Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Formins are complex proteins with multiple effects on actin filament assembly, including nucleation, capping with processive elongation, and bundling. Determining which of these activities is important for a given biological process and normal cellular function is a major challenge.

Here, the authors study the formin FHOD3L, which is essential for normal sarcomere assembly in muscle cells. They identify point mutants of FHOD3L in which formin nucleation and elongation/bundling activities are functionally separated. Expression of these mutants in neonatal rat ventricular myocytes shows that the control of actin filament elongation by formin is the major activity required for the normal assembly of functional sarcomeres.

Strengths:

The strength of this work is to combine sensitive biochemical assays with excellent work in neonatal rat ventricular myocytes. This combination of approaches is highly effective for analyzing the function of proteins with multiple activities in vitro.

Weaknesses:

FHOD3L does not seem to be the easiest formin to study because of its relatively weak nucleation activity and the short duration of capping events. This difficulty imposes rigorous biochemical analysis and careful interpretation of the data, which should be improved in this work.

We thank the reviewer for their praise and appreciation of our work. Indeed, FHOD3L is a challenging formin to work with.

Important points are raised here and below regarding the brief elongation events we reported. As suggested, we performed more rigorous analysis of the data and present it in the revised manuscript. We now report that from 45 dim regions analyzed, in three independent experiments with wild type FHOD3L, we detected 40 bursts. (The remaining five could be formin falling off too quickly to detect or the dim spots could be regions of inhomogeneity in intensity, not due to formin.) For comparison to the presented data with FHOD3L-CT, we analyzed the filaments in TIRF assays with no formin present. As the reviewers point out, inhomogeneities in filament intensity are normal. Thus, we examined any dim spots for pauses and/or bursts. As is now reported in Figure 2G,H, the velocity of growth of these dim spots is indistinguishable from the velocity of the rest of the filament. We acknowledge that our numbers may not be perfectly accurate, due to the noise in our system, we believe that the difference of 3-4 fold increase versus no change in rate is substantial and convincing.

We also determined the number of dim spots per length of filament. We found a higher frequency when FHOD3L-CT or FHOD3S-CT was present vs no formin, as now shown in Figure 2 – supplements 1G and 2E.

We were asked about the pauses we observe before bursts of elongation and how we know they are functionally relevant. The short answer is that we do not know. We reported them because they were so common: Of the 40 bursts, pauses preceded the burst in 38 cases. We cannot rule out that this pause reflects an interaction with the surface but might expect the frequency to be lower if it were. We revise the text to make our conclusions about pauses more circumspect.

We are convinced that the brief dim events we observed in the presence of FHOD3L-CT, in fact, reflect formin-mediated elongation and worked hard to improve their presentation, in addition to the added analysis. We include new kymographs, including examples from FHOD3L, FHOD3S, K1193L, and actin alone. We hope that the reviewers are also convinced.

This does not preclude our interest in the microfluidics and two-color assays, which will be pursued in the future. We have reached out to a colleague who is set up to repeat these measurements with microfluidics-assisted TIRF. The noise should be greatly reduced and the system is also optimal for directly visualizing labeled FHOD3, as suggested. We expect these experimental approaches will provide additional insights.

Reviewer #2 (Public review):

This article elucidates the biochemical and cellular mechanisms by which the FHOD-family of formins, particularly FHOD3, contributes to sarcomere formation and contractility in cardiomyocytes. Formins are mainly known to nucleate and elongate actin filaments, with certain family members also exhibiting capping, severing, and bundling activities. Although FHOD3 has been well-established as essential for sarcomere assembly in cardiomyocytes, its precise biochemical functions and contributions to actin dynamics remain poorly understood.

In this study, the authors combine in vitro biochemical assays with cellular experiments to dissect FHOD3's roles in actin assembly and sarcomere formation. They demonstrate that FHOD3 nucleates actin filaments and acts as a transient elongator, pausing elongation after an initial burst of filament growth. Using separation-of-function mutants, they show thatFHOD3's elongation activity - rather than its nucleation, capping, or bundling capabilities - is key for its sarcomeric function.

The experiments have been conducted rigorously and well-analyzed, and the paper is clearly written. The data presented support the authors' conclusions. I appreciate the detailed description and rationale behind the FHOD3 constructs used in this study.

We are happy to hear others find paper to be clearly written and well described.

However, I was somewhat surprised and a bit disappointed that while the authors conducted single-color TIRF experiments to observe the effects of FHOD3 on single filaments, they did not use fluorescently labeled FHOD3 to directly visualize its behavior. Incorporating such experiments would significantly strengthen their conclusions regarding FHOD3's bursts of elongation interspersed with capping activity. While I understand this might require a few additional weeks of experiments, these data would add considerable value by directly testing the proposed mechanism.

We appreciate the suggestion and hope to incorporate a two-color approach soon. As noted, FHOD3L is not always easy to work with and we do not have a functional labeled copy of the protein at this time.

There is a typo in the word "required" in line number 30. The authors also use fit data to extract parameters in several panels (e.g., Figures 2b, 2d, 3a, and 3b). While these fit functions may be intuitive to actin experts, explicitly describing the fit functions in the figure legends or methods would greatly benefit the broader readership.

Thank you for these comments. We updated the indicated figures and described the analysis in greater detail.

Reviewer #3 (Public review):

Valencia et al. aim to elucidate the biochemical and cellular mechanisms through which the human formin FHOD3 drives sarcomere assembly in cardiomyocytes. To do so, they combined rigorous in vitro biochemical assays with comprehensive in vivo characterizations, evaluating two wild-type FHOD3 isoforms and two function-separating mutants. Surprisingly, they found that both wild-type FHOD3 isoforms can nucleate new actin filaments, as well as elongate existing actin filaments in conjunction with profilin following barbed-end capping. This is in addition to FHOD3's proposed role as an actin bundler. Next, the authors asked whether FHOD3L promotes sarcomere assembly in cardiomyocytes through its activity in actin nucleation or rather elongation. With two function-separating mutants, the authors evaluated the numbers and morphology of sarcomeres, as well as their ability to beat and generate cardiac rhythm. The authors found that while the wild-type FHOD3L and the K1193L mutant can rescue sarcomere morphology and physiology, the GS-FH1 mutant fails to do so. Given that in GS-FH1 mainly elongation activity is compromised, the authors concluded that the elongation activity of FHOD3 is essential for its role in sarcomere assembly in cardiomyocytes, while its nucleator activity is dispensable. Overall, this important study provided a broadened view on the biochemical activities of FHOD3, and a pioneering view on a possible cellular mechanism of how FHOD3L drives sarcomere assembly. If further validated, this can lead to new mechanistic models of sarcomere assembly and potentially new therapeutic targets of cardiomyopathy.

The conclusions of this paper are mostly well supported by the comprehensive biochemical analyses performed by the authors. However, the sarcomere assembly defect phenotype in the GS-FH1 rescue condition requires further investigation, as the extremely low level of GS-FH1 signal in transfected cells in Figure 6A may reflect a failure of actin-binding by this construct in vivo, rather than its inability to drive elongation. Though the authors do show in Figure 6 that GS-FH1 can bind to normal-looking sarcomeres when they are present, this may be due to a lack of siRNA activity in these cells, such that endogenous FHOD3L is still present. In this possible scenario, GS-FH1 may dimerize with endogenous FHOD3L. The authors should demonstrate that GS-FH1 alone can indeed interact with existing actin filaments in vivo. While this has been clearly demonstrated in vitro, given the more complex biochemical environment in vivo where additional unknown binding partners may present, cautions should be made when extrapolating findings from the former to the latter.

The reviewer is concerned about the low protein levels in the GS-FH1 rescue experiments as reflected in the HA fluorescence intensity distributions shown in Fig. 5 Supplement 2A. While the scenario proposed could explain our observations with the GSFH1 rescues it is quite complex. Nor does the scenario preclude the conclusion that the FH1 domain is critical. We agree that the observed sarcomeres are likely to be residual in cells with incomplete RNAi. We now include the image of a cell that is still full of sarcomeres and note that the GH-FH1 is expressed at a relatively high level and striated throughout the cell. We interpret this as evidence that GS-FH1 is stable when suitable binding sites are available. We cannot exclude that there is more GS-FH1 because there was more endogenous FHOD3L with which to heterodimerize. If the GS-FH1 heterodimer were simply poisoning the wild type protein, we do not expect that it would be bound correctly to sarcomeres. If, instead, heterodimers have some activity, it seems far from sufficient to rescue sarcomere formation, suggesting that two functional FH1 domains are critical.

Furthermore, we do not see evidence of correlation between protein levels and rescue at the level present in these cells (addressed below). Unfortunately, the proposed IP to test whether FHOD3L binds actin in vivo would only potentially report on filament side binding (both direct and indirect). It would not address whether the GS-FH1 mutant functions as a nucleator, elongator, bundler and/or capping protein in vivo.

The critical question that we can address is whether the phenotype is due to low protein levels, assuming the protein present is functional, or due to loss of elongation activity by FHOD3L. To address this question, we returned to our data.

First, we plotted the distributions of the intensities of the cells we analyzed further, in addition to the automated readout of all of the cells in the dish (Fig. 4 supplement 1). These cells were selected randomly and, as should be the case, the distributions of their intensities agree well with the original distributions for the three different rescue constructs: FHOD3L, K1193L, and GS-FH1 (Fig. 6 supplement 1). We then asked whether there was any correlation in HA intensities with the sarcomere metrics. As seen in our pilot data, no correlation is evident in any of the three cases across the range of intensities we collected (400 – 2700 a.u.) (old Fig. 6 supplement C,D,E). We now replace the data from pilot experiments with analysis of HA intensities and sarcomere metrics from the data sets included in the paper (new Fig 6. Supplement 1). Again, little to no correlation was observed (the single highest r-squared value is 0.2 and the remaining eight values are less than or equal to 0.08).

To more specifically address the question of whether low HA fluorescence intensity is likely to reflect sufficient protein levels to build sarcomeres we re-examined two data sets from the FHOD3L WT rescue data. We found that, by chance, the first replicate of data from the wild type rescue has a comparable intensity distribution to that of the GSFH1 rescues (580 +/- 261 / cell vs. 548 +/- 105 / cell). In addition, we collected all of the data from cells with intensity levels <720, designed to mimic the distribution of the GS-FH1 cells (Fig. 6 supplement 3). We then compared the sarcomere metrics (sarcomere number, sarcomere length, sarcomere width) between the full data set and the two low intensity subsets:

• Sarcomere number is the only non-normal metric. We therefore used the Mann Whitney U test, which shows no difference between all 3 WT distributions.

• We compared Z-line lengths by one-way ANOVA and Tukey's post hoc tests, again finding no significant difference for all distributions.

• Sarcomere length shows a weakly significant difference (p=0.038) between the whole WT data set and bio rep 1, but no difference between the whole WT data set and the HA<720 group.

Thus, cells expressing wild type FHOD3L at levels comparable to levels detected in GS-FH1 mutant rescues, are fully rescued. Based on these findings we conclude that the expression levels in the GS-FH1 are high enough to rescue the FHOD3 knock down, supporting our conclusion that the defect is due to loss of elongation activity. We have added this analysis and discussion to the revised manuscript.

Recommendations for the authors:

Reviewing Editor Comments:

You will see that the 3 reviewers are very positive about your work and appreciate the elegant combination of biochemical assays and functional tests in cardiomyocytes. We've had a long discussion with them and we all agree that two experiments deserve further effort to make the conclusions of your paper more convincing.

Thank you.

The first experiment is the TIRF elongation assay, where the two biochemist Reviewers remain doubtful that these short events are really due to the presence of a formin at the end of the filament. One of them suggests that two-color imaging with a labeled formin should clearly prove this point.

We agree that the elongation assays can be improved. Given the similarity of processivity of Fhod3L, Fhod3S and Drosophila FhodA (measured by a distinct method), we are inclined to believe them. However, the reviewer raises an excellent point about the accuracy of the measurements given the resolution (and noise) of the data. We are interested in the two-color imaging assay but do not believe it will necessarily simplify the analysis. We suspect that Fhod spends more time at/near the barbed end than is apparent based on elongation rates. The fact that we see repeated events on individual filaments at such low concentrations of FHOD3L (0.1 nM) supports this idea. Otherwise, the likelihood of FHOD3L finding barbed ends so often is really quite low.

We will return to these experiments, using alternate methods, curious to see what else we learn. In the meantime, we conducted more thorough analysis, including controls, and improved visualization of example traces. Data for elongation analysis and kymographs were acquired with Jfilament. We stretched the x-axis (time) in kymographs for FHOD3L-CT (Fig. 2F), FHOD3S-CT (Fig. 2, supplement 2C), FHOD3L-CT K1193L (Fig. 3, supplement 1A), and actin alone (Fig 2G), and highlighted regions of analysis. The slopes for these regions, separated based on intensity, were fit to the data in KaleidaGraph. The fits are offset from the data such that they do not obscure the filaments and corresponding rates are given. The fact that we never see fast dim regions when FHOD3 is not present, as shown in Fig. 2H and that the frequency of dim events is markedly increased (Fig. 2-supplements 1G and 2E) give us confidence that the events are real. We acknowledge in the text that the precise values of the short events may be inaccurate due to the resolution of our experiments. We hope the reviewers are convinced by the improved analysis.

The second experiment is the sarcomere assembly defect phenotype in the GS-FH1 rescue condition. This requires further investigation, as the extremely low level of GS-FH1 signal in transfected cells in Figure 6A may reflect a failure of actin-binding/nucleation in vivo, rather than its inability to elongate F-actin. Although you show that GS-FH1 can bind to sarcomeres when they are present, this may be due to a lack of siRNA activity in these cells, such that endogenous FHOD3L is still present. In this possible scenario, GS-FH1 could dimerize with endogenous FHOD3L.

We agree that the sarcomeres we see are likely to be residual and could reflect some remaining endogenous FHOD3. The reviewers are concerned about the low protein levels in the GSFH1 rescues. First, we do not agree that the levels are “extremely” low. Through careful analysis, we established that 3xHA-FHOD3L intensities between 300 and 3000 a.u./um² were sufficient for full rescue. The mean for the GSFH1 experiments is 533 +/- 93, which is well within this range. Furthermore, we did not observe correlation between sarcomere number, length, or width and HA intensity over the full range collected for wild type FHOD3L or within the GS-FH1 data. We previously showed pilot data but now show correlation analysis for every analyzed cell (Fig. 4 – figure supplement 1 D-F). We conducted this analysis on all of the mutant rescue experiments (Fig. 6-supplement 1). Finally, we identified two subpopulations of the wildtype rescue data. One is all of the cells with HA intensity < 720, which gives a distribution of mean 545 +/- 85. The second set is the first biological replicate of wild type rescue, which has a distribution of mean 560 +/- 160. Again correlation shows little relationship between HA levels and sarcomere metrics. Nevertheless, we show intensity level matched images in Fig 6, as opposed to images reflecting average intensities.

The critical question remains whether the phenotype is due to low protein levels or due to loss of elongation by FHOD3L. Notably, we now show a cell that is full of sarcomeres and has relatively high FHOD3L levels as well, consistent with available binding sites stabilizing mutant protein but not ruling out heterodimerization (Fig. 6 – figure supplement 2C). Others have expressed mutant FHOD3L in a wild type background in mice. They observed poisoning, consistent with heterodimerization. Thus, it is possible that, as suggested, the FHOD3L-GSFH1 detected in sarcomeres is in fact heterodimerized with residual endogenous FHOD3L. In this case, we would still conclude that the protein is not functional enough to rescue, supporting a role for the FH1 domain.

In the future, we plan to perform experiments with compromised, but not inactive, FH1 domains, as we discuss in the paper.

We hope that you will find these comments useful.

Yes, the comments were thoughtful and helped us write a better paper. Thank you.

Reviewer #1 (Recommendations for the authors):

Some experiments should be described and analyzed more carefully. This lack of clarity calls into question the interpretation of some experiments. Overall, this study is not yet as convincing as it should be.

Main recommendations:

(1) Formin elongation phases in the TIRF experiment are not convincing. They are rare and it is difficult to see any significant difference between the control movie without FHOD3L-CT and the movie with FHOD3L-CT. Filaments assembled in the absence of FHOD3L-CT also show some fluorescence inhomogeneity (which is normal), and measurements of formin elongation rates and capping times are not convincing (for example, the kymograph of the control profilin-actin situation in Figure 2F also shows a fast elongation phase on the right).

Please see response above. We conducted more thorough analysis and created improved visualizations. We hope the data are more convincing now.

It is also difficult to understand how an accurate measurement can be made from these noisy kymographs, and the method section should explain that precisely.

This is a valid point. We added details of analysis to the methods section and we discuss the fact that the measurements are at the limit of our resolution in the paper. We rely on the large (~3-fold) difference in elongation, more than specific elongation rates for our interpretation.

One of the problems is that these events are too transient to quantify well with noisy data. I noticed that the formin concentration used in these movies is quite low (0.1 nM FHOD3L-CT). Is there a reason for this? Is it possible to increase the formin concentration to increase the number of formin capping/elongation events and provide more convincing movies?

We acknowledge that the data are noisy. We felt that it was necessary to perform experiments with filaments only tethered at one end, leaving the growing end free. We did so, in part, because when we did experiments with biotinylated actin to anchor the filaments down, we observed pauses in the absence of formin. Ultimately, we compromised, using anchored seeds and a relatively low concentration of NEM-myosin to decrease motion of the actin filaments.

The experiments were performed with such low FHOD3L-CT because it was a potent nucleator in TIRF assays, making data analysis nearly impossible with more formin present. FHOD3S-CT and FHOD3L-CT K1193L behaved somewhat differently between these experiments and we were able to perform them with 1 nM formin.

Not seeing formin at the tip of the filaments is an additional difficulty because we do not know if these pauses occur because formin is stuck to the coverslips (which could very well happen with these sticky proteins) or freely bound at the end of a filament as the text suggests. Is there any argument in favor of one scenario over the other?

This will be an important experiment. As described above, we suspect that Fhod spends more time at/near the barbed end than is apparent based on elongation data. The fact that we see repeated events on individual filaments at such low concentrations of FHOD3L (0.1 nM) supports this idea. Otherwise, the likelihood of FHOD3L finding barbed ends so often is really quite low. In order to address the question about the cause of pauses, we reviewed our data, finding that 38 of 40 bursts were preceded by pauses. We do, however, discuss that we cannot rule out non-specific interactions with the surface.

(2) Pyrene elongation assays in the presence of profilin are actually more convincing to test the elongation ability of formins. However, such an assay is not presented for all mutants. It should be.

While we agree to some extent with this comment, we did not include the pyrene data for all of the mutants because the shapes of the curves were even more complicated than those seen with wild type FHOD3L-CT rendering them uninterpretable.

(3) Some experiments (e.g. in Figure 2E) are performed with yeast profilin, while others (e.g. in Figure 2F) are performed with human profilin. Obviously, both profilins could modulate formin activity differently and the side-by-side interpretation of both experiments is difficult. Could the authors stick to human profilin for all experiments?

We used to always perform pyrene assays with yeast profilin because it was known to be insensitive to pyrene. These data were collected before we realized that the affinity of human profilin for actin is so high that we could probably do everything with this profilin. We have compared the two profilins for other formins, e.g. Delphilin, Capu, and did not observe detectable differences.

Minor recommendations:

(1) The pyrene assays with the light blue colored curve choice are not ideal. I have difficulties seeing some of the curves.

Thank you. We added symbols to a subset of the traces to make them more visible.

(2) In the same curves, I can't understand what the +3.75 and 0.078 numbers mean. Could these results be plotted in a clearer way?

These values are the lowest concentrations in the range tests. They were matching light blue with black outline for visibility. We added symbols and changed the color of the numbering for improved visibility/understanding.

(3) In Figure 2D, is the Kd of I1163A really determined only from 2 experimental data points?

Of course not. We now show the figure with extended axes in Fig. 2 - figure supplement 1C.

(4) In Figure 2C, the shape of the curves suggests that this is not a pure capping assay, but a mix of capping and nucleation. It's not dramatic but could lead to an under-estimation of the capping efficiency.

We agree with the reviewer that the complicated shapes confound interpretation. Our analysis is based on the earliest slopes, in part, for this reason. We added discussion of this complication to the text.

Reviewer #3 (Recommendations for the authors):

Suggestions for additional experiments:

(1) To evaluate whether GS-FH1 alone can indeed interact with existing actin filaments in vivo, the authors may consider performing immunoprecipitation assays with GS-FH1 extracted from rescued NRVMs.

An IP of GS-FH1 from cells could show actin filament side binding but, unfortunately, will not provide any information about filament end binding, which is of much greater interest.

It will be helpful to show phalloidin staining in GS-FH1 rescues in a similar manner as in Figure 6-supplement 1, panel B, and compare that with mock rescue in Figure 4 panel D. It will be essential to prove this prior to concluding that actin elongation activity is essential for sarcomere assembly.

This is an excellent suggestion. We now include images of phalloidin stained cells from both K1193L and GS-FH1 rescues (Fig. 6A’ – supplement 2A,B). We were intrigued to see small actin punctae that were sometimes aligned. We speculate that these could be pre-premyofibrils and suggest that this is further evidence that the GS-FH1 protein is not completely unstable.

(2) Prior to sarcomere assembly, a-actinin is known to form short bundles with actin filaments (I-Z-I complex) without clearly defined periodicity. This semi-ordered state then transforms into the more ordered sarcomeres with periodic spacing. It will be valuable to show the phalloidin staining in addition to the a-actinin IF consistently across all conditions. This may lead to further insights into the defects of sarcomere assembly. Along the same vein, higher magnification images showcasing several sarcomeres will help the readers evaluate these defects.

We agree that there are additional valuable measurements to be made. In order to favor synchronized contraction, we plated the cells at too high a density to reliably identify IZI complexes. We have included some zoomed in images of the phalloidin staining.

Recommendations for improving the writing:

The authors mentioned the interaction between cardiac MyBP-C and FHOD3L as essential for the localization of FHOD3L to the C-line of the sarcomere. Can they discuss whether this interaction is important for the role of FHOD3L in sarcomere assembly? If so, how?

This is a very interesting question that we cannot answer at this time.

Minor corrections to the text and figures:

In the legend of Figure 2-Figure Supplement 1, the labels of (F) and (E) are swapped.

Thank you for catching this.

Author response:

eLife Assessment

This useful study presents Altair-LSFM, a solid and well-documented implementation of a light-sheet fluorescence microscope (LSFM) designed for accessibility and cost reduction. While the approach offers strengths such as the use of custom-machined baseplates and detailed assembly instructions, its overall impact is limited by the lack of live-cell imaging capabilities and the absence of a clear, quantitative comparison to existing LSFM platforms. As such, although technically competent, the broader utility and uptake of this system by the community may be limited.

We thank the reviewers and editors for their thoughtful evaluation of our work and for recognizing the technical strengths of the Altair-LSFM platform, including the custom-machined baseplates and detailed documentation provided to support accessibility and reproducibility. We respectfully disagree, however, with the assessment that the system lacks live-cell imaging capabilities. We are fully confident in the system’s suitability for live-cell applications and will demonstrate this by including representative live-cell imaging data in the revised manuscript, along with detailed instructions for implementing environment control. Moreover, we will expand our discussion to include a broader, more quantitative comparison to existing LSFM platforms—highlighting trade-offs in cost, performance, and accessibility—to better contextualize Altair’s utility and adaptability across diverse research settings.

Public Reviews:

Reviewer #1 (Public review):

Summary:

The article presents the details of the high-resolution light-sheet microscopy system developed by the group. In addition to presenting the technical details of the system, its resolution has been characterized and its functionality demonstrated by visualizing subcellular structures in a biological sample.

Strengths:

(1) The article includes extensive supplementary material that complements the information in the main article.

(2) However, in some sections, the information provided is somewhat superficial.

Our goal was to make the supplemental content as comprehensive and useful as possible. In addition to the materials provided with the manuscript, our intention is for the online documentation (available at thedeanlab.github.io/altair) to serve as a living resource that evolves in response to user feedback. For this reason, we are especially interested in identifying and expanding any sections that are perceived as superficial, and we would greatly appreciate the reviewer’s guidance on which areas would benefit from further elaboration.

Weaknesses:

(1) Although a comparison is made with other light-sheet microscopy systems, the presented system does not represent a significant advance over existing systems. It uses high numerical aperture objectives and Gaussian beams, achieving resolution close to theoretical after deconvolution. The main advantage of the presented system is its ease of construction, thanks to the design of a perforated base plate.

We appreciate the reviewer’s assessment and the opportunity to clarify our intent. Our primary goal was not to introduce new optical functionality beyond that of existing high-performance light-sheet systems, but rather to reduce the barrier to entry for non-specialist labs.

(2) Using similar objectives (Nikon 25x and Thorlabs 20x), the results obtained are similar to those of the LLSM system (using a Gaussian beam without laser modulation). However, the article does not mention the difficulties of mounting the sample in the implemented configuration.

We agree that there are practical challenges associated with handling 5 mm diameter coverslips. However, the Nikon 25x can readily be replaced by a Zeiss W Plan-Apochromat 20x/1.0 objective, which eliminates the need for the 5 mm coverslip[1]. In the revised manuscript, we will more explicitly detail the practical challenges in handling a 5 mm coverslip and mention the alternative detection objective.

(3) The authors present a low-cost, open-source system. Although they provide open source code for the software (navigate), the use of proprietary electronics (ASI, NI, etc.) makes the system relatively expensive. Its low cost is not justified.

We understand the reviewer’s concern regarding the use of proprietary control hardware such as the ASI Tiger Controller and NI data acquisition cards. While lower-cost alternatives for analog and digital control (e.g., microcontroller-based systems) do exist, our choice was intentional. By relying on a unified and professionally supported platform, we minimize the complexity of sourcing, configuring, and integrating components from disparate vendors—each of which would otherwise demand specialized technical expertise. Moreover, in future releases, we aim to further streamline the system by eliminating the need for the NI card, consolidating all optoelectronic control through the ASI Tiger Controller. This approach allows users to purchase a fully assembled and pre-configured system that can be operational with minimal effort.

It is worth noting that the ASI components are not the primary cost driver. The full set—including XYZ and focusing stages, a filter wheel, a tube lens, the Tiger Controller, and basic optomechanical adapters—costs approximately $27,000, or ~18% of the total system cost. Additional cost reductions are possible. For example, replacing the motorized sample positioning and focusing stages with manual alternatives could reduce the cost by ~$12,000. However, this would eliminate key functionality such as autofocusing, 3D tiling, and multi-position acquisition. Open-source mechanical platforms such as OpenFlexure could in principle be adapted, but they would require custom assembly and would need to be integrated into our control software. Similarly, the filter wheel could be omitted in favor of a multi-band emission filter, reducing the cost by ~$5,000. However, this comes at the expense of increased spectral crosstalk, often necessitating spectral unmixing. An industrial CMOS camera—such as the Ximea MU196CR-ON, recently demonstrated in a Direct View Oblique Plane Microscopy configuration[2]—could substitute for the sCMOS cameras typically used in high-end imaging. However, these industrial sensors often exhibit higher noise floors and lower dynamic range, limiting sensitivity for low-signal imaging applications.

While a $150,000 system represents a significant investment, we consider it relatively cost-effective in the context of advanced light-sheet microscopy. For comparison, commercially available systems with similar optical performance—such as LLSM systems from 3i or Zeiss—are several-fold more expensive.

(4) The fibroblast images provided are of exceptional quality. However, these are fixed samples. The system lacks the necessary elements for monitoring cells in vivo, such as temperature or pH control.

We thank the reviewer for their positive comment regarding the quality of our fibroblast images. As noted, the current manuscript focuses on the optical design and performance characterization of the system, using fixed specimens to validate resolution and imaging stability. We acknowledge the importance of environmental control for live-cell imaging. Temperature regulation is routinely implemented in our lab using flexible adhesive heating elements paired with a power supply and PID controller. For pH stabilization in systems that lack a 5% CO₂ atmosphere, we typically supplement the imaging medium with 10–25 mM HEPES buffer. In the revised manuscript, we will introduce a modified sample chamber capable of maintaining user-specified temperatures, along with detailed assembly instructions. We will also include representative live-cell imaging data to demonstrate the feasibility of in vitro imaging using this system.

Reviewer #2 (Public review):

Summary:

The authors present Altair-LSFM (Light Sheet Fluorescence Microscope), a high-resolution, open-source microscope, that is relatively easy to align and construct and achieves sub-cellular resolution. The authors developed this microscope to fill a perceived need that current open-source systems are primarily designed for large specimens and lack sub-cellular resolution or are difficult to construct and align, and are not stable. While commercial alternatives exist that offer sub-cellular resolution, they are expensive. The authors' manuscript centers around comparisons to the highly successful lattice light-sheet microscope, including the choice of detection and excitation objectives. The authors thus claim that there remains a critical need for high-resolution, economical, and easy-to-implement LSFM systems.

Strengths:

The authors succeed in their goals of implementing a relatively low-cost (~ USD 150K) open-source microscope that is easy to align. The ease of alignment rests on using custom-designed baseplates with dowel pins for precise positioning of optics based on computer analysis of opto-mechanical tolerances, as well as the optical path design. They simplify the excitation optics over Lattice light-sheet microscopes by using a Gaussian beam for illumination while maintaining lateral and axial resolutions of 235 and 350 nm across a 260-um field of view after deconvolution. In doing so they rest on foundational principles of optical microscopy that what matters for lateral resolution is the numerical aperture of the detection objective and proper sampling of the image field on to the detection, and the axial resolution depends on the thickness of the light-sheet when it is thinner than the depth of field of the detection objective. This concept has unfortunately not been completely clear to users of high-resolution light-sheet microscopes and is thus a valuable demonstration. The microscope is controlled by an open-source software, Navigate, developed by the authors, and it is thus foreseeable that different versions of this system could be implemented depending on experimental needs while maintaining easy alignment and low cost. They demonstrate system performance successfully by characterizing their sheet, point-spread function, and visualization of sub-cellular structures in mammalian cells, including microtubules, actin filaments, nuclei, and the Golgi apparatus.

We thank the reviewer for their thoughtful summary of our work. We are pleased that the foundational optical principles, design rationale, and emphasis on accessibility came through clearly. We agree that the approach used to construct the microscope is highly modular, and we anticipate that these design principles will serve as the basis for additional system variants tailored to specific biological samples and experimental contexts. To support this, we provide all Zemax simulations and CAD files openly on our GitHub repository, enabling advanced users to build upon our design and create new functional variants of the Altair system.

Weaknesses:

There is a fixation on comparison to the first-generation lattice light-sheet microscope, which has evolved significantly since then:

(1) The authors claim that commercial lattice light-sheet microscopes (LLSM) are "complex, expensive, and alignment intensive", I believe this sentence applies to the open-source version of LLSM, which was made available for wide dissemination. Since then, a commercial solution has been provided by 3i, which is now being used in multiple cores and labs but does require routine alignments. However, Zeiss has also released a commercial turn-key system, which, while expensive, is stable, and the complexity does not interfere with the experience of the user. Though in general, statements on ease of use and stability might be considered anecdotal and may not belong in a scientific article, unreferenced or without data.

The referee is correct that our comparisons reference the original LLSM design, which was simultaneously disseminated as an open-source platform and commercialized by 3i. While we acknowledge that newer variants of LLSM have been developed—including systems incorporating adaptive optics[3] and the MOSAIC platform (which remains unpublished)—the original implementation remains the most widely described and cited in the literature. It is therefore the most appropriate point of comparison for contextualizing Altair’s performance, complexity, and accessibility. Importantly, this version of LLSM is far from obsolete; it continues to be one of the most commonly used imaging systems at Janelia Research Campus’s Advanced Imaging Center.

We acknowledge that more recent commercial implementation by Zeiss has addressed several of the practical limitations associated with the original design. In particular, we agree that the Zeiss Lattice Lightsheet 7 system, which integrates a meniscus lens to facilitate oblique imaging through a coverslip, offers a user-friendly experience—albeit with a modest tradeoff in resolution (reported deskewed resolution: 330 nm × 330 nm × 500–1000 nm).

While we recognize that statements on usability and stability can be subjective, one objective proxy for system complexity is the number of optical elements that require precise alignment during assembly. The original LLSM setup includes approximately 29 optical components that must each be carefully positioned laterally, angularly, and coaxially along the optical path. In contrast, the first-generation Altair system contains only 9 such elements. By this metric, Altair is considerably simpler to assemble and align, supporting our overarching goal of making high-resolution light-sheet imaging more accessible to non-specialist laboratories. In the revised manuscript, we will clarify the scope of our comparison and provide more precise language about what we mean by complexity (e.g., number of optical elements needed to align).

(2) One of the major limitations of the first generation LLSM was the use of a 5 mm coverslip, which was a hinderance for many users. However, the Zeiss system elegantly solves this problem, and so does Oblique Plane Microscopy (OPM), while the Altair-LSFM retains this feature, which may dissuade widespread adoption. This limitation and how it may be overcome in future iterations is not discussed.

We agree that the use of 5 mm diameter coverslips, while enabling high-NA imaging in the current Altair-LSFM configuration, may serve as an inconvenience for many users. We will discuss this more explicitly in the revised manuscript. Specifically, we note that changing the detection objective is sufficient to eliminate the need for a 5 mm coverslip. For example, as demonstrated in Moore et al., Lab Chip 2021, pairing the Zeiss W Plan-Apochromat 20x/1.0 objective with the Thorlabs TL20X-MPL allows imaging beyond the physical surfaces of both objectives, removing the constraint imposed by small-format coverslips[1]. In the revised manuscript, we will propose this modification as a straightforward path for increasing compatibility with more conventional sample mounting formats.

(3) Further, on the point of sample flexibility, all generations of the LLSM, and by the nature of its design, the OPM, can accommodate live-cell imaging with temperature, gas, and humidity control. It is unclear how this would be implemented with the current sample chamber. This limitation would severely limit use cases for cell biologists, for which this microscope is designed. There is no discussion on this limitation or how it may be overcome in future iterations.

We appreciate the reviewer’s emphasis on the importance of environmental control for live-cell imaging applications. It is worth noting that the original LLSM design, including the system commercialized by 3i, provided temperature control only, without integrated gas or humidity regulation. Despite this, it has been successfully used by a wide range of scientists to generate important biological insights.

We agree that both OPM and the Zeiss implementation of LLSM offer clear advantages in terms of environmental control, as we previously discussed in detail in Sapoznik et al., eLife, 2020[4]. However, assembly of high numerical aperture OPM systems is highly technical, and no open-source variant of OPM delivers sub-cellular scale resolution yet.

(4) The authors' comparison to LLSM is constrained to the "square" lattice, which, as they point out, is the most used optical lattice (though this also might be considered anecdotal). The LLSM original design, however, goes far beyond the square lattice, including hexagonal lattices, the ability to do structured illumination, and greater flexibility in general in terms of light-sheet tuning for different experimental needs, as well as not being limited to just sample scanning. Thus, the Alstair-LSFM cannot compare to the original LLSM in terms of versatility, even if comparisons to the resolution provided by the square lattice are fair.

We thank the reviewer for this comment. It is true that our discussion focused primarily on the square lattice implementation of LLSM. While this could be viewed as a subset of the system’s broader capabilities, we chose this focus intentionally, as the square lattice remains by far the most commonly used variant in practice. Even in the original LLSM publication, 16 out of 20 figure subpanels utilized the square lattice, with only one panel each representing the hexagonal lattice in SIM mode, a standard Bessel beam in incoherent SIM mode, a hex lattice in dithered mode, and a single Bessel in dithered mode. This usage pattern largely reflects the operational simplicity of the square lattice: it minimizes sidelobe growth and enables more straightforward alignment and data processing compared to hexagonal or structured illumination modes.

In 2019, we performed an exhaustive accounting of published illumination modes in LLSM and found that the SIM mode had only been used in two additional peer-reviewed publications at that time. We will consider updating this table in the revised manuscript and will expand our discussion to acknowledge the broader flexibility of the LLSM platform—including its capacity for structured illumination and alternative light-sheet geometries. However, we will also emphasize that, despite these advanced capabilities, the square lattice remains the dominant mode used by the community and therefore serves as a fair and practical benchmark for comparison.

(5) There is no demonstration of the system's live-imaging capabilities or temporal resolution, which is the main advantage of existing light-sheet systems.

In the revised manuscript, we will include a demonstration of live-cell imaging to directly validate the system’s suitability for dynamic biological applications. We will also characterize the temporal resolution of the system. As a sample-scanning microscope, the imaging speed is primarily limited by the performance of the Z-piezo stage. For simplicity and reduced optoelectronic complexity, we currently power the piezo through the ASI Tiger Controller. We will expand the supplementary material to describe the design criteria behind this choice, including potential trade-offs, and provide data quantifying the achievable volume rates under typical operating conditions.

While the microscope is well designed and completely open source, it will require experience with optics, electronics, and microscopy to implement and align properly. Experience with custom machining or soliciting a machine shop is also necessary. Thus, in my opinion, it is unlikely to be implemented by a lab that has zero prior experience with custom optics or can hire someone who does. Altair-LSFM may not be as easily adaptable or implementable as the authors describe or perceive in any lab that is interested, even if they can afford it. The authors indicate they will offer "workshops," but this does not necessarily remove the barrier to entry or lower it, perhaps as significantly as the authors describe.

We appreciate the reviewer’s perspective and agree that building any high-performance custom microscope—Altair-LSFM included—requires a baseline familiarity with optics and instrumentation. Our goal is not to eliminate this requirement entirely, but to significantly reduce the technical and logistical barriers that typically accompany custom light-sheet microscope construction.

Importantly, no machining experience or in-house fabrication capabilities are required—users can simply submit provided design files and specifications directly to the vendor. We will make this process as straightforward as possible by supplying detailed instructions, recommended materials, and vendor-ready files. Additionally, we draw encouragement from the success of related efforts such as mesoSPIM, which has seen over 30 successful implementations worldwide using a similar model of exhaustive online documentation, open-source control software, and community support through user meetings and workshops.

We recognize that documentation alone is not always sufficient, and we are committed to further lowering barriers to adoption. To this end, we are actively working with commercial vendors to streamline procurement and reduce the logistical burden on end users. Additionally, Altair-LSFM is supported by a Biomedical Technology Development and Dissemination (BTDD) grant, which provides dedicated resources for hosting workshops, offering real-time community support, and generating supplementary materials such as narrated video tutorials. We will expand our discussion in the revised manuscript to better acknowledge these implementation challenges and outline our ongoing strategies for supporting a broad and diverse user base.

There is a claim that this design is easily adaptable. However, the requirement of custom-machined baseplates and in silico optimization of the optical path basically means that each new instrument is a new design, even if the Navigate software can be used. It is unclear how Altair-LSFM demonstrates a modular design that reduces times from conception to optimization compared to previous implementations.

We appreciate the reviewer’s comment and agree that our language regarding adaptability may have been too strong. It was not our intention to suggest that the system can be easily modified without prior experience. Meaningful adaptations of the optical or mechanical design would require users to have expertise in optical layout, optomechanical design, and alignment.

That said, for labs with sufficient expertise, we aim to facilitate such modifications by providing comprehensive resources—including detailed Zemax simulations, CAD models, and alignment documentation. These materials are intended to reduce the development burden for those seeking to customize the platform for specific experimental needs.

In the revised manuscript, we will clarify this point and explicitly state in the discussion what technical expertise is required to modify the system. We will also revise our language around adaptability to better reflect the intended audience and realistic scope of customization.

Reviewer #3 (Public review):

Summary:

This manuscript introduces a high-resolution, open-source light-sheet fluorescence microscope optimized for sub-cellular imaging.

The system is designed for ease of assembly and use, incorporating a custom-machined baseplate and in silico optimized optical paths to ensure robust alignment and performance. The authors demonstrate lateral and axial resolutions of ~235 nm and ~350 nm after deconvolution, enabling imaging of sub-diffraction structures in mammalian cells.

The important feature of the microscope is the clever and elegant adaptation of simple gaussian beams, smart beam shaping, galvo pivoting and high NA objectives to ensure a uniform thin light-sheet of around 400 nm in thickness, over a 266 micron wide Field of view, pushing the axial resolution of the system beyond the regular diffraction limited-based tradeoffs of light-sheet fluorescence microscopy.

Compelling validation using fluorescent beads and multicolor cellular imaging highlights the system's performance and accessibility. Moreover, a very extensive and comprehensive manual of operation is provided in the form of supplementary materials. This provides a DIY blueprint for researchers who want to implement such a system.

Strengths:

(1) Strong and accessible technical innovation: With an elegant combination of beam shaping and optical modelling, the authors provide a high-resolution light-sheet system that overcomes the classical light-sheet tradeoff limit of a thin light-sheet and a small field of view. In addition, the integration of in silico modelling with a custom-machined baseplate is very practical and allows for ease of alignment procedures. Combining these features with the solid and super-extensive guide provided in the supplementary information, this provides a protocol for replicating the microscope in any other lab.

(2) Impeccable optical performance and ease of mounting of samples: The system takes advantage of the same sample-holding method seen already in other implementations, but reduces the optical complexity. At the same time, the authors claim to achieve similar lateral and axial resolution to Lattice-light-sheet microscopy (although without a direct comparison (see below in the "weaknesses" section). The optical characterization of the system is comprehensive and well-detailed. Additionally, the authors validate the system imaging sub-cellular structures in mammalian cells.

(3) Transparency and comprehensiveness of documentation and resources: A very detailed protocol provides detailed documentation about the setup, the optical modeling, and the total cost.

Weaknesses:

(1) Limited quantitative comparisons: Although some qualitative comparison with previously published systems (diSPIM, lattice light-sheet) is provided throughout the manuscript, some side-by-side comparison would be of great benefit for the manuscript, even in the form of a theoretical simulation. While having a direct imaging comparison would be ideal, it's understandable that this goes beyond the interest of the paper; however, a table referencing image quality parameters (taken from the literature), such as signal-to-noise ratio, light-sheet thickness, and resolutions, would really enhance the features of the setup presented. Moreover, based also on the necessity for optical simplification, an additional comment on the importance/difference of dual objective/single objective light-sheet systems could really benefit the discussion.

In the revised manuscript, we will expand our discussion to include a broader range of light-sheet microscope designs and imaging modes, including both single- and dual-objective configurations. We agree that highlighting the trade-offs between these approaches—such as working distance, sample geometry constraints, and alignment complexity—will enhance the overall context and utility of the manuscript.

To further aid comparison, we will include a summary table referencing key image quality parameters such as lateral and axial resolution, and illumination beam NA for Altair-LSFM. Where available, we will reference values from published work—such as the axial resolution reported in Valm et al. (Nature, 2017)—to provide a clearer benchmark. Because such comparisons can be technically nuanced, especially when comparing across systems with different geometries and sample mounting constraints, we will also include a supplementary note outlining the assumptions and limitations of these comparisons.

(2) Limitation to a fixed sample: In the manuscript, there is no mention of incubation temperature, CO₂ regulation, Humidity control, or possible integration of commercial environmental control systems. This is a major limitation for an imaging technique that owes its popularity to fast, volumetric, live-cell imaging of biological samples.

We thank the reviewer for highlighting this important consideration. In the revised manuscript, we will provide a detailed description of how temperature control can be implemented using flexible adhesive heating elements, a power supply, and a PID controller. Step-by-step assembly instructions and recommended components will be included to facilitate adoption by users interested in live-cell imaging. We also note that most light-sheet microscopy systems capable of sub-cellular resolution—including the original LLSM design, diSPIM, and ASLM—typically do not incorporate integrated CO₂ or humidity control. These systems often rely on HEPES-buffered media to maintain pH stability, which is generally sufficient for short- to intermediate-term imaging. While full environmental control may be necessary for extended time-lapse studies, it is not a prerequisite for high-resolution volumetric imaging in many applications. Nonetheless, we will include a discussion of the challenges associated with adding CO₂ and humidity control to open or semi-enclosed architectures like Altair-LSFM, and outline potential future paths for integration with commercial incubation systems.

(3) System cost and data storage cost: While the system presented has the advantage of being open-source, it remains relatively expensive (considering the 150k without laser source and optical table, for example). The manuscript could benefit from a more direct comparison of the performance/cost ratio of existing systems, considering academic settings with budgets that most of the time would not allow for expensive architectures. Moreover, it would also be beneficial to discuss the adaptability of the system, in case a 30k objective could not be feasible. Will this system work with different optics (with the obvious limitations coming with the lower NA objective)? This could be an interesting point of discussion. Adaptability of the system in case of lower budgets or more cost-effective choices, depending on the needs.

We thank the reviewer for raising this important point. First, we would like to clarify that the quoted $150k cost estimate includes the optical table and laser source. We apologize for any confusion and will communicate this more effectively in the revised manuscript.

We agree that adaptability is a key concern, especially in academic settings with limited budgets. The detection path can be readily altered depending on experimental needs and cost constraints. For example, in our discussion of alternatives to the 5 mm coverslip geometry, we will describe how switching to a Zeiss W Plan-Apochromat 20x/1.0 in combination with a compatible excitation objective allows high-resolution imaging while accommodating more conventional sample formats. We will expand this to include cost-effective alternatives as well.

We will also expand our discussion on cost-reduction strategies and the associated trade-offs. These include replacing motorized stages with manual ones, omitting the filter wheel in favor of a multi-band emission filter, or using industrial-grade cameras in place of scientific CMOS detectors. While each change entails some loss in functionality or sensitivity, such modifications allow users to tailor the system to their specific budget and application.

Finally, we recognize the challenge in communicating exact costs of commercial systems due to variability in configuration and pricing. Nonetheless, we will include approximate figures where possible and note that comparable commercial systems—such as LLSM platforms from 3i and Zeiss—are several-fold more expensive than the system presented here.

Last, not much is said about the need for data storage. Light-sheet microscopy's bottleneck is the creation of increasingly large datasets, and it could be beneficial to discuss more about the storage needs and the quantity of data generated.

Data storage is indeed a critical consideration in light-sheet microscopy. In the revised manuscript, we will provide a note outlining typical volume dimensions for live-cell imaging experiments along with the associated data overhead. This will include estimates for voxel counts, bit depth, time-lapse acquisitions, and multi-channel datasets to help users anticipate storage needs. We will also briefly discuss strategies for managing large datasets, file types and compression formats.

Conclusion:

Altair-LSFM represents a well-engineered and accessible light-sheet system that addresses a longstanding need for high-resolution, reproducible, and affordable sub-cellular light-sheet imaging. While some aspects-comparative benchmarking and validation, limitation for fixed samples-would benefit from further development, the manuscript makes a compelling case for Altair-LSFM as a valuable contribution to the open microscopy scientific community.

References

(1) Moore, R. P. et al. A multi-functional microfluidic device compatible with widefield and light sheet microscopy. Lab Chip 22, 136-147 (2021). https://doi.org/10.1039/d1lc00600b

(2) Lamb, J. R., Mestre, M. C., Lancaster, M. & Manton, J. D. Direct-view oblique plane microscopy. Optica 12, 469-472 (2025). https://doi.org/10.1364/OPTICA.558420

(3) Liu, T. L. et al. Observing the cell in its native state: Imaging subcellular dynamics in multicellular organisms. Science 360 (2018). https://doi.org/10.1126/science.aaq1392

(4) Sapoznik, E. et al. A versatile oblique plane microscope for large-scale and high-resolution imaging of subcellular dynamics. eLife 9 (2020). https://doi.org/10.7554/eLife.57681

(5) Huisken, J. & Stainier, D. Y. Even fluorescence excitation by multidirectional selective plane illumination microscopy (mSPIM). Opt Lett 32, 2608-2610 (2007). https://doi.org/10.1364/ol.32.002608

(6) Ricci, P. et al. Removing striping artifacts in light-sheet fluorescence microscopy: a review. Prog Biophys Mol Biol 168, 52-65 (2022). https://doi.org/10.1016/j.pbiomolbio.2021.07.003

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Mollá-Albaladejo et al. investigate the neurons downstream of GR64f and Gr66a, called G2Ns. They identify downstream neurons using trans-Tango labeling with RFP and then perform bulk RNA-seq on the RFP-sorted cells. Gene expression is up- or downregulated between the cell populations and between fed and starved states. They specifically identify Leukocinin as a neuropeptide that is upregulated in starved Gr66a cells. Leucokinin cells, identified by a GAL4 line indeed show higher expression when starved, especially in the SEZ. Furthermore, Leucokinin cells colocalize with the transTango signal from downstream neurons of both GRs. This connection is confirmed with GRASP. According to EM data, Leucokinin cells in the SEZ receive a lot of input and connect to many downstream neurons. In behavior experiments performed with flies lacking Leucokinin neurons, flies show reduced responsiveness to sugar and bitter mixtures when starved. The authors suggest that Leucokinin neurons integrate bitter and sugar tastes and that their output is modified by a hunger state.

Strengths:

The authors use a multitude of tools to identify SELK neurons downstream of taste sensory neurons and as starvation-sensitive cells. This study provides an example of how combining genetic labeling, RNA-seq, and EM analysis can be combined to investigate neural circuits.

Weaknesses:

The authors do not show a functional connection between sensory neurons and SELK neurons. Additionally, data from RNA seq, anatomical studies, and EM analysis are sometimes contradictory in terms of connectivity. GRASP signal is not foolproof that cells are synaptically connected.

We appreciate the reviewer’s comments. Unfortunately, we have not successfully demonstrated a functional response of SELK neurons using in vivo calcium imaging with UAS-GCaMP7 (we tried f, m, and s versions), primarily due to challenges in obtaining stable signals. We stimulated GRNs using sucrose, caffeine, or a mixture of both, and maybe even if the concentrations were high, they were not enough to induce a response.

Regarding GRASP, we acknowledge its limitations as a standalone technique for establishing genuine synaptic connections between neurons, as some signals may reflect false positives resulting from the mere proximity of the candidate neurons. To strengthen our findings, we complemented these results by demonstrating the positive colocalization of the Leucokinin antibody signal over the Gr66aGal4>trans-TANGO and Gr64f-Gal4>trans-TANGO (Figure 4), confirming that Leucokinin neurons are indeed postsynaptic to both sweet and bitter GRNs. Moreover, we incorporated BacTrace data to highlight the direct connectivity between sweet and bitter GRNs (now Figure 5E).

In the revised manuscript, we have introduced the active-GRASP technique (Macpherson et al., 2015). In this version of GRASP, the presynaptic half of GFP (GFP 1-10) is fused to synaptobrevin, which becomes accessible in the membrane of the presynaptic neuron within the synaptic cleft upon presynaptic stimulation (in our case, by stimulating with sucrose sweet Gr64f^GRNs and with caffeine the bitter Gr66a^GRNs). Utilizing this technique, we successfully demonstrated (see new Figure 5B and 5D) that when presented with water, no signal was detected in the Gr66a-LexA, Lk-Gal4 > active-GRASP, or Gr64f-LexA, Lk-Gal4 > active-GRASP transgene flies. However, in the presence of caffeine, Gr66aLexA, Lk-Gal4 > active-GRASP transgene flies exhibited a clear signal in the SEZ, and similarly, sucrose presentation to Gr64f-LexA, Lk-Gal4 > active-GRASP transgene flies yielded a detectable signal. The results obtained from active-GRASP provide additional evidence supporting the connectivity between SELK neurons and both Gr64f^GRNs and Gr66a^GRNs, further indicating the functional connectivity of the GRNs and SELK neurons.

The authors describe a behavioral phenotype when flies are starved, however, they do not use a specific driver for the described cell type, thus they should also tone down their claims.

We agree with the reviewer that the Lk-Gal4 driver line used labels SELK, LHLK, and ABLK neurons. The behavior examined in this paper, the Proboscis Extension Response (PER), measures the initiation of feeding. Although the neural circuit involved in this behavior is primarily confined to the SEZ where SELK neurons are located, we cannot rule out the possibility that other Lk neurons may also play a role in the process. To restrict expression of the Tetanus Toxin, we have utilized the tsh-Gal80 (Clyne et al., 2008) transgene in combination with the Lk-Gal4>UAS-TNT and Lk-Gal4>UAS-TNT^imp constructs to prevent the expression of the Tetanus Toxin in ABLK neurons, thereby restricting its expression to the SELK and LHLK neurons in the central brain. The new results (Sup Figure 7A) indicate that ABLK neurons do not play a role in integrating sweet and bitter information. However, we acknowledge the reviewer's point that we are still silencing LHLK neurons, so we have adjusted our claims to align more closely with our data

Generally, the authors do not provide a big advancement to the field and some of the results are contradictory with previous publications.

We believe our work does not contradict previous findings, nor does it invalidate the role of ABLK neurons in water homeostasis or the role of LHLK neurons in regulating sleep via starvation. We provide additional information on the possible role of SELK neurons in integrating gustatory information. The location of SELK neurons in the SEZ suggests that they may play a role in feeding behavior, and we have demonstrated that these neurons are indeed involved in integrating gustatory information to influence feeding decisions. We consider we have contributed by highlighting a new role for the Leucokinin neuropeptide in feeding behavior.

Reviewer #2 (Public review):

Summary:

A core task of the brain is processing sensory cues from the environment. The neural mechanisms of how sensory information is transmitted from peripheral sense organs to subsequent being processing in defined brain centers remain an important topic in neuroscience. The taste system hereby assesses the palatability of food by evaluating the chemical composition and nutrient content while integrating the current need for energy by assessing the satiation level of the organism. The current manuscript provides insights into the early circuits of gustatory coding using the fruit fly as a model. By combining trans-tango and FACS- based bulk RNAseq to assess the target neurons of sweet sensing (using Gr64fGal4) and bitter sensing (using Gr66a-Gal4) in a first set of experiments the authors investigate genes that are differentially expressed or co-expressed in normal and starved conditions. With a focus on neuropeptides and neurotransmitters, different expressions in the different conditions were assessed resulting in the identification of Leucokinin as a potentially interesting gene. The notion is further supported by RNAseq of Lk- Gal4>mCD8:GFP sorted cells and immunostainings. GRASP and BacTrace experiments further support that the two Lk- expressing cells in the SEZ should indeed be postsynaptic to both types of sensories. Using EM-based connectomics data (based on a previous publication by Engert et al.), the authors also look for downstream targets of the bitter versus sweet gustatory neurons to identify the Lk-neurons. Based on the morphology they identify candidates and further depict the potential downstream neurons in the connectome, which appears largely in agreement with GRASP experiments. Finally silencing the Lk- neurons shows an increased PER response in starved flies (when combined with bitter compounds) as well as increased feeding neurons shows an increased PER response in starved flies (when combined with bitter compounds) as well as increased feeding in a FlyPad assay. Strengths:

Overall this is an intriguing manuscript, which provides insight into the organization of 2nd order gustatory neurons. It specifically provides strong evidence for the Lk-neurons as a target of sweet and bitter GRNs and provides evidence for their role in regulating sweet vs bitter-based behavioral responses. Particularly the integration of different techniques and datasets in an elegant fashion is a strong side of the manuscript. Moreover to put the known LK-neurons into the context of 2nd order gustatory signalling is strengthening the knowledge about this pathway.

Weaknesses:

I do not see any major weakness in the current manuscript. Novelty is to some degree lessened by the fact, that the RNAseq approach did not identify new neurons but rather put the known LK-neurons as major findings. Similarly, the final behavioral section is not very deep and to some degree corroborates the previous publication by the Keene and Nässel labs - that said, the model they propose is indeed novel (but lacks depth in analyses; e.g. there is no physiology that would support the modulation of Lk neurons by either type of GRN). The connectomic section appears a bit out of place and after reading it it's not really clear what one should make of the potential downstream neurons (particularly since the Lk-receptor expression has been previously analyzed); here it might have been interesting to address if/how Lk-neurons may signal directly via a classical neurotransmitter (an information that might be found easily in the adult brain single-cell data).

We thank the reviewer for the comment. Indeed, we attempted in vivo Ca imaging but were unsuccessful. We have rewritten the connectomic section to better integrate it with the rest of the text and have reanalyzed the data obtained. We considered gathering data from the single-cell adult dataset, but this dataset includes the entire adult fly brain, encompassing SELK and LHLK neurons, making it impossible to differentiate between the two types of Lk neurons. Any further analysis will require transcriptomic analysis of SELK via scRNAseq under the different metabolic conditions tested in this study work.

Reviewer #3 (Public review):

Summary:

To make feeding decisions, animals need to process three types of information: positive cues like sweetness, negative cues like bitterness, and internal states such as hunger or satiety. This study aims to identify where the information is integrated into the fruit fly brain. The authors applied RNA sequencing on second-order gustatory neurons responsible for sweet and bitter processing, under fed and starved conditions. The sequencing data reveal significant changes in gene expression across sweet vs. bitter pathways and fed vs. starved states. The authors focus on the neuropeptide Leucokinin (Lk), whose expression is dependent on the starvation state. They identify a pair of neurons, named SELK neurons, which express Lk and receive direct input from both sweet and bitter gustatory neurons. These SELK neurons are ideal candidates to integrate gustatory and internal state information. Behavioral experiments show that blocking these neurons in starved flies alters their tolerance to bitter substances during feeding.

Strengths:

(1) The study employs a well-designed approach, targeting specific neuronal populations, which is more efficient and precise compared to traditional large-scale genetic screening methods.

(2) The RNAseq results provide valuable data that can be utilized in future studies to explore other molecules beyond Lk.

(3) The identification of SELK neurons offers a promising avenue for future research into how these neurons integrate conflicting gustatory signals and internal state information.

Weaknesses:

(1) Unfortunately, due to technical challenges, the authors were unable to directly image the functional activity of SELK neurons.

(2) In the behavioral experiments, tetanus toxin was used to block SELK neurons. Since these neurons may release multiple neurotransmitters or neuropeptides, the results do not specifically demonstrate that Leucokinin (Lk) is the critical factor, as suggested in Figure 8. To address this, I recommend using RNAi to inhibit Lk expression in SELK neurons and comparing the outcomes to wild-type controls via the PER assay.

We appreciate the author's comments and suggestions. As noted, Tetanus Toxin silences the neuron’s activity, affecting the functioning of various neurotransmitters and neuropeptides released by the targeted neuron. In response to the reviewer's recommendation, we employed an RNAi line specifically designed to silence Leucokinin production in Lk-expressing neurons.

The results presented in Supplementary Figure 7B demonstrate that knocking down Leucokinin in Lk neurons significantly reduces the flies' tolerance to caffeine in sweet food.

It is crucial to highlight that the sucrose concentration used in Figure 7C was 50mM, whereas in Supplementary Figure 7B, it was increased to 100mM. This adjustment was necessary because the Lk-Gal4, UAS-RNAi, and Lk-Gal4>UAS-RNAi transgenic lines exhibited reduced sensitivity to sucrose compared to the Lk-Gal4>UAS-TNT or Lk-Gal4>UAS-TNT^imp lines. We aimed to establish a sucrose concentration that would elicit a 50% Proboscis Extension Response (PER) without adding any other compound, thereby allowing us to evaluate the additional effect of caffeine in the food.

However, according to the data derived from the connectome, SELK neurons might be cholinergic, and this neurotransmitter might be involved in controlling also the behavior of the flies.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

To get more evidence for connections between sensory cells and SELK neurons, could the authors also analyze a second available EM data set? Would setting a different threshold (>5 synapses) reveal connections to both sensories? Comparisons between SELK in- and outputs from EM data and Tango labeling also seem to differ quite a lot based on provided images - can the authors count cell bodies in the stainings? Further proof would be to provide functional imaging data that shows that SELK neurons respond to sugar and bitter compounds.

In this study, we utilized the recently published EM dataset for the Drosophila central brain connectome (Dorkenwald et al., 2024; Flywire.ai). Changing the number of synapses affects the counts of pre- and postsynaptic neurons. We set a threshold of more than five synapses, as recommended by Flywire, to avoid false positives (Dorkenwald et al., 2024). This threshold has been widely used in recent papers (Engert et al., 2022; Shiu et al., 2022; Walker et al., 2025).

The neuron counts in the connectomic data differ from those in the trans- and retro-TANGO experiments. In our initial trans-TANGO experiment, which labeled postsynaptic neurons in the Gr64fGal4 and Gr66a-Gal4 transgenic lines, we counted the labeled neurons (see Supplementary Figure 1C) and observed considerable variability between different brains. Due to anticipated variability, we did not count the labeled neurons from trans-TANGO and retro-TANGO techniques in the Leucokinin neurons. Furthermore, neither technique labels all postsynaptic or presynaptic neurons, respectively. A recent study on the retro-TANGO technique (Sorkac et al., 2023) found a minimum threshold: the presynaptic neuron must form a certain number of synapses with the neuron of interest to be adequately labeled. According to this paper, the established threshold is 17 synapses. It is likely that the trans-TANGO technique also has a threshold relating to the number of labeled neurons, contingent on the synapse count. This would explain the discrepancy between the two results.

Unfortunately, we have not been able to provide functional data pointing to the activation of SELK neurons by sucrose or caffeine. However, our active-GRASP data indicates that the connectivity between Gr64f^GRNs and Gr66a^GRNs with SELK neurons is present and functional.

How many Leucokinin-positive cells are in the SEZ? Does the RNA-seq data provide further information about the SELK neurons? Potential receptor candidates for how they integrate hunger signals? AMPKa was described to be required in LHLK neurons.

There are two SELK neurons in the SEZ. Due to the nature of our bulk RNA sequencing (RNAseq), we cannot link any additional gene expressions detected in our transcriptomic analysis specifically to the SELK neurons regarding the integration of various signaling processes. Furthermore, the single-cell RNA sequencing (scRNAseq) data available from the Drosophila brain, as reported by Li et al. (2022), does not allow accurate differentiation between SELK and LHLK neurons. To understand how these neurons integrate both metabolic and sensory information, it is crucial to conduct a focused RNAseq study specifically on the SELK neurons to understand how these neurons integrate both metabolic and sensory information. This targeted analysis would provide the necessary insights to elucidate their functional roles better. However, according to the data derived from the connectome, SELK neurons might be cholinergic, and this neurotransmitter might be involved in controlling also the behavior of the flies.

According to previous studies (Yurgel et al., 2019), the Lk-GAL4 line is also expressed in the VNC, thus the authors could make use of the tsh-GAL80 tool to clean up the line. This study also performed GCaMP imaging in fed and 24h starved animals in SELK and couldn't find a difference, can the authors explain this discrepancy?

We thank the reviewer for this suggestion. We have now added a new piece of data using the tsh-Gal80 transgene in our PER experiments (Supplementary Figure 7A). Blocking the expression of TNT in the ABLK neurons does not affect the main conclusion of the behavioral results. As stated previously, we were unable to obtain in vivo Ca imaging responses in SELK neurons upon exposure to sucrose, caffeine, or mixtures of sucrose and caffeine. We do not believe this is a discrepancy with previous works like Yurgel et al., 2019. It is likely that we faced technical issues regarding expression stability and that the stimulation was possibly too weak to detect changes in GFP levels

Reviewer #2 (Recommendations for the authors):

As mentioned above I do not have any major comments on the manuscript, but there are a few points that I feel should be considered:

(1) The identification of the Lk-candidate neurons in the connectome remains a bit mysterious. In the method sections, this reads as follows "manual and visual criteria were applied to identify the neurons of interest ". a) What precisely was done to get to the candidates?b) Are there alternative candidates that may be Lk-neurons? c) How would another neuron affect the conclusion of the downstream analysis?

We thank the reviewer for this comment. We have now modified and added new information in the connectomic section, reinforcing our conclusions and correcting the results obtained.

Our GRASP, BacTRace, and immunohistochemistry experiments pointed to SELK neurons as postsynaptic to both Gr64f^GRNs (sweet) and Gr66a^GRNs (bitter). To identify which neurons in the connectome could be the SELK neurons, we utilized a previously described set of GRNs already identified in the connectome (Shiu et al., 2022). We extracted all postsynaptic neurons to the sweet and bitter GRNs identified and intersected both datasets, retaining only those candidate hits receiving simultaneous input from sweet and bitter GRNs. This process yielded a total of 333 hits. Through visual inspection, we discarded all hits that were merely neuronal fragments or neurons that clearly were not our candidates. We narrowed the list down to a final set of 17 candidate neurons whose arborization was located in the SEZ. We reduced the candidates to two final entries from this list: ID 720575940623529610 (GNG.276) and ID 720575940630808827 (GNG.685). The GNG.276 neuron had a counterpart in the SEZ identified as GNG.246. Both of these neurons were annotated as DNg70 in the Flywire database. GNG.685 had a counterpart identified as GNG.595, and these two neurons were classified as DNg68. In both cases, the neuronal candidates, DNg70 and DNg68, were classified as descending neurons, a characteristic of previously described SELK neurons (Nässel et al., 2021). In our initial analysis published in bioRxiv and sent for revision, we identified DNg70 as potentially the SELK neurons based solely on the morphology of the neurons via visual inspection. However, we employed a better method to determine which candidate is more likely to be the SELK neurons, concluding that DNg68, rather than DNg70, represents the SELK neurons. Briefly, we performed an immunohistochemistry for GFP in the Lk-Gal4>UAS-CD8:GFP flies. We aligned the resulting image in a Drosophila reference brain (JRC2018 U) using the CMTK Registration plugin in ImageJ. The resulting image was skeletonized using the Single Neurite Tracer plugin in ImageJ and later uploaded to the Flywire Gateway platform to compare the structure of the aligned and skeletonized SELK neurons to our candidates. This comparison clearly indicated that the DNg68 neurons are the best candidates for representing the SELK neurons, rather than DNg70. We have updated the text and Figures 6 and Supplementary Figure 6 to reflect the new results. These new results do not alter the conclusions of the paper.

(2) In the transcriptomic experiments It seems that the raw transcripts are reporters, rather than normalised data. Why?

All transcriptomic data is normalized. In Figure 1 the differential expression was calculated using Deseq2 normalized counts. In Figure 2, Transcripts Per Million (TPM) were calculated using the Salmon package and normalized for the gene length.

(3) The expression of nAChRbeta1 in the transcriptomic data is rather striking. However, this remains currently not addressed: is this expression real?

We have not confirmed the upregulation or downregulation in gene expression for other but for Leucokinin, which is our main interest. We found the presence of nAChRbeta1 interesting, as GRNs are cholinergic (Jaeger et al., 2018), suggesting that it would make sense to find cholinergic receptors in G2Ns. However, it is possible that these receptors are expressed in all G2Ns and serve as a common means of communication.

(4) The description of the behavioural experiments in the results section is rather brief. I had a hard time following it since the genotypes are not repeated nor is it stated what is different in the experimental group vs control (but instead simply what changes in the experimental group, in a rather discussion-like fashion).

We thank the reviewer for the comment, we have rewritten this section to improve its clarity.

(5) If I understand the genetics for the behavioural experiments correctly it addresses the entire Lk-Gal4 expressing population, thus it is not possible to describe the role of the two SEZ neurons, but rather LkGal4 neurons. This should be clarified.

We thank the reviewer for this comment. Indeed, the Lk-Gal4 driver we used drives expression in all Leucokinin neurons, making it impossible to distinguish between the SELK, LHLK, or ABLK neurons. We have added a new piece of behavioral data by using the tsh-Gal80 transgene to prevent the expression of TNT in the ABLK neurons (Supplementary Figure 7A), but still we cannot distinguish between SELK and LHLK. We have rewritten the text to clarify this fact.

Reviewer #3 (Recommendations for the authors):

Overall, the manuscript is well-written, I only have one minor suggestion for improvement. In Figure 8C, please clarify the use of TNT to block Lk release.

We thank the reviewer for the comment, we have clarified the use of TNT in the text.

References Clyne, J. D. & Miesenböck, G. Sex-Specific Control and Tuning of the Pattern Generator for Courtship Song in Drosophila. Cell 133, 354–363 (2008).

Dorkenwald, S. et al. Neuronal wiring diagram of an adult brain. Nature 634, 124–138 (2024).

Engert, S., Sterne, G. R., Bock, D. D. & Scott, K. Drosophila gustatory projections are segregated by taste modality and connectivity. Elife 11, e78110 (2022).

Jaeger, A. H. et al. A complex peripheral code for salt taste in Drosophila. Elife 7, e37167 (2018).

Macpherson, L. J. et al. Dynamic labelling of neural connections in multiple colours by trans-synaptic fluorescence complementation. Nat Commun 6, 10024 (2015).

Nässel, D. R. Leucokinin and Associated Neuropeptides Regulate Multiple Aspects of Physiology and Behavior in Drosophila. Int J Mol Sci 22, 1940 (2021).

Shiu, P. K., Sterne, G. R., Engert, S., Dickson, B. J. & Scott, K. Taste quality and hunger interactions in a feeding sensorimotor circuit. eLife 11, e79887 (2022).

Walker, S. R., Peña-Garcia, M. & Devineni, A. V. Connectomic analysis of taste circuits in Drosophila. Sci. Rep. 15, 5278 (2025).

Author response:

Reviewer #1:

As this code was developed for use with a 4096 electrode array, it is important to be aware of double-counting neurons across the many electrodes. I understand that there are ways within the code to ensure that this does not happen, but care must be taken in two key areas. Firstly, action potentials traveling down axons will exhibit a triphasic waveform that is different from the biphasic waveform that appears near the cell body, but these two signals will still be from the same neuron (for example, see Litke et al., 2004 "What does the eye tell the brain: Development of a System for the Large-Scale Recording of Retinal Output Activity"; figure 14). I did not see anything that would directly address this situation, so it might be something for you to consider in updated versions of the code.

We thank the reviewer for this insightful comment. We agree that signals from the same neuron may be collected by adjacent channels. To address this concern in our software, we plan to add a routine to SpikeMAP that allows users to discard nearby channels where spike count correlations exceed a pre-determined threshold. Because there is no ground truth to map individual cells to specific channels on the hd-MEA, a statistical approach is warranted.

Secondly, spike shapes are known to change when firing rates are high, like in bursting neurons (Harris, K.D., Hirase, H., Leinekugel, X., Henze, D.A. & Buzsáki, G. Temporal interaction between single spikes and complex spike bursts in hippocampal pyramidal cells. Neuron 32, 141-149 (2001)). I did not see this addressed in the present version of the manuscript.

This is a valid concern. To ensure that firing rates are relatively constant over the duration of a recording, we will plot average spike rates using rolling windows of a fixed duration. We expect that population firing rates will remain relatively stable across the duration of recordings.

Another area for possible improvement would be to build on the excellent validation experiments you have already conducted with parvalbumin interneurons. Although it would take more work, similar experiments could be conducted for somatostatin and vasoactive intestinal peptide neurons against a background of excitatory neurons. These may have different spike profiles, but your success in distinguishing them can only be known if you validate against ground truth, like you did for the PV interneurons.

We agree that further cycles of experiments could be performed with SOM, VIP, and other neuronal subtypes, and we hope that researchers will take advantage of SpikeMAP too. We will clarify this possibility in the Discussion section of the manuscript.

Reviewer #2:

Summary:

While I find that the paper is nicely written and easy to follow, I find that the algorithmic part of the paper is not really new and should have been more carefully compared to existing solutions. While the GT recordings to assess the possibilities of a spike sorting tool to distinguish properly between excitatory and inhibitory neurons are interesting, spikeMAP does not seem to bring anything new to state-of-the-art solutions, and/or, at least, it would deserve to be properly benchmarked. I would suggest that the authors perform a more intensive comparison with existing spike sorters.

We thank the reviewer for this comment. As detailed in Table 1, SpikeMAP is the only method that performs E/I sorting on large-scale multielectrodes, hence a comparison to competing methods is not currently possible. That being said, many of the pre-processing steps of SpikeMAP (Figure 1) involve methods that are already well-established in the literature and available under different packages. To highlight the contribution of our work and facilitate the adoption of SpikeMAP, we plan to provide a “modular” portion of SpikeMAP that is specialized in performing E/I sorting and can be added to the pipeline of other packages such as KiloSort more clearly.  This modularized version of the code will be shared freely along with the more complete version already available.

Weaknesses:

(1) The global workflow of spikeMAP, described in Figure 1, seems to be very similar to that of Hilgen et al. 2020 (10.1016/j.celrep.2017.02.038). Therefore, the first question is what is the rationale of reinventing the wheel, and not using tools that are doing something very similar (as mentioned by the authors themselves). I have a hard time, in general, believing that spikeMAP has something particularly special, given its Methods, compared to state-of-the-art spike sorters.

We agree with the reviewers that there are indeed similarities between our work and the Hilgen et al. paper. However, while the latter employs optogenetics to stimulate neurons on a large-scale array, their technique does not specifically target inhibitory (e.g., PV) neurons as described in our work. We will clarify our paper accordingly.

This is why, at the very least, the title of the paper is misleading, because it lets the reader think that the core of the paper will be about a new spike sorting pipeline. If this is the main message the authors want to convey, then I think that numerous validations/benchmarks are missing to assess first how good spikeMAP is, with reference to spike sorting in general, before deciding if this is indeed the right tool to discriminate excitatory vs inhibitory cells. The GT validation, while interesting, is not enough to entirely validate the paper. The details are a bit too scarce for me, or would deserve to be better explained (see other comments after).

The title of our work will be edited to make it clear that while elements of the pipeline are well-established and available from other packages, we are the first to extend this pipeline to E/I sorting on large-scale arrays.

(2) Regarding the putative location of the spikes, it has been shown that the center of mass, while easy to compute, is not the most accurate solution [Scopin et al, 2024, 10.1016/j.jneumeth.2024.110297]. For example, it has an intrinsic bias for finding positions within the boundaries of the electrodes, while some other methods, such as monopolar triangulation or grid-based convolution, might have better performances. Can the authors comment on the choice of the Center of Mass as a unique way to triangulate the sources?

We agree with the reviewer and will point out limits of the center-of-mass algorithm based on the article of Scopin et al (2024). Further, we will augment the existing code library to include monopolar triangulation or grid-based convolution as options available to end-users.

(3) Still in Figure 1, I am not sure I really see the point of Spline Interpolation. I see the point of such a smoothing, but the authors should demonstrate that it has a key impact on the distinction of Excitatory vs. Inhibitory cells. What is special about the value of 90kHz for a signal recorded at 18kHz? What is the gain with spline enhancement compared to without? Does such a value depend on the sampling rate, or is it a global optimum found by the authors?

We will clarify these points. Specifically, the value of 90kHz was chosen because it provided a reasonable temporal characterization of spikes; this value, however, can be adjusted within the software based on user preference.

(4) Figure 2 is not really clear, especially panel B. The choice of the time scale for the B panel might not be the most appropriate, and the legend filtered/unfiltered with a dot is not clear to me in Bii.

We will re-check Fig.2B which seems to have error in rendering, likely due to conversion from its original format.

In panel E, the authors are making two clusters with PCA projections on single waveforms. Does this mean that the PCA is only applied to the main waveforms, i.e. the ones obtained where the amplitudes are peaking the most? This is not really clear from the methods, but if this is the case, then this approach is a bit simplistic and does not really match state-of-the-art solutions. Spike waveforms are quite often, especially with such high-density arrays, covering multiple channels at once, and thus the extracellular patterns triggered by the single units on the MEA are spatio-temporal motifs occurring on several channels. This is why, in modern spike sorters, the information in a local neighbourhood is often kept to be projected, via PCA, on the lower-dimensional space before clustering. Information on a single channel only might not be informative enough to disambiguate sources. Can the authors comment on that, and what is the exact spatial resolution of the 3Brain device? The way the authors are performing the SVD should be clarified in the methods section. Is it on a single channel, and/or on multiple channels in a local neighbourhood?

Here, the reviewer is suggesting that it may be better to perform PCA on several channels at once, since spikes can occur at several channels at the same time. To address this concern, small routine will be written allowing users to choose how many nearby channels to be selected for PCA.

(5) About the isolation of the single units, here again, I think the manuscript lacks some technical details. The authors are saying that they are using a k-means cluster analysis with k=2. This means that the authors are explicitly looking for 2 clusters per electrode? If so, this is a really strong assumption that should not be held in the context of spike sorting, because, since it is a blind source separation technique, one cannot pre-determine in advance how many sources are present in the vicinity of a given electrode. While the illustration in Figure 2E is ok, there is no guarantee that one cannot find more clusters, so why this choice of k=2? Again, this is why most modern spike sorting pipelines do not rely on k-means, to avoid any hard-coded number of clusters. Can the authors comment on that?

It is true that k=2 is a pre-determined choice in our software. In practice, we found that k>2 leads to poorly defined clusters. However, we will ensure that this parameter can be adjusted in the software. Furthermore, if the user chooses not to pre-define this value, we will provide the option to use a Calinski-Harabasz criterion to select k.

(6) I'm surprised by the linear decay of the maximal amplitude as a function of the distance from the soma, as shown in Figure 2H. Is it really what should be expected? Based on the properties of the extracellular media, shouldn't we expect a power law for the decay of the amplitude? This is strange that up to 100um away from the soma, the max amplitude only dropped from 260 to 240 uV. Can the authors comment on that? It would be interesting to plot that for all neurons recorded, in a normed manner V/max(V) as function of distances, to see what the curve looks like.

We share the reviewer’s concern and will add results that include a population of neurons to assess the robustness of this phenomenon.

(7) In Figure 3A, it seems that the total number of cells is rather low for such a large number of electrodes. What are the quality criteria that are used to keep these cells? Did the authors exclude some cells from the analysis, and if yes, what are the quality criteria that are used to keep cells? If no criteria are used (because none are mentioned in the Methods), then how come so few cells are detected, and can the authors convince us that these neurons are indeed "clean" units (RPVs, SNRs, ...)?

We applied stringent criteria to exclude cells, and we will revise the main text to be clear about these criteria, which include a minimum spike rate and the use of LDA to separate out PCA clusters. For the cells that were retained, we will include SNR estimates.

(8) Still in Figure 3A, it looks like there is a bias to find inhibitory cells at the borders, since they do not appear to be uniformly distributed over the MEA. Can the authors comment on that? What would be the explanation for such a behaviour? It would be interesting to see some macroscopic quantities on Excitatory/Inhibitory cells, such as mean firing rates, averaged SNRs... Because again, in Figure 3C, it is not clear to me that the firing rates of inhibitory cells are higher than Excitatory ones, whilst they should be in theory.       

We will include a comparison of firing rates for E and I neurons. It is possible that I cells are located at the border of the MEA due to the site of injections of the viral vector, and not because of an anatomical clustering of I cells per se. We will clarify the text accordingly.

(9) For Figure 3 in general, I would have performed an exhaustive comparison of putative cells found by spikeMAP and other sorters. More precisely, I think that to prove the point that spikeMAP is indeed bringing something new to the field of spike sorting, the authors should have compared the performances of various spike sorters to discriminate Exc vs Inh cells based on their ground truth recordings. For example, either using Kilosort [Pachitariu et al, 2024, 10.1038/s41592-024-02232-7], or some other sorters that might be working with such large high-density data [Yger et al, 2018, 10.7554/eLife.34518].

As mentioned previously, Kilosort and related approaches do not address the problem of E/I identification (see Table 1). However, they do have pre-processing steps in common with SpikeMAP. We will add some specific comparison points – for instance, the use of k-means and PCA (which is more common across packages) and the use of cubic spline interpolation (which is less common). Further, we will provide a stand-alone E/I sorting module that can be added to the pipeline of other packages, so that users can use this functionality without having to migrate their entire analysis.

(10) Figure 4 has a big issue, and I guess the panels A and B should be redrawn. I don't understand what the red rectangle is displaying.

We apologize for this issue. It seems there was a rendering problem when converting the figure from its original format. We will address this issue in the revised version of the manuscript.

(11) I understand that Figure 4 is only one example, but I have a hard time understanding from the manuscript how many slices/mice were used to obtain the GT data? I guess the manuscript could be enhanced by turning the data into an open-access dataset, but then some clarification is needed. How many flashes/animals/slices are we talking about? Maybe this should be illustrated in Figure 4, if this figure is devoted to the introduction of the GT data.

We will mention how many flashes/animals/slices were employed in the GT data and provide open access to these data.

(12) While there is no doubt that GT data as the ones recorded here by the authors are the most interesting data from a validation point of view, the pretty low yield of such experiments should not discourage the use of artificially generated recordings such as the ones made in [Buccino et al, 2020, 10.1007/s12021-020-09467-7] or even recently in [Laquitaine et al, 2024, 10.1101/2024.12.04.626805v1]. In these papers, the authors have putative waveforms/firing rate patterns for excitatory and inhibitory cells, and thus, the authors could test how good they are in discriminating the two subtypes.

We thank the reviewer for the suggestion that SpikeMAP could be tested on artificially generated spike trains and will add the citation of the two papers mentioned. We hope future efforts will employ SpikeMAP on both synthetic and experimental data to explore the neural dynamics of E and I neurons in healthy and pathological circuits of the brain.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The Authors investigated the anatomical features of the excitatory synaptic boutons in layer 1 of the human temporal neocortex. They examined the size of the synapse, the macular or the perforated appearance and the size of the synaptic active zone, the number and volume of the mitochondria, the number of the synaptic and the dense core vesicles, also differentiating between the readily releasable, the recycling and the resting pool of synaptic vesicles. The coverage of the synapse by astrocytic processes was also assessed, and all the above parameters were compared to other layers of the human temporal neocortex. The Authors conclude that the subcellular morphology of the layer 1 synapses is suitable for the functions of the neocortical layer, i.e. the synaptic integration within the cortical column. The low glial coverage of the synapses might allow the glutamate spillover from the synapses enhancing synaptic crosstalk within this cortical layer.

Strengths:

The strengths of this paper are the abundant and very precious data about the fine structure of the human neocortical layer 1. Quantitative electron microscopy data (especially that derived from the human brain) are very valuable, since this is a highly time- and energy consuming work. The techniques used to obtain the data, as well as the analyses and the statistics performed by the Authors are all solid, strengthen this manuscript, and support the conclusions drawn in the discussion.

Comments on latest version:

The third version of this paper has been substantially improved. The English is significantly better, there are only few paragraphs and sentences which are hard to understand (see my comments and suggestions below). Almost all of my suggestions were incorporated.

We would like to thank the reviewer for the comments and incorporated the suggestions within the latest version of the manuscript.

Remaining minor concerns:

About epileptic and non-epileptic (non-affected) tissue. I am aware that temporal lobe neocortical tissue derived from epileptic patients is regarded as non-affected by many groups, and they are quite similar to the cortex of non-epileptic (tumour) patients in their electrophysiological properties and synaptic physiology. But please, note, that one paper you cited did not use samples from epileptic patients, but only tissue from non-epileptic tumor patients (Molnár et al. PLOS 2008).

When you look deeper, and make thorough comparison of tissues derived from epileptic and non-epileptic patients, there are differences in the fine structure, as well as in several electrophysiological features. See for example Tóth et al., J Physiol, 2018, where higher density of excitatory synapses were found in L2 of neocortical samples derived from epileptic patients compared to non-epileptic (tumor) patients. Furthermore, the appearance of population bursts is similar, but their occurrence is more frequent and their amplitude is higher in tissue from epileptic compared to non-epileptic patients. So, I still cannot agree, that temporal neocortex of epileptic patients with the seizure focus in the hippocampus would be non-affected. Therefore I suggested to use the term biopsy tissue.

We are thankful for this comment on using non-epileptic tissue also by others. We are also aware that Molnár et al. 2008 worked with tumor tissue.

It is still not emphasized in the first paragraph of the Discussion, that only excitatory axon terminals were investigated.

We now mentioned in the first paragraph of the discussion that only excitatory synaptic boutons were investigated.

The text in the Results and the Discussion are somewhat inconsistent.

The last two paragraphs of the Results section ends with several sentences which should be part of the discussion, such as line 328: This finding strongly supports multivesicular release... or line 344: --- pointing towards a layer-specific regulation of the putative RRP. Moreover, the results suggest that... and line 370: ... it is most likely... Please, correct this.

We disagree with the reviewer on these points because these sentences summarizes the findings.

The first paragraph of the Discussion summarizes the work of the quantitative EM work and gives one conclusion about the astrocytic coverage. This last sentence is inconsistent with the other parts of the paragraph. I would either write that "astrocytic coverage was also investigated" (or something similar), or move this sentence to the paragraph which discusses the astrocytic coverage.

Results line 180-183. "Special connections" between astrocytic processes and synaptic boutons are mentioned, but not shown. Either show these (but then prove with staining!), or leave out this paragraph.

We deleted this paragraph as suggested.

Reviewer #2 (Public review):

Summary:

The study of Rollenhagen et al examines the ultrastructural features of Layer 1 of human temporal cortex. The tissue was derived from drug-resistant epileptic patients undergoing surgery, and was selected as further from the epilepsy focus, and as such considered to be non-epileptic. The analyses has included 4 patients with different age, sex, medication and onset of epilepsy. The manuscript is a follow-on study with 3 previous publications from the same authors on different layers of the temporal cortex:

Layer 4 - Yakoubi et al 2019 eLife

Layer 5 - Yakoubi et al 2019 Cerebral Cortex,

Layer 6 - Schmuhl-Giesen et al 2022 Cerebral Cortex

They find, the L1 synaptic boutons mainly have single active zone a very large pool of synaptic vesicles and are mostly devoid of astrocytic coverage.

Strengths:

The MS is well written easy to read. Result section gives a detailed set of figures showing many morphological parameters of synaptic boutons and surrounding glial elements. The authors provide comparative data of all the layers examined by them so far in the Discussion. Given that anatomical data in human brain are still very limited, the current MS has substantial relevance. The work appears to be generally well done, the EM and EM tomography images are of very good quality. The analyses is clear and precise.

Weaknesses:

The authors made all the corrections required and answered all of my concerns, included additional data sets, and clarified statements where needed.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Minor suggestions:

Synaptic density, lines 189-193. If you say "comparatively" high, then compare to something (cite your own work for the other layers, and tell the approximative values for the other layers). Same in line 194 comparably high to what? Other option: say "relatively high".

We corrected the sentences as suggested by the reviewer.

Line 206: When present, mitochondria (comma missing)

Corrected as suggested by the reviewer.

Line 265: Dot is missing at the end of the sentence (after Shapira et al. 2003)

Corrected as suggested by the reviewer.

Lines 300-301: Check the English for this sentence: significant difference BETWEEN TWO sublaminae and not significant difference for both sublaminae.

Corrected as suggested by the reviewer.

Lines 304-305: Check the sentence, please, it is not understandable without the text in parenthesis.

Corrected as suggested by the reviewer.

Line 354 Dot missing at the end of the sentence (after Figure 6A, B)

Corrected as suggested by the reviewer.

Line 354-358: Please rephrase this sentence (too complicated, not understandable). I do not understand why results of the L4, L5, L6 are described here. What does it mean "Astrocytes and their fine processes formed a relatively dense, but a comparably loose network within the neuropil in L1"? Dense or loose?

In the experiment measuring the volume fraction of astrocytic processes (Figure 6C), all six cortical layers were analyzed, thus we compared the values obtained for L1 with the results for L4, L5 and L6. For more clarity, we rephrased the sentence: “Astrocytes and their fine processes formed a relatively dense network in L4 and L5, but a comparably loose one within the neuropil in L1…” We also rephrased other sentences in this paragraph (as also suggested below).

Lines 359-369: Please rephrase this paragraph. The sentences are too complicated, have too many parentheses, and are not understandable. I suggest to write first how many synapses were examined in L1 and L4, then how many of them were on spine and on dendrites (either n or %). Then give the values how many (n or %) of them were "tripartite synapses", out of spine synapses and of dendritic synapses in both layers. How many of them were partially covered in both layers. Please, write the data in a systematic way. The best would be to give the values in a table as well. This way it will be more understandable (now, it is chaotic, hard to follow).

We rephrased the paragraph and added a new table (3).

Line 383: Dot missing from the end of the sentence.

Corrected as suggested by the reviewer.

Line 436: Reconsider "comparably low compared to". The comparably means what in this case? The whole paragraph is hard to understand, please, check and review for improvements to the use of English or use chatGPT to check it.

We corrected the sentence according to the reviewer’s suggestion.

Line 487: Same thing again: "The comparably largest size of the RP in L1 when compared..." What would you like to say with "comparably"? Check the meaning of this word in a dictionary, please. I have the feeling that you are using this word instead of "relatively".

Corrected as suggested by the reviewer.

Line 488 "and TO that found fot L4 and L5 in rodents..."

Corrected as suggested by the reviewer.

Line 493-495: Same again, comparably when compared, correct, please.

Corrected as suggested by the reviewer.

Supplemental figures: Now I do understand why Hu-01 and Hu-02 are twice, and I think, 3 patients were examined for L1a and three for L1b. But which side is which on the subfigures? Left side (Hu-01, 02 03) was used for L1a, or L1b? Could you write this in the legend, or mark on the figure (at least at one subfigure), please?

We implemented a comment for clarity.

Author response:

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

Public reviews:

Concerning the grounding in experimental phenomenology, it would be beneficial to identify specific experiments to strengthen the model. In particular, what evidence supports reversible beta cell inactivation? This could potentially be tested in mice, for instance, by using an inducible beta cell reporter, treating the animals with high glucose levels, and then measuring the phenotype of the marked cells. Such experiments, if they exist, would make the motivation for the model more compelling.

There is some direct evidence of reversible beta cell inactivation in rodent / in vitro models. We had already mentioned this in the discussion, but we have added some text emphasizing / clarifying the role of this evidence (lines 359–362).

Others have also argued that some analyses of insulin treatment in conventional T2D, which has a stronger effect in patients with higher glucose before treatment, provides indirect evidence of reversal of glucotoxicity. We have also mentioned this in the revised paper (lines 284–285).

For quantitative experiments, the authors should be more specific about the features of beta cell dysfunction in KPD. Does the dysfunction manifest in fasting glucose, glycemic responses, or both? Is there a ”pre-KPD” condition? What is known about the disease’s timescale?

The answers to some of these questions are not entirely clear—patients present with very high glucose, and thus must be treated immediately. Due to a lack of antecedent data it is not entirely clear what the pre-KPD condition is, but there is some evidence that KPD is at least not preceded by diabetes symptoms. This point is already noted in the introduction of the paper and Table 1. However, we have added a small note clarifying that this does not rule out mild hyperglycemia, as in prediabetes (and indeed, as our model might predict) (lines 76–77). Similarly, due to the necessity of immediate insulin treatment, it is not clear from existing data whether the disorder manifests more strongly in fasting glucose or glucose response, although it is likely in both. (We might infer this since continuous insulin treatment does not produce fasting hypoglycemia, and the complete lack of insulin response to glucose shortly after presentation should produce a strong effect in glycemic response.) We believe our existing description of KPD lists all of the relevant timescales, however we have also slightly clarified this description in response to the first referee’s comments (lines 66–73, 83)

The authors should also consider whether their model could apply to other conditions besides KPD. For example, the phenomenology seems similar to the ”honeymoon” phase of T1D. Making a strong case for the model in this scenario would be fascinating.

This is an excellent idea, which had not occurred to us. We have briefly discussed this possibility in the remission (lines 281–291), but plan to analyze it in more detail in a future manuscript.

Reviewer #1 (Recommendations for the author):

Whenever simulation results are presented, parameter values should be specified right there in the figure captions.

We have added the values of glucotoxicity parameters to the caption of Figure 2. In other figures, we have explicitly mentioned which panel of Figure 2 the parameters are taken from. Description of the non-glucotoxicity parameters is a bit cumbersome (there are a lot of them, but our model of fast dynamics is slightly different from Topp et al. so it does not suffice to simply say we took their parameters) so we have referred the reader to the Materials and Methods for those.

I was confused by the language in Figure 4. Could the authors clarify whether they argue that: (1) the observed KPD behaviour is the result of the system switching from one stable state to another when perturbed with high glucose intake? (2) the observed KPD behaviour is the result of one of the steady states disappearing with high glucose intake?

What we mean to say is that during a period of high sugar intake or exogeneous insulin treatment, one of the fixed points is temporarily removed—it is still a fixed point of the “normal” dynamics, but not a fixed point of the dynamics with the external condition added. Since when glucose (insulin) intake is high enough, only the low (high)-β fixed point is present, under one of these conditions the dynamics flow toward that fixed point. When the external influx of glucose/insulin is turned off, both fixed points are present again—but if the dynamics have moved sufficiently far during the external forcing, the fixed point they end up in will have switched from one fixed point to the other. We have edited the text to make this clearer (lines 153–185). Do note, however, that in response to both referee’s comments (see below), Figures 3 and 4 have been replaced with more illuminating ones. This specific point is now addressed by the new Figure 3.

The adaptation of the prefactor ’c’ was confusing to me. I think I understood it in the end, but it sounded like, ”here’s a complication, but we don’t explain it because it doesn’t really matter”. I think the authors can explain this better (or potentially leave out the complication with ’c’ altogether?).

Indeed, the existence of an adaptation mechanism is important for our overall picture of diabetes pathogenesis, but not for many of our analyses, which assume prediabetes. Nonetheless, we agree that the current explanation of it’s role is confusing because of its vagueness. We have elaborated the explanation of the type of dynamics we assume for c, adding an equation for its dynamics to the “Model” section of the Materials and methods, explained in lines 456–465. We have also amended Figure 1 to note this compensation.

I expect the main impact of this work will be to get clinical practitioners and biomedical researchers interested in the intermediate timescale dynamics of β-cells and take seriously the possibility that reversible inactive states might exist. But this impact will only be achieved when the results are clearly and easily understandable by an audience that is not familiar with mathematical modelling. I personally found it difficult to understand what I was supposed to see in the figures at first glance. Yes, the subtle points are indeed explained in the figure captions, but it might be advantageous to make the points visually so clear that a caption is barely needed. For example, when claiming that a change in parameters leads to bistability, why not plot the steady state values as a function of that parameter instead of showing curves from which one has to infer a steady state?

I would advise the authors to reconsider their visual presentation by, e.g., presenting the figures to clinical practitioners or biomedical researchers with just a caption title to test whether such an audience can decipher the point of the figure! This is of course merely a personal suggestion that the authors may decide to ignore. I am making this suggestion only because I believe in the quality of this work and that improving the clarity of the figures and the ease with which one can understand the main points would potentially lead to a much larger impact on the presented results.

This is a very good point. We have made several changes. Firstly, we have added smaller panels showing the dynamics of β to Figure 2; previously, the reader had to infer what was happening to β from G(t). Secondly, we have completely replaced the two figures showing dβ/dt, and requiring the reader to infer the fixed points of β, with bifurcation diagrams that simply show the fixed points of G and β. The new figures show through bifurcation diagrams how there are multiple fixed points in KPD, how glucose or insulin treatment force the switching of fixed points, and how the presence of bistability depends on the rate of glucotoxicity. (These new figures are Fig. 3–5 in the revised manuscript.)

Could the authors explicitly point out what could be learned from their work for the clinic? At the moment treatment consists of giving insulin to patients. If I understand correctly, nothing about the current treatment would change if the model is correct. Is there maybe something more subtle that could be relevant to devising an optimal treatment for KPD patients?

This is another very good point. We have added a new figure (Fig. 7) in our results section showing how this model, or one like it, can be analyzed to suggest an insulin treatment schedule (once parameters for an individual patient can be measured), and added some discussion of this point (lines 224–240) as well as lifestyle changes our model might suggest for KPD patients to the discussion (lines 413–425).

Similarly, could the authors explicitly point out how their model could be experimentally tested? For example, are the functions f(G) and g(G) experimentally accessible? Related to that, presumably the shape of those functions matters to reproduce the observed behaviour. Could the authors comment on that / analyze how reproducing the observed behaviour puts constraints on the shape of the used functions and chosen parameter values?

g(G) has not been carefully measured in cellular data, however it could be in more quantative versions of existing experiments. Further, our model indeed requires some general features for the forms of f(G) and g(G) to produce KPD-like phenomena. We have added some comment on this to the discussion section of the revised manuscript (lines 367–372).

Could the authors explicitly spell out which parameters they think differ between individual KPD patients, and which parameters differ between KPD patients and ’regular’ type 2 diabetics?

In general we expect all parameters should vary both among KPD patients and between KPD / “conventional” T2D. The primary parameter determining whether KPD and conventional T2D, is seen, however, is the ratio kIN/kRE. We have elaborated on both these points in the revised mansuscript. (Lines 186–192, 250–257.)

I was confused about the timescale of remission. At one point the authors write “KPD patients can often achieve partial remission: after a few weeks or months of treatment with insulin” but later the authors state that “the duration of the remission varies from 6 months to 10 years”.

The former timescale is the typical timescale achieve remission. After remission is reached, however, it may or may not last—patients may experience a relapse, where their condition worsens and they again require insulin. We have edited the text to clarify this distinction (lines 66–73).

When the authors talk about intermediate timescales in the main text could they specify an actual unit of time, such as days, weeks, or months as it would relate to the rate constants in their model for those transitions?

We have done so (lines 86–87, figure 1 caption, figure 2 caption). Getting KPD-like behavior requires (at high glucose) the deactivation process to be somewhat faster than the reactivation process, so the relevant scales are between weeks (reactivation) and days (deactivation at high G).

The authors state ”Our simple model of β-cell adaptation also neglects the known hyperglycemiainduced leftward shift in the insulin secretion curve f(G) in Eq. (2)) ”. This seems an important consideration. Could the authors comment on why they did not model this shift, and/or explicitly discuss how including it is expected to change the model dynamics?

We agree that this process seems potentially relevant, as it seems to happen on a relatively fast timescale compared to glucose-induced β-cell death. It is, however, not so well characterized quantitatively that including it is a simple matter of putting in known values—we would be making assumptions that would complicate the interpretation of our results.

It is clear that this effect will need to be considered when quanitatively modelling real patient data. However, it is also straightforward to argue that this effect by itself cannot produce KPD-like symptoms, and will only tend to reduce the rate of glucotoxocity necessary to produce bibstability. We have added a discussion of this in the revisions (lines 307–315). We have also, in general, expanded the discussion of the effects that each neglected detail we have mentioned is expected to have (lines 292–315).

The authors end with a statement that their results may “contribute to explanation of other observations that involve rapid onset or remission of diabetes-like phenomena, such as during pregnancy or for patients on very low calorie diets.” Could the authors spell out exactly how their model potentially relates to these phenomena?

Our thinking is that, even when another direct cause, such as loss of insulin resistance, is implicated in reversal of diabetes, some portion of the effect may be explained by reversal of glucotoxicity. This is indeed at this point just a hypothesis, but we have expanded on it briefly in the revision. (Lines 281–291.)

Minor typos:

In Figure 2.D the last zero of 200 on the axis was cut off.

Line 359 - there is a missing word ”in the analysis”.

We have fixed these typos, thanks.

Reviewer #2 (Recommendations for the author):

The manuscript could be significantly improved in two key areas: the presentation of the analysis, and the relation with experimental phenomenology.

Regarding the analysis presentation, the figures could be substantially enhanced with minimal effort from the authors. At present, they are sparse, lack legends, and offer only basic analysis. The authors should consider presenting, for example, a bifurcation diagram for beta cell mass and fasting glucose levels as a function of kIN, and how insulin sensitivity and average meal intake modulate this relationship. The goal should be to present clear, testable predictions in an intuitive manner. Currently, the specific testable predictions of the model are unclear.

The response to this question is copied from the reponses to related questions from the first referee.

This is a very good point. We have made several changes. Firstly, we have added smaller panels showing the dynamics of β to Figure 2; previously, the reader thad to infer what was happening to β from G(t). Secondly, we have completely replaced the two figures showing dβ/dt, and requiring the reader to infer the fixed points of β, with bifurcation diagrams that simply show the fixed points of G and β. The new figures show through bifurcation diagrams how there are multiple fixed points in KPD, how glucose or insulin treatment force the switching of fixed points, and how the presence of bistability depends on the rate of glucotoxicity. We have also supplemented our phase diagram that shows the effects of SI and the total beta cell population with bifurcation diagrams showing β as SI and βTOT are varied. (These new figures are Fig. 3–5 in the present manuscript.) Finally, we have added another figure analyzing the model’s predictions for the optimal insulin treatment and the resulting time needed to achieve remission (Fig. 7)

Author response:

Reviewer #1 (Public review):

The manuscript titled "The distinct role of human PIT in attention control" by Huang et al. investigates the role of the human posterior inferotemporal cortex (hPIT) in spatial attention. Using fMRI experiments and resting-state connectivity analyses, the authors present compelling evidence that hPIT is not merely an object-processing area, but also functions as an attentional priority map, integrating both top-down and bottom-up attentional processes. This challenges the traditional view that attentional control is localized primarily in frontoparietal networks.

The manuscript is strong and of high potential interest to the cognitive neuroscience community. Below, I raise questions and suggestions to help with the reliability, methodology, and interpretation of the findings.

Thank you for a nice summary of the key points of our study. Below you will find our responses to your questions.

(1) The authors argue that hPIT satisfies the criteria for a priority map, but a clearer justification would strengthen this claim. For example, how does hPIT meet all four widely recognized criteria, such as spatial selectivity, attentional modulation, feature invariance, and input integration, when compared to classical regions such as LIP or FEF? A more systematic summary of how hPIT meets these benchmarks would be helpful. Additionally, to what extent are the observed attentional modulations in hPIT independent of general task difficulty or behavioral performance?

Great suggestions! For the first suggestion, we will include a clearer justification in the revised manuscript. For the second one, all participants received task practice prior to scanning, and task accuracy exceeded 90% (we will explicitly report the accuracy rate in revision), suggesting the tasks were not overly demanding. Although ceiling effects limit the interpretability of behavioral-performance correlations, we argue that higher task demands would likely require greater attentional effort, leading to stronger modulation in hPIT, which aligns with our findings when we manipulated the attentional load.

(2) The authors report that hPIT modulation is invariant to stimulus category, but there appear to be subtle category-related effects in the data. Were the face, scene, and scrambled images matched not only in terms of luminance and spatial frequency, but also in terms of factors such as semantic familiarity and emotional salience? This may influence attentional engagement and bias interpretation.

The response of hPIT is generally insensitive to stimulus category, however, the reviewer is correct in noticing that attentional modulation in hPIT is slightly stronger to faces than scenes and scrambled images. Although faces used in the task had neutral expressions and the scene pictures were also neutral, it is indeed possible that potential semantic familiarity or emotional salience may contribute to the subtle category-related effects in the results of experiment 3. This point will be noted in the revised manuscript.

(3) The result that attentional load modulates hPIT is important and adds depth to the main conclusions. However, some clarifications would help with the interpretation. For example, were there observable individual differences in the strength of attentional modulation? How consistent were these effects across participants?

Yes, individual differences exist. In the revised manuscript, we will include individual subject data points in the figure 6B.

(4) The resting-state data reveal strong connections between hPIT and both dorsal and ventral attention networks. However, the analysis is correlational. Are there any complementary insights from task-based functional connectivity or latency analyses that support a directional flow of information involving hPIT? In addition, do the authors interpret hPIT primarily as a convergence hub receiving input from both DAN and VAN, or as a potential control node capable of influencing activity in these networks? Also, were there any notable differences between hemispheres in either the connectivity patterns or attentional modulation?

We agree that besides resting-state connection, task-based functional connectivity analyses would have the potential to provide additional information about whether hPIT serves as a convergence node or a control hub. While fMRI data are not the best to generate directional flow of information due to the low temporal resolution, we will conduct task-based functional connectivity analyses.

We also observed modest hemispheric asymmetries in connectivity—for instance, both left and right hPIT showed stronger connectivity with right-hemisphere attention nodes. This will be described in the revised supplement.

(5) A few additional questions arise regarding the anatomical characteristics of hPIT: How consistent were its location and size across participants? Were there any cases where hPIT could not be reliably defined? Given the proximity of hPIT to FFA and LOp, how was overlap avoided in ROI definition? Were the functional boundaries confirmed using independent contrasts?

The size and location of hPIT are generally consistent across subjects, as shown in Supplementary Figure 1. The consistency is also supported by figure 4C. The hPIT is defined by conjunction maps across three tasks and then manually delineated avoiding overlapping voxels with FFA and LOp. The FFA was defined using an independent contrast (Exp3 contrast [face-scene]) and the Lop location was defined by anatomical parcellation (Glasser et al., 2016).

Reviewer #2 (Public review):

Summary

This study investigates the role of the human posterior inferotemporal cortex (hPIT) in attentional control, proposing that hPIT serves as an attentional priority map that integrates both top-down (endogenous) and bottom-up (exogenous) attentional processes. The authors conducted three types of fMRI experiments and collected resting-state data from 15 participants. In Experiment 1, using three different spatial attention tasks, they identified the hPIT region and demonstrated that this area is modulated by attention across tasks. In Experiment 2, by manipulating the presence or absence of visual stimuli, they showed that hPIT exhibits strong attentional modulation in both conditions, suggesting its involvement in both bottom-up and top-down attention. Experiment 3 examined the sensitivity of hPIT to stimulus features and attentional load, revealing that hPIT is insensitive to stimulus category but responsive to task load - further supporting its role as an attentional priority map. Finally, resting-state functional connectivity analyses showed that hPIT is connected to both dorsal and ventral attention networks, suggesting its potential role as a bridge between the two systems. These findings extend prior work on monkey PITd and provide new insights into the integration of endogenous and exogenous attention.

Strengths

(1) The study is innovative in its use of specially designed spatial attention tasks to localize and validate hPIT, and in exploring the region's role in integrating both endogenous and exogenous attention, as prior works focus primarily on its involvement in endogenous attention.

(2) The authors provided very comprehensive experiment designs with clear figures and detailed descriptions.

(3) A broad range of analyses was conducted to support the hypothesis that hPIT functions as an attentional priority map -- including experiments of attentional modulation under both top-down and bottom-up conditions, sensitivity to stimulus features and task load, and resting-state functional connectivity. These analyses showed consistent results.

(4) Multiple appropriate statistical analyses - including t-tests, ANOVAs, and post-hoc tests - were conducted, and the results are clearly reported.

Thank you for a nice summary of the key points and strengths of our study.

Weaknesses

(1) The sample size is relatively small (n = 15), and inter-subject variability is big in Figures 5 and 6, as seen in the spread of individual data points and error bars. The analysis of attention-modulated voxel map intersections appears to be influenced by multiple outliers.

We agree that the sample size (n = 15) is not ideal, and we acknowledge that some data points in Figures 5 and 6 appear to be potential outliers. However, according to conventional outlier detection criteria, all data points are within three standard deviations of the group mean and were therefore retained for analysis. Moreover, the attention-modulated voxel intersection map shown in Figure 4C is insensitive to outliers, because the intersection map plotted is based on the number of subjects.

(2) The authors acknowledge important limitations, including the lack of exploration of feature-based attention and the temporal constraints inherent to fMRI.

Yes, we hope to address these limitations in future studies.

(3) Prior research has established that regions such as the prefrontal cortex (PFC) and posterior parietal cortex (PPC) are involved in both endogenous and exogenous attention and have been proposed as attentional priority maps. It remains unclear what is uniquely contributed by hPIT, how it functionally interacts with these classical attentional hubs, and whether its role is complementary or redundant. The study would benefit from more direct comparisons with these regions.

In this study, we define the ROI base on intersection across three different types of spatial attention tasks, and the hPIT stands out in showing spatial attentional modulation across tasks. This could be due to the weak lateralized responses in PFC/PPC. To evaluate whether a region qualifies as a priority map, we applied four criteria (as mentioned in introduction). While dorsal and ventral attention network (DAN and VAN) regions can be considered important components of the priority map system, our findings suggest that among the regions tested, hPIT meets all four criteria. In Experiment 2, we included regions such as VFC (as part of PFC) and IPS (as part of PPC), and our findings suggest these areas are more involved in top-down attention. We agree with the reviewer’s suggestion and will perform additional analysis on PPC and PFC.

(4) The functional connectivity analysis is only performed on resting-state data, and this approach does not capture context-dependent interactions. Task-based data analysis can provide stronger evidence.

We acknowledge that resting-state FC is limited in assessing task-specific communication. To further investigate the role of hPIT, we plan to conduct task-based functional connectivity analyses.

(5) The study does not report whether attentional modulation in hPIT is consistent across the two hemispheres. A comparison of hemispheric effects could provide important insight into lateralization and inter-individual variability, especially given the bilateral localization of hPIT.

We thank the reviewer for this suggestion. hPIT was localized bilaterally using the same intersection-based method in Experiment 1. We have now performed additional analysis and found in Experiment 3, the difference in attentional modulation between high and low load conditions was significant in the right hPIT but not in the left. This result will be reported in the revised manuscript.

Author response:

Below, we will address point by point any and all concerns of the reviewers.

Reviewer #1:

There are no major concerns, but some material could be added for clarity and to make the work more accessible to a more general scientific audience.

We will add text for clarity and to make the work more accessible to a general audience per this comment and similar suggestions of the other reviewers.

(1.1) A figure clearly showing the habituation protocol and the use of the dishabituators would be a good addition, even if the procedure has been done before and is cited. There can always be readers who are seeing this for the first time.

We do think this is a good idea as the time scales of the experiment will be clearly marked as well and we plan to generate one in the revised manuscript.

(1.2) It would also be nice to comment on other ways dishabituation can happen (for example, when the stimulus is removed for a short time and returns) and what their time scales are.

If the stimulus is withheld, spontaneous recovery occurs, a process distinct from dishabituation and worth exploring on its own. In a previous publication (Semelidou et al. eLife 2018;7:e39569), we have shown that in this habituation paradigm with 4 min exposure either to the aversive Octanol, or the attractive Ethyl Acetate, spontaneous recovery occurs on or after 6 minutes after the habituated stimulus is withheld. This contrasts the immediate effect of the single dishabituating stimulus, delivered for a few seconds at the end of exposure to the habituator. Granted that per Thomson (Neurobiol Learn Mem. 2009), spontaneous recovery is a characteristic of habituation, we will work this point in the text.

(1.3) And more generally, the paper could perhaps improve by making a stronger case for why the results are important not just for flies but for neuroscience in general.

Thank you for the encouragement. We will try to rationally generalize our findings.

Reviewer #2:

(2.1) However, the claim that this represents a fundamental difference between homosensory and heterosensory pathways for dishabituation is overstated.

We had no intention of stating more than the fact that footshock and yeast odor dishabituators relay these stimuli to the mushroom bodies via distinct dopaminergic neurons, hence differentiating distinct dishabituating stimuli via the mechanosensory (footshock) and olfactory (yeast odor) modalities as they engage the mushroom bodies. As the reviewer suggests we will use more measured and specific language to state the above.

(2.2) The introductory section does not adequately present current broad models for habituation and dishabituation.

This was not done intentionally, but rather because we aimed at a less extended introductory section and ostensibly this resulted in brief and possibly inadequate presentation of current habituation models. We will present a much more detailed introduction and detail of habituation and dishabituation models in the revised manuscript (Also see reply to point 3.5 below).

(2.3) There are many different time scales, even for Drosophila olfactory habituation. These, as well as potential underlying mechanistic differences, need to be acknowledged; any claim should be specifically qualified for the time scales being studied here.

We understand and appreciate the point of the reviewer, as well as its significance and we will address this both in the revised text, but also by the paradigm figure we will add as stated above (point 1.1), where the time scales will be explicitly included and emphasized.

(2.4) Additionally, there are several unclear, vague, and inaccurate sections and statements. A more careful, precise, and considered presentation of current views, as well as more measured claims of the impact of the findings, would substantially enhance my enthusiasm.

We will address these concerns of course, though pointing out the specific offending parts would ascertain addressing them thoroughly. As stated above, we will incorporate current views in the introduction and when discussing our results and their impact.

Reviewer #3:

(3.1) The key issue is that the main concepts of this manuscript appear to be based on a misunderstanding/misinterpretation of the literature. As the authors set out to settle the debate "whether the novel dishabituating stimulus elicits sensitization of the habituated circuits, or it engages distinct neuronal routes to bypass habituation reinstating the naïve response", it seems that the authors based their investigation on the premise that "sensitization" is mediated by a facilitatory process within the S-R pathway, and "dishabituation" by a facilitatory process outside the S-R pathway. This is not the status quo in the field, particularly with the prevailing theory like the Dual-Process Theory.

We appreciate the reviewer’s comment and the opportunity to clarify the conceptual framework of our work. Our intention was in fact to test the Groves and Thomson hypothesis (Neurobiol Learn Mem. 2009), in our olfactory habituation system. As such, dishabituation could have been the result of a facilitatory process within the S-R pathway, or from mechanisms outside of it. Our experimental design allowed to distinguish these possibilities and our results clearly show that dishabituation involves circuitry outside the S-R pathway. We do thank the reviewer for pointing out that we have not articulated clearly this intention and we will take care to communicate this effectively in the revised manuscript.

(3.2) The original version of Dual-Process Theory (Groves and Thompson 1970, but also see Thompson 2008, Neurobiol Learn Mem) already hypothesized that habituation happens within the specific S-R pathway, and sensitization occurs separately in an "organism-wide" state system that modulates the output of all S-R pathways.

As mentioned above, we are aware of the Dual-Process hypothesis. In fact, our data demonstrate that activity outside the olfactory S-R pathway, engaging novel neuronal circuits, mediates dishabituation. Unlike habituation, these circuits mediating dishabituation include at minimum, the mushroom bodies, the dopaminergic system and the APL neurons. In our view this does not support the “organism-wide state” system, but rather particular circuits that in agreement with the Groves and Thomson hypothesis, are outside the S-R pathway and modulate its behavioral output. We will work these concepts in the discussion section of the revised manuscript.

(3.3) Dishabituation is recognized by the Dual-Process Theory as sensitization (organism-wide facilitation) manifested on top of existing habituation (depressed S-R pathway). This notion has been supported by a wide range of studies, including cat spinal cord reflex (e.g. Spencer et al. 1966) and work in Aplysia on heterosynaptic facilitation for both sensitization and dishabituation. Therefore, simply showing that the newly identified facilitatory pathways are outside the S-R habituation pathway is insufficient to demonstrate dishabituation.

We respectfully disagree with the concluding sentence here. In all of our experiments, we observe a clear recovery of olfactory avoidance after exposure to the footshock, or yeast odor dishabituators. Moreover, the dishabituators are emulated by (photo)activation of particular neuronal circuits and the recovery of olfactory avoidance is blocked when these circuits are silenced. Regardless of whether this recovery is classified as dishabituation via sensitization or another facilitatory process, the key point is that the habituated response is reliably reinstated contingent upon the dishabituating stimulus. We believe this meets the established criteria for dishabituation.

(3.4) As behavioral facilitation of a habituated response can be achieved by dishabituating (specific recovery of the S-R pathway) and/or superimposed sensitizing (organism-wide) processes, dishabituation and sensitization of this olfactory response must be first dissociated; however, the study provided no evidence for the dissociation. Without this piece of evidence, the claim of this paper that the newly identified pathways mediate dishabituation is not fully supported.

We agree with the reviewer that we have not provided specific evidence dissociating dishabituation and sensitization of the particular olfactory response beyond the evidence implicating particular circuitry in the outcome of facilitation of the olfactory response.

It should be noted that in photoactivation of the implicated circuitries in naïve flies, we do not observe enhanced octanol avoidance, suggesting that activation of these circuits alone does not induce sensitization. Moreover, our results show that neither footshock nor yeast odor drive an organism-wide sensitization, as silencing specific circuits was sufficient to block dishabituation—something that would not be expected if a global sensitization process was responsible of reinstating the olfactory response.

Nonetheless, we will also attempt to dissociate sensitization from dishabituation using mutants previously reported deficient in sensitization (Duerr and Quinn, PNAS 1982), assuming these mutants retain normal olfactory habituation. We will also try sensitization protocols in the case of within-modal dishabituation to further clarify the underlying mechanisms. In principle, this includes using diluted Octanol as the habituating stimulus and attempt dishabituation with concentrated octanol.

(3.5) The literature review of this manuscript has some discrepancies. In the introduction, the authors wrote "initial studies in Aplysia were consistent with the "dual-process theory" (Groves and Thompson 1979), where response recovery due to dishabituation appeared to result from sensitization superimposed on habituation, thus driving reversal of the attenuated response (Carew, Castellucci et al. 1971, Hochner, Klein et al. 1986, Marcus, Nolen et al. 1988, Ghirardi, Braha et al. 1992, Cohen, Kaplan et al. 1997, Antonov, Kandel et al. 1999, Hawkins, Cohen et al. 2006)." Hochner 1986 and Marcus 1988 in fact indicated otherwise. Hochner 1986 suggests that dishabituation and sensitization involve different molecular processes, while Marcus 1988 showed that dishabituation and sensitization have different behavioral characteristics. Therefore, the authors' statement is not supported by the cited literature.

We are grateful to the reviewer for pointing out these significant discrepancies, consequent of multiple rounds of edits followed by our own oversight. These important publications for this manuscript will be referenced properly in the revised version of the manuscript.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

This manuscript describes the role of PRDM16 in modulating BMP response during choroid plexus (ChP) development. The authors combine PRDM16 knockout mice and cultured PRDM16 KO primary neural stem cells (NSCs) to determine the interactions between BMP signaling and PRDM16 in ChP differentiation.

They show PRDM16 KO affects ChP development in vivo and BMP4 response in vitro. They determine genes regulated by BMP and PRDM16 by ChIP-seq or CUT&TAG for PRDM16, pSMAD1/5/8, and SMAD4. They then measure gene activity in primary NSCs through H3K4me3 and find more genes are co-repressed than co-activated by BMP signaling and PRDM16. They focus on the 31 genes found to be co-repressed by BMP and PRDM16. Wnt7b is in this set and the authors then provide evidence that PRDM16 and BMP signaling together repress Wnt activity in the developing choroid plexus.

Strengths:

Understanding context-dependent responses to cell signals during development is an important problem. The authors use a powerful combination of in vivo and in vitro systems to dissect how PRDM16 may modulate BMP response in early brain development.

We thank the reviewer for the thoughtful summary and positive feedback. We appreciate the recognition of our integrative in vivo and in vitro approach. We're glad the reviewer found our findings on context-dependent gene regulation and developmental signaling valuable.

Main weaknesses of the experimental setup:

(1) Because the authors state that primary NSCs cultured in vitro lose endogenous Prdm16 expression, they drive expression by a constitutive promoter. However, this means the expression levels are very different from endogenous levels (as explicitly shown in Supplementary Figure 2B) and the effect of many transcription factors is strongly dose-dependent, likely creating differences between the PRDM16-dependent transcriptional response in the in vitro system and in vivo.

We acknowledge that our in vitro experiments may not ideally replicate the in vivo situation, a common limitation of such experiments, our primary aim was to explore the molecular relationship between PRDM16 and BMP signaling in gene regulation. Such molecular investigations are challenging to conduct using in vivo tissues. In vitro NSCs treated with BMP4 has been used a model to investigate NSC proliferation and quiescence, drawing on previous studies (e.g., Helena Mira, 2010; Marlen Knobloch, 2017). Crucially, to ensure the relevance of our in vitro findings to the in vivo context, we confirmed that cultured cells could indeed be induced into quiescence by BMP4, and this induction necessitated the presence of PRDM16. Furthermore, upon identifying target genes co-regulated by PRDM16 and SMADs, we validated PRDM16's regulatory role on a subset of these genes in the developing Choroid Plexus (ChP) (Fig. 7 and Suppl.Fig7-8). Only by combining evidence from both in vitro and in vivo experiments could we confidently conclude that PRDM16 serves as an essential co-factor for BMP signaling in restricting NSC proliferation.

(2) It seems that the authors compare Prdm16_KO cells to Prdm16 WT cells overexpressing flag_Prdm16. Aside from the possible expression of endogenous Prdm16, other cell differences may have arisen between these cell lines. A properly controlled experiment would compare Prdm16_KO ctrl (possibly infected with a control vector without Prdm16) to Prdm16_KO_E (i.e. the Prdm16_KO cells with and without Prdm16 overexpression.)

We agree that Prdm16 KO cells carrying the Prdm16-expressing vector would be a good comparison with those with KO_vector. However, despite more than 10 attempts with various optimization conditions, we were unable to establish a viable cell line after infecting Prdm16 KO cells with the Prdm16-expressing vector. The overall survival rate for primary NSCs after viral infection is low, and we observed that KO cells were particularly sensitive to infection treatment when the viral vector was large (the Prdm16 ORF is more than 3kb).

As an alternative oo assess vector effects, we instead included two other control cell lines, wt and KO cells infected with the 3xNLS_Flag-tag viral vector, and presented the results in supplementary Fig 2.  When we compared the responses of the four lines — wt, KO, wt infected with the Flag vector, KO infected with the Flag vector — to the addition and removal of BMP4, we confirmed that the viral infection itself has no significant impacts on the responses of these cells to these treatments regarding changes in cell proliferation and Ttr induction.

Given that wt cells and the KO cells, with or without viral backbone infection behave quite similarly in terms of cell proliferation, we speculate that even if we were successful in obtaining a cell line with Prdm16-expressing vector in the KO cells, it may not exhibit substantial differences compared to wt cells infected with Prdm16-expressing vector.

Other experimental weaknesses that make the evidence less convincing:

(1) The authors show in Figure 2E that Ttr is not upregulated by BMP4 in PRDM16_KO NSCs. Does this appear inconsistent with the presence of Ttr expression in the PRDM16_KO brain in Figure1C?

The reviwer’s point is that there was no significant increase in Ttr expression in Prdm16_KO cells after BMP4 treatment (Fig. 2E), but there remained residule Ttr mRNA signals in the Prdm16 mutant ChP (Fig. 1C). We think the difference lies in the measuable level of Ttr expression between that induced by BMP4 in NSC culture and that in the ChP. This is based on our immunostaining expreriment in which we tried to detect Ttr using a Ttr antibody. This antibody could not detect the Ttr protein in BMP4-treated Prdm16_expressing NSCs but clearly showed Ttr signal in the wt ChP. This means that although Ttr expression can be significantly increased by BMP4 in vitro to a level measurable by RT-qPCR, its absolute quantity even in the Prdm16_expressing condition is much lower compared to that in vivo. Our results in Fig 1C and Fig 2E, as well as Fig 7B, all consistently showed that Prdm16 depletion significantly reduced Ttr expression in in vitro and in vivo.

(2) Figure 3: The authors use H3K4me3 to measure gene activity. This is however, very indirect, with bulk RNA-seq providing the most direct readout and polymerase binding (ChIP-seq) another more direct readout. Transcription can be regulated without expected changes in histone methylation, see e.g. papers from Josh Brickman. They verify their H3K4me3 predictions with qPCR for a select number of genes, all related to the kinetochore, but it is not clear why these genes were picked, and one could worry whether these are representative.

H3K4me3 has widely been used as an indicator of active transcription and is a mark for cell identity genes. And it has been demonstrated that H3K4me3 has a direct function in regulating transciption at the step of RNApolII pausing release. As stated in the text, there are advantages and disadvantages of using H3K4me3 compared to using RNA-seq. RNA-seq profiles all gene products, which are affected by transcription and RNA stability and turnover. In contrast, H3K4me3 levels at gene promoter reflects transcriptional activity. In our case, we aimed to identify differential gene expression between proliferation and quiescence states. The transition between these two states is fast and dynamic. RNA-seq may not be able to identify functionally relevant genes but more likely produces false positive and negative results. Therefore, we chose H3K4me3 profiling.

We agree that transcription may change without histone methylation changes. This may cause an under-estimation of the number of changed genes between the conditions. 

We validated 7 out of 31 genes (Wnt7b, Id3, Mybl2, Spc24, Spc25, Ndc80 and Nuf2). We chose these genes based on two critira: 1) their function is implicated in cell proliferation and cell-cycle regulation based on gene ontology analysis; 2) their gene products are detectable in the developing ChP based on the scRNA-seq data. Three of these genes (Wnt7b, Id3, Mybl2) are not related to the kinetochore. We now clarify this description in the revised text.

(3) Line 256: The overlap of 31 genes between 184 BMP-repressed genes and 240 PRDM16-repressed genes seems quite small.

This result indicates that in addition to co-repressing cell-cycle genes, BMP and PRDM16 have independent fucntions. For example, it was reported that BMP regulates neuronal and astrocyte differentiation (Katada, S. 2021), while our previous work demonstrated that Prdm16 controls temporal identity of NSCs (He, L. 2021).

(4) The Wnt7b H3K4me3 track in Fig. 3G is not discussed in the text but it shows H3K4me3 high in _KO and low in _E regardless of BMP4. This seems to contradict the heatmap of H3K4me3 in Figure 3E which shows H3K4me3 high in _E no BMP4 and low in _E BMP4 while omitting _KO no BMP4. Meanwhile CDKN1A, the other gene shown in 3G, is missing from 3E.

The track in Fig 3G shows the absolute signal of H3K4me3 after mapping the sequencing reads to the genome and normaliz them to library size. Compare the signal in Prdm16_E with BMP4 and that in Prdm16_E without BMP4, the one with BMP4 has a lower peak. The same trend can be seen for the pair of Prdm16_KO cells with or without BMP4.  The heatmap in Fig. 3E shows the relative level of H3K4me3 in three conditions. The Prdm16_E cells with BMP4 has the lowest level, while the other two conditions (Prdm16_KO with BMP4 and Prdm16_E without BMP4) display higher levels. These two graphs show a consistent trend of H3K4me3 changes at the Wnt7b promoter across these conditions. Figure 3E only includes genes that are co-repressed by PRDM16 and BMP. CDKN1A’s H3K4me3 signals are consistent between the conditions, and thus it is not a PRDM16- or BMP-regulated gene. We use it as a negative control. 

(5) The authors use PRDM16 CUT&TAG on dissected dorsal midline tissues to determine if their 31 identified PRDM16-BMP4 co-repressed genes are regulated directly by PRDM16 in vivo. By manual inspection, they find that "most" of these show a PRDM16 peak. How many is most? If using the same parameters for determining peaks, how many genes in an appropriately chosen negative control set of genes would show peaks? Can the authors rigorously establish the statistical significance of this observation? And why wasn't the same experiment performed on the NSCs in which the other experiments are done so one can directly compare the results? Instead, as far as I could tell, there is only ChIP-qPCR for two genes in NSCs in Supplementary Figure 4D.

In our text, we indicated the genes containing PRDM16 binding peaks in the figures and described them as “Text in black in Fig. 6A and Supplementary Fig. 5A”. We will add the precise number “25 of these genes” in the main text to clarify it. We used BMP-only repressed 184-31 =153 genes (excluding PRDM16-BMP4 co-repressed) as a negative control set of genes. By computationally determine the nearest TSS to a PRDM16 peak, we identified 24/31 co-repressed genes and 84/153 BMP-only-repressed genes, containing PRDM16 peaks in the E12.5 ChP data. Fisher’s Exact Test comparing the proportions yields the P-value = 0.015.

We are confused with the second part of the comment “And why wasn't the same experiment performed on the NSCs in which the other experiments are done so one can directly compare the results? Instead, as far as I could tell, there is only ChIP-qPCR for two genes in NSCs in Supplementary Figure 4D.” If the reviewer meant why we didn’t sequence the material from sequential-ChIP or validate more taget genes, the reason is the limitation of the material. Sequential ChIP requires a large quantity of the antibodies, and yields little material barely sufficient for a few qPCR after the second round of IP. This yielded amount was far below the minimum required for library construction. The PRDM16 antibody was a gift, and the quantity we have was very limited. We made a lot of efforts to optimize all available commercial antibodies in ChIP and Cut&Tag, but none of them worked in these assays.

(6) In comparing RNA in situ between WT and PRDM16 KO in Figure 7, the authors state they use the Wnt2b signal to identify the border between CH and neocortex. However, the Wnt2b signal is shown in grey and it is impossible for this reviewer to see clear Wnt2b expression or where the boundaries are in Figure 7A. The authors also do not show where they placed the boundaries in their analysis. Furthermore, Figure 7B only shows insets for one of the regions being compared making it difficult to see differences from the other region. Finally, the authors do not show an example of their spot segmentation to judge whether their spot counting is reliable. Overall, this makes it difficult to judge whether the quantification in Figure 7C can be trusted.

In the revised manuscript we have included an individal channel of Wnt2b and mark the boundaries. We also provide full-view images and examples of spot segmentation in the new supplementary figure 8. 

(7) The correlation between mKi67 and Axin2 in Figure 7 is interesting but does not convincingly show that Wnt downstream of PRDM16 and BMP is responsible for the increased proliferation in PRDM16 mutants.

We agree that this result (the correlation between mKi67 and Axin2) alone only suggests that Wnt signaling is related to the proliferation defect in the Prdm16 mutant, and does not necessarily mean that Wnt is downstream of PRDM16 and BMP. Our concolusion is backed up by two additional lines of evidences:  the Cut&Tag data in which PRDM16 binds to regulatory regions of Wnt7b and Wnt3a; BMP and PRDM16 co-repress Wnt7b in vitro.

An ideal result is that down-regulating Wnt signaling in Prdm16 mutant can rescue Prdm16 mutant phenotype. Such an experiment is technically challenging. Wnt plays diverse and essential roles in NSC regulation, and one would need to use a celltype-and stage-specific tool to down-regulate Wnt in the background of Prdm16 mutation. Moreover, Wnt genes are not the only targets regulated by PRDM16 in these cells, and downregulating Wnt may not be sufficient to rescue the phenotype. 

Weaknesses of the presentation:

Overall, the manuscript is not easy to read. This can cause confusion.

We have revised the text to improve clarity.

Reviewer #1 (Recommendations for the authors):

(1) Overall, the manuscript is not easy to read. Here are some causes of confusion for which the presentation could be cleaned up:

We are grateful for the reviewer’s suggestion. In the revised manuscript, we have made efforts to improve the clarity of the text.

(a) Part of the first section is confusing in that some statements seem contradictory, in particular:

"there is no overall patterning defect of ChP and CH in the Prdm16 mutant" (line 125)

"Prdm16 depletion disrupted the transition from neural progenitors into ChP epithelia" (line 144)

It would be helpful if the authors could reformulate this more clearly.

We modified the text to clarify that while the BMP-patterned domain is not affected, the transition of NSCs into ChP epithelial cells is compromised in the Prdm16 mutant.

(b) Flag_PRDM16, PRDM16_expressing, PRDM16_E, PRDM16 OE all seem to refer to the same PRDM16 overexpressing cells, which is very confusing. The authors should use consistent naming. Moreover, it would be good if they renamed these all to PRDM16_OE to indicate expression is not endogenous but driven by a constitutive promoter.

We appreciate the comment and agree that the use of multiple terms to refer to the same PRDM16-overexpressing condition was confusing. Our original intention in using Prdm16_E was to distinguish cells expressing PRDM16 from the two other groups: wild-type cells and Prdm16_KO cells, which both lack PRDM16 protein expression. However, we acknowledge that Prdm16_E could be misinterpreted as indicating expression from the endogenous Prdm16 promoter. To avoid this confusion and ensure consistency, we have now standardized the terminology and refer to this condition as Prdm16_OE, indicating Flag-tagged PRDM16 expression driven by a constitutive promoter.

(c) Line 179 states "generated a cell line by infecting Prdm16_KO cells with the same viral vector, expressing 3xNSL_Flag". Do the authors mean 3xNLS_Flag_Prdm16, so these are the Prdm16_KO_E cells by the notation suggested above? Or is this a control vector with Flag only? The following paragraph refers to Supplementary Figure 2C-F where the same construct is called KO_CDH, suggesting this was an empty CDH vector, without Flag, or Prdm16. This is confusing.

We appreciate the reviewer’s careful reading and helpful comment. We acknowledge the confusion caused by the inconsistent terminology. To clarify: in line 179, we intended to describe an attempt to generate a Prdm16_KO cell line expressing 3xNLS_Flag_Prdm16, not a control vector with Flag only. However, despite repeated attempts, we were unable to establish this line due to low viral efficiency and the vulnerability of Prdm16_KO cells to infection with the large construct. Therefore, these cells were not included in the subsequent analyses.

The term KO_CDH refers to Prdm16_KO cells infected with the empty CDH control vector, which lacks both Flag and Prdm16. This is the line used in the experiments shown in Supplementary Fig. 2C–F. We have revised the text throughout the manuscript to ensure consistent use of terminology and to avoid this confusion.

(2) The introductory statements on lines 53-54 could use more references.

Thanks for the suggestion. We have now included more references.

(3) It would be helpful if all structures described in the introduction and first section were annotated in Figure 1, or otherwise, if a cartoon were included. For example, the cortical hem, and fourth ventricle.

Thanks for the suggestion. We have now indicated the structures, ChP, CH and the fourth ventricle, in the images in Figure 1 and Supplementary Figure 1.

(4) In line 115, "as previously shown.." - to keep the paper self-contained a figure illustrating the genetics of the KO allele would be helpful.

Thanks for the suggestion. We have now included an illustration of the Prdm16 cGT allele in Figure 1B.

(5) In Figure 1D as costain for a ChP marker would be helpful because it is hard to identify morphologically in the Prdm16 KO.

Appoligize for the unclarity. The KO allele contains a b-geo reporter driven by Prdm16 endogenous promoter. The samples were co-stained for EdU, b-Gal and DAPI. To distingquish the ChP domain from the CH, we used the presence of b b-Gal as a marker. We indicated this in the figure legend, but now have also clarified this in the revised text.

(6) The details in Figure 1E are hard to see, a zoomed-in inset would help.

A zoomed-in inset is now included in the figure.

(7) Supplementary Figure 2A does not convincingly show that PRDM16 protein is undetectable since endogenous expression may be very low compared to the overexpression PRDM16_E cells so if the contrast is scaled together it could appear black like the KO.

We appreciate the reviewer’s point and have carefully considered this concern. We concluded that PRDM16 protein is effectively undetectable in cultured wild-type NSCs based on direct comparison with brain tissue. Both cultured NSCs and brain sections were processed under similar immunostaining and imaging conditions. While PRDM16 showed robust and specific nuclear localization in embryonic brain sections (Fig. 1B and Supplementary Fig. 1A), only a small subset of cultured NSCs exhibited PRDM16 signal, primarily in the cytoplasm (middle panel of Fig. 2A). This stark contrast supports our conclusion that endogenous PRDM16 protein is either absent or significantly downregulated in vitro. Because of this limitation, we turned to over-expressing Prdm16 in NSC culture using a constitutive promoter. 

(9) Line 182 "Following the washout step" - no such step had been described, maybe replace by "After washout of BMP".

Yes, we have revised the text.

(8) Line 214: "indicating a modest level" - what defines modest? Compared to what? Why is a few thousand moderate rather than low? Does it go to zero with inhibitors for pathways?

Here a modest level means a lower level than to that after adding BMP4. To clarify this, we revised the description to “indicating endogenous levels of …”

(9) The way qPCR data are displayed makes it difficult to appreciate the magnitude of changes, e.g. in Supplementary Figure 2B where a gap is introduced on the scale. Displaying log fold change / relative CT values would be more informative.

We used a segmented Y-axis in Supplementary Figure 2B because the Prdm16 overexpression samples exhibited much higher experssion levels compared to other conditions. In response to this suggestion, we explored alternative ways to present the result, including ploting log-transformed values and log fold changes. However, these methods did not enhance the clarity of the differences – in fact, log scaling made the magnitude of change appear less apparent. To address this, we now present the overexpression samples in a separate graph, thereby eliminating the need for a broken Y-axis and improving the overall readability of the data.

(10) Writing out "3 days" instead of 3D in Figure 2A would improve clarity. It would be good if the used time interval is repeated in other figures throughout the paper so it is still clear the comparison is between 0 and 3 days.

We have changed “3D” to “3 days”. All BMP4 treatments in this study were 3 days.

(11) Line 290: "we found that over 50% of SMAD4 and pSMAD1/5/8 binding peaks were consistent in Prdm16_E and Prdm16_KO cells, indicating that deletion of Prdm16 does not affect the general genomic binding ability of these proteins" - this only makes sense to state with appropriate controls because 50% seems like a big difference, what is the sample to sample variability for the same condition? Moreover, the next paragraph seems to contradict this, ending with "This result suggests that SMAD binding to these sites depends on PRDM16". The authors should probably clarify the writing.

We appreciate the reviwer’s comment and agree that clarification was needed. Our point was that SMAD4 and pSMAD1/5/8 retain the ability to bind DNA broadly in the Prdm16 KO cells, with more than half of the original binding sites still occupied. This suggests that deletion of Prdm16 does not globally impair SMAD genomic binding. Howerever, our primary interest lies in the subset of sites that show differential by SMAD binding between wt and Prdm16 KO conditions, as thse are likely to be PRDM16-dependent. 

In the following paragraph, we focused specifically on describing SMAD and PRDM16 co-bound sites. At these loci, SMAD4 and pSMAD1/5/8 showed reduced enrichment in the absence of PRDM16, suggesting PRDM16 facilitates SMAD binding at these particular regions. We have revised the text in the manuscript to more clearly distinguish between global SMAD binding and PRDM16-dependent sites.

(12) Much more convincing than ChIP-qPCR for c-FOS for two loci in Figures 5F-G would be a global analysis of c-FOS ChIP-seq data.

We agree that a global c-FOS ChIP-seq analysis would provide a more comprehensive view of c-FOS binding patterns. However, the primary focus of this study is the interaction between BMP signaling and PRDM16. The enrichment of AP-1 motifs at ectopic SMAD4 binding sites was an unexpected finding, which we validated using c-FOS ChIP-qPCR at selected loci. While a genome-wide analysis would be valuable, it falls beyond the current scope. We agree that future studies exploring the interplay among SMAD4/pSMAD, PRDM16, and AP-1 will be important and informative.

(13) Figure 6A is hard to read. A heatmap would make it much easier to see differences in expression. Furthermore, if the point is to see the difference between ChP and CH, why not combine the different subclusters belonging to those structures? Finally, why are there 28 genes total when it is said the authors are evaluating a list of 31 genes and also displaying 6 genes that are not expressed (so the difference isn't that unexpressed genes are omitted)?

For the scRNA-seq data, we chose violin plots because they display both gene expression levels and the number of cells that express each gene. However, we agree that the labels in Figure 6A were too small and difficult to read. We have revised the figure by increasing the font size and moved genes with low expression to  Supplementary Figure 5A. Figure 6A includes 17 more highly expressed genes together with three markers, and  Supplementary Figure 5A contains 13 lowly expressed genes. One gene Mrtfb is missing in the scRNA-seq data and thus not included. We have revised the description of the result in the main text and figure legends.

Reviewer #2 (Public review):

Summary:

This article investigates the role of PRDM16 in regulating cell proliferation and differentiation during choroid plexus (ChP) development in mice. The study finds that PRDM16 acts as a corepressor in the BMP signaling pathway, which is crucial for ChP formation.

The key findings of the study are:

(1) PRDM16 promotes cell cycle exit in neural epithelial cells at the ChP primordium.

(2) PRDM16 and BMP signaling work together to induce neural stem cell (NSC) quiescence in vitro.

(3) BMP signaling and PRDM16 cooperatively repress proliferation genes.

(4) PRDM16 assists genomic binding of SMAD4 and pSMAD1/5/8.

(5) Genes co-regulated by SMADs and PRDM16 in NSCs are repressed in the developing ChP.

(6) PRDM16 represses Wnt7b and Wnt activity in the developing ChP.

(7) Levels of Wnt activity correlate with cell proliferation in the developing ChP and CH.

In summary, this study identifies PRDM16 as a key regulator of the balance between BMP and Wnt signaling during ChP development. PRDM16 facilitates the repressive function of BMP signaling on cell proliferation while simultaneously suppressing Wnt signaling. This interplay between signaling pathways and PRDM16 is essential for the proper specification and differentiation of ChP epithelial cells. This study provides new insights into the molecular mechanisms governing ChP development and may have implications for understanding the pathogenesis of ChP tumors and other related diseases.

Strengths:

(1) Combining in vitro and in vivo experiments to provide a comprehensive understanding of PRDM16 function in ChP development.

(2) Uses of a variety of techniques, including immunostaining, RNA in situ hybridization, RT-qPCR, CUT&Tag, ChIP-seq, and SCRINSHOT.

(3) Identifying a novel role for PRDM16 in regulating the balance between BMP and Wnt signaling.

(4) Providing a mechanistic explanation for how PRDM16 enhances the repressive function of BMP signaling. The identification of SMAD palindromic motifs as preferred binding sites for the SMAD/PRDM16 complex suggests a specific mechanism for PRDM16-mediated gene repression.

(5) Highlighting the potential clinical relevance of PRDM16 in the context of ChP tumors and other related diseases. By demonstrating the crucial role of PRDM16 in controlling ChP development, the study suggests that dysregulation of PRDM16 may contribute to the pathogenesis of these conditions.

We thank the reviewer for the thorough and thoughtful summary of our study. We’re glad the key findings and significance of our work were clearly conveyed, particularly regarding the role of PRDM16 in coordinating BMP and Wnt signaling during ChP development. We also appreciate the recognition of our integrated approach and the potential implications for understanding ChP-related diseases.

Weaknesses:

(1) Limited investigation of the mechanism controlling PRDM16 protein stability and nuclear localization in vivo. The study observed that PRDM16 protein became nearly undetectable in NSCs cultured in vitro, despite high mRNA levels. While the authors speculate that post-translational modifications might regulate PRDM16 in NSCs similar to brown adipocytes, further investigation is needed to confirm this and understand the precise mechanism controlling PRDM16 protein levels in vivo.

While mechansims controlling PRDM16 protein stability and nuclear localization in the developing brain are interesting, the scope of this paper is revealing the function of PRDM16 in the choroid plexus and its interaction with BMP signaling. We will be happy to pursuit this direction in our next study.

(2) Reliance on overexpression of PRDM16 in NSC cultures. To study PRDM16 function in vitro, the authors used a lentiviral construct to constitutively express PRDM16 in NSCs. While this approach allowed them to overcome the issue of low PRDM16 protein levels in vitro, it is important to consider that overexpressing PRDM16 may not fully recapitulate its physiological role in regulating gene expression and cell behavior.

As stated above, we acknowledge that findings from cultured NSCs may not directly apply to ChP cells in vivo. We are cautious with our statements. The cell culture work was aimed to identify potential mechanisms by which PRDM16 and SMADs interact to regulate gene expression and target genes co-regulated by these factors. We expect that not all targets from cell culture are regulated by PRDM16 and SMADs in the ChP, so we validated expression changes of several target genes in the developing ChP and now included the new data in Fig. 7 and Supplementary Fig. 7. Out of the 31 genes identified from cultured cells, four cell cycle regulators including Wnt7b, Id3, Spc24/25/nuf2 and Mybl2, showed de-repression in Prdm16 mutant ChP. These genes can be relevant downstream genes in the ChP, and other target genes may be cortical NSC-specific or less dependent on Prdm16 in vivo.

(3) Lack of direct evidence for AP1 as the co-factor responsible for SMAD relocation in the absence of PRDM16. While the study identified the AP1 motif as enriched in SMAD binding sites in Prdm16 knockout cells, they only provided ChIP-qPCR validation for c-FOS binding at two specific loci (Wnt7b and Id3). Further investigation is needed to confirm the direct interaction between AP1 and SMAD proteins in the absence of PRDM16 and to rule out other potential co-factors.

We agree that the finding of the AP1 motif enriched at the PRDM16 and SMAD co-binding regions in Prdm16 KO cells can only indirectly suggest AP1 as a co-factor for SMAD relocation. That’s why we used ChIP-qPCR to examine the presence of C-fos at these sites. Although we only validated two targets, the result confirms that C-fos binds to the sites only in the Prdm16 KO cells but not Prdm16_expressing cells, suggesting AP1 is a co-factor.  Our results cannot rule out the presence of other co-factors.

Reviewer #2 (Recommendations for the authors):

Minor typo: [7, page 3] "sicne" should be "since".

We appreciate the reviewer’s careful reading. We have now corrected the typo and revised some part of the text to improve clarity.

Reviewer #3 (Public review):

Summary:

Bone morphogenetic protein (BMP) signaling instructs multiple processes during development including cell proliferation and differentiation. The authors set out to understand the role of PRDM16 in these various functions of BMP signaling. They find that PRDM16 and BMP co-operate to repress stem cell proliferation by regulating the genomic distribution of BMP pathway transcription factors. They additionally show that PRDM16 impacts choroid plexus epithelial cell specification. The authors provide evidence for a regulatory circuit (constituting of BMP, PRDM16, and Wnt) that influences stem cell proliferation/differentiation.

Strengths:

I find the topics studied by the authors in this study of general interest to the field, the experiments well-controlled and the analysis in the paper sound.

We thank the reviewer for their positive feedback and thoughtful summary. We appreciate the recognition of our efforts to define the role of PRDM16 in BMP signaling and stem cell regulation, as well as the soundness of our experimental design and analysis.

Weaknesses:

I have no major scientific concerns. I have some minor recommendations that will help improve the paper (regarding the discussion).

We have revised the discussion according to the suggestions.

Reviewer #3 (Recommendations for the authors):

Specific minor recommendations:

Page 18. Line 526: In a footnote, the authors point out a recent report which in parallel was investigating the link between PRDM16 and SMAD4. There is substantial non-overlap between these two papers. To aid the reader, I would encourage the authors to discuss that paper in the discussion section of the manuscript itself, highlighting any similarities/differences in the topic/results.

Thanks for the suggestion. We now included the comparison in the discussion. One conclusion between our study and this publication is consistent, that PRDM16 functions as a co-repressor of SMAD4. However, the mechanims are different. Our data suggests a model in which PRDM16 facilitates SMAD4/pSMAD binding to repress proliferation genes under high BMP conditions. However, the other report suggests that SMAD4 steadily binds to Prdm16 promoter and switches regulatory functions depending on the co-factors. Together with PRDM16, SMAD4 represses gene expression, while with SMAD3 in response to high levels of TGF-b1, it activates gene expression. These differences could be due to different signaling (BMP versus TGF-b), contexts (NSCs versus Pancreatic cancers) etc.

Page 3. Line 65: typo 'since'

We appreciate the reviewer’s careful reading. We have now corrected the typo and revised the text to improve clarity.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public Review):

This manuscript describes a series of experiments documenting trophic egg production in a species of harvester ant, Pogonomyrmex rugosus. In brief, queens are the primary trophic egg producers, there is seasonality and periodicity to trophic egg production, trophic eggs differ in many basic dimensions and contents relative to reproductive eggs, and diets supplemented with trophic eggs had an effect on the queen/worker ratio produced (increasing worker production).

The manuscript is very well prepared and the methods are sufficient. The outcomes are interesting and help fill gaps in knowledge, both on ants as well as insects, more generally. More context could enrich the study and flow could be improved.

We thank the reviewer for these comments. We agree that the paper would benefit from more context. We have therefore greatly extended the introduction.

Reviewer #2 (Public Review):

The manuscript by Genzoni et al. provides evidence that trophic eggs laid by the queen in the ant Pogonomyrmex rugosis have an inhibitory effect on queen development. The authors also compare a number of features of trophic eggs, including protein, DNA, RNA, and miRNA content, to reproductive eggs. To support their argument that trophic eggs have an inhibitory effect on queen development, the authors show that trophic eggs have a lower content of protein, triglycerides, glycogen, and glucose than reproductive eggs, and that their miRNA distributions are different relative to reproductive eggs. Although the finding of an inhibitory influence of trophic eggs on queen development is indeed arresting, the egg cross-fostering experiment that supports this finding can be effectively boiled down to a single figure (Figure 6). The rest of the data are supplementary and correlative in nature (and can be combined), especially the miRNA differences shown between trophic and reproductive eggs. This means that the authors have not yet identified the mechanism through which the inhibitory effect on queen development is occurring. To this reviewer, this finding is more appropriate as a short report and not a research article. A full research article would be warranted if the authors had identified the mechanism underlying the inhibitory effect on queen development. Furthermore, the article is written poorly and lacks much background information necessary for the general reader to properly evaluate the robustness of the conclusions and to appreciate the significance of the findings.

We thank the reviewer for these comments. We agree that the paper would benefit by having more background information and more discussion. We have followed this advice in the revision.

Reviewer #3 (Public Review):

In "Trophic eggs affect caste determination in the ant Pogonomyrmex rugosus" Genzoni et al. probe a fundamental question in sociobiology, what are the molecular and developmental processes governing caste determination? In many social insect lineages, caste determination is a major ontogenetic milestone that establishes the discrete queen and worker life histories that make up the fundamental units of their colonies. Over the last century, mechanisms of caste determination, particularly regulators of caste during development, have remained relatively elusive. Here, Genzoni et al. discovered an unexpected role for trophic eggs in suppressing queen development - where bi-potential larvae fed trophic eggs become significantly more likely to develop into workers instead of gynes (new queens). These results are unexpected, and potentially paradigm-shifting, given that previously trophic eggs have been hypothesized to evolve to act as an additional intracolony resource for colonies in potentially competitive environments or during specific times in colony ontogeny (colony foundation), where additional food sources independent of foraging would be beneficial. While the evidence and methods used are compelling (e.g., the sequence of reproductive vs. trophic egg deposition by single queens, which highlights that the production of trophic eggs is tightly regulated), the connective tissue linking many experiments is missing and the downstream mechanism is speculative (e.g., whether miRNA, proteins, triglycerides, glycogen levels in trophic eggs is what suppresses queen development). Overall, this research elevates the importance of trophic eggs in regulating queen and worker development but how this is achieved remains unknown.

We thank the reviewer for these comments and agree that future work should focus on identifying the substances in trophic eggs that are responsible for caste determination.  

Recommendations For The Authors:

Reviewer #1 (Recommendations For The Authors):

Introduction:

The context for this study is insufficiently developed in the introduction - it would be nice to have a more detailed survey of what is known about trophic eggs in insects, especially social insects. The end of the introduction nicely sets up the hypothesis through the prior work described by Helms Cahan et al. (2011) where they found JH supplementation increased trophic egg production and also increased worker size. I think that the introduction could give more context about egg production in Pogonomyrmex and other ants, including what is known about worker reproduction. For example, Suni et al. 2007 and Smith et al. 2007 both describe the absence of male production by workers in two different harvester ants. Workers tend to have underdeveloped ovaries when in the presence of the queen. Other species of ants are known to have worker reproduction seemingly for the purpose of nutrition (see Heinze and Hölldober 1995 and subsequent studies on Crematogaster smithi). Because some ants, including Pogonomyrmex, lack trophallaxis, it has been hypothesized that they distribute nutrients throughout the nest via trophic eggs as is seen in at least one other ant (Gobin and Ito 2000). Interestingly, Smith and Suarez (2009) speculated that the difference in nutrition of developing sexual versus worker larvae (as seen in their pupal stable isotope values) was due to trophic egg provisioning - they predicted the opposite as was found in this study, but their prediction was in line with that of Helms Cahan et al. (2011). This is all to say that there is a lot of context that could go into developing the ideas tested in this paper that is completely overlooked. The inclusion of more of what is known already would greatly enrich the introduction.

We agree that it would be useful to provide a larger context to the study. We now provide more information on the life-history of ants and explained under what situations queens and workers may produce trophic eggs. We also mentioned that some ants such as Crematogaster smithi have a special caste of “large workers” which are morphologically intermediate between winged queens and small workers and appear to be specialized in the production of unfertilized eggs. We now also mention the study of Goby and Ito (200) where the authors show that trophic eggs may play an important role in food distribution withing the colony, in particular in species where trophallaxis is rare or absent.

Methods:

L49: What lineage is represented in the colonies used? The collection location is near where both dependent-lineage (genetic caste determining) P. rugosus and "H" lineage exist. This is important to know. Further, depending on what these are, the authors should note whether this has relevance to the study. Not mentioning genetic caste determination in a paper that examines caste determination is problematic.

This is a good point. We have now provided information at the very beginning of the material and method section that the queens had been collected in populations known not to have dependentlineage (genetic caste determining) mechanisms of caste determination.

L63 and throughout: It would be more efficient to have a paragraph that cites R (must be done) and RStudio once as the tool for all analyses. It also seems that most model construction and testing was done using lme4 - so just lay this out once instead of over and over.

We agree and have updated the manuscript accordingly.

L95: 'lenght' needs to be 'length' in the formula.

Thanks, corrected.

L151: A PCA was used but not described in the methods. This should be covered here. And while a Mantel test is used, I might consider a permANOVA as this more intuitively (for me, at least) goes along with the PCA.

We added the PCA description in the Material and Method section.

Results:

I love Fig. 3! Super cool.

Thanks for this positive comment.

Discussion:

It would be good to have more on egg cannibalism. This is reasonably well-studied and could be good extra context.

We have added a paragraph in the discussion to mention that egg cannibalism is ubiquitous in ants.

Supp Table 1: P. badius is missing and citations are incorrectly attributed to P. barbatus.

P. badius was present in the Table but not with the other Pogonomyrmex species. For some genera the species were also not listed in alphabetic order. This has been corrected.

Reviewer #2 (Recommendations For The Authors):

Comments on introduction:

The introduction is missing information about caste determination in ants generally and Pogonomyrmex rugosis specifically. This is important because some colonies of Pogonomyrmex rugosis have been shown to undergo genetic caste determination, in which case the main result would be rendered insignificant. What is the evidence that caste determination in the lineages/colonies used is largely environmentally influenced and in what contexts/environmental factors? All of this should be made clear.

This is a good point. We have expanded the introduction to discuss previous work on caste determination in Pogonomyrmex species with environmental caste determination and now also provide evidence at the beginning of the Material and Method section that the two populations studied do not have a system of genetic caste determination.

Line 32 and throughout the paper: What is meant exactly by 'reproductive eggs'? Are these eggs that develop specifically into reproductives (i.e., queens/males) or all eggs that are non-trophic? If the latter, then it is best to refer to these eggs as 'viable' in order to prevent confusion.

We agree and have updated the manuscript accordingly.

Figure 1/Supp Table 1: It is surprising how few species are known to lay trophic eggs. Do the authors think this is an informative representation of the distribution of trophic egg production across subfamilies, or due to lack of study? Furthermore, the branches show ant subfamilies, not families. What does the question mark indicate? Also, the information in the table next to the phylogeny is not easy to understand. Having in the branches that information, in categories, shown in color for example, could be better and more informative. Finally, having the 'none' column with only one entry is confusing - discuss that only one species has been shown to definitely not lay trophic eggs in the text, but it does not add much to the figure.

Trophic eggs are probably very common in ants, but this has not been very well studied. We added a sentence in the manuscript to make this clear.

Thanks for noticing the error family/subfamily error. This has been corrected in Figure 1 and Supplementary Table 1.

The question mark indicates uncertainty about whether queens also contribute to the production of trophic eggs in one species (Lasius niger). We have now added information on that in the Figure legend.

We agree with the reviewer that it would be easier to have the information on whether queens and workers produce trophic on the branches of the Tree. However, having the information on the branches would suggest that the “trait” evolved on this part of the tree. As we do not know when worker or queen production of trophic eggs exactly evolved, we prefer to keep the figure as it is.

Finally, we have also removed the none in the figure as suggested by the reviewer and discussed in the manuscript the fact that the absence of trophic eggs has been reported in only one ant species (Amblyopone silvestrii: Masuko 2003_)._

Comments on materials and methods:

Why did they settle on three trophic eggs per larva for their experimental setup?

We used three trophic eggs because under natural conditions 50-65% of the eggs are trophic. The ratio of trophic eggs to viable eggs (larvae) was thus similar natural condition.

Line 50: In what kind of setup were the ants kept? Plaster nests? Plastic boxes? Tubes? Was the setup dry or moist? I think this information is important to know in the context of trophic eggs.

We now explain that colonies were maintained in plastic boxes with water tubes.

Line 60: Were all the 43 queens isolated only once, or multiple times?

Each of the 43 queens were isolated for 8 hours every day for 2 weeks, once before and once after hibernation (so they were isolated multiple times). We have changed the text to make clear that this was done for each of the 43 queens.

Could isolating the queen away from workers/brood have had an effect on the type of eggs laid?

This cannot be completely ruled out. However, it is possible to reliably determine the proportion of viable and trophic eggs only by isolating queens. And importantly the main aim of these experiments was not to precisely determine the proportion viable and trophic eggs, but to show that this proportion changes before and after hibernation and that queens do not lay viable and trophic eggs in a random sequence.

Since it was established that only queens lay trophic eggs why was the isolation necessary?

Yes this was necessary because eggs are fragile and very difficult to collect in colonies with workers (as soon as eggs are laid they are piled up and as soon as we disturb the nest, a worker takes them all and runs away with them). Moreover, it is possible that workers preferentially eat one type of eggs thus requiring to remove eggs as soon as queens would have laid them. This would have been a huge disturbance for the colonies.

Line 61: Is this hibernation natural or lab induced? What is the purpose of it? How long was the hibernation and at what temperature? Where are the references for the requirement of a diapause and its length?

The hibernation was lab induced. We hibernated the queens because we previously showed that hibernation is important to trigger the production of gynes in P. rugosus colonies in the laboratory (Schwander et al 2008; Libbrecht et al 2013). Hibernation conditions were as described in Libbrecht et al (2013).  

Line 73: If the queen is disturbed several times for three weeks, which effect does it have on its egg-laying rate and on the eggs laid? Were the eggs equally distributed in time in the recipient colonies with and without trophic eggs to avoid possible effects?

It is difficult to respond what was the effect of disturbance on the number and type of eggs laid. But again our aim was not to precisely determine these values but determine whether there was an effect of hibernation on the proportion of trophic eggs. The recipient colonies with and without trophic eggs were formed in exactly the same way. No viable eggs were introduced in these colonies, but all first instar larvae have been introduced in the same way, at the same time, and with random assignment. We have clarified this in the Material and Method section.

Line 77: Before placing the freshly hatched larvae in recipient colonies, how long were the recipient colonies kept without eggs and how long were they fed before giving the eggs? Were they kept long enough without the queen to avoid possible effects of trophic eggs, or too long so that their behavior changed?

The recipient colonies were created 7 to 10 days before receiving the first larvae and were fed ad libitum with grass seeds, flies and honey water from the beginning. Trophic eggs that would have been left over from the source colony should have been eaten within the first few days after creating the recipient colonies. However, even if some trophic eggs would have remained, this would not influence our conclusion that trophic eggs influence caste fate, given the fully randomized nature of our treatments and the considerable number of independent replicates. The same applies to potential changes in worker behavior following their isolation from the queen.

Line 77: Is it known at what stage caste determination occurs in this species? Here first instar larvae were given trophic eggs or not. Does caste-determination occur at the first instar stage? If not, what effect could providing trophic eggs at other stages have on caste-determination?

A previous study showed that there is a maternal effect on caste determination in the focal species (Schwander et al 2008). The mechanism underlying this maternal effect was hypothesized to be differential maternal provisioning of viable eggs. However, as we detail in the discussion, the new data presented in our study suggests that the mechanism is in fact a different abundance of trophic eggs laid by queens. There is currently no information when exactly caste determination occurs during development

Comments on results:

Line 65: How does investigating the order of eggs laid help to "inform on the mechanisms of oogenesis"?

We agree that the aim was not to study the mechanism of oogenesis. We have changed this sentence accordingly: “To assess whether viable and trophic eggs were laid in a random order, or whether eggs of a given type were laid in clusters, we isolated 11 queens for 10 hours, eight times over three weeks, and collected every hour the eggs laid”

Figure 2: There is no description/discussion of data shown in panels B, C, E, and F in the main text.

We have added information in the main text that while viable eggs showed embryonic development at 25 and 65 hours (Fig 12 B, C) there was no such development for trophic eggs (Fig. 2 E,F).

Line 172: Please explain hibernation details and its significance on colony development/life cycle.

We have added this information in the Material and Method section.

Figure 6: How is B plotted? How could 0% of gynes have 100% survival?

The survival is given for the larvae without considering caste. We have changed the de X axis of panel B and reworded the Figure legend to clarify this.

Is reduced DNA content just an outcome of reduced cell number within trophic eggs, i.e., was this a difference in cell type or cell number? Or is it some other adaptive reason?

It is likely to be due to a reduction in cell number (trophic eggs have maternal DNA in the chorion, while viable eggs have in addition the cells from the developing zygote) but we do not have data to make this point.

Is there a logical sequence to the sequence of egg production? The authors showed that the sequence is non-random, but can they identify in what way? What would the biological significance be?

We could not identify a logical sequence. Plausibly, the production of the two types of eggs implies some changes in the metabolic processes during egg production resulting in queens producing batches of either viable or trophic eggs. This would be an interesting question to study, but this is beyond the scope of this paper.

Figure 6b is difficult to follow, and more generally, legends for all figures can be made clearer and more easy to follow.

We agree. We have now improved the legends of Fig 6B and the other figures.

Lines 172-174: "The percentage of eggs that were trophic was higher before hibernation...than after. This higher percentage was due to a reduced number of reproductive eggs, the number of trophic eggs laid remained stable" - are these data shown? It would be nice to see how the total egglaying rate changes after hibernation. Also, is the proportion of trophic eggs laid similar between individual queens?

No the data were not shown and we do not have excellent data to make this point. We have therefore removed the sentence “This higher percentage was due to a reduced number of reproductive eggs, the number of trophic eggs laid remained stable” from the manuscript.

Figure 6B: Do several colonies produce 100% gynes despite receiving trophic eggs? It would be interesting if the authors discussed why this might occur (e.g., the larvae are already fully determined to be queens and not responsive to whatever signal is in the trophic eggs).

The reviewer is correct that 4 colonies produced 100% gynes despite receiving trophic eggs. However, the number of individuals produced in these four colonies was small (2,1,2,1, see supplementary Table 2). So, it is likely that it is just by chance that these colonies produced only gynes.

Figure 5: Why a separation by "size distribution variation of miRNA"? What is the relevance of looking at size distributions as opposed to levels?

We did that because there many different miRNA species, reflected by the fact that there is not just one size peak but multiple one. This is why we looked at size distribution

Figure 2: The image of the viable embryo is not clear. If possible, redo the viable to show better quality images.

Unfortunately, we do not anymore have colonies in the laboratory so this is not possible.

Comments on discussion:

Lines 236-247: Can an explanation be provided as to why the effect of trophic eggs in P. rugosus is the opposite of those observed by studies referenced in this section? Could P. rugosus have any life history traits that might explain this observation?

In the two mentioned studies there were other factors that co-varied with variation in the quantity of trophic eggs. We mentioned that and suggested that it would be useful to conduct experimental manipulation of the quantity of trophic eggs in the Argentine ant and P. barbatus (the two species where an effect of trophic eggs had been suggested).

The discussion should include implications and future research of the discovery.

We made some suggestions of experiments that should be performed in the future

The conclusion paragraph is too short and does not represent what was discussed.

We added two sentences at the end of the paragraph to make suggestions of future studies that could be performed.

Lines 231 to 247: Drastically reduce and move this whole part to the introduction to substantiate the assumption that trophic eggs play a nutritional role.

We moved most of this paragraph to the introduction, as suggested by the reviewer.

Reviewer #3 (Recommendations For The Authors):

I would like to commend the authors on their study. The main findings of the paper are individually solid and provide novel insight into caste determination and the nature of trophic eggs. However, the inferences made from much of the data and connections between independent lines of evidence often extend too far and are unsubstantiated.

We thank the reviewer for the positive comment. We made many changes in the manuscript to improve the discussion of our results.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In the manuscript submission by Zhao et al. entitled, "Cardiac neurons expressing a glucagon-like receptor mediate cardiac arrhythmia induced by high-fat diet in Drosophila" the authors assert that cardiac arrhythmias in Drosophila on a high-fat diet are due in part to adipokinetic hormone (Akh) signaling activation. High-fat diet induces Akh secretion from activated endocrine neurons, which activate AkhR in posterior cardiac neurons. Silencing or deletion of Akh or AkhR blocks arrhythmia in Drosophila on a high-fat diet. Elimination of one of two AkhR-expressing cardiac neurons results in arrhythmia similar to a high-fat diet.

Strengths:

The authors propose a novel mechanism for high-fat diet-induced arrhythmia utilizing the Akh signaling pathway that signals to cardiac neurons.

Weaknesses:

Major comments:

(1) The authors state, "Arrhythmic pathology is rooted in the cardiac conduction system." This assertion is incorrect as a blanket statement on arrhythmias. There are certain arrhythmias that have been attributable to the conduction system, such as bradycardic rhythms, heart block, sinus node reentry, inappropriate sinus tachycardia, AV nodal reentrant tachycardia, bundle branch reentry, fascicular ventricular tachycardia, or idiopathic ventricular fibrillation to name a few. However the etiological mechanism of many atrial and ventricular arrhythmias, such as atrial fibrillation or substrate-based ventricular tachycardia, are not rooted in the conduction system. The introduction should be revised to reflect a clear focus (away from?) on atrial fibrillation (AF). In addition, AF susceptibility is known to be modulated by autonomic tone, which is topically relevant (irrelevant?) to this manuscript.

Thank you for the helpful comment. We rephrased the sentence as “Arrhythmic pathology is often rooted in the cardiac conduction system”.

(2) The authors state that "HFD led to increased heartbeat and an irregular rhythm." In representative examples shown, HFD resulted in pauses, slower heart rate, and increased irregularity in rhythm but not consistently increased heart rate (Figures 1B, 3A, and 4C). Based on the cited work by Ocorr et al (https://doi.org/10.1073/pnas.0609278104), Drosophila heart rate is highly variable with periods of fast and slow rates, which the authors attributed to neuronal and hormonal inputs. Ocorr et al then describe the use of "semi-intact" flies to remove autonomic input to normalize heart rate. Were semi-intact flies used? If not, how was heart rate variability controlled? And how was heart rate "increase" quantified in high-fat diet compared to normal-fat diet? Lastly, how does one measure "arrhythmia" when there is so much heart rate variability in normal intact flies?

We also observed that fly heart rate is highly variable with periods of fast and slow rates. To control heart rate variability, Ocorr et al. used semi-intact flies to record the heartbeat  (https://doi.org/10.1073/pnas.0609278104). We consider it a rigorous method to get highly consistent results with high quality videos/images. Since our work has a focus on the neuronal inputs to the heart, we did not use the semi-intact method. Our concern is that it is likely to disrupt the neuronal processes during the dissection. Using OCT, we recorded the heartbeat of intact flies in an 8 s time window, when the heartbeat was relatively stable. The different groups of flies, which were fed on a high-fat diet or a normal-fat diet, were recorded using the same method. Thus, we could compare the differences in heart rate.

(3) The authors state, "to test whether the HFD-induced increase in Akh in the APC affects APC neuron activity, we used CaLexA (https://doi.org/10.3109/01677063.2011.642910)." According to the reference, CaLexA is a tool to map active neurons and would not indicate, as the authors state, whether Akh affects APC neuron activity specifically. It is equally possible that APC neurons may be activated by HFD and produce more Akh. Please clarify this language.

Thank you for clarifying the calcium reporter, CaLexA. We rephrased this sentence to “to test whether HFD affects APC neuron activity, we used CaLexA”.

(4) Are the AkhR+ neurons parasympathetic or sympathetic? Please provide additional experimentation that characterizes these neurons. The AkhR+ neurons appear to be anti-arrhythmic. Please expand the discussion to include a working hypothesis of the overall findings on Akh, AkhR, and AkhR+ neurons.

Noyes et al. showed that Akh treatment increases heartbeat (Noyes, B. E., F. N. Katz, and M. H. Schaffer. 1995. “Identification and Expression of the Drosophila Adipokinetic Hormone Gene.” Molecular and Cellular Endocrinology 109 (2): 133–41.), suggesting that AkhR+ neurons are sympathetic. We showed that high-fat diet induced Akh expression and secretion, which led to stimulation of AkhR+ neuron and increased heart rate, supporting the sympathetic role of the AkhR+ neurons. Additional explanation on the sympathetic & anti-arrhythmic role of the Akh, AkhR, and AkhR+ neurons were added to the discussion.

(5) The authors state, "Heart function is dependent on glucose as an energy source." However, the heart's main energy source is fatty acids with minimal use of glucose (doi: 10.1016/j.cbpa.2006.09.014). Glucose becomes more utilized by cardiomyocytes under heart failure conditions. Please amend/revise this statement.

Thank you for pointing this out and providing the reference. We rephrased this sentence “Heart function is dependent on continuous ATP production. Cardiac ATP in Drosophila might come from fatty acids, glucose, and lactate (Kodde et al., 2007), as well as trehalose.”

Reviewer #2 (Public Review):

This manuscript explores mechanisms underlying heart contractility problems in metabolic disease using Drosophila as a model. They confirm, as others have demonstrated, that a high-fat diet (HFD) induces cardiac problems in flies. They showed that a high-fat diet increased Akh mRNA levels and calcium levels in the Akh-producing cells (APC), suggesting there is increased production and release of this hormone in a HFD context. When they knock down Akh production in the APCs using RNAi they see that cardiac contractility problems are abolished. They similarly show that levels of the Akh receptor (Akhr) are increased on a HFD and that loss of Akhr also rescues contractility problems on a HFD.

One highlight of the paper was the identification of a pair of neurons that express a receptor for the metabolic hormone Akh, and showing initial data that these neurons innervate the cardiac muscle. They then overexpress cell death gene reaper (rpr) in all Akhr-positive cells with Akhr-GAL4 and see that cardiac contractility becomes abnormal.

However, this paper contains several findings that have been reported elsewhere and it contains key flaws in both experimental design and data interpretation. There is some rationale for doing the experiments, and the data and images are of good quality. However, others have shown that HFD induces cardiac contractility problems (Birse 2010), that Akh mRNA levels are changed with HFD (Liao 2021) that Akh modulates cardiac rhythms (Noyes 1995), so Figures 1-4 are largely a confirmation of what is already known. This limits the overall magnitude of the advances presented in these figures. Overall, the stated concerns limit the impact of the manuscript in advancing our understanding of heart contractility.

We thank the reviewer for the positive comments and appreciate the reviewer for the instructive suggestions. Birse 2010 (PMID: 21035763) was cited in our manuscript. Liao 2021 showed that Akh mRNA levels are changed with HFD. We added the reference to the revised manuscript and modified the text as: “In consistent with a previous work (Liao et al., 2020), we showed that the expression of Akh was significantly up-regulated in the flies fed a HFD, compared to NFD-fed flies (Figure 2B)”. Our qPCR verified Liao’s results. On top of this, we investigated the calcium levels in the Akh producing cells (APCs) and showed elevated calcium levels in the APC in HFD fed flies. In the revised version, we added more data to show that Akh protein levels were increased with HFD (Figure 2E-F). In line with Noyes' discovery, which showed that Akh injection caused cardioaccelation in prepupae, we showed that genetic manipulation of Akh expression affected heartbeat in the adults.   

Reviewer #3 (Public Review):

Zhao et al. provide new insights into the mechanism by which a high-fat diet (HFD) induces cardiac arrhythmia employing Drosophila as a model. HFD induces cardiac arrhythmia in both mammals and Drosophila. Both glucagon and its functional equivalent in Drosophila Akh are known to induce arrhythmia. The study demonstrates that Akh mRNA levels are increased by HFD and both Akh and its receptor are necessary for high-fat diet-induced cardiac arrhythmia, elucidating a novel link. Notably, Zhao et al. identify a pair of AKH receptor-expressing neurons located at the posterior of the heart tube. Interestingly, these neurons innervate the heart muscle and form synaptic connections, implying their roles in controlling the heart muscle. The study presented by Zhao et al. is intriguing, and the rigorous characterization of the AKH receptor-expressing neurons would significantly enhance our understanding of the molecular mechanism underlying HFD-induced cardiac arrhythmia.

Many experiments presented in the manuscript are appropriate for supporting the conclusions while additional controls and precise quantifications should help strengthen the authors' augments. The key results obtained by loss of Akh (or AkhR) and genetic elimination of the identified AkhR-expressing cardiac neurons do not reconcile, complicating the overall interpretation.

It is intriguing to see an increase in Akh mRNA levels in HFD-fed animals. This is a key result for linking HFD-induced arrhythmia to Akh. Thus, demonstrating that HFD also increases the Akh protein levels and Akh is secreted more should significantly strengthen the manuscript.

Thank you for the positive comments and the instructive suggestions. We performed immunostaining to show that Akh protein levels increased, which is consistent with elevated Akh mRNA expression in HFD-fed flies. The data was added to Figure 2, panels E and F. Akh secretion from the APCs is regulated by APC activity (https://doi.org/10.1038/s41586-019-1675-4). We used a calcium reporter CaLexA (https://doi.org/10.3109/01677063.2011.642910) to monitor APC activity and showed that HFD increased APC activity (Figure 2, C-D).

The experiments employing an AkhR null allele nicely demonstrate its requirement for HFD-induced cardiac arrhythmia. Depletion of Akh in Akh-expressing cells recapitulates the consequence of AkhR knockout, supporting that both Akh and its receptor are required for HFD-induced cardiac arrhythmia. Given that RNAi is associated with off-target effects and some RNAi reagents do not work, testing multiple independent RNAi lines is the standard procedure. It is also important to show the on-target effect of the RNAi reagents used in the study.

Indeed, RNAi approaches can suffer from off-target effects. For Akh experiments, we used an RNAi line BL_34960, which was generated using artificial microRNAs shRNA (DOI: 10.1038/nmeth.1592). In comparison to long-hairpin constructs, shRNA constructs are expected to be advantageous, e.g., more efficient and minimized off-target. We performed immunostaining to determine Akh-Gal4>UAS-Akh-RNAi efficiency. We showed that anti-Akh fluorescence diminished in Akh-Gal4>UAS-Akh-RNAi APCs. The data was added to Figure 3-figure supplement 1.

The most exciting result is the identification of AkhR-expressing neurons located at the posterior part of the heart tube (ACNs). The authors attempted to determine the function of ACNs by expressing rpr with AkhR-GAL4, which would induce cell death in all AkhR-expressing cells, including ACNs. The experiments presented in Figure 6 are not straightforward to interpret. Moreover, the conclusion contradicts the main hypothesis that elevated Akh is the basis of HFD-induced arrhythmia. The results suggest the importance of AkhR-expressing cells for normal heartbeat. However, elimination of Akh or AkhR restores normal rhythm in HFD-fed animals, suggesting that Akh and AkhR are not important for maintaining normal rhythms. If Akh signaling in ACNs is key for HFD-induced arrhythmia, genetic elimination of ACNs should unalter rhythm and rescue the HFD-induced arrhythmia. An important caveat is that the experiments do not test the specific role of ACNs. ACNs should be just a small part of the cells expressing AkhR. The experiments presented in Figure 6 cannot justify the authors' conclusion. Specific manipulation of ACNs will significantly improve the study. Moreover, the main hypothesis suggests that HFD may alter the activity of ACNs in a manner dependent on Akh and AkhR. Testing how HFD changes calcium, possibly by CaLexA (Figure 2) and/or GCaMP, in wild-type and AkhR mutants could be a way to connect ACNs to HFD-induced arrhythmia. Moreover, optogenetic manipulation of ACNs will allow for specific manipulation of ACNs, which is crucial for studying the specific role of ACNs in controlling cardiac rhythms.

Thank you for the insightful comments. We have been trying to find a way to only target the AkhR neurons using split-Gal4. Up to now, it’s not successful. Akh/AkhR signaling shall play a key role in the ACNs, however, we cannot rule out the possibility that ACNs also receive signals other than Akh in the modulation of heartbeat.

Interestingly, expressing rpr with AkhR-GAL4 was insufficient to eliminate both ACNs. It is not clear why it didn't eliminate both ACNs. Given the incomplete penetrance, appropriate quantifications should be helpful. Additionally, the impact on other AhkR-expressing cells should be assessed. Adding more copies of UAS-rpr, AkhR-GAL4, or both may eliminate all ACNs and other AkhR-expressing cells. The authors could also try UAS-hid instead of UAS-rpr.

We added more data to show that AkhR+ neurons are positive in anti-Akh staining, indicating the AkhR+ neurons indeed receive Akh.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Typo in line 765: "increased Akh section into the circulation." Section should be secretion.

Thank you for finding the typo. We changed section to secretion.

Reviewer #2 (Recommendations For The Authors):

One interesting extension to our knowledge in Figures 3 & 4 is that loss of Akhr and loss of Akh both block the cardiac contractility defects that accompany a HFD. The main concern I have with the Akh finding is that the authors use only a GAL4 control and no UAS alone control. Metabolic phenotypes often show strain-specific effects, so to make conclusions it is essential that the authors include a UAS alone control alongside the other genotypes to be sure it does not rescue the cardiac contractility defects that accompany a HFD by itself.

I am interested in the authors' identification of a pair of Akhr-positive neurons that innervate the cardiac muscle. I am not aware of any other studies identifying these neurons, or revealing their function. The contents of Figure 5 therefore represent the largest advance in the study. However, the characterization of these neurons is very superficial, and a lot more work to understand their regulation and function in a HFD context is needed to make conclusions about their role in any HFD-induced cardiac contractility problems. Or to determine how Akh influences the function of these specific neurons in an HFD context.

The reason I say this is that the authors ablate all Akhr-positive cells in Figure 6 and show that this disturbs normal cardiac contractility. While studies on the one pair of Akhr-positive neurons would be really interesting, ablating all Akhr-positive cells, which includes the fat and many other cell types in the fly, is not a scientifically rigorous approach to answering this question. As a result, the authors are only able to make the claim that ablating many cell types throughout the animal disrupts cardiac contractility, which does not advance our understanding of mechanisms underlying heart contractility problems. In addition, because the experiments they designed did not test whether it was Akh binding to Akhr on those neurons that regulate cardiac contractility problems in a HFD context, their experiments do not support their model in Figure 7.

The authors also make conclusions that are fairly speculative around Line 231 when describing their model in Figure 7. These claims are simply not supported by the data they present and must be removed. For example, the authors have not identified an endocrine-heart axis, they simply showed that changes in Akh can influence the heart, but this is not necessarily a direct effect on a specific cell type. They do not show data that Akh binds the newly identified Akhr-positive neuron pair to mediate the effects of HFD-induced contractility defects - they just ablate all Akhr-positive cells (fat, neurons, and other types) and show cardiac defects. If those neurons did mediate the abnormal cardiac rhythm promoted by Akh, then ablating those neurons (and not a large number of additional tissues) should rescue HFD-induced heart defects just like reducing Akhr or Akh did (but this is the opposite of what they see). Overall, concerns with experimental design, data interpretation, and relatively few findings that aren't reported elsewhere reduce the impact of this paper.

We appreciate the positive comments and helpful suggestions. Indeed, it is important to get clean genetic access to the cardiac neurons. We intended to use split Gal4 system to target the AkhR cardiac neurons. We have tried to build a split Gal4 driver AkhR-p65.AD. Two rounds of injection were carried out. However, we did not recover a transgenic line.

In the revised version, we performed immunostaining using Akh antibodies to show that anti-Akh fluorescence was observed in AkhR neurons (Figure 5-figure supplement 1), indicating an endocrine-heart axis.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript by Duilio M. Potenza et al. explores the role of Arginase II in cardiac aging, majorly using whole-body arg-ii knock-out mice. In this work, the authors have found that Arg-II exerts non-cell-autonomous effects on aging cardiomyocytes, fibroblasts, and endothelial cells mediated by IL-1b from aging macrophages. The authors have used arg II KO mice and an in vitro culture system to study the role of Arg II. The authors have also reported the cell-autonomous effect of Arg-II through mitochondrial ROS in fibroblasts that contribute to cardiac aging. These findings are sufficiently novel in cardiac aging and provide interesting insights. While the phenotypic data seems strong, the mechanistic details are unclear. How Arg II regulates the IL-1b and modulates cardiac aging is still being determined. The authors still need to determine whether Arg II in fibroblasts and endothelial contributes to cardiac fibrosis and cell death. This study also lacks a comprehensive understanding of the pathways modulated by Arg II to regulate cardiac aging.

We sincerely appreciate the valuable feedback provided by the reviewer. It's gratifying to hear that our work provided novel information on the role of arginase-II in cardiac aging which is a complex process involving various cell types and mechanisms. We have devoted considerable effort by performing new experiments to address the reviewer's comments and to delineate more detailed mechanisms of Arg-II in cardiac aging. Please, see below our specific answers to each point of the reviewers.

Strengths:

This study provides interesting information on the role of Arg II in cardiac aging.

The phenotypic data in the arg II KO mice is convincing, and the authors have assessed most of the aging-related changes.

The data is supported by an in vitro cell culture system.

We appreciate this reviewer’s positive assessment on the strength of our study.

Weaknesses:

The manuscript needs more mechanistic details on how Arg II regulates IL-1b and modulates cardiac aging.

We made great effort and have performed new experiments in human monocyte cell line (THP1) in which iNOS is not expressed and not inducible by LPS and arg-ii gene was knocked out by CRISPR technology. Moreover, murine bone-marrow derived macrophages in which inos gene was ablated, is also use for this purpose. We found that in the human THP1 monocytes in which Arg-II but not iNOS is induced by LPS (100 ng/mL for 24 hours) (Suppl. Fig. 6A), mRNA and protein levels of IL-1b precursor are markedly reduced in arg-ii knockout THP1^{arg-ii^-/-} as compared to the THP1^wt cells (Suppl. Fig. 6B and 6C), further confirming that Arg-II promotes IL-1b production as also shown in RAW264.7 macrophages (Suppl. Fig. 5A and 5C). Moreover, in the mouse bone-marrow-derived macrophages, LPS-induced IL-1b production is inhibited by inos deficiency (BMDM^inos-/- vs BMDM^wt) (Suppl. Fig. 6D and 6E), while Arg-II levels are slightly enhanced in the BMDM^inos-/- cells (Suppl. Fig. 6D and 6F). All together, these results suggest that iNOS slightly reduces Arg-II expression. Arg-II and iNOS can be upregulated by LPS independently. Both Arg-II and iNOS are required for IL-1b production upon LPS stimulation as illustrated in Suppl. Fig. 6G. For detailed results and discussion, please see answers to the comments point 2 or point 6 raised by this reviewer.

The authors used whole-body KO mice, and the role of macrophages in cardiac aging is not studied in this model. A macrophage-specific arg II Ko would be a better model.

We fully agree with this comment of the reviewer. Unfortunately, this macrophage specific arg-ii knockout animal model is not available, yet. Future research shall develop the macrophage-specific arg-ii^-/- mouse model to confirm this conclusion with aging animals. Since Arg-II is also expressed in fibroblasts and endothelial cells and exerts cell-autonomous and paracrine functions, aging mouse models with conditional arg-ii knockout in the specific cell types would be the next step to elucidate cell-specific function of Arg-II in cardiac aging. We have pointed out this aspect for future research on page 19, lines 2 to 6.

Experiments need to validate the deficiency of Arg II in cardiomyocytes.

As pointed out by this reviewer in the comment point 10, Arg-II was previously reported to be expressed in isolated cardiomyocytes from in rats (PMID: 16537391). Unfortunately, negative controls. i.e., arg-ii^-/- samples were not included in the study to avoid any possible background signals. We made great effort to investigate whether Arg-II is present in the cardiomyocytes from different species including mice, rats and humans and have included old arg-ii^-/- mouse samples as a negative control. This allows to validate the antibody specificity and background noises beyond any reasonable doubt. The new experiments in Suppl. Fig. 4 confirms the specificity of the antibody against Arg-II in old mouse kidney which is known to express Arg-II in the S3 proximal tubular cells (Huang J, et al. 2021). To exclude the possible species-specific different expression of Arg-II in the cardiomyocytes, aged mouse and rat heart tissues were used for cellular localization of Arg-II by confocal immunofluorescence staining. As shown in Suppl. Fig. 4B and 4C, both species show Arg-II expression only in non-cardiomyocytes (cells between striated cardiomyocytes) (red arrows) but not in striated cardiomyocytes. Even in the rat myocardial infarction tissues, Arg-II was not found in cardiomyocytes but in endocardium cells (Suppl. Fig. 4B). In isolated cardiomyocytes exposed to hypoxia, a well know strong stimulus for Arg-II protein levels, no Arg-II signals could be detected, while in fibroblasts from the same animals, an elevated Arg-II levels under hypoxia is demonstrated (Fig. 5B). Furthermore, even RT-qPCR could not detect arg-ii mRNA in cardiomyocytes but in non-cardiomyocytes (Fig. 5C). All together, these results demonstrate that Arg-II are not expressed or at negligible levels in cardiomyocytes but expressed in non-cardiomyocytes. This new experiments with rat heart are included in the method section on page 20, the 1st paragraph. The results are described on page 7, the 1st paragraph, and discussed on page 12, the 2nd paragraph. Legend to Suppl. Fig. 4 is included in the file “Suppl. figure legend_R”.

The authors have never investigated the possibility of NO involvement in this mice model.

As above mentioned, we made great effort and have performed new experiments in human monocyte cell line (THP1) in which iNOS is not expressed and not inducible by LPS and arg-ii gene was knocked out by CRISPR technology. Moreover, murine bone-marrow derived macrophages in which inos gene was ablated, is also use for this purpose. The results show that Arg-II and iNOS can be upregulated by LPS independent of each other and iNOS slightly reduces Arg-II expression. However, both Arg-II and iNOS are required for IL-1b production upon LPS stimulation. For detailed results and discussion, please see answers to the comments point 2 or point 6 raised by this reviewer.

A co-culture system would be appropriate to understand the non-cell-autonomous functions of macrophages.

We appreciate the suggestion by this reviewer regarding the co-culture system to test the non-cell autonomous role of Arg-II. We think that our current model, which involves treating cells with conditioned media, is a well-established and effective method for demonstrating the non-cell autonomous role of Arg-II. This approach allows us to observe the effects of Arg-II on surrounding cells through the factors present in the conditioned media released from macrophages. The co-culture system could be considered, if the released factor in the conditioned medium is not stable. This is however not the case. Therefore, we are confident that our experimental model with conditioned medium is sufficiently enough to demonstrate a paracrine effect of cell-cell interaction (please also see answers to the comment point 16.

The Myocardial infarction data shown in the mice model may not be directly linked to cardiac aging.

As we have introduced and discussed in the manuscript, aging is a predominant risk factor for cardiovascular disease (CVD). Studies in experimental animal models and in humans provide evidence demonstrating that aging heart is more vulnerable to stressors such as ischemia/reperfusion injury and myocardial infarction as compared to the heart of young individuals. Even in the heart of apparently healthy individuals of old age, chronic inflammation, cardiomyocyte senescence, cell apoptosis, interstitial/perivascular tissue fibrosis, endothelial dysfunction and endothelial-mesenchymal transition (EndMT), and cardiac dysfunction either with preserved or reduced ejection fraction rate are observed. Our study is aimed to investigate the role of Arg-II in cardiac aging phenotype and age-associated cardiac vulnerability to stressors. Therefore, cardiac functional changes and myocardial infarction in response to ischemia/reperfusion injury are suitable surrogate parameters for the purpose.

Reviewer #2 (Public Review):

Summary:

The results from this study demonstrated a cell-specific role of mitochondrial enzyme arginase-II (Arg-II) in heart aging and revealed a non-cell-autonomous effect of Arg-II on cardiomyocytes, fibroblasts, and endothelial cells through the crosstalk with macrophages via inflammatory factors, such as by IL-1b, as well as a cell-autonomous effect of Arg-II through mtROS in fibroblasts contributing to cardiac aging phenotype. These findings highlight the significance of non-cardiomyocytes in the heart and bring new insights into the understanding of pathologies of cardiac aging. It also provides new evidence for the development of therapeutic strategies, such as targeting the ArgII activation in macrophages.

We're grateful for the reviewer's positive feedback, acknowledging the significant findings of our study on the role of arginase-II (Arg-II) in cardiac aging. We appreciate this reviewer’s insight into the therapeutic potential of targeting Arg-II activation in macrophages and are excited about the implications for future interventions in age-related cardiac pathologies. Thank you for recognizing the importance of our work in advancing our understanding of cardiac aging and potential therapeutic strategies.

Strengths:

This study targets an important clinical challenge, and the results are interesting and innovative. The experimental design is rigorous, the results are solid, and the representation is clear. The conclusion is logical and justified.

We thank this reviewer for the positive comment.

Weaknesses:

The discussion could be extended a little bit to improve the realm of the knowledge related to this study.

We appreciate this comment and have added and revised our discussion on this aspect accordingly at the end of the discussion section on page 19.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

I have several critical concerns, specifically about the mechanism of how Arg-II plays a role in cardiac aging.

My major concerns are:

(1) The authors have shown non-cell-autonomous effects on aging cardiomyocytes, fibroblasts, and endothelial cells mediated by IL-1b from aging macrophages. A macrophage-specific Arg-II knock-out mouse model is a suitable and necessary control to establish claims.

We fully agree with this comment of the reviewer. Unfortunately, this macrophage specific arg-ii knockout animal model is not available, yet. Future research shall develop the macrophage-specific arg-ii^-/- mouse model to confirm this conclusion with aging animals. Since Arg-II is also expressed in fibroblasts and endothelial cells and exerts cell-autonomous and paracrine functions, aging mouse models with conditional arg-ii knockout in the specific cell types would be the next step to elucidate cell-specific function of Arg-II in cardiac aging. We have pointed out this aspect for future research on page 19, lines 2 to 6.

(2) This study suggests that Arg-II exerts its effect through IL-1b in cardiac ageing. However, all experiments performed to demonstrate the link between ArgII and IL-1β are correlative at best. The underlying molecular mechanism, including transcription factors involved in the regulation of IL-1β by arg-ii, has not been demonstrated.

We sincerely appreciate this reviewer’s comment on the aspect! To make it clear, a causal role of Arg-II in promoting IL-1β production in macrophages is evidenced by the experimental results showing that old arg-ii^-/- mouse heart has lower IL-1β levels than the age-matched wt mouse heart (Fig. 6A to 6D). We further showed that the cellular IL-1β protein levels and release are reduced in old arg-ii^-/- mouse splenic macrophages as compared to the wt cells (Fig. 7A, 7C, and 7D). This result is further confirmed with the mouse macrophage cell line RAW264.7 (Suppl. Fig. 5A and suppl. Fig. 5C), in which we demonstrate that silencing arg-ii reduces IL-1β levels stimulated with LPS.

According to this reviewer’s comment (see comment point 6), we made further effort to investigate possible involvement of iNOS in Arg-II-regulated IL-1β production in macrophages stimulated with LPS. We performed new experiments in human monocyte cell line (THP1) in which iNOS is not expressed and not inducible by LPS and arg-ii gene was knocked out by CRISPR technology in the cells.

Moreover, murine bone-marrow derived macrophages in which inos gene was ablated, is also use for this purpose. We found that in the human THP1 monocytes in which Arg-II but not iNOS is induced by LPS (100 ng/mL for 24 hours) (Suppl. Fig. 6A), mRNA and protein levels of IL-1b are markedly reduced in arg-ii knockout THP1^{arg-ii^-/-} as compared to the THP1^wt cells (Suppl. Fig. 6B and 6C), further confirming that Arg-II promotes IL-1b production as also shown in RAW264.7 macrophages (Suppl. Fig. 5A and 5C). The results suggest that Arg-II promotes IL-1b production independently of iNOS. Moreover, the role of iNOS in IL-1b production was also studied in the mouse bone-marrow-derived macrophages in which inos gene is ablated. The results demonstrate that LPS-induced IL-1b production is inhibited by inos deficiency (BMDM^inos-/- vs BMDM^wt) (Suppl. Fig. 6D and 6E), while Arg-II levels are slightly enhanced in the BMDM^inos-/- cells (Suppl. Fig. 6D and 6F). Since arginase and iNOS share the same metabolic substrate L-arginine, ^inos-/- is expected to increase IL-1b production. This is however not the case. A strong inhibition of IL-1β production in ^inos-/- macrophages is observed. These results implicate that iNOS promotes IL-1β production independently of Arg-II and the inhibiting effect of IL-1β by inos deficiency is dominant and able to counteract Arg-II’s stimulating effect on IL-1β production. Hence, our results demonstrate that Arg-II promotes IL-1β production in macrophages independently of iNOS. All together, these results suggest that iNOS slightly reduces Arg-II expression. Arg-II and iNOS can be upregulated by LPS independently. Both Arg-II and iNOS are required for IL-1b production upon LPS stimulation (This concept is illustrated in the Suppl. Fig. 6G). The new results are described on page 8, the last paragraph and page 9, the 1st paragraph, presented in Suppl. Fig.6. The legend to Suppl. Fig. 6 is described in the file “Supplementary figure legend-R”. The related experimental methods are updated on page 23, the last two paragraphs and page 26 the last paragraph. The results are discussed o page 14, the last paragraph and page 15, the first two paragraphs.

(3) Figure 2: The authors have not validated the whole-body Arg-II knock-out mice for arg-ii ablation.

Thanks for pointing out this missing information! We have added the information regarding genotyping of the mice in the method section on page 20, first paragraph. Moreover, Fig. 5C also confirms the genotyping of the non-cardiomyocyte cells isolated from wt and arg-ii^-/- animals.

(4) It is unclear why the authors have chosen to focus on IL-1β specifically, among other pro-inflammatory cytokines that were also downregulated in Arg-II-/- mice as demonstrated in Fig. 2A-D.

We appreciate the reviewer's question, which provides an opportunity to delve deeper into our findings. In our investigation, we observed that aging is accompanied by elevated levels of various proinflammatory markers. Intriguingly, our data revealed that tnf-α remained unaffected by the ablation of arg-ii during aging in the heart tissues, while Il-1β showed a significant reduction in arg-ii^-/- animals compared to age-matched wild-type (wt) mice (Fig. 2). Mcp1 is however a chemoattractant for macrophages and F4-80 serves as a pan marker for macrophages. Moreover, our previous studies demonstrate a relationship between Arg-II and IL-1β in vascular disease and obesity and age-associated renal and pulmonary fibrosis. Finally, IL-1β has been shown to play a causal role in patients with coronary atherosclerotic heart disease as shown by CANTOS trials. Therefore, we have focused on IL-1β in this study. We have now explained and strengthened this aspect in the manuscript on page 7, the last two lines and page 8, the 1st paragraph as following:

“Taking into account that our previous studies demonstrated a relationship of Arg-II and IL-1β in vascular disease and obesity (Ming et al., 2012) and in age-associated organ fibrosis such as renal and pulmonary fibrosis (Huang et al., 2021; Zhu et al., 2023), and IL-1β has been shown to play a causal role in patients with coronary atherosclerotic heart disease as shown by CANTOS trials (Ridker et al., 2017), we therefore focused on the role of IL-1β in crosstalk between macrophages and cardiac cells such as cardiomyocytes, fibroblasts and endothelial cells”.

(5) Although macrophages are shown to be involved in cardiac ageing in the arg-ii mouse model, the authors have not estimated macrophage infiltration and expression of inflammatory or senescence markers in the hearts of these mice.

Thank you very much for raising this important point! Taking the comments of the reviewer into account, we have performed new experiments, i.e., multiple immunofluorescent staining to analyze the infiltrated (CCR2<sup>+</sip>/F4-80⁺) and resident (LYVE1⁺/F4-80⁺) macrophage populations and to investigate to which extent that Arg-II affects the infiltrated and resident macrophage populations in the aging heart and whether this is regulated by arg-ii^-/-. The results show an age-associated increase in the numbers of F4/80⁺ cells in the wt mouse heart, which is reduced in the age-matched arg-ii^-/- animals (Fig. 2G). This result is in accordance with the result of f4/80 gene expression shown in Fig. 2A, demonstrating that arg-ii gene ablation reduces macrophage accumulation in the aging heart. Interestingly, resident macrophages as characterized by LYVE1⁺/F4-80⁺ cells (Fig. 2E and 2H) are predominant in the aging heart as compared to the infiltrated CCR2⁺/F4-80⁺ cells (Fig. 2F and 2I). The increase in both LYVE1⁺/F4-80⁺ and CCR2⁺/F4-80⁺ macrophages in aging heart is reduced in arg-ii^-/- mice (Fig. 2E, 2F, 2H, and 2I). These new results are described on page 6, the 1st paragraph, presented in Fig. 2E to 2I, and discussed on page 13, the 2nd, paragraph. The legend to Fig. 2 is revised. The method for this additional experiment is included on page 22, the 1st paragraph.

Moreover, the aged-associated accumulation of the senescence cells as demonstrated by p16^ink4 positive cells is significantly reduced in arg-ii^-/- animals. This new result is incorporated in the Fig. 1 as Fig. 1G and 1H and described / discussed on page 5, the 2nd paragraph and page 14, the 2nd last sentences of the 1st paragraph. The method of p16^ink4 staining is included in the method section on page 22, the 1st paragraph, line 7. The legend to Fig. 1 is revised accordingly.

(6) Previously, Arg-II has been reported to serve a crucial role in ageing associated with reduced contractile function in rat hearts by regulating Nitric Oxide Synthase (PMID: 22160208). Elevated NO and superoxide have been shown to play crucial roles in the etiology of cardiovascular diseases (PMID: 24180388). Therefore, it is important to assess whether Nitric Oxide (NO) is involved in the aging-related phenotype in this mouse model.

Following the reviewer's suggestion, we conducted new experiments to investigate the role of nitric oxide (NO) in the context of the effect of Arg-II-induced IL-1b production in macrophages. We have addressed this question in the response to the comment point 2.

(7) Based on the results demonstrated in the study, ablation of Arg-II can be expected to cause a reduction in inflammation-associated phenotypes throughout the body at the multi-organ level. The observed improved cardiac phenotype could be an outcome of whole-body Arg-II ablation. It would be fruitful to develop a cardiac-specific Arg-II knockout mouse model to establish the role of Arg-II in the heart, independent of other organ systems.

We agree with the comment of the reviewer on this point. Unfortunately, as explained above (see point 1), it is currently not possible for us to perform the requested experiments, due to lack of cardiac specific arg-ii-knockout mouse model. Moreover, such an approach is complicated by the absence of Arg-II in cardiomyocytes and the expression of Arg-II in multiple cells including endothelial cells, fibroblasts and macrophage of different origin (resident and monocyte-derived infiltrating cells). It’s thus difficult to generate a cardiac-specific gene knockout mouse. One shall investigate roles of cell-specific Arg-II in cardiac aging by generating cell-specific arg-ii^-/- mice. We appreciate very this important aspect and have discussed issue on page 19, the lines 2 to 6.

(8) Contrary to the findings in this paper, Arg-II has previously been reported to be essential for IL-10-mediated downregulation of pro-inflammatory cytokines, including IL-1β (PMID: 33674584).

Thank you very much for mentioning this study! We have now discussed thoroughly the controversies as the following on page 15, the last paragraph and page 16, the 1st paragraph;

“It is of note that a study reported that Arg-II is required for IL-10 mediated-inhibition of IL-1b in mouse BMDM upon LPS stimulation (Dowling et al., 2021), which suggests an anti-inflammatory function of Arg-II. The results of our present study, however, demonstrate that LPS enhances Arg-II and IL-1b levels in macrophages and knockout or silencing Arg-II reduces IL-1b production and release, demonstrating a pro-inflammatory effect of Arg-II. Our findings are supported by the study from another group, which shows decreased pro-inflammatory cytokine production including IL-6 and IL-1b in arg-ii^-/- BMDM most likely through suppression of NFkB pathway, since arg-ii^-/- BMDM reveals decreased activation of NFkB and IL-1b levels upon LPS stimulation (Uchida et al., 2023). Most importantly, our previous study also showed that re-introducing arg-ii gene back to the arg-ii^-/- macrophages markedly enhances LPS-stimulated pro-inflammatory cytokine production (Ming et al., 2012), providing further evidence for a pro-inflammatory role of arg-ii under LPS stimulation. In support of this conclusion, chronic inflammatory diseases such as atherosclerosis and type 2 diabetes (Ming et al., 2012), inflammaging in lung (Zhu et al., 2023), kidney (Huang et al., 2021) and pancreas (Xiong, Yepuri, Necetin, et al., 2017) of aged animals or acute organ injury such as acute ischemic/reperfusion or cisplatin-induced renal injury are reduced in the arg-ii^-/- mice (Uchida et al., 2023). The discrepant findings between these studies and that with IL-10 may implicate dichotomous functions of Arg-II in macrophages, depending on the experimental context or conditions. Nevertheless, our results strongly implicate a pro-inflammatory role of Arg-II in macrophages in the inflammaging in aging heart”.

(9) The authors have only performed immunofluorescence-based experiments to show fibrotic and apoptotic phenotypes throughout this study. To verify these findings, we suggest that they additionally perform RT-PCR or western blotting analysis for fibrotic markers and apoptotic markers.

The fibrotic aspect was analyzed not only by microscopy but also by using a quantitative biochemical assay such as hydroxyproline content assessment. Hydroxyproline is a major component of collagen and largely restricted to collagen. Therefore, the measurement of hydroxyproline levels can be used as an indicator of collagen content as previous investigated in the lung (Zhu et al., 2023). We have also measured collagen genes expression by RT-qPCR as suggested by the reviewer and found an age-related decline of collagen mRNA expression levels in both wt and arg-ii^-/- mice, suggesting that the age-associated cardiac fibrosis and prevention in arg-ii^-/- mice is due to alterations of translational and/or post-translational regulations, including collagen synthesis and/or degradation. The results are in accordance with that reported by other studies published in the literature. We have pointed out this aspect on page 5, the 2nd paragraph:

“The increased cardiac fibrosis in aging is however, associated with decreased mRNA levels of collagen-Ia (col-Ia) and collagen-IIIa (col-IIIa), the major isoforms of pre-collagen in the heart (Suppl. Fig. 2A and 2B), which is a well-known phenomenon in cardiac fibrotic remodelling (Besse et al., 1994; Horn et al., 2016). The results demonstrate that age-associated cardiac fibrosis and prevention in arg-ii^-/- mice is due to alterations of translational and/or post-translational regulations including collagen synthesis and/or degradation”.

The results are presented in Suppl. Fig. 2, legend to Suppl. Fig. 2 is included in the file “Suppl. figure legend_R”. Suppl. table 2 for primers is revised accordingly.

We did not use additional markers to perform apoptotic assays with whole heart, since Fig. 3 shows good evidence that the aging is associated with increased apoptotic cells in the heart and significantly reduced in the arg-ii^-/- mice. The reduction of TUNEL positive (apoptotic) cells in aged arg-ii^-/- mice is mainly due to decrease in apoptotic cardiomyocytes. With the histological analysis, the apoptotic cell types can be well analysed. Moreover, biochemical assay for apoptosis such as caspase-3 cleavage with whole heart tissues can not distinguish apoptotic cell types and may not be sensitive enough for aging heart, due to relatively low numbers of apoptotic cells in aging heart as compared to myocardial infarct model.  

(10) Figure 4: arg-ii has previously been reported to be expressed in rat cardiomyocytes (PMID: 16537391). We strongly suggest the authors verify the expression of Arg-II via immunostaining in isolated cardiomyocytes (using published protocols), and by using multiple different cardiomyocyte-specific markers for colocalization studies to prove the lack of arg-ii expression beyond a reasonable doubt.

As pointed out by this reviewer, Arg-II was previously reported to be expressed in isolated cardiomyocytes from in rats (PMID: 16537391). Unfortunately, negative controls. i.e., arg-ii^-/- samples were not included in the study to avoid any possible background signals. We made great effort to investigate whether Arg-II is present in the cardiomyocytes from different species including mice, rats and humans and have included old arg-ii^-/- mouse samples as a negative control. This allows to validate the antibody specificity and background noises beyond any reasonable doubt. The new experiments in Suppl. Fig. 4 confirms the specificity of the antibody against Arg-II in old mouse kidney which is known to express Arg-II in the S3 proximal tubular cells (Huang J, et al. 2021). To exclude the possible species-specific different expression of Arg-II in the cardiomyocytes, aged mouse and rat heart tissues were used for cellular localization of Arg-II by confocal immunofluorescence staining. As shown in Suppl. Fig. 4B and 4C, both species show Arg-II expression only in non-cardiomyocytes (cells between striated cardiomyocytes) (red arrows) but not in striated cardiomyocytes. Even in the rat myocardial infarction tissues, Arg-II was not found in cardiomyocytes but in endocardium cells (Suppl. Fig. 4B). In isolated cardiomyocytes exposed to hypoxia, a well know strong stimulus for Arg-II protein levels, no Arg-II signals could be detected, while in fibroblasts from the same animals, an elevated Arg-II levels under hypoxia is demonstrated (Fig. 5B). Furthermore, RT-qPCR could not detect arg-ii mRNA in cardiomyocytes but in non-cardiomyocytes (Fig. 5C). All together, these results demonstrate that Arg-II are not expressed or at negligible levels in cardiomyocytes but expressed in non-cardiomyocytes. This new experiments with rat heart are included in the method section on page 20, the 1st paragraph. The results are described on page 7, the 1st paragraph, and discussed on page 12, the 2nd paragraph. Legend to Suppl. Fig. 4 is included in the file “Suppl. figure legend_R”.

(11) Figure 6G: It may be worthwhile to supplement arg-ii^-/- old cells with IL-1beta to see if there is an increase in TUNEL-positive cells.

IL-1b is a well known pro-inflammatory cytokine that causes apoptosis in various cell types including cardiomyocytes (Shen Y., et al., Tex Heart Inst J. 2015;42:109–116. doi: 10.14503/THIJ-14-4254; Liu Z. et. al., Cardiovasc Diabetol 2015;14,125. doi: 10.1186/s12933-015-0288-y; Li. Z., et al., Sci Adv 2020;6:eaay0589. doi: 10.1126/sciadv.aay0589). We appreciate very much the interesting idea of this reviewer to investigate the apoptotic responses of cardiomyocytes from arg-ii^-/- mice to IL-1b. We agree that it is possible that cardiomyocytes from wt from arg-ii^-/- mice react differently to IL-1b, although the cardiomyocytes do not express Arg-II as demonstrated in our present study. If this is true, it must be due to non-cell autonomous effects of different aging microenvironment in the heart or epigenetic modulations of the myocytes. We found that this is a very interesting aspect and requires further extensive investigation. Since our current study focused on the effect of wt and arg-ii^-/- macrophages on cardiomyocytes and non-cardiomyocytes, we prefer not to include this suggested aspect in our manuscript and would like to explore it in the following study.

(12) Figures 4-9: It would be interesting to see if the effect of ArgII in cardiac ageing is gender-specific. It is recommended to include experimental data with male mice in addition to the results demonstrated in female mice.

As pointed out in the manuscript, we have focused on female mice, because an age-associated increase in arg-ii expression is more pronounced in females than in males (Fig. 1A). As suggested by this reviewer, we performed additional experiments investigating effects of arg-ii deficiency in male mice during aging, focusing on pathophysiological outcomes of ischemia/reperfusion injury in ex vivo experiments. The ex vivo functional analytic experiments with Langendorff system were performed in aged male mice (see Suppl. Fig. 9). Following ischemia/reperfusion injury, wt male mice display reduced left ventricular developed pressure (LVDP), as well as the inotropic and lusitropic states (expressed as dP/dt max and dP/dt min, respectively). As previously reported (Murphy et al., 2007), we also found that old male mice are more prone to I/R injury than age-matched female animals. Specifically, 15 minutes of ischemia are enough to significantly affect the left ventricle contractile function in the male mice (Suppl. Fig. 9). As opposite, age-matched old female mice are relatively resistant to I/R injury, and at least 20 min of ischemia are necessary to induce a significant impairment of the contractile function (Fig. 10). Similar to females, the post I/R recovery of cardiac function is also significantly improved in the male arg-ii^-/- mice as compared to age-matched wt animals. In addition to functional recovery, triphenyl tetrazolium chloride (TTC) staining (myocardial infarction) upon I/R-injury in males is significantly reduced in the age-matched male arg-ii^-/- animals (Suppl. Fig. 9C and 9D). All together, these results reveal a role for Arg-II in heart function impairment during aging in both genders with a higher vulnerability to stress in the males. These new results are presented in Suppl. Fig. 9, described on page 10, the last paragraph and page 11. The results are discussed on page 18, the 2nd paragraph as following:

“The fact that aged females have higher Arg-II but are more resistant to I/R injury seems contradictory to the detrimental effect of Arg-II in I/R injury. It is presumable that cardiac vulnerability to injuries stressors depends on multiple factors/mechanisms in aging. Other factors/mechanisms associated with sex may prevail and determine the higher sensitivity of male heart to I/R injury, which requires further investigation. Nevertheless, the results of our study show that Arg-II plays a role in cardiac I/R injury also in males”.

The information on the experimental methods in the male animals is included on page 20, the last paragraph and page 21, the 1st paragraph. Legend to Suppl. Fig. 9 is included in the file “Suppl. figure legend_R”.

(13) Figure 6G: cardiomyocytes from wild-type mice, when treated with macrophages, show 0% TUNEL-positive cells. Since it is unlikely to obtain no TUNEL staining in a cell population, there may be an experimental or analytical error.

Now it is Fig. 7F and 7G. This is due to our specific experimental procedure. After tissue digestion, cardiomyocytes were plated on laminin-coated dishes. Laminin promotes the adhesion of survived cells. Following plating, we conducted a deep washing process to remove damaged and partially adherent cells. This step ensures that only well-shaped, viable, and strongly adherent cells remain as bioassay cells. These “healthy” cells are then selected for the experiments. the apoptotic cells are removed by washing out, reflecting the high viability of the bioassay cells. We have added this detailed information in the method section on page 24, the 2nd paragraph.

(14) Figure 7J: Please assess whether arg-ii depletion also affects the mtROS phenotype.

According to the suggestion of this reviewer, we performed new experiments which show that human cardiac fibroblasts (HCFs) exposed to hypoxia (1% O₂, 48 hours), a known physiological trigger of Arg-II up-regulation, exhibit increased mtROS generation, which involves Arg-II (new Fig. 8M to 8P). We found that Arg-II protein level as well as mtROS (assessed by mitoSOX staining) were both enhanced, accompanied by increased levels of HIF1α (Fig 8M). Moreover, mito-TEMPO pre-incubation reduces mtROS, confirming the mitochondrial origin of the ROS. Silencing of arg-ii with rAd-mediated shRNA, significantly reduces mtROS levels demonstrating a role of Arg-II in the production of mitochondrial ROS in cardiac fibroblasts (Fig 8M to 8P). We have included these results on page 9, the last paragraph and discussed the results on page 17, the 1st paragraph. The related method is described on page 26, the 2nd paragraph. Legend to Fig. 8 is updated on page 32.

(15) Figure 8A-E: The authors have treated human-origin endothelial cells with mice-origin macrophage-conditioned media. It would be more suitable to treat the endothelial cells with human-origin macrophage-conditioned media.

We acknowledge the concern regarding the use of mouse-origin macrophage-conditioned media on human-origin endothelial cells. It is to note, the biological cross-reactivity of cytokines from one species on cells from a different species has been reported in the literature. It was observed that there is quite a strict threshold of 60% amino acid identity, above which cytokines tend to cross-react and statistically, cytokines would tend to cross-react more often as their % amino acid identity increases (Scheerlinck JPY. Functional and structural comparison of cytokines in different species. Vet Immunol Immunopathol. 1999; 72:39-44. https://doi.org/10.1016/S0165-2427(99)00115-4). Taking IL-1b as an example, the 17.5 kDa mature mouse and human IL-1b share 92% aa sequence identity, suggesting a high cross-reactivity. Indeed, human IL-1b has shown biological cross-reactivity in mouse cells (Ledesma E., et al. Interleukin-1 beta (IL-1β) induces tumor necrosis factor alpha (TNF-α) expression on mouse myeloid multipotent cell line 32D cl3 and inhibits their proliferation. Cytokine. 2004; 26:66-72. https://doi.org/10.1016/j.cyto.2003.12.009). Moreover, our results also support the reported cross-reactivity between human and mouse IL-1b. The CM from mouse macrophage indeed showed biological function in human endothelial cells. The observed effects of the conditioned media from aged wild-type macrophages on endothelial cells were specifically mediated through IL-1β. This conclusion is supported by our data showing that the upregulation induced by the conditioned media was significantly reduced by the addition of an IL-1β receptor blocker.

(16) The co-culture system would be more interesting to test the non-cell autonomous role of Arg II.

We appreciate the suggestion by this reviewer regarding the co-culture system to test the non-cell autonomous role of Arg-II. We believe that our current model, which involves treating cells with conditioned media, is a well-established and effective method for demonstrating the non-cell autonomous role of Arg-II. This approach allows us to observe the effects of Arg-II on surrounding cells through the factors present in the conditioned media. The co-culture system could be considered, if the released factor in the conditioned medium is not stable. This is however not the case. So we are confident that our experimental model with conditioned medium is good enough to demonstrate a paracrine effect of cell-cell interaction.

Reviewer #2 (Recommendations For The Authors):

Some minor comments may be considered to improve the realm of the knowledge related to this study.

We appreciate this comment and have added and revised our discussion on this aspect accordingly at the end of the discussion section on page 19, the last 6 lines.

(1) The current study showed strong evidence demonstrating the key role of cardiac macrophages in pathologies of cardiac aging, particularly, the macrophages (MФ) from the circulating blood (hematogenous). It is known that the heart is among the minority of organs in which substantial numbers of yolk-sac MФ persist in adulthood and play a crucial role in maintaining cardiac function. Thus, the adult mammalian heart contains two separate and discrete cardiac MФ subgroups, i.e., the resident MФs originated from yolk sac-derived progenitors and the hematogenous MФs recruited from circulating blood monocytes. These two subtypes of MФs may play distinctive roles in the aging heart and the response to cardiac injury. The author could extend the discussion on the possibility of the resident MФs in aging hearts, which could be further investigated in the future.

We appreciate the suggestion and agree that it provides valuable insight into the study. Taking the comments of the reviewer 1 into account, we have performed new experiments, i.e., co- immunostaining to analyze the infiltrated (CCR2⁺/F4-80⁺) and resident (LYVE1⁺/F4-80⁺) macrophage populations and to investigate to which extent that Arg-II affects infiltrated and resident macrophage populations in the aging heart. We found that in line with the gene expression of f4/80, immunofluorescence staining reveals an age-associated increase in the numbers of F4/80⁺ cells in the wt mouse heart, which is reduced in the age-matched arg-ii^-/- animals (Fig. 2E, F, G), demonstrating that arg-ii gene ablation reduces macrophage accumulation in the aging heart. Interestingly, resident macrophages as characterized by LYVE1⁺/F4-80⁺ cells (Fig. 2E and 2H) are predominant in the aging heart as compared to the infiltrated CCR2⁺/F4-80⁺ cells (Fig. 2F and 2I). The increase in both LYVE1⁺/F4-80⁺ and CCR2⁺/F4-80⁺ macrophages in aging heart is reduced in arg-ii^-/- mice (Fig. 2E, 2F, 2H, and 2I). These new results are described on page 6, the 1st paragraph, presented in Fig. 2E to 2I, and discussed on page 13, the 2nd, paragraph. The legend to Fig. 2 is revised. The method for this additional experiment is included on page 22, the 1st paragraph.

(2) It would be beneficial to the readers if the author could provide some explanation about why ArgII could not be detected in VSMCs in the mouse heart and the species difference between humans and mice. In addition, the author may provide an assumption on the possibility that there may also be a cross-talk between macrophages and VSMCs in the aging heart. A little bit more explanation in the Discussion will be helpful.

We acknowledge and appreciate the suggestion and have discussed these points on page 19 as the following:

“In this context, another interesting aspect is the cross-talk between macrophages and vascular SMC in the aging heart. In our present study, we could not detect Arg-II in vascular SMC of mouse heart but in that of human heart. This could be due to the difference in species-specific Arg-II expression in the heart or related to the disease conditions in human heart which is harvested from patients with cardiovascular diseases. Indeed, in the apoe^-/- mouse atherosclerosis model, aortic SMCs do express Arg-II (Xiong et al., 2013). It is interesting to note that rodents hardly develop atherosclerosis as compared to humans. Whether this could be partly contributed by the different expression of Arg-II in vascular SMC between rodents and humans requires further investigation. In our present study, the aspect of the cross-talk between macrophages and vascular SMC is not studied. Since the crosstalk between macrophages and vascular SMC has been implicated in the context of atherogenesis as reviewed (Gong et al., 2025), further work shall investigate whether Arg-II expressing macrophages could interact with vascular SMC in the coronary arteries in the heart and contribute to the development of coronary artery disease and/or vascular remodelling and the underlying mechanisms“.

(3) Please clarify the arrows in Figure 9C that indicate the infarct area in each splicing section from one heart.

The arrows in Figure 9C (now Fig. 10C) are indeed utilized to indicate the sections displaying the infarcted area within each splicing section from one heart. We have explained the arrow in the figure legend (now Fig. 10 and also new Suppl. Fig. 9).

Author response:

Our response aims to address the following:

The lack of pleiotropy is an unconfirmable assumption of MR, and the addition of those models is therefore quite important, as this is a primary weakness of the MR approach. Given that concern, I read the sensitivity analyses using pleiotropy-robust models as the main result, and in that case, they can't test their hypotheses as these models do not show a BMI instrumental variable association. The other weakness, which might be remedied, is that the power of the tests here is not described. When a hypothesis is tested with an under-powered model, the apparent lack of association could be due to inadequate sample size rather than a true null. Typically, when a statistically significant association is reported, power concerns are discounted as long as the study is not so small as to create spurious findings. That is the case with their primary BMI instrumental variable model - they find an association so we can presume it was adequately powered. But the primary models they share are not the pleiotropy-robust methods MR-Egger, weighted median, and weighted mode. The tests for these models are null, and that could mean a couple of things: (1) the original primary significant association between the BMI genetic instrument was due to pleiotropy, and they therefore don't have a robust model to explore the effects of the tobacco genetic instrument. (2) The power for the sensitivity analysis models (the pleiotropy-robust methods) is inadequate, and the authors share no discussion about the relative power of the different MR approaches. If they do have adequate power, then again, there is no need to explore the tobacco instrument.

We would like to highlight that post-hoc power calculations are often considered redundant since the statistical power estimated for an observed association is directly related to its p-value[1]. In other words, the uncertainty of the association is already reflected in its 95% confidence interval. However, we understand power calculations may still be of interest to the reader, so we will incorporate them in the revised manuscript.

The reason we use inverse variance weighted (IVW) Mendelian randomization (MR) to obtain our main results rather than the pleiotropy-robust methods mentioned by the reviewer/editors (i.e., MR-Egger, weighted median and weighted mode) is that the former has greater statistical power than the latter[2]. Hence, instead of focussing on the statistical significance of the pleiotropy-robust analyses, we consider it is of more value to compare the consistency of the effect sizes and direction of the effect estimates across methods. Any evidence of such consistency increases our confidence in our main findings, since each method relies on different assumptions. As we cannot be sure about the presence and nature of horizontal pleiotropy, it is useful to compare results across methods even though they are not equally powered. It is true that our results for the genetically predicted effects of body mass index (BMI) on the risk of head and neck cancer (HNC) differ across methods. This is precisely what led us to question the validity of our main finding (suggesting a positive effect of BMI on HNC risk). We will clarify this in the discussion section of the revised manuscript as advised.

We understand that the reviewer/editors are concerned that we do not have a robust model to explore the role of tobacco consumption in the link between BMI and HNC. However, we have a different perspective on the matter. If indeed, the main IVW finding for BMI and HNC is due to pleiotropy (since some of the pleiotropy-robust methods suggest conflicting results), then the IVW multivariable MR method is a way to explore the potential source of this bias[3]. We were particularly interested in exploring the role of smoking in the observed association because smoking and adiposity are known to influence each other [4-9] and share a genetic basis[10, 11].

References:

(1) Heinsberg LW, Weeks DE: Post hoc power is not informative. Genet Epidemiol 2022, 46(7):390-394.

(2) Burgess S, Butterworth A, Thompson SG: Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 2013, 37(7):658-665.

(3) Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, Hartwig FP, Kutalik Z, Holmes MV, Minelli C et al: Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019, 4:186.

(4) Morris RW, Taylor AE, Fluharty ME, Bjorngaard JH, Asvold BO, Elvestad Gabrielsen M, Campbell A, Marioni R, Kumari M, Korhonen T et al: Heavier smoking may lead to a relative increase in waist circumference: evidence for a causal relationship from a Mendelian randomisation meta-analysis. The CARTA consortium. BMJ Open 2015, 5(8):e008808.

(5) Taylor AE, Morris RW, Fluharty ME, Bjorngaard JH, Asvold BO, Gabrielsen ME, Campbell A, Marioni R, Kumari M, Hallfors J et al: Stratification by smoking status reveals an association of CHRNA5-A3-B4 genotype with body mass index in never smokers. PLoS Genet 2014, 10(12):e1004799.

(6) Taylor AE, Richmond RC, Palviainen T, Loukola A, Wootton RE, Kaprio J, Relton CL, Davey Smith G, Munafo MR: The effect of body mass index on smoking behaviour and nicotine metabolism: a Mendelian randomization study. Hum Mol Genet 2019, 28(8):1322-1330.

(7) Asvold BO, Bjorngaard JH, Carslake D, Gabrielsen ME, Skorpen F, Smith GD, Romundstad PR: Causal associations of tobacco smoking with cardiovascular risk factors: a Mendelian randomization analysis of the HUNT Study in Norway. Int J Epidemiol 2014, 43(5):1458-1470.

(8) Carreras-Torres R, Johansson M, Haycock PC, Relton CL, Davey Smith G, Brennan P, Martin RM: Role of obesity in smoking behaviour: Mendelian randomisation study in UK Biobank. BMJ 2018, 361:k1767.

(9) Freathy RM, Kazeem GR, Morris RW, Johnson PC, Paternoster L, Ebrahim S, Hattersley AT, Hill A, Hingorani AD, Holst C et al: Genetic variation at CHRNA5-CHRNA3-CHRNB4 interacts with smoking status to influence body mass index. Int J Epidemiol 2011, 40(6):1617-1628.

(10) Thorgeirsson TE, Gudbjartsson DF, Sulem P, Besenbacher S, Styrkarsdottir U, Thorleifsson G, Walters GB, Consortium TAG, Oxford GSKC, consortium E et al: A common biological basis of obesity and nicotine addiction. Transl Psychiatry 2013, 3(10):e308.

(11) Wills AG, Hopfer C: Phenotypic and genetic relationship between BMI and cigarette smoking in a sample of UK adults. Addict Behav 2019, 89:98-103.

Author response:

The following is the authors’ response to the previous reviews

In response to Reviewer #1, we have replaced the original images in Figure 1A with new immunofluorescence data showing matched DAPI staining density between control and AD patient samples. We also have updated the PINK1 staining images of mouse brain sections in Figure 1C to eliminate potential non-specific signals. These revisions provide clearer evidence supporting our conclusions about PINK1/pUb’s role in neurodegeneration.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary

In this beautiful paper the authors examined the role and function of NR2F2 in testis development and more specifically on fetal Leydig cells development. It is well known by now that FLC are developed from an interstitial steroidogenic progenitors at around E12.5 and are crucial for testosterone and INSL3 production during embryonic development, which in turn shapes the internal and external genitalia of the male. Indeed, lack of testosterone or INSL3 are known to cause DSD as well as undescended testis, also termed as cryptorchidism. The authors first characterized the expression pattern of the NR2R2 protein during testis development and then used two cKO systems of NR2F2, namely the Wt1-creERT2 and the Nr5a1-cre to explore the phenotype of loss of NR2F2. They found in both cases that mice are presenting with undescended testis and major reduction in FLC numbers. They show that NR2F2 has no effect on the amount and expression of the progenitor cells but in its absence, there are less FLC and they are immature.

The effect of NR2F2 is cell autonomous and does not seem to affect other signalling pathways implemented in Leydig cell development as the DHH, PDGFRA and the NOTCH pathway.

Overall, this paper is excellent, very well written, fluent and clear. The data is well presented, and all the controls and statistics are in place. I think this paper will be of great interest to the field and paves the way for several interesting follow up studies as stated in the discussion

Reviewer #2 (Public review):

The major conclusion of the manuscript is expressed in the title: "NR2F2 is required in the embryonic testis for Fetal Leydig Cell development" and also at the end of the introduction and all along the result part. All the authors' assertions are supported by very clear and statistically validated results from ISH, IHC, precise cell counting and gene expression levels by qPCR. The authors used two different conditional Nr2f2 gene ablation systems that demonstrate the same effects at the FLC level. They also showed that the haplo-insufficiency of Wt1 in the first system (knock-in Wt1-cre-ERT2) aggravated the situation in FLC differentiation by disturbing the differentiation of Sertoli cells and their secretion of pro-FLC factors, which had a confounding effect and encouraged them to use the second system. This demonstrates the great rigor with which the authors interpreted the results. In conclusion, all authors' claims and conclusions are justified by their high-quality results.

Recommendations for the authors:

We thank the reviewers for their comments which have improved and strengthened our manuscript. Please see our responses to specific comments below in blue.

Reviewer #1 (Recommendations for the authors):

I have several small comments:

(1) There has been recently a preprint from the Yao lab about the role of NR2F2 is steroidogenic cells (https://www.biorxiv.org/content/10.1101/2024.09.16.613312v1). They performed cKO of NR2F2 using the Wt1creERT2 and found similar results. You should present and discuss this paper in light of your results.

Estermann et al., report a very similar phenotype of FLC hypoplasia in an independent mouse model of Nr2f2 conditional mutation. We have now referred to this article in the discussion of our manuscript as suggested.

(2) In the introduction I think it is important to mention that the steroidogenic progenitors are derived from Wnt5a positive cells (https://pubmed.ncbi.nlm.nih.gov/35705036/).

We have mentioned this point in the introduction as suggested.

(3) In both models you show a decrease in the number of FLC (60% or 40%) and yet they both present with undescended testis. It is important to discuss the fact that there is no need for a complete ablation of testosterone and INSL3 in order to get cryptorchidism.

We have mentioned this point in the discussion as suggested.

The fact that you get only partial reduction in FLC is likely due to redundancy with additional factors, possibly the ARX like you stated in the discussion and it will be interesting to explore that in the future but is beyond the scope of the current paper.

We agree with the reviewer, this question could be addressed by analyzing Arx,Nr2f2 double mutants.

(4) In page 8 line 11 you mention data not shown- not sure if this is allowed in the journal .

The data is now shown in Figure S5A as suggested.

(5) In Figure 2- it will be good if you add a schematic model of the mouse strains used as well as the experimental and control mice next to the Tam scheme. Similar scheme should be in figure 3 for Nr5a1-cre.

We have modified Figures 2 and 3 as suggested.

(6) There is a clear and pronounced effect of the testis cords number and size. It will be good if you could qualify testis cord numbers/ diameter in the mutants even if you do not follow in detail the effect on Sertoli cells

We have quantified testis cords numbers and area in E14.5 Control and Wt1^CreERT2/+; Nr2f2^flox/flox testes. This data is now shown in Figure S2M.

(7) It will be good to present the undescended testis in the Wt1-cre model in figure 2 and not in the supp figure

The data is now shown in Figure 2H-I as suggested.

(8) Please add labelling of the testis, kidney, bladder, vas deferens in figure 3 N+O and in the Wt1-cre model

We have added the labels in Figures 2 and 3 as suggested.

(9) In figure 5 which present both models- it will be good to use the scheme I suggested before to highlight which results refer to which ko model.

We have modified Figure 5 as suggested.

Reviewer #2 (Recommendations for the authors):  

The work presented in this manuscript gave me food for thought. I have always been intrigued by the fact that of the large number of interstitial cells in the testis, a minority differentiate into mature androgen-producing Leydig cells. In other words, how is the number of functional steroidogenic cells defined from a large pool of progenitor cells (ARX and NR2F2 positive ones)? This may have a link with the levels of androgens produced (a kind of feedback control) or the effectiveness of these androgens on the target tissues (i.e.: as spermatogenesis efficiency in adults). In addition, there must be specific signals (probably linked to gonadotropins) that induce the recruitment of Leydig cells from the progenitor pool. Perhaps the genetic models generated in this study could help to address these questions. I leave it to the authors to judge.

We agree with the reviewer. How NR2F2 (and other factors) integrate extrinsic cues to regulate the recruitment of a subset of interstitial steroidogenic progenitors along the Leydig cell differentiation pathway is a fascinating question beyond the scope of this work.

In addition to this reflection, I propose a few minor modifications likely to improve the quality of the manuscript:

(1) Page 3, lane 3: I suggest to replace "growth" by "differentiation"

We have modified the text as suggested.

(2) Page 3, lane 4: the "scrotum" is missing in the parenthesis. Please add it before "and penis"

We have modified the text as suggested.

(3) Page 5, lanes 21-24: kidney hypoplasia is also evident on Fig S2H (stated in the figure legend). It could be also mentioned in this sentence and it implies "...that NR2F2 function is required for testicular and kidney development."

We have modified the text as suggested.

(4) Page 5, lanes 28-30. In addition to the reduction in the number of HSD3B-positive cells, HSD3B staining seems clearly more faint in mutant FLC (Fig 2M) compared to adrenal cells on the same section or FLC in control gonads. This fits well with other results on the level of steroidogenic enzymes (Fig 2O) and those presented thereafter (Fig S4 I-J and Fig 5). Perhaps the author could mention this fact.

We have modified the text as suggested in the results section “NR2F2 is required for FLC maturation” (Page 8).

(5) Page 5, lanes 31-34: testicular descent is hugely sensible to INSL3 in the mouse (by contrast with other species where androgens seem to be more critical). I was wondering if you can check a better phenotypic marker for the absence (or reduction) of androgens like the differentiation of epididymides by HE staining or the anogenital distance at birth.

We have measured the anogenital distance at P0 and P1 as suggested and have included the corresponding graph in Fig. S3P

(6) Page 8, lanes 21-22: "HSD3B positive FLC were smaller and more elongated". It is clear on Fig 5F but not evident on Fig 5D. Could the authors propose another image?

We have modified Figure 5 as suggested and provide now another example of HSD3B positive FLCs in a Nr5a1Cre; Nr2f2^flox/flox mutant gonad (Fig. 5D) and the corresponding control littermate (Fig. 5C).

(7) Page 14, lane 12: "(arrow in I)" should be "(arrow in H)"

We have modified the text as suggested. Please note that ACTA 2 expression is now shown in Figure S2 G-H.

(8) Page 15, lane 6: "Arrows indicate NR5A1 positive FLC". There is no arrow on Fig4 C,D; but a kind of scale bar on the enlargement shown in C.

We have modified Figure 4 as suggested.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

This paper provides a computational model of a synthetic task in which an agent needs to find a trajectory to a rewarding goal in a 2D-grid world, in which certain grid blocks incur a punishment. In a completely unrelated setup without explicit rewards, they then provide a model that explains data from an approach-avoidance experiment in which an agent needs to decide whether to approach or withdraw from, a jellyfish, in order to avoid a pain stimulus, with no explicit rewards. Both models include components that are labelled as Pavlovian; hence the authors argue that their data show that the brain uses a Pavlovian fear system in complex navigational and approach-avoid decisions.

Thanks to the reviewer’s comments, we have now added the following text to our Discussion section (Lines 290-302):

“When it comes to our experiments, both the simulation and VR experiment models are related and derived from the same theoretical framework maintaining an algebraic mapping. They differ only in task-specific adaptations i.e. differ in action sets and differ in temporal difference learning rules - multi-step decisions in the grid world vs. Rescorla-Wagner rule for single-step decisions in the VR task. This is also true for Dayan et al. [2006] who bridge Pavlovian bias in a Go-No Go task (negative auto-maintenance pecking task) and a grid world task. A further minor difference between the simulation and VR experiment models is the use of a baseline bias in the human experiment's RL and the RLDDM model, where we also model reaction times with drift rates which is not a behaviour often simulated in the grid world simulations. As mentioned previously, we use the grid world tasks for didactic purposes, similar to Dayan et al. [2006] and common to test-beds for algorithms in reinforcement learning [Sutton et al., 1998]. The main focus of our work is on Pavlovian fear bias in safe exploration and learning, rather than on its role in complex navigational decisions. Future work can focus on capturing more sophisticated safe behaviours, such as escapes [Evans et al., 2019, Sporrer et. al., 2023] and model-based planning, which span different aspects of the threat-imminence continuum [Mobbs et al., 2020].”

In the first setup, they simulate a model in which a component they label as Pavlovian learns about punishment in each grid block, whereas a Q-learner learns about the optimal path to the goal, using a scalar loss function for rewards and punishments. Pavlovian and Q-learning components are then weighed at each step to produce an action. Unsurprisingly, the authors find that including the Pavlovian component in the model reduces the cumulative punishment incurred, and this increases as the weight of the Pavlovian system increases. The paper does not explore to what extent increasing the punishment loss (while keeping reward loss constant) would lead to the same outcomes with a simpler model architecture, so any claim that the Pavlovian component is required for such a result is not justified by the modelling. 

Thanks to the reviewer’s comments, we have now added the following text to our Discussion section (Line 303-313):

“In our simulation experiments, we assume the coexistence of the Pavlovian fear system and the instrumental system to demonstrate the emergent safety-efficiency trade-off from their interaction. It is possible that similar behaviours could be modelled using an instrumental system alone, with higher punishment sensitivity, therefore we do not argue for the necessity for the Pavlovian fear system here. Instead, the Pavlovian fear system itself could be a potential biologically plausible implementation of punishment sensitivity. Unlike punishment sensitivity (scaling of the punishments), which has not been robustly mapped to neural substrates in fMRI studies; the neural substrates for the Pavlovian fear system are well known (e.g., the limbic loop and amygdala, further see Supplementary Fig. 16). Additionally, Pavlovian fear system provides a separate punishment memory that cannot be erased by greater rewards like [Elfwing and Seymour, 2017, Wang et al., 2018]. This fundamental point can be observed in our simple T-maze simulations, where the Pavlovian fear system encourages avoidance behaviour and the agent chooses the smaller reward instead of the greater reward.”

In the second setup, an agent learns about punishments alone. "Pavlovian biases" have previously been demonstrated in this task (i.e. an overavoidance when the correct decision is to approach). The authors explore several models (all of which are dissimilar to the ones used in the first setup) to account for the Pavlovian biases. 

Thanks to the reviewer’s comments, we have now added a paragraph in our Discussion section (Line 290-302) explaining the similarity of our models and their integrated interpretation. We hope this addresses the reviewer’s concerns.

Strengths: 

Overall, the modelling exercises are interesting and relevant and incrementally expand the space of existing models. 

Weaknesses: 

I find the conclusions misleading, as they are not supported by the data. 

First, the similarity between the models used in the two setups appears to be more semantic than computational or biological. So it is unclear to me how the results can be integrated. 

Thanks to the reviewer’s comments, we have now added a paragraph in our Discussion section (Line 290-302 onwards) explaining the similarity of our models and their integrated interpretation. We hope this addresses the reviewer’s concerns.

Secondly, the authors do not show "a computational advantage to maintaining a specific fear memory during exploratory decision-making" (as they claim in the abstract). Making such a claim would require showing an advantage in the first place. For the first setup, the simulation results will likely be replicated by a simple Q-learning model when scaling up the loss incurred for punishments, in which case the more complex model architecture would not confer an advantage. The second setup, in contrast, is so excessively artificial that even if a particular model conferred an advantage here, this is highly unlikely to translate into any real-world advantage for a biological agent. The experimental setup was developed to demonstrate the existence of Pavlovian biases, but it is not designed to conclusively investigate how they come about. In a nutshell, who in their right mind would touch a stinging jellyfish 88 times in a short period of time, as the subjects do on average in this task? Furthermore, in which real-life environment does withdrawal from a jellyfish lead to a sting, as in this task? 

Crucially, simplistic models such as the present ones can easily solve specifically designed lab tasks with low dimensionality but they will fail in higher-dimensional settings. Biological behaviour in the face of threat is utterly complex and goes far beyond simplistic fight-flight-freeze distinctions (Evans et al., 2019). It would take a leap of faith to assume that human decision-making can be broken down into oversimplified sub-tasks of this sort (and if that were the case, this would require a meta-controller arbitrating the systems for all the sub-tasks, and this meta-controller would then struggle with the dimensionality j). 

Thanks to the reviewer’s comments, we have now mentioned this point in Lines 299-302.

On the face of it, the VR task provides higher "ecological validity" than previous screen-based tasks. However, in fact, it is only the visual stimulation that differs from a standard screen-based task, whereas the action space is exactly the same. As such, the benefit of VR does not become apparent, and its full potential is foregone. 

If the authors are convinced that their model can - then data from naturalistic approach-avoidance VR tasks is publicly available, e.g. (Sporrer et al., 2023), so this should be rather easy to prove or disprove. In summary, I am doubtful that the models have any relevance for real-life human decision-making. 

Finally, the authors seem to make much broader claims that their models can solve safety-efficiency dilemmas. However, a combination of a Pavlovian bias and an instrumental learner (study 1) via a fixed linear weighting does not seem to be "safe" in any strict sense. This will lead to the agent making decisions leading to death when the promised reward is large enough (outside perhaps a very specific region of the parameter space). Would it not be more helpful to prune the decision tree according to a fixed threshold (Huys et al., 2012)? So, in a way, the model is useful for avoiding cumulatively excessive pain but not instantaneous destruction. As such, it is not clear what real-life situation is modelled here. 

We hope our additions to the Discussion section, from Line 290 to Line 313 address the reviewer’s concerns.  

A final caveat regarding Study 1 is the use of a PH associability term as a surrogate for uncertainty. The authors argue that this term provides a good fit to fear-conditioned SCR but that is only true in comparison to simpler RW-type models. Literature using a broader model space suggests that a formal account of uncertainty could fit this conditioned response even better (Tzovara et al., 2018). 

We have now added a line discussing this. (Line 356-358)

“Future work could also use a formal account of uncertainty which could fit the fear-conditioned skin-conductance response better than Pearce-Hall associability [Tzovara et al., 2018].”

Reviewer #2 (Public review): 

Summary: 

The authors tested the efficiency of a model combining Pavlovian fear valuation and instrumental valuation. This model is amenable to many behavioral decision and learning setups - some of which have been or will be designed to test differences in patients with mental disorders (e.g., anxiety disorder, OCD, etc.). 

Strengths: 

(1) Simplicity of the model which can at the same time model rather complex environments. 

(2) Introduction of a flexible omega parameter. 

(3) Direct application to a rather advanced VR task. 

(4) The paper is extremely well written. It was a joy to read. 

Weaknesses: 

Almost none! In very few cases, the explanations could be a bit better. 

Thank you, we have added further explanations in the discussion section. We have further improved the writing in abstract, introduction and Methods section taking into account recommendations from reviewer #2 and #3.

Reviewer #2 (Recommendations for the authors): 

(1) Why is there no flexible omega in Figures 3B and 3C? Did I miss this? 

Thank you. We have now added additional text to explain our motivation in Experiment 2, which only varies the fixed omega and omits the flexible omega (Lines 136-140).

“In this set of results, we wish to qualitatively tease apart the role of a Pavlovian bias in shaping and sculpting the instrumental value and also provide more insight into the resulting safety-efficiency trade-off. Having shown the benefits of a flexible ω in the previous section, here we only vary the fixed ω to illustrate the effect of a constant bias and are not concerned with the flexible bias in this experiment.”

We encourage the reader to consider this akin to an additional study that will explain how Pavlovian bias to withdraw can play a role in avoiding punishments similar to that of punishment sensitivity. This is particularly important as we do have neural correlates for Pavlovian biases but lack a clear neural correlation for punishment sensitivity so far, as mentioned in our new additions to the Discussion section (Lines 303-313).

(2) The introduction of the flexible omega and the PAL agent in the results is a bit sudden. Some more details are needed to understand this during the first read of this passage. 

We thank reviewer #2 for bringing this to our notice. We have attempted to refine our passage by including sentences like - 

“The standard (rational) reinforcement learning system is modelled as the instrumental learning system. The additional Pavlovian fear system biases the withdrawal actions to aid in safe exploration, in line with our hypothesis.”

“Both systems learn using a basic temporal difference updating rule (or in instances, its special case, the Rescorla-Wagner rule)”

“We implement the flexible ω using Pearce-Hall associability (see equation 15 in Methods). The Pearce-Hall associability maintains a running average of absolute temporal difference errors (δ) as per equation 14. This acts as a crude but easy-to-compute metric for outcome uncertainty which gates the influence of the Pavlovian fear system, in line with our hypothesis. This implies that higher the outcome uncertainty, as is the case in early exploration, the more cautious our agent will be, resulting in safer exploration”

(3) In my view, the possibility of modeling moving predators is extremely interesting. I would include Figure 8D and the corresponding explanation in the main text. 

Response with revision: We thank the reviewer for finding our simulation on moving predators extremely interesting. Unfortunately, since our instrumental system is not model-based, and especially is not explicitly modelling the predator dynamics, our simulation might not be a very accurate representation of real moving predator environments. As pointed out by Reviewer #1, perhaps several other systems other than Pavlovian fear responses are necessary for safe behaviour in such environments and we hope to address these in future studies. Thanks again for taking an interest in our simulations.

(4) The VR experiment should be mentioned more clearly in the abstract and the introduction. It should be mentioned a bit more clearly why VR was helpful and why the authors did not use a simple bird's eye grid world task. 

I cannot assess the RLDDM and I did not check the code. 

Thank you, we have now mentioned the VR experiment more clearly in the abstract and the introduction. We also now further mention that the VR experiment “builds upon previous Go-No Go studies studying Pavlovian-Instrumental transfer (Guitart-Masip et al, 2012; Cavanagh et al, 2013). The virtual-reality approach confers a greater ecological validity and the immersive nature may contribute better fear conditioning, making it easier to distinguish the aversive components.”

A bird’s eye grid world may not invoke a strong withdrawal response, as seen in these immersive approach-withdrawal tasks where we can clearly distinguish a Pavlovian fear-based withdrawal response. We did include immersive VR maze results in the supplementary materials, but future work is needed to isolate the different systems at play in such a complex behaviour.

Reviewer #3 (Public review): 

Summary: 

This paper aims to address the problem of exploring potentially rewarding environments that contain the danger, based on the assumption that an independent Pavlovian fear learning system can help guide an agent during exploratory behaviour such that it avoids severe danger. This is important given that otherwise later gains seem to outweigh early threats, and agents may end up putting themselves in danger when it is advisable not to do so. 

The authors develop a computational model of exploratory behaviour that accounts for both instrumental and Pavlovian influences, combining the two according to uncertainty in the rewards. The result is that Pavlovian avoidance has a greater influence when the agent is uncertain about rewards. 

Strengths: 

The study does a thorough job of testing this model using both simulations and data from human participants performing an avoidance task. Simulations demonstrate that the model can produce "safe" behaviour, where the agent may not necessarily achieve the highest possible reward but ensures that losses are limited. Interestingly, the model appears to describe human avoidance behaviour in a task that tests for Pavlovian avoidance influences better than a model that doesn't adapt the balance between Pavlovian and instrumental based on uncertainty. The methods are robust, and generally, there is little to criticise about the study. 

Weaknesses: 

The extent of the testing in human participants is fairly limited but goes far enough to demonstrate that the model can account for human behaviour in an exemplar task. There are, however, some elements of the model that are unrealistic (for example, the fact that pre-training is required to select actions with a Pavlovian bias would require the agent to explore the environment initially and encounter a vast amount of danger in order to learn how to avoid the danger later). The description of the models is also a little difficult to parse. 

Thank you, we have now attempted to clarify these points in the Discussion section by adding the following text (Lines 313-321):

“ We next discuss the plausibility of pre-training to select the hardwired actions In the human experiment, the withdrawal action is straightforwardly biased, as noted, while in the grid world, we assume a hardwired encoding of withdrawal actions for each state/grid. This innate encoding of withdrawal actions could be represented in the dPAG [Kim et al., 2013]. We implement this bias using pre-training, which we assume would be a product of evolution. Alternatively, this could be interpreted as deriving from an appropriate value initialization where the gradient over initialized values determines the action bias. Such aversive value initialization, driving avoidance of novel and threatening stimuli, has been observed in the tail of the striatum in mice, which is hypothesised to function as a Pavlovian fear/threat learning system [Menegas et al., 2018].”

Reviewer #3 (Recommendations for the authors): 

I have relatively little to suggest, as in my view the paper is robust, thorough, and creative, and does enough to support the primary argument being made at the most fundamental level. My suggestions for improvement are as follows: 

(1) Some aspects of the model are potentially unrealistic (as described in the public review), and the paper may benefit from some discussion of these issues or attempts to make the model more realistic - i.e., to what extent is this plausible in explaining more complex avoidance behaviour? Primarily, the fact that pre-training is required to identify actions subject to Pavlovian bias seems unlikely to be effective in real-world situations - is there a better way to achieve this in cases where there isn't necessarily an instinctual Pavlovian response? 

Thank you, we agree that the advantage of Pavlovian bias is restricted to the bias/instinctual Pavlovian response conferred by evolution. Future work is needed to model more complex avoidance behaviour such as escapes. We hope to have made this more clear with our edits to the Discussion (Lines 299-302) in our response to Reviewer #1’s comments, specifically:

“The main focus of our work is on Pavlovian fear bias in safe exploration and learning, rather than on its role in complex navigational decisions. Future work can focus on capturing more sophisticated safe behaviours, such as escapes [Evans et al., 2019, Sporrer et. al., 2023] and model-based planning which span different aspects of the threat-imminence continuum [Mobbs et al., 2020]”  

(2) The description of the model in the method can be a little hard to follow and would benefit from further explanation of certain parameters. In general, it would be good to ensure that all terms mentioned in equations are described clearly in the text (for example, in Equation1 it isn't clear what k refers to). 

Thank you, we have now added further information on all of the parameters in Equation 1 and overall improved the Methods section writing, for instance using time subscript for less confusion while introducing the parameters. We use the standard notation used in Sutton and Barto textbook. k refers to the timesteps into the future, and is now explained better in the Methods section.

(3) Another point of clarification in Equation 1 - does the policy account for the Pavlovian influence or is this purely instrumental? 

Thank you, Equation 1 is purely instrumental. We have now specifically mentioned this. The Pavlovian influence follows later. They are combined into propensities for action as per equations 11-13.

(4) I was curious whether similar outcomes could be achieved by more complex instrumental models without the need for Pavlovian influences. For example, could different risk-sensitive decision rules (e.g., conditional value at risk) that rely only on the instrumental system afford safe behaviour without the need for an additional Pavlovian system? 

Thank you for your comment. Yes, CVaR can achieve safe exploration/cautious behaviour in choices similar to Pavlovian avoidance learning. But we think both differ in the following ways:

(1) CVaR provides the correct solution to the wrong problem (objective that only maximises the lower tail of the distribution of outcomes)

(2) Pavlovian bias provides the wrong solution to the right problem (normative objective, but a Pavlovian bias which may be vestige of evolution)

Here we use the “wrong problem, wrong solution, wrong environment” categorisation terminology from Huys et al. 2015.

Huys, Q. J., Guitart-Masip, M., Dolan, R. J., & Dayan, P. (2015). Decision-theoretic psychiatry. Clinical Psychological Science, 3(3), 400-421.

Secondly, we find an effect of Pavlovian bias on reaction times - slowing down of approach responses and faster withdrawal responses. We do not think this can be best explained in a CVaR type model and is a direction for future work. We think such model-based methods are slower to compute, but Pavlovian withdrawal bias is quicker response.

We have now included this in brief in Lines 280-288.

(5) Figure 5 would benefit from a clearer caption as it is not necessarily clear from the current one that the left panels refer to choices and the right panels to reaction times. 

Thank you, we have improved the caption for Fig. 5.

(6) It would be good to include some indication of the quality of the model fits for the human behavioural study (i.e., diagnostics such as R-hat) to ensure that differences in model fit between models are not due to convergence issues with different models. This would be especially helpful for the RLDDM models as these can be difficult to fit successfully.

Thank you, we observed that all Rhat values were strictly less than 1.05 (most parameters were less than 1.01 and generally close to 1), indicating that the models converged. We have now added this line to the results (Line 246-248). Thanks to the reviewer’s comments, we have now added the following text to our Discussion section (Lines 290-302): “When it comes to our experiments, both the simulation and VR experiment models are related and derived from the same theoretical framework maintaining an algebraic mapping. They differ only in task-specific adaptations i.e. differ in action sets and differ in temporal difference learning rules - multi-step decisions in the grid world vs. Rescorla-Wagner rule for single-step decisions in the VR task. This is also true for Dayan et al. [2006] who bridge Pavlovian bias in a Go-No Go task (negative auto-maintenance pecking task) and a grid world task. A further minor difference between the simulation and VR experiment models is the use of a baseline bias in the human experiment's RL and the RLDDM model, where we also model reaction times with drift rates which is not a behaviour often simulated in the grid world simulations. As mentioned previously, we use the grid world tasks for didactic purposes, similar to Dayan et al. [2006] and common to test-beds for algorithms in reinforcement learning [Sutton et al., 1998]. The main focus of our work is on Pavlovian fear bias in safe exploration and learning, rather than on its role in complex navigational decisions. Future work can focus on capturing more sophisticated safe behaviours, such as escapes [Evans et al., 2019, Sporrer et. al., 2023] and model-based planning, which span different aspects of the threat-imminence continuum [Mobbs et al., 2020].” In the first setup, they simulate a model in which a component they label as Pavlovian learns about punishment in each grid block, whereas a Q-learner learns about the optimal path to the goal, using a scalar loss function for rewards and punishments. Pavlovian and Q-learning components are then weighed at each step to produce an action. Unsurprisingly, the authors find that including the Pavlovian component in the model reduces the cumulative punishment incurred, and this increases as the weight of the Pavlovian system increases. The paper does not explore to what extent increasing the punishment loss (while keeping reward loss constant) would lead to the same outcomes with a simpler model architecture, so any claim that the Pavlovian component is required for such a result is not justified by the modelling.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this manuscript, Azlan et al. identified a novel maternal factor called Sakura that is required for proper oogenesis in Drosophila. They showed that Sakura is specifically expressed in the female germline cells. Consistent with its expression pattern, Sakura functioned autonomously in germline cells to ensure proper oogenesis. In Sakura KO flies, germline cells were lost during early oogenesis and often became tumorous before degenerating by apoptosis. In these tumorous germ cells, piRNA production was defective and many transposons were derepressed. Interestingly, Smad signaling, a critical signaling pathway for GSC maintenance, was abolished in sakura KO germline stem cells, resulting in ectopic expression of Bam in whole germline cells in the tumorous germline. A recent study reported that Bam acts together with the deubiquitinase Otu to stabilize Cyc A. In the absence of sakura, Cyc A was upregulated in tumorous germline cells in the germarium. Furthermore, the authors showed that Sakura co-immunoprecipitated Otu in ovarian extracts. A series of in vitro assays suggested that the Otu (1-339 aa) and Sakura (1-49 aa) are sufficient for their direct interaction. Finally, the authors demonstrated that the loss of otu phenocopies the loss of sakura, supporting their idea that Sakura plays a role in germ cell maintenance and differentiation through interaction with Otu during oogenesis.

Strengths:

To my knowledge, this is the first characterization of the role of CG14545 genes. Each experiment seems to be well-designed and adequately controlled.

Weaknesses:

However, the conclusions from each experiment are somewhat separate, and the functional relationships between Sakura's functions are not well established. In other words, although the loss of Sakura in the germline causes pleiotropic effects, the cause-and-effect relationships between the individual defects remain unclear.

Reviewer #2 (Public review):

In this study, the authors identified CG14545 (and named it Sakura), as a key gene essential for Drosophila oogenesis. Genetic analyses revealed that Sakura is vital for both oogenesis progression and ultimate female fertility, playing a central role in the renewal and differentiation of germ stem cells (GSC).

The absence of Sakura disrupts the Dpp/BMP signaling pathway, resulting in abnormal bam gene expression, which impairs GSC differentiation and leads to GSC loss. Additionally, Sakura is critical for maintaining normal levels of piRNAs. Also, the authors convincingly demonstrate that Sakura physically interacts with Otu, identifying the specific domains necessary for this interaction, suggesting a cooperative role in germline regulation. Importantly, the loss of otu produces similar defects to those observed in Sakura mutants, highlighting their functional collaboration.

The authors provide compelling evidence that Sakura is a critical regulator of germ cell fate, maintenance, and differentiation in Drosophila. This regulatory role is mediated through the modulation of pMad and Bam expression. However, the phenotypes observed in the germarium appear to stem from reduced pMad levels, which subsequently trigger premature and ectopic expression of Bam. This aberrant Bam expression could lead to increased CycA levels and altered transcriptional regulation, impacting piRNA expression. Given Sakura's role in pMad expression, it would be insightful to investigate whether overexpression of Mad or pMad could mitigate these phenotypic defects (UAS-Mad line is available at Bloomington Drosophila Stock Center).

As suggested reviewer 1, we tested whether overexpression of Mad could rescue or mitigate the loss of sakura phenotypic defects, by using nos-Gal4-VP16 > UASp-Mad-GFP in the background of sakura^null. As shown in Fig S11, we did not observe any mitigation of defects.

Then, we also tested whether expressing a constitutive active form of Tkv, by using UAS-Dcr2, NGT-Gal4 > UASp-tkv.Q235D in the background of sakura^RNAi. As shown in Fig S12, we did not observe any mitigation of defects by this approach either.

A major concern is the overstated role of Sakura in regulating Orb. The data does not reveal mislocalized Orb; rather, a mislocalized oocyte and cytoskeletal breakdown, which may be secondary consequences of defects in oocyte polarity and structure rather than direct misregulation of Orb. The conclusion that Sakura is necessary for Orb localization is not supported by the data. Orb still localizes to the oocyte until about stage 6. In the later stage, it looks like the cytoskeleton is broken down and the oocyte is not positioned properly, however, there is still Orb localization in the ~8-stage egg chamber in the oocyte. This phenotype points towards a defect in the transport of Orb and possibly all other factors that need to localize to the oocyte due to cytoskeletal breakdown, not Orb regulation directly. While this result is very interesting it needs further evaluation on the underlying mechanism. For example, the decrease in E-cadherin levels leads to a similar phenotype and Bam is known to regulate E-cadherin expression. Is Bam expressed in these later knockdowns?

We examined Bam and DE-Cadherin expression in later RNAi knockdowns driven by ToskGal4. As shown in Fig S9, Bam was not expressed in these later knockdowns compared with controls. DE-Cadherin staining suggested a disorganized structure in late-stage egg chambers.

We agree that we overstated a role of Sakura in regulating Orb in the initial manuscript. We changed the text to avoid overstating.

The manuscript would benefit from a more balanced interpretation of the data concerning Sakura's role in Orb regulation. Furthermore, a more expanded discussion on Sakura's potential role in pMad regulation is needed. For example, since Otu and Bam are involved in translational regulation, do the authors think that Mad is not translated and therefore it is the reason for less pMad? Currently the discussion presents just a summary of the results and not an extension of possible interpretation discussed in context of present literature.

We changed the text to avoid overstating a role of Sakura in regulating Orb localization.

Based on our newly added results showing that transgenic overexpression of Mad could not rescue or mitigate the phenotypic defects of sakura^null mutant (Fig S11), we do not think the reason for less pMad is less translation of Mad.

Reviewer #3 (Public review):

In this very thorough study, the authors characterize the function of a novel Drosophila gene, which they name Sakura. They start with the observation that sakura expression is predicted to be highly enriched in the ovary and they generate an anti-sakura antibody, a line with a GFP-tagged sakura transgene, and a sakura null allele to investigate sakura localization and function directly. They confirm the prediction that it is primarily expressed in the ovary and, specifically, that it is expressed in germ cells, and find that about 2/3 of the mutants lack germ cells completely and the remaining have tumorous ovaries. Further investigation reveals that Sakura is required for piRNA-mediated repression of transposons in germ cells. They also find evidence that sakura is important for germ cell specification during development and germline stem cell maintenance during adulthood. However, despite the role of sakura in maintaining germline stem cells, they find that sakura mutant germ cells also fail to differentiate properly such that mutant germline stem cell clones have an increased number of "GSC-like" cells. They attribute this phenotype to a failure in the repression of Bam by dpp signaling. Lastly, they demonstrate that sakura physically interacts with otu and that sakura and otu mutants have similar germ cell phenotypes. Overall, this study helps to advance the field by providing a characterization of a novel gene that is required for oogenesis. The data are generally high-quality and the new lines and reagents they generated will be useful for the field. However, there are some weaknesses and I would recommend that they address the comments in the Recommendations for the authors section below.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

General Comments:

(1) The gene nomenclature: As mentioned in the text, Sakura means cherry blossom and is one of the national flowers of Japan. I am not sure whether the phenotype of the CG14545 mutant is related to Sakura or not. I would like to suggest the authors reconsider the naming.

The striking phenotype of sakura mutant­ is tumorous and germless ovarioles. The tumorous phenotype, exhibiting lots of round fusome in germarium visualized by anti-Hts staining, looks like cherry blossom blooming to us. Also, the germless phenotype reminds us falling of the cherry blossom, especially considering that the ratio of tumorous phenotype decreases and that of germless decreases over fly age. Furthermore, “Sakura” symbolizes birth and renewal in Japanese culture (the last author of this manuscript is Japanese). Our findings indicated that the gene sakura is involved in regulation of renewal and differentiation of GSCs (which leads to birth). These are the reasons for the naming, which we would like to keep.

(2) In many of the microscopic photographs in the figures, especially for the merged confocal images, the resolution looks low, and the images appear blurred, making it difficult to judge the authors' claims. Also, the Alpha Fold structure in Figure 10A requires higher contrast images. The magnification of the images is often inadequate (e.g. Figures 3A, 3B, 5E, 7A, etc). The authors should take high-magnification images separately for the germarium and several different stages of the egg chambers and lay out the figures.

We are very sorry for the low-resolution images. This was caused when the original PDF file with high-resolution images was compressed in order to meet the small file size limit in the eLife submission portal. In the revised submission, we used high-resolution images.

Specific Comments

(1) How Sakura can cooperate with Otu remains unanswered. Sakura does not regulate deubiquitinase activity in vitro. Both sakura and otu appear to be involved in the Dpp-Smad signaling pathway and in the spatial control of Bam expression in the germarium, whereas Otu has been reported to act in concert with Bam to deubiquitinate and stabilize Cyc A for proper cystoblast differentiation. Therefore, it is plausible that the stabilization of Cyc A in the Sakura mutant is an indirect consequence of Bam misexpression and independent of the Sakura-Otu interaction. The authors may need to provide much deeper insight into the mechanism by which Sakura plays roles in these seemingly separable steps to orchestrate germ cell maintenance and differentiation during early oogenesis.

Yes, it is possible that the stabilization of CycA in the sakura mutant is an indirect consequence of Bam misexpression and independent of the Sakura-Otu interaction. To test the significance and role of the Sakura-Otu interaction, we have attempted to identify Sakura point mutants that lose interaction with Otu. If such point mutants were successfully obtained, we were planning to test if their transgene expression could rescue the phenotypes of sakura mutant as the wild-type transgene did. However, after designing and testing the interaction of over 30 point mutants with Otu, we could not obtain such mutant version of Sakura yet. We will continue making efforts, but it is beyond the scope of the current study. We hope to address this important point in future studies.

(2) Figure 3A and Figure 4: The authors show that piRNA production is abolished in Sakura KO ovaries. It is known that piRNA amplification (the ping-pong cycle) occurs in the Vasa-positive perinuclear nuage in nurse cells. Is the nuage normally formed in the absence of Sakura? The authors provide high-magnification images in the germarium expressing Vas-GFP. How does Sakura, and possibly Out, contribute to piRNA production? Are the defects a direct or indirect consequence of the loss of Sakura?

We provided higher magnification images of germarium expressing Vasa-EGFP in sakura mutant background (Fig 3A and 3B). The nuage formation does not seem to be dysregulated in sakura mutant. Currently, we do not know if the piRNA defects are direct or indirect consequence of the loss of Sakura. This question cannot be answered easily. We hope to address this in future studies.

(3) Figure 7 and Figure 12: The authors showed that Dpp-Smad signaling was abolished in Sakura KO germline cells. The same defects were also observed in otu mutant ovaries (Figure 12B). How does the Sakura-Otu axis contribute to the Dpp-Smad pathway in the germline?

As we mentioned in the response to comment (1), we attempted to test the significance and role of the Sakura-Otu interaction, including in the Dpp-Smad pathway in the germline, but we have not yet been able to obtain loss-of-interaction mutant(s) of Sakura. We hope to address this in future studies.

(4) Figure 9 and Fig 10: The authors raised antibodies against both Sakura and Otu, but their specificities were not provided. For Western blot data, the authors should provide whole gel images as source data files. Also, the authors argue that the Otu band they observed corresponds to the 98-kDa isoform (lines 302-304). The molecular weight on the Western blot alone would be insufficient to support this argument.

When we submitted the initial manuscript, we also submitted original, uncropped, and unmodified whole Western blot images for all gel images to the eLife journal, as requested. We did the same for this revised submission. I believe eLife makes all those files available for downloading to readers.

In the newly added Fig S13B, we used very young 2-5 hours ovaries and 3-7 days ovaries. 2-5 days ovaries contain only mostly pre-differentiated germ cells. Older ovaries (3-7 days in our case here) contain all 14 stages of oogenesis and later stages predominate in whole ovary lysates.

As reported in previous literature (Sass et al. 1995), we detected a higher abundance of the 104 kDa Otu isoform than the 98 kDa isoform in from 2-5 hours ovaries and predominantly the 98 kDa isoform in 3-7 days ovaries (Fig S13B). These results confirmed that the major Otu isoform we detected in Western blot, all of which uses old ovaries except for the 2-5 hours ovaries in Fig S13B, is the 98 kDa isoform.

(5) Otu has been reported to regulate ovo and Sxl in the female germline. Is Sakura involved in their regulation?

We examined sxl alternative splicing pattern in sakura mutant ovaries. As shown in Fig S6, we detected the male-specific isoform of sxl RNA and a reduced level of the female-specific sxl isoform in sakura mutant ovaries. Thus Sakura seems to be involved in sxl splicing in the female germline, while further studies will be needed to understand whether Sakura has a direct or indirect role here.

(6) Lines 443-447: The GSC loss phenotype in piwi mutant ovaries is thought to occur in a somatic cell-autonomous manner: both piwi-mutant germline clones and germline-specific piwi knockdown do not show the GSC-loss phenotype. In contrast, the authors provide compelling evidence that Sakura functions in the germline. Therefore, the Piwi-mediated GSC maintenance pathway is likely to be independent of the Sakura-Otu axis.

We changed the text accordingly.

Reviewer #2 (Recommendations for the authors):

Overall, this is a cleanly written manuscript, with some sentences/sections that are confusing the way they are constructed (i.e. Line 37-38, 334, section on Flp/FRT experiments).

We rewrote those sections to avoid confusion.

Comment for all merged image data: the quality of the merged images is very poor - the individual channels are better but should also be reprocessed for more resolved image data sets. Also, it would be helpful to have boundaries drawn in an individual panel to identify the regions of the germarium, as cartooned in Figure S1A (which should be brought into Figure 1) F-actin or Vsg staining would have helped throughout the manuscript to enhance the visualization of described phenotypes.

We are very sorry for the low-resolution images. This was caused when the original PDF file with high-resolution images was compressed in order to meet the small file size limit in the eLife submission portal. In the revised submission, we used high-resolution images.

We outlined the germarium in Fig 1E.

We brought the former FigS1 into Fig 1A.

We provided Phalloidin (F-Actin) staining images in Fig S7.

All p-values seem off. I recommend running the data through the student t-test again.

We used the student t-test to calculate p-values and confirmed that they are correct. We don’t understand why the reviewer thinks all p-values seem off.

In the original manuscript, as we mentioned in each figure legends, we used asterisk (*) to indicate p-value <0.05, without distinguishing whether it’s <0.001, <0.01< or <0.05.

Probably reviewer 2 is suggesting us to use ***, **, and *, to indicate p-value of <0.001, <0.01, and <0.05, respectively? If so, we now followed reviewer2’s suggestions.

Figure 1

(1) Within the text, C is mentioned before A.

We updated the text and now we mentioned Fig 1A before Fig 1C.

(2) B should be the supplemental figure.

We moved the former Fig 1B to Supplemental Figure 1.

(3) C - How were the different egg chamber stages selected in the WB? Naming them 'oocytes' is deceiving. Recommend labeling them as 'egg chambers', since an oocyte is claimed to be just the one-cell of that cyst.

We changed the labeling to egg chambers.

(4) Is the antibody not detecting Sakura in IF? There is no mention of this anywhere in the manuscript.

While our Sakura antibody detects Sakura in IF, it seems to detect some other proteins as well. Since we have Sakura-EGFP fly strain (which fully rescues sakura^null phenotypes) to examine Sakura expression and localization without such non-specific signal issues, we relied on Sakura-EGFP rather than anti-Sakura antibodies for IF.

(5) Expand on the reliance of the sakura-EGFP fly line. Does this overexpression cause any phenotypes?

sakura-EGFP does not cause any phenotypes in the background of sakura[+/+] and sakura[+/-].

(6) Line 95 "as shown below" is not clear that it's referencing panel D.

We now referenced Fig 1D.

(7) Re: Figures 1 E and F. There is no mention of Hts or Vasa proteins in the text.<br /> "Sakura-EGFP was not expressed in somatic cells such as terminal filament, cap cells, escort cells, or follicle cells (Figure 1E). In the egg chamber, Sakura-EGFP was detected in the cytoplasm of nurse cells and was enriched in developing oocytes (Figure 1F)". Outline these areas or label these structures/sites in the images. The color of Merge labels is confusing as the blue is not easily seen.

We mentioned Hts and Vasa in the text. We labeled the structures/sites in the images and updated the color labeling.

Figure 2

(1) Entire figure is not essential to be a main figure, but rather supplemental.

We don’t agree with the reviewer. We think that the female fertility assay data, where sakura null mutant exhibits strikingly strong phenotype, which was completely rescued by our Sakura-EGFP transgene, is very important data and we would like to present them in a main figure.

(2) 2A- one star (*) significance does not seem correct for the presented values between 0 and 100+.

In the original manuscript, as we mentioned in each figure legends, we used asterisk (*) to indicate p-value <0.05, without distinguishing whether it’s <0.001, <0.01< or <0.05.

Probably reviewer 2 is suggesting us to use ***, **, and *, to indicate p-value of <0.001, <0.01, and <0.05, respectively? If so, we now followed reviewer2’s suggestions.

(3) 2C images are extremely low quality. Should be presented as bigger panels.

We are very sorry for the low-resolution images. This was caused when the original PDF file with high-resolution images was compressed in order to meet the small file size limit in the eLife submission portal. In the revised submission, we used high-resolution images. We also presented as bigger panels.

Figure 3

(1) "We observed that some sakura^null /null ovarioles were devoid of germ cells ("germless"), while others retained germ cells (Fig 3A)" What is described is, that it is hard to see. Must have a zoomed-in panel.

We provided zoomed-in panels in Fig 3B

(2) C - The control doesn't seem to match. Must zoom in.

We provided matched control and also zoomed in.

(3) For clarity, separate the tumorous and germless images.

In the new image, only one tumorous and one germless ovarioles are shown with clear labeling and outline, for clarity.

(4) Use arrows to help clearly indicate the changes that occur. As they are presented, they are difficult to see.

We updated all the panels to enhance clarity.

(5) Line 158 seems like a strong statement since it could be indirect.

We softened the statement.

Figure 4

(1) Line 188-189 - Conclusion is an overstatement.

We softened the statement.

(2) Is the piRNA reduction due to a change in transcription? Or a direct effect by Sakura?

We do not know the answers to these questions. We hope to address these in future studies.

Figure 5

(1) D - It might make more sense if this graph showed % instead of the numbers.

We did not understand the reviewer’s point. We think using numbers, not %, makes more sense.

(2) Line 213 - explain why RNAi 2 was chosen when RNAi 1 looks stronger.

Fly stock of RNAi line 2 is much healthier than RNAi line 1 (without being driven Gal4) for some reasons. We had a concern that the RNAi line 1 might contain an unwanted genetic background. We chose to use the RNAi 2 line to avoid such an issue.

(3) In Line 218 there's an extra parenthesis after the PGC acronym.

We corrected the error.

(4) TOsk-Gal4 fly is not in the Methods section.

We mentioned TOsk-Gal4 in the Methods.

Figure 6:

(1) The FLP-FRT section must be rewritten.

We rewrote the FLP-FRT section.

(2) A - include statistics.

We included statistics using the chi-square test.

(3) B - is not recalled in the Results text.

We referred Fig 6B in the text.

(4) Line 232 references Figure 3, but not a specific panel.

We referred Fig 3A, 3C, 3D, and 3E, in the text.

Figure 7/8 - can go to Supplemental.

We moved Fig 8 to supplemental. However, we think Fig 7 data is important and therefore we would like to present them as a main figure.

(1) There should be CycA expression in the control during the first 4 divisions.

Yes, there is CycA expression observed in the control during the first 4 divisions, while it’s much weaker than in sakura^null clone.

(2) Helpful to add the dotted lines to delineate (A) as well.

We added a dotted outline for germarium in Fig 7A.

(3) Line 263 CycA is miswritten as CyA.

We corrected the typo.

Figure 9

(1) Otu antibody control?

We validated Otu antibody in newly added Fig 10C and Fig S13A.

(2) Which Sakura-EGFP line was used? sakura het. or null background? This isn't mentioned in the text, nor legend.

We used Sakura-EGFP in the background of sakura[+/+]. We added this information in the methods and figure legend.

(3) C - Why the switch to S2 cells? Not able to use the Otu antibody in the IP of ovaries?

We can use the Otu antibody in the IP of ovaries. However, in anti-Sakura Western after anti-Otu IP, antibody light chain bands of the Otu antibodies overlap with the Sakura band. Therefore, we switched to S2 cells to avoid this issue by using an epitope tag.

Figure 10

(1) A- The resolution of images of the ribbon protein structure is poor.

We are very sorry for the low-resolution images. This was caused when the original PDF file with high-resolution images was compressed in order to meet the small file size limit in the eLife submission portal. In the revised submission, we used high-resolution images.

(2) A table summarizing the interactions between domains would help bring clarity to the data presented.

We added a table summarizing the fragment interaction results.

(3) Some images would be nice here to show that the truncations no longer colocalize.

We did not understand the reviewer’s points. In our study, even for the full-length proteins.

We have not shown any colocalization of Sakura and Otu in S2 cells or in ovaries, except that they both are enriched in developing oocytes in egg chambers.

Figure 12

(1) A - control and RNAi lines do not match.

We provided matched images.

(2) In general, since for Sakura, only its binding to Otu was identified and since they phenocopy each other, doesn't most of the characterization of Sakura just look at Otu phenotypes? Does Sakura knockdown affect Otu localization or expression level (and vice versa)?

We tested this by Western (Fig S15) and IF (Fig 12). Sakura knockdown did not decrease Otu protein level, and Otu knockdown did not decrease Sakura protein level (Fig S15). In sakura^null clone, Otu level was not notably affected (Fig 12). In sakura^null clone, Otu lost its localization to the posterior position within egg chambers.

Figure S6

(1) It is Luciferase, not Lucifarase.

We corrected the typo.

Reviewer #3 (Recommendations for the authors):

(1) It is interesting that germless and tumorous phenotypes coexist in the same population of flies. Additional consideration of these essentially opposite phenotypes would significantly strengthen the study. For example, do they co-exist within the same fly and are the tumorous ovarioles present in newly eclosed flies or do they develop with age? The data in Figure 8 show that bam knockdown partially suppresses the germless phenotype. What effect does it have on the tumorous phenotype? Is transposon expression involved in either phenotype? Do Sakura mutant germline stem cell clones overgrow relative to wild-type cells in the same ovariole? Does sakura RNAi driven by NGT-Gal4 only cause germless ovaries or does it also cause tumorous phenotypes? What happens if the knockdown of Sakura is restricted to adulthood with a Gal80ts? It may not be necessary to answer all of these questions, but more insight into how these two phenotypes can be caused by loss of sakura would be helpful.

We performed new experiments to answer these questions.

do they co-exist within the same fly and are the tumorous ovarioles present in newly eclosed flies or do they develop with age?

Tumorous and germless ovarioles coexist in the same fly (in the same ovary). Tumorous ovarioles are present in very young (0-1 day old) flies, including newly eclosed (Fig S5). The ratio of germless ovarioles increases and that of tumorous ovarioles decreases with age (Fig S5).

The data in Figure 8 show that bam knockdown partially suppresses the germless phenotype. What effect does it have on the tumorous phenotype?

bam knockdown effect on tumorous phenotype is shown in Fig S10. bam knockdown increased the ratio of tumorous ovarioles and the number of GSC-like cells.

Is transposon expression involved in either phenotype?

Since our transposon-piRNA reporter uses germline-specific nos promoter, it is expressed only in germ line cells, so we cannot examine in germless ovarioles.

Do Sakura mutant germline stem cell clones overgrow relative to wild-type cells in the same ovariole?

Yes, Sakura mutant GSC clones overgrow. Please compare Fig 6C and Fig S8.

Does sakura RNAi driven by NGT-Gal4 only cause germless ovaries or does it also cause tumorous phenotypes?

Fig S10 and Fig S12 show the ovariole phenotypes of sakura RNAi driven by NGT-Gal4. It causes both germless and tumorous phenotypes.

What happens if the knockdown of Sakura is restricted to adulthood with a Gal80ts?

Our mosaic clone was induced at the adult stage, so we already have data of adulthood-specific loss of function. Gal80ts does not work well with nos-Gal4.

(2) The idea that the excessive bam expression in tumorous ovaries is due to a failure of bam repression by dpp signaling is not well-supported by the data. Dpp signaling is activated in a very narrow region immediately adjacent to the niche but the images in Figure 7A show bam expression in cells that are very far away from the niche. Thus, it seems more likely to be due to a failure to turn bam expression off at the 16-cell stage than to a failure to keep it off in the niche region. To determine whether bam repression in the niche region is impaired, it would be important to examine cells adjacent to the niche directly at a higher magnification than is shown in Figure 7A.

We provided higher magnification images of cells adjacent to the niche in new Fig 7A.

We found that cells adjacent to the niche also express Bam-GFP.

That said, we agree with the reviewer. A failure to turn bam expression off at the 16-cell stage may be an additional or even a main cause of bam misexpression in sakura mutant. We added this in the Discussion.

(3) In addition, several minor comments should be addressed:

a. Does anti-Sakura work for immunofluorescence?

While our Sakura antibody detects Sakura in IF, it seems to detect some other proteins as well. Since we have Sakura-EGFP fly strain to examine Sakura expression and localization without such non-specific signal issues, we relied on Sakura-EGFP rather than anti-Sakura antibodies.

b. Please provide insets to show the phenotypes indicated by the different color stars in Figure 3C more clearly.

We provided new, higher-magnification images to show the phenotypes more clearly.

c. Please indicate the frequency of the expression patterns shown in Figure 4D (do all ovarioles in each genotype show those patterns or is there variable penetrance?).

We indicated the frequency.

d. An image showing TOskGal4 driving a fluorophore should be provided so that readers can see which cells express Gal4 with this driver combination.

It has been already done in the paper ElMaghraby et al, GENETICS, 2022, 220(1), iyab179, so we did not repeat the same experiment.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

Mallimadugula et al. combined Molecular Dynamics (MD) simulations, thiol-labeling experiments, and RNA-binding assays to study and compare the RNA-binding behavior of the Interferon Inhibitory Domain (IID) from Viral Protein 35 (VP35) of Zaire ebolavirus, Reston ebolavirus, and Marburg marburgvirus. Although the structures and sequences of these viruses are similar, the authors suggest that differences in RNA binding stem from variations in their intrinsic dynamics, particularly the opening of a cryptic pocket. More precisely, the dynamics of this pocket may influence whether the IID binds to RNA blunt ends or the RNA backbone.

Overall, the authors present important findings to reveal how the intrinsic dynamics of proteins can influence their binding to molecules and, hence, their functions. They have used extensive biased simulations to characterize the opening of a pocket which was not clearly seen in experimental results - at least when the proteins were in their unbound forms. Biochemical assays further validated theoretical results and linked them to RNA binding modes. Thus, with the combination of biochemical assays and state-of-the-art Molecular Dynamics simulations, these results are clearly compelling.

Strengths:

The use of extensive Adaptive Sampling combined with biochemical assays clearly points to the opening of the Interferon Inhibitory Domain (IID) as a factor for RNA binding. This type of approach is especially useful to assess how protein dynamics can affect its function.

Weaknesses:

Although a connection between the cryptic pocket dynamics and RNA binding mode is proposed, the precise molecular mechanism linking pocket opening to RNA binding still remains unclear.

Reviewer #2 (Public review):

Summary:

The authors aimed to determine whether a cryptic pocket in the VP35 protein of Zaire ebolavirus has a functional role in RNA binding and, by extension, in immune evasion. They sought to address whether this pocket could be an effective therapeutic target resistant to evolutionary evasion by studying its role in dsRNA binding among different filovirus VP35 homologs. Through simulations and experiments, they demonstrated that cryptic pocket dynamics modulate the RNA binding modes, directly influencing how VP35 variants block RIG-I and MDA5-mediated immune responses.

The authors successfully achieved their aim, showing that the cryptic pocket is not a random structural feature but rather an allosteric regulator of dsRNA binding. Their results not only explain functional differences in VP35 homologs despite their structural similarity but also suggest that targeting this cryptic pocket may offer a viable strategy for drug development with reduced risk of resistance.

This work represents a significant advance in the field of viral immunoevasion and therapeutic targeting of traditionally "undruggable" protein features. By demonstrating the functional relevance of cryptic pockets, the study challenges long-standing assumptions and provides a compelling basis for exploring new drug discovery strategies targeting these previously overlooked regions.

Strengths:

The combination of molecular simulations and experimental approaches is a major strength, enabling the authors to connect structural dynamics with functional outcomes. The use of homologous VP35 proteins from different filoviruses strengthens the study's generality, and the incorporation of point mutations adds mechanistic depth. Furthermore, the ability to reconcile functional differences that could not be explained by crystal structures alone highlights the utility of dynamic studies in uncovering hidden allosteric features.

Weaknesses:

While the methodology is robust, certain limitations should be acknowledged. For example, the study would benefit from a more detailed quantitative analysis of how specific mutations impact RNA binding and cryptic pocket dynamics, as this could provide greater mechanistic insight. This study would also benefit from providing a clear rationale for the selection of the amber03 force field and considering the inclusion of volume-based approaches for pocket analysis. Such revisions will strengthen the robustness and impact of the study.

Reviewer #3 (Public review):

Summary:

The authors suggest a mechanism that explains the preference of viral protein 35 (VP35) homologs to bind the backbone of double-stranded RNA versus blunt ends. These preferences have a biological impact in terms of the ability of different viruses to escape the immune response of the host.

The proposed mechanism involves the existence of a cryptic pocket, where VP35 binds the blunt ends of dsRNA when the cryptic pocket is closed and preferentially binds the RNA double-stranded backbone when the pocket is open.

The authors performed MD simulation results, thiol labelling experiments, fluorescence polarization assays, as well as point mutations to support their hypothesis.

Strengths:

This is a genuinely interesting scientific question, which is approached through multiple complementary experiments as well as extensive MD simulations. Moreover, structural biology studies focused on RNA-protein interactions are particularly rare, highlighting the importance of further research in this area.

Weaknesses:

- Sequence similarity between Ebola-Zaire (94% similarity) explains their similar behaviour in simulations and experimental assays. Marburg instead is a more distant homolog (~80% similarity relative to Ebola/Zaire). This difference is sequence and structure can explain the propensities, without the need to involve the existence of a cryptic pocket.  

- No real evidence for the presence of a cryptic pocket is presented, but rather a distance probability distribution between two residues obtained from extensive MD simulations. It would be interesting to characterise the modelled RNA-protein interface in more detail

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Before assessing the overall quality and significance of this work, this reviewer needs to specify the context of this review. This reviewer's expertise lies in biased and unbiased molecular dynamics simulations and structural biology. Hence, while this reviewer can overall understand the results for thiol-labeling and RNA-binding assays, this review will not assess the quality of these biochemical assays and will mainly focus on the modelling results.

Overall, the authors present important findings to reveal how the intrinsic dynamics of proteins can influence their binding to molecules and, hence, their functions. They have used extensive biased simulations to characterize the opening of a pocket which was not clearly seen in experimental results - at least when the proteins were in their unbound forms. Biochemical assays further validated theoretical results and linked them to RNA binding modes. Thus, with the combination of biochemical assays and state-of-the-art Molecular Dynamics simulations, these results are clearly compelling.

Beyond the clear qualities of this work, I would like to mention a few points that may help to better contextualize and rationalize the results presented here.

- First, both the introduction and discussion sections seem relatively condensed. Extending them to, for example, better describe the methodological context and discuss the methodological limitations and potential future developments related to biased simulations may help the reader get a better idea of the significance of this work.

- The authors presented 3 homologs in this study: IIDs of Reston, Zaire, and Marburg viruses. While Zaire and Reston are relatively similar in terms of sequence (Figure S1). The sequences clearly differ between Marburg and the two other viruses. Can the author indicate a similarity/identity score for each sequence alignment and extend Figure S1 to really compare Marburg sequence with Reston and Zaire? Can they also discuss how these differences may impact the comparison of the three IIDs? This may also help the reader to understand why sometimes the authors compare the three viruses and why sometimes they are focusing only on comparing Zaire and Reston.

We would like to thank the reviewer for raising this point and we agree that additional details about the sequence comparison provide more context for the choices of substitutions we made. Therefore, we have updated Fig S1 to include a detailed pairwise comparison of all the IID sequences including the percentage sequence similarity and identity. We have also added the following sentences to the results section where we first introduced the substitutions between Zaire and Reston IIDs

“While the sequence of Marburg IID differs significantly from Reston and Zaire IIDs with a sequence identity of 42% and 45% respectively (Fig S1), the sequences of Reston and Zaire IID are 88% identical and 94% similar. Particularly, substitutions between these homologs are all distal to the RNA-binding interfaces and all the residues known to make contacts with dsRNA from structural studies are identical. Therefore, we reasoned that comparing these two homologs would help us identify minimal substitutions that control pocket opening probability and allow us to study its effect on dsRNA binding with minimal perturbation of other factors.”

- In this work, the authors mentioned the cryptic pocket but only illustrated the opening of this pocket by using a simple distance between residues (Figure 2) and a SASA of one cysteine (Figure 3). In previous work done by the authors (Cruz et al. , Nature Communications, 2022), they better characterized residues involved in RNA binding and forming the cryptic pocket. Thus, would it be possible to better described this cryptic pocket (residues involved, volume, etc ..) and better explain how, structurally speaking, it can affect RNA binding mode (blunt ends vs backbone) ?

We thank the reviewer for pointing out the need for clarification on the residues involved in RNA binding and pocket opening and the mechanism linking them. We have performed the CARDS analysis on Reston and Marburg IID simulations as we had done on Zaire IID simulations in Cruz et al, 2022. The results are shown in Fig S3 and discussed in the main text in the first results section.

- As a counter-example, the authors used C315 for SASA calculation and thiol labeling (Figure 3). This cysteine is mainly buried as seen by SASA for Reston and Marburg and thiol labelling (Figure 3 E,G,H). Would it be possible to also get thiol labeling rates for Cystein 264 in Reston and its equivalent to see a case where the residue is solvent exposed?

We have shown the SASA for C264 from the simulations in Fig S4 and the thiol labeling rates for all 4 cysteines in Reston IID in Fig S6. Comparing these rates to the rates of all 4 cysteines obtained for Zaire IID (Fig 4 in Cruz et Al, 2022), we observe that the rates for C264, which is expected to be exposed are significantly faster than those of C315 which is largely buried in all variants.  

- I strongly support here the will of the authors to share their data by depositing them in an OSF repository. These data help this reviewer to assess some of the results produced by the authors and help to better understand the dynamics of their respective systems. I have just a few comments that need to be addressed regarding these data: o While there are data for WT Reston and Marburg, there is no data for Zaire. Is this because these data correspond to the previous work (Cruz et al. 2022) (in this case, it would be good to make this clear in the main text) or is it an omission? o There is no center.xtc file in the Marburg-MSM directory o There is no protmasses.pdb in the Reston-MSM directory

- In general, if possible, it would be good to use the same name for each type of file presented in each directory to help a potential user understand a bit more how to use these data.

- If possible, adding a bit more of metadata and explanations on the OSF webpage would be very beneficial to help find these data. To help in this direction, the authors may have a look to the guidelines presented at the end of this article: https://elifesciences.org/articles/90061

We thank the reviewer for pointing out the omissions from the OSF repository. We have added the missing files and followed a uniform naming convention. We have also added documentation in the metadata section of the OSF repository to help others use the data.  

Indeed, the simulation data used for Zaire IID is available on the OSF repository corresponding to Cruz et al. 2022 at https://osf.io/5pg2a. We have also clarified this in the data availability section of the main text.  

Minor point:

In Figure 2, there is a slight bump for the 225-295 distance around 1 nm for Reston. Can the author comment it ? As these results are based on long AS, even if very small, do the authors think this population is significant?

Comparing the probability distributions obtained from bootstrapping the frames used to calculate the MSM equilibrium probabilities (Revised Fig1), we observe that the bump for the Reston IID distribution is persistent in all bootstraps indicating that it might indeed be significant. This is also consistent with our observation that the cysteine 296 does get fully labeled in our thiol labeling experiments, albeit significantly slowly compared to the other homologs.  

Reviewer #2 (Recommendations for the authors):

I recommend that the authors implement moderate revisions prior to the publication of this research article, addressing the identified weaknesses (see below).

The authors should provide a rationale for their selection of the amber03 force field (Duan et al., JCTC 24, 1999-2012, 2003) for molecular dynamics simulations, particularly given the availability of more recent and optimized versions of the AMBER force fields. These newer force fields may offer improved parameterization for biomolecular systems, potentially enhancing the accuracy and reliability of the simulation results.

We chose the Amber03 force field because it has performed well in much of our past work, including the original prediction of the cryptic pocket that we study in this manuscript. The results presented in this manuscript also demonstrate the predictive power of Amber03.

Additionally, while the authors utilized solvent-accessible surface area (SASA) for cryptic pocket analysis, volume-based approaches may be more suitable for this purpose. Several studies (e.g., Sztain et al. J. Chem. Inf. Model. 2021, 61, 7, 3495-3501) have demonstrated the utility of volume analysis in identifying and characterizing cryptic pockets. The authors could consider incorporating such methodologies to provide a more comprehensive assessment of pocket dynamics.

The authors propose that the cryptic pocket is not merely a random structural feature but functions as an allosteric regulator of dsRNA binding. To further substantiate this claim, an in-depth analysis of this allosteric effect using for instance network analysis could significantly enhance the study. Such an approach could identify key residues and interaction networks within the protein that mediate the allosteric regulation. This type of mechanistic insight would not only provide a stronger theoretical framework but also offer valuable information for the rational design of therapeutic interventions targeting the cryptic pocket.  

We thank the reviewer for pointing out the need for clarification on the molecular mechanism linking the opening of the cryptic pocket to RNA binding. We have performed the CARDS analysis on Reston and Marburg IID simulations as was done on Zaire IID simulations in Cruz et al, 2022. The results are shown in Fig S3 and discussed in the main text in the first results section. Briefly, we do find a community (blue) comprising the pocket residues in Reston and Marburg IIDs as we did in Zaire. Similarly, we find that many of the RNA binding residues fall into the orange and green communities as in Zaire. However, there are differences in exactly which residues are clustered into which of these two communities. There are also differences in how strongly connected these communities are in the three homologs. Therefore, while we can conclude that pocket residues likely have varying influence on the RNA binding residues in the homologs, it is hard to say exactly what that variation is from this analysis alone.  

Reviewer #3 (Recommendations for the authors):

- MD simulations: All simulations were initialised from the 3 crystal structures, is it correct? In all cases, RNA ds was not included in simulations, right? Were crystallographic MG ions in the vicinity of the binding site included? these are known to influence structural dynamics to a large extent.

All simulations were indeed initialized using only protein atoms from the crystal structures 3FKE, 4GHL, and 3L2A. Therefore, crystallographic Mg ions were not included in the simulations. However, we do agree with the reviewer and think that the effect of parameters such as salt concentration, specifically Mg ions which are known to be important for the stability of dsRNA, on the pocket opening equilibrium merits detailed study in future work.

- Figure 2: Would it be possible to perform e.g. a block error analysis and show the statistical errors of the distributions?

We agree that showing the statistical variation in the MSM equilibrium probabilities is important for comparing the different distributions. Therefore, we have updated Figs 2 and 5 to show the distributions obtained from MSMs constructed using 100 and 10 random samples of the data respectively to indicate the extent of the statistical variability in the MSM construction.  

- More detailed structural biology experiments (such as NMR or HDX-MS) could potentially shed more light on the differential behaviour of the three different homologs, providing more evidence for the presence of the cryptic pocket.

We agree that NMR and HDX-MS are powerful means to study dynamics and are actively exploring these approaches for our future work.

Author response:

Reviewer #1:

We appreciate the Reviewer's positive feedback on the strengths of our study.

The timescales of the peptide recognition and unbinding process are much longer than what can be sampled from unbiased simulations. Therefore, the proposed mechanism of recognition should only be considered a hypothesis based on the results presented here. For example, peptides that do not dissociate within one one-microsecond MD simulation are considered to be stable binders. However, they may not have a viable way to bind to the narrow protein cleft in the first place.

We thank the Reviewer for this valuable feedback. We agree with the Reviewer. Our work on the IRE1 cLD activation mechanism is focused on generating hypotheses of the binding mechanism driven by MD simulations. We recognize the limitations in defining a stable binder due to the time scales sampled. However, our primary focus was to sample and characterize a possible binding pose in the center of the cLD dimer. We will contextualize our statements about stable binders and limit our claims to stating that the protein-peptide complex is stable within 1 μs-long simulations. However, we believe that our finding that the cLD dimer groove is not able to accommodate peptides is solid, as the steric impediment described is present in all our replicas, both with and without peptides, in a cumulative sampling time of 72 μs. Additionally, we will include a plot showing the distribution of groove width across all replicas.

Oftentimes, representative structures sampled from MD simulation are used to draw conclusions (e.g., Figure 4 about the role of R161 mutation in binding affinity). This is not appropriate as one unbinding event being observed or not observed in a microsecond-long trajectory does not provide sufficient information about the binding strength of the free energy difference.

We thank the Reviewer for the insightful comment. As explained in the previous point, we believe that our simulations provide useful hypotheses, and we agree that we do not currently have data to comment on binding affinity. We will, therefore, remove all references to this term. We are aware of the limitations due to the timescale and agree that these limitations cannot be overcome with standard equilibrium simulations. To address these limitations, we plan to use orthogonal methods, namely MM/PB(GB)SA calculations for calculating binding free energies from existing trajectories (as performed by https://doi.org/10.1021/acs.jcim.4c00975). We will add predictions of all the peptides using AlphaFold 3, to confirm the binding region.

Reviewer #2:

We thank the Reviewer for their positive feedback.

Improving presentation to include more computational details.

We thank the Reviewer for raising this critical point. We agree that the manuscript is tailored for a biology audience, as the data are particularly relevant for that community. Nevertheless, we also understand the importance of providing sufficient methodological detail for computational readers. We will add appropriate computational information in the main text.

More quantitative analysis in addition to visual structures.

We will add an uncertainty estimate for the HDX calculations using bootstrapping and include additional information on bond distances for Y161. We will also incorporate time-series data showing the distance of the peptide from the groove across all replicas.

Reviewer #3:

We appreciate the Reviewer's positive feedback on our work.

A potential weakness of the study is the usage of equilibrium (unbiased) molecular dynamics simulations so that processes and conformational changes on the microsecond time scale can be probed. Furthermore, there can be inaccuracies and biases in the description of unfolded peptides and protein segments due to the protein force fields. Here, it should be noted that the authors do acknowledge these possible limitations of their study in the conclusions.

We appreciate the Reviewer's thoughtful comment. As noted in our response to Reviewer 1, we plan to address the concern about sampling by applying orthogonal methods. We agree with the Reviewer that some form of enhanced sampling is necessary if we want to assess binding in a more quantitative way, e.g., via free energy calculations. However, we also realize that applying any enhanced sampling scheme to our system is very challenging, given its large size and the complex peptide-protein interactions, which are not easily captured in a few collective variables. After a careful assessment and some preliminary tests, we decided that estimating free energies using enhanced sampling would necessitate a separate paper due to both the conceptual complexity of the project and the size of the necessary sampling campaign.

Author response:

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

We wanted to clarify Reviewer #1’s latest comment in the last round of review, “Furthermore, the referee appreciates that the authors have echoed the concern regarding the limited statistical robustness of the observed scrambling events.” We appreciate the follow up information provided from Reviewer #1 that their comment is specifically about the low count alternative pathway events that we view at the dimer interface, and not the statistics of the manuscript overall as they believe that “the study presents a statistically rigorous analysis of lipid scrambling events across multiple structures and conformations (Reviewer #1)”. We agree with the Reviewer and acknowledge that overall our coarse-grained study represents the most comprehensive single manuscript of the entire TMEM16 family to date.

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

Public Review:

Reviewer #1 (Public review):

Summary:

The manuscript investigates lipid scrambling mechanisms across TMEM16 family members using coarse-grained molecular dynamics (MD) simulations. While the study presents a statistically rigorous analysis of lipid scrambling events across multiple structures and conformations, several critical issues undermine its novelty, impact, and alignment with experimental observations.

Critical issues:

(1) Lack of Novelty:

The phenomenon of lipid scrambling via an open hydrophilic groove is already well-established in the literature, including through atomistic MD simulations. The authors themselves acknowledge this fact in their introduction and discussion. By employing coarse-grained simulations, the study essentially reiterates previously known findings with limited additional mechanistic insight. The repeated observation of scrambling occurring predominantly via the groove does not offer significant advancement beyond prior work.

We agree with the reviewer’s statement regarding the lack of novelty when it comes to our observations of scrambling in the groove of open Ca2+-bound TMEM16 structures. However, we feel that the inclusion of closed structures in this study, which attempts to address the yet unanswered question of how scrambling by TMEM16s occurs in the absence of Ca2+, offers new observations for the field. In our study we specifically address to what extent the induced membrane deformation, which has been theorized to aid lipids cross the bilayer especially in the absence of Ca2+, contributes to the rate of scrambling (see references 36, 59, and 66). There are also several TMEM16F structures solved under activating conditions (bound to Ca2+ and in the presence of PIP2) which feature structural rearrangements to TM6 that may be indicative of an open state (PDB 6P48) and had not been tested in simulations. We show that these structures do not scramble and thereby present evidence against an out-of-the-groove scrambling mechanism for these states. Although we find a handful of examples of lipids being scrambled by Ca2+-free structures of TMEM16 scramblases, none of our simulations suggest that these events are related to the degree of deformation.

(2) Redundancy Across Systems:

The manuscript explores multiple TMEM16 family members in activating and non-activating conformations, but the conclusions remain largely confirmatory. The extensive dataset generated through coarse-grained MD simulations primarily reinforces established mechanistic models rather than uncovering fundamentally new insights. The effort, while statistically robust, feels excessive given the incremental nature of the findings.

Again, we agree with the reviewer’s statement that our results largely confirm those published by other groups and our own. We think there is however value in comparing the scrambling competence of these TMEM16 structures in a consistent manner in a single study to reduce inconsistencies that may be introduced by different simulation methods, parameters, environmental variables such as lipid composition as used in other published works of single family members. The consistency across our simulations and high number of observed scrambling events have allowed us to confirm that the mechanism of scrambling is shared by multiple family members and relies most obviously on groove dilation.

(3) Discrepancy with Experimental Observations:

The use of coarse-grained simulations introduces inherent limitations in accurately representing lipid scrambling dynamics at the atomistic level. Experimental studies have highlighted nuances in lipid permeation that are not fully captured by coarse-grained models. This discrepancy raises questions about the biological relevance of the reported scrambling events, especially those occurring outside the canonical groove.

We thank the reviewer for bringing up the possible inaccuracies introduced by coarse graining our simulations. This is also a concern for us, and we address this issue extensively in our discussion. As the reviewer pointed out above, our CG simulations have largely confirmed existing evidence in the field which we think speaks well to the transferability of observations from atomistic simulations to the coarse-grained level of detail. We have made both qualitative and quantitative comparisons between atomistic and coarse-grained simulations of nhTMEM16 and TMEM16F (Figure 1, Figure 4-figure supplement 1, Figure 4-figure supplement 5) showing the two methods give similar answers for where lipids interact with the protein, including outside of the canonical groove. We do not dispute the possible discrepancy between our simulations and experiment, but our goal is to share new nuanced ideas for the predicted TMEM16 scrambling mechanism that we hope will be tested by future experimental studies.

(4) Alternative Scrambling Sites:

The manuscript reports scrambling events at the dimer-dimer interface as a novel mechanism. While this observation is intriguing, it is not explored in sufficient detail to establish its functional significance. Furthermore, the low frequency of these events (relative to groove-mediated scrambling) suggests they may be artifacts of the simulation model rather than biologically meaningful pathways.

We agree with the reviewer that our observed number of scrambling events in the dimer interface is too low to present it as strong evidence for it being the alternative mechanism for Ca2+-independent scrambling. This will require additional experiments and computational studies which we plan to do in future research. However, we are less certain that these are artifacts of the coarse-grained simulation system as we observed a similar event in an atomistic simulation of TMEM16F.

Conclusion:

Overall, while the study is technically sound and presents a large dataset of lipid scrambling events across multiple TMEM16 structures, it falls short in terms of novelty and mechanistic advancement. The findings are largely confirmatory and do not bridge the gap between coarse-grained simulations and experimental observations. Future efforts should focus on resolving these limitations, possibly through atomistic simulations or experimental validation of the alternative scrambling pathways.

Reviewer #2 (Public review):

Summary:

Stephens et al. present a comprehensive study of TMEM16-members via coarse-grained MD simulations (CGMD). They particularly focus on the scramblase ability of these proteins and aim to characterize the "energetics of scrambling". Through their simulations, the authors interestingly relate protein conformational states to the membrane's thickness and link those to the scrambling ability of TMEM members, measured as the trespassing tendency of lipids across leaflets. They validate their simulation with a direct qualitative comparison with Cryo-EM maps.

Strengths:

The study demonstrates an efficient use of CGMD simulations to explore lipid scrambling across various TMEM16 family members. By leveraging this approach, the authors are able to bypass some of the sampling limitations inherent in all-atom simulations, providing a more comprehensive and high-throughput analysis of lipid scrambling. Their comparison of different protein conformations, including open and closed groove states, presents a detailed exploration of how structural features influence scrambling activity, adding significant value to the field. A key contribution of this study is the finding that groove dilation plays a central role in lipid scrambling. The authors observe that for scrambling-competent TMEM16 structures, there is substantial membrane thinning and groove widening. The open Ca2+-bound nhTMEM16 structure (PDB ID 4WIS) was identified as the fastest scrambler in their simulations, with scrambling rates as high as 24.4 {plus minus} 5.2 events per μs. This structure also shows significant membrane thinning (up to 18 Å), which supports the hypothesis that groove dilation lowers the energetic barrier for lipid translocation, facilitating scrambling.

The study also establishes a correlation between structural features and scrambling competence, though analyses often lack statistical robustness and quantitative comparisons. The simulations differentiate between open and closed conformations of TMEM16 structures, with open-groove structures exhibiting increased scrambling activity, while closed-groove structures do not. This finding aligns with previous research suggesting that the structural dynamics of the groove are critical for scrambling. Furthermore, the authors explore how the physical dimensions of the groove qualitatively correlate with observed scrambling rates. For example, TMEM16K induces increased membrane thinning in its open form, suggesting that membrane properties, along with structural features, play a role in modulating scrambling activity.

Another significant finding is the concept of "out-of-the-groove" scrambling, where lipid translocation occurs outside the protein's groove. This observation introduces the possibility of alternate scrambling mechanisms that do not follow the traditional "credit-card model" of groove-mediated lipid scrambling. In their simulations, the authors note that these out-of-the-groove events predominantly occur at the dimer interface between TM3 and TM10, especially in mammalian TMEM16 structures. While these events were not observed in fungal TMEM16s, they may provide insight into Ca2+-independent scrambling mechanisms, as they do not require groove opening.

Weaknesses:

A significant challenge of the study is the discrepancy between the scrambling rates observed in CGMD simulations and those reported experimentally. Despite the authors' claim that the rates are in line experimentally, the observed differences can mean large energetic discrepancies in describing scrambling (larger than 1kT barrier in reality). For instance, the authors report scrambling rates of 10.7 events per μs for TMEM16F and 24.4 events per μs for nhTMEM16, which are several orders of magnitude faster than experimental rates. While the authors suggest that this discrepancy could be due to the Martini 3 force field's faster diffusion dynamics, this explanation does not fully account for the large difference in rates. A more thorough discussion on how the choice of force field and simulation parameters influence the results, and how these discrepancies can be reconciled with experimental data, would strengthen the conclusions. Likewise, rate calculations in the study are based on 10 μs simulations, while experimental scrambling rates occur over seconds. This timescale discrepancy limits the study's accuracy, as the simulations may not capture rare or slow scrambling events that are observed experimentally and therefore might underestimate the kinetics of scrambling. It's however important to recognize that it's hard (borderline unachievable) to pinpoint reasonable kinetics for systems like this using the currently available computational power and force field accuracy. The faster diffusion in simulations may lead to overestimated scrambling rates, making the simulation results less comparable to real-world observations. Thus, I would therefore read the findings qualitatively rather than quantitatively. An interesting observation is the asymmetry observed in the scrambling rates of the two monomers. Since MARTINI is known to be limited in correctly sampling protein dynamics, the authors - in order to preserve the fold - have applied a strong (500 kJ mol-1 nm-2) elastic network. However, I am wondering how the ENM applies across the dimer and if any asymmetry can be noticed in the application of restraints for each monomer and at the dimer interface. How can this have potentially biased the asymmetry in the scrambling rates observed between the monomers? Is this artificially obtained from restraining the initial structure, or is the asymmetry somehow gatekeeping the scrambling mechanism to occur majorly across a single monomer? Answering this question would have far-reaching implications to better describe the mechanism of scrambling.

The main aim of our computational survey was to directly compare all relevant published TMEM16 structures in both open and closed states using the Martini 3 CGMD force field. Our standardized simulation and analysis protocol allowed us to quantitatively compare scrambling rates across the TMEM16 family, something that has never been done before. We do acknowledge that direct comparison between simulated versus experimental scrambling rates is complicated and is best to be interpreted qualitatively. In line with other reports (e.g., Li et al, PNAS 2024), lipid scrambling in CGMD is 2-3 orders of magnitude faster than typical experimental findings. In the CG simulation field, these increased dynamics due to the smoother energy landscape are a well known phenomenon. In our view, this is a valuable trade-off for being able to capture statistically robust scrambling dynamics and gain mechanistic understanding in the first place, since these are currently challenging to obtain otherwise. For example, with all-atom MD it would have been near-impossible to conclude that groove openness and high scrambling rates are closely related, simply because one would only measure a handful of scrambling events in (at most) a handful of structures.

Considering the elastic network: the reviewer is correct in that the elastic network restrains the overall structure to the experimental conformation. This is necessary because the Martini 3 force field does not accurately model changes in secondary (and tertiary) structure. In fact, by retaining the structural information from the experimental structures, we argue that the elastic network helped us arrive at the conclusion that groove openness is the major contributing factor in determining a protein’s scrambling rate. This is best exemplified by the asymmetric X-ray structure of TMEM16K (5OC9), in which the groove of one subunit is more dilated than the other. In our simulation, this information was stored in the elastic network, yielding a 4x higher rate in the open groove than in the closed groove, within the same trajectory.

Notably, the manuscript does not explore the impact of membrane composition on scrambling rates. While the authors use a specific lipid composition (DOPC) in their simulations, they acknowledge that membrane composition can influence scrambling activity. However, the study does not explore how different lipids or membrane environments or varying membrane curvature and tension, could alter scrambling behaviour. I appreciate that this might have been beyond the scope of this particular paper and the authors plan to further chase these questions, as this work sets a strong protocol for this study. Contextualizing scrambling in the context of membrane composition is particularly relevant since the authors note that TMEM16K's scrambling rate increases tenfold in thinner membranes, suggesting that lipid-specific or membrane-thickness-dependent effects could play a role.

Considering different membrane compositions: for this study, we chose to keep the membranes as simple as possible. We opted for pure DOPC membranes, because it has (1) negligible intrinsic curvature, (2) forms fluid membranes, and (3) was used previously by others (Li et al, PNAS 2024). As mentioned by the reviewer, we believe our current study defines a good, standardized protocol and solid baseline for future efforts looking into the additional effects of membrane composition, tension, and curvature that could all affect TMEM16-mediated lipid scrambling.

Reviewer #3 (Public review):

Strengths:

The strength of this study emerges from a comparative analysis of multiple structural starting points and understanding global/local motions of the protein with respect to lipid movement. Although the protein is well-studied, both experimentally and computationally, the understanding of conformational events in different family members, especially membrane thickness less compared to fungal scramblases offers good insights.

We appreciate the reviewer recognizing the value of the comparative study. In addition to valuable insights from previous experimental and computational work, we hope to put forward a unifying framework that highlights various TMEM16 structural features and membrane properties that underlie scrambling function.

Weaknesses:

The weakness of the work is to fully reconcile with experimental evidence of Ca²⁺-independent scrambling rates observed in prior studies, but this part is also challenging using coarse-grain molecular simulations. Previous reports have identified lipid crossing, packing defects, and other associated events, so it is difficult to place this paper in that context. However, the absence of validation leaves certain claims, like alternative scrambling pathways, speculative.

Answer: It is generally difficult to quantitatively compare bulk measurements of scrambling phenomena with simulation results. The advantage of simulations is to directly observe the transient scrambling events at a spatial and temporal resolution that is currently unattainable for experiments. The current experimental evidence for the precise mechanism of Ca2+-independent scrambling is still under debate. We therefore hope to leverage the strength of MD and statistical rigor of coarse-grained simulations to generate testable hypotheses for further structural, biochemical, and computational studies.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The findings are largely confirmatory and do not bridge the gap between coarse-grained simulations and experimental observations. Future efforts should focus on resolving these limitations, possibly through atomistic simulations or experimental validation of the alternative scrambling pathways.

While we agree with what the reviewer may be hinting at regarding limitations of coarse-grained MD simulations, we believe that our study holds much more merit than this comment suggests. We have provided something that has yet to be done in the field: a comprehensive study that directly compares the scrambling rates of multiple TMEM16 family members in different conformations using identical simulation conditions. Our work clearly shows that a sufficiently dilated grooves is the major structural feature that enables robust scrambling for all TMEM16 scramblases members with solved structures. While all TMEM16s cause significant distortion and thinning of the membrane, we assert that the extreme thinning observed around open grooves is significantly enhanced by the lipid scrambling itself as the two leaflets merge through lipid exchange.  We saw no evidence that membrane thinning/distortion alone, in the absence of an open groove, could support scrambling at the rates observed under activating conditions or even the low rates observed in Ca2+-independent scrambling. Moreover, our handful of observations of scrambling events outside of the groove, which has not yet been reported in any study, opens an exciting new direction for studying alternative scrambling mechanisms. That said, we are currently following up on many of the observations reported here such as: scrambling events outside the groove, the kinetics of scrambling, the possibility that lipids line the groove of non-scramblers like TMEM16A, etc. This is being done experimentally with our collaborators through site directed mutagenesis and with all-atom MD in our lab. Unfortunately, it is well beyond the scope of the current study to include all of this in the current paper.

Reviewer #2 (Recommendations for the authors):

Major comments and questions:

(1) Line 214 and Figure 1- Figure Supplement 1: why have you only compared the final frame of the trajectory to the cryo-EM structure? Even if these comparisons are qualitative, they should be representative of the entire trajectory, not a single frame.

We thank the reviewer for this suggestion and replaced the single-frame snapshots in Figure 1-figure supplement 1 for ensemble-averaged head groups densities. The overall agreement between membrane shapes in CGMD and cryo-EM was not affected by this change.

(2) Lines 228-231: You comment 'Residues in this site on nhTMEM16 and TMEMF also seem to play a role in scrambling but the mechanism by which they do so is unclear.' This is something you could attempt to quantify in the simulations by calculating the correlation between scrambling and protein-membrane interactions/contacts in this site. Can you speculate on a mechanism that might be a contributing factor?

We probed the correlation between these residues and scrambling lipids, as suggested by the reviewer, and interestingly not all scrambling lipids interact with these residues. Yet there is strong lipid density in this vicinity (see insets in Figure 1 and Figure 4-figure supplement 2). These observations lead us to suspect these residues impact scrambling indirectly through influencing the conformation of the protein or flexibility and shape of the membrane. This interpretation fits with mutagenesis studies highlighting a role for these residues in scrambling (see refs 59, 62, and 67). Specifically, Falzone et al. 2022 (ref 59) suggested that they may thin the membrane near the groove, but this has not been tested via structure determination and a detailed model of how they impact scrambling is missing. We could address this question with in silico mutations; however, CG simulation is not an appropriate method to study large scale protein dynamics, and AA simulations are likely best, but beyond the scope of this paper.

(3) Lines 240-245 and Figure 1B: This section discusses the coupling between membrane distortions and the sinusoidal curve around the protein, however, Figure 1B only shows snapshots of the membrane distortions. Is it possible to understand how these two collective variables are correlated quantitatively (as opposed to the current qualitative analysis)?

We believe that it may be possible to quantitatively capture these two key features of the membrane, as we did previously with nhTMEM16 using our continuum elasticity-based model of the membrane (Bethel and Grabe 2016). Our model agreed with all atom MD surfaces to within ~1 Å, hence showing good quantitative agreement throughout the entire membrane. However, we doubt that we could distill the essence of our model down to a simple functional relationship between the sinusoidal wave and pinching, which we think the reviewer is asking. Rather, we believe that the large-scale sinusoidal distortion (collective variable 1) and pinching/distortion (collective variable 2) near the groove arise from the interplay of the specific protein surface chemistry for each protein (patterning of polar and non-polar residues) and the membrane. This is why we chose to simply report the distinct patterns that the family members impose on the surrounding membrane, which we think is fascinating. Specifically, Fig. 1B shows that different TMEM16 family members distort the membrane in different ways. Most notably, fungal TMEM16s feature a more pronounced sinusoidal deformation, whereas the mammalian members primarily produce local pinching. Then, in Fig. 3A we show that the thinning at the groove happens in all structures and is more pronounced in open, scrambling-competent conformations. In other words, proteins can show very strong thinning (e.g. TMEM16K, 5OC9) even though the membrane generally remains flat.

(4) Lines 257-258: Authors comment that TMEM16A lacks scramblase activity yet can achieve a fully lipid-lined groove (note the typo - should be lipid-lined, not lipid-line). Is a fully lipid-lined groove a prerequisite for scramblase activity? Are lipid-lined grooves the only requirement for scramblase activity? Could the authors clarify exactly what the prerequisite for scramblase activity is to avoid any confusion; this will be useful for later descriptions (i.e. line 295) where scrambling competence is again referred to. Additionally, the associated figure panel (Figure 1D) shows a snapshot of this finding but lacks any statistical quantifications - is a fully lipid-lined groove a single event? Perhaps the additional analyses, such as the groove-lipid contacts, may be useful here.

The definition of lipid scrambling is that a lipid fully transitions from one membrane leaflet to the other. While a single lipid could transition through the groove on its own, it is well documented in both atomistic and CG MD simulations, that lipid scrambling typically happens through a lipid-lined groove, as shown in Fig. 1A-B. The lipids tend to form strong choline-to-phosphate interactions with nearest neighbors that make this energetically favorable. That said, lipid-lined grooves are not sufficient for robust scrambling, which is what we show in Fig. 1D where the non-scrambler TMEM16A did in fact feature a lipid-lined groove. As suggested, we performed contact analysis and found that residue K645 on TM6 in the middle of the groove contacts lipids in 9.2% of the simulation frames.

To get a better understanding of how populated the TM4-TM6 pathway is with lipids across all simulated structures, we determined for every simulation frame how many headgroup beads resided in the groove. This indicates that the ion-conductive state of TMEM16A (5OYB*, Fig. 1D) only had 1 lipid in the pathway, on average, meaning that the configuration shown Fig. 1D is indeed exceptional. As a reference, our strongest scrambler nhTMEM16 4WIS, had an average of 2.8 lipids in the groove. We added a table containing the means and standard deviations that resulted from this analysis as Figure 1-Table supplement 1.

(5) Lines 295-298 : The scrambling rates of the Ca²⁺-bound and Ca²⁺-free structures fall within overlapping error margins, it becomes difficult to definitively state that Ca²⁺ binding significantly enhances scrambling activity. This undermines the claim that the Ca²⁺-bound structure is the strongest scrambler. The authors should conduct statistical analyses to determine if the difference between the two conditions is statistically significant.

In contrast to the reviewer’s comment, we do not claim that Ca2+-binding itself enhances lipid scrambling. Instead, what we show is that WT structures that are solved in an open confirmation (all of which are Ca2+-bound, except 6QM6) are robust scramblers. For nhTMEM16, we did not observe any scrambling events for the closed-groove proteins, making further statistical analysis redundant.

(6) The authors claim that the scrambling rates derived from their MD simulations are in "excellent agreement" with experimental findings (lines 294-295), despite significant discrepancy between simulated and experimentally measured rates. For example, the simulated rate of 24.4 {plus minus} 5.2 events/µs for the open, Ca²⁺-bound fungal nhTMEM16 (PDB ID 4WIS) corresponds to approximately 24 million events per second, which is vastly higher than experimental rates. Experimental studies have reported scrambling rate constants of ~0.003 s⁻¹ for TMEM16 family members in the absence of Ca²⁺, measured under physiological conditions (https://doi.org/10.1038/s41467-019-11753-1 ). Even with Ca²⁺ activation, scrambling rates remain several orders of magnitude lower than the rates observed in simulations. Moreover, this highlights a larger problem: lipid scrambling rates occur over timescales that are not captured by these simulations. While the authors elude to these discrepancies (lines 605-606), they should be emphasised in the text, as opposed to the table caption. These should also be reconducted to differences between the membrane compositions of different studies.

We agree with the spirit of the reviewer’s comment, and because of that, we were very careful not to claim that we reproduce experimental scrambling rates, just that the trends (scrambling-competent, or not) are correct. On lines 294-295, we actually said that the scrambling rates in our simulations excellently agree with “the presumed scrambling competence of each experimental structure”, which is true. 

As explained extensively in the discussion section of our paper (and by many others), direct comparison between MD (e.g., Martini 3, but also atomistic force fields) dynamics and experimental measurements is challenging. The primary goal of our paper is to quantify and compare the scrambling capacity of different TMEM16 family members and different states, within a CGMD context.

That said, we agree with the reviewer that we may have missed rare or long-timescale events (as is the case in any MD experiment) and added this point to the discussion.

(7) To address these discrepancies, the authors should: i) emphasize that simulated rates serve as qualitative indicators of scrambling competence rather than absolute values comparable to experimental findings and ii) discuss potential reasons for the divergence, such as simulation timescale limitations or lipid bilayer compositions that may favor scrambling and force field inaccuracies.

Please see our answer to question 6. Within the context of our CGMD survey, we confidently call our results quantitative. However, we agree with the reviewer that comparison with experimental scrambling rates is qualitative and should be interpreted with caution. To reflect this, we rewrote the first sentence of the relevant paragraph in the discussion section.

(8) Line 310: Can the authors provide a rationale as to why one monomer has a wider groove than the other? Perhaps a contact analysis could be useful. See the comment above about ENM.

The simulation of Ca2+-bound TMEM16K was initiated from an asymmetric X-ray structure in which chain B features a more dilated groove than chain A (PDB 5OC9). The backbones of TM4 and TM6 in the closed groove (A) are close enough together to be directly interconnected by the elastic network. In contrast, TM4 and TM6 in the more dilated subunit (B) are not restricted by the elastic network and, as a consequence, display some “breathing” behavior (Fig. 3B and Fig. 3-Suppl. 6A), giving rise to a ~4x higher scrambling rate. We explicitly added the word “cryo-EM” and the PDB ID to the sentence to emphasize that the asymmetry stems from the original experimental structure.

When answering this question, we also corrected a mislabeled chain identifier which was in the original manuscript ‘chain A’ when it is actually ‘chain B’ in Fig.2-Suppl. 3A.

(9) Line 312: Authors speculate that increased groove width likely accounts for increased scrambling rates. For statistical significance, authors should attempt to correlate scrambling rates and groove width over the simulation period.

The Reviewer is referring to our description of scrambling rates we measured for TMEM16K where we noted that on average the groove with the highest scrambling rate is also on average wider than the opposite subunit which is below 6 Å. We do not suggest that the correlation between scrambling and groove width is continuous, as the Reviewer may have interpreted from our original submission, but we think it is a binary outcome – lipids cannot easily enter narrow grooves (< 6 Å) and hence scrambling can only occur once this threshold is reached at which point it occurs at a near constant rate. We showed this for 4 different family members in the original Fig. 3B, where scrambling events (black dots) were much more likely during, or right after, groove dilation to distances > 6 Å. 

(10) Line 359: Authors have plotted the minimum distance between residues TM4 and TM6 in Fig. 3A/B, claiming that a wide groove is required for scrambling. Upon closer examination, it is clear that several of these distributions overlap, reducing the statistical significance of these claims. Statistical tests (i.e. KS-tests) should be performed to determine whether the differences in distributions are significant.

The Reviewer appears to be asking for a statistical test between the six distance distributions represented by the data in Fig. 3A for the scrambling competent structures (6QP6*, 8B8J, 6QM6, 7RXG, 4WIS, 5OC9), and we think this is being asked because it is believed that we are making a claim that the greater the distance, the greater the scrambling rate. If we have interpreted this comment correctly, we are not making this claim. Rather, we are simply stating that we only observe robust scrambling when the groove width regularly separates beyond 6 Å. The full distance distributions can now be found in Figure 3-figure supplement 6B, and we agree there is significant overlap between some of these distributions. However, the distinguishing characteristic of the 6 distributions from scrambling competent proteins is that they all access large distances, while the others do not. Notably, TMEM16F proteins (6QP6*, 8B8J) are below the 6 Å threshold on average, but they have wide standard deviations and spend well over ¼ of their time in the permissive regime (the upper error bar in the whisker plots in Fig. 3A is the 75% boundary).

(11) Line 363-364: The authors state that all TMEM16 structures thin the membrane. Could the authors include a description of how membrane thinning is calculated, for instance, is the entire membrane considered, or is thinning calculated on a membrane patch close to the protein? Do membrane patches closer to the transmembrane protein increase or decrease thickness due to hydrophobic packing interactions? The latter question is of particular concern since Martini3 has been shown to induce local thinning of the membrane close to transmembrane helices, yielding thicknesses 2-3 Å thinner than those reported experimentally (https://doi.org/10.1016/j.cplett.2023.140436). This could be an important consideration in the authors' comparison to the bulk membrane thickness (line 364). Finally, how is the 'bulk membrane thickness' measured (i.e., from the CG simulations, from AA simulations, or from experiments)?

Regarding the calculation of thinning and bulk membrane thickness, as described in Method “Quantification of membrane deformations”, the minimal membrane thickness, or thinning, is defined as the shortest distance between any two points from the interpolated upper and lower leaflet surfaces constructed using the glycerol beads (GL1 and GL2). Bulk membrane thickness is calculated by taking the vertical distance between the averaged glycerol surfaces at the membrane edge.

The concern of localized membrane deformation due to force field artifacts is well-founded. However, the sinusoidal deformations shown here are much greater than 2-3 Å Martini3 imperfections, and they extend for up to 10 Å radially away from the protein into the bulk membrane (see Figure 3-figure supplement 1-5 for more of a description). Most importantly, the sinusoidal wave patterns set up by the proteins is very similar to those described in the previous continuum calculation and all-atom MD for nhTMEM16 (https://www.pnas.org/doi/full/10.1073/pnas.1607574113).

(12) Line 374: The authors state a 'positive correlation' between membrane thinning/groove opening and scrambling rates. To support this claim, the authors should report. the correlation coefficients.

We have removed any discussion concerning correlations between the magnitude of the scrambling rate and the degree of membrane thinning/groove opening. Rather we simply state that opening beyond a threshold distance is required for robust scrambling, as shown in our analysis in Fig. 3A.

Concerning the relation between thinning and scrambling: Instantaneous membrane thinning is poorly defined (because it is governed by fluctuations of single lipids), and therefore difficult to correlate with the timing of individual scrambling events in a meaningful way.  Moreover, as we state later in that same section, “we argue that the extremely thin membranes are likely correlated with groove opening, rather than being an independent contributing factor to lipid scrambling”.

(13) Line 396: It is stated that TMEM16A is not a scramblase but the simulating scrambling activity is not zero. How can you be sure that you are monitoring the correct collective variable if you are getting a false positive with respect to experiments?

We only observe 2 scrambling events in 10 ms, which is a very small rate compared to the scrambling competent states. In a previous large survey Martini CG simulation study that inspired our protocol (Li et al, PNAS 2024), they employed a 1 event/ms cut-off to distinguish scramblers from non-scramblers. Hence, they would have called TMEM16A a non-scrambler as well. We expect that false negatives in this context might be an artifact of the CG forcefield, or it could be that TMEM16A can scramble but too slowly to be experimentally detected. Regarding the collective variable for lipid flipping, it is correct, and we know that this lipid actually flipped.

(14) Line 402: Distance distributions for the electrostatic interactions between E633 and K645 should be included in the manuscript. This is also the case for the interactions between E843-K850 (lines 491-492).

Our description of interactions between lipid headgroups and E633 and K645 in TMEM16A (5OYB*) are based on qualitative observations of the MD trajectory, and we highlight an example of this interaction in Figure 3-video 4. The video clearly shows that the lipid headgroups in the center of the groove orient themselves such that the phosphate bead (red) rests just above K645 (blue) and at other times the choline bead (blue) rests just below E633 (red). We do not think an additional plot with the distance distributions between lipids and these residues will add to our understanding of how lipids interact residues in the TMEM16A pore.

We made a similar qualitative observation for the interaction between the POPC choline to E843 and POPC phosphate to K850 while watching the AAMD simulation trajectory of TMEM16F (PDB ID 6QP6). Given that this was a single observation, and the same interactions does not appear in CG simulation of the same structure (see simulation snapshots in Figure 4-figure supplement 5) we do not think additional analysis would add significantly to our understanding of which residues may stabilize lipids in the dimer interface.

(15) Lines 450-451: 'As the groove opens, water is exposed to the membrane core and lipid headgroups insert themselves into the water-filled groove to bridge the leaflets.' Is this a qualitative observation? Could the authors report the correlation between groove dilation and the number of water permeation events?

Yes, this is qualitative, and it sketches the order of events during scrambling, and we revised the main text starting at line 450 to indicate this. As illustrated by the density isosurfaces in Appendix 1-Figure 2A, the amount of water found in the closed versus open grooves is striking – there is a significant flood of water that connects the upper and lower solutions upon groove opening. Moreover, Appendix 1-Figure 2B shows much greater water permeation for open structures (4WIS, 7RXG, 5OC9, 8B8J, …) compared to closed structures (6QMB, 6QMA, 8B8Q, and many of the non-labeled data in the figure that all have closed grooves and near 0 water permeation). A notable exception is TMEM16A (7ZK3*8), which has water permeation but a closed groove and little-to-no lipid scrambling.

Minor Comments:

(1) Inconsistent use of '10' and 'ten' throughout.

We like to kindly point out that we do not find examples of inconsistent use.

(2) Line 32: 'TM6 along with 3, 4 and 5...' should be 'TM6 along with TM3, TM4 and TM5...'. Same in line 142. Naming should stay consistent.

Changes are reflected in the updated manuscript.

(3) Line 141: do you mean traverse (i.e. to travel across)? Or transverse (i.e. to extend across the membrane)?

This is a typo. We meant “traverse”. Thanks for pointing it out.

(4) Line 142: 'greasy' should be 'strongly hydrophobic'.

Changes are reflected in the updated manuscript.

(5) Line 143-144: "credit card mechanism" requires quotation marks.

Changes are reflected in the updated manuscript.

(6) Line 144: state if Nectria haematococca is mammalian or fungal, this is not obvious for all readers.

Changes are reflected in the updated manuscript.

(7) Line 147-148: Is TMEM16A/TMEM16K fungal or mammalian? What was the residue before the mutation and which residue is mutated? Perhaps the nomenclature should read as TMEM16X10Y where X=the residue prior to the mutation, 10 is a placeholder for the residue number that is mutated and Y=the new residue following mutation.

“TMEM16” is the protein family. “A” denotes the specific homolog rather than residue.  

(8) Lines 157-158: same as 10, it is unclear if these are fungal or mammalian.

Clarifications added.

(9) Line 184: "...CGMD simulation" should be "...CGMD simulations".

Changes made.

(10) Line 191-192: It would help to create a table of all of the mutants (including if they are mammalian or fungal) summarizing the salt concentrations, lipid and detergent environments, the presence of modulators/activators, etc.

We added this information to Appendix 1-Table 1 in the supplemental information. We did not specify NaCl concentrations, because they all experimental procedures used standard physiological values for this (100-150 mM).

(11) Line 210: inconsistencies with 'CG' and 'coarse-grain'.

Changes made.

(12) Figure 1 caption: '...totaling ~2μs (B)...' is missing the fullstop after 2μs.

Changes made.

(13) Figure 1B: it may be useful to label where the Ca2+ ion binds or include a schematic.

We updated Fig. 1A to illustrate where Ca2+ binds.

(14) Line 311: Are these mean distances? The authors should add standard deviations.

Yes, they are. We added the standard deviations to the text.

(15) Line 321-322: Perhaps a schematic in Figure 2 would be useful to visualize the structural features described here.

We would kindly refer interested readers to reference [60].

(16) Line 377: '...are likely a correlate of groove opening...' should read as: '...are likely correlated to groove opening...'.

Thank you for pointing it out. Changes made.

(17) Line 398: the '...empirically determined 6Å threshold for scrambling.' Was this determined from the simulations or from experiments? What does "empirically" mean here? Please state this.

This value was determined from the simulations. Based on our analysis of the correlation between scrambling rate and groove dilation, we found that the minimal TM4/6 distance of 6 Å can distinguish between the high and low activity scramblers. The exact numerical value is somewhat arbitrary as there is a range of values around 6 Å that serve to distinguish scramblers from non-scramblers.

(18) Figure 4: This figure should be labelled as A, B, C and D, with the figure caption updated accordingly.

We updated Figure 4 and its caption.

Reviewer #3 (Recommendations for Authors):

The authors must do additional simulations to further validate their claim with different lipids and further substantiate dimer interface independent of Ca2+ ions.

Thank you for the suggestion. We completely agree that studying scrambling in the context of a diverse lipid environment is an exciting area to explore. We are indeed actively working on a project that shares the similar idea. We decided not to include that study because we think the additional discussion involved would be excessive for the current manuscript. We, however, look forward to publishing our findings in a separate manuscript in the near future. In terms of Ca2+-independent scrambling, we are planning with our experimental collaborator for mutagenesis studies that target the residues we identified along the dimer interface.

Since calcium ions are critical for the stability of these structures, authors should show that they were placed throughout the simulations consistently.

As stated in the method section “Coarse-grained system preparation and simulation detail”, all Ca2+ ions are manually placed into the coarse-grained structure from the beginning of the simulation at their identical corresponding position in the experimental structure and harmonically bonded to adjacent acidic residues throughout the duration of simulation. We have also added a label to Fig 1A to indicate where the two Ca2+ ions are located.

The comparison with experimental structures should be consistent with complete simulation, and not the last structure of the trajectory. Depending on the conformational variability, this might be misleading.

We agree and updated Fig. 1-supplement figure 1 accordingly. The overall agreement between membrane shapes in CGMD and cryo-EM was not affected by this change.

Author response:

The following is the authors’ response to the original reviews

Public Review:

Reviewer #1 (Public review):

Summary:

Meteorin proteins were initially described as secreted neurotrophic factors. In this manuscript, Eggeler et al. demonstrate a novel role for Meteorins in establish left-right axis formation in the zebrafish embryo. The authors generated null mutations in each of the three zebrafish meteorin genes - metrn, metrnla, and metrnlab. Triple mutant embryos displayed phenotypes strongly associated with left-right defects such as heart looping and visceral organ placement, and disrupted expression of Nodal-responsive genes, as did single mutants for metrn and metrnla. The authors then go on to demonstrate that these defects in left-right asymmetry are likely to due to defects in Kupffer's Vesicle and the progenitor dorseal forerunner cells including impaired lumen formation and reduced fluid flow, reduced clustering among DFCs, impaired DFC migration, mislocalization of apical proteins ZO-1 and aPKC, and detachment of DFCs from the EVL. Notably, the authors found that expression of marker genes sox32 and sox17 were not affected, suggesting Meteorins are required for DFC/KV morphogenesis but not necessarily fate specification. Finally, the authors show genetic interaction between Meteorins and integrin receptors, which were previously implicated in left-right patterning. In a supplemental figure, the manuscript also presents data showing expression of meteorin genes around the chick Hensen's node, suggesting that the left-right patterning functions may be conserved among vertebrates.

Strengths:

Strengths of this study include the generation of a triple mutant line that targets all known zebrafish meteorin family members. The experiments presented in this study were rigorous, especially with respect to quantification and statistical analysis.

Weaknesses:

Although the authors convincingly demonstrate a role for Meteorins in zebrafish left-right patterning, data supporting a conserved role in other vertebrates is compelling but limited to one supplemental figure.

We thank the reviewer for their thoughtful summary of our study and for highlighting the strengths of our work, including the generation of the triple mutant line and the rigor of our experimental design and quantitative analyses. We also appreciate the constructive feedback regarding the limited functional data supporting the conservation of Meteorin function in other vertebrates. We agree that this is an important aspect that could be further explored. While functional studies in additional species are beyond the current scope, we will consider such experiments in future work.

We would like to highlight the phylogenetic analysis of Meteorin proteins we have already performed and included in the manuscript (Fig. S7D), which illustrates the evolutionary conservation of this protein family and supports the possibility of a conserved role in left-right patterning.

Additionally, we have expanded the methods and discussion to include: (1) details on zebrafish viability in contrast to reported embryonic lethality in metrn mutant mice, (2) the background strains used in our study, (3) observed variability in DFC number and potential batch effects and (4) clarification of our 'convergence ratio' quantification approach.

Reviewer #2 (Public review):

Summary:

In this manuscript the authors describe their study on the role of meteorins in establishing the left-right organizer. The left-right organizer is a transient organ in vertebrate embryos in which rotating cilia cause a fluid flow that breaks the left-right symmetry and coordinates lateralization of internal organs such as gut and heart. In zebrafish, the left-right organizer (also named Kupffer's vesicle) is formed by dorsal forerunner cells, but very little is known about how dorsal forerunner cells coalles and form this ciliated vesicle in the embryo. The authors mutated the three meteorin-coding genes in zebrafish and observed that mutations in each one of these causes laterality defects with the strongest defects observed in the triple mutant. Loss of meteorins affects nodal gene expression, which play essential roles in establishing organ laterality. Meteorins are widely expressed in developing embryos and expression in lateral plate mesoderm and dorsal forerunner cells was observed. The meteorin triple mutant embryos display defects in the migration and clustering of the dorsal forerunner cells impairing kupffer's vesicle formation and cilia rotation. Finally, the authors show that meteorins genetically interact with integrins.

Strengths:

- These authors went through the lengthy process of generating triple mutants affecting all three meteorin genes. This provides robust genetic evidence on the role of meteorins in establishing organ laterality and circumvented that interpretation of the results would be hard due to redundant functions of meteorins.

- The use of life imaging on triple mutants is appreciated

- High-quality imaging of dorsal forerunner to quantify cell migrations and its relation to Kupffer's vesicle formation.

Weaknesses:

- Lack of a model how meteorins regulate dorsal forerunner cell migration.

- Only genetic data to suggest a link between meteorins and integrins

- Besides its role in DFC migration, meteorins may also play a more direct role in regulating Nodal signaling, which is not addressed here.

We appreciate the recognition of the strengths of our study, particularly the generation of the triple meteorin mutants and the use of high-resolution imaging to quantify DFC behavior and Kupffer’s vesicle formation—both of which were central to providing robust evidence for Meteorins' role in left-right patterning.

We also value the reviewer’s comments on areas that need further exploration, including the need for a mechanistic model explaining how Meteorins regulate DFC migration, the genetic interaction with integrins, and the potential direct involvement of Meteorins in Nodal signaling.

We agree that deeper mechanistic insights would strengthen the study. While our findings suggest that Meteorins influence DFC migration and clustering through integrin pathways, a detailed mechanistic dissection, particularly regarding the yet unidentified Meteorin receptor, lies beyond the current scope. However, we consider this a key aspect for future research and have discussed it further in the revised discussion section.

In response to the reviewer’s suggestions, we have expanded the discussion to address the limitations of the current data linking Meteorins and integrins, including relevant citations to studies that implicate integrins in similar contexts. Additionally, we have added a more detailed discussion of the potential for Meteorins to directly influence Nodal signaling, and we cite a relevant study to support this possibility.

Once again, we thank the reviewer for their insightful and constructive comments. These points raise important directions for future investigation that will further advance our understanding of Meteorin function in left-right axis formation.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

In the Results section (p. 9), the authors state, "...a reduced ZO-1 enrichment at the apical junctions of triplMUT GFP-positive DFCs could be detected." However, in Fig. 4F-G, the areas of ZO-1 enrichment indicated by arrowheads appear quite far from the DFCs themselves, making it unclear if these ZO-1-enriched areas are apical DFC junctions (as stated in the text) or instead are part of the EVL. Is it possible to include an additional cell membrane marker or other landmarks? In addition, the differences in ZO-1 accumulation between mutants and WT appear relatively modest. Is it possible to provide quantification of this effect?

We appreciate the reviewer’s request for additional stainings and further clarification and we would like to highlight the requested quantifications of ZO-1 accumulation, including statistical analysis, are already provided in Fig. S5E.

In mouse, loss of Meteorin is embryonic lethal yet the zebrafish triple mutants are viable. Could the authors discuss this discrepancy?

We have expanded the discussion to address this point, suggesting that species-specific differences in compensatory mechanisms may explain the observed differences in viability. We would like to reiterate that while one study has reported embryonic lethality in metrn mutant mice, this specific mouse line has not been further investigated in any recent publications. Additionally, in collaboration with the lab of Alain Chédotal, we generated independent metrn and metrnl mutant mouse lines, which did not exhibit the phenotype described in the previously mentioned study.

It has been reported that TL and AB strains exhibit variable numbers of DFCs and thus laterality defects (Moreno-Ayala et al., 2021, Cell Reports 34(2):108606). Would it be possible for the authors to report background stains used in this study and those used to generate the meteorin knock-outs?

We appreciate the comment highlighting the importance of specifying the background strains used in our study. We have now included this information in the methods section, detailing the zebrafish strains utilized throughout our experiments.

For statistical analysis, would be possible for the authors to report the number of clutches examined to control for batch effects (especially given the wide variability in DFC numbers as noted above)?

For further clarification, we have now included additional explanation on number of clutches in the methods section.

In the Methods section (p. 19), the description of how the convergence ratio was computed was somewhat unclear. Could the authors provide a citation or include a diagram/schematic?

We have revised the Methods section to provide a clearer definition of the convergence ratio and have included a schematic (Fig. 4D) to illustrate how it was calculated.

Reviewer #2 (Recommendations for the authors):

- Meteorins are widely expressed in the embryo. Can the authors comment on whether meteorin expression is required in the dorsal forerunner cells (DFCs) or in other cells? This could be addressed by knockdown experiments in DFCs as described by others (PMID: 15716348)

We thank the reviewer for this important comment. In our study, we have shown that Meteorins are not required for the identity of DFCs, as several DFC-specific markers remain expressed in the respective cells within the meteorin mutant background (see Fig. S4).

- In fig1d and 1e the authors use heterotaxy to describe visceral organ placement. The embryo shown in 1d seems to display situs inversus instead of heterotaxy, which is defined as discordance in organ position. The authors should clarify this.

We agree with the reviewer and have revised the figures and figure legends to clarify the distinction between situs inversus and heterotaxy.

- In Fig2 the authors show that nodal pathway genes are reduced, suggesting reduced Nodal signaling. How do they explain this as loss of cilia rotation generally leads to randomization of Nodal signaling but not a reduction in signaling.

Following this suggestion we have now added a further discussion on the possibility that Meteorins could directly regulate Nodal signaling in addition to their role in DFC migration and have cited a relevant study.

- Reduced Nodal signaling in the LPM leads to organ laterality defects. Most anterior tissues like the heart are more sensitive to perturbation in Nodal signaling in the LPM compared to more posterior organs like gut (see also PMID: 25684355). Since in triple mutants the position of the heart is more affected than the position of the visceral organs this suggests that meteorins play an additional role in Nodal signaling in the LPM. As others have shown that meteorins regulate nodal activity (PMID: 24558432), the authors should address this further.

As described above, we have now added a further discussion on the possibility that Meteorins could directly regulate Nodal signaling in addition to their role in DFC migration and have cited a relevant study. Further investigation into a possible direct role of Meteorins in Nodal signaling will be pursued in future work.

- The term 'convergence ratio' is not clearly described and confusing as convergence is also used for the movement of LPM cells towards the midline.

As noted in response to Reviewer #1, we have revised the Methods section and included a schematic in Fig. 4D to better explain this parameter.

We are grateful for the thoughtful critiques from both reviewers, which have been very constructive and improved the clarity of our study. We believe that the revisions we have made address the concerns raised, and we look forward to your evaluation of our revised manuscript.

Author response:

Reviewer #1 (Public Review):

In this manuscript, Tran et al. investigate the interaction between BICC1 and ADPKD genes in renal cystogenesis. Using biochemical approaches, they reveal a physical association between Bicc1 and PC1 or PC2 and identify the motifs in each protein required for binding. Through genetic analyses, they demonstrate that Bicc1 inactivation synergizes with Pkd1 or Pkd2 inactivation to exacerbate PKD-associated phenotypes in Xenopus embryos and potentially in mouse models. Furthermore, by analyzing a large cohort of PKD patients, the authors identify compound BICC1 variants alongside PKD1 or PKD2 variants in trans, as well as homozygous BICC1 variants in patients with early-onset and severe disease presentation. They also show that these BICC1 variants repress PC2 expression in cultured cells.

Overall, the concept that BICC1 variants modify PKD severity is plausible, the data are robust, and the conclusions are largely supported. However, several aspects of the study require clarification and discussion:

(1) The authors devote significant effort to characterizing the physical interaction between Bicc1 and Pkd2. However, the study does not examine or discuss how this interaction relates to Bicc1's well-established role in posttranscriptional regulation of Pkd2 mRNA stability and translation efficiency.

The reviewer is correct that the present study has not addressed the downstream consequences of this interaction considering that Bicc1 is a posttranscriptional regulator of Pkd2 (and potentially Pkd1). We think that the complex of Bicc1/Pkd1/Pkd2 retains Bicc1 in the cytoplasm and thus restrict its activity in participating in posttranscriptional regulation. As we do not have yet experimental data to support this model, we have not included this model in the manuscript. Yet, we will update the discussion of the manuscript to further elaborate on the potential mechanism of the Bicc1/Pkd1/Pkd2 complex.

(2) Bicc1 inactivation appears to downregulate Pkd1 expression, yet it remains unclear whether Bicc1 regulates Pkd1 through direct interaction or by antagonizing miR-17, as observed in Pkd2 regulation. This should be further examined or discussed.

This is a very interesting comment. The group of Vishal Patel published that PKD1 is regulated by a mir-17 binding site in its 3’UTR (PMID: 35965273). We, however, have not evaluated whether BICC1 participates in this regulation. A definitive answer would require us utilize some of the mice described in above reference, which is beyond the scope of this manuscript. We, however, will revise the discussion to elaborate on this potential mechanism.

(3) The evidence supporting Bicc1 and ADPKD gene cooperativity, particularly with Pkd1, in mouse models is not entirely convincing, likely due to substantial variability and the aggressive nature of Bpk/Bpk mice. Increasing the number of animals or using a milder Bicc1 strain, such as jcpk heterozygotes, could help substantiate the genetic interaction.

We have initially performed the analysis using our Bicc1 complete knockout, we previously reported on (PMID 20215348) focusing on compound heterozygotes. Yet, like the Pkd1/Pkd2 compound heterozygotes (PMID 12140187) no cyst development was observed until we sacrificed the mice at P21. Our strain is similar to the above mentioned jcpk, which is characterized by a short, abnormal transcript thought to result in a null allele (PMID: 12682776). We thank the reviewer for pointing use to the reference showing the heterozygous mice show glomerular cysts in the adults (PMID: 7723240). This suggestion is an interesting idea we will investigate. In general, we agree with the reviewer that the better understanding the contribution of Bicc1 to the adult PKD phenotype will be critical. To this end, we are currently generating a floxed allele of Bicc1 that will allow us to address the cooperativity in the adult kidney, when e.g. crossed to the Pkd1^RC/RC mice. Yet, these experiments are unfortunately beyond the scope of this manuscript.

Reviewer #2 (Public Review):

Tran and colleagues report evidence supporting the expected yet undemonstrated interaction between the Pkd1 and Pkd2 gene products Pc1 and Pc2 and the Bicc1 protein in vitro, in mice, and collaterally, in Xenopus and HEK293T cells. The authors go on to convincingly identify two large and non-overlapping regions of the Bicc1 protein important for each interaction and to perform gene dosage experiments in mice that suggest that Bicc1 loss of function may compound with Pkd1 and Pkd2 decreased function, resulting in PKD-like renal phenotypes of different severity. These results led to examining a cohort of very early onset PKD patients to find three instances of co-existing mutations in PKD1 (or PKD2) and BICC1. Finally, preliminary transcriptomics of edited lines gave variable and subtle differences that align with the theme that Bicc1 may contribute to the PKD defects, yet are mechanistically inconclusive.

These results are potentially interesting, despite the limitation, also recognized by the authors, that BICC1 mutations seem exceedingly rare in PKD patients and may not "significantly contribute to the mutational load in ADPKD or ARPKD". The manuscript has several intrinsic limitations that must be addressed.

As mentioned above, the study was designed to explore whether there is an interaction between BICC1 and the PKD1/PKD2 and whether this interaction is functionally important. How this translates into the clinical relevance will require additional studies (and we have addressed this in the discussion of the manuscript).

The manuscript contains factual errors, imprecisions, and language ambiguities. This has the effect of making this reviewer wonder how thorough the research reported and analyses have been.

We respectfully disagree with the reviewer on the latter interpretation. The study was performed with rigor. We have carefully assessed the critiques raised by the reviewer. Most of the criticisms raised by the reviewer will be easily addressed in the revised version of the manuscript. Yet, none of the critiques raised by the reviewer seems to directly impact the overall interpretation of the data.

Reviewer #3 (Public Review):

Summary:

This study investigates the role of BICC1 in the regulation of PKD1 and PKD2 and its impact on cytogenesis in ADPKD. By utilizing co-IP and functional assays, the authors demonstrate physical, functional, and regulatory interactions between these three proteins.

Strengths:

(1) The scientific principles and methodology adopted in this study are excellent, logical, and reveal important insights into the molecular basis of cystogenesis.

(2) The functional studies in animal models provide tantalizing data that may lead to a further understanding and may consequently lead to the ultimate goal of finding a molecular therapy for this incurable condition.

(3) In describing the patients from the Arab cohort, the authors have provided excellent human data for further investigation in large ADPKD cohorts. Even though there was no patient material available, such as HUREC, the authors have studied the effects of BICC1 mutations and demonstrated its functional importance in a Xenopus model.

Weaknesses:

This is a well-conducted study and could have been even more impactful if primary patient material was available to the authors. A further study in HUREC cells investigating the critical regulatory role of BICC1 and potential interaction with mir-17 may yet lead to a modifiable therapeutic target.

This is an excellent suggestion. We agree with the reviewer that it would have been interesting to analyze HUREC material from the affected patients. Unfortunately, besides DNA and the phenotypic analysis described in the manuscript neither human tissue nor primary patient-derived cells collected before the two patients with the BICC1 p.Ser240Pro mutation passed away. To address this missing link, we have – as a first pass - generated HEK293T cells carrying the BICC1 p.Ser240Pro variant. While these admittingly are not kidney epithelial cells, they indeed show a reduced level of PC2 expression. These data are shown in the manuscript. We have not yet addressed how this relates to its crosstalk with miR-17.

Conclusion:

The authors achieve their aims. The results reliably demonstrate the physical and functional interaction between BICC1 and PKD1/PKD2 genes and their products.

The impact is hopefully going to be manifold:

(1) Progressing the understanding of the regulation of the expression of PKD1/PKD2 genes.

(2) Role of BiCC1 in mir/PKD1/2 complex should be the next step in the quest for a modifiable therapeutic target.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Filamentous fungi are established workhorses in biotechnology, with Aspergillus oryzae as a prominent example with a thousand-year history. Still, the cell biology and biochemical properties of the production strains is not well understood. The paper of the Takeshita group describes the change in nuclear numbers and correlates it to different production capacities. They used microfluidic devices to really correlate the production with nuclear numbers. In addition, they used microdissection to understand expression profile changes and found an increase in ribosomes. The analysis of two genes involved in cell volume control in S. pombe did not reveal conclusive answers to explain the phenomenon. It appears that it is a multi-trait phenotype. Finally, they identified SNPs in many industrial strains and tried to correlate them to the capability of increasing their nuclear numbers.

The methods used in the paper range from high-quality cell biology, Raman spectroscopy, to atomic force and electron microscopy, and from laser microdissection to the use of microfluidic devices to study individual hyphae.

This is a very interesting, biotechnologically relevant paper with the application of excellent cell biology. I have only minor suggestions for improvement.

We sincerely appreciate your fair and positive evaluation of our work. Thank you for your suggestions for improvement. We respond to each of them appropriately.

Reviewer #2 (Public review):

Summary:

In the study presented by Itani and colleagues, it is shown that some strains of Aspergillus oryzae - especially those used industrially for the production of sake and soy sauce - develop hyphae with a significantly increased number of nuclei and cell volume over time. These thick hyphae are formed by branching from normal hyphae and grow faster and therefore dominate the colonies. The number of nuclei positively correlates with the thicker hyphae and also the amount of secreted enzymes. The addition of nutrients such as yeast extract or certain amino acids enhanced this effect. Genome and transcriptome analyses identified genes, including rseA, that are associated with the increased number of nuclei and enzyme production. The authors conclude from their data involvement of glycosyltransferases, calcium channels, and the tor regulatory cascade in the regulation of cell volume and number of nuclei. Thicker hyphae and an increased number of nuclei were also observed in high-production strains of other industrially used fungi such as Trichoderma reesei and Penicillium chrysogenum, leading to the hypothesis that the mentioned phenotypes are characteristic of production strains, which is of significant interest for fungal biotechnology.

Strengths:

The study is very comprehensive and involves the application of diverse state-of-the-art cell biological, biochemical, and genetic methods. Overall, the data are properly controlled and analyzed, figures and movies are of excellent quality.

The results are particularly interesting with regard to the elucidation of molecular mechanisms that regulate the size of fungal hyphae and their number of nuclei. For this, the authors have discovered a very good model: (regular) strains with a low number of nuclei and strains with a high number of nuclei. Also, the results can be expected to be of interest for the further optimization of industrially relevant filamentous fungi.

Weaknesses:

There are only a few open questions concerning the activity of the many nuclei in production strains (active versus inactive), their number of chromosomes (haploid/diploid), and whether hyper-branching always leads to propagation of nuclei.

We are very grateful for your recognition of our findings, the proposed model, and their significance for future applications. We are grateful for the questions, which contribute to a more accurate understanding.

Our responses to each are provided below. Necessary experiments are in progress.

Reviewer #3 (Public review):

Summary:

The authors seek to determine the underlying traits that support the exceptional capacity of Aspergillus oryzae to secrete enzymes and heterologous proteins. To do so, they leverage the availability of multiple domesticated isolates of A. oryzae along with other Aspergillus species to perform comparative imaging and genomic analysis.

Strengths:

The strength of this study lies in the use of multifaceted approaches to identify significant differences in hyphal morphology that correlate with enzyme secretion, which is then followed by the use of genomics to identify candidate functions that underlie these differences.

Weaknesses:

There are aspects of the methods that would benefit from the inclusion of more detail on how experiments were performed and data interpreted.

Overall, the authors have achieved their aims in that they are able to clearly document the presence of two distinct hyphal forms in A. oryzae and other Aspergillus species, and to correlate the presence of the thicker, rapidly growing form with enhanced enzyme secretion. The image analysis is convincing. The discovery that the addition of yeast extract and specific amino acids can stimulate the formation of the novel hyphal form is also notable. Although the conclusions are generally supported by the results, this is perhaps less so for the genetic analysis as it remains unclear how direct the role of RseA and the calcium transporters might be in supporting the formation of the thicker hyphae.

The results presented here will impact the field. The complexity of hyphal morphology and how it affects secretion is not well understood despite the importance of these processes for the fungal lifestyle. In addition, the description of approaches that can be used to facilitate the study of these different hyphal forms (i.e., stimulation using yeast extract or specific amino acids) will benefit future efforts to understand the molecular basis of their formation.

We are very grateful for your fair and thoughtful evaluation of our work. We agree that the genetic analysis in the latter part is relatively weaker compared to the imaging analysis in the first half. Rather than a single mutation causing a dramatic phenotypic change, we believe that the accumulation of various mutations through breeding leads to the observed phenotype, making it difficult to clearly demonstrate causality. Since transcriptome and SNP analyses have revealed key pathways and phenotypes, it would be gratifying if these insights could contribute to future applications utilizing filamentous fungi.

Author Response:

We sincerely thank the reviewers and the editorial team for their thoughtful and constructive evaluation of our manuscript. We are very pleased that both reviewers and the Reviewing Editor found the work to be compelling and of interest to the community studying membrane-associated condensates. Below we outline our planned revisions in response to the public reviews.

Reviewer #1

We appreciate Reviewer #1’s positive evaluation of the study’s significance and the utility of our theoretical framework.

1. Understandably, the authors used one system to test their theory (ZO-1). However, to establish a theoretical framework, this is sufficient.

Response: We acknowledge this limitation. While we agree that additional systems would strengthen the generality of our theory, we note that the focus of this work is to introduce and validate a theoretical framework. As the reviewer notes, this is sufficient for establishing the framework. Nonetheless, we are open to further collaborations or future studies to test the model with other systems.

Reviewer #2

We are grateful for Reviewer #2’s detailed comments and will address each of the points as follows:

1. In the theoretical section, what has previously been known, compared to which equations are new, should be made more clear.

Response: We will revise the theory section to clearly distinguish previously established formulations from novel contributions.

1. Some assumptions in the model are made purely for convenience and without sufficient accompanying physical justification. E.g., the authors should justify, on physical grounds, why binding rate effects are/could be larger than the other fluxes.

Response: We will expand the discussion to provide key physical justification, especially to explain why binding rate effects are/could be larger than the other fluxes.

1. I feel that further mechanistic explanation as to why bulk phase separation widens the regime of surface phase separation is warranted.

Response: We will elaborate on the mechanism underlying this coupling.

1. The major advantage of the non-dilute theory as compared with a best parameterized dilute (or homogenous) theory requires further clarification/evidence with respect to capturing the experimental data.

Response: We will clarify this comparison more explicitly and highlight how the non-dilute model captures key nonlinear behaviors and concentration-dependent adsorption phenomena that the dilute model fails to reproduce.

1. Discrete (particle-based) molecular modelling could help to delineate the quantitative improvements that the non-dilute theory has over the previous state-of-the-art. Also, this could help test theoretical statements regarding the roles of bulk-phase separation, which were not explored experimentally.

Response:  We appreciate the suggestion and agree that such modeling would be valuable. However, this is beyond the scope of the current study. We will add a discussion on how discrete simulations could be used to further test our theory in future work.

1. Discussion of the caveats and limitations of the theory and modelling is missing from the text.

Response:  We will add a paragraph outlining caveats and limitations of the modelling.

We believe these changes will significantly improve the clarity and impact of our manuscript, and we thank the reviewers again for their valuable input.

Author response:

We thank the reviewers for their thoughtful and constructive feedback. As the reviewers noted, dissecting the contributions of Gtr1/2 and Pib2 to TORC1 signaling across diverse nutrient states is a technically and conceptually challenging problem. Indeed, many of the issues raised—including the interpretation of non-canonical TORC1 readouts (e.g., Rps6, Par32), the influence of strain auxotrophy and media composition, and the limitations of phosphoproteomic analysis performed under a single growth condition—underscore the challenges of working with the TORC1 signaling system.

In response to the reviewers’ comments, we have undertaken a broader and more systematic analysis of TORC1 regulation across defined nitrogen transitions, building directly on the signaling framework established in Figures 6 and 8 of this manuscript. This work, which includes expanded phosphoproteomic profiling and the use of refined genetic tools, supports and extends the key conclusions of Cecil et. al. Specifically, it reinforces the existence of a Pib2-dependent TORC1 output under nitrogen-limited conditions and further clarifies the physiological relevance of the intermediate TORC1 activity state. Due to the scope and depth of this expanded work, we are reporting those findings in a separate publication. Nonetheless, we view the data presented here as a key foundational step in establishing a non-redundant framework for Gtr1/2- and Pib2-dependent control of TORC1.

We have therefore made minor changes to the manuscript to clarify our use of different growth media and to temper our conclusions where appropriate. These changes, together with the context of ongoing work, should reinforce the value of Cecil et. al. in advancing our understanding of TORC1 and nutrient signaling in eukaryotes.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public Review):

In this work Jeong and colleagues focus on exploring the role of the acyltransferase ZDHHC9 in myelinating OLs in particular in the palmitoylation of several myelin proteins. After confirming the specific enrichment of the Zdhhc9 transcript in mouse and human OLs, the authors examine the subcellular localization of the protein in vitro and observed that in comparison with other isoforms, ZDHHC9 localizes at OLs cell bodies and at discrete puncta in the processes. These observations (Figures 1 and 2) led the authors to hypothesize that ZDHHC9 plays an important role in myelination. No gross changes were detected in OL development in Zdhhc9 KO mice and analyses from P28 Zdhhc9 KO mice crossed with Mobp-EGFP reporter mice did not show changes in EGFP+ OL differentiation (Figure 3).

However, and given the observed subcellular localization of ZDHHC9 in OL processes (Figure 2) and the observation that the percentage of unmyelinated axons is increased in Zdhhc9 KO (Figure 6), early time points to examine the differentiated pools of OLs and their capacity to extend processes/contact axons need to be considered.

We appreciate this point, but due to the order in which experiments were performed, the ZDHHC9 KO mouse colony that we maintained after initial submission of this work contains homozygous MOBP-EGFP, but not the mT/mG transgene that would be most optimal for the proposed experiment. We hope the reviewer appreciates that it would take considerable time and effort regarding mouse breeding to cross out the MOBP and add back the mT/mG. We nonetheless appreciate the importance of the point raised and therefore examined an earlier developmental time point (P21, 3 weeks) to quantify OLs and NG2+ OPCs. In our updated Fig 3C1-C3, we use Mobp-EGFP mice to show that Zdhhc9 KO does not significantly affect the number of EGFP+ OLs at this time point in the cortex, corpus callosum and spinal cord. We also show that in corpus callosum, Zdhhc9 KO does not significantly affect the number of NG2+ OPCs at this earlier time point (Fig 3D, E). Furthermore, immunostaining to detect BCAS1, a marker of pre-mature OLs, also revealed no qualitative difference with ZDHHC9 loss at P21. We show representative images from these BCAS1 experiments in an updated Fig S3. While these new experiments do not address the morphology of OLs in Zdhhc9 KO, they do provide further evidence that deficits in myelination in young Zdhhc9 KO mice (Figure 6) are not likely due to gross differences in OPC or OL numbers during development.

Maturation of OL in Zdhhc9 KO was examined by crossing Zdhhc9 KO with Pdgfra-CreER;R26- EGFP and following the newly EGFP-labelled OPCs following tamoxifen administration. No changes in the numbers of EGFP+ OL were detected. The authors concluded that the loss of ZDHHC9 does not alter oligodendrogenesis in either the young or mature CNS. The authors observed defects in Zdhhc9 KO OL protrusions that they attributed to abnormal OL membrane expansion (Fig 4 and 5). Can they show evidence for this?

This is an important point, and we appreciate the opportunity to explain the reasoning behind our initial statement more fully, while noting that other explanations are possible. Fig 5B (an Imaris-assisted reconstruction using the EGFP cell fill/morphology marker) highlights large spheroid-like distensions along OL processes. We reason that these spheroids are enclosed by the OL lipid membrane because if the membrane were ruptured, the EGFP signal would likely diffuse. This in turn suggests that the caliber of the OL process at the position of the spheroid is grossly abnormal i.e. the membrane has hyper-expanded. Given that OL membrane growth during myelination extends in two directions, i.e., spiral growth to the axonal surface and longitudinal growth along the axon, it is possible that spheroid-like structures are formed by uneven myelin growth. We recognize that we cannot yet conclude whether and how spheroid formation might be linked to the myelination deficit that we observe in Zdhhc9 KO mice. However, defining the subcellular mechanism for spheroid formation may provide further insights into this issue. We have therefore largely retained the original statement but have added the reasoning above to our revised Discussion.

The authors report that Zdhhc9 KO primary and secondary branches in OL were longer, some contained spheroid-like swellings and the OL protrusion complexity was higher. However, these data is partially contradictory to what they show in OL differentiation experiments in vitro (Fig 7). There is also no evidence for increased membrane expansion in Zdhhc9 knockdown myelin forming cells in culture. How to reconcile this? 

We appreciate the reviewer’s interest in this issue. Several non-mutually exclusive factors could account for the differences in OL morphology in vitro versus in vivo caused by Zdhhc9 loss. First, morphology in vivo may well be influenced by the axons and/or other extrinsic components around each OL that are not present in our primary cultures. Second, OL growth in vivo is highly 3-dimensional, whereas growth in culture is largely 2-dimensional – it may be difficult to support formation of spheroids (by definition, a 3-dimensional structure) in the latter situation. Finally, Zdhhc9 is absent in vivo from the beginning of development until the time points examined, whereas in our cultured OL experiments, Zdhhc9 shRNA is virally delivered to OPC cultures at DIV2 and likely acutely affects Zdhhc9 expression predominantly in committed OLs (following the switch to differentiation medium at DIV3). These differences may also affect the ability of other PATs or, potentially, palmitoylation-independent subcellular processes, to compensate for Zdhhc9 loss. We have more fully explained these points in our revised Discussion. 

Reviewer #2 (Public Review):

This study provides an in-depth exploration of the impact of X-linked ZDHHC9 gene mutations on cognitive deficits and epilepsy, with a particular focus on the expression and function of ZDHHC9 in myelin-forming oligodendrocytes (OLs). These findings offer crucial insights into understanding ZDHHC9-related X-linked intellectual disability (XLID) and shed light on the regulatory mechanisms of palmitoylation in myelination. The experimental design and analysis of results are convincing, providing a valuable reference for further research in this field. However, upon careful review, I believe the article still needs further improvement and supplementation in the following aspects:

(1) Regarding the subcellular localization experiment of ZDHHC9 mutants in OL, it is currently limited to in vitro cultured OL, lacking validation in vivo OL or myelin sheath. Additionally, it is necessary to investigate whether the abnormal subcellular localization of ZDHHC9 mutants affects their enzyme activity and palmitoylation modification of substrate proteins.

This is an important point but is technically challenging to address in vivo as it would likely require delivery of AAV to express ZDHHC9wt and XLID mutants specifically in OLs, preferably in the absence of endogenous ZDHHC9. We hope the reviewers would agree that this experiment is beyond the scope of the current study. However, we did compare the ability of ZDHHC9wt and XLID mutants to palmitoylate MBP, and to autopalmitoylate (sometimes used as a surrogate measure of PAT activity) in transfected heterologous cells. Although we recognize that this over-expression system is less physiological than a native OL, it has the benefit of being able to readily compare transfected wt vs mutant forms of ZDHHC9 with minimal contribution from endogenous ZDHHC9. Intriguingly, using this system, we found that autopalmitoylation activity of the XLID ZDHHC9-P150S mutant does not differ significantly from that of ZDHHC9wt, and that this mutant is still capable of palmitoylating MBP. Moreover, the R96W mutant, while impaired in autopalmitoylation, still palmitoylated MBP approximately 50% as effectively as ZDHHC9wt in our cell-based assay. These findings suggest that ZDHHC9-P150S and, probably, ZDHHC9-R96W mutants might still be able to palmitoylate substrates in OLs if they were properly localized. This possibility in turn suggests that impaired subcellular targeting in addition to, or instead of, impaired catalytic activity, may be a key factor in certain cases of ZDHHC9-associated XLID. We have expanded our Figure 8 (new panels 8E-G) to show these additional experiments and have summarized the conclusions above in our revised Discussion. We thank the reviewer for suggesting that we further investigate this issue.

(2) The experimental period (P21+21 days) using genetic labeling to track the development of myelinating cells may not be long enough. It is recommended to extend the observation time and analyze at more time points to more comprehensively reflect the impact of Zdhhc9 KO.

We appreciate this point from the reviewer but, regrettably, we did not maintain the PdgfraCreER; R26-EGFP; Zdhhc9 KO mouse line and hope the reviewer appreciates that it would take considerable time and effort to rederive this line and then perform the suggested extended time course experiments. However, we note for the reviewer that our preliminary studies did not reveal any effect of Zdhhc9 KO on the number of MOBP-EGFP+ OLs in 6-month-old mice (not shown), consistent with a model in which Zdhhc9 loss does not affect OPC-OL commitment per se.

(3) The author speculates that Zdhhc9 may regulate myelination by affecting the membrane localization of specific myelin proteins, but lacks direct experimental evidence to support this. It is suggested to detect the expression and distribution of relevant proteins in the myelin of Zdhhc9 KO mice.

We share the reviewer’s interest in this point but realized that it is more technically challenging to address than might be initially thought. The main protein we would implicate and seek to test is MBP, but we already found that there is no gross change in MBP distribution in vivo in Zdhhc9 KO mice (Fig 3A). However, an anti-MBP antibody recognizes all forms of MBP, not just the specific splice variants whose palmitoylation is affected by ZDHHC9 loss. Specifically assessing nanoscale distribution of these splice variants would require a way (e.g. anti-MBP splice form-specific antibodies that are compatible with immuno-EM) to distinguish these variants from other, non-palmitoylated forms of MBP. Although such an antibody could be an important tool, we hope the reviewers would agree that developing and characterizing such a reagent is beyond the scope of the current study.

We do, however, note that the lack of gross change in MBP distribution and levels in Zdhhc9 KO mice is consistent with the relatively mild phenotype of these mice, compared with shiverer (shi/shi) mice, in which MBP is completely lost. In shiverer, CNS compact myelin is almost absent (PMID: 671037; PMID: 88695; PMID: 460693) and, as the name suggests, mice display a shivering gait, and exhibit seizures and early death. In contrast, Zdhhc9 mice show only subtle behavioral deficits (PMID: 29944857). These differences are all consistent with a model in which Zdhhc9 KO mice, despite their significantly reduced MBP palmitoylation (Fig 8) have grossly normal distribution and levels of MBP when all splice variants are assessed (Fig 3, Fig 8). It is not inconceivable that Zdhhc9 KO mice have a nanoscale change in the distribution of MBP, particularly of specific palmitoylated splice variants, within myelin that profoundly affects myelin ultrastructure, without grossly altering MBP distribution. However, an alternative and not mutually exclusive possibility is that aberrant palmitoylation of other Zdhhc9 substrates accounts for, or contributes to, the abnormalities in myelin at the ultrastructural level. Addressing this issue would require a multi-pronged approach, not just to assess palmitoylation and distribution of such proteins in Zdhhc9 KO, but also to test whether they are direct Zdhhc9 substrates, in order to rule out indirect effects. We hope reviewers would agree that this is best left to a separate study. However, in our revised Discussion we now summarize what can be inferred regarding Zdhhc9-dependent effects on total and splicevariant specific distribution and levels of MBP.  

(4) Although the article mentions the association of Zdhhc9 with intellectual disabilities, it does not involve behavioral analysis of Zdhhc9 KO mice. It is recommended to supplement some behavioral experimental data to support the important role of Zdhhc9 in maintaining normal cognitive function, enhancing the clinical relevance of the article.

We appreciate this point from the reviewer. The behavior of the same ZDHHC9 KO mouse line that we used was reported in PMID: 31747610 and in PMID: 29944857. In the former study, Zdhhc9 KO mice were reported to display seizures reminiscent of phenotypes in human patients with ZDHHC9 mutation. The latter study assessed performance of Zddhc9 KO mice in several tasks that test cognitive function. Specifically the KO mice were reported to display “altered behaviour in the open-field test, elevated plus maze and acoustic startle test that is consistent with a reduced anxiety level; a reduced hang time in the hanging wire test that suggests underlying hypotonia but which may also be linked to reduced anxiety [and] deficits in the Morris water maze test of hippocampal-dependent spatial learning and memory.”. We have incorporate these findings in our revised Discussion, where we summarize how these phenotypes are common, not just to human patients with ZDHHC9 mutation, but also to other human neurodevelopmental conditions and mouse models in which ID is a common feature.

(5) For the abnormal myelination observed in Zdhhc9 KO mice, including unmyelinated large-diameter axons and excessively myelinated small-diameter axons, the article lacks indepth research and explanation on the exact mechanism and mode of action of ZDHHC9 in regulating myelination.

We share the reviewer’s interest in this point but again note that gaining definitive insights into this issue is far from trivial. Convincing evidence of a causative mechanism would require an exhaustive identification of ZDHHC9 in vivo substrates, followed by point mutation of substrate palmitoylation site(s) to determine the extent to which palmitoylation of such protein(s) phenocopies ZDHHC9 loss. Nonetheless, it is possible to break this question down and to summarize what we do and do not know. For example, our experiments in cultured OLs show that ZDHHC9 loss causes call-autonomous deficits in morphological maturation of these cells. We also know that ZDHHC9 loss results in impaired palmitoylation of MBP, a direct substrate for ZDHHC9. Moreover, loss of ZDHHC9 at Golgi outposts in OLs (a phenotype observed with several XLID-associated mutant forms of ZDHHC9, even those with no significant loss of catalytic activity) correlates with intellectual disability. Together, these findings are consistent with a model in which ZDHHC9 action at OL Golgi outposts is critical for normal myelination. However, it is yet to be determined whether the key substrates of ZDHHC9 include MBP, other palmitoyl-proteins that are key constituents of CNS myelin, or proteins whose palmitoylation is important for myelin protein trafficking and targeting. Another non-mutually exclusive possibility is that ZDHHC9 acts at Golgi outposts but indirectly, for example to drive the expression of myelin protein genes. Future experiments, including but not limited to palmitoyl-proteomics in ZDHHC9 (OL-specific) KO mice, will be needed to provide more definitive insights into this issue. We have expanded our Discussion of links between ZDHHC9 mutation and impaired myelination to summarize the above points.

(6) The function of ZDHHC9 in OL may be related to the Golgi apparatus, but its exact role in these structures is still unclear. It is suggested to discuss in more detail the role of ZDHHC9 in the Golgi apparatus in the discussion section.

We appreciate this point, which we considered as related to point (5) above. In our revised Discussion we highlight how ZDHHC9 action at Golgi outposts may involve direct palmitoylation of myelin proteins, palmitoylation of proteins that direct myelin proteins to the myelin membrane and/or activation of gene expression programs that serve to drive myelination. We further note that these possibilities are not mutually exclusive.

(7) More experimental support and in-depth research are needed on the detailed mechanism of how ZDHHC9 and Golga7 cooperatively regulate MBP palmitoylation, and how this decrease in palmitoylation level leads to myelination defects.

This is another important point – our new experiments suggest that, although some XLID mutations markedly affect ZDHHC9’s ability to palmitoylate MBP, others do not, yet all of the mutant forms fail to localize to Golgi outposts. These findings are consistent with a model in which the subcellular location at which ZDHHC9 palmitoylates MBP, and potentially other substrates, is critical for normal myelination. Interestingly, despite their marked differences in basal catalytic activity (as assessed by autopalmitoylation), wt and all XLID forms of ZDHHC9 appear to show enhanced activity (measured by both auto- and MBP palmitoylation) in the presence of ZDHHC9, suggesting that the association with Golga7 (which also localizes to Golgi outposts) is central to ZDHHC9 activity. This model is also highly consistent with the biased expression of Golga7 in OLs, compared to other CNS cell types (Fig 1E, 1F). Moreover, XLID-associated mutant forms of ZDHHC9 also show reduced protein stability and are impaired in their ability to form complexes with Golga7 (also known as Golgi Complex Protein 16kDa; GCP16; PMID: 37035671). Failure of ZDHHC9 XLID mutants to localize to Golgi outposts may thus be due to aberrant trafficking of mutant ZDHHC9 per se, but may also involve impaired association/stabilization of ZDHHC9/Golga7 complexes at these locations. Again, it is possible that either or both of these mechanisms, which are not mutually exclusive, contribute to impaired MBP palmitoylation and/or myelination deficits. We summarize these points in our revised Discussion.

In summary, it is recommended that the authors address the above issues through additional experiments and improved discussions to further strengthen the credibility and clinical relevance of the article.

Recommendations for the authors:  

Reviewer #1 (Recommendations For The Authors):

No gross changes were detected in OL development in Zdhhc9 KO mice and analyses from P28 Zdhhc9 KO mice crossed with Mobp-EGFP reporter mice did not show changes in EGFP+ OL differentiation (Figure 3). However, and given the observed subcellular localization of ZDHHC9 in OL processes (Figure 2) and the observation that the percentage of unmyelinated axons is increased in Zdhhc9 KO (Figure 6), ***early time points to examine the differentiated pools of OLs and their capacity to extend processes/contact axons need to be considered***.

We appreciate this point, but due to the order in which experiments were performed, the ZDHHC9 KO mouse colony that we maintained after initial submission of this work contains homozygous MOBP-EGFP, but not the mT/mG transgene that would be most optimal for the proposed experiment. We hope the reviewer appreciates that it would take considerable time and effort regarding mouse breeding to cross out the MOBP and add back the mT/mG. We nonetheless appreciate the importance of the point raised and therefore examined an earlier developmental time point (P21, 3 weeks) to quantify OLs and NG2+ OPCs. In our updated Fig 3C1-C3, we use Mobp-EGFP mice to show that Zdhhc9 KO does not significantly affect the number of EGFP+ OLs at this time point in the cortex, corpus callosum and spinal cord. We also show that in corpus callosum, Zdhhc9 KO does not significantly affect the number of NG2+ OPCs at this earlier time point (Fig 3D, E). Furthermore, immunostaining to detect BCAS1, a marker of pre-mature OLs, also revealed no qualitative difference with ZDHHC9 loss at P21. We show representative images from these BCAS1 experiments in an updated Fig S3. While these new experiments do not address the morphology of OLs in Zdhhc9 KO, they do provide further evidence that deficits in myelination in young Zdhhc9 KO mice (Figure 6) are not likely due to gross differences in OPC or OL numbers during development.

The authors observed defects in Zdhhc9 KO OL protrusions that they attributed to abnormal OL membrane expansion (Fig 4 and 5). Can they show evidence for this?

This is an important point, and we appreciate the opportunity to explain the reasoning behind our initial statement more fully, while noting that other explanations are possible. Fig 5B (an Imaris-assisted reconstruction using the EGFP cell fill/morphology marker) highlights large spheroid-like distensions along OL processes. We reason that these spheroids are enclosed by the OL lipid membrane because if the membrane were ruptured, the EGFP signal would likely diffuse. This in turn suggests that the caliber of the OL process at the position of the spheroid is grossly abnormal i.e. the membrane has hyper-expanded. Given that OL membrane growth during myelination extends in two directions, i.e., spiral growth to the axonal surface and longitudinal growth along the axon, it is possible that spheroid-like structures are formed by uneven myelin growth. We recognize that we cannot yet conclude whether and how spheroid formation might be linked to the myelination deficit that we observe in Zdhhc9 KO mice.

However, defining the subcellular mechanism for spheroid formation may provide further insights into this issue. We have therefore largely retained the original statement but have added the reasoning above to our revised Discussion.

The authors report that Zdhhc9 KO primary and secondary branches in OL were longer, some contained spheroid-like swellings and the OL protrusion complexity was higher. However, these data is partially contradictory to what they show in OL differentiation experiments in vitro (Fig 7). There is also no evidence for increased membrane expansion in Zdhhc9 knockdown myelin forming cells in culture. How do they reconcile these different findings?

We appreciate the reviewer’s interest in this issue. Several non-mutually exclusive factors could account for the differences in OL morphology in vitro versus in vivo caused by Zdhhc9 loss. First, morphology in vivo may well be influenced by the axons and/or other extrinsic components around each OL that are not present in our primary cultures. Second, OL growth in vivo is highly 3-dimensional, whereas growth in culture is largely 2-dimensional – it may be difficult to support formation of spheroids (by definition, a 3-dimensional structure) in the latter situation. Finally, Zdhhc9 is absent in vivo from the beginning of development until the time points examined, whereas in our cultured OL experiments, Zdhhc9 shRNA is virally delivered to OPC cultures at DIV2 and likely acutely affects Zdhhc9 expression predominantly in committed OLs (following the switch to differentiation medium at DIV3). These differences may also affect the ability of other PATs or, potentially, palmitoylation-independent subcellular processes, to compensate for Zdhhc9 loss. We have more fully explained these points in our revised Discussion. 

Page 7: "The OL processes in this culture condition correspond to large lipid-rich membranous sheets that form spiral membrane expansion on axons in vivo (49)." At which stage are authors referring to? OL processes are extended in culture before membrane formation and this is not clear here. In a 3-days differentiation culture, most OLs have not yet formed a myelin sheath (eg., Figure 2 in Zuchero et al., 2015, Dev Cell).

We appreciate the reviewer highlighting this point. We first note that our oligodendrocyte (OL) culture conditions differ from the immunopanning method used by Zuchero et al., 2015 (original reference (Emery and Dugas, 2013)), which may affect the time course and progression of OL process elaboration and/or myelin sheath formation. We further note that in our cultures most EGFP+ processes are also MBP+ at the time point examined (strictly 3 days plus 9 hours post-differentiation). It thus seems likely that these MBP+ structures largely correspond to the MBP+ wrapping sheaths that occur in vivo, so we have therefore retained our original statement but have added this further explanation.

Minor: Figure 6 (Legend): Time points should be indicated throughout the panels.

We have added this information as requested

Reviewer 2 Recommendations for the Authors:

(1) Regarding the subcellular localization experiment of ZDHHC9 mutants in OL, it is currently limited to in vitro cultured OL, lacking validation in vivo OL or myelin sheath. Additionally, it is necessary to investigate whether the abnormal subcellular localization of ZDHHC9 mutants affects their enzyme activity and palmitoylation modification of substrate proteins.

We thank the reviewer for raising this point. New data in our revised Figure 8 compares autopalmitoylation (sometimes used as a surrogate measure of PAT activity) of ZDHHC9wt and XLID mutants, and their ability to palmitoylate MBP in transfected cells. Intriguingly, we found that autopalmitoylation activity of the ZDHHC9-P150S mutant does not differ significantly from that of ZDHHC9wt, and that this mutant is still capable of palmitoylating MBP. Moreover, the R96W mutant, while impaired in autopalmitoylation, still palmitoylated MBP approximately 50% as effectively as ZDHHC9wt in our cell-based assay. These findings suggest that ZDHHC9-P150S and, probably, ZDHHC9-R96W mutants might still be able to palmitoylate substrates in OLs if they were properly localized. This possibility in turn suggests that impaired subcellular targeting in addition to, or instead of, impaired catalytic activity, may be a key factor in certain cases of ZDHHC9-associated XLID. We have expanded our Figure 8 to show these new experiments and have summarized the conclusions above in our revised Discussion. We thank the reviewer for suggesting that we further investigate this issue.

(2) The experimental period (P21+21 days) using genetic labeling to track the development of myelinating cells may not be long enough. It is recommended to extend the observation time and analyze at more time points to more comprehensively reflect the impact of Zdhhc9 KO.

We appreciate this point from the reviewer but, regrettably, we did not maintain the PdgfraCreER; R26-EGFP; Zdhhc9 KO mouse line and hope the reviewer appreciates that it would take considerable time and effort to rederive this line and then perform the suggested extended time course experiments. However, we note for the reviewer that our preliminary studies did not reveal any effect of Zdhhc9 KO on the number of MOBP-EGFP+ OLs in 6-month-old mice (not shown), consistent with a model in which Zdhhc9 loss does not affect OPC-OL commitment per se.

(3) The author speculates that Zdhhc9 may regulate myelination by affecting the membrane localization of specific myelin proteins, but lacks direct experimental evidence to support this. It is suggested to detect the expression and distribution of relevant proteins in the myelin of Zdhhc9 KO mice.

We share the reviewer’s interest in this point but realized that it is more technically challenging to address than might be initially thought. The main protein we would implicate and seek to test is MBP, but we already found that there is no gross change in MBP distribution in vivo in Zdhhc9 KO mice (Fig 3A). However, an anti-MBP antibody recognizes all forms of MBP, not just the specific splice variants whose palmitoylation is affected by ZDHHC9 loss. Specifically assessing nanoscale distribution of these splice variants would require a way (e.g. am anti-MBP splice form-specific antibody that is compatible with immuno-EM) to distinguish these variants from other, non-palmitoylated forms of MBP. Although such an antibody could be an important tool we hope the reviewers would agree that developing and characterizing such a reagent is beyond the scope of the current study.

We do, however, note that the lack of gross change in MBP distribution and levels in Zdhhc9 KO mice is consistent with the relatively mild phenotype of these mice, compared with shiverer (shi/shi) mice, in which MBP is completely lost. In shiverer, CNS compact myelin is almost absent (PMID: 671037; PMID: 88695; PMID: 460693) and, as the name suggests, mice display a shivering gait, and exhibit seizures and early death. In contrast, Zdhhc9 mice show only subtle behavioral deficits (PMID: 29944857). These differences are all consistent with a model in which Zdhhc9 KO mice, despite their significantly reduced MBP palmitoylation (Fig 8) have grossly normal distribution and levels of MBP when all splice variants are assessed (Fig 3, Fig 8). It is not inconceivable that Zdhhc9 KO mice have a nanoscale change in the distribution of MBP, particularly of specific palmitoylated splice variants, within myelin that profoundly affects myelin ultrastructure, without grossly altering MBP distribution. However, an alternative and not mutually exclusive possibility is that aberrant palmitoylation of other

Zdhhc9 substrates accounts for, or contributes to, the abnormalities in myelin at the ultrastructural level. Addressing this issue would require a multi-pronged approach, not just to assess palmitoylation and distribution of such proteins in Zdhhc9 KO, but also to test whether they are direct Zdhhc9 substrates, in order to rule out indirect effects. We hope reviewers would agree that this is best left to a separate study. However, in our revised Discussion we now summarize what can be inferred regarding Zdhhc9-dependent effects on total and splicevariant specific distribution and levels of MBP.  

(4) Although the article mentions the association of Zdhhc9 with intellectual disabilities, it does not involve behavioral analysis of Zdhhc9 KO mice. It is recommended to supplement some behavioral experimental data to support the important role of Zdhhc9 in maintaining normal cognitive function, enhancing the clinical relevance of the article.

We appreciate this point from the reviewer. The behavior of the same ZDHHC9 KO mouse line that we used was reported in PMID: 31747610 and in PMID: 29944857. In the former study, Zdhhc9 KO mice were reported to display seizures reminiscent of phenotypes in human patients with ZDHHC9 mutation. The latter study assessed performance of Zddhc9 KO mice in several tasks that test cognitive function. Specifically the KO mice were reported to display “altered behaviour in the open-field test, elevated plus maze and acoustic startle test that is consistent with a reduced anxiety level; a reduced hang time in the hanging wire test that suggests underlying hypotonia but which may also be linked to reduced anxiety [and] deficits in the Morris water maze test of hippocampal-dependent spatial learning and memory.”. We have incorporate these findings in our revised Discussion, where we summarize how these phenotypes are common, not just to human patients with ZDHHC9 mutation, but also to other human neurodevelopmental conditions and mouse models in which ID is a common feature.

(5) For the abnormal myelination observed in Zdhhc9 KO mice, including unmyelinated large-diameter axons and excessively myelinated small-diameter axons, the article lacks indepth research and explanation on the exact mechanism and mode of action of ZDHHC9 in regulating myelination.

We share the reviewer’s interest in this point but again note that gaining definitive insights into this issue is far from trivial. Convincing evidence of a causative mechanism would require an exhaustive identification of ZDHHC9 in vivo substrates, followed by point mutation of substrate palmitoylation site(s) to determine the extent to which palmitoylation of such protein(s) phenocopies ZDHHC9 loss. Nonetheless, it is possible to break this question down and to summarize what we do and do not know. For example, our experiments in cultured OLs show that ZDHHC9 loss causes call-autonomous deficits in morphological maturation of these cells. We also know that ZDHHC9 loss results in impaired palmitoylation of MBP, a direct substrate for ZDHHC9. Moreover, loss of ZDHHC9 at Golgi outposts in OLs (a phenotype observed with several XLID-associated mutant forms of ZDHHC9, even those with no significant loss of catalytic activity) correlates with intellectual disability. Together, these findings are consistent with a model in which ZDHHC9 action at OL Golgi outposts is critical for normal myelination. However, it is yet to be determined whether the key substrates of ZDHHC9 include MBP, other palmitoyl-proteins that are key constituents of CNS myelin, or proteins whose palmitoylation is important for myelin protein trafficking and targeting. Another non-mutually exclusive possibility is that ZDHHC9 acts at Golgi outposts but indirectly, for example to drive the expression of myelin protein genes. Future experiments, including but not limited to palmitoyl-proteomics in ZDHHC9 (OL-specific) KO mice, will be needed to provide more definitive insights into this issue. We have expanded our Discussion of links between ZDHHC9 mutation and impaired myelination to summarize the above points.

(6) The function of ZDHHC9 in OL may be related to the Golgi apparatus, but its exact role in these structures is still unclear. It is suggested to discuss in more detail the role of ZDHHC9 in the Golgi apparatus in the discussion section.

We appreciate this point, which we considered as related to point (5) above. In our revised Discussion we highlight how ZDHHC9 action at Golgi outposts may involve direct palmitoylation of myelin proteins, palmitoylation of proteins that direct myelin proteins to the myelin membrane and/or activation of gene expression programs that serve to drive myelination. We further note that these possibilities are not mutually exclusive.

(7) More experimental support and in-depth research are needed on the detailed mechanism of how ZDHHC9 and Golga7 cooperatively regulate MBP palmitoylation, and how this decrease in palmitoylation level leads to myelination defects.

This is another important point – our new experiments suggest that, although some XLID mutations markedly affect ZDHHC9’s ability to palmitoylate MBP, others do not, yet all of the mutant forms fail to localize to Golgi outposts. These findings are consistent with a model in which the subcellular location at which ZDHHC9 palmitoylates MBP, and potentially other substrates, is critical for normal myelination. Interestingly, despite their marked differences in basal catalytic activity (as assessed by autopalmitoylation), wt and all XLID forms of ZDHHC9 appear to show enhanced activity (measured by both auto- and MBP palmitoylation) in the presence of ZDHHC9, suggesting that the association with Golga7 (which also localizes to Golgi outposts) is central to ZDHHC9 activity. This model is also highly consistent with the biased expression of Golga7 in OLs, compared to other CNS cell types (Fig 1E, 1F). Moreover, XLID-associated mutant forms of ZDHHC9 also show reduced protein stability and are impaired in their ability to form complexes with Golga7 (also known as Golgi Complex Protein 16kDa; GCP16; PMID: 37035671). Failure of ZDHHC9 XLID mutants to localize to Golgi outposts may thus be due to aberrant trafficking of mutant ZDHHC9 per se, but may also involve impaired association/stabilization of ZDHHC9/Golga7 complexes at these locations. Again, it is possible that either or both of these mechanisms, which are not mutually exclusive, contribute to impaired MBP palmitoylation and/or myelination deficits. We summarize these points in our revised Discussion.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #2 (Public review):

This manuscript determines how PA28g, a proteasome regulator that is overexpressed in tumors, and C1QBP, a mitochondrial protein for maintaining oxidative phosphorylation that plays a role in tumor progression, interact in tumor cells to promote their growth, migration and invasion. Evidence for the interaction and its impact on mitochondrial form and function was provided although it is not particularly strong.

The revised manuscript corrected mislabeled data in figures and provides more details in figure legends. Misleading sentences and typos were corrected. However, key experiments that were suggested in previous reviews were not done, such as making point mutations to disrupt the protein interactions and assess the consequence on protein stability and function. Results from these experiments are critical to determine whether the major conclusions are fully supported by the data.

The second revision of the manuscript included the proximity ligation data to support the PA28g-C1QBP interaction in cells. However, the method and data were not described in sufficient detail for readers to understand. The revision also includes the structural models of the PA28g-C1QBP complex predicted by AlphaFold. However, the method and data were not described with details for readers to understand how this structural modeling was done, what is the quality of the resulting models, and the physical nature of the protein-protein interaction such as what kind of the non-covalent interactions exist in the interface of the protein complexes. Furthermore, while the interactions mediated by the protein fragments were tested by pull-down experiments, the interactions mediated by the three residues were not tested by mutagenesis and pull-down experiments. In summary, the revision was improved, but further improvement is needed.

Thank you very much for your comments.

(1) Based on your suggestion, we predicted the possible interaction sites using AlphaFold 3 and found that mutations in amino acids 76 and 78 of C1QBP affect the interaction with PA28γ (Revised Appendix Figure 1J). Subsequently, pulldown experiment also found that after mutating the amino acids at the two aforementioned sites (T76A, G78N), C1QBP that could bind to PA28γ decreased (Revised Figure 1J). The above results confirm that PA28γ could interacts with C1QBP, in a manner dependent on the N-terminus of C1QBP. These findings are now included in the revised manuscript “In addition, we employed AlphaFold 3 to perform energy minimization and predict hydrogen bonds between the C1QBP N-terminus (amino acids 1-167) and the PA28γ protein interaction region. The results suggest that the T76 and G78 residues of C1QBP may be key contributors to the interaction. Consistently, coimmunoprecipitation analysis demonstrated that mutations at these sites (C1QBPT76A and C1QBPG78N) significantly reduced the binding ability to PA28γ (Fig. 1J and Appendix Fig. 1J)”, specifically in results section. We believe this additional validation strengthens the robustness of our findings.

(2) According to your suggestion, we have added a description of the results of PLA in the figure legend (Revised Figure 1C) and the method of PLA in the appendix file (Revised Appendix file, Part “Proximity Ligation Assay”). The revised text reads as follows: (C) PLA image of UM1 cells shows the interaction between C1QBP and PA28γ in both cytoplasm and nucleus (red fluorescence).

(3) In the light of your suggestion, we have enriched the description of AlphaFold 3 analysis in the appendix file (Revised Appendix file, Page 10-11). The revised text reads as follows:

“Prediction and Analysis of Protein Interactions

Protein Sequence Retrieval and Structure Prediction

The protein sequences of C1QBP and PA28γ were obtained from the AlphaFold Protein Structure Database. Structural predictions of the protein-protein interaction between C1QBP and PA28γ were conducted using AlphaFold 3. The plDDT (predicted local distance difference test) values were utilized to assess the confidence of the predicted models. Models with a plDDT score above 70 were considered confident, while those with a score above 90 were categorized as very high confidence. These values were annotated in the figures to indicate the reliability of the structural predictions.”

“Protein Preparation and Structure Optimization

The best-scored model for the C1QBP-PA28γ interaction predicted by AlphaFold 3 was selected for further analysis. The model was imported into MOE 2022 (Molecular Operating Environment) software for protein preparation. This process included the removal of water molecules and other heteroatoms, followed by the addition of hydrogen atoms to the structure. This step was essential for optimizing the protein’s 3D conformation and ensuring the correctness of the protonation states at physiological pH.”

“Energy Minimization and Hydrogen Bond Prediction

The protein structure was subjected to energy minimization using the Amber10: EHT (Effective Hamiltonian Theory) force field, with R-field 1: 80 settings to refine the model’s geometry. The minimization process was performed to optimize the protein’s internal energy and ensure stable conformation, followed by calculation of hydrogen bond interactions. The interaction energies and hydrogen bonds were analyzed to identify potential binding sites and stabilize the predicted protein-protein complex.”

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public Review):

Astrocytes are known to express neuroligins 1-3. Within neurons, these cell adhesion molecules perform important roles in synapse formation and function. Within astrocytes, a significant role for neuroligin 2 in determining excitatory synapse formation and astrocyte morphology was shown in 2017. However, there has been no assessment of what happens to synapses or astrocyte morphology when all three major forms of neuroligins within astrocytes (isoforms 1-3) are deleted using a well characterized, astrocyte specific, and inducible cre line. By using such selective mouse genetic methods, the authors here show that astrocytic neuroligin 1-3 expression in astrocytes is not consequential for synapse function or for astrocyte morphology. They reach these conclusions with careful experiments employing quantitative western blot analyses, imaging and electrophysiology. They also characterize the specificity of the cre line they used. Overall, this is a very clear and strong paper that is supported by rigorous experiments. The discussion considers the findings carefully in relation to past work. This paper is of high importance, because it now raises the fundamental question of exactly what neuroligins 1-3 are actually doing in astrocytes. In addition, it enriches our understanding of the mechanisms by which astrocytes participate in synapse formation and function. The paper is very clear, well written and well illustrated with raw and average data.

Comments on revisions:

My previous comments have been addressed. I have no additional points to make and congratulate the authors.

Thank you for your acceptance.

Reviewer #2 (Public Review):

In the present manuscript, Golf et al. investigate the consequences of astrocyte-specific deletion of Neuroligin (Nlgn) family cell adhesion proteins on synapse structure and function in the brain. Decades of prior research had shown that Neuroligins mediate their effects at synapses through their role in the postsynaptic compartment of neurons and their transsynaptic interaction with presynaptic Neurexins. More recently, it was proposed for the first time that Neuroligins expressed by astrocytes can also bind to presynaptic Neurexins to regulate synaptogenesis (Stogsdill et al. 2017, Nature). However, several aspects of the model proposed by Stogsdill et al. on astrocytic Neuroligin function conflict with prior evidence on the role of Neuroligins at synapses, prompting Golf et al. to further investigate astrocytic Neuroligin function in the current study. Using postnatal conditional deletion of Nlgn1-3 specifically from astrocytes in mice, Golf et al. show that virtually no changes in the expression of synaptic proteins or in the properties of synaptic transmission at either excitatory or inhibitory synapses are observed. Moreover, no alterations in the morphology of astrocytes themselves were found. To further extend this finding, the authors additionally analyzed human neurons co-cultured with mouse glia lacking expression of Nlgn1-4. No difference in excitatory synaptic transmission was observed between neurons cultured in the present of wildtype vs. Nlgn1-4 conditional knockout glia. The authors conclude that while Neuroligins are indeed expressed in astrocytes and are hence likely to play some role there, this role does not include any direct consequences on synaptic structure and function, in direct contrast to the model proposed by Stogsdill et al.

Overall, this is a strong study that addresses a fundamental and highly relevant question in the field of synaptic neuroscience. Neuroligins are not only key regulators of synaptic function, they have also been linked to numerous psychiatric and neurodevelopmental disorders, highlighting the need to precisely define their mechanisms of action. The authors take a wide range of approaches to convincingly demonstrate that under their experimental conditions, Nlgn1-3 are efficiently deleted from astrocytes in vivo, and that this deletion does not lead to major alterations in the levels of synaptic proteins or in synaptic transmission at excitatory or inhibitory synapses, or in the morphology of astrocytes. While the co-culture experiments are somewhat more difficult to interpret due to lack of a control for the effect of wildtype mouse astrocytes on human neurons, they are also consistent with the notion that deletion of Nlgn1-4 from astrocytes has no consequences for the function of excitatory synapses. Together, the data from this study provide compelling and important evidence that, whatever the role of astrocytic Neuroligins may be, they do not contribute substantially to synapse formation or function under the conditions investigated.

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

The authors have fully addressed my concerns, and have in particular conducted a very elegant and compelling analysis of the degree of deletion of astrocytic Nlgn1-3/4 in their models. This greatly strengthens the main claims of their study and the fundamental nature of their conclusions for the field of synapse biology.

I am somewhat less convinced by the newly added experiment to investigate deletion of Nlgns1-4 from glia in glia-neuron co-cultures. The authors provide no evidence to show that either WT or cKO glia have any effect on synapse formation or function in human neurons, and therefore, the current lack of a difference could equally result from the fact that both WT and cKO glia were non-functional altogether. The authors cite two studies to state that human neurons do not form synapses in the absence of astrocytes, Zhang et al. 2013 and Huang et al. 2017, but neither seem to be listed in the references (unless Zhang et al. 2014 was meant), making it difficult to assess the relevance of these data. However, since the data on astrocytic Nlgn1-3 deletion in vivo are compelling on their own, I do not see the co-culture experiment as essential for the main conclusions of the study.

Minor comment:

Please add the information on the strain background of the mice to the methods section of the manuscript. Strain background can have a significant impact on many aspects of neuronal function, and this information is therefore essential for the interpretation of potential differences to other studies.

We deeply apologize for forgetting to include the two important references mentioned by the reviewer in the reference list. We understand that the reviewer as a result could not assess the validity of our statement that co-culture of glia is required for efficient synapse formation by human neurons that are induced from ES or iPS cells. Note that this conclusion does not postulate that all synapse formation requires glia, since the cited papers demonstrate that human neurons induced by our protocol still form scarce synapses without glia. This observation has been confirmed in many different experiments that were performed after the data presented in the cited papers. As a result of this extensive prior documentation that human neurons produced by forced expression of Ngn2 require coculture of glia for efficient synapse formation, we do not feel that we need to repeat this basic characterization of our culture system again to validate multiple previous papers and hope the reviewer will concur. We have additionally added the relevant mouse strain information to the methods section.

Author response:

Reviewer #1:

Point 1

Not many weaknesses, but probably validation at more enhancers could have made the paper stronger.

We experimentally validated two sets of enhancers from two distinct tissues and observed similar effects. While this supports the idea that the TEAD-tissue-specific TF interaction we observe is not restricted to a single tissue, we agree that testing additional enhancers from a third tissue would strengthen our conclusions. We will acknowledge in the discussion that including a third tissue could provide additional support for the generality of our findings.

Reviewer #2:

Point 1

The authors propose a mechanism of a TF trio (TEAD - CHD4 - tissue-specific TFs). However, only one validation experiment checked CHD4. CHD4 binding was not mentioned at all in the other cases.

Indeed, CHD4 binding was experimentally validated at only one enhancer. This was a deliberate decision based on two key considerations:

(1) Consistent functional response across enhancers: We tested multiple enhancers (n =8) for functional response to the TEAD+YAP and GATA4/6 combination. All enhancers tested exhibited the same trend—attenuation of GATA-mediated activation upon co-expression of TEAD or TEAD/YAP. This consistent pattern supports a shared mechanism across these elements.

(2) Substantial prior evidence supporting CHD4 recruitment by both GATA4 and YAP: Specifically, CHD4 recruitment by GATA4 has been described in the context of cardiovascular development[1], and CHD4 can also be recruited by TEAD coactivator YAP2. Furthermore, published genomic occupancy data from embryonic heart tissue show widespread co-binding of GATA4, TEAD, and CHD4[1,3], including at most of the cardiac enhancers we functionally tested (4 out of 5).

Given the consistent enhancer responses and the supporting literature and genomic data indicating TEAD-CHD4 co-occupancy, we chose to validate CHD4 binding at a representative enhancer as a proof of concept.

We will clarify this rationale in the revised manuscript to better address this concern.

Reviewer #2:

Point 2

The authors integrated E12.5 TEAD binding with E11.5 acetylation data, and it would be important to show that this experimental approach is valid or otherwise qualify its limitations.

We will provide additional evidence in support of this approach in the revised manuscript or alternatively acknowledge its limitations.

Reviewer #2:

Point 3

Motif co-occurrence analysis was extended to claiming TF interactions without further validation.

We thank the reviewer for pointing out this important distinction. We reviewed the manuscript and identified seven instances where TF interactions were mentioned. Four of these correctly refer to previously established protein-protein interactions. For the remaining instances, we will adjust the wording to reflect the level of evidence, e.g.  describe combinatorial binding based on motif co-occurrence, rather than implying direct interaction.

Reviewer #3:

Point 1

Much of this manuscript focuses on confirming transcription factor relationships that have been reported previously. For example, it is well known that GATA4 interacts with MEF2 in the ventricle. There are limited new or unexpected associations discussed and tested.

We thank the reviewer for this important observation and see the recurrence of known interactions, such as GATA4-MEF2, not as a drawback, but as an important validation of our methodology.

The identification of novel TF-TF combinations was geared toward uncovering shared regulatory principles across diverse human developmental tissues. While analysing 13 heterogeneous embryonic tissues introduced limitations, such as cellular complexity that may obscure rare interactions, it also allowed the identification of robust, recurrent patterns across tissues.  Indeed, using this approach, we identified the widespread combinatorial effect of TEAD in partnership with lineage-specific TFs, which is explored more in depth in the manuscript.

Another main goal of the study was to develop and demonstrate a generalizable strategy for identifying combinatorial TF binding patterns that underlie tissue-specific gene regulation. Given the inherent heterogeneity of the embryonic organs analysed, the approach is naturally biased toward recovering the most prevalent, and often well-characterized, TF combinations. While we fully acknowledge this limitation, we believe that the ability to robustly recover well-established TF partnerships across multiple organs provides a valuable proof of concept. The next step will be to apply this strategy to single-cell RNA datasets, in order to define TF relationships at higher resolution, for example, resolving associations down to specific family members that cooperate within distinct lineages or cell types, and identifying less frequent or underrepresented TF-TF relationships.

In this context, we believe that our strategy has successfully highlighted shared enhancer logic and offers a framework for future high-resolution dissection of TF cooperativity at the single-cell level. The rationale for analysing heterogeneous tissues, along with its limitations, will be addressed in the revised version.

Reviewer #3:

Point 2

Embryonic tissues are highly heterogeneous, limiting the utility of the bulk ChIP-seq employed in these analyses. Does the cellular heterogeneity explain the discrepancy between TEAD binding and histone acetylation? Similarly, how does conservation between species affect the TF predictions?

We thank the reviewer for raising these important points. We acknowledge the limitations of using bulk ChIP-seq data in the context of complex embryonic tissues (see also previous point). We cannot exclude that the discrepancy between TEAD binding and histone acetylation is an effect of cellular heterogeneity. Indeed, we mention in the results “Our ventricle-specific enhancers were sampled at a single time point and likely represent enhancers that are selectively active in different cell types and developmental stages, given the heterogeneity of cell types in the ventricle”. The limitation of bulk ChIP-seq will be addressed in the discussion. In the specific case of the enhancers selected for validation, the binding site sequences are conserved between species, suggesting that the cis-regulatory activity is likely to be similar in both.

Reviewer #3:

Point 3

Some of the interpretations should also be fleshed out a bit more to clarify the advantage of the analyses presented here. For example, if Gata4 and Foxa2 transcripts are expressed during different stages of development, then it's likely that (as stated by the authors) these motifs are not used during the same stage of development. But examining the flanking regions wasn't necessary to make that statement. This type of conclusion seems tangential to the benefit of this analysis, which is to understand which TFs work together in a single organ at a single time point.

We appreciate the reviewer’s comment and the opportunity to clarify our interpretation. The reviewer refers to the finding that GATA4 and FOXA2 motifs are flanked by different sets of motifs in liver enhancers, suggesting that these TFs operate within distinct regulatory contexts.

Our aim was not to state that GATA4 and FOXA2 do not function simultaneously—this can indeed be inferred from their non-overlapping expression patterns. Rather, we intended to highlight the potential of our approach, even when applied to bulk data, to resolve distinct regulatory modules that may act in different subpopulations of cells or developmental windows within the same tissue.

We will revise the relevant section of the manuscript to make this interpretative point clearer.

Reviewer #3:

Point 4

This manuscript hinges on luciferase assays whose results can be difficult to translate to complex gene regulation networks. Many motifs are often clustered together, which makes designing experiments at endogenous loci important in studies such as this one.

We agree with the Reviewer that luciferase assays represent an oversimplified model of gene regulation and do not fully capture the complexity of endogenous regulatory networks. We will explicitly acknowledge this limitation in the discussion.

Mutagenesis of TEAD and tissue-specific TF motifs at endogenous loci would provide more conclusive evidence. However, our goal was to test the generality of TEAD effect across multiple enhancers and tissues. Despite its limitations, a luciferase-based assay was the most feasible approach, as an endogenous strategy would not have allowed us to assess a broader set of enhancers efficiently. Additionally, the presence of recurrent motifs and the potential functional redundancy among enhancers targeting the same gene can complicate the interpretation of single-locus perturbations.

References

(1) Robbe ZL, Shi W, Wasson LK, Scialdone AP, Wilczewski CM, Sheng X, et al. CHD4 is recruited by GATA4 and NKX2-5 to repress noncardiac gene programs in the developing heart. Genes Dev. 2022 Apr 1;36(7–8):468–82.

(2) Kim M, Kim T, Johnson RL, Lim DS. Transcriptional Co-repressor Function of the Hippo Pathway Transducers YAP and TAZ. Cell Rep. 2015 Apr;11(2):270–82.

(3) Akerberg BN, Gu F, VanDusen NJ, Zhang X, Dong R, Li K, et al. A reference map of murine cardiac transcription factor chromatin occupancy identifies dynamic and conserved enhancers. Nat Commun. 2019 Oct 28;10(1):4907.

Author response:

The following is the authors’ response to the original reviews

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

The authors state the study's goal clearly: "The goal of our study was to understand to what extent animal individuality is influenced by situational changes in the environment, i.e., how much of an animal's individuality remains after one or more environmental features change." They use visually guided behavioral features to examine the extent of correlation over time and in a variety of contexts. They develop new behavioral instrumentation and software to measure behavior in Buridan's paradigm (and variations thereof), the Y-maze, and a flight simulator. Using these assays, they examine the correlations between conditions for a panel of locomotion parameters. They propose that inter-assay correlations will determine the persistence of locomotion individuality.

Strengths: 

The OED defines individuality as "the sum of the attributes which distinguish a person or thing from others of the same kind," a definition mirrored by other dictionaries and the scientific literature on the topic. The concept of behavioral individuality can be characterized as: 

(1) a large set of behavioral attributes, 

(2) with inter-individual variability, that are 

(3) stable over time. 

A previous study examined walking parameters in Buridan's paradigm, finding that several parameters were variable between individuals, and that these showed stability over separate days and up to 4 weeks (DOI: 10.1126/science.aaw718). The present study replicates some of those findings and extends the experiments from temporal stability to examining the correlation of locomotion features between different contexts.

The major strength of the study is using a range of different behavioral assays to examine the correlations of several different behavior parameters. It shows clearly that the inter-individual variability of some parameters is at least partially preserved between some contexts, and not preserved between others. The development of highthroughput behavior assays and sharing the information on how to make the assays is a commendable contribution.

We thank the reviewer for his exceptionally kind assessment of our work!

Weaknesses: 

The definition of individuality considers a comprehensive or large set of attributes, but the authors consider only a handful. In Supplemental Fig. S8, the authors show a large correlation matrix of many behavioral parameters, but these are illegible and are only mentioned briefly in Results. 

We have now uploaded a high-resolution PDF to the Github Address: https://github.com/LinneweberLab/Mathejczyk_2024_eLife_Individuality/blob/main/S8.pdf, and this is also mentioned in the figure legend for Fig. S8

Why were five or so parameters selected from the full set? How were these selected? 

The five parameters (% of time walked, walking speed, vector strength, angular velocity, and centrophobicity) were selected because they describe key aspects of the investigated behaviors that can be compared directly across assays. Importantly, several parameters we typically use (e.g., Linneweber et al., 2020) cannot be applied under certain conditions, such as darkness or the absence of visual cues. Furthermore, these five parameters encompass three critical aspects of navigation across standard visual behavioral arenas: (1) The “exploration” category is characterized by parameters describing the fly’s activity. (2) Parameters related to “attention” reflect heightened responses to visual cues, but unlike commonly used metrics such as angle or stripe deviations (e.g., Coulomb, 2012; Linneweber et al., 2020), they can also be measured in absence of visual cues and are therefore suitable for cross-assay comparisons. (3) The parameter “centrophobicity,” used as a potential indicator of anxiety, is conceptually linked to the open-field test in mice, where the ratio of wall-to-open-field activity is frequently calculated as a measurement of anxiety (see for example Carter, Sheh, 2015, chapter 2. https://www.sciencedirect.com/book/9780128005118/guide-to-researchtechniques-in-neuroscience). Admittedly, this view is frequently challenged in mice, but it has a long history which is why we use it.

Do the correlation trends hold true across all parameters? For assays in which only a subset of parameters can be directly compared, were all of these included in the analysis, or only a subset? 

As noted above, we only included a subset of parameters in our final analysis, as many were unsuitable for comparison across assays while still providing valuable assayspecific information which are important to relate these results to previous publications.

The correlation analysis is used to establish stability between assays. For temporal retesting, "stability" is certainly the appropriate word, but between contexts, it implies that there could be 'instability'. Rather, instead of the 'instability' of a single brain process, a different behavior in a different context could arise from engaging largely (or entirely?) distinct context-dependent internal processes, and have nothing to do with process stability per se. For inter-context similarities, perhaps a better word would be "consistency". 

Thank you for this suggestion. During the preparation of the manuscript, we indeed frequently alternated between the terms “stability” and “consistency.” And decided to go with “stability” as the only descriptor, to keep it simple. We now fully agree with the reviewer’s argument and have replaced “stability” by “consistency” throughout the current version of the manuscript in order to increase clarity and coherence.

The parameters are considered one by one, not in aggregate. This focuses on the stability/consistency of the variability of a single parameter at a time, rather than holistic individuality. It would appear that an appropriate measure of individuality stability (or individuality consistency) that accounts for the high-dimensional nature of individuality would somehow summarize correlations across all parameters. Why was a multivariate approach (e.g. multiple regression/correlation) not used? Treating the data with a multivariate or averaged approach would allow the authors to directly address 'individuality stability' and analyses of single-parameter variability stability.

We agree with the reviewer that a multivariate analysis adds clear advantages in terms of statistical power, in addition to our chosen approach. On one hand, we believe that the simplicity of our initial analysis, both for correlational and mean data, makes easy for readers to understand and reproduce our data. While preparing the previous version of the manuscript we were skeptical since more complex analyses often involve numerous choices, which can complicate reproducibility. For instance, a recent study in personality psychology (Paul et al., 2024) highlighted the risks of “forking paths” in statistical analysis, showing that certain choices of statistical methods could even reverse findings—a concern mitigated by our simplistic straightforward approach. Still, in preparation of this revised version of the manuscript, we accepted the reviewer’s advice and reanalyzed the data using a generalized linear model. This analysis nicely recapitulates our initial findings and is now summarized in a single figure (Fig. 9).

The correlation coefficients are sometimes quite low, though highly significant, and are deemed to indicate stability. For example, in Figure 4C top left, the % of time walked at 23{degree sign}C and 32{degree sign}C are correlated by 0.263, which corresponds to an R2 of 0.069 i.e. just 7% of the 32{degree sign}C variance is predictable by the 23{degree sign}C variance. Is it fair to say that a 7% determination indicates parameter stability? Another example: "Vector strength was the most correlated attention parameter... correlations ranged... to -0.197," which implies that 96% (1 - R2) of Y-maze variance is not predicted by Buridan variance. At what level does an r value not represent stability?

We agree that this is an important question. Our paper clearly demonstrates that individuality always plays a role in decision-making (and, in this context, any behavioral output can be considered a decision). However, the non-linear relationship between certain situations and the individual’s behavior often reduces the predictive value (or correlation) across contexts, sometimes quite drastically.

For instance, temperature has a relatively linear effect on certain behavioral parameters, leading to predictable changes across individuals. As a result, correlations across temperature conditions are often similar to those observed across time within the same situation. In contrast, this predictability diminishes when comparing conditions like the presence or absence of visual stimuli, the use of different arenas, or different modalities.

For this reason, we believe that significance remains the best indicator for describing how measurable individuality persists, even across vastly different situations.

The authors describe a dissociation between inter-group differences and interindividual variation stability, i.e. sometimes large mean differences between contexts, but significant correlation between individual test and retest data. Given that correlation is sensitive to slope, this might be expected to underestimate the variability stability (or consistency). Is there a way to adjust for the group differences before examining the correlation? For example, would it be possible to transform the values to in-group ranks prior to correlation analysis?  

We thank the reviewer for this suggestion, and we have now addressed this point. To account for slope effects, we have now introduced in-group ranks for our linear model computation (see Fig. 9). 

What is gained by classifying the five parameters into exploration, attention, and anxiety? To what extent have these classifications been validated, both in general and with regard to these specific parameters? Is the increased walking speed at higher temperatures necessarily due to an increased 'explorative' nature, or could it be attributed to increased metabolism, dehydration stress, or a heat-pain response? To what extent are these categories subjective?

We agree that grouping our parameters into traits like exploration, attention, and anxiety always includes subjective decisions. The classification into these three categories is even considered partially controversial in the mouse specific literature, which uses the term “anxiety” in similar experiments (see for exampler Carter, Sheh, 2015, chapter 2 . https://www.sciencedirect.com/book/9780128005118/guide-to-research-techniquesin-neuroscience). Nevertheless, we believe that readers greatly benefit from these categories, since they make it easier to understand (beyond mathematical correlations) which aspects of the flies’ individuality can be considered consistent across situations. Furthermore, these categories serve as a bridge to compare insight from very distinct models.

The legends are quite brief and do not link to descriptions of specific experiments. For example, Figure 4a depicts a graphical overview of the procedure, but I could not find a detailed description of this experiment's protocol.

We assume the reviewer is referring to Figure 3a. The detailed experimental protocol can be found in the Materials and Methods section under Setup 2: IndyTrax Multi-Arena Platform. We have now clarified this in the mentioned figure legend.

Using the current single-correlation analysis approach, the aims would benefit from rewording to appropriately address single-parameter variability stability/consistency (as distinct from holistic individuality). Alternatively, the analysis could be adjusted to address the multivariate nature of individuality, so that the claims and the analysis are in concordance with each other.

The reviewer raises an important point about hierarchies within the concept of animal individuality or personality. We agree that this is best addressed by first focusing on single behavioral traits/parameters and then integrating several trait properties into a cohesive concept of animal personality (holistic individuality). To ensure consistency throughout the text, we have now thoroughly reviewed the entire manuscript clearly distinguish between single-parameter variability stability/consistency and holistic individuality/personality.

The study presents a bounty of new technology to study visually guided behaviors. The GitHub link to the software was not available. To verify the successful transfer of open hardware and open-software, a report would demonstrate transfer by collaboration with one or more other laboratories, which the present manuscript does not appear to do. Nevertheless, making the technology available to readers is commendable.

We have now uploaded all codes and materials to GitHub and made them available as soon as we received the reviewers’ comments. All files and materials can be accessed at https://github.com/LinneweberLab/Mathejczyk_2024_eLife_Individuality, which is now frequently mentioned throughout the revised manuscript.

The study discusses a number of interesting, stimulating ideas about inter-individual variability, and presents intriguing data that speaks to those ideas, albeit with the issues outlined above.

While the current work does not present any mechanistic analysis of inter-individual variability, the implementation of high-throughput assays sets up the field to more systematically investigate fly visual behaviors, their variability, and their underlying mechanisms. 

We thank the reviewer again for the extensive and constructive feedback.

Reviewer #2 (Public Review): 

Summary: 

The authors repeatedly measured the behavior of individual flies across several environmental situations in custom-made behavioral phenotyping rigs.

Strengths: 

The study uses several different behavioral phenotyping devices to quantify individual behavior in a number of different situations and over time. It seems to be a very impressive amount of data. The authors also make all their behavioral phenotyping rig design and tracking software available, which I think is great and I'm sure other folks will be interested in using and adapting it to their own needs.

We thank the reviewer for highlighting the strengths of our study.

Weaknesses/Limitations: 

I think an important limitation is that while the authors measured the flies under different environmental scenarios (i.e. with different lighting and temperature) they didn't really alter the "context" of the environment. At least within behavioral ecology, context would refer to the potential functionality of the expressed behaviors so for example, an anti-predator context, a mating context, or foraging. Here, the authors seem to really just be measuring aspects of locomotion under benign (relatively low-risk perception) contexts. This is not a flaw of the study, but rather a limitation to how strongly the authors can really say that this demonstrates that individuality is generalized across many different contexts. It's quite possible that rank order of locomotor (or other) behaviors may shift when the flies are in a mating or risky context. 

We agree with the reviewer that the definition of environmental context can differ between fields and that behavioral context is differently defined, particularly in ecology. Nevertheless, we highlight that our alternations of environmental context are highly stereotypic, well-defined, and unbiased from any interpretation (we only modified what we stated in the experimental description without designing a specific situation that might be again perceived individually differently. E.g., comparing a context with a predator and one without might result in a binary response because one fraction of the tested individuals might perceive the predator in the predator situation, and the other half does not. 

The analytical framework in terms of statistical methods is lacking. It appears as though the authors used correlations across time/situations to estimate individual variation; however, far more sophisticated and elegant methods exist. The paper would be a lot stronger, and my guess is, much more streamlined if the authors employ hierarchical mixed models to analyse these data these models could capture and estimate differences in individual behavior across time and situations simultaneously. Along with this, it's currently unclear whether and how any statistical inference was performed. Right now, it appears as though any results describing how individuality changes across situations are largely descriptive (i.e. a visual comparison of the strengths of the correlation coefficients?). 

The reviewer raises an important point, also raised by reviewer #1. On one hand, we agree with both reviewers that a more aggregated analysis has clear advantages like more statistical power and has the potential to streamline our manuscript, which is why we added such an analysis (see below). On the other hand, we would also like to defend the initial approach we took, since we think that the simplicity of the analysis for both correlational and mean data is easy to understand and reproduce. More complex analyses necessarily include the selection of a specific statistical toolbox by the experimenters and based on these decisions, different analyses become less comparable and more and more complicated to reproduce, unless the entire decision tree is flawlessly documented. For instance, a recent personality psychology paper investigated the relationship between statistical paths within the decision tree (forking analysis) and their results, leading to very surprising results (Paul et al., 2024), since some paths even reversed their findings. Such a variance in conclusions is hardly possible with the rather simplistic and easily reproducible analysis we performed. One of the major strengths of our study is the simple experimental design, allowing for rather simple and easy to understand analyses.

We nevertheless took the reviewer’s advice very seriously and reanalyzed the data using a generalized linear model, which largely recapitulated the findings of our previously performed “low-tech” analysis in a single figure (Fig. 9).

Another pretty major weakness is that right now, I can't find any explicit mention of how many flies were used and whether they were re-used across situations. Some sort of overall schematic showing exactly how many measurements were made in which rigs and with which flies would be very beneficial. 

We apologize for this inconvenience. A detailed overview of male and female sample sizes has been listed in the supplemental boxplots next to the plots (e.g, Fig S6). Apparently, this was not visible enough. Therefore, we have now also uniformly added the sample sizes to the main figure legends.

I don't necessarily doubt the robustness of the results and my guess is that the author's interpretations would remain the same, but a more appropriate modeling framework could certainly improve their statistical inference and likely highlight some other cool patterns as these methods could better estimate stability and covariance in individual intercepts (and potentially slopes) across time and situation.

As described above, we have now added the suggested analyses. We hope that the reviewer will appreciate the new Fig. 9, which, in our opinion, largely confirms our previous findings using a more appropriate generalized linear modelling framework.

Reviewer #3 (Public Review): 

This manuscript is a continuation of past work by the last author where they looked at stochasticity in developmental processes leading to inter-individual behavioural differences. In that work, the focus was on a specific behaviour under specific conditions while probing the neural basis of the variability. In this work, the authors set out to describe in detail how stable the individuality of animal behaviours is in the context of various external and internal influences. They identify a few behaviours to monitor (read outs of attention, exploration, and 'anxiety'); some external stimuli (temperature, contrast, nature of visual cues, and spatial environment); and two internal states (walking and flying).

They then use high-throughput behavioural arenas - most of which they have built and made plans available for others to replicate - to quantify and compare combinations of these behaviours, stimuli, and internal states. This detailed analysis reveals that:

(1) Many individualistic behaviours remain stable over the course of many days. 

(2) That some of these (walking speed) remain stable over changing visual cues. Others (walking speed and centrophobicity) remain stable at different temperatures.

(3) All the behaviours they tested failed to remain stable over the spatially varying environment (arena shape).

(4) Only angular velocity (a readout of attention) remains stable across varying internal states (walking and flying).

Thus, the authors conclude that there is a hierarchy in the influence of external stimuli and internal states on the stability of individual behaviours.

The manuscript is a technical feat with the authors having built many new highthroughput assays. The number of animals is large and many variables have been tested - different types of behavioural paradigms, flying vs walking, varying visual stimuli, and different temperatures among others. 

We thank the reviewer for this extraordinary kind assessment of our work!

Recommendations for the authors:  

Reviewing Editor (Recommendations For The Authors): 

While appreciating the effort and quality of the work that went into this manuscript, the reviewers identified a few key points that would greatly benefit this work.

(1) Statistical methods adopted. The dataset produced through this work is large, with multiple conditions and comparisons that can be made to infer parameters that both define and affect the individualistic behaviour of an animal. Hierarchical mixed models would be a more appropriate approach to handle such datasets and infer statistically the influence of different parameters on behaviours. We recommend that the authors take this approach in the analyses of their data.

(2) Brevity in the text. We urge the authors to take advantage of eLife's flexible template and take care to elaborate on the text in the results section, the methods adopted, the legends, and the guides to the legends embedded in the main text. The findings are likely to be of interest to a broad audience, and the writing currently targets the specialist.

Reviewer #2 (Recommendations For The Authors): 

I want to start by saying this seems like a really cool study! It's an impressive amount of work and addressing a pretty basic question that is interesting (at least I think so!)

We thank the reviewer again for this assessment!

That said, I would really strongly recommend the authors embrace using mixed/hierarchical models to analyze their data. They're producing some really impressive data and just doing Pearson correlation coefficients across time points and situations is very clunky and actually losing out on a lot of information. The most up-todate, state-of-the-art are mixed models - these models can handle very complex (or not so complex) random structures which can estimate variance and importantly, covariance, in individual intercepts both over time and across situations. I actually think this could add some really cool insights into the data and allow you to characterize the patterns you're seeing in far more detail. It's datasets exactly like this that are tailormade for these complex variance partitioning models! 

As mentioned before, we have now adopted a more appropriate GLM-based data analysis (see above).

Regardless of which statistical methods you decide to use, please explicitly state in your methods exactly what analyses you did. That is completely lacking now and was a bit frustrating. As such, it's completely unclear whether or how statistical inference was performed. How did you do the behavioral clustering? 

We apologize that these points were not clearly represented in the previous version of the manuscript. We have now significantly extended the methods section to include a separate paragraph on the statistical methods used, in order to address this critique and hope that the revised version is clear now.

Also, I could not for the life of me figure out how many flies had been measured. Were they reused across the situation? Or not?

We reused the same flies across situations whenever possible. However, having one fly experience all assays consecutively was not feasible due to their fragility. Instead, individual flies were exposed to at least 2 of the 3 groups of assays used here: in the Indytrax setup ,  the Buridan arenas and variants thereof, and the virtual arenas Hence, we have compared flies across entirely different setups, but the number of times flies can be retested is limited (as otherwise, sample sizes will drop over time, and the flies will have gone through too many experimental alternations). To make this more clear, we have elaborated on this point in the main text, and we added group sample sizes to figure legends r.

What are these "groups" and "populations" that are referred to in the results (e.g. lines 384, 391, 409)?

We apologize for using these two terms somewhat interchangeably without proper introduction/distinction. We have now made this more clear in at the beginning of the results in the main text, by focusing on the term ‘group’. By ‘group’ we refer to the average of all individuals tested in the same situation. Sample sizes in the figure legends now indicate group/population sizes to make this clearer.

Some of the rationale for the development of the behavioral rigs would have actually been nice to include in the intro, rather than in the results.

This rationale is introduced at the beginning of the last paragraph of the introduction. We hope that this now becomes clear in the revised version of the manuscript.

Reviewer #3 (Recommendations For The Authors): 

This manuscript would do well to take advantage of eLife's flexible word limit. I sense that it has been written in brevity for a different journal but I would urge the authors to revisit this and unpack the language here - in the text, in the figure legends, in references to the figures within the text. The way it's currently written, though not misleading, will only speak to the super-specialist or the super-invested :). But the findings are nice, and it would be nice to tailor it to a broader audience.

We appreciate this suggestion. Initially, we were hoping that we had described our results as clearly and brief as possible. We apologize if that was not always the case. The comments and requests of all three reviewers now led to a series of additions to both main text and methods, leading to a significantly expanded manuscript. We hope that these additons are helpful for the general, non-expert audience.

Author response:

The following is the authors’ response to the original reviews

Overview of changes in the revision

We thank the reviewers for the very helpful comments and have extensively revised the paper. We provide point-by-point responses below and here briefly highlight the major changes:

(1) We expanded the discussion of the relevant literature in children and adults.

(2) We improved the contextualization of our experimental design within previous reinforcement studies in both cognitive and motor domains highlighting the interplay between the two.

(3) We reorganized the primary and supplementary results to better communicate the findings of the studies.

(4) The modeling has been significantly revised and extended. We now formally compare 31 noise-based models and one value-based model and this led to a different model from the original being the preferred model. This has to a large extent cleaned up the modeling results. The preferred model is a special case (with no exploration after success) of the model proposed in Therrien et al. (2018). We also provide examples of individual fits of the model, fit all four tasks and show group fits for all, examine fits vs. data for the clamp phases by age, provide measures of relative and absolute goodness of fit, and examine how the optimal level of exploration varies with motor noise.

Reviewer #1 (Public review):

Summary:

Here the authors address how reinforcement-based sensorimotor adaptation changes throughout development. To address this question, they collected many participants in ages that ranged from small children (3 years old) to adulthood (1 8+ years old). The authors used four experiments to manipulate whether binary and positive reinforcement was provided probabilistically (e.g., 30 or 50%) versus deterministically (e.g., 100%), and continuous (infinite possible locations) versus discrete (binned possible locations) when the probability of reinforcement varied along the span of a large redundant target. The authors found that both movement variability and the extent of adaptation changed with age.

Thank you for reviewing our work. One note of clarification. This work focuses on reinforcementbased learning throughout development but does not evaluate sensorimotor adaptation. The four tasks presented in this work are completed with veridical trajectory feedback (no perturbation).

The goal is to understand how children at different ages adjust their movements in response to reward feedback but does not evaluate sensorimotor adaptation. We now explain this distinction on line 35.

Strengths:

The major strength of the paper is the number of participants collected (n = 385). The authors also answer their primary question, that reinforcement-based sensorimotor adaptation changes throughout development, which was shown by utilizing established experimental designs and computational modelling.

Thank you.

Weaknesses:

Potential concerns involve inconsistent findings with secondary analyses, current assumptions that impact both interpr tation and computational modelling, and a lack of clearly stated hypotheses.

(1) Multiple regression and Mediation Analyses.

The challenge with these secondary analyses is that:

(a) The results are inconsistent between Experiments 1 and 2, and the analysis was not performed for Experiments 3 and 4,

(b) The authors used a two-stage procedure of using multiple regression to determine what variables to use for the mediation analysis, and

(c)The authors already have a trial-by-trial model that is arguably more insightful.

Given this, some suggested changes are to:

(a) Perform the mediation analysis with all the possible variables (i.e., not informed by multiple regression) to see if the results are consistent.

(b) Move the regression/mediation analysis to Supplementary, since it is slightly distracting given current inconsistencies and that the trial-by-trial model is arguably more insightful.

Based on these comments, we have chosen to remove the multiple regression and mediation analyses. We agree that they were distracting and that the trial-by-trial model allows for differentiation of motor noise from exploration variability in the learning block.

(2) Variability for different phases and model assumptions:

A nice feature of the experimental design is the use of success and failure clamps. These clamped phases, along with baseline, are useful because they can provide insights into the partitioning of motor and exploratory noise. Based on the assumptions of the model, the success clamp would only reflect variability due to motor noise (excludes variability due to exploratory noise and any variability due to updates in reach aim). Thus, it is reasonable to expect that the success clamps would have lower variability than the failure clamps (which it obviously does in Figure 6), and presumably baseline (which provides success and failure feedback, thus would contain motor noise and likely some exploratory noise).

However, in Figure 6, one visually observes greater variability during the success clamp (where it is assumed variability only comes from motor noise) compared to baseline (where variability would come from: (a) Motor noise.

(b) Likely some exploratory noise since there were some failures.

(c) Updates in reach aim.

Thanks for this comment. It made us realize that some of our terminology was unintentionally misleading. Reaching to discrete targets in the Baseline block was done to a) determine if participants could move successfully to targets that are the same width as the 100% reward zone in the continuous targets and b) determine if there are age dependent changes in movement precision. We now realize that the term Baseline Variability was misleading and should really be called Baseline Precision.

This is an important distinction that bears on this reviewer's comment. In clamp trials, participants move to continuous targets. In baseline, participants move to discrete targets presented at different locations. Clamp Variability cannot be directly compared to Baseline Precision because they are qualitatively different. Since the target changes on each baseline trial, we would not expect updating of desired reach (the target is the desired reach) and there is therefore no updating of reach based on success or failure. The SD we calculate over baseline trials is the endpoint variability of the reach locations relative to the target centers. In success clamp, there are no targets so the task is qualitatively different.

We have updated the text to clarify terminology, expand upon our operational definitions, and motivate the distinct role of the baseline block in our task paradigm (line 674).

Given the comment above, can the authors please:

(a) Statistically compare movement variability between the baseline, success clamp, and failure clamp phases.

Given our explanation in the previous point we don't think that comparing baseline to the clamp makes sense as the trials are qualitatively different.

(b) The authors have examined how their model predicts variability during success clamps and failure clamps, but can they also please show predictions for baseline (similar to that of Cashaback et al., 2019; Supplementary B, which alternatively used a no feedback baseline)?

Again, we do not think it makes sense to predict the baseline which as we mention above has discrete targets compared to the continuous targets in the learning phase.

(c) Can the authors show whether participants updated their aim towards their last successful reach during the success clamp? This would be a particularly insightful analysis of model assumptions.

We have now compared 31 models (see full details in next response) which include the 7 models in Roth et al. (2023). Several of these model variants have updating even after success with so called planning noise). We also now fit the model to the data that includes the clamp phases (we can't easily fit to success clamp alone as there are only 10 trials). We find that the preferred model is one that does not include updating after success.

(d) Different sources of movement variability have been proposed in the literature, as have different related models. One possibility is that the nervous system has knowledge of 'planned (noise)' movement variability that is always present, irrespective of success (van Beers, R.J. (2009). Motor learning is optimally tuned to the properties of motor noise. Neuron, 63(3), 406-417). The authors have used slightly different variations of their model in the past. Roth et al (2023) directly Rill compared several different plausible models with various combinations of motor, planned, and exploratory noise (Roth A, 2023, "Reinforcement-based processes actively regulate motor exploration along redundant solution manifolds." Proceedings of the Royal Society B 290: 20231475: see Supplemental). Their best-fit model seems similar to the one the authors propose here, but the current paper has the added benefit of the success and failure clamps to tease the different potential models apart. In light of the results of a), b), and c), the authors are encouraged to provide a paragraph on how their model relates to the various sources of movement variability and ther models proposed in the literature.

Thank you for this. We realized that the models presented in Roth et al. (2023) as well as in other papers, are all special cases of a more general model. Moreover, in total there are 30 possible variants of the full model so we have now fit all 31 models to our larger datasets and performed model selection (Results and Methods). All the models can be efficiently fit by Kalman smoother to the actual data (rather than to summary statistics which has sometimes been done). For model selection, we fit only the 100 learning trials and chose the preferred model based on BIC on the children's data (Figure 5—figure Supplement 1). After selecting the preferred model we then refit this model to all trials including the clamps so as to obtain the best parameter estimates.

The preferred model was the same whether we combined the continuous and discrete probabilistic data or just examin d each task separately either for only the children or for the children and adults combined. The preferred model is a pecial case (no exploration after success) of the one proposed in Therrien et al. (2018) and has exploration variability (after failure) and motor noise with full updating with exploration variability (if any) after success. This model differs from the model in the original submission which included a partial update of the desired reach after exploration this was considered the learning rate. The current model suggests a unity learning rate.

In addition, as suggested by another reviewer, we also fit a value-based model which we adapted from the model described in Giron et al. (2023). This model was not preferred.

We have added a paragraph to the Discussion highlighting different sources of variability and links to our model comparison.

(e) line 155. Why would the success clamp be composed of both motor and exploratory noise? Please clarify in the text

This sentence was written to refer to clamps in general and not just success clamps. However, in the revision this sentence seemed unnecessary so we have removed it.

(3) Hypotheses:

The introduction did not have any hypotheses of development and reinforcement, despite the discussion above setting up potential hypotheses. Did the authors have any hypotheses related to why they might expect age to change motor noise, exploratory noise, and learning rates? If so, what would the experimental behaviour look like to confirm these hypotheses? Currently, the manuscript reads more as an exploratory study, which is certainly fine if true, it should just be explicitly stated in the introduction. Note: on line 144, this is a prediction, not a hypothesis. Line 225: this idea could be sharpened. I believe the authors are speaking to the idea of having more explicit knowledge of action-target pairings changing behaviour.

We have included our hypotheses and predictions at two points in the paper In the introduction we modified the text to:

"We hypothesized that children's reinforcement learning abilities would improve with age, and depend on the developmental trajectory of exploration variability, learning rate (how much people adjust their reach after success), and motor noise (here defined as all sources of noise associated with movement, including sensory noise, memory noise, and motor noise). We think that these factors depend on the developmental progression of neural circuits that contribute to reinforcement learning abilities (Raznahan et al., 2014; Nelson et al., 2000; Schultz, 1998)."

In results we modified the sentence to:

"We predicted that discrete targets could increase exploration by encouraging children to move to a different target after failure.”

Reviewer #2 (Public review):

Summary:

In this study, Hill and colleagues use a novel reinforcement-based motor learning task ("RML"), asking how aspects of RML change over the course of development from toddler years through adolescence. Multiple versions of the RML task were used in different samples, which varied on two dimensions: whether the reward probability of a given hand movement direction was deterministic or probabilistic, and whether the solution space had continuous reach targets or discrete reach targets. Using analyses of both raw behavioral data and model fits, the authors report four main results: First, developmental improvements reflected 3 clear changes, including increases in exploration, an increase in the RL learning rate, and a reduction of intrinsic motor noise. Second, changes to the task that made it discrete and/or deterministic both rescued performance in the youngest age groups, suggesting that observed deficits could be linked to continuous/probabilistic learning settings. Overall, the results shed light on how RML changes throughout human development, and the modeling characterizes the specific learning deficits seen in the youngest ages.

Strengths:

(1) This impressive work addresses an understudied subfield of motor control/psychology - the developmental trajectory of motor learning. It is thus timely and will interest many researchers.

(2) The task, analysis, and modeling methods are very strong. The empirical findings are rather clear and compelling, and the analysis approaches are convincing. Thus, at the empirical level, this study has very few weaknesses.

(3) The large sample sizes and in-lab replications further reflect the laudable rigor of the study.

(4) The main and supplemental figures are clear and concise.

Thank you.

Weaknesses:

(1) Framing.

One weakness of the current paper is the framing, namely w/r/t what can be considered "cognitive" versus "non-cognitive" ("procedural?") here. In the Intro, for example, it is stated that there are specific features of RML tasks that deviate from cognitive tasks. This is of course true in terms of having a continuous choice space and motor noise, but spatially correlated reward functions are not a unique feature of motor learning (see e.g. Giron et al., 2023, NHB). Given the result here that simplifying the spatial memory demands of the task greatly improved learning for the youngest cohort, it is hard to say whether the task is truly getting at a motor learning process or more generic cognitive capacities for spatial learning, working memory, and hypothesis testing. This is not a logical problem with the design, as spatial reasoning and working memory are intrinsically tied to motor learning. However, I think the framing of the study could be revised to focus in on what the authors truly think is motor about the task versus more general psychological mechanisms. Indeed, it may be the case that deficits in motor learning in young children are mostly about cognitive factors, which is still an interesting result!

Thank you for these comments on the framing of our study. We now clearly acknowledge that all motor tasks have cognitive components (new paragraph at line 65). We also explain why we think our tasks has features not present in typical cognitive tasks.

(2) Links to other scholarship.

If I'm not mistaken a common observation in tudies of the development of reinforcement learning is a decrease in exploration over-development (e.g., Nussenbaum and Hartley, 2019; Giron et al., 2023; Schulz et al., 2019); this contrasts with the current results which instead show an increase. It would be nice to see a more direct discussion of previous findings showing decreases in exploration over development, and why the current study deviates from that. It could also be useful for the authors to bring in concepts of different types of exploration (e.g. "directed" vs "random"), in their interpretations and potentially in their modeling.

We recognize that our results differ from prior work. The optimal exploration pattern differs from task to task. We now discuss that exploration is not one size fits all, it's benefits vary depending upon the task. We have added the following paragraphs to the Discussion section:

"One major finding from this study is that exploration variability increases with age. Some other studies of development have shown that exploration can decrease with age indicating that adults explore less compared to children (Schulz et al., 2019; Meder et al., 2021; Giron et al., 2023). We believe the divergence between our work and these previous findings is largely due to the experimental design of our study and the role of motor noise. In the paradigm used initially by Schulz et al. (2019) and replicated in different age groups by Meder et al. (2021) and Giron et al. (2023), participants push buttons on a two-dimensional grid to reveal continuous-valued rewards that are spatially correlated. Participants are unaware that there is a maximum reward available and therefore children may continue to explore to reduce uncertainty if they have difficulty evaluating whether they have reached a maxima. In our task by contrast, participants are given binary reward and told that there is a region in which reaches will always be rewarded. Motor noise is an additional factor which plays a key role in our reaching task but minimal if any role in the discretized grid task. As we show in simulations of our task, as motor noise goes down (as it is known to do through development) the optimal amount of exploration goes up (see Figure 7—figure Supplement 2 and Appendix 1). Therefore, the behavior of our participants is rational in terms of R230 increasing exploration as motor noise decreases.

A key result in our study is that exploration in our task reflects sensitivity to failure. Older children make larger adjustments after failure compared to younger children to find the highly rewarded zone more quickly. Dhawale et al. (2017) discuss the different contexts in which a participant may explore versus exploit (i.e., stick at the same position). Exploration is beneficial when reward is low as this indicates that the current solution is no longer ideal, and the participant should search for a better solution. Konrad et al. (2025) have recently shown this behavior in a real-world throwing task where 6 to 12 year old children increased throwing variability after missed trials and minimized variability after successful trials. This has also been shown in a postural motor control task where participants were more variable after non-rewarded trials compared to rewarded trials (Van Mastrigt et al., 2020). In general, these studies suggest that the optimal amount of exploration is dependent on the specifics of the task."

(3) Modeling.

First, I may have missed something, but it is unclear to me if the model is actually accounting for the gradient of rewards (e.g., if I get a probabilistic reward moving at 45°, but then don't get one at 40°, I should be more likely to try 50° next then 35°). I couldn't tell from the current equations if this was the case, or if exploration was essentially "unsigned," nor if the multiple-trials-back regression analysis would truly capture signed behavior. If the model is sensitive to the gradient, it would be nice if this was more clear in the Methods. If not, it would be interesting to have a model that does "function approximation" of the task space, and see if that improves the fit or explains developmental changes.

The model we use (similar to Roth et al. (2023) and Therrien et al. (2016, 2018)) does not model the gradient. Exploration is always zero-mean Gaussian. As suggested by the reviewer, we now also fit a value-based model (described starting at line 810) which we adapted from the model presented in Giron et al. (2023). We show that the exploration and noise-based model is preferred over the value-based model.

The multiple-trials-back regression was unsigned as the intent was to look at the magnitude and not the direction of the change in movement. We have decided to remove this analysis from the manuscript as it was a source of confusion and secondary analysis that did not add substantially to the findings of these studies.

Second, I am curious if the current modeling approach could incorporate a kind of "action hysteresis" (aka perseveration), such that regardless of previous outcomes, the same action is biased to be repeated (or, based on parameter settings, avoided).

In some sense, the learning rate in the model in the original submission is highly related to perseveration. For example if the learning rate is 0, then there is complete perseveration as you simply repeat the same desired movement. If the rate is 1, there is no perseveration and values in between reflect different amounts of perseveration. Therefore, it is not easy to separate learning rate from perseveration. Adding perseveration as another parameter would likely make it and the learning unidentifiable. However, we now compare 31 models and those that have a non-unity learning rate are not preferred suggesting there is little perseveration.

(4) Psychological mechanisms. There is a line of work that shows that when children and adults perform RL tasks they use a combination of working memory and trial-by-trial incremental learning processes (e.g., Master et al., 2020; Collins and Frank 2012). Thus, the observed increase in the learning rate over development could in theory reflect improvements in instrumental learning, working memory, or both. Could it be that older participants are better at remembering their recent movements in short-term memory (Hadjiosif et al., 2023; Hillman et al., 2024)?

We agree that cognitive processes, such as working memory or visuospatial processing, play a role in our task and describe cognitive elements of our task in the introduction (new paragraph at line 65). However, the sensorimotor model we fit to the data does a good job of explaining the variation across age, which suggests that that age-dependent cognitive processes probably play a smaller role.

Reviewer #3 (Public review):

Summary:

The study investigates reinforcement learning across the lifespan with a large sample of participants recruited for an online game. It finds that children gradually develop their abilities to learn reward probability, possibly hindered by their immature spatial processing and probabilistic reasoning abilities. Motor noise, reinforcement learning rate, and exploration after a failure all contribute to children's subpar performance.

Strengths:

(1) The paradigm is novel because it requires continuous movement to indicate people's choices, as opposed to discrete actions in previous studies.

(2) A large sample of participants were recruited.

(3) The model-based analysis provides further insights into the development of reinforcement learning ability.

Thank you.

Weaknesses:

(1 ) The adequacy of model-based analysis is questionable, given the current presentation and some inconsistency in the results.

Thank you for raising this concern. We have substantially revised the model from our first submission. We now compare 31 noise-based models and 1 value-based model and fit all of the tasks with the preferred model. We perform model selection using the two tasks with the largest datasets to identify the preferred model. From the preferred model, we found the parameter fits for each individual dataset and simulated the trial by trial behavior allowing comparison between all four tasks. We now show examples of individual fits as well as provide a measure of goodness of fit. The expansion of our modeling approach has resolved inconsistencies and sharpened the conclusions drawn from our model.

(2) The task should not be labeled as reinforcement motor learning, as it is not about learning a motor skill or adapting to sensorimotor perturbations. It is a classical reinforcement learning paradigm.

We now make it clear that our reinforcement learning task has both motor and cognitive demands, but does not fall entirely within one of these domains. We use the term motor learning because it captures the fact that participants maximize reward by making different movements, corrupted by motor noise, to unmarked locations on a continuous target zone. When we look at previous ublications, it is clear that our task is similar to those that also refer to this as reinforcement motor learning Cashaback et al. (2019) (reaching task using a robotic arm in adults), Van Mastrigt et al. (2020) (weight shifting task in adults), and Konrad et al. (2025) (real-world throwing task in children). All of these tasks involve trial-by-trial learning through reinforcement to make the movement that is most effective for a given situation. We feel it is important to link our work to these previous studies and prefer to preserve the terminology of reinforcement motor learning.

Recommendations for the authors:

Reviewing Editor Comments:

Thank you for this summary. Rather than repeat the extended text from the responses to the reviewers here, we point the Editor to the appropriate reviewer responses for each issue raised.

The reviewers and editors have rated the significance of the findings in your manuscript as "Valuable" and the strength of evidence as "Solid" (see eLife evalutation). A consultancy discussion session to integrate the public reviews and recommendations per reviewer (listed below), has resulted in key recommendations for increasing the significance and strength of evidence:

To increase the Significance of the findings, please consider the following:

(1) Address and reframe the paper around whether the task is truly getting at a motor learning process or more generic cognitive decision-making capacities such as spatial memory, reward processing, and hypothesis testing.

We have revised the paper to address the comments on the framing of our work. Please see responses to the public review comments of Reviewers #2 and #3.

(2) It would be beneficial to specify the differences between traditional reinforcement algorithms (i.e., using softmax functions to explore, and build representations of state-action-reward) and the reinforcement learning models used here (i.e., explore with movement variability, update reach aim towards the last successful action), and compare present findings to previous cognitive reinforcement learning studies in children.

Please see response to the public review comments of Reviewer #1 in which we explain the expansion of our modeling approach to fit a value-based model as well as 31 other noise-based models. In our response to the public review comments of Reviewer #2, we comment on our expanded discussion of how our findings compare with previous cognitive reinforcement learning studies.

To move the "Strength of Evidence" to "Convincing", please consider doing the following:

(1 ) Address some apparently inconsistent and unrealistic values of motor noise, exploration noise, and learning rate shown for individual participants (e.g., Figure 5b; see comments reviewers 1 and take the following additional steps: plotting r squares for individual participants, discussing whether individual values of the fitted parameters are plausible and whether model parameters in each age group can extrapolate to the two clamp conditions and baselines.

We have substantially updated our modeling approach. Now that we compare 31 noise-based models, the preferred model does not show any inconsistent or unrealistic values (see response to Reviewer #3). Additionally, we now show example individual fits and provide both relative and absolute goodness of fit (see response to Reviewer #3).

(2) Relatedly, to further justify if model assumptions are met, it would be valuable to show that the current learning model fits the data better than alternative models presented in the literature by the authors themselves and by others (reviewer 1). This could include alternative development models that formalise the proposed explanations for age-related change: poor spatial memory, reward/outcome processing, and exploration strategies (reviewer 2).

Please see response to public review comments of Reviewer #1 in which we explain that we have now fit a value-based model as well as 31 other noise-based models providing a comparison of previous models as well as novel models. This led to a slightly different model being preferred over the model in the original submission (updated model has a learning rate of unity). These models span many of the processes previously proposed for such tasks. We feel that 32 models span a reasonable amount of space and do not believe we have the power to include memory issues or heuristic exploration strategies in the model.

(3) Perform the mediation analysis with all the possible variables (i.e., not informed by multiple regression) to see if the results are more consistent across studies and with the current approach (see comments reviewer 1).

Please see response to public review comments of Reviewer #1. We chose to focus only on the model based analysis because it allowed us to distinguish between exploration variability and motor noise.

Please see below for further specific recommendations from each reviewer.

Reviewer #1 (Recommendations for the author):

(1) In general, there should be more discussion and contextualization of other binary reinforcement tasks used in the motor literature. For example, work from Jeroen Smeets, Katinka van der Kooij, and Joseph Galea.

Thank you for this comment. We have edited the Introduction to better contextualize our work within the reinforcement motor learning literature (see line 67 and line 83).

(2) Line 32. Very minor. This sentence is fine, but perhaps could be slightly improved. “select a location along a continuous and infinite set of possible options (anywhere along the span of the bridge)"

Thank you for this comment. We have edited the sentence to reflect this suggestion.

(3) Line 57. To avoid some confusion in successive sentences: Perhaps, "Both children over 12 and adolescents...".

Thank you for this comment. We have edited the sentence to reflect this suggestion.

(4) Line 80. This is arguably not a mechanistic model, since it is likely not capturing the reward/reinforcement machinery used by the nervous system, such as updating the expected value using reward predic tion errors/dopamine. That said, this phenomenological model, and other similar models in the field, do very well to capture behaviour with a very simple set of explore and update rules.

We use mechanistic in the standard use in modeling, as in Levenstein et al. (2023), for example. The contrast is not with neural modeling, but with normative modeling, in which one develops a model to optimize a function (or descriptive models as to what a system is trying to achieve). In mechanistic modeling one proposes a mechanism and this can be at a state-space level (as in our case) or a neural level (as suggested my the reviewer) but both are considered mechanistic, just at different levels. Quoting Levenstein "... mechanistic models, in which complex processes are summarized in schematic or conceptual structures that represent general properties of components and their interactions, are also commonly used." We now reference the Levenstein paper to clarify what we mean by mechanistic.

(5) Figure 1. It would be useful to state that the x-axis in Figure 1 is in normalized units, depending on the device.

Thank you for this comment. We have added a description of the x-axis units to the Fig. 1 caption.

(6) Were there differences in behaviour for these different devices? e.g., how different was motor noise for the mouse, trackpad, and touchscreen?

Thank you for this question. We did not find a significant effect of device on learning or precision in the baseline block. We have added these one way ANOVA results for each task in Supplementary Table 1.

(7) Line 98. Please state that participants received reinforcement feedback during baseline.

Thank you for this comment. We have updated the text to specify that participants receive reward feedback during the baseline block.

(8) Line 99. Did the distance from the last baseline trial influence whether the participant learned or did not learn? For example, would it place them too far from the peak success location such that it impacted learning?

Thank you for this question. We looked at whether the position of movement on the last baseline block trial was correlated with the first movement position in the learning block. We did not find any correlations between these positions for any of the tasks. Interestingly, we found that the majority of participants move to the center of the workspace on the first trial of the learning block for all tasks (either in the presence of the novel continuous target scene or the presentation of 7 targets all at once). We do not think that the last movement in the baseline block "primed" the participant for the location of the success zone in the learning block. We have added the following sentence to the Results section:

"Note that the reach location for the first learning trial was not affected by (correlated with) the target position on the last baseline trial (p > 0.3 for both children and adults, separately)."

(9) The term learning distance could be improved. Perhaps use distance from target.

Thank you for this comment. We appreciate that learning distance defined with 0 as the best value is counter intuitive. We have changed the language to be "distance from target" as the learning metric.

(10) Line 188. This equation is correct, but to estimate what the standard deviation by the distribution of changes in reach position is more involved. Not sure if the authors carried out this full procedure, which is described in Cashaback et al., 2019; Supplemental 2.

There appear to be no Supplemental 2 in the referenced paper so we assume the reviewer is referring to Supplemental B which deals with a shuffling procedure to examine lag-1 correlations.

In our tasks, we are limited to only 9 trials to analyze in each clamp phase so do not feel a shuffling analysis is warranted. In these blocks, we are not trying to 'estimate what the standard deviation by the distribution of changes in reach position' but instead are calculating the standard deviation of the reach locations and comparing the model fit (for which the reviewer says the formula is correct) with the data. We are unclear what additional steps the reviewer is suggesting. In our updated model analysis, we fit the data including the clamp phases for better parameter estimation. We use simulations to estimate s.d. in the clamp phase (as we ensure in simulations the data does not fall outside the workspace) making the previous analytic formulas an approximation that are no longer used.

(11) Line 197-199. Having done the demo task, it is somewhat surprising that a 3-year-old could understand these instructions (whose comprehension can be very different from even a 5-year old).

Thank you for raising this concern. We recognize that the younger participants likely have different comprehension levels compared to older participants. However, we believe that the majority of even the youngest participants were able to sufficiently understand the goal of the task to move in a way to get the video clip to play. We intentionally designed the tasks to be simple such that the only instructions the child needed to understand were that the goal was to get the video clip to play as much as possible and the video clip played based on their movement. Though the majority of younger children struggled to learn well on the probabilistic tasks, they were able to learn well on the deterministic tasks where the task instructions were virtually identical with the exception of how many places in the workspace could gain reward. On the continuous probabilistic task, we did have a small number (n = 3) of 3 to 5 year olds who exhibited more mature learning ability which gives us confidence that the younger children were able to understand the task goal.

(12) Line 497: Can the authors please report the F-score and p-value separately for each of these one-way ANOVA (the device is of particular interest here).

Thank you for this request. We have added ina upplementarytable (Supplementary Table 1) with the results of these ANOVAs.

(13) Past work has discussed how motivation influences learning, which is a function of success rate (van der Kooij, K., in 't Veld, L., & Hennink, T. (2021). Motivation as a function of success frequency. Motivation and Emotion, 45, 759-768.). Can the authors please discuss how that may change throughout development?

Thank you for this comment. While motivation most probably plays a role in learning, in particular in a game environment, this was out of the scope of the direct focus of this work and not something that our studies were designed to test. We have added the following sentence to the discussion section to address this comment:

"We also recognize that other processes, such as memory and motivation, could affect performance on these tasks however our study was not designed to test these processes directly and future work would benefit from exploring these other components more explicitly."

(14) Supplement 6. This analysis is somewhat incomplete because it does not consider success.

Pekny and collegues (2015) looked at 3 trials back but considered both success and reward. However, their analysis has issues since successive time points are not i.i.d., and spurious relationships can arise. This issue is brought up by Dwahale (Dhawale, A. K., Miyamoto, Y. R., Smith, M. A., & R475 Ölveczky, B. P. (2019). Adaptive regulation of motor variability. Current Biology, 29(21), 3551-3562.). Perhaps it is best to remove this analysis from the paper.

Thank you for this comment. We have decided to remove this secondary analysis from the paper as it was a source of confusion and did not add to the understanding and interpretation of our behavioral results.

Reviewer #2 (Recommendations for the author):

(1 ) the path length ratio analyses in the supplemental are interesting but are not mentioned in the main paper. I think it would be helpful to mention these as they are somewhat dramatic effects

Thank you for this comment. Path length ratios are defined in the Methods and results are briefly summarized in the Results section with a point to the supplementary figures. We have updated the text to more explicitly report the age related differences in path length ratios.

(2) The second to last paragraph of the intro could use a sentence motivating the use ofthe different task features (deterministic/probabilistic and discrete/continuous).

Thank you for this comment. We have added an additional motivating sentence to the introduction.

Reviewer #3 (Recommendations for the author):

The paper labeled the task as one for reinforcement motor learning, which is not quite appropriate in my opinion. Motor learning typically refers to either skill learning or motor adaptation, the former for improving speed-accuracy tradeoffs in a certain (often new) motor skill task and the latter for accommodating some sensorimotor perturbations for an existing motor skill task. The gaming task here is for neither. It is more like a

decision-making task with a slight contribution to motor execution, i.e., motor noise. I would recommend the authors label the learning as reinforcement learning instead of reinforcement motor learning.

Thank you for this comment. As noted in the response to the public review comments, we agree that this task has components of classical reinforcement learning (i.e. responding to a binary reward) but we specifically designed it to require the learning of a movement within a novel game environment. We have added a new paragraph to the introduction where we acknowledge the interplay between cognitive and motor mechanisms while also underscoring the features in our task that we think are not present in typical cognitive tasks.

My major concern is whether the model adequately captures subjects' behavior and whether we can conclude with confidence from model fitting. Motor noise, exploration noise, and learning rate, which fit individual learning patterns (Figure 5b), show some quite unrealistic values. For example, some subjects have nearly zero motor noise and a 100% learning rate.

We have now compared 31 models and the preferred model is different from the one in the first submission. The parameter fits of the new model do not saturate in any way and appear reasonable to us. The updates to the model analysis have addressed the concern of previously seen unrealistic values in the prior draft.

Currently, the paper does not report the fitting quality for individual subjects. It is good to have an exemplary subject's fit shown, too. My guess is that the r-squared would be quite low for this type of data. Still, given that the children's data is noisier, it might be good to use the adult data to show how good the fitting can be (individual fits, r squares, whether the fitted parameters make sense, whether it can extrapolate to the two clamp phases). Indeed, the reliability of model fitting affects how we should view the age effect of these model parameters.

We now show fits to individual subjects. But since this is a Kalman smoother it fits the data perfectly by generating its best estimate of motor noise and exploration variability on each trial to fully account for the data — so in that sense R² is always 1 so that is not helpful.

While the BIC analysis with the other model variants provides a relative goodness of fit, it is not straightforward to provide an absolute goodness of fit such as standard R² for a feedforward simulation of the model given the parameters (rather than the output of the Kalman smoother). There are two problems. First, there is no single model output. Each time the model is simulated with the fit parameters it produces a different output (due to motor noise, exploration variability and reward stochasticity). Second, the model is not meant to reproduce the actual motor noise, exploration variability and reward stochasticity of a trial. For example, the model could fit pure Gaussian motor noise across trials (for a poor learner) by accurately fitting the standard deviation of motor noise but would not be expected to actually match each data point so would have a traditional R² of O.

To provide an overall goodness of fit we have to reduce the noise component and to do so we exam ined the traditional R² between the average of all the children's data and the average simulation of the model (from the median of 1000 simulations per participant) so as to reduce the stochastic variation. The results for the continuous probabilistic and discrete probabilistic task are R² of 0.41 and 0.72, respectively.

Not that variability in the "success clamp" doe not change across ages (Figure 4C) and does not contribute to the learning effect (Figure 4F). However, it is regarded as reflecting motor noise (Figure SC), which then decreases over age from the model fitting (Figure 5B). How do we reconcile these contradictions? Again, this calls the model fitting into question.

For the success clamp, we only have 9 trials to calculate variability which limits our power to detect significance with age. In contrast, the model uses all 120 trials to estimate motor noise. There is a downward trend with age in the behavioral data which we now show overlaid on the fits of the model for both probabilistic conditions (Figure 5—figure Supplement 4) and Figure 6—figure Supplement 4). These show a reasonable match and although the variance explained is 1 6 and 56% (we limit to 9 trials so as to match the fail clamp), the correlations are 0.52 and 0.78 suggesting we have reasonable relation although there may be other small sources of variability not captured in the model.

Figure 5C: it appears one bivariate outlier contributes a lot to the overall significant correlation here for the "success clamp".

Recalculating after removing that point in original Fig 5C was still significant and we feel the plots mentioned in the previous point add useful information to this issue. With the new model this figure has changed.

It is still a concern that the young children did not understand the instructions. Nine 3-to-8 children (out of 48) were better explained by the noisy only model than the full model. In contrast, ten of the rest of the participants (out of 98) were better explained by the noisy-only model. It appears that there is a higher percentage of the "young" children who didn't get the instruction than the older ones.

Thank you for this comment. We did take participant comprehension of the task into consideration during the task design. We specifically designed it so that the instructions were simple and straight forward. The child simply needs to understand the underlying goal to make the video clip play as often as possible and that they must move the penguin to certain positions to get it to play. By having a very simple task goal, we are able to test a naturalistic response to reinforcement in the absence of an explicit strategy in a task suited even for young children.

We used the updated reinforcement learning model to assess whether an individual's performance is consistent with understanding the task. In the case of a child who does not understand the task, we expect that they simply have motor noise on their reach, and crucially, that they would not explore more after failure, nor update their reach after success. Therefore, we used a likelihood ratio test to examine whether the preferred model was significantly better at explaining each participant's data compared to the model variant which had only motor noise (Model 1). Focusing on only the youngest children (age 3-5), this analysis showed that that 43, 59, 65 and 86% of children (out of N = 21, 22, 20 and 21 ) for the continuous probabilistic, discrete probabilistic, continuous deterministic, and discrete deterministic conditions, respectively, were better fit with the preferred model, indicating non-zero exploration after failure. In the 3-5 year old group for the discrete deterministic condition, 18 out of 21 had performance better fit by the preferred model, suggesting this age group understands the basic task of moving in different directions to find a rewarding location.

The reduced numbers fit by the preferred model for the other conditions likely reflects differences in the task conditions (continuous and/or probabilistic) rather than a lack of understanding of the goal of the task. We include this analysis as a new subsection at the end of the Results.

Supplementary Figure 2: the first panel should belong to a 3-year-old not a 5-year-old? How are these panels organized? This is kind of confusing.

Thank you for this comment. Figure 2—figure Supplement 1 and Figure 2—figure Supplement 2 are arranged with devices in the columns and a sample from each age bin in the rows. For example in Figure 2—figure Supplement 1, column 1, row 1 is a mouse using participant age 3 to 5 years old while column 3, row 2 is a touch screen using participant age 6 to 8 years old. We have edited the labeling on both figures to make the arrangement of the data more clear.

Line 222: make this a complete sentence.

This sentence has been edited to a complete sentence.

Line 331: grammar.

This sentence has been edited for grammar.

Author response:

The following is the authors’ response to the original reviews

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

This article investigates the phenotype of macrophages with a pathogenic role in arthritis, particularly focusing on arthritis induced by immune checkpoint inhibitor (ICI) therapy. 

Building on prior data from monocyte-macrophage coculture with fibroblasts, the authors hypothesized a unique role for the combined actions of prostaglandin PGE2 and TNF. The authors studied this combined state using an in vitro model with macrophages derived from monocytes of healthy donors. They complemented this with single-cell transcriptomic and epigenetic data from patients with ICI-RA, specifically, macrophages sorted out of synovial fluid and tissue samples. The study addressed critical questions regarding the regulation of PGE2 and TNF: Are their actions co-regulated or antagonistic? How do they interact with IFN-γ in shaping macrophage responses? 

This study is the first to specifically investigate a macrophage subset responsive to the PGE2 and TNF combination in the context of ICI-RA, describes a new and easily reproducible in vitro model, and studies the role of IFNgamma regulation of this particular Mф subset. 

Strengths: 

Methodological quality: The authors employed a robust combination of approaches, including validation of bulk RNA-seq findings through complementary methods. The methods description is excellent and allows for reproducible research. Importantly, the authors compared their in vitro model with ex vivo single-cell data, demonstrating that their model accurately reflects the molecular mechanisms driving the pathogenicity of this macrophage subset. 

Weaknesses: 

Introduction: The introduction lacks a paragraph providing an overview of ICI-induced arthritis pathogenesis and a comparison with other types of arthritis. Including this would help contextualize the study for a broader audience.

Thank you for this suggestion, we have added a paragraph on ICI-arthritis to intro (pg. 4, middle paragraph).  

Results Section: At the beginning of the results section, the experimental setup should be described in greater detail to make an easier transition into the results for the reader, rather than relying just on references to Figure 1 captions.

We have clarified the experimental setup (pg. 5).  

There is insufficient comparison between single-cell RNA-seq data from ICI-induced arthritis and previously published single-cell RA datasets. Such a comparison may include DEGs and GSEA, pathway analysis comparison for similar subsets of cells. Ideally, an integration with previous datasets with RA-tissue-derived primary monocytes would allow for a direct comparison of subsets and their transcriptomic features.

We thank the Reviewer for this suggestion, which has increased the impact of our data and analysis. A computationally rigorous representation mapping approach showed that ICI-arthritis myeloid subsets predominantly mapped onto 4 previously defined RA subsets including IL-1β+ cells. This result was corroborated using a complementary data integration approach. Analysis of (TNF + PGE)-induced gene sets (TP signatures) in ICI-arthritis myeloid cells projected onto the RA subsets using the AUCell package showed elevated TP gene expression in similar ICI-arthritis and RA monocytic cell subsets. We also found mutually exclusive expression of TP and IFN signatures in distinct RA and ICI-arthritis myeloid cell subsets, which supports that the opposing cross-regulation between IFN-γ and PGE2 pathways that we identified in vitro also functions similarly in vivo. This analysis is shown in the new Fig. 3, described on pg. 7, and discussed on pp. 13-14.

While it's understandable that arthritis samples are limited in numbers and myeloid cell numbers, it would still be interesting to see the results of PGE2+TNF in vitro stimulation on the primary RA or ICI-RA macrophages. It would be valuable to see RNA-Seq signatures of patient cell reactivation in comparison to primary stimulation of healthy donor-derived monocytes.

We agree that this would be interesting but given limited samples and distribution of samples amongst many studies and investigators this is beyond the scope of the current study.  

Discussion: Prior single-cell studies of RA and RA macrophage subpopulations from 2019, 2020, 2023 publications deserve more discussion. A thorough comparison with these datasets would place the study in a broader scientific context. 

Creating an integrated RA myeloid cell atlas that combines ICI-RA data into the RA landscape would be ideal to add value to the field. 

As one of the next research goals, TNF blockade data in RA and ICI-RA patients would be interesting to add to such an integrated atlas. Combining responders and non-responders to TNF blockade would help to understand patient stratification with the myeloid pathogenic phenotypes. It would be great to read the authors' opinion on this in the Discussion section. 

Please see our response to point 3 above. This point is addressed in Fig. 3, pg. 7, and pp. 13-14, which includes a discussion of responders and nonresponders and patient stratification.  

Conclusion: The authors demonstrated that while PGE2 maintains the inflammatory profile of macrophages, it also induces a distinct phenotype in simultaneous PGE2 and TNF treatment. The study of this specific subset in single-cell data from ICI-RA patients sheds light on the pathogenic mechanisms underlying this condition, however, how it compares with conventional RA is not clear from the manuscript. 

Given the substantial incidence of ICI-induced autoimmune arthritis, understanding the unique macrophage subsets involved for future targeting them therapeutically is an important challenge. The findings are significant for immunologists, cancer researchers, and specialists in autoimmune diseases, making the study relevant to a broad scientific audience. 

Reviewer #2 (Public review): 

Summary/Significance of the findings: 

The authors have done a great job by extensively carrying out transcriptomic and epigenomic analyses in the primary human/mouse monocytes/macrophages to investigate TNF-PGE2 (TP) crosstalk and their regulation by IFN-γ in the Rheumatoid arthritis (RA) synovial macrophages. They proposed that TP induces inflammatory genes via a novel regulatory axis whereby IFN-γ and PGE2 oppose each other to determine the balance between two distinct TNF-induced inflammatory gene expression programs relevant to RA and ICI-arthritis. 

Strengths: 

The authors have done a great job on RT-qPCR analysis of gene expression in primary human monocytes stimulated with TNF and showing the selective agonists of PGE2 receptors EP2 and EP4 22 that signal predominantly via cAMP. They have beautifully shown IFN-γ opposes the effects of PGE2 on TNF-induced gene expression. They found that TP signature genes are activated by cooperation of PGE2-induced AP-1, CEBP, and NR4A with TNF-induced NF-κB activity. On the other hand, they found that IFN-γ suppressed induction of AP-1, CEBP, and NR4A activity to ablate induction of IL-1, Notch, and neutrophil chemokine genes but promoted expression of distinct inflammatory genes such as TNF and T cell chemokines like CXCL10 indicating that TP induces inflammatory genes via IFN-γ in the RA and ICI-arthritis. 

Weaknesses: 

(1) The authors carried out most of the assays in the monocytes/macrophages. How do APCcells like Dendritic cells behave with respect to this TP treatment similar dosing? 

We agree that this is an interesting topic especially as TNF + PGE2 is one of the standard methods of maturing in vitro generated human DCs and promoting antigen-presenting function. As DC maturation is quite different from monocyte activation this would represent a new study and is beyond the scope of the current manuscript. We have instead added a paragraph to the discussion (pg. 12) and cited the literature on DC maturation by TNF + PGE2 including one of our older papers (PMID: 18678606; 2008)  

(2) The authors studied 3h and 24h post-treatment transcriptomic and epigenomic. What happens to TP induce inflammatory genes post-treatment 12h, 36h, 48h, 72h. It is critical to see the upregulated/downregulated genes get normalised or stay the same throughout the innate immune response.

We now clarify that subsets of inducible genes showed distinct kinetics of induction with transient expression at 3 hr versus sustained expression over the 24 hr stimulation period as shown in Supplementary Fig. 1 (pg. 5).

(3) The authors showed IL1-axis in response to the TP-treatment. Do other cytokine axes get modulated? If yes, then how do they cooperate to reduce/induce inflammatory responses along this proposed axis?

This is an interesting question, which we approached using a combination of pathway analysis and targeted inspection of pathways important pathogenesis of RA, which is the inflammatory condition most relevant for this study. In addition to genes in the IL-1-NF-κB core inflammatory pathway, pathway analysis of genes induced by TP co-stimulation showed enrichment of genes related to leukocyte chemotaxis, in particular neutrophil migration. Accordingly, TP costimulation increased expression of CSF3, which plays a key role in mobilizing neutrophils from the bone marrow, and major neutrophil chemokines CXCL1, CXCL2, CXCL3 and CXCL5 that recruit neutrophils to sites of inflammation including in inflammatory arthritis. Analysis of the late response to TNF similarly showed enrichment of genes important in chemotaxis, and suppression of genes in the cholesterol biosynthetic pathway, which we and others have previously linked to IFN responses. Targeted inspection of genes in additional pathways implicated in RA pathogenesis showed increased expression of genes in the Notch pathway. We believe that these pathways work together with the IL-1 pathway to increase immune cell recruitment and activation in inflammatory responses; these results are described on pp. 5-6 and are incorporated into Figures 1, 2 and Supplementary Fig. 2. 

Overall, the data looks good and acceptable but I need to confirm the above-mentioned criticisms. 

Recommendations for the authors: 

Reviewer #1 (Recommendations for the authors):   

The discussion section of the manuscript claims: "In this study, we utilized transcriptomics to demonstrate a 'TNF + PGE2' (TP) signature in RA and ICI-arthritis IL-1β+ synovial macrophages." This statement is misleading, as no new transcriptomic data from RA synovial samples were generated in this study. To support such a claim, the authors would need to compare primary monocytes or macrophages from RA patients using bulk RNA-seq or singlecell RNA-seq. Based on the current data, the comparison is limited to bulk RNA-seq findings from the authors' in vitro model and prior monocyte-fibroblast coculture studies. 

We have modified the abstract and discussion (pg. 10) to reflect that we have compared an in vitro generated TP signature with gene expression in previously identified RA macrophage subsets.

Author response:

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

We extend our sincere thanks to the editor, referees for eLife, and other commentators who have written evaluations of this manuscript, either in whole or in part. Sources of these comments were highly varied, including within the bioRxiv preprint server, social media (including many comments received on X/Twitter and some YouTube presentations and interviews), comments made by colleagues to journalists, and also some reviews of the work published in other academic journals. Some of these are formal and referenced with citations. Others were informal but nonetheless expressed perspectives that helped enable us to revise the manuscript with the inclusion of broader perspectives than the formal review process. It is beyond the scope of this summary to list every one of these, which have often been brought to the attention of different coauthors, but we begin by acknowledging the very wide array of peer and public commentary that have contributed to this work. The reaction speaks to a broad interest in open discussion and review of preprints. 

As we compiled this summary of changes to the manuscript, we recognized that many colleagues made comments about the process of preprint dissemination and evaluation rather than the data or analyses in the manuscript. Addressing such comments is outside the scope of this revised manuscript, but we do feel that a broader discussion of these comments would be valuable in another venue. Many commentators have expressed confusion about the eLife system of evaluation of preprints, which differs from the editorial acceptance or rejection practiced in most academic journals. As authors in many different nations, in varied fields, and in varied career stages, we ourselves are still working to understand how the academic publication landscape is changing, and how best to prepare work for new models of evaluation and dissemination. 

The manuscript and coauthor list reflect an interdisciplinary collaboration. Analyses presented in the manuscript come from a wide range of scientific disciplines. These range from skeletal inventory, morphology, and description, spatial taphonomy, analysis of bone fracture patterns and bone surface modifications, sedimentology, geochemistry, and traditional survey and mapping. The manuscript additionally draws upon a large number of previous studies of the Rising Star cave system and the Dinaledi Subsystem, which have shaped our current work. No analysis within any one area of research stands alone within this body of work: all are interpreted in conjunction with the outcomes of other analyses and data from other areas of research. Any single analysis in isolation might be consistent with many different hypotheses for the formation of sediments and disposition of the skeletal remains. But testing a hypothesis requires considering all data in combination and not leaving out data that do not fit the hypothesis. We highlight this general principle at the outset because a number of the comments from referees and outside specialists have presented alternative hypotheses that may arguably be consistent with one kind of analysis that we have presented, while seeming to overlook other analyses, data, or previous work that exclude these alternatives. In our revision, we have expanded all sections describing results to consider not only the results of each analysis, but how the combination of data from different kinds of analysis relate to hypotheses for the deposition and subsequent history of the Homo naledi remains. We address some specific examples and how we have responded to these in our summary of changes below. 

General organization:

The referee and editor comments are mostly general and not line-by-line questions, and we have compiled them and treated them as a group in this summary of changes, except where specifically noted. 

The editorial comments on the previous version included the suggestion that the manuscript should be reorganized to test “natural” (i.e. noncultural) hypotheses for the situations that we examine. The editorial comment suggested this as a “null hypothesis” testing approach. Some outside comments also viewed noncultural deposition as a null hypothesis to be rejected. We do not concur that noncultural processes should be construed as a null hypothesis, as we discuss further below. However, because of the clear editorial opinion we elected to revise the manuscript to make more explicit how the data and analyses test noncultural depositional hypotheses first, followed by testing of cultural hypotheses. This reorganization means that the revised manuscript now examines each hypothesis separately in turn. 

Taking this approach resulted in a substantial reorganization of the “Results” section of the manuscript. The “Results” section now begins with summaries of analyses and data conducted on material from each excavation area. After the presentation of data and analyses from each area, we then present a separate section for each of several hypotheses for the disposition and sedimentary context of the remains. These hypotheses include deposition of bodies upon a talus (as hypothesized in some previous work), slow sedimentary burial on a cave floor or within a natural depression, rapid burial by gravity-driven slumping, and burial of naturally mummified remains. We then include sections to test the hypothesis of primary cultural burial and secondary cultural burial. This approach adds substantial length to the Results. While some elements may be repeated across sections, we do consider the new version to be easier to take piece by piece for a reader trying to understand how each hypothesis relates to the evidence. 

The Results section includes analyses on several different excavation areas within the Dinaledi Subsystem. Each of these presents somewhat different patterns of data. We conceived of this manuscript combining these distinct areas because each of them provides information about the formation history of the Homo naledi-associated sediments and the deposition of the Homo naledi remains. Together they speak more strongly than separately. In the previous version of the manuscript, two areas of excavation were considered in detail (Dinaledi Feature 1 and the Hill Antechamber Feature), with a third area (the Puzzle Box area) included only in the Discussion and with reference to prior work. We now describe the new work undertaken after the 2013-2014 excavations in more detail. This includes an overview of areas in the Hill Antechamber and Dinaledi Chamber that have not yielded substantial H. naledi remains and that thereby help contextualize the spatial concentration of H. naledi skeletal material. The most substantial change in the data presented is a much expanded reanalysis of the Puzzle Box area. This reanalysis provides greater clarity on how previously published descriptions relate to the new evidence. The reanalysis also provides the data to integrate the detailed information on bone identification fragmentation, and spatial taphonomy from this area with the new excavation results from the other areas. 

In addition to Results, the reorganization also affected the manuscript’s Introduction section. Where the previous version led directly from a brief review of Pleistocene burial into the description of the results, this revised manuscript now includes a review of previous studies of the Rising Star cave system. This review directly addresses referee comments that express some hesitation to accept previous results concerning the structure and formation of sediments, the accessibility of the Dinaledi Subsystem, the geochronological setting of the H. naledi remains, and the relation of the Dinaledi Subsystem to nearby cave areas. Some parts of this overview are further expanded in the Supplementary Information to enable readers to dive more deeply into the previous literature on the site formation and geological configuration of the Rising Star cave system without needing to digest the entirety of the cited sources. 

The Discussion section of the revised manuscript is differentiated from Results and focuses on several areas where the evidence presented in this study may benefit from greater context. One new section addresses hypothesis testing and parsimony for Pleistocene burial evidence, which we address at greater length in this summary below. The majority of the Discussion concerns the criteria for recognizing evidence for burial as applied in other studies. In this research we employ a minimal definition but other researchers have applied varied criteria. We consider whether these other criteria have relevance in light of our observations and whether they are essential to the recognition of burial evidence more broadly. 

Vocabulary:

We introduce the term “cultural burial” in this revised manuscript to refer to the burial of dead bodies as a mortuary practice. “Burial” as an unmodified term may refer to the passive covering of remains by sedimentary processes. Use of the term “intentional burial” would raise the question of interpreting intent, which we do not presume based on the evidence presented in this research. The relevant question in this case is whether the process of burial reflects repeated behavior by a group. As we received input from various colleagues it became clear that burial itself is a highly loaded term. In particular there is a common assumption within the literature and among professionals that burial must by definition be symbolic. We do not take any position on that question in this manuscript, and it is our hope that the term “cultural burial” may focus the conversation around the extent that the behavioral evidence is repeated and patterned. 

Sedimentology and geochemistry of Dinaledi Feature 1:

Reviewer 4 provided detailed comments on the sedimentological and geochemical context that we report in the manuscript. One outside review (Foecke et al. 2024) included some of the points raised by reviewer 4, and additionally addressed the reporting of geochemical and sedimentological data in previous work that we cite. 

To address these comments we have revised the sedimentary context and micromorphology of sediments associated with Dinaledi Feature 1. In the new text we demonstrate the lack of microstratigraphy (supported by grain size analysis) in the unlithified mud clast breccia (UMCB), while such a microstratigraphy is observed in the laminated orange-red mudstones (LORM) that contribute clasts to the UMCB. Thus, we emphasize the presence and importance of a laterally continuous layer of LORM nature occurring at a level that appears to be the maximum depth of fossil occurrence. This layer is severely broken under extensive accumulation of fossils such as Feature 1 and only evidenced by abundant LORM clasts within and around the fossils. 

We have completely reworked the geochemical context associated with Feature 1 following the comments of reviewer 4. We described the variations and trends observed in the major oxides separate from trace and rare-earth elements. We used Harker variations plots to assess relationships between these element groups with CaO and Zn, followed by principal component analysis of all elements analyzed. The new geochemical analysis clearly shows that Feature 1 is associated with localized trace element signatures that exist in the sediments only in association with the fossil bones, which suggests lack of postdepositional mobilization of the fossils and sediments. We additionally have included a fuller description of XRF methods. 

To clarify the relation of all results to the features described in this study, we removed the geochemical and sedimentological samples from other sites within the Dinaledi Subsystem. These localities within the fissure network represent only surface collection of sediment, as no excavation results are available from those sites to allow for comparison in the context of assessing evidence of burial. These were initially included for comparison, but have now been removed to avoid confusion.  

Micromorphology of sediments:

Some referees (1, 3, and 4) and other commentators (including Martinón-Torres et al. 2024) have suggested that the previous manuscript was deficient due to an insufficient inclusion of micromorphological analysis of sediments. Because these commentators have emphasized this kind of evidence as particularly important, we review here what we have included and how our revision has addressed this comment. Previous work in the Dinaledi Chamber (Dirks et al., 2015; 2017) included thin section illustrations and analysis of sediment facies, including sediments in direct association with H. naledi remains within the Puzzle Box area. The previous work by Wiersma and coworkers (2020) used micromorphological analysis as one of several approaches to test the formation history of Unit 3 sediments in the Dinaledi Subsystem, leading to the interpretation of autobrecciation of earlier Unit 1 sediment. In the previous version of this manuscript we provided citations to this earlier work. The previous manuscript also provided new thin section illustrations of Unit 3 sediment near Dinaledi Feature 1 to place the disrupted layer of orange sediment (now designated the laminated orange silty mudstone unit) into context. 

In the new revised manuscript we have added to this information in three ways. First, as noted above in response to reviewer 4, we have revised and added to our discussion of micromorphology within and adjacent to the Dinaledi Feature 1. Second, we have included more discussion in the Supplementary Information of previous descriptions of sediment facies and associated thin section analysis, with illustrations from prior work (CC-BY licensed) brought into this paper as supplementary figures, so that readers can examine these without following the citations. Third, we have included Figure 10 in the manuscript which includes six panels with microtomographic sections from the Hill Antechamber Feature. This figure illustrates the consistency of sub-unit 3b sediment in direct contact with H. naledi skeletal material, including anatomically associated skeletal elements, with previous analyses that demonstrate the angular outlines and chaotic orientations of LORM clasts. It also shows density contrasts of sediment in immediate contact with some skeletal elements, the loose texture of this sediment with air-filled voids, and apparent invertebrate burrowing activity. To our knowledge this is the first application of microtomography to sediment structure in association with a Pleistocene burial feature. 

To forestall possible comments that the revised manuscript does not sufficiently employ micromorphological observations, or that any one particular approach to micromorphology is the standard, we present here some context from related studies of evidence from other research groups working at varied sites in Africa, Europe, and Asia. Hodgkins et al. (2021) noted: “Only a handful of micromorphological studies have been conducted on human burials and even fewer have been conducted on suspected burials from Paleolithic or hunter-gatherer contexts.” In that study, one supplementary figure with four photomicrographs of thin sections of sediments was presented. Interpretation of the evidence for a burial pit by Hodgkins et al. (2021) noted the more open microstructure of sediment but otherwise did not rely upon the thin section data in characterizing the sediments associated with grave fill. Martinón-Torres et al. (2021) included one Extended Data figure illustrating thin sections of sediments and bone, with two panels showing sediments (the remainder showing bone histology). The micromorphological analysis presented in the supplementary information of that paper was restricted to description of two microfacies associated with the proposed “pit” in that study. That study did carry out microCT scanning of the partially-prepared skeletal remains but did not report any sediment analysis from the microtomographic results. Maloney et al. (2022) reported no micromorphological or thin section analysis. Pomeroy et al. (2020a) included one illustration of a thin section; this study may be regarded as a preliminary account rather than a full description of the work undertaken. Goldberg et al. (2017) analyzed the geoarchaeology of the Roc de Marsal deposits in which possible burial-associated sediments had been fully excavated in the 1960s, providing new morphological assessments of sediment facies; the supplementary information to this work included five scans (not microscans) of sediment thin sections and no microphotographs. Fewlass et al. (2023) presented no thin section or micromorphological illustrations or methods. In summary of this research, we note that in one case micromorphological study provided observations that contributed to the evidence for a pit, in other cases micromorphological data did not test this hypothesis, and many researchers do not apply micromorphological techniques in their particular contexts. 

Sediment micromorphology is a growing area of research and may have much to provide to the understanding of ancient burial evidence as its standards continue to develop (Pomeroy et al. 2020b). In particular microtomographic analysis of sediments, as we have initiated in this study, may open new horizons that are not possible with more destructive thin-section preparation. In this manuscript, the thin section data reveals valuable evidence about the disruption of sediment structure by features within the Dinaledi Chamber, and microtomographic analysis further documents that the Hill Antechamber Feature reflects similar processes, in addition to possible post-burial diagenesis and invertebrate activity. Following up in detail on these processes will require further analysis outside the scope of this manuscript. 

Access into the Dinaledi Subsystem:

Reviewer 1 emphasizes the difficulty of access into the Dinaledi Subsystem as a reason why the burial hypothesis is not parsimonious. Similar comments have been made by several outside commentators who question whether past accessibility into the Dinaledi Subsystem may at one time have been substantially different from the situation documented in previous work. Several pieces of evidence are relevant to these questions and we have included some discussion of them in the Introduction, and additionally include a section in the Supplementary Information (“Entrances to the cave system”) to provide additional context for these questions. Homo naledi remains are found not only within the Dinaledi Subsystem but also in other parts of the cave system including the Lesedi Chamber, which is similarly difficult for non-expert cavers to access. The body plan, mass, and specific morphology of H. naledi suggest that this species would be vastly more suited to moving and climbing within narrow underground passages than living people. On this basis it is not unparsimonious to suggest that the evidence resulted from H. naledi activity within these spaces. We note that the accessibility of the subsystem is not strictly relevant to the hypothesis of cultural burial, although the location of the remains does inform the overall context which may reflect a selection of a location perceived as special in some way. 

Stuffing bodies down the entry to the subsystem:

Reviewer 3 suggests that one explanation for the emplacement of articulated remains at the top of the sloping floor of the Hill Antechamber is that bodies were “stuffed” into the chute that comprises the entry point of the subsystem and passively buried by additional accumulation of remains. This was one hypothesis presented in earlier work (Dirks et al. 2015) and considered there as a minimal explanation because it did not entail the entry of H. naledi individuals into the subsystem. The further exploration (Elliott et al. 2021) and ongoing survey work, as well as this manuscript, all have resulted in data that rejects this hypothesis. The revised manuscript includes a section in the results “Deposition upon a talus with passive burial” that examines this hypothesis in light of the data. 

Recognition of pits:

Referee 3 and 4 and several additional commentators have emphasized that the recognition of pit features is necessary to the hypothesis of burial, and questioned whether the data presented in the manuscript were sufficient to demonstrate that pits were present. We have revised the manuscript in several ways to clarify how all the different kinds of evidence from the subsystem test the hypothesis that pits were present. This includes the presentation of a minimal definition of burial to include a pit dug by hominins, criteria for recognizing that a pit was present, and an evaluation of the evidence in each case to make clear how the evidence relates to the presence of a pit and subsequent infill. As referee 3 notes, it can be challenging to recognize a pit when sediment is relatively homogeneous. This point was emphasized in the review by Pomeroy and coworkers (2020b), who reflected on the difficulty seeing evidence for shallow pits constructed by hominins, and we have cited this in the main text. As a result, the evidence for pits has been a recurrent topic of debate for most Pleistocene burial sites. However in addition to the sedimentological and contextual evidence in the cases we describe, the current version also reflects upon other possible mechanisms for the accumulation of bones or bodies. The data show that the sedimentary fill associated with the H. naledi remains in the cases we examine could not have passively accumulated slowly and is not indicative of mass movement by slumping or other high-energy flow. To further put these results into context, we added a section to the Discussion that briefly reviews prior work on distinguishing pits in Pleistocene burial contexts, including the substantial number of sites with accepted burial evidence for which no evidence of a pit is present. 

Extent of articulation and anatomical association:

We have added significantly greater detail to the descriptions of articulated remains and orientation of remains in order to describe more specifically the configuration of the skeletal material. We also provide 14 figures in main text (13 of them new) to illustrate the configuration of skeletal remains in our data. For the Puzzle Box area, this now includes substantial evidence on the individuation of skeletal fragments, which enables us to illustrate the spatial configuration of remains associated with the DH7 partial skeleton, as well as the spatial position of fragments refitted as part of the DH1, DH2, DH3, and DH4 crania. For Dinaledi Feature 1 and the Hill Antechamber Feature we now provide figures that key skeletal parts as identified, including material that is unexcavated where possible, and a skeletal part representation figure for elements excavated from Dinaledi Feature 1. 

Archaeothanatology:

Reviewer 2 suggests that a greater focus on the archaeothanatology literature would be helpful to the analysis, with specific reference to the sequence of joint disarticulation, the collapse of sediment and remains into voids created by decomposition, and associated fragmentation of the remains. In the revised manuscript we have provided additional analysis of the Hill Antechamber Feature with this approach in mind. This includes greater detail and illustration of our current hypothesis for individuation of elements. We now discuss a hypothesis of body disposition, describe the persistent joints and articulation of elements, and examine likely decomposition scenarios associated with these remains. Additionally, we expand our description and illustration of the orientation of remains and degree of anatomical association and articulation within Dinaledi Feature 1. For this feature and for the Hill Antechamber Feature we have revised the text to describe how fracturing and crushing patterns are consistent with downward pressure from overlying sediment and material. In these features, postdepositional fracturing occurred subsequent to the decomposition of soft tissue and partial loss of organic integrity of the bone. We also indicate that the loss by postdepositional processes of most long bone epiphyses, vertebral bodies, and other portions of the skeleton less rich in cortical bone, poses a challenge for testing the anatomical associations of the remaining elements. This is a primary reason why we have taken a conservative approach to identification of elements and possible associations. 

A further aspect of the site revealed by our analysis is the selective reworking of sediments within the Puzzle Box area subsequent to the primary deposition of some bodies. The skeletal evidence from this area includes body parts with elements in anatomical association or articulation, juxtaposed closely with bone fragments at varied pitch and orientation. This complexity of events evidenced within this area is a challenge for approaches that have been developed primarily based on comparative data from single-burial situations. In these discussions we deepen our use of references as suggested by the referee.   

Burial positions:

Reviewer 2 further suggests that illustrations of hypothesized burial positions would be valuable. We recognize that a hypothesized burial position may be an appealing illustration, and that some recent studies have created such illustrations in the context of their scientific articles. However such illustrations generally include a great deal of speculation and artist imagination, and tend to have an emotive character. We have added more discussion to the manuscript of possible primary disposition in the case of the Hill Antechamber Feature as discussed above. We have not created new illustrations of hypothesized burial positions for this revision. 

Carnivore involvement:

Referee 1 suggests that the manuscript should provide further consideration of whether carnivore activity may have introduced bones or bodies into the cave system. The reorganized Introduction now includes a review of previous work, and an expanded discussion within the Supplementary Information (“Hypotheses tested in previous work”). This includes a review of literature on the topic of carnivore accumulation and the evidence from the Dinaledi and Lesedi Chamber that rejects this hypothesis. 

Water transport and mud:

The eLife referees broadly accepted previous work showing that water inundation or mass flow of water-saturated sediment did not occur within the history of Unit 2 and 3 sediments, including those associated with H. naledi remains. However several outside commentators did refer specifically to water flow or mud flow as a mechanism for slumping of deposits and possible sedimentary covering of the remains. To address these comments we have added a section to the

Supplementary Information (“Description of the sedimentary deposits of the Dinaledi Subsystem”) that reviews previous work on the sedimentary units and formation processes documented in this area. We also include a subsection specifically discussing the term “mud” as used in the description of the sedimentology within the system, as this term has clearly been confusing for nonspecialists who have read and commented on the work. We appreciate the referees’ attention to the previous work and its terminology.  

Redescription of areas of the cave system:

Reviewer 1 suggests that a detailed reanalysis of all portions of the cave system in and around the Dinaledi Subsystem is warranted to reject the hypothesis that bodies entered the space passively and were scattered from the floor by natural (i.e. noncultural) processes. The referee suggests that National Geographic could help us with these efforts. To address this comment we have made several changes to the manuscript. As noted above, we have added material in Supplementary Information to review the geochronology of the Dinaledi Subsystem and nearby Dragon’s Back Chamber, together with a discussion of the connections between these spaces. 

Most directly in response to this comment we provide additional documentation of the possibility of movement of bodies or body parts by gravity within the subsystem itself. This includes detailed floor maps based on photogrammetry and LIDAR measurement, where these are physically possible, presented in Figures 2 and 3. In some parts of the subsystem the necessary equipment cannot be used due to the extremely confined spaces, and for these areas our maps are based on traditional survey methods. In addition to plan maps we have included a figure showing the elevation of the subsystem floor in a cross-section that includes key excavation areas, showing their relative elevation. All figures that illustrate excavation areas are now keyed to their location with reference to a subsystem plan. These data have been provided in previous publications but the visualization in the revised manuscript should make the relationship of areas clear for readers. The Introduction now includes text that discusses the configuration of the Hill Antechamber, Dinaledi Chamber, and nearby areas, and also discusses the instances in which gravity-driven movement may be possible, at the same time reviewing that gravity-driven movement from the entry point of the subsystem to most of the localities with hominin skeletal remains is not possible. 

Within the Results, we have added a section on the relationship of features to their surroundings in order to assist readers in understanding the context of these bone-bearing areas and the evidence this context brings to the hypothesis in question. We have also included within this new section a discussion of the discrete nature of these features, a question that has been raised by outside commentators. 

Passive sedimentation upon a cave floor or within a natural depression:

Reviewer 3 suggests that the situation in the Dinaledi Subsystem may be similar to a European cave where a cave bear skeleton might remain articulated on a cave floor (or we can add, within a hollow for hibernation), later to be covered in sediment. The reviewer suggests that articulation is therefore no evidence of burial, and suggests that further documentation of disarticulation processes is essential to demonstrating the processes that buried the remains. We concur that articulation by itself is not sufficient evidence of cultural burial. To address this comment we have included a section in the Results that tests the hypothesis that bodies were exposed upon the cave floor or within a natural depression. To a considerable degree, additional data about disarticulation processes subsequent to deposition are provided in our reanalysis of the Puzzle Box area, including evidence for selective reworking of material after burial. 

Postdepositional movement and floor drains:

Reviewer 3 notes that previous work has suggested that subsurface floor drains may have caused some postdepositional movement of skeletal remains. The hypothesis of postdepositional slumping or downslope movement has also been discussed by some external commentators (including Martinón-Torres et al. 2024). We have addressed this question in several places within the revised manuscript. As we now review, previous discussion of floor drains attempted to explain the subvertical orientation of many skeletal elements excavated from the Puzzle Box area. The arrangement of these bones reflects reworking as described in our previous work, and without considering the possibility of reworking by hominins, one mechanism that conceivably might cause reworking was downward movement of sediments into subsurface drains. Further exploration and mapping, combined with additional excavation into the sediments beneath the Puzzle Box area provided more information relevant to this hypothesis. In particular this evidence shows that subsurface drains cannot explain the arrangement of skeletal material observed within the Puzzle Box area. As now discussed in the text, the reworking is selective and initiated from above rather than below. This is best explained by hominin activity subsequent to burial. 

In a new section of the Results we discuss slumping as a hypothesis for the deposition of the remains. This includes discussion of downslope movement within the Hill Antechamber and the idea that floor drains may have been a mechanism for sediment reworking in and around the Puzzle Box area and Dinaledi Feature 1. As described in this section the evidence does not support these hypotheses. 

Hypothesis testing and parsimony:

Referees 1 and 3 and the editorial guidance all suggested that a more appropriate presentation would adopt a null hypothesis and test it. The specific suggestion that the null hypothesis should be a natural sedimentary process of deposition was provided not only by these reviewers but also by some outside commentators. To address this comment, we have edited the manuscript in two ways. The first is the addition of a section to the Discussion that specifically discusses hypothesis testing and parsimony as related to Pleistocene evidence of cultural burial. This includes a brief synopsis of recent disciplinary conversations and citation of work by other groups of authors, none of whom adopted this “null hypothesis” approach in their published work. 

As we now describe in the manuscript, previous work on the Dinaledi evidence never assumed any role for H. naledi in the burial of remains. Reading the reviewer reports caused us to realize that this previous work had followed exactly the “null hypothesis” approach that some suggested we follow. By following this null hypothesis approach, we neglected a valuable avenue of investigation. In retrospect, we see how this approach impeded us from understanding the pattern of evidence within the Puzzle Box area. Thus in the revised manuscript we have mentioned this history within the Discussion and also presented more of the background to our previous work in the Introduction. Hopefully by including this discussion of these issues, the manuscript will broaden conversation about the relation of parsimony to these issues. 

Language and presentation style:

Reviewer 4 criticizes our presentation, suggesting that the text “gives the impression that a hypothesis was formulated before data were collected.” Other outside commentators have mentioned this notion also, including Martinón-Torres et al. (2024) who suggest that the study began from a preferred hypothesis and gathered data to support it. The accurate communication of results and hypotheses in a scientific article is a broader issue than this one study. Preferences about presentation style vary across fields of study as well as across languages. We do not regret using plain language where possible. In any study that combines data and methods from different scientific disciplines, the use of plain language is particularly important to avoid misunderstandings where terms may mean different things in different fields. 

The essential question raised by these comments is whether it is appropriate to present the results of a study in terms of the hypothesis that is best supported. As noted above, we read carefully many recent studies of Pleistocene burial evidence. We note that in each of these studies that concluded that burial is the best hypothesis, the authors framed their results in the same way as our previous manuscript: an introduction that briefly reviews background evidence for treatment of the dead, a presentation of results focused on how each analysis supports the hypothesis of burial for the case, and then in some (but not all) cases discussion of why some alternative hypotheses could be rejected. We do not infer from this that these other studies started from a presupposition and collected data only to confirm it. Rather, this is a simple matter of presentation style. 

The alternative to this approach is to present an exhaustive list of possible hypotheses and to describe how the data relate to each of them, at the end selecting the best. This is the approach that we have followed in the revised manuscript, as described above under the direction of the reviewer and editorial guidance. This approach has the advantage of bringing together evidence in different combinations to show how each data point rejects some hypotheses while supporting others. It has the disadvantage of length and repetition. 

Possible artifact:

We have chosen to keep the description of the possible artifact associated with the Hill Antechamber Feature in the Supplementary Information. We do this while acknowledging that this is against the opinion of reviewer 4, who felt the description should be removed unless the object in question is fully excavated and physically analyzed. The previous version of the manuscript did not rely upon the stone as positive evidence of grave goods or symbolic content, and it noted that the data do not test whether the possible artifact was placed or was intentionally modified. However this did not satisfy reviewer 4, and some outside commentators likewise asserted that the object must be a “geofact” and that it should be removed. 

We have three arguments against this line of thinking. First, we do not omit data from our reporting. Whether Homo naledi shaped the rock or not, used it as a tool or not, whether the rock was placed with the body or not, it is unquestionably there. Omitting this one object from the report would be simply dishonest. Second, the data on this rock are at 16 micron resolution. While physical inspection of its surface may eventually reveal trace evidence and will enable better characterization of the raw material, no mode of surface scanning will produce better evidence about the object’s shape. Third, the position of this possible artifact within the feature provides significant information about the deposition of the skeletal material and associated sediments. The pitch, orientation, and position of the stone is not consistent with slow deposition but are consistent with the hypothesis that the surrounding sediment was rapidly emplaced at the same time as the articulated elements less than 2 cm away. 

In the current version, we have redoubled our efforts to provide information about the position and shape of this stone while not presupposing the intentionality of its shape or placement. We add here that the attitude expressed by referee 4 and other commentators, if followed at other sites, would certainly lead to the loss or underreporting of evidence, which we are trying to avoid.  

Consistency versus variability of behavior:

As described in the revised manuscript, different features within the Dinaledi Subsystem exhibit some shared characteristics. At the same time, they vary in positioning, representation of individuals and extent of commingling. Other localities within the subsystem and broader cave system present different evidence. Some commentators have questioned whether the patterning is consistent with a single common explanation, or whether multiple explanations are necessary. To address this line of questioning, we have added several elements to the manuscript. We created a new section on secondary cultural burial, discussing whether any of the situations may reflect this practice. In the Discussion, we briefly review the ways in which the different features support the involvement of H. naledi without interpreting anything about the intentionality or meaning of the behavior. We further added a section to the Discussion to consider whether variation among the features reflects variation in mortuary practices by H. naledi. One aspect of this section briefly cites variation in the location and treatment of skeletal remains at other sites with evidence of burial. 

Grave goods:

Some commentators have argued that grave goods are a necessary criterion for recognizing evidence of ancient burial. We added a section to the Discussion to review evidence of grave goods at other Pleistocene sites where burial is accepted. 

References:

• Dirks, P. H., Berger, L. R., Roberts, E. M., Kramers, J. D., Hawks, J., Randolph-Quinney, P. S., Elliott, M., Musiba, C. M., Churchill, S. E., de Ruiter, D. J., Schmid, P., Backwell, L. R., Belyanin, G. A., Boshoff, P., Hunter, K. L., Feuerriegel, E. M., Gurtov, A., Harrison, J. du G., Hunter, R., … Tucker, S. (2015). Geological and taphonomic context for the new hominin species Homo naledi from the Dinaledi Chamber, South Africa. eLife, 4, e09561. https://doi.org/10.7554/eLife.09561

• Dirks, P. H., Roberts, E. M., Hilbert-Wolf, H., Kramers, J. D., Hawks, J., Dosseto, A., Duval, M., Elliott, M., Evans, M., Grün, R., Hellstrom, J., Herries, A. I., Joannes-Boyau, R., Makhubela, T. V., Placzek, C. J., Robbins, J., Spandler, C., Wiersma, J., Woodhead, J., & Berger, L. R. (2017). The age of Homo naledi and associated sediments in the Rising Star Cave, South Africa. eLife, 6, e24231. https://doi.org/10.7554/eLife.24231

• Elliott, M., Makhubela, T., Brophy, J., Churchill, S., Peixotto, B., FEUERRIEGEL, E., Morris, H., Van Rooyen, D., Ramalepa, M., Tsikoane, M., Kruger, A., Spandler, C., Kramers, J., Roberts, E., Dirks, P., Hawks, J., & Berger, L. R. (2021). Expanded Explorations of the Dinaledi Subsystem,Rising Star Cave System, South Africa. PaleoAnthropology, 2021(1), 15–22. https://doi.org/10.48738/2021.iss1.68

• Fewlass, H., Zavala, E. I., Fagault, Y., Tuna, T., Bard, E., Hublin, J.-J., Hajdinjak, M., & Wilczyński, J. (2023). Chronological and genetic analysis of an Upper Palaeolithic female infant burial from Borsuka Cave, Poland. iScience, 26(12). https://doi.org/10.1016/j.isci.2023.108283

• Foecke, Kimberly K., Queffelec, Alain, & Pickering, Robyn. (n.d.). No Sedimentological Evidence for Deliberate Burial by Homo naledi – A Case Study Highlighting the Need for Best Practices in Geochemical Studies Within Archaeology and Paleoanthropology. PaleoAnthropology, 2024. https://doi.org/10.48738/202x.issx.xxx

• Goldberg, P., Aldeias, V., Dibble, H., McPherron, S., Sandgathe, D., & Turq, A. (2017). Testing the Roc de Marsal Neandertal “Burial” with Geoarchaeology. Archaeological and Anthropological Sciences, 9(6), 1005–1015. https://doi.org/10.1007/s12520-013-0163-2

• Maloney, T. R., Dilkes-Hall, I. E., Vlok, M., Oktaviana, A. A., Setiawan, P., Priyatno, A. A. D., Ririmasse, M., Geria, I. M., Effendy, M. A. R., Istiawan, B., Atmoko, F. T., Adhityatama, S., Moffat, I., Joannes-Boyau, R., Brumm, A., & Aubert, M. (2022). Surgical amputation of a limb 31,000 years ago in Borneo. Nature, 609(7927), 547–551. https://doi.org/10.1038/s41586-022-05160-8

• Martinón-Torres, M., d’Errico, F., Santos, E., Álvaro Gallo, A., Amano, N., Archer, W., Armitage, S. J., Arsuaga, J. L., Bermúdez de Castro, J. M., Blinkhorn, J., Crowther, A., Douka, K., Dubernet, S., Faulkner, P., Fernández-Colón, P., Kourampas, N., González García, J., Larreina, D., Le Bourdonnec, F.-X., … Petraglia, M. D. (2021). Earliest known human burial in Africa. Nature, 593(7857), Article 7857. https://doi.org/10.1038/s41586021-03457-8

• Martinón-Torres, M., Garate, D., Herries, A. I. R., & Petraglia, M. D. (2023). No scientific evidence that Homo naledi buried their dead and produced rock art. Journal of Human Evolution, 103464. https://doi.org/10.1016/j.jhevol.2023.103464

• Pomeroy, E., Bennett, P., Hunt, C. O., Reynolds, T., Farr, L., Frouin, M., Holman, J., Lane, R., French, C., & Barker, G. (2020a). New Neanderthal remains associated with the ‘flower burial’ at Shanidar Cave. Antiquity, 94(373), 11–26. https://doi.org/10.15184/aqy.2019.207

• Pomeroy, E., Hunt, C. O., Reynolds, T., Abdulmutalb, D., Asouti, E., Bennett, P., Bosch, M., Burke, A., Farr, L., Foley, R., French, C., Frumkin, A., Goldberg, P., Hill, E., Kabukcu, C., Lahr, M. M., Lane, R., Marean, C., Maureille, B., … Barker, G. (2020b). Issues of theory and method in the analysis of Paleolithic mortuary behavior: A view from Shanidar Cave. Evolutionary Anthropology: Issues, News, and Reviews, 29(5), 263–279. https://doi.org/10.1002/evan.21854

• Robbins, J. L., Dirks, P. H. G. M., Roberts, E. M., Kramers, J. D., Makhubela, T. V., HilbertWolf, H. L., Elliott, M., Wiersma, J. P., Placzek, C. J., Evans, M., & Berger, L. R. (2021). Providing context to the Homo naledi fossils: Constraints from flowstones on the age of sediment deposits in Rising Star Cave, South Africa. Chemical Geology, 567, 120108. https://doi.org/10.1016/j.chemgeo.2021.120108

• Wiersma, J. P., Roberts, E. M., & Dirks, P. H. G. M. (2020). Formation of mud clast breccias and the process of sedimentary autobrecciation in the hominin-bearing (Homo naledi) Rising Star Cave system, South Africa. Sedimentology, 67(2), 897–919. https://doi.org/10.1111/sed.12666

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

This work tried to map the synaptic connectivity between the inputs and outputs of the song premotor nucleus, HVC in zebra finches to understand how sensory (auditory) to motor circuits interact to coordinate song production and learning. The authors optimized the optogenetic technique via AAV to manipulate auditory inputs from a specific auditory area one-by-one and recorded synaptic activity from a neuron with whole-cell recording from slice preparation with identification of the projection area by retrograde neuronal tracing. This thorough and detailed analysis provides compelling evidence of synaptic connections between 4 major auditory inputs (3 forebrain and 1 thalamic region) within three projection neurons in the HVC; all areas give monosynaptic excitatory inputs and polysynaptic inhibitory inputs, but proportions of projection to each projection neuron varied. They also find specific reciprocal connections between mMAN and Av. Taken together the authors provide the map of the synaptic connection between intercortical sensory to motor areas which is suggested to be involved in zebra finch song production and learning.

Strengths:

The authors optimized optogenetic tools with eGtACR1 by using AAV which allow them to manipulate synaptic inputs in a projection-specific manner in zebra finches. They also identify HVC cell types based on projection area. With their technical advance and thorough experiments, they provided detailed map synaptic connections.

Weaknesses:

As it is the study in brain slice, the functional implication of synaptic connectivity is limited. Especially as all the experiments were done in the adult preparation, there could be a gap in discussing the functions of developmental song learning.

We thank the reviewer for their appreciation of our work. Although we agree that there can be limitations to brain slice preparations, the approaches used here for synaptic connectivity mapping are well-designed to identify long-range synaptic connectivity patterns. Optogenetic stimulation of axon terminals in brain slices does not require intact axons and works well when axons are cut, allowing identification of all inputs expressing optogenetic channels from aXerent regions. Terminal stimulation in slices yields stable post-synaptic responses for hours without rundown, assuring that polysynaptic and monosynaptic connections can be reliably identified in our brain slices.  Additionally, conducting similar types of experiments in vivo can run into important limitations. First, the extent of TTX and 4-AP diXusion, which is necessary for identification of long-range monosynaptic connections, can be diXicult to verify in vivo - potentially confounding identification of monosynaptic connectivity.  Second, conducting whole-cell patch-clamp experiments in vivo, particularly in deeper brain regions, is technically challenging, and would limit the number of cells that can be patched and increase the number of animals needed. 

We agree that there may well be important diXerences between adult connectivity and connectivity patterns in the juvenile brain. Indeed, learning and experience during development almost certainly shape connectivity patterns and these patterns of connectivity may change incrementally and/or dynamically during development. Ultimately, adult connectivity patterns are the result of changes in the brain that accrue over development. Given that this is the first study mapping long-range connectivity of HVC input-output pathways, we reasoned that the adult connectivity would provide a critical reference allowing future studies to map diXerent stages of juvenile connectivity and the changes in connectivity driven by milestones like forming a tutor song memory, sensorimotor learning, and song crystallization.

In this revision we worked to better highlight the points raised above and thank the reviewer for their comments.

Reviewer #2 (Public review):

Summary:

The manuscript describes synaptic connectivity in the Songbird cortex's four main classes of sensory neuron aXerents onto three known classes of projection neurons of the pre-motor cortical region HVC. HVC is a region associated with the generation of learned bird songs. Investigators here use all male zebra finches to examine the functional anatomy of this region using patch clamp methods combined with optogenetic activation of select neuronal groups.

Strengths:

The quality of the recordings is extremely high and the quantity of data is on a very significant scale, this will certainly aid the field.

Weaknesses:

The authors could make the figures a little easier to navigate. Most of the figures use actual anatomical images but it would be nice to have this linked with a zebra finch atlas in more of a cartoon format that accompanied each fluro image. Additionally, for the most part, figures showing the labeling lack scale bar values (in um). These should be added not just shown in the legends.

The authors could make it clear in the abstract that this is all male zebra finches - perhaps this is obvious given the bird song focus, but it should be stated. The number of recordings from each neuron class and the overall number of birds employed should be clearly stated in the methods (this is in the figures, but it should say n=birds or cells as appropriate).

The authors should consider sharing the actual electrophysiology records as data.

We thank the reviewer for their assessment of our research and suggestions. We have implemented many of these suggestions and provide details in our response to their specific Recommendations. Additionally, we are organizing our data and will make it publicly available with the version of record.

Reviewer #3 (Public review):

Nucleus HVC is critical both for song production as well as learning and arguably, sitting at the top of the song control system, is the most critical node in this circuit receiving a multitude of inputs and sending precisely timed commands that determine the temporal structure of song. The complexity of this structure and its underlying organization seem to become more apparent with each experimental manipulation, and yet our understanding of the underlying circuit organization remains relatively poorly understood. In this study, Trusel and Roberts use classic whole-cell patch clamp techniques in brain slices coupled with optogenetic stimulation of select inputs to provide a careful characterization and quantification of synaptic inputs into HVC. By identifying individual projection neurons using retrograde tracer injections combined with pharmacological manipulations, they classify monosynaptic inputs onto each of the three main classes of glutamatergic projection neurons in HVC (RA-, Area X- and Av-projecting neurons). This study is remarkable in the amount of information that it generates, and the tremendous labor involved for each experiment, from the expression of opsins in each of the target inputs (Uva, NIf, mMAN, and Av), the retrograde labelling of each type of projection neuron, and ultimately the optical stimulation of infected axons while recording from identified projection neurons. Taken together, this study makes an important contribution to increasing our identification, and ultimately understanding, of the basic synaptic elements that make up the circuit organization of HVC, and how external inputs, which we know to be critical for song production and learning, contribute to the intrinsic computations within this critic circuit.

This study is impressive in its scope, rigorous in its implementation, and thoughtful regarding its limitations. The manuscript is well-written, and I appreciate the clarity with which the authors use our latest understanding of the evolutionary origins of this circuit to place these studies within a larger context and their relevance to the study of vocal control, including human speech. My comments are minor and primarily about legibility, clarification of certain manipulations, and organization of some of the summary figures.

We thank the reviewer for their thoughtful assessment of our research.

Recommendations for the authors:

The following recommendations were considered by all reviewers to be important to incorporate for improving this paper:

(1) Clarify the site of viral injection and the possibility of labeling other structures a) Show images of viral injection sites.

We provide a representative image of viral expression for each pathway studied in this manuscript. Please see panel A in Figures 2-3 and 5-6 showing our viral expression in Uva, NIf, mMAN, and Av respectively.  

b) Include in discussion caveats that the virus may spread beyond the boundaries of structures (e.g. especially injections into NIF could spread into Field L).

For each HVC aXerent nucleus we have now included a sentence describing the possible spread of viral infection in surrounding structures in the Results. We also now expanded the image from the Av section to include NIf, to showcase lack of viral expression in NIf (see Fig. 6A).

(2) Clarify the logic and precise methods of the TTX and 4-AP experiments

a) Please see the detailed issue raised by Reviewer 3, Major Point 1 below.

The TTX and 4AP application is the gold-standard of opsin-assisted synaptic circuit interrogation, pioneered by the Svoboda lab in 2009 (Petreanu, Mao et al. 2009) and widely used to assess monosynaptic connectivity in multiple brain circuits, as summarized in a recent review(Linders, Supiot et al. 2022). We now better describe the logic of this approach in the second paragraph of the Results section and cite the first description of this method from the Svoboda lab and a recent review weighing this method with other optogenetic methods for tracing synaptic connections in the brain.

(3) Include caveats in discussion

a) Note that there may be other inputs to HVC that were not examined in this study (e.g. CMM, Field L)

In our original manuscript we did state “Although a complete description of HVC circuitry will require the examination of other potential inputs (i.e. RA_HVC PNs, A11 glutamatergic neurons(Roberts, Klein et al. 2008, Ben-Tov, Duarte et al. 2023)) and a characterization of interneuron synaptic connectivity, here we provide a map of the synaptic connections between the 4 best described aPerents to HVC and its 3 populations of projection neurons” in the last paragraph of the Discussion. We have now edited this sentence to include the projection from NCM to HVC and cited Louder et al., 2024.

We have extensively mapped input pathways to HVC, and consistent with Vates (Vates, Broome et al. 1996) we have not found evidence that Field L projects to HVC. Rather that it projects to the shelf region outside of HVC. Consistent with this, we do not see retrogradely labeled neurons in Field L following tracer injections confined to HVC (see Fig. 3G). Additionally, we find that CM projections to HVC arise from the nucleus Avalanche (Roberts, Hisey et al. 2017) which we specifically examine in this study. We do not dispute that there may be other pathways projecting to HVC that will need to be examined in the future, including known projections from neuromodulatory regions and RA, from developmentally restricted pathway(s) like NCM (Louder, Kuroda et al. 2024), and from yet unidentified pathways.

b) Also note that birds in this study were adults and that some inputs to HVC likely to be important for learning may recede during development (e.g. Louder et al, 2024).

In the second to last paragraph of the Discussion we now state: While our opsin-assisted circuit mapping provides us with a new level of insight into HVC synaptic circuitry, there are limitations to this research that should be considered. All circuit mapping in this study was carried out in brain slices from adult male zebra finches. Future studies will be needed to examine how this adult connectivity pattern relates to patterns of connectivity in juveniles during sensory or sensorimotor phases of vocal learning and connectivity patterns in female birds.   

(4) Consider cosmetic changes to figures as suggested by Reviewers 2-3 below.

We thank the reviewers for their suggestions and have implemented the changes as best we can.

(5) Address all minor issues raised below.

Reviewer #1 (Recommendations for the authors):

I see this study is well designed to answer the author's specific question, mapping synaptic auditorymotor connections within HVC. Their experiments with advanced techniques of projection-specific optogenetic manipulation of synaptic inputs and retrograde identification of projection areas revealed input-output combination selective synaptic mapping.

As I found this study advanced our knowledge with the compelling dataset, I have only some minor comments here.

(1) One technical concern is we don't see how much the virus infection was focused on the target area and if we can ignore the eXect of synaptic connectivity from surrounding areas. As the amount of virus they injected is large (1.5ul) and target areas are small, we assume the virus might spread to the surrounding area, such as field L which also projects to HVC when targeting Nif. While I think the majority of the projections were from their target areas, it would be better to mention (also the images with larger view areas) the possibilities of projections of surrounding areas.

We agree with the reviewer about the concern about specificity of viral expression. For this reason, we included sample images of the viral expression in each target area (panel A in Fig. 2,3,5,6). We have now also included a sentence at the beginning of each subsection of our Result to describe how we have ensured interpretability of the results. Uva and mMAN’s surrounding areas are not known to project to HVC. Possible cross-infection is an issue for Av and NIf, and we checked each bird’s injection site to ensure that eGtACR1+ cells were not visible in the unintended HVC-projecting areas.

As mentioned in our response the public comment, consistent with Vates (Vates, Broome et al. 1996) we do not see evidence that Field L projects directly to HVC (see Fig. 3G).

(2) Another concern about the technical issue is the damage to axonal projections. While I understand the authors stimulated axonal terminals axonal projections were assumed to be cut and their ability to release neurotransmitters would be reduced especially after long-term survival or repeated stimulation. Mentioning whether projection pathways were within their 230um-thick slice (probably depends on input sites) or not and the eXect of axonal cut would be helpful.

We agree that slice electrophysiology has limitations. However, we disagree with the claim of reduced reliability or stability of the evoked response. We and others find that electrical and optogenetic repeated terminal stimulation in slices can yield stable post-synaptic responses for tens of minutes and even hours (Bliss and Gardner-Medwin 1973, Bliss and Lomo 1973, Liu, Kurotani et al. 2004, Pastalkova, Serrano et al. 2006, Xu, Yu et al. 2009, Trusel, Cavaccini et al. 2015, Trusel, Nuno-Perez et al. 2019). Indeed, long-term synaptic plasticity experiments in most preparations and across brain areas rely on such stability of the presynaptic machinery for synaptic release, despite axons being severed from their parent soma. Our assumption is the vast majority, if not all, connections between axon terminals and their cell body in the aXerent regions have been cut in our preparations. Nonetheless, the diversity of outcomes we report (currents returning after TTX+4AP or not, depending on the specific combination of input and HVCPN class) is consistent with the robustness of the synaptic interrogation method. 

(3) While I understand this study focused on 4 major input areas and the authors provide good pictures of synaptic HVC connections from those areas, HVC has been reported to receive auditory inputs from other areas as well (CMM, FieldL, etc.). It is worth mentioning that there are other auditory inputs and would be interesting to discuss coordination with the inputs from other areas.

We have extensively mapped input pathways to HVC, and consistent with Vates (Vates, Broome et al. 1996) we have not found evidence that Field L projects to HVC. Rather that it projects to the shelf region outside of HVC. Consistent with this, we do not see retrogradely labeled neurons in Field L following tracer injections confined to HVC (see Fig. 3G). Additionally, we find that CM projections to HVC arise from the nucleus Avalanche (Roberts, Hisey et al. 2017) which we specifically examine in this study. We do not dispute that there may be other pathways projecting to HVC that will need to be examined in the future, including known projections from neuromodulatory regions and RA, from developmentally restricted pathway(s) like NCM (Louder, Kuroda et al. 2024), and from yet unidentified pathways.

(4) The HVC local neuronal connections have been reported to be modified and a recent study revealed the transient auditory inputs into HVC during song learning period. The author discusses the functions of HVC synaptic connections on song learning (also title says synaptic connection for song learning), however, the experiments were done in adults and dp not discuss the possibility of diXerent synaptic connection mapping in juveniles in the song learning period. Mentioning the neuronal activities and connectivity changes during song learning is important. Also, it would be helpful for the readers to discuss the potential diXerences between juveniles/adults if they want to discuss the functions of song learning.

We now mention in the Discussion that this is an important caveat of our research and that future studies will be needed to examine how these adult connectivity patterns relate to connectivity patterns in juveniles during sensory or sensorimotor phases of vocal learning and connectivity patterns in female birds. Nonetheless, the title and abstract cite song learning because it is important for the broader public to understand that at least some of these aXerent brain regions carry an essential role in song learning (Foster and Bottjer 2001, Roberts, Gobes et al. 2012, Roberts, Hisey et al. 2017, Zhao, Garcia-Oscos et al. 2019, Koparkar, Warren et al. 2024).

Reviewer #2 (Recommendations for the authors):

The work is very detailed and will be an important resource to those working in the field. The recordings are of a high quality and lots of information is included such as measures of response kinetics amplitude and pharmacological confirmation of excitatory and inhibitory synaptic responses. In general, I feel the quality is extremely high and the quantity of data is on a very significant exhaustive scale that will certainly aid the field. I have come at this conclusion as a non zebra finch person but I feel the connection information shown will be of benefit given its high quality.

Figure 7 is a nice way of showing the overall organization. Optional suggestion, consider highlighting anything in Figure 7 that results in a new understanding of the song system as compared to previous work on anatomy and function.

We thank the reviewer for the kind comments about our research. We have highlighted our newly found connection between mMAN and Av and all the connections onto the HVC PNs in Panel B are newly identified in this study.

Reviewer #3 (Recommendations for the authors):

Major points

(1) Clarification regarding methods for determining monosynaptic events:

One of the manipulations that I struggled the most with was those describing the use of TTX + 4AP to isolate monosynaptic events. Initially, not being as familiar with the use of optically based photostimulation of axons to release transmitter locally, I was initially confused by statements such as "we found that oEPSC returned after application of TTX+4AP". This might be clear to someone performing these manipulations, but a bit more clarification would be helpful. Should I assume that an existing monosynaptic EPSC would be masked by co-occurring polysynaptic IPSCs which disappear following application of TTX + 4AP, thereby unmasking the monosynaptic EPSC, thereby causing the EPSC to "return"? A word that I am not sure works. Continuing my confusion with these experiments, I am unsure how this cocktail of drugs is added, if it is even added as a cocktail, which is what I initially assumed. The methods and the results are not so clear if they are added in sequence and why and if traces are recorded after the addition of both drugs or if they are recorded for TTX and then again for TTX + 4AP. Finally, looking at the traces in the experimental figures (e.g. Figures 2F, 3F, 5F, and 6F), it is diXicult to see what is being shown, at least for me. First, the authors need to describe better in the results why they stimulate twice in short succession and why they seem to use the response to the second pulse (unless I am mistaken) to measure the monosynaptic event. Second, I was confused by the traces (which are very small) in the presence of TTX. I would have expected to see a response if there was a monosynaptic EPSC but I only seem to see a flat line.  

The confusion that I list above might be due in part to my ignorance, but it is important in these types of papers not to assume too much expertise if you want readers with a less sophisticated understanding of synaptic physiology to understand the data. In other words, a little bit more clarity and hand-holding would be welcome.

We understand the reviewer’s confusion about the methodology.  In Voltage clamp, the amplifier injects current through the electrode maintaining the membrane voltage to -70mV, where the equilibrium potential for Cl- is near equilibrium, and therefore the only synaptic current evoked by light stimulation is due to cation influx, mainly through AMPA receptors (see Fig. 1).  Therefore, cooccurring polysynaptic IPSCs wouldn’t be visible. We examine those holding the membrane voltage at +10mV, see Fig. 1. TTX application suppresses V-dependent Na+ channels and therefore stops all neurotransmission. We show the traces upon TTX to show that currents we were recording prior to TTX application were of synaptic origin, and not due to accidental expression of opsin in the patched cell. Also, this ensures that any current visible after 4AP application is due to monosynaptic transmission and not to a failure of TTX application.

After recording and light stimulation with TTX, we then add 4AP, which is a blocker of presynaptic K+ channels. This prevents the repolarization of the terminals that would occur in response to opsinmediated local depolarization. 4AP application, therefore, allows local opsin-driven depolarizations to reach the threshold for Ca2+-dependent vesicle docking and release. This procedure selectively reveals or unmasks the monosynaptic currents because any non-monosynaptically connected neuron would still need V-dependent Na+ channels to eXectively produce indirect neurotransmission onto the patched cell. The TTX and 4AP application is the gold-standard of opsinassisted synaptic circuit interrogation, pioneered by the Svoboda lab in 2009 and widely used to assess monosynaptic connectivity in multiple brain circuits, as summarized in a recent review (Linders et al., 2022). We now include 2 more sentences near the beginning of the Results to clarify this process and directly point to the Linders review for researchers wanting a deeper explanation of this technique. 

The double stimulation is unrelated to our testing of monosynaptic connections. We originally conducted the experiments by delivering 2 pulses of light separated by 50ms, a common way to examine the pair-pulse ratio (PPR) – a physiological measure which is used to probe synapses for short-term plasticity and release probability. However, through discussions with colleagues we realized that the slow decay time of eGtACR1 may complicate interpretation of the response to the second light pulse. Thus, we elected to not report these results and indicated this in the Methods section:  “We calculated the paired-pulse ratio (PPR) as the amplitude of the second peak divided by the amplitude of the first peak elicited by the twin stimuli, however due to slow kinetics of eGtACR1 the results would be diPicult to interpret, and therefore we are not currently reporting them.” 

(2) Suggestions for improving summary figures:

Summary Figure 1a: The circuit diagram (schematic to the right of 1a) is OK but I initially found it a bit diXicult to interpret. For example, it is not clear why pink RA projecting neurons don't reach as far to the right as X or Av projecting neurons, suggesting that they are not really projection neurons. Also, the big question marks in the intermediate zone are not entirely intuitive. It seems there might be a better way of representing this. It might also be worth stating in the figure legend that the interconnectivity patterns shown in the figure between PNs in HVC are based on specific prior studies.

We thank the reviewer for the constructive criticism. We have modified the figure to extend the RA projection line and mentioned in the figure legend that connectivity between PNs is based on prior studies.

Summary Figure 1a: I am not sure I love this figure. There are a few minor issues. First, there are too many browns [Nif/AV and mMAN] which makes it more challenging to clearly disambiguate the diXerent projections. Second, it is unclear why this figure does not represent projections from RA to HVC. My biggest concern with this figure is that it oversimplifies some of the findings. From the figure, one gets the impression that Uva only projects to RA-PNs and that Av only projects to X-PNs even though the authors show connections to other PNs. With the small sample size in this current study for each projection and each PN type, one really cannot rule out that these "minority" projections are not important. I, therefore, suggest that the authors qualitatively represent the strength/probability of connections by weighting with thickness of aXerent connections.

We assume the reviewer is commenting on our summary figure panel 7B. We agree with the referee that this is a simplified representation of our findings. We had indeed indicated in the legend that this was just a “Schematic of the HVC aXerent connectivity map resulting from the present work” and that “For conceptualization purposes, aXerent connectivity to HVC-PNs is shown only when the rate of monosynaptic connectivity reaches 50% of neurons examined”. We have added a title to highlight that this is but a simplification. We have now adjusted the colors to make the figure easier to follow. Based on the reviewers critique we searched for a better method for summarizing the complex connectivity patterns described in this research. We settled on a Sankey diagram of connectivity. This is now Figure 7C. In this diagram, we are able to show the proportion of connections from each input pathway onto each class of neuron and if these connections are poly or monosynaptic. We find this to a straightforward way of displaying all of the connectivity patterns identified in our figure 2-3 and 4-5 look forward to understanding if the reviewers find this a useful way of illustrating our findings.

Minor points:

(1) Line 50 - typo - song circuits.

Thank you for catching this.

(2) Line 106 - 111 - The findings suggest that 100% of Uva projections onto HVCRA neurons are monosynaptic. However, because the authors only tested 6 neurons their statements that their findings are so diXerent from other studies, should be somewhat tempered since these other studies (e.g. Moll et al.) looked at 251 neurons in HVC and sampling bias could still somewhat explain the diXerence.

We observed oEPSCs in 43 of 51 (84.3%) HVC-RA neurons recorded (mean rise time = 2.4 ms) and monosynaptic connections onto 100% of the HVC-RA neurons tested (n = 6). Moll et al. combined electrical stimulation of Uva with two-photon calcium imaging (GCaMP6s) of putative HVC-RA neurons (n = 251 neurons). We should note that these are putative HVC-RA neurons because they were not visually identified using retrograde tracing or using some other molecular handle. They report that only ~16% of HVC-RA neurons showed reliable calcium responses following Uva stimulation. Although the experiments by Moll et al are technically impressive, calcium imaging is an insensitive technique for measuring post-synaptic responses, particularly subthreshold responses, when compared to whole-cell patch-clamp recordings. This approach cannot identify monosynaptic connections and is likely limited to only be sensitive suprathreshold activity that likely relies on recruitment of other polysynaptic inputs onto the neurons in HVC. Furthermore, as indicated in the Discussion, our opsin-mediated synaptic interrogation recruits any eGtACR1+ Uva terminal in the slice and therefore will have great likelihood of revealing any existing connections. 

A limitation of whole-cell patch-clamp recordings is that it is a laborious low throughput technique. Future experiments using better imaging approaches, like voltage imaging, may be able to weigh in on diXerences between what we report here using whole-cell patch-clamp recordings from visually identified HVC-RA neurons combined with optogenetic manipulations of Uva terminals and the calcium imaging results reported by Moll. Nonetheless, whole-cell patch-clamp recordings combined with optogenetic manipulations is likely to remain the most sensitive method for identifying synaptic connectivity.

(3) Figure 2G - the significance of white circles is not clear.

The figure legend indicates that those highlight and mark the position of “retrogradely labeled HVCprojecting neurons in Uva (cyan, white circles)” to facilitate identification of colocalization with the in-situ markers.

(4) Line 135 - Cardin et al. (J. Neurophys. 2004) is the first to show that song production does not require Nif.

We thank the reviewer pointing this out and we have cited this important study. 

(5) Line 183 - This is a confusing sentence because I initially thought that mMAN-mMANHVC PNs was a category!

We switched the dash with a colon.

(6) Figure 4d could use some arrows to identify what is shown. It is assumed that the box represents mMAN. Should it be assumed that Av is not in the plane of this section? If not, this should be stated in the legend. It is also unclear where the anterograde projections are. Is this the dork highway that goes from the box to the dorsal surface? If yes this should be indicated but it should also be made clear why the projections go both in the dorsal as well as the ventral directions.

The inset, as indicated by the lines around it, is a magnification of the terminal fields in Av. We added an explanation of the inset.

(7) Discussion. In the introduction, the authors mention projections from RA to HVC but never end up studying them in the current manuscript which seems like a missed opportunity and perhaps even a weakness of the study. In the discussion, it would certainly be good for the authors to at least discuss the possible significance of these projections and perhaps why they decided not to study them.

We thank the reviewer for the comment. Unfortunately, we couldn’t reliably evoke interpretable currents from RA, and we elected to publish the current version of the paper with these 4 major inputs. Nonetheless, we have indicated in the Introduction and in the Discussion that more inputs (e.g. RA, A11, NCM) remain to be evaluated. 

(8) Line 622 - Is this reference incomplete?

We thank the reviewer. We have corrected the reference.

• Ben-Tov, M., F. Duarte and R. Mooney (2023). "A neural hub for holistic courtship displays." Curr Biol 33(9): 1640-1653 e1645.

• Bliss, T. V. and A. R. Gardner-Medwin (1973). "Long-lasting potentiation of synaptic transmission in the dentate area of the unanaestetized rabbit following stimulation of the perforant path." J Physiol 232(2): 357-374.

• Bliss, T. V. and T. Lomo (1973). "Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path." J Physiol 232(2): 331-356.

• Foster, E. F. and S. W. Bottjer (2001). "Lesions of a telencephalic nucleus in male zebra finches: Influences on vocal behavior in juveniles and adults." J Neurobiol 46(2): 142-165.

• Koparkar, A., T. L. Warren, J. D. Charlesworth, S. Shin, M. S. Brainard and L. Veit (2024). "Lesions in a songbird vocal circuit increase variability in song syntax." Elife 13.

• Linders, L. E., L. F. Supiot, W. Du, R. D'Angelo, R. A. H. Adan, D. Riga and F. J. Meye (2022). "Studying Synaptic Connectivity and Strength with Optogenetics and Patch-Clamp Electrophysiology." Int J Mol Sci 23(19).

• Liu, H. N., T. Kurotani, M. Ren, K. Yamada, Y. Yoshimura and Y. Komatsu (2004). "Presynaptic activity and Ca2+ entry are required for the maintenance of NMDA receptor-independent LTP at visual cortical excitatory synapses." J Neurophysiol 92(2): 1077-1087.

• Louder, M. I. M., M. Kuroda, D. Taniguchi, J. A. Komorowska-Muller, Y. Morohashi, M. Takahashi, M. Sanchez-Valpuesta, K. Wada, Y. Okada, H. Hioki and Y. Yazaki-Sugiyama (2024). "Transient sensorimotor projections in the developmental song learning period." Cell Rep 43(5): 114196.

• Pastalkova, E., P. Serrano, D. Pinkhasova, E. Wallace, A. A. Fenton and T. C. Sacktor (2006). "Storage of spatial information by the maintenance mechanism of LTP." Science 313(5790): 1141-1144.

• Petreanu, L., T. Mao, S. M. Sternson and K. Svoboda (2009). "The subcellular organization of neocortical excitatory connections." Nature 457(7233): 1142-1145.

• Roberts, T. F., S. M. Gobes, M. Murugan, B. P. Olveczky and R. Mooney (2012). "Motor circuits are required to encode a sensory model for imitative learning." Nat Neurosci 15(10): 1454-1459.

• Roberts, T. F., E. Hisey, M. Tanaka, M. G. Kearney, G. Chattree, C. F. Yang, N. M. Shah and R. Mooney (2017). "Identification of a motor-to-auditory pathway important for vocal learning." Nat Neurosci 20(7): 978-986.

• Roberts, T. F., M. E. Klein, M. F. Kubke, J. M. Wild and R. Mooney (2008). "Telencephalic neurons monosynaptically link brainstem and forebrain premotor networks necessary for song." J Neurosci 28(13): 3479-3489.

• Trusel, M., A. Cavaccini, M. Gritti, B. Greco, P. P. Saintot, C. Nazzaro, M. Cerovic, I. Morella, R. Brambilla and R. Tonini (2015). "Coordinated Regulation of Synaptic Plasticity at Striatopallidal and Striatonigral Neurons Orchestrates Motor Control." Cell Rep 13(7): 1353-1365.

• Trusel, M., A. Nuno-Perez, S. Lecca, H. Harada, A. L. Lalive, M. Congiu, K. Takemoto, T. Takahashi, F. Ferraguti and M. Mameli (2019). "Punishment-Predictive Cues Guide Avoidance through Potentiation of Hypothalamus-to-Habenula Synapses." Neuron 102(1): 120-127.e124.

• Vates, G. E., B. M. Broome, C. V. Mello and F. Nottebohm (1996). "Auditory pathways of caudal telencephalon and their relation to the song system of adult male zebra finches." Journal of Comparative Neurology 366(4): 613-642.

• Xu, T., X. Yu, A. J. Perlik, W. F. Tobin, J. A. Zweig, K. Tennant, T. Jones and Y. Zuo (2009). "Rapid formation and selective stabilization of synapses for enduring motor memories." Nature 462(7275): 915-919.

• Zhao, W., F. Garcia-Oscos, D. Dinh and T. F. Roberts (2019). "Inception of memories that guide vocal learning in the songbird." Science 366: 83 - 89.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Wang et al., recorded concurrent EEG-fMRI in 107 participants during nocturnal NREM sleep to investigate brain activity and connectivity related to slow oscillations (SO), sleep spindles, and in particular their co-occurrence. The authors found SO-spindle coupling to be correlated with increased thalamic and hippocampal activity, and with increased functional connectivity from the hippocampus to the thalamus and from the thalamus to the neocortex, especially the medial prefrontal cortex (mPFC). They concluded the brain-wide activation pattern to resemble episodic memory processing, but to be dissociated from task-related processing and suggest that the thalamus plays a crucial role in coordinating the hippocampal-cortical dialogue during sleep.

The paper offers an impressively large and highly valuable dataset that provides the opportunity for gaining important new insights into the network substrate involved in SOs, spindles, and their coupling. However, the paper does unfortunately not exploit the full potential of this dataset with the analyses currently provided, and the interpretation of the results is often not backed up by the results presented. I have the following specific comments.

Thank you for your thoughtful and constructive feedback. We greatly appreciate your recognition of the strengths of our dataset and findings Below, we address your specific comments and provide responses to each point you raised to ensure our methods and results are as transparent and comprehensible as possible. We hope these revisions address your comments and further strengthen our manuscript. Thank you again for the constructive feedback.

(1) The introduction is lacking sufficient review of the already existing literature on EEG-fMRI during sleep and the BOLD-correlates of slow oscillations and spindles in particular (Laufs et al., 2007; Schabus et al., 2007; Horovitz et al., 2008; Laufs, 2008; Czisch et al., 2009; Picchioni et al., 2010; Spoormaker et al., 2010; Caporro et al., 2011; Bergmann et al., 2012; Hale et al., 2016; Fogel et al., 2017; Moehlman et al., 2018; Ilhan-Bayrakci et al., 2022). The few studies mentioned are not discussed in terms of the methods used or insights gained.

We acknowledge the need for a more comprehensive review of prior EEG-fMRI studies investigating BOLD correlates of slow oscillations and spindles. However, these articles are not all related to sleep SO or spindle. Articles (Hale et al., 2016; Horovitz et al., 2008; Laufs, 2008; Laufs, Walker, & Lund, 2007; Spoormaker et al., 2010) mainly focus on methodology for EEG-fMRI, sleep stages, or brain networks, which are not the focus of our study. Thank you again for your attention to the comprehensiveness of our literature review, and we will expand the introduction to include a more detailed discussion of the existing literature, ensuring that the contributions of previous EEG-fMRI sleep studies are adequately acknowledged.  

Introduction, Page 4 Lines 62-76

“Investigating these sleep-related neural processes in humans is challenging because it requires tracking transient sleep rhythms while simultaneously assessing their widespread brain activation. Recent advances in simultaneous EEG-fMRI techniques provide a unique opportunity to explore these processes. EEG allows for precise event-based detection of neural signal, while fMRI provides insight into the broader spatial patterns of brain activation and functional connectivity (Horovitz et al., 2008; Huang et al., 2024; Laufs, 2008; Laufs, Walker, & Lund, 2007; Schabus et al., 2007; Spoormaker et al., 2010). Previous EEG-fMRI studies on sleep have focused on classifying sleep stages or examining the neural correlates of specific waves (Bergmann et al., 2012; Caporro et al., 2012; Czisch et al., 2009; Fogel et al., 2017; Hale et al., 2016; Ilhan-Bayrakcı et al., 2022; Moehlman et al., 2019; Picchioni et al., 2011). These studies have generally reported that slow oscillations are associated with widespread cortical and subcortical BOLD changes, whereas spindles elicit activation in the thalamus, as well as in several cortical and paralimbic regions. Although these findings provide valuable insights into the BOLD correlates of sleep rhythms, they often do not employ sophisticated temporal modeling (Huang et al., 2024), to capture the dynamic interactions between different oscillatory events, e.g., the coupling between SOs and spindles.”

(2) The paper falls short in discussing the specific insights gained into the neurobiological substrate of the investigated slow oscillations, spindles, and their interactions. The validity of the inverse inference approach ("Open ended cognitive state decoding"), assuming certain cognitive functions to be related to these oscillations because of the brain regions/networks activated in temporal association with these events, is debatable at best. It is also unclear why eventually only episodic memory processing-like brain-wide activation is discussed further, despite the activity of 16 of 50 feature terms from the NeuroSynth v3 dataset were significant (episodic memory, declarative memory, working memory, task representation, language, learning, faces, visuospatial processing, category recognition, cognitive control, reading, cued attention, inhibition, and action).

Thank you for pointing this out, particularly regarding the use of inverse inference approaches such as “open-ended cognitive state decoding.” Given the concerns about the indirectness of this approach, we decided to remove its related content and results from Figure 3 in the main text and include it in Supplementary Figure 7. We will refocus the main text on direct neurobiological insights gained from our EEG-fMRI analyses, particularly emphasizing the hippocampal-thalamocortical network dynamics underlying SO-spindle coupling, and we will acknowledge the exploratory nature of these findings and highlight their limitations.

Discussion, Page 17-18 Lines 323-332

“To explore functional relevance, we employed an open-ended cognitive state decoding approach using meta-analytic data (NeuroSynth: Yarkoni et al. (2011)). Although this method usefully generates hypotheses about potential cognitive processes, particularly in the absence of a pre- and post-sleep memory task, it is inherently indirect. Many cognitive terms showed significant associations (16 of 50), such as “episodic memory,” “declarative memory,” and “working memory.” We focused on episodic/declarative memory given the known link with hippocampal reactivation (Diekelmann & Born, 2010; Staresina et al., 2015; Staresina et al., 2023). Nonetheless, these inferences regarding memory reactivation should be interpreted cautiously without direct behavioral measures. Future research incorporating explicit tasks before and after sleep would more rigorously validate these potential functional claims.”

(3) Hippocampal activation during SO-spindles is stated as a main hypothesis of the paper - for good reasons - however, other regions (e.g., several cortical as well as thalamic) would be equally expected given the known origin of both oscillations and the existing sleep-EEG-fMRI literature. However, this focus on the hippocampus contrasts with the focus on investigating the key role of the thalamus instead in the Results section.

We appreciate your insight regarding the relative emphasis on hippocampal and thalamic activation in our study. We recognize that the manuscript may currently present an inconsistency between our initial hypothesis and the main focus of the results. To address this concern, we will ensure that our Introduction and Discussion section explicitly discusses both regions, highlighting the complementary roles of the hippocampus (memory processing and reactivation) and the thalamus (spindle generation and cortico-hippocampal coordination) in SO-spindle dynamics.

Introduction, Page 5 Lines 87-103

“To address this gap, our study investigates brain-wide activation and functional connectivity patterns associated with SO-spindle coupling, and employs a cognitive state decoding approach (Margulies et al., 2016; Yarkoni et al., 2011)—albeit indirectly—to infer potential cognitive functions. In the current study, we used simultaneous EEG-fMRI recordings during nocturnal naps (detailed sleep staging results are provided in the Methods and Table S1) in 107 participants. Although directly detecting hippocampal ripples using scalp EEG or fMRI is challenging, we expected that hippocampal activation in fMRI would coincide with SO-spindle coupling detected by EEG, given that SOs, spindles, and ripples frequently co-occur during NREM sleep. We also anticipated a critical role of the thalamus, particularly thalamic spindles, in coordinating hippocampal-cortical communication.

We found significant coupling between SOs and spindles during NREM sleep (N2/3), with spindle peaks occurring slightly before the SO peak. This coupling was associated with increased activation in both the thalamus and hippocampus, with functional connectivity patterns suggesting thalamic coordination of hippocampal-cortical communication. These findings highlight the key role of the thalamus in coordinating hippocampal-cortical interactions during human sleep and provide new insights into the neural mechanisms underlying sleep-dependent brain communication. A deeper understanding of these mechanisms may contribute to future neuromodulation approaches aimed at enhancing sleep-dependent cognitive function and treating sleep-related disorders.”

Discussion, Page 16-17 Lines 292-307

“When modeling the timing of these sleep rhythms in the fMRI, we observed hippocampal activation selectively during SO-spindle events. This suggests the possibility of triple coupling (SOs–spindles–ripples), even though our scalp EEG was not sufficiently sensitive to detect hippocampal ripples—key markers of memory replay (Buzsáki, 2015). Recent iEEG evidence indicates that ripples often co-occur with both spindles (Ngo, Fell, & Staresina, 2020) and SOs (Staresina et al., 2015; Staresina et al., 2023). Therefore, the hippocampal involvement during SO-spindle events in our study may reflect memory replay from the hippocampus, propagated via thalamic spindles to distributed cortical regions.

The thalamus, known to generate spindles (Halassa et al., 2011), plays a key role in producing and coordinating sleep rhythms (Coulon, Budde, & Pape, 2012; Crunelli et al., 2018), while the hippocampus is found essential for memory consolidation (Buzsáki, 2015; Diba & Buzsá ki, 2007; Singh, Norman, & Schapiro, 2022). The increased hippocampal and thalamic activity, along with strengthened connectivity between these regions and the mPFC during SO-spindle events, underscores a hippocampal-thalamic-neocortical information flow. This aligns with recent findings suggesting the thalamus orchestrates neocortical oscillations during sleep (Schreiner et al., 2022). The thalamus and hippocampus thus appear central to memory consolidation during sleep, guiding information transfer to the neocortex, e.g., mPFC.”

(4) The study included an impressive number of 107 subjects. It is surprising though that only 31 subjects had to be excluded under these difficult recording conditions, especially since no adaptation night was performed. Since only subjects were excluded who slept less than 10 min (or had excessive head movements) there are likely several datasets included with comparably short durations and only a small number of SOs and spindles and even less combined SO-spindle events. A comprehensive table should be provided (supplement) including for each subject (included and excluded) the duration of included NREM sleep, number of SOs, spindles, and SO+spindle events. Also, some descriptive statistics (mean/SD/range) would be helpful.

We appreciate your recognition of our sample size and the challenges associated with simultaneous EEG-fMRI sleep recordings. We acknowledge the importance of transparently reporting individual subject data, particularly regarding sleep duration and the number of detected SOs, spindles, and SO-spindle events. To address this, we will provide comprehensive tables in the supplementary materials, contains descriptive information about sleep-related characteristics (Table S1), as well as detailed information about sleep waves at each sleep stage for all 107 subjects(Table S2-S4), listing for each subject:(1)Different sleep stage duration; (2)Number of detected SOs; (3)Number of detected spindles; (4)Number of detected SO-spindle coupling events; (5)Density of detected SOs; (6)Density of detected spindles; (7)Density of detected SO-spindle coupling events.

However, most of the excluded participants were unable to fall asleep or had too short a sleep duration, so they basically had no NREM sleep period, so it was impossible to count the NREM sleep duration, SO, spindle, and coupling numbers.

Supplementary Materials, Page 42-54, Table S1-S4

(5) Was the 20-channel head coil dedicated for EEG-fMRI measurements? How were the electrode cables guided through/out of the head coil? Usually, the 64-channel head coil is used for EEG-fMRI measurements in a Siemens PRISMA 3T scanner, which has a cable duct at the back that allows to guide the cables straight out of the head coil (to minimize MR-related artifacts). The choice for the 20-channel head coil should be motivated. Photos of the recording setup would also be helpful.

Thank you for your comment regarding our choice of the 20-channel head coil for EEG-fMRI measurements. We acknowledge that the 64-channel head coil is commonly used in Siemens PRISMA 3T scanners; however, the 20-channel coil was selected due to specific practical and technical considerations in our study. In particular, the 20-channel head coil was compatible with our EEG system and ensured sufficient signal-to-noise ratio (SNR) for both EEG and fMRI acquisition. The EEG electrode cables were guided through the lateral and posterior openings of the head coil, secured with foam padding to reduce motion and minimize MR-related artifacts. Moreover, given the extended nature of nocturnal sleep recordings, the 20-channel coil allowed us to maintain participant comfort while still achieving high-quality simultaneous EEG-fMRI data.

We have made this clearer in the revised manuscript. 

Methods, Page 20 Lines 385-392

“All MRI data were acquired using a 20-channel head coil on a research-dedicated 3-Tesla Siemens Magnetom Prisma MRI scanner. Earplugs and cushions were provided for noise protection and head motion restriction. We chose the 20-channel head coil because it was compatible with our EEG system and ensured sufficient signal-to-noise ratio (SNR) for both EEG and fMRI acquisition. The EEG electrode cables were guided through the lateral and posterior openings of the head coil, secured with foam padding to reduce motion and minimize MR-related artifacts. Moreover, given the extended nature of nocturnal sleep recordings, the 20-channel coil helped maintain participant comfort while still achieving high-quality simultaneous EEG-fMRI data.”

(6) Was the EEG sampling synchronized to the MR scanner (gradient system) clock (the 10 MHz signal; not referring to the volume TTL triggers here)? This is a requirement for stable gradient artifact shape over time and thus accurate gradient noise removal.

Thank you for raising this important point. We confirm that the EEG sampling was synchronized to the MR scanner’s 10 MHz gradient system clock, ensuring a stable gradient artifact shape over time and enabling accurate artifact removal. This synchronization was achieved using the standard clock synchronization interface of the EEG amplifier, minimizing timing jitter and drift. As a result, the gradient artifact waveform remained stable across volumes, allowing for more effective artifact correction during preprocessing. We appreciate your attention to this critical aspect of EEG-fMRI data acquisition.

We have made this clearer in the revised manuscript. 

Methods, Page 19-20 Lines 371-383

“EEG was recorded simultaneously with fMRI data using an MR-compatible EEG amplifier system (BrainAmps MR-Plus, Brain Products, Germany), along with a specialized electrode cap. The recording was done using 64 channels in the international 10/20 system, with the reference channel positioned at FCz. In order to adhere to polysomnography (PSG) recording standards, six electrodes were removed from the EEG cap: one for electrocardiogram (ECG) recording, two for electrooculogram (EOG) recording, and three for electromyogram (EMG) recording. EEG data was recorded at a sample rate of 5000 Hz, the resistance of the reference and ground channels was kept below 10 kΩ, and the resistance of the other channels was kept below 20 kΩ. To synchronize the EEG and fMRI recordings, the BrainVision recording software (BrainProducts, Germany) was utilized to capture triggers from the MRI scanner. The EEG sampling was synchronized to the MR scanner’s 10 MHz gradient system clock, ensuring a stable gradient artifact shape over time and enabling accurate artifact removal. This was achieved via the standard clock synchronization interface of the EEG amplifier, minimizing timing jitter and drift.”

(7) The TR is quite long and the voxel size is quite large in comparison to state-of-the-art EPI sequences. What was the rationale behind choosing a sequence with relatively low temporal and spatial resolution?

We acknowledge that our chosen TR and voxel size are relatively long and large compared to state-of-the-art EPI sequences. This decision was made to optimize the signal-to-noise ratio (SNR) and reduce susceptibility-related distortions, which are particularly critical in EEG-fMRI sleep studies where head motion and physiological noise can be substantial. A longer TR allowed us to sample whole-brain activity with sufficient coverage, while a larger voxel size helped enhance BOLD sensitivity and minimize partial volume effects in deep brain structures such as the thalamus and hippocampus, which are key regions of interest in our study. We appreciate your concern and hope this clarification provides sufficient rationale for our sequence parameters.

We have made this clearer in the revised manuscript. 

Methods, Page 20-21 Lines 398-408

“Then, the “sleep” session began after the participants were instructed to try and fall asleep. For the functional scans, whole-brain images were acquired using k-space and steady-state T2*-weighted gradient echo-planar imaging (EPI) sequence that is sensitive to the BOLD contrast. This measures local magnetic changes caused by changes in blood oxygenation that accompany neural activity (sequence specification: 33 slices in interleaved ascending order, TR = 2000 ms, TE = 30 ms, voxel size = 3.5 × 3.5 × 4.2 mm3, FA = 90°, matrix = 64 × 64, gap = 0.7 mm). A relatively long TR and larger voxel size were chosen to optimize SNR and reduce susceptibility-related distortions, which are critical in EEG-fMRI sleep studies where head motion and physiological noise can be substantial. The longer TR allowed whole-brain coverage with sufficient temporal resolution, while the larger voxel size helped enhance BOLD sensitivity and minimize partial volume effects in deep brain structures (e.g., the thalamus and hippocampus), which are key regions of interest in this study.”

(8) The anatomically defined ROIs are quite large. It should be elaborated on how this might reduce sensitivity to sleep rhythm-specific activity within sub-regions, especially for the thalamus, which has distinct nuclei involved in sleep functions.

We appreciate your insight regarding the use of anatomically defined ROIs and their potential limitations in detecting sleep rhythm-specific activity within sub-regions, particularly in the thalamus. Given the distinct functional roles of thalamic nuclei in sleep processes, we acknowledge that using a single, large thalamic ROI may reduce sensitivity to localized activity patterns. To address this, we will discuss this limitation in the revised manuscript, acknowledging that our approach prioritizes whole-structure effects but may not fully capture nucleus-specific contributions.

Discussion, Page 18 Lines 333-341

“Despite providing new insights, our study has several limitations. First, our scalp EEG did not directly capture hippocampal ripples, preventing us from conclusively demonstrating triple coupling. Second, the combination of EEG-fMRI and the lack of a memory task limit our ability to parse fine-grained BOLD responses at the DOWN- vs. UP-states of SOs and link observed activations to behavioral outcomes. Third, the use of large anatomical ROIs may mask subregional contributions of specific thalamic nuclei or hippocampal subfields. Finally, without a memory task, we cannot establish a direct behavioral link between sleep-rhythm-locked activation and memory consolidation. Future studies combining techniques such as ultra-high-field fMRI or iEEG with cognitive tasks may refine our understanding of subregional network dynamics and functional significance during sleep.”

(9) The study reports SO & spindle amplitudes & densities, as well as SO+spindle coupling, to be larger during N2/3 sleep compared to N1 and REM sleep, which is trivial but can be seen as a sanity check of the data. However, the amount of SOs and spindles reported for N1 and REM sleep is concerning, as per definition there should be hardly any (if SOs or spindles occur in N1 it becomes by definition N2, and the interval between spindles has to be considerably large in REM to still be scored as such). Thus, on the one hand, the report of these comparisons takes too much space in the main manuscript as it is trivial, but on the other hand, it raises concerns about the validity of the scoring.

We appreciate your concern regarding the reported presence of SOs and spindles in N1 and REM sleep and the potential implications. Our detection method for detecting SO, spindle, and coupling were originally designed only for N2&N3 sleep data based on the characteristics of the data itself, and this method is widely recognized and used in the sleep research (Hahn et al., 2020; Helfrich et al., 2019; Helfrich et al., 2018; Ngo, Fell, & Staresina, 2020; Schreiner et al., 2022; Schreiner et al., 2021; Staresina et al., 2015; Staresina et al., 2023). While, because the detection methods for SO and spindle are based on percentiles, this method will always detect a certain number of events when used for other stages (N1 and REM) sleep data, but the differences between these events and those detected in stage N23 remain unclear. We will acknowledge the reasons for these results in the Methods section and emphasize that they are used only for sanity checks.

Methods, Page 25 Lines 515-524

“We note that the above methods for detecting SOs, spindles, and their couplings were originally developed for N2 and N3 sleep data, based on the specific characteristics of these stages. These methods are widely recognized in sleep research (Hahn et al., 2020; Helfrich et al., 2019; Helfrich et al., 2018; Ngo, Fell, & Staresina, 2020; Schreiner et al., 2022; Schreiner et al., 2021; Staresina et al., 2015; Staresina et al., 2023). However, because this percentile-based detection approach will inherently identify a certain number of events if applied to other stages (e.g., N1 and REM), the nature of these events in those stages remains unclear compared to N2/N3. We nevertheless identified and reported the detailed descriptive statistics of these sleep rhythms in all sleep stages, under the same operational definitions, both for completeness and as a sanity check. Within the same subject, there should be more SOs, spindles, and their couplings in N2/N3 than in N1 or REM (see also Figure S2-S4, Table S1-S4).”

(10) Why was electrode F3 used to quantify the occurrence of SOs and spindles? Why not a midline frontal electrode like Fz (or a number of frontal electrodes for SOs) and Cz (or a number of centroparietal electrodes) for spindles to be closer to their maximum topography?

We appreciate your suggestion regarding electrode selection for SO and spindle quantification. Our choice of F3 was primarily based on previous studies (Massimini et al., 2004; Molle et al., 2011), where bilateral frontal electrodes are commonly used for detecting SOs and spindles. Additionally, we considered the impact of MRI-related noise and, after a comprehensive evaluation, determined that F3 provided an optimal balance between signal quality and artifact minimization. We also acknowledge that alternative electrode choices, such as Fz for SOs and Cz for spindles, could provide additional insights into their topographical distributions.

(11) Functional connectivity (hippocampus -> thalamus -> cortex (mPFC)) is reported to be increased during SO-spindle coupling and interpreted as evidence for coordination of hippocampo-neocortical communication likely by thalamic spindles. However, functional connectivity was only analysed during coupled SO+spindle events, not during isolated SOs or isolated spindles. Without the direct comparison of the connectivity patterns between these three events, it remains unclear whether this is specific for coupled SO+spindle events or rather associated with one or both of the other isolated events. The PPIs need to be conducted for those isolated events as well and compared statistically to the coupled events.

We appreciate your critical perspective on our functional connectivity analysis and the interpretation of hippocampus-thalamus-cortex (mPFC) interactions during SO-spindle coupling. We acknowledge that, in the current analysis, functional connectivity was only examined during coupled SO-spindle events, without direct comparison to isolated SOs or isolated spindles. To address this concern, we have conducted PPI analyses for all three ROIs(Hippocampus, Thalamus, mPFC) and all three event types (SO-spindle couplings, isolated SOs, and isolated spindles). Our results indicate that neither isolated SOs nor isolated Spindles yielded significant connectivity changes in all three ROIs, as all failed to survive multiple comparison corrections. This suggests that the observed connectivity increase is specific to SO-spindle coupling, rather than being independently driven by either SOs or spindles alone.

Results, Page 14 Lines 248-255

“Crucially, the interaction between FC and SO-spindle coupling revealed that only the functional connectivity of hippocampus -> thalamus (ROI analysis, t(106) = 1.86, p = 0.0328) and thalamus -> mPFC (ROI analysis, t(106) = 1.98, p = 0.0251) significantly increased during SO-spindle coupling, with no significant changes in all other pathways (Fig. 4e). We also conducted PPI analyses for the other two events (SOs and spindles), and neither yielded significant connectivity changes in the three ROIs, as all failed to survive whole-brain FWE correction at the cluster level (p < 0.05). Together, these findings suggest that the thalamus, likely via spindles, coordinates hippocampal-cortical communication selectively during SO-spindle coupling, but not isolated SOs or spindle events alone.”

(12) The limited temporal resolution of fMRI does indeed not allow for easily distinguishing between fMRI activation patterns related to SO-up- vs. SO-down-states. For this, one could try to extract the amplitudes of SO-up- and SO-down-states separately for each SO event and model them as two separate parametric modulators (with the risk of collinearity as they are likely correlated).

We appreciate your insightful comment regarding the challenge of distinguishing fMRI activation patterns related to SO-up vs. SO-down states due to the limited temporal resolution of fMRI. While our current analysis does not differentiate between these two phases, we acknowledge that separately modeling SO-up and SO-down states using parametric modulators could provide a more refined understanding of their distinct neural correlates. However, as you notes, this approach carries the risk of collinearity, and there is indeed a high correlation between the two amplitudes across all subjects in our results (r=0.98). Future studies could explore more on leveraging high-temporal-resolution techniques. While implementing this in the current study is beyond our scope, we will acknowledge this limitation in the Discussion section.

Discussion, Page 17 Lines 308-322

“An intriguing aspect of our findings is the reduced DMN activity during SOs when modeled at the SO trough (DOWN-state). This reduced DMN activity may reflect large-scale neural inhibition characteristic of the SO trough. The DMN is typically active during internally oriented cognition (e.g., self-referential processing or mind-wandering) and is suppressed during external stimuli processing (Yeshurun, Nguyen, & Hasson, 2021). It is unlikely, however, that this suppression of DMN during SO events is related to a shift from internal cognition to external responses given it is during deep sleep time. Instead, it could be driven by the inherent rhythmic pattern of SOs, which makes it difficult to separate UP- from DOWN-states (the two temporal regressors were highly correlated, and similar brain activation during SOs events was obtained if modelled at the SO peak instead, Fig. S5). Since the amplitude at the SO trough is consistently larger than that at the SO peak, the neural activation we detected may primarily capture the large-scale inhibition from DOWN-state. Interestingly, no such DMN reduction was found during SO-spindle coupling, implying that coupling may involve distinct neural dynamics that partially re-engage DMN-related processes, possibly reflecting memory-related reactivation. Future research using high-temporal-resolution techniques like iEEG could clarify these possibilities.”

Discussion, Page 18 Lines 333-341

“Despite providing new insights, our study has several limitations. First, our scalp EEG did not directly capture hippocampal ripples, preventing us from conclusively demonstrating triple coupling. Second, the combination of EEG-fMRI and the lack of a memory task limit our ability to parse fine-grained BOLD responses at the DOWN- vs. UP-states of SOs and link observed activations to behavioral outcomes. Third, the use of large anatomical ROIs may mask subregional contributions of specific thalamic nuclei or hippocampal subfields. Finally, without a memory task, we cannot establish a direct behavioral link between sleep-rhythm-locked activation and memory consolidation. Future studies combining techniques such as ultra-high-field fMRI or iEEG with cognitive tasks may refine our understanding of subregional network dynamics and functional significance during sleep.”

(13) L327: "It is likely that our findings of diminished DMN activity reflect brain activity during the SO DOWN-state, as this state consistently shows higher amplitude compared to the UP-state within subjects, which is why we modelled the SO trough as its onset in the fMRI analysis." This conclusion is not justified as the fact that SO down-states are larger in amplitude does not mean their impact on the BOLD response is larger.

We appreciate your concern regarding our interpretation of diminished DMN activity reflecting the SO down-state. We acknowledge that the current expression is somewhat misleading, and our interpretation of it is: it could be driven by the inherent rhythmic pattern of SOs, which makes it difficult to separate UP- from DOWN-states (the two temporal regressors were highly correlated, and similar brain activation during SOs events was obtained if modelled at the SO peak instead). Since the amplitude at the SO trough is consistently larger than that at the SO peak, the neural activation we detected may primarily capture the large-scale inhibition from DOWN-state. And we will make this clear in the Discussion section.

Discussion, Page 17 Lines 308-322

“An intriguing aspect of our findings is the reduced DMN activity during SOs when modeled at the SO trough (DOWN-state). This reduced DMN activity may reflect large-scale neural inhibition characteristic of the SO trough. The DMN is typically active during internally oriented cognition (e.g., self-referential processing or mind-wandering) and is suppressed during external stimuli processing (Yeshurun, Nguyen, & Hasson, 2021). It is unlikely, however, that this suppression of DMN during SO events is related to a shift from internal cognition to external responses given it is during deep sleep time. Instead, it could be driven by the inherent rhythmic pattern of SOs, which makes it difficult to separate UP- from DOWN-states (the two temporal regressors were highly correlated, and similar brain activation during SOs events was obtained if modelled at the SO peak instead, Fig. S5). Since the amplitude at the SO trough is consistently larger than that at the SO peak, the neural activation we detected may primarily capture the large-scale inhibition from DOWN-state. Interestingly, no such DMN reduction was found during SO-spindle coupling, implying that coupling may involve distinct neural dynamics that partially re-engage DMN-related processes, possibly reflecting memory-related reactivation. Future research using high-temporal-resolution techniques like iEEG could clarify these possibilities.”

(14) Line 77: "In the current study, while directly capturing hippocampal ripples with scalp EEG or fMRI is difficult, we expect to observe hippocampal activation in fMRI whenever SOs-spindles coupling is detected by EEG, if SOs- spindles-ripples triple coupling occurs during human NREM sleep". Not all SO-spindle events are associated with ripples (Staresina et al., 2015), but hippocampal activation may also be expected based on the occurrence of spindles alone (Bergmann et al., 2012).

We appreciate your clarification regarding the relationship between SO-spindle coupling and hippocampal ripples. We acknowledge that not all SO-spindle events are necessarily accompanied by ripples (Staresina et al., 2015). However, based on previous research, we found that hippocampal ripples are significantly more likely to occur during SO-spindle coupling events. This suggests that while ripple occurrence is not guaranteed, SO-spindle coupling creates a favorable network state for ripple generation and potential hippocampal activation. To ensure accuracy, we will revise the manuscript to delete this misleading sentence in the Introduction section and acknowledge in the Discussion that our results cannot conclusively directly observe the triple coupling of SO, spindle, and hippocampal ripples.

Discussion, Page 18 Lines 333-341

“Despite providing new insights, our study has several limitations. First, our scalp EEG did not directly capture hippocampal ripples, preventing us from conclusively demonstrating triple coupling. Second, the combination of EEG-fMRI and the lack of a memory task limit our ability to parse fine-grained BOLD responses at the DOWN- vs. UP-states of SOs and link observed activations to behavioral outcomes. Third, the use of large anatomical ROIs may mask subregional contributions of specific thalamic nuclei or hippocampal subfields. Finally, without a memory task, we cannot establish a direct behavioral link between sleep-rhythm-locked activation and memory consolidation. Future studies combining techniques such as ultra-high-field fMRI or iEEG with cognitive tasks may refine our understanding of subregional network dynamics and functional significance during sleep.”

Reviewer #2 (Public review):

In this study, Wang and colleagues aimed to explore brain-wide activation patterns associated with NREM sleep oscillations, including slow oscillations (SOs), spindles, and SO-spindle coupling events. Their findings reveal that SO-spindle events corresponded with increased activation in both the thalamus and hippocampus. Additionally, they observed that SO-spindle coupling was linked to heightened functional connectivity from the hippocampus to the thalamus, and from the thalamus to the medial prefrontal cortex-three key regions involved in memory consolidation and episodic memory processes.

This study's findings are timely and highly relevant to the field. The authors' extensive data collection, involving 107 participants sleeping in an fMRI while undergoing simultaneous EEG recording, deserves special recognition. If shared, this unique dataset could lead to further valuable insights. While the conclusions of the data seem overall well supported by the data, some aspects with regard to the detection of sleep oscillations need clarification.

The authors report that coupled SO-spindle events were most frequent during NREM sleep (2.46 [plus minus] 0.06 events/min), but they also observed a surprisingly high occurrence of these events during N1 and REM sleep (2.23 [plus minus] 0.09 and 2.32 [plus minus] 0.09 events/min, respectively), where SO-spindle coupling would not typically be expected. Combined with the relatively modest SO amplitudes reported (~25 µV, whereas >75 µV would be expected when using mastoids as reference electrodes), this raises the possibility that the parameters used for event detection may not have been conservative enough - or that sleep staging was inaccurately performed. This issue could present a significant challenge, as the fMRI findings are largely dependent on the reliability of these detected events.

Thank you very much for your thorough and encouraging review. We appreciate your recognition of the significance and relevance of our study and dataset, particularly in highlighting how simultaneous EEG-fMRI recordings can provide complementary insights into the temporal dynamics of neural oscillations and their associated spatial activation patterns during sleep. In the sections that follow, we address each of your comments in detail. We have revised the text and conducted additional analyses wherever possible to strengthen our argument, clarify our methodological choices. We believe these revisions improve the clarity and rigor of our work, and we thank you for helping us refine it.

We appreciate your insightful comments regarding the detection of sleep oscillations. Our methods for detecting SOs, spindles, and their couplings were originally developed for N2 and N3 sleep data, based on the specific characteristics of these stages. These methods are widely recognized in sleep research (Hahn et al., 2020; Helfrich et al., 2019; Helfrich et al., 2018; Ngo, Fell, & Staresina, 2020; Schreiner et al., 2022; Schreiner et al., 2021; Staresina et al., 2015; Staresina et al., 2023). However, because this percentile-based detection approach will inherently identify a certain number of events if applied to other stages (e.g., N1 and REM), the nature of these events in those stages remains unclear compared to N2/N3. We nevertheless identified and reported the detailed descriptive statistics of these sleep rhythms in all sleep stages, under the same operational definitions, both for completeness and as a sanity check. Within the same subject, there should be more SOs, spindles, and their couplings in N2/N3 than in N1 or REM. We will acknowledge the reasons for these results in the Methods section and emphasize that they are used only for sanity checks.

Regarding the reported SO amplitudes (~25 µV), during preprocessing, we applied the Signal Space Projection (SSP) method to more effectively remove MRI gradient artifacts and cardiac pulse noise. While this approach enhances data quality, it also reduces overall signal power, leading to systematically lower reported amplitudes. Despite this, our SO detection in NREM sleep (especially N2/N3) remain physiologically meaningful and are consistent with previous fMRI studies using similar artifact removal techniques. We appreciate your careful evaluation and valuable suggestions.

In addition, we will provide comprehensive tables in the supplementary materials, contains descriptive information about sleep-related characteristics (Table S1), as well as detailed information about sleep waves at each sleep stage for all 107 subjects(Table S2-S4), listing for each subject:(1)Different sleep stage duration; (2)Number of detected SOs; (3)Number of detected spindles; (4)Number of detected SO-spindle coupling events; (2)Density of detected SOs; (3)Density of detected spindles; (4)Density of detected SO-spindle coupling events.

Methods, Page 25 Lines 515-524

“We note that the above methods for detecting SOs, spindles, and their couplings were originally developed for N2 and N3 sleep data, based on the specific characteristics of these stages. These methods are widely recognized in sleep research (Hahn et al., 2020; Helfrich et al., 2019; Helfrich et al., 2018; Ngo, Fell, & Staresina, 2020; Schreiner et al., 2022; Schreiner et al., 2021; Staresina et al., 2015; Staresina et al., 2023). However, because this percentile-based detection approach will inherently identify a certain number of events if applied to other stages (e.g., N1 and REM), the nature of these events in those stages remains unclear compared to N2/N3. We nevertheless identified and reported the detailed descriptive statistics of these sleep rhythms in all sleep stages, under the same operational definitions, both for completeness and as a sanity check. Within the same subject, there should be more SOs, spindles, and their couplings in N2/N3 than in N1 or REM (see also Figure S2-S4, Table S1-S4).”

Supplementary Materials, Page 42-54, Table S1-S4

Reviewer #3 (Public review):

Summary:

Wang et al., examined the brain activity patterns during sleep, especially when locked to those canonical sleep rhythms such as SO, spindle, and their coupling. Analyzing data from a large sample, the authors found significant coupling between spindles and SOs, particularly during the upstate of the SO. Moreover, the authors examined the patterns of whole-brain activity locked to these sleep rhythms. To understand the functional significance of these brain activities, the authors further conducted open-ended cognitive state decoding and found a variety of cognitive processing may be involved during SO-spindle coupling and during other sleep events. The authors next investigated the functional connectivity analyses and found enhanced connectivity between the hippocampus, the thalamus, and the medial PFC. These results reinforced the theoretical model of sleep-dependent memory consolidation, such that SO-spindle coupling is conducive to systems-level memory reactivation and consolidation.

Strengths:

There are obvious strengths in this work, including the large sample size, state-of-the-art neuroimaging and neural oscillation analyses, and the richness of results.

Weaknesses:

Despite these strengths and the insights gained, there are weaknesses in the design, the analyses, and inferences.

Thank you for your detailed and thoughtful review of our manuscript. We are delighted that you recognize our advanced analysis methods and rich results of neuroimaging and neural oscillations as well as the large sample size data. In the following sections, we provide detailed responses to each of your comments. And we have revised the text and conducted additional analyses to strengthen our arguments and clarify our methodological choices. We believe these revisions enhance the clarity and rigor of our work, and we sincerely appreciate your thoughtful feedback in helping us refine the manuscript.

(1) A repeating statement in the manuscript is that brain activity could indicate memory reactivation and thus consolidation. This is indeed a highly relevant question that could be informed by the current data/results. However, an inherent weakness of the design is that there is no memory task before and after sleep. Thus, it is difficult (if not impossible) to make a strong argument linking SO/spindle/coupling-locked brain activity with memory reactivation or consolidation.

We appreciate your suggestion regarding the lack of a pre- and post-sleep memory task in our study design. We acknowledge that, in the absence of behavioral measures, it is hard to directly link SO-spindle coupling to memory consolidation in an outcome-driven manner. Our interpretation is instead based on the well-established role of these oscillations in memory processes, as demonstrated in previous studies. We sincerely appreciate this feedback and will adjust our Discussion accordingly to reflect a more precise interpretation of our findings.

Discussion, Page 18 Lines 333-341

“Despite providing new insights, our study has several limitations. First, our scalp EEG did not directly capture hippocampal ripples, preventing us from conclusively demonstrating triple coupling. Second, the combination of EEG-fMRI and the lack of a memory task limit our ability to parse fine-grained BOLD responses at the DOWN- vs. UP-states of SOs and link observed activations to behavioral outcomes. Third, the use of large anatomical ROIs may mask subregional contributions of specific thalamic nuclei or hippocampal subfields. Finally, without a memory task, we cannot establish a direct behavioral link between sleep-rhythm-locked activation and memory consolidation. Future studies combining techniques such as ultra-high-field fMRI or iEEG with cognitive tasks may refine our understanding of subregional network dynamics and functional significance during sleep.”

(2) Relatedly, to understand the functional implications of the sleep rhythm-locked brain activity, the authors employed the "open-ended cognitive state decoding" method. While this method is interesting, it is rather indirect given that there were no behavioral indices in the manuscript. Thus, discussions based on these analyses are speculative at best. Please either tone down the language or find additional evidence to support these claims.

Moreover, the results from this method are difficult to understand. Figure 3e showed that for all three types of sleep events (SO, spindle, SO-spindle), the same mental states (e.g., working memory, episodic memory, declarative memory) showed opposite directions of activation (left and right panels showed negative and positive activation, respectively). How to interpret these conflicting results? This ambiguity is also reflected by the term used: declarative memory and episodic memories are both indexed in the results. Yet these two processes can be largely overlapped. So which specific memory processes do these brain activity patterns reflect? The Discussion shall discuss these results and the limitations of this method.

We appreciate your critical assessment of the open-ended cognitive state decoding method and its interpretational challenges. Given the concerns about the indirectness of this approach, we decided to remove its related content and results from Figure 3 in the main text and include it in Supplementary Figure 7. 

Due to the complexity of memory-related processes, we acknowledge that distinguishing between episodic and declarative memory based solely on this approach is not straightforward. We will revise the Supplementary Materials to explicitly discuss these limitations and clarify that our findings do not isolate specific cognitive processes but rather suggest general associations with memory-related networks.

Discussion, Page 17-18 Lines 323-332

“To explore functional relevance, we employed an open-ended cognitive state decoding approach using meta-analytic data (NeuroSynth: Yarkoni et al. (2011)). Although this method usefully generates hypotheses about potential cognitive processes, particularly in the absence of a pre- and post-sleep memory task, it is inherently indirect. Many cognitive terms showed significant associations (16 of 50), such as “episodic memory,” “declarative memory,” and “working memory.” We focused on episodic/declarative memory given the known link with hippocampal reactivation (Diekelmann & Born, 2010; Staresina et al., 2015; Staresina et al., 2023). Nonetheless, these inferences regarding memory reactivation should be interpreted cautiously without direct behavioral measures. Future research incorporating explicit tasks before and after sleep would more rigorously validate these potenial functional claims.”

(3) The coupling strength is somehow inconsistent with prior results (Hahn et al., 2020, eLife, Helfrich et al., 2018, Neuron). Specifically, Helfrich et al. showed that among young adults, the spindle is coupled to the peak of the SO. Here, the authors reported that the spindles were coupled to down-to-up transitions of SO and before the SO peak. It is possible that participants' age may influence the coupling (see Helfrich et al., 2018). Please discuss the findings in the context of previous research on SO-spindle coupling.

We appreciate your concern regarding the temporal characteristics of SO-spindle coupling. We acknowledge that the SO-spindle coupling phase results in our study are not identical to those reported by Hahn et al. (2020); Helfrich et al. (2018). However, these differences may arise due to slight variations in event detection parameters, which can influence the precise phase estimation of coupling. Notably, Hahn et al. (2020) also reported slight discrepancies in their group-level coupling phase results, highlighting that methodological differences can contribute to variability across studies. Furthermore, our findings are consistent with those of Schreiner et al. (2021), further supporting the robustness of our observations.  

That said, we acknowledge that our original description of SO-spindle coupling as occurring at the "transition from the lower state to the upper state" was not entirely precise. The -π/2 phase represents the true transition point, while our observed coupling phase is actually closer to the SO peak rather than strictly at the transition. We will revise this statement in the manuscript to ensure clarity and accuracy in describing the coupling phase.  

Discussion, Page 16 Lines 283-291

“Our data provide insights into the neurobiological underpinnings of these sleep rhythms. SOs, originating mainly in neocortical areas such as the mPFC, alternate between DOWN- and UP-states. The thalamus generates sleep spindles, which in turn couple with SOs. Our finding that spindle peaks consistently occurred slightly before the UP-state peak of SOs (in 83 out of 107 participants), concurs with prior studies, including Schreiner et al. (2021). Yet it differs from some results suggesting spindles might peak right at the SO UP-state (Hahn et al., 2020; Helfrich et al., 2018). Such discrepancies could arise from differences in detection algorithms, participant age (Helfrich et al., 2018), or subtle variations in cortical-thalamic timing. Nonetheless, these results underscore the importance of coordinated SO-spindle interplay in supporting sleep-dependent processes.”

(4) The discussion is rather superficial with only two pages, without delving into many important arguments regarding the possible functional significance of these results. For example, the author wrote, "This internal processing contrasts with the brain patterns associated with external tasks, such as working memory." Without any references to working memory, and without delineating why WM is considered as an external task even working memory operations can be internal. Similarly, for the interesting results on SO and reduced DMN activity, the authors wrote "The DMN is typically active during wakeful rest and is associated with self-referential processes like mind-wandering, daydreaming, and task representation (Yeshurun, Nguyen, & Hasson, 2021). Its reduced activity during SOs may signal a shift towards endogenous processes such as memory consolidation." This argument is flawed. DMN is active during self-referential processing and mind-wandering, i.e., when the brain shifts from external stimuli processing to internal mental processing. During sleep, endogenous memory reactivation and consolidation are also part of the internal mental processing given the lack of external environmental stimulation. So why during SO or during memory consolidation, the DMN activity would be reduced? Were there differences in DMN activity between SO and SO-spindle coupling events?

We appreciate your concerns regarding the brevity of the discussion and the need for clearer theoretical arguments. We will expand this section to provide more in-depth interpretations of our findings in the context of prior literature. Regarding working memory (WM), we acknowledge that our phrasing was ambiguous. We will modify this statement in the Discussion section.

For the SO-related reduction in DMN activity, we recognize the need for a more precise explanation. This reduced DMN activity may reflect large-scale neural inhibition characteristic of the SO trough. The DMN is typically active during internally oriented cognition (e.g., self-referential processing or mind-wandering) and is suppressed during external stimuli processing (Yeshurun, Nguyen, & Hasson, 2021). It is unlikely, however, that this suppression of DMN during SO events is related to a shift from internal cognition to external responses given it is during deep sleep time. Instead, it could be driven by the inherent rhythmic pattern of SOs, which makes it difficult to separate UP- from DOWN-states (the two temporal regressors were highly correlated, and similar brain activation during SOs events was obtained if modelled at the SO peak instead). Since the amplitude at the SO trough is consistently larger than that at the SO peak, the neural activation we detected may primarily capture the large-scale inhibition from DOWN-state.

To address your final question, we have conducted the additional post hoc comparison of DMN activity between isolated SOs and SO-spindle coupling events. Our results indicate that

DMN activation during SOs was significantly lower than during SO-spindle coupling (t(106) = -4.17, p < 1e-4). This suggests that SO-spindle coupling may involve distinct neural dynamics that partially re-engage DMN-related processes, possibly reflecting memory-related reactivation. We appreciate your constructive feedback and will integrate these expanded analyses and discussions into our revised manuscript.

Results, Page 11 Lines 199-208

“Spindles were correlated with positive activation in the thalamus (ROI analysis, t(106) = 15.39, p < 1e-4), the anterior cingulate cortex (ACC), and the putamen, alongside deactivation in the DMN (Fig. 3c). Notably, SO-spindle coupling was linked to significant activation in both the thalamus (ROI analysis, t(106) \= 3.38, p = 0.0005) and the hippocampus (ROI analysis, t(106) \= 2.50, p = 0.0070, Fig. 3d). However, no decrease in DMN activity was found during SO-spindle coupling, and DMN activity during SO was significantly lower than during coupling (ROI analysis, t(106) \= -4.17, p < 1e-4). For more detailed activation patterns, see Table S5-S7. We also varied the threshold used to detect SO events to assess its effect on hippocampal activation during SO-spindle coupling and observed that hippocampal activation remained significant when the percentile thresholds for SO detection ranged between 71% and 80% (see Fig. S6).”

Discussion, Page 17-18 Lines 308-332

“An intriguing aspect of our findings is the reduced DMN activity during SOs when modeled at the SO trough (DOWN-state). This reduced DMN activity may reflect large-scale neural inhibition characteristic of the SO trough. The DMN is typically active during internally oriented cognition (e.g., self-referential processing or mind-wandering) and is suppressed during external stimuli processing (Yeshurun, Nguyen, & Hasson, 2021). It is unlikely, however, that this suppression of DMN during SO events is related to a shift from internal cognition to external responses given it is during deep sleep time. Instead, it could be driven by the inherent rhythmic pattern of SOs, which makes it difficult to separate UP- from DOWN-states (the two temporal regressors were highly correlated, and similar brain activation during SOs events was obtained if modelled at the SO peak instead, Fig. S5). Since the amplitude at the SO trough is consistently larger than that at the SO peak, the neural activation we detected may primarily capture the large-scale inhibition from DOWN-state. Interestingly, no such DMN reduction was found during SO-spindle coupling, implying that coupling may involve distinct neural dynamics that partially re-engage DMN-related processes, possibly reflecting memory-related reactivation. Future research using high-temporal-resolution techniques like iEEG could clarify these possibilities.

To explore functional relevance, we employed an open-ended cognitive state decoding approach using meta-analytic data (NeuroSynth: Yarkoni et al. (2011)). Although this method usefully generates hypotheses about potential cognitive processes, particularly in the absence of a pre- and post-sleep memory task, it is inherently indirect. Many cognitive terms showed significant associations (16 of 50), such as “episodic memory,” “declarative memory,” and “working memory.” We focused on episodic/declarative memory given the known link with hippocampal reactivation (Diekelmann & Born, 2010; Staresina et al., 2015; Staresina et al., 2023). Nonetheless, these inferences regarding memory reactivation should be interpreted cautiously without direct behavioral measures. Future research incorporating explicit tasks before and after sleep would more rigorously validate these potential functional claims.”

Recommendations for the authors:

Reviewing Editor Comment:

The reviewers think that you are working on a relevant and important topic. They are praising the large sample size used in the study. The reviewers are not all in line regarding the overall significance of the findings, but they all agree the paper would strongly benefit from some extra work, as all reviewers raise various critical points that need serious consideration.

We appreciate your recognition of the relevance and importance of our study, as well as your acknowledgment of the large sample size as a strength of our work. We understand that there are differing perspectives regarding the overall significance of our findings, and we value the constructive critiques provided. We are committed to addressing the key concerns raised by all reviewers, including refining our analyses, clarifying our interpretations, and incorporating additional discussions to strengthen the manuscript. Below, we address your specific recommendations and provide responses to each point you raised to ensure our methods and results are as transparent and comprehensible as possible. We believe that these revisions will significantly enhance the rigor and impact of our study, and we sincerely appreciate your thoughtful feedback in helping us improve our work.

Reviewer #1 (Recommendations for the authors):

(1) The phrase "overnight sleep" suggests an entire night, while these were rather "nocturnal naps". Please rephrase.

Response: Thank you for pointing this out. We have revised the phrasing in our manuscript to "nocturnal naps" instead of "overnight sleep" to more accurately reflect the duration of the sleep recordings.

(2) Sleep staging results (macroscopic sleep architecture) should be provided in more detail (at least min and % of the different sleep stages, sleep onset latency, total sleep duration, total recording duration), at least mean/SD/range.

Thank you for this suggestion. We will provide comprehensive tables in the supplementary materials, contains descriptive information about sleep-related characteristics. This information will help provide a clearer overview of the macroscopic sleep architecture in our dataset.

Reviewer #2 (Recommendations for the authors):

In order to allow for a better estimation of the reliability of the detected sleep events, please:

(1) Provide densities and absolute numbers of all detected SOs and spindles (N1, NREM, and REM sleep).

Thank you for pointing this out. We will provide comprehensive tables in the supplementary materials, contains detailed information about sleep waves at each sleep stage for all 107 subjects (Table S2-S4), listing for each subject:1) Different sleep stage duration; 2) Number of detected SOs; 3) Number of detected spindles; 4) Number of detected SO-spindle coupling events; 5) Density of detected SOs; 6) Density of detected spindles; 7) Density of detected SO-spindle coupling events.

Supplementary Materials, Page 43-54, Table S2-S4

(2) Show ERPs for all detected SOs and spindles (per sleep stage).

Thank you for the suggestion. We will provide ERPs for all detected SOs and spindles, separated by sleep stage (N1, N2&N3, and REM) in supplementary Fig. S2-S4. These ERP waveforms will help illustrate the characteristic temporal profiles of SOs and spindles across different sleep stages.

Methods, Page 25, Line 525-532

“Event-related potentials (ERP) analysis. After completing the detection of each sleep rhythm event, we performed ERP analyses for SOs, spindles, and coupling events in different sleep stages. Specifically, for SO events, we took the trough of the DOWN-state of each SO as the zero-time point, then extracted data in a [-2 s to 2 s] window from the broadband (0.1–30 Hz) EEG and used [-2 s to -0.5 s] for baseline correction; the results were then averaged across 107 subjects (see Fig. S2a). For spindle events, we used the peak of each spindle as the zero-time point and applied the same data extraction window and baseline correction before averaging across 107 subjects (see Fig. S2b). Finally, for SO-spindle coupling events, we followed the same procedure used for SO events (see Fig. 2a, Figs. S3–S4).”

(3) Provide detailed info concerning sleep characteristics (time spent in each sleep stage etc.).

Thank you for this suggestion. Same as the response above, we will provide comprehensive tables in the supplementary materials, contains descriptive information about sleep-related characteristics.

Supplementary Materials, Page 42, Table S1 (same as above)

(4) What would happen if more stringent parameters were used for event detection? Would the authors still observe a significant number of SO spindles during N1 and REM? Would this affect the fMRI-related results?

Thank you for this suggestion. Our methods for detecting SOs, spindles, and their couplings were originally developed for N2 and N3 sleep data, based on the specific characteristics of these stages. These methods are widely recognized in sleep research (Hahn et al., 2020; Helfrich et al., 2019; Helfrich et al., 2018; Ngo, Fell, & Staresina, 2020; Schreiner et al., 2022; Schreiner et al., 2021; Staresina et al., 2015; Staresina et al., 2023). However, because this percentile-based detection approach will inherently identify a certain number of events if applied to other stages (e.g., N1 and REM), the nature of these events in those stages remains unclear compared to N2/N3. We nevertheless identified and reported the detailed descriptive statistics of these sleep rhythms in all sleep stages, under the same operational definitions, both for completeness and as a sanity check. Within the same subject, there should be more SOs, spindles, and their couplings in N2/N3 than in N1 or REM (see also Figure S2-S4, Table S1-S4).

Furthermore, in order to explore the impact of this on our fMRI results, we conducted an additional sensitivity analysis by applying different detection parameters for SOs. Specifically, we adjusted amplitude percentile thresholds for SO detection (the parameter that has the greatest impact on the results). We used the hippocampal activation value during N2&N3 stage SO-spindle coupling as an anchor value and found that when the parameters gradually became stricter, the results were similar to or even better than the current results. However, when we continued to increase the threshold, the results began to gradually decrease until the threshold was increased to 80%, and the results were no longer significant. This indicates that our results are robust within a specific range of parameters, but as the threshold increases, the number of trials decreases, ultimately weakening the statistical power of the fMRI analysis.

Thank you again for your suggestions on sleep rhythm event detection. We will add the results in Supplementary and revise our manuscript accordingly.

Results, Page 11, Line 199-208

“Spindles were correlated with positive activation in the thalamus (ROI analysis, t(106) = 15.39, p < 1e-4), the anterior cingulate cortex (ACC), and the putamen, alongside deactivation in the DMN (Fig. 3c). Notably, SO-spindle coupling was linked to significant activation in both the thalamus (ROI analysis, t(106) \= 3.38, p = 0.0005) and the hippocampus (ROI analysis, t(106) \= 2.50, p = 0.0070, Fig. 3d). However, no decrease in DMN activity was found during SO-spindle coupling, and DMN activity during SO was significantly lower than during coupling (ROI analysis, t(106) \= -4.17, p < 1e-4). For more detailed activation patterns, see Table S5-S7. We also varied the threshold used to detect SO events to assess its effect on hippocampal activation during SO-spindle coupling and observed that hippocampal activation remained significant when the percentile thresholds for SO detection ranged between 71% and 80% (see Fig. S6).”

Finally, we sincerely thank all again for your thoughtful and constructive feedback. Your insights have been invaluable in refining our analyses, strengthening our interpretations, and improving the clarity and rigor of our manuscript. We appreciate the time and effort you have dedicated to reviewing our work, and we are grateful for the opportunity to enhance our study based on your recommendations.  

References:

Bergmann, T. O., Mölle, M., Diedrichs, J., Born, J., & Siebner, H. R. (2012). Sleep spindle-related reactivation of category-specific cortical regions after learning face-scene associations. NeuroImage, 59(3), 2733-2742. 

Buzsáki, G. (2015). Hippocampal sharp wave‐ripple: A cognitive biomarker for episodic memory and planning. Hippocampus, 25(10), 1073-1188. 

Caporro, M., Haneef, Z., Yeh, H. J., Lenartowicz, A., Buttinelli, C., Parvizi, J., & Stern, J. M. (2012). Functional MRI of sleep spindles and K-complexes. Clinical neurophysiology, 123(2), 303-309. 

Coulon, P., Budde, T., & Pape, H.-C. (2012). The sleep relay—the role of the thalamus in central and decentral sleep regulation. Pflügers Archiv-European Journal of Physiology, 463, 53-71. 

Crunelli, V., Lőrincz, M. L., Connelly, W. M., David, F., Hughes, S. W., Lambert, R. C., Leresche, N., & Errington, A. C. (2018). Dual function of thalamic low-vigilance state oscillations: rhythm-regulation and plasticity. Nature Reviews Neuroscience, 19(2), 107-118. 

Czisch, M., Wehrle, R., Stiegler, A., Peters, H., Andrade, K., Holsboer, F., & Sämann, P. G. (2009). Acoustic oddball during NREM sleep: a combined EEG/fMRI study. PloS one, 4(8), e6749. 

Diba, K., & Buzsáki, G. (2007). Forward and reverse hippocampal place-cell sequences during ripples. Nature Neuroscience, 10(10), 1241. 

Diekelmann, S., & Born, J. (2010). The memory function of sleep. Nature Reviews Neuroscience, 11(2), 114-126. 

Fogel, S., Albouy, G., King, B. R., Lungu, O., Vien, C., Bore, A., Pinsard, B., Benali, H., Carrier, J., & Doyon, J. (2017). Reactivation or transformation? Motor memory consolidation associated with cerebral activation time-locked to sleep spindles. PloS one, 12(4), e0174755. 

Hahn, M. A., Heib, D., Schabus, M., Hoedlmoser, K., & Helfrich, R. F. (2020). Slow oscillation-spindle coupling predicts enhanced memory formation from childhood to adolescence. Elife, 9, e53730. 

Halassa, M. M., Siegle, J. H., Ritt, J. T., Ting, J. T., Feng, G., & Moore, C. I. (2011). Selective optical drive of thalamic reticular nucleus generates thalamic bursts and cortical spindles. Nature Neuroscience, 14(9), 1118-1120. 

Hale, J. R., White, T. P., Mayhew, S. D., Wilson, R. S., Rollings, D. T., Khalsa, S., Arvanitis, T. N., & Bagshaw, A. P. (2016). Altered thalamocortical and intra-thalamic functional connectivity during light sleep compared with wake. NeuroImage, 125, 657-667. 

Helfrich, R. F., Lendner, J. D., Mander, B. A., Guillen, H., Paff, M., Mnatsakanyan, L., Vadera, S., Walker, M. P., Lin, J. J., & Knight, R. T. (2019). Bidirectional prefrontal-hippocampal dynamics organize information transfer during sleep in humans. Nature Communications, 10(1), 3572. 

Helfrich, R. F., Mander, B. A., Jagust, W. J., Knight, R. T., & Walker, M. P. (2018). Old brains come uncoupled in sleep: slow wave-spindle synchrony, brain atrophy, and forgetting. Neuron, 97(1), 221-230. e224. 

Horovitz, S. G., Fukunaga, M., de Zwart, J. A., van Gelderen, P., Fulton, S. C., Balkin, T. J., & Duyn, J. H. (2008). Low frequency BOLD fluctuations during resting wakefulness and light sleep: A simultaneous EEG‐fMRI study. Human brain mapping, 29(6), 671-682. 

Huang, Q., Xiao, Z., Yu, Q., Luo, Y., Xu, J., Qu, Y., Dolan, R., Behrens, T., & Liu, Y. (2024). Replay-triggered brain-wide activation in humans. Nature Communications, 15(1), 7185. 

Ilhan-Bayrakcı, M., Cabral-Calderin, Y., Bergmann, T. O., Tüscher, O., & Stroh, A. (2022). Individual slow wave events give rise to macroscopic fMRI signatures and drive the strength of the BOLD signal in human resting-state EEG-fMRI recordings. Cerebral Cortex, 32(21), 4782-4796. 

Laufs, H. (2008). Endogenous brain oscillations and related networks detected by surface EEG‐combined fMRI. Human brain mapping, 29(7), 762-769. 

Laufs, H., Walker, M. C., & Lund, T. E. (2007). ‘Brain activation and hypothalamic functional connectivity during human non-rapid eye movement sleep: an EEG/fMRI study’—its limitations and an alternative approach. Brain, 130(7), e75. 

Margulies, D. S., Ghosh, S. S., Goulas, A., Falkiewicz, M., Huntenburg, J. M., Langs, G., Bezgin, G., Eickhoff, S. B., Castellanos, F. X., & Petrides, M. (2016). Situating the default-mode network along a principal gradient of macroscale cortical organization. Proceedings of the National Academy of Sciences, 113(44), 12574-12579. 

Massimini, M., Huber, R., Ferrarelli, F., Hill, S., & Tononi, G. (2004). The sleep slow oscillation as a traveling wave. Journal of Neuroscience, 24(31), 6862-6870. 

Moehlman, T. M., de Zwart, J. A., Chappel-Farley, M. G., Liu, X., McClain, I. B., Chang, C., Mandelkow, H., Özbay, P. S., Johnson, N. L., & Bieber, R. E. (2019). All-night functional magnetic resonance imaging sleep studies. Journal of neuroscience methods, 316, 83-98. 

Molle, M., Bergmann, T. O., Marshall, L., & Born, J. (2011). Fast and slow spindles during the sleep slow oscillation: disparate coalescence and engagement in memory processing. Sleep, 34(10), 1411-1421. 

Ngo, H.-V., Fell, J., & Staresina, B. (2020). Sleep spindles mediate hippocampal-neocortical coupling during long-duration ripples. Elife, 9, e57011. 

Picchioni, D., Horovitz, S. G., Fukunaga, M., Carr, W. S., Meltzer, J. A., Balkin, T. J., Duyn, J. H., & Braun, A. R. (2011). Infraslow EEG oscillations organize large-scale cortical– subcortical interactions during sleep: a combined EEG/fMRI study. Brain research, 1374, 63-72. 

Schabus, M., Dang-Vu, T. T., Albouy, G., Balteau, E., Boly, M., Carrier, J., Darsaud, A., Degueldre, C., Desseilles, M., & Gais, S. (2007). Hemodynamic cerebral correlates of sleep spindles during human non-rapid eye movement sleep. Proceedings of the National Academy of Sciences, 104(32), 13164-13169. 

Schreiner, T., Kaufmann, E., Noachtar, S., Mehrkens, J.-H., & Staudigl, T. (2022). The human thalamus orchestrates neocortical oscillations during NREM sleep. Nature communications, 13(1), 5231. 

Schreiner, T., Petzka, M., Staudigl, T., & Staresina, B. P. (2021). Endogenous memory reactivation during sleep in humans is clocked by slow oscillation-spindle complexes. Nature Communications, 12(1), 3112. 

Singh, D., Norman, K. A., & Schapiro, A. C. (2022). A model of autonomous interactions between hippocampus and neocortex driving sleep-dependent memory consolidation. Proceedings of the National Academy of Sciences, 119(44), e2123432119. 

Spoormaker, V. I., Schröter, M. S., Gleiser, P. M., Andrade, K. C., Dresler, M., Wehrle, R., Sämann, P. G., & Czisch, M. (2010). Development of a large-scale functional brain network during human non-rapid eye movement sleep. Journal of Neuroscience, 30(34), 11379-11387. 

Staresina, B. P., Bergmann, T. O., Bonnefond, M., van der Meij, R., Jensen, O., Deuker, L., Elger, C. E., Axmacher, N., & Fell, J. (2015). Hierarchical nesting of slow oscillations, spindles and ripples in the human hippocampus during sleep. Nature Neuroscience, 18(11), 1679-1686. 

Staresina, B. P., Niediek, J., Borger, V., Surges, R., & Mormann, F. (2023). How coupled slow oscillations, spindles and ripples coordinate neuronal processing and communication during human sleep. Nature Neuroscience, 1-9. 

Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C., & Wager, T. D. (2011). Large-scale automated synthesis of human functional neuroimaging data. Nature methods, 8(8), 665-670. 

Yeshurun, Y., Nguyen, M., & Hasson, U. (2021). The default mode network: where the idiosyncratic self meets the shared social world. Nature Reviews Neuroscience, 1-12.

Author response:

The following is the authors’ response to the original reviews

Main revision made to the manuscript

The main revision made to the manuscript is to reconcile our findings with the line attractor model. The revision is based on Reviewer 1’s comment on reinterpreting our results as a superposition of an attractor model with fast timescale dynamics. We expanded our analysis regime to the start of a trial and characterized the overall within-trial dynamics to reinterpret our findings.

We first acknolwedge that our results are not in contradiction with evidence integration on a line attractor. As pointed out by the reviewers, our finding that the integration of reward outcome explains the reversal probability activity x_rev (Figure 3) is compatible with the line attractor model. However, the reward integration equation is an algebraic relation and does not characterize the dynamics of reversal probability activity. So a closer analysis on the neural dynamics is needed to assess the feasibility of line attractor.

In the revised manuscript, we show that x_rev exhibits two different activity modes (Figure 4). First, x_rev has substantial non-stationary dynamics during a trial, and this non-stationary activity is incompatible with the line attractor model, as claimed in the original manuscript. Second, we present new results showing that x_rev is stationary (i.e., constant in time) and stable (i.e., contracting) at the start of a trial. These two properties of x_rev support that it is a point attractor at the start of a trial and is compatible with the line attractor model. 

We further analyze how the two activity modes are linked (Figure 4, Support vector regression). We show that the non-stationary activity is predictable from the stationary activity if the underlying dynamics can be inferred. In other words, the non-stationary activity during a trial is generated by an underlying dynamics with the initial condition provided by the stationary state at the start of trial.

These results suggest an extension of the line attractor model where an attractor state at the start of a trial provides an initial condition from which non-stationary activity is generated during a trial by an underlying dynamics associated with task-related behavior (Figure 4, Augmented model). 

The separability of non-stationary trajectories (Figure 5 and 6) is a property of the non-stationary dynamics that allows separable points in the initial stationary state to remain separable during a trial, thus making it possible to represent distinct probabilistic values in non-stationary activity.

This revised interpretation of our results (1) retains our original claim that the non-stationary dynamics during a trial is incompatible with the line attractor model and (2) introduces attractor state at the start of a trial which is compatible with the line attractor model. Our anlaysis shows that the two activity modes are linked by an underlying dynamics, and the attractor state serves as initial state to launch the non-stationary activity.

Responses to the Public Reviews:

Reviewer # 1:

(1) To provide better explanation of the reversal learning task and network training method, we added detailed description of RNN and monkey task structure (Result Section 1), included a schematic of target outputs (Figure1B), explained the rationale behind using inhibitory network model (Method Section 1) and explained the supervised RNN training scheme (Result Section 1). This information can also be found in the Methods.

(2) Our understanding is that the augmented model discussed in the previous page is aligned with the model suggested by Reviewer 1: “a curved line attractor, with faster timescale dynamics superimposed on this structure”. It is likely that the “fast” non-stationary activity observed during the trial is driven by task-related behavior, thus is transient. For instance, we do not observe such non-stationary activity in the inter-trial-interval when the task-related behavior is absent. For this reason, the non-stationary trajectories were not considered to be part of the attractor. Instead, they are transient activity generated by the underlying neural dynamics associated with task-related behavior. We believe such characterization of faster timescale dynamics is consistent with Reviewer 1’s view and wanted to clarify that there are two different activity modes.

(3) We appreciate the reviewers (Reviewer 1 and Reviewer 2) comment that TDR may be limited in isolating the neural subspace of interest. Our study presents what could be learned from TDR but is by no means the only way to interpret the neural data. It would be of future work to apply other methods for isolating task-related neural activities.

We would appreciate it if the reviewers could share thoughts on what other alternative methods could better isolate the reversal probability activity.

Reviewer # 2:

(1) (i) We respectfully disagree with Reviewer 2’s comment that “no action is required to be performed by neurons in the RNN”. In our network setup, the output of RNN learns to choose a sign (+ or -), as Reviewer 2 pointed out, to make a choice. This is how the RNN takes an action. It is unclear to us what Reviewer 2 has intended by “action” and how reaching a target value (not just taking a sign) would make a significant difference in how the network performs the task. 

(ii)  From Reviewer 2’s comment that “no intervening behavior is thus performed by neurons”, we noticed that the term “intervening behavior” has caused confusion. It refers to task-related behavior, such as making choices or receiving reward, that the subject must perform across trials before reversing its preferred choice. These are the behaviors that intervene the reversal of preferred choice. To clarify its meaning, in the revised manuscript, we changed the term to “task-related behavior” and put them in context. For example, in the Introduction we state that “However, during a trial, task-related behavior, such as making decisions or receiving feedback, produced …”

(iii) As pointed out by Reviewer 2, the lack of fixation period in the RNN could make differences in the neural dynamics of RNN and PFC, especially at the start of a trial. We demonstrate this issue in Result Section 4 where we analyze the stationary activity at the start of a trial. We find that fixating the choice output to zero before making a choice promotes stationary activity and makes the RNN activity more similar to the PFC activity.

Reviewer #3:

(1) (i) In the previous study (Figure 1 in [Bartolo and Averbeck ‘20]), it was shown that neural activity can predict the behavioral reversal trial. This is the reason we examined the neural activity in the trials centered at the behavioral reversal trial. We explained in Result Section 2 that we followed this line of analysis in our study.

(ii) We would like to emphasize that the main point of Figures 4 and 5 is to show the separability of neural trajectories: the entire trajectory shifts without overlapping. It is not obvious that high-dimensional neural population activity from two trials should remain separated when their activities are compressed into a one-dimensional subspace. The onedimensional activities can easily collide since their activities are compressed into a lowdimensional space. We revised the manuscript to bring out these points. We added an opening paragraph that discusses separability of trajectories and revised the main text to bring out the findings on separability. 

(iii) We agree with Reviewer 3 that it would be interesting to look at what happens in other subspace of neural activity that are not related to reversal probability and characterize how different neural subspace interact with each. However, the focus of this paper was the reversal probability activity, and we’d consider these questions out of the scope of current paper. We point out that, using the same dataset, neural activity related to other experimental variables were analyzed in other papers [Bartolo and Averbeck ’20; Tang, Bartolo and Averbeck ‘21] 

(2) (i) In the revised manuscript, we added explanation on the rational behind choosing inhibitory network as a simplified model for the balanced state. In brief, strong inhibitory recurrent connections with strong excitatory external input operates in the balanced state, as in the standard excitatory-inhibitory network. We included references that studied this inhibitory network. We also explained the technical reason (GPU memory) for choosing the inhibitory model.

(ii) We thank the reviewer for pointing out that the original manuscript did not mention how the feedback and cue were initialized. They were random vectors sample from Gaussian distribution. We added this information in the revised manuscript. In our opinion, it is common to use random external inputs for training RNNs, as it is a priori unclear how to choose them. In fact, it is possible to analyze the effects of random feedback on one-dimensional x_rev dynamics by projecting the random feedback vector to the reversal probability vector. This is shown in Figure 4F.

(iii) We agree that it would be more natural to train the RNN to solve the task without using the Bayesian model. We point out this issue in the Discussion in the revised manuscript.

Recommendations for the authors:

Reviewer #1:

(1) My understanding of network training was that a Bayesian ideal observer signaled target output based on previous reward outcomes. However, the authors never mention that networks are trained by supervised learning in the main text until the last paragraph of the discussion. There is no mention that there was an offset in the target based on the behavior of the monkeys in the main text. These are really important things to consider in the context of the network solution after training. I couldn't actually find any figure that presents the target output for the network. Did I miss something key here?

In Result Section 1, we added a paragraph that describes in detail how the RNN is trained. We explained that the network is first simulated and then the choice outputs and reward outcomes are fed into the Bayesian model to infer the scheduled reversal trial. A few trials are added to the inferred reversal trial to obtain the behavioral reversal trial, as found in a previous study [Bartolo and Averbeck ‘20]. Then the network weights are updated by backpropagation-through-time via supervised learning. 

In the original manuscript, the target output for the network was described in Methods Section 2.5, Step 4. To make this information readily accessible, we added a schematic in Figure 1B that shows the scheduled, inferred and behavioral reversal trials. It also shows how the target choice ouputs are defined. They switch abruptly at the behavioral reversal trial.

(2) The role of block structure in the task is an important consideration. What are the statistics of block switches? The authors say on average the reversals are every 36 trials, but also say there are random block switches. The reviewer's notes suggest that both the networks and monkeys may be learning about the typical duration of blocks, which could influence their expectations of reversals. This aspect of the task design should be explained more thoroughly and considered in the context of Figure 1E and 5 results.

We provided more detailed description of the reversal learning task in Result Section 1. We clarified that (1) a task is completed by executing a block of fixed number of trials and (2) reversal of reward schedule occurrs at a random trial around the mid-trial in a block. The differences in the number of trials in a block that the RNNs (36) and the monkeys (80) perform are also explained. We also pointed out the differences in how the reversal trial is randomly sampled.

However, it is unclear what Reviewer 1 meant by random block switches. Our reversal learning task is completed when a block of fixed number of trials is executed. Reversal of reward schedule occurs only once on a randomly selected trial in the block, and the reversed reward schedule is maintained until the end of a block. It is different from other versions of reveral learning where the reward schedule switches multiple times across trials. We clarified this point in Result Section 1.

(3) The relationship between the supervised learning approach used in the RNNs and reinforcement learning was confused in the discussion. "Although RNNs in our study were trained via supervised learning, animals learn a reversal-learning task from reward feedback, making it into a reinforcement learning (RL) problem." This is fundamentally not true. In the case of this work, the outcome of the previous trial updates the target output, rather than the trial and error type learning as is typical in reinforcement learning. Networks are not learning by reinforcement learning and this statement is confusing.

We agree with Reviewer 1’s comment that the statement in the original manuscript is confusing. Our intention was to point out that our study used supervised learning, and this is different from animals learn by reinforcement learning in rea life. We revised the sentence in Discussion as follows:

“The RNNs in our study were trained via supervised learning. However, in real life, animals learn a reversal learning task via reinforcement learning (RL), i.e., learn the task from reward outcomes.”

(4) The distinction between line attractors and the dynamic trajectories described by the authors deserves further investigation. A significant concern arises from the authors' use of targeted dimensionality reduction (TDR), a form of regression, to identify the axis determining reversal probability. While this approach can reveal interesting patterns in the data, it may not necessarily isolate the dimension along which the RNN computes reversal probability. This limitation could lead to misinterpretation of the underlying neural dynamics.

a) This manuscript cites work described in "Prefrontal cortex as a meta-reinforcement learning system," which examined a similar task. In that study, the authors identified a v-shaped curve in the principal component space of network states, representing the probability of choosing left or right.

Importantly, this curve is topologically equivalent to a line and likely represents a line attractor. However, regressing against reversal probability in such a case would show that a single principal component (PC2) directly correlates with reversal probability.

b) The dynamics observed in the current study bear a striking resemblance to this structure, with the addition of intervening loops in the network state corresponding to within-trial state evolution. Crucially, these observations do not preclude the existence of a line attractor. Instead, they may reflect the network's need to produce fast timescale dynamics within each trial, superimposed on the slower dynamics of the line attractor.

c) This alternative interpretation suggests that reward signals could function as inputs that shift the network state along the line attractor, with information being maintained across trials. The fast "intervening behaviors" observed by the authors could represent faster timescale dynamics occurring on top of the underlying line attractor dynamics, without erasing the accumulated evidence for reversals.

d) Given these considerations, the authors' conclusion that their results are better described by separable dynamic trajectories rather than fixed points on a line attractor may be premature. The observed dynamics could potentially be reconciled with a more nuanced understanding of line attractor models, where the attractor itself may be curved and coexist with faster timescale dynamics.

We appreciate the insightful comments on (1) the similarity of the work by Wang et al ’18 with our findings and (2) an alternative interpretation that augments the line attractor with fast timescale dynamics. 

(1) We added a discussion of the work by Wang et al ’18 in Result Section 2 to point out the similarity of their findings in the principal component space with ours in the x_rev and x_choice space. We commented that such network dynamics could emerge when learning to perform the reversal learning the task, regardless of the training schemes. 

We also mention that the RL approach in Wang et al ’18 does not consider within-trial dynamics, therefore lacks the non-stationary activity observed during the trial in the PFC of monkeys and our trained RNNs.

(2) We revised our original manuscript substantially to reconcile the line attractor model with the nonstationary activity observed during a trial. 

Here are the highlights of the revised interpretation of the PFC and the RNN network activity

- The dynamics of x_rev consists of two activity modes, i.e., stationary activity at the start of a trial and non-stationary activity during the trial. Schematic of the augmented model that reconciles two activity modes is shown in Figure 4A. Analysis of the time derivative (dx_reverse / dt) and contractivity of the stationary state are shown in Figure 4B,C to demonstrate two activity modes.

- We discuss in Result Section 4 main text that the stationary activity is consistent with the line attractor model, but the non-stationary activity deviates from the model. 

- The two activity modes are linked dynamically. There is an underlying dynamics that can map the stationary state to the non-stationary trajectory. This is shown by predicting the nonstationary trajectory with the stationary state using a support vector regression model. The prediction results are shown in Figure 4D,E,F.

- We discuss in Result Section 4 an extension of the standard line attractor model: points on the line attractor can serve as initial states that launch non-stationary activity associated with taskrelated behavior.

- The separability of neural trajectories presented in Result Section 5 is framed as a property of the non-stationary dynamics associated with task-related behavior.

To strengthen their claims, the authors should:

(1) Provide a more detailed description of their RNN training paradigm and task structure, including clear illustrations of target outputs.

(2) Discuss how their findings relate to and potentially extend previous work on similar tasks, particularly addressing the similarities and differences with the v-shaped state organization observed in reinforcement learning contexts. (https://www.nature.com/articles/s41593-018-0147-8 Figure1).

(3) Explore whether their results could be consistent with a curved line attractor model, rather than treating line attractors and dynamic trajectories as mutually exclusive alternatives.

Our response to these three comments is described above.

Addressing these points would significantly enhance the impact of the study and provide a more nuanced understanding of how reversal probabilities are represented in neural circuits.

In conclusion, while this study provides interesting insights into the neural representation of reversal probability, there are several areas where the methodology and interpretations could be refined.

Additional Minor Concerns:

(1) Network Training and Reversal Timing: The authors mention that the network was trained to switch after a reversal to match animal behavior, stating "Maximum a Posterior (MAP) of the reversal probability converges a few trials past the MAP estimate." More explanation of how this training strategy relates to actual animal behavior would enhance the reader's understanding of the meaning of the model's similarity to animal behavior in Figure 1.

In Method Section 2.5, we described how our observation that the running estimate of MAP converges a few trials after the actual MAP is analogous to the animal’s reversal behavior.

“This observation can be interpreted as follows. If a subject performing the reversal learning task employs the ideal observer model to detect the trial at which reward schedule is reversed, the subject can infer the reversal of reward schedule a few trials past the actual reversal and then switch its preferred choice. This delay in behavioral reversal, relative to the reversal of reward schedule, is analogous to the monkeys switching their preferred choice a few trials after the reversal of reward schedule.”

In Step 4, we also mentioned that the target choice outputs are defined based on our observation in Step 3.

“We used the observation from Step 3 to define target choice outputs that switch abruptly a few trials after the reversal of reward schedule, denoted as $t^*$ in the following. An example of target outputs are shown in Fig.\,\ref{fig_behavior}B.”

(2) How is the network simulated in step 1 of training? Is it just randomly initialized? What defines this network structure?

The initial state at the start of a block was random. We think the initial state is less relevant as the external inputs (i.e., cue and feedback) are strong and drive the network dynamics. We mentioned these setup and observation in Step 1 of training.

“Step 1. Simulate the network starting from a random initial state, apply the external inputs, i.e., cue and feedback inputs, at each trial and store the network choices and reward outcomes at all the trials in a block. The network dynamics is driven by the external inputs applied periodically over the trials.”

(3) Clarification on Learning Approach: More description of the approach in the main text would be beneficial. The statement "Here, we trained RNNs that learned from a Bayesian inference model to mimic the behavioral strategies of monkeys performing the reversal learning task [2, 4]" is somewhat confusing, as the model isn't directly fit to monkey data. A more detailed explanation of how the Bayesian inference model relates to monkey behavior and how it's used in RNN training would improve clarity.

We described the learning approach in more detail, but also tried to be concise without going into technical details.

We revised the sentence in Introduction as follows:

“We sought to train RNNs to mimic the behavioral strategies of monkeys performing the reversal learning task. Previous studies \cite{costa2015reversal, bartolo2020prefrontal} have shown that a Bayesian inference model can capture a key aspect of the monkey's behavioral strategy, i.e., adhere to the preferred choice until the reversal of reward is detected and then switch abruptly. We trained the RNNs to replicate this behavioral strategy by training them on target behaviors generated from the Bayesian model.”

We also added a paragraph in Result Section 1 that explains in detail how the training approach works.

(4) In Figure 1B, it would be helpful to show the target output.

We added a figure in Fig1B that shows a schematic of how the target output is generated.

(5) An important point to consider is that a line attractor can be curved while still being topologically equivalent to a line. This nuance makes Figure 4A somewhat difficult to interpret. It might be helpful to discuss how the observed dynamics relate to potentially curved line attractors, which could provide a more nuanced understanding of the neural representations.

As discussed above, we interpret the “curved” activity during the trial as non-stationary activity. We do not think this non-stationary activity would be characterized as attractor. Attractor is (1) a minimal set of states that is (2) invariant under the dynamics and (3) attracting when perturbed into its neighborhood [Strogatz, Nonlinear dynamics and chaos]. If we consider the autonomous system without the behavior-related external input as the base system, then the non-stationary states could satisfy (2) and (3) but not (1), so they are not part of the attractor. If we include the behavior-related external input to the autonomous dynamics, then it may be possible that the non-stationary trajectories are part of the attractor. We adopted the former interpretation as the behavior-related inputs are external and transient.

(6) The results of the perturbation experiments seem to follow necessarily from the way x_rev was defined. It would be valuable to clarify if there's more to these results than what appears to be a direct consequence of the definition, or if there are subtleties in the experimental design or analysis that aren't immediately apparent.

The neural activity x_rev is correlated to the reversal probability, but it is unclear if the activity in this neural subspace is causally linked to behavioral variables, such as choice output. We added this explanation at the beginning of Results Section 7 to clarify the reason for performing the perturbation experiments.

“The neural activity $x_{rev}$ is obtained by identifying a neural subspace correlated to reversal probability. However, it remains to be shown if activity within this neural subspace is causally linked to behavioral variables, such as choice output.”

Reviewer #2:

Below is a list of things I have found difficult to understand, and been puzzled/concerned about while reading the manuscript:

(1) It would be nice to say a bit more about the dataset that has been used for PFC analysis, e.g. number of neurons used and in what conditions is Figure 2A obtained (one has to go to supplementary to get the reference).

We added information about the PFC dataset in the opening paragraph of Result Section 2 to provide an overview of what type of neural data we’ve analyzed. It includes information about the number of recorded neurons, recording method and spike binning process.

(2) It would be nice to give more detail about the monkey task and better explain its trial structure.

In Result Section 1 we added a description of the overall task structure (and its difference with other versions of revesal learning task), the RNN / monkey trial structure and differences in RNN and monkey tasks.

(3) In the introduction it is mentioned that during the hold period, the probability of reversal is represented. Where does this statement come from?

The fact that neural activity during a hold period, i.e., fixation period before presenting the target images, encodes the probability of reversal was demonstrated in a previous study (Bartolo and Averbeck ’20). 

We realize that our intention was to state that, during the hold period, the reversal probability activity is stationary as in the line attractor model, instead of focusing on that the probability of reversal is represented during this period. We revised the sentence to convey this message. In addition, we revised the entire paragraph to reinterpret our findings: there are two activity modes where the stationary activity is consistent with the line attractor model but the non-stationary activity deviates from it.

(4) "Around the behavioral reversal trial, reversal probabilities were represented by a family of rankordered trajectories that shifted monotonically". This sentence is confusing and hard to understand.

Thank you for point this out. We rewrote the paragraph to reflect our revised interpretation. This sentence was removed, as it can be considered as part of the result on separable trajectories.

(5) For clarity, in the first section, when it is written that "The reversal behavior of trained RNNs was similar to the monkey's behavior on the same task" it would be nice to be more precise, that this is to be expected given the strategy used to train the network.

We removed this sentence as it makes a blanket statement. Instead, we compared the behavioral outputs of the RNNs and the monkeys one by one.

We added a sentence in Result Section 1 that the RNN’s abrupt behavioral reversal is expected as they are trained to mimic the target choice outputs of the Bayesian model.

“Such abrupt reversal behavior was expected as the RNNs were trained to mimic the target outputs of the Bayesian inference model.”

(6) What is the value of tau used in eq (1), and how does it compare to trial duration?

We described the value of time constant tau in Eq (1) and also discussed in Result Section 1 that tau=20ms is much faster than trial duration 500ms, thus the persistent behavior seen in trained RNNs is due to learning.

(7) It would be nice to expand around the notion of « temporally flexible representation » to help readers grasp what this means.

Instead of stating that the separable dynamic trajectories have “temporally flexible representation”, we break down in what sense it is temporally flexible: separable dynamic trajectories can accommodate the effects that task-related behavior have on generating non-stationary neural dynamics.

“In sum, our results show that, in a probabilistic reversal learning task, recurrent neural networks encode reversal probability by adopting, not only stationary states as in a line attractor, but also separable dynamic trajectories that can represent distinct probabilistic values while accommodating non-stationary dynamics associated with task-related behavior.”

Reviewer #3:

(1) Data:

It would be useful to describe the experimental task, recording setup, and analyses in much more detail - both in the text and in the methods. What part of PFC are the recordings from? How many neurons were recorded over how many sessions? Which other papers have they been used in? All of these things are important for the reader to know, but are not listed anywhere. There are also some inconsistencies, with the main text e.g. listing the 'typical block length' as 36 trials, and the methods listing the block length as 24 trials (if this is a difference between the biological data and RNN, that should be more explicit and motivated).

We provided more detailed description of the monkey experimental task and PFC recordings in Result Section 1. We also added a new section in Methods 2.1 to describe the monkey experiment.

The experimental analyses should be explained in more detail in the methods. There is e.g. no detailed description of the analysis in Figure 6F.

We added a new section in Methods 6 to describe how the residual PFC activity is computed. It also describes the RNN perturbation experiments.

Finally, it would be useful for more analyses of monkey behaviour and performance, either in the main text or supplementary figures.

We did not pursue this comment as it is unclear how additional behavioral analyses would improve the manuscript.

(2) Model:

When fitting the network, 'step 1' of training in 2.3 seems superfluous. The posterior update from getting a reward at A is the same as that from not getting a reward at B (and vice versa), and it is therefore completely independent of the network choice. The reversal trial can therefore be inferred without ever simulating the network, simply by generating a sample of which trials have the 'good' option being rewarded and which trials have the 'bad' option being rewarded.

We respectfully disagree with Reviewer 3’s comment that the reversal trial can be inferred without ever simulating the network. The only way for the network to know about the underlying reward schedule is to perform the task by itself. By simulating the network, it can sample the options and the reward outcomes. 

Our understanding is that Review 3 described a strategy that a human would use to perform this task. Our goal was to train the RNN to perform the task.

Do the blocks always start with choice A being optimal? Is everything similar if the network is trained with a variable initial rewarded option? E.g. in Fig 6, would you see the appropriate swap in the effect of the perturbation on choice probability if choice B was initially optimal?

Thank you for pointing out that the initial high-value option can be random. When setting up the reward schedule, the initial high-value option was chosen randomly from two choice outputs and, at the scheduled reversal, it was switched to the other option. We did not describe this in the original manuscript.

We added a descrption in Training Scheme Step 4 that the the initial high-value option is selected randomly. This is also explained in Result Section 1 when we give an overview of the RNN training procedure.

(3) Content:

It is rarely explained what the error bars represent (e.g. Figures 3B, 4C, ...) - this should be clear in all figures.

We added that the error bars represent the standard error of mean.

Figure 2A: this colour scheme is not great. There are abrupt colour changes both before and after the 'reversal' trial, and both of the extremes are hard to see.

We changed the color scheme to contrast pre- and post-reversal trials without the abrupt color change.

Figure 3E/F: how is prediction accuracy defined?

We added that the prediction accuracy is based on Pearson correlation.

Figure 4B: why focus on the derivative of the dynamics? The subsequent plots looking at the actual trajectories are much easier to understand. Also - what is 'relative trial' relative to?

The derivative was analyzed to demonstrate stationarity or non-stationarity of the neural activity. We think it will be clearer in the revised manuscript that the derivative allows us to characterize those two activity modes.

Relative trial number indicate the trial position relative to the behavioral reversal trial. We added this description to the figures when “relative trial” is used.

Figure 4C: what do these analyses look like if you match the trial numbers for the shift in trajectories? As it is now, there will presumably be more rewarded trials early and late in each block, and more unrewarded trials around the reversal point. Does this introduce biases in the analysis? A related question is (i) why the black lines are different in the top and bottom plots, and (ii) why the ends of the black lines are discontinuous with the beginnings of the red/blue lines.

We could not understand what Reviewer 3 was asking in this comment. It’d help if Review 3 could clarify the following question:

“Figure 4C: what do these analyses look like if you match the trial numbers for the shift in trajectories?”

Question (i): We wanted to look at how the trajectory shifts in the subsequent trial if a reward is or is not received in the current trial. The top panel analyzed all the trials in which the subsquent trial did not receive a reward. The bottom panel analyzed all the trials in which the subsequent trial received a reward. So, the trials analyzed in the top and bottom panels are different, and the black lines (x_rev of “current” trial) in the top and bottom panels are different.

Question (ii): Black line is from the preceding trial of the red/blue lines, so if trials are designed to be continuous with the inter-trial-interval, then black and red/blue should be continuous. However, in the monkey experiment, the inter-trial-intervals were variable, so the end of current trial does not match with the start of next trial. The neural trajectories presented in the manuscript did not include the activity in this inter-trial-interval.

Figure 6C: are the individual dots different RNNs? Claiming that there is a decrease in Delta x_choice for a v_+ stimulation is very misleading.

Yes individual dots are different RNN perturbations. We added explanation about the dots in Figure7C caption. 

We agree with the comment that \Delta x_choice did not decrease. This sentence was removed. Instead, we revised the manuscript to state that x_choice for v_+ stimulation was smaller than the x_choice for v_- stimulation. We performed KS-test to confirm statistical significance.

Discussion: "...exhibited behaviour consistent with an ideal Bayesian observer, as found in our study". The RNN was explicitly trained to reproduce an ideal Bayesian observer, so this can only really be considered an assumption (not a result) in the present study.

We agree that the statement in the original manuscript is inaccurate. It was revised to reflect that, in the other study, behavior outputs similar to a Bayesian observer emerged by simply learning to do the task, intead of directly mimicking the outputs of Bayesian observer as done in our study.

“Authors showed that trained RNNs exhibited behavior outputs consistent with an ideal Bayesian observer without explicitly learning from the Bayesian observer. This finding shows that the behavioral strategies of monkeys could emerge by simply learning to do the task, instead of directly mimicking the outputs of Bayesian observer as done in our study.”

Methods: Would the results differ if your Bayesian observer model used the true prior (i.e. the reversal happens in the middle 10 trials) rather than a uniform prior? Given the extensive literature on prior effects on animal behaviour, it is reasonable to expect that monkeys incorporate some non-uniform prior over the reversal point.

Thank you for pointing out the non-uniform prior. We haven’t conducted this analysis, but would guess that the convergence to the posterior distribution would be faster. We’d have to perform further analysis, which is out of the scope of this paper, to investigate whether the posteior distribution would be different from what we obtained from uniform prior.

Making the code available would make the work more transparent and useful to the community.

The code is available in the following Github repository: https://github.com/chrismkkim/LearnToReverse

Author response:

Reviewer #1 (Public review):

This study investigates the sex determination mechanism in the clonal ant Ooceraea biroi, focusing on a candidate complementary sex determination (CSD) locus-one of the key mechanisms supporting haplodiploid sex determination in hymenopteran insects. Using whole genome sequencing, the authors analyze diploid females and the rarely occurring diploid males of O. biroi, identifying a 46 kb candidate region that is consistently heterozygous in females and predominantly homozygous in diploid males. This region shows elevated genetic diversity, as expected under balancing selection. The study also reports the presence of an lncRNA near this heterozygous region, which, though only distantly related in sequence, resembles the ANTSR lncRNA involved in female development in the Argentine ant, Linepithema humile (Pan et al. 2024). Together, these findings suggest a potentially conserved sex determination mechanism across ant species. However, while the analyses are well conducted and the paper is clearly written, the insights are largely incremental. The central conclusion - that the sex determination locus is conserved in ants - was already proposed and experimentally supported by Pan et al. (2024), who included O. biroi among the studied species and validated the locus's functional role in the Argentine ant. The present study thus largely reiterates existing findings without providing novel conceptual or experimental advances.

Although it is true that Pan et al., 2024 demonstrated (in Figure 4 of their paper) that the synteny of the region flanking ANTSR is conserved across aculeate Hymenoptera (including O. biroi), Reviewer 1’s claim that that paper provides experimental support for the hypothesis that the sex determination locus is conserved in ants is inaccurate. Pan et al., 2024 only performed experimental work in a single ant species (Linepithema humile) and merely compared reference genomes of multiple species to show synteny of the region, rather than functionally mapping or characterizing these regions.

Other comments:

The mapping is based on a very small sample size: 19 females and 16 diploid males, and these all derive from a single clonal line. This implies a rather high probability for false-positive inference. In combination with the fact that only 11 out of the 16 genotyped males are actually homozygous at the candidate locus, I think a more careful interpretation regarding the role of the mapped region in sex determination would be appropriate. The main argument supporting the role of the candidate region in sex determination is based on the putative homology with the lncRNA involved in sex determination in the Argentine ant, but this argument was made in a previous study (as mentioned above).

Our main argument supporting the role of the candidate region in sex determination is not based on putative homology with the lncRNA in L. humile. Instead, our main argument comes from our genetic mapping (in Fig. 2), and the elevated nucleotide diversity within the identified region (Fig. 4). Additionally, we highlight that multiple genes within our mapped region are homologous to those in mapped sex determining regions in both L. humile and Vollenhovia emeryi, possibly including the lncRNA.

In response to the Reviewer’s assertion that the mapping is based on a small sample size from a single clonal line, we want to highlight that we used all diploid males available to us. Although the primary shortcoming of a small sample size is to increase the probability of a false negative, small sample sizes can also produce false positives. We used two approaches to explore the statistical robustness of our conclusions. First, we generated a null distribution by randomly shuffling sex labels within colonies and calculating the probability of observing our CSD index values by chance (shown in Fig. 2). Second, we directly tested the association between homozygosity and sex using Fisher’s Exact Test (shown in Supplementary Fig. S2). In both cases, the association of the candidate locus with sex was statistically significant after multiple-testing correction using the Benjamini-Hochberg False Discovery Rate. These approaches are clearly described in the “CSD Index Mapping” section of the Methods.

We also note that, because complementary sex determination loci are expected to evolve under balancing selection, our finding that the mapped region exhibits a peak of nucleotide diversity lends orthogonal support to the notion that the mapped locus is indeed a complementary sex determination locus.

The fourth paragraph of the results and the sixth paragraph of the discussion are devoted to explaining the possible reasons why only 11/16 genotyped males are homozygous in the mapped region. The revised manuscript will include an additional sentence (in what will be lines 384-388) in this paragraph that includes the possible explanation that this locus is, in fact, a false positive, while also emphasizing that we find this possibility to be unlikely given our multiple lines of evidence.

In response to Reviewer 1’s suggestion that we carefully interpret the role of the mapped region in sex determination, we highlight our careful wording choices, nearly always referring to the mapped locus as a “candidate sex determination locus” in the title and throughout the manuscript. For consistency, the revised manuscript version will change the second results subheading from “The O. biroi CSD locus is homologous to another ant sex determination locus but not to honeybee csd” to “O. biroi’s candidate CSD locus is homologous to another ant sex determination locus but not to honeybee csd,” and will add the word “candidate” in what will be line 320 at the beginning of the Discussion, and will change “putative” to “candidate” in what will be line 426 at the end of the Discussion.

In the abstract, it is stated that CSD loci have been mapped in honeybees and two ant species, but we know little about their evolutionary history. But CSD candidate loci were also mapped in a wasp with multi-locus CSD (study cited in the introduction). This wasp is also parthenogenetic via central fusion automixis and produces diploid males. This is a very similar situation to the present study and should be referenced and discussed accordingly, particularly since the authors make the interesting suggestion that their ant also has multi-locus CSD and neither the wasp nor the ant has tra homologs in the CSD candidate regions. Also, is there any homology to the CSD candidate regions in the wasp species and the studied ant?

In response to Reviewer 1’s suggestion that we reference the (Matthey-Doret et al. 2019) study in the context of diploid males being produced via losses of heterozygosity during asexual reproduction, the revised manuscript will include the following sentence: “Therefore, if O. biroi uses CSD, diploid males might result from losses of heterozygosity at sex determination loci (Fig. 1C), similar to what is thought to occur in other asexual Hymenoptera that produce diploid males (Rabeling and Kronauer 2012; Matthey-Doret et al. 2019).”

We note, however, that in their 2019 study, Matthey-Doret et al. did not directly test the hypothesis that diploid males result from losses of heterozygosity at CSD loci during asexual reproduction, because the diploid males they used for their mapping study came from inbred crosses in a sexual population of that species.

We address this further below, but we want to emphasize that we do not intend to argue that O. biroi has multiple CSD loci. Instead, we suggest that additional, undetected CSD loci is one possible explanation for the absence of diploid males from any clonal line other than clonal line A. In response to Reviewer 1’s suggestion that we reference the (Matthey-Doret et al. 2019) study in the context of multilocus CSD, the revised manuscript version will include the following additional sentence in the fifth paragraph of the discussion: “Multi-locus CSD has been suggested to limit the extent of diploid male production in asexual species under some circumstances (Vorburger 2013; Matthey-Doret et al. 2019).”

Regarding Reviewer 2’s question about homology between the putative CSD loci from the (Matthey-Doret et al. 2019) study and O. biroi, we note that there is no homology. The revised manuscript version will have an additional Supplementary Table (which will be the new Supplementary Table S3) that will report the results of this homology search. The revised manuscript will also include the following additional sentence in the Results: “We found no homology between the genes within the O. biroi CSD index peak and any of the genes within the putative L. fabarum CSD loci (Supplementary Table S3).”

The authors used different clonal lines of O. biroi to investigate whether heterozygosity at the mapped CSD locus is required for female development in all clonal lines of O. biroi (L187-196). However, given the described parthenogenesis mechanism in this species conserves heterozygosity, additional females that are heterozygous are not very informative here. Indeed, one would need diploid males in these other clonal lines as well (but such males have not yet been found) to make any inference regarding this locus in other lines.

We agree that a full mapping study including diploid males from all clonal lines would be preferable, but as stated earlier in that same paragraph, we have only found diploid males from clonal line A. We stand behind our modest claim that “Females from all six clonal lines were heterozygous at the CSD index peak, consistent with its putative role as a CSD locus in all O. biroi.” In the revised manuscript version, this sentence (in what will be lines 199-201) will be changed slightly in response to a reviewer comment below: “All females from all six clonal lines (including 26 diploid females from clonal line B) were heterozygous at the CSD index peak, consistent with its putative role as a CSD locus in all O. biroi.”

Reviewer #2 (Public review):

The manuscript by Lacy et al. is well written, with a clear and compelling introduction that effectively conveys the significance of the study. The methods are appropriate and well-executed, and the results, both in the main text and supplementary materials, are presented in a clear and detailed manner. The authors interpret their findings with appropriate caution.

This work makes a valuable contribution to our understanding of the evolution of complementary sex determination (CSD) in ants. In particular, it provides important evidence for the ancient origin of a non-coding locus implicated in sex determination, and shows that, remarkably, this sex locus is conserved even in an ant species with a non-canonical reproductive system that typically does not produce males. I found this to be an excellent and well-rounded study, carefully analyzed and well contextualized.

That said, I do have a few minor comments, primarily concerning the discussion of the potential 'ghost' CSD locus. While the authors acknowledge (line 367) that they currently have no data to distinguish among the alternative hypotheses, I found the evidence for an additional CSD locus presented in the results (lines 261-302) somewhat limited and at times a bit difficult to follow. I wonder whether further clarification or supporting evidence could already be extracted from the existing data. Specifically:

We agree with Reviewer 2 that the evidence for a second CSD locus is limited. In fact, we do not intend to advocate for there being a second locus, but we suggest that a second CSD locus is one possible explanation for the absence of diploid males outside of clonal line A. In our initial version, we intentionally conveyed this ambiguity by titling this section “O. biroi may have one or multiple sex determination loci.” However, we now see that this leads to undue emphasis on the possibility of a second locus. In the revised manuscript, we will split this into two separate sections: “Diploid male production differs across O. biroi clonal lines” and “O. biroi lacks a tra-containing CSD locus.”

(1) Line 268: I doubt the relevance of comparing the proportion of diploid males among all males between lines A and B to infer the presence of additional CSD loci. Since the mechanisms producing these two types of males differ, it might be more appropriate to compare the proportion of diploid males among all diploid offspring. This ratio has been used in previous studies on CSD in Hymenoptera to estimate the number of sex loci (see, for example, Cook 1993, de Boer et al. 2008, 2012, Ma et al. 2013, and Chen et al., 2021). The exact method might not be applicable to clonal raider ants, but I think comparing the percentage of diploid males among the total number of (diploid) offspring produced between the two lineages might be a better argument for a difference in CSD loci number.

We want to re-emphasize here that we do not wish to advocate for there being two CSD loci in O. biroi. Rather, we want to explain that this is one possible explanation for the apparent absence of diploid males outside of clonal line A. We hope that the modifications to the manuscript described in the previous response help to clarify this.

Reviewer 2 is correct that comparing the number of diploid males to diploid females does not apply to clonal raider ants. This is because males are vanishingly rare among the vast numbers of females produced. We do not count how many females are produced in laboratory stock colonies, and males are sampled opportunistically. Therefore, we cannot report exact numbers. However, we will add the following sentence to the revised manuscript: “Despite the fact that we maintain more colonies of clonal line B than of clonal line A in the lab, all the diploid males we detected came from clonal line A.”

(2) If line B indeed carries an additional CSD locus, one would expect that some females could be homozygous at the ANTSR locus but still viable, being heterozygous only at the other locus. Do the authors detect any females in line B that are homozygous at the ANTSR locus? If so, this would support the existence of an additional, functionally independent CSD locus.

We thank the reviewer for this suggestion, and again we emphasize that we do not want to argue in favor of multiple CSD loci. We just want to introduce it as one possible explanation for the absence of diploid males outside of clonal line A.

The 26 sequenced diploid females from clonal line B are all heterozygous at the mapped locus, and the revised manuscript will clarify this in what will be lines 199-201. Previously, only six of those diploid females were included in Supplementary Table S2, and that will be modified accordingly.

(3) Line 281: The description of the two tra-containing CSD loci as "conserved" between Vollenhovia and the honey bee may be misleading. It suggests shared ancestry, whereas the honey bee csd gene is known to have arisen via a relatively recent gene duplication from fem/tra (10.1038/nature07052). It would be more accurate to refer to this similarity as a case of convergent evolution rather than conservation.

In the sentence that Reviewer 2 refers to, we are representing the assertion made in the (Miyakawa and Mikheyev 2015) paper in which, regarding their mapping of a candidate CSD locus that contains two linked tra homologs, they write in the abstract: “these data support the prediction that the same CSD mechanism has indeed been conserved for over 100 million years.” In that same paper, Miyakawa and Mikheyev write in the discussion section: “As ants and bees diverged more than 100 million years ago, sex determination in honey bees and V. emeryi is probably homologous and has been conserved for at least this long.”

As noted by Reviewer 2, this appears to conflict with a previously advanced hypothesis: that because fem and csd were found in Apis mellifera, Apis cerana, and Apis dorsata, but only fem was found in Mellipona compressipes, Bombus terrestris, and Nasonia vitripennis, that the csd gene evolved after the honeybee (Apis) lineage diverged from other bees (Hasselmann et al. 2008). However, it remains possible that the csd gene evolved after ants and bees diverged from N. vitripennis, but before the divergence of ants and bees, and then was subsequently lost in B. terrestris and M. compressipes. This view was previously put forward based on bioinformatic identification of putative orthologs of csd and fem in bumblebees and in ants [(Schmieder et al. 2012), see also (Privman et al. 2013)]. However, subsequent work disagreed and argued that the duplications of tra found in ants and in bumblebees represented convergent evolution rather than homology (Koch et al. 2014). Distinguishing between these possibilities will be aided by additional sex determination locus mapping studies and functional dissection of the underlying molecular mechanisms in diverse Aculeata.

Distinguishing between these competing hypotheses is beyond the scope of our paper, but the revised manuscript will include additional text to incorporate some of this nuance. We will include these modified lines below:

“A second QTL region identified in V. emeryi (V.emeryiCsdQTL1) contains two closely linked tra homologs, similar to the closely linked honeybee tra homologs, csd and fem (Miyakawa and Mikheyev 2015). This, along with the discovery of duplicated tra homologs that undergo concerted evolution in bumblebees and ants (Schmieder et al. 2012; Privman et al. 2013) has led to the hypothesis that the function of tra homologs as CSD loci is conserved with the csd-containing region of honeybees (Schmieder et al. 2012; Miyakawa and Mikheyev 2015). However, other work has suggested that tra duplications occurred independently in honeybees, bumblebees, and ants (Hasselmann et al. 2008; Koch et al. 2014), and it remains to be demonstrated that either of these tra homologs acts as a primary CSD signal in V. emeryi.”

(4) Finally, since the authors successfully identified multiple alleles of the first CSD locus using previously sequenced haploid males, I wonder whether they also observed comparable allelic diversity at the candidate second CSD locus. This would provide useful supporting evidence for its functional relevance.

As is already addressed in the final paragraph of the results and in Supplementary Fig. S4, there is no peak of nucleotide diversity in any of the regions homologous to V.emeryiQTL1, which is the tra-containing candidate sex determination locus (Miyakawa and Mikheyev 2015). In the revised manuscript, the relevant lines will be 307-310. We want to restate that we do not propose that there is a second candidate CSD locus in O. biroi, but we simply raise the possibility that multi-locus CSD *might* explain the absence of diploid males from clonal lines other than clonal line A (as one of several alternative possibilities).

Overall, these are relatively minor points in the context of a strong manuscript, but I believe addressing them would improve the clarity and robustness of the authors' conclusions.

Reviewer #3 (Public review):

Summary:

The sex determination mechanism governed by the complementary sex determination (CSD) locus is one of the mechanisms that support the haplodiploid sex determination system evolved in hymenopteran insects. While many ant species are believed to possess a CSD locus, it has only been specifically identified in two species. The authors analyzed diploid females and the rarely occurring diploid males of the clonal ant Ooceraea biroi and identified a 46 kb CSD candidate region that is consistently heterozygous in females and predominantly homozygous in males. This region was found to be homologous to the CSD locus reported in distantly related ants. In the Argentine ant, Linepithema humile, the CSD locus overlaps with an lncRNA (ANTSR) that is essential for female development and is associated with the heterozygous region (Pan et al. 2024). Similarly, an lncRNA is encoded near the heterozygous region within the CSD candidate region of O. biroi. Although this lncRNA shares low sequence similarity with ANTSR, its potential functional involvement in sex determination is suggested. Based on these findings, the authors propose that the heterozygous region and the adjacent lncRNA in O. biroi may trigger female development via a mechanism similar to that of L. humile. They further suggest that the molecular mechanisms of sex determination involving the CSD locus in ants have been highly conserved for approximately 112 million years. This study is one of the few to identify a CSD candidate region in ants and is particularly noteworthy as the first to do so in a parthenogenetic species.

Strengths:

(1) The CSD candidate region was found to be homologous to the CSD locus reported in distantly related ant species, enhancing the significance of the findings.

(2) Identifying the CSD candidate region in a parthenogenetic species like O. biroi is a notable achievement and adds novelty to the research.

Weaknesses

(1) Functional validation of the lncRNA's role is lacking, and further investigation through knockout or knockdown experiments is necessary to confirm its involvement in sex determination.

See response below.

(2) The claim that the lncRNA is essential for female development appears to reiterate findings already proposed by Pan et al. (2024), which may reduce the novelty of the study.

We do not claim that the lncRNA is essential for female development in O. biroi, but simply mention the possibility that, as in L. humile, it is somehow involved in sex determination. We do not have any functional evidence for this, so this is purely based on its genomic position immediately adjacent to our mapped candidate region. We agree with the reviewer that the study by Pan et al. (2024) decreases the novelty of our findings. Another way of looking at this is that our study supports and bolsters previous findings by partially replicating the results in a different species.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors have used full-length single-cell sequencing on a sorted population of human fetal retina to delineate expression patterns associated with the progression of progenitors to rod and cone photoreceptors. They find that rod and cone precursors contain a mix of rod/cone determinants, with a bias in both amounts and isoform balance likely deciding the ultimate cell fate. Markers of early rod/cone hybrids are clarified, and a gradient of lncRNAs is uncovered in maturing cones. Comparison of early rods and cones exposes an enriched MYCN regulon, as well as expression of SYK, which may contribute to tumor initiation in RB1 deficient cone precursors.

Strengths:

(1) The insight into how cone and rod transcripts are mixed together at first is important and clarifies a long-standing notion in the field.

(2) The discovery of distinct active vs inactive mRNA isoforms for rod and cone determinants is crucial to understanding how cells make the decision to form one or the other cell type. This is only really possible with full-length scRNAseq analysis.

(3) New markers of subpopulations are also uncovered, such as CHRNA1 in rod/cone hybrids that seem to give rise to either rods or cones.

(4) Regulon analyses provide insight into key transcription factor programs linked to rod or cone fates.

(5) The gradient of lncRNAs in maturing cones is novel, and while the functional significance is unclear, it opens up a new line of questioning around photoreceptor maturation.

(6) The finding that SYK mRNA is naturally expressed in cone precursors is novel, as previously it was assumed that SYK expression required epigenetic rewiring in tumors.

We thank the reviewer for describing the study’s strengths, reflecting the major conclusions of the initially submitted manuscript.  However, based on new analyses – including the requested analyses of other scRNA-seq datasets, our revision clarifies that:

-  related to point (1), cone and rod transcripts do not appear to be mixed together at first (i.e., in immediately post-mitotic immature cone and rod precursors) but appear to be coexpressed in subsequent cone and rod precursor stages; and 

- related to point (3), CHRNA1 appears to mark immature cone precursors that are distinct from the maturing cone and rod precursors that co-express cone- and rod-related RNAs (despite the similar UMAP positions of the two populations in our dataset). 

Weaknesses:

(1) The writing is very difficult to follow. The nomenclature is confusing and there are contradictory statements that need to be clarified.

(2) The drug data is not enough to conclude that SYK inhibition is sufficient to prevent the division of RB1 null cone precursors. Drugs are never completely specific so validation is critical to make the conclusion drawn in the paper.

We thank the reviewer for noting these important issues. Accordingly, in the revised manuscript:

(1) We improve the writing and clarify the nomenclature and contradictory statements, particularly those noted in the Reviewer’s Recommendations for Authors. 

(2) We scale back claims related to the role of SYK in the cone precursor response to RB1 loss, with wording changes in the Abstract, Results, and Discussion, which now recognize that the inhibitor studies only support the possibility that cone-intrinsic SYK expression contributes to retinoblastoma initiation, as detailed in our responses to Reviewer’s Recommendations for Authors. We agree and now mention that genetic perturbation of SYK is required to prove its role.  

Reviewer #2 (Public review):

Summary:

The authors used deep full-length single-cell sequencing to study human photoreceptor development, with a particular emphasis on the characteristics of photoreceptors that may contribute to retinoblastoma.

Strengths:

This single-cell study captures gene regulation in photoreceptors across different developmental stages, defining post-mitotic cone and rod populations by highlighting their unique gene expression profiles through analyses such as RNA velocity and SCENIC. By leveraging fulllength sequencing data, the study identifies differentially expressed isoforms of NRL and THRB in L/M cone and rod precursors, illustrating the dynamic gene regulation involved in photoreceptor fate commitment. Additionally, the authors performed high-resolution clustering to explore markers defining developing photoreceptors across the fovea and peripheral retina, particularly characterizing SYK's role in the proliferative response of cones in the RB loss background. The study provides an in-depth analysis of developing human photoreceptors, with the authors conducting thorough analyses using full-length single-cell RNA sequencing. The strength of the study lies in its design, which integrates single-cell full-length RNA-seq, longread RNA-seq, and follow-up histological and functional experiments to provide compelling evidence supporting their conclusions. The model of cell type-dependent splicing for NRL and THRB is particularly intriguing. Moreover, the potential involvement of the SYK and MYC pathways with RB in cone progenitor cells aligns with previous literature, offering additional insights into RB development.

We thank the reviewer for summarizing the main findings and noting the compelling support for the conclusions, the intriguing cell type-dependent splicing of rod and cone lineage factors, and the insights into retinoblastoma development.  

Weaknesses:

The manuscript feels somewhat unfocused, with a lack of a strong connection between the analysis of developing photoreceptors, which constitutes the bulk of the manuscript, and the discussion on retinoblastoma. Additionally, given the recent publication of several single-cell studies on the developing human retina, it is important for the authors to cross-validate their findings and adjust their statements where appropriate.

We agree that the manuscript covers a range of topics resulting from the full-length scRNAseq analyses and concur that some studies of developing photoreceptors were not well connected to retinoblastoma. However, we also note that the connection to retinoblastoma is emphasized in several places in the Introduction and throughout the manuscript and was a significant motivation for pursuing the analyses. We suggest that it was valuable to highlight how deep, fulllength scRNA-seq of developing retina provides insights into retinoblastoma, including i) the similar biased expression of NRL transcript isoforms in cone precursors and RB tumors, ii) the cone precursors’ co-expression of rod- and cone-related genes such as NR2E3 and GNAT2, which may explain similar co-expression in RB cells, and iii) the expression of  SYK in early cones and RB cells.  While the earlier version had mainly highlighted point (iii), the revised Discussion further refers to points (i) and (ii) as described further in the response to the Reviewer’s Recommendations for Authors. 

We address the Reviewer’s request to cross-validate our findings with those of other single-cell studies of developing human retina by relating the different photoreceptor-related cell populations identified in our study to those characterized by Zuo et al (PMID 39117640), which was specifically highlighted by the reviewer and is especially useful for such cross-validation given the extraordinarily large ~ 220,000 cell dataset covering a wide range of retinal ages (pcw 8–23) and spatiotemporally stratified by macular or peripheral retina location. Relevant analyses of the Zuo et al dataset are shown in Supplementary Figures S3G-H, S10B, S11A-F, and S13A,B. 

Reviewer #3 (Public review):

Summary:

The authors use high-depth, full-length scRNA-Seq analysis of fetal human retina to identify novel regulators of photoreceptor specification and retinoblastoma progression.

Strengths:

The use of high-depth, full-length scRNA-Seq to identify functionally important alternatively spliced variants of transcription factors controlling photoreceptor subtype specification, and identification of SYK as a potential mediator of RB1-dependent cell cycle reentry in immature cone photoreceptors.

Human developing fetal retinal tissue samples were collected between 13-19 gestational weeks and this provides a substantially higher depth of sequencing coverage, thereby identifying both rare transcripts and alternative splice forms, and thereby representing an important advance over previous droplet-based scRNA-Seq studies of human retinal development.

Weaknesses:

The weaknesses identified are relatively minor. This is a technically strong and thorough study, that is broadly useful to investigators studying retinal development and retinoblastoma.

We thank the reviewer for describing the strengths of the study. Our revision addresses the concerns raised separately in the Reviewer’s Recommendations for Authors, as detailed in the responses below.  

Recommendations for the authors:

Reviewing Editor Comments:

The reviewers have completed their reviews. Generally, they note that your work is important and that the evidence is generally convincing. The reviewers are in general agreement that the paper adds to the field. The findings of rod/cone fate determination at a very early stage are intriguing. Generally, the paper would benefit from clarifications in the writing and figures. Experimentally, the paper would benefit from validation of the drug data, for example using RNAi or another assay. Alternatively, the authors could note the caveats of the drug experiments and describe how they could be improved. In terms of analysis, the paper would be improved by additional comparisons of the authors' data to previously published datasets.

We thank the reviewing editor for this summary. As described in the individual reviewer responses, we clarify the writing and figures and provide comparisons to previously published datasets (in particular, the large snRNA-seq dataset of Zuo et al., 2024 (PMID 39117640).  With regard to the drug (i.e., SYK inhibitor) studies, we opted to provide caveats and describe the need for genetic approaches to validate the role of SYK, owing to the infeasibility of completing genetic perturbation experiments in the appropriate timeframe.  We are grateful for the opportunity to present our findings with appropriate caveats. 

Reviewer #1 (Recommendations for the authors):

Shayler cell sort human progenitor/rod/cone populations then full-length single cell RNAseq to expose features that distinguish paths towards rods or cones. They initially distinguish progenitors (RPCs), immature photoreceptor precursors (iPRPs), long/medium wavelength (LM) cones, late-LM cones, short wavelength (S) cones, early rods (ER) and late rods (LR), which exhibit distinct transcription factor regulons (Figures 1, 2). These data expose expected and novel enriched genes, and support the notion that S cones are a default state lacking expression of rod (NRL) or cone (THRB) determinants but retaining expression of generic photoreceptor drivers (CRX/OTX2/NEUROD1 regulons). They identify changes in regulon activity, such as increasing NRL activity from iPRP to ER to LR, but decreasing from iPRP to cones, or increasing RAX/ISL2/THRB regulon activity from iPRP to LM cones, but decreasing from iPRP to S cones or rods.

They report co-expression of rod/cone determinants in LM and ER clusters, and the ratios are in the expected directions (NRLTHRB or RXRG in ER). A novel insight from the FL seq is that there are differing variants generated in each cell population. Full-length NRL (FL-NRL) predominates in the rod path, whereas truncated NRL (Tr-NRL) does so in the cone path, then similar (but opposite) findings are presented for THRB (Fig 3, 4), whereas isoforms are not a feature of RXRG expression, just the higher expression in cones.

The authors then further subcluster and perform RNA velocity to uncover decision points in the tree (Figure 5). They identify two photoreceptor precursor streams, the Transitional Rods (TRs) that provide one source for rod maturation and (reusing the name from the initial clustering) iPRPs that form cones, but also provide a second route to rods. TR cells closest to RPCs (immediately post-mitotic) have higher levels of the rod determinant NR2E3 and NRL, whereas the higher resolution iPRPs near RPCs lack NR2E3 and have higher levels of ONECUT1, THRB, and GNAT2, a cone bias. These distinct rod-biased TR and cone-biased high-resolution iPRPs were not evident in published scRNAseq with 3′ end-counting (i.e. not FL seq). Regulon analysis confirmed higher NRL activity in TR cells, with higher THRB activity in highresolution iPRP cells.

Many of the more mature high-resolution iPRPs show combinations of rod (GNAT1, NR2E3) and cone (GNAT2, THRB) paths as well as both NRL and THRB regulons, but with a bias towards cone-ness (Figure 6). Combined FISH/immunofluorescence in fetal retina uncovers cone-biased RXRG-protein-high/NR2E3-protein-absent cone-fated cells that nevertheless expressed NR2E3 mRNA. Thus early cone-biased iPRP cells express rod gene mRNA, implying a rod-cone hybrid in early photoreceptor development. The authors refer to these as "bridge region iPRP cells".

In Figure 7, they identify CHRNA1 as the most specific marker of these bridge cells (overlapping with ATOH7 and DLL3, previously linked to cone-biased precursors), and FISH shows it is expressed in rod-biased NRL protein-positive and cone-biased RXRG proteinpositive cones at fetal week 12.

Figure 8 outlines the graded expression of various lncRNAs during cone maturation, a novel pattern.

Finally (Figure 9), the authors identify differential genes expressed in early rods (ER cluster from Figure 1) vs early cones (LM cluster, excluding the most mature opsin+ cells), revealing high levels of MYCN targets in cones. They also find SYK expression in cones. SYK was previously linked to retinoblastoma, so intrinsic expression may predispose cone precursors to transformation upon RB loss. They finish by showing that a SYK inhibitor blocks the proliferation of dividing RB1 knockdown cone precursors in the human fetal retina.

Overall, the authors have uncovered interesting patterns of biased expression in cone/rod developmental paths, especially relating to the isoform differences for NRL and THRB which add a new layer to our understanding of this fate choice. The analyses also imply that very soon after RPCs exit the cell cycle, they generate post-mitotic precursors biased towards a rod or cone fate, that carry varying proportions of mixed rod/cone determinants and other rod/cone marker genes. They also introduce new markers that may tag key populations of cells that precede the final rod/cone choice (e.g. CHRNA1), catalogue a new lncRNA gradient in cone maturation, and provide insight into potential genes that may contribute to retinoblastoma initiation, like SYK, due to intrinsic expression in cone precursors. However, as detailed below, the text needs to be improved considerably, and overinterpretations need to be moderated, removed, or tested more rigorously with extra data.

Major Comments

The manuscript is very difficult to follow. The nomenclature is at times torturous, and the description of hybrid rod/cone hybrid cells is confusing in many aspects.

(1) A single term, iPRP, is used to refer to an initial low-resolution cluster, and then to a subset of that cluster later in the paper.

We agree that using immature photoreceptor precursor (iPRP) for both high-resolution and lowresolution clusters was confusing. We kept this name for the low-resolution cluster (which includes both immature cone and immature rod precursors), renamed the high-resolution iPRP cluster immature cone precursors (iCPs). and renamed their transitional rod (TR) counterparts immature rod precursors (iRPs). These designations are based on 

- the biased expression of THRB, ONECUT1, and the THRB regulon in iCPs (Fig. 5D,E);

- the biased expression of NRL, NR2E3, and NRL regulon iRPs (Fig. 5D,E);

- the partially distinct iCP and iRP UMAP positions (Figure 5C); and 

- the evidence of similar immature cone versus rod precursor populations in the Zuo et al 3’ snRNA-seq dataset, as noted below and described in two new paragraphs starting at the bottom of p. 12.

(2) To complicate matters further, the reader needs to understand the subset within the iPRP referred to as bridge cells, and we are told at one point that the earliest iPRPs lack NR2E3, then that they later co-express NR2E3, and while the authors may be referring to protein and RNA, it serves to further confuse an already difficult to follow distinction. I had to read and re-read the iPRP data many times, but it never really became totally clear.

We agree that the description of the high-resolution iPRP (now “iCP”) subsets was unclear, although our further analyses of a large 3’ snRNA-seq dataset in Figure S11 support the impression given in the original manuscript that the earliest iCPs lack NR2E3 and then later coexpress NR2E3 while the earliest iRPs lack THRB and then later express THRB. As described in new text in the Two post-mitotic immature photoreceptor precursor populations section (starting on line 7 of p. 13): 

When considering only the main cone and rod precursor UMAP regions, early (pcw 8 – 13) cone precursors expressed THRB and lacked NR2E3 (Figure S11D,E, blue arrows), while early (pcw 10 – 15) rod precursors expressed NR2E3 and lacked THRB (Figure S11D,E, red arrows), similar to RPC-localized iCPs and iRPs in our study (Figure 5D).

Next, as summarized in new text in the Early cone and rod precursors with rod- and conerelated RNA co-expression section (new paragraph at top of p. 16): 

Thus, a 3’ snRNA-seq analysis confirmed the initial production of immature photoreceptor precursors with either L/M cone-precursor-specific THRB or rod-precursor-specific NR2E3 expression, followed by lower-level co-expression of their counterparts, NR2E3 in cone precursors and THRB in rod precursors. However, in the Zuo et al. analyses, the co-expression was first observed in well-separated UMAP regions, as opposed to a region that bridges the early cone and early rod populations in our UMAP plots. These findings are consistent with the notion that cone- and rod-related RNA co-expression begins in already fate-determined cone and rod precursors, and that such precursors aberrantly intermixed in our UMAP bridge region due to their insufficient representation in our dataset.  

Importantly, and as noted in our ‘Public response’ to Reviewer 1, “CHRNA1 appears to mark immature cone precursors that are distinct from the maturing cone and rod precursors that coexpress cone- and rod-related RNAs (despite the similar UMAP positions of the two populations in our dataset).” In support of this notion, the immature cone precursors expressing CHRNA1  and other  populations did not overlap in UMAP space in the Zuo et al dataset. We hope the new text cited above along with other changes will significantly clarify the observations.

(3) The term "cone/rod precursor" shows up late in the paper (page 12), but it was clear (was it not?) much earlier in this manuscript that cone and rod genes are co-expressed because of the coexpressed NRL and THRB isoforms in Figures 3/4.

We thank the reviewer for noting that the differential NRL and THRB isoform expression already implies that cone and rod genes are co-expressed. However, as we now state, the co-expression of RNAs encoding an additional cone marker (GNAT2) and rod markers (GNAT1, NR2E3) was 

“suggestive of a proposed hybrid cone/rod precursor state more extensive than implied by the coexpression of different THRB and NRL isoforms” (first paragraph of “Early cone and rod …” section on p. 14; new text underlined). 

(4) The (incorrect) impression given later in the manuscript is that the rod/cone transcript mixture applies to just a subset of the iPRP cells, or maybe just the bridge cells (writing is not clear), but actually, neither of those is correct as the more abundant and more mature LM and ER populations analyzed earlier coexpress NRL and THRB mRNAs (Figures 2, 3). Overall, the authors need to vastly improve the writing, simplify/clarify the nomenclature, and better label figures to match the text and help the reader follow more easily and clearly. As it stands, it is, at best, obtuse, and at worst, totally confusing.

We thank the reviewer for bringing the extent of the confusing terminology and wording to our attention. We revised the terminology (as in our response to point 1) and extensively revised the text.  We also performed similar analyses of the Zuo et al. data (as described in more detail in our response to Reviewer 2), which clarifies the distinct status of cells with the “rod/cone transcript mixture” and cells co-expressing early cone and rod precursor markers.  

To more clearly describe data related to cells with rod- and cone-related RNA co-expression, we divided the former Figure 6 into two figures, with Figure 6 now showing the cone- and rodrelated RNA co-expression inferred from scRNA-seq and Figure 7 showing GNAT2 and NR2E3 co-expression in FISH analyses of human retina plus a new schematic in the new panel 7E.

To separate the conceptually distinct analyses of cone and rod related RNA co-expression and the expression of early photoreceptor precursor markers (which were both found in the so-called bridge region – yet now recognized to be different subpopulations), we separated the analyses of the early photoreceptor precursor markers to form a new section, “Developmental expression of photoreceptor precursor markers and fate determinants,” starting on p. 16. 

Additionally, we further review the findings and their implications in four revised Discussion paragraphs starting at the bottom of p. 23).

(5) The data showing that overexpressing Tr-NRL in murine NIH3T3 fibroblasts blocks FL-NRL function is presented at the end of page 7 and in Figure 3G. Subsequent analysis two paragraphs and two figures later (end page 8, Figure 5C + supp figs) reveal that Tr-NRL protein is not detectable in retinoblastoma cells which derive from cone precursors cells and express Tr-NRL mRNA, and the protein is also not detected upon lentiviral expression of Tr-NRL in human fetal retinal explants, suggesting it is unstable or not translated. It would be preferable to have the 3T3 data and retinoblastoma/explant data juxtaposed. E.g. they could present the latter, then show the 3T3 that even if it were expressed (e.g. briefly) it would interfere with FL-NRL. The current order and spacing are somewhat confusing.

We thank the reviewer for this suggestion and moved the description of the luciferase assays to follow the retinoblastoma and explant data and switched the order of Figure panels 3G and 3H.  

(6) On page 15, regarding early rod vs early cone gene expression, the authors state: "although MYCN mRNA was not detected....", yet on the volcano plot in Figure S14A MYCN is one of the marked genes that is higher in cones than rods, meaning it was detected, and a couple of sentences later: "Concordantly, the LM cluster had increased MYCN RNA". The text is thus confusing.

With respect, we note that the original text read, “although MYC RNA was not detected,” which related to a statement in the previous sentence that the gene ontology analysis identified “MYC targets.” However, given that this distinction is subtle and may be difficult for readers to recognize, we revised the text (now on p. 19) to more clearly describe expression of MYCN (but not MYC) as follows:

“The upregulation of MYC target genes was of interest given that many MYC target genes are also targets of MYCN, that MYCN protein is highly expressed in maturing (ARR3+) cone precursors but not in NRL+ rods (Figure 10A), and that MYCN is critical to the cone precursor proliferative response to pRB loss8–10.  Indeed, whereas MYC RNA was not detected, the LM cone cluster had increased MYCN RNA …”

(7) The authors state that the SYK drug is "highly specific". They provide no evidence, but no drug is 100% specific, and it is possible that off-target hits are important for the drug phenotype. This data should be removed or validated by co-targeting the SYK gene along with RB1.

We agree that our data only show the potential for SYK to contribute to the cone proliferative response; however, we believe the inhibitor study retains value in that a negative result (no effect of the SYK inhibitor) would disprove its potential involvement. To reflect this, we changed wording related to this experiment as follows:

In the Abstract, we changed:

(1) “SYK, which contributed to the early cone precursors’ proliferative response to RB1 loss” To: “SYK, which was implicated in the early cone precursors’ proliferative response to RB1 loss.”  

(2) “These findings reveal … and a role for early cone-precursor-intrinsic SYK expression.” To:  “These findings reveal … and suggest a role for early cone-precursor-intrinsic SYK expression.”

In the last paragraph of the Results, we changed:

(1) “To determine if SYK contributes…” To:  “To determine if SYK might contribute…”

(2) “the highly specific SYK inhibitor” To:  “the selective SYK inhibitor”  

(3)  “indicating that cone precursor intrinsic SYK activity is critical to the proliferative response” To: “consistent with the notion that cone precursor intrinsic SYK activity contributes to the proliferative response.”

In the Results, we added a final sentence: 

“However, given potential SYK inhibitor off-target effects, validation of the role of SYK in retinoblastoma initiation will require genetic ablation studies.”

In the Discussion (2nd-to-last paragraph), we changed: 

“SYK inhibition impaired pRB-depleted cone precursor cell cycle entry, implying that native SYK expression rather than de novo induction contributes to the cone precursors’ initial proliferation.” To: “…the pRB-depleted cone precursors’ sensitivity to a SYK inhibitor suggests that native SYK expression rather than de novo induction contributes to the cone precursors’ initial proliferation, although genetic ablation of SYK is needed to confirm this notion.” In the Discussion last sentence, we changed:

“enabled the identification of developmental stage-specific cone precursor features that underlie retinoblastoma predisposition.” To: “enabled the identification of developmental stage-specific cone precursor features that are associated with the cone precursors’ predisposition to form retinoblastoma tumors.”

Minor/Typos

Figure 7 legend, H should be D.

We corrected the figure legend (now related to Figure 8).

Reviewer #2 (Recommendations for the authors):

(1) The author should take advantage of recently published human fetal retina data, such as PMID:39117640, which includes a larger dataset of cells that could help validate the findings. Consequently, statements like "To our knowledge, this is the first indication of two immediately post-mitotic photoreceptor precursor populations with cone versus rod-biased gene expression" may need to be revised.

We thank the reviewer for noting the evidence of distinct immediately post-mitotic rod and cone populations published by others after we submitted our manuscript. In response, we omitted the sentence mentioned and extensively cross-checked our results including:

- comparison of our early versus late cone and rod maturation states to the cone and rod precursor versus cone and rod states identified by Zuo et al (new paragraph on the top half of p. 6 and new figure panels S3G,H);

- detection of distinct immediately post-mitotic versus later cone and rod precursor populations (two new paragraphs on pp. 12-13 and new Figures S10B and S11A-E); 

- identification of cone and rod precursor populations that co-express cone and rod marker genes (two new paragraphs starting at the bottom of p. 15 and new Figures S11D-F);

- comparison of expression patterns of immature cone precursor (iCP) marker genes in our and the Zuo et al dataset (new paragraph on top half of p. 17 and new Figure S13).

We also compare the cell states discerned in our study and the Zuo et al. study in a new Discussion paragraph (bottom of p. 23) and new Figure S17.

(2) The data generated comes from dissociated cells, which inherently lack spatial context. Additionally, it is unclear whether the dataset represents a pool of retinas from multiple developmental stages, and if so, whether the developmental stage is known for each cell profiled. If this information is available, the authors should examine the distribution of developmental stages on the UMAP and trajectory analysis as part of the quality control process. 

We thank the reviewer for highlighting the importance of spatial context and developmental stage. 

Related to whether the dataset represents a pool of retinae from multiple developmental stages, the different cell numbers examined at each time point are indicated in Figure S1A. To draw the readers’ attention to this detail, Figure S1A is now cited in the first sentence of the Results. 

Related to the age-related cell distributions in UMAP plots, the distribution of cells from each retina and age was (and is) shown in Fig. S1F. In addition, we now highlight the age distributions by segregating the FW13, FW15-17, and FW17-18-19 UMAP positions in the new Figure 1C. We describe the rod temporal changes in a new sentence at the top of  p. 5:

“Few rods were detected at FW13, whereas both early and late rods were detected from FW15-19 (Figure 1C), corroborating prior reports [15,20].”  

We describe the cone temporal changes and note the likely greater discrimination of cell state changes that would be afforded by separately analyzing macula versus peripheral retina at each age in a new sentence at the bottom of p. 5:

“L/M cone precursors from different age retinae occupied different UMAP regions, suggesting age-related differences in L/M cone precursor maturation (Figure 1C).”

Moreover, they should assess whether different developmental stages impact gene expression and isoform ratios. It is well established that cone and rod progenitors typically emerge at different developmental times and in distinct regions of the retina, with minimal physical overlap. Grouping progenitor cells based solely on their UMAP positioning may lead to an oversimplified interpretation of the data.

(2a) We agree that different developmental stages may impact gene expression and isoform ratios, and evaluated stages primarily based on established Louvain clustering rather than UMAP position. However, we also used UMAP position to segregate so-called RPC-localized and nonRPC-localized iCPs and iRPs, as well as to characterize the bridge region iCP sub-populations. In the revision, we examine whether cell groups defined by UMAP positions helped to identify transcriptomically distinct populations and further examine the spatiotemporal gene expression patterns of the same genes in the Zuo et al. 3’ snRNA-seq dataset. 

(2b) Related to analyses of immediately post-mitotic iRPs and iCPs, the new Figure S10A expanded the violin plots first shown in Figure 5D to compare gene expression in RPC-localized versus non-RPC-localized iCPs and iRPs and subsequent cone and rod precursor clusters (also presented in response to Reviewer 3). The new Figure S10C, shows a similar analysis of UMAP region-specific regulon activities. These figures support the idea that there are only subtle UMAP region-related differences in the expression of the selected gene and regulons. 

To further evaluate early cone and rod precursors, we compared expression patterns in our cluster- and UMAP-defined cell groups to those of the spatiotemporally defined cell groups in the Zuo et al. 3’ snRNA-seq study. The results revealed similar expression timing of the genes examined, although the cluster assignments of a subset of cells were brought into question, especially the assigned rod precursors at pcw 10 and 13, as shown in new Figures S10B (grey columns) and S11, and as described in two new paragraphs starting near the bottom of p.12. 

(2c) Related to analyses of iCPs in the so-called bridge region, our analyses of the Zuo et al dataset helped distinguish early cone and rod precursor populations (expressing early markers such as ATOH7 and CHRNA1) from the later stages exhibiting rod- and cone-related gene coexpression, which had intermixed in the UMAP bridge region in our dataset. Further parsing of early cone precursor marker spatiotemporal expression revealed intriguing differences as now described in the second half of a new paragraph at the top of p. 17, as follows:

“Also, different iCP markers had different spatiotemporal expression: CHRNA1 and ATOH7 were most prominent in peripheral retina with ATOH7 strongest at pcw 10 and CHRNA1 strongest at pcw 13; CTC-378H22.2 was prominently expressed from pcw 10-13 in both the macula and the periphery; and DLL3 and ONECUT1 showed the earliest, strongest, and broadest expression (Figure S13B). The distinct patterns suggest spatiotemporally distinct roles for these factors in cone precursor differentiation.”

(3) I would commend the authors for performing a validation experiment via RNA in situ to validate some of the findings. However, drawing conclusions from analyzing a small number of cells can still be dangerous. Furthermore, it is not entirely clear how the subclustering is done. Some cells change cell type identities in the high-resolution plot. For example, some iPRP cells from the low-resolution plots in Figure 1 are assigned as TR in high-resolution plots in Figure 5.

The authors should provide justification on the identifies of RPC localized iPRP and TR.

Comparison of their data with other publicly available data should strengthen their annotation

We agree that drawing conclusions from scRNA-seq or in situ hybridization analysis of a small number of cells can be dangerous and have followed the reviewer’s suggestion to compare our data with other publicly available data, focusing on the 3’ snRNA-seq of Zuo et al. given its large size and extensive annotation. Our analysis of  the Zuo et al. dataset helped clarify cell identities by segregating cone and rod precursors with similar gene expression properties in distinct UMAP regions. However, we noted that the clustering of early cone and rod precursors likely gave numerous mis-assigned cells (as noted in response 2b above and shown in the new Figure S11). It would appear that insights may be derived from the combination of relatively shallow sequencing of a high number of cells and deep sequencing of substantially fewer cells. 

Related to how subclustering was done, the Methods state, “A nearest-neighbors graph was constructed from the PCA embedding and clusters were identified using a Louvain algorithm at low and high resolutions (0.4 and 1.6)[70],” citing the Blondel et al reference for the Louvain clustering algorithm used in the Seurat package.  To clarify this, the results text was revised such that it now indicates the levels used to cluster at low resolution (0.4, p. 4, 2nd paragraph) and at high resolution (1.6, top of p. 11) .

Related to the assignment of some iPRP cells from the low-resolution plots in Figure 1 to the TR cluster (now called the ‘iRP’ ‘cluster) in the high-resolution plots in Figure 5, we suggest that this is consistent with Louvain clustering, which does not follow a single dendrogram hierarchy. 

The justification for referring to these groups as RPC-localized iCPs and iRPs relates to their biased gene and regulon expression in Fig. 5D and 5E, as stated on p. 12: 

“In the RPC-localized region, iCPs had higher ONECUT1, THRB, and GNAT2, whereas iRPs trended towards higher NRL and NR2E3 (p= 0.19, p=0.054, respectively).”

(4) Late-stage LM5 cluster Figure 9 is not defined anywhere in previous figures, in which LM clusters only range from 1 to 4. The inconsistency in cluster identification should be addressed.

We revised the text related to this as follows: 

“Indeed, our scRNA-seq analyses revealed that SYK RNA expression increased from the iCP stage through cluster LM4, in contrast to its minimal expression in rods (Figure 10E).  Moreover, SYK expression was abolished in the five-cell group with properties of late maturing cones (characterized in Figure 1E), here displayed separately from the other LM4 cells and designated LM5 (Figure 10E).”  (p. 19-20)

(5) Syk inhibitor has been shown to be involved in RB cell survival in previous studies. The manuscript seems to abruptly make the connection between the single-cell data to RB in the last figure. The title and abstract should not distract from the bulk of the manuscript focusing on the rod and cone development, or the manuscript should make more connection to retinoblastoma.

We appreciate the reviewer’s concern that the title may seem to over-emphasize the connection to retinoblastoma based solely on the SYK inhibitor studies. However, we suggest the title also emphasizes the identification and characterization of early human photoreceptor states, per se, and that there are a number of important connections beyond the SYK studies that could warrant the mention of cell-state-specific retinoblastoma-related features in the title.

Most importantly, a prior concern with the cone cell-of-origin theory was that retinoblastoma cells express RNAs thought to mark retinal cell types other than cones, especially rods. The evidence presented here, that cone precursors also express the rod-related genes helps resolve this issue. The issue is noted numerous times in the manuscript, as follows:  

In the Introduction, we write:

“However, retinoblastoma cells also express rod lineage factor NRL RNAs, which – along with other evidence – suggested a heretofore unexplained connection between rod gene expression and retinoblastoma development[12,13]. Improved discrimination of early photoreceptor states is needed to determine if co-expression of rod- and cone-related genes is adopted during tumorigenesis or reflects the co-expression of such genes in the retinoblastoma cell of origin.” (bottom, p. 2) And: 

“In this study, we sought to further define the transcriptomic underpinnings of human  photoreceptor development and their relationship to retinoblastoma tumorigenesis.” (last paragraph, p. 3)

The Discussion also alluded to this issue and in the revised Discussion, we aimed to make the connection clearer.  We previously ended the 3rd-to-last paragraph with,  

“iPRP [now iCP] and early LM cone precursors’ expression of NR2E3 and NRL RNAs suggest that their presence in retinoblastomas[12,13] reflects their normal expression in the L/M cone precursor cells of origin.” 

We now separate and elaborate on this point in a new paragraph as follows: 

“Our characterization of cone and rod-related RNA co-expression may help resolve questions about the retinoblastoma cell of origin. Past studies suggested that retinoblastoma cells co-express RNAs associated with rods, cones, or other retinal cells due to a loss of lineage fidelity[12]. However, the early L/M cone precursors’ expression of NR2E3 and NRL RNAs suggest that their presence in retinoblastomas[12,13] reflects their normal expression in the L/M cone precursor cells of origin. This idea is further supported by the retinoblastoma cells’ preferential expression of cone-enriched NRL transcript isoforms (Figure S5B).” (middle of p. 24) Based on the above, we elected to retain the title.  

Minor comments:

(1) It is difficult to see the orange and magenta colors in the Fig 3E RNA-FISH image. The colors should be changed, or the contrast threshold needs to be adjusted to make the puncta stand out more.

We re-assigned colors, with red for FL-NRL puncta and green for Tr-NRL puncta. 

(2) Figure 5C on page 8 should be corrected to Supplementary Figure 5C.

We thank the reviewer for noting this error and changed the figure citation.

Reviewer #3 (Recommendations for the authors):

(1) Minor concerns

a. Abbreviation of some words needs to be included, example: FW. 

We now provide abbreviation definitions for FW and others throughout the manuscript.  

b. Cat # does not matches with the 'key resource table' for many reagents/kits. Some examples are: CD133-PE mentioned on Page # 22 on # 71, SMART-Seq V4 Ultra Low Input RNA Kit and SMARTer Ultra Low RNA Kit for the Fluidigm C1 Sytem on Page # 22 on # 77, Nextera XT DNA Library preparation kit on Page # 23 on # 77.

We thank the reviewer for noting these discrepancies. We have now checked all catalog numbers and made corrections as needed.

c. Cat # and brand name of few reagents & kits is missing and not mentioned either in methods or in key resource table or both. Eg: FBS, Insulin, Glutamine, Penicillin, Streptomycin, HBSS, Quant-iT PicoGreen dsDNA assay, Nextera XT DNA LibraryPreparation Kit, 5' PCR Primer II A with CloneAmp HiFi PCR Premix. 

Catalog numbers and brand names are now provided for the tissue culture and related reagents within the methods text and for kits in the Key Resources Table. Additional descriptions of the primers used for re-amplification and RACE were added to the Methods (p. 28-29).

d. Spell and grammar check is needed throughout the manuscript is needed. Example. In Page # 46 RXRγlo is misspelled as RXRlo.

Spelling and grammar checks were reviewed.

(2) Methods & Key Resource table.

a. In Page # 21, IRB# needs to be stated.      

The IRB protocols have been added, now at top of p. 26.

b. In Page # 21, Did the authors dissociate retinae in ice-cold phosphate-buffered saline or papain?   

The relevant sentence was corrected to “dissected while submerged in ice-cold phosphatebuffered saline (PBS) and dissociated as described10.” ( p. 26)

c. In Page # 21, How did the authors count or enumerate the cell count? Provide the details.

We now state, “… a 10 µl volume was combined with 10 µl trypan blue and counted using a hemocytometer” (top of p. 27)

d. Why did the authors choose to specifically use only 8 cells for cDNA preparation in Page # 22? State the reason and provide the details.

The reasons for using 8 cells (to prevent evaporation and to manually transfer one slide-worth of droplets to one strip of PCR tubes) and additional single cell collection details are now provided as follows (new text underlined): 

“Single cells were sorted on a BD FACSAria I at 4°C using 100 µm nozzle in single-cell mode into each of eight 1.2 µl lysis buffer droplets on parafilm-covered glass slides, with droplets positioned over pre-defined marks … .  Upon collection of eight cells per slide, droplets were transferred to individual low-retention PCR tubes (eight tubes per strip) (Bioplastics K69901, B57801) pre-cooled on ice to minimize evaporation. The process was repeated with a fresh piece of parafilm for up to 12 rounds to collect 96 cells). (p. 27, new text underlined)

e. Key resource table does not include several resources used in this study. Example - NR2E3 antibody.

We added the NR2E3 antibody and checked for other omissions.

(3) Results & Figures & Figure Legends

a. Regulon-defined RPC and photoreceptor precursor states

i. On page # 4, 1 paragraph - Clarify the sentence 'Exclusion of all cells with <100,000 cells read and 18 cells.........Emsembl transcripts inferred'. Did the authors use 18 cells or 18FW retinae? 

The sentence was changed to:

“After sequencing, we excluded all cells with <100,000 read counts and 18 cells expressing one or more markers of retinal ganglion, amacrine, and/or horizontal cells (POU4F1, POU4F2, POU4F3, TFAP2A, TFAP2B, ISL1) and concurrently lacking photoreceptor lineage marker OTX2. This yielded 794 single cells with averages of 3,750,417 uniquely aligned reads, 8,278 genes detected, and 20,343 Ensembl transcripts inferred (Figure S1A-C).” (p. 4, new words underlined)

To clarify that 18 retinae were used, the first sentence of the Results was revised as follows:

“To interrogate transcriptomic changes during human photoreceptor development, dissociated RPCs and photoreceptor precursors were FACS-enriched from 18 retinae, ages FW13-19 …” (p. 4).

Why did the authors 'exclude cells lacking photoreceptor lineage marker OTX2' from analysis especially when the purpose here was to choose photoreceptor precursor states & further results in the next paragraph clearly state that 5 clusters were comprised of cells with OTX2 and CRX expression. This is confusing.

We apologize for the imprecise diction. We divided the evidently confusing sentence into two sentences to more clearly indicate that we removed cells that did not express OTX2, as in the first response to the previous question.

ii. In Page # 5, the authors reported the number of cell populations (363 large and 5 distal) identified in the THRB+ L/M-cone cluster. What were the # of cell populations identified in the remaining 5 clusters of the UMAP space?

We added the cell numbers in each group to Fig. 1B. We corrected the large LM group to 366 cells (p. 5) and note 371 LM cells , which includes the five distal cells, in Figure 1B.

b. Differential expression of NRL and THRB isoforms in rod and cone precursors

i. In Figure 3B, the authors compare and show the presence of 5 different NRL isoforms for all the 6 clusters that were defined in 3A. However, in the results, the ENST# of just 2 highly assigned transcript isoforms is given. What are the annotated names of the three other isoforms which are shown in 3B? Please explain in the Results.

As requested, we now annotate the remaining isoforms as encoding full-length or truncated NRL in Fig. 3B and show isoform structures in new Supplementary Figure S4B.  We also refer to each transcript isoform in the Results (p. 7, last paragraph) and similarly evaluate all isoforms in RB31 cells (Fig. S5B).

ii. What does the Mean FPM in the y-axis of Fig 3C refer to?

Mean FPM represents mean read counts (fragments per million, FPM) for each position across Ensembl NRL exons for each cluster, as now stated in the 6th line of the Fig. 3 legend.

iii. A clear explanation of the results for Figures 3E-3F is missing.

We revised the text to more clearly describe the experiment as follows:

“The cone cells’ higher proportional expression of Tr-NRL first exon sequences was validated by RNA fluorescence in situ hybridization (FISH) of FW16 fetal retina in which NRL immunofluorescence was used to identify rod precursors, RXRg immunofluorescence was used to identify cone precursors, and FISH probes specific to truncated Tr-NRL exon 1T or FL-NRL exons 1 and 2 were used to assess Tr-NRL and FL-NRL expression (Figure 3E,F).” (p. 8, new text underlined).

c. Two post-mitotic photoreceptor precursor populations

i. Although deep-sequencing and SCENIC analysis clarified the identities of four RPC-localized clusters as MG, RPC, and iPRP indicative of cone-bias and TR indicative of rod-bias. It would be interesting to see the discriminating determinant between the TR and ER by SCENIC and deep-sequencing gene expression violin/box plots.

We agree it is of interest to see the discriminating determinant between the TR [now termed iRP] and ER clusters by SCENIC and deep-sequencing gene expression violin/box plots. We now provide this information for selected genes and regulons of interest in the new Supplementary Figures S10A and S10C, along with a similar comparison between the prior high-resolution iPRP (now termed iCP) cluster and the first high-resolution LM cluster, LM1, as described for gene expression on p. 12:

“Notably, THRB and GNAT2 expression did not significantly change while ONECUT1 declined in the subsequent non-RPC-localized iCP and LM1 stages, whereas NR2E3 and NRL dramatically increased on transitioning to the ER state (Figure S10A).”

And as described for regulon activities on pp. 13-14:

“Finally, activities of the cone-specific THRB and ISL2 regulons, the rod-specific NRL regulon, and the pan-photoreceptor LHX3, OTX2, CRX, and NEUROD1 regulons increased to varying extents on transitioning from the immature iCP or iRP states to the early-maturing LM1 or ER states (Figure 10C).”

We also show expression of the same genes for spatiotemporally grouped cells from the Zuo et al. dataset in the new Figure S10B, which displays a similar pattern (apart from the possibly mixed pcw 10 and pcw13 designated rod precursors).

d. Early cone precursors with cone- and rod-related RNA expression

i. On page #12, the last paragraph where the authors explain the multiplex RNA FISH results of RXRγ and NR2E3 by citing Figure S8E. However, in Fig S8E, the authors used NRL to identify the rods. Please clarify which one of the rod markers was used to perform RNA FISH?

Figure S8E (where NRL was used as a rod marker) was cited to remind readers that RXRg has low expression in rods and high expression in cones, rather than to describe the results of this multiplex FISH section. To avoid confusion on this point, Figure S8E is now cited using “(as earlier shown in Figure S8E).” With this issue clarified, we expect the markers used in the FISH + IF analysis will be clear from the revised explanation, 

“… we examined GNAT2 and NR2E3 RNA co-expression in RXRg+ cone precursors in the outermost NBL and in RXRg+ rod precursors in the middle NBL … .” (p. 14-15).

To provide further clarity, we provide a diagram of the FISH probes, protein markers, and expression patterns in the new Figure 7E.

ii. The Y-axis of Fig 6G-6H needs to be labelled.

The axes have been re-labeled from “Nb of cells” to “Number of RXRg+ outermost NBL cells in each region” (original Fig. 6G, now Fig. 7C) and “Number of RXRg+ middle NBL cells in each region” (original Fig. 6H, now Fig. 7D).

iii. The legends of Figures 6G and 6H are unclear. In the Figure 6G legend, the authors indicate 'all cells are NR2E3 protein-'. Does that imply the yellow and green bars alone? Similarly, clarify the Figure 6H legend, what does the dark and light magenta refer to? What does the light magenta color referring to NR2E3+/ NR2E3- and the dark magenta color referring to NR2E3+/ NR2E3+ indicate? 

We regret the insufficient clarity. We revised the Fig. 6G (now Fig. 7C) key, which now reads

“All outermost NBL cells are NR2E3 protein-negative.”  We added to the figure legend for panel 7C,D “(n.b., italics are used for RNAs, non-italics for proteins).”  The new scheme in Figure 7E shows the RNAs in italics proteins in non-italics. We hope these changes will clarify when RNA or protein are represented in each histogram category.

Overall, the results (on page # 13) reflecting Figures 6E-6H & Figure S11 are confusing and difficult to understand. Clear descriptions and explanations are needed.

We revised this results section described in the paragraph now spanning p. 14:

-  We now refer to the bar colors in Figures 7C and 7D that support each statement. 

-  We provide an illustration of the findings in Figure 7E.

iv. Previously published literature has shown that cells of the inner NBL are RXRγ+ ganglion cells. So, how were these RXRγ+ ganglion cells in the inner NBL discriminated during multiplex RNA FISH (in Fig 6E-6H and in Fig S11)?

We thank the reviewer for requesting this clarification. We agree that “inner NBL” is the incorrect term for the region in which we examined RXRg+ photoreceptor precursors, as this could include RXRγ+ nascent RGCs. We now clarify that 

“we examined GNAT2 and NR2E3 RNA co-expression in RXRg+ cone precursors in the outermost NBL and in RXRg+ rod precursors in the middle NBL … .”  (p. 14-15) We further state, 

“Limiting our analysis to the outer and middle NBL allowed us to disregard RXRγ+ retinal ganglion cells in the retinal ganglion cell layer or inner NBL (top of p. 15)”

Figure 7E is provided to further aid the reader in understanding the positions examined, and the legend states “RXRg+ retinal ganglion cells in the inner NBL and ganglion cell layer not shown. 

v. In Figure 6E, what marker does each color cell correspond to?

In this figure (now panel 7A), we declined to provide the color key since the image is not sufficiently enlarged to visualize the IF and FISH signals. The figure is provided solely to document the regions analyzed and readers are now referred to “see Figure S12 for IF + FISH images” (2nd line, p. 15), where the marker colors are indicated.

vi. In Figure S11 & 6E, Protein and RNA transcript color of NR2E3, GNAT2 are hard to distinguish. Usage of other colors is recommended.  

We appreciate the reviewer’s concern related to the colors (in the now redesignated Figure S12 and 7A); however, we feel this issue is largely mitigated by our use of arrows to point to the cells needed to illustrate the proposed concepts in Figure S12B. All quantitation was performed by examining each color channel separately to ensure correct attribution, which is now mentioned in the Methods (2nd-to-last line of Quantitation of FISH section, p. 35).

vii. 

With due respect, we suggest that labeling each box (now in Figure 8B) makes the figure rather busy and difficult to infer the main point, which is that boxed regions were examined at various distanced from the center (denoted by the “C” and “0 mm”) with distances periodically indicated. We suggest the addition of such markers would not improve and might worsen the figure for most readers.    

e. An early L/M cone trajectory marked by successive lncRNA expression

i. In Figure 8C - color-coded labelling of LM1-4 clusters is recommended.

We note Fig. 8C (now 9C) is intended to use color to display the pseudotemporal positions of each cell. We recognize that an additional plot with the pseudotime line imposed on LM subcluster colors could provide some insights, yet we are unaware of available software for this and are unable to develop such software at present. To enable readers to obtain a visual impression of the pseudotime vs subcluster positions, we now refer the reader to Figure 5A in the revised figure legend, as follows:  (“The pseudotime trajectory may be related to LM1-LM4 subcluster distributions in Figure 5A.”).

ii. In Figure 8G - what does the horizontal color-coded bar below the lncRNAs name refer to? These bars are similar in all four graphs of the 8G figure.

As stated in the Fig. 8G (now 9G) legend, “Colored bars mark lncRNA expression regions as described in the text.”  We revised the text to more clearly identify the color code. (p. 18-19)   

f. Cone intrinsic SYK contributions to the proliferative response to pRB loss

i. In Fig 9F - The expression of ARR3+ cells (indicated by the green arrow in FW18) is poorly or rarely seen in the peripheral retina.

We thank the reviewer for finding this oversight. In panel 9F (now 10F), we removed the green arrows from the cells in the periphery, which are ARR3- due to the immaturity of cones in this region. 

ii. In Figure 9F - Did the authors stain the FW16 retina with ARR3?

Unfortunately, we did not stain the FW16 retina for ARR3 in this instance.

iii. Inclusion of DAPI staining for Fig 9F is recommended to justify the ONL & INL in the images.

We regret that we are unable to merge the DAPI in this instance due to the way in which the original staining was imaged.  A more detailed analysis corroborating and extending the current results is in progress. 

iv. Immunostaining images for Figure 9G are missing & are required to be included. What does shSCR in Fig 9G refer to?

We now provide representative immunostaining images below the panel (now 10G). The legend was updated: “Bottom: Example of Ki67, YFP, and RXRg co-immunostaining with DAPI+ nuclei (yellow outlines). Arrows: Ki67+, YFP+, RXRg+ nuclei.”  The revised legend now notes that shSCR refers to the scrambled control shRNA.

v. For Figure 9H - Is the presence and loss of SYK activity consistent with all the subpopulations (S & LM) of early maturing and matured cones?

We appreciate the reviewer’s question and interest (relating to the redesignated Figure 10H); however, we have not yet completed a comprehensive evaluation of SYK expression in all the subpopulations (S & LM) of early maturing and matured cones and will reserve such data for a subsequent study. We suggest that this information is not critical to the study’s major conclusions.

vi. Figure 9A is not explained in the results. Why were MYCN proteins assessed along with ARR3 and NRL? What does this imply?

We thank the reviewer for noting that this figure (now Figure 10A) was not clearly described. 

As per the response to Reviewer 1, point 6 , the text now states,  

“The upregulation of MYC target genes was of interest given that many MYC target genes are also MYCN targets, that MYCN protein is highly expressed in maturing (ARR3+) cone precursors but not in NRL+ rods (Figure 10A), and that MYCN is critical to the cone precursor proliferative response to pRB loss [8–10].” (middle, p. 19, new text underlined).

Hence, the figure demonstrates the cone cell specificity of high MYCN protein.  This is further noted in the Fig. 10a legend: “A. Immunofluorescent staining shows high MYCN in ARR3+ cones but not in NRL+ rods in FW18 retina.”

Author response:

Reviewer #1 (Public review):

Functional lateralization between the right and left hemispheres is reported widely in animal taxa, including humans. However, it remains largely speculative as to whether the lateralized brains have a cognitive gain or a sort of fitness advantage. In the present study, by making use of the advantages of domestic chicks as a model, the authors are successful in revealing that the lateralized brain is advantageous in the number sense, in which numerosity is associated with spatial arrangements of items. Behavioral evidence is strong enough to support their arguments. Brain lateralization was manipulated by light exposure during the terminal phase of incubation, and the left-to-right numerical representation appeared when the distance between items gave a reliable spatial cue. The light-exposure induced lateralization, though quite unique in avian species, together with the lack of intense inter-hemispheric direct connections (such as the corpus callosum in the mammalian cerebrum), was critical for the successful analysis in this study. Specification of the responsible neural substrates in the presumed right hemisphere is expected in future research. Comparable experimental manipulation in the mammalian brain must be developed to address this general question (functional significance of brain laterality) is also expected.

We sincerely appreciate the Reviewer's insightful feedback and his/her recognition of the key contributions of our study.

Reviewer #2 (Public review):

Summary:

This is the first study to show how a L-R bias in the relationship between numerical magnitude and space depends on brain lateralisation, and moreover, how is modulated by in ovo conditions.

Strengths:

Novel methodology for investigating the innateness and neural basis of an L-R bias in the relationship between number and space.

We would like to thank the Reviewer for their valuable feedback and for highlighting the key contributions of our study.

Weaknesses:

I would query the way the experiment was contextualised. They ask whether culture or innate pre-wiring determines the 'left-to-right orientation of the MNL [mental number line]'.

We thank the Reviewer for raising this point, which has allowed us to provide a more detailed explanation of this aspect. Rather than framing the left-to-right orientation of the mental number line (MNL) as exclusively determined by either cultural influences or innate pre-wiring, our study highlights the role of environmental stimulation. Specifically, prenatal light exposure can shape hemispheric specialization, which in turn contributes to spatial biases in numerical processing. Please see lines 115-118.

The term, 'Mental Number Line' is an inference from experimental tasks. One of the first experimental demonstrations of a preference or bias for small numbers in the left of space and larger numbers in the right of space, was more carefully described as the spatialnumerical association of response codes - the SNARC effect (Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental representation of parity and numerical magnitude. Journal of Experimental Psychology: General, 122, 371-396).

We have refined our description of the MNL and SNARC effect to ensure conceptual accuracy in the revised manuscript; please see lines 53-59.

This has meant that the background to the study is confusing. First, the authors note, correctly, that many other creatures, including insects, can show this bias, though in none of these has neural lateralisation been shown to be a cause. Second, their clever experiment shows that an experimental manipulation creates the bias. If it were innate and common to other species, the experimental manipulation shouldn't matter. There would always be an LR bias. Third, they seem to be asserting that humans have a left-to-right (L-R) MNL. This is highly contentious, and in some studies, reading direction affects it, as the original study by Dehaene et al showed; and in others, task affects direction (e.g. Bachtold, D., Baumüller, M., & Brugger, P. (1998). Stimulus-response compatibility in representational space. Neuropsychologia, 36, 731-735, not cited). Moreover, a very careful study of adult humans, found no L-R bias (Karolis, V., Iuculano, T., & Butterworth, B. (2011), not cited, Mapping numerical magnitudes along the right lines: Differentiating between scale and bias. Journal of Experimental Psychology: General, 140(4), 693-706). Indeed, Rugani et al claim, incorrectly, that the L-R bias was first reported by Galton in 1880. There are two errors here: first, Galton was reporting what he called 'visualised numerals', which are typically referred to now as 'number forms' - spontaneous and habitual conscious visual representations - not an inference from a number line task. Second, Galton reported right-to-left, circular, and vertical visualised numerals, and no simple left-to-right examples (Galton, F. (1880). Visualised numerals. Nature, 21, 252-256.). So in fact did Bertillon, J. (1880). De la vision des nombres. La Nature, 378, 196-198, and more recently Seron, X., Pesenti, M., Noël, M.-P., Deloche, G., & Cornet, J.-A. (1992). Images of numbers, or "When 98 is upper left and 6 sky blue". Cognition, 44, 159-196, and Tang, J., Ward, J., & Butterworth, B. (2008). Number forms in the brain. Journal of Cognitive Neuroscience, 20(9), 1547-1556.

We sincerely appreciate the opportunity to discuss numerical spatialization in greater detail. We have clarified that an innate predisposition to spatialize numerosity does not necessarily exclude the influence of environmental stimulation and experience. We have proposed an integrative perspective, incorporating both cultural and innate factors, suggesting that numerical spatialization originates from neural foundations while remaining flexible and modifiable by experience and contextual influences. Please see lines 69–75.

We have incorporated the Reviewer’s suggestions and cited all the recommended papers; please see lines 47–75.

If the authors are committed to chicks' MN Line they should test a series of numbers showing that the bias to the left is greater for 2 and 3 than for 4, etc. 

What does all this mean? I think that the paper should be shorn of its misleading contextualisation, including the term 'Mental Number Line'. The authors also speculate, usefully, on why chicks and other species might have a L-R bias. I don't think the speculations are convincing, but at least if there is an evolutionary basis for the bias, it should at least be discussed.

In the revised version of the manuscript, we have resorted to adopt the Spatial Numerical Association (SNA). We thank the Reviewer for this valuable comment.

We appreciated the Reviewer’s suggestion regarding the evolutionary basis of lateralization and have included considerations of its relevance in chicks and other species; please see lines 143-151 and 381-386.

This paper is very interesting with its focus on why the L-R bias exists, and where and why it does not.

We wish to thank the Reviewer again for his/her work.

Author response:

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

Reviewer 1:

(1) In several instances the paper does not address apparent inconsistencies between the prior literature and the findings. For example, the first main finding is that recalled items have more differentiated lateral temporal cortex representations within lists than not recalled items. This seems to be the opposite of the prediction from temporal context models that are used to motivate the paper-context models would predict that greater contextual similarity within a list should lead to greater memory through enhanced temporal clustering in recall. This is what El-Kalliny et al (2019) found, using a highly similar design (free recall, intracranial recordings from the lateral temporal lobe). The authors never address this contradiction in any depth to reconcile it with the previous literature and with the motivating theoretical model. 

Figure 2 supports the findings from El-Kalliny and colleagues because it shows the relationship of each list item relative to the first item (El-Kalliny et al. 2019). Items encoded adjacent to SP1 show the highest spectral similarity supporting the idea of overlapping context predicted by the Temporal Context Model. However, our figure characterizes how increasing inter-item distance affects spectral similarity. It shows that two items successfully recalled from temporally distant serial positions show reduced spectral similarity. These findings align with the predictions of the temporal context model because two temporally distant items would lack significant contextual overlap and therefore would have more distinct spectral representations.

El-Kalliny and colleagues do use a similar experimental set-up however the authors define drift differently. They identified patients with a tendency to temporally cluster, and observed those patients tend to drift less between temporally clustered items however they do not specify drift relative to a constant serial position as we do in our analysis. They define drift as spectral change between two adjacent items which is a more relative measure between any two items rather than in relation to a fixed point like SP1. Finally, our analysis focuses only on gamma activity while El-Kalliny and colleagues identified drift across a much broader set of frequency bands.

(2) The way that the authors conduct the analysis of medial parietal neural similarity at boundaries leads to results that cannot be conclusively interpreted. The authors report enhanced similarity across lists for the first item in each list, which they interpret as reflecting a qualitatively distinct boundary signal. However, this finding can readily be explained by contextual drift if one assumes that whatever happens at the start of each list is similar or identical across lists (for example, a get ready prompt or reminder of instructions). The authors do not include analyses to rule this out, which undermines one of the main findings. 

Extensions of the temporal context model (Lohnas et al. 2015) predict context at the beginning of a list will be most similar to the end of the prior list. The theory assumes a single-context state, consisting of a recency-weighted average of prior items, that is updated, even across different encoding periods.

However, our results show a boundary item representation is most similar to the prior lists first item rather than the last item. Our results conflict with the extension of TCM because the shared similarity of boundary items suggests the context state for the first item in the list is not a recency-weighted average of the items presented immediately prior. The same boundary sensitive signal is not present in other regions, namely the hippocampus and lateral temporal cortex. Those regions do not show similarity between items at the beginning of each list.  

Our main conclusion from these data was that the medial parietal lobe activity seems to be specifically sensitive to task boundaries, defined by the first event or the get ready prompt, while other regions are not.

(3) Although several previous studies have linked hippocampal fMRI and electrophysiological activity at event boundaries with memory performance, the authors do not find similar relationships between hippocampal activity, event boundaries, and memory There are potential explanations for why this might be the case, including the distinction between item vs. associative memory, which has been a prominent feature of previous work examining this question. However, the authors do not address these potential explanations (or others) to explain their findings' divergence from prior work -this makes it difficult to interpret and to draw conclusions from the data about the hippocampus' mechanistic role in forming event memories.

The following text was added and revised in the discussion to discuss hippocampal activity shown in our results and its lack of sensitivity to boundaries.  

“Spectral activity in the medial parietal lobe aligned closely with boundaries. Drift between item pairs seemed to reset at each boundary, leading to renewed similarity after each boundary. This observation aligns with previous work suggesting boundaries reset temporal context.  In the temporal cortex, our findings extend prior studies which suggest the temporal lobe may play a role in associating adjacently presented items (Yaffe et al. 2014, ElKalliny et al 2019). We found items encoded in distant serial positions, but within the same list, drifted significantly more than items from adjacent serial positions (Figure 2C). Consistent with the predictions of the temporal context model, the reduced similarity between distant items may reflect reduced contextual overlap proportional to the time elapsed between them. However, across task boundaries, our study did not detect a robust change in drift rate in the medial or lateral temporal cortex. This finding contrasts with significant work (Ben-Yakov et al. 2018, Ezzyat et al.  2014; Griffiths et al. 2020) which shows hippocampal sensitivity to event-boundaries. One interpretation would be that boundary representations in the hippocampus are quite sparse and represented by populations of time-sensitive cells whose activity is indexed to task-related boundaries (Umbach et al. 2020). While the sparse representations may not be detectable in gamma activity, perhaps it suggests drift in these regions represents a more abstract set of contextual features accumulated from multiple brain regions.”

(4) There is a similar absence of interpretation with respect to the previous literature for the data showing enhanced boundary-related similarity in the medial parietal cortex. The authors’ interpretation seems to be that they have identified a boundary-specific signal that reflects a large and abrupt change in context, however, another plausible interpretation is that enhanced similarity in the medial parietal cortex is related to a representation of a schema for the task structure that has been acquired across repeated instances. 

We agree our results could suggest the MPL creates a generalized situational model or schematic of the task. Unfortunately, our behavioral task does not allow us to differentiate between these ideas and pure boundary representation. However, given boundaries are a component in defining situational models, we chose to interpret our results conservatively as a form of boundary representation.  

(5) The authors do not directly compare their model to other models that could explain how variability in neural activity predicts memory. One example is the neural fatigue hypothesis, which the authors mention, however there are no analyses or data to suggest that their data is better fit by a boundary/contextual drift mechanism as opposed to neural fatigue. 

The study by Lohnas and colleagues does find higher HFA was greater for recalled items but does not describe a serial position specific trend (Lohnas et al. 2020). For our study, we stringently controlled for recall success in each of our analyses. Our main finding of boundary similarity compares recalled boundary items to recalled items in each of the other serial positions. We also show the similarity of nonrecalled items in all serial positions to demonstrate the lack of boundary representation in first list items, when neural fatigue is presumably least present.

In addition, their study demonstrated neural fatigue in the hippocampus. They did not find evidence of fatigue in the DLPFC, suggesting region-specific mechanisms of neural fatigue. Our results are focused on the medial parietal lobe, and we were not able to find a fatigue model of the region for further comparison. While our results do not rule out the possibility of neural fatigue driving a drifting or boundary signal, we focus on the relevance of the signal to memory performance.

(6) P2. Line 65 cites Polyn et al (2009b) as an example where ‘random’ boundary insertions improve subsequent memory. However, the boundaries in that study always occurred at the same serial position and were therefore completely predictable and not random.

The citation was removed from the corresponding sentence.

(7) P2. Line 74 cites Pu et al. (2022) as an example of medial temporal lobe ‘regional activity’ showing sensitivity to event boundaries; however, this paper reported behavioral and computational modeling results and did not include measurement of neural activity. 

The citation was removed from the corresponding sentence.

(8) P.3 Line 117, Hseih et al (2014) and Hseih and Ranganath (2015) are cited as evidence that ‘spectral’ relatedness decreases as a function of distance, but neither of these studies examined ‘spectral’ activity (fMRI univariate and multivariate). The manuscript would benefit from a careful review and updating of how the prior literature is cited, which will increase the impact of the findings for readers. 

The text has been updated to reflect this distinction by modifying the statement to:  “Previous work consistent with temporal context models suggests neural pattern similarity reduces as a function of distance between related memories.”

(9) Several previous studies have found hippocampal activity at event boundaries correlates with memory performance (Ben-Yakov et al 2011, 2018; Baldassano et al 2017), yet here the authors do not find evidence for hippocampal activity at event boundaries related to memory. Does this difference reflect something important about how the hippocampus vs. medial parietal cortex vs. lateral temporal cortex contribute to memory formation? Currently, there is not much discussion about how to interpret the differences between brain regions. Previous work has suggested that hippocampal pattern similarity at event boundaries specifically supports associative memory across events (Ezzyat & Davachi, 2014; Griffiths & Fuentemilla, 2020; Heusser et al., 2016), which may help explain their findings. In any case the authors could increase the impact of their paper by further situating their findings within the previous literature. 

We would not suggest there is no boundary-related activity in the hippocampus. Similar to an earlier point made by the reviewer, to clarify our interpretation of regional differences, the following text has been added to the discussion.  

“However, across task boundaries, our study did not detect a robust change in drift rate in the medial or lateral temporal cortex. This finding contrasts with significant work (Ezzyat and Davachi, 2014; Griffiths and Fuentemilla, 2020) which shows hippocampal sensitivity to event-boundaries. One interpretation would be that boundary representations in the hippocampus are quite sparse and represented by populations of time-sensitive cells whose activity is indexed to task-related boundaries (Umbach et al 2020). While the sparse representations may not be detectable in gamma activity, perhaps it suggests drift in these regions represents a more abstract set of contextual features accumulated from multiple brain regions (Baldassano et al. 2017). “

(10) The authors mention neural fatigue as an alternative theory to explain the primacy effect (Serruya et al., 2014), however there are no analyses or data to suggest that their data is better fit by a boundary mechanism as opposed to neural fatigue. Previous studies have shown that gamma activity in the hippocampus changes with serial position and with encoding history (Serruya et al 2014; Lohnas et al 2020). Here, the authors could compare the reported pattern similarity results to control analyses that replicate this prior work, which would strengthen their argument that there is unique information at boundaries that is distinct from a neural fatigue signal. 

The serial position effects described by Serruya and colleagues describe decreasing HFA with increasing serial position in the MTL, lateral temporal cortex and prefrontal cortex (Serruya et al. 2014). Despite their findings, we do not observe a strong boundary effect in those regions (see Supp Fig 3 a,b). The lack of boundary effect in regions where HFA is selectively increased for primacy items suggests the global neural fatigue model does not account for our results.

Notably, the authors do not characterize HFA trends in the MPL. Nevertheless, their findings do not rule out the possibility of a boundary effect driving the HFA. We demonstrate boundary-relevant HFA only in the MPL but not in other regions. In addition, we show a correlation between SP1 recalls and boundary representation strength, as well as a conserved similarity of multiple boundary-adjacent items.  

Next, the neural fatigue study by Lohnas and colleagues does find higher HFA was greater for recalled items but does not describe a serial position specific trend (Lohnas et al. 2015). For our study, we stringently controlled for recall success in each of our analyses. Our main finding of boundary similarity compares recalled boundary items to recalled items in each of the other serial positions. We also show the similarity of non-recalled items in all serial positions to demonstrate the lack of boundary representation in the first list items, when neural fatigue is presumably least present.

In addition, their study demonstrated neural fatigue in the hippocampus. They did not find evidence of fatigue in the DLPFC, suggesting region-specific mechanisms of neural fatigue. Our results are focused on the medial parietal lobe, and we were not able to find a fatigue model of the region for further comparison. While our results do not rule out the possibility of neural fatigue driving a drifting or boundary signal, we focus on the relevance of the signal to memory performance.

(11) For the analyses that examine cross-list similarity (e.g. the medial parietal analysis in Figure 3), how did the authors choose the number of lists over which similarity was calculated? Was the selection of this free parameter cross-validated to ensure that it is not overfitting the data? Given that there were 25 lists per session, using the three succeeding lists seems arbitrary. Why not use every list across the whole session? 

Given the volume of data, number of patients, and computational time available at our facility, we extended the analysis as far as we could to characterize the observed trend.

(12) P4. Line 155 says that Figure 3C shows example subject data, but it looks like it is actually Figure 3D. 

The text was updated to reference the correct figure.

(13) The t-tests on P.4 Line 159 have two sets of degrees of freedom but should only have one. 

The t-tests described by Figure 3B represent the mean parameter estimate of the predictor for boundary proximity contrasted by region for all item pairs. The statistical test in this case was an unpaired t-test between parameter estimates for patients with electrodes in each of the regions. The numbers within parentheses represent the sample size, or number of subjects, contributing electrodes to each region.

Reviewer 2:

(1) Because this is not a traditional event boundary study, the data are not ideally positioned to demonstrate boundary specific effects. In a typical study investigating event boundary effects, a series of stimuli are presented and within that series occurs an event boundary – for instance, a change in background color. The power of this design is that all aspects between stimuli are strictly controlled – in particular, the timing – meaning that the only difference between boundary-bridging items is the boundary itself. The current study was not designed in this manner, thus it is not possible to fully control for effects of time or that multiple boundaries occur between study lists (study to distractor, distractor to recall, recall to study). Each list in a free recall study can be considered its own “mini” experiment such that the same mechanisms should theoretically be recruited across any/all lists. There are multiple possible processes engaged at the start of a free recall study list which may not be specific to event boundaries per se. For example, and as cited by the authors, neural fatigue/attentional decline (and concurrent gamma power decline) may account for serial position effects. Thus, SP1 on all lists will be similar by virtue of the fact that attention/gamma decrease across serial position, which may or may not be a boundaryspecific effect. In an extreme example, the analyses currently reported could be performed on an independent dataset with the same design (e.g. 12 word delayed free recall) and such analyses could potentially reveal high similarity between SP1-list1 in the current study and SP1-list1 in the second dataset, effects which could not be specifically attributed to boundaries.

The neural fatigue study by Lohnas and colleagues does find higher HFA was greater for recalled items but does not describe a serial position specific trend (Lohnas et al. 2020). For our study, we stringently controlled for recall success in each of our analyses. Our main finding of boundary similarity compares recalled boundary items to recalled items in each of the other serial positions. We also show the similarity of non-recalled items in all serial positions to demonstrate the lack of boundary representation in the first list items, when neural fatigue is presumably least present.

In addition, their study demonstrated neural fatigue in the hippocampus. They did not find evidence of fatigue in the DLPFC, suggesting region-specific mechanisms of neural fatigue. Our results are focused on the medial parietal lobe, and we were not able to find a fatigue model of the region for further comparison. While our results do not rule out the possibility of neural fatigue driving a drifting or boundary signal, we focus on the relevance of the signal to memory performance.

(2) Comparisons of recalled "pairs" does not account for the lag between those items during study or recall, which based on retrieved context theory and prior findings (e.g. Manning et al., 2011), should modulate similarity between item representations. Although the GLM will capture a linear trend, it will not reveal serial position specific effects. It appears that the betas reported for the SP12 analyses are driven by the fact that similarity with SP12 generally increases across serial position, rather a specific effect of "high similarity to SP12 in adjacent lists" (Page 5, excluding perhaps the comparison with list x+1). It is also unclear how the SP12 similarity analyses support the statement that "end-list items are represented more distinctly, or less similarly, to all succeeding items" (Page 5). It is not clear how the authors account for the fact that the same participants do not contribute equally to all ROIs or if the effects are consistent if only participants who have electrodes in all ROIs are included.

In our study, all pairs are defined by the lag between a reference and target item. The results in Figure 3 show the similarity between each serial position in relation to SP1; Figure 4 shows lag between each serial position relative to SP2 and 3; and Figure 5 shows lag relative to SP12. Each statistical model accounts for the lag by ordering the data by increased inter-item distance. Further, our definition of lag is significantly more rigorous than that used by Manning and colleagues. Our similarity results for Figures 3-5 characterize the change in similarity relative to a constant reference point, such as SP1, rather than a relative reference point, such as +1 lag, which aggregates similarity between pairs such as SP1 to SP2 with SP4 to SP5, which maybe recalled via different memory mechanisms.  

In Figure 5, we agree your characterization that ‘similarity with SP12 generally increases across serial position’ is a more accurate description of the trend. The text has been updated to reflect this by changing the interpretation to “later serial positions in adjacent lists shared a gradually increasing similarity to SP12.”  

Next, we clarify the statement "end-list items are represented more distinctly, or less similarly, to all succeeding items". When recalling SP12, the subsequent items recalled exhibit significantly lower similarity to SP12 (see Figure 5D, pink). Consequently, the spectral representation of successfully recalled end-list items appears more distinct from later items in similar serial positions. This stands in contrast to our observations illustrated in Figures 3 and 4, where successfully recalled start-list items demonstrate greater similarity to later items in similar serial positions.

(3) The authors use the term "perceptual" boundary which is confusing. First, "perceptual boundary" seems to be a specific subset of the broader term "event boundary," and it is unclear why/how the current study is investigating "perceptual" boundaries specifically. Second and relatedly, the current study does not have a sole "perceptual" boundary (as discussed in point 1 above), it is really a combination of perceptual and conceptual since the task is changing (from recalling the words in the previous list to studying the words in the current list OR studying the words in the current list to solving math problems in the current list) in addition to changes in stimulus presentation. 

We agree with the statement that ‘perceptual’ as a modifier to the boundaries described here does not add significant information. Therefore, we have removed all reference to perceptual boundaries.

(4) Although the results show that item-item similarity in the gamma band decreases across serial position, it is unclear how the present findings further describe "how gamma activity facilitates contextual associations" (Page 5). As mentioned in point 1 above, such effects could be driven by attentional declines across serial position -- and a concurrent decline in gamma power -- which may be unrelated to, and actually potentially impair, the formation of contextual associations, given evidence from the literature that increased gamma power facilitates binding processes.

We agree that our study does not elucidate a mechanistic relationship between gamma power and contextual associations. The referenced sentence has been changed to: “how gamma activity is associated with context”.

Please see our response to point 1 above. In addition, studies demonstrating decreasing gamma power with increasing serial position focus primarily on the MTL, lateral temporal cortex and prefrontal cortex (Serruya et al. 2012). Despite their findings, we do not observe a strong boundary effect in those regions (see Supp Fig 3 a,b). The lack of boundary effect in regions where HFA is selectively increased for primacy items suggests the global attentional decline or neural fatigue model does not account for our results.

Notably, HFA trends in the MPL are poorly described. Further, gamma power decline does not rule out the possibility of a boundary effect driving the HFA. We demonstrate boundary-relevant HFA only in the MPL but not in other regions. In addition, we show a correlation between SP1 recalls and boundary representation strength, as well as a conserved similarity of multiple boundary-adjacent items.

(5) Some of the logic and interpretations are inconsistent with the literature. For example, the authors state that "The temporal context model (TCM) suggests that gradual drift in item similarity provides context information to support recovery of individual items" however, this does not seem like an accurate characterization of TCM. According to TCM, context is a recency-weighted average of previous experience. Context "drifts" insofar as information is added to/removed from context. Context drift thus influences item similarity -- it is not that item similarity itself drifts, but that any change in item-item similarity is due to context drift. 

The current findings do not appear at odds with the conceptualization of drift and context in current version of the context maintenance and retrieval model. Furthermore, the context representation is posited to include information beyond basic item representations. Two items, regardless of their temporal distance, can be associated with similar contexts if related information is included in both context representations, as predicted and shown for multiple forms of relatedness including semantic relatedness (Manning & Kahana, 2012) and task relatedness (Polyn et al., 2012).

We revised the sentence and encompassing paragraph to describe the temporal context model more accurately and emphasize how our findings align with the stated version of CMR. The revised text is below:  

“Next, we asked how gamma spectral activity reflects contextual association between items. In the medial parietal lobe, we observed recurring similarity between items distant in time but adjacent to boundaries. This pattern suggests spectral activity may carry information about an item's relationship to a boundary. These observations align with the Context Maintenance and Retrieval model which extends the predictions of TCM to encompass broader relationships among items. Our results demonstrate boundaries as an important aspect of context and specify the spectral and regional properties of these boundary-related contextual features.”

(6) Lohnas et al. (2020) Neural fatigue influences memory encoding in the human hippocampus, Neuropsychologia, should be cited when discussing neural fatigue

Thank you for your suggestion. The citation has been added to the text.

(7) A within-list, not an across list, similarity analysis should be used to test the interpretation that end-of-list items are more distinct than other list items.

We believe this recommendation refers to the following line in our text: “These findings suggest end-list items are represented more distinctly, or less similarly, to all succeeding items.” Our statement compares list x, SP12 to all succeeding items (in list x+1, x+2, etc.). Therefore, this statement refers to items in the next lists which is why we performed an across list analysis rather than within-list one.

(8) It is unclear why it is necessary to use PCA to estimate similarity between items.

PCA was used to reduce the dimensionality of the time-frequency matrix for the gamma band. This technique allowed us to compare predominant trends in gamma between items. In addition, we added a figure showing 3 example subjects in Figure 3 – supplementary figure 2D to show unique time-frequency components contribute to signal reconstructed from the PCs for each subject. Therefore, the boundary representation may be represented differently for each patient.

(9) Lags are listed as -4, 4 (Page 8), however with a list length of 12, possible lags should be 11, 11.

The listed parenthetical statement ‘(-4 to 4)’ referred to Figure 1 where Lag CRP is shown for transitions from -4 to 4. However, we did calculate lag CRP for all possible transitions. Therefore, the referenced phrase was changed to: “Lagged CRP was calculated for all possible transitions (-11 to 11).”

(10) Hsieh et al. 2014 and Hsieh & Ranganath (2015) are fMRI studies and as such, do not support the statement "Previous work consistent with temporal context models suggests spectral relatedness reduces as a function of distance between words" (Page 3). 

The statement has been revised to: “Previous work consistent with temporal context models suggests neural pattern similarity reduces as a function of distance between related memories.”

(11) Although statistically one can measure "How item-item similarity is affected by recollection" (Page 3), this is logically backwards, given that similarity during study necessarily precedes performance during free recall. Additionally, it is erroneous to assume that recalled words are "recollected" without additional measurements (e.g. Mickes et al. (2013) Rethinking familiarity: Remember/Know judgments in free recall, JML).

The statement was changed to “item-item similarity is affected based on successful recall” given recollection cannot be determined in our paradigm.

Reviewer 3:

(1) My primary confusion in the current version of this paper is that the analyses don't seem to directly compare the two proposed models illustrated in Fig 1B, i.e. the temporal context model (with smooth drifts between items, including across lists) versus the boundary model (with similarities across all lists for items near boundaries). After examining smooth drift in the within-list analysis (Fig 2), the across-list analyses (Figs 3-5) use a model with two predictors (boundary proximity and list distance), neither of which is a smoothlydrifting context. Therefore there does not appear to be a quantitative analysis supporting the conclusion that in lateral temporal cortex "drift exhibits a relationship with elapsed time regardless of the presences of intervening boundaries" (lines 272-3).

We could not use a smoothly drifting regressor due to its collinearity with any model of boundary similarity. Therefore, we chose our two regressors: boundary proximity, which models intra-list changes in similarity and list distance, which models a stepwise decrease in similarity from adjacent lists.

However, we agree with the comment that the presented data does not directly support the lateral temporal cortex drifts independent of intervening boundaries. Therefore, we amended the statement to: “We found successfully recalled items encoded in distant serial positions drifted significantly more than items from adjacent serial positions (Figure 2C)”. Consistent with the predictions of the temporal context model, the reduced similarity between distant items may reflect reduced contextual overlap proportional time elapsed between them.”

(2) The feature representation used for the neural response to each item is a gamma power time-frequency matrix. This makes it unclear what characteristics of the neural response are driving the observed similarity effects. It appears that a simple overall scaling of the response after boundaries (stronger responses to initial items during the beginning portion of the 1.6s time window) would lead to the increased cosine similarity between initial items, but wouldn't necessarily reflect meaningful differences in the neural representation or context of these items.

Our study aims to draw the connection between the neural response after boundaries with neural representation and context of these items. Prior studies (Manning et al. 2011, El Kalliny et al. 2017) have interpreted similarity in neural spectra as a memory relevant phenomenon. We use very similar methods to perform our analysis.  

In addition, we compare the fit of our boundary similarity model to behavioral performance to show increased boundary representation correlates with improved boundary item recall.

While our study does not specify which time-frequency components underly the increased similarity, we do limit our analysis to the gamma band. Traditional analyses include log-scaled, broadband time-frequency data (eg. 3-100hz) from which we specify the relevance of a much narrower spectral band.  

Finally, we tried to study which time–frequency components contributed to the increased similarity, but it varied greatly between patients (see Figure 3 – supplementary figure 2D). Hence, we opted to use principal component analyses to compare the features showing the most variation for each given participant. This added analytical step allows us to detect boundary effects across patients despite individual variability in boundary representation.

(3) The specific form of the boundary proximity models is not well justified. For initial items, a model of e^(1-d) is used (with d being serial position), but it is not stated how the falloff scale of this model was selected (as opposed to e.g. e^((1-d)/2)). For final items, a different model of d/#items is used, which seems to have a somewhat different interpretation (about drift between boundaries, rather than an effect specific to items near a final boundary). The schematic in Fig 1B appears to show a hypothesis which is not tested, with symmetric effects at initial and final boundaries.

The boundary proximity models were chosen empirically. Our model was intended to quantify a decreasing relationship across many patients. We acknowledge the constants and variables may not definitively describe underlying neural processes.  

For start- and end-list boundaries, we used different models because primacy and recency effects are unique phenomena. Primacy memory is classically thought to arise from rehearsal during the encoding time (Polyn et al. 2009, Lohnas et al. 2015). Alternatively, recency memory is thought to arise from strong contextual cues of recency items during recall due to their temporal proximity. Therefore, we have a limited basis on which to assume their spectral representation in relation to task boundaries would be symmetric.

(4) The main text description of Fig 2 only describes drift effects in lateral temporal cortex, but Fig 2 - supplement 1 shows that there is also drift and a significant subsequent memory effect in the other two ROIs as well. There is not a significant memory x drift slope interaction in these regions; are the authors arguing that the lack of this interaction (different drift rates for remembered versus forgotten items) is critical for interpreting the roles of lateral temporal cortex versus medial parietal and hippocampal regions?

Yes. Fig 2- Supplement 1 shows that drift occurs in both the HC and MPL. However, the interaction term is not significant, which suggests that the rate of drift between recalled and non-recalled items is not significantly different.  

In contrast, Fig 2C shows that recalled pairs drift at a higher rate than non-recalled pairs. For the LTC, the interaction term is negative in magnitude and statistically significant. This suggests successfully encoded item pairs encoded far apart share more distinct spectral representations, specifically in the LTC. These findings lead to our interpretation in the discussion that “elevated drift rate might allow the representations of recalled items to remain distinct but ordered in memory.”

(5) The parameter fits for the "list distance" regressor are not shown or analyzed, though they do appear to be important for the observed similarity structure (e.g. Fig 3E). I would interpret this regressor as also being "boundary-related" in the sense that it assumes discrete changes in similarity at boundaries.

Parameter fits for the ‘list distance’ regressor are now shown in the supplementary portion of Figures 3 and Figure 5. The difference between regions is non-significant.

(6) To make strong claims about temporal context versus boundary models as implied by Fig 1B, these two regressors should be fit within the same model to explain across-list similarity. The temporal context model could be based on the number of intervening items (as in Fig 1B) or actual time elapsed between items. The relationship between the smoothly drifting temporal context model and the discretely-jumping list distance models should also be clarified.

We could not use a smoothly drifting regressor due to its collinearity with any model of boundary similarity. A model which included a ‘temporal context regressor’ would not be able to account for the presence of a boundary effect and would not allow us to demonstrate a boundary representation in the presence of drift. Therefore, we chose our two regressors: boundary proximity, which models intra-list changes in similarity and list distance, which models a stepwise decrease in similarity from adjacent lists. These regressors allow the model to differentiate between intra-list changes (the boundary regressor) verses inter-list changes (the list distance regressor).  

(7) The features of the time-frequency matrix that are driving similarity between events could be visualized to provide a better understanding of the boundary-related signals. The analysis could also be re-run with reduced versions of the feature space in order to determine the critical components of this signal; for example, responses could be averaged across time to examine only differences across frequencies, or across frequencies to examine purely temporal changes across the 1.6 second window.

Figure 3 – supplementary figure 2 A-C has been added to show varying the number of principal components (PCs) does not change the trend of boundary sensitivity in the MPL. In addition, we included 3 example subjects in Figure 3 – supplementary figure 2D to show unique time-frequency components contribute to signal reconstructed from the PCs for each subject. Therefore, the boundary representation may be represented differently for each patient.

(8) If the authors are considering a space of multiple models as "boundary proximity models" (e.g. linear models and exponential models with different scale factors), this should be part of the model-fitting process rather than a single model being selected posthoc.

We agree with the reviewer’s suggestion that the most ideal way to fit a model to the trend would be using a model-fitting process. However, due to a limitation on the amount of computational resources available, we were not able to perform it given the size of our dataset.

(9) The interpretation of region differences in the results in Fig 2 and Fig 2 - supplement 1 should be clarified. 

In discussion, we have added the following text to clarify our interpretation of the regional differences shown in the mentioned figures.  

“However, across task boundaries, our study did not detect a robust change in drift rate in the medial or lateral temporal cortex. This finding contrasts with significant work (Ezzyat and Davachi, 2014; Griffiths and Fuentemilla, 2020) which shows hippocampal sensitivity to event-boundaries. One interpretation would be that boundary representations in the hippocampus are quite sparse and represented by populations of time-sensitive cells whose activity is indexed to task-related boundaries (Umbach et al 2018). While the sparse representations may not be detectable in gamma activity, perhaps it suggests drift in these regions represents a more abstract set of contextual features accumulated from multiple brain regions (Baldassano et al. 2017). “

(10) Whether there are significant fits for the list distance regressor, and whether these fits vary across regions, could be stated. The list distance regressor could also be directly compared (in the same model) to a temporal-context regressor, which predicts graded changes in similarity between items rather than the discrete changes between lists.

We have added parameter fits for the ‘list distance’ regressor in the supplementary portion of Figures 3 and Figure 5. The difference between regions is non-significant. Therefore, our results show very similar stepwise decrease in similarity across lists between regions (list distance regressor; Figure 3 —supplementary figure 1B).

We could not compare these parameters to a separate model which includes a smoothly drifting ‘temporal-context’ regressor due to the regressors collinearity with any representation of boundary. See our response to Reviewer 3 –comment 6.  

(11) The authors should clarify their interpretation of the results, and whether they are proposing a tweak to the temporal context model or a substantially different organizational system. 

In the disucssion we include the following statements to clarify what we suggest regarding the temporal context model.  

“Our findings suggest a broader scope of contextual association than just prior items, where temporal proximity as well as task structure in the form of boundaries, play intertwined roles in contextual construction. Our data therefore have implications for updated iterations of the temporal context model incorporating (perhaps) specific terms for boundary information. This may in turn provide a more systematic prediction of primacy effects in behavioral data.”  

(12) Minor typos and corrections: 

52: using -> use 

108: patients -> patients'  156: list -> lists 

The list distance plot is described as "pink" in Fig 3 and Fig 5 - supplement 1, but appears gray in the figures.

Each of these corrections has been corrected in the text.

Author response:

The following is the authors’ response to the original reviews

We thank the reviewers for their very constructive and helpful comments on the previous version of this manuscript. They have focused on some important issues and have raised many valuable questions that we expect to answer as research begins on these markings. As has been often the case with preprints, a number of experts beyond the four reviewers and editor have provided comments, questions, and suggestions, and we have taken these on board in our revision of the manuscript. In particular, Martinón-Torres et al. (2024) focused several comments upon this manuscript and raise some points that were not considered by the reviewers, and so we discuss those points here in addition to the reviewer comments.

Some of us have been engaged in other aspects of the possible cultural activities of Homo naledi. After the discovery of these markings we considered it indefensible to publish further research on the activity of H. naledi within this part of the cave system without making readers aware that the H. naledi skeletal remains occur in a spatial context near markings on cave walls. Of course, the presence of markings leaves many questions open. A spatial context does not answer all questions about the temporal context. The situation of the Dinaledi Subsystem does entail some constraints that would not apply to markings within a more open cave or rock wall, and we discuss those in the text.

We find ourselves in agreement with most of the reviewers on many points. As reflected by several of the reviewers, and most pointedly in the remarks by reviewer 1, the purpose of this preprint is a preliminary report on the observation of the markings in a very distinctive location. This initial report is an essential step to enable further research to move forward. That research requires careful planning due to the difficulty of working within the Dinaledi Subsystem where the markings are located. This pattern of initial publication followed by more detailed study is common with observations of rock art and other markings identified in South Africa and elsewhere. We appreciate that the reviewers have understood the role of this initial study in that process of research.

Because of this, the revised manuscript represents relatively minimal changes, and all those at the advice of reviewers. Many thanks to all the reviewers for noting various typographic errors, missed references and other issues that we have done our best to fix in the revised manuscript.

Expertise of authors. Reviewer 4 mentions that the expertise of the authors does not include previous publication history on the identification of rock art, and other reviewers briefly comment that experts in this area would enhance the description. AF does have several publications on ancient engravings and other markings; LRB has geological training and field experience with rock art. Notwithstanding this, we do take on board the advice to include a wider array of subject experts in this research, and this is already underway.

Image enhancement. We appreciate the suggestions of some reviewers for possible strategies to use software filters to bring out details that may not be obvious even with our cross-polarization lighting and filtering. These are great ideas to try. In this manuscript we thought that going very far into software editing or image enhancement might be perceived by some readers as excessive manipulation, particularly in an age of AI. In future work we will experiment with the suggested approaches. 

Natural weathering. In the process of review and commentary by experts and the public there has been broad acceptance that many of the markings illustrated in this paper are artificial and not a product of natural weathering of the dolomite rock. We deeply appreciate this. At the same time, we accept the comments from reviewers that some markings may be difficult to differentiate from natural weathering, and that some natural features that were elaborated or altered may be among the markings we recognize. On pages 3 and 4 we present a description of the process of natural subaerial weathering of dolomite, which we have rooted in several references as well as our own observations of the natural weathering visible on dolomite cave walls in the Rising Star cave system. This includes other cave walls within the Dinaledi Subsystem. We discuss the “elephant skin” patterning of natural dolomite surface weathering, how that patterning emerges, and how that differs from the markings that are the subject of this manuscript.

Animal claw marks. Martinón-Torres et al. 2024 accept that some of the markings illustrated on Panel A are artificial, but they offer the hypothesis that some of those markings may be consistent with claw marks from carnivores or other mammals. They provide a photo of claw marks within a limestone cave in Europe to illustrate this point. On pages 5 and 6 of the revised manuscript we discuss the hypothesis of claw marks. We discuss the presence of animals in southern Africa that may dig in caves or mark surfaces. However the key aspect of the Malmani dolomite caves is that the hardness of dolomitic limestone rock is much greater than many of the limestone caves in other regions such as Europe and Australia, where claw marks have been noted in rock walls. As we discuss, we have not been able to find evidence of claw marks within the dolomite host bedrock of caves in this region, although carnivores, porcupines, and other animals dig into the soft sediments within and around caves. The form of the markings themselves also counter-indicates the hypothesis that they are claw marks. 

Recent manufacture. One comment that occurs within the reviews and from other readers of the preprint is that recent human visitors to the cave, either in historic or recent prehistoric times, may have made these marks. We discuss this hypothesis on page 6 of the revised manuscript. The simple answer is that no evidence suggests that any human groups were in the Dinaledi Subsystem between the presence of H. naledi and the entry of explorers within the last 25 years. The list of all explorers and scientific visitors to have entered this portion of the cave system is presented in a table. We can attest that these people did not make the marks. More generally, such marks have not been known to be made by cavers in other contexts within southern Africa.

Panels B and C. We have limited the text related to these areas, other than indicating that we have observed them. The analysis of these areas and quantification of artificial lines does not match what we have done for the Panel A area and we leave these for future work. 

Presence of modern humans. We have observed no evidence of modern humans or other hominin populations within the Dinaledi Subsystem, other than H. naledi. Several reviewers raise the question of whether the absence of evidence is evidence of absence of modern humans in this area. This is connected by two of the reviewers to the observation that the investigation of other caves in recent years has shown that markings or paintings were sometimes made by different groups over tens of thousands of years, in some cases including both Neanderthals and modern humans. We have decided it is best for us not to attempt to prove a negative. It is simple enough to say that there is no evidence for modern humans in this area, while there is abundant evidence of H. naledi there.

Association with H. naledi. Reviewer 2 made an incisive point that the previous version contained some text that appeared contradictory: on the one hand we argued that modern humans were not present in the subsystem due to the absence of evidence of them, yet we accepted that H. naledi may have been present for a longer time than currently established by geochronological methods.

We appreciate this comment because it helped us to think through the way to describe the context and spatial association of these markings and the skeletal remains, and how it may relate to their timeline. Other reviewers also raised similar questions, whether the context by itself demonstrates an association with H. naledi. We have revised the text, in particular on pages 5 and 7, to simply state that we accept as the most parsimonious alternative at present the hypothesis that the engravings were made by H. naledi, which is the only hominin known to be present in this space.

Age of H. naledi in the system. At one place in the previous manuscript we indicated that we cannot establish that H. naledi was only active in the cave system within the constraints of the maximum and minimum ages for the Dinaledi Subsystem skeletal remains (viz., 335 ka – 241 ka), because some localities with skeletal material are undated. We have adjusted this paragraph on page 7 to be clear that we are discussing this only to acknowledge uncertainty about the full range of H. naledi use of the cave system.

Geochronological methods. Several reviewers discuss the issue of geochronology as applied to these markings. This is an area of future investigation for us after the publication of this initial report. As some reviewers note, the prospects for successful placement of these engraved features and other markings with geochronological methods depends on factors that we cannot predict without very high-resolution investigation of the surfaces. We have included greater discussion of the challenges of geochronological placement of engravings on page 6, including more references to previous work on this topic. We also briefly note the ethical problems that may arise as we go further with potentially  invasive, destructive or contact studies of these engravings, which must be carefully considered by not just us, but the entire academy.

Title. Some reviewers suggested that the title should be rephrased because this paper does not use chronological methods to derive date constraints for the markings. We have rephrased the title to reflect less certainty while hopefully retaining the clear hypothesis discussed in the paper.

Author response:

The following is the authors’ response to the previous reviews

Reviewer #1 (Public review):

Summary:

The present study aims to associate reproduction with age-related disease as support of the antagonistic pleiotropy hypothesis of ageing predominantly using Mendelian Randomization. The authors found evidence that early-life reproductive success is associated with advanced ageing.

Strengths:

Large sample size. Many analyses.

Weaknesses:

Still a number of doubts with regard to some of the results and their interpretation.

Reviewer #1 (Recommendations for the authors):

Thank you for the opportunity to review a revised version.

I still have serious doubts with regard to a number of datasets presented. For example, the results on essential hypertension and cervical cancer show very small effect sizes, but according to the authors still reach the level of statistical significance. This is unlikely to be accurate. For MR analyses, this is nearly impossible. The analyses of these data and the statistical analysis need to be checked for errors and repeated. While BOLT-LLM might not be relevant here, there might be other things happening here. The authors should therefore always interpret the results also with regard to the observed effect sizes instead of only looking at the p-values (0.999 means that there is a 0.1% lower risk).

Thank you for your suggestions. We have updated the results for essential hypertension, GAD, and cervical cancer in results, figures, and supplemental tables (lines 65-89, Figure 1, Tables S3-S4).

Reviewer #2 (Public review):

Summary:

The authors present an interesting paper where they test the antagonistic pleiotropy theory. Based on this theory they hypothesize that genetic variants associated with later onset of age at menarche and age at first birth may have a positive effect on a multitude of health outcomes later in life, such as epigenetic aging and prevalence of chronic diseases. Using a mendelian randomization and colocalization approach, the authors show that SNPs associated with later age at menarche are associated with delayed aging measurements, such as slower epigenetic aging and reduced facial aging and a lower risk of chronic diseases, such as type 2 diabetes and hypertension. Moreover, they identify 128 fertility-related SNPs that associate with age-related outcomes and they identified BMI as a mediating factor for disease risk, discussing this finding in the context of evolutionary theory.

Strengths:

The major strength of this manuscript is that it addresses the antagonistic pleiotropy theory in aging. Aging theories are not frequently empirically tested although this is highly necessary. The work is therefore relevant for the aging field as well as beyond this field, as the antagonistic pleiotropy theory addresses the link between fitness (early life health and reproduction) and aging.

The authors addressed the remarks on the previous version very well. Addressing the two points below would further increase the quality of the manuscript.

(1) In the previous version the authors mentioned that their results are also consistent with the disposable soma theory: "These results are also consistent with the disposable soma theory that suggests aging as an outcome tradeoff between an organism's investment in reproduction and somatic maintenance and repair."

Although the antagonistic pleiotropy and disposable soma theories describe different mechanisms, both provide frameworks for understanding how genes linked to fertility influence health. The antagonistic pleiotropy theory posits that genes enhancing fertility early in life may have detrimental effects later. In contrast, the disposable soma theory suggests that energy allocation involves a trade-off, where investment in fertility comes at the expense of somatic maintenance, potentially leading to poorer health in later life.

To strengthen the manuscript, a discussion section should be added to clarify the overlap and distinctions between these two evolutionary theories and suggest directions for future research in disentangling their specific mechanisms.

Thank you for your suggestions to clarify the overlap and distinctions between the antagonistic pleiotropy and disposable soma theories. While our primary focus is on the antagonistic pleiotropy framework, we acknowledge that the disposable soma theory also provides a relevant perspective on the trade-offs between reproduction and somatic maintenance.

To address this, we have expanded the discussion section to highlight how both theories contribute to our understanding of the relationship between fertility-related traits and aging-related health outcomes. We also suggested potential future research directions, such as integrating genetic data with biomarkers of somatic to further explore the mechanisms underlying these trade-offs (lines 213-223).

(2) In response to the question why the authors did not include age at menopause in addition to the already included age at first child and age at menarche the following explanation was provided: "Our manuscript focuses on the antagonistic pleiotropy theory, which posits that inherent trade-off in natural selection, where genes beneficial for early survival and reproduction (like menarche and childbirth) may have costly consequences later. So, we only included age at menarche and age at first childbirth as exposures in our research."

It remains, however, unclear why genes beneficial for early survival and reproduction would be reflected only in age at menarche and age at first childbirth, but not in age at menopause. While age at menarche marks the onset of fertility, age at menopause signifies its end. Since evolutionary selection acts directly until reproduction is no longer possible (though indirect evolutionary pressures persist beyond this point), the inclusion of additional fertility-related measures could have strengthened the analysis. A more detailed justification for focusing exclusively on age at menarche and first childbirth would enhance the clarity and rigor of the manuscript.

Thank you for your question regarding the age at menopause in our analysis. Our decision was based on the theoretical framework of antagonistic pleiotropy, which emphasizes early-life reproductive advantages that may have trade-offs later in life. Age at menarche and age at first childbirth are direct markers of early reproductive investment, which align closely with this framework.

While age at menopause marks the cessation of reproductive capability, its evolutionary role is distinct. The selective pressures acting on menopause are complex and may involve post-reproductive contributions rather than direct reproductive fitness benefits. Moreover, the genetic architecture of menopause may be influenced by different biological pathways compared to early reproductive traits.

Nonetheless, we acknowledge that including age at menopause could provide additional insights into reproductive aging. Several papers1,2 were already published regarding age at menopause and age-related outcomes, including diabetes, AD, osteoporosis, cancers, and cardiovascular diseases.

Reviewing Editor (Recommendations for the authors):

Above/below you will find the remaining comments from the reviewers. One of the main issues remaining is that some of the data seems to be incorrectly analysed and some of the findings may not be correct. To clarify this a lot more, I asked the reviewer for some details and received the following:

- In Figure 1B one of their main outcomes is "age of menopause", but they report the data as an odds ratio. This is not correct and should be fixed (it seems the authors can run the right analysis, but just reported it with the wrong heading in the figure). This likely also applies to the outcome "facial aging". Also the heading in Figure 1A should be Beta instead of OR.

We have updated the figures to ensure that the beta values of continuous outcomes and odds ratio values of categorical outcomes are presented in Figure 1.

- With essential hypertension, GAD and cervical cancer, the estimates are so small that they need to re-review their results. The current MR analysis is not sufficiently powered to have such small confidence intervals. Essential hypertension was based on data from UK biobank, although I was also unable to find what program was used to generate the GWAS results, I have strong thoughts this was also BOLT-LLM. Same for cervical cancer. Both datasets used familial-related samples, so they are very likely derived with BOLT-LLM.

I hope this will help to solve this issue.

Based on published paper, gastrointestinal or abdominal disease (GAD) (GWAS ID: ebi-a-GCST90038597) is after BOLT-LLM. Based on MRC IEU UK Biobank GWAS pipeline, version 1 and 2, essential hypertension (GWAS ID: ukb-b-12493) and cervical cancer (GWAS ID: ukb-b-8777) are after BOLT-LLM. We have updated the MR analysis results and figures (lines 65-89, Figure 1, Tables S3-S4) as well as the following IPA analysis (lines 106-162 and 255-280, Figures 2-3).

(1) Magnus, M. C., Borges, M. C., Fraser, A. & Lawlor, D. A. Identifying potential causal effects of age at menopause: a Mendelian randomization phenome-wide association study. Eur J Epidemiol 37, 971-982 (2022). https://doi.org:10.1007/s10654-022-00903-3

(2) Zhang, X., Huangfu, Z. & Wang, S. Review of mendelian randomization studies on age at natural menopause. Front Endocrinol (Lausanne) 14, 1234324 (2023). https://doi.org:10.3389/fendo.2023.1234324

Author response:

The following is the authors’ response to the original reviews

Public Reviews: 

Reviewer #1 (Public Review): 

This manuscript presents insights into biased signaling in GPCRs, namely cannabinoid receptors. Biased signaling is of broad interest in general, and cannabinoid signaling is particularly relevant for understanding the impact of new drugs that target this receptor. Mechanistic insight from work like this could enable new approaches to mitigate the public health impact of new psychoactive drugs. Towards that end, this manuscript seeks to understand how new psychoactive substances (NPS, e.g. MDMB-FUBINACA) elicit more signaling through βarrestin than classical cannabinoids (e.g. HU-210). The authors use an interesting combination of simulations and machine learning. 

We thank the reviewer for the comments. We have provided point by point response to the reviewer’s comment below and incorporated the suggestions in our revised manuscript. Modified parts of manuscripts are highlighted in yellow.   

Comments:

(1) The caption for Figure 3 doesn't explain the color scheme, so it's not obvious what the start and end states of the ligand are. 

We thank the reviewer to point this out. We have added the color scheme in the figure caption. 

(2) For the metadynamics simulations were multiple Gaussian heights/widths tried to see what, if any, impact that has on the unbinding pathway? That would be useful to help ensure all the relevant pathways were explored.  

We thank the reviewer for the suggestion. We agree with the reviewer that gaussian height/width may impact unbinding pathway. However, we like to point out that we used a well-tempered version of the metadynamics. In well-tempered metadynamics, the effective gaussian height decreases as bias deposition progresses. Therefore, we believe that the gaussian height/width should have minimal impact on the unbinding pathway. To address the reviewer's suggestion, we conducted additional well-tempered metadynamics simulations varying key parameters such as bias height, bias factor, and the deposition rate, all of which can influence the sampling space. Parameter values for bias height, bias factor and deposition rate that we originally used in the paper are 0.4 kcal/mol, 15 and 1/5 ps^-1, respectively. We explored different values for these parameters and projected the sampled space on top of previously sampled region (Figure S4). We observed that new simulations sample similar unbinding pathway in the extracellular direction and discover similar space in the binding pocket as well. 

Results and Discussion (Page 10)

“We also performed unbinding simulations using well-tempered metadynamics parameters (bias height, bias deposition rate and bias factor) to confirm the existence of alternative pathways (Figure S4). However, the simulations show that ligands follow the similar pathway for all

metadynamics runs.”

(3) It would be nice to acknowledge previous applications of metadynamics+MSMs and (separately) TRAM, such as the Simulation of spontaneous G protein activation... (Sun et al. eLife 2018) and Estimation of binding rates and affinities... (Ge and Voelz JCP 2022). 

We appreciate the reviewer's feedback. We have incorporated additional citations of studies demonstrating the use of TRAM as an estimator for both kinetics and thermodynamics (e.g. Ligand binding: Ge, Y. and Voelz, V.A., JCP, 2022[1]; Peptide-protein binding kinetics: Paul, F. et al., Nat. Commun., 2017[2], Ge, Y. et al., JCIM, 2021[3]). Additionally, we have included references to studies where biased simulations were initially used to explore the conformational space, and the results were then employed to seed unbiased simulations for building a Markov state model. (Metadynamics: Sun, X. et al., elife, 2018[4]; Umbrella Sampling: Abella, J. R. et al., PNAS, 2020[5]; Replica Exchange: Paul, F. et al., Nat. Commun., 2017[2]).

(4) What is KL divergence analysis between macrostates? I know KL divergence compares probability distributions, but it is not clear what distributions are being compared. 

We apologize for this confusion. The KL divergence analysis was performed on the probability distributions of the inverse distances between residue pairs from any two macrostates. Each macrostate was represented by 1000 frames that were selected proportional to the TRAM stationary density. All possible pair-wise inverse distances were calculated per frame for the purpose of these calculations. Although KL divergence is inherently asymmetric, we symmetrized the measurement by calculating the average. Per-residue K-L divergence, which is shown in the main figures as color and thickness gradient, was calculated by taking the sum of all pairs corresponding to the residue. We have included a detailed discussion of K-L divergence in Methods section.  We have also modified the result section to add a brief discussion of K-L divergence methodology.

Results and Discussion (Page 15)

“We further performed Kullback-Leibler divergence (K-L divergence) analysis between inverse distance of residue pairs of two macrostates to highlight the protein region that undergoes high conformational change with ligand movement.”

Methods (Page 33)

“Kullback–Leibler divergence (K-L divergence) analysis was performed to show the structural differences in protein conformations in different macrostates[4,114] . In this study, this technique was used to calculate the difference in the pairwise inverse distance distributions between macrostates. Each macrostate was represented by 1000 frames that were selected proportional to their TRAM weighted probabilities. Although K-L divergence is an asymmetric measurement, for this study, we used a symmetric version of the K-L divergence by taking the average between two macrostates. Per residue contribution of K-L divergence was calculated by taking the sum of all the pairwise distances corresponding to that residue. This analysis was performed by inhouse Python code.”  

(5) I suggest being more careful with the language of universality. It can be "supported" but "showing" or "proving" its universal would require looking at all possible chemicals in the class. 

We thank the reviewer for the suggestion. In response, we have revised the manuscript to ensure that the language reflects that our findings are based on observations from a limited set of ligands, namely one NPS and one classical cannabinoid. We have replaced references to ligand groups (such as NPS or classical cannabinoid) with the specific ligand names (such as MDMB-FUBINACA or HU-210) to avoid claims of universality and prevent any potential confusion.

Results and Discussion (Page 19)

“In this work, we trained the network with the NPS (MDMB-FUBINACA), and classical cannabinoid (HU-210) bound unbiased trajectories (Method Section). Here, we compared the allosteric interaction weights between the binding pocket and the NPxxY motif which involves in triad interaction formation. Results show that each binding pocket residue in MDMBFUBINACA bound ensemble shows higher allosteric weights with the NPxxY motif, indicating larger dynamic interactions between the NPxxY motif and binding pocket residues(Figure S9).  The probability of triad formation was estimated to observe the effect of the difference in allosteric control. TRAM weighted probability calculation showed that MDMB-FUBINACA bound CB1 has the higher probability of triad formation (Figure 8A). Comparison of the pairwise interaction of the triad residues shows that interaction between Y397^7.53-T210^3.46 is relatively more stable in case of MDMB-FUBINACA bound CB1, while other two inter- actions have similar behavior for both systems (Figures S10A, S10B, and S10C). Therefore, higher interaction between Y397^7.53 and T210^3.46 in MDMB-FUBINACA bound receptor causes the triad interaction to be more probable. 

Furthermore, we also compared TM6 movement for both ligand bound ensemble which is another activation metric involved in both G-protein and β-arrestin binding. Comparison of TM6 distance from the DRY motif of TM3 shows similar distribution for HU-210 and MDMBFUBINACA (Figure 8B). These observations support that NPS binding causes higher β-arrestin signaling by allosterically controlling triad interaction formation.” 

Reviewer #2 (Public Review): 

Summary: 

The investigation provides computational as well as biochemical insights into the (un)binding mechanisms of a pair of psychoactive substances into cannabinoid receptors. A combination of molecular dynamics simulation and a set of state-of-the art statistical post-processing techniques were employed to exploit GPCR-ligand dynamics. 

Strengths: 

The strength of the manuscript lies in the usage and comparison of TRAM as well as Markov state modelling (MSM) for investigating ligand binding kinetics and thermodynamics. Usually, MSMs have been more commonly used for this purpose. But as the authors have pointed out, implicit in the usage of MSMs lies the assumption of detailed balance, which would not hold true for many cases especially those with skewed binding affinities. In this regard, the author's usage of TRAM which harnesses both biased and unbiased simulations for extracting the same, provides a more appropriate way out. 

Weaknesses: 

(1) While the authors have used TRAM (by citing MSM to be inadequate in these cases), the thermodynamic comparisons of both techniques provide similar values. In this case, one would wonder what advantage TRAM would hold in this particular case. 

We thank the reviewer for the comment. While we agree that the thermodynamic comparisons between MSM and TRAM provide similar values in this instance, we would like to emphasize the underlying reasoning behind our choice of TRAM.

MSM can struggle to accurately estimate thermodynamic and kinetic properties in cases where local state reversibility (detailed balance) is not easily achieved with unbiased sampling. This is especially relevant in ligand unbinding processes, which often involve overcoming high free energy barriers. TRAM, by incorporating biased simulation data (such as umbrella sampling) in addition to unbiased data, can better achieve local reversibility and provide more robust estimates when unbiased sampling is insufficient.

The similarity in thermodynamic estimates between MSM and TRAM in our study can be attributed to the relatively long unbiased sampling period (> 100 µs) employed. With sufficient sampling, MSM can approach detailed balance, leading to results comparable to those from TRAM. However, as we demonstrated in our manuscript (Figure 4D), when the amount of unbiased sampling is reduced, the uncertainties in both the thermodynamics and kinetics estimates increase significantly for MSM compared to TRAM. Thus, while MSM and TRAM perform similarly under the conditions of extensive sampling, TRAM's advantage lies in its robustness when unbiased sampling is limited or difficult to achieve. 

(2) The initiation of unbiased simulations from previously run biased metadynamics simulations would almost surely introduce hysteresis in the analysis. The authors need to address these issues. 

We thank the reviewer for the comment. We acknowledge that biased simulations could potentially introduce hysteresis or result in the identification of unphysical pathways. However, we believe this issue is mitigated using well-tempered metadynamics, which gradually deposit a decaying bias. This approach enables the simulation to explore orthogonal directions of collective variable (CV) space, reducing the likelihood of hysteresis effects(Invernizzi, M. and Parrinello, M., JCTC, 2019[6]).

Furthermore, there is precedent for using metadynamics-derived pathways to initiate unbiased simulations for constructing Markov State Models (MSMs). This methodology has been successfully applied in studying G-protein activation (Sun, X. et al., elife, 2018[4]).

Additional support to our observation can be found in two independent binding/unbinding studies of ligands from cannabinoid receptors, which have discovered similar pathway using different CVs (Saleh, et al., Angew. Chem., 2018[7]; Hua, T. et al., Cell, 2020[8]).   

(3) The choice of ligands in the current work seems very forced and none of the results compare directly with any experimental data. An ideal case would have been to use the seminal D.E. Shaw research paper on GPCR/ligand binding as a benchmark and then show how TRAM, using much lesser biased simulation times, would fare against the experimental kinetics or even unbiased simulated kinetics of the previous report 

We would like to address the reviewer's concerns regarding the choice of ligands, lack of direct experimental comparison, and the use of TRAM, and clarify our rationale point by point:

Ligand Choice: The ligands selected for this study were chosen due to their relevance and well characterized binding properties. MDMB-FUBINACA is well-known NPS ligand with documented binding properties. This ligand is still the only NPS ligand with experimentally determined CB1 bound structure (Krishna Kumar, K. et al., Cell, 2019[9]). Similarly, the classical cannabinoid (HU-210) used in this study has established binding characteristics and is one of earliest known synthetic classical cannabinoid. Therefore, these ligands serve as representative compounds within their respective categories, making them suitable for our comparative analysis.

Experimental Comparison: We have indeed compared our simulation results to experimental data, particularly focusing on binding free energies. In the result section, we have shown that the relative binding free energy estimated from our simulation aligns closely with the experimentally measured values. Additionally, Absolute binding energy estimates are also within ~3 kcal/mol of the experimentally predicted value.

TRAM Performance: TRAM estimated free energies, and rates have been benchmarked against experimental predictions for various studies along with our study (Peptide-protein binding: Paul, F. et al., Nat. Commun., 2017[2]; Ligand unbinding: Wu, H. et al., PNAS, 2016[10]) . As the primary goal of this study is to compare ligand unbinding mechanism, we believe benchmarking against other datasets, such as the D.E. Shaw GPCR/ligand binding paper, is not essential for this work.

(4) The method section of the manuscript seems to suggest all the simulations were started from a docked structure. This casts doubt on the reliability of the kinetics derived from these simulations that were spawned from docked structure, instead of any crystallographic pose. Ideally, the authors should have been more careful in choosing the ligands in this work based on the availability of the crystallographic structures. 

We thank the reviewer for the comment. We would like to clarify that we indeed used an experimentally derived pose for one of the ligands (MDMB-FUBINACA) as the cryo-EM structure of MDMB-FUBINACA bound to the protein was available (PDB ID: 6N4B) (Krishna Kumar K. et al., Cell, 2019[9]). However, as the cryo-EM structure had missing loops, we modeled these regions using Rosetta. We apologize for this confusion and have modified our method section to make this point clearer. 

Regarding HU-210, we acknowledge that a crystallographic or cryo-EM structure for this specific ligand was not available. We selected HU-210 because it is most commonly used example of classical cannabinoid in the literature with extensively studied thermodynamic properties. Importantly, our docking results for HU-210 align closely with previously experimentally determined poses for other classical cannabinoids (Figure S11) and replicate key polar interactions, such as those with S383^7.39, which are characteristic of this class of compounds. 

System Preparation (Page 22)

“Modeling of this membrane proximal region was also performed Remodel protocol of Rosetta loop modeling. A distance constraint is added during this modeling step between C98N−term and C107N−term to create the disulfide bond between the residues. [74,76] 

As the cryo-EM structure of MDMB-FUBINACA was known, ligand coordinate of MDMB- FUBINACA was added to the modeled PDB structure. The “Ligand Reader & Modeler” module of CHARMM-GUI was used for ligand (e.g., MDMB-Fubinaca) parameterization using CHARMM General Force Field (CGenFF).[77]”

(5) The last part of using a machine learning-based approach to analyze allosteric interaction seems to be very much forced, as there are numerous distance-based more traditional precedent analyses that do a fair job of identifying an allosteric job. 

We thank the reviewer for the valuable comment. Neural relational inference method, which leverages a VAE (Variational Autoencoder) architecture, attempts to reconstruct the conformation (X) at time t + τ based on the conformation at time t. In doing so, it captures the non-linear dynamic correlations between residues in the VAE latent space. We chose this method because it is not reliant on specific metrics such as distance or angle, making it potentially more robust in predicting allosteric effects between the binding pocket residues and the NPxxY motif.

In response to the reviewer's suggestion, we have also performed a more traditional allosteric analysis by calculating the mutual information between the binding pocket residues and the NPxxY motif. Mutual information was computed based on the backbone dihedral angles, as this provides a metric that is independent of the relative distances between residues. Our results indicate that the mutual information between the binding pocket residues and the NPxxY motif is indeed higher for the NPS binding simulation (Figure S11).

Method

Mutual information calculation

Mutual information was calculated on same trajectory data as NRI analysis. Python package MDEntropy was used for estimating mutual information between backbone dihedral angles of two residues. 

Results and Discussion (Page 21)

“To further validate our observations, we estimated allosteric weights between the binding pocket and the NPxxY motif by calculating mutual information between residue movements. Mutual information analysis reaffirms that allosteric weights between these residues are indeed higher for the MDMB-FUBINACA bound ensemble (Figure S11).”

Mutual Information Estimation (Page 37)

“Mutual information between dynamics of residue pairs was computed based on the backbone dihedral angles, as this provides a metric that is independent of the relative distances between residues. The calculations were done on same trajectory data as NRI analysis. Python package MDEntropy was used for estimating mutual information between backbone dihedral angles of two residues.[124]”

(6) While getting busy with the methodological details of TRAM vs MSM, the manuscript fails to share with sufficient clarity what the distinctive features of two ligand binding mechanisms are. 

We thank the reviewer for the insightful comment. In the manuscript, we discussed that the overall ligand (un)binding pathways are indeed similar for both ligands. Therefore, they interact with similar residues during the unbinding process. However, we have focused on two key differences in unbinding mechanism between the two ligands:

(1) MDMB-FUBINACA exhibits two distinct unbinding mechanisms. In one, the linked portion of the ligand exits the receptor first. In the other mechanism, the ligand rotates within the pocket, allowing the tail portion to exit first. By contrast, for HU-210, we observe only a single unbinding mechanism, where the benzopyran ring leads the ligand out of the receptor. We have highlighted these differences in the Figure 6 and 7 and talked about the intermediate states appear along these different unbinding mechanisms. For further clarification of these differences, we have added arrows in the free energy landscapes to highlight these distinct pathways.

(2) In the bound state, a significant difference is observed in the interaction profiles. HU-210, a classical cannabinoid, forms strong polar interactions with TM7, while MDMB-FUBINACA shows weaker polar interactions with this region.

We have discussed these differences in the Results and Discussion section (Page 13-18) & conclusion section (Page 23-24).

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors): 

(1) The authors should choose at least one case where the ligand's crystallographic pose is known and show how TRAM works in comparison to MSM or experimental report. 

We thank the reviewer for the comment. We have used the experimentally determined cryo-EM pose for one of the ligands (i.e. MDMB-FUBINACA).  We have modified the manuscript to avoid confusion. (Please refer to the response of comment 4 of reviewer 2)

(2) The authors should consider existing traditional methods that are used to detect allostery and compare their machine-learning-based approach to show its relevance. 

We appreciate the reviewer’s comment. We have performed the traditional analysis by calculating mutual information between residue dynamics. We have shown that the traditional analysis matches with Machine learning based NRI calculation. (Please refer to the response of comment 5 of reviewer 2)

(3) Figure 3 doesn't provide a guide on the pathway of ligand. Without a proper arrow, it is difficult to surmise what is the start and end of the pathway. The figures should be improved. 

We appreciate the reviewer’s suggestion. In response, we have revised Figure 3 to clearly indicate the ligand’s unbinding pathway by adding directional arrows and labeling the bound pose. Additionally, we have updated the figure caption to better clarify the color scheme used in the illustration. 

(4) The Figure 5 presentation of free energetics has a very similar shape for the two ligands. More clarity is required on how these two ligands are different. 

We thank the reviewer for the comment. While the overall shapes of the free energy profiles for the two ligands are indeed similar, this is expected as both ligands dissociate from the same pocket and follow a comparable pathway. However, key differences in their unbinding mechanisms arise due to variations in the ligand motion within the pocket. Specifically, the intermediate metastable minima in the free energy landscapes reflect these differences. For instance, in the NPS unbinding free energy landscape, the intermediate metastable state I1 corresponds to a conformation where the NPS ligand maintains a polar interaction with TM7, while the tail of the ligand has shifted away from TM5. This intermediate state is absent in the classical cannabinoid unbinding pathway, where no equivalent conformation appears in the landscape.  

(6) Page 30: TICA is wrongly expressed as 'Time-independent component analysis'. It is not a time-independent process. Rather it is 'Time structured independent component analysis'. 

We thank the reviewer for pointing this out. TICA should be expressed as Time-lagged independent component analysis or Time-structure independent component analysis. We have used the first expression and modified the manuscript accordingly.  

(7) The manuscript's MSM theory part is quite well-known which can be removed and appropriate papers can be cited. 

We thank the reviewer for the comment. We have removed the theory discussion of MSM and cited relevant papers.

“Markov State Model

Markov state model (MSM) is used to estimate the thermodynamics and kinetics from the unbiased simulation.[56,91] MSM characterizes a dynamic process using the transition probability matrix and estimates its relevant thermodynamics and kinetic properties from the eigendecomposition of this matrix. This matrix is usually calculated using either maximum likelihood or Bayesian approach.[56,97] The prevalence of MSM as a post-processing technique for MD simulations was due to its reliance on only local equilibration of MD trajectories to predict the global equilibrium properties.[92,93] Hence, MSM can combine information from distinct short trajectories, which can only attain the local equilibrium.[94–96]  

The following steps are taken for the practical implementation of the MSM from the MD data. [4,17,98–100]”

(8) A proper VAMP score-based analysis should be provided to show confidence in MSM's clustering metric and other hyperparameters. 

We thank the reviewer for the recommendation. VAMP-2 score based analysis had been discussed in the method section.  We estimated VAMP-2 score of MSM built with different cluster number and input TIC dimensions (Figure S15). Model with best VAMP-2 was selected for comparison with TRAM result.

Author response:

We thank the reviewers for the valuable and constructive reviews. Thanks to these, we believe the article will be considerably improved. We have organized our response to address points that are relevant to both reviewers first, after which we address the unique concerns of each individual reviewer separately. We briefly paraphrase each concern and provide comments for clarification, outlining the precise changes that we will make to the text.

Common Concerns (Reviewer 1 & Reviewer 2):

Can you clarify how NREM and REM sleep relate to the oneirogen hypothesis?

Within the submission draft we tried to stay agnostic as to whether mechanistically similar replay events occur during NREM or REM sleep; however, upon a more thorough literature review, we think that there is moderately greater evidence in favor of Wake-Sleep-type replay occurring during REM sleep which is related to classical psychedelic drug mechanisms of action.

First, we should clarify that replay has been observed during both REM and NREM sleep, and dreams have been documented during both sleep stages, though the characteristics of dreams differ across stages, with NREM dreams being more closely tied to recent episodic experience and REM dreams being more bizarre/hallucinatory (see Stickgold et al., 2001 for a review). Replay during sleep has been studied most thoroughly during NREM sharp-wave ripple events, in which significant cortical-hippocampal coupling has been observed (Ji & Wilson, 2007). However, it is critical to note that the quantification methods used to identify replay events in the hippocampal literature usually focus on identifying what we term ‘episodic replay,’ which involves a near-identical recapitulation of neural trajectories that were recently experienced during waking experimental recordings (Tingley & Peyrach, 2020). In contrast, our model focuses on ‘generative replay,’ where one expects only a statistically similar reproduction of neural activity, without any particular bias towards recent or experimentally controlled experience. This latter form of replay may look closer to the ‘reactivation’ observed in cortex by many studies (e.g. Nguyen et al., 2024), where correlation structures of neural activity similar to those observed during stimulus-driven experience are recapitulated. Under experimental conditions in which an animal is experiencing highly stereotyped activity repeatedly, over extended periods of time, these two forms of replay may be difficult to dissociate.

Interestingly, though NREM replay has been shown to couple hippocampal and cortical activity, a similar study in waking animals administered psychedelics found hippocampal replay without any obvious coupling to cortical activity (Domenico et al., 2021). This could be because the coupling was not strong enough to produce full trajectories in the cortex (psychedelic administration did not increase ‘alpha’ enough), and that a causal manipulation of apical/basal influence in the cortex may be necessary to observe the increased coupling. Alternatively, as Reviewer 1 noted, it may be that psychedelics induce a form of hippocampus-decoupled replay, as one would expect from the REM stage of a recently proposed complementary learning systems model (Singh et al., 2022). 

Evidence in favor of a similarity between the mechanism of action of classical psychedelics and the mechanism of action of memory consolidation/learning during REM sleep is actually quite strong. In particular, studies have shown that REM sleep increases the activity of soma-targeting parvalbumin (PV) interneurons and decreases the activity of apical dendrite-targeting somatostatin (SOM) interneurons (Niethard et al., 2021), that this shift in balance is controlled by higher-order thalamic nuclei, and that this shift in balance is critical for synaptic consolidation of both monocular deprivation effects in early visual cortex (Zhou et al., 2020) and for the consolidation of auditory fear conditioning in the dorsal prefrontal cortex (Aime et al., 2022). These last studies were not discussed in the present manuscript–we will add them, in addition to a more nuanced description of the evidence connecting our model to NREM and REM replay.

Can you explain how synaptic plasticity induced by psychedelics within your model relates to learning at a behavioral level?

While the Wake-Sleep algorithm is a useful model for unsupervised statistical learning, it is not a model of reward or fear-based conditioning, which likely occur via different mechanisms in the brain (e.g. dopamine-dependent reinforcement learning or serotonin-dependent emotional learning). The Wake-Sleep algorithm is a ‘normative plasticity algorithm,’ that connects synaptic plasticity to the formation of structured neural representations, but it is not the case that all synaptic plasticity induced by psychedelic administration within our model should induce beneficial learning effects. According to the Wake-Sleep algorithm, plasticity at apical synapses is enhanced during the Wake phase, and plasticity at basal synapses is enhanced during the Sleep phase; under the oneirogen hypothesis, hallucinatory conditions (increased ‘alpha’) cause an increase in plasticity at both apical and basal sites. Because neural activity is in a fundamentally aberrant state when ‘alpha’ is increased, there are no theoretical guarantees that plasticity will improve performance on any objective: psychedelic-induced plasticity within our model could perhaps better be thought of as ‘noise’ that may have a positive or negative effect depending on the context.

In particular, such ‘noise’ may be beneficial for individuals or networks whose synapses have become locked in a suboptimal local minimum. The addition of large amounts of random plasticity could allow a system to extricate itself from such local minima over subsequent learning (or with careful selection of stimuli during psychedelic experience), similar to simulated annealing optimization approaches. If our model were fully validated, this view of psychedelic-induced plasticity as ‘noise’ could have relevance for efforts to alleviate the adverse effects of PTSD, early life trauma, or sensory deprivation; it may also provide a cautionary note against repeated use of psychedelic drugs within a short time frame, as the plasticity changes induced by psychedelic administration under our model are not guaranteed to be good or useful in-and-of themselves without subsequent re-learning and compensation.

We should also note that we have deliberately avoided connecting the oneirogen hypothesis model to fear extinction experimental results that have been observed through recordings of the hippocampus or the amygdala (Bombardi & Giovanni, 2013; Jiang et al., 2009; Kelly et al., 2024; Tiwari et al., 2024). Both regions receive extensive innervation directly from serotonergic synapses originating in the dorsal raphe nucleus, which have been shown to play an important role in emotional learning (Lesch & Waider, 2012); because classical psychedelics may play a more direct role in modulating this serotonergic innervation, it is possible that fear conditioning results (in addition to the anxiolytic effects of psychedelics) cannot be attributed to a shift in balance between apical and basal synapses induced by psychedelic administration. We will provide a more detailed review of these results in the text, as well as more clarity regarding their relation to our model.

Reviewer 1 Concerns:

Is it reasonable to assign a scalar parameter ‘alpha’ to the effects of classical psychedelics? And is your proposed mechanism of action unique to classical psychedelics? E.g. Could this idea also apply to kappa opioid agonists, ketamine, or the neural mechanisms of hallucination disorders?

We will clarify that within our model ‘alpha’ is a parameter that reflects the balance between apical and basal synapses in determining the activity of neurons in the network. For the sake of simplicity we used a single ‘alpha’ parameter, but realistically, each neuron would have its own ‘alpha’ parameter, and different layers or individual neurons could be affected differentially by the administration of any particular drug; therefore, our scalar ‘alpha’ value can be thought of as a mean parameter for all neurons, disregarding heterogeneity across individual neurons.

There are many different mechanisms that could theoretically affect this ‘alpha’ parameter, including: 5-HT2a receptor agonism, kappa opioid receptor binding, ketamine administration, or possibly the effects of genetic mutations underlying the pathophysiology of complex developmental hallucination disorders. We focused exclusively on 5-HT2a receptor agonism for this study because the mechanism is comparatively simple and extensively characterized, but similar mechanisms may well be responsible for the hallucinatory symptoms of a variety of drugs and disorders.

Can you clarify the role of 5-HT2a receptor expression on interneurons within your model?

While we mostly focused on the effects of 5-HT2a receptors on the apical dendrites of pyramidal neurons, these receptors are also expressed on soma-targeting parvalbumin (PV) interneurons. This expression on PV interneurons is consistent with our proposed psychedelic mechanism of action, because it could lead to a coordinated decrease in the influence of somatic and proximal dendritic inputs while increasing the influence of apical dendritic inputs. We will elaborate on this point, and will move the discussion earlier in the text.

Discussions of indigenous use of psychedelics over millenia may amount to over-romanticization.

We will take great care to conduct a more thorough literature review to reevaluate our statement regarding indigenous psychedelic use (including the citations you suggested), and will either provide a more careful statement or remove this discussion from our introduction entirely, as it has little bearing on the rest of the text. The Ethics Statement will also be modified accordingly.

You isolate the 5-HT2a agonism as the mechanism of action underlying ‘alpha’ in your model, but there exist 5-HT2a agonists that do not have hallucinatory effects (e.g. lisuride). How do you explain this?

Lisuride has much-reduced hallucinatory effects compared to other psychedelic drugs at clinical doses (though it does indeed induce hallucinations at high doses; Marona-Lewicka et al., 2002), and we should note that serotonin (5-HT) itself is pervasive in the cortex without inducing hallucinatory effects during natural function. Similarly, MDMA is a partial agonist for 5-HT2a receptors, but it has much-reduced perceptual hallucination effects relative to classical psychedelics (Green et al., 2003) in addition to many other effects not induced by classical psychedelics.

Therefore, while we argue that 5-HT2a agonism induces an increase in influence of apical dendritic compartments and a decrease in influence of basal/somatic compartments, and that this change induces hallucinations, we also note that there are many other factors that control whether or not hallucinations are ultimately produced, so that not all 5-HT2a agonists are hallucinogenic. We will discuss two such factors in our revision: 5-HT receptor binding affinity and cellular membrane permeability.

Importantly, many 5-HT2a receptor agonists are also 5-HT1a receptor agonists (e.g. serotonin itself and lisuride), while MDMA has also been shown to increase serotonin, norepinephrine, and dopamine release (Green et al., 2003). While 5-HT2a receptor agonism has been shown to reduce sensory stimulus responses (Michaiel et al., 2019), 5-HT1a receptor agonism inhibits spontaneous cortical activity (Azimi et al., 2020); thus one might expect the net effect of administering serotonin or a nonselective 5-HT receptor agonist to be widespread inhibition of a circuit, as has been observed in visual cortex (Azimi et al., 2020). Therefore, selective 5-HT2a agonism is critical for the induction of hallucinations according to our model, though any intervention that jointly excites pyramidal neurons’ apical dendrites and inhibits their basal/somatic compartments across a broad enough area of cortex would be predicted to have a similar effect. Lisuride has a much higher binding affinity for 5-HT1a receptors than, for instance, LSD (Marona-Lewicka et al., 2002).

Secondly, it has recently been shown that both the head-twitch effect (a coarse behavioral readout of hallucinations in animals) and the plasticity effects of psychedelics are abolished when administering 5-HT2a agonists that are impermeable to the cellular membrane because of high polarity, and that these effects can be rescued by temporarily rendering the cellular membrane permeable (Vargas et al., 2023). This suggests that the critical hallucinatory effects of psychedelics (apical excitation according to our model) may be mediated by intracellular 5-HT2a receptors. Notably, serotonin itself is not membrane permeable in the cortex.

Therefore, either of these two properties could play a role in whether a given 5-HT2a agonist induces hallucinatory effects. We will provide a considerably extended discussion of these nuances in our revision.

Your model proposes that an increase in top-down influence on neural activity underlies the hallucinatory effects of psychedelics. How do you explain experimental results that show increases in bottom-up functional connectivity (either from early sensory areas or the thalamus)?

Firstly, we should note that our proposed increase in top-down influence is a causal, biophysical property, not necessarily a statistical/correlative one. As such, we will stress that the best way to test our model is via direct intervention in cortical microcircuitry, as opposed to correlative approaches taken by most fMRI studies, which have shown mixed results with regard to this particular question. Correlative approaches can be misleading due to dense recurrent coupling in the system, and due to the coarse temporal and spatial resolution provided by noninvasive recording technologies (changes in statistical/functional connectivity do not necessarily correspond to changes in causal/mechanistic connectivity, i.e. correlation does not imply causation).

There are two experimental results that appear to contradict our hypothesis that deserve special consideration in our revision. The first shows an increase in directional thalamic influence on the distributed cortical networks after psychedelic administration (Preller et al., 2018). To explain this, we note that this study does not distinguish between lower-order sensory thalamic nuclei (e.g. the lateral and medial geniculate nuclei receiving visual and auditory stimuli respectively) and the higher-order thalamic nuclei that participate in thalamocortical connectivity loops (Whyte et al., 2024). Subsequent more fine-grained studies have noted an increase in influence of higher order thalamic nuclei on the cortex (Pizzi et al., 2023; Gaddis et al., 2022), and in fact extensive causal intervention research has shown that classical psychedelics (and 5-HT2a agonism) decrease the influence of incoming sensory stimuli on the activity of early sensory cortical areas, indicating decoupling from the sensory thalamus (Evarts et al., 1955; Azimi et al., 2020; Michaiel et al. 2019). The increased influence of higher-order thalamic nuclei is consistent with both the cortico-striatal-thalamo-cortical (CTSC) model of psychedelic action as well as the oneirogen hypothesis, since higher-order thalamic inputs modulate the apical dendrites of pyramidal neurons in cortex (Whyte et al., 2024).

The second experimental result notes that DMT induces traveling waves during resting state activity that propagate from early visual cortex to deeper cortical layers (Alamia et al., 2020). There are several possibilities that could explain this phenomenon: 1) it could be due to the aforementioned difficulties associated with directed functional connectivity analyses, 2) it could be due to a possible high binding affinity for DMT in the visual cortex relative to other brain areas, or 3) it could be due to increases in apical influence on activity caused by local recurrent connectivity within the visual cortex which, in the absence of sensory input, could lead to propagation of neural activity from the visual cortex to the rest of the brain. This last possibility is closest to the model proposed by (Ermentrout & Cowan, 1979), and which we believe would be best explained within our framework by a topographically connected recurrent network architecture trained on video data; a potentially fruitful direction for future research.

Shouldn’t the hallucinations generated by your model look more ‘psychedelic,’ like those produced by the DeepDream algorithm?

We believe that the differences in hallucination visualization quality between our algorithm and DeepDream are mostly due to differences in the scale and power of the models used across these two studies. We are confident that with more resources (and potentially theoretical innovations to improve the Wake-Sleep algorithm’s performance) the produced hallucination visualizations could become more realistic, but we believe this falls outside the scope of the present study.

We note that more powerful generative models trained with backpropagation are able to produce surreal images of comparable quality (Rezende et al., 2014; Goodfellow et al., 2020; Vahdat & Kautz, 2020), though these have not yet been used as a model of psychedelic hallucinations. However, the DeepDream model operates on top of large pretrained image processing models, and does not provide a biologically mechanistic/testable interpretation of its hallucination effects. When training smaller models with a local synaptic plasticity rule (as opposed to backpropagation), the hallucination effects are less visually striking due to the reduced quality of our trained generative model, though they are still strongly tied to the statistics of sensory inputs, as quantified by our correlation similarity metric (Fig. 5b). We will provide a more detailed explanation of this phenomenon when we discuss our model limitations in our revised manuscript.

Your model assumes domination by entirely bottom-up activity during the ‘wake’ phase, and domination entirely by top-down activity during ‘sleep,’ despite experimental evidence indicating that a mixture of top-down and bottom-up inputs influence neural activity during both stages in the brain. How do you explain this?

Our use of the Wake-Sleep algorithm, in which top-down inputs (Sleep) or bottom-up inputs (Wake) dominate network activity is an over-simplification made within our model for computational and theoretical reasons. Models that receive a mixture of top-down and bottom-up inputs during ‘Wake’ activity do exist (in particular the closely related Boltzmann machine (Ackley et al., 1985)), but these models are considerably more computationally costly to train due to a need to run extensive recurrent network relaxation dynamics for each input stimulus. Further, these models do not generalize as cleanly to processing temporal inputs. For this reason, we focused on the Wake-Sleep algorithm, at the cost of some biological realism, though we note that our model should certainly be extended to support mixed apical-basal waking regimes. We will make sure to discuss this in our ‘Model Limitations’ section.

Your model proposes that 5-HT2a agonism enhances glutamatergic transmission, but this is not true in the hippocampus, which shows decreases in glutamate after psychedelic administration.

We should note that our model suggests only compartment specific increases in glutamatergic transmission; as such, our model does not predict any particular directionality for measures of glutamatergic transmission that includes signaling at both apical and basal compartments in aggregate, as was measured in the provided study (Mason et al., 2020).

You claim that your model is consistent with the Entropic Brain theory, but you report increases in variance, not entropy. In fact, it has been shown that variance decreases while entropy increases under psychedelic administration. How do you explain this discrepancy?

Unfortunately, ‘entropy’ and ‘variance’ are heavily overloaded terms in the noninvasive imaging literature, and the particularities of the method employed can exert a strong influence on the reported effects. The reduction in variance reported by (Carhart-Harris et al., 2016) is a very particular measure: they are reporting the variance of resting state synchronous activity, averaged across a functional subnetwork that spans many voxels; as such, the reduction in variance in this case is a reduction in broad, synchronous activity. We do not have any resting state synchronous activity in our network due to the simplified nature of our model (particularly an absence of recurrent temporal dynamics), so we see no reduction in variance in our model due to these effects.

Other studies estimate ‘entropy’ or network state disorder via three different methods that we have been able to identify. 1) (Carhart-Harris et al., 2014) uses a different measure of variance: in this case, they subtract out synchronous activity within functional subnetworks, and calculate variability across units in the network. This measure reports increases in variance (Fig. 6), and is the closest measure to the one we employ in this study. 2) (Lebedev et al., 2016) uses sample entropy, which is a measure of temporal sequence predictability. It is specifically designed to disregard highly predictable signals, and so one might imagine that it is a measure that is robust to shared synchronous activity (e.g. resting state oscillations). 3) (Mediano et al., 2024) uses Lempel-Ziv complexity, which is, similar to sample entropy, a measure of sequence diversity; in this case the signal is binarized before calculation, which makes this method considerably different from ours. All three of the preceding methods report increases in sequence diversity, in agreement with our quantification method. Our strongest explanation for why the variance calculation in (Carhart-Harris et al., 2016) produces a variance reduction is therefore due to a reduction in low-rank synchronous activity in subnetworks during resting state.

As for whether the entropy increase is meaningful: we share Reviewer 1’s concern that increases in entropy could simply be due to a higher degree of cognitive engagement during resting state recordings, due to the presence of sensory hallucinations or due to an inability to fall asleep. This could explain why entropy increases are much more minimal relative to non-hallucinating conditions during audiovisual task performance (Siegel et al., 2024; Mediano et al., 2024). However, we can say that our model is consistent with the Entropic Brain Theory without including any form of ‘cognitive processing’: we observe increases in variability during resting state in our model, but we observe highly similar distributions of activity when averaging over a wide variety of sensory stimulus presentations (Fig. 5b-c). This is because variability in our model is not due to unstructured noise: it corresponds to an exploration of network states that would ordinarily be visited by some stimulus. Therefore, when averaging across a wide variety of stimuli, the distribution of network states under hallucinating or non-hallucinating conditions should be highly similar.

One final point of clarification: here we are distinguishing Entropic Brain Theory from the REBUS model–the oneirogen hypothesis is consistent with the increase in entropy observed experimentally, but in our model this entropy increase is not due to increased influence of bottom-up inputs (it is due instead to an increase in top-down influence). Therefore, one could view the oneirogen hypothesis as consistent with EBT, but inconsistent with REBUS.

You relate your plasticity rule to behavioral-timescale plasticity (BTSP) in the hippocampus, but plasticity has been shown to be reduced in the hippocampus after psychedelic administration. Could you elaborate on this connection?

When we were establishing a connection between our ‘Wake-Sleep’ plasticity rule and BTSP learning, the intended connection was exclusively to the mathematical form of the plasticity rule, in which activity in the apical dendrites of pyramidal neurons functions as an instructive signal for plasticity in basal synapses (and vice versa): we will clarify this in the text. Similarly, we point out that such a plasticity rule tends to result in correlated tuning between apical and basal dendritic compartments, which has been observed in hippocampus and cortex: this is intended as a sanity check of our mapping of the Wake-Sleep algorithm to cortical microcircuitry, and has limited further bearing on the effects of psychedelics specifically.

Reduction in plasticity in the hippocampus after psychedelic administration could be due to a complementary learning systems-type model, in which the hippocampus becomes partly decoupled from the cortex during REM sleep (Singh et al., 2022); were this to be the case, it would not be incompatible with our model, which is mostly focused on the cortex. Notably, potentiating 5HT-2a receptors in the ventral hippocampus does not induce the head-twitch response, though it does produce anxiolytic effects (Tiwari et al., 2024), indicating that the hallucinatory and anxiolytic effects of classical psychedelics may be partly decoupled. 

Reviewer 2 Concerns:

Could you provide visualizations of the ‘ripple’ phenomenon that you’re referring to?

We will do this! For now, you can get a decent understanding of what the ‘ripple effect’ looks like from the ‘eyes closed’ hallucination condition for networks trained on CIFAR10 (Fig. 2d). The ripple effect that we are referring to is very similar, except it is superimposed on a naturalistic image under ordinary viewing conditions; to give a higher quality visualization of the ripple phenomenon itself, we will subtract out the static contribution of the image itself, leaving only the ripple phenomenon.

Could you provide a more nuanced description of alternative roles for top-down feedback, beyond being used exclusively for learning as depicted in your model?

For the sake of simplicity, we only treat top-down inputs in our model as a source of an instructive teaching signal, the originator of generative replay events during the Sleep phase, and as the mechanism of hallucination generation. However, as discussed in a response to a previous question, in the cortex pyramidal neurons receive and respond to a mixture of top-down and bottom-up processing.

There are a variety of theories for what role top-down inputs could play in determining network activity. To name several, top-down input could function as: 1) a denoising/pattern completion signal (Kadkhodaie & Simoncelli, 2021), 2) a feedback control signal (Podlaski & Machens, 2020), 3) an attention signal (Lindsay, 2020), 4) ordinary inputs for dynamic recurrent processing that play no specialized role distinct from bottom-up or lateral inputs except to provide inputs from higher-order association areas or other sensory modalities (Kar et al., 2019; Tugsbayar et al., 2025). Though our model does not include these features, they are perfectly consistent with our approach.

In particular, denoising/pattern completion signals in the predictive coding framework (closely related to the Wake-Sleep algorithm) also play a role as an instructive learning signal (Salvatori et al., 2021); and top-down control signals can play a similar role in some models (Gilra & Gerstner, 2017; Meulemans et al., 2021). Thus, options 1 and 2 are heavily overlapping with our approach, and are a natural consequence of many biologically plausible learning algorithms that minimize a variational free energy loss (Rao & Ballard, 1997; Ackley et al., 1985). Similarly, top-down attentional signals can exist alongside top-down learning signals, and some models have argued that such signals can be heavily overlapping or mutually interchangeable (Roelfsema & van Ooyen, 2005). Lastly, generic recurrent connectivity (from any source) can be incorporated into the Wake-Sleep algorithm (Dayan & Hinton, 1996), though we avoided doing this in the present study due to an absence of empirical architecture exploration in the literature and the computational complexity associated with training on time series data.

To conclude, there are a variety of alternative functions proposed for top-down inputs onto pyramidal neurons in the cortex, and we view these additional features as mutually compatible with our approach; for simplicity we did not include them in our model, but we believe that these features are unlikely to interfere with our testable predictions or empirical results.

Author response:

The following is the authors’ response to the original reviews

We thank the reviewers for their positive and constructive comments on the manuscript. In the revised manuscript we addressed these comments, which we believe have improved the quality of our work.

In summary:

(1) We acknowledge the reviewer's suggestion to incorporate open-source segmentation and tracking functionalities, increasing its accessibility to a wider user base; however, these additions fall outside the primary scope of our current work, which is to provide an analytical framework for IVM data after segmentation and tracking. Developing open-source segmentation and tracking tools represents a substantial undertaking in its own right, which has been comprehensively explored in other studies (e.g. https://doi.org/10.4049/jimmunol.2100811; https://doi.org/10.7554/eLife.60547; https://doi.org/10.1016/j.media.2022.102358; https://doi.org/10.1038/s41592024-02295-6 - now cited in our revised manuscript). 

In our analyses, we used data processed with Imaris, a commercial software that, despite its limitations, is widely used by the intravital microscopy community due to its user-friendly platform for 3D image visualization and analysis. Nevertheless, recognizing the need for compatibility with tracking data from various pipelines, we have modified our tool to accept other data formats, such as those generated by open-source Fiji plugins like TrackMate, MTrackJ, ManualTracking (https://github.com/imAIgene-Dream3D/BEHAV3D_Tumor_Profiler?tab=readme-ov-file#data-input). These updates are available in our GitHub repository and are described in the revised manuscript. 

(2) We appreciate the reviewer #3 suggestion to incorporate additional features into our analytical pipeline. In response, we have already updated the GitHub repository to allow users to input and select which features (dynamic, morphological, or spatial) they wish to include in the analysis (https://github.com/imAIgene-Dream3D/BEHAV3D_Tumor_Profiler?tab=readmeov-file#feature-selection ). In the revised manuscript, we highlighted this new functionality and provided examples using alternative datasets to demonstrate the application of these features.

(3)  We appreciate the constructive feedback of reviewers #1 and #2 regarding the statistical analysis and interpretation of the data presented in Figures 3 and 4. We understand the importance of clarity and rigor in data analysis and presentation, and we addressed the concerns raised in the revised version of the manuscript.

(4) We appreciate reviewer #1's suggestion regarding the inclusion of demo data, as we believe it would greatly enhance the usability of our pipeline. We acknowledge that this was an oversight on our part. To address this, we have now added demos to our GitHub repository (https://github.com/imAIgene-

Dream3D/BEHAV3D_Tumor_Profiler/tree/BEHAV3D_TP-v2.0/demo_datasets). In the revised manuscript, we referenced this addition and present new figures with examples of these demo’s processing different IVM dataset (2D/3D, different tumors and healthy tissues). Additionally, we have provided processed DMG IVM movie samples in an imaging repository.

(5) Finally, we made some small changes to the manuscript based on the reviewers’ feedback.

Below we provide a point-by-point response to the reviewers’ comments

Reviewer #1 (Public review):

Comment #1: A key limitation of the pipeline is that it does not overcome the main challenges and bottlenecks associated with processing and extracting quantitative cellular data from timelapse and longitudinal intravital images. This includes correcting breathing-induced movement artifacts, automated registration of longitudinal images taken over days/weeks, and accurate, automated segmentation and tracking of individual cells over time. Indeed, there are currently no standardised computational methods available for IVM data processing and analysis, with most laboratories relying on custom-built solutions or manual methods. This isn't made explicit in the manuscript early on (described below), and the researchers rely on expensive software packages such as IMARIS for image processing and data extraction to feed the required parameters into their pipeline. This limitation unfortunately reduces the likely impact of BEHAV3D-TP on the IVM field. 

As highlighted above, the tool does not facilitate the extraction of quantitative kinetic cellular parameters (e.g. speed, directionality, persistence, and displacement) from intravital images. Indeed, to use the tool researchers must first extract dynamic cellular parameters from their IVM datasets, requiring access to expensive software (e.g. IMARIS as used here) and/or above-average computational expertise to develop and use custom-made open-source solutions. This limitation is not made explicit or discussed in the text.

We acknowledge the reviewer's suggestion to incorporate open-source segmentation and tracking functionalities, increasing its accessibility to a wider user base; however, these additions fall outside the primary scope of our current work and represent a substantial undertaking in their own right. Several studies (e.g., Diego Ulisse Pizzagalli et al., J Immunol (2022); Aby Joseph et al., eLife (2020); Molina-Moreno et al., Medical Image Analysis (2022); Hidalgo-Cenalmor et al., Nat Methods (2024); Ershov et al., Nat Methods (2022)) have comprehensively addressed these topics, and we now reference them in the revised manuscript to provide readers with relevant background.

The objective of our manuscript is not to develop a complete segmentation or tracking pipeline but rather to introduce an analytical framework capable of extracting enhanced insights from the data generated by existing tools. This goal arises from our observations of the field: despite significant investment in image processing, researchers often rely on simplistic approaches, such as averaging single parameters across conditions, which can obscure tumor heterogeneity and spatial behavioral dynamics within the tumor microenvironment.

Our current tool focuses on providing this much-needed analytical capability. For our analysis we used Imaris, a widely utilized software in the intravital microscopy (IVM) community, known for its intuitive 3D visualization and analysis platform despite certain limitations. 

In our own literature search of recent IVM studies published by leading laboratories in high-impact journals, we found that close to half used Imaris, while the remainder primarily relied on manual workflows with Fiji plugins. Thus, we consider it valuable to offer a pipeline compatible with such commonly used software, given its prevalence in the field.

However, following the suggestion of the reviewer, and to enhance the tool’s flexibility and compatibility, we have expanded the pipeline to accept data formats generated by open-source Fiji plugins, such as TrackMate, MTrackJ, and ManualTracking. These updates are detailed in the revised manuscript and are implemented in our GitHub repository (https://github.com/imAIgene-Dream3D/BEHAV3D_Tumor_Profiler?tab=readme-ov-file#data-input ), where we also provide several demos using TrackMate and Imaris processed data. This addition demonstrates our tool's capability to integrate with segmented and tracked datasets from diverse platforms, increasing its applicability to a broader range of researchers using both commercial and open-source pipelines.

Comment #2: The number of cells (e.g. per behavioural cluster), and the number of independent mice, represented in each result figure, is not included in the figure legends and are difficult to ascertain from the methods.

We appreciate the reviewer's constructive feedback regarding the clarity of the number and type of replicates used in our analyses. In the revised manuscript, we have included detailed information in the figure legends and the number of independent mice represented in each figure legend to ensure transparency. Regarding the number

of cells, we have indicated the total number of processed cells in Figure 2b legend (953 cells). Additionally, we have now included figures (Sup Fig 4c, Sup Fig 5e-g, Fig 5c,e, Sup Fig 6 c,d) for each cluster, where individual dots represent the individual cell tracks with color indicating the position and the shape indicating individual mice.

Comment #3: The data used to test the pipeline in this manuscript is currently not available, making it difficult to assess its usability. It would be important to include this for researchers to use as a 'training dataset'.

As stated above we acknowledge that this was an oversight on our part and thank the reviewer for pointing this out. To address this, we have now added demo data to our GitHub repository (BEHAV3D_Tumor_Profiler/demo_datasets at main · imAIgeneDream3D/BEHAV3D_Tumor_Profiler · GitHub). In the revised manuscript we have referenced this addition in the Data availability section. Since we included now processing with Fiji as well, we provide 4 demo datasets (https://github.com/imAIgeneDream3D/BEHAV3D_Tumor_Profiler/tree/main/demo_datasets), one processed with Imaris in 3D; and one with CellPose2.0 and Trackmate in 2D; one processed with µSAM and Trackmate in 3D and one manually processed with MtrackJ in 2D . Moreover, we now provide Imaris-processed DMG IVM movie samples in an open-source repository.

Comment #4: Precisely how the BEHAV3D-TP large-scale phenotyping module can map large-scale spatial phenotyping data generated using LSR-3D imaging data and Cytomap to 3D intravital imaging movies is unclear. Further details in the text and methods would be beneficial to aid understanding.

We appreciate the reviewer’s comment and in the revised manuscript we have now provided details in the methods section “Tumor large-scale spatial phenotyping with Cytomap” to clarify how the BEHAV3D-TP module maps LSR-3D and Cytomap data to 3D intravital imaging movies:

“To map the assigned regions onto IVM movies, a 3D image of the cluster distribution within the tumor was generated and exported for each sample (Figure Supplement 5a). Next, regions within the IVM movies were visually matched to the corresponding regions identified by the Large-Scale Phenotyping module of Cytomap (Figure 3c). For each mouse, at least one or two representative positions per matched region type were selected, cropped, and analyzed to assess tumor cell behavior, following the previously described cell tracking methodology (Imaris Cell tracking).”

Moreover, we updated Figure 3 c to further clarify these steps.

Comment #5: The analysis provides only preliminary evidence in support of the authors' conclusions on DMG cell migratory behaviours and their relationship with components of the tumour microenvironment. Conclusions should therefore be tempered in the absence of additional experiments and controls. 

We appreciate the reviewer’s comment and acknowledge that our conclusions should be tempered due to the preliminary nature of our evidence. In the revised version of the manuscript we have revised our conclusions accordingly and emphasize the necessity for additional experiments and controls to further validate our findings on DMG cell migratory behaviors and their relationship with the tumor microenvironment.

In discussion: “While our findings suggest that microenvironmental factors may influence tumor cell migration, further studies will be necessary to establish causal relationships. Additional experimental validation, such as macrophage ablation experiments, could help clarify the specific contributions of these factors.”

Reviewer #1 (Recommendations for the authors): 

(1) To test the ability of the pipeline to identify relevant patterns of migratory behaviours additional 'control' experiments would be helpful e.g. comparing non-invasive vs invasive tumour cell lines, artificially controlling migratory behaviours of cells such as implanting beads soaked in factors that would attract/repel cells? 

(2) Does the pipeline work well for a variety of cell types/contexts? e.g. can it identify and cluster more subtle migratory behaviours such as non-tumour cells during tissue development or regeneration conditions? 

We appreciate the reviewer’s valuable suggestions. In the revised manuscript, we have included additional examples demonstrating the capability of our pipeline to investigate heterogeneous cell behavior across two additional experimental setups:

(1) We have now evaluated our BEHAV3D TP heterogeneity module using IVM data from breast cancer cell lines with varying migratory capacities (DOI: 10.1016/j.yexcr.2019.04.009). In these datasets, our pipeline extends beyond predefined characteristics based solely on speed, enabling the identification of distinct cell populations. Notably, our analysis reveals that the breast cancer lines exhibit different proportions of different migratory behaviors such as Fast, Intermediate, Very slow and Static (Supplementary Fig 1).

(2) We have now evaluated our BEHAV3D TP heterogeneity module using IVM data from healthy breast epithelial cells (DOI: 10.1016/j.celrep.2024.115073), where we identify distinct morhophynamic epithelial cell populations in the terminal end but of the mammary gland that have a distinct distribution among Hormone receptor (HR) + and HR- terminal end but cells.

(3) To support biological conclusions could the authors show that ablating tumourassociated macrophages or vasculature alters the migratory patterns of nearby tumour cells? 

We appreciate the reviewer's suggestion regarding the potential effects of ablating tumor-associated macrophages or vasculature on the migratory patterns of nearby tumor cells. While these experiments would functionally validate the observations made by our method, we would like to clarify that the primary focus of our study was on the development and application of computational tools for behavioral analysis and thus we consider that delving deeper in understanding the biology behind our observation is out of the scope of the current study. However, as mentioned previously, we have carefully tempered our conclusions to acknowledge the limitations of our current study. In the revised manuscript, we explicitly highlight that experiments involving the ablation of tumor-associated macrophages or vasculature would be crucial for further understanding the biological relevance of our findings.

Minor corrections to text: 

(4) Line 63 - are references formatted correctly?

Thank you for pointing out this error. We have corrected it in the revised manuscript.

(5) Lines 161 -162 - 'intravitally imaged' used twice in a sentence.

Thank you for pointing out the typo. We have corrected it in the revised manuscript.

Reviewer #2 (Public review):

Comment#1: The strength of democratizing this kind of analysis is undercut by the reliance upon Imaris for segmentation, so it would be nice if this was changed to an open-source option for track generation.

As noted in our previous response to Reviewer #1, we would like to point out that although Imaris is a commercial software, it is widely used in the intravital microscopy community due to its user-friendly interface. We conducted a literature review to evaluate this aspect and below we include references from leading laboratories in the IVM field that utilize Imaris. One of its key advantages, which we also utilized, is semi-automated data tracking that allows for manual corrections in 3D—a process that can be more challenging in other open-source software with less effective data visualization.

However, we recognize that enhancing our pipeline's compatibility with open-source options is important. To this end, we have updated our tool to support 2D and 3D data formats generated by open-source Fiji plugins like TrackMate, MTrackJ, and ManualTracking, improving compatibility with various segmentation and tracking pipelines (https://github.com/imAIgene-Dream3D/BEHAV3D_Tumor_Profiler?tab=readme-ov-file#data-input ). In the revised manuscript, we describe the new functionality and demonstrate the operation of the BEHAV3D-TP heterogeneity module across various IVM datasets, processed in both 2D and 3D with different processing pipelines (Supplementary Fig 1-3). This includes CellPose 2.0 and the novel 'Segment Anything' model, followed by TrackMate tracking, applied to both tumor and healthy IVM data. Moreover we have developed a new web application that integrates morphological and tracking information from Segment Anything segmentation and Trackmate tracking, depicted in Supplementary Fig 3 a (https://morphotrack-merger.streamlit.app/ ). Additionally, we have updated the introduction to better clarify the scope of our study and include references to existing image processing solutions.

Comment#2: The main issue is with the interpretation of the biological data in Figure 3 where ANOVA was used to analyse the proportional distribution of different clusters. Firstly the n is not listed so it is unclear if this represents an n of 3 where each mouse is an individual or whether each track is being treated as a test unit. If the latter this is seriously flawed as these tracks can't be treated as independent. Also, a more appropriate test would be something like a Chi-squared test or Fisher's exact test. Also, no error bars are included on the stacked bar graphs making interpretation impossible. Ultimately this is severely flawed and also appears to show very small differences which may be statistically different but may not represent biologically important findings. This would need further study.

We appreciate the reviewer’s insightful comments regarding the interpretation of the biological data in Figure 3. 

To clarify, each imaged position is considered an independent biological replicate (n = 18 from a total of 6 mice). We acknowledge that the description of the statistical methods and the experimental units was not sufficiently clear in the previous version. In our original submission, we used an ANOVA to test whether the proportion of each behavioral cluster differed across the tumor microenvironment regions. Post hoc pairwise comparisons were performed using Tukey’s test, with the results shown in Supplementary Figure 2d (currently Fig 3d). However, we agree with the reviewer that this approach may be misleading when paired with stacked bar plots that lack error bars, as it can obscure individual variability and does not explicitly represent statistical uncertainty.

In the revised manuscript, we present the data as boxplots with individual data points, where each dot represents an imaged position, and the shape corresponds to a specific mouse. In Figure 3 d the y-axis displays the normalized percentage of each cluster across TME regions, expressed as z-scores. This normalization corrects for inter-mouse variability and facilitates a comparison of the relative distribution of clusters across TME regions, independent of the overall abundance differences between mice. We performed an ANOVA with Tukey's post hoc test for each individual behavioral cluster to assess differences across TME regions. Additionally, for transparency, in Supplementary Figure 5 d we provide the raw percentage values. The legends provide the number of positions and mice included in the analysis. 

Comment#3:  Figure 4 has similar statistical issues in that the n is not listed and, again, it is unclear whether they are treating each cell track as independent which, again, would be inappropriate. The best practice for this type of data would be the use of super plots as outlined in Lord et al. (2020) JCI - SuperPlots: Communicating reproducibility and variability in cell biology.

We appreciate the reviewer’s comments and suggestions regarding Figure 4. In this case as we are comparing overall the behavioral clusters features, each individual cell is treated as a unit. In the revised manuscript, we have clarified this point in the figure legend and incorporated plots in Figure 4c and 4e, indicating the mouse and imaging position each data point originates from. This enhances the visualization of reproducibility and variability in our data, demonstrating that the results are consistent across multiple mice and positions and are not driven by a single mouse or imaging position.

Comment#4: The main issue that this raises is that the large-scale phenotyping module and the heterogeneity module appear designed to produce these statistical analyses that are used in these figures and, if they are based on the assumption that each track is independent, then this will produce inappropriate analyses as a default.

We appreciate the reviewer’s comment, although we are unclear about the specific concern being raised. To clarify, in our large-scale phenotyping analysis, each position is assigned to a TME niche based on the CytoMAP analysis and the workflow outlined in Figure 3c. Multiple positions are imaged per mouse. For each position, we measure the proportion of tumor cells exhibiting a specific behavioral phenotype, and these proportions are subsequently used for statistical analysis (Figure 3 d). 

In contrast, in Supplementary Fig. 5e-g, we treat each cell track as an individual unit, grouping them by their assigned large-scale region. Here, we assess whether differences between regions can be detected using a conventional single-feature analysis—a more traditional approach. However, we find that this method loses important behavioral patterns and distinctions that BEHAV3D-TP captures.

We hope that this explanation, along with the modifications made to the figures and figure legends, provides greater clarity.  

Reviewer #3 (Public review):

Comment #1: The most challenging task of analyzing 3D time-lapse imaging data is to accurately segment and track the individual cells in 3D over a long time duration. BEHAV3D Tumor Profiler did not provide any new advancement in this regard, and instead relies on commercial software, Imaris, for this critical step. Imaris is known to have a very high error rate when used for analyzing 3D time-lapse data. In the Methods section, the authors themselves stated that "Tumor cell tracks were manually corrected to ensure accurate tracking". Based on our own experience of using Imaris, such manual correction is tedious and often required for every time step of the movie. Therefore, Imaris is not a satisfactory tool for analyzing 3D time-lapse data. Moreover, Imaris is expensive and many research labs probably can't afford to buy it. The fact that BEHAV3D Tumor Profiler critically depends on the faulty ImarisTrack module makes it unclear whether the BEHAV3D tool or the results are reliable.

If the authors want to "democratize the analysis of heterogeneous cancer cell behaviors", they should perform image segmentation and tracking using open-source codes (e.g., Cellpose, Stardisk & 3DCellTracker) and not rely on the expensive and inaccurate ImarisTrack Module for the image analysis step of BEHAV3D.

We appreciate the reviewer’s comments on the challenges of segmenting and tracking individual cells in 3D time-lapse imaging data. As mentioned previously (please refer to comment #1 to reviewer #1), our primary focus is to develop an analytical tool for comprehensive data analysis rather than developing tools for image processing. However to enhance accessibility, we have updated our tool to support data formats from open-source Fiji plugins, such as TrackMate, which will benefit users without access to commercial software (https://github.com/imAIgeneDream3D/BEHAV3D_Tumor_Profiler?tab=readme-ov-file#data-input ). In Supplementary Figures 1, 2, and 3, we present IVM data from different sources, processed using three distinct methods: MTrackJ (Supplementary Fig. 1), Cellpose + TrackMate (Supplementary Fig. 2), and µSAM + TrackMate (Supplementary Fig. 3). The latter two represent state-of-the-art deep learning approaches.

On the other hand, while we recognize the limitations of Imaris, it remains widely used in the intravital microscopy community due to its user-friendly interface for 3D visualization and semi-automated segmentation capabilities. Since no perfect tracking method currently exists, we initially utilized Imaris for its ability to allow manual correction of faulty tracks, ensuring the reliability of our results. This approach, not only widely used (see above) but was the best available option when we began our analysis, allowing us to obtain accurate results efficiently.

In the revised manuscript, we clarify the scope of our study and provide information on both Imaris and alternative processing options to strengthen the reliability of our findings:

In introduction: “While significant efforts have been made to develop opensource segmentation and tracking tools for live imaging data, including IVM22–27 fewer tools exist for the unbiased analysis of tumor dynamics. One major barrier is that implementing such analytical methods often requires substantial computational expertise, limiting accessibility for many biomedical researchers conducting IVM experiments. To bridge this gap, we present BEHAV3D Tumor Profiler (BEHAV3D-TP)  by providing a robust, user-friendly tool that allows researchers to extract meaningful insights from dynamic cellular behaviors without requiring advanced programming skills.”

In the Methods, we describe now describe not only Imaris processing pipeline, but also the µSAM segmentation pipelines and reference to CellPose IVM processing, which are combined with TrackMate for tracking. Additionally, to integrate morphological information from µSAM with tracking data from TrackMate, we developed a web tool to merge the outputs from both processing steps: https://morphotrack-merger.streamlit.app/  

Comment #2: The authors developed a "Heterogeneity module" to extract distinctive tumor migratory phenotypes from the cell tracks quantified by Imaris. The cell tracks of the individual tumor cells are all quite short, indicating relatively low motility of the tumor cells. It's unclear whether such short migratory tracks are sufficient to warrant the PCA analysis to identify the 7 distinctive migratory phenotypes shown in Figure 2d. It's also unclear whether these 7 migratory phenotypes correspond to unique functional phenotypes.  

For the 7 distinctive motility clusters, the authors should provide a more detailed analysis of the differences between them. It's unclear whether the difference in retreating, slow retreating, erratic, static, slow, slow invading, and invading correspond to functional difference of the tumor cells.

While some tumor cells exhibit limited motility, indicated by short tracks, others demonstrate significant migratory capabilities (Figure 2 Invading and Retreating cells). This variability in tumor cell behavior is a central focus of our analysis, and our tool is specifically designed to identify and distinguish these differences. Our PCA analysis effectively captures this variability, as illustrated in Figure 2 d-f. It differentiates between cells exhibiting varying degrees of migratory behavior, including both highly and less migratory phenotypes, as well as their directionality relative to the tumor core and the persistence of their movements. Thus, we believe that our approach provides valuable insights into the distinct migratory phenotypes within the tumor microenvironment. 

While our current manuscript does not provide explicit evidence linking each motility cluster to functional differences among the tumor cells, it is important to note that the state of the field supports the idea that cell dynamics can predict cell states and phenotypes. Research conducted by ourselves (Dekkers, Alieva et al., Nat Biotech, 2023) and others, such as Craiciuc et al. (Nature, 2022) and Freckmann et al. (Nat Comm, 2022) has shown that variations in cell motility patterns are indicative of underlying functional characteristics. For instance, cell morphodynamic features have been shown to reflect differences in cell types, T cell targeting states (Dekkers, Alieva et al., Nat Biotech, 2023), immune cell types (Crainiciuc et al. (Nature, 2022)), tumor metastatic potential, and drug resistance states (Freckmann et al. (Nat Comm, 2022)). In the revised manuscript, we have referenced relevant studies to underscore the biological significance of these behaviors. By doing so, we hope to clarify the potential implications of our findings and strengthen the overall narrative of our research:

In discussion: “While our current study does not provide direct functional validation of the distinct motility clusters identified, existing literature strongly supports the notion that cell dynamics can serve as a proxy for functional states and phenotypic heterogeneity. Prior work, including studies by our group[19,66]  as well as Crainiciuc et al.[35] and Freckmann et al.[20], has demonstrated that variations in cell motility patterns can reflect underlying functional characteristics. Specifically, cell morpho-dynamic features have been shown to correlate with differences in cell type identity, T-cell engagement, metastatic potential, and drug resistance states. This growing body of evidence suggests that tumor cell behavior, as captured by BEHAV3D-TP, may serve as a predictive tool for deciphering functional tumor heterogeneity. Future studies integrating transcriptomic or proteomic profiling of motility-defined subpopulations could further elucidate the biological significance of these behavioral phenotypes.”

Comment #3: Using only motility to classify tumor cell behaviours in the tumor microenvironment (TME) is probably not sufficient to capture the tumor cell difference. There are also other non-tumor cell types in the TME. If the authors aim to develop a computational tool that can elucidate tumor cell behaviors in the TME, they should consider other tumor cell features, e.g., morphology, proliferation state, and tumor cell interaction with other cell types, e.g., fibroblasts and distinct immune cells.

The authors should expand the scale of tumor behavior features to classify the tumor phenotype clusters, e.g., to include tumor morphology, proliferation state, and tumor cell interaction with other TME cell types.

We believe that using dynamic features alone is sufficient to capture differences in tumor behavior, as demonstrated by our results in Figure 2. However, we appreciate the reviewer’s suggestion to consider additional features, such as cell morphology, to finetune our analyses. To this end, we have adapted our pipeline to be compatible with any dynamic, morphologic or spatial features present in the data. In the revised manuscript we showcase this new addition with the analyses of two new dataset: 2D IVM data from healthy epithelial breast cells (Supplementary Fig 2) and 3D IVM data from adult gliomas (Supplementary Fig 3). These analyses identified cells with specific morphodynamic characteristics, which exhibited distinct kinetic behaviors or spatial distributions.

However, we would like to point out that not all features may provide informative insights and that a wide range of features can instead introduce biologically irrelevant noise, making interpretation more challenging. For instance, in 3D microscopy, the zaxis resolution is typically lower, which can lead to artifacts like elongation in that direction. Adding morphological features that capture this may skew the analysis. Therefore, we believe that incorporating additional features should be approached with caution. We clarify these considerations in the revised manuscript to better guide users in utilizing our computational tool effectively:

In discussion: “In addition to motility-based classification, features such as tumor cell morphology, proliferation state, and interactions with the tumor microenvironment can further refine tumor phenotyping. BEHAV3D-TP allows for the selection of diverse feature types, supporting datasets that include both dynamic, morphological and spatial parameters. However, we recognize that expanding the feature set may introduce biologically irrelevant noise, particularly in 3D microscopy data where limited z-axis resolution can lead to morphological artifacts. This highlights the potential need in the future to include unbiased feature selection strategies, such as bootstrapping methods67, to ensure the identification of meaningful and biologically relevant parameters. Careful consideration of these aspects is key to maximizing the interpretability and predictive value of analyses performed with BEHAV3D-TP.”

Comment #4: The authors have already published two papers on BEHAV3D [Alieva M et al. Nat Protoc. 2024 Jul;19(7): 2052-2084; Dekkers JF, et al. Nat Biotechnol. 2023 Jan;41(1):60-69]. Although the previous two papers used BEHAV3D to analyze T cells, the basic pipeline and computational steps are similar, in particular regarding cell segmentation and tracking. The addition of a "Heterogeneity module" based on PCA analysis does not make a significant advancement in terms of image analysis and quantification.

We want to emphasize that we have no intention of duplicating our previous publications. In this manuscript, we have consistently cited our foundational papers, where BEHAV3D was first developed for T cell migratory analysis in in vitro settings. In the introduction, we clearly state that our earlier work inspired us to adopt a similar approach for analyzing cell behavior in intravital microscopy (IVM) data, addressing the specific needs and complexities of analyzing tumor cell behaviors in the tumor microenvironment.

Importantly, our new work provides several key advancements: 1) a pipeline specifically adapted for intravital microscopy (IVM) data; 2) integration of spatial characteristics from both large-scale and small-scale phenotyping; and 3) a zero-code approach designed to empower researchers without coding skills to effectively utilize the tool. We believe that these enhancements represent meaningful progress in the analysis of cell behaviors within the tumor microenvironment which will be valuable for the IVM community. We ensure that these points are clearly articulated in the revised manuscript:

In introduction: “In line with this concept of characterizing cellular dynamic properties for cell classification, we have previously developed an analytical platform termed BEHAV3D 19,21 allowing to perform behavioral phenotyping of engineered T cells targeting cancer. While BEHAV3D was initially developed to analyze T cell migratory behavior under controlled in vitro conditions, we sought to expand its application to investigate tumor cell behaviors in IVM data, where the complexity of the TME presents distinct analytical challenges. This manuscript builds on our foundational work but represents a significant advancement by adapting the pipeline specifically for IVM datasets.”

Reviewer #3 (Recommendations for the authors): 

(1) If the authors want to "democratize the analysis of heterogeneous cancer cell behaviors", they should perform image segmentation and tracking using open-source codes (e.g., Cellpose, Stardisk & 3DCellTracker) and not rely on the expensive and inaccurate ImarisTrack Module for the image analysis step of BEHAV3D. 

We thank the reviewer for this recommendation and as stated above we recognize that enhancing our pipeline's compatibility with open-source options is important. To this end, we have updated our tool to support data formats generated by open-source Fiji plugins like TrackMate, MTrackJ, and ManualTracking, improving compatibility with various segmentation and tracking pipelines (https://github.com/imAIgeneDream3D/BEHAV3D_Tumor_Profiler?tab=readme-ov-file#data-input ). In the revised manuscript, we detail this new functionality and demonstrate the operation of the BEHAV3D-TP heterogeneity module using an example dataset of glioma tumors.

Additionally, we have updated the introduction to better clarify the scope of our study (See comment #1 from Review #3) and include references to existing image processing solutions.

(2) For the 7 distinctive motility clusters, the authors should provide a more detailed analysis of the differences between them. It's unclear whether the difference in retreating, slow retreating, erratic, static, slow, slow invading, and invading correspond to functional difference of the tumor cells. 

As noted in the comment above, the revised manuscript now incorporates references to relevant literature that support our understanding that behavioral differences among cells are driven by their underlying functional differences (See comment #2 from Reviewer #3). Additionally, we would like to point to Figure 2d and Supplementary Fig 4 c that provide evidence of the functional distinctions between the identified clusters.

(3) The authors should expand the scale of tumor behavior features to classify the tumor phenotype clusters, e.g., to include tumor morphology, proliferation state, and tumor cell interaction with other TME cell types.

We thank the reviewer for this valuable suggestion. In the revised manuscript, we have added the flexibility to incorporate a wide range of features, including morphological ones, and enabled users to select the specific features they wish to include in their analysis. To illustrate this functionality, we have included 2 example dataset analyzed using this approach (See comment #3 from Reviewer #3). Additionally, as indicated above we emphasize the importance of careful selection and interpretation of features, as improper choices may lead to biologically irrelevant results. This clarification is intended to ensure that users apply the tool thoughtfully and derive meaningful insights.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

We thank reviewer 1 for the helpful comments. As indicated in the responses below, we have taken all comments and suggestions into consideration in this revised version of the manuscript.

Weaknesses:

While this study convincingly describes the phenotype seen upon Drp1 loss, my major concern is that the mechanism underlying these defects in zygotes remains unclear. The authors refer to mitochondrial fragmentation as the mechanism ensuring organelle positioning and partitioning into functional daughters during the first embryonic cleavage. However, could Drp1 have a role beyond mitochondrial fission in zygotes? I raise these concerns because, as opposed to other Drp1 KO models (including those in oocytes) which lead to hyperfused/tubular mitochondria, Drp1 loss in zygotes appears to generate enlarged yet not tubular mitochondria. Lastly, while the authors discard the role of mitochondrial transport in the clustering observed, more refined experiments should be performed to reach that conclusion.

It would be difficult to answer from this study whether Drp1 plays a role beyond mitochondrial fission in zygotes. However, the reasons why Drp1 KO zygotes differ from the somatic Drp1 KO model can be discussed as follows.

First, the reviewer mentioned that the loss of Drp1 in oocytes leads to hyperfused/tubular mitochondria, but in fact, unlike in somatic cells, the EM images in Drp1 KO oocytes show enlarged mitochondria rather than tubular structures (Udagawa et al., Curr Biol. 2014, PMID: 25264261, Fig. 2C and Fig. S1B-D), as in the case of zygotes in this study. Mitochondria in oocytes/zygotes have the shape of a small sphere with an irregular cristae located peripherally. These structural features may be the cause of insensitivity or resistance to inner membrane fusion the resultant failure to form tubular mitochondria as seen in somatic cell models. Nonetheless, quantitative analysis of EM images in the revised version confirmed that the mitochondria of Drp1-depleted embryos were not only enlarged but also significantly elongated (Figure 2J-2M). Therefore, in Drp1-depleted embryos, significant structural and functional (e.g., asymmetry between daughters) changes in mitochondria were observed, and these are expected to lead to defects in the embryonic development.

As for mitochondrial transport, we do not fully understand the intent of this question, but we do not entirely rule out mitochondrial transport. At least clustered mitochondria did not disperse again, but how mitochondria behave through the cytoskeleton within clusters will require further study, as the reviewer pointed out.

Reviewer #1 (Recommendations For The Authors):

(1) The authors show no effect of Myo19 Trim-Away, yet it remains unclear whether myo19 is involved in the positioning of mitochondria around the spindle. Judging by their co-localization during that stage, it might be. Therefore, in the absence of myo19, mitochondria might remain evenly distributed throughout mitosis, thus passively resulting in equal partitioning to daughter cells, with no severe developmental defects. Could the authors show a video of the whole process and discuss it?

We have newly performed live imaging of mitochondria and chromosomes in Myo19 Trim-Away zygotes (n=13). As shown in Figure 1-figure supplement 2 and Figure 1-Video 2, there were no obvious changes in mitochondrial (and chromosomal) dynamics throughout the first cleavage and no significant mitochondrial asymmetry was observed, Therefore, we conclude that depletion of Myo19 does not cause mitochondrial asymmetry during embryonic cleavage. These results are described in the revised manuscript (Line 218-221).

(2) Mitochondrial aggregation upon Drp1 depletion should be characterized in more detail: for example, % of mitochondria free, % in small clusters (> X diameter), and % in big clusters (>Y diameter).

In the revised version, mitochondrial aggregation has been quantified by comparing the cluster size and number in control, Drp1 Trim-Away and Drp1 Trim-Away embryos expressing exogenous Drp1 (mCh-Drp1) (Figure 2G, 2H). In control embryos, mitochondria were interspersed in a large number of small clusters, while in Drp1-depleted embryos, mitochondria became highly aggregated into a small number of large clusters that was reversed by expression of mCh-Drp1. These results are described in the revised manuscript (Line 242-245).

(3) The discrepancies with parthenogenetic embryos derived from Drp1 (-/-) parthenotes should be commented on. Quantification of the dimensions of the clusters would help establish the degree of similarity/difference. Could the authors comment on their hypothesis as to why the clusters are remarkably larger in Drp1 depleted zygotes?

In the revised version, we have quantified the mitochondrial aggregation in Drp1 KO parthenotes (Figure 2-figure supplement 1; the data for Drp1 KO parthenotes has been reorganized into the supplemental figure, due to lack of space in figure 2 caused by the addition of quantitative data for Drp1 Trim-Away embryos). The size of mitochondrial clusters in Drp1 KO parthenotes was significantly increased compared to controls, but as the reviewer noted, mitochondrial aggregation appears to be moderate compared to that in Drp1-depleted embryos. The phenotypic discrepancies in two Drp1-deficient embryo models is discussed below.

First, it is clear that phenotypic severity of Drp1 KO oocytes is dependent on the age of the female. Indeed, oocytes collected from 8-week-old female arrested meiosis after NEB, mainly due to marked mitochondrial aggregation (Udagawa et al., Curr Biol. 2014, PMID: 25264261), whereas oocytes from juvenile female completed meiosis (Adhikari et al., Sci Adv. 2022, PMID: 35704569), and thus Drp1 KO pathenotes were obtained from juvenile female in the present study. Comparison of mitochondrial morphology in Drp1 KO oocytes in both papers also suggests that mitochondrial aggregation in adult mice is more intense (Udagawa et al., Curr Biol. Fig. 2A) than in juvenile mice (Adhikari et al., Sci Adv. 2022: Fig. 1G, 1H), and appears to be similar to Drp1-depleted embryos in this study (Figure 2E). There may be differences in the level of Drp1 depletion in these Drp1-deficient oocytes/zygotes. Similar results occurring between juvenile and adult KO female have been reported in a previous paper (Yueh et al., Development 2021, PMID: 34935904), as adult-derived Smac3<sup>Δ/Δ<?sup> zygotes arrested at the 2-cell stage, whereas juvenile-derived Smac3<sup>Δ/Δ<?sup> zygotes have developmental competence comparable to the wild type. Remarkably, the SMC3 protein levels in juvenile Smac3<sup>Δ/Δ<?sup> oocytes was also comparable to Smc3^fl/fl. The authors surmised that the decline maternal SMC3 between juvenile and sexual maturity is probably due to the continuous induction of the promoter-Cre driver, suggesting that similar induction may also occur in Drp1 KO oocytes. In addition, we also observed not only age differences but also batch differences in Drp1 KO oocytes (and resulting embryos) such that little mitochondrial aggregation was observed in oocytes collected from some juvenile KO colonies. Therefore, for KO models showing age (sexual maturation)-dependent gradual phenotypic changes, Trim-way may be an approach that provides more reproducible results as it induces acute degradation of maternal proteins.

(4) Mitochondrial clusters in Drp1 trim-away zygotes resemble those seen when defects in mitochondrial positioning are obtained by TRAK2 induction (PMID: 38917013), pointing again to a role of actin in the clustering process. Could the authors explore the role of actin further?

TRAK2 and microtubule-dependent mechanisms may also be involved in mitochondrial dynamics during the first cleavage division, possibly in association with migration of two pronuclei. Although the mitochondrial aggregation induced by TRAK2 overexpression is similar to that in Drp1-depleted embryos, it is unlikely that changes at the EM level occurred as seen in Drp1-depleted embryos (enlarged mitochondria, etc.). In addition, in TRAK2-overexpressing embryos, rather than uneven partitioning of mitochondria, the daughter blatomeres themselves were uneven in size after cleavage, making it difficult to precisely assess the similarity between the two models.

Regarding the role of F-actin, we show that the subcellular distribution of cytoplasmic actin overlaps with that of mitochondria throughout the first cleavage and seems to accumulate in aggregated mitochondria, particularly during the mitotic phase, as higher correlation was observed (Figure 1E). Although it was not observed that actin and the myo19 motor regulate mitochondrial partitioning, as reported in somatic cell-based studies, it is possible that actin accumulated in mitochondria may be indirectly involved in mitochondrial dynamics via mitochondrial fission. For example, inverted formin 2 (INF2) enhance actin polymerization and is required for efficient mitochondrial fission as an upstream function of Drp1 (Korobova et al., Science 2013, PMID: 23349293). In the revised manuscript, we have added the description on this point. (Line 452-456)

(5) Electron microscopy images showed indeed aberrant morphology of the mitochondria, yet not a hyperfused morphology. Aspect ratio (long/short axis) quantification should be included, besides the current measurement, since mitochondria in Drp1 trim-away look bigger yet as round as in the control.

In the revised version, detailed quantitative data on EM images has been added (Figure 2J-2M). In Drp1 depleted embryos, significant increases were observed in both the major and minor axes of mitochondria. As the reviewer noted, we also assumed that mitochondria in depleted embryos were enlarged rather than elongated, but the quantification of aspect ratio shows that significant elongation occurred. These results has been described in the revised manuscript (Line 252-256).

(6) Why are mitochondria in golgi-mcherry-expressing cells showing a different morphology of the clusters?

As noted by the reviewer, compared to other mitochondrial images, Drp1-depleted embryos expressing Golgi-mCherry appear to have less mitochondrial aggregation. The exact reason is not known, but may be due to inter-lot variation of Trim21 mRNA used in this experiment. Nevertheless, substantial mitochondrial aggregation was observed compared to the control, which does not seem to affect the conclusion.

(7) Authors comment on ROS being enriched (highly accumulated) in mitochondria. However, while quantification is missing, it might seem that ROS are equally distributed in control or Drp1 Trim-Away embryos. Could the authors quantify ROS signal inside and outside of the mitochondria, perhaps using a mask drawn by mitotracker? Furthermore, it would make these data more convincing to artificially induce/deplete ROS to validate the sensitivity of the technique to variations. Also, why is ROS pattern referred to as ectopic?

Thank you for your useful suggestions. In the revised version, masked binary images were created from mitochondrial images to quantify ROS levels inside and outside mitochondria (Line 734-741). The result shows the accumulation of ROS to mitochondria in Drp1-depleted embryos (Figure 4-figure supplement 1E). The term ectopic was used to mean excessive accumulation of ROS in the mitochondria compared to normal embryos, but has been deleted as it is not very accurate.

Minor comments:

(A) Video 1: images at t=-00:20 and t=00:00 of the mtGFP are actually the same images as H2B-mCherry.

Probably a faulty filter/shutter control failed to capture GFP fluorescence at these times. It appears that the autocontrast function detected a small amount of mCherry fluorescence leakage. It would be possible to replace it with another video, but as the relevant frame were unrelated to the analysis, the previous video was used as is. The same problem also occurs in the newly added Myo19-depleted zygote movie (Figure 1-Video 2, 03:15).

(B) Could you calculate the degree of colocalization between mt-GFP and ER-mCherry in ctrl and Drp1 trim-away? While it is apparent that ER is somehow more associated with mitochondrial clusters, it would be informative to quantify it.

Since the ER is partially confined to the mitochondrial aggregation site, it was difficult to calculate correlation coefficients from fluorescence images of mt-GFP and ER-mCherry to quantitatively assess colocalization. Instead, line scan analysis of whole mitochondrial clumps showed that the peak of the ER-mCherry signal overlaps with that of mt-GFP, but this is not the case for Golgi-mCherry or peroxisome-mCherry (Figure 2-figure supplement 2A-2C).

(C) Regarding the developmental arrest: The quantification of the different stages at each developmental time could be more informative. For example, at E4.5 how many embryos are at each stage (2-cell, 4-cell, ... blastocyst)? Also, could the authors comment on the reduction in developmental competence in Figure 4C, regarding the blastocyst stage?

Many arrested embryos do not maintain their morphologies and undergo a unique degenerative process over time, known as cell fragmentation. Therefore, it is difficult to accurately determine the number of each developmental stage at, for example, E4.5 days. In this study, the 2-cell stage was observed at E1.5, the 4-8 cell at E2.5-E3.0, morula at E3.5 and the blastocyst at E4.5.

Although the rate of embryos reaching the blastocyst stage was reduced compared to that of normal embryos, the overexpression of mCh-Drp1 may explain the failure of complete restoration of developmental competence, since embryos injected solely with mCh-Drp1 mRNA also showed reduced developmental competence. For rescue experiments, the comparison with internal controls is more important and therefore we described below. This is a specific effect of Drp1 deletion because none of the internal control conditions increased arrest at the 2-cell stage and arrest was completely reversed by microinjecting Trim-away insensitive exogenous mCh-Drp1 mRNA (Line 337-340).

(D) In lines 103 to 105, proliferation should be changed to division or development.

In the revised version, proliferation has been changed to division (Line 103).

(E) Could the authors reference the statement in lines 168-169?

The following 3 references have been added (Hardy et al., 1993, PMID: 8410824; Meriano et al., 2004, PMID: 15588469; Seikkula et al., 2018, PMID: 29525505).

(F) Line 448: "Cells lacking Drp1 have highly elongated mitochondria that cannot be divided into transportable units,..." This is clearly not the case for zygotes, so why are then these mitochondria still clustering and not transported elsewhere?

Although it is difficult to answer this reviewer's question precisely, EM images of Drp1-depleted embryos suggest that individual mitochondria appear not only to be enlarged but also to have increased outer membrane attachment due to excessive aggregation. Thus, these large mitochondrial clumps may therefore be preventing transport.

Reviewer #2 (Public review):

We thank reviewer 2 for the helpful comments. As indicated in the responses below, we have taken all comments and suggestions into consideration in this revised version of the manuscript.

Weaknesses:

The authors first describe the redistribution of mitochondria during normal development, followed by alterations induced by Drp1 depletion. It would be useful to indicate the time post-hCG for imaging of fertilised zygotes (first paragraph of the results/Figure 1) to compare with subsequent Drp1 depletion experiments.

In the revised version, the time after hCG has been indicated (Line 176-182). In subsequent Drp1 depletion experiments, the revised version notes that “no significant delay in cell cycle progression was observed following Drp1 depletion (data not shown) compared to control embryos (Figure 1A)” (Line 291-193). There was a slight discrepancy in the time post-hCG between live imaging and immunofluorescence analysis (Figure 1-figure supplement 1A), which may be due to manipulation of zygotes outside incubator during the microinjection of mRNA.

It is noted that Drp1 protein levels were undetectable 5h post-injection, suggesting earlier times were not examined, yet in Figure 3A it would seem that aggregation has occurred within 2 hours (relative to Figure 1).

As the reviewer pointed out, the depletion of Drp1 is likely to have occurred at an earlier stage. In this study, due to the injection of various mRNAs to visualize organelles such as mitochondria and chromosomes, observations were started after about 5 h of incubation for their fluorescent proteins to be sufficiently expressed. Therefore, for the Western blot analysis, samples were prepared according to the time of the start of the observation.

Mitochondria appear to be slightly more aggregated in Drp1 fl/fl embryos than in control, though comparison with untreated controls does not appear to have been undertaken. There also appears to be some variability in mitochondrial aggregation patterns following Drp1 depletion (Figure 2-suppl 1 B) which are not discussed.

In the revised version, mitochondrial aggregation has been quantified by comparing the cluster size and number in control, Drp1 Trim-Away and Drp1 Trim-Away embryos expressing exogenous Drp1 (mCh-Drp1) (Figure 2G, 2H). We have also quantified the mitochondrial aggregation in Drp1^fl/fl and Drp1^Δ/Δ parhenotes (Figure 2-figure supplement 1; note that the data for Drp1 KO parthenotes has been reorganized into the supplemental figure, due to lack of space in figure 2 caused by the addition of quantitative data for Drp1 Trim-Away embryos). Mitochondria appear to be slightly more aggregated in Drp1^fl/fl embryos than in control, but no significant differences in cluster size or number were observed (data not shown). On the other hand, mitochondrial clusters in Drp1 Trim-Away embryos were remarkably larger than Drp1^Δ/Δ parhenotes, Please refer to the response to reviewer 1's comment (3) for discussion of this discrepancy.

As noted by the reviewer, compared to other mitochondrial images, Drp1-depleted embryos expressing Golgi-mCherry appear to have less mitochondrial aggregation. The exact reason is not known, but may be due to inter-lot variation of Trim21 mRNA used in this experiment. Nevertheless, substantial mitochondrial aggregation was observed compared to the control, which does not seem to affect the conclusion.

The authors use western blotting to validate the depletion of Drp1, however do not quantify band intensity. It is also unclear whether pooled embryo samples were used for western blot analysis.

In the revised version, the band intensities in Western blot analysis were quantified and validated the previous results (Figure 1H for Myo19 depletion, Figure 2B for Drp1 expression during preimplantation development, Figure 2D for Drp1 depletion). The number of embryos analyzed was described in Figure legends (Pooled samples ranging from 20 to 100 were used).

Likewise, intracellular ROS levels are examined however quantification is not provided. It is therefore unclear whether 'highly accumulated levels' are of significance or related to Drp1 depletion.

In the revised version, masked binary images were created from mitochondrial images to quantify ROS levels inside and outside mitochondria (Line 734-741). The result shows the accumulation of ROS to mitochondria in Drp1-depleted embryos (Figure 4-figure supplement 1E).

In previous work, Drp1 was found to have a role as a spindle assembly checkpoint (SAC) protein. It is therefore unclear from the experiments performed whether aggregation of mitochondria separating the pronuclei physically (or other aspects of mitochondrial function) prevents appropriate chromosome segregation or whether Drp1 is acting directly on the SAC.

In the revised manuscript, we have discussed this reference (Zhou et al., Nature Communications, PMID: 36513638) (Line 482-483).

Reviewer #2 (Recommendations For The Authors):

The authors report that disruption of F-actin organization led to asymmetry in mitochondrial inheritance, however depletion of Myo19 does not impact inheritance. The authors note in the discussion that loss of another mitochondrial motor protein, Miro, has been shown to affect mitochondrial inheritance. They suggest this may be due to reduced levels of Myo19, despite data from the present study suggesting a lack of involvement of Myo19. Given that Miro1 also interacts with microtubules, and crosstalk between actin filaments and microtubules has been reported, have the authors considered whether other motor proteins, such as KIF5, may be involved in mitochondrial movement in the zygote and therefore inheritance? Myo19 also plays a role in mitochondrial architecture. Were any differences noted at the EM level?

During oocyte meiosis and early embryonic cleavage, kinesin-5 has been reported to be important for the formation of bipolar spindles (Fitzharris, Curr Biol., 2009, PMID: 19465601) and may have some involvement in mitochondrial dynamics. Given that the migration of two pronuclei towards the zygotic centre is dynein-dependent manner (Scheffler Nat Commun. 2021PMID: 33547291), dynein may also be involved in the process of mitochondrial accumulation around the pronuclei. Nevertheless, whether microtubule-dependent mechanisms regulate mitochondrial partitioning remains controversial. Mitochondria basically diverge from microtubules at the onset of mitosis, and indeed Miro1-deleted zygotes did not show the asymmetric mitochondrial partitioning (Lee et al., Front Cell Dev Biol. 2022, PMID: 36325364). More recently, it was reported that overexpression of TRAK2 causes significant mitochondrial aggregation in embryos (Lee et al., Proc Natl Acad Sci U S A. 2024, PMID: 36325364), but since overexpression might disrupt a regulatory balance by other motors/adaptor complexes, further investigation using TRAK2-deficient embryos is expected.

As noted by the reviewer, myo19 seems to be important for the maintenance of mitochondrial cristae architecture and, consequently, for the regulation of mitochondrial function (Shi et al., Nat Commun. 2022, PMID: 35562374). We have not observed the EM images in myo19-depleted embryos, but we examined their membrane potential and ROS by TMRM and H2DCF staining, respectively, and confirmed that they were comparable to control embryos (data not shown). The loss of myo19 in zygotes/embryos did not cause any functional changes in mitochondria, suggesting that mitochondrial architecture may not be substantially affected either.

Transcriptomic analysis would be useful to identify alterations in cell cycle checkpoint regulators, as well as immunofluorescence to identify changes in spindle assembly checkpoint protein recruitment.

The present results showed that the majority of Drp1-depleted embryos arrest at the G2 stage, possibly due to cell cycle checkpoint mechanisms. Transcriptome analysis would certainly be beneficial, but eventually more detailed analysis of proteins and their phosphorylation modifications, etc. is needed for accurate assessment. These studies will be the subject of future work.

Minor comments:

There are many instances where the English could be improved, particularly the overuse of the word 'the'.

We have checked the manuscript again carefully and hopefully it has been improved some.

Line 144: replace 'took' with 'take'.

We have corrected this in the revised version (Line 140).

Line 157: it is unclear what is meant by 'hinders the functional importance of Drp1 in mature oocytes and embryos'.

This description has been corrected to “complicates the functional analysis of Drp1 in mature oocytes and embryos” (Line 152-153)

Line 198: replace with 'displayed a mitochondrial distribution pattern closely associated with'

We have corrected this in the revised version (Line 195-196).

Line 200: provide a time to clarify when the cytoplasmic meshwork was 'subsequently reorganized'

In the revised version, “at the metaphase” has been added (Line 198).

Line 204: replace 'to' with 'for'

We have corrected this in the revised version (Line 203).

Lines 285-87: consider rearranging the text to improve the flow.

To improve the flow of text before and after, the following sentence has been added; We postulated that this asymmetry was due to non-uniformity in the distribution of mitochondria around the spindle (Line 295-297)

Line 418: replace 'central' with 'centre'

We have corrected this in the revised version (Line 430).

Line 427: replace 'pertaining' with 'partitioning'

We have corrected this in the revised version (Line 438).

Line 574: clarify to what '1-5% of that of the oocytes' refers

We have corrected it to “1-5% of the total volume of the zygote.” (Line 587-588).

Line 619: indicate the dilution used

We apologize for the previous incorrect description. We used a part of the extract as the template, not a dilution, and have corrected it to be accurate (Line 631-632).

Line 634: replace 'on' with 'in' and detail in which medium embryos were mounted.

We have corrected this in the revised version (Line 647).

Please check all spelling in the figures.

Figure 1J - inheritance is spelt incorrectly.

Figure-Suppl 1, D: Interphase (PN) and (2-cell) is spelt incorrectly. G: inheritance is spelt incorrectly.

Figure 5F - bottom section prior to cytokinesis, spindle is spelt 'spincle'

Ensure consistency in abbreviation use (e.g. use of NEB and NEBD).

Thank you for your careful correction of typographical errors. In the revised version, all points raised by the reviewers have been corrected.

Reviewer #3 (Public review):

We thank reviewer 2 for the helpful comments. As indicated in the responses below, we have taken all comments and suggestions into consideration in this revised version of the manuscript.

Seemingly, there are few apparent shortcomings. Following are the specific comments to activate the further open discussion.

Line 246: Comments on cristae morphology of mitochondria in Drp1-depleted embryos would better be added.

In the revised manuscript, we have added the following comment; swollen or partially elongated mitochondria with lamella cristae structures in the inner membrane were observed in Drp1 depleted embryos. In addition, the quantification of aspect ratio (long/short axis) shows that significant mitochondrial elongation was occurred (Figure 2M). These results has been described in the revised manuscript (Line 251-256).

- Regarding Figure 2H: If possible, a representative picture of Ateam would better be included in the figure. As the authors discussed in line 458, Ateam may be able to detect whether any alterations of local energy demand occurred in the Drp1-depleted embryos.

Thank you for your very useful comments. Although it would be interesting to investigate whether alterations in ATP levels occurred in localized areas (e.g., around the spindle), the present study used conventional fluorescence microscope instead of confocal laser microscopy to observe ATeam fluorescence in order to quantify the fluorescence intensity in the whole embryo (or whole blastomere) and thus we currently cannot provide the images that reviewer expected. As shown in Figure-figure supplement 1C, the ATP levels tend to be higher at the cell periphery in control and at the mitochondrial aggregation areas in Drp1-depleted embryos, but it would need high resolution images using confocal microscopy to show it clearly.

- Line 282: In Figure 3-Video 1, mitochondria were seemingly more aggregated around female pronucleus. Is it OK to understand that there is no gender preference of pronuclei being encircled by more aggregated mitochondria?

Review of multiple videos shows that aggregated mitochondria were localized toward the cell center, but did not exhibit the behavior of preferentially concentrating near the female pronucleus.

- Line 317: A little more explanation of the "variability" would be fine. Does that basically mean that the Ca²⁺ response in both Drp1-depleted blastomeres were lower than control and blastomere with more highly aggregated mitochondria show severer phenotype compared to the other blastomere with fewer mito?

We think that the reviewer's comments are mostly correct. It is clear that there is a bias in Ca²⁺ store levels between blastomeres of Drp1 depleted embryos, However, since mitochondria were not stained simultaneously in this experiment, we cannot draw conclusions in detail, such that daughter blastomere that inherit more mitochondria have higher Ca²⁺ stores, or that blastomere with more aggregated mitochondria have lower Ca²⁺ stores.

- Regarding Figure 5B (& Figure 1-figure supplement 1B): Do authors think that there would be less abnormalities in the embryos if Drp1 is trim-awayed after 2-cell or 4-cell, in which mitochondria are less involved in the spindle?

The marked centration of mitochondrial clusters in Drp1-depleted embryos appears to be associated with migration of the pronuclei toward the cell center, which is unique to the first embryonic cleavage. Since the assembly of the male and female pronuclei at the cell center is also unique to the first cleavage, binucleation due to mitochondrial misplacement was observed only in the first cleavage. Therefore, if Drp1 is depleted at the 2-cell or 4-cell stage, chromosome segregation errors may be less frequent. However, since unequal partitioning of mitochondria is thought to occur, some abnormalities in embryonic development is likely to be observed.

Reviewer #3 (Recommendations For The Authors):

Specific comments

- Line 262: "Since mitochondrial dynamics are spatially coordinated at the ER-mitochondria MCSs," adequate ref. would better be added.

We have added an adequate reference to the revised manuscript (Friedman et al., 2011, PMID: 21885730).

- Line 333-336: "...as assessed by the presence of the nuclear envelope." Do authors show the data? In Figure 4-figure supplement 1A, the difference of the phosphoH3-ser10 signal between control and Trim-Away group might be weak. For clarity, it would be helpful if authors indicate the different points to note in the figure.

Although the data is not shown, nuclear staining of arrested 2-cell stage embryos exhibited clear nuclear membranes, similar to the DAPI image in Figure 4-figure supplement 1A. We have indicated that the data is not shown in the revised version (Line 345). Based on a report that phosphorylated histone H3 (Ser10) localizes in pericentromeric heterochromatin that hat can be visualized by DAPI staining in late G2 interphase cell (Hendzel et al., 1997, Chromosoma, PMID: 9362543), this study qualitatively estimated the G2 phase from the phosphorylated histone H3 signal and the DAPI counterstained images. We have noted this point in the revised figure legend (Line 1012-1014).

Typos or points for reword/rephrase

- Line 149: "molecular identification" may better be " molecular characteristics".

We have corrected this in the revised version (Line 145).

- Line 157: "hinders the functional importance" would be "implies the functional importance" or "complicates the functional analysis".

We have corrected this in the revised version (Line 152-153).

- Line 208: "Since the role of F-actin in many cellular events, such as cytokinesis, preclude them as targets for experimentally manipulating mitochondrial distribution, " may better be "Given many cellular roles, disruption of F-actin per se was unsuitable as a strategy for manipulating mitochondrial distribution", for example.

We have corrected this in the revised version (Line 207-208).

- Line 260: "with MCSs with the plasma.." may better be "with MCSs such as with the plasma..".

We have corrected this in the revised version (Line 267-268).

- Line 312: "distribution and segregation" may better be "distribution and the resulting segregation of the inter-organelle contacts".

We have corrected this in the revised version (Line 324-325).

- Line 427: "pertaining" might be "partitioning".

We have corrected this in the revised version (Line 438).

Line 463: "loss of Drp1 induced mitochondrial aggregation disturbs" may better be "mitochondrial aggregation induced by the loss of Drp1 disturbs".

We have corrected this in the revised version (Line 478-479).

- Line 752: "endoplasmic reticulum (pink) " would be " endoplasmic reticulum (aqua) ".

We have corrected this in the revised version (Line 780).

- Figure 5E: "(Noma 2-cell embryos)" would be "(Nomal 2-cell embryos)".

- Figure 5F: "Mitochondrial centration prevents dual spincle assembly" would be "Mitochondrial centration prevents dual spindle assembly".

Thank you for your careful correction of typographical errors. We have corrected all the words/expressions the reviewer pointed out in the revised version.