for how spherical cells can readily evolve from a rod-shaped ancestor but leaves open how rods evolved in the first place. Can the authors speculate as to how the complex, coordinated system leading to rods first evolved? Or why not all cells have lost rod shape and become spherical, if it is so easy to achieve? These are important evolutionary questions that remain unaddressed. The manuscript could be improved by at least flagging these as unanswered questions deserving of further attention. 

These are interesting points, but our capacity to comment is entirely speculative. Nonetheless, we have added an additional paragraph to the Discussion that expresses an opinion that has yet to receive attention:

“Given the complexity of the cell wall synthesis machinery that defines rod-shape in bacteria, it is hard to imagine how rods could have evolved prior to cocci. However, the cylindrical shape offers a number of advantages. For a given biomass (or cell volume), shape determines surface area of the cell envelope, which is the smallest surface area associated with the spherical shape. As shape sets the surface/volume ratio, it also determines the ratio between supply (proportional to the surface) and demand (proportional to cell volume). From this point of view, it is more efficient to be cylindrical (Young 2006). This also holds for surface attachment and biofilm formation (Young 2006). But above all, for growing cells, the ratio between supply and demand is constant in rod shaped bacteria, whereas it decreases for cocci. This requires that spherical cells evolve complex regulatory networks capable of maintaining the correct concentration of cellular proteins despite changes in surface/volume ratio. From this point of view, rod-shaped bacteria offer opportunities to develop unsophisticated regulatory networks.”

why not all cells have lost rod shape and become spherical.

Please see Kevin Young’s 2006 review on the adaptive significance of cell shape

The value of this paper stems both from the insight it provides on the underlying molecular model for cell shape and from what it reveals about some key features of the evolutionary process. The paper, as it currently stands, provides more on which to chew for the molecular side than the evolutionary side. It provides valuable insights into the molecular architecture of how cells grow and what governs their shape. The evolutionary phenomena emphasized by the authors - the importance of loss-of-function mutations in driving rapid compensatory fitness gains and that multiple genetic and molecular routes to high fitness are often available, even in the relatively short time frame of a few hundred generations - are well understood phenomena and so arguably of less broad interest. The more compelling evolutionary questions concern the nature and cause of stabilizing selection (in this case cell volume) and the evolution of complexity. The paper misses an opportunity to highlight the former and, while claiming to shed light on the latter, provides rather little useful insight. 

Thank you for these thoughts and comments. However, we disagree that the experimental results are an overlooked opportunity to discuss stabilising selection. Stabilising selection occurs when selection favours a particular phenotype causing a reduction in underpinning population-level genetic diversity. This is not happening when selection acts on SBW25 ∆mreB leading to a restoration of fitness. Driving the response are biophysical factors, primarily the critical need to balance elongation rate with rate of septation. This occurs without any change in underlying genetic diversity.  

Recommendations for the authors:  

Reviewer 1 (Recommendations for the Authors): 

Hereby my suggestion for improvement of the quantification of the data, the figures, and the text. 

-  p 14, what is the unit of elongation rate?  

At first mention we have made clear that the unit is given in minutes^-1

-  p 14, please give an error bar for both p=0.85 and f=0.77, to be able to conclude they are different 

Error on the probability p is estimated at the 95% confidence interval by the formula:1.96 , where N is the total number of cells. This has been added in the paragraph p »probability » of the Image Analysis section in the Material and Methods. 

We also added errors on p measurement in the main text.

-  p 14, all the % differences need an errorbar 

The error bars and means are given in Fig 3C and 3D.

-  Figure 1B adds units to compactness, and what does it represent? Is the cell size the estimated volume (that is mentioned in the caption)? Shouldn't the datapoints have error bars? 

Compactness is defined in the “Image Analysis” section of the Material and Methods. It is a dimensionless parameter. The distribution of individual cell shapes / sizes are depicted in Fig 1B. Error does arise from segmentation, but the degree of variance (few pixels) is much smaller than the representations of individual cells shown.

-  Figure 1C caption, are the 50.000 cells? 

Correct. Figure caption has been altered.

-  Figure 1D, first the elongation rate is described as a volume per minute, but now, looking at the units it is a rate, how is it normalized? 

Elongation rate is explained in the Materials and Methods (see the image analysis section) and is not volume per minute. It is dV/dt = r*V (the unit of r is min^-1). Page 9 includes specific mention of the unit of r.

-  Figure 1E, how many cells (n) per replicate? 

Our apologies. We have corrected the figure caption that now reads:

“Proportion of live cells in ancestral SBW25 (black bar) and ΔmreB (grey bar) based on LIVE/DEAD BacLight Bacterial Viability Kit protocol. Cells were pelleted at 2,000 x g for 2 minutes to preserve ΔmreB cell integrity. Error bars are means and standard deviation of three biological replicates (n>100).”

-  Figure 1G, how does this compare to the wildtype 

The volume for wild type SBW25 is 3.27µm^3 (within the “white zone”). This is mentioned in the text.

-  Figure 2B, is this really volume, not size? And can you add microscopy images? 

The x-axis is volume (see Materials and Methods, subsection image analysis). Images are available in Supp. Fig. 9.

-  Figure 3A what does L1, L4 and L7 refer too? Is it correct that these same lines are picked for WT and delta_mreB 

Thank you for pointing this out. This was an earlier nomenclature. It was shorthand for the mutants that are specified everywhere else by genotype and has now been corrected. 

-  Figure 3c: either way write out p, so which probability, or you need a simple cartoon that is plotted. 

The value p is the probability to proceed to the next generation and is explained in Materials and Methods  subsection image analysis.  We feel this is intuitive and does not require a cartoon. We nonetheless added a sentence to the Materials and Methods to aid clarity.

-  Figure 4B can you add a ladder to the gel? 

No ladder was included, but the controls provide all the necessary information. The band corresponding to PBP1A is defined by presence in SBW25, but absence in SBW25 ∆pbp1A.

-  Figure 4c, can you improve the quantification of these images? How were these selected and how well do they represent the community? 

We apologise for the lack of quantitative description for data presented in Fig 4C. This has now been corrected. In brief, we measured the intensity of fluorescent signal from between 10 and 14 cells and computed the mean and standard deviation of pixel intensity for each cell. To rule out possible artifacts associated with variation of the mean intensity, we calculated the ratio of the standard deviation divided by the square root of the mean. These data reveal heterogeneity in cell wall synthesis and provide strong statistical support for the claim that cell wall synthesis in ∆mreB is significantly more heterogeneous than the control. The data are provided in new Supp. Fig. 21. 

Minor comments: 

-  It would be interesting if the findings of this experimental evolution study could be related to comparative studies (if these have ever been executed).  

Little is possible, but Hendrickson and Yulo published a portion of the originally posted preprint separately. We include a citation to that paper. 

-  p 13, halfway through the page, the second paragraph lacks a conclusion, why do we care about DNA content? 

It is a minor observation that was included by way of providing a complete description of cell phenotype.  

-  p 17, "suggesting that ... loss-of-function", I do no not understand what this is based upon. 

We show that the fitness of a pbp1A deletion is indistinguishable from the fitness of one of the pbp1A point mutants. This fact establishes that the point mutation had the same effects as a gene deletion thus supporting the claim that the point mutations identified during the course of the selection experiment decrease (or destroy) PBP1A function.

-  p 25, at the top of the page: do you have a reference for the statement that a disorganized cell wall architecture is suited to the topology of spherical cells? 

The statement is a conclusion that comes from our reasoning. It stems from the fact that it is impossible to entirely map the surface of a sphere with parallel strands.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

Evading predation is of utmost importance for most animals and camouflage is one of the predominant mechanisms. Wu et al. set out to test the hypothesis of a unique camouflage system in leafhoppers. These animals coat themselves with brochosomes, which are spherical nanostructures that are produced in the Malpighian tubules and are distributed on the cuticle after eclosion. Based on previous findings on the reflectivity properties of brochosomes, the authors provide very good evidence that these nanostructures indeed reduce the reflectivity of the animals thereby reducing predation by jumping spiders. Further, they identify four proteins, which are essential for the proper development and function of brochosomes. In RNAi experiments, the regular brochosome structure is lost, the reflectivity reduced and the respective animals are prone to increased predation. Finally, the authors provide some phylogenetic sequence analyses and speculate about the evolution of these essential genes.

Strengths:

The study is very comprehensive including careful optical measurements, EM and TM analysis of the nanoparticles and their production line in the malphigian tubules, in vivo predation tests, and knock-down experiments to identify essential proteins. Indeed, the results are very convincingly in line with the starting hypothesis such that the study robustly assigns a new biological function to the brochosome coating system.

A key strength of the study is that the biological relevance of the brochosome coating is convincingly shown by an in vivo predation test using a known predator from the same habitat.

Another major step forward is an RNAi screen, which identified four proteins, which are essential for the brochosome structure (BSMs). After respective RNAi knock-downs, the brochosomes show curious malformations that are interesting in terms of the self-assembly of these nanostructures. The optical and in vivo predation tests provide excellent support for the model that the RNAi knock-down leads to a change of brochosomes structure, which reduces reflectivity, which in turn leads to a decrease of the antipredatory effect.

Thank you very much for your positive feedback and insightful comments on our manuscript. We are delighted that you acknowledge the efforts we have made in studying the components and functions of Brochosomal proteins. We have carefully considered your suggestions and have thoroughly revised the manuscript to address the shortcomings identified in our original submission. We hope that the revised version meets with your approval. Below, please find our detailed point-by-point responses.

Weaknesses:

The reduction of reflectivity by aberrant brochosomes or after ageing is only around 10%. This may seem little to have an effect in real life. On the other hand, the in vivo predation tests confirm an influence. Hence, this is not a real weakness of the study - just a note to reconsider the wording for describing the degree of reflectivity.

Thank you for your valuable suggestions. Based on your recommendations, we have revised the manuscript accordingly. Although the absolute reduction in light reflection due to Brochosomal coverage is approximately 10%, the relative decrease in light reflection on the leafhopper's surface is nearly 30%. Specifically, in the ultraviolet region, the reflection is reduced from about 30% to 20%, and in the visible light region, it is reduced from 20% to 10%. For detailed revisions, please refer to lines 151-156 of the revised manuscript.

The single gene knockdowns seemed to lead to a very low penetrance of malformed brochosomes (Figure Supplement 3). Judging from the overview slides, less than 1% of brochosomes may have been affected. A quantification of regular versus abnormal particles in both, wildtype and RNAi treatments would have helped to exclude that the shown aberrant brochosomes did not just reflect a putative level of "normal" background defects. Of note, the quadruple knock-down of all BSMs seemed to lead to a high penetrance (Figure 4), which was already reflected in the microtubule production line. While the data shown are convincing, a quantification might strengthen the argument.

While the RNAi effects seemed to be very specific to brochosomes and therefore very likely specific, an off-target control for RNAi was still missing. Finding the same/similar phenotype with a non-overlapping dsRNA fragment in one off-target experiment is usually considered required and sufficient. Further, the details of the targeted sequence will help future workers on the topic.

Thank you for your valuable suggestions. Based on your recommendations, we have synthesized dsRNA targeting two non-overlapping regions of the coding sequences for four Brochosomal structural protein genes. These dsRNAs were injected individually and in combination for each gene. Our RNAi experiments for each BSM gene demonstrated that both individual and combined injections significantly suppressed the expression of the target genes, with the combined injection yielding slightly better silencing efficiency. Statistical analysis of the SEM observations revealed that the combined injection of dsRNAs targeting two non-overlapping regions led to a 60-70% reduction in the surface area coverage of Brochosomes. Additionally, approximately 20% of the remaining Brochosomes exhibited significant morphological changes. For detailed revisions, please refer to lines 199-211 of the revised manuscript, as well as Figures 3A and 3C, and Supplementary Figures 4 and 5.

The main weakness in the current manuscript may be the phylogenetic analysis and the model of how the genes evolved. Several aspects were not clearly or consistently stated such that I felt unsure about what the authors actually think. For instance: Are all the 4 BSMs related to each other or only BSM2 and 3? If so, not only BSM2 and 3 would be called "paralogs" but also the other BSMs. If they were all related, then a phylogenetic tree including all BSMs should be shown to visualize the relatedness (including the putative ancestral gene if that is the model of the authors). Actually, I was not sure about how the authors think about the emergence of the BSMs. Are they real orphan genes (i.e. not present outside the respective clade) or was there an ancestral gene that was duplicated and diverged to form the BSMs? Where in the phylogeny does the first of the BSMs or ancestral proteins emerge (is the gene found in Clastoptera arizonana the most ancestral one?)? Maybe, the evolution of the BSMs would have to be discussed individually for each gene as they show somewhat different patterns of emergence and loss (BSM4 present in all species, the others with different degrees of phylogenetic restriction).

Thank you very much for your constructive feedback on our phylogenetic analysis and the modeling of gene evolution. We fully agree with your insights and acknowledge that the evolutionary analysis of BSM genes remains somewhat ambiguous. This ambiguity is primarily due to the limited research on the precise structural protein composition of Brochosomes. While proteomics studies have analyzed and discussed the structural proteins of Brochosomes, the accurate composition of these proteins is still poorly understood. In this study, we identified four BSM proteins, but given the intricate structure of Brochosomes as proteinaceous spheres, we believe there may be additional BSM genes that have not yet been identified. Moreover, despite the presence of over ten thousand species within the Cicadomorpha, only three species have genome sequences available, and fewer than a hundred species have transcriptome sequencing data. The scarcity of research on Brochosomes, as well as the limited availability of genomic and transcriptomic data, poses significant challenges for our phylogenetic analysis and understanding of BSM gene evolution.

Based on your suggestions, we have revised the manuscript accordingly. Specifically, we have updated Figure 5C by including ten additional species from Cereopoidea, Cicadoidea, and Fulgoroidea to better illustrate that BSM genes are true orphan genes. We have also added a phylogenetic tree of BSM genes within Cicadidae in Supplementary Figure 3. Additionally, we have expanded the discussion of BSM gene evolution in the manuscript (lines 503-556). For detailed revisions, please refer to Figure 5C, Supplementary Figure 3, and lines 507-585 of the revised manuscript.

Related to these questions I remained unsure about some details in Figure 5. On what kind of analysis is the phylogeny based? Why are some species not colored, although they are located on the same branch as colored ones? What is the measure for homology values - % identity/similarity? The homology labels for Nephotetix cincticeps and N. virescens seem to be flipped: the latter is displayed with 100% identity for all genes with all proteins while the former should actually show this. As a consequence of these uncertainties, I could not fully follow the respective discussion and model for gene evolution.

Thank you very much for your insightful comments and suggestions. We have carefully considered your feedback and have thoroughly revised our manuscript accordingly. Specifically, we have enhanced the description of the phylogenetic analysis process to provide greater clarity and transparency, with the detailed methods now included in lines 789-798. Regarding Figure 5C, we appreciate your attention to the coloring scheme. We would like to clarify that the family Cicadellidae comprises 25 subfamilies, many of which are represented by only one species in our figure. To ensure clarity and meaningful representation, we have chosen to color only those subfamilies with more than three species, thereby avoiding visual clutter and emphasizing the most relevant taxonomic groups. Additionally, we have corrected the inverted homology labels for Nephotetix cincticeps and Nephotetix virescens to ensure the accuracy and consistency of our data presentation.

Conclusion:

The authors successfully tested their hypothesis in a multidisciplinary approach and convincingly assigned a new biological function to the brochosomes system. The results fully support their claims - only the quantification of the penetrance in the RNAi experiments would be helpful to strengthen the point. The author's analysis of the evolution of BSM genes remained a bit vague and I remained unsure about their respective conclusions.

The work is a very interesting study case of the evolutionary emergence of a new system to evade predators. Based on this study, the function of the BSM genes could now be studied in other species to provide insights into putative ancestral functions. Further, studying the self-assembly of such highly regular complex nano-structures will be strongly fostered by the identification of the four key structural genes.

Reviewer #1 (Recommendations for the authors):

Main manuscript:

Please consider the annotated pdf with suggestions for wording and comments at the authors' discretion:

Thank you very much for your detailed suggestions and comments provided in the annotated PDF. We have carefully reviewed each of your points and have revised the manuscript accordingly. All changes have been highlighted in red text for your convenience. The revised manuscript with tracked changes is available for your review. We believe these revisions have improved the clarity and quality of our manuscript. Thank you again for your valuable feedback.

Supplementary Figure 2 C:

Y-axes:

- label: "surface coverage in %"

- there are different scale values for the different days (e.g. 80-105 for day 5 and 0-80 at day 25). As a comparison between days is interesting, it would help to have the same scale values for all. That would show the decrease more intuitively.

Thank you very much for your suggestion regarding the Y-axis in Supplementary Figure 2C. We agree that using a consistent scale across all time points is essential for clear and intuitive comparison. In the revised manuscript, we have standardized the Y-axis scale for Supplementary Figure 2C to a uniform range of 0-100% for all days. This change allows for a more straightforward visualization of the decreasing trend in surface coverage over time.

Reviewer #3 (Public review):

Summary:

In this manuscript, the authors investigate the optical properties of brochosomes produced by leafhoppers. They hypothesize that brochosomes reduce light reflection on the leafhopper's body surface, aiding in predator avoidance. Their hypothesis is supported by experiments involving jumping spiders. Additionally, the authors employ a variety of techniques including micro-UV-Vis spectroscopy, electron microscopy, transcriptome and proteome analysis, and bioassays. This study is highly interesting, and the experimental data is well-organized and logically presented.

Strengths:

The use of brochosomes as a camouflage coating has been hypothesized since 1936 (R.B. Swain, Entomol. News 47, 264-266, 1936) with evidence demonstrated by similar synthetic brochosome systems in a number of recent studies (S. Yang, et al. Nat. Commun. 8:1285, 2017; L. Wang, et al., PNAS. 121: e2312700121, 2024). However, direct biological evidence or relevant field studies have been lacking to directly support the hypothesis that brochosomes are used for camouflage. This work provides the first biological evidence demonstrating that natural brochosomes can be used as a camouflage coating to reduce the leafhoppers' observability of their predators. The design of the experiments is novel.

We are extremely grateful for your positive feedback and insightful comments on our manuscript. We are delighted that you have recognized the efforts we have put into our research on how brochosomes serve as a camouflage coating to reduce the detectability of leafhoppers to their predators. We have carefully considered your suggestions and have thoroughly revised the manuscript to address the shortcomings of the original version. We hope that the revised version meets with your approval. Below, please find our detailed point-by-point responses.

Weaknesses:

(1) The observation that brochosome coatings become sparse after 25 days in both male and female leafhoppers, resulting in increased predation by jumping spiders, is intriguing. However, since leafhoppers consistently secrete and groom brochosomes, it would be beneficial to explore why brochosomes become significantly less dense after 25 days.

Thank you very much for your valuable suggestions. We appreciate your interest in the reduction of brochosomal density on the surface of leafhoppers after 25 days.We believe that the primary reason for the decreased density of brochosomes on the leafhopper surface after 25 days is the reduced synthesis and secretion of brochosomes. The Malpighian tubules are the main sites for brochosome synthesis. As shown in Figure 2D and Supplementary Figure 1, the thick glandular segments of the Malpighian tubules in both male and female leafhoppers begin to atrophy 15 days after reaching adulthood. This indicates a gradual decline in brochosome synthesis and secretion after day 15 of adulthood. Following your suggestion, we have revised the discussion section of the manuscript to elaborate on this observation. The detailed changes can be found in lines 474-491 of the revised manuscript.

(2) The authors demonstrate that brochosome coatings reduce UV (specular) reflection compared to surfaces without brochosomes, which can be attributed to the rough geometry of brochosomes as discussed in the literature. However, it would be valuable to investigate whether the proteins forming the brochosomes are also UV absorbing.

Thank you very much for your valuable suggestions. Following your advice, we have successfully expressed four BSM genes in a prokaryotic system, purified the corresponding proteins, and applied them to quartz glass surfaces. We then measured the light reflectance of the quartz glass surfaces coated with these purified proteins. The results showed that the purified BSM proteins did not exhibit better antireflective properties compared to the control GST protein. For more details, please refer to Supplementary Figure 8 in the revised manuscript.  We believe that the excellent antireflective properties of brochosomes are fundamentally due to their unique geometric shapes. The hollow pores within the brochosomes, with diameters of approximately 100 nm, are significantly smaller than most wavelengths in the visible spectrum. When light passes through these tiny pores, diffraction occurs, while light passing through the ridges of the brochosomes causes scattering. The interference between the diffracted and scattered light from these pores and ridges results in the observed extinction characteristics of brochosomes. We have incorporated these insights into the discussion section of the revised manuscript (lines 416-425 and lines 432-442 of the revised manuscript).

(3) The experiments with jumping spiders show that brochosomes help leafhoppers avoid predators to some extent. It would be beneficial for the authors to elaborate on the exact mechanism behind this camouflage effect. Specifically, why does reduced UV reflection aid in predator avoidance? If predators are sensitive to UV light, how does the reduced UV reflectance specifically contribute to evasion?

Thank you very much for your valuable suggestions. Based on your advice, we have included a detailed discussion on how reducing ultraviolet (UV) reflection can help insects avoid predation. The revised content can be found in lines 445-460 of the revised manuscript.

“UV light serves as a crucial visual cue for various insect predators, enhancing foraging, navigation, mating behavior, and prey identification (Cronin & Bok, 2016; Morehouse et al., 2017; Silberglied, 1979). Predators such as birds, reptiles, and predatory arthropods often rely on UV vision to detect prey (Church et al., 1998; Li & Lim, 2005; Zou et al., 2011). However, UV reflectance from insect cuticles can disrupt camouflage, increasing the risk of detection and predation, as natural backgrounds like leaves, bark, and soil typically reflect minimal UV light (Endler, 1997; Li & Lim, 2005; Tovee, 1995). To mitigate this risk, insects often possess anti-reflective cuticular structures that reduce UV and broad-spectrum light reflectance. This strategy is widespread among insects, including cicadas, dragonflies, and butterflies, and has been shown to decrease predator detection rates (Hooper et al., 2006; Siddique et al., 2015; Zhang et al., 2006). For example, the compound eyes of moths feature hexagonal protuberances that reduce UV reflectance, aiding nocturnal concealment (Blagodatski et al., 2015; Stavenga et al., 2005). In butterflies, UV reflectance from eyespots on wings can attract predators, but reducing UV reflectance or eyespot size can lower predation risk and enhance camouflage (Chan et al., 2019; Lyytinen et al., 2004). Hence, the reflection of ultraviolet light from the insect cuticle surface increases the risk of predation by disrupting camouflage (Tovee, 1995)”

(4) An important reference regarding the moth-eye effect is missing. Please consider including the following paper: Clapham, P. B., and M. C. Hutley. "Reduction of lens reflection by the 'Moth Eye' principle." Nature 244: 281-282 (1973).

Thank you very much for pointing out the omission of the important reference on the “moth eye” effect. We sincerely apologize for the oversight. Based on your suggestion, we have now included the seminal paper by Clapham and Hutley (1973) in the revised manuscript. The reference has been added to both the Introduction and Discussion sections to provide a more comprehensive context for our discussion on anti-reflective structures in insects.

(5) The introduction should be revised to accurately reflect the related contributions in literature. Specifically, the novelty of this work lies in the demonstration of the camouflage effect of brochosomes using jumping spiders, which is verified for the first time in leafhoppers. However, the proposed use of brochosome powder for camouflage was first described by R.B. Swain (R.B. Swain, Notes on the oviposition and life history of the leafhopper Oncometopta undata Fabr. (Homoptera: Cicadellidae), Entomol. News. 47: 264-266 (1936)). Recently, the antireflective and potential camouflage functions of brochosomes were further studied by Yang et al. based on synthetic brochosomes and simulated vision techniques (S. Yang, et al. "Ultra-antireflective synthetic brochosomes." Nature Communications 8: 1285 (2017)). Later, Lei et al. demonstrated the antireflective properties of natural brochosomes in 2020 (C.-W. Lei, et al., "Leafhopper wing-inspired broadband omnidirectional antireflective embroidered ball-like structure arrays using a nonlithography-based methodology." Langmuir 36: 5296-5302 (2020)). Very recently, Wang et al. successfully fabricated synthetic brochosomes with precise geometry akin to those natural ones, and further elucidated the antireflective mechanisms based on the brochosome geometry and their role in reducing the observability of leafhoppers to their predators (L. Wang et al. "Geometric design of antireflective leafhopper brochosomes." Proceedings of the National Academy of Sciences 121: e2312700121 (2024)).

Thank you very much for your valuable suggestions regarding the revision of the introduction to accurately reflect the relevant contributions in the literature. Based on your feedback, we have thoroughly revised the introduction and added the suggested references to provide a comprehensive context for our study. The details of these revisions can be found in lines 84-94 of the revised manuscript.

Reviewer #3 (Recommendations for the authors):

(1) In Figure 2E, the data for Male-5d appears to be missing. Please verify and ensure all relevant data is included.

Thank you for pointing out the issue regarding the data presentation in Figure 2E.We apologize for any confusion caused by the overlapping data points and the less conspicuous color choice for Male-5d. We have carefully reviewed the data and confirmed that all relevant data points, including Male-5d, are indeed present in the dataset. In the revised manuscript, we have adjusted the color scheme for Male-5d and Female-5d in Figure 2E to ensure that both curves are clearly distinguishable, even in areas where they overlap. This adjustment should facilitate a more accurate and convenient observation of the data trends. We appreciate your attention to detail, and we believe these revisions have improved the clarity and readability of the figure.

(2) In Figure 6, please clarify the reflectance data in the inset. Clearly explain what the blue and light blue curves represent.

Thank you for your suggestion regarding Figure 6.We have revised the figure to improve clarity. The light blue curve now represents the reflectance measurements of leafhoppers with higher brochosome coverage, while the dark blue curve corresponds to those with lower coverage. These changes, along with updated labels in the figure legend, ensure that the data are clearly distinguishable and easy to interpret. We appreciate your feedback and believe these revisions have enhanced the overall clarity of the figure.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Weaknesses (clarifications needed):

(1) Experimental Design:

The study does not mention whether the authors examined sex differences or any measures of attractiveness or hierarchy among participants (e.g., students vs. teachers). Including these variables could provide a more nuanced understanding of group dynamics.

We are grateful to the reviewer for pointing out this valuable question. We have clarified that future studies should include sex differences or any measures of attractiveness or hierarchy among participants (e.g., students vs. teachers) (p. 27).

“Finally, future research should investigate additional variables, including sex differences and measures of attractiveness or hierarchy among participants, such as students versus teachers.”  p. 27

(2) fNIRS Data Acquisition:

The authors' approach to addressing individual differences in anatomy is lacking in detail. Understanding how they identified the optimal channels for synchrony between participants would be beneficial. Was this done by averaging to find the location with the highest coherence?

We apologize for missing some details here. We have included the following information in the fNIRS data acquisition and fNIRS data analyses to clarify the details (pp. 8 and 12).

We employed the one-sample t-test method to assess the GNS disparity between the baseline and task sessions, identifying particular channels of interest. This analysis did not ascertain the maximum coherence level, but rather pinpointed the channel exhibiting significant divergence between the two sessions, which we designated as pertinent to the group decision-making task. Furthermore, we selected the PFC and left TPJ as our reference brain regions, guided by existing literature.

“Two optode probe sets were used to cover each participant's prefrontal and left TPJ regions (Figure S1). The DLPFC plays a crucial role in group decision-making processes, with findings suggesting that individuals exhibiting reduced prefrontal activity were more prone to out-group exclusion and demonstrated stronger in-group preferences (Goupil et al., 2021; Jankovic, 2014; Yang et al., 2020). Similarly, the left TPJ has been previously reported to be associated with decision-making and information exchange (Freitas et al., 2019; Tindale et al., 2019).”  p. 8

“Time-averaged GNS (also averaged across channels in each group) was compared between the baseline session (i.e., the resting phase) and the task session (from reading information to making decisions) using a series of one-sample t-tests. Here, p-values were thresholded by controlling for FDR (p < 0.05; Benjamini & Hochberg, 1995). When determining the frequency band of interest, the time-averaged GNS was also averaged across channels. After that, we analyzed the time-averaged GNS of each channel. Then, channels showing significant GNS were regarded as regions of interest and included in subsequent analyses.” p. 12

(3) Behavioral Analysis:

For group identification, the analysis currently uses a dichotomous approach. Introducing a regression model to capture the degree of identification could offer more granular insights into how varying levels of group identification affect collective behavior and performance.

Thank you for your suggestion. As suggested, we have conducted the regression model to examine how varying levels of group identification affect collective performance, with the score of group identification being the independent variable and collective performance as the dependent variable (pp.9 and 15).

“Moreover, we employed a regression model to examine how varying levels of group identification affect collective performance, using group identification scores as the independent variable and collective performance as the dependent variable.”  p.9

“The results from the regression model highlighted a significant association between the degree of group identification and collective performance (β \= 0.45, t = 4.56, p \= 0.019).”  p.15

(4) Single Brain Activation Analysis:

The application of the General Linear Model (GLM) is unclear, particularly given the long block durations and absence of multiple trials. Further explanation is needed on how the GLM was implemented under these conditions.

Thank you for your suggestion, we have added more details in this section (p.11).

“In the GLM model analysis, HbO was the dependent variable, and the regression amount was set for different task stages (a. Reading information, b. Sharing private information, c. Discussing information, d. Decision). After that, we convolved the regression factor with the Hemodynamic Response Function (HRF) and obtained the brain activation β value of each participant in each channel at different task stages through regression analysis.’  p.11

(5) Within-group neural Synchrony (GNS) Calculation:

The method for calculating GNS could be improved by using mutual information instead of pairwise summation, as suggested by Xie et al. (2020) in their study on fMRI triadic hyperscanning. Additionally, the explanation of GNS calculation is inconsistent. At one point, it is mentioned that GNS was averaged across time and channels, while elsewhere, it is stated that channels with the highest GNS were selected. Clarification on this point is essential.

We appreciate the reviewer for highlighting this inquiry. We utilized a conventional GNS calculation approach, as detailed in Line 296 of the manuscript, where the GNS was determined in pairs after the WTC computation, and then averaged. Further details regarding the second question have been provided in the article (p.12).

(6) Placement of fNIRS Probes:

The probes were only placed in the frontal regions, despite literature suggesting that the superior temporal sulcus (STS) and temporoparietal junction (TPJ) regions are crucial for triadic team performance. A justification for this choice or inclusion of these regions in future studies would be beneficial.

The original manuscript clearly stated the use of two optode probe sets to encompass the prefrontal and left TPJ regions of each participant (see Figure S1, p. 8).

(7) Interpretation of fNIRS Data:

Given that fNIRS signals are slow, similar to BOLD signals in fMRI, the interpretation of Figure 6 raises concerns. It suggests that it takes several minutes (on the order of 4-5 minutes) for people to collaborate, which seems implausible. More context or re-evaluation of this interpretation is needed.

The question you have pointed out is very pertinent, and we have added more explanation for this result (pp. 25-26).

As previous studies have shown, the BOLD signal collected by fNIRS is slowly increasing compared to neuronal activity, which means that it has hysteresis (Turner et al., 1998). In social interactions such as group decision-making, the time of neural synchronization is delayed because people need to spend time increasing the number of dialogues to improve collaboration efficiency and form the same preference (Zhang et al., 2019). For example, the study of group consensus found that participants would show significant neural alignment after completing a period of dialogue (Sievers et al., 2024). In the task of cooperation, with the improvement of tacit understanding between two participants, the higher degree of neural synchronization (Cui et al., 2012). Therefore, the generation of neural synchronization depends on the interaction over a period of time. Therefore, we believe that the 4-5 minutes of collaboration time shown in Figure 6 may be related to establishing consensus and the same preference of team members, which is reflected in the dynamic time change of neural synchronization.

Moreover, previous studies on neural synchronization during social interaction and group decision-making revealed that substantial neural synchronization occurred around 50-55 seconds into a teaching task involving prior knowledge (Liu et al., 2019) and persisted approximately 6 minutes into the discussion period (Xie et al., 2023). These results collectively validate the suitability of utilizing fNIRS signal response time in our study (pp. 25-26).

“Our study also has demonstrated significant increases in single-brain activation, DLPFC-OFC functional connectivity, and GNS at 7, 12, and 17 minutes, respectively, following task initiation. The significant increase in these neural activities together constructs the two-in-one neural model that explains how group identification influences the collective performance we proposed. As previous studies have shown, the BOLD signal collected by fNIRS is slowly increasing compared to neuronal activity, which means that it has hysteresis (Turner et al., 1998). In social interactions such as group decision-making, the time of neural synchronization is delayed because people need to spend time increasing the number of dialogues to improve collaboration efficiency and form the same preference (Zhang et al., 2019). For example, participants would exhibit significant neural alignment, but only after they had completed a period of dialogue (Sievers et al., 2024). In the task of cooperation, with the improvement of cooperation efficiency between two participants, the higher degree of neural synchronization (Cui et al., 2012). Therefore, the generation of neural synchronization depends on the interaction over a period of time, which can affect the estimation of collaboration time. Prior research has shown that when the teaching task with prior knowledge began 50-55 seconds, significant neural synchronization could be generated between teacher and students, which meant that students and teacher achieved the same goal of learning knowledge (Liu et al., 2019). Moreover, a noteworthy increase in GNS was observed approximately 6 minutes into the group discussion period for better discussing and solving the problem (Xie et al., 2023). These findings are similar to ours. Therefore, the time points we found could reflect the dynamic time change of the neural process of team collaboration.’ pp.25-26

Reviewer #2 (Public review):

Weaknesses:

The authors need to clearly articulate their hypothesis regarding why neural synchronization occurs during social interaction. For example, in line 284, it is stated that "It is plausible that neural synchronization is closely associated with group identification and collective performance...", but this is far from self-evident. Neural synchronization can occur even when people are merely watching a movie (Hasson et al., 2004), and movie-watchers are not engaged in collective behavior. There is no direct link between the IBS and collective behavior. The authors should explain why they believe inter-brain synchronization occurs in interactive settings and why they think it is related to collective behavior/performance.

Thank you for bringing these points to our attention, we have clarified the relationship between neural synchronization and collective behavior in the Introduction section. (p.4). Moreover, in order to investigate whether neural synchronization stems from a common task or environment, we pseudo-randomized all pairs of subjects and created a null distribution consisting of 1,000 pseudo-groups, as described in Lines 311-315. This approach enabled us to eliminate neural synchronization resulting from factors other than social interaction, allowing us to identify neural patterns associated with collective performance (p.12).

“Moreover, Ni et al. (2024) indicated that neural synchronization was linked to the strength of social-emotional communication and connections between individuals. An increase in neural synchronization has also been shown to predict the coordination and cooperation abilities of group members (Lu et al., 2023). Therefore, we hypothesize that neural synchronization may be related to group performance.” p.4

“After that, the nonparametric permutation test was conducted on the observed interaction effects on GNS of the real group against the 1,000 permutation samples. By pseudo-randomizing the data of all participants, a null distribution of 1000 pseudo-groups was generated (e.g., time series from member 1 in group 1 were grouped with member 2 in group 2 & member 3 in group 3). The GNS of 1,000 reshuffled pseudo-groups was computed, and the GNS of the real groups was assessed by comparing it with the values generated by 1000 reshuffled pseudo-groups.” p.12

The authors state that "GNS in the OFC was a reliable neuromarker, indicating the influence of group identification on collective performance," but this claim is too strong. Please refer to Figure 4B. Do the authors really believe that collective performance can be predicted given the correlation with the large variance shown? There is a significant discrepancy between observing a correlation between two variables and asserting that one variable is a predictive biomarker for the other.

Thank you for your suggestion, we have revised the relevant statement (p.18).

“Through correlation and regression model analysis, we found that in group decision-making, the increase in group identity would affect group performance by improving GNS in the OFC brain region.”  p.18

Why are the individual answers being analyzed as collective performance (See, L-184)? Although these are performances that emerge after the group discussion, they seem to be individual performances rather than collective ones. Typically, wouldn't the result of a consensus be considered a collective performance? The authors should clarify why the individual's answer is being treated as the measure of collective performance.

We appreciate the insightful comment provided by the reviewer. The decision to utilize individual responses as a metric of overall performance is based on several key considerations. Previous studies on various hidden profile tasks have utilized averaged individual scores to represent collective performance (e.g., Stasser et al., 1995; Wittenbaum et al., 1996; Brockner et al., 2022). Secondly, while consensus outcomes are typically regarded as collective expressions, we argue that in the context of this study, individual responses are not independent entities but rather extensions of the group decision-making process. The collective deliberation process significantly influenced individual thinking and decision-making in this study. Through group discussions, members shared perspectives, adjusted their stances, and formulated their responses based on collective insights. The responses provided by participants in this study were molded by the dynamics of group conversations, serving as an indirect measure of group performance and potentially indicating the efficacy of collective deliberations.

Performing SPM-based mapping followed by conducting a t-test on the channels within statistically significant regions constitutes double dipping, which is not an acceptable method (Kriegeskorte et al., 2011). This issue is evident in, for example, Figures 3A and 4A.

Please refer to the following source: https://www.nature.com/articles/nn.2303

We have carefully reviewed the articles provided by the reviewer, and we acknowledge the concerns regarding selective analysis and double dipping in our statistical approach. To address this, we believe it is important to clarify this issue further in the Discussion section (pp.26-27).

Our study introduces a novel perspective while utilizing conventional fNIRS-based hyperscanning analyses (Liu et al., 2019; Pärnamets et al., 2020; Reinero et al., 2021; Számadó et al., 2021; Solansky, 2011), methods that are widely endorsed within the field. In our analysis, significant channels were first identified using a one-sample t-test, followed by additional analyses including ANOVA, independent samples t-tests, and other procedures. We would like to emphasize that the statistical assumptions underlying the one-sample t-test and paired-sample t-test in our study maintain a level of independence. Moreover, to further mitigate concerns about the potential for double dipping, we employed permutation testing to validate the robustness of our results and ensure that our findings are not influenced by biases inherent in the selection of significant regions.

We recognize the importance of rigorous statistical practices and are committed to upholding the highest standards of analysis. As such, we have revisited our methodology and included a more detailed explanation of the steps taken to avoid double dipping and ensure the integrity of our analyses in the revised manuscript.

“Although our study has found a new perspective, the analysis method still refers to and uses the traditional fNIR-based hyperscanning analyses (Liu et al., 2019; P¨arnamets et al., 2020; Reinero et al., 2021; Számadó et al., 2021; Solansky, 2011), which is generally accepted by the majority of fNIR-based hyperscanning researchers. For example, we would first identify significant channels through a one-sample t-test and then conduct further analyses, such as ANOVA or independent samples t-tests. Selective analysis is a powerful tool and is perfectly justified whenever the results are statistically independent of the selection criterion under the null hypothesis (Kriegeskorte et al., 2019). However, it may lead to double dipping and missing information. In this study, the absence of statistically significant TPJ activation in the analyzed data led to the TPJ being ignored. In the future, it should be made explicit in the analysis, and the reliability of the results should be ensured by appropriate statistical methods (e.g., cross-validation, independent data sets, or techniques to control for selective bias).” p.26-27

In several key analyses within this study (e.g., single-brain activation in the paragraph starting from L398, neural synchronization in the paragraph starting from L393), the TPJ is mentioned alongside the DLPFC. However, in subsequent detailed analyses, the TPJ is entirely ignored.

We thank the reviewer for your careful review and valuable comment. TPJ is referenced in certain analyses within this paper (as detailed in paragraphs L414 and L440); however, its role remains inadequately investigated and expounded upon in subsequent more intricate analyses. This is due to the absence of statistically significant TPJ activation in the analyzed data. As pointed out by the reviewer, limitations may exist in pursuing further analyses through ROIs, a point we also have addressed in the Discussion section (p.27).

The method for analyzing single-brain activation is unclear. Although it is mentioned that GLM (generalized linear model) was used, it is not specified what regressors were prepared, nor which regressor's β-values are reported as brain activity. Without this information, it is difficult to assess the validity of the reported results.

We have revised the relevant description to clarify the analyses of single-brain activation (p. 11)

While the model illustrated in Figure 7 seems to be interesting, for me, it seems not to be based on the results of this study. This is because the study did not investigate the causal relationships among the three metrics. I guess, Figure 5D might be intended to explain this, but the details of the analysis are not provided, making it unclear what is being presented.

We regret the confusion that has arisen. Firstly, as highlighted by the reviewer, the model depicted in Figure 7 is not directly derived from the causal analysis conducted in this study. Our investigation did not directly explore the causal relationships among the three indicators; instead, we constructed a model based on correlations and potential mechanisms. In the revised manuscript, we have explicitly stated that Figure 7 represents a descriptive model (p.22).

Regarding Figure 5D, the reviewer noted that while it may offer some explanatory value, it lacks the necessary analytical detail to elucidate the chart's significance clearly. We have clarified the details of the analysis in Figure 5 (pp.13-14). The model in Figure 5D suggested that the connection between the similarity in individual-collective performance and the correlation of brain activation, as well as whether the impact of each individual’s single-brain activation on the corresponding group’s GNS was regulated by their brain activation connectivity.

“Finally, we employed correlation and mediation analyses to assess if brain activation connectivity could explain the connection between individuals’ single-brain activation and the related group’s GNS. We examined the connection between the similarity in individual-collective performance and the correlation of brain activation, as well as whether the impact of each individual’s single-brain activation on the corresponding group’s GNS was regulated by their brain activation connectivity. We utilized the PROCESS tool in SPSS to investigate the proposed moderation effect. Specifically, we applied Model 1 with 5000 bootstrap resamples to examine the interaction between the independent variable (i.e., single-brain activation) and the moderator (i.e., brain activation connectivity) in predicting the dependent variable (i.e., GNS). It is noteworthy that prior to analysis, all variables in the moderation model were mean-centered to reduce multicollinearity and improve the interpretability of interaction terms.”  p.13-14

“Building on the above results, we have developed a two-in-one neural model that explains how group identification influences collective performance. This descriptive model aims to illustrate the potential interrelationships among these indicators and establish a conceptual framework to inspire forthcoming research endeavors.”  p.21

The details of the experiment are not described at all. While I can somewhat grasp what was done abstractly, the lack of specific information makes it impossible to replicate the study.

As suggested, we have clarified the details of the experiment in the manuscript.

(1) As stated in the public review, the details of the experiment are not described at all and while I can somewhat grasp what was done abstractly, the lack of specific information makes it impossible to replicate the study. In points a-e below, I list the aspects that I could not fully understand, but I am not asking for direct answers to these points. Instead, please provide a detailed description of the experiment so that it can be replicated.

Thank you for your suggestion; we have responded to each question sequentially and elaborated on the experiment specifics to ensure replicability.

(a) Please provide more detailed information about the Group Identification Task. How much did each participant speak (was there any asymmetry in the amount of speaking, and was there any possibility that the asymmetry influenced the identification rating)? Did the three participants interact in person, or online? Are they isolated from experimenters? How was the rating conducted, what I mean is that it's a PC-based rating?

We apologize for the lack of detail in our description of the procedures for the experiment.

For the first question, we draw upon previous studies concerning the manipulation of group identity while controlling the content of pre-task conversations. Specifically, the high-identity group engaged in self-introductions and identified similarities among the three members, whereas the low-identity group discussed topics related to the current semester's classes (Xie et al., 2023; Yang et al., 2020). Both discussions were conducted for the same duration of three minutes, ensuring that the number of exchanges between the two groups remained comparable. There was almost no asymmetry in the amount of speaking. We also conducted a manipulation check, which confirmed the effectiveness of our identity manipulation(pp.5-6).

Xie, E., Li, K., Gu, R., Zhang, D., & Li, X. (2023). Verbal information exchange enhances collective performance through increasing group identification. NeuroImage, 279, 120339.

Yang, J., Zhang, H., Ni, J., De Dreu, C. K., & Ma, Y. (2020). Within-group synchronization in the prefrontal cortex associates with intergroup conflict. Nature neuroscience, 23(6), 754-760.

“Both discussions were conducted for the same duration of three minutes, ensuring that the number of exchanges between the two groups remained comparable.”  p.5-6

For the second question，the three participants interacted offline in a face-to-face setting, while the experimenter remained outside the laboratory (p.6).

“The three participants conducted face-to-face offline interaction throughout the manipulation process.” p.6

For the third question, at the beginning of the experimental task, participants were isolated from the experimenters (p.6).

“In addition to explaining the next phase of the task and controlling the timer, experimenters would be isolated from participants.” p.6

For the last question, the rating of group identification was conducted through a questionnaire presented on participants’ phones (p.6).

“The questionnaire was presented on participants’ phones.” p.6

(b) The procedures of the Main Task are also unclear. For the Reading Information (5 min): How was the information presented? PC-based or paper-based? How were the participants seated? Did they read it independently?

We apologize for the missing details. We have included the following information in the article.

For the first and last question, each participant would get a piece of paper, which presents the common information and private information. They read independently. (p.6)

“Each participant would get a piece of paper, which presented the information. Participants could read independently.” p.6

About how the participants sat, the three participants sat around a table without partitions between each other. Only in the discussion stage, they could communicate face-to-face (p.6).

“They sat around a table without partitions between each other.” p.6

“In this process of discussion, the participants were able to communicate face-to-face and verbally.” p.6

(c) For Sharing Private Information: The authors stated they share text messages using Tencent Meeting. If so, how and with what devices? How was the information displayed on the screen? Were the participants even in the same room?

Thank you for your reminder. We have added more details now (p.6). Firstly, the experimenter sent the Tencent Meeting link to the participants. After the participants entered the meeting through their mobile phones, they could text the information they wanted to share in the chat box of the meeting. They were in the same room, with Tencent Meeting recording shared information, the participants could view them at any time.

“During the group sharing, participants entered Tencent Meeting via their mobile phones and were able to text their private information in the chat box to their group members for 5 minutes.” p.6

(d) For Discussing Information: It's a verbal interaction. How did they interact with others? What is the distance between them? I found a very small picture in Figure 8, but that is all information about experiment settings, that is provided by the authors.

We are sorry about the missing details. As we have explained in the article it’s a verbal communication, so participants could talk face to face in one room. We have included the following information in the article (p.6).

“Participants were sitting and communicating around a table. The distance between adjacent participants was about 15 cm, and the distance between face-to-face participants was about 40 cm. In this process of discussion, the participants were able to communicate face-to-face and verbally.” p.6

(e) For the Decision Process (5 min): How did they answer (What I mean is verbally, writing, or computer-based input), and how did the experimenters record these answers?

The questions were presented on paper, so the participants could write down their answers and experimenters could count the answers on paper. We have included the following information in the article(p.7).

“After discussion, all triads were given 5 minutes to answer the following questions (i) the probability of three suspects, 0%-100% for each suspect; (ii) the motivation and tool of crime; and (iii) deduced the entire process of crime. The three questions were presented on paper, allowing participants to write their answers directly on the same sheet. Subsequently, three independent raters used these paper questionnaires to record and calculate the scores for each group.” p.7

(2) I find the model presented in Figure 7 to be intriguing. Understanding why inter-brain synchronization occurs and how it is supported by specific single-brain activations or intra-brain functional connectivity is indeed a critical area for researchers conducting hyperscanning studies to explore. However, the content depicted in this model is not based on the results of this study. This is because the study did not investigate the causal relationships among the three metrics. I guess, Figure 5D might be intended to explain this, but the details of the analysis are not provided, making it unclear what is being presented. Please include a detailed explanation.

The specific answers are available on page 5 of our response letter.

(3) The analysis of single-brain activation analysis (and probably other analyses) focuses on the period from reading to making decisions (L237). Why was this entire interval chosen for analysis? Reading does not involve social interaction. As mentioned in a previous comment, the details of the tasks are unclear, so it's difficult to understand what was actually done in the reading period. Anyway, why were these different phases combined as the focus of analysis? Please clarify the reasoning behind this choice.

Thank you for your feedback. The decision to analyze the entire interval, spanning from reading to decision-making, was primarily made to grasp the continuum of information processing comprehensively. While reading itself lacks social interaction, it serves as the foundation for subsequent decision-making, during which participants' cognitive states and affective responses gradually evolve. Therefore, examining these two phases collectively enables a more thorough investigation into how information influences decision-making. Furthermore, considering the task details remain ambiguous, we aim to uncover the underlying cognitive and affective mechanisms through a holistic analysis.

(4) The method for analyzing single-brain activation is unclear. Please provide a detailed description of the analysis methods.

Thank you for your suggestion, we have added more details in the Method section (p.11).

“In the GLM model analysis, HbO was the dependent variable, and the regression amount was set to different task stages (a. Reading information, b. Sharing private information, c. Discussion information, d. Decision). After that, we convolved the regression factor with the Hemodynamic Response Function (HRF), and obtained the brain activation β value of each participant in each channel at different task stages through regression analysis.”  p.11

(5) In the periods of Reading Information and Sharing Private Information, there appears to be no social interaction between participants (Figure1D). However, Figure 6 shows an increase in brain activity correlation even during the first 10 minutes (it corresponds to the Reading and Sharing period). Why does inter-brain correlation (GNS, in this study) increase even though there is no interaction between participants? Please provide an explanation.

Sharing private information fosters interactive engagement, necessitating its exchange during Tencent Meetings to facilitate sharing. Previous research suggests that heightened correlations in brain activity can be attributed to (1) intrinsic cognitive processes, wherein participants display similar cognitive and emotional responses, fostering shared cognitive processing and brain activity synchronization despite limited external interaction; (2) emotional connections, as divulging private information elicits emotional responses that can be neurally correlated among individuals; and (3) environmental influences, where shared environments and contexts prompt neural interaction among participants even in the absence of direct social engagement. These factors collectively contribute to increased brain activity correlations without active interaction. Our primary focus, however, lies in the phase characterized by significant synchronized brain activity.

Minor Comments:

(6) Equation 1 Explanation: There is no explanation of Equation 1. It mentions Yi as the collective score, but what constitutes the collective score Yi is not defined in the manuscript. Additionally, while "i" is referred to as an item (in Line 196), the meaning of "item" is not clear. Therefore, the meaning of this equation is not understood.

We apologize for this confusion. We have added a description in the manuscript (p.9).

“In Eq.1, x is the individual score, y is the collective score (y is calculated from the three per capita scores), and i stands for the group number for the item. So, x_i means the individual score of participants in the _i group, and y_i means the collective score of the _i group. _d (x, y) r_epresents the distance from the individual to the collective score.”  p.9

(7) Equation 2 Explanation: There is no explanation for Equation 2. Please provide descriptions for all variables such as S, t, and w.

We have clearly stated the meaning of s, t, and w in the first edition of the manuscript article (p.12).

As shown in L291-293: Here, t denotes the time, s denotes the wavelet scale, 〈⋅〉 represents a smoothing operation in time, and W is the continuous wavelet transform (Grinsted, Moore, & Jevrejeva, 2004).

(8) Acronyms: Please define all acronyms upon their first appearance (e.g., CFI, TLI, RMSEA in L380).

We apologize for these mistakes, and we have added full explanations for abbreviations upon their first use (p.16).

“The mediation model demonstrated a satisfactory fit (CFI = 0.93, TLI = 0.93, RMSEA = 0.04) (CFI-Comparative Fit Index; TLI-Tucker-Lewis index; RMSEA-Root-Mean-Square Error of Approximation), suggesting that the perceived group identification of each individual affected the alterations in single-brain activations in the DLPFC, consequently leading to variations in their performance (β_a = 0.16, t = 2.20, p = 0.030; β_b = 0.26, t = 3.56, p < 0.001; β_c = 0.18, t = 2.34, p = 0.020) (Figure 3C).”  p.16

(9) Hyperscanning fMRI Studies: Since there are hyperscanning fMRI studies analyzing communication among three people (e.g., Xie et al., 2020, PNAS), it would be beneficial to cite this research. pnas.org/doi/pdf/10.1073/pnas.1917407117.

As suggested, we have cited this paper. (p.4)

(10) Line 272; Line 275: Should these references be to Benjamini & Hochberg (1995)?

As suggested, we have revised our citation.

(11) Research Objectives: The authors' aim seems to be understanding the relationship between Group Identification Level (High or Low), collective performance, and inter-brain synchronization (GNS). If so, shouldn't the results shown in Figure 6 illustrate how these differ between High and Low groups?

We are grateful to the reviewer for your insightful comment. This study aimed to investigate the impact of group identity levels on collective performance and interbrain synchronization. Our analysis primarily focused on inter-group disparities to elucidate the potential influence of varying levels of group identification on collective behavior and neural synchrony, as highlighted by the reviewer. It is important to note that the relationship between group identification levels and collective performance, as well as neural synchronization, may represent a continuous or correlational process, rather than a binary comparison between two distinct groups. Notably, we treated group identification as a continuous variable and, consequently, Figure 6 was designed to illustrate trends in the association between group identification levels and both collective performance and neural synchronization, without conducting significance tests between groups. We are confident that the depiction in Figure 6 effectively captures the evolving dynamics between group identification levels and both collective performance and neural synchronization.

(12) Figure 6 Star-Marker: What is the star marker shown in Figure 6? Please provide an explanation.

We apologize for this confusion. We have added this explanation to the article. (p.21)

“The red star sign indicates that at this time point, the neural signal began to increase significantly.” p.21

(13) Pearson's Correlation: Use "Pearson's correlation" instead of "Pearson correlation."

Thanks for your comments, we've changed Pearson correlation to Pearson's Correlation for a total of 10 places in the original text (pp. 9,11,13, 15,16, 19,23).

“Moreover, the Pearson’s correlation was used to examine the relationship between group identification_2 and collective performance.” p.9

“Subsequently, we used Pearson’s correlation analyses to investigate the relationship between single-brain activation and individual performance.” p.11

“Second, the Pearson’s correlation between GNS and collective performance was performed.” p.13

“Following that, we analyzed Pearson’s correlations between the original HbO data in the region related to individual and collective performance, denoted as brain activation connectivity (Lu et al., 2010).” p.13

“Subsequently, the Pearson’s correlation between the quality of information exchange and collective performance was assessed.” p.15

“Furthermore, the results of the Pearson’s correlation indicated that groups with higher group identification were more likely to exhibit better collective performance (r \= 0.38, p \= 0.003) (Figure 2B).” p.15

“The Pearson’s correlation and its associated analyses were based on the data from group identification_2. *p < 0.05.” p.16

“We first extracted the HbO brain activities related to individual performance (e.g., DLPFC, CH4) and collective performance (e.g., OFC, CH21) of each group member and conducted a Pearson’s correlation between the two.” p.19

“Subsequently, Pearson’s correlation was used to test whether individual differences in the similarity in individual-collective performance were reflected by DLPFC-OFC connectivity.” p.19

“Pearson’s correlation showed that the higher quality of information exchange, the better collective performance (r \= 0.36, p \= 0.007) (Figure 8C).” p.23

(14) MNI Coordinates: The MNI coordinates for each channel are listed in the supporting information. How were these coordinates measured? Were they consistent for all participants? Was MRI conducted for each participant to obtain these coordinates?

Thank you for your reminder, we have included the necessary instructions in the revised version. First, we need to clarify that we referred to previous literature to determine the placement of the optical probe plates. Following the completion of data collection, we utilized the Vpen positioning system to accurately locate the detection light poles, ultimately obtaining the MNI positioning coordinates. These coordinates were basically consistent for each participant. (p.8)

“For each participant, one 3 × 5 optode probe set (8 emitters and 7 detectors forming 22 measurement points with 3 cm optode separation, see Table S1 for detailed MNI coordinates) was placed over the prefrontal cortex (reference optode is placed at Fpz, following the international 10-20 system for positioning). The other 2 × 4 probe set (4 emitters and 4 detectors forming 10 measurement points with 3 cm optode separation, see Table S2 for detailed MNI coordinates) was placed over the left TPJ (reference optode is placed at T3, following the international 10-20 system for positioning). The probe sets were examined and adjusted to ensure consistency of the positions across the participants. After the completion of data collection, we utilized the Vpen positioning system to accurately locate the detection light poles, ultimately obtaining the MNI positioning coordinates.”  p.8

Author response:

The following is the authors’ response to the previous reviews

Joint Public Reviews:

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.

The short answer to the OEC bridges quantification is that in our last 2 studies combined, we observed bridges in 3/13 OB-OEC-transplanted rats versus 0/16 control rats (p=0.042 by two-sample proportion test; Thornton et al., 2018, Dixie, 2019). In addition to the new data on bridge formation shown in the current manuscript, our previous and most impressive data of serotonergic axons (5-HT-labeled, red) that crossed the entire lesion site is shown below (from Thornton et al., 2018). The image together with Supplemental video 1 (https://ars.els-cdn.com/content/image/1-s2.0-S0014488618302632-mmc1.mp4) show a reconstruction of multiple sections containing serotonergic axons that bridge the injury site in one OEC-transplanted, completely transected rat (1/5 OEC vs. 0/5 fibroblast-transplanted rat). The video also shows retrogradely-labeled Pseudo-rabies virus taken up by a few scattered neurons (green dots) within and above the lesion site, additional evidence suggesting axonal regeneration.

In addition to adding bridge quantification in the Results section, we now discuss quantified results on physiological and anatomical evidence of axon regeneration across the injury site from five of the six large spinal cord injury (SCI) studies conducted by the Phelps and Edgerton laboratories. Our studies used the most difficult SCI model, a complete, thoracic spinal cord transection in adult rats, followed by OB-OEC transplantation. This is the only model in which axon regeneration can be differentiated from axon sparing found in incomplete SCIs. An introductory paragraph now summarizes and references data generated from these studies that specifically addresses questions about how OECs modify the injury site and facilitate axonal outgrowth into and across into the lesion core. While relatively few axons cross the entire injury site to reach the caudal spinal cord, many more axons project into the injury site of OEC-transplanted rats compared to those in control rats. Quantification of axonal outgrowth into the lesion site of completely transected, OEC-transplanted rats from three previous long-term studies is now discussed in the Introduction. Based on both physiological and anatomical evidence reviewed from our previous work, we hope the editors and Reviewer agree that our previous studies have shown that OECs promote axonal outgrowth and modify the injury site.

Page 5, Introduction:

“Together with collaborators, we conducted six spinal cord injury studies in adult rats with a completely transected, thoracic spinal cord model followed by OB-OEC transplantation (Kubasak et al., 2008; Takeoka et al., 2011; Ziegler et al., 2011; Khankan et al., 2016; Thornton et al., 2018; Dixie, 2019). Results from five of our six studies showed physiological and anatomical evidence of axonal regeneration into and occasionally across the injury site. In 6-8-month-long studies, Takeoka et al. (2011) and Ziegler et al. (2011) reported physiological evidence of motor connectivity across the transection in OEC- but not media-transplanted rats. These experiments used transcranial electric stimulation of the motor cortex or brainstem to detect motor-evoked potentials (MEPs) with EMG electrodes in hindlimb muscles at 4- and 7-months post-transection. After 7 months, 70% of OEC-treated rats responded to stimulation with hindlimb MEPs (motor cortex, 5/20; brainstem 12/20; Takeoka et al, 2011). A complete re-transection above the original transection was carried out one month later and all MEPs in OEC-injected rats were eliminated. These results provide physiological evidence of axon conductivity across the injury site in OEC-treated rats. Additionally, three of our long-term studies evaluated anatomical axonal outgrowth of the descending serotonergic Raphespinal pathway into and through the injury site. Significantly more serotonergic-labeled axons crossed the rostral inhibitory scar border (Takeoka et al., 2011) or occupied a larger area within the injury site core (Thornton et al., 2018, Dixie, 2019) in OEC-transplanted rats than in fibroblast or media controls. In addition, significantly more neurofilament-labeled axons were found within the lesion core of OEC-transplanted versus control rats (Thornton et al., 2018, Dixie, 2019).”

Page 7, Results: We revised the sentence below and added additional information.

“We note, however, that such bridge formation is rare following severe spinal cord injury in adult mammals and was detected in 2 out of 8 OEC-transplanted rats and 0/11 media or fibroblast-transplanted controls in this study (Dixie, 2019). Combined with the 1/5 OEC-transplanted rats with axons crossing the injury and 0/5 fibroblast controls in our previous study (Thornton, 2018), we observed bridges in 3/13 OEC-transplanted rats vs 0/16 controls (p=0.042, two-sample proportion test). Bridge formation, in conjunction with the additional physiological and anatomical evidence of axonal connections across the injury site presented in our previous studies, strongly supports the capacity of OECs in neural repair.”

Page 46, Figure legend 1: We added statistical data to the legend

“Bridge formation across the injury site was observed in 2 of 8 OEC-transplanted and 0 of 11fibroblast- or media-transplanted spinal cord transected rats. Combined with the 1/5 OEC-transplanted rats with axons crossing the injury and 0/5 fibroblast controls in our previous study (Thornton, 2018), we observed bridges in 3/13 OEC-transplanted rats vs 0/16 controls (p=0.042, two-sample proportion test).”

(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.

Response: We agree that studying SCI repair in male rodents is very important as most people with these injuries are male. We did find one publication by Walker et al. (2019, Journal of Neurotrauma 36:1974-1984) that looked at sex differences in aged-matched male and female rats after a moderate contusion SCI. They examined a number of histological and functional features, and did not find many differences between the genders. Compared to studies of moderate SCI, studies using a completely transected spinal cord model must carry out manual bladder expressions a minimum of twice a day throughout the entire 5 to 7-month study in order to maintain kidney health. Because male urethras are much longer than those of females, males are much more likely that females to die from kidney disease during a complicated, long-term studies such as ours. Fortunately, most SCIs in humans are contusions rather than complete transections so an incomplete contusion model is most appropriate for studying sex differences. We modified the previous statement in our Discussion section as below.

Page 25, Discussion

“We acknowledge that in humans, males account for ~80% of spinal cord injuries (National Spinal Cord Injury Statistical Center, 2024) and sustain more serious urinary tract issues than females. We examined females in the current study due to practical experimental considerations, but it is necessary to examine males in future studies.”

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

(1) It is strongly recommended that some sort of quantification of bridging be included in the figures or in a table, whether this is the number of rats showing bridges, the percent area of OECs near the lesion site, or some other meaningful analysis.

As discussed in the response in Challenge section (1) above, we observed bridges in 3/13 OEC transplanted rats vs 0/16 controls across our two most recent studies. In addition, we added evidence of physiological and anatomical axonal connections across the injury site from our previous studies. We have added the additional information in the Introduction, Results, and Figure legend 1.

(2) It is recommended that clinical sex differences in spinal cord injury (with ~80% occurring in men) be acknowledged in the Discussion. This clinical fact could be directly mentioned without much justification.

See Challenge (2) above and addition to the Discussion on page 25.

(3) Figs. 1, 5, 6: There is still no quantification included for these figures, which detracts from the ability of readers to understand the context and importance of these results. It is recommended to include quantification for these figures.

Response regarding quantification associated with Figures 1, 5 and 6:

Regarding Figure 1: We have discussed the additions to the text of the Introduction, Results and the legend of Figure 1 in detail on pages 2-3 of this response. These are important new additions to our paper.

Regarding Figure 5: We added quantitative information regarding the analysis of Connective Tissue Growth Factor (Ctgf) expression in the injury site.

Page 10-11, Results:

“We found high levels of Ctgf expression in GFP-OECs (n=4 rats) that bridged much of the injury site and also detected Ctgf on near-by cells (Figure 5d, d1-2). GFP-labeled fibroblast transplantations (n=3 rats) served as controls and also expressed Ctgf.”

Page 36, Methods:

“To examine Ctgf expression in the spinal cord lesion site, we processed 1 slide per animal with ~6 equally-spaced sagittal sections throughout spinal cord from the Khankan et al. (2016) study. Our aim was to assess if transplanted OECs (n=4 rats) and transplanted fibroblasts (n=3 rats) express CTGF in the injury site.”

Regarding Figure 6: The statistics for Figure 6 are found on page 13 of the Results section and page 38 of the Methods section. We now added the statistics to the Figure 6 legend on page 49.

Page 13, Results:

“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 (~3/4^ths) 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 certain degrees of overlap, indicating similarities but also differences between the two classification methods.”

Page 38, Methods: Morphological analyses of Ki67 OEC subtypes

“To determine if OEC progenitor cells marked with Ki67 immunoreactivity have a distinctive morphology, purified and fixed OEC cultures from 4 rats were processed with anti- Ngfr^p75, anti-Ki67 and counterstained with Hoechst (Bis-benzimide, 1:500, Sigma-Aldrich, #B2261). Images were acquired from 7-10 randomly selected fields/sample using an Olympus AX70 microscope and Zen image processing and analysis software (Carl Zeiss). We distinguished the larger, flat ‘astrocyte-like’ OECs from the smaller, fusiform ‘Schwann cell-like’ OECs, and recorded their expression of Ngfr^p75 and Ki67. Cell counts from each field were averaged per rat and then averaged into a group mean ± SEM. A Student t-test was conducted to compare the effect of Ngfr^p75-labeled cell morphology and the proliferative marker Ki67. Statistical significance was determined by p < 0.05.”

Page 49, Figure 6 legend:

“Of the OEC progenitors that express Ki67, 76% ± 8 of them display low levels of Ngfr^p75 immunoreactivity and a “flat” morphology (g2, h2; green nuclei, arrowheads). The remainder of Ki67-expressing OECs express high levels of Ngfr^p75 and are fusiform in shape (24% ± 8%, n=4 cultures, Student-t test, p= 0.023).”

(4) Fig. 9: Quantification is still not included in the figure for these Western blots, although it is appreciated that the authors included some quantification in their response letter. Including this in the figure would provide clarification for the reader.

Thank you for your suggestion. We now add the quantification to figure 9, together with the methods used for western blot quantification and the figure legend.

Page 32, Methods:

“For quantification, ImageJ software (NIH) was used to analyze the densitometric data. Western blot images at 400, 300, and 150 kDa resolution were converted to grayscale followed by manually defining a Region of Interest (ROI) frame that captured the entire band in each lane using the "Rectangular" tool. The area of each selected band was measured by employing the same ROI frame around the band to record the integrated density, “Grey Mean Value”. Background measurements were similarly quantified, and background subtraction was performed by deducting the inverted background from the inverted band value. For relative quantification, target protein bands were normalized to the corresponding loading control (GAPDH) to derive normalized protein expression (fold change). Band intensities were quantified in triplicate for each sample. Data were analyzed with the Mann-Whitney U test to compare normalized protein expression between the Reln^-/- group and the other groups. A one-sided p-value was calculated to test the hypothesis that protein expression levels in the other groups are greater than those in the Reln^-/- group (negative control). Statistical significance was determined at p < 0.05. Analysis was performed using GraphPad Prism (version 9).”

Page 52, Figure legend 9f:

“(f) Quantitation of multiple isoforms of Reelin from 4-15% gradient gels. Positive and negative controls are Reln^+/+ and Reln^-/- mouse cortices. Both rat tissue from the ONL (n=3) and CM (n=9) contain more 400 and 300 kDa Reelin compared to the Reln^-/- mouse. Bars represent the standard deviation of the mean. One-sided Mann-Whitney U test was used to test that protein expression levels in the other groups are greater than those in the Reln^-/- group, indicative of significant expression of Reln in the test groups. *p < 0.05.”

Author response:

The following is the authors’ response to the previous reviews

Reviewer #1 (Public review):

Summary:

This study examines to what extent this phenomenon varies based on the visibility of the saccade target. Visibility is defined as the contrast level of the target with respect to the noise background, and it is related to the signal-to-noise ratio of the target. A more visible target facilitates the oculomotor behavior planning and execution, however, as speculated by the authors, it can also benefit foveal prediction even if the foveal stimulus visibility is maintained constant. Remarkably, the authors show that presenting a highly visible saccade target is beneficial for foveal vision as detection of stimuli with an orientation similar to that of the saccade target is improved, the lower is the saccade target visibility, the less prominent is this effect.

Strengths:

The results are convincing and the research methodology is technically sound.

Weaknesses:

It is still unclear why the pre-saccadic enhancement would oscillate for targets with higher opacity levels, and what would be the benefit of this oscillatory pattern. The authors do not speculate too much on this and loosely relate it to feedback processes, which are characterized by neural oscillations in a similar range.

We thank the reviewer for their assessment. We intentionally decided to describe the oscillatory pattern without claiming to be able to pinpoint its origin. The finding was incidental and, based on psychophysical data alone, we would not feel comfortable doing anything but loosely relating it to potential mechanisms on an explicitly speculative basis. In the potential explanation we provide in the manuscript, the oscillatory pattern would likely not serve a benefit–rather, it would constitute an innate consequence and, thus, a coincidental perceptual signature of potential feedback processes.

Reviewer #2 (Public review):

Summary:

In this manuscript, the authors ran a dual task. Subjects monitored a peripheral location for a target onset (to generate a saccade to), and they also monitored a foveal location for a foveal probe. The foveal probe could be congruent or incongruent with the orientation of the peripheral target. In this study, the authors manipulated the conspicuity of the peripheral target, and they saw changes in performance in the foveal task. However, the changes were somewhat counterintuitive.

We regret that our findings remain counterintuitive to the reviewer even after our extensive explanations in the previous revision round and the corresponding changes in the manuscript. We repeat that both the decrease in foveal Hit Rates and the increase in foveal enhancement with increasing target contrast were expected and preregistered prior to data collection.

Strengths:

The authors use solid analysis methods and careful experimental design.

Comments on revisions:

The authors have addressed my previous comments.

One minor thing is that I am confused by their assertion that there was no smoothing in the manuscript (other than the newly added time course analysis). Figure 3A and Figure 6 seem to have smoothing to me.

When the reviewer suggested that the “data appear too excessively smoothed” in the first revision, we assumed that they were referring to pre-saccadic foveal Hit and False Alarm rates, not to fitted distributions. As we state in the legend of Figure 3A (as well as in Figures 6 and S1), the “smoothed” curves constitute the probability density distributions of our raw data. Concerning the energy maps resulting from reverse correlation analyses, we described our proceeding in detail in our initial article (Kroell & Rolfs, 2022): 

“Using this method, we obtained filter responses for 260 SF*ori combinations per noise image (Figure 6 in Materials and methods, ‘Stimulus analysis’). SFs ranged from 0.33 to 1.39 cpd (in 20 equal increments). Orientations ranged from –90–90° (in 13 equal increments). To normalize the resulting energy maps, we z-transformed filter responses using the mean and standard deviation of filter responses from the set of images presented in a certain session. To obtain more fine-grained maps, we applied 2D linear interpolations by iteratively halving the interval between adjacent values 4 times in each dimension. To facilitate interpretability, we flipped the energy maps of trials in which the target was oriented to the left. In all analyses and plots,+45° thus corresponds to the target’s orientation while –45° corresponds to the other potential probe orientation. Filter responses for all response types are provided at https://osf.io/v9gsq/.”

We have added a pointer to this explanation to the current manuscript (see line 836).

Another minor comment is related to the comment of Reviewer 1 about oscillations. Another possible reason for what looks like oscillations is saccadic inhibition. when the foveal probe appears, it can reset the saccade generation process. when aligned to saccade onset, this appears like a characteristic change in different parameters that is time-locked to saccade onset (about a 100 ms earlier). So, maybe the apparent oscillation is a manifestation of such resetting and it's not really an oscillation. so, I agree with Reviewer 1 about removing the oscillation sentence from the abstract.

While we understand that a visible probe will result in saccadic inhibition (White & Rolfs, 2016), we are unsure how a resetting of the saccade generation process should manifest in increased perceptual enhancement of a specific, peripheral target orientation in the presaccadic fovea. Moreover, as we describe in our initial article (Kroell & Rolfs, 2022), we updated the background noise image every 50 ms and embedded our probe stimulus into the surrounding noise using smooth orientation filters and raised cosine masks to avoid a disruptive influence of probe appearance on movement planning and execution (Hanning, Deubel, & Szinte, 2019). And indeed, we demonstrated that the appearance of the foveal probe did not disrupt saccade preparation, that is, did not increase saccade latencies compared to ‘probe absent’ trials in which no foveal probe was presented (see Kroell & Rolfs, 2022; sections “Parameters of included saccades in Experiment 1” and “Parameters of included saccades in Experiment 2”). In the current submission, saccade latencies in ‘probe present’ trials exceeded saccade latencies in ‘probe absent’ trials by a mere 4.7±2.3 ms. Additionally, to inspect the variation of saccade execution frequency directly, we aligned the number of saccade generation instances to the onset of the foveal probe stimulus (see Author response image 1). In line with what we described in a previous paradigm employing flickering bandpass filtered noise patches (Kroell & Rolfs, 2021; 10.1016/j.cortex.2021.02.021), we observed a regular variation in saccade execution frequency that reflected the duration of an individual background noise image (50 ms in this investigation). In other words, the repeated dips in saccadic frequency are likely caused by the flickering background noise and not the onset of the foveal probe which would produce a single dip ~100 ms after probe onset. Given these results, we do not see a straight-forward explanation for how the variation of saccade execution frequency in 20 Hz intervals would boost peripheral-to-foveal feature prediction before the saccade in ~10 Hz intervals. Nonetheless, we removed the sentence referencing oscillations from the Abstract.

Author response image 1.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Overall, The authors did a good job in addressing the points I raised. Two new sections were added to the manuscript, one to address how the mechanisms of foveal predictions would play out in natural viewing conditions, and another one examining more in depth the potential neural mechanisms implicated in foveal predictions. I found these two sections to be quite speculative, and at points, a bit convoluted but could help the reader get the bigger picture. I still do not have a clear sense of why the pre-saccadic enhancement would oscillate for targets with higher opacity levels, and what would be the benefit of this oscillatory pattern. The authors do not speculate too much on this and loosely relate it to feedback processes, which are characterized by neural oscillations in a similar range.  

Please see our response to ‘Weaknesses’.

I still find this a loose connection and would suggest removing the following phrase from the abstract "Interestingly, the temporal frequency of these oscillations corresponded to the frequency range typically associated with neural feedback signaling". 

We have removed this phrase.

Finally, the authors should specify how much of this oscillation is due to oscillations in HR of cong vs. oscillations in HR of incongruent trials or both.

We fitted separate polynomials to congruent and incongruent Hit Rates instead of their difference. Peaks in enhancement relied on both, oscillatory increases in congruent Hit Rates and simultaneous decreases in incongruent Hit Rates. In other words, enhancement peaks appear to reflect a foveal enhancement of target-congruent feature information along with a concurrent suppression of target-incongruent features. We added this paragraph and Figure 4 to the Results section.

Additional changes:

Two figures had accidentally been labeled as Figure 5 in our first revision. We corrected the figure legends and all corresponding figure references in the text.

Author response:

The following is the authors’ response to the original reviews

Recommendations for the authors:

Reviewer #1 (Recommendations for the Authors):

The interpretation of results obtained with opto-Treacle (related to Figure 2C) may be expanded.

We thank the reviewer for their insightful comment regarding the interpretation of the results obtained with opto-Treacle. We understand the concern that the difference in the size of the condensates formed by opto-Treacle (Figure 2C) compared to Treacle-2S or other constructs may raise questions about the role of tetramerization in driving condensate formation, as 2S is known to tetramerize while FusionRed is not susceptible to multimerization.

To address this concern, we emphasize that we have demonstrated that overexpressed Treacle forms large condensates even in the absence of any fluorescent protein, as included in the revised manuscript. This observation supports the conclusion that Treacle's ability to form condensates is intrinsic and does not depend on the multimerization capacity of the fluorescent tag.

We believe that the observed difference in condensate size between opto-Treacle and Treacle-2S, Treacle-GFP, or untagged Treacle arises primarily from the time available for condensate assembly. Opto-Treacle condensation occurs rapidly, within approximately 10 seconds of blue light illumination, whereas Treacle-2S, Treacle-GFP, or untagged Treacle undergo condensation over the extended period of 24–48 hours of protein overexpression. This temporal difference likely accounts for the disparity in condensate size, as longer assembly times allow for larger and more mature condensates to form.

Given this reasoning, we consider it unnecessary to further emphasize the size differences in the main text of the article, as we believe the underlying explanation is clear and supported by the data. Nonetheless, we are open to incorporating additional clarifications if the reviewer deems it necessary.

The authors might reconsider referring to Treacle as a scaffold. Ultimately, the scaffold for the nucleolus is the rDNA with its bound proteins. Scaffold proteins, by definition, bind multiple protein partners and facilitate the formation of multiprotein complexes, a role not really attributed to homotypic LLPS.

We thank the reviewer for raising this important point regarding the use of the term "scaffold" in relation to Treacle. We fully acknowledge that rDNA, along with its associated protein complexes, serves as the primary structural scaffold for the nucleolus. However, we believe that referring to Treacle as a scaffold is appropriate and justified within the specific context of our study.

First, we emphasize that we describe Treacle as a scaffold specifically for nucleolar fibrillar centers (FCs), rather than for the nucleolus as a whole. This distinction is important, as our work focuses on the role of Treacle in organizing FC components, rather than the broader structural organization of the nucleolus.

Second, as the reviewer notes, scaffold proteins are defined by their ability to bind multiple protein partners and facilitate the formation of multiprotein complexes. Our findings demonstrate that Treacle's condensation properties promote the binding and retention of key rDNA-associated protein partners, including RPA194, UBF, and Fibrillarin, within the FCs. This activity aligns with the functional definition of a scaffold protein, as Treacle supports the spatial organization and cooperative interactions of FC components essential for rRNA transcription and processing. Therefore, while we appreciate the reviewer's observation regarding the central role of rDNA as a nucleolar scaffold, we maintain that the use of the term "scaffold" to describe Treacle's role in organizing FCs is consistent with its demonstrated functional properties.

If authors decide to add the "Ideas and Speculation" subsection to their Discussion, it may be interesting to discuss the following outstanding questions: does Treacle undergo homotypic or heterotypic LLPS? Does its overexpression favor homotypic interactions? How does it segregate FC and DFC compartments -by exclusion? How does phase-separated Treacle interact with other proteins?

We thank the reviewer for these insightful questions. While we believe that adding a dedicated "Ideas and Speculation" subsection would be redundant, we have already addressed the questions regarding Treacle’s homotypic or heterotypic LLPS and its interactions with other proteins in the revised "Discussion" section. Additionally, we have included a new section in the manuscript specifically focused on investigating the role of Treacle condensation in its interactions with protein partners, further expanding on these points.

In Materials and Methods, smFISH section -"probes were designed as described (Yao et al, 2019) and labeled with FITS on the 3'ends" - was it meant to say FITC (i.e. Fluorescein)?

We thank the reviewer for catching this error. This was indeed a typo, and we have corrected it to "FITC (i.e., Fluorescein)" in the revised text.

Reviewer #2 (Recommendations for the Authors):

Regarding recombinant Treacle, the main concern is that the authors may not be observing the condensation of Treacle itself. The quality of the purchased recombinant Treacle is unclear (this reviewer could not find Treacle listed on the vendor website despite using the supplied catalog number or vapors search terms). Furthermore, it is not clear if the condensates observed are Treacle or potentially the Dextran crowder. Only small percentages (>1%-5%) of either Dextran or PEG are needed to induce phase separation in two-component mixtures of these polymers. PEG may be in the Treacle storage butter. In addition to clarifying the State of recombinant Treacle, these concerns could be further assuaged by direct visualizing of Treacle forming condensates (via fluorescent n-terminal tagging) and filling in more of the phase space to observe the loss of condensates at a threshold concentration of Treacle. In general, the gold standard for establishing condensation of a given protein is mapping the full binodal phase diagram diagram of the protein. Understanding that protein is a limited resource, most groups simply map the lower concentration arm of the binodal, and this is sufficient to characterize a protein as having intrinsic condensation behavior. A similar mapping effort of Treacle would be welcomed. 

We thank the reviewer for their thoughtful comments and for highlighting concerns regarding the interpretation of our experiments with commercial recombinant Treacle. We recognize the importance of ensuring that the observed condensation properties are intrinsic to Treacle and not influenced by potential contaminants, storage buffer components, or tags on the protein.

To address these concerns, we have re-evaluated the condensation properties of Treacle using a recombinant fragment independently purified in our laboratory. Specifically, we expressed and purified a Treacle fragment (amino acids 291–426), which includes two S/E-rich low-complexity regions (LCRs) and two linker regions, in E. coli. The protein was expressed as a TEV-cleavable maltose-binding protein (MBP) fusion, purified under native conditions via amylose resin, and subjected to TEV cleavage. This was followed by ion-exchange chromatography and extensive dialysis to remove any remaining impurities. These additional steps ensured that the purified Treacle fragment was of high purity and free from confounding components, such as polyethylene glycol (PEG). We have included detailed descriptions of this protocol in the revised manuscript.

Using this purified Treacle fragment, we confirmed its intrinsic condensation behavior in vitro. In the presence of 5% PEG8000 as a crowding agent, the fragment formed liquid-like condensates that exhibited spherical morphology and dynamic fusion events, key hallmarks of liquid-liquid phase separation (LLPS). Additionally, we demonstrated that the condensation of this Treacle fragment was sensitive to changes in pH and salt concentration but unaffected by 1,6-hexanediol treatment, suggesting that the condensates are stabilized predominantly by electrostatic interactions (Fig. 4B of the revised manuscript). Importantly, these findings provide robust evidence that Treacle possesses intrinsic phase-separation properties. All results from the commercial Treacle protein used in the initial version of the manuscript have been replaced with data obtained using this independently purified recombinant fragment.

We undestand that the condensation behavior of the fragment may not fully capture the behavior of full-length Treacle. Nevertheless, the in vitro experiments provide valuable mechanistic insights into the biophysical properties of Treacle. Furthermore, as emphasized in the revised manuscript, our study primarily focuses on understanding the condensation and functional role of Treacle in a cellular context, where we observe its critical involvement in organizing nucleolar structure and regulating rRNA transcription. These cellular experiments highlight the biological relevance of Treacle’s condensation behavior.

With regard to mapping the binodal phase diagram of Treacle, we concur with the reviewer that such an effort would be ideal for a more comprehensive characterization of Treacle’s condensation properties. However, the limited availability of purified protein currently precludes a detailed mapping effort. Despite this limitation, we believe the qualitative assessments of Treacle’s condensation under varying conditions, now included in the revised manuscript, sufficiently demonstrate its intrinsic ability to phase-separate.

In conclusion, we are grateful for the reviewer’s feedback, which has allowed us to refine our methodology and strengthen the evidence supporting the intrinsic condensation properties of Treacle. We are confident that the revised manuscript provides a robust and thorough characterization of Treacle’s phase-separation behavior and its functional role in the cell, addressing the reviewer’s concerns. Thank you for your constructive recommendations, which have significantly improved the quality of our work.

Replacing 'liquid-phase' and 'liquid' with 'liquid-like' would make the language consistent with other papers in the field and more accurately reflect the degree of material state analysis carried out in the study.

We thank the reviewer for this insightful recommendation. In response to the suggestion, we have revised the manuscript to replace the terms "liquid-phase" and "liquid" with "liquid-like" throughout the text. This change ensures consistency with terminology commonly used in the field and more accurately reflects the degree of material state analysis performed in our study. We believe this adjustment improves the clarity and precision of our findings, aligning the manuscript with standard practices in the field. Thank you for helping us enhance the quality of the presentation.

The 'unclear' nature of the condensation behavior of the FC phase of the nucleolus is listed as a motivation for carrying out the study in the introduction; the authors could note here two recent papers that have investigated the nature of FC condensation: Jaberi-Lashkari et al. 2023 and King et al. 2024. The reviewer notes that while these were both pre-printed in late 2022, they were only recently published.

We thank the reviewer for bringing these recent studies to our attention. In response to the suggestion, we have cited the papers by Jaberi-Lashkari et al. (2023) and King et al. (2024) in both the introduction and discussion sections of the revised manuscript. These references are highly relevant to the context of our study and provide valuable insights into the condensation behavior of the FC phase of the nucleolus. We agree that incorporating these works strengthens the framing of our study and situates it more effectively within the broader field. Thank you for this constructive recommendation.

The statement that Treacle is "the main molecule present in the FC" is a substantial claim that does not need to be made to promote the author's case, nor is it well supported by the provided reference (Gal et al., 2022).

We thank the reviewer for pointing out this overstatement in our original manuscript. In response, we have revised the text to provide a more accurate and well-supported description. Specifically, we have replaced the claim that Treacle is "the main molecule present in the FC" with a statement highlighting its direct interactions with UBF and RNA Pol I, as well as its colocalization with these proteins within the FC. This revision ensures alignment with the provided references and more accurately reflects the current understanding of Treacle's role in the FC. We appreciate the reviewer's attention to this detail, which has helped us improve the clarity and accuracy of our manuscript.

The statement that "Treacle is one of the most intrinsically disordered proteins" is vague and unnecessarily grand. Treacle is a fully intrinsically disordered protein; these comprise 5% of the human proteome (Tsang et al. 2020), so Treacle is, indeed, unusual in that regard.

We thank the reviewer for highlighting the vague and unnecessarily broad nature of the original statement. In response, we have revised the text to provide a more precise and accurate description of Treacle's structural properties. Specifically, we replaced the claim that "Treacle is one of the most intrinsically disordered proteins" with the statement that "According to protein structure predictors (e.g., AlphaFold, IUPred2, PONDR, and FuzDrop), Treacle is a fully intrinsically disordered protein." This wording reflects the unique nature of Treacle while remaining scientifically accurate and supported by reliable computational predictions. We appreciate the reviewer's feedback, which has allowed us to improve the rigor and clarity of our manuscript.

A comment on the implications of the immobile pool of Treacle (which appears to be ~50% in WT and across a range of mutants) would be welcome. Additionally, the limitations of FRAP for interrogating material properties of condensed material in living systems are provided in Goetz and Mahamid, 2020. In this paper, the authors review instances where the ultrastructure of condensate is known and where FRAP data is available. They show that crystalline assemblies can recover faster than apparently liquid, spherical assemblies. A comment in the text about how these limitations apply to this study would be welcome.

We appreciate the reviewer’s insightful comments regarding the interpretation of the immobile pool of Treacle and the limitations of FRAP for characterizing material properties in living systems. As noted in our response to the public review, we believe the ~50% recovery rate after photobleaching observed in our experiments is best explained by the redistribution of Treacle molecules within the condensate, rather than significant exchange with the surrounding phase. This interpretation is strongly supported by the full- and half-FRAP analyses included in the revised manuscript, which demonstrated internal mixing dynamics within the condensates.

There appears to be a typo in the following sentence: "The highly positively charged CD serves as the nucleation center for RD but exhibits ambivalent phase properties, transitioning from LLPS to LSPS in the absence of rRNA." The LLPS to LSPS behavior was observed for mutants to the central domain (RD), not the c-terminal domain (CD).

Throughout the authors report single snapshots of representative cells and single line traces. Analysis of the key morphological feature across the population of cells would help the reader understand how widespread the observed phenotype is.

We thank the reviewer for raising this important point regarding the representation of morphological features across the cell population. To address this concern, we have included widefield micrographs of cell fields in the revised figures to provide a more comprehensive view of the phenotypes observed.

The statement that "The phase behavior of polymers is determined by interactions through associative motifs, referred to as stickers, separated by spacers, which are not the primary driving forces for phase separation" could be improved by pointing out that this is potentially incomplete for describing the kind of condensation that highly charged polymers undergo. The high charge and charge segregation of Treacle suggest that it is a blocky polyampholyte and that it condenses by coacervation. Models of associative polymers can be useful for describing coacervation, however, the driving forces for coacervation are less understood and have been proposed to include an entropic component (see Sathyavageeswaran et al. 2024, Sing and Perry 2020 and work from their groups as well as the Obermayer (Columbia) and Terrell (U. Chicago) Groups).

We thank the reviewer for highlighting this important aspect of the phase behavior of charged polymers and for suggesting relevant references. In response, we have revised the discussion section of the manuscript to include a more nuanced explanation of the condensation mechanisms for highly charged polymers such as Treacle. Specifically, we now describe Treacle as a blocky polyampholyte, suggesting that its condensation behavior may be driven by coacervation mechanisms.The relevant references have been added to the discussion section of the revised manuscript.

In addition to the above, the authors may consider citing two recent publications from the Pappu group (King et al. Cell 2024 and King et al. Nucleus 2024) that directly investigate the condensation potential of K-rich and E/D-rich' grammars' on nucleolar proteins and show that, like the authors, the K-rich region is essential for localization and is conserved across nucleolar proteins.

We thank the reviewer for bringing these relevant publications to our attention. The suggested references from the Pappu group (King et al., Cell 2024, and King et al., Nucleus 2024) have been added to the introduction and discussion sections of the revised manuscript, and their findings have been appropriately integrated into our analysis.

The authors could consider replacing the use of LLPS with a more generic term such as "condensation" or "biomolecular condensation." LLPS of polymers is a segregative transition driven by its incompatibility with the surrounding solvent. As indicated, Treacle is likely to be undergoing some form of coacervation (which is predominantly an associative tradition), which can be genetically described as condensation. See Pappu et al. 2023 for more details.

We thank the reviewer for their insightful suggestion. Following the reviewer's recommendation, we have replaced the term "LLPS" with "condensation" or "coacervation" throughout the manuscript, where appropriate. Additionally, we have referenced Pappu et al. (2023) and other to provide further context and clarity regarding the distinctions between these terms.

The authors cite Yao et al. 2019, but do not cite the follow-up study (Wu et al. 2021) or provide a statement on how the Chan group finds a role for the RGG domain of FBL in keeping the certain canonical markers of the FC and DFC de-mixed.

We thank the reviewer for pointing out these important references. The relevant citations, including Wu et al. (2021), have been added to the manuscript.

Reviewer #3 (Recommendations for the Authors):

The following comment is true but could be broadened to include examples of structured regions promoting biomolecular condensation. "In biological systems, phase separation is mainly a characteristic of multivalent or intrinsically disordered proteins (Banani et al, 2017; Shin & Brangwynne,2017; Uversky, 2019)."

We have expanded the statement as recommended by the reviewer: "In biological systems, phase separation is facilitated by a combination of multivalent interactions mediated by intrinsically disordered proteins and site-specific interactions that drive percolation."

Related to Figure 1.

The authors report Treacle-dependent EU incorporation (Figure 1D), but are there any changes more broadly to nucleolar number or size as a consequence? How do the authors interpret that the quantitative effect of AMD treatment is more extreme than Treacle depletion (Figure 1E).

We thank the reviewer for raising these important points. Regarding nucleolar number and morphology, we did not observe a change in the number of nucleoli upon Treacle depletion. However, nucleoli appeared more regularly rounded under these conditions, which we interpret as a consequence of the decreased rDNA transcription activity caused by Treacle depletion. A similar rounding of nucleoli is also observed upon actinomycin D (AMD) treatment, which is consistent with reduced transcriptional activity.

As for the more pronounced effect of AMD compared to Treacle depletion on EU incorporation, this can be explained by the fundamentally different mechanisms through which these conditions affect transcription. Treacle depletion reduces the local concentration of transcription factors at rDNA sites, thereby impairing transcription initiation and elongation to a certain extent. However, under Treacle depletion, RNA polymerase I still retains the ability to bind to the promoter and support a residual level of transcription. In contrast, AMD acts as a potent intercalator in GC-rich regions of rDNA, physically blocking the ability of RNA polymerase I to move along rDNA, resulting in near-complete cessation of rRNA synthesis.

Related to Figure 2.

The authors observe that AMD leads to coalescence of individual Treacle-2S+ bodies (e.g. Figure 2E) - does this suggest that ongoing rRNA transcription is required to prevent such events?

Thank you for your thoughtful question. Indeed, our observations strongly suggest that ongoing rRNA transcription is required to prevent the coalescence of Treacle-2S+ bodies, as observed upon AMD treatment. This interpretation aligns with the findings of Tetsuya Yamamoto et al., who demonstrated that nascent ribosomal RNA (pre-rRNA) acts as a surfactant to suppress the growth and fusion of fibrillar centers (FCs) in the nucleolus. Their work highlighted that nucleolar condensates formed via liquid-liquid phase separation (LLPS) tend to grow to minimize surface energy, provided sufficient components are available. However, the transcription of prerRNA stabilizes FCs by maintaining multiple microphases, preventing coalescence unless transcription is inhibited.

According to Yamamoto et al., nascent pre-rRNAs tethered to FC surfaces by RNA Polymerase I generate lateral pressure that counteracts interfacial tensions, effectively suppressing FC fusion. This activity is analogous to the surfactant properties of molecules in physical systems. When transcription is inhibited (e.g., by AMD), the loss of nascent rRNA allows condensates to coalesce, consistent with the behavior we observe.

We further propose that the AMD-induced coalescence of Treacle-2S+ bodies reflects the loss of this surfactant-like effect, as transcriptional activity ceases. This theory is also supported by the observation that Treacle condensates in the nucleoplasm, where rRNA transcription is absent, form larger structures. Collectively, these insights highlight the critical role of ongoing rRNA transcription in maintaining the structural integrity and dynamic organization of nucleolar substructures.

Related to Figure 3.

In the figure panels B-H the DAPI signal in gray obscures the Treacle localization, especially in Figure 3H. A non-merged image for each of these examples for the Treacle localization would be very helpful.

We thank the reviewer for this observation. To address this, we have included wide-field images without the DAPI overlay for the deletion mutant lacking the 1121-1488 region. These are now presented in Supplementary Figure S5G of the revised manuscript.

Related to Figure 5.

Only a single representative nucleus is shown in the PLA analysis presented in Figure 5B.

Quantification to assess the robustness of this response with the addition of VP16 is needed. The authors use ChIP and immunocytochemistry as orthogonal methods but it would be best to therefore show both for each manipulation that is performed - the immunostaining of TOPBP1 in the Treacle KD cells in S5A should be in the main Figure 5 to complement transformation of constructs as in Figure 5D.

We appreciate the reviewer’s comment. To address this, we performed a quantitative analysis of PLA fluorescence signals in control and etoposide-treated cells, and the results are now presented in Supplementary Figure S8C. Additionally, as recommended, we have transferred the results of the immunocytochemistry of TOPBP1 in Treacle KD and Treacle KN cells to the main figure, now included as Figures 7D-E in the revised manuscript.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public Review):

Summary.

In this meticulously conducted study, the authors show that Drosophila epidermal cells can modulate escape responses to noxious mechanical stimuli. First, they show that activation of epidermal cells evokes many types of behaviors including escape responses. Subsequently, they demonstrate that most somatosensory neurons are activated by activation of epidermal cells, and that this activation has a prolonged effect on escape behavior. In vivo analyses indicate that epidermal cells are mechanosensitive and require stored-operated calcium channel Orai. Altogether, the authors conclude that epidermal cells are essential for nociceptive sensitivity and sensitization, serving as primary sensory noxious stimuli.

Strengths.

The manuscript is clearly written. The experiments are logical and complementary. They support the authors' main claim that epidermal cells are mechanosensitive and that epidermal mechanically evoked calcium responses require the stored-operated calcium channel Orai. Epidermal cells activate nociceptive sensory neurons as well as other somatosensory neurons in Drosophila larvae, and thereby prolong escape rolling evoked by mechanical noxious stimulation.

Weaknesses.

Core details are missing in the protocols, including the level of LED intensity used, which are necessary for other researchers to reproduce the experiments. For most experiments, the epidermal cells are activated for 60 s, which is long when considering that nocifensive rolling occurs on a timescale of milliseconds. It would be informative to know the shortest duration of epidermal cell activation that is sufficient for observing the behavioral phenotype (prolongation of escape behavior) and activation of sensory neurons.

(1) We agree with the reviewer that the LED intensity is an important detail of the experimental paradigm. We updated the methods to include intensity measurements for the stimuli used throughout the manuscript.

(2) The Reviewer asks about the shortest duration of epidermal cell activation sufficient for observing the behavior phenotype. We note in the manuscript that behavioral responses to optogenetic epidermal stimulation are apparent within 2 seconds of stimulus (see Figure 2F); this is consistent with our calcium imaging data in which C4da response reaches its maximum within 2-3 sec of stimulation.

Reviewer #1 (Recommendations):

(1) The epidermal cells in this study are activated for 60 s. In the real world, the nociceptive stimulation (a poke, such as penetration by the ovipositor of a parasitic wasp) that evokes escape rolling is short. Does optogenetic activation of 1 s or less still evoke rolling? For example, it is unclear in Figure 4K how long the epidermal cells need to be activated before the poke stimulus prolongs rolling. Is it possible to test behavior and GCaMP activity in sensory neurons when epidermal cells are briefly (1 second) activated?

As described above, behavioral responses to optogenetic epidermal stimulation are apparent within 2 seconds of stimulus (see Figure 2F); this is consistent with our calcium imaging data in which C4da response reaches its maximum within 2-3 sec of stimulation. The kinetics are consistent with a role for epidermal cells in modulating neuronal responses to nocifensive stimuli, and similar to the response kinetics observed in mammalian epidermal cells that modulate neuronal touch and pain responses  (Maksimovic et al., 2014; Woo et al., 2014; Mikesell et al., 2022).

(2) The protocol for optogenetic screening states that the authors used a 488-nm LED. Why was a 488-nm LED used instead of the 610-nm LED for Chrimson activation? No information (except figure 4K) about the light intensity is provided in the figure legend or the protocol section. Please state the LED intensity used for all optogenetic experiments (GCaMP imaging, behavioral experiments, etc.).

We used 488 nm light for the initial screen for technical reasons. The screen was conducted by students at the MBL Neurobiology course (hence the affiliation; student authors are included in the manuscript), and the only LED available to us at that time delivered insufficient illumination at longer wavelenths to be useful. We chose to include the student’s data because (1) we found that the 488 nm light alone did not induce rolling in our setup, (2) we repeated and extended the studies with the epidermal drivers using a higher resolution imaging platform and longer wavelength stimulation (all studies other than Fig. 1), and (3) we observed qualitatively similar results when we repeated stimulation with all drivers using 561 nm light.

We agree that the LED intensity is an important detail of the experimental paradigm. We updated the methods to include intensity measurements for the stimuli used throughout the manuscript. We also include the intensities here:

- 30 μW/mm^2 for calcium imaging experiments Fig 3B-E, Fig 4A, Fig 3S1A-D, Fig 4S1A

- 300 μW/mm^2 for behavior studies in Fig 2B-E, Fig 1S6, Fig 2S1, Fig 3E-F, Fig 3S2A-C

- 25 μW/mm^2 for behavior studies in Fig 4E-J

- 1.16 μW/mm^2 for behavior studies in Fig 4K

(3) Lines 150 - 152: Although the authors refer to "a stereotyped behavior sequence" in Fig 2D, there are no data supporting this claim in Fig 2. Rather, the data appear to represent proportions of different types of behavior at each time point, rather than behavior sequences. If the authors wish to claim that the data show stereotyped behavior sequences, they should analyze the data using a different method (e.g., Markov models).

We agree that in the absence of additional analysis we should avoid commenting on stereotypy of behavior sequences; we therefore adjusted the text to reflect the tendency of nociceptive behaviors to precede non-nociceptive behaviors. The raster plots shown in Supplemental Fig. 2A illustrate this point: in larvae exhibiting nociceptive behaviors, these behaviors appear first, followed by backing and frequently freezing. As one quantitative readout of this sequence we show that the latency of rolling (nociceptive) is shorter compared with backing or freezing (non-nociceptive) (Fig. 2F, Fig. S2G).

(4) Figure 3A-E: a cursory glance at the data suggests that the most responsive sensory neurons are C1da, with all sensory neurons activated. However, at the behavioral level, only some sensory neurons are activated. If all sensory were activated by Chrimson, what behavioral phenotypes would the authors expect to see? Would it be the same as epidermal activation?

The Reviewer raises an interesting question, but we intentionally avoid comparing the response properties among sensory neurons because of differences in driver strength. Likewise, extrapolating “activation” at the behavioral level is exceedingly difficult if/when multiple sensory neurons are simultaneously activated. In response to the Reviewer’s specific question, when all da neurons are activated simultaneously, larvae largely exhibited hunching rather than rolling (Hwang et al., 2007). We find that epidermal stimulation rarely elicits hunching; instead, epidermal stimulation generally triggers nocifensive behaviors followed by non-nocifensive behaviors such as backing and freezing, suggesting an order or priority in neurons activated by epidermal cells (or different response times). Defining the mechanisms by which epidermal cells communicate with different types of sensory neurons is therefore a top priority for future studies.

(5) Figure 3S2; The behavior phenotypes between Fig. 3E, F and Fig 3S2 seems a slightly different. I suggest adding some comments in different behavior phenotype depending on the different GAL4. Specifically, is there increased freezing in some genotypes (e.g., ppk-LexA or NompC-lexA)? Can you show this without TNT data? Is this a background effect or specific GAL4 phenotype?

We currently do not have the driver-only control for this experiment, but our effector-only control experiment (see Fig. 3S2A) suggests that larvae carrying the AOP-TNT insertion exhibit enhanced nociceptive behavioral responses. This point is addressed in our manuscript by the following (copied from the figure legend):

“We note that although baseline rolling probability is elevated in all genetic backgrounds containing the AOP-LexA-TnT insertion, silencing C4da and C3da neurons significantly attenuates responses to epidermal stimulation.”

(6) Calcium-free solution is used in Figure 3. Why do the authors still observe calcium influx? Does this mean that internal calcium stores are released? If so, does the calcium influx represent an action potential? How do the authors focus their LED stimulation to activate epidermal cells and avoid activation of the imaging laser?

The specimens were imaged in calcium-free solution to minimize movement artifacts. However, the CNS is wrapped by glial cells and over short timescales such as those used for the imaging we speculate that extracellular calcium persists in the CNS.

(7) It is unclear when animals begin to crawl after the epidermal cells are mechanically stimulated. How do the authors distinguish between peristaltic crawling and a poke by Orai receptors? Although the in vitro experiments beautifully show radial tensions, it is unclear to what extent A-P axis tension (peristaltic crawling) and radial tension (poke) differ. It might be helpful to explain in the discussion section how epidermal cells are selectively activated.

The Reviewer raises an interesting question about the types and thresholds of forces required to elicit epidermal responses. We cannot eliminate the possibility that peristaltic crawling (or crawling through a 3D substrate) stimulates epidermal cells to a certain degree. Indeed, our results demonstrate a dose-dependent response of Drosophila epidermal cells and human keratinocytes to radial stretch. However, we do not have any information about selectivity in response to different stimuli, though we agree that this is an intriguing avenue for future studies. For example, we don't know whether stretch-responsive cells are more or less responsive to poke. But, a salient feature of our studies is the recruitment of greater numbers of responders with increasing stimulus intensity, therefore we added the following statement to the discussion to clarify our model:

“Finally, we find that epidermal cells exhibit a dose-dependent response to radial stretch; we therefore anticipate that the output of epidermal cells is likewise dependent on the stimulus intensity.  Hence, rather than a fixed threshold beyond which epidermal cells are selectively activated, we hypothesize that increasing stimulus intensities drive increasing signal outputs to neurons.”

(8) Some Protocols are missing. For example, in Figure 4, many stimulus combinations were used to test behavior. How were stimuli of different modalities applied to the animals? Further details need to be provided in the protocols.

We thank the Reviewer for identifying this oversight. The methods section of our original submission detailed most of the stimulus combinations but omitted the opto + mechano combination (4F). We updated our methods to correct these omissions.

(9) It might be helpful if the authors could provide a sample video for each behavior to clarify how they were each defined.

Our manuscript includes a table with a detailed description of the behaviors (Table S2), and we added two annotated videos that show representative behavioral responses to optogenetic nociceptor or epidermis stimulation.

(10) A supplementary summary table of genotypes might be helpful for the reader.

Experimental genotypes are provided in the figure legends, and a detailed list of all alleles used in the study as well as their source is provided in supplemental table S1.

Reviewer #2 (Public Review):

Summary.

The authors provide compelling evidence that stimulation of epidermal cells in Drosophila larvae results in the stimulation of sensory neurons that evoke a variety of behavioral responses. Further, the authors demonstrate that epidermal cells are inherently mechanoresponsive and implicate a role for store-operated calcium entry (mediated by Stim and Orai) in the communication to sensory neurons.

Strengths.

The study represents a significant advance in our understanding of mechanosensation. Multiple strengths are noted. First, the genetic analyses presented in the paper are thorough with appropriate consideration to potential confounds. Second, behavioral studies are complemented by sophisticated optogenetics and imaging studies. Third, identification of roles for store-operated calcium entry is intriguing. Lastly, conservation of these pathways in vertebrates raise the possibility that the described axis is also functional in vertebrates.

Weaknesses.

The study has a few conceptual weaknesses that are arguably minor. The involvement of store-operated calcium entry implicates ER calcium store release. Whether mechanical stimulation evokes ER calcium release in epidermal cells and how this might come about (e.g., which ER calcium channels, roles for calcium-induced calcium release etc.) remains unaddressed. On a related note, the kinetics of store-operated calcium entry is very distinct from that required for SV release. The link between SOC and epidermal cells-neuron transmission is not reconciled. Finally, it is not clear how optogenetic stimulation of epidermal cells results in the activation of SOC.

(1) The involvement of store-operated calcium entry implicates ER calcium store release. Whether mechanical stimulation evokes ER calcium release in epidermal cells and how this might come about (e.g., which ER calcium channels, roles for calcium-induced calcium release etc.) remains unaddressed.

Our studies suggest that mechanically evoked responses in epidermal cells involve both ER calcium release and store-operated calcium entry. Notably, we show that depletion of ER calcium stores before mechanical stimulation, by treating with thapsigargin, reduces (but does not eliminate) mechanically evoked calcium responses in fly epidermal cells (Fig. 6C-6F). Likewise, fly epidermal cells and human keratinocytes both exhibit mechanically evoked calcium responses in the absence of extracellular calcium (10mM EGTA to chelate all free calcium ions). These data support a model whereby mechanical stimuli trigger calcium release from ER stores and influx. Indeed, several cell types have been shown to display mechanically evoked release of calcium from stores. For example, mechanical stimulation of enteroendocrine cells of the gut epithelium results in both calcium release from ER stores and calcium influx across the plasma membrane (Knutson et al., 2023). Similar to our findings, Knutson et al found that depleting stores decreased mechanically evoked calcium signals by over 70% in these gut epithelial stores. In our revised manuscript we have more clearly emphasized these points.

We agree with the reviewer that deciphering the mechanisms by which mechanical stimuli promote ER calcium release and subsequent store-operated calcium entry is an exciting topic to explore. One potential mechanism is the activation of a mechanosensitive receptor that promotes calcium release from the ER via calcium-induced calcium release or IP3 production, as has been proposed for enteroendocrine cells. A recent paper demonstrated that the ER itself is mechanosensitive and that mechanical stimuli promotes calcium release via the opening of calcium-permeable ion channels in the ER membrane (Song et al., 2024). Determining the relative contributions of store-operated calcium entry and ER calcium release and deciphering their underlying mechanisms will require a thorough investigation of ER calcium channels and receptors, thus we believe this would be beyond the scope of the present manuscript and merits publication on its own. However, we now include this in our discussion as an exciting new direction we aim to pursue.

(2) The kinetics of store-operated calcium entry is very distinct from that required for SV release. The link between SOC and epidermal cells-neuron transmission is not reconciled.

The Reviewer raises an interesting point regarding the mode of epidermal cell-neuronal communication. We demonstrated a requirement for dynamin-dependent vesicle release from epidermal cells in mechanical sensitization. However, the nature of the vesicular pool, the mode and kinetics of release, and the type of neuromodulator released remain to be characterized. Hence, it’s not clear that kinetics of synaptic vesicle release is an appropriate comparison. Our studies do demonstrate that behavioral responses to optogenetic epidermal stimulation are relatively slow – on the order of seconds – which is not incompatible with the kinetics of store-operated calcium entry. Furthermore, the primary functional output we define for epidermal mechanosensory responses, mechanical nociceptive sensitization, is apparent 10 sec following the stimulus and persists for minutes in our behavior assays. Consistent with this model, studies of the mammalian touch dome have shown that touch-sensitive Merkel cells secrete neurotransmitters to modulate neurons and promote sustained action potential firing on a similar timescale. Likewise, mechanically evoked ER calcium-release promotes sustained secretion of serotonin from enterochromaffin cells.

(3) It is not clear how optogenetic stimulation of epidermal cells results in the activation of SOC.

We appreciate the opportunity to clarify our results. We demonstrate that optogenetic epidermal stimulation elicits behavioral responses in larvae and calcium responses in somatosensory neurons, but we do not claim that optogenetic epidermal stimulation elicits SOC. Our optogenetic studies demonstrate the capacity for epidermal stimulation to modulate somatosensory function, but we characterize contributions of SOC only to mechanical stimuli which are more physiologically relevant. However, it is worth noting that CsChrimson is a calcium-permeable channel, suggesting that an increase in intracellular calcium may trigger epidermal-evoked neuronal responses and behaviors during optogenetic stimulation.

References

Hwang, RY, Zhong, L, Xu, Y, Johnson, T, Zhang, F, Deisseroth, K, and Tracey, WD (2007). Nociceptive neurons protect Drosophila larvae from parasitoid wasps. Curr Biol 17, 2105–2116.

Knutson, KR, Whiteman, ST, Alcaino, C, Mercado-Perez, A, Finholm, I, Serlin, HK, Bellampalli, SS, Linden, DR, Farrugia, G, and Beyder, A (2023). Intestinal enteroendocrine cells rely on ryanodine and IP3 calcium store receptors for mechanotransduction. J Physiol 601, 287–305.

Maksimovic, S, Nakatani, M, Baba, Y, Nelson, AM, Marshall, KL, Wellnitz, SA, Firozi, P, Woo, S-H, Ranade, S, Patapoutian, A, et al. (2014). Epidermal Merkel cells are mechanosensory cells that tune mammalian touch receptors. Nature 509, 617–621.

Mikesell, AR, Isaeva, O, Moehring, F, Sadler, KE, Menzel, AD, and Stucky, CL (2022). Keratinocyte PIEZO1 modulates cutaneous mechanosensation. Elife 11, e65987.

Song, Y, Zhao, Z, Xu, L, Huang, P, Gao, J, Li, J, Wang, X, Zhou, Y, Wang, J, Zhao, W, et al. (2024). Using an ER-specific optogenetic mechanostimulator to understand the mechanosensitivity of the endoplasmic reticulum. Dev Cell 59, 1396-1409.e5.

Woo, S-H, Ranade, S, Weyer, AD, Dubin, AE, Baba, Y, Qiu, Z, Petrus, M, Miyamoto, T, Reddy, K, Lumpkin, EA, et al. (2014). Piezo2 is required for Merkel-cell mechanotransduction. Nature 509, 622–626.

Author response:

We appreciate the reviewers’ constructive comments and suggestions. We plan the following revisions to address the public reviews.

Regarding model selection (from Reviewers 1 and 3)

We will test whether the latent cause model has a better explanatory power for the observed reinstatement data compared with at least two other models, including the Rescorla-Wagner model. For each model, the prediction errors across all trials and those in the test 3 trial (reinstatement) will be calculated for individual animals. The explanatory power of the models will be discussed based on these results. 

Regarding model validation (from Reviewers 1, 2, and 3)

We acknowledge the reviewers’ concerns about potential parameter overfitting and misinterpretation. First, the simulation in the latent cause model will be run under other possible conditions to test whether our original condition can be justified, then clarify how certain parameters affect the predicted CR. Second, we will confirm if the prediction errors are comparable between experimental groups, present the correlation between parameters, and discuss this result in the revision. 

To evaluate the effect of context in explaining reinstatement in the latent cause model, simulations of CR in test 3 when only context or tone is presented will also be performed and discussed with the behavioral data.

Regarding the interpretation of the behavioral data (from Reviewers 1, 2, and 3) We will clarify our interpretation of the behavioral data by incorporating the additional analyses mentioned above; for example, to clarify the contribution of context in test 3, we will provide data on the CR before the tone presentation in our revision. In addition, how we expected and interpreted the reversal Barnes maze results from the memory modification characteristics estimated in the reinstatement test will be further discussed.

Regarding the application of the latent cause model to the reversal Barnes maze task (from Reviewers 1, 2)

We acknowledge the reviewers’ suggestions to apply the latent cause model to our Barnes maze results to strengthen the link and consistency. To further clarify the reason for including Barnes maze results, we will explicitly discuss how associative learning is involved in spatial learning in the revision. However, we will not be able to directly apply the latent cause model for the Barnes maze data for the following reasons. As we noted in the Results and Discussion, the latent cause model was built on associative learning and cannot be directly applied to the Barnes maze data. The cognitive processes in the Barnes maze task involve maintaining spatial representation of the environment, integrating own position and expected goal, and evaluating potential actions. Importantly, the chosen actions in this task directly affect subsequent observations, while an animal’s response based on an expected outcome typically does not alter future observation in a simple associative learning paradigm. 

Thus, although associative learning (e.g., associations between the spatial cue and the location of the escape box) is certainly a critical building block and contributes to performance in the Barnes maze task, this mechanism alone cannot fully explain the animal’s navigation in the maze. We agree that having solid modeling results in the reversal Barnes maze task is an important direction, but extending the latent cause model for this purpose is beyond the scope of this study. We have suggested some possible approaches in the Discussion and will elaborate further on these conceptual distinctions and how latent cause framework assists in the interpretation of results.

Author response:

We thank the reviewers for their insightful feedback. Incorporating their recommendations will greatly enhance our manuscript for resubmission. Based on the review, it seems a major challenge to the interpretation of our study surrounds whether locomotion, itself, is responsible for increased ACC activity during our task. This was a shared concern for us during our analysis. We included data in our initial submission hoping to address these concerns. Specifically, we show that post-action activity outlasts movement termination, in most cases, on the order seconds after termination (Supplementary Fig 2). Likewise, post-action activity is not tied to shuttle initiations as ACC activity onset can vary greatly before and after initiation (Supplementary Fig 2). Lastly, the unique nature of action content neurons further supports a distinction from locomotor activity. They selectively fire for specific directions and, as a result, do not fire during movement in opposite directions. Despite these findings, we agree with reviews that inclusion of additional analyses, such as examining firing rates in respect to locomotion speed and acceleration/deceleration, will greatly strengthen our claim of ACC’s role in post-action activity. In our resubmission, we will seek to perform such an analysis, among others, to elucidate completely the role of locomotion in ACC post-action activity.

Reviewers also pointed out an overall lack of details surrounding our task, analysis, statistical methods and experimental approaches. We will consider all the recommendations from the reviewers and integrate them into our resubmission to provide more detailed information. Notably, we will adjust our approach in describing our task. Reviewers discussed some criticism regarding the perceived novelty of the task as it shares many similarities with previous discrimination-avoidance tasks. The distinction with our task is regarding the nuance of how the meaning (safety vs shock) of the context and sensory stimuli dynamically changes based on the current environment (context x sound). This requires not only the discrimination of contextual and sensory stimuli but also the inter-modal integration of stimuli, which varies throughout the task. Sound A/B leads to different outcomes depending on the context, and similarly, the meaning of the context shifts in a sound-dependent manner.

Lastly, in our follow-up submission we will work to include more robust analyses to utilize our temporal sensitivity of our recordings. We also will provide greater clarity on how each individual animal contributes to our overall findings. To conclude, we would like to once again thank our reviewers for their feedback and evaluation of our manuscript. We look forward to making the necessary adjustments for our future submission.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors tackled the public concern about E-cigarettes among young adults by examining the lung immune environment in mice using single-cell RNA sequencing, discovering a subset of Ly6G- neutrophils with reduced IL-1 activity and increased CD8 T cells following exposure to tobacco-flavored e-cigarettes. Preliminary serum cotinine (nicotine metabolite) measurements validated the effective exposure to fruit, menthol, and tobacco-flavored e-cigarettes with air and PG/VG serving as control groups. They also highlighted the significance of metal leaching, which fluctuated over different exposure durations to flavored e-cigarettes, underscoring the inherent risks posed by these products. The scRNAseq analysis of e-cig exposure to flavors and tobacco demonstrated the most notable differences in the myeloid and lymphoid immune cell populations. Differentially expressed genes (DEGs) were identified for each group and compared against the air control. Further sub-clustering revealed a flavor-specific rise in Ly6G- neutrophils and heightened activation of cytotoxic T cells in response to tobacco-flavored e-cigarettes. These effects varied by sex, indicating that immune changes linked to e-cig use are dependent on gender. By analyzing the expression of various genes and employing gene ontology and gene enrichment analysis, they identified key pathways involved in this immune dysregulation resulting from flavor exposure. Overall, this study affirmed that e-cigarette exposure can suppress the neutrophil-mediated immune response, subsequently enhancing T cell toxicity in the lung tissue of mice.

Strengths:

This study used single-cell RNA sequencing to comprehensively analyze the impact of e-cigarettes on the lung. The study pinpointed alterations in immune cell populations and identified differentially expressed genes and pathways that are disrupted following e-cigarette exposure. The manuscript is well written, the hypothesis is clear, the experiments are logically designed with proper control groups, and the data is thoroughly analyzed and presented in an easily interpretable manner. Overall, this study suggested novel mechanisms by which e-cigs impact lung immunity and created a dataset that could benefit the lung immunity field.

We thank the reviewer for identifying the strengths of our work.

Weaknesses:

The authors included a valuable control group - the PG/VG group, since PG/VG is the foundation of the e-liquid formulation. However, most of the comparative analyses use the air group as the control. Further analysis comparing the air group to the PG/VG group, and the PG/VG group to the individual flavored e-cig groups will provide more clear insights into the true source of irritation. This is done for a few analyses but not consistently throughout the paper. Flavor-specific effects should be discussed in greater detail. For example, Figure 1E shows that the Fruit flavor group exhibits more severe histological pathology, but similar effects were not corroborated by the single-cell data.

We thank the reviewer for this query. We agree that PG/VG group is the foundation of the e-liquid formulation and hence comparisons with this group is of significance to understand the effect of individual flavors on the cell population. Though we compared the flavored e-cig groups with PG/VG group, we did not discuss it in detail within the manuscript to avoid confusions in interpretation for such a big dataset. However, we will include the comparisons with the PG/VG group as a Supplement File in our revised manuscript to facilitate proper interpretation of our omics data to interested readers.

While we agree that flavor-specific effects might be of interest, we did not delve into exploring them in detail as the fruit flavored e-liquids have now been regulated for sale in the US. Thus, from regulatory point of view, the effects of tobacco- and menthol-flavored e-liquids hold most interest. Since at the time of conducting this study, fruit flavors were in the market, we have still included the data. However, studying it further was not the focus of this work. Nevertheless, interested readers of our manuscript can have access to our dataset to allow further analyses and interpretation of our results.

The characterization of Ly6g+ vs Ly6g- neutrophils is interesting and potentially very impactful. Key results like this from scRNAseq analyses should be validated by qPCR and flow cytometry.

Also, a recent study by Ruscitti et al reported Ly6g+ macrophages in the lung which can potentially confound the cell type analysis. A more detailed marker gene and sub-population analysis of the myeloid clusters could rule out this potential confounding factor.

We agree with the reviewer that the loss of Ly6G on neutrophils is a very interesting find and we are in process of designing neutrophil specific experiments to study the impact of e-cig exposure on neutrophil maturation and function which will be discussed in subsequent work by our group. However, to address the concerns raised by the reviewer, we are staining the lung tissue samples from air-and differently flavored e-cig aerosol exposed mouse lungs with Ly6G and S100A8 (universal marker for neutrophil) to see the infiltration of Ly6g+ vs Ly6g- neutrophils within the lungs of exposed and unexposed mice. This would also address the question if these populations were neutrophils or belong to another myeloid origin as suggested by recent publications. We will share the results from our findings in the revised manuscript and update our interpretations accordingly with better validations.

Reviewer #2 (Public review):

This study provides some interesting observations on how different flavors of e-cigarettes can affect lung immunology, however there are numerous flaws including a low number of replicates and a lack of effective validation methods which reduces the robustness and rigor of the findings.

Strengths:

The strength of the study is the successful scRNA-seq experiment which gives good preliminary data that can be used to create new hypotheses in this area.

We appreciate the reviewer for recognizing the strength of this work.

Weaknesses:

The major weakness is the low number of replicates and the limited analysis methods. Two biological n per group is not acceptable to base any solid conclusions. Any validatory data was too little (only cell % data) and did not always support the findings (e.g. Figure 4D does not match 4C). Often n seems to be combined and only one data point is shown, it is not at all clear how the groups were analyzed and how many cells in each group were compared.

We thank the reviewer for the critique to allow us to improve our analyses. We understand that the low number of replicates in this work makes the analyses difficult to draw solid conclusions, but this was a pilot study to understand the changes in the mouse lung upon acute exposures to flavored e-cig aerosols at a single cell level. So far, the e-cig field has been primarily focused on conducting toxicological studies to help regulatory bodies to set standards and enforce laws to better regulate the manufacture, sale and distribution of e-cig products. However, adolescents and young adults are still getting access to these products, and there is little to no understanding of how this may affect the lung health upon acute and chronic exposures. Single cell technology is a powerful tool to analyze the gene expression changes within cell populations to study cell heterogeneity and function. Yet, it is a costly tool, owing to which, conducting such analyses on large sample sizes is not ideal. This pilot study was designed to get some initial leads for future studies involving larger sample sizes and chronic exposures. Further, we still intend to share our results with the scientific community due to the value of such a dataset for a wider audience interested in learning about the mechanistic underpinnings of e-cig exposures in vivo.

We understand that the validations are limited in our current work and so we are in process of conducting some immunostaining to validate a few targets made through this work. We also want to add here that validating single cell findings using any of the classical methods of experimentation including ELISA, qPCR or flow cytometry is sometimes difficult as many of these techniques still investigate the tissue while the changes shown in single cell analyses are mainly pertaining to a single cell type. This could be a probable reason for the scRNA seq results not aligning with our findings from flow cytometry. The data/findings from this pilot study have now allowed us to be better informed to design an effective flow panel for our future studies. In terms of the statistics and the number of cells for each analysis, we will share the detailed account and information for each to allow better interpretation of our results.

Only 71,725 cells means only 7,172 per group, which is 3,586 per animal - how many of these were neutrophils, T-cells, and macrophages? This was not shown and could be too low.

We do agree that the number of cells could be too low, but to avoid this we never studied the gene expression variations at the finest level of cell identity. We classified the cell clusters into general annotations -myeloid, lymphoid, endothelial, stromal and epithelial- and identified the changes in the gene expressions. Of these, only two clusters (myeloid and lymphoid) with more than ~1000 cells per cell type per group were studied in detail. We will include the cell count information to allow better interpretation of our results in the revised manuscript.

The dynamic range of RNA measurement using scRNAseq is known to be limited - how do we know whether genes are not expressed or just didn't hit detection? This links into the Ly6G negative neutrophil comment, but in general, the lack of gene expression in this kind of data should be viewed with caution, especially with a low n number and few cells.

This is a well-made point, and we thank the reviewer for this comment. We agree that the dynamic range RNA measurement is limited and for low cell numbers that could lead to bias. We are in process of validating the findings regarding the presence of Ly6G+ and Ly6G- cells in our control and treated lungs, the outcome of which will be discussed in the revised manuscript. We will also provide the cell number for the Ly6G- cell cluster for each sample with more detailed discussion of our findings. Due to the small sample size and cell capture, few limitations are hard to overcome which will be further elaborated upon in our revisions.

There is no rigorous quantification of Ly6G+ and Ly6G- cells in the flow cytometry data.

We understand that flow-based quantification of our scRNA seq findings would be interesting. However, flow cytometry and single cell suspension to perform sequencing were performed parallelly for this study. We used a basic flow panel using single markers to identify individual immune cell type. We did identify changes in the Ly6G population in our treated and control samples using scRNA seq and intend to include it as a marker for our future studies using flow cytometry. But unfortunately, the same analyses could not be performed for the current batch of samples. We will still include results from IHC staining to identify the Ly6G+ and Ly6G- population in the lung tissues from control and treated mice in revised manuscript to address some of the concerns raised here.

Eosinophils are heavily involved in lung biology but are missing from the analysis.

We used RBC lysis buffer to remove the excess RBCs during lung digestion for preparation of single cell suspension for scRNA seq in this study. Reports suggest that RBC lysis could adversely affect the eosinophil number and function. We did not identify any cell cluster, representing markers for eosinophils through our scRNA seq data and we believe that our lung digestion protocol could be the reason for the same. We have studied the eosinophil number changes through flow cytometry in these samples and have found significant changes as well. However due to our inability to find cell clusters for eosinophil through scRNA seq data, we did not include these results in the final manuscript. To avoid confusions and maintain transparency we will include our results from flow cytometry experiments in the revised manuscript.

The figures had no titles so were difficult to navigate.

We will make necessary adjustments to the data representation and include the titles to enable easy navigation of the Figures.

PG/VG is not defined and not introduced early enough.

We agree that PG/VG is an important control to compare in e-cig studies. This was the reason why this group was included, and we performed comparisons with this group for scRNA seq studies as well. However, to reduce the complexity of the study, we only shared the comparisons with Air control in this manuscript. We will include the comparisons made with PG/VG group as a Supplementary File in the revised manuscript to allow the interested readers have access to the study results and make necessary interpretations for future research.

Neutrophils are not well known to proliferate, so any claims about proliferation need to be accompanied by validation such as BrdU or other proliferation assays.

We thank the reviewer for this suggestion; however, we cannot perform the BrDU or other proliferation assay on neutrophils for now. We are planning to include these in the study designs of our future work, however we have limitations of funds to continue further experimentation to support this claim for this study. We mention clearly that this is only a scRNA seq finding and requires further study to avoid over-interpretation of our results.

It was not clear how statistics were chosen and why Table S2 had a good comparison (two-way ANOVA with gender as a variable) but this was not used for other data particularly when looking at more functional RNA markers (Table S2 also lacks the interaction statistic which is most useful here).

We thank the reviewer for bringing this concern. We understand that this is a valid point and will include all the necessary information regarding the statistics and other related parameters in the revised manuscript.

Many statistics are only vs air control, but it would be more useful as a flavor comparison to see these vs PG/VG. In some cases, the carrier PG/VG looks worse than some of the flavors (which have nicotine).

We will include the comparisons with PG/VG as supplementary file in our revised manuscript, however we do not intend to describe all those changes in detail in the main manuscript.

The n number is a large issue, but in Figures such as 4, 6, and 7 it could be a bigger factor. The number of significant genes identified has been determined by chance rather than any real difference, e.g. Is Il1b not identified in Fruit flavor vs air because there wasn't enough n, while in Air vs Tobacco, it randomly hit the significance mark. This is but an example of the problems with the analysis and conclusions.

While we agree in part with the concern raised here, we wish to point out that there are limitations to every experiment. In our opinion, an omics study is not necessarily aimed to find the changes at transcript level with absolute certainty, rather to identify probable cell and gene targets to validate with subsequent work. We never claim that our findings are absolute outcomes but rather add the limitation of sample number and need for further research at every step. The strength of this work is to be the first study of its kind looking at changes in the lung cell population at single cell level upon e-cig aerosol exposure. This study has provided us with interesting gene and cell targets that we are now validating with future work. We still strongly believe that a dataset like this is a useful resource for a wider audience to allow efficient study designs and hence it is befitting to be published and discussed amongst our peers.

The data in Figure 7A is confusing, if this is a comparison to air, then why does air vs air not equal 1? Even if this was the comparison to the average of air between males and females, then this doesn't explain why CCL12 is >1 in both. Is this z-score instead? Regardless the data is difficult to interpret in this format.

We thank the reviewer for pointing this out. We realize that the data might be difficult to understand due to scaling of the color codes for the heatmap. We will change the graphical representation and include actual number for fold change in our revised manuscript to allow easy interpretation of these results.

Individual n was not shown for almost all experiments - e.g. Figure 1D - what is this representative of? Figure 2D - is this bulk-grouped data for all cells and all mice? The heatmaps are also pooled from 2n and don't show the variability.

While we have included a pictorial representation of the n number in Figure 1A and mentioned n number in the Figure legends for each figure, we understand that it maybe difficult to navigate. We will attempt to address this in a better manner in the revised manuscript.

However, with respect to the second comment we would like to differ from the reviewer’s opinion. Each scRNA seq data had 2 samples – one for male and another for female which has been clearly shown in the current figures. The pooling of cells as mentioned in the comment happened at the stage of preparation of cell suspension from each sex/group at the start of the sequencing. We do not have any means to show the variability amongst pooled samples, which we acknowledge as a shortcoming of our work. So, in terms of representation of the heatmaps and data analyses we have included all the needed information to uphold transparency of our study design and data visualization for each figure and would like to stick to the current representations.

Reviewer #3 (Public review):

This work aims to establish cell-type specific changes in gene expression upon exposure to different flavors of commercial e-cigarette aerosols compared to control or vehicle. Kaur et al. conclude that immune cells are most affected, with the greatest dysregulation found in myeloid cells exposed to tobacco-flavored e-cigs and lymphoid cells exposed to fruit-flavored e-cigs. The up-and-down-regulated genes are heavily associated with innate immune response. The authors suggest that a Ly6G-deficient subset of neutrophils is found to be increased in abundance for the treatment groups, while gene expression remains consistent, which could indicate impaired function. Increased expression of CD4+ and CD8+ T cells along with their associated markers for proliferation and cytotoxicity is thought to be a result of activation following this decline in neutrophil-mediated immune response.

Strengths:

(1) Single-cell sequencing data can be very valuable in identifying potential health risks and clinical pathologies of lung conditions associated with e-cigarettes considering they are still relatively new.

(2) Not many studies have been performed on cell-type specific differential gene expression following exposure to e-cig aerosols.

(3) The assays performed address several factors of e-cig exposure such as metal concentration in the liquid and condensate, coil composition, cotinine/nicotine levels in serum and the product itself, cell types affected, which genes are up- or down-regulated and what pathways they control.

(4) Considerations were made to ensure clinical relevance such as selecting mice whose ages corresponded with human adolescents so that the data collected was relevant.

We thank the reviewer for identifying the key strengths of our work and listing it in a concise and well-rounded fashion.

Weaknesses:

The exposure period of 1 hour a day for 5 days is not representative of chronic use and this time point may be too short to see a full response in all cell types. The experimental design is not well-supported based on the literature available for similar mouse models.

This study was not designed to study the effects of chronic exposures on lung tissues. We were interested in delineating the effect of acute exposures for which the proposed study design was chosen. Previous work by our group has performed similar exposures and has been well received by the community. We understand that chronic exposures will be interesting to look at, however that was not the purpose of this pilot study. We will now explicitly mention this aspect in the revised manuscript.

Several claims lack supporting evidence or use data that is not statistically significant. In particular, there were no statistical analyses to compare results across sex, so conclusions stating there is a sex bias for things like Ly6G+ neutrophil percentage by condition are observational.

We thank the reviewer for this observation, and we will include the necessary validations and details of the sex-based statistical analyses in the revised version of this manuscript.

Statistical analyses lack rigor and are not always displayed with the most appropriate graphical representation.

We thank the reviewer and will include all the necessary statistical details with more details in the revised manuscript.

Overall, the paper and its discussion are relatively limited and do not delve into the significance of the findings or how they fit into the bigger picture of the field.

We are in process of performing a few validatory experiments and intend to include few other pieces of data to this manuscript to add to the overall merit of our findings. However as pointed out by the reviewer themselves the strength of this work is in the first ever scRNA seq analyses of mouse exposed to differently flavored e-cig aerosols in vivo. We also show cell-specific differential gene expression and address some of the major queries made around e-cig research including release of metals on a day-to-day basis from the same coil. The limited sample number make it difficult to draw solid conclusions from this work, which has been discussed as a shortcoming. However the major strength of this work is not in identifying specific trends but rather to explore the possible cell and gene targets to expand the study for longer (chronic) exposures with a larger sample group.

The manuscript lacks validation of findings in tissue by other methods such as staining.

We are conducting some studies and will include the validatory experiments and staining in the revised manuscript to support our findings.

This paper provides a foundation for follow-up experiments that take a closer look at the effects of e-cig exposure on innate immunity. There is still room to elaborate on the differential gene expression within and between various cell types.

We thank the reviewer for this observation. The cell numbers for some cell clusters (especially epithelial cells) were too low. So, though we have performed the differential gene expression analyses on all the cell clusters, we refrained from discussing it in the manuscript to avoid over interpretation of our results. Only clusters with high enough (~1000) cells per sex per group were used to plot the heatmaps. We will also include the cell numbers for each cell type in the revisions to allow better interpretation of our data. Furthermore, the raw data from this study will be freely available to the public upon publication of this manuscript. This would enable the interested readers to access the raw data and study the cell types of interest in detail based on their study requirements. This data will be a useful resource for all in this community to inform and design future studies.

Author response:

Reviewer #1:

A) The presentation of the paper must be strengthened. Inconsistencies, mislabelling, duplicated text, typos, and inappropriate colour code should be changed.

We will revise the manuscript to correct the abovementioned issues.

B) Some claims are not supported by the data. For example, the sentence that says that "adolescent mice showed lower discrimination performance than adults (l.22) should be rewritten, as the data does not show that for the easy task (Figure 1F and Figure 1H).

We will carefully review, verify claims, and correct conclusions where needed.

C) In Figure 7 for example, are the quantified properties not distinct across primary and secondary areas?

We will analyse the data in Figure 7 separately for AUDp and secondary auditory cortices to test regional differences. Additionally, we will provide a table summarizing key neuronal firing properties for each area during passive recordings to clarify how activity varies across cortical subregions and developmental stages.

D) Some analysis interpretations should be more cautious. (..) A lower lick rate in general could reflect a weaker ability to withhold licking- as indicated on l.164, but also so many other things, like a lower frustration threshold, lower satiation, more energy, etc).

We will address issues around lick bias including alternative explanations, such as differences in motivation or impulsivity.

Reviewer #2:

A) For some of the analyses that the authors conducted it is unclear what the rationale behind them is and, consequently, what conclusion we can draw from them.

We will edit the discussion and clarify these points. In addition, we will adjust and extend the methodology section to clarify the rationale of our analysis.

B) The results of the optogenetic manipulation, while very interesting, warrant a more in-depth discussion.

We agree that the effects observed in our optogenetic manipulation warrant further discussion. We will extend on the analysis and discussion of ACx silencing.

Reviewer #3:

A) One fact that could help shed light on this would be to know how often the animals licked the spout in between trials. Finally, for the head-fixed version of the task, only d' values are reported. Without the corresponding hit and false alarm rates (and frequency of licking in the intertrial interval), it's hard to know what exactly the animals were doing.

We recognize the need for a more nuanced analysis for the head-fixed version of the task. We will extend the behavioral analysis and provide more details to clarify these points.

B) There are some instances where the citations provided do not support the preceding claim. For example, in lines 64-66, the authors highlight the fact that the critical period for pure tone processing in the auditory cortex closes relatively early (by ~P15). However, one of the references cited (ref 14) used FM sweeps, not pure tones, and even provided evidence that the critical period for this more complex stimulus occurred later in development (P31-38). Similarly, on lines 72-74, the authors state that "ACx neurons in adolescents exhibit high neuronal variability and lower tone sensitivity as compared to adults." The reference cited here (ref 4) used AM noise with a broadband carrier, not tones.

We appreciate the reviewer pointing out instances where our citations may not fully support our claims. We will carefully review the relevant citations and revise them to ensure they accurately reflect the findings of the cited studies. We will update references in lines 64–66 and 72–74 to better align with the specific stimulus types and developmental timelines discussed.

C) Given that the authors report that neuronal firing properties differ across auditory cortical subregions (as many others have previously reported), why did the authors choose to pool neurons indiscriminately across so many different brain regions?

We agree that pooling neurons from multiple auditory cortical regions could potentially obscure region-specific differences. However, we addressed this concern by analyzing regional differences in neuronal firing properties, as shown in Supplementary Figures S4-1 and S4-2, and Supplementary Tables 2 and 3. Additionally, we examined stimulus-related and choice-related activity across regions and found no significant differences, as presented in Supplementary Figure S4-3. Please see our response to Reviewer 1, where we further elaborate on this point.

D) And why did they focus on layers 5/6? (Is there some reason to think that age-related differences would be more pronounced in the output layers of the auditory cortex than in other layers?)

We acknowledge that other cortical layers are also of interest and may contribute differently to auditory processing across development. Our focus on layers 5/6 was motivated by both methodological considerations and biological relevance. These layers contain many of the principal output neurons of the auditory cortex, and are therefore well positioned to influence downstream decision-making circuits. We will clarify this rationale in the revised manuscript and note the limitations of our approach.

Author response:

Reviewer #1 (Public Review):

The work of Umetani et al. monitors the death of about 100,000 cells caused by lethal antibiotic treatments in a microfluidic device. They observe that the surviving bacteria are either in a dormant or in a non-dormant state prior to the antibiotic treatment. They then study the relative abundances of these different persister cells when varying the physiological state of the culture. In agreement with previous observations, they observe that late stationary phase cultures harbor a high number of dormant persister cells and that this number goes down as the culture is more exponential but remains non-zero, suggesting that cultures at the exponential phase contain different types of persister bacteria. These results were qualitatively similar in a rich and poor medium. Further characterization of the growing persister bacteria shows that they often form Lforms, have low RpoS-mcherry expression levels and grow only slightly more slowly than the non-persister bacteria. Taken together, these results draw a detailed view of persister bacteria and the way they may survive extensive antibiotic treatments. However, in order to represent a substantial advance on previous knowledge, a deeper analysis of the persister bacteria should be done.

We thank the reviewer for suggesting the addition of more detailed analyses of persister cells. As we wrote in our response to Essential Revision 1, we now include a new section titled “Response of growing persisters to Amp exposure is heterogeneous” (Page 11-12) and present the results of the detailed analyses of single-cell dynamics of growth and cell morphology over the course of the pre-exposure, exposure, and post-exposure periods (Fig. 2D and H, Fig. 4B and D, Fig. 4 – figure supplement 1 and 2, Fig. 5B and D, Fig. 5 – figure supplement 1, Fig. 8B and D, and Figure 8 – figure supplement 1). The new results characterize differential responses to Amp treatment among growing persister cells (Fig. 4A-D, Fig. 4 – figure supplement 1, Fig. 4 – figure supplement 2A, Fig. 5A-D, and Fig. 5 – figure supplement 1), comparable division rates of MG1655 between non-surviving cells and persister cells growing prior to antibiotic treatments (Fig. 4E and Fig. 8E), except for the post-exponential phase cell populations of MF1 to Amp treatment in the LB medium and the post-exponential phase cell populations of MG1655 to Amp treatment in the M9 medium (Fig. 4 – figure supplement 2B and Fig. 5E) and the presence of persister cells to CPFX that avoid filamentation after the treatment (Fig. 8C and D, and Fig. 8 – figure supplement 1). We believe that these new analyses would provide new insights into the diverse dynamics and survival modes of antibiotic persistence at the single-cell level and represent important contributions to the field.

Reviewer #2 (Public Review):

The main question asked by Umenati et al. is whether persister cells to ampicillin arise preferentially from dormant, non-dividing cells or from cells that are actively growing before antibiotic exposure. The authors tracked persister cells generated from populations at different growth phases and culture media using a microfluidic device coupled to fluorescence microscopy, which is a challenge due to the low frequency of these persister cells. One of the main conclusions is that the majority of persisters arising in exponentially-growing populations originated from actively-dividing cells before the antibiotic treatment, reinforcing the idea that dormancy is not a prerequisite for persister formation. The authors made use of a fluorescent reporter monitoring RpoS activity (RpoS-mCherry fusion) and observed that RpoS levels in these persister cells were low. In the few lineages that exhibited no growth before the ampicillin treatment, RpoS levels were low as well, indicating that RpoS is not a predictive marker for persistence. By performing the same experiment with early and late stationary phase cultures, the authors observed that the proportion of persister cells that originated from dormant cells before the ampicillin treatment is significantly increased under these conditions. In the late stationary phase condition, dormant cells were expressing high levels of RpoS. The authors suggested that RpoS-mCherry proteins form aggregates which were suggested by the authors to be a characteristic of 'deep dormancy'. These cells were mostly unable to restart growth after the antibiotic removal while others with the lowest levels of RpoS tended to be persister. Confirming that these cells indeed contain protein aggregates as well as determining the physiological state of these cells appears to be crucial.

We thank reviewer #2 for pointing out the critical issue with the RpoS-mCherry fusion that we used to quantify RpoS expression levels in single cells in the original manuscript. As explained in our reply to the comments below, we performed a suggested experiment and confirmed that the RpoS function was impaired by tagging it with mCherry. To resolve this issue, we repeated almost all the experiments using the wild-type strain MG1655 and confirmed the reproducibility of the main results (Fig. 3, Fig. 3 – figure supplement 1, and Fig. 7). Due to this change of the main strain used in this study, we removed the results on the correlation between RpoS expression and the persistence trait in the revised manuscript because it may not reflect the relationship of intact RpoS. However, we decided to still keep and show some of the results with the MF1 strain, such as the population killing curves and the survival mode analyses, because they also provide insight into the role of RpoS in antibiotic persistence. In particular, we found both beneficial and detrimental effects of RpoS on antibiotic persistence, depending on culture conditions and duration of antibiotic treatment (Fig. 1 – figure supplement 3 and Fig. 6 – figure supplement 1). Therefore, we have included these results and related discussions in the revised manuscript.

Reviewer #3 (Public Review):

In their manuscript, Umetani, et al. address the question of the origin of persister bacteria using single-cell approaches. Persistence refers to a physiological state where bacteria are less sensitive to antibiotherapy, although they have not acquired a resistance mutation; importantly, the concept of persistence has been refined in the past decade to distinguish it from tolerance where bacteria are only transiently insensitive. Since persister cells are very rare in growing populations (typically 1e-5 or 1e-6), it is very challenging to observe them directly. It had been proposed that individual cells surviving antibiotics are not growing at the start of the treatment, but recent studies (nicely reviewed in the introduction) where persister bacteria were observed directly do not support this link. Following a similar line, the authors nonetheless still aim at "investigating whether non-growing cells are predominantly responsible for bacterial persistence". Based on new experimental data, they claim the contrary that most surviving cells were "actively growing before drug exposure" and that their work "reveals diverse survival pathways underlying antibiotic persistence".

We thank the reviewer for this helpful comment, which suggested to us that some revisions in our Introduction would better place our study in the context of previous understanding of antibiotic persistence. As mentioned in our response to Essential Revision 4 and the second comment of Reviewer 1's Recommendations for the authors, we have modified the Introduction to more appropriately place our study in the context of the field.

The main strengths of the manuscript are in my opinion:

- To report on direct observation of E. coli persisters to ampicillin (200µg/mL) in 5 different growth media (typically 20 persisters or more per condition, one condition with 12 only), which constitutes without a doubt an experimental tour de force.

- To aim at bridging the population level and the single-cell level by measuring relevant variables for each and analyzing them jointly.

- To demonstrate that in most conditions a large fraction of surviving cells was actively growing before drug exposure.

In addition, although it is well-known that E. coli doesn't need to maintain its rod shape for surviving and dividing, I found very remarkable in their data the extent to which morphology can be affected in persister cells and their progeny, since this really challenges our understanding of E. coli's "lifestyle" (these swimming amoeba-like cells in Supp Video 11 are mind-blowing!).

We are grateful to the reviewer for the articulation of the strength of this study. 

Unfortunately, these positive aspects are counter-balanced by several shortcomings in the way experiments are analyzed and interpreted, which I explain below. Moreover, the manuscript is written in a way that makes it very hard to find important information on how experiments are done and is likely to leave the reader with an impression of confusion about what the main findings actually are.

We thank the reviewer for pointing out these important issues regarding the original manuscript. Please see our replies below regarding how we corresponded to each specific comment to resolve the issue. To make the experimental methods and procedures more accessible and interpretable, we have added more explanations of the experimental details to the Results and Methods sections. Furthermore, since we understood that some of the confusions came from the insufficient explanation of the preculture procedures for the microfluidic experiments, we have modified the schematic illustration of the method shown in Fig. S1 in the original manuscript and moved it as the first main figure in the revised manuscript (Fig. 1C and D). We have also added an illustration that explains the cultivation procedures for the batch culture experiments as Fig.

6A. 

My major concerns are the following:

(1) The main interpretation framework proposed by the authors is to assess whether cells not growing before drug exposure (so-called "dormant") are more or less likely to survive the treatment than growing ones ("non-dormant"). Fig 2A and Fig 3G show the main conclusions of the article from this perspective, that growing cells can survive the treatment and that the fraction of persisters in a given condition is not explained by the fraction of "dormant" cells, respectively. With this analysis, the authors essentially assume that "dormant" cells are of the same type in their different conditions, which ignores the progress in this field over the last decade (Balaban et al. 2019). I argue on the contrary that the observation of "diverse modes of survival in antibiotic persistence" is expected from their experimental design. In particular, the sensitivity of E. coli to beta-lactams such as ampicillin is expected to be much lower during the lag out of the stationary phase, a phenomenon which has been coined "tolerance"; hence in the Late Stationary condition, two subpopulations coexist for which different response to ampicillin is expected. I propose steps toward a more compelling interpretation of the experimental data. Should this point be taken seriously by the authors, it, unfortunately, implies a major rewriting of the article, including its title.

We thank the reviewer for bringing to our attention the point that may have caused confusion in the original manuscript. 

The primary purpose of this manuscript was not to assess whether non-growing cells prior to drug exposure are more or less likely to survive treatment than growing cells. Rather, we wanted to examine how different persister cell dynamics emerge at the single-cell level depending on previous cultivation history, growth media, and antibiotic types. We believe that this point is clearer in the revised manuscript with the newly added single-cell dynamics data (Fig. 2D, 2H, 4B, 4D, Fig. 4 – figure supplement 1 and 2A, Fig. 5B, 5D, Fig. 5 – figure supplement 1, Fig. 8B, 8D, and Fig. 8 – figure supplement 1). 

We also did not mean to imply that "dormant cells" were of the same type under different conditions, as we were aware of the diversity of cellular states of non-growing cells, as well as the reduced sensitivity of cells to antibiotics during the lag out of stationary phase. We believe that one of the reasons this point may have been unclear is that in the previous version we had referred to all cells that were not growing prior to antibiotic treatment as "dormant cells", a term that is often used in a more restricted way to refer to cells under prolonged growth arrest. Therefore, in the revised manuscript, we have avoided the term "dormant cells" and instead simply referred to these as "non-growing cells". Accordingly, we have changed the title of the paper from "Observation of non-dormant persister cells reveals diverse modes of survival in antibiotic persistence" to "Observation of persister cell histories reveals diverse modes of survival in antibiotic persistence".

To further address these points, we have improved the description of the experimental procedures for the single-cell measurements (see the reviewer's next comment as well). The nongrowing persisters of the MF1 strain found in the post-exponential phase cell populations must be of a different type than those found in the post-early and post-late stationary phase cell populations due to the experimental design. All early and late stationary phase cells were maintained in a non-growing state by flowing conditioned media prepared from the early and late stationary phase cultures until the start of the time-lapse measurements. Thus, aside from potential physiological heterogeneity, the non-growing cells prior to drug treatment are all long lagging cells. On the other hand, for the post-exponential phase condition, we maintained exponential growth conditions during the period from the start of the second pre-culture to the start of antibiotic treatment, including the period during sample preparation for time-lapse measurements. Given the exponential dilution by growth of cell populations, the non-growing persisters are unlikely to be long lagging cells (see our response to Reviewer 2's third comment  in "Recommendations for the authors"). We now describe these experimental procedures in more detail in the Results section (L161-178, L287-297). In addition, we discuss the diversity of cellular states of both non-growing and growing cells in Discussion, citing literature (L545-557).

(2) The way the authors describe their experiments with bacteria in the stationary phase is very problematic. For instance, they write that they "sampled cells from early and late stationary phases (...) and exposed them to 200 μg/mL of Amp in both batch and single-cell cultures." For any reader in a hurry (hence skipping methods and/or supplementary figure), this leads to believe that bacteria sampled in the stationary phase were exposed to the drug right away (either by adding the drug to the stationary phase sample, or more classically by transferring cells to fresh media with antibiotics). However, it turns out that, after sampling and loading in the microfluidic device, bacteria are grown 2 h in LB (or 4 h in M9) - I don't know what to think of such a blatant omission. The names chosen for each condition should reflect their most important aspects, here "stationary" is simply not appropriate - maybe something like "post early stationary" instead. In any case, I believe that this point highlights further the misconception pointed out in 1 and implies that the average reader will be at best confused, and probably misled.

We again thank the reviewer for pointing out the insufficient explanation of the method for the single-cell measurements and the helpful recommendation regarding our nomenclature for different conditions. As mentioned above, we now present the previous supplementary figure that schematically explains the experimental procedure as the first main figure to clarify how we prepared the cells loaded into the microfluidic device for single-cell measurements (Fig. 1C and D). Also, following the reviewer's suggestion, we now refer to the conditions as "post-exponential phase," "post-early stationary phase," and "post-late stationary phase" in the revised manuscript. 

We included a 2-hour (or 4-hour in M9) cultivation period in fresh medium in batch cultures for measuring killing curves to make the cultivation conditions prior to antibiotic treatment as similar as possible between batch and microfluidic experiments. We have clarified the presence of preexposure cultivation of post-early stationary and post-late stationary phase cell populations in the fresh medium before treating them with antibiotics (L264-269, Fig. 6A), so that readers can more easily recognize the experimental conditions.

(3) Figures 4 and 5 are of very minor significance, and the methodology used in Fig 4 is questionable. The authors measure the abundance of an Rpos-mCherry translational fusion because its "high expression has been suggested to predict persistence". The rationale for this (that an RpoS-mCherry fusion would be a proxy for intracellular ppGpp levels, and in turn predict persistence) has never been firmly established, and the standards used in the article where this reporter was introduced (Maisonneuve, Castro-Camargo, and Gerdes 2013) are notoriously low (which eventually led to its retraction) - I don't know what to think of the fact that the authors cite a review by this group rather than their retracted article. While transcriptional fusions of promoters regulated by RpoS have been proposed to measure its regulatory activity (Patange et al. 2018), the combination of self-regulation and complex post-translational regulation of rpoS makes the physical meaning of the reporter used here completely unclear. Moreover, this translational fusion is introduced without doing any of the necessary controls to demonstrate that the activity of RpoS is not impaired by the addition of the fluorescent protein. Fig 5 simply reports the existence of persisters to ciprofloxacin growing before the treatment. This might be a new observation but it is not unexpected given that a similar observation has been made with a similar drug, ofloxacin (Goormaghtigh and van Melderen 2019), as pointed out in the introduction. There is no further quantitative claim on this.

We thank the reviewer for pointing out the issue of the RpoS-mCherry fusion. As we mentioned in our response to Essential Revision 2 and also to the comment from reviewer #2, we have tested the sensitivity of this fluorescent reporter strain to oxidative stress and confirmed that it is as sensitive as the rpoS strain (Fig. 1 – figure supplement 1C). Therefore, the RpoS function seems to be defective in this strain, as now explained in Results (L69-79). After confirming the problem with the RpoS-mCherry fusion, we removed all analyses and related arguments that relied on the RpoS expression level (previous Figure 4). In addition, we repeated almost all the experiments with the original MG1655 strain to confirm that the observed results are not specific to the problematic reporter strain. 

Regarding the experiments with CPFX, we have added a more detailed analysis of single cell dynamics and found that, contrary to the reported results for ofloxacin, not all persistent cells show filamentation after drug withdrawal (Fig. 8C and D, Fig. 8 – figure supplement 1). In addition, we performed new microfluidic experiments in which we treated post-late stationary phase cells with CPFX (Fig. 3). In contrast to the Amp treatment result and the previous study that reported the persistence of post-stationary phase cell populations to ofloxacin (ref. 20), all the persisters for which we identified the pre-exposure growth traits in this condition grew normally prior to CPFX treatment. These newly added analyses and experiments clarify the significance of the CPFX experiments. 

(4) The authors don't mention the dead volume nor the speed of media exchange in their device. Hopefully, it is short compared to the duration of the treatment; however, it is challenging to remove all antibiotics after the treatment and only 1e-3 or 1e-4 of the treatment concentration is already susceptible to affecting regrowth in fresh media. If this is described in another article, it would be worth adding a comment in the main text.

We thank the reviewer for bringing up this important point. We have added the perfusion chamber volume and medium flow rate information in the Methods section (L809-817).   

In the study in which two of the authors participated, the medium exchange rate across the semipermeable membrane was evaluated in a similar device with similar microchamber dimensions (ref. 26). There, we confirmed that the medium exchange was completed within 5 min, which is much shorter than the period of antibiotic treatment and post-antibiotic treatment periods for observing regrowth. We have also included this information in the main text with the reference (L58-63).

Despite the relatively high medium exchange rate, we cannot formally exclude the possibility that a small amount of antibiotic may remain in the device, e.g. due to non-specific adsorption on the internal surface of the microchambers. In such cases, the residual antibiotics may influence the physiological states of the cells and the regrowth kinetics in the post-exposure periods, as suggested by the reviewer. However, the frequencies of persister cells in the cell populations in our single-cell measurements are comparable to those in the batch culture measurements. Therefore, the removal of antibiotic drugs in our device is at least as efficient as in the batch culture assay. To clarify this point, we have added a paragraph to the Discussion with a reference that reviews the influence of antibiotics at concentrations significantly lower than the MICs (L482-

489).    

(5) Fig 2A supports the main finding that a significant fraction of bacteria surviving the treatment are growing before drug exposure, but it uses a poorly chosen representation.

- In order to compare between conditions, one would like to see the fraction of each type in the population.

- The current representation (of a fraction of each type among surviving cells) requires a side-byside comparison with a random sample (which will practically be equivalent to the fraction of each type among killed cells) in order to be informative.

We have changed the style of the previous Fig. 2A to show the fraction of each type in the population instead of the fraction of each type among surviving cells (Fig. 3 and Fig. 3-figure supplement 1).

Author response:

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

Reviewer #1 (Public review):

Weiler, Teichert, and Margrie systematically analyzed long-range cortical connectivity, using a retrograde viral tracing strategy to identify layer and region-specific cortical projections onto the primary visual, primary somatosensory, and primary motor cortices. Their analysis revealed several hundred thousand inputs into each region, with inputs originating from almost all cortical regions but dominated in number by connections within cortical sub-networks (e.g. anatomical modules). Generally, the relative areal distribution of contralateral inputs followed the distribution of corresponding ipsilateral inputs. The largest proportion of inputs originated from layer 6a cells, and this layer 6 dominance was more pronounced for contralateral than ipsilateral inputs, which suggests that these connections provide predominantly feedback inputs. The hierarchical organization of input regions was similar between ipsi- and contralateral regions, except for within-module connections, where ipsilateral connections were much more feed-forward than contralateral. These results contrast earlier studies which suggested that contralateral inputs only come from the same region (e.g. V1 to V1) and from L2/3 neurons. Thus, these results provide valuable data supporting a view of interhemispheric connectivity in which layer 6 neurons play an important role in providing modulatory feedback.

The conclusions of this paper are mostly well-supported by the data and analysis, but additional consideration of possible experimental biases is needed.

We thank the reviewer for their positive feedback on our manuscript.

Further discussion or analysis is needed about possible biases in uptake efficiency for different cell types. Is it possible that the nuclear retro-AAV has a tropism for layer 6 axons? Quantitative comparisons with results obtained with alternative methods such as rabies virus (Yao et al., 2023) or anterograde tracing (Harris et al., 2019) may be helpful for this.

We appreciate this technical comment. For the reasons indicated below we are confident that our AAV approach successfully and rather comprehensively labels inputs to the three target areas. Firstly, in the brains in which we injected our retrograde nuclear-AAV tracer into VISp, SSp-bfd or MOp we found several instances where layer 5 and/or layer 2/3 as was the dominant cortical projection layer (please see e.g. Figure 3 heatmaps). This was true for both ipsilateral and contralateral projection. 

Secondly, by way of comparison Yao et al., 2023 injected rabies virus into VISp (but not in SSp-bfd or MOp) and their results show notable similarities to ours: 1) They show that contralateral inputs to VISp (and higher visual areas) were mainly located in Layers 5 and 6. 2) Retrogradely labelled neurons in higher visual areas revealed anatomical hierarchy that reflects the known functional hierarchy of the mouse cortical visual system and that shown by our retro-AAV approach. Thus, as AAV and rabies based tracing lead to similar results, this is further evidence against bias via tropism of our AAV tracer. That said, direct comparisons of the results between our study and the Yao et al., 2023 study should be viewed with some caution since Yao et. al.  injected rabies virus into specific Cre-driver lines in which the rabies virus targets individual genetically defined cell types in specific layers. Importantly, because of the lack of a specific cre-driver line, L6 cortico-cortical (L6 CC) cells could not be targeted by their approach. Thus, the dataset in Yao et al., overlook the contribution of L6 CCs due to the lack of available Cre-lines. 

Thirdly, in a recent study (Weiler et al., 2024) we found that in a specific pathway (SSp-bfd→ VISp) both retro-AAV and the more traditional non-viral tracer cholera toxin subunit B (CTB) identified neurons in Layer 6 as the main source of projection neurons. The same results for the same pathway was shown by Bieler et al., 2019 (Bieler et al., 2017) using Fluorogold for retrograde tracing. Thus, the described dominance of Layer 6 projection neurons in specific pathways is likely not the result of a tropism of retro-AAV tracers. 

Please also see that we have now further extended the summary of these points in our revised manuscript in the discussion section (e.g. lines 457-463): 

Quantitative analysis of the injection sites should be included to account for possible biases. For example, L6 neurons are known to be the main target of contralateral inputs into the visual cortex (Yao et al., 2023). Thus, if the injections are biased towards or against layer 6 neurons, this may change the layer distribution of retrogradely labeled input cells. Comparison across biological replicates may help reveal sensitivity to particular characteristics of the injections.

In response to the reviewers' feedback, please see we have now quantified the injection volume per cortical layer, as shown in the revised Fig. S3D. Our results indicate that the injections were not biased toward Layer 6. Instead, the injected tracer volumes in Layers 1, 4, 5, and 6 were similar across all animals and injected areas. However, we observed that the injected tracer volume in Layer 2/3 tended to be higher than in other layers. Although the tracer volumes in Layers 2/3 appeared to be higher, the proportion of input neurons located in Layers 2/3 for most of the cortical projection areas was consistently lower than that from Layer 6. These findings provide strong evidence against injection bias towards L6 inputs.

The possibility of labelling axons of passage within the white matter should be addressed. This could potentially lead to false positive connections, contributing to the broad connectivity from most cortical regions that were observed.

For clarification, please see Fig.S2B in our revised manuscript. In this panel we plot the average percentage volume of the viral boli in the target areas and in all other nearby structures including the white matter. The percentage of virus injected into the white matter (WM) was 0.0824 ± 0.0759% for VISp and 0.0650 ± 0.0481 for SSp-bfd injections. Notably, injections into MOp showed no leakage into white matter (0%). These minimal volumes of virus in the white matter are unlikely to significantly influence the observed profile of widespread connectivity. Please see we have added a sentence to the Results section (lines 84-86) where we state that we only used brains that had a transduction of the white matter below 0.1%.

Reviewer #2 (Public review):

Summary:

Weiler et al use retrograde tracers, two-photon tomography, and automatic cell detection to provide a detailed quantitative description of the laminar and area sources of ipsi- and contralateral cortico-cortical inputs to two primary sensory areas and a primary motor area. They found considerable bilateral symmetry in the areas providing cortico-cortical inputs. However, although the same regions in both hemispheres tended to supply inputs, a larger proportion of inputs from contralateral areas originated from deeper layers (L5 and L6).

Strengths:

The study applies state-of-the-art anatomical methods, and the data is very effectively presented and carefully analyzed. The results provide many novel insights into the similarities and differences of inputs from the two hemispheres. While over the past decade there have been many studies quantitatively and comprehensively describing cortico-cortical connections, by directly comparing inputs from the ipsi and contralateral hemispheres, this study fills in an important gap in the field. It should be of great utility and an important reference for future studies on inter-hemispheric interactions.

We thank the reviewer for this encouraging feedback on our manuscript.

Weaknesses:

Overall, I do not find any major weakness in the analyses or their interpretation. However, one must keep in mind that the study only analyses inputs projecting to three areas. This is not an inherent flaw of the study; however, it warrants caution when extrapolating the results to callosal projections terminating in other areas. As inputs to two primary sensory areas and one is the primary motor cortex are studied, some of the conclusions could potentially be different for inputs terminating in high-order sensory and motor areas. Given that primary areas were injected, there are few instances of feedforward connections sampled in the ipsilateral hemisphere. The study finds that while ipsi-projections from the visual cortex to the barrel cortex are feedforward given its fILN values, those from the contralateral visual cortex are feedback instead. One is left to wonder whether this is due to the cross-modal nature of these particular inputs and whether the same rule (that contralateral inputs consistently exhibit feedback characteristics regardless of the hierarchical relationship of their ipsilateral counterparts with the target area,) would also apply to feedforward inputs within the same sensory cortices.

We acknowledge that what we find for primary sensory and motor target areas may not hold for other functionally different areas such as anterior cingulate cortex, retrosplenial cortex or frontal lobe that might be expected to receive strong feedforward cortical input. To begin to understand the organization of the global cortical input we have however first explored with primary sensory and motor areas. Please see that we have now added a sentence to the Discussion section of our manuscript that highlights the importance of investigating the hierarchical organization of intra and interhemispheric input onto higher cortical areas or within subregions of a given sensory area.

Another issue that is left unexplored is that, in the current analyses the barrel and primary visual cortex are analyzed as a uniform structure. It is well established that both the laminar sources of callosal inputs and their terminations differ in the monocular and binocular areas of the visual cortex (border with V2L). Similarly, callosal projections differ when terminating the border of S1 (a row of whiskers), and then in other parts of S1. Thus, some of the conclusions regarding the laminar sources of callosal inputs might depend on whether one is analyzing inputs terminating or originating in these border regions.

The aim of the present study was to analyse the global projectome to the VISp, SSp-bfd and MOp, irrespective of which subregions were included. Importantly, we purposely injected rather large bolus volumes to achieve large sample sizes of target neurons in each cortical layer.  For SSp-bfd, we utilised our previously reconstructed barrel map (Weiler et al., 2024) to precisely map our viral injection sites onto the barrels (Author response image 1). Analysis revealed that the six injection sites consistently encompassed 7–13 barrels (Author response image 1, three exemplary injection sites). Additionally, we determined the centres of mass for each injection site and mapped them onto the barrel map. Four of the injection sites were located in the lateral part of SSp-bfd, two in the central region, and none in the medial part. Notably, the injection sites within SSp-bfd exhibited significant overlap. As a result, a selective analysis of callosal projections targeting these injection sites would likely not yield distinct projection patterns, as the projectomes would inevitably include projections to surrounding barrels, leading to contamination.

Author response image 1.

Left: exemplary Injection sites mapped onto the 3D barrel map of SSp-bfd within the Mouse Allen Brain Atlas. Barrels were reconstructed using a specialized software as described previously (Weiler et al., 2024) Right: Centres of mass of all SSp-bfd injection sites mapped onto the 3D barrel map.

Due to the fact we covered a significant proportion of the respective target primary sensory area any further subdivision of these data is not possible and requires more tailored injections into clearly defined subareas. Investigating the separate projectomes onto these subregions (e.g. onto V1M and V1B) remains an important interesting research question that we, at least in part, will address in a future study.

Finally, while the paper emphasizes that projections from L6 "dominate" intra and contralateral cortico-cortical inputs, the data shows a more nuanced scenario. While it is true that the areas for which L6 neurons are the most common source of cortico-cortical projections are the most abundant, the picture becomes less clear when considering the number of neurons sending these connections. In fact, inputs from L2/3 and L5 combined are more abundant than those from L6 (Figure 3B), challenging the view that projections from L6 dominate ipsi- and contralateral projecting cortico-cortical inputs.

We agree in the case of the barrel cortex, layer 5 significantly contributes in terms of the number of brain areas projecting from within the ipsilateral and contralateral hemispheres. Please see we have replaced the term “dominates” in the title, abstract and in the manuscript where relevant.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors):

The sections analyzing the role of L6 towards feedback (pg. 11-13, Figure 6) were a bit verbose and confusing to me. Three possible models are proposed:

(1) a decrease in L23 projections, (2) an increase in L56 projections, or (3) both.

However, what is being quantified appears to be the fractions inputs, with L23. L5, and L6 summing to 1. Thus, a decrease in L23 would necessarily result in an increase in L56 projections. It seems like it would make more sense to quantify the percent change in the total number of inputs (rather than fractional) from each layer so that the 3 models are actually independent possibilities.

The issue with the suggested analysis is that, with one exception (one area projecting to MOp), the number of projection neurons in contralateral areas is always ~60-80% lower compared to their ipsilateral counterparts. Consequently, this is also true for the number of projection neurons in the different cortical layers. Thus, quantifying the percentage change from the ipsilateral to the contralateral hemisphere in the total number of inputs from each layer will always result in negative values. 

Nevertheless, we addressed the reviewer’s issue by calculating the preservation index (1(ipsi-contra)/(ipsi+contra)) for the sensory-motor areas independently for the absolute number of neurons within L2/3, 5 and 6 for the cortical areas projecting to VISp, SSp-bfd and MOp (see Author response image 2). When analysing the shift from the ipsilateral to the contralateral hemisphere, we observed that significantly more projection neurons were preserved in L6 compared to L2/3 for VISp and SSp-bfd. This shows that the number of L6 projection neurons declines less from the ipsilateral to the contralateral hemisphere compared to L2/3. However, our focus was on the fraction of projection neurons within each layer relative to the other layers per hemisphere (see Fig.6 of our manuscript). This measure is critical for distinguishing between feedforward and feedback connectivity. Calculating the change for each layer independently unfortunately does not provide insights into this comparison, as it does not capture the relative distribution of projection neurons across layers, which is central to our analysis. Therefore, we chose to present the data as layer fractions normalised within each hemisphere separately, enabling a comparison of relative changes between hemispheres, as shown in Fig.6 in the manuscript. We agree that with our approach a decrease in the fraction of L2/3 neurons would necessarily lead to an increase in the fraction of L5+6 neurons. However, as we analysed the fractional change for L5 and L6 separately, we found that the fraction of projection neurons in L5 generally showed only minor changes, while the fraction of L6 projection neurons increased substantially (Fig.6C). In addition, excluding L5 from the ipsi- or contralateral default network had significant effects on the fILN in only a relatively small number of projection areas. Excluding L6 resulted in significant changes in many more projection areas than layer 5.

Author response image 2.

Preservation index for L2/3, L5 and L6 of the 24 sensory-motor areas projecting onto the three target areas VISp, SSp-bfd and MOp.

Reviewer #2 (Recommendations for the authors):

I feel that there are a few conclusions that could be strengthened in the paper:

(1) The laminar sources of callosal inputs and their terminations differ in the monocular and binocular areas of the visual cortex (border with V2L. Similarly, callosal inputs are different close to the border of S1 with S2 than in the rest of the barrel cortex. From the methods sections and Figure S2, it seems that some injections targeted the V1 binocular zone while others were aimed at the monocular zone. Thus, it would be of interest to compare the laminar distribution and fILM of the contra inputs in inputs to the binocular and monocular zones (and S1 border vs the rest, if possible within this dataset).

Please see the answer for the reviewer’s second point in the public review (above).

(2) The results are currently a bit unclear on whether the contra inputs reflect the cortical hierarchy. Figure 4E-F makes it clear that the ipsi and contra fILMs do not always match. However, it seems from the plots in Figure 4D and Figure S6 that, while the contra fILM values are always higher, there might be a correlation between the ipsi and contra fILM. This could be addressed by directly plotting contra vs ipsi fILM.

Similarly, it would be useful to directly address if there is any hint of the visual hierarchy, as calculated in Figure S5 for the contra inputs.

Regarding the first point of the reviewer: We appreciate this comment. We do indeed find a positive correlation between the fILN ipsilateral and fILN contralateral across the individual cortical areas for all three targets. (please see Author response image 3 below). This is indeed an interesting observation that indicates a high degree of preservation concerning the rank order of the anatomical hierarchy within the input arising from both hemispheres. Please see we have included this new figure 4F into the manuscript and added a sentence in the results (lines 282-288): 

Regarding the second point of the reviewer: For visual hierarchy, although weaker, we find that the hierarchical ranking of the higher cortical visual areas is preserved for the contralateral hemisphere (see Author response image 3 below). 

Author response image 3.

Rank ordered average fILN values (± sem) of higher visual cortical areas of the ventral and dorsal visual stream for the ipsilateral and contralateral hemisphere.

(3) I find the emphasis in the title and other parts of the paper on Layer 6 corticocortical cells dominating the anatomical organization of intra and interhemispheric feedback a bit of an overstatement. While it is true that the areas for which L6 is the most abundant source of cortico-cortical projections are the most abundant (Figure 3C), when just focusing on the number of neurons sending corticocortical connections (Figure 3B), this is less clear. Ipsi connections are roughly divided 1/3, 1/3 , 1/3 between L2/3 , L5 and L6. In the contra, while projections from L6 neurons are the most abundant, there are not a majority and are less than those of L2/3 and L5 together. I suggest revising the statement about L6 cells dominating cortico-cortical connections to more accurately reflect these nuances.

(4) The observations from Figure 3 discussed above suggest that L6 inputs dominate in areas with less abundant projections to the injected areas. Is this the case? Is the fraction of L6 inputs inversely correlated with the number of inputs from that area?

Please see the following correlation plots for the total number of inputs versus the fraction of L6 inputs per area for both the ipsilateral and contralateral hemisphere. We do find on the ipsilateral hemisphere a negative correlation between the total number of inputs and the L6 input fraction for VISp and to a lesser degree for SSp-bfd. Interestingly, we find the opposite correlation for the ipsilateral MOp, contralateral VISp, SSp-bfd and MOp (Author response image 4, Author response table 1). While this is an interesting finding, the correlations often appeared to be weak and often absent within the individual animals and across the three target areas (Author response table 1). Thus, these correlations are seemingly not a general feature of cortical connectivity.

Author response image 4.

Total number of cells versus fraction of cells within L6 per cortical brain area (average across animals) for the ipsilateral (top) and contralateral (bottom) hemisphere for the three target areas VISp, SSp-bfd and MOp.

Author response table 1: Respective correlations between total numbers of cells and fraction of cells within L6 per cortical brain area for the ipsilateral and contralateral hemisphere for the three target areas (significant correlations highlighted with green).

Minor issues:

(5) Where was the mouse in Figure 3A injected?

In this exemplary mouse the retrograde tracer was injected into VISp. We added this information in the Figure legend of Figure 3A1. 

(6) Clarify in panel 4F that the position of the circle corresponds to the area location.

Done as suggested. 

References

Bieler M, Sieben K, Cichon N, Schildt S, Röder B, Hanganu-Opatz IL. 2017. Rate and Temporal Coding Convey Multisensory Information in Primary Sensory Cortices. eNeuro 4. doi:10.1523/ENEURO.0037-17.2017

Weiler S, Rahmati V, Isstas M, Wutke J, Stark AW, Franke C, Graf J, Geis C, Witte OW, Hübener M, Bolz J, Margrie TW, Holthoff K, Teichert M. 2024. A primary sensory cortical interareal feedforward inhibitory circuit for tacto-visual integration. Nat Commun 15:3081. doi:10.1038/s41467-024-47459-2

Yao S, Wang Q, Hirokawa KE, Ouellette B, Ahmed R, Bomben J, Brouner K, Casal L, Caldejon S, Cho A, Dotson NI, Daigle TL, Egdorf T, Enstrom R, Gary A, Gelfand E, Gorham M, Griffin F, Gu H, Hancock N, Howard R, Kuan L, Lambert S, Lee EK, Luviano J, Mace K, Maxwell M, Mortrud MT, Naeemi M, Nayan C, Ngo N-K, Nguyen T, North K, Ransford S, Ruiz A, Seid S, Swapp J, Taormina MJ, Wakeman W, Zhou T, Nicovich PR, Williford A, Potekhina L, McGraw M, Ng L, Groblewski PA, Tasic B, Mihalas S, Harris JA, Cetin A, Zeng H. 2023. A whole-brain monosynaptic input connectome to neuron classes in mouse visual cortex. Nat Neurosci 26:350–364. doi:10.1038/s41593-022-01219-x

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study aims to identify the proteins that compose the electrical synapse, which are much less understood than those of the chemical synapse. Identifying these proteins is important to understand how synaptogenesis and conductance are regulated in these synapses. The authors identified more than 50 new proteins and used immunoprecipitation and immunostaining to validate their interaction of localization. One new protein, a scaffolding protein, shows particularly strong evidence of being an integral component of the electrical synapse. However, many key experimental details are missing (e.g. mass spectrometry), making it difficult to assess the strength of the evidence.

Strengths:

One newly identified protein, SIPA1L3, has been validated both by immunoprecipitation and immunohistochemistry. The localization at the electrical synapse is very striking.<br /> A large number of candidate interacting proteins were validated with immunostaining in vivo or in vitro.

Weaknesses:

There is no systematic comparison between the zebrafish and mouse proteome. The claim that there is "a high degree of evolutionary conservation" was not substantiated.

We agree that we should have included a comprehensive comparison of proteins captured in the different species.  We are assembling this table and it will be included in the revised manuscript.  There is, indeed, significant conservation of many of the proteins enriched in both species.

No description of how mass spectrometry was done and what type of validation was done.

Since the mass spec was outsourced to a core facility, we had not included methodological details.  We have requested these and will include full details in the revised version of the manuscript.  In terms of “validation,” enrichment of proteins at electrical synapses was determined based on capture relative to control samples (non-transgenic zebrafish retinas or non-transgenic mouse retinas infected with the dGBP-TurboID virus) captured and processed at the same time.  Actual validations based on protein co-localization and pull-downs is the subject of the rest of the manuscript, and could only be done for a fraction of the identified proteins.  This type of validation can be pursued in many future studies. 

The threshold for enrichment seems arbitrary.

Yes, the thresholds are somewhat arbitrary.  This is due to the fact that experiments that captured larger total amounts of protein (mouse retina samples) had higher signal-to-noise ratio than those that captured smaller total amounts of protein (zebrafish retina).  This allowed us to use a more stringent threshold in the mouse dataset to focus on high probability captured proteins. 

Inconsistent nomenclature and punctuation usage.

We have scanned through the manuscript and updated terms that were used inconsistently in the interim revision of the manuscript.

To describe the mass spec procedure, we will get in touch with the mass spec facility and provide the details in the next round of submission.

The description of figures is very sparse and error-prone (e.g. Figure 6).

In Figure 1B, there is very broad non-specific labeling by avidin in zebrafish (In contrast to the more specific avidin binding in mice, Figure 2B). How are the authors certain that the enrichment is specific at the electrical synapse?

The enrichment of the proteins we identified is specific for electrical synapses because we compared the abundance of all candidates between Cx35b-V5-TurboID and wildtype retinas. Proteins that are components of electrical synapses, will only show up in the Cx35b-V5-TurboID condition. The western blot (Strep-HRP) in figure 1C shows the differences in the streptavidin labeling and hence the enrichment of proteins that are part of electrical synapses. Moreover, while the background appears to be quite abundant in sections, biotinylation is a rare posttranslational modification and mainly occurs in carboxylases: The two intense bands that show up above 50 and 75 kDa.  The background mainly originates from these two proteins.

In Figure 1E, there is very little colocalization between Cx35 and Cx34.7. More quantification is needed to show that it is indeed "frequently associated."

We agree that “frequently associated” is too strong as a statement. We corrected this and instead wrote “that Cx34.7 was only expressed in the outer plexiform layer (OPL) where it was associated with Cx35b at some gap junctions” in line 150. There are many gap junctions at which Cx35b is not colocalized with Cx34.7. 

Expression of GFP in HCs would potentially be an issue, since GFP is fused to Cx36 (regardless of whether HC expresses Cx36 endogenously) and V5-TurboID-dGBP can bind to GFP and biotinylate any adjacent protein.  

Thank you for this suggestion! There should be no Cx36-GFP expression in horizontal cells, which means that the nanobody cannot bind to anything in these cells. Moreover, to recognize specific signals from non-specific background, we included wild type retinas throughout the entire experiments. This condition controls for non-specific biotinylation.

Figure 7: the description does not match up with the figure regarding ZO-1 and ZO-2.

It appears that a portion of the figure legend was left out of the submitted version of the manuscript.  We have put the legend for panels A through C back into the manuscript in the interim revision.

Reviewer #2 (Public review):

Summary:

This study aimed to uncover the protein composition and evolutionary conservation of electrical synapses in retinal neurons. The authors employed two complementary BioID approaches: expressing a Cx35b-TurboID fusion protein in zebrafish photoreceptors and using GFP-directed TurboID in Cx36-EGFP-labeled mouse AII amacrine cells. They identified conserved ZO proteins and endocytosis components in both species, along with over 50 novel proteins related to adhesion, cytoskeleton remodeling, membrane trafficking, and chemical synapses. Through a series of validation studies¬-including immunohistochemistry, in vitro interaction assays, and immunoprecipitation - they demonstrate that novel scaffold protein SIPA1L3 interacts with both Cx36 and ZO proteins at electrical synapse. Furthermore, they identify and localize proteins ZO-1, ZO-2, CGN, SIPA1L3, Syt4, SJ2BP, and BAI1 at AII/cone bipolar cell gap junctions.

Strengths:

The study demonstrates several significant strengths in both experimental design and validation approaches. First, the dual-species approach provides valuable insights into the evolutionary conservation of electrical synapse components across vertebrates. Second, the authors compare two different TurboID strategies in mice and demonstrate that the HKamac promoter and GFP-directed approach can successfully target the electrical synapse proteome of mouse AII amacrine cells. Third, they employed multiple complementary validation approaches - including retinal section immunohistochemistry, in vitro interaction assays, and immunoprecipitation-providing evidence supporting the presence and interaction of these proteins at electrical synapses.

Weaknesses:

The conclusions of this paper are supported by data; however, some aspects of the quantitative proteomics analysis require clarification and more detailed documented. The differential threshold criteria (>3 log2 fold for mouse vs >1 log2 fold for zebrafish) will benefit from biological justification, particularly given the cross-species comparison. Additionally, providing details on the number of biological or technical replicates used in this study, along with analyses of how these replicates compare to each other, would strengthen the confidence in the identification of candidate proteins. Furthermore, including negative controls for the histological validation of proteins interacting with Cx36 could increase the reliability of the staining results.

While the study successfully characterized the presence of candidate proteins at the electrical synapses between AII amacrine cells and cone bipolar cells, it did not compare protein compositions between the different types of electrical synapses within the circuit. Given that AII amacrine cells form both homologous (AII-AII) and heterologous (AII-cone bipolar cell) electrical synapses-connections that serve distinct functional roles in retinal signaling processing-a comparative analysis of their molecular compositions could have provided important insights into synapse specificity.

Reviewer #3 (Public review):

Summary:

This study by Tetenborg S et al. identifies proteins that are physically closely associated with gap junctions in retinal neurons of mice and zebrafish using BioID, a technique that labels and isolates proteins proximal to a protein of interest. These proteins include scaffold proteins, adhesion molecules, chemical synapse proteins, components of the endocytic machinery, and cytoskeleton-associated proteins. Using a combination of genetic tools and meticulously executed immunostaining, the authors further verified the colocalizations of some of the identified proteins with connexin-positive gap junctions. The findings in this study highlight the complexity of gap junctions. Electrical synapses are abundant in the nervous system, yet their regulatory mechanisms are far less understood than those of chemical synapses. This work will provide valuable information for future studies aiming to elucidate the regulatory mechanisms essential for the function of neural circuits.

Strengths:

A key strength of this work is the identification of novel gap junction-associated proteins in AII amacrine cells and photoreceptors using BioID in combination with various genetic tools. The well-studied functions of gap junctions in these neurons will facilitate future research into the functions of the identified proteins in regulating electrical synapses.

Thank you for these comments.

Weaknesses:

I do not see major weaknesses in this paper. A minor point is that, although the immunostaining in this study is beautifully executed, the quantification to verify the colocalization of the identified proteins with gap junctions is missing. In particular, endocytosis component proteins are abundant in the IPL, making it unclear whether their colocalization with gap junction is above chance level (e.g. EPS15l1, HIP1R, SNAP91, ITSN in Figure 3B).

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors explore associations between plasma metabolites and glaucoma, a primary cause of irreversible vision loss worldwide. The study relies on measurements of 168 plasma metabolites in 4,658 glaucoma patients and 113,040 controls from the UK Biobank. The authors show that metabolites improve the prediction of glaucoma risk based on polygenic risk score (PRS) alone, albeit weakly. The authors also report a "metabolomic signature" that is associated with a reduced risk (or "resilience") for developing glaucoma among individuals in the highest PRS decile (reduction of risk by an estimated 29%). The authors highlight the protective effect of pyruvate, a product of glycolysis, for glaucoma development and show that this molecule mitigates elevated intraocular pressure and optic nerve damage in a mouse model of this disease.

Strengths:

This work provides additional evidence that glycolysis may play a role in the pathophysiology of glaucoma. Previous studies have demonstrated the existence of an inverse relationship between intraocular pressure and retinal pyruvate levels in animal models (Hader et al. 2020, PNAS 117(52)) and pyruvate supplementation is currently being explored for neuro-enhancement in patients with glaucoma (De Moraes et al. 2022, JAMA Ophthalmology 140(1)). The study design is rigorous and relies on validated, standard methods. Additional insights gained from a mouse model are valuable.

We thank the reviewer for these supportive comments.

Weaknesses:

Caution is warranted when examining and interpreting the results of this study. Among all participants (cases and controls) glaucoma status was self-reported, determined on the basis of ICD codes or previous glaucoma laser/surgical therapy. This is problematic as it is not uncommon for individuals in the highest PRS decile to have undiagnosed glaucoma (as shown in previous work by some of the authors of this article). The authors acknowledge a "relatively low glaucoma prevalence in the highest decile group" but do not explore how undiagnosed glaucoma may affect their results. This also applies to all controls selected for this study. The authors state that "50 to 70% of people affected [with glaucoma] remain undiagnosed". Therefore, the absence of self-reported glaucoma does not necessarily indicate that the disease is not present. Validation of the findings from this study in humans is, therefore, critical. This should ideally be performed in a well-characterized glaucoma cohort, in which case and control status has been assessed by qualified clinicians.

We appreciate the comment regarding the challenges of glaucoma ascertainment in UK Biobank. This is a valid limitation, as glaucoma in UK Biobank is based on self-reports and hospital records rather than comprehensive ophthalmologic examinations for all participants. To the best of our knowledge, there is no comparably sized dataset where all participants have undergone standardized glaucoma assessments, comprehensive metabolomic profiling, and high-throughput genotyping. Work is currently ongoing to link UK Biobank data to ophthalmic EMR data, which will help confirm self-reported diagnoses. This work is not complete, and the coverage of the cohort from such linkage is uncertain at present. Nonetheless, several factors speak to the validity of our findings. The top members of the metabolomic signature associated with resilience in the top decile of glaucoma polygenic risk score (PRS) decile—lactate (P=8.8E-12) and pyruvate (P=1.9E-10) —had robust values for statistical significance after appropriate adjustment for multiple comparisons, with additional validation for pyruvate in a human-relevant, glaucoma mouse model. Strikingly, the glaucoma odds ratio (OR) for subjects in the highest quartile of glaucoma PRS and metabolic risk score (MRS) was 25-fold, using participants in the lowest quartile of glaucoma PRS and MRS as the reference group. An effect size this large for a putative glaucoma determinant has only been seen for intraocular pressure (IOP), which is now largely accepted to be in the causal pathway of the disease.

The Discussion now contains the following statement: “A second limitation is that glaucoma ascertainment in the UK Biobank is based on self-reported diagnoses and hospital records rather than comprehensive ophthalmologic examinations. Nonetheless, it is reassuring that the prevalence of glaucoma in our sample (~4%) is similar to a directly performed disease burden estimate in a comparable, albeit slightly older, United Kingdom sample (2.7%)(79)”. (Lines 379-382)

The authors indicate that within the top decile of PRS participants with glaucoma are more likely to be of white ethnicity, while they are more likely to be of Black and Asian ethnicity if they are in the bottom half of PRS. Have the authors explored how sensitive their predictions are to ethnicity? Since their cohort is predominantly of European ancestry (85.8%), would it make sense to exclude other ethnicities to increase the homogeneity of the cohort and reduce the risk for confounders that may not be explicitly accounted for?

Comparing data in Tables 3 and 4 of the manuscript, we observe that, on a percentage basis, more individuals have glaucoma in the highest 10th percentile of risk compared to the lowest 50th percentile of risk across all ancestral groups.  We recently reported that the risk of glaucoma increases with each standard deviation increase in the glaucoma PRS across ancestral groups in the UK Biobank, utilizing a slightly different sample size (see Author response table 1 below). (1)Since the PRS is applicable across ancestral groups, we aim to make our results as generalizable as possible; therefore, we prefer to report our findings for all ethnic groups and not restrict our results to Europeans.

Author response table 1.

Performance of the mtGPRS Across Ancestral Groups in the UK Biobank

Abbreviations: mtGPRS, multitrait analysis of GWAS polygenic risk score; OR, odds ratio; CI, confidence interval.(1)

UK Biobank ancestry was genetically inferred based on principal component analysis. The OR represents the risk associated with each standard deviation change in mtGRS and is adjusted for multiple covariates including age, sex, and medical comorbidities.

In the discussion, we stated that “... we chose to analyze Europeans and non-Europeans together to make the results as generalizable as possible.” (Lines 378-379)

The authors discuss the importance of pyruvate, and lactate for retinal ganglion cell survival, along with that of several lipoproteins for neuroprotection. However, there is a distinction to be made between locally produced/available glycolysis end products and lipoproteins and those circulating in the blood. It may be useful to discuss this in the manuscript, and for the authors to explore if plasma metabolites may be linked to metabolism that takes place past the blood-retinal barrier.

As the reviewer points out, it is crucial to interpret the results for lipoproteins within the context of their access to the blood-retinal barrier. Even for smaller metabolites like pyruvate and lactate, it is essential to consider local production versus serum-derived molecules in mediating any neuroprotective effects. Our murine data suggest that exogenous pyruvate contributed to neuroprotection. However, for the other glycolysis-related metabolites (lactate and citrate), we cannot rule out the possibility that locally produced metabolites may also contribute to neuroprotection. None of the lipoproteins identified as potential resilience biomarkers had an adjusted P-value of less than 0.05. Nevertheless, HDL analytes can cross blood-ocular barriers to enter the aqueous humor.(2) Therefore, it is also possible for serum-derived HDL to influence retinal ganglion cell homeostasis. Overall, much more research is needed to clarify the roles of locally produced versus serum-derived factors in conferring resilience to genetic predisposition to glaucoma.

We have added the following sentences to the discussion:

“Notably, although our validation data confirm the neuroprotective effects of exogenous pyruvate, it remains possible that endogenously produced pyruvate within ocular tissues may also contribute to RGC protection.” (Lines 329-331)

“Furthermore, as HDL analytes can cross blood-ocular barriers,(78) there is a plausible route for serum-derived HDL to influence RGC homeostasis. Nonetheless, the relative contributions of circulating lipoproteins versus local synthesis within ocular tissues remain unclear and warrant further investigation.” (Lines 355-358)

“Incorporating ocular physiology and blood-retinal barrier considerations when interpreting lipoproteins as potential resilience biomarkers will be critical for future studies aimed at understanding and therapeutically targeting increased glaucoma risk.” (Lines 360-363)

Reviewer #2 (Public review):

Summary

The authors have used the UK Biobank data to interrogate the association between plasma metabolites and glaucoma.

(1) They initially assessed plasma metabolites as predictors of glaucoma: The addition of NMR-derived metabolomic data to existing models containing clinical and genetic data was marginal.

(2) They then determined whether certain metabolites might protect against glaucoma in individuals at high genetic risk: Certain molecules in bioenergetic pathways (lactate, pyruvate, and citrate) conferred protection.

(3) They provide support for protection conferred by pyruvate in a murine model.

Strengths

(1) The huge sample size supports a powerful statistical analysis and the opportunity for the inclusion of multiple covariates and interactions without overfitting the models.

(2) The authors have constructed a robust methodology and statistical design.

(3) The manuscript is well written, and the study is logically presented.

(4) The figures are of good quality.

(5) Broadly, the conclusions are justified by the findings.

We thank the reviewer for these supportive comments.

Weaknesses

(1) Although it is an invaluable treasure trove of data, selection bias and self-reporting are inescapable problems when using the UK Biobank data for glaucoma research. The high-impact glaucoma-related GWAS publications (references 26 and 27) referenced in support of the method suffer the same limitations. This doesn't negate the conclusions but must be taken into consideration. The authors might note that it is somewhat reassuring that the proportion of glaucoma cases (4%) is close to what would be expected in a population-based study of 40-69-year-olds of predominantly white ethnicity.

While there are limitations when open-angle glaucoma (OAG) is ascertained by self-report, as discussed above, we agree with the reviewer that the prevalence of glaucoma is consistent with data from population-based studies of Europeans who are 40-69 years of age. 

We also want to point out that references 26 and 27 indicate glaucoma self-reports can be an acceptable surrogate for OAG that is ascertained by clinical evaluation. Consider the methodologic details for each study:

Reference 26 is a 4-stage genome-wide meta-analysis to identify loci for OAG from 21 independent populations. The phenotypic definition of OAG was based on clinical assessment in the discovery stage, and 7286 glaucoma self-reports from the UK Biobank served as an effective replication set.  It is also important to note that 120 out of the 127 discovered OAG loci were nominally replicated in 23andMe, where glaucoma was ascertained entirely by self-report.

Reference 27 is a genome-wide meta-analysis to identify IOP genetic loci, an important endophenotype for OAG. The study identified 112 loci for IOP. These loci were incorporated into a glaucoma prediction model in the NEIGHBORHOOD study and the UK Biobank. The area under the receiver operator curve was 0.76 and 0.74, respectively, in these studies. While the AUCs were similar, OAG was ascertained clinically in NEIGHBORHOOD and largely by self-report in UK Biobank. 

Finally, a strength of the UK Biobank is that selection bias is minimized. Patients need not be insured or aligned to the study for any reason aside from being a UK resident. There is indeed a healthy bias in the UK Biobank. Ambulatory patients who tend to be health conscious and willing to donate their time and provide biological specimens tend to participate. We agree with the reviewer that the use of self-reported cases does not negate the conclusions, and hopefully, future iterations of the UK Biobank where clinical validation of self-reports are performed will confirm these findings, which already have some validation in a preclinical glaucoma model.

We add the following sentence to the first action item above regarding our case ascertainment method. “Nonetheless, it is reassuring that the prevalence of glaucoma in our sample (~4%) is similar to a directly performed disease burden estimate in a comparable, albeit slightly older, United Kingdom sample (2.7%)..”(3) (Lines 381-383)

(2) As noted by the authors, a limitation is the predominantly white ethnicity profile that comprises the UK Biobank. 

(3) Also as noted by the authors, the study is cross-sectional and is limited by the "correlation does not imply causation" issue.

While the epidemiological arm of our study was cross-sectional, the studies testing the ability of pyruvate to mitigate the glaucoma phenotype in mice with the Lmxb1 mutation were prospective.

We already pointed out in the discussion that pyruvate supplementation reduced glaucoma incidence in a human-relevant genetic mouse model.

(4) The optimal collection, transport, and processing of the samples for NMR metabolite analysis is critical for accurate results. Strict policies were in place for these procedures, but deviations from protocol remain an unknown influence on the data.

Comments 4 and 5 are related and will be addressed after comment 5.

(5) In addition, all UK Biobank blood samples had unintended dilution during the initial sample storage process at UK Biobank facilities. (Julkunen, H. et al. Atlas of plasma NMR biomarkers for health and disease in 118,461 individuals from the UK Biobank. Nat Commun 14, 604 (2023) Samples from aliquot 3, used for the NMR measurements, suffered from 5-10% dilution. (Allen, Naomi E., et al. Wellcome Open Research 5 (2021): 222.) Julkunen et al. report that "The dilution is believed to come from mixing of participant samples with water due to seals that failed to hold a system vacuum in the automated liquid handling systems. While this issue is likely to have an impact on some of the absolute biomarker concentration values, it is expected to have limited impact on most epidemiological analyses."

We thank the reviewer for making us aware of the unintended sample dilution issue from aliquot 3, used for NMR metabolomics in UK Biobank participants. While ~98% of samples experienced a 5-10% dilution, this would not affect our reported results, which did not rely on absolute biomarker concentrations. All metabolites in the main tables were probit transformed and used as continuous variables per 1 standard deviation increase.  Nonetheless, in supplemental material, we show the unadjusted median levels of pyruvate (in mmol/L) were higher in participants without glaucoma vs those with glaucoma, both in the population overall and in those in the top 10 percentile of glaucoma risk. 

Furthermore, we see no evidence in the literature that unidentified protocol deviations might impact metabolite results in UK Biobank participants. For example, a recent study evaluated the relationship between a weighted triglyceride-raising polygenic score (TG.PS) and type 3 hyperlipidemia (T3HL) in the Oxford Biobank (OBB) and the UK Biobank. In both biobanks, metabolomics was performed on the Nightingale NMR platform. A one standard deviation increase in TG.PS was associated with a 13% and 15.2% increased risk of T3HL in the OBB and UK Biobank, respectively.(4) Replication of the OBB result in the UK Biobank suggests there are no additional concerns regarding the processing of the UK Biobank for NMR metabolomics. Of course, we remain vigilant for protocol deviations that might call our results into question and will seek to validate our findings in other biobanks in future research.

Impact

The findings advance personalized prognostics for glaucoma that combine metabolomic and genetic data. In addition, the protective effect of certain metabolites influences further research on novel therapeutic strategies.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Given the uncertainty in the proportion of controls with undiagnosed glaucoma, it may be appropriate to include a sensitivity analysis in the manuscript. The authors could then provide the readers with an estimate of how sensitive their predictions are to the proportion of undiagnosed individuals among controls.

Since UK Biobank participants did not undergo standardized clinical assessments, it is not possible to perform sensitivity analyses as we don’t know which controls might have glaucoma, although we can offer the following comments.

We are performing a cross-sectional, prospective, detailed glaucoma assessment of participants in the top and bottom 10 percent of genetic predisposition recruited from BioMe at Icahn School of Medicine at Mount Sinai and Mass General Brigham Biobank at Harvard Medical School. We find that 21% of people in the top decile of genetic risk have glaucoma,(5) which compares reasonably well to the 15% of people in the top 10% of genetic risk in the UK Biobank. This underscores the assertion that our definition of glaucoma in the UK Biobank, while not ideal, is a reasonable surrogate for a detailed clinical assessment.

Currently, 10,077 subjects in the top decile of glaucoma genetic predisposition did not meet our definition of glaucoma. If we assume that the glaucoma prevalence is 3% and 50% of people with glaucoma are undiagnosed, then that would translate to an additional 150 cases misclassified as controls, which could either drive our result to the null, have no impact on our current result or contribute to a false positive result, depending on their pyruvate (and other metabolite) levels.   

We have already addressed the issue of a lack of standardized exams in the UK Biobank and the need for more studies to confirm our findings.

Reviewer #2 (Recommendations for the authors):

(1) I am curious about the proposed reason for some individuals having metabolic profiles conferring resilience. Plasma pyruvate levels are normally distributed. Is it simply the case that some individuals with naturally high levels of pyruvate are fortuitously protected against glaucoma? Some sort of self-regulation mechanism seems unlikely.

Thank you for your insightful question regarding the potential mechanism underlying the association between pyruvate levels and glaucoma resilience. There may be modest inter-individual differences which can have significant physiological implications, particularly in the context of neurodegeneration and metabolic stress. One possibility is that individuals with naturally higher pyruvate levels may benefit from pyruvate's known neuroprotective and metabolic support functions(6–8), which could confer resilience against the oxidative and bioenergetic challenges associated with glaucoma. Pyruvate is important for cellular metabolism, redox balance, and mitochondrial function - processes that are increasingly implicated in glaucomatous neurodegeneration. (9)Elevated pyruvate levels support mitochondrial ATP production(10), buffer oxidative stress,(11) and impact metabolic flux(12,13) through pathways such as the tricarboxylic acid cycle and NAD+/NADH homeostasis. This is consistent with prior studies suggesting that mitochondrial dysfunction contributes to retinal ganglion cell vulnerability in glaucoma.(14–17) While a direct self-regulation mechanism may seem unlikely, both genetic and environmental factors can influence pyruvate metabolism, which could lead to subtle but clinically meaningful variations in its levels. Our findings are supported by validation in a mouse model, which suggests that the association is less likely fortuitous, but there may be an underlying biological process that merits further mechanistic investigation. Future studies incorporating longitudinal metabolic profiling and functional validation in human-derived models will help better understand this relationship.

(2) Conceivably, the higher levels of pyruvate and lactate may have resulted from recent exercise and may reflect individuals with high levels of exercise that confers resilience against glaucoma by independent mechanisms such as improved blood flow. Any way to rule that out from the UK Biobank data?

Thank you for raising this important point. To account for the potential confounding effects of physical activity, we adjusted for metabolic equivalents of task (METs) in our models, a standardized measure of physical activity available in the UK Biobank. By incorporating METs as a covariate, we aimed to minimize the influence of individual exercise levels on plasma pyruvate and lactate levels. This helps us ascertain that the observed associations are not solely attributable to differences in physical activity. However, we do acknowledge that longitudinal analysis of exercise patterns would provide further clarity on this relationship. 

(3) It may be worth mentioning that the retinal ganglion cells contain a plasma membrane monocarboxylate transporter that supports pyruvate and lactate uptake from the extracellular space.

Thank you for this extremely insightful suggestion on retinal ganglion cell (RGC) expression of monocarboxylate transporters, which can facilitate the uptake of pyruvate and lactate from the extracellular space. This is relevant for our study, given the high metabolic demands of RGCs and their reliance on both glycolytic and oxidative metabolism for neuroprotection and survival.

We acknowledged this in the discussion section of the manuscript by adding the following statement: "RGCs express monocarboxylate transporters, which facilitate the uptake of extracellular pyruvate and lactate, improving energy homeostasis, neuronal metabolism, and survival.” (Lines 309-311)

(4) The mechanism of protection in the mice, at least in part, is likely due to the lower IOP in the pyruvate-treated animals. Did the authors investigate the influence of pyruvate on IOP in the UK Biobank data (about 110,000 individuals had IOP measurements)?

Thank you for your suggested investigation. We ran the suggested analysis among 68,761 individuals with IOP measurements and metabolomic profiling. Imputed pretreatment IOP values for participants using ocular hypotensive agents were calculated by dividing the measured IOP by 0.7, based on the mean IOP.

We plotted the relationship between IOP and pyruvate levels (probit transformed). We compared participants with pyruvate levels +2 standard deviations, above the mean (red dashed line), which has a probit-transformed value of 2 and an absolute concentration of 0.15 mmol/L. We found a statistically significant difference between the groups (p=0.017) using the Welch two-sample t-test. We have not added this analysis to the manuscript, but readers can find the data here as the reviews are public. We acknowledge and addressed the dilutional issue above, where we utilized probit-transformed metabolite levels analyzed as continuous variables per 1 SD increase, rather than absolute concentrations.

Author response image 1.

(5) Line 88: I suggest changing "patients" to "affected individuals". The term "patients" tends to imply that the individual has already been diagnosed, but the idea being conveyed is about underdiagnosis in the population.

Thank you for your suggestion.

We have added the change from "patients" to "affected individuals" in the introduction. (Line 90)

(6) Line 93: "However, glaucoma is also significantly affected by environmental and lifestyle factors,10-14". Although lifestyle risk factors such as caffeine intake, alcohol, smoking, and air pollution have been reported, the associations are generally weak and inconsistently reported. Consider modifying this notion to stress the emerging evidence around gene-environment interactions (reference 14) rather than environmental factors per se, with the implication that genes + metabolism may be greater than the sum of the parts.

Thank you for this thoughtful suggestion to highlight gene-environment interactions, where genetic susceptibility may amplify or mitigate the impact of metabolic and environmental influences on glaucoma progression. We have revised the statement to better reflect the synergistic effects of genetics and metabolism rather than considering environmental factors in isolation.

Revised sentence for inclusion in the introduction of the manuscript: "Glaucoma risk is influenced by both genetic and metabolic factors, with emerging evidence suggesting that gene-environment interactions may play a greater role in conferring disease risk than independent exposures alone.” (Lines 95-97)

(7) Lines 156-161: In model 4, rather than stating that the very small increase in AUC with the addition of metabolic data compared to clinical and genetic data alone, "modestly enhances the prediction of glaucoma", it may be better interpreted as a marginal difference that was statistically significant due to the very large sample size but not clinically significant.

Thank you for your suggested comment.

We have adjusted the wording by changing “modestly” to “marginally” to address that the statistical significance is in the context of the study’s large sample size in the results section (Line 162) and throughout the manuscript.

NB: We made other minor edits to correct minor grammatical errors, improve clarity, and streamline the revised manuscript. Furthermore, the paragraph regarding slit lamp examination in the Methods was inadvertently omitted but is added back in the revised manuscript (Lines 571-579).

References:

(1) Kim J, Kang JH, Wiggs JL, et al. Does Age Modify the Relation Between Genetic Predisposition to Glaucoma and Various Glaucoma Traits in the UK Biobank? Invest Ophthalmol Vis Sci. 2025;66(2):57. doi:10.1167/iovs.66.2.57

(2) Cenedella RJ. Lipoproteins and lipids in cow and human aqueous humor. Biochim Biophys Acta BBA - Lipids Lipid Metab. 1984;793(3):448-454. doi:10.1016/0005-2760(84)90262-5

(3) Minassian DC, Reidy A, Coffey M, Minassian A. Utility of predictive equations for estimating the prevalence and incidence of primary open angle glaucoma in the UK. Br J Ophthalmol. 2000;84(10):1159-1161. doi:10.1136/bjo.84.10.1159

(4) Pieri K, Trichia E, Neville MJ, et al. Polygenic risk in Type III hyperlipidaemia and risk of cardiovascular disease: An epidemiological study in UK Biobank and Oxford Biobank. Int J Cardiol. 2023;373:72-78. doi:10.1016/j.ijcard.2022.11.024

(5) Zhao H, Pasquale LR, Zebardast N, et al. Screening by glaucoma polygenic risk score to identify primary open-angle glaucoma in two biobanks: An updated report. ARVO 2025 meeting. Published online 2025.

(6) Zilberter Y, Gubkina O, Ivanov AI. A unique array of neuroprotective effects of pyruvate in neuropathology. Front Neurosci. 2015;9. doi:10.3389/fnins.2015.00017

(7) Quansah E, Peelaerts W, Langston JW, Simon DK, Colca J, Brundin P. Targeting energy metabolism via the mitochondrial pyruvate carrier as a novel approach to attenuate neurodegeneration. Mol Neurodegener. 2018;13(1):28. doi:10.1186/s13024-018-0260-x

(8) Gray LR, Tompkins SC, Taylor EB. Regulation of pyruvate metabolism and human disease. Cell Mol Life Sci. 2014;71(14):2577-2604. doi:10.1007/s00018-013-1539-2

(9) Harder JM, Guymer C, Wood JPM, et al. Disturbed glucose and pyruvate metabolism in glaucoma with neuroprotection by pyruvate or rapamycin. Proc Natl Acad Sci. 2020;117(52):33619-33627. doi:10.1073/pnas.2014213117

(10) Kim MJ, Lee H, Chanda D, et al. The Role of Pyruvate Metabolism in Mitochondrial Quality Control and Inflammation. Mol Cells. 2023;46(5):259-267. doi:10.14348/molcells.2023.2128

(11) Wang X, Perez E, Liu R, Yan LJ, Mallet RT, Yang SH. Pyruvate Protects Mitochondria from Oxidative Stress in Human Neuroblastoma SK-N-SH Cells. Brain Res. 2007;1132(1):1-9. doi:10.1016/j.brainres.2006.11.032

(12) Tilton WM, Seaman C, Carriero D, Piomelli S. Regulation of glycolysis in the erythrocyte: role of the lactate/pyruvate and NAD/NADH ratios. J Lab Clin Med. 1991;118(2):146-152.

(13) Li X, Yang Y, Zhang B, et al. Lactate metabolism in human health and disease. Signal Transduct Target Ther. 2022;7(1):305. doi:10.1038/s41392-022-01151-3

(14) Zhang ZQ, Xie Z, Chen SY, Zhang X. Mitochondrial dysfunction in glaucomatous degeneration. Int J Ophthalmol. 2023;16(5):811-823. doi:10.18240/ijo.2023.05.20

(15) Ju WK, Perkins GA, Kim KY, Bastola T, Choi WY, Choi SH. Glaucomatous optic neuropathy: Mitochondrial dynamics, dysfunction and protection in retinal ganglion cells. Prog Retin Eye Res. 2023;95:101136. doi:10.1016/j.preteyeres.2022.101136

(16) Jassim AH, Inman DM, Mitchell CH. Crosstalk Between Dysfunctional Mitochondria and Inflammation in Glaucomatous Neurodegeneration. Front Pharmacol. 2021;12. doi:10.3389/fphar.2021.699623

(17) Yang TH, Kang EYC, Lin PH, et al. Mitochondria in Retinal Ganglion Cells: Unraveling the Metabolic Nexus and Oxidative Stress. Int J Mol Sci. 2024;25(16):8626. doi:10.3390/ijms25168626

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

By way of background, the Jiang lab has previously shown that loss of the type II BMP receptor Punt (Put) from intestinal progenitors (ISCs and EBs) caused them to differentiate into EBs, with a concomitant loss of ISCs (Tian and Jiang, eLife 2014). The mechanism by which this occurs was activation of Notch in Put-deficient progenitors. How Notch was upregulated in Put-deficient ISCs was not established in this prior work. In the current study, the authors test whether a very low level of Dl was responsible. But co-depletion of Dl and Put led to a similar phenotype as depletion of Put alone. This result suggested that Dl was not the mechanism. They next investigate genetic interactions between BMP signaling and Numb, an inhibitor of Notch signaling. Prior work from Bardin, Schweisguth and other labs has shown that Numb is not required for ISC self-renewal. However the authors wanted to know whether loss of both the BMP signal transducer Mad and Numb would cause ISC loss. This result was observed for RNAi depletion from progenitors and for mad, numb double mutant clones. Of note, ISC loss was observed in 40% of mad, numb double mutant clones, whereas 60% of these clones had an ISC. They then employed a two-color tracing system called RGT to look at the outcome of ISC divisions (asymmetric (ISC/EB) or symmetric (ISC/ISC or EB/EB)). Control clones had 69%, 15% and 16%, respectively, whereas mad, numb double mutant clones had much lower ISC/ISC (11%) and much higher EB/EB (37%). They conclude that loss of Numb in moderate BMP loss of function mutants increased symmetric differentiation which lead caused ISC loss. They also reported that numb¹⁵ and numb⁴ clones had a moderate but significant increase in ISC-lacking clones compared to control clones, supporting the model that Numb plays a role in ISC maintenance. Finally, they investigated the relevance of these observation during regeneration. After bleomycin treatment, there was a significant increase in ISC-lacking clones and a significant decrease in clone size in numb⁴ and numb¹⁵ clones compared to control clones. Because bleomycin treatment has been shown to cause variation in BMP ligand production, the authors interpret the numb clone under bleomycin results as demonstrating an essential role of Numb in ISC maintenance during regeneration.

Strengths:

(i) Most data is quantified with statistical analysis

(ii) Experiments have appropriate controls and large numbers of samples

(iii) Results demonstrate an important role of Numb in maintaining ISC number during regeneration and a genetic interaction between Mad and Numb during homeostasis.

Weaknesses:

(i) No quantification for Fig. 1

Quantification of Fig.1 has been added. 

(ii) The premise is a bit unclear. Under homeostasis, strong loss of BMP (Put) leads to loss of ISCs, presumably regardless of Numb level (which was not tested). But moderate loss of BMP (Mad) does not show ISC loss unless Numb is also reduced. I am confused as to why numb does not play a role in Put mutants. Did the authors test whether concomitant loss of Put and Numb leads to even more ISC loss than Put-mutation alone.

We have tested the genetic interaction between put and numb using Put RNAi and Numb RNAi driven by esg^ts. According to the results in this study and our previously published data, put mutant clone or esg^ts > Put-RNAi induced a rapid loss of ISC (whin 8 days). We did not observe further enhancement of stem cell loss phenotype in Put and Numb double RNAi guts.

(iii) I think that the use of the word "essential" is a bit strong here. Numb plays an important role but in either during homeostasis or regeneration, most numb clones or mad, numb double mutant clones still have ISCs. Therefore, I think that the authors should temper their language about the role of Numb in ISC maintenance.

We have revised the language and changed “essential” to important”.

Reviewer #2 (Public review):

Summary:

This work assesses the genetic interaction between the Bmp signaling pathway and the factor Numb, which can inhibit Notch signalling. It follows up on the previous studies of the group (Tian, Elife, 2014; Tian, PNAS, 2014) regarding BMP signaling in controlling stem cell fate decision as well as on the work of another group (Sallé, EMBO, 2017) that investigated the function of Numb on enteroendocrine fate in the midgut. This is an important study providing evidence of a Numb-mediated back up mechanism for stem cell maintenance.

Strengths:

(1) Experiments are consistent with these previous publications while also extending our understanding of how Numb functions in the ISC.

(2) Provides an interesting model of a "back up" protection mechanism for ISC maintenance.

Weaknesses:

(1) Aspects of the experiments could be better controlled or annotated:

(a) As they "randomly chose" the regions analyzed, it would be better to have all from a defined region (R4 or R2, for example) or to at least note the region as there are important regional differences for some aspects of midgut biology.

Thank you for the suggestion. In fact, we conducted all the analyses in region 4, we have added statement to clarify this in the revised manuscript.

(b) It is not clear to me why MARCM clones were induced and then flies grown at 18{degree sign}C? It would help to explain why they used this unconventional protocol.

We kept the flies at 18°C to avoid spontaneous clone.

(2) There are technical limitations with trying to conclude from double-knockdown experiments in the ISC lineage, such as those in Figure 1 where Dl and put are both being knocked down: depending on how fast both proteins are depleted, it may be that only one of them (put, for example) is inactivated and affects the fate decision prior to the other one (Dl) being depleted. Therefore, it is difficult to definitively conclude that the decision is independent of Dl ligand.

In our hand, Dl-RNAi is very effective and exhibited loss of N pathway activity (as determined by the N pathway reporter Su(H)-lacZ ) after RNAi for 8 days (Fig. 1D). Therefore, the ectopic Su(H)-lacZ expression in Punt Dl double RNAi (fig. 1E) is unlikely due to residual Dl expression. Nevertheless, we have changed the statement “BMP signaling blocks ligand-independent N activity” to” Loss of BMP signaling results in ectopic N pathway activity even when Dl is depleted”

(3) Additional quantification of many phenotypes would be desired.

(a) It would be useful to see esg-GFP cells/total cells and not just field as the density might change (2E for example).

We focused on R4 region for quantification where the cell density did not exhibit apparent change in different experimental groups. In addition, we have examined many guts for quantification. It is very unlikely that the difference in the esg-GFP+ cell number is caused by change in cell density.

(b) Similarly, for 2F and 2G, it would be nice to see the % of ISC/ total cell and EB/total cell and not only per esgGFP+ cell.

Unfortunately, we didn’t have the suggested quantification. However, we believe that quantification of the percentage of ISC or EB among all progenitor cells, as we did here, provides a meaningful measurement of the self-renewal status of each experimental group.

(c) Fig1: There is no quantification - specifically it would be interesting to know how many esg+ are su(H)lacZ positive in Put- Dl- condition compared to WT or Put- alone. What is the n?

Quantification of Fig.1 has been added. 

(d) Fig2: Pros + cells are not seen in the image? Are they all DllacZ+?

Anti-Pros and anti-E(spl)mβ-CD2 were stained in the same channel (magenta).  Pros+ exhibited “dot-like” nuclear staining while CD2 staining outlined the cell membrane of EBs. We have clarified this in the revised figure legend.

(e) Fig3: it would be nice to have the size clone quantification instead of the distribution between groups of 2 cell 3 cells 4 cell clones.

Because of the heterogeneity of clone size for each genotype, we chose to group clones based on their sizes ( 2, 3-6, 6-8, >8 cells) and quantified the distribution of individual groups for each genotype, which clearly showed an overall reduction in clone size for mad numb double mutant clones. We and others have used the same clone size analysis in previous studies (e.g., Tian and Jiang, eLife 2014).

(f) How many times were experiments performed?

All experiments were performed at least 3 times.

(4) The authors do not comment on the reduction of clone size in DSS treatment in Figure 6K. How do they interpret this? Does it conflict with their model of Bleo vs DSS?

Guts containing numb⁴ clones treated with DSS exhibited a slight reduction of clone size, evident by a higher percentage of 2-cell clones and lower percentage of > 8 cell clones. This reduction is less significant in guts containing numb¹⁵ clones. However, the percentage of Dl⁺-containing clones is similar between DSS and mock-treated guts. It is possible that ISC proliferation is lightly reduced due to numb⁴ mutation or the genetic background of this stock.

(5) There is probably a mistake on sentence line 314 -316 "Indeed, previous studies indicate that endogenous Numb was not undetectable by Numb antibodies that could detect Numb expression in the nervous system".

We have modified the sentence.

Reviewer #3 (Public review):

Summary:

The authors provide an in-depth analysis of the function of Numb in adult Drosophila midgut. Based on RNAi combinations and double mutant clonal analyses, they propose that Numb has a function in inhibiting Notch pathway to maintain intestinal stem cells, and is a backup mechanism with BMP pathway in maintaining midgut stem cell mediated homeostasis.

Strengths:

Overall, this is a carefully constructed series of experiments, and the results and statistical analyses provides believable evidence that Numb has a role, albeit weak compared to other pathways, in sustaining ISC and in promoting regeneration especially after damage by bleomycin, which may damage enterocytes and therefore disrupt BMP pathway more. The results overall support their claim.

The data are highly coherent, and support a genetic function of Numb, in collaborating with BMP signaling, to maintain the number and proliferative function of ISCs in adult midguts. The authors used appropriate and sophisticated genetic tools of double RNAi, mutant clonal analysis and dual marker stem cell tracing approaches to ensure the results are reproducible and consistent. The statistical analyses provide confidence that the phenotypic changes are reliable albeit weaker than many other mutants previously studied.

Weaknesses:

In the absence of Numb itself, the midgut has a weak reduction of ISC number (Fig. 3 and 5), as well as weak albeit not statistically significant reduction of ISC clone size/proliferation. I think the authors published similar experiments with BMP pathway mutants. The mad^1-2 allele used here as stated below may not be very representative of other BMP pathway mutants. Therefore, it could be beneficial to compare the number of ISC number and clone sizes between other BMP experiments to provide the readers with a clearer picture of how these two pathways individually contribute (stronger/weaker effects) to the ISC number and gut homeostasis.

Thanks for the comment. We have tested other components of BMP pathway in our previously study (Tian et al., 2014). More complete loss of BMP signaling (for example, Put clones, Put RNAi, Tkv/Sax double mutant clones or double RNAi) resulted in ISC loss regardless the status of numb, suggesting a more predominant role of BMP signaling in ISC self-renewal compared with Numb. We speculate that the weak stem cell loss phenotype associated with numb mutant clones in otherwise wild type background could be due to fluctuation of BMP signaling in homeostatic guts.

The main weakness of this manuscript is the analysis of the BMP pathway components, especially the mad^1-2 allele. The mad RNAi and mad^1-2 alleles (P insertion) are supposed to be weak alleles and that might be suitable for genetic enhancement assays here together with numb RNAi. However, the mad^1-2 allele, and sometimes the mad RNAi, showed weakly increased ISC clone size. This is kind of counter-intuitive that they should have a similar ISC loss and ISC clone size reduction.

We used mad^1-2 and mad RNAi here to test the genetic interaction with numb because our previous studies showed that partial loss of BMP signaling under these conditions did not cause stem cell loss, therefore, may provide a sensitized background to determine the role of Numb in ISC self-renewal. The increased proliferation of ISC/ clone size associated with mad^1-2 and mad RNAi is due to the fact that reduction of BMP signaling in either EC or EB non-autonomously induces stem cell proliferation. However, in mad numb double mutant clones, there was a reduction in clone size due to loss of ISC in many clones.

A much stronger phenotype was observed when numb mutants were subject to treatment of tissue damaging agents Bleomycin, which causes damage in different ways than DSS. Bleomycin as previously shown to be causing mainly enterocyte damage, and therefore disrupt BMP signaling from ECs more likely. Therefore, this treatment together with loss of numb led to a highly significant reduction of ISC in clones and reduction of clone size/proliferation. One improvement is that it is not clear whether the authors discussed the nature of the two numb mutant alleles used in this study and the comparison to the strength of the RNAi allele. Because the phenotypes are weak and more variable, the use of specific reagents is important.

We have included information about the two numb alleles in the “Materials and Methods”. numb¹⁵ is a null allele, and the nature of numb⁴ has not been elucidated. According to Domingos, P.M. et al., numb¹⁵ induced a more severe phenotype than numb⁴ did. Consistently, we also found that more numb¹⁵ mutant clones were void of stem cell than numb⁴ mutant clones.

Furthermore, the use of possible activating alleles of either or both pathways to test genetic enhancement or synergistic activation will provide strong support for the claims.

Activation of BMP (esgts>Tkv^CA) alone induced stem cell tumor (Tian et al., 2014) whereas overexpression of Numb did not induce increase stem cell number although overexpression of Numb in wing discs produced phenotypes indictive of inhibition of N (our unpublished observation), making it difficult to test the synergistic effect of activating both BMP and Numb.

Reviewer #1 (Recommendations for the authors):

- Cartoon of RGT in Fig 4 needs to be improved. We need to know what chromosome harbors the esgts. It is not sufficient to simply put the location of the ubi-GFP and ubi-RFP (on 19A) and not show the location of other components of the RGT system.

Thank you for the suggestion. We have revised the cartoon in Fig. 4 to include all three pairs of chromosomes and indicate where the esgts driver and UAS-RNAi are located. In addition, we have included the genotypes for all the genetic experiments in the Method section.

- Quantification of the results in Fig. 1

Quantification of Fig.1 has been added. 

- The authors need to explain the premise more carefully (see above) and explain whether or not they tested put, numb double knockdowns.

We have explained why not testing put numb double RNAi (see above).

Reviewer #2 (Recommendations for the authors):

The number of times the experiments have been performed would be useful to include.

This information has been added in the figure legends.

Author response:

We thank the reviewers for their thoughtful comments on our submitted manuscript.

The major point from all three reviewers was that the sensory inputs may be more complex than simply ASH and AWC, since mutations in osm-9 and tax-4 will affect many more sensory neurons. We fully agree. The differential effects of osm-9 and ta_x-_4 allowed us to recognize that there were two distinct afferent pathways operating simultaneously, mediating repulsion and attraction separately. However, it remains to be determined which sensory neurons are contributing to each pathway. We have planned a full analysis of the sensory inputs, not limited to just ASH and AWC, using neuron-specific rescue and neuron-specific chemogenetic inactivation (using HisCl1). While this analysis falls outside the scope of the present study, we will perform the inactivations of ASH and AWC and include the data for the revised version of this study. We expect to demonstrate whether ASH and AWC inputs are sufficient or whether other sensory neurons make significant contributions. Additionally, we will include chemotaxis dose-response data for osm-9 mutants as part of this analysis and make the minor corrections in data presentation requested.

Author response:

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

We are disappointed that the reviewers do not acknowledge that our data constitute a major step forward for the field. We will prepare a revised version that takes care of the remaining small issues concerning the technical descriptions and a detailed response to the current round of comments. We will also add a summary of the major new findings of our study.

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

We appreciate the time of the reviewers and their detailed comments, which have helped to improve the manuscript.

Our study presents the largest systematic dataset so far on the evolution of sex-biased gene expression in animals. It is also the first that explores the patterns of individual variation in sex-biased gene expression and the SBI is an entirely new procedure to directly visulize these variance patterns in an intuitive way.

Also, we should like to point out that our study contradicts recent conclusions that had suggested that a substantial set of sex-biased genes has conserved functions between humans and mice and that mice can therefore be informative for gender-specific medicine studies. Our data suggest that only a very small set of genes are conserved in their sex-biased expression between mice and humans in more than one organ.

In the revised version we have made the following major updates:

- added a rate comparison of gene regulation turnover between sex-biased and non-sex-biased genes

- added additional statistics to the variance comparisons and selection tests

- added a regulatory module analysis that shows that much of the gene turnover happens within modules

- added a mosaic pattern analysis that shows the individual complexity of sex-biased patterns

- extended introduction and discussion

Reviewer #1 (Public Review):<br /> The authors describe a comprehensive analysis of sex-biased expression across multiple tissues and species of mouse. Their results are broadly consistent with previous work, and their methods are robust, as the large volume of work in this area has converged toward a standardized approach.

I have a few quibbles with the findings, and the main novelty here is the rapid evolution of sex-biased expression over shorter evolutionary intervals than previously documented, although this is not statistically supported. The other main findings, detailed below, are somewhat overstated.

(1) In the introduction, the authors conflate gametic sex, which is indeed largely binary (with small sperm, large eggs, no intermediate gametic form, and no overlap in size) with somatic sexual dimorphism, which can be bimodal (though sometimes is even more complicated), with a large variance in either sex and generally with a great deal of overlap between males and females. A good appraisal of this distinction is at . This distinction in gene expression has been recognized for at least 20 years, with observations that sex-biased expression in the soma is far less than in the gonad.

For example, the authors frame their work with the following statement:

"The different organs show a large individual variation in sex-biased gene expression, making it impossible to classify individuals in simple binary terms. Hence, the seemingly strong conservation of binary sex-states does not find an equivalent underpinning when one looks at the gene-expression makeup of the sexes"

The authors use this conflation to set up a straw man argument, perhaps in part due to recent political discussions on this topic. They seem to be implying one of two things. a) That previous studies of sex-biased expression of the soma claim a binary classification. I know of no such claim, and many have clearly shown quite the opposite, particularly studies of intra-sexual variation, which are common - see https://doi.org/10.1093/molbev/msx293, https://doi.org/10.1371/journal.pgen.1003697, https://doi.org/10.1111/mec.14408, https://doi.org/10.1111/mec.13919, https://doi.org/10.1111/j.1558-5646.2010.01106.x for just a few examples. Or b) They are the first to observe this non-binary pattern for the soma, but again, many have observed this. For example, many have noted that reproductive or gonad transcriptome data cluster first by sex, but somatic tissue clusters first by species or tissue, then by sex (https://doi.org/10.1073/pnas.1501339112, https://doi.org/10.7554/eLife.67485)

Figure 4 illustrates the conceptual difference between bimodal and binary sexual conceptions. This figure makes it clear that males and females have different means, but in all cases the distributions are bimodal.

I would suggest that the authors heavily revise the paper with this more nuanced understanding of the literature and sex differences in their paper, and place their findings in the context of previous work.

We are sorry that our introduction seems to have been too short to make our points sufficiently clear. Of course, overlapping somatic variation has been shown for morphological characters, but we were aiming to assess this at the sex-biased transcriptome level. Previous studies looking at sex-biased genes were usually limited by the techniques that were available at their times, resulting in a focus on gonads in most studies and almost all have too few individuals included to study within-group variation. We detail this below for the papers that are mentioned by the referee. In view of this, we cite them now as examples for the prevalent focus on gonadal comparisons in most studies. Only Scharmann et al. 2021 on plant leaf dimorphism is indeed relevant for our study with respect to its general findings and we make now extensive reference to it. In addition, we have generally modified the introduction and substantially extended the discussion to make our points clear.

Snell-Rood 2010: the paper focuses on sex-specific morphological structures in beetles. It samples six somatic tissues for four individuals each of each class. Analysis is done via microarray hybridizations. While categorial differences were traced, variability between individuals was not discussed. By today´s standards, microarrays have anyway too much technical variability to even consider such a discussion.

Pointer et al. 2013: this paper studies three sexual phenotypes in a bird species, females, dominant males and subordinate males. Tissues include telencephalon, spleen and left gonad. The focus of the analysis is on the gonads, since only few sex-biased genes were found in spleen and brain (according to suppl. Table S1, 0 for the spleen and 2 for the brain). No inferences could be made on somatic variation.

Harrison 2015: this paper focuses on gonads plus spleen in six bird species with between 2-6 individuals for each sex collected. In the spleen, only one female biased gene and no male biased gene was detected. Hence, the data do not allow to infer patterns of somatic variation.

Dean et al. 2016: this paper compares four categories of fish caught around nests, with four to seven individuals per category. Only gonads were analyzed, hence no inferences could be made about somatic variability between individuals.

Cardoso et al. 2017: this paper test categories of fish with alternative reproductive tactics based on brain transcriptomes. While it uses 9-10 individuals per category, it uses pools for sequencing with two pools per category. This does not allow to make any inference on individual variation.

Todd et al 2017: this paper focuses on three categories of a fish species, females and dominant and sneaker males. It uses brain and gonads as samples with five individuals each for each category. For the brain, more different genes were found between the two types of males, rather than between females and males (3 and 9 respectively). The paper focuses on individual gene descriptions and does not mention somatic variation.

Scharmann 2021: the paper focuses on 10 species of plants with sexually dimorphic leafs. 5-6 individuals were sampled per sex. The major finding is that sex-biased gene expression does not correlate with the degree of sexual dimorphism of the leafes. The study shows also a fast evolution of sex-biased expression and states that signatures of adaptive evolution are weak. But it does not discuss variance patterns within populations.

(2) The authors also claim that "sexual conflict is one of the major drivers of evolutionary divergence already at the early species divergence level." However, making the connection between sex-biased genes and sexual conflict remains fraught. Although it is tempting to use sex-biased gene expression (or any form of phenotypic dimorphism) as an indicator of sexual conflict, resolved or not, as many have pointed out, one needs measures of sex-specific selection, ideally fitness, to make this case (https://doi.org/10.1086/595841, 10.1101/cshperspect.a017632). In many cases, sexual dimorphism can arise in one sex only without conflict (e.g. 10.1098/rspb.2010.2220). As such, sex-biased genes alone are not sufficient to discriminate between ongoing and resolved conflict.

We imply sexual conflict as a driver of genomic divergence patterns in a similar way as it has been done by many authors before (e.g. Mank 2017a, Price et al. 2023, Tosto et al. 2023). While we fully appreciate the point of the referee, we do not really see where we deviate from the standard wording that is used in the context of genomic data. In such data, it is of course usually assumed that they represent solved conflicts (Figure 1D in Cox and Calsbeek) where selection differentials would not be measurable anyway. (Please note also that the phylogenetic approach used in Oliver and Monteiro 2010 becomes rather problematic in view of introgressive hybridization patterns in butterflies), We have extended the discussion to address this.

(3) To make the case that sex-biased genes are under selection, the authors report alpha values in Figure 3B. Alpha value comparisons like this over large numbers of genes often have high variance. Are any of the values for male- female- and un-biased genes significantly different from one another? This is needed to make the claim of positive selection.

Sorry, we had accidentally not included the statistics in the final version of the figure. We have added this now in the supplementary table but have also generally changed the statistical approach and the design of the figure.

Reviewer #2 (Public Review):

The manuscript by Xie and colleagues presents transcriptomic experiments that measure gene expression in eight different tissues taken from adult female and male mice from four species. These data are used to make inferences regarding the evolution of sex-biased gene expression across these taxa. The experimental methods and data analysis are appropriate; however, most of the conclusions drawn in the manuscript have either been previously reported in the literature or are not fully supported by the data.

We are not aware of any study that has analyzed somatic sex-biased expression in such a large and taxonomically well resolved closely related taxa of animals. Only the study by Scharman et al. 2021 on plant leaves comes close to it, but even this did not specifically analyze the intragroup variation aspects. Of course, some of our results confirm previous conclusions, but we should still like to point out that they go far beyond them.

There are two ways the manuscript could be modified to better strengthen the conclusions.

First, some of the observed differences in gene expression have very little to no effect on other phenotypes, and are not relevant to medicine or fitness. Selectively neutral gene expression differences have been inferred in previous studies, and consistent with that work, sex-biased and between-species expression differences in this study may also be enriched for selectively neutral expression differences. This idea is supported by the analysis of expression variance, which indicates that genes that show sex-biased expression also tend to show more inter-individual variation. This perspective is also supported by the MK analysis of molecular evolution, which suggests that positive selection is more prevalent among genes that are sex-biased in both mus and dom, and genes that switch sex-biased expression are under less selection at the level of both protein-coding sequence and gene expression.

We have now revisited these points by additional statistical analysis of the variance patterns and an extended discussion under the heading "Neutral or adaptive?". 

As an aside, I was confused by (line 176): "implying that the enhanced positive selection pressure is triggered by their status of being sex-biased in either taxon." - don't the MK values suggest an excess of positive selection on genes that are sex-biased in both taxa?

There are different sets of genes that are sex-biased in these two taxa - hence this observation is actually a strong argument for selection on these genes. We have changed the correspondiung text to make this clearer.

Without an estimate of the proportion of differentially expressed genes that might be relevant for broader physiological or organismal phenotypes, it is difficult to assess the accuracy and relevance of the manuscript's conclusions. One (crude) approach would be to analyze subsets of genes stratified by the magnitude of expression differences; while there is a weak relationship between expression differences and fitness effects, on average large gene expression differences are more likely to affect additional phenotypes than small expression differences.

We agree that it remains a challenge to show functional effects for the sex-biased genes. The argument that they should have a function is laid out above (and stated in many reviews on the topic). To use the expression level as a proxy of function does not seem justified, given the current literature. For example, genes that are highly conected in modules are not necessrily highly expressed (e.g. transcription factors). Also, genes may be highly expressed in a rare cell type of an organ and have an important funtion there, but this would not show up across the RNA of the whole organ. The most direct functional relationship between sex-biased expression and phenotype comes from the human data in Naqvi et al. 2019 - which we had cited.

Another perspective would be to compare the within-species variance to the between-species variance to identify genes with an excess of the latter relative to the former (similar logic to an MK test of amino acid substitutions).

Such an analysis was actually our intial motivation for this study. However, the new (and surprising!) result is that the status of being sex-biased shows such a high turnover that not many genes are left per organ where one could even try to make such a test. However, we have extended the variance analysis with reciprocal gene sets (as we had done it for the MK test) and extended the discussion on the topic, including citation of our prior work on these questions.

Second, the analysis could be more informative if it distinguished between genes that are expressed across multiple tissues in both sexes that may show greater expression in one sex than the other, versus genes with specialized function expressed solely in (usually) reproductive tissues of one sex (e.g. ovary-specific genes). One approach to quantify this distinction would be metrics like those used defined by [Yanai I, et al. 2005. Genome-wide midrange transcription profiles reveal expression-level relationships in human tissue specification. Bioinformatics 21:650-659.] These approaches can be used to separate out groups of genes by the extent to which they are expressed in both sexes versus genes that are primarily expressed in sex-specific tissue such as testes or ovaries. This more fine-grained analysis would also potentially inform the section describing the evolution/conservation of sex-biased expression: I expect there must be genes with conserved expression specifically in ovaries or testes (these are ancient animal structures!) but these may have been excluded by the requirement that genes be sex-biased and expressed in at least two organs.

Given that our study focuses on somatic sex-biased genes, we refrain from a comparative analysis of genes that are only expressed in the sex-organs in this paper. With respect to sharing of sex-biased gene expresssion between the somatic tissues, we show in Figure 8 that there are only very few of them (8 female-biased and 3 male-biased). A separate statistical treatment is not possible for this small set of genes.

There are at least three examples of statements in the discussion that at the moment misinterpret the experimental results.

The discussion frames the results in the context of sexual selection and sexually antagonistic selection, but these concepts are not synonymous. Sexual selection can shape phenotypes that are specific to one sex, causing no antagonism; and fitness differences between males and females resulting from sexually antagonistic variation in somatic phenotypes may not be acted on by sexual selection. Furthermore, the conditions promoting and consequence of both kinds of selection can be different, so they should be treated separately for the purposes of this discussion.

We cannot make such a distinction for gene expression patterns - and we are not aware that this was done before in the literature (except gene expression was directly linked to a morphological structure). We have updated this discussion accordingly.

The discussion claims that "Our data show that sex-biased gene expression evolves extremely fast" but a comparison or expectation for the rate of evolution is not provided. Many other studies have used comparative transcriptomics to estimate rates of gene expression evolution between species, including mice; are the results here substantially and significantly different from those previous studies? Furthermore, the experimental design does not distinguish between those gene expression phenotypes that are fixed between species as compared to those that are polymorphic within one or more species which prevents straightforward interpretation of differences in gene expression as interspecific differences.

Our statement was in relation to the comparison between somatic and gondadal gene turnover, as well as the comparison to humans. We have now included an additional analysis for a direct comparison with non-sex-biased genes in the same populations (Figure 2B). Note that gene expression variances cannot get fixed anyway, they can only become different in average and magnitude.

The conclusion that "Our results show that most of the genetic underpinnings of sex differences show no long-term evolutionary stability, which is in strong contrast to the perceived evolutionary stability of two sexes" - seems beyond the scope of this study. This manuscript does not address the genetic underpinnings of sex differences (this would involve eQTL or the like), rather it looks at sex differences in gene expression phenotypes.

This comes back to the points discussed above about the validity to infer function from sex-biased expression. We have updated the text to clarify this.

Simply addressing the question of phenotypic evolutionary stability would be more informative if genes expressed specifically in reproductive tissues were separated from somatic sex-biased genes to determine if they show similar patterns of expression evolution.

Our study is generally focused on somatic gene expression. The comparison with reproductive tissues serves merely as a reference. Since they are of course very different tissues, they should not be compared with each other in the same way. We have now specifically addressed this point in the discussion.

Reviewer #3 (Public Review):

This manuscript reports some interesting and important patterns. The results on sex-bias in different tissues and across four taxa would benefit from alternative (or additional) presentation styles. In my view, the most important results are with respect to alpha (fraction of beneficial amino acid changes) in relation to sex-bias (though the authors have made this as a somewhat minor point in this version).

The part that the authors emphasize I don't find very interesting (i.e., the sexes have overlapping expression profiles in many nongonadal tissues), nor do I believe they have the appropriate data necessary to convincingly demonstrate this (which would require multiple measures from the same individual).

This is the first study that reports such overlaps and we show that this is not always the case (e.g. liver and kidney data in mice). We are not aware of any preditions of how such patterns would look like and how they would evolve - why should such a new finding not be interesting? Concerning the appropriateness of the data we do not agree with the point the referee makes - see response below.

This study reports several interesting patterns with respect to sex differences in gene expression across organs of four mice taxa. An alternative presentation of the data would yield a clearer and more convincing case that the patterns the authors claim are legitimate.

I recommend that the authors clarify what qualifies as "sex-bias".

This is defined by the statistical criteria that we have applied, following the general standard of papers on this topic.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) "However, already Darwin has pointed out that the phenotypes of the sexes should evolve fast". I think the authors mean that Darwin was quick to point out that sex-specific phenotypes evolve quickly".

We have modified this text part.

(2) Non-gonadal is more often referred to as somatic. I would encourage the authors to use this more common term for accessibility.

We have adopted this term

(3) Figure 5 is interesting, however, it is difficult to know whether the decreased bimodality in humans compared to mice is biological or technical due to the differences in the underlying data. For example, the mouse samples tightly controlled age and environmental conditions within each species. It is not possible to do that with human samples, and there are very good reasons to think that these factors will affect variance in both sexes.

Yes, this is certainly true and we know this also from other comparative data between mice and humans. Still, this is human reality vs mouse artificialness. We pick this now up in the discussion.

(4) Line 273. The large numbers of cells needed for single-cell analysis require that most studies pool multiple samples, however these pools are helpful in themselves. This approach was used by https://doi.org/10.1093/evlett/qrad013 to quantify the degree of sex-bias within cell types across multiple tissues and to compare how bulk and single-cell sex-bias measures compare. Sex-bias in some somatic cell types was very high, even when bulk sex-bias in those tissues was not. This suggests that the bulk data the authors use in this study may in fact obscure the pattern of sex-bias.

Yes, we agree, and this is exactly how we did the analysis and interpretation, based on the cited paper.

(5)- Line 379 "Total RNAs were" should be "Total RNA was"

Corrected

References cited in this review and which should be included in the manuscript :

Sam L Sharpe, Andrew P Anderson, Idelle Cooper, Timothy Y James, Alexandra E Kralick, Hans Lindahl, Sara E Lipshutz, J F McLaughlin, Banu Subramaniam, Alicia Roth Weigel, A Kelsey Lewis, Sex and Biology: Broader Impacts Beyond the Binary, Integrative, and Comparative Biology, Volume 63, Issue 4, October 2023, Pages 960-967.

Included

Masculinization of Gene Expression Is Associated with Exaggeration of Male Sexual Dimorphism Pointer MA, Harrison PW, Wright AE, Mank JE (2013) Masculinization of Gene Expression Is Associated with Exaggeration of Male Sexual Dimorphism. PLOS Genetics 9(8): e1003697.

Included

Erica V Todd, Hui Liu, Melissa S Lamm, Jodi T Thomas, Kim Rutherford, Kelly C Thompson, John R Godwin, Neil J Gemmell, Female Mimicry by Sneaker Males Has a Transcriptomic Signature in Both the Brain and the Gonad in a Sex-Changing Fish, Molecular Biology and Evolution, Volume 35, Issue 1, January 2018, Pages 225-241.

Included

Cardoso SD, Gonçalves D, Goesmann A, Canário AVM, Oliveira RF. Temporal variation in brain transcriptome is associated with the expression of female mimicry as a sequential male alternative reproductive tactic in fish. Mol Ecol. 2018; 27: 789-803.

Included

Dean, R., Wright, A.E., Marsh-Rollo, S.E., Nugent, B.M., Alonzo, S.H. and Mank, J.E. (2017), Sperm competition shapes gene expression and sequence evolution in the ocellated wrasse. Mol Ecol, 26: 505-518.

Included

Emilie C. Snell‐Rood, Amy Cash, Mira V. Han, Teiya Kijimoto, Justen Andrews, Armin P. Moczek, DEVELOPMENTAL DECOUPLING OF ALTERNATIVE PHENOTYPES: INSIGHTS FROM THE TRANSCRIPTOMES OF HORN‐POLYPHENIC BEETLES, Evolution, Volume 65, Issue 1, 1 January 2011.

Not included, since its technical approach is not really comparable

Harrison PW, Wright AE, Zimmer F, Dean R, Montgomery SH, Pointer MA, Mank JE (2015) Sexual selection drives evolution and rapid turnover of male gene expression. Proceedings of the National Academy of Sciences, USA 112: 4393-4398.

Included

Mathias Scharmann, Anthony G Rebelo, John R Pannell (2021) High rates of evolution preceded shifts to sex-biased gene expression in Leucadendron, the most sexually dimorphic angiosperms eLife 10:e67485.

Included

Sexually Antagonistic Selection, Sexual Dimorphism, and the Resolution of Intralocus Sexual Conflict. Robert M. Cox and Ryan Calsbeek , The American Naturalist 2009 173:2, 176-187.

Included

Ingleby FC, Flis I, Morrow EH. Sex-biased gene expression and sexual conflict throughout development. Cold Spring Harb Perspect Biol. 2014 Nov 6;7(1):a017632.

Included

Oliver JC, Monteiro A 2011. On the origins of sexual dimorphism in butterflies. Proc Biol Sci 278: 1981-1988.

Included

Iulia Darolti, Judith E Mank, Sex-biased gene expression at single-cell resolution: cause and consequence of sexual dimorphism, Evolution Letters, Volume 7, Issue 3, June 2023, Pages 148-156.

Included

Reviewer #2 (Recommendations For The Authors):

I am concerned the smoothed density plots in Figure 4 may be providing a misleading sense of the distributions since each distribution is inferred from only 9 values. A boxplot might better represent the data to the reader.

Boxplots with 9 values are much more difficult to interpret for a reader, this is the very reason why one tends to smoothen them. In this way, they also become similar to the standard plots that are used for showing morphological variation between the sexes. Note that the original data are availble for the individual values, if these are of special interest in some cases. In addition, our new “mosaic” analysis (Figure 6) provides another presentation for readers.

Line 235: "the overall numbers are lower" I assume this is the number of genes included in the analyses, but this should be explicitly stated.

Clarified in the text

The analysis of gene expression from different brain regions in control individuals from the Alzheimer's study (line 273) suffers from low power and it is not clear to me how much taking samples from different brain regions eliminates the issue of different cell types within a sample (the stated motivation for this analysis). While I support publishing negative results, this section does not feel like it adds much to the manuscript and could be cut in my opinion.

This is actually a study on single cell types, differentiating each of them. We are sorry that the text was apparently unclear about this. Given that there are studies that show the importance of looking at single cell data, we still think that is a suitable analysis. We have updated the text to make it clearer.

It might be useful to separate out X-linked genes from autosomal genes to see if they show consistent patterns with regard to sex-bias.

We have added this information in suppl. Table S2 and include some description in the text.

Reviewer #3 (Recommendations For The Authors):

Comments follow the order of the Results section:

(1) The latter half of this line in the Methods is too vague to be helpful: "We have explored a range of cutoffs and found that a sex-bias ratio of 1.25-fold difference of MEDIAN expression values combined with a Wilcoxon rank sum test and Benjamini-Hochberg FDR correction (using FDR <0.1 as cutoff) (Benjamini & Hochberg, 1995) yields the best compromise between sensitivity and specificity". What precisely is meant by "the best compromise between sensitivity and specificity"?

We explain now that this was based on pre-tests with comparing randomized with actual data. However, we agree that this is in the end a subjective decision, but there is no single standard used in the literature, especially when somatic organs are included. We consider our criteria as rather stringent.

(2) The 1.25 number for sex bias is, ultimately, an arbitrary cut-off. It is common in this literature to choose some arbitrary level and, in this sense, the authors are following common practice. The choice of 1.25 should be stated in the main text as it is a lower (but not reasonable) value than has been used in many other papers.

It is not only the cutoff, but also the Wilcoxon test and FDR correction that defines the threshold. See also comment above.

(3) In truth, dimorphism is continuous rather than discrete (i.e, greater or less than 1.25 fold different). Thus, where possible it would be useful to present results in a fashion that allows readers to see the continuous range of ratios rather than having to worry about whether the patterns are due to the rather arbitrary choices of how genes were binned into sex-bias categories.

It is necessary to work with cutoffs in such cases - and this is the usual practice for any such paper. But we provide now in Figure 1 Figure supplement 1 plots with the female/male ratio distributions.

a) Number of genes that are female- / male-biased. I would like to be able to see a version of Figure 1 showing the full distribution of TPM ratios rather than bar graphs of the numbers of (arbitrarily defined) female- and male-biased genes. This will be, of course, a larger figure (a full distribution rather than 2 bars for each species for each organ) and so could be relegated to Supplementary Material (assuming the message of that figure is the same as the current Figure 1).

This is a very unusual request, given that no other paper has done this either. It would indeed result in a non-managable figure size, or many separate figures that would be difficult to scrutinize. Note that there would be one plot of two (female and male) TPM distributions for each sex-biased gene in each organ and each taxon, leading to hundreds of thousands of plots. We think that by providing the general distributions as plots (see above), and the original data as supplements is sufficient.

b) Turnover of genes with sex bias. This important issue is addressed in Figure 2. First, it is not precisely clear what "percentages of sums of shared genes for any pairwise comparison" in Figure 2 legend means and no further detail is given in the Methods; this must be made clearer or the info in Figure 2 is meaningless. Regardless, this approach again relies heavily on the arbitrary criterion of defining sex-bias. Thus, I would like to see correlation plots of the log(TPM ratio) between taxa as done in the classic multispecies fly paper of Zhang et al. 2007. In Figure 2 it is quite clear that male-biased genes evolve with respect to sex bias more rapidly than female-biased genes.

We have provided a better explanation of this analysis. Note that the Zhang et al. 2007 paper was not focussing on somatic expression and covers a much broader evolutionary spectrum. Hence, the results are not comparable. Also, we doubt that it would be so helpful to generate a huge figure with all these plots.

(4) Is there a simpler explanation for the results in the "Variance patterns" section? The total variance for any variable can be decomposed into the variance within and among "groups". If we use "sex" as the group, then there are genes - labelled sex-biased genes - that were identified as such, in essence, because they have high among-group variance. Given that we then know a priori at the start of this section of sex-biased genes have high among-group variance, is it at all surprising that they have higher total variance than the unbiased genes (which we know a priori have low among-group variance)? Perhaps I misunderstood the point of this section. Maybe it would be more meaningful to examine the WITHIN-SEX variance (averaged across the two sexes) instead.

We did calculate IQR/median (“normalized variance”) with the nine mice for each gene and each sex in each organ, hence sex is not a variance factor in this calculation. The algorithm steps are outlined in suppl. Table S17. We have now also added a variance calculation for reciprocal gene sets and added an extended discussion of these results.

(5) Analysis of alpha for sex-biased genes. This was the most interesting part of this manuscript to me.

(a) More information about what SNVs were used is required.

i. Were only sites where SPR was fixed used? (If not, how was polarization done?)

ii. Were sites only considered diverged if they were fixed for different bases in DOM and MUS? (If not, what was the criteria?)

iii. Using, say, DOM as the focal species, a site must be polymorphic in DOM. But did its status (polymorphic/fixed) in MUS matter?

We have added a more detailed description on this in the Methods section. For the direct answers of the three questions: (i) yes; (ii) yes; (iii) no, considering that DOM and MUS are two subspecies of Mus musculus separating recently, a variant might occur before separating and there might be gene flow between them.

(b) A particularly interesting part of the analysis is the investigation of alpha for genes that are NOT sex-biased in one taxa but are sex-biased in the other. At the moment (as I understand it), alpha is only calculated for these genes in the taxa where they are NOT sex-biased (and this alpha value can be compared to the alpha of sex-biased genes and of unbiased genes in that taxa). I would like to see both sets of genes (set 1: those sex-biased in MUS and not in DOM; set 2: those sex-biased DOM and not in MUS) analyzed in each of the 2 species, with results presented in a 2x2 table.

By definition of these categories, these genes are sex-biased in the respective other taxon, hence the values are already in the table. They are named as “reciprocal”.

(c) No confidence intervals are given for the alpha values, despite the legend of Figure 3 referring to them.

These were accidentally omitted - we now included the full table in suppl. Table S6; Figure 3 was modified to show violin plots of the bootstrap distributions

The author's creation and use of a "sex-bias index" (SBI). My greatest skepticism of this manuscript is with respect to the value of their manufactured index, SBI. Of course, it is possible to create such an index but does this literature really need this index or does this just add to the "clutter" in the literature for this field? Is it helping to illuminate important patterns? This index is presumably some attempt to quantify how "male-like" or "female-like" overall expression is for a given individual (for a given organ). It is calculated as SBI = (MEDIAN of all female-biased tpm) - (MEDIAN of all male-biased tpm).

(6) A main result that comes from this is that the sexes tend to overlap for these values for most nongonad tissues but are clearly distinct for gonadal tissues. I do not think this result would come as a surprise to almost anyone and I'm far from convinced that this metric is a good way to quantify that point. Let's consider testes vs. ovaries. Compared to non-gonadal tissues, I am reasonably certain that not only are there many more genes that are classified as "sex-biased" in gonads but also the magnitude of sex-bias among these genes is typically much greater than it is for the so-called sex-biased genes in nongonadal tissue (density plots requested in #3a would make this clear). In other words, males and females are, on average, very different with respect to expression in gonads so even allowing for variation within each sex will still result in a clear separation of all individuals of the two sexes. In contrast, males and females are, on average, much less different in, say, heart so when we consider the variation within each sex, there is overlap. One could imagine a variety of different metrics which could be used to make this point. The merits of "SBI" are unclear. It is a novel metric and its properties are poorly understood. (A simple alternative would be looking at individual scores along the axis separating mean/median males and females; almost certainly, for gonads, this would be very similar to PC scores for PC1.)

As throughout the text, we use gonadal comparisons only as general reference, not as the main result. The main result that we are stressing is the fast turnover of these patterns, including from binary to overlapping for kidney and liver in mouse. We consider this as a new finding. If it comes "not to a surprise to anyone", isn´t it great that one does not have to guess anymore but has finally real data on this?

We have now also added a mosaic analysis to show that the SBI can be used as summary measure in different presentations.

The use of a single PC axis is no good alternative, since it throws away the information from the other axis.

We have now included an explicit discussion on the usefulness of the SBI.

(7) For simplicity, let's assume all males are identical and all females are identical. Let's imagine that heart and kidney have the exact same set of sex-biased genes. There are 20 female-biased genes; they all happen to be identical in expression level (within tissue) and look like this:

Female TPM Male TPM TPM ratio (F:M)

Heart 4 2 2

Kidney 40 20 2

And there are 20 male-biased genes that look like this:

Female TPM Male TPM TPM ratio (F:M)

Heart 1 3 1/3

Kidney 10 30 1/3

Most people would describe these two tissues as equally sex-biased.

However, the SBIs would be:

Female SBI Male SBI Sex difference (F - M)

Heart 4-1 = 3 2 - 3 = -1 4

Kidney 40-10 =30 20-30 = -10 40

Is it a desirable property that by this metric these two tissues have wildly different SBI values for each sex as well as for the difference between sexes? (At the very least, shouldn't you make readers aware of these strange properties of SBI so they can decide how much value they put into them?)

Actually, in this example the simple ratio between the expression levels has a strange property, since it does not reflect a much higher expression of the relevant genes in the kidney. The SBI is actually more suitable for making such cases clear. Of course, this is under the assumption that expression level has a meaning for the phenotype, but this is the general assumption for all RNA-Seq experiment comparisons.

(8) With respect to Figure 4, why do females often have mean SBI values close to zero or even negative (e.g., kidney, mammary glands)? Is this simply because the female-biased genes tend to have lower TPM than the male-biased genes? It seems that the value zero for this metric is really not very biologically meaningful because this metric is a difference of two things that are not necessarily expected to be equal.

This is the extra information about the expression levels that is gained via the SBI values (see comment above). However, we noticed that people can get confused about this. We have now added a re-scaling step to focus completely on the variance information in these plots.

(9) Interpreting variances. A substantial fraction of the latter half of the manuscript focuses on interpreting variances among individual samples. This is problematic because there is no replication within individuals (i.e.., "repeatability"), thus it is impossible to infer the extent of observed variance among individuals of a given group (e.g., among females) is due to true biological differences among individuals or is simply due to noise (i.e., "measurement error" in the broad sense). Is the larger variance for mammary glands than liver or gonads just due to measurement error? What is the evidence?

This point was of course a major issue during the times where microarrays were used for transcriptome studies. However, the first systematic RNA-Seq studies showed already that the technical replicability is so high, that technical replicates are not required. In fact, practically all RNA-Seq studies are done without technical replicates for this reason.

(10) Because I have little confidence in the SBI metric (#7-8) and in interpreting within sex variances (#9), I found little value in the human results and how SBI distributions (and degree of overlap between sexes) compare between humans and mice.

We disagree - the current published status is that there are thousands of sex-biased gene in humans and this has implications for gender-specific medicine (Oliva et al. 2020). Our results show a much more nuanced picture in this respect.

(11) I found even less value in the single-cell data. It too suffers from the issues above. Further, as the authors more or less state, the data are too limited to say much of value here. It is impossible to tell to what extent the results are simply due to data limitations.

We have pointed out that it is still valuable to have them. They are good enough to exclude the possibility that only a small set of cells drives the overall pattern across an organ. We have further clarified this in the text.

(12) The code for data analysis should be posted on GitHub or some other repository.

The code for the sex-biased gene detection and analysis has been posted on GitHub (see Code availability in the manuscript).

Author response:

The following is the authors’ response to the original reviews

Public reviews:

Reviewer #1:

Weaknesses:

As this paper only uses anatomical analyses, no functional interpretations of cell function are tested.

The aim of this paper was to describe the ultrastructural organization of compound eyes in the extremely small wasp Megaphragma viggianii. The authors successfully achieved this aim and provided an incredibly detailed description of all cell types with respect to their location, volume, and dimensions. As this is the first of its kind, the results cannot easily be compared with previous work. The findings are likely to be an important reference for future work that uses similar techniques to reconstruct the eyes of other insect species. The FIB-SEM method used is being used increasingly often in structural studies of insect sensory organs and brains and this work demonstrates the utility of this method.

We thank you for your high assessment of our work. Unfortunately, it is hard to test our functional interpretations and check them with electrophysiological methods due to the extremely small size of the animal. Studies on three-dimensional ultrastructural datasets obtained using vEM have just started to appear, and we hope that a lot of data will become available for comparison in the nearest future.

Reviewer #2:

Thank you for your work and for your high assessment of our manuscript.

Reviewer #3:

Weaknesses:

The claim that the large dorsal part of the eye is the dorsal rim area (DRA), supported by anatomical data on rhabdomere geometry and connectomics in authors' earlier work, would eventually greatly benefit from additional evidence, obtained by immunocytochemical staining, that could also reveal a putative substrate for colour vision. The cell nuclei that are located in the optical path in the DRA crystalline cone have only a putative optical function, which may be either similar to pore canals in hymenopteran DRA cornea (scattering) or to photoreceptor nuclei in camera-type eyes (focussing), both explanations being mutually exclusive.

We thank the Reviewer for high assessment of our study and for detailed analysis of our manuscript. Your comments and recommendations are very valued and helped us to improve the text. We understand that immunocytochemical methods could improve our findings and supply additional evidence, but there is no technical possibility for this in present. Megaphragma is a very complicated model organism for such methods. We are currently working on the optimization of the protocol for staining, which is needed because of the high level of autoluminescence and because of insufficient penetration of dyes into the samples.

Recommendations for the authors:

Reviewer #1:

I do not have any major concerns about the content of the paper.

There are some minor spelling and grammatical errors throughout the text but these can be identified most readily using a spelling/grammar check.

We have revised the text, checked the spelling, and fixed the grammatical errors throughout the text.

I suggest consistency when referring to the capitalization of the term 'non-DRA' as it is sometimes 'Non-DRA' in the text.

We have fixed the term “non-DRA” throughout the text. Thank you.

Also, check carefully the spelling of headings in the tables as there are a few mistakes in Table 1 and 5 in particular.

The grammar errors have been fixed.

Figure 7 legend: an explanation of the abbreviation RPC should be added.

We have done so.

Reviewer #2:

(1) The paper presents the data in great detail, however, since this is the first time the technique has been applied to get whole insect eyes, even if on a small insect, it would be worth outlining in the methods section what innovations in the staining/ scanning or sample preparation allowed these improvements and a roadmap for extending this method to larger insects if possible.

The whole method, including sample preparation, staining, and scanning, was described in our previous paper (Polilov et al., 2021), where it was presented in every detail. Due to the complicated methodology we suppose that it is not necessary to include all the stages of the technique in the present paper, and thus described it more briefly.

(2) The optical modelling needs a statement in the discussion providing a disclaimer on parameters like sensitivity, anatomical measurements can provide limits and some measure, but the inherent optics are also key and it is worth qualifying these as only estimates and measurements that give a sense of the variation in morphology, only coupled with optical and potentially neural measurements could one confirm the true sensitivity and acceptance angle.

In the absence of experimental data or precise computational models of Megaphragma vision, we try to discuss rather carefully the functions of structures based on their morphology, ultrastructure, first-order visual connectome, and analogies with other species. This is reflected in the methods and those sections of our paper that contain functional interpretations.

Reviewer #3

(1) The finding that the CNS neurons are enucleated, while the compound eye contains cell nuclei, deserves another word. I would confidentially say that the optical demands of a miniaturized compound eye (the minimal size of the optics due to diffraction, the rhabdomere size, and the minimal thickness of optically insulating granules) are such that further cellular miniaturization is not possible, and the minimal sizes even render the cells that build the eye sufficiently large to accommodate cell nuclei. This is in my opinion a parsimonious explanation, yet speculative and I leave it up to you to embrace it or not.

We agree with the Reviewer and understand the limiting factors and the optical demands of a miniaturized compound eye. According to our data, nuclei occupy a considerable volume in the eye (in the cells of compound eye there are more nuclei than in the whole brain), and on average the cell volume is larger than in Trichogramma, which is minute, but larger than Megaphragma. But as the Reviewer rightly assumed, it is speculative; therefore, we would like to avoid it.

(2) Our current understanding of DRA optics and function is limited and I claim that your interpretation of the cell nuclei in the DRA dioptrical apparatuses is inappropriate. Please consider a few articles on hymenopteran DRA, starting with the one below and the citing literature:

Meyer, E.P., Labhart, T. Pore canals in the cornea of a functionally specialized area of the honey bee's compound eye. Cell Tissue Res. 216, 491-501 (1981). https://doi.org/10.1007/BF00238646

Honebyee DRA has a milky appearance under a stereomicroscope and can be discerned from the outside. This is due to pore canals in the cornea. I happen to be studying this exact structure and its function right now. I found that the result of those canals is not so much the extended receptor acceptance angles, but rather a minimized light gain. This is counterintuitive, but think of the following. The DRA photoreceptors must encode the limited range of polarization contrasts with a maximal working dynamic range (= voltage) of the photoreceptors, which results in a very steep stimulus-response curve.

Physiologically such a curve is due to very high transduction gain and a high cell input resistance. In most of the retina, small contrasts are transcoded by LMC neurons, but DRA receptors are long visual fibres and must do the job themselves. The skylight intensity (especially antisolar, where the polarized pattern is maximal) varies little during the day. Hence, the DRA receptors work almost at a fixed intensity range. In order to prevent receptor saturation and keep steep contrast coding, the corneal lenses in DRA have a built-in diffusor ring, which diminishes the light influx. Unfortunately, I have yet to publish this and I may be wrong, of course. But if I look into your data, I see consistently smaller corneal lenses and crystalline cones in the DRA, plus the cell nuclei obstructing the incident light. I think this is similar to the optics of honeybee DRA.

You do not support your claim that the nuclei additionally focus light by optical calculations, but cite literature on camera-type eyes, which is not OK.

In any case, I think it is fair to limit the discussion by saying that the nuclei may have an optical role. Further evidence from hymenopteran and vertebrate literature is controversial. “so that the nuclei act as extra collecting lenses, as was reported for rod cells of nocturnal vertebrates (Solovei et al., 2009; Błaszczak et al., 2014)” - please consider omitting this.

We thank the Reviewer for this piece of advice. And we have rewritten the text, to omit the comparison with vertebrates, but left the citation as an illustration of the fact that nuclei could perform the optical role.

“Since the nuclei in DRA and non-DRA ommatidia are arranged differently in cone cells, we suggest that the nuclei of the cone cells of DRA ommatidia in M. viggianii perform some optical role, facilitating the specialization of this group of ommatidia. The optical function for nuclei was described for rod cells of nocturnal vertebrates, where chromatin inside the cell nucleus has a direct effect on light propagation (Solovei et al., 2009; Błaszczak et al., 2014; Feodorova et al., 2020).”

(3) Please consider comparing the structure and function of ectopic receptors with the eyelet in Drosophila (i.e. https://doi.org/10.1523/JNEUROSCI.22-21-09255.2002 )

We thank the Reviewer for this advice and have included the comparison fragment into the text:

“The position of ePR, their morphology and synaptic targets look similar to the eyelet (extraretinal photoreceptor cluster) discovered in Drosophila (Helfrich-Förster et al., 2002). Eyelets are remnants of the larval photoreceptors, Bolwig’s organs in Drosophila (Hofbauer, Buchner, 1989). Unlike Drosophila, Trichogrammatidae are egg parasitoids and their central nervous system differentiation is shifted to the late larva and even early pupa (Makarova et al., 2022). According to the available data on the embryonic development of Trichogrammatidae, no photoreceptors cells were found during the larval stages (Ivanova-Kazas, 1954, 1961).”

According to this, the analogy question remains open.

(4) Minor remarks:

“but also to trace the pathways that connect the analyzer with the brain.” - I find the word analyzer a bit stretched here; sure, the DRA is polarization analyzer, but if the main retina was monochromatic, it would only be a detector, not an analyzer.

The sentence was changed according to the Reviewer’s advice.

Table I: thikness -> thickness, wigth -> width

We have fixed these misprints.

“The cross-section of Non-DRA ommatidia has a strongly spherical shape” - perhaps circular, not spherical. And not necessary to say “strongly”

The spelling was changed according to the Reviewer’s advice.

“which can be rarely visualized in the cell's projections not far from the basement membrane.” - I'd suggest saying “which are nearly absent in retinula axons”

The spelling was changed according to the Reviewer’s advice.

“The pigment granules of the retinula cells have an elongated nearly oval shape” - please consider replacing 'elongated nearly oval' with 'prolate' (try googling for “prolate” or “oblate spheroids”; the adjective describes precisely what you wanted to say)

We thank the Reviewer for this piece of advice but prefer to leave our original phrasing, because it is more readily understandable.

“The results of our morphological analysis of all ommatidia in Megaphragma are consistent with the light-polarization related features in Hymenoptera and other insects” - please add citations, see my comment on the DRA above.

We have added the citations according to the Reviewer’s advice.

“The group of short PRs (R1-R6)” - please consider renaming into “short visual fibre photoreceptors” (as opposed to “long visual fibre PRs”; hence SVFs and LVFs). This naming is quite common.

The naming was changed according to the Reviewer’s advice.

“The total rhabdom shortening in M. viggianii ommatidia probably favors polarization and absolute sensitivity,” - please see comments on DRA. Wide rhabdom means also a wider acceptance angle.

Shortening of DRA rhabdoms does not result in their widening compared to other rhabdoms, so it is difficult to say how this may be related to sensitivity. The comments on DRA given earlier have been taken into account.

“Ommatidia located across the diagonal area of the eye are more sensitive to light” - I don't understand what is diagonal area.

We have deleted the sentence.

“Estimated optical sensitivity of the eyes very close to those reported for diurnal hymenopterans with apposition eyes (Greiner et al., 2004; Gutiérrez et al., 2024) and possess around 0.19 {plus minus} 0.04 μm2 sr. M. viggianii have reasonably huge values of acceptance angle Δρ, and thus should result in a low spatial resolution” - please correct English here. “eyes IS very close”, “should result in a low”

The grammatical errors were fixed.

Table 6 legend: “SPC - secondary pigment cells.” -> “SPC – secondary pigment cells.”

Citation “(Makarova et al., 2025).” - probably 2015

The typos were fixed.

Methods, FIB-SEM: I can't understand the sentence “The volumetric data of lenses and cones, some linear measurements (lens thickness, cone length, cone width, curvature radius) and to visualize the complete 3D-model of eye we use (measure or reconstruct) the elements from another eye (left).”

The sentence is a continuation of the previous one. We have rewritten it as follows to clarify the meaning and move it to the 3D reconstruction section:

“The right eye, on which the reconstruction was performed, has several damaged regions from milling (see Appendix 1С), which hinder the complete reconstructions of lenses and cones on a few ommatidia. According to this, for the volumetric data on lenses and cones, some linear measurements (lens thickness, cone length, cone width, curvature radius), we use (measure or reconstruct) the corresponding elements from the other (left) eye.”

“The cells of single interfacet bristles were not reconstructed, because of damaging on right eye and worst quality of section on the left.” - please change to “The cells of the single interfacet bristle were not reconstructed, because of damage to the right eye and inferior quality of the sections of the left eye.”

The text has been changed as follows:

“The cells of single interfacet bristles were not reconstructed, because of the damage present in the right eye and because of the generally lower quality of this region on the left eye.”

“Morphometry. Each ommatidia was” -> “Morphometry. Each ommatidium was”

The grammatical error has been fixed.

Author response:

The following is the authors’ response to the original reviews

Reviewer #3 (Recommendations for the authors):

Major concerns:

P.6, lines 223-224: The sentence sounds like the authors produced all the OVGP1s by themselves in their laboratories, which is not completely true. The recombinant human and mouse OVGP1s were purchased from OriGene. It is suggested that the authors should state and explain clearly here which OVGP1 is produced by their laboratories and that recombinant human and mouse OVGP1s were obtained and purchased from Origene.

It is already clearly included in the M&M.

P6, lines 227-229: The authors stated that "Western blots of the three OVGP1recombinants indicated expected sizes based on those of the proteins: 75 kDa for human and murine OVGP1 and around 60 kDa for bovine OVGP1 (Fig. 4B and S6)." I pointed out in my last review report that the size of the recombinant human OVGP1shown by the authors in their manuscript is not in agreement with what has been published previously in literature regarding the molecular weight of native human OVGP1 as well as that of recombinant human OVGP1. The authors did not address the above concern adequately. In fact, recombinant human OVGP1 has been produced a few years ago (Reproduction (2016) 152:561-573) and it has been previously demonstrated that a single protein band of approximately 110-130 kDa was detected for both native human OVGP1 (see Microscopy Research and Technique (1995) 32:57-69) and recombinant human OVGP1 (Reproduction (2016) 152:561-573; Carbohydrate Research (2012) 358:47-55) using antibodies specific for human OVGP1. Molecular weight of the protein core or polypeptide of human OVGP1 is approximately 75 kDa, but the glycosylated form of native human OVGP1 and recombinant human OVGP1 is approximately 110-130 kDa. Therefore, the authors might have been using the recombinant core protein of human OVGP1 instead of the fully glycosylated recombinant OVGP1 in their study. The same concern also applies to the commercially obtained mouse recombinant OVGP1 used by the authors in their study. I would also like to mention that the mature and fully glycosylated OVGP1s in mammals vary in molecular weight (90-95 kDa in domestic animals; 110-150 kDa in primates; 160-350 kDa in rodents). Again, the 75kDa of mouse OVGP1 detected by the authors could be the core protein or polypeptide of mouse OVGP1 instead of the fully glycosylated mouse OVGP1.

In our study, as previously mentioned, we included commercially available recombinant proteins from Origene for human and murine OVGP1, which are produced in mammalian cells, and we also produced and purified bovine OVGP1 in mammalian cells. Therefore, these proteins should be properly glycosylated. Moreover, we performed Western blot assays favouring the blotting of higher molecular weight proteins, ensuring the optimal conditions for the assay. Additionally, we tested the size of OVGP1 from murine and bovine oviductal fluids on the same blot. During oestrus, the size of OVGP1 from oviductal fluids matches that of the recombinant proteins, and this band is downregulated during anoestrus, confirming the proper size of recombinant protein.

P.7, lines 236 and 237: Please provide a figure or source to support the statement "...as confirmed by proteomics of the bands along with PEAKS Studio v11.5 search engine peptide identification software."

It is included in the text the amount of unique peptides obtained by Proteomics for OVGP1 identification over all protein groups identified.

P.7, lines 243 to 245: The statement "...using rabbit polyclonal antibody to human OVGP1 for bOVGP1 and endogenous OVGP1, and mouse monoclonal antibody against Flag (DDK)-tag for hOVGP1 and mOVGP1." is confusing and might be inaccurate. First of all, I wondered why the authors did not use an antibody against bovine OVGP1 for the recombinant bOVGP1 instead of using a rabbit polyclonal antibody to human OVGP1. Secondly, what does the "endogenous OVGP1" refer to in the statement? Thirdly, the authors in their study used the commercially available recombinant human OVGP1 and recombinant mouse OVGP1 purchased from Origene. Based on the data sheet provided by Origene, the tag used for both recombinant human OVGP1 and recombinant mouse is C-Myc/DDK-tag and not Flag-tag. Can the authors explain these discrepancies?

Firstly, for the recombinant protein of bOVGP1 we used the same antibody that we used in the Western blot for all the proteins and oviductal fluids because we do not have anti-His tag working for Immunofluorescence (the one we had only worked for Western blot) and neither we do not have any antibody against bovine OVGP1. In the case of human and murine since we had anti-Flag antibody that worked for Western blot and for immunofluorescence, we used this one. However, as has been shown in our figure and supplementary material, the antibody against human OVGP1 works properly for both techniques (Western blot and Immunofluorescence). Secondly, endogenous OVGP1 is referred to the OVGP1 present in the oviductal fluid. Thirdly, as you can see in the datasheet of the protein, the recombinant proteins purchased from Origene contains a c-myc tag (EQKLISEEDL) some amino acids and a ddk-tag (DYKDDDDK). The sequence of ddk is the same of Flag-tag (DYKDDDDK). Since the proteins have both tags we used the antibody against Flag (or ddk) epitope.

P12, lines 429-432: The newly added statement at the end of the Discussion saying "Additionally, future studies would be valuable to investigate whether incubating oocytes with oviductal fluid (or OVGP1) could reduce polyspermy in porcine IVF and whether ZPs could be leveraged to naturally enhance sperm selection in human ICSI" is very concerning and requires further attention. The statement reflects that the authors do not keep pace with and do not pay attention to what has been published in literature regarding porcine and human OVGP1s. In fact, porcine oviduct-specific glycoprotein (OVGP1) has already been reported to reduce the incidence of polyspermy in pig oocytes (Biology of Reproduction (2000) 63:242-250). Porcine oviductal fluid, used in porcine IVF, has also been found to exert a beneficial effect on oocytes by reducing the incidence of polyspermy without decreasing the penetration rate. (Theriogenology (2016) 86:495-502). Therefore, the studies deemed valuable by the authors to be investigated in the future have, in fact, already been carried out two decades ago by several other laboratories. I am surprised the authors were not aware of these published work in literature. All the above should have been incorporated in the Discussion.

This sentence is modified in the discussion and the references are included.

Furthermore, as mentioned earlier, recombinant human OVGP1 has also been produced (Reproduction (2016) 152:561-573), and recombinant human OVGP1 has been found to increase tyrosine phosphorylation of sperm proteins, a biochemical hallmark of sperm capacitation, and potentiate the subsequent acrosome reaction (Reproduction (2016) 152:561-573) as well as increase sperm-zona binding (Journal of Assisted Reproduction and Genetics (2019) 36:1363-1377). These earlier findings should be incorporated into the Discussion.

Thank you for your comment, but in this work we had not performed any experimental setting related to tyrosine phosphorylation and despite is a very interesting topic is not directly related to this work.

P.19, lines 678-683: Since the human and mouse recombinant oviductin proteins were purchased from Origene, the authors should be aware of the fact that these commercially available recombinant OVGP1s might not be fully glycosylated. While I appreciate the fact that the authors wanted to briefly describe how the human and mouse recombinant OVGP1s were prepared by the manufacturer, I strongly suggest that the authors should contact Origene, the manufacturer, for all information regarding the procedures for producing the human and mouse recombinant oviductin proteins. For example, the authors stated on lines 680-681 that "A sequence expressing FLAG-tagged epitope proteins (DYKDDDDK) was cloned into an expression vector." According to the data sheet provided by Origene, it appears that both human and recombinant oviductin proteins are C-Myc/DDK-tagged and not FLAG-tagged.

Thank you for your comment, as according to the sequence of Flag-tag it is matching with the sequence of the tag in the datasheet corresponding to DDK (this is in detail in previous comment). Besides, the protein is tagged also by C-Myc tag. Among both tags, the antibody selected to detect it was anti-Flag tag.

P.19, lines 692-697: The description of the primary and secondary antibodies used for detection of the various recombinant OVGP1s is also very confusing and not clearly presented. For example, it is mentioned here that "...membranes were...incubated with anti-OVGP1 rabbit monoclonal antibody for OVGP1,..". What specifically does "OVGP1" refer to here? The authors then stated that anti-Histamine Tag antibody was used to detect bOVGP1 and mOVGP1 and anti-Flag antibody was used to detect hOVGP1. As pointed out earlier, the human and mouse recombinant OVGP1s were produced using C-Myc/DDK tag and not His-tag or Flag-tag. Can the authors clarify these discrepancies?

We apologise for the complexity of the antibodies, we included in this paragraph the ones used to Western blot for both figures: anti- human OVGP1 was used for the principal figure that contains the three recombinant proteins and oviductal fluids; and the anti-Histidine and anti-Flag antibodies that are included in supplementary figure, specifically for recombinant bovine OVGP1 (Histidine tag) and for recombinant murine and human OVGP (DDK tag). A clarifying sentence has been included in the text.

P.31, lines 1143-1149: Figure 10 is not mentioned anywhere in the main text of the manuscript. Rewrite the second half of the sentence "...; being this specificity lost when OVGP1 is heterologous to the ZP (right diagram)." Which sounds awkward and grammatically not correct.

The figure is already mentioned in the text, thank you for your comment. The sentence is also corrected.

Other comments: P.1, the statement of "All authors contributed equally to this work" on line 14 can be deleted because detailed and specific contributions from each authors are listed in lines 1009-1017 on page 27.

Both authors contributed equally to this work, now is clear in authors contribution section.

P.2, lines 43 and 44: Do the authors mean "sperm-oocyte binding protein" instead of "sperm-oocyte fusion protein" in the sentence? "Fusion protein" is a protein composed of two or more domains encoded by different genes, or a hybrid molecule created by combining two different proteins for various purposes. I believe the term "fusion protein" is wrongly used in the sentence which should be rephrased with a proper term.

Done.

P2, line 73: Remove the comma after the word "Both".

Done.

P.5, line 179: "...mice ZP..." should be written as "...mouse ZP...".  

Done.

P.6, heading of 3rd paragraph on line 207: The term "binding" will be a better term than "fusion" used in the heading because the results do not actually show the fusion of the OVGP1 proteins with the ZP glycoprotein. Instead, binding of the OVGP1 proteins to the ZP occurred.

Done.

P.6, lines 215-217: Authors, please provide a reference or references to support the statement "Region A, corresponding to the amino acid end, shows high identity among monotremes, marsupials and placentals."

In the text was indicated a review (29) which includes the supporting idea of this statement for Figure 4. Moreover, we have included some if the references used for the description of the domains when performing the sequence alignment of Figure S5.

P.6, line 230 and line 233 on P.7: Authors, please be consistent in the use of either American English or British English. The word "oestrus" is British English whereas "estrus" is American English.

Done.

P.7, line 264: The word "sticking" used here means non-specific binding. I believe the author means specific binding here. If so, a more appropriate word should be used here instead of "sticking".

Done.

P.7, lines 267-269: This newly added sentence sounds very awkward and should be completely rewritten.

Done.

P.8, line 288: This reviewer finds it difficult to understand the meaning of the heading. The heading should be rephrased to bring out exactly what the authors want to say in well-written English.

Done.

P.8, line 290: The word "would" should be replaced by "could" in the sentence.

Done.

P.13, line 437: Authors, please provide the location of Sigma-Aldrich.

Done.

P.13, line 457: Here, the authors used "1800 rpm" to indicate the centrifugation speed but used the g-force elsewhere in the Materials and Methods. Please be consistent. The g-force is preferred.

Done.

P.14, lines 483-485: The procedure of sacrificing the cats should be provided in the Materials and Methods

Cats weren’t sacrificed they were vasectomized. It is now included in the text.

P.17, line 628: "...the ZPs were exposed or no exposed to..." should be written as "...the ZPs were either exposed or not exposed to...".

Done.

P.17, line 629: "...each groups were incubated with..." should be "...each group was incubated with...".

Done.

P.19, line 700: "As loading control, was used the primary antibody....." is not a complete sentence and it needs to be rewritten.

Done.

P.20, lines 744-754: For scanning electron microscopy and image processing, the procedures of prior treatment of the oocytes with and without oviductal fluid and OVGP1 should be included here.

Done.

P.21, line 756: It is stated here that "Two hundred isolated ZPs were treated with Clostridium perfringens neuraminidase....". However, it is not clear whether two hundred isolated ZPs of both porcine and murine ZPs were treated. Authors, please clarify.

We used 200 isolated ZPs of each specie, bovine and murine. It is classified in the text.

P.28, lines 1039 and 1040: The author only mentioned the use of bovine and murine sperm here. What about human sperm?

Done.

P.29, line 1076: "...in mammalian cells..." is very vague. Be specific what exactly the mammalian cells were.

Done.

P.29, line 1079: "Oviductal fluid from ovulated cows or anoestrus cows." is not a complete sentence and it needs to be rewritten.

Done.

Author response:

Conflation of control, difficulty and reward rate

In response to the comment of control being conflated with task difficulty (and thus reward rate) that the reviewer feels is not adequately discussed in the paper, we will add more to this point in our discussion, especially in relation to previous literature. It is important to note, however, that our measure of perceived difficulty was included in analyses assessing the fluctuations in stress and control. Subjective control still had a unique effect on the experience of stress over and above perceived difficulty, suggesting that subjective control explains variance in stress beyond what is accounted for by perceived difficulty. We will also include additional analyses in which we include the win rate (i.e. percentage of all trials won) as a covariate when assessing the relationship between subjective control, perceived difficulty and subjective stress, which shows that win rate does not predict stress, but subjective control and perceived difficulty still uniquely predict subjective stress. The results of this will be added and elaborated further in the discussion.

Neutral video condition

In response to the comment of the neutral video condition not being active enough, we believe that any task with action-outcome contingencies would have a degree of controllability. To better distinguish experiences of control (WS task) to an experience of no/neutral control (i.e., neither high nor low controllability), we decided to use a task in which no actions were required during the task itself, although concentration was still required (attention checks regarding the content of the videos and ratings of the videos).

The suggestion of having a high arousal video condition would indeed be interesting to test how experiencing ‘neutral’ control and high(er) stress levels preceding the stressor task influences stress buffering and stress relief. This is a good suggestion for future work that we can include in the discussion section.

The TSST version (online and anticipatory)

We will add more information regarding prior literature that the Trier Social Anticipatory Stress test has found physiological and psychological correlates (e.g. Nasso et al., 2019, Schlatter et al., 2021, Steinbeis et al., 2015), suggesting that the anticipation is still a valid stress manipulation despite participants not performing the actual speech task. Further, the TSST had a significant impact on subjective stress in the expected direction demonstrating that it was effective at eliciting subjective stress.

Internal consistency

We will parcellate the timepoints differently (not just odd/even sliders) to test the internal consistency, for example a random split or first half/second half.

Effect of win-loss domain in Study 2

We will run additional analyses testing the interaction of Domain (win or loss) with stressor intensity when predicting the stress buffering and stress relief effects. To test whether the loss domain is more valuable at mitigating experiences of stress than the win condition, we will run additional analyses with just the high control conditions (WS task) to test for a Domain*Time interaction, as we cannot test a Control*Domain*Time interaction in the full model given that we do not have ‘Domain’ for the video (neutral control) condition.

Stress relief analyses

Regarding the stress relief analyses (timepoints 2 and 3) and ‘baseline’ stress (timepoint 1), we will add to the manuscript that there is no significant difference in stress ratings between the high control and neutral control (collapsed across stress and domain) after the WS/video task, hence why we do not think it’s necessary to include in the stress relief model. Nevertheless, we will include a sensitivity analysis in the supplementary material to test the Timepoint*Control interaction (of stress relief – timepoints 2 and 3) when including timepoint 1 stress as a covariate.

Clarity

We will add more clarity in the methods section regarding within- and between-subject manipulations. We will also add Figure S4 to the main manuscript and expand Figure 1 to include both Studies 1 and 2 and a timeline of when subjective stress was assessed throughout the experiment.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

Busch and Hansel present a morphological and histological comparison between mouse and human Purkinje cells (PCs) in the cerebellum. The study reveals species- specific differences that have not previously been reported despite numerous observations of these species. While mouse PCs show morphological heterogeneity and occasional multi-innervation by climbing fibers (CFs), human PCs exhibit a widespread, multi-dendritic structure that exceeds expectations based on allometric scaling. Specifically, human PCs are significantly larger, and exhibit increased spine density, with a unique cluster-like morphology not found in mice.

Strengths:

The manuscript provides an exceptionally detailed analysis of PC morphology across species, surpassing any prior publication. Major strengths include a systematic and thorough methodology, rigorous data analysis, and clear presentation of results. This work is likely to become the go-to resource for quantitation in this field. The authors have largely achieved their aims, with the results effectively supporting their conclusions.

We are grateful to this reviewer for their thoughtful assessment that this work will be a go-to resource for the field.

Weaknesses:

There are a few concerns that need to be addressed, specifically related to details of the methodology as well as data interpretation based on the limits of some experimental approaches. Overall, these weaknesses are minor.

We thank this reviewer for their careful reading of the manuscript and for highlighting limitations and weaknesses in the methodology. We are in full agreement that while interpretation is somewhat limited, there is still value in their description. As detailed below in response to this reviewer’s recommendations, we provide more description of our imaging resolution. This additional detail clarifies that our quantitation is appropriate for the scale of the objects being measured and provides critical information to help readers assess the findings as they may pertain to their own work.

Reviewer #2 (Public review):

Summary:

This manuscript aims to follow up on a previously published paper (Busch and Hansel 2023) which proposed that the morphological variation of dendritic bifurcation in Purkinje cells in mice and humans is indicative of the number of climbing fiber inputs, with dendritic bifurcation at the level of the soma resulting in a proportion of these neurons being multi-innervated. The functional and anatomical climbing fiber data was obtained solely from mice since all human tissue was embalmed and fixed, and the extension of these findings to human Purkinje cells was indirect. The current comparative anatomy study aims to resolve this question in human tissue more directly and to further analyse in detail the properties of adult human Purkinje cell dendritic morphology.

Strengths:

The authors have carried out a meticulous anatomical quantification of human Purkinje cell dendrites, in tissue preparations with a better signal-to-noise ratio than their previous study, comparing them with those from mice. Importantly, they now present immunolabelling results that trace climbing fiber axons innervating human PCs. As well as providing detailed analyses of spine properties and interesting new findings of human PC dendritic length and spine types, the work confirms that human PCs that have two clearly distinct dendritic branches have an approximately x% chance of receiving more than one CF input, segregated across the two branches. Albeit entirely observational, the data will be of widespread interest to the cerebellar field, in particular, those building computational models of Purkinje cells.

We thank this reviewer for their positive and considered assessment of our work. We enthusiastically agree that while these data are descriptive in nature, they may be of interest across modalities of cerebellar research and will provide a more detailed framework for cross-species comparisons and single cell computational modeling, which remains a critical tool to explore the human case given the inaccessibility of physiological experimentation.

Weaknesses:

The work is, by necessity, purely anatomical. It remains to be seen whether there are any functional differences in ion channel expression or functional mapping of granule inputs to human PCs compared with the mouse that might mitigate the major differences in electronic properties suggested.

We are in full agreement with the reviewer that the focused anatomical description of this manuscript could not make strong assertions about function given that cellular and circuit physiology is determined by many additional factors that remain unexamined. We appreciate that the reviewer acknowledges that this is out of necessity as those factors are inaccessible to experimentation at the current time; however, we are enthusiastic that our current findings will motivate future work that will shed light on these critical additional features of the system, both in rodents and humans.

Reviewer 1 (Recommendations for the authors):

PCs are now known to be genetically diverse, with unique PC types found only in humans. Could this cellular diversity contribute to the differences observed between species in this study? This possibility should be at least discussed in the context of the findings.

We agree that this is a fascinating possibility. The perhaps most detailed recent study (Sepp et al., Nature 625, 2024) – in a conservative assessment – describes four developmental PC subtypes in mice that are identical in humans. The study points out that the subtype ratio changes over the course of development, though. Taken together with the possibility of additional human-specific subtypes, a genetic basis for morphological as well as physiological diversity arises. This is now discussed on p. 7. It needs to be kept in mind, however, that other factors, such as push-pull influences during tissue growth, might also play a role.

The human tissue used in this study was obtained from elderly individuals, while the mouse tissue was not. It is unclear whether the age difference might influence the findings, and this warrants further discussion or control.

We share this concern, in particular regarding the spine / spine cluster analysis as here tissue quality and or degenerative effects might play a role. We additionally analyzed a tissue sample from a 37 year-old human, and observed the same spine clusters as in the other human brains. This is now described on p. 4 of the revised manuscript.

The study includes spine size comparisons, but it is not clear if the point spread function (PSF) of the microscope provides the necessary resolution for these quantitative assessments. For instance, are multi-headed spines truly multi-headed, or could this be an artifact of limited resolution?

This is an important point. We addressed it by calculating the Rayleigh limit (more conservative than the Abbe limit) as 248.4nm for the equipment and conditions used (Methods, p. 22). On pages 3-5, we updated our Results section accordingly to point out what quantifications are well supported and discuss the limitations (p. 3-5).

Reviewer 2 (Recommendations for the authors):

This is nice work which must have been very time-consuming. It would be good to make sure that the technical details are properly discussed, to quantify the data properly. Please include details of how you measured the resolution of the microscope used to evaluate spine size.

See our response to the last comment of Referee 1 above.

The figure panels are mostly satisfactory, but they are exceptionally crowded and will probably be difficult to read at the final size. Some work tidying these would be worth it. In Figure 3B, include mention of open and blue triangles in legend. In 3E, the dendritic branches are shown at a different gray scale. You have not done this elsewhere, so probably good to mention it in the legend.

Figure 3 and its legend have been updated / improved accordingly.

The definition of horizontal and vertical is not absolutely clear. Perhaps re-assess this bit of the text. Does it mean that you did not include cells that were neither vertical nor horizontal?

We categorized those PCs as ‘vertical’ that have a >30° angle relative to the PC layer, and those as ‘horizontal’ that have a <30° angle relative to the PC layer. All PCs are covered by these categories. This is now described on p. 5.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study from Belato, Knight, and co-workers, the authors investigated the Rec domain of a thermophilic Cas9 from Geobacillus stearothermophilus (GeoCas9). The authors investigated three constructs, two individual subdomains of Rec (Rec1 and Rec2) and the full Rec domain. This domain is involved in binding to the guide RNA of Cas9, as well as the RNA-DNA duplex that is formed upon target binding. The authors performed RNA binding and relaxation experiments using NMR for the wild-type domain as well as two-point mutants. They observed differences in RNA binding activities as well as the flexibility of the domain. The authors also performed experiments on fulllength GeoCas9 to determine whether these biophysical differences affect the RNA binding or cleavage activity. Although the authors observed some changes in the thermal stability of the mutant GeoCas9-gRNA complex, they did not observe substantial differences in the cleavage activities of the mutant GeoCas9 variants.

Overall, this manuscript provides a detailed biophysical analysis of the GeoCas9 Rec domain. The NMR assignments for this construct should prove very useful, and the results may provide the grounds for future engineering of higher fidelity variants of GeoCas9. While the NMR results are generally well presented, it is unclear how the results on the isolated Rec domain related to the overall function of full-length GeoCas9. In addition, some conclusions are overstated and not fully supported by the evidence provided. The following major points should be addressed by the authors.

(1) Many of the results rely on the backbone resonance assignments of the three constructs that were used, and the authors have done an excellent job of assigning the Rec1 and Rec2 constructs. However, it is unclear from the descriptions in the text how the full-length Rec construct was assigned. Were these assignments made based on assignments for the individual domains? The authors state that the spectra of individual domains and RecFL overlay very well, but there appear to be many resonances that have chemical shift differences or are only present in one construct. As it stands, it is unclear how the resonances were assigned for residues whose chemical shifts were perturbed, making it difficult to interpret many of the results.

The Reviewer raises an important oversight. In Lines 491-493, we clarify that we were able to transfer the assignments using spectral overlays of the individual domains with GeoRec (i.e. careful analysis of the data in Figure S3). We also cite two new references where a similar approach was applied to Cas9.

(2) The minimal gRNA that was used for the Rec-gRNA binding experiments is unlikely to be a good mimic for the full-length gRNA, as it lacks any of the secondary structure that is most specifically recognized by the REC lobe and the rest of the Cas9 protein. The majority of this RNA is a "spacer" sequence, but spacers are variable, so this sequence is arbitrary. Thus, the interactions that the authors are observing most likely represent non-specific interactions between the Rec domains and RNA. The authors also map chemical shift perturbations and line broadening on structural models with an RNA-DNA duplex bound, but this is not an accurate model for how the Rec domain binds to a single-stranded RNA (for which there is no structural model). Thus, many of the conclusions regarding the RNA binding interface are overstated.

The Reviewer again raises an important point. We have added a section of text explaining the rationale for truncating the gRNA for binding experiments with NMR (Lines 223-235). We chose the 5’end of the gRNA containing the spacer sequence based on crystal structures of NmeCas9 and SpCas9 that show the Rec lobe interacting with this section of nucleic acid. The newly published GeoCas9 cryo-EM structure bound to gRNA, which overlaid well with the NmeCas9 structure, also suggested that this portion of the gRNA could interact with Rec.

Figures S11 and S12 show our gradual truncation of the gRNA and Rec construct to achieve useful atomic detail. Ultimately, a 39nt gRNA containing a 23 base pair spacer sequence was chosen for this study to retain the NMR signal of the complex and because several structures suggested this 39nt sequence would be long enough to interact with the entire Rec lobe.

To investigate the effect of the spacer sequence, we have now measured binding affinities via MST between GeoRec and a 39nt Tnnt2 gRNA and a 39nt gRNA from PDB: 8UZA, containing a different spacer sequence used in the very recent GeoCas9 cryo-EM structure. The observed trends for each gRNA are consistent across the samples. We also measured WT, K267E, and R332A GeoCas9 affinity for the full-length Tnnt2 and PDB:8UZA gRNAs.

Lastly, we used a new cryo-EM structure of GeoCas9 bound to gRNA (PDB: 8JTR) to better define the interface for NMR CSPs and line broadening and have adjusted the language in this section.

(3) The authors include microscale thermophoresis (MST) data for the Rec constructs binding to the minimal gRNA. These data suggest that all three Rec variants have very similar Kd's for the RNA. Given these similarities, it is unclear why the RNA titration experiments by NMR yielded such different results. Moreover, in the Discussion, the authors state that the NMR titration data are consistent with the MST-derived Kd values. This conclusion appears to be overstated given the very small differences in affinities measured by MST.

MST and NMR experiments describing the weakened binding affinity of GeoRec and GeoRec2 for the Tnnt2 gRNA agree with each other (Figure 5). However, additional MST experiments with a different gRNA sequence (from PDB: 8UZA) and with fulllength GeoCas9 (new Figure 7) have provided new insight for us to soften and reframe the Discussion to avoid overstatement. See Lines 263-282 and 375-385.

(4) While the authors have performed some experiments to help place their findings on the isolated Rec domain in the context of the full-length protein, these experiments do not fully support the conclusions that the authors draw about the meaning of their NMR results. The two Cas9 variants that were explored via NMR have no effect on Cas9 cleavage activity, and it is unclear from the data provided whether they have any effect on GeoCas9 binding to the full sgRNA. This makes it difficult to determine whether the observed differences in RNA binding and dynamics of the isolated Rec domain have any consequence in the context of the full protein.

We have since measured the binding affinities of full-length GeoCas9 to full-length gRNA. (new Figure 7) We have also added a comment in the Discussion section describing how both GeoRec and GeoRec2 domain variants bind the truncated RNA with weaker affinity than the WT, but this biophysical effect does not translate to GeoCas9 with its full-length gRNA. We describe this finding as an explanation for why the single-point mutants have minimal effect of GeoCas9 cleavage activity. See Lines 375-385.

(5) The authors state in multiple places that the K267E/R332A mutant enhanced GeoCas9 specificity. Improved specificity refers to a situation in which the efficiency of cleavage of a perfectly matched target improves in comparison to a mismatched target. This is not what the authors observed for the double mutant. Instead, the cleavage of the perfect target was drastically reduced, in some cases to a larger degree than for the mismatched target. The double mutant does not appear to have improved specificity, it has simply decreased cleavage efficiency of the enzyme.

The conclusion has been reframed to suggest that the K267E/R332A double mutant has decreased cleavage efficiency of the enzyme but does not enhance GeoCas9 specificity. We discuss an interesting contrast, namely that mutations in the SpCas9 Rec lobe alter its specificity, which is at times accompanied by a loss of overall activity. We also speculate on why this may not be the case in GeoCas9, considering some very recent (unpublished at the time of initial submission) structural and biochemical data. See Lines 414-418.

Reviewer #2 (Public Review):

Summary:

The manuscript from Belato et al. used advanced NMR approaches and a mutagenesis campaign to probe the conformational dynamics of the recognition lobe (Rec) of the CRISPR Cas9 enzyme from G. stearothermophilus (GeoCas9). Using truncated and full-length constructs they assess the impacts of two different point mutations have on the redistribution and timescale of these motions and assess gRNA recognition and specificity. Single point mutations in the Rec domain in a Cas9 from a related species had profound impacts on- and off-target DNA editing, therefore the authors reasoned analogous mutations in GeoCas9 would have similar effects. However, despite a redistribution of local motions and changes in global stability, their chosen mutations had little impact on DNA editing in the context of the full-length enzyme. Their studies highlight the species-specific complexity of interdomain communication and allosteric mechanisms used by these multi-domain endonucleases. Despite these negative results, their study is highly rigorous, and their approach will broadly support understanding how the activity and specificity of these enzymes can be engineered to tune activity and limit off-target cleavage by these enzymes.

Strengths:

(1) Atomistic investigation of the conformational dynamics of the GeoCas9 gRNA recognition lobe (GeoRec), probing dynamics on a broad range of timescales from ps to ms using advanced NMR approaches will be broadly interesting to both the structural biology and CRISPR engineering communities.

(2) Highly rigorous biophysical studies that push the boundaries of current techniques, provide insight into local dynamics of the GeoRec domain that serve to propagate allosteric information and potentially regulate enzymatic activity.

(3) The study highlights the complexities of understanding interdomain communication in Cas9 enzymes since analogous mutations in different species have different effects on target recognition and cleavage.

(4) The type of structural and dynamic insights derived from this study design could serve as foundational information to guide a rational design strategy aimed at improving the selectivity and reducing the off-target effects of Cas9 enzymes.

Weaknesses:

(1) Despite the rigor of the experiments, the mutations chosen by the authors do not have a profound effect on the overall substrate affinity or activity of GeoCas9 rendering little mechanistic insight into allosteric communication in this particular Cas9. However, the double mutant K267E/R332A has a more pronounced effect on the cleavage of WT and mismatched (at nucleotides 19 and 20) DNA substrates while minimally affecting the cleavage of mismatched (at nucleotides 5 and 6), suggesting more could be learned about the allosteric mechanism from the detailed characterization of this mutant.

We thank the Reviewer for this comment. While we have included new binding experiments with full-length GeoCas9 and gRNAs (new Figure 7), the addition of new MD simulations (new Figure 6) better address this point. MD examined our single and double mutants, as well as the recently published high-specificity iGeoCas9, and reported the degree of conformational sampling and nucleic acid contacts and binding energies.

The simulations show that our mutations induce some, but not the full extent of the effect of iGeoCas9 (with one mutation in GeoRec and many others in the adjacent WED domain), implying that further engineering of GeoRec to mimic iGeoCas9’s properties can have profound functional outcomes. Future efforts to mutate GeoRec will be leverage this strategy. See Lines 309-342.

(2) Follow-up experiments with other residues that were identified as being highly dynamic might affect substrate recognition and cleavage activity in different ways providing additional insight.

The Reviewer is correct. While beyond this initial scope, new MD simulations (see the response directly above) and NMR resonances distally affect by gRNA (via CSP or relaxation dispersion) will be used identify the primary targets for this analysis.

(3) Details regarding the authors' experimental approach are incomplete such as a description of the model used to fit the CD data, a detailed explanation of the global fitting of the relaxation dispersion data describing how the best-fit model was selected, and the description of the ModelFree fitting of fast timescale dynamics is incomplete.

We thank the Reviewer for pointing out these oversights. We have now included the fitting equation in the CD Methods section.

We included new Figures S8-S10 with the individual relaxation dispersion curves and note in the Methods that global fits were deemed superior based on the Akaike Information Criterion. For WT, the AIC showed the global fit to be ~10-fold better. For K267E, the global model was 4-fold better, and for R332A, the global model was 6-fold better.

We have included a more detailed description of CPMG and Model-free fitting. See Lines 520-526.

Reviewer #3 (Public Review):

The authors explore the role of Rec domains in a thermophilic Cas9 enzyme. They report on the crystal structure of part of the recognition lobe, its dynamics from NMR spin relaxation and relaxation-dispersion data, its interaction mode with guide RNA, and the effect of two single-point mutations hypothesised to enhance specificity. They find that mutations have small effects on Rec domain structure and stability but lead to significant rearrangement of micro- to milli-second dynamics which does not translate into major changes in guide RNA affinity or DNA cleavage specificity, illustrating the inherent tolerance of GeoCas9. The work can be considered as a first step towards understanding motions in GeoCas9 recognition lobe, although no clear hotspots were discovered with potential for future rational design of enhanced Cas9 variants.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Suggestions for improved or additional experiments, data, or analyses

(1) Please update the sentences on lines 100-105 and the Methods to clarify how the RecFL assignments were obtained. If RecFL was assigned based on the assignments for Rec1 and Rec2, please describe in the Methods how the shifted resonances were handled. Please also provide chemical shift perturbation profiles for the truncated constructs versus the full-length Rec construct.

We have now added text (Lines 491-493) and two new references explaining the GeoRec (full-length) assignment.

We appreciate this point. We have now provided a new Figure S9 with analysis of CSPs and line broadening in truncated constructs (GeoRec2 only). See also Lines 263-282. We also show a similar structural response to mutation in full-length GeoRec and GeoRec2 NMR CSPs (Figure 2 and Figure S5).

We have provided the CSPs for each construct, relative to the full-length GeoRec domain, Author response image 1. In most cases, the largest CSPs occur at resonances on the periphery of the spectra, retaining the ability to unambiguously assign it.

Author response image 1.

(2) It is unclear whether the differences in Kd's for the Rec-gRNA interactions are statistically significant, given the errors associated with the values. Can the authors further analyze these data to determine statistical significance? If they are not found to be significantly different, the authors should soften all conclusions related to the observed differences.

Statistical significance was calculated for all MST data and Figures 5 and 7 have been updated to reflect this

(3) As mentioned above, it seems likely that the Rec-RNA binding that is observed is non-specific. Have the authors tried MST with another 39 nt RNA? Are there differences in affinities for the Rec constructs?

We have done MST with another 39nt RNA. The affinity for each gRNA (Tnnt2 vs 8UZA) is similar for WT and K267E, and a factor of ~4 weaker for R332A with 8UZA gRNA. The trend is the same, that WT Rec has a (statistically significant) stronger affinity for the gRNA compared to the mutants.

(4) Have the authors tried MST with full-length GeoCas9 and the sgRNA? The current data on the thermal stability of the RNP's is interesting, but a more direct measurement of the affinity of the Cas9-sgRNA complexes would provide stronger evidence of the effects of the mutations.

The Reviewer makes an excellent suggestion. We have now generated Cy5-labeled full-length gRNAs and conducted MST with full-length GeoCas9 (new Figure 7). The binding affinities to multiple guides do not vary significantly. We have discussed this, and its implications, in Lines 376-385.

(5) One potential issue with not observing differences between the three Cas9 variants' cleavage activity is that the activity of these purified proteins appears to be very low in comparison to previous studies of GeoCas9. There are significant differences in the expression protocol used by the authors of the current study and previous studies. Have the authors attempted to replicate the expression and purification protocol of previous reports? This may improve the enzymatic activity and allow for a more detailed investigation of cleavage between the three variants (e.g. by performing time-course cleavage assays).

The expression protocol of GeoCas9 is identical to those of previous studies. This was a written mistake on our part, which has now been corrected in the methods section. We apologize for this oversight.

Recommendations for improving the writing and presentation

The introduction of the manuscript is reasonable for specialists who are very familiar with Cas9 function, but it does not contain important details that may be unknown to most readers. The authors do not introduce the domains of Cas9 in the Introduction section. A brief description of the domains that are important to this work should be provided. For example, what is the role of the Rec lobe? This is not introduced until lines 110-111, after some discussion of the authors' initial work on these domains. For a broad audience, it would also be helpful to define the two catalytic domains of the protein. A paragraph describing the general architecture of Cas9 and the overall mechanism of Cas9, including allostery and domain movement, would be very helpful to a general audience. There are elements of this throughout the manuscript, but it would be better to have everything described in a single location at the beginning of the Introduction.

The Reviewer makes an excellent point. We have added significant clarifying text to the Introduction (Lines 42-47, 52-58, and 61-66). We have also amended Figure 1 to highlight the domain arrangement of GeoCas9 and construct domain boundaries.

Minor corrections to the text

(1) Lines 37-38: The statement about GeoCas9 activity should reference citation.

We have added two references here.

(2) Line 39-40: "The widely-studied SpCas9, as well as GeoCas9, are Type-II CRISPR systems". Cas9 is only a single component of a larger system that contains other proteins and DNA elements, so it would be more appropriate to say "are effectors of type II CRISPR systems" or "are signature proteins of type II CRISPR systems". Also, please define the organism from which SpCas9 is derived. It may be more appropriate to use the three-letter abbreviation "SpyCas9" to be consistent with the abbreviation used for GeoCas9.

We have revised the initial suggestion and specified the organisms. We have, however, chosen to keep “SpCas9” for consistency with our prior work and the work of many several others, including Doudna et al and Zhang et al.

(3) Lines 39-42: "only the Type II-C class to which GeoCas9 belongs has been rigorously validated for mammalian genome editing". SpCas9 is from a type II-A system and is by far the most commonly used ortholog for genome editing, including in ongoing clinical trials. It is unlikely that any of the type II-C Cas9 orthologs have been more rigorously validated than SpCas9. The reference cited in this sentence also does not support this statement and is a review written in 2017, so would be unlikely to reflect the current state of the art. Please revise this sentence.

We have softened and revised this text (Lines 42-47).

(4) Lines 48-52: It would be helpful to describe the dynamic movement of the HNH domain (and cite appropriate references) prior to describing the authors' previous work. As it stands, it is unclear how this sentence would be understood by a non-specialist.

We have added text in Lines 61-68

(5) Lines 44-45: The wording is a little unclear, as it sounds like the guide RNA, rather than the nuclease domains, is responsible for dsDNA cleavage. The sentence could be adjusted to remove "and cleave". Cleavage by the HNH and RuvC domains could be described in a separate sentence.

We have revised this text. See Lines 49-50.

(6) Lines 46-48: This segment of the sentence suggests that PAM recognition triggers the allosteric events that result in the movement of the nuclease domain (HNH). This is misleading, as HNH movement is triggered by the complete formation of an R-loop, rather than initial PAM recognition. Please revise this sentence.

We have revised the text in Lines 52-58.

(7) Lines 62-65: The first sentence is unclear. The specificity of many protein-nucleic acid complexes is well understood and is also readily quantified by several wellestablished methods. Are the authors specifically referring to the structural basis for Cas9 specificity? Although Cas9 specificity is highly complex, it has been studied structurally in great detail and should not be described as "poorly understood" without some discussion of what is already known. These sentences also elide the fact that Cas9 specificity has been successfully altered via rational design, based on our general framework for understanding protein-nucleic acid interactions. Please clarify these statements.

The Reviewer makes an important point. We have softened this statement (Lines 8081). We have clarified that we intended to refer to structural characterization of large, multidomain proteins and nucleic acid complexes via NMR. We agree that many critical structural studies comment on Cas9 dynamics and specificity in great detail, including at the domain-level.

(8) Lines 62-68: It seems like the citations do not match up with the references in this section. The references for citations 8-10 are not about DNA repair complexes, references 11-14 are not papers about the directed evolution of Cas9 (should these be 16-17?), and the references for the HNH domain movements should be for citations 1821.

We apologize for the confusion, and the references have been updated

(9) Lines 116-119: The description of the RNAs used is unclear, as the segments that are described add up to 141 not 101. Also, what is meant by "110-nt guide sequence intrinsic to GeoCas9"? Is this referring to the tracrRNA segment? It may be helpful if the RNA sequences shown in the accompanying figures were replaced with cartoons of the RNAs that were used, with the different segments labeled.

We now describe the gRNA sequences in detail in new Table S4. We also expanded a bit in the text (Lines 224-235).

(10) Line 121-123: This sentence should contain reference(s).

We have changed the sentence.

(11) Line 156-158: Reference 19 did not report or investigate any higher specificity SpCas9 variants, is this citation correct?

We have removed the reference from this line. Ref. 19 (now Ref 23, Slaymaker et al) should be correct.

(12) Lines 162-166: Please provide a sequence and structural alignment for SpCas9 and GeoCas9 to support the claim that the amino acid substitutions are equivalent between the two orthologs.

We have updated Figure 1 to display the similarity in domain arrangement between SpCas9 and GeoCas9 and have noted similarity in structure and sequence of these proteins in Figure S1.

(13) Lines 234-236: There is insufficient evidence to conclude that the alterations in protein dynamics caused the changes in gRNA interaction. The substitutions are charge swap substitutions, and it is equally (if not more) feasible that these substitutions decrease the potential for favorable electrostatic interactions.

(14) Lines 261-265: While the RNP stability for R332A is clearly decreased in comparison to WT, the authors' conclusions regarding K267E seem overstated. The difference in Tm for the K267E mutant and WT RNPs is not very large and may be within error, especially given that the CD data are noisy. Similarly, on lines 321-322, only one of the mutations really impacted the stability of the full-length RNP.

We have softened this text in Lines 303-305.

(15) Lines 336-338: HiFi-SpCas9 does not contain four mutations, it is a single R691A point mutation, as reported in reference 17. This sentence and subsequent sentences should be updated.

Here, the “final form” of HiFi SpCas9 contains the R691A and three additional mutations. The Reviewer is correct, though, that the R691A mutation alone was enough to enhance the specificity of WT SpCas9. We have clarified this point on Line 156.

Minor corrections to the figures

(16) The cryo-EM structures of GeoCas9 have recently been released on the PDB. The authors may now update figures to include the experimentally determined structure, rather than an AlphaFold model and update the text accordingly.

We have made this change.

(17) For Figure S4, please describe what the red dashed lines are in the top three graphs. Are these the Tm values determined for the two individual Rec domains? How do these compare to the inflection points for the two transitions in the full Rec construct (could be determined by plotting the first derivative data)? Please provide information in the Methods on how the temperature-dependent CD spectral data were fit and Tm's were determined.

We have made these changes in the Figure S4 caption and Methods section.

(18) The blue box denoting the unassigned region is missing from Figure 2C-D, although it is mentioned in the figure legend.

We have added the blue box denoting the unassigned linker.

Reviewer #2 (Recommendations For The Authors):

The manuscript is well-written and generally clear and concise. The following recommendations will help improve the readability and include details important for interpreting the results.

(1) In general, the figures are too small and difficult to interpret, it was hard to discern the differences described in the text (e.g. Figure 1A, E, 4A, etc.), the text labels are illegible in several panels (e.g. Figure 4A, S8B, C, etc.), the chosen colors were difficult to interpret in the structures (Figure 4C, S8G, H, etc.), as well as residues with motion (as balls) were difficult to make out due to size and color usage. Similar story for the dispersion curves (Fig 3A), the plots are chaotically crowded, and it is impossible to interpret (or see) the undelaying data.

We apologize for these difficulties. We have now revised the Figures in several ways. First, we greatly simplified Figure 1, such that it now includes only the domain arrangement, structure, and initial NMR details for GeoRec (essentially A-B of the old Figure 1).

Second, we have reformatted Figure 3 to make the structure maps a bit easier to see.

We certainly appreciate the point made by the Reviewer about the dispersion curves. Our intent here is to illustrate the number of curves that can be fit globally, which substantially increase for K267E and R332A GeoRec3, versus WT. As a compromise, we have included the individual dispersion curves in the SI for each variant. We have also thinned the line weights for each fit, and added NMR order parameters to the main figure to round out the discussion of dynamics.

Third, we have compiled the gRNA titration into Figure 4, removing the CD analysis (to SI), MST data (new Fig 5), and unclear structure maps to focus only on the NMR spectra here.

Fourth, we have created a new Figure 5 focusing on MST studies of two gRNAs with GeoRec, which now include bar charts of affinities with appropriate statistics.

Much of the data trimmed from the prior version of the manuscript figures has been moved to Supporting Information. We have also created two new main text Figures (6 & 7) based on MD simulations and MST studies of full-length GeoCas9 and gRNAs to provide additional context for interpreting the results in prior figures.

(2) Line 39 - this sentence is awkward, could you rephrase it?

We have rephrased this sentence.

(3) There is inconsistent labeling, in Figure S2 the full-length construct is referred to as GeoRecFL while in other places in the text and in Figure 1 it is called GeoRec.

We have changed all references to the intact Rec lobe to “GeoRec.”

(4) It would be helpful to include a cartoon of the domain organization of GeoCas9 and indicate the truncation mutants that were studied in this manuscript.

We included the domain organization in Figure 1A and indicated the amino acid boundaries for each construct on the figure and in the Methods section.

(5) There is significant line broadening that occurs during the titration, not all line broadening is due to changes in rotational correlation time, and differential line broadening may reveal interactions of residues that are in the intermediate regime, certainly, uM affinities measured by the authors, would suggest this, therefore, a plot of I/Io might inform on binding sites, and it might be useful to look at differential broadening as a function of titrant added.

The Reviewer makes a very good point. In addition to the data in Figure 4, which show a clear reduction in gRNA-induced line broadening in larger GeoRec constructs, we included new titration data on smaller GeoRec2 domains (Figure S12). Here, we conducted an I/I0 analysis and added some clarifying language about the possible nature of line broadening in these samples. See new Figure S12 and Lines 268-274.

(6) Line 126 "Importantly, many resonances are also minimally impacted." This statement is unclear since from the plots shown in Figure 1D, it seems that many of the residues are impacted by RNA titration, see the point about differential broadening above, this sort of plot may help pick apart residues that broaden due to RNA contacts (rather than changing rotational correlation).

We have removed this statement, in addition to our revisions above regarding the line broadening.

(7) Line 137 - I am not sure that a max chemical shift of 0.15 ppm constitutes "strong chemical shift perturbations"

The Reviewer makes a good point. We have changed “strong” to “significant” which refers to 1 standard deviation above the 10% trimmed mean of the data. See Line 237.

(8) Line 144 - change to "...experimentally determined structure...".

We have added new lines 135-136 to make this point clear. We reinforced that initial predictions were based on the Alphafold2, since an experimental structure was lacking, but we have now discussed the mutations in context of the new structural data.

(9) The section from lines 150 - 166, comparison of the effect of different mutations in different Cas9 seems more appropriate for the discussion section.

We have added additional text on this point in the Discussion section, within several new paragraphs.

(10) In Figure S6, chemical shifts are observed at the distal site away from the mutations, could the authors discuss?

The Reviewer makes an important observation. Indeed, the CSPs caused by K267E and R332A extend beyond the mutation site. These shifts are mostly close in 3D space to the mutation, and consistent in Figures 2 and S5. New titrations of gRNA into isolated GeoRec2 also activate some distal sites, and new MD simulations suggests the mutations disrupt RNA and DNA contacts, where these distal effects may play a role with full-length gRNAs.

We agree it would be worth mutating distal sites undergoing CSPs to examine their impact on function, but two complicating factors are 1) the lack of substantial gRNA affinity differences in experiments with full-length GeoCas9 and 2) the lack of functional changes in the mutants. In this initial study, it appears difficult to assign an effect to these distal sites in GeoCas9 (beyond speculation). We do have a brief discussion of the distal sites (Lines 293-298) and will follow up this work with more comprehensive mutagenesis studies of these sites.

(11) It appears that the authors fitted the Tm data to some model although this is not mentioned in the text, figure captions, or methods. In the caption for Figure 4D the authors refer to "Fitted thermal denaturation profiles...".

We have added the relevant Equation in the Methods and referenced it in Figure S6 and S14 captions.

(12) Details of the ModelFree fitting are needed, how many residues fit with the minimal models, and how many invoked Rex and other terms? How does the statement in line 191 about the elevated S2 values arising from global tumbling compare with an experimental estimation of rotational correlation eg. from R2/R1 ratios?

We have included an expanded description of the Model-free protocol (Lines 521-527). The best diffusion tensor was an ellipsoid model. The number of residues utilizing Rex was 81, though Rex contribution was very small. The mean and errors for the fast motion (S²_f), slow motion (S²_z) and generalized order parameter were 0.97 ± 0.15, 0.84 ± 0.14, and 0.91 ± 0.20, respectively.