Author Response

Reviewer #1 (Public Review):

This paper combines a number of cutting-edge approaches to explore the role of a specific mouse retinal ganglion cell type in visual function. The approaches used include calcium imaging to measure responses of RGC populations to a collection of visual stimuli and CNNs to predict the stimuli that maximally activate a given ganglion cell type. The predictions about feature selectivity are tested and used to generate a hypothesized role in visual function for the RGC type identified as interesting. The paper is impressive; my comments are all related to how the work is presented.

We thank the reviewer for appreciating our study and for the interesting comments.

Is the MEI approach needed to identify these cells?

To briefly summarize the approach, the paper fits a CNN to the measured responses to a range of stimuli, extracts the stimulus (over time, space, and color) that is predicted to produce a maximal response for each RGC type, and then uses these MEIs to investigate coding. This reveals that G28 shows strong selectivity for its own MEI over those of other RGC types. The feature of the G28 responses that differentiate it appears to be its spatially-coextensive chromatic opponency. This distinguishing feature, however, should be relatively easy to discover using more standard approaches.

The concern here is that the paper could be read as indicating that standard approaches to characterizing feature selectivity do not work and that the MEI/CNN approach is superior. There may be reasons why the latter is true that I missed or were not spelled out clearly. I do think the MEI/CNN approach as used in the paper provides a very nice way to compare feature selectivity across RGC types - and that it seems very well suited in this context. But it is less clear that it is needed for the initial identification of the distinguished response features of the different RGC types. What would be helpful for me, and I suspect for many readers, is a more nuanced and detailed description of where the challenges arise in standard feature identification approaches and where the MEI/CNN approaches help overcome those challenges.

Thank you for the opportunity for clarification. In fact, the MEI (or an alternative nonlinear approach) is strictly necessary to discover this selectivity: as we show above (response #1 to editorial summary), the traditional linear filter approach does not reveal the color opponency. We realize that this fact was not made sufficiently clear in the initial submission. In the revised manuscript, we now include this analysis. Moreover, throughout the manuscript, we added explanations on the differences between MEIs and standard approaches and more intuitions about how to interpret MEIs. We also added a section to the discussion dedicated to explaining the advantages and limitations of the MEI approach.

Interpretation of MEI temporal structure

Some aspects of the extracted MEIs look quite close to those that would be expected from more standard measurements of spatial and temporal filtering. Others - most notably some of the temporal filters - do not. In many of the cells, the temporal filters oscillate much more than linear filters estimated from the same cells. In some instances, this temporal structure appears to vary considerably across cells of the same type (Fig. S2). These issues - both the unusual temporal properties of the MEIs and the heterogeneity across RGCs of the same type - need to be discussed in more detail. Related to this point, it would be nice to understand how much of the difference in responses to MEIs in Figure 4d is from differences in space, time, or chromatic properties. Can you mix and match MEI components to get an estimate of that? This is particularly relevant since G28 responds quite well to the G24 MEI.

One advantage of the MEI approach is that it allows to distinguish between transient and sustained cells in a way that is not possible with the linear filter approach: Because we seek to maximize activity over an extended period of time, transient cells need to be repetitively stimulated whereas sustained cells will also respond in the absence of multiple contrast changes. In the revised manuscript, we add a section explaining this, together with Figure 3-supplement 2, illustrating this point by showing that oscillations disappear when we optimize the MEI for a short time window. The benefit of a longer time window lies in the increased discriminability between transient and sustained cells, which is also shown in the new supplementary figure.

Regarding the heterogeneity of MEIs, this is most likely due to heterogeneity within the RGC group: “The mixed non-direction-selective groups G17 and G31 probably contain more than one type, as supported by multiple distinct morphologies and genetic identities (for example, G31,32, Extended Data Fig. 5) or response properties (for example, G17, see below)” (Baden et al. Nature 2016). We added a paragraph in the Results section.

Concerning the reviewer’s last point: We agree that it is important to know whether the defining feature - i.e., the selectivity for chromatic contrast - is robust against variations in other stimulus properties. New electrophysiological data included in the manuscript (Fig. 6e,f) offers some insights here. We probed G28/tSbC cells with full-field flashed stimuli that varied in chromatic contrast. Despite not matching the cell’s preferred spatial and temporal properties, this stimulus still recovered the cell’s preference for chromatic contrast. While we think it is an interesting direction to systematically quantify the relative importance of temporal, spatial and chromatic MEI properties for an RGC type’s responses, we think this is beyond the scope of this manuscript.

Explanation of RDM analysis

I really struggled with the analysis in Figure 5b-c. After reading the text several times, this is what I think is happening. Starting with a given RGC type (#20 in Figure 5b), you take the response of each cell in that group to the MEI of each RGC type, and plot those responses in a space where the axes correspond to responses of each RGC of this type. Then you measure euclidean distance between the responses to a pair of MEIs and collect those distances in the RDM matrix. Whether correct or not, this took some time to arrive at and meant filling in some missing pieces in the text. That section should be expanded considerably.

We appreciate the reviewer’s efforts to understand this analysis and confirm that they interpreted it correctly. However, we decided to remove the analysis. The point we were trying to make with this analysis is that the transformation implemented by G28/tSbC cells “warps” stimulus space and increases the discriminability of stimuli with similar characteristics like the cell’s MEI. We now make this point in a - we think - more accessible manner by the new analysis about the nonlinearity of G28/tSbC cell’s color opponency (see above).

Centering of MEIs

How important is the lack of precise centering of the MEIs when you present them? It would be helpful to have some idea about that - either from direct experiments or using a model.

In the electrophysiological experiments, the MEIs were centered precisely (now Fig. 5 in revised manuscript) and these experiments yielded almost identical results to the 2P imaging experiments, where the MEIs were presented on a grid to approach the optimal position for the recorded cells. Additionally, all model simulations work with perfectly centered MEIs. We hence conclude that our grid-approach at presenting stimuli provided sufficient precision in stimulus positioning.

We added this information to the revised manuscript.

Reviewer #2 (Public Review):

This paper uses two-photon imaging of mouse ganglion cells responding to chromatic natural scenes along with convolutional neural network (CNN) models fit to the responses of a large set of ganglion cells. The authors analyze CNN models to find the most effective input (MEI) for each ganglion cell as a novel approach to identifying ethological function. From these MEIs they identify chromatic opponent ganglion cells, and then further perform experiments with natural stimuli to interpret the ethological function of those cells. They conclude that a type of chromatic opponent ganglion cell is useful for the detection of the transition from the ground to the sky across the horizon. The experimental techniques, data, and fitting of CNN models are all high quality. However, there are conceptual difficulties with both the use of MEIs to draw conclusions about neural function and the ethological interpretations of experiments and data analyses, as well as a lack of comparison with standard approaches. These bear directly both on the primary conclusions of the paper and on the utility of the new approaches.

We thank the reviewer for the detailed comments.

1) Claim of feature detection.

The color opponent cells are cast as a "feature detector" and the term 'detector' is in the title. However insufficient evidence is given for this, and it seems likely a mischaracterization. An example of a ganglion cell that might qualify as a feature detector is the W3 ganglion cell (Zhang et al., 2012). These cells are mostly silent and only fire if there is differential motion on a mostly featureless background. Although this previous work does not conduct a ROC analysis, the combination of strong nonlinearity and strong selectivity are important here, giving good qualitative support for these cells as participating in the function of detecting differential motion against the sky. In the present case, the color opponent cells respond to many stimuli, not just transitions across the horizon. In addition, for the receiver operator characteristic (ROC) analysis as to whether these cells can discriminate transitions across the horizon, the area under the curve (AUC) is on average 0.68. Although there is not a particular AUC threshold for a detector or diagnostic test to have good discrimination, a value of 0.5 is chance, and values between 0.5 and 0.7 are considered poor discrimination, 'not much better than a coin toss' (Applied Logistic Regression, Hosmer et al., 2013, p. 177). The data in Fig. 6F is also more consistent with a general chromatic opponent cell that is not highly selective. These cells may contribute information to the problem of discriminating sky from ground, but also to many other ethologically relevant visual determinations. Characterizing them as feature detectors seems inappropriate and may distract from other functional roles, although they may participate in feature detection performed at a higher level in the brain.

The reviewer apparently uses a rather narrow definition of a feature detector. We, however, argue for a broader definition, which, in our view, better captures the selectivities described for RGCs in the literature. For example, while W3 cells have been quite extensively studied, one can probably agree on that so far only a fraction of the possible stimulus space has been explored. Therefore, it cannot be excluded that W3 cells respond also to other features than small dark moving dots, but we (like the reviewer) still refer to it as a feature detector. Or, for instance, direction-selective (DS) RGCs are commonly considered feature detectors (i.e., responsive to a specific motion direction), although they also respond to flashes and spike when null-direction motion is paused (Barlow & Levick J Physiol 1965).

The G28/tSbC cells’ selectivity for full-field changes in chromatic contrast enables them to encode ground-sky horizon transitions reliably across stimulus parameters (e.g., see new Fig. 7i panel). This cell type is thus well-suited to contribute to detecting context changes, as elicited by ground-sky transitions.

Therefore, we think that the G28/tSbC RGC can be considered a feature detector and as such, could be used at a higher level in the brain to quickly detect changes in visual context (see also Kerschensteiner Annu Rev Vis Sci 2022). Still, their signals may also be useful for other computations (e.g., defocus, as discussed in our manuscript).

Regarding the ROC analysis, we acknowledge that an average AUC of .68 may seem comparatively low; however, this is based on the temporally downsampled information (i.e., by way of Ca2+ imaging) gathered from the activity of a single cell. A downstream area would have access to the activity of a local population of cells. This AUC value should therefore be considered a lower bound on the discrimination performance of a downstream area. We now comment on this in the manuscript.

2) Appropriateness of MEI analysis for interpretations of the neural code.

There is a fundamental incompatibility between the need to characterize a system with a complex nonlinear CNN and then characterizing cells with a single MEI. MEIs represent the peak in a complex landscape of a nonlinear function, and that peak may or may not occur under natural conditions. For example, MEIs do not account for On-Off cells, On-Off direction selectivity, nonlinear subunits, object motion sensitivity, and many other nonlinear cell properties where multiple visual features are combined. MEIs may be a useful tool for clustering and distinguishing cells, but there is not a compelling reason to think that they are representative of cell function. This is an open question, and thus it should not be assumed as a foundation for the study. This paper potentially speaks to this issue, but there is more work to support the usefulness of the approach. Neural networks enable a large set of analyses to understand complex nonlinear effects in a neural code, and it is well understood that the single-feature approach is inadequate for a full understanding of sensory coding. A great concern is that the message that the MEI is the most important representative statistic directs the field away from the primary promise of the analysis of neural networks and takes us back to the days when only a single sensory feature is appreciated, now the MEI instead of the linear receptive field. It is appropriate to use MEI analyses to create hypotheses for further experimental testing, and the paper does this (and states as much) but it further takes the point of view that the MEI is generally informative as the single best summary of the neural code. The representation similarity analysis (Fig. 5) acts on the unfounded assumption that MEIs are generally representative and conveys this point of view, but it is not clear whether anything useful can be drawn from this analysis, and therefore this analysis does not support the conclusions about changes in the representational space. Overall this figure detracts from the paper and can safely be removed. In addition, in going from MEI analysis to testing ethological function, it should be made much more clear that MEIs may not generally be representative of the neural code, especially when nonlinearities are present that require the use of more complex models such as CNNs, and thus testing with other stimuli are required.

The reviewer correctly characterizes MEIs as representing the peak in a nonlinear loss landscape that, in this case, describes the neurons’ tuning. As such, the MEI approach is indeed capable of characterizing nonlinear neuronal feature selectivities that are captured by a nonlinear model, such as the CNN we used here. We therefore disagree with the suggestion that MEIs should not be used “when nonlinearities are present that require the use of more complex models such as CNNs”. It is unclear what other “analysis of neural networks” the reviewer refers to. One approach to analyze the predictive neural network are MEIs.

We also want to clarify that, while the reviewer is correct in stating that the MEI approach as used here only identifies a single peak, this does not mean that it cannot capture neuronal selectivities for a combination of features, as long as this combination of features can be described as a point in high-dimensional stimulus space. In fact, this is demonstrated in our manuscript for the case of G28/tSbC cell’s selectivity for large or full-field, sustained changes in chromatic contrast (a combination of spatial, temporal, and chromatic features). While approaches similar to the one used here generate several diverse exciting inputs (Ding et al. bioRxiv 2023) and could therefore also fully capture On-Off selectivities, we pointed out the limitation of MEIs when describing On-Off cells in the manuscript (both original and revised).

Regarding the reviewer’s concern that “[...] the message that the MEI is the most important representative statistic [...] takes us back to the days when only a single sensory feature is appreciated”. It was certainly not our intention to proclaim MEIs as the ultimate representation of a cell’s response features and we have clarified this in the revised manuscript. However, we also think that (i) in applying a nonlinear method to extract chromatic, temporal, and spatial response properties from natural movie responses, we go beyond many characterizations that use linear methods to extract spatial or temporal only, achromatic response properties from static, white-noise stimuli. This said, we agree that (ii) expanding around the peak is desirable, and we do that in an additional analysis (new Fig. 6); but that reducing complexity to a manageable degree (at least, at first) is useful and even necessary when discovering novel response properties.

Concerning the representational similarity analysis (RSA): the point we were trying to make with this analysis is that the transformation implemented by G28 “warps” stimulus space and increases the discriminability of stimuli with similar characteristics like the cell’s MEI. We now made this point in a more accessible fashion through the above-mentioned analysis, where we extended the estimate around the peak. We therefore agree to remove the RSA from the paper.

In the revised manuscript, we (a) discuss the advantages and limitations of the MEI approach in more detail (in Results and Discussion; see also our reply #1) and (b) replaced the RSA analysis.

3) Usefulness of MEI approach over alternatives. It is claimed that analyzing the MEI is a useful approach to discovering novel neural coding properties, but to show the usefulness of a new tool, it is important to compare results to the traditional technique. The more standard approach would be to analyze the linear receptive field, which would usually come from the STA of white noise measurement, but here this could come from the linear (or linear-nonlinear) model fit to the natural scene response, or by computing an average linear filter from the natural scene model. It is important to assess whether the same conclusion about color opponency can come from this standard approach using the linear feature (average effective input), and whether the MEIs are qualitatively different from the linear feature. The linear feature should thus be compared to MEIs for Fig. 3 and 4, and the linear feature should be compared with the effects of natural stimuli in terms of chromatic contrast (Fig. 6b). With respect to the representation analysis (Fig. 5), although I don't believe this is meaningful for MEIs, if this analysis remains it should also be compared to a representation analysis using the linear feature. In fact, a representation analysis would be more meaningful when performed using the average linear feature as it summarizes a wider range of stimuli, although the most meaningful analysis would be directly on a broader range of responses, which is what is usually done.

We agree that the comparison with a linear model is an important validation. Therefore, we performed an additional analysis (see also reply #1, as well as Fig. 6 and corresponding section in the manuscript) which demonstrates that an LN model does not recover the chromatic feature selectivity. This finding supports our claims about the usefulness of the MEI approach over linear approaches.

Regarding the comment on the representation analysis, as mentioned above, we consider it replaced by the analysis comparing results from an LN model and a nonlinear CNN.

4) Definition of ethological problem. The ethological problem posed here is the detection of the horizon. The stimuli used do not appear to relate to this problem as they do not include the horizon and only include transitions across the horizon. It is not clear whether these stimuli would ever occur with reasonable frequency, as they would only occur with large vertical saccades, which are less common in mice. More common would be smooth transitions across the horizon, or smaller movements with the horizon present in the image. In this case, cells which have a spatial chromatic opponency (which the authors claim are distinct from the ones studied here) would likely be more important for use in chromatic edge detection or discrimination. Therefore the ethological relevance of any of these analyses remains in question.

It is further not clear if detection is even the correct problem to consider. The horizon is always present, but the problem is to determine its location, a conclusion that will likely come from a population of cells. This is a distinct problem from detecting a small object, such as a small object against the background of the sky, which may be a more relevant problem to consider.

Thank you for giving us the opportunity to clear these things up. First, we would like to clarify that we propose that G28/tSbC cells contribute to detecting context changes, such as transitions across the horizon from ground to sky, not to detecting the horizon itself. We acknowledge that we were not clear enough about this in the manuscript and corrected this. To back-up our hypothesis that G28 RGCs contribute to detecting context changes, we performed an additional simulation analysis, which is described in our reply #3 (see above).

5) Difference in cell type from those previously described. It is claimed that the chromatic opponent cells are different from those previously described based on the MEI analysis, but we cannot conclude this because previous work did not perform an MEI analysis. An analysis should be used that is comparable to previous work, the linear spatiotemporal receptive field should be sufficient. However, there is a concern that because linear features can change with stimulus statistics (Hosoya et al., 2005), a linear feature fit to natural scenes may be different than those from previous studies even for the same cell type. The best approach would likely be presenting a white noise stimulus to the natural scenes model to compute a linear feature, which still carries the assumption that this linear feature from the model fit to a natural stimulus would be comparable to previous studies. If the previous cells have spatial chromatic opponency and the current cells only have chromatic opponency in the center, there should be both types of cells in the current data set. One technical aspect relating to this is that MEIs were space-time separable. Because the center and surround have a different time course, enforcing this separability may suppress sensitivity in the surround. Therefore, it would likely be better if this separability were not enforced in determining whether the current cells are different than previously described cells. As to whether these cells are actually different than those previously described, the authors should consider the following uncited work; (Ekesten Gouras, 2005), which identified chromatic opponent cells in mice in approximate numbers to those here (~ 2%). In addition, (Yin et al., 2009) in guinea pigs and (Michael, 1968) in ground squirrels found color-opponent ganglion cells without effects of a spatial surround as described in the current study.

First of all, we did not intend to claim to have discovered a completely new type of color-opponent tuning in general; what we were trying to say is that tSbC cells display spatially co-extensive color opponency, a feature selectivity previously not described in this mouse RGC type, and which may be used to signal context changes as elicited by ground-sky transitions.

Concerning the reviewer’s first argument about a lack of comparability of our results to results previously obtained with a different approach: We think that this is now addressed by the new analysis (new Fig. 6), where we show why linear methods are limited in their capability to recover the type of color opponency that we discovered with the MEI approach.

Regarding the argument about center-surround opponency, we agree that “if the previous cells have spatial chromatic opponency and the current cells only have chromatic opponency in the center, there should be both types of cells in the current data set”. We did not focus on analyzing center-surround opponency in the present study, but from the MEIs, it is visible that many cells have a stronger antagonistic surround in the green channel compared to the UV channel (see Fig. 4a, example RGCs of G21, G23, G24; Figure 3-supplement 1 example RGCs of G21, G23, G24, G31, G32). Importantly, the MEIs shown in Fig. 4a were also shown in the verification experiment, and had G28 RGCs preferred this kind of stimulus, they would have responded preferentially to these MEIs, which was not the case (Fig. 4f).

It should also be noted here that, while the model’s filters were space-time separable, we did not impose a restriction on the MEIs to be space-time separable during optimization. However, we analyzed only the rank 1 components of the MEIs (see Methods section Validating MEIs experimentally). since our analysis focused on aspects of retinal processing not contingent on spatiotemporal interactions in the stimulus.

In summary, we are convinced that our finding of center-opponency in G28 is not an artifact of the methodology.

We discuss this in the manuscript and add the references mentioned by the reviewer to the respective part of the Discussion.

Reviewer #3 (Public Review):

This study aims to discover ethologically relevant feature selectivity of mouse retinal ganglion cells. The authors took an innovative approach that uses large-scale calcium imaging data from retinal ganglion cells stimulated with both artificial and natural visual stimuli to train a convolutional neural network (CNN) model. The resulting CNN model is able to predict stimuli that maximally excite individual ganglion cell types.

The authors discovered that modeling suggests that the "transient suppressed-by-contrast" ganglion cells are selectively responsive to Green-Off, UV-On contrasts, a feature that signals the transition from the ground to the sky when the animal explores the visual environment. They tested this hypothesis by measuring the responses of these suppressed-by-contrast cells to natural movies, and showed that these cells are preferentially activated by frames containing ground-to-sky transitions and exhibit the highest selectivity of this feature among all ganglion cell types. They further verified this novel feature selectivity by single-cell patch clamp recording.

This work is of high impact because it establishes a new paradigm for studying feature selectivity in visual neurons. The data and analysis are of high quality and rigor, and the results are convincing. Overall, this is a timely study that leverages rapidly developing AI tools to tackle the complexity of both natural stimuli and neuronal responses and provides new insights into sensory processing.

We thank the reviewer for appreciating our study.

Author Response

Reviewer #3 (Public Review):

This manuscript uses ASO to inhibit the self-cleaving ribozyme within CPEB intron 3 and test its effect on CPEB3 expression and memory consolidation. The authors conclude that the intronic ribozyme negatively affects CPEB3 mRNA splicing and expression, and suggests its implications for experience-induced gene expression underlying learning and memory.

The strength of the manuscript is in its exploration of a potentially novel mechanism of regulating CPEB3 expression in learning and memory, a combination of both biochemical and behavioral approaches to gain a wide perspective of this regulatory mechanism, and the application of ASO in this context. The introduction is sufficiently detailed. Statistics are thorough and appropriate. If the results could be more robust, the mechanism would provide a novel target and venue to modify learning and memory paradigm.

The weakness of the manuscript is that the magnitude of the activity-dependent regulation of ribozyme, the effects of ASOs on CPEB3 expression (mRNA and protein) and downstream target gene expression, in vitro and in vivo, are generally weak, raising concerns about the robustness of the result. This may have caused some of the inconsistencies between the data presentation (see below). Also unclear is whether the ribozyme activity is physiologically regulated by experience without ASO interference.

While the statistics tests support corresponding figure panels and their conclusions. The manuscript can be significantly strengthened by additional evidence, clarification of some methodologies, and reconciling some inconsistent results.

The premise of a comparable timescale between transcription and ribozyme activity as the foundation of the whole thesis was based on in vitro measurement of self-scission half-life and a broadly generalized transcription rate (which actually varies significantly between genes). This premise is weak and needs direct experimental support.

The physiological relevance of the proposed mechanism has yet to be demonstrated without ASO interference.

Fig2b: how were total and uncleaved Ribozymes measured by qRT-PCR? Where are the primers' locations? If the two products were amplified using different primers, their subtraction to derive % cleavage would not be appropriate.

We thank the reviewer for the thoughtful review. We measured the levels of the total ribozyme by measuring a 220-bp amplicon that starts 18 nts downstream from the ribozyme cleavage site. The uncleaved ribozyme levels were measured using oligos that amplify a region of the intron that starts 45 nts upstream and ends 238 nts downstream of the ribozyme cleavage site. We added this information to the Table of primers in the manuscript. For all PCR oligos we established independent standard curves and calculated RNA levels independently of other amplicons, as noted in the Methods section and now specified in the Results section as well (Page 15). The measurements were thus appropriate for the calculation of the cleaved ribozyme fractions in the various experiments. The fraction ribozyme cleaved was calculated from the uncleaved fraction as the difference between uncleaved fraction and unity (1 – fraction uncleaved), now specified on page 16 of the manuscript. Fraction uncleaved was calculated as [uncleaved ribozyme]/[total ribozyme], as was done previously (see Salehi-Ashtiani et al. Science 313:1788-1792 or Webb et al. Science 326:953).

Line 400-403: shouldn't ribozyme-blocking ASO prevent ribozyme self-cleavage, and as a result should further increase ribozyme levels? This would contradict the result in fig3a.

We showed that the ribozyme is inhibited in vitro (Fig. 1F and 1G) and all our data are consistent with ASO inhibition of the ribozyme in cellulo and in vivo. However, we do not have direct evidence for this ribozyme inhibition in vivo, because such an experiment would require a single-molecule FRET-type sensitivity in cells and this assay has not been developed for ribozyme cleavage in cellulo or in vivo. We measured the ribozyme levels by RT-qPCR and observed lower ribozyme levels in presence of ASO in cultured neurons (Fig. 3A) as well as in vivo (Fig. 5B), which is nominally in contrast to the observations in vitro. However, in these situations we do not measure the co-transcriptional fate of the intron or the ribozyme; rather, we measure the levels of the intron after splicing (evidenced by the increased levels of spliced exons 2–3) when the intron is likely already being degraded. We also do not know what effect the ribozyme ASO has on the intron stability once splicing occurs. Understandably, this is a weakness of the study—and we are fully open about this result— however, given the abundance of evidence that the ribozyme ASO leads to increase of CPEB3 mRNA under all conditions tested, we feel that there is strong, if indirect, evidence that our model for the ribozyme function is correct. Future studies will examine this issue closer, but a definitive experimental investigation for the mechanism and timing of ribozyme inhibition and intron degradation is out of scope of this study.

Author Response

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

Reviewer 1 (Public Review):

Weakness: Although the cross-links stimulate ATP hydrolysis, further controls are needed to convince me that the TM1 conformations observed in the structures are physiologically relevant, since they have been trapped by "large" substrates covalently-tethered by crosslinks.

Our response: Reviewer 1 raised concerns about the relatively large size of our covalently attached AAC substrate that would potentially distort TM1 in Pgp. We would like to clarify that AAC has a molecular weight of 462 Da, which, in comparison to many known Pgp substrates ranging from 250 to over 1,000 Da, is not a large compound. For instance, the few other Pgp substrates mentioned in our manuscript all have a comparable or larger size: verapamil, 455 Da; doxorubicin, 544 Da; FK506, 804 Da; valinomycin, 1,111 Da; cyclosporin A, 1,203 Da.

Furthermore, AAC was strategically attached to a site distant from TM1 in the inwardfacing Pgp conformation. After it was exported to the outward-facing state, several TM helices accommodate the compound. The observation that only TM1 exhibited significant conformational changes suggests its potential role in the transport mechanism. This hypothesis is supported by our findings, where a conservative substitution (G72A) in TM1 resulted in a dramatic loss of transport function for various drug substrates and impaired verapamil-stimulated ATPase activity.

Reviewer 1 (Recommendations for the Authors):

I understand the need for an unconventional approach to understanding the translocation pathway. What would help to support this model is to cross-link a much smaller substrate, as the one used is quite large and could potentially distort TM1 in the outward-state when cross-linked.

Our response: We thank the reviewer for this recommendation, and we have outlined plans for future experiments involving other substrates, including smaller ones, to further investigate our proposed model. However, it is important to acknowledge that conducting these studies will require a significant amount of effort and resources, which we believe extend beyond the scope of our current manuscript.

In unbiased MD simulations starting from the IF state are there any simulations where the substrate follows the same path as proposed here?

Our response: All our MD simulations were performed in the outward-facing state to focus on potential substrate release pathways. Starting MD simulations from the inwardfacing state would introduce complexities in capturing the necessary domain motions and nucleotide binding and hydrolysis required for substrate translocations. Therefore, we opted not to perform MD studies starting from the inward-facing state.

Reviewer 2 (Public Review):

Weakness: There is much to like about the experimental work here but I am less sanguine on the interpretation. The main idea is to covalently link via disulfide bonds a model tripeptide substrate under different conditions that mimic transport and then image the resulting conformations. The choice of the Pgp cysteine mutants here is critical but also poses questions regarding the interpretation. What seems to be missing, or not reported, is a series of control experiments for further cysteine mutations.

Our response: Reviewer 2 raised concerns about the interpretation of our results and suggested the need for additional mutant designs to validate our proposed TM1 mechanism. Firstly, we believe that the observed TM1 conformational changes are valid in our cryoEM structures, despite the use of different conditions and several mutants to capture Pgp in the outward-facing state.

Regarding the G72A mutant, we consider it conclusive that this single point mutation in the TM1 has a profound effect. Importantly, the G72A mutant was readily expressed and purifiable as a stable protein. We were able to resolve a high-resolution structure of the G72A mutant (without the substrate), confirming that the protein is not generally destabilized but properly folded.

Above all, we appreciate the Reviewer’s suggestion to explore additional mutations and intend to do so in future studies.

Reviewer 2 (Recommendations for the Authors):

I am sold on the results regarding TM1 conformational changes as they are evident in the cryoEM structures. However, the set of states compared between mutants are not biochemically equivalent: for 335 and 978 they used an ATP-impaired Pgp whereas for 971 they used what appears to be WT, and the conformation was imaged presumably subsequent to ATP hydrolysis and Vanadate trapping. This is significant if the authors were unable to trap the OF in the impaired mutant background and should be highlighted. I have to believe that they tried that condition but I could be wrong.

Our response: We acknowledge the point made by the Reviewer about the biochemical equivalence of mutant states and the potential significance of using an ATP-impaired mutant for trapping the outward-facing conformation of 971. We have not yet attempted to use the ATPase-deficient 971C mutant for crosslinking and intend to address this question in future studies.

In our current approach, we used the ATPase-active 971C for two specific reasons:

1) Our biochemistry data, as shown in Fig 1C, indicates that 971C only crosslinks in the presence of ATP hydrolysis conditions. Vanadate trapping was employed to stabilize the outward-facing conformation.

2) Based on our experience, we have observed that the conformations of ATP-bound (mutant) and vanadate-trapped states of an ABC transporter are structurally equivalent at this resolution level of our study (see ref. 21: Hoffmann et al. NATURE 2019).

The authors propose a new model for substrate translocation. It is based on three mutants and a number of structures. If the authors were not challenging the current dogma I would not have written the next comment. Considering the impact of the findings, I would have designed a couple more cysteine mutants based on their model. For instance, this pathway has a number of stabilizing interactions, can't they make a mutant that preserves conformational switching but eliminates substrate translocation? I like the G97A mutant result but I am worried that the effect could just be a general destabilization or misfolding as part of the cryoEM particles seem to suggest. The authors advance one interpretation of the disorder observed in this mutant but it could easily be my interpretation.

Our response: We thank the reviewer for the suggestion to design additional mutants to further validate our proposed model for substrate translocation. We agree that this would be highly valuable, considering the potential impact of our findings. However, given the time-intensive nature of our approach, we believe that presenting these additional designs in a future study is a reasonable course of action.

Regarding the G72A mutation, we believe that our current data fully supports our model and the role of TM1 in regulating the Pgp activity. Importantly, we would like to emphasize that the G72A mutant was readily expressed and purifiable as a stable protein. Additionally, our cryoEM structural determination of the G72A mutant at high resolution confirmed that the protein is not generally destabilized but properly folded.

There are a couple of troubling methodological questions that I want the authors to address or clarify:

1. In the methods they report that the final sample for cryoEM was prepared on a SEC devoid of detergent. It is obvious that the sample was folded but I was wondering why the detergent was removed? Was that critical for observing these structures with multiple ligands? Did they observe any lipids in their cryoEM?

Our response: We avoid detergent in the buffer on final SEC purification. This step is to remove free detergent from the background which helps during cryoEM imaging. Of course, this cannot be done with every detergent but due to the very low CMC of LMNG it is possible. By now, we have verified this method for several other transporters with the same success. While this procedure helps us to obtain better images it is not necessary to obtain specific conformations or ligand bound states, nor does it affect these states or conformations.

In our cryoEM structures , we did observe multiple cholesterol hemisuccinate (CHS) molecules on the outer transmembrane surface of Pgp.

1. Can the authors comment on why labeling was carried out in the presence of ATP? Does it matter if the substrate was added prior to ATP and incubated for a few minutes?

Our response: For every dataset, we first added the substrate to be cross-linked and afterwards added the ATP. In the cases of 335C and 978C, labeling was successful before ATP was added, as evidenced by the inward-facing structures with cross-linked substrate. However, for 971C, cross-linking only occurred after the addition of ATP. We interpret this data to suggest that the 971 site is inaccessible to the substrate in the inward-facing state, and cross-linking can only occur after the transporter transitions to outward-facing state. This is in line with our inward-facing structure which does not show a cross-linked substrate, and our biochemical data shown in Fig 1C, where 971C only crosslinked in the presence of ATP.

1. I am not an expert on MD simulations and I understand that carrying out simulations at higher temperatures used to be a trick to accelerate the process. Is this still necessary? Why didn't the author use approaches such as WESTPA?

Our response: Most so-called enhanced sampling methods, including WESTPA, explicitly define a reaction coordinate for the process of interest, usually based on intuition or prior studies. If this coordinate is chosen poorly, enhanced sampling usually fails, either because the sampling becomes inefficient or because the sampling biases the transition pathway (or both). Lacking reliable intuition or prior knowledge on which motions would result in substrate release, we chose temperature to speed up the process. High temperature largely avoids the introduction of an any bias through the definition of a progress coordinate. By contrast, the weighted ensemble method underlying WESTPA is a great method to simulate unbiased dynamics of a process with a known progress coordinate, but unfortunately requires to choose a progress coordinate prior to the simulation and will then mostly sample the process along this progress coordinate, because this is the only direction in which sampling is improved. High temperature MD on the other hand accelerates all processes in the system under study. Indeed, we have now confirmed that the pathway found at high temperature is also feasible at near-ambient conditions.

In new simulations, we have now observed a similar release pathway at T=330 K. As the only difference, the substrate has not fully dissociated from the protein after 2.5 us, with weak interactions persisting at the top part of TM1 from the extracellular side. Importantly, this is a configuration observed also in higher temperature simulations but with much shorter lifetime.

In response, we now included these new findings and a new Extended Data Fig. 15 in the revised manuscript.

1. One way to show that the two substrates binding mode is biochemically relevant is to measure Vmax at different substrate concentrations. One would expect a cooperative transition if that interaction is mechanistically important.<br /> Our response: We have measured Vmax as a function of QZ-Ala concentration in a previous report (ref. 24), supporting positive cooperativity for binding to two sites.

Reviewer 3 (Public Review):

We thank Reviewer 3 for recommending the acceptance of our manuscript as is.

Reviewer 3 (Recommendations for the Authors):

Page 4, last line: Pgp302 should be Pgp1302. In addition, I can only encourage the authors to add an additional table to the manuscript. Here, the mutation, the obtained structure(s), IF or OF, the resolution, and the main message should be summarized.

Our response: Following the reviewer’s suggestion, we have added Extended Data Table 2 summarizing the Pgp mutants and respective structural data in the revised manuscript.<br /> We verified that Pgp302 is the correct term on Page 4, last line.

Pg. 5, section 'Covalent ligand design for Pgp labeling', it is mentioned that even in the presence of Mg2+ATP, Pgp302 could not react with AAC-DNPT. Maybe it would be worthwhile to add the data either in Supplementary Information or state 'data not shown'.

Our response: We stated ‘data not shown’ in the text.

Pg. 47, last line : A space is missing between M68, and M74.

Our response: Space was added.

Pg. 7, line 2: The authors mention that a single dataset of ATP-bound Pgp335 revealed three different OF conformations: ligand-free, single-ligand-bound, and double-ligandbound. However, the percentage fraction of each dataset sums up to be more than 100%. Would request the authors to recalculate the fraction size of each conformation.

Our response: We have corrected the error in our calculation, based on the particle distribution in our dataset (OF335-nolig: 1,437,110 particles, 40.4%; OF335-1lig: 1,184,253 particles, 33.3%; and OF335-2lig: 939,924 particles, 26.4%).

Pg 53, Figure legend of Extended Data Fig. 11: Please include the color coding for the helix TM1 and also the residues colored plum.

Our response: We added the color coding for TM1 and other residues in the figure legend.

Pg. 8, line 3: While referring to the structure of OF971-1lig, the authors nicely point towards the conserved residues M74 and F78 which coordinate the ligand. However, in Fig. 3b, residues M74 and F78 should also be indicated.

Our response: We updated Fig. 3b by adding arrows pointing towards the residues M74 and F78.

Pg. 54, Extended data Fig. 12: The authors should adopt a single writing style. In some places, Pgp is referred to as P-gp while in others as Pgp.

Our response: We updated the protein labels in Extended Data Fig. 12.

Pg. 54, Extended data Fig. 12: The authors should clearly mention which OF335 structure (1st panel) was used for visualizing the interactions.

Our response: To clarify, we added the following sentences in the figure legend: “Pgp335 OF in the top panel refers to OF335-1lig. In the bottom panel describing OF335-2lig, the left and right diagrams refer to the binding positions of non-covalent and covalent ligand, respectively”.

Pg. 18, section 'synthesis of dipeptide 8': In the text it is mentioned that for the synthesis of thiazole acid 6, compound 3 was dissolved in a mixture of THF/MeOH/H2O (3:1:1), while in the corresponding figure (Extended Data Fig. 1), the ratio is stated as 5:1:2.

Our response: 3:1:1 ratio is correct. We made the correction in Extended Data Fig. 1.

Pg. 19, section 'synthesis of linear tripeptide 10': Same as above for compounds 10 and 4, respectively.

Our response: We corrected the conditions in the Extended Data Fig. 1 accordingly.

Pg. 20, section 'Synthesis of cyclic peptide 11': There seems to be a discrepancy in the synthesis protocol between the text and the extended figure 1, especially regarding the use of THF/MeOH/H20, followed by NaOH and TFA or only NaOH and TFA.

Our response: we further clarified the conditions of using NaOH in THF/MeOH/H2O (3:1:1) and TFA in DCM in the text for synthesis and Extend Data Fig. 1.

Pg. 40, Extended Data Fig. 1: In the bottom last panel showing the synthesis of peptide 11, the authors have missed showing peptide 10 as the starting material for the reaction.

Our response: Label for the peptide 10 was added following the suggestion.

Pg. 26, third last line: 'o' is missing from the last word cry'o'

Our response: We corrected the typo.

Pg. 63 and 64, Extended Data Table 1: The Cryo-EM data collection, refinement, and validation statistics for OF971-1lig, IF971-1lig, OF978-1lig, and IF978-2lig are mentioned twice in the table.

Our response: This was now corrected in the revision.

Author Response

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

Recommendations for the authors:

Reviewer #1 (Recommendations for the Authors):

The authors have addressed my recommendations in the previous review round in a satisfactory way. I only have one additional comment to the authors:

In the manuscript abstract lines 31-32, the author state that: "Using NIH data for the period 2006-2022, we report that ~230 K99 awards were made every year, representing ~$25 million annually."-- The "~$25 million" is under-stating the actual funds spent because this sum is just money spent on the first year of some k99s while the NIH is paying years 2,3,4 etc for others for k99 awards (~90% conversion rate to R00) awarded in previous years for a given year. The NIH is actually spending ~$230-$250 million a year on the k99 award mechanism in a given year. so the authors need to amend the stated amount in the manuscript.

Thank you for pointing this out. The reviewer is correct, that we had incorrectly only calculated the investment $ in new K99 awards made. We have corrected this in the revised manuscript. We appreciate your careful reading of our manuscript and the edits made based on your comments have improved the final version.

Reviewer #2 (Recommendations for the Authors):

Thank you for taking the time to revise this important work. I learned a lot reading this paper a second time, and appreciate the improvements you have made.

My only major thought while re-reading this is that I wish you all had written two papers! I see two themes in this work: one looking at faculty hiring networks from the Wapman et al. dataset, and another at K99/R00 conversions by institution, gender, and researcher mobility and its impact on subsequent funding success. After reading, I felt like I had many follow-up questions about both analyses, but it would be impractical for me to suggest all these follow-up analyses without making your paper unreasonably long.

Thank you for these comments. We agree that there are 2 general themes in this paper. While we feel that significantly expanding on both themes will be important in future research. Our hope is that this work continues to inspire others to critically examine funding practices and inequity in the same way that the work of Wapman, Pickett, etc. inspired the present work.

For example, regarding the results that more R00 are activated at different institutions, and that moving institutions improves subsequent funding success, I wonder: Do proportionally more women or men move institutions? Do proportionally more K99 awardees at less-funded places move for their R00, or less? The Cox proportional hazard models illustrate the impact of various characteristics on subsequent funding success, but they do not illustrate disparate impacts of mobility on different groups (if I am understanding them correctly). (You sort of dive into these questions in the very interesting subsection, "K99/R00 awardee self-hires are more common at institutions with top NIH funding." I wanted to read more!)

Thank you for these kind comments. These are fantastic follow-up questions. We do not feel that we can adequately address them within the present manuscript without potentially splitting it into 2 separate manuscripts. However, we may examine these in future analyses. We are particularly interested in examining additional aspects such as how the K99 MOSAIC funding mechanism may differ from the traditional K99 mechanism. Since the K99 MOSAIC mechanism is newer, there may not be enough K99 MOSAIC awards made for a thorough exploration.

As another example, for your analysis on faculty hiring networks, the prevalence of self-hiring amongst institutions and regions was one finding. However, this finding seems somewhat at odds with the previous takeaway about how researcher mobility improves subsequent funding success. Are institutions doing themselves a disfavor by hiring their own, then? I suspect there is more to say here about this pattern... maybe there are important differences between PhD institution and postdoc institution and its impact on hiring/subsequent funding success? Or is this a story about upward mobility into the top 25 well-funded NIH institutions?

Again, these are very insightful comments and follow-up questions. We hope to address these in potential future manuscripts. We also hope that others may become interested in finding answers to these questions by exploring our dataset as well as other publicly available datasets such as the Wapman et al. dataset.

I can completely understand how combining the faculty hiring network analysis with the K99/R00 conversions would seem like a natural fit, but I personally feel - emphasis on this being a personal opinion - that there would have been benefits to giving more space to the details of both analyses separately. Perhaps this is a "hindsight is 20/20" issue. Or an issue with the current times in which ones' brain can only hold so many main takeaways from a single body of work. (For example, I struggled to summarize your paper in my public review because I find so many takeaways important.)

I suppose this is all to say that I find your work important enough to warrant additional follow-up work! :)

Thank you for these very kind remarks. This work evolved over 8-10 months as evidenced by the updates to the biorXiv preprint. With unlimited time and foresight, it would probably be best to have separated the 2 themes into separate manuscripts and expanded both. Given current constraints, we plan to make some changes/updates to the present manuscript and hopefully include more in-depth analyses on each theme in future works. Thank you again for the thoughtful reading and critique of both our original manuscript and the revised version.

Minor comments/questions:

"K99 to R00 conversions are increasing in time"

• Assuming I am interpreting the figures correctly, in my opinion, the most important takeaway is that the number of R00 awards have increased, but only for awardees moving to another institution. This key result, best illustrated by panels A and C of Figure 1, is buried in the long paragraph in this section. The organization of content in this section could be improved and more focused. Consider renaming this subsection to be more declarative: "K99 tR00 conversions have increased, but only for awardees moving to another institution."

This is a very concise interpretation of this data. We have edited the paragraph referenced by the reviewer, split it into 2 paragraphs, and changed the title to “K99 awardees increasingly move to other institutions for R00 awards from 2008 to 2022” and the final sentence to “Thus, the number of K99 to R00 conversions is consistent over time, but increasingly more R00 awardees have moved to other institutions since 2013”

• Similarly, I personally found the current title of the subsection, "K99 to R00 conversions are increasing with time" is mildly confusing. An R00 award indicates a successful conversion, so why not simply call this an R00 award instead of saying K99-to-R00 conversion? Also, when I look at Figure 1B and exclude the conversion rates for 2007 and 2008 (because this is a 3 year rolling average), I see that conversion rates (or R00 awards) have remained stagnant. This comment is very much in-the-weeds and is mainly to do with clarity of language.

Thank you for these comments. We had “K99 to R00 conversion” to highlight the unique nature of this award mechanism that a person can only receive an R00 if they previously had a K99 award. Nevertheless, we have edited the text to “R00 awards” and “R00 awardees” to simplify things. We also want to note that we did not compute a 3-year rolling average. The function we used was: (X/(Y -1))x100 where X is the number of R00 awards made in a year and Y is the number of K99 awards made in a year. We did note an error in our calculation in the previous version of the manuscript. Previously, we included all R00 awards and K99 awards for each year from the NIH Reporter dataset; however, this is a flawed methodology. NIH reporter includes only extramural K99 award data and extramural R00 awards, but intramural K99 awardees can receive extramural R00 awards and thus are only included in the R00 dataset. There were 141 R00 awardees in our dataset from NIH Reporter that did not have K99 data, so we assume these are intramural K99 awards since it is required to have a K99 to be eligible for the R00 award. Since we do not know the awarding year for intramural K99 awardees or have data on intramural K99 awardees that fail to activate the R00 award (or stay internal at NIH), we have excluded these 141 R00 awardees. In the previous version, this mis-calculation exaggerated rolling conversion rate (we had correctly calculated the 78% total conversion rate). We re-analyzed our rolling conversion rate and found the average is 81.8% (excluding the first 2 years of the K99 program and the last 2 years).

This is a long explanation, but essentially, we overestimated the number of R00 awards which inadvertently increased the rolling conversion rate. We have corrected this and simplified the first 2 paragraphs of the Results section.

• I was also mildly confused looking at Figure 1c. The caption says that the percentages represent the K99 awardees that stayed at the same institution for the R00 activation, but the percentages are next to the solid circles which the legend labels as "different institution." Perhaps another or different way to show this is a stacked bar chart, where one bar represents the percentage of R00 awards activated at the same institution and another bar represents the percentage of R00 awards activated at a different institution. The bars always add to 100% but the change in proportions illustrates that proportionally fewer awards are being made to those remaining at the same institution.

Great idea. We have included a stacked bar chart here. Since the stacked bar chart is percentages, we felt it was important to also show the total numbers so we still included the previous chart also but removed the percentage numbers from it. We also changed the departmental analysis to stacked bar charts. This shows the stark difference between 2008-2012 and 2013 onward. These changes were made in the revised Fig. 1.

• Minor question: I would love to see Table 3 and Table 4 as a time-series. Has the proportion of recipients at various institution types changed with time?

This is a great suggestion and we felt it fit best in Figure 5, so we’ve added it there.

• Table 3 is useful but only indirectly addresses my first "Recommendation to the Authors" from my previous review. I did some number crunching myself from the data provided. Assuming I did this correctly: If you're a K99 awardee at a private institute, you had a 76.3% change of getting an R00 compared to 80.4% for a K99 awardee at a public institution. If you're a K99 awardee at a top-funded institution, you had a 76.8% chance of R00 compared to 78.6% for a lower-funded institution. I would have liked to see more figures and tables to illustrate conversion rates by institution type in this way. Interestingly, to me, these data suggest that there are not enormous conversion rate differences by institution type (though looking at these now, I am confused at the 89% statistic in line 174 and where that comes form, since it is much higher than what I've calculated).

Thank you for this suggestion and these comments. Please see above where we describe how we incorrectly overestimated the 89% statistic. This has been corrected. As the reviewer suggested, we now show yearly percent of grants to specific institution types in the revised Figure 5. We agree with the reviewer that showing the conversion rate by institution type is interesting; however, it is fairly obvious from the new panels in Figure 5 that there is not much difference in conversion rate. Thus, to avoid crowding too many panels into the figure, we opted to keep the stacked bar plot.

Reviewer #3 (Recommendations for the Authors):

-One minor change to Figure 1C would be to switch the color coding for the lines so that they match with 1D whereby "same institution" would be white circles, or whatever the authors decide would be best for consistency since they are similar comparisons.

Thank you for this suggestion. We have corrected this to be consistent.

-Minor note for lines 459-461: I would suggest changing the wording to "intersectional inequalities" as it is not that a scientist's identities impact their careers as much as how those identities are positioned within an unequal opportunity structure and differentially treated that produce varying career trajectories and experiences of marginalization and cumulative (dis)advantages.

Thank you and we agree with you. We have made this correction.

-To carry forward a suggestion for the authors in my previous review, future research that more fully explores the research infrastructure of institutions for how top NIH funded institutions continue to be top funded institutions year after year could help clarify some of the career mobility and same/similar institution hiring found in the data. Rather than hand coding institutions for some of the infrastructure, the National Center for Education Statistics' Integrated Postsecondary Education Data System (IPEDS) has data on colleges and universities including whether they operate a hospital, have a medical degree, and many other interesting data about student and faculty demographics, institutional expenditures (including research budgets), and degrees awarded in different fields of study (undergrad and grad) that may be helpful to the authors as they continue their research stream in this area.

Thank you very much. We will look into this data set as we continue our investigations in this area.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

The discussion seems to imply that the ball-and-chain peptide is or is related to the common gate. (Although it isn't stated explicitly, it is implied based on the presentation of the gating model in Figure 8 immediately after the discussion of common gating, and the simultaneous opening of both pores in Figure 8). What does the asymmetric structure say about the relationship between the N-term peptide and common gating in ClC-2? It seems like this structure suggests that the CTDs can independently rotate, and independently bind N-terminal peptide, which might not be expected to impact both pores. Some additional clarification and/or discussion of these ideas could be helpful here.

We thank the reviewer for raising these very important points. We agree we should have been more explicit and have now expanded our discussion on this topic, highlighting the independent movement of the N-term peptide and CTDs and clarifying that it is currently unknown whether CLC-2 has a common gate (lines 431484).

Discussion of "Revised Framework for CLC-2 gating": I think this would be a little easier to follow if most of the legend from Figure 8 was in the main text at the end of that section. Also, additional labels in Figure 8 (of the glutamates, the N-terminal peptide, and what the CTD arrows represent).

We have revised this section of the text and added labels to the (revised) Figure as suggested.

Line 261: typo, misspelling of "hydrogen"

Fixed. (Now line 279.)

Figure 6 - supplement 2B: Looks like an error in numbering y-axis - should be 90/120/150, I think. Can you show the three data points for the WT initial current rectification? Can you clarify whether the 3 that you are analyzing are the ones where AK42 the AK42 "zero current" level is not more than the initial positive current?

We apologize for this error, which arose from the Y-axis label overlapping the tick labels, so 90/120/150 showed as 90/20/50. We have fixed this error and have added a new panel (C) to show three data points for the WT initial current rectification. In the Figure legend to panel C, we clarify that the 3 experiments we analyzed are the ones where the AK-42 current level is not more than the initial current at 80 mV.

Reviewer #2 (Recommendations For The Authors):

1. It appears from a close inspection of Figure 2 that the TM dimer is not quite symmetric, but I couldn't tell for sure from the figures as presented. No comment is made in the methods about symmetry imposed, and the authors explicitly comment on asymmetry in the cytoplasmic domain. It would be useful to have an explicit discussion of the TM dimer symmetry.

We have now explicitly stated that the TM dimer is symmetric, and we have clarified the wording in the Methods:

Main text, line 81: "The TM region of CLC-2 displays a typical CLC family symmetric homodimeric structure, with each subunit containing an independent Cl– pathway (Figure 2A, B)."

Methods (lines 557-558): "The following ab initio reconstruction and 3D refinement (for all structures presented in this paper) were performed with C1 symmetry (no symmetry imposed)."

1. For the simulations in Figure 5 Supplement 2, the N terminus flexibility is shown, but this of course can't be compared to a control. However, given the structural results, one might expect the JK helix to show changes in flexibility/mobility in the apo vs inactivated structures. Is this observed?

We agree that the structures strongly suggest the JK-helix is not as stable without the N-terminus bound. We did not perform comparative simulations on the JK helix in the apo vs inactivated structures. While we agree this could be of interest, we don’t think it is essential to our conclusions, and the simulations might need to be quite long to adequately capture dynamics of the JK helix. [In the simulation results shown in Figure 5 Supplement 2, our aim was to test the validity of the structure by determining whether the N-terminus remains bound to the channel in simulations. The plot shows that the N-terminus stays in the same binding pose with an average RMSD (to the initial structure) of less than 2 angstroms, which is generally considered to be relatively stable.]

1. I find the section "revised framework for ClC-2 gating" to be wanting. The ideas are illustrated in the cartoon, but should also be laid out in the text. In what ways are you revising the framework, and in what aspects are you carrying through ideas already proposed?

Thank you for raising this point, which was also raised by Reviewer 1. We have revised this section and the accompanying Figure (Figure 8 and Lines 431-484).

1. The authors mention in passing the idea that the hairpin could contribute to inward rectification (lines 227/8), but also suggest a role for the gating glutamate in this process. They also mention the idea of a common gate, but don't flesh out its function very much. These possibilities are very interesting and should be substantially fleshed out in the "framework" section, even if they cannot be fully answered yet.

We have expanded on these points in the “framework” section.

1. Figure 6E. points representing individual experiments should be shown.

We added points representing individual experiments for Delta N (normalized to WT) in the surface-expression experiments in Figure 6E. Individual data points for the electrophysiology experiments are in panel C; we did not replot these in panel E because some of the points would have been off scale.

1. The density in Figure 2A is hard to see, is there a better way to display it? Also, the orientation of the rightmost panel in Figure 2C is difficult to interpret.

We revised 2A to make the density easier to see. We revised Figure 2C so that the middle and rightmost panels have the same orientation.

1. P6. Line 87. This sentence is a little confusing, and perhaps could be a little clearer-the density is consistent with a Cl- ion, but no experiments have been done to support this, no?

We have clarified the wording as suggested (now line 89) and added references supporting Clˉ binding to the Sext site in CLCs (line 90).

1. P6 lines 89-98. Two lines of evidence, the conformation of the gate and the pinch point, both point to the structure representing a closed state. The wording as presented is a little hard to follow.

We have revised the wording in this paragraph (lines 92-111)

1. It's hard to distinguish water protons and oxygens in the lower right panel (QQQ).

We revised this panel (in Figure 3 – figure supplement 2) to better distinguish the water protons and oxygens.

Reviewer #3 (Recommendations For The Authors):

A few points to consider for improving the manuscript

1. It is intriguing that in the AK-42 structure, there is no density for the hairpin loop even though the CTD is in a symmetrical conformation as the apo. The authors could perhaps comment on whether there is any difference in the rectification properties of currents (or run-up) upon unblocking of AK-42 which may suggest that the hairpin binding is prevented by AK-42.

We have not yet performed the suggested experiment nor any experiments to examine state-dependence, though we agree such experiments would be informative. We have added a note on this point in the discussion, lines 334-337.

1. Although the conformation-dependent placement of the hairpin loop is convincing based on the density, the sequence assigned to this region is not conclusive.

To strengthen our conclusion concerning the hairpin assignment, we investigated fits of peptide segments from the disordered sections of the C-terminal cytoplasmic domain to the hairpin density. We found that these fits are not as good as that with the N-terminal peptide. This analysis is described in lines 179-181 and a new figure (Figure 5 – figure supplement 1). We appreciate the reviewer’s point that it is extremely difficult to conclusively assign residues that are not contiguous with the rest of the structure. Nevertheless, given the wide variety of evidence all pointing to the conclusion that the hairpin loop corresponds to residues 14-28, we think the assignment is on strong footing. We respectfully ask that you consider removing this criticism from the public review, as we think it will hinder the casual reader from recognizing the strength of the evidence: (1) of unresolved regions in CLC-2, residues 14-28 fit best; (2) residues 14-28 were previously identified as part of the ball blocking region (lines 158-161); (3) MD simulations support that the N-terminal residues stay stably bound (Figure 5 – figure supplement 4) (4) gain-of-function disease causing mutations map onto either the Nterminal residues or interacting residues on the TM domain (Figure 5 – figure supplement 6). Thank you for considering this request.

1. The authors should comment on the physiological relevance of the CBS domain rearrangements during gating.

We have added this sentence (lines 131-133): “The physiological relevance of C-terminal domain rearrangements is suggested by disease-causing mutations that alter channel gating (Estevez et al., 2004; Brenes et al., 2023).”

1. For the figures with cryo-EM maps, indicate the contour levels.

Contour levels are now indicated in the Figure legends.

1. It will be useful to the electrostatic map of the N-terminal peptide and the docking site.

This is now shown in Figure 5 – figure supplement 3 and Video 5.

1. Include a comment on the recent CLC-2 /AK-42 structure and if there are any differences in the structural features.

We added this text to lines 273-274: “The RMSD between our CLC2-TM-AK42 structure and that of Ma et al. is 0.655 Å, and the RMSD between the apo TM structures is 0.756 Å.”

Author Response

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

eLife assessment

The paper contains some useful analysis of existing data but there are concerns regarding the conclusion that there might be alternative mechanisms for determining the location of origins of DNA replication in human cells compared to the well known mechanism known from many eukaryotic systems, including yeast, Xenopus, C. elegans and Drosophila. The lack of overlap between binding sites for ORC1 and ORC2, which are known to form a complex in human cells, is a particular concern and points to the evidence for the accurate localization of their binding sites in the genome being incomplete.

Public Reviews:

Reviewer #1 (Public Review):

In the best genetically and biochemically understood model of eukaryotic DNA replication, the budding yeast, Saccharomyces cerevisiae, the genomic locations at which DNA replication initiates are determined by a specific sequence motif. These motifs, or ARS elements, are bound by the origin recognition complex (ORC). ORC is required for loading of the initially inactive MCM helicase during origin licensing in G1. In human cells, ORC does not have a specific sequence binding domain and origin specification is not specified by a defined motif. There have thus been great efforts over many years to try to understand the determinants of DNA replication initiation in human cells using a variety of approaches, which have gradually become more refined over time.

In this manuscript Tian et al. combine data from multiple previous studies using a range of techniques for identifying sites of replication initiation to identify conserved features of replication origins and to examine the relationship between origins and sites of ORC binding in the human genome. The authors identify a) conserved features of replication origins e.g. association with GC-rich sequences, open chromatin, promoters and CTCF binding sites. These associations have already been described in multiple earlier studies. They also examine the relationship of their determined origins and ORC binding sites and conclude that there is no relationship between sites of ORC binding and DNA replication initiation. While the conclusions concerning genomic features of origins are not novel, if true, a clear lack of colocalization of ORC and origins would be a striking finding. However, the majority of the datasets used do not report replication origins, but rather broad zones in which replication origins fire. Rather than refining the localisation of origins, the approach of combining diverse methods that monitor different objects related to DNA replication leads to a base dataset that is highly flawed and cannot support the conclusions that are drawn, as explained in more detail below.

Response: We are using the narrowly defined SNS-seq peaks as the gold standard origins and making sure to focus in on those that fall within the initiation zones defined by other methods. The objective is to make a list of the most reproducible origins. Unlike what the reviewer states, this actually refines the dataset to focus on the SNS origins that have also been reproduced by the other methods in multiple cell lines. We have changed the last box of Fig. 1A to make this clearer: Shared origins = reproducible SNS-seq origins that are contained in initiation zones defined by Repli-seq, OK-seq and Bubble-seq. This and the Fig. 2B (as it is) will make our strategy clearer.

Methods to determine sites at which DNA replication is initiated can be divided into two groups based on the genomic resolution at which they operate. Techniques such as bubble-seq, ok-seq can localise zones of replication initiation in the range ~50kb. Such zones may contain many replication origins. Conversely, techniques such as SNS-seq and ini-seq can localise replication origins down to less than 1kb. Indeed, the application of these different approaches has led to a degree of controversy in the field about whether human replication does indeed initiate at discrete sites (origins), or whether it initiates randomly in large zones with no recurrent sites being used. However, more recent work has shown that elements of both models are correct i.e. there are recurrent and efficient sites of replication initiation in the human genome, but these tend to be clustered and correspond to the demonstrated initiation zones (Guilbaud et al., 2022).

These different scales and methodologies are important when considering the approach of Tian et al. The premise that combining all available data from five techniques will increase accuracy and confidence in identifying the most important origins is flawed for two principal reasons. First, as noted above, of the different techniques combined in this manuscript, only SNS-seq can actually identify origins rather than initiation zones. It is the former that matters when comparing sites of ORC binding with replication origin sites, if a conclusion is to be drawn that the two do not co-localise.

Response: We agree. So the reviewer should agree that our method of finding SNS-seq peaks that fall within initiation zones actually refines the origins to find the most reproducible origins. We are not losing the spatial precision of the SNS-seq peaks.

Second, the authors give equal weight to all datasets. Certainly, in the case of SNS-seq, this is not appropriate. The technique has evolved over the years and some earlier versions have significantly different technical designs that may impact the reliability and/or resolution of the results e.g. in Foulk et al. (Foulk et al., 2015), lambda exonuclease was added to single stranded DNA from a total genomic preparation rather than purified nascent strands), which may lead to significantly different digestion patterns (ie underdigestion). Curiously, the authors do not make the best use of the largest SNS-seq dataset (Akerman et al., 2020) by ignoring these authors separation of core and stochastic origins. By blending all data together any separation of signal and noise is lost. Further, I am surprised that the authors have chosen not to use data and analysis from a recent study that provides subsets of the most highly used and efficient origins in the human genome, at high resolution (Guilbaud et al., 2022).

Response: 1) We are using the data from Akerman et al., 2020: Dataset GSE128477 in Supplemental Table 1. We have now separately examined the core origins defined by the authors to check its overlap with ORC binding (Supplementary Fig. S8b)

2) To take into account the refinement of the SNS-seq methods through the years, we actually included in our study only those SNS-seq studies after 2018, well after the lambda exonuclease method was introduced. Indeed, all 66 of SNS-seq datasets we used were obtained after the lambda exonuclease digestion step. To reiterate, we recognize that there may be many false positives in the individual origin mapping datasets. Our focus is on the True positives, the SNS-seq peaks that have some support from multiple SNS-seq studies AND fall within the initiation zones defined by the independent means of origin mapping (described in Fig. 1A and 2B). These True positives are most likely to be real and reproducible origins and should be expected to be near ORC binding sites.

We have changed the last box of Fig. 1A to make this clearer: Shared origins = reproducible SNS-seq origins that are contained in initiation zones defined by Repli-seq, OK-seq or Bubble-seq.

Ini-seq by Torsten Krude and co-workers (Guillbaud, 2022) does NOT use Lambda exonuclease digestion. So using Ini-seq defined origins is at odds with the suggestion above that we focus only on SNS-seq datasets that use Lambda exonuclease. However, Ini-seq identifies a much smaller subset of SNS-seq origins, so, as requested, we have also done the analysis with just that smaller set of origins, and it does show a better proximity to ORC binding sites, though even then the ORC proximate origins account for only 30% of the Ini-seq2 origins (Supplementary Fig. S8d). Note Ini-seq2 identifies DNA replication initiation sites seen in vitro on isolated nuclei.

Update in response to authors' comments on the original review:

While the authors have clarified their approach to some aspects of their analysis, I believe they and I are just going to have to disagree about the methodology and conclusions of this work. I do not find the authors responses sufficiently compelling to change my mind about the significance of the study or veracity of the conclusions. In my opinion, the method for identification of strong origins is not robust and of insufficient resolution. In addition, the resolution and the overlap of the MCM Chip-seq datasets is poor. While the conclusion of the paper would indeed be striking and surprising if true, I am not at all persuaded that it is based on the presented data.

Reviewer #2 (Public Review):

Tian et al. performed a meta-analysis of 113 genome-wide origin profile datasets in humans to assess the reproducibility of experimental techniques and shared genomics features of origins. Techniques to map DNA replication sites have quickly evolved over the last decade, yet little is known about how these methods fare against each other (pros and cons), nor how consistent their maps are. The authors show that high-confidence origins recapitulate several known features of origins (e.g., correspondence with open chromatin, overlap with transcriptional promoters, CTCF binding sites). However, surprisingly, they find little overlap between ORC/MCM binding sites and origin locations.

Overall, this meta-analysis provides the field with a good assessment of the current state of experimental techniques and their reproducibility, but I am worried about: (a) whether we've learned any new biology from this analysis; (b) how binding sites and origin locations can be so mismatched, in light of numerous studies that suggest otherwise; and (c) some methodological details described below.

• I understand better the inclusion/exclusion logic for the samples. But I'm still not sure about the fragments. As the authors wrote, there is both noise and stochasticity; the former is not important but the latter is essential to include. How can these two be differentiated, and what may be the expected overlap as a function of different stochasticity rates?

It is difficult to separate the effect of noise from the effect of stochastic firing of origins. We therefore took the simplest approach: focus only on the most reproducible origins (shared origins) and ignore the non-reproducible origins. At least the most reproducible origins can be used to test the hypotheses regarding origin firing.

• Many of the major genomic features analyzed have already been found to be associated with origin sites. For example, the correspondence with TSS has been reported before:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6320713/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6547456/

• Line 250: The most surprising finding is that there is little overlap between ORC/MCM binding sites and origin locations. The authors speculate that the overlap between ORC1 and ORC2 could be low because they come from different cell types. Equally concerning is the lack of overlap with MCM. If true, these are potentially major discoveries that butts heads with numerous other studies that have suggested otherwise.

The key missing dataset is ORC1 and ORC2 CHiP-seq from the same cell type. This shouldn't be too expensive to perform, and I hope someone performs this test soon. Without this, I remain on the fence about how much existing datasets are "junk" vs how much the prevailing hypothesis about replication needs to be revisited. Nonetheless, the authors do perform a nice analysis showing that existing techniques should be carefully used and interpreted.

We agree that a thorough set of ChIP-seq data (with multiple antibodies or with equivalent techniques that do not use antibodies) for all six subunits of ORC in mammalian cells will be very useful for the field. Note, though, that just by simple cell lysis, it is very easy to divide human ORC into at least three different parts: ORC1, ORC2-5, and ORC6. The subunits do not form as robust a complex as seen in the yeasts and in flies.

Reviewer #3 (Public Review):

Summary: The authors present a thought-provoking and comprehensive re-analysis of previously published human cell genomics data that seeks to understand the relationship between the sites where the Origin Recognition Complex (ORC) binds chromatin, where the replicative helicase (Mcm2-7) is loaded, and where DNA replication actually beings (origins). The view that these should coincide is influenced by studies in yeast where ORC binds site-specifically to dedicated nucleosome-free origins where Mcm2-7 can be loaded and remains stably positioned for subsequent replication initiation. However, this is most certainly not the case in metazoans where it has already been reported that chromatin bindings sites of ORC and Mcm2-7 do not necessarily overlap, nor do they always overlap with origins. This is likely due to Mcm2-7 possessing linear mobility on DNA (i.e., it can slide) such that other chromatin-contextualized processes can displace it from the site in which it was originally loaded. Additionally, Mcm2-7 is loaded in excess and thus only a fraction of Mcm2-7 would be predicted to coincide with replication start sites. This study reaches a very similar conclusion of these previous studies: they find a high degree of discordance between ORC, Mcm2-7, and origin positions in human cells.

Strengths: The strength of this work is its comprehensive and unbiased analysis of all relevant genomics datasets. To my knowledge, this is the first attempt to integrate these observations. It also is an important cautionary tale to not confuse replication factor binding sites with the genomic loci where replication actually begins, although this point is already widely appreciated in the field. Response: Thank you for recognizing the comprehensive and unbiased nature of our analysis. Our findings will prevent the unwise adoption of ORC or MCM binding sites as surrogate markers of origins and will stimulate the field to try and improve methods of identifying ORC or MCM binding until the binding sites are found to be proximal to the most reproducible origins. The last possibility is that there are ORC- or MCM-independent modes of defining origins, but we have no evidence of that.

Weaknesses: The major weakness of this paper is the lack of novel biological insight and that the comprehensive approach taken failed to provide any additional mechanistic insight regarding how and why ORC, Mcm2-7, and origin sites are selected or why they may not coincide.

Response: we agree that we cannot provide a novel biological insight from this kind of meta-analysis. The importance of this study is in highlighting that there is either significant problems with the data collected till now (preventing the co-localization of ORC or MCM binding sites with the most reproducible origins) or ORC and MCM binding sites are often far away from where the most reproducible origins fire, which should make us consider ways in which origins could be activated kilobases away from ORC and MCM binding sites.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

All suggestions and recommendations were described in a previous review.

Reviewer #3 (Recommendations For The Authors):

The most significant omission is a contextualization of the results in the discussion and an explanation of why these results matter for the biology of replication, disease, and/or our confidence in the genomic techniques reported on in this study. As written, the discussion simply restates the results without any interpretation towards novel insight. I suggest that the authors revise their discussion to fill this important gap.

A second important, unresolved point is whether replication origins identified by the various methods differ due to technical reasons or because different cell types were analyzed. Given the correlation between TSS and origins (reported in this study but many others too), it is somewhat expected that origins will differ between cell types as each will have a distinct transcriptional program. This critique is partly addressed in Figure S1C. However, given the conclusion that the techniques are only rarely in agreement (only 0.27% origins reproducibly detected by the four techniques), a more in-depth analysis of cell type specific data is warranted. Specifically, I would suggest that cell type-specific data be reported wherever origins have been defined by at least two methods in the same cell type, specifically reporting the percent of shared origins amongst the datasets. This type of analysis may also inform on whether one or more techniques produces the highest (or lowest) quality list of true origins.

We have done what has been suggested: used K562 cell type-specific data because here the origins have been defined by at least two methods in the same cell type and reported the percent of shared origins amongst the datasets (Supp. Fig. S4).

Other MINOR comments include:

• Line 215: the authors show that shared origins overlap with TF binding hotspots more often than union origins, which they claim suggests "that they are more likely to interact with transcription factors." As written, it sounds like the authors are proposing that ORC may have some direct physical interaction with transcription factors. Is this intended? If so, what support is there for this claim?

The reviewer is correct. We have rephrased because we have no experimental support for this claim.

• In the text, Figure 3G is discussed before Figure 3F. I suggest switching the order of these panels in Figure 3.

Done.

• It's not clear what Figure 5H to Figure 6 accomplishes. What specifically is added to the story by including these data? Is there something unique about the high confidence origins? If there is nothing noteworthy, I would suggest removing these data.

We want to keep them to highlight the small number of origins that meet the hypothesis that ORC and MCM must bind at or near reproducible origins. These would be the origins that the field can focus in on for testing the hypothesis rigorously. They also show the danger of evaluating proximity between ORC or MCM binding sites with origins based on a few browser shots. If we only showed this figure, we could conclude that ORC and MCM binding sites are very close to reproducible origins.

• Line 394: "Since ORC is an early factor for initiating DNA replication, we expected that shared human origins will be proximate to the reproducible ORC binding sites." This is only expected if one disbelieves the prior literature that shows that ORC and origins are not, in many cases, proximal. This statement should be revised, or the previous literature should be cited, and an explanation provided about why this prior work may have missed the mark.

We do not know of any genome-wide study in mammalian cell lines where ORC binding sites and MCM binding have been compared to highly reproducible origins, or that show that these binding sites and highly reproducible origins are mostly not proximal to each other. Most studies cherry pick a few origins and show by ChIP-PCR that ORC and/or MCM bind near those sites. Alternatively, studies sometimes show a selected browser shot, without a quantitative measure of the overlap genome wide and without doing a permutation test to determine if the observed overlap or proximity is higher than what would be expected at random with similar numbers of sites of similar lengths. In the revised manuscript we have discussed Dellino, 2013; Kirstein, 2021; Wang, 2017; Mas, 2023. None of them have addressed what we are addressing, is the small subset of the most reproducible origins proximal to ORC or MCM binding sites?

• Line 402-404: given the lack of agreement between ORC binding sites and origins the authors suggest as an explanation that "MCM2-7 loaded at the ORC binding sites move much further away to initiate origins far from the ORC binding sites, or that there are as yet unexplored mechanisms of origin specification in human cancer cells". The first part of this statement has been shown to be true (Mcm2-7 movement) and should be cited. But what do the authors mean by the second suggestion of "unexplored mechanisms"? Please expand.

We have addressed this point in the revised manuscript.

• The authors should better reference and discuss the previous literature that relates to their work, some of these include Gros et al., 2015 Mol Cell, Powell et al., 2015 EMBO J, Miotto et al., 2016 PNAS, but likely there are many others.

We have addressed this point in the revised manuscript.

Note for authors:

Line 107: The introduction discusses the mechanism for yeast ORC recognizes specific origins and discusses the Orc4 contribution, but it is known that Orc2 also binds DNA on a base-specific manner (see PMID 33056978). Thus Lee et al. did not "humanize ORC" as stated.

Done

Lines 117-119: Two of the cited papers are on endo-reduplication and not on initiation in a normal cell cycle and this should be pointed out. Second, there is contradictory evidence that ORC is essential in human cells and this should be cited (PMID 33522487)

Done

Author Response

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

Based on the reviewer comments (see below) and subsequent discussion between the reviewers and the Reviewing Editor, I would like to invite the authors to make major revisions, including new experiments. However, if major new experiments are not feasible, as may be the case, then at a minimum, I would urge the authors to:

1. Tone down the language regarding a causative role for changes in GH/IGF-I signaling in mediating the effects of Tmem63 on the skeleton, and also be very open in acknowledging the lack of mechanistic insight into how Tmem regulates GH signaling.

Response: We toned down the language as suggested and also acknowledged the lack of mechanistic insights into how Tmem263 regulates GH signaling.

1. Revise/redo or if not possible, then delete the problematic experiment in Fig. 5E.

Response: We have included additional Western blot data in Figure 5 from control WT and KO male mice without exogenous GH injection. In the absence of GH injection, we could not detect Jak2 and Stat5 phosphorylation in the liver of male WT and KO mice.

1. Address the comments about liver feminization.

Response: We have performed additional analysis as suggested by reviewer # 3. We have now included additional data to address the issue of liver feminization (new Fig. 6G-I and Figure 6-figure supplement 1). We plan to expand on this very topic in future studies as this is an interesting transcriptional phenomenon.

1. Revise the manuscript to address as many of the recommendations for the authors as possible, many of which can be addressed by textual edits. Response: We have addressed as many of the textual changes as suggested in the revised manuscript.

Reviewer #2 (Recommendations for The Authors):

TMEM263 has been suggested to be associated with bone mineral density and growth in humans and mice, but the functional role of this transmembrane protein in the regulation of bone metabolism is unknown. With the knockout mouse approach, this manuscript demonstrates that Tmem263 is essential for longitudinal bone growth in the mouse as deletion of Tmem263 in knockout (KO) mice developed severe postnatal growth impairment and proportional dwarfism. It is determined that the dwarfism was caused by a substantial reduction in liver expression of growth hormone receptor (GHR), a slight increase in serum GH, and a reduction in serum IGF-I, which resulted in disruptive of GH/IGF-I regulatory axis of endochondral bone formation.

The study was relatively well designed, and the results in general are supportive of the conclusions. While this study discloses new and intriguing functional information about a novel cytoplasmic membrane gene, there are a few minor issues that the authors may wish to address. These issues are listed in the following:

1. One of the intriguing findings of this manuscript is that deletion of a gene encoding a small cytoplasmic membrane protein could cause a substantial reduction in the expression and protein levels of GHR. Inasmuch as a couple of potential explanations were offered in the Discussion section (first complete paragraph of page 10), there has been no attempt to test any of the suggested causes, since many of these potential mechanisms can readily be tested experimentally. Accordingly, the lack of mechanistic investigation into this intriguing effect renders the manuscript largely descriptive in nature.

Response: The point made by the reviewer is well taken. We do plan to have follow up studies to establish which among the mechanisms we highlighted in the discussion is contributing to the reduction in GHR transcript and protein level. Our present study is the first functional characterization of this enigmatic novel membrane protein. We anticipate that multiple follow-up studies are needed to gain a deeper understanding of the biology of Tmem263. We believe that our present study represents an important first step.

1. Because a major conclusion is that the bone phenotype of Tmem263 KO mice was caused by deficient hepatic expression and/or action of GHR, it would be helpful to (or strengthen) the conclusion if a brief comparison of the bone phenotype between GHR KO mice and Tmem263 KO mice is included in the Discussion section.

Response: We have now included this information in the revised manuscript.

1. In Figure 3, the cortical bone parameters (i.e., Tt.Ar, Ct.Ar, and Ct.Th), but none of the trabecular bone parameters (i.e., BV/TV, Tb.N, Tb.Th), were normalized against femur length. The authors did not provide a rationale for this differential treatment with the cortical bone parameters from the trabecular bone parameters. If the reason to normalize the cortical bone parameters against bone length was to demonstrate that the reduced cortical bone mass in mutants was related to the impaired longitudinal bone growth, then why did the authors not also assess whether the observed reduction in these trabecular bone parameters in KO mutants was proportional to reduced longitudinal bone growth?

Response: We actually made the exact adjustments that the reviewer refers to, as stated in the methods section. Please see page 14. The regions of interest (ROIs) of both the trabecular bone analysis and the cortical analysis in the mutants was reduced proportional to the length of the bone (40% smaller). The normalization to Tt Ar to femur length in Figure 3I was only meant to show that the reduction in Tt Ar in the mutants was proportional. We have modified the text in our result section for clarity.

1. Elements described in Fig. 5A have been well documented. Therefore, Fig. 5A is unnecessary and can be deleted.

Response: We felt that Figure 5A should remain. It helps orient readers that are not familiar with the literature to be aware that both liver- and bone-derived IGF-1 contribute to longitudinal bone growth.

1. Figure 6 was performed with male KO mice. Were the altered gene expression profiles in female KO mice any different from male KO mice?

Response: We plan to perform RNA-seq in female mouse liver in our follow-up studies. We do not know, at present, whether and to what extent the liver transcriptomic profile would be different between male and female KO mice. As far as dwarfism and deficiency in skeletal acquisition, both male and female KO mice showed the same phenotypes.

1. The number of animals (or samples) per group in some of the Figures (i.e., Fig. 2G, 2I, 2J, 3A to J, the entire Fig. 4, 5D, 5F, and Suppl Fig. 1) is needed to be provided in the legends.

Response: We have included this information in the figure legends.

Reviewer #3 (Recommendations for The Authors):

1. Explain the discrepancy between the impact of KO on serum Igfbp3 (= decreased) vs. hepatic Igfbp3 (= unchanged).

Response: We do not have a plausible mechanism, at present, that can explain the reduction in circulating serum Igfbp3 level without an apparent reduction in Igfbp3 transcript level in the liver. In human studies, typically only serum IGFBP3 levels are measured but not the hepatic IGFBP3 transcript level. Therefore, it is unclear whether the circulating levels of IGFBP3 is being regulated at the posttranscriptional level, an issue that can be explored in future studies.

1. Line 215, 221, and elsewhere - Foxa1 does not show significant male-biased expression in mouse liver.

Response: We have removed Foxa1 from the text.

1. Line 225- According to the abstract of Ref. #45, Cux2 regulates a subset of sex-biased genes in the liver. The authors should compare the genes dysregulated by TMEM263-KO (Fig. 6) to those altered by Cux2 loss (Ref. #45) to ascertain whether the results of Fig. 6 are partially or entirely explained by Cux2 overexpression.

Response: We agree that this is a great area of future study. We do feel this, however, would be better explored in a more in-depth follow-up article. We felt, given the current direction of the paper it made more sense to include differential expression comparisons of male vs female, hypophysectomized vs sham control, and Stat5b-KO vs WT mouse liver gene expression data. Our future work will explore the transcriptomes of male and female WT and Tmem263-KO liver gene expression in the context of the observed physiology.

1. Line 262- "lower transcription of Ghr gene". A decrease in mRNA levels does NOT equate with a decrease in transcription per se. Altered mRNA splicing, poly A, export, cytoplasmic stability, etc. are all potential contributors.

Response: We have included these possibilities highlighted by the reviewer in our revised Discussion section.

1. Line 273, "TMEM263... most highly expressed in liver" Not correct - see Fig. 1C for TMEM263 RNA levels in mouse tissues.

Response: We have corrected the text on page 11.

1. Line 425 - Include GEO accession number.

Response: We have already uploaded our RNA-seq data to the NCBI Sequence Read Archive (SRA), and the data can be accessed under accession number # PRJNA938158.

1. Fig. 6 - Line 796 - Specify the age and sex of mice analyzed.

Response: We have included the information in the revised figure 6 legend.

1. Fig.2 - Suppl 1- Specify age of mice.

Response: We have included the information in the revised Figure 2-figure supplement 2.

1. Fig.2G -Specify the sex of the mice.

Response: For the P1 to P21 pups’ data, we did not separate by sex, as gender determination of pups at P1 and P7 can be challenging. We now indicated this in the figure legend.

1. Fig. 6A and 6C-6F: Which of these genes shows sex-dependent expression in wild-type liver? Use color to highlight gene names for genes that show male-biased or female-biased expression.

Response: We agree with the reviewer that additional labels on Figure 6A and 6C-F would be helpful to show genes of sex-bias. However, this is not the primary point of the paper. This topic deserves a much more in-depth analysis in follow up studies focused on defining the exact type and degree of transcript feminization in the liver of Tmem263-KO mice, as well as, its physiologic consequences. For readers interested in this topic, we have included the subfigures G-I in Figure 6 and for greater transcript level detail, figure 6 supplement 1.

Author Response

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

Reviewer #1

Recommendation 1: The authors reasoned upon the presence of a differential basal hydraulic stress in waves' valleys vs hills at first from the observation of "domes" formation upon 48h cultivation. I suggest performing a quantification to support the statement as a good scientific practice. Furthermore, it would strengthen the concept when the formation of domes was compared between the waves' dimensions as a different grade of cell extrusion was quantified. i.e., 50, 100, and 200 µm.

Response 1: Upon seeing the phenomenon (Author response image 1 A), we performed a count for domes on the 100 µm and saw a significant effect. We refrained from including the results as it is the subject of ongoing research in our lab. In response to the reviewer’s suggestion, we have included a graph (Author response image 1 B) showing the increasing number of domes over 48 hours from three 100 µm wave samples.

We have updated Figure 2A and B in the manuscript to include the new graph.

Author response image 1.

(A) shows dome (white arrows) over a 100 µm wave substrate. (B) is the number of accumulated domes in valley and hill regions, for 3 independent samples, over 48 hours.

Recommendation 2: Using RICM microscopy to quantify the cell basal separation with the substrate and hydraulic stress is very clever. Nevertheless, I am in doubt if the different intensity reported for the hills vs valley (Fig. 2G and H) is a result of the signal reduction at deeper Z levels. Since there is no difference in extrusion and forces between valleys and hills in the 200 µm waves but only in 50µm and 100µm, I would add this to the quantification. I would expect no intensity difference from RICM for the 200 µm sample if this is not an artefact of imaging.

Response 2: We performed additional experiments on blank wave substrates (both 100 and 200 µm) to ascertain the extent of reflection intensity drop (Author response image 2A). And, as correctly pointed out by Reviewer #1, there was a drop in intensity even without cells. On the 100 µm waves, hill reflections are on average ~27 % dimmer than valley reflections. Whereas, on the 200 µm waves, hill reflections are on average ~39 % dimmer.

Using this information, we performed a calibration on the RICM results obtained from both the 100 and 200 µm waves (Author response image 3B). The calibrated 100 µm data showed residual signatures of difference, whereas the calibrated 200 µm distributions appeared very similar. We noticed large cross- sample variations in the registered intensities, which will negatively impact effect size if not accounted for. To do this, we subsequently normalized both hill and valley intensities against planar region intensities for each sample. As shown by the final output (Author response image 3C), we were able to remove the skewness in the distributions. Moreover, 1-way ANOVA followed by a post hoc analysis with BH correction revealed a significant reduction in 100 µm hill/flat intensity ratio compared to 100 µm valley/flat intensity ratios (Δ~-23 %). Conversely, no significance was observed for the same comparison on the 200 µm waves.

Author response image 2.

(A). RICM from blank wave samples reveal a reduction in reflection intensity in hill regions compared to flat and valley regions.

Author response image 3.

(B) shows the RICM intensities after adjusting for the inherent reflection intensity drop shown in (A). (C) show the RICM intensities after normalization against planar region signals; this removes cross-sample variations and improve effect size of differences.

We have updated the manuscript Figure 2I and text accordingly. The blank wave results are included in Figure 2-figure supplement 1 along with updated text and summary data table in Supplementary File 4.

Recommendation 3: To measure 3D forces on top of the hills and valleys, the use of PAA gels is necessary. Since in Fig 3B, the authors show a difference in cell extrusion number between substrates and stiffnesses, I think it is necessary to confirm the presence of more extrusion in valleys vs hills on PAA gels. This would ensure the conclusion between normal forces and extrusion.

Response 3: We do have time-lapse data with monolayers on the PAA waves. However, we felt results from the flat regions were sufficient in supporting the point being made in the text. Specifically, our original intention with PAA gels was to show that the extrusion reductions seen in osmotic perturbations were by virtue of removing basal stress and not some cryptic osmotic response. Hydrogels were chosen because they can effectively dilute basal solute concentration and thereby reduce the osmotically induced water transport. Moreover, as fluid could freely move within the gel, the fluid stress can quickly equilibrate across the basal surface. In contrast, poorly water/solute permeable substrates could lead to localized spikes in solute concentration and transient basal regions with high fluid stress.

To get a sense of the potential difference in basal solute concentration between the two materials, we can do a quick hand-waving estimation. For monolayers on non-water/solute permeable PDMS of 20x20 mm and using the laser wavelength (640 nm) for RICM as an extreme estimate of basal separation, we should expect ~0.25 µl of total basal water content. On the other hand, we typically produce our PAM gel slabs using ~150 µl of precursor solutions. This means that, given similar amounts of solute, PAM gels will lead to monolayer basal osmolarity that is around 3 orders of magnitude lower than monolayers on PDMS, producing significantly lower osmotic potential. This implies from the outset that we should expect high survivability of cells on these substrates irrespective of curvature domains. Indeed, later immunoblotting experiments showed MDCKs exhibiting hyper activated FAK and Akt on PAM gels.

In response to Reviewer #1’s suggestion then, we have added another supporting time-lapse (Video 19) showing typical response of MDCK monolayers on 100 µm PAA waves (Author response image 4). Evident from the time-lapses, like the planar regions, cell extrusions were very rare. This supports the idea that on PAM gels the effects of basal hydraulic stress and asymmetric forces are marginal against the strong survival signals. And the response is similar to hyper-osmotic perturbations; there, we did not see a significant difference between valley and hill extrusions.

Author response image 4.

Time-lapse snapshot showing negligible MDCK extrusions 24 hours after confluency over PAM gel wave substrates.

Recommendation 4: Before proceeding with the FAK inhibitor experiment, the authors should better justify why the 4.1 wt % sucrose vs DMSO or NaCl is the most inert treatment. This can be done by citing relevant papers or showing time-lapses (as it is done for the higher FAKI14 dose).

Response 4: Although some cells have recently been shown to be able to transport and utilize sucrose, mammalian cells generally cannot directly take up polysaccharides for metabolism and this is frequently mentioned in literature: see (Ref. R1) for example. Without special enzymes to break sucrose down into monosaccharides, such as sucrase found in the gut, the sugars should remain spectators in the culture medium, contributing only to osmotic effects.

DMSO on the other hand, besides changing osmolarity, can also be integrated into cell membrane and pass through cells over time. It has been reported to chronically affect cell membrane properties and gene expressions (Ref. R2).

Finally, it is well known that both sodium and chloride ions are readily taken up and transported by cells (Ref R3). They help to regulate the transmembrane potential, which in turn can affect membrane bound proteins and biochemical reactions within a cell.

Hence, comparing the 3 hyper-osmotic perturbations, adding sucrose should have the least off- target effects on both the inhibitor study and the subsequent immunoblotting. And, in response to the reviewer’s recommendation, we have updated the text accordingly and included new references to support our statement.

Ref R1. H. Meyer, O. Vitavska, H. Wieczorek; Identification of an animal sucrose transporter. Journal of Cell Science 124, 1984–1991 (2011). Doi: 10.1242/jcs.082024

Ref R2. B. Gironi, Z. Kahveci, B. McGill, B.-D. Lechner, S. Pagliara, J. Metz, A. Morresi, F. Palombo, P. Sassi, P. G. Petrov; Effect of DMSO on the Mechanical and Structural Properties of Model and Biological Membranes. Biophysical Journal 119, 274-286 (2020). Doi: doi.org/10.1016/j.bpj.2020.05.037

Ref R3. X. Zhang, H. Li; Interplay between the electrostatic membrane potential and conformational changes in membrane proteins. Protein Science 28, 502-512 (2019). Doi: 10.1002/pro.3563

Recommendation 5: The data showing a FAK-dependent phosphorylation of AKT responsible for a higher cell survival rate in the hills is not yet completely convincing. Please show a reduced AKT phosphorylation level after FAK inhibition in high osmolarity levels. Furthermore, the levels of AKT activation seem to increase slightly upon substrate softening independently of FAK activation or osmotic pressure (i.e., Fig. 4E, Soft PDMS). The authors should comment on this in connection with the results shown for PAA gels.

Response 5: For the additional immunoblotting experiments, work is currently underway. We could not, however, complete these experiments in time for this revision, as both Cheng-Kuang and Xianbin will shortly be taking on new jobs elsewhere. David will continue with the immunoblotting studies and should be able to include the results in an update in the coming months. As for the apparent elevated levels of AKT seen on soft silicones, we speculate that it is because we cannot immunoblot cells that have died and were inevitably washed out at the start of the procedure. Inferring from the higher extrusion rates on these soft substrates, we could be missing a significant portion of stats. Specifically, we are missing all the cells that would have lowered AKT activation but died, and had we been able to collect those statistics, perhaps both the FAK and AKT should have shown lower levels. We risk committing survival bias on the results if we read too much into the data as is.

Alternatively, another explanation could be that, by virtue of survival of the fittest, we might have effectively selected a subpopulation of cells that were able to survive on lower FAK signals, or completely irrespectively of it.

At any rate, to prove our foregoing hypothesis would require us to perform comprehensive immunoblotting and total transcriptome analysis over different duration conditions. Unfortunately, we do not have the time to do that for the current article, but it could be developed into a stand-alone molecular biology investigation in future. We have included similar discussion in the main text.

Recommendation 6: In the discussion, the authors suggest the reported findings be especially relevant for epithelia that significantly separate compartments and regulate water and soluble transport. These are for example kidney epithelia (i.e., MDCK is the best experimental choice), retinal epithelium or intestinal epithelium. I would suggest that some proof-of-concept experiments could be done to support this concept. For example, I would expect keratinocytes (i.e., HaCaT) not to show a strong difference in extrusion rate between valleys and hills since the monolayer is not so sealed as kidney epithelium. In general, this kind of experiment would significantly strengthen the finding of this work.

Response 6: As recommended, we tracked the behavior of retina pigment epithelial cells (hTERT RPE-1 from ATCC) which do not form tight monolayers like MDCKs (Ref. R4). We did not detect extrusion events occurring from monolayers of these cells (Author response image 5). This is true even for portions of monolayers over waved regions.

Author response image 5.

Time-lapse snapshot showing non-existent o cell extrusions from RPE monolayers confluent for over 21 hours.

We have updated these findings in the main text discussions and included a new supporting time- lapse (Video 15) in our article.

Ref R4 F. Liu, T. Xu, S. Peng, R. A. Adelman, L. I. Rizzolo; Claudins regulate gene and protein expression of the retinal pigment epithelium independent of their association with tight junctions. Experimental Eye Research 198, 108157 (2020). Doi: 10.1016/j.exer.2020.108157

Recommendation 7 (minor point): Figure S1 needs to have clear notes indicating in each step what is what. i.e., where is glass, PDMS, NOA73, etc? A more detailed caption will help the figure's comprehension. Also "Cy52" should be changed to "soft silicone" to be consistent with the text (or Cy52 should be mentioned in the text).

Response 7 (minor point): Changes were made to Figure 1-figure supplement 1 to improve comprehension accordingly. CY52 was added to the main-text, next to the first appearance of the word soft silicone, to be consistent with the figures.

Recommendation 8 (minor point): The authors often mentioned that epithelial monolayers are denser on PAA gels. Please add a reference(s) to this statement.

Response 8 (minor point): The statement is an inference from visually comparing monolayers on PAM gels and PDMS. The difference is quite evident (Author response image 6). The density difference is in spite of the fact that the substrates share similar starting cell numbers.

To address the reviewer’s comment, we have combined time-lapses of monolayers on silicones and PAM gels side-by-side in Video 17 to facilitate convenient comparisons.

Author response image 6.

Time-lapse snapshot at 24 hours after confluence, showing conspicuously higher density of MDCK monolayers on PAM gel compared to those on silicon elastomer.

Reviewer #2

Recommendation 1: The sinusoidal wavy substrate that the authors use in their investigation is interesting and relevant, but it is important to realize that this is a single-curved surface (also known as a developable surface). This means that the Gaussian curvature is zero and that monolayers need to undergo (almost) no stretching to conform to the curvature. The authors should at least discuss other curved surfaces as an option for future research, and highlight how the observations might change. Convex and concave hemispherical surfaces, for example, might induce stronger differences than observed on the sinusoidal substrates, due to potentially higher vertical resultant forces that the monolayer would experience. The authors could discuss this geometry aspect more in their manuscript and potentially link it to some other papers exploring cell-curvature interactions in more complex environments (e.g. non-zero Gaussian curvature).

Response 1: In response to reviewer #2’s recommendation we have highlighted in the discussion of our text that our waves constitute a developable surface and that cells will experience little stretching for the most part. Based on our knowledge of how curvature can modulate forces and thus osmotic effects, we included some rudimentary analysis of what one would expect on hemispherical surfaces of two types: one that is periodic and contiguous (Ref. R5), and another with delineating flat regions (Ref. R6).

For epithelial monolayers in the first scenario, and on poorly solute/water permeable substrates, we should also expect to see a relatively higher likelihood of extrusions from concave regions compared to convex ones. Moreover, as the surfaces are now curved in both principal directions (producing larger out-of-plane forces), we should see the onset of differential extrusions seen in this study, but at larger length scales. For example, the effects seen on 100 µm hemicylindrical waves might now happen at larger feature size for hemispherical waves. Furthermore, as this kind of surface would invariably contain hyperbolic regions (saddle points), we might expect an intermediate response from these locations. If the forces in both principal directions offset each other, the extrusion response may parallel planar regions. On the other hand, if one dominates over the other, we may see extrusion responses tending to the dominating curvature (concave of convex).

On the other hand, on curved landscapes with discrete convex or concave regions, we should expect, within the curved surface, extrusion behaviors paralleling findings in this study. What would be interesting would be to see what happens at the rims (or skirt regions) of the features. At these locations we effectively have hyperbolically curved surfaces, and like before, we should expect some sort of competing effect between the forces generated from the principal directions. So, for dome skirts, we should see fewer extrusions when the domes are small, and vice versa, when they are larger. Meanwhile, for pit rims, we should see a reversed behavior. It should also be noted that the transitioning curvature between convex/concave and planar regions would also modulate the effect.

These effects might have interesting developmental implications. For instance, in developing pillar like tissues (e.g., villi) structures, the strong curvatures of nascent lumps would favor accumulation of cell numbers. However, once the size of the lumps reaches some critical value, epithelial cell extrusions might begin to appear at the roots of the developing structures, offsetting cell division, and eventually halting growth.

Ref R5. L. Pieuchot, J. Marteau, A. Guignandon, T. Dos Santos, I. Brigaud, P. Chauvy, T. Cloatre, A. Ponche, T. Petithory, P. Rougerie, M. Vassaux, J. Milan, N. T. Wakhloo, A. Spangenberg, M. Bigerelle, K. Anselme, Curvotaxis directs cell migration through cell-scale curvature landscapes. Nature Communications 9, 3995 (2018). Doi: 10.1038/s41467-018-06494-6

Ref R6. M. Werner, S. B.G. Blanquer, S. P. Haimi, G. Korus, J. W. C. Dunlop, G. N. Duda, D. W. Grijpma, A. Petersen, Surface curvature differentially regulates stem cell migration and differentiation via altered attachment morphology and nuclear deformation. Advanced Science 4, 1–11 (2017). Doi: 10.1002/advs.201600347

Recommendation 2: The discussion of the experiments on PAM gels is rather limited. The authors describe that cells on the PAM gels experience fewer extrusions than on the PDMS substrates, but this is not discussed in sufficient detail (e.g. why is this the case). Additionally, the description of the 3D traction force microscopy and its validation is quite limited and should be extended to provide more convincing evidence that the measured force differences are not an artefact of the undulations of the surface.

Response 2: We first saw a significant reduction in cell extrusions when we performed hyper-osmotic perturbations, and to eliminate possible off-target effects of the compounds used to increase osmolarity, we used three different compounds to be sure. In spite of this, we felt it would further support our argument, that basal accumulation of fluid stress was responsible for the extrusions, if we had some other independent means of removing fluid stress without directly tuning osmolarity through addition of extraneous solutes. We hence thought of culturing MDCK monolayers on hydrogels.

Hydrogels were chosen because they can effectively dilute basal solute concentration (for reference ions (Na+) are continuously pumped out basally by the monolayer) and thereby reduce the associated osmotically induced water transport. Moreover, as fluid could freely move within the gel, the fluid stress can quickly equilibrate across the basal surface. In contrast, poorly water/solute permeable substrates will lead to localized spikes in solute concentration and transient basal regions with high fluid stress.

To get a sense of the extent of difference in basal solute concentration between the two materials, we can do a quick hand-waving estimation. For monolayers on non-water-permeable PDMS of 20x20 mm, and using the laser wavelength (640 nm) for RICM as an extreme estimate of basal separation, we should expect ~0.25 µl of total basal water content. On the other hand, we typically produce our PAM gel slabs using ~150 µl of precursor solutions. This means that, given similar amounts of solute, PAM gels will lead to monolayer basal osmolarity that is around 3 orders of magnitude lower than monolayers on PDMS, producing significantly lower osmotic potential. This implies from the outset that we should expect high survivability of cells on these substrates. Indeed, later immunoblotting experiments showed MDCKs exhibiting hyper activated FAK and Akt on PAM gels.

As for the 3D TFM used in this study, it is actually implemented from a well-established finite element method to solve inverse problems in engineering and has been repeatedly validated in larger scale engineering contexts (Ref. R7). The novelty and contribution of our article is in its adaptation to reconstruct cellular forces at microscopic scales.

In brief, soft materials, such as hydrogels used in our case, are doped with fluorescent particles, coated with ECM, and then seeded with cells. The cells would exert forces that deform the soft substrate, thereby displacing the fluorescent particles from their equilibrium positions. This particle displacement can be extracted by producing an image pair with microscopy; first one with the cells, and subsequent one of relaxed gel after removal of cells with acutely cytotoxic reagents, such as SDS. There are several ways in which the displacement field can be extracted from the image pair. These include particle tracking velocimetry, particle image velocimetry, digital volume correlation, and optical flow.

We employed 3D Farneback optical flow in our study for its superior computational performance. The method was validated using synthetically generated images from Sample 14 of the Society for Experimental Mechanics DIC challenge. The accuracy of the calculated displacements using the 3D Farneback optical flow was then compared to the provided ground truth displacements. For the highest frequency displacement image pairs, an x-component root-mean-square-error (RMSE) value of 0.0113 was observed. This was lower than the 0.0141 RMSE value for the Augmented Lagrangian Digital Volume Correlation method. This suggested that the 3D Farneback optical flow is capable of accurately calculating the displacement between two bead images.

The displacement fields are then fed into a finite element suite (ANSYS in our case) along with the model and mesh of the underlying substrate structure to obtain node specific displacements. This is required because mech nodes do not typically align with voxel positions of displacements. With these node specific displacements, we subsequently solve the inverse problem for the forces using Tikhonov regularization (Ref. R8). The outcome is a vector of node specific forces.

In light of the above, to physically validate the method in our context would require the generation of a known ground truth force on the scale of pico- to nano-newtons and subsequently image the particle displacements from this force using confocal microscopy. The force must then be released in situ in order for the relaxed gel to be imaged again. This is not a straightforward feat at this scale, and a method that immediately springs to mind is magnetic tweezers. Unfortunately, this is a tool that we cannot develop within reasonable timeframes, as the method will have to be seamlessly integrated with our spinning-disk confocal. However, as a compromise, we have included an in-silico validation with our revised manuscript.

Specifically, given a finite element model with a predefined curvature, a known force was applied to the surface of the model (Author response image 7A). The resulting displacements were then calculated from the finite element solution. A 10% random noise is then added to the resulting displacement. The traction force recovery (Fig. R2-1 B) was then performed using the in-silico noisy displacements. To evaluate the accuracy of the recovery, the cosine similarity along with the mean norm of the force vectors were calculated. A value closer to 1 for both evaluation metrics indicates a more accurate reconstruction of the simulated traction force. The cosine similarity of the recovered traction forces to the original applied force was 0.977±0.056 while the norm of the recovered traction forces as a proportion of the original applied force was 1.016±0.165. As both values are close to 1 (i.e., identical), this suggested that the traction forces could be satisfactorily recovered using the finite-element based method.

In response to the reviewer’s recommendations then, additional content has been included in the main text to explain the use of PAM gels and the workings of our 3D TFM pipeline.

Ref R7. James F. Doyle, Modern Experimental Stress Analysis: Completing the Solution of Partially Specified Problems (John Wiley & Sons, Chichester, 2004).

Ref R8. Per Christian Hansen, Discrete Inverse Problems: Insight and Algorithms (siam, Philadelphia, 2010).

Author response image 7.

(A) shows simulated force field to generate simulated displacements. (B) shows force field reconstructed from simulated displacements with noise.

Recommendation 3: The authors show nuclear deformation on the hills and use this as evidence for a resultant downward-pointing force vector. This has, indeed, also been observed in other works referenced by the authors (e.g. Werner et al.), and could be interesting evidence to support the current observations, provided the authors also show a nuclear shape on the concave and flat regions. The authors could potentially also characterize this shape change better using higher-resolution data.

Response 3: We characterized nucleus deformation using Hoechst-stained samples as per recommendation. The deformation is estimated by dividing segmented nuclei volumes by best-fit ellipsoid volumes of same objects. In this way, objects exhibiting minimal bending will lead to values close to 1.0. The obtained graph is shown in figure Author response image 8B (and manuscript Figure 3D).

Author response image 8.

(A) an example of deformed nuclei on 50 µm wave hill region. (B) a Violin plot of calculated nuclear deformations across dimensions and features using segmented volume normalized against best-fit ellipsoid volume.

Our quantifications show a statistically significant difference in nuclei deformation measure medians between hill and valley cells on the 50 µm (0.973 vs 0.982) and 100 µm (0.971 vs 0.979) waves; this indicates that cells on the hills tend to have more deformed nuclei compared to cells in the valleys. Meanwhile, no significant difference was found for a similar comparison on 200 µm (0.978 vs 0.978) samples. For reference, the median found for cells pooled from planar regions was 0.975.

In response to the reviewer’s suggestions Figure 3 of our manuscript has been updated to include the new results on nuclei deformation. The text has also been updated to account for the new information to support our claims. The statistics are included in a new summary data table in Supplementary File 6.

Recommendation 4: The U-net for extrusion detection is a central tool used within this study, though the explanation and particularly validation of the tool are somewhat lacking. More clarity in the explanation and more examples of good (or bad) detections would help establish this tool as a more robust component of the data collection (on all geometries).

Response 4: The architecture of the neural network used in this study is outlined in supplementary figure S5a. To validate the performance of the model, a test dataset consisting of 200 positive examples and 100 negative examples were fed into the network and the resulting prediction was obtained from model. The confusion matrix of the model is shown in supplementary figure S5c. The weighted precision and recall of the model are 0.958 and 0.953 respectively.

Additionally, we have included examples of false positive and false negative detections in Figure 1-figure supplement 5 (Author response image 8). For false positive detections, these were typically observed to be extrusions that were labelled to have occurred the frame prior to the frame of interest (Author response image 9 bottom sequence). However, as the extrusion process is incomplete in the prior frame, there are still changes in the extruded cell body and the network falsely predicts this as a detection.

Author response image 9.

Examples of false negative and false positive extrusions registration.

Recommendation 5: The authors study the involvement of FAK in the observed curvature-dependent and hydraulic stress-dependent spatial regulation of cell extrusion. In one of the experiments, the authors supplement the cell medium with FAK inhibitors, though only in a hyper-osmotic medium. They show that FAK inhibition counteracts the extrusion-suppressing effect of a hyper-osmotic medium. However, no data is shown on the effect of FAK inhibitors within the control medium. Would the extrusion rates be even higher then?

Response 4: We proceeded, as suggested by the reviewer, to explore the effects of the FAK inhibitor on MDCK monolayers in our control medium. The results revealed that, at the 3 µM FAK concentration, where cells in sucrose media showed an elevated extrusion rate, monolayers in control medium quickly suffered massive cell death (Author response image 10) similar to what was seen when 6 µM FAK was introduced to sucrose medium.

This finding suggests that osmolarity protects against FAK inhibitors in a dose dependent manner. Moreover, as cell extrusions require an intact monolayer, its rates cannot increase indefinitely: a point will be reached where an intact monolayer can no longer be maintained.

We have updated the main text of our article to mention this observation, and also included a new time-lapse (Video 22) to demonstrate the effect.

Author response image 10.

Timelapse snapshot of MDCK monolayers over waves 4 hours after inclusion of focal adhesion kinase inhibitor.

Recommendation 6: The supplementary videos show two fields of view next to each other, which is not immediately clear to the viewer. I strongly advise the authors to add a clear border between the two panels, so that it is clear that the cells from one panel are not migrating into the next panel.

Response 6: A distinctive border has been added to the movies to separate panels showing different focal planes of the same stack.

Recommendation 7: The general quality and layout of the figures could be improved. Some figures would benefit from higher-resolution or larger cell images (e.g. Figure 2A, C, D), and the organisation of subpanels could be improved (e.g. especially in Figure 2). The box plots and bar graphs are also not consistent throughout the manuscript in terms of colouring and style, which should be improved.

Response 7: We have enlarged the figures in question accordingly, at the cost of reducing some information. However, the full scope of the sub-figures remains accessible in the supplementary movies. We have also tried to change the placement of the panels to improve readability. We have also adjusted the valley, hill, and flat coloring scheme for the extrusion boxplots in Figures 1 and 2 to make them consistent.

Recommendation 8: The graphs in Figures 3E and F are confusing and difficult to interpret. The x-axis states "Position along curve in radians" but it is unclear how to relate this to the position on the wavy substrate. The graphs also have a second vertical axis on the right ("valley-interface-hill"), which adds to the confusion. I would recommend the authors provide more explanation and consider a different approach of plotting this.

Response 8: We have removed the confusing plot of cross-sectional profile from the force graphs. To indicate positions on the waves, we have augmented radian values with Hill, Interface, and Valley accordingly.

Recommendation 9: Specify which silicone was used for the low-stiffness silicone substrates in the methods and in the main text.

Response 9: CY52 has been added to the main-text, next to the first appearance of the word soft silicone, to be consistent with the figures.

Recommendation 10: The flow lines that are plotted over the RICM data make it difficult to see the underlying RICM images. I would advise to also show the RICM images without the flow lines.

Response 10: The original movie S15 (now Video 16) showing the RICM overlapped with optical flow paths has now been replaced by a movie showing the same, but with the flow paths and RICM in separate panels.

Recommendation 11: In the first paragraph of the discussion, the authors write: "And this difference was both dependent on the sense (positive or negative)...". This is superfluous since the authors already mentioned earlier in the paragraph that the convex and concave regions (i.e. different signs of curvature) show differences in extrusion rates.

Response 11: The sentence has been changed to “And this difference was also dependent on the degree of curvature.”

Recommendation 12: In the second paragraph of the discussion, the authors mention that "basal fluid spaces under monolayers in hill regions were found consistently smaller than those in valley regions". Is this data shown in the figures of the manuscript? If so, a reference should be made because it was unclear to me.

Response 12: This statement is an inference from the comparison of the hill and valley RICM grey values. Specifically, RICM intensities are direct surrogates for basal separations (i.e., fluid space (as there cannot be a vacuum)) by virtue of the physics underlying the effect. To be more precise then, “inferred from RICM intensity differences (Figure 2I)” has been added to support the statement.

Recommendation 13: On page 7 of the discussion, the authors talk about positively and negatively curved surfaces. This type of description should be avoided, as this depends on the definition of the surface normal (i.e. is positive convex or concave?). Rather use convex and concave in this context.

Response 13: The wording has been changed accordingly.

Recommendation 14: The label of Table 8 reads "Table 2".

Response 14: The error has been corrected.

Reviewer #3

Recommendation 1: The central finding seems to be opposite to an earlier report (J Cell Sci (2019) 132, jcs222372), where MDCK cells in curved alginate tubes exhibit increased extrusion on a convex surface. I suggest that you comment on possible explanations for the different behaviors.

Response 1: The article in question primarily reported the phenomenon of MDCK and J3B1A monolayers detaching from the concave alginate tube walls coated with Matrigel. The authors attributed this to the curvature induced out-of-plane forces towards the center of the tubes. Up to this point, the findings and interpretation are consistent with our current study where we also find a similar force trend in concave regions.

To further lend support to the importance of curvature in inducing detachment, the authors cleverly bent the tubes to introduce asymmetry in curvature between outer and inner surfaces. Specifically, the outside bend is concave in both principal directions, whereas the inside bend is convex in one of its principal directions. As expected, the authors found that detachment rates from the outer surface were much larger compared to the inner one. Again, the observations and interpretations are consistent with our own findings; the convex direction will generate out-of-plane forces pointing into the surface, serving to stabilize the monolayer against the substrate. It should be noted however, since the inner-side tube is characterized by both convex and concave curvatures in its two principal directions, the resulting behavior of overlaying monolayers will depend on which of the two resulting forces become dominant. So, for gradual bends, one should expect the monolayers to still be able to detach from the inner tube surface. This is what was reported in their findings.

For their extrusion observations, I am surprised. Because their whole material (hydrogels) is presumably both solute and water permeable, I would be more inclined to expect very few extrusions irrespective of curvature. This is indeed the case with our study of MDCKs on PAM hydrogels, where the hydrogel substrate effectively buffers against the quick build-up of solute concentration and basal hydraulic stress. Without the latter, concave monolayer forces alone are unlikely to be able to disrupt cell focal adhesions. Indeed, the detachments seen in their study are more likely by exfoliation of Matrigel rather than pulling cells off Matrigel matrix entirely.

My guess is that the extrusions seen in their study are solely of the canonical crowding effect. If this was the case, then the detached monolayer on the outside bend could buffer against crowding pressure by buckling. Meanwhile, the monolayer on the inside bend, being attached to the surface, can only regulate crowding pressure by removing cells through extrusions. This phenomenon should be particular to soft matrices such as Matrigel. Using stiffer and covalently bonded ECM should be sufficient to prevent monolayers from detaching, leading to similar extrusion behaviors. In response to the reviewer’s recommendation then, we have included a short paragraph to state the points discussed in this response.

Recommendation 2: Fig 3E, F: The quantities displayed on the panels are not forces, but have units of pressure (or stress).

Response 2: we have changed “force” to “stress” according to the reviewer’s suggestion. The reason we kept the use of force in the original text was due to the fact that we were reconstructing forces. Due to discretization, the resulting forces will inevitably be assigned to element nodes. In between the nodes, in the faces, there will be no information. So, in order to have some form of continuity to plot, the face forces are obtained by averaging the 4 nodes around the element face. Unfortunately, element face areas are not typically of the same size, therefore the average forces obtained needs to be further normalized against the face area, leading to a quantity that has units of stress.

Recommendation 3: Fig 2D: Asterisks are hard to see.

Response 3: the color of the asterisks has been changed to green for better clarity against a B&W background.

Recommendation 4: p 19, l 7: Word missing in "the of molding"

Response 4: the typo has been amended to “the molding of”.

Author response

Reviewer #1 (Public Review):

Loss of skeletal muscle tissue from traumatic injury is debilitating. Restoring muscle mass and function remains a challenge. Using a mouse model, the authors performed punch biopsy injuries of the tibialis anterior in which the volume of muscle loss was varied to result in either successful muscle regeneration with a smaller injury or the unsuccessful outcome of fibrosis with a larger injury. For both conditions, a novel lipidomic profiling approach was used to evaluate pro-inflammatory and anti-inflammatory lipids at key time points post-injury with respect to collagen deposition, macrophage infiltration, muscle fiber regeneration, and force produced during isometric contractions. A key finding was that while all lipids increased at 3 days post-injury (dpi) and then declined through 14 dpi, pro-inflammatory lipids remained elevated during recovery from greater muscle loss which led to fibrosis. Maresin 1 was identified as an anti-inflammatory lipid that, when injected into injured muscle, reduced fibrosis, improved muscle regeneration, and partially restored the strength of contraction.

Strengths: The metabolipidomic profiling demonstrated here represents a novel approach to identifying pro-inflammatory and anti-inflammatory mediators of successful vs unsuccessful skeletal muscle regeneration. These findings may translate into a new therapeutic approach for promoting successful regeneration following volumetric muscle loss.

Weaknesses: Certain aspects of the data are overinterpreted; while some measures appear to have an adequate sample size to make sound conclusions, other measures are likely to lack sufficient statistical power given their variability. Presentation of the results would be strengthened by adhering to consistent terminology and labeling of figures throughout; specific examples are identified in recommendations to the authors. Several of the images used to illustrate differences between treatments are unconvincing because differences are not readily.

We agree with the reviewer and have scaled back some of the interpretation as well as clarified the sample sizes. We have also amended the text to maintain a consistent terminology.

Reviewer #2 (Public Review):

The study is novel and valuable to the field and provides new and important insights into the role of lipid mediators in VML injuries. By expanding our understanding of the mechanisms that regulate muscle regeneration following VML injuries, the study has the potential to guide the development of novel therapeutic interventions that promote tissue repair and recovery. The data presented in the manuscript is of good quality. The findings and conclusions are supported by a variety of different analyses (e.g., gene expression, histology, flow cytometry).

Despite the strengths of the study, some limitations are identified. Specifically, the impact of maresin 1 on macrophage phenotypes (M1/M2) could have been explored in more detail using histological or protein expression analysis. Moreover, additional data are needed to substantiate the claims about increased muscle regeneration. Lastly, the study does not address myofiber innervation, myofiber-type transitions, or motor unit remodeling.

We thank the reviewer for the suggestions and have performed a more in-depth exploration of macrophage phenotypes through additional scRNA-sequencing analysis. We have also included additional data describing how Maresin 1 impacts muscle stem cells through cyclic AMP. Respectfully, profiling myofiber innervation, motor unit remodeling and myofiber-type transitions are beyond the scope of this manuscript.

Author Response

Reviewer #1 (Public Review):

In this work George et al. describe RatInABox, a software system for generating surrogate locomotion trajectories and neural data to simulate the effects of a rodent moving about an arena. This work is aimed at researchers that study rodent navigation and its neural machinery.

Strengths:

• The software contains several helpful features. It has the ability to import existing movement traces and interpolate data with lower sampling rates. It allows varying the degree to which rodents stay near the walls of the arena. It appears to be able to simulate place cells, grid cells, and some other features.

• The architecture seems fine and the code is in a language that will be accessible to many labs.

• There is convincing validation of velocity statistics. There are examples shown of position data, which seem to generally match between data and simulation.

Weaknesses:

• There is little analysis of position statistics. I am not sure this is needed, but the software might end up more powerful and the paper higher impact if some position analysis was done. Based on the traces shown, it seems possible that some additional parameters might be needed to simulate position/occupancy traces whose statistics match the data.

Thank you for this suggestion. We have added a new panel to figure 2 showing a histogram of the time the agent spends at positions of increasing distance from the nearest wall. As you can see, RatInABox is a good fit to the real locomotion data: positions very near the wall are under-explored (in the real data this is probably because whiskers and physical body size block positions very close to the wall) and positions just away from but close to the wall are slightly over explored (an effect known as thigmotaxis, already discussed in the manuscript).

As you correctly suspected, fitting this warranted a new parameter which controls the strength of the wall repulsion, we call this “wall_repel_strength”. The motion model hasn’t mathematically changed, all we did was take a parameter which was originally a fixed constant 1, unavailable to the user, and made it a variable which can be changed (see methods section 6.1.3 for maths). The curves fit best when wall_repel_strength ~= 2. Methods and parameters table have been updated accordingly. See Fig. 2e.

• The overall impact of this work is somewhat limited. It is not completely clear how many labs might use this, or have a need for it. The introduction could have provided more specificity about examples of past work that would have been better done with this tool.

At the point of publication we, like yourself, also didn’t know to what extent there would be a market for this toolkit however we were pleased to find that there was. In its initial 11 months RatInABox has accumulated a growing, global user base, over 120 stars on Github and north of 17,000 downloads through PyPI. We have accumulated a list of testimonials[5] from users of the package vouching for its utility and ease of use, four of which are abridged below. These testimonials come from a diverse group of 9 researchers spanning 6 countries across 4 continents and varying career stages from pre-doctoral researchers with little computational exposure to tenured PIs. Finally, not only does the community use RatInABox they are also building it: at the time of writing RatInABx has received logged 20 GitHub “Issues” and 28 “pull requests” from external users (i.e. those who aren’t authors on this manuscript) ranging from small discussions and bug-fixes to significant new features, demos and wrappers.

Abridged testimonials:

● “As a medical graduate from Pakistan with little computational background…I found RatInABox to be a great learning and teaching tool, particularly for those who are underprivileged and new to computational neuroscience.” - Muhammad Kaleem, King Edward Medical University, Pakistan

● “RatInABox has been critical to the progress of my postdoctoral work. I believe it has the strong potential to become a cornerstone tool for realistic behavioural and neuronal modelling” - Dr. Colleen Gillon, Imperial College London, UK

● “As a student studying mathematics at the University of Ghana, I would recommend RatInABox to anyone looking to learn or teach concepts in computational neuroscience.” - Kojo Nketia, University of Ghana, Ghana

● “RatInABox has established a new foundation and common space for advances in cognitive mapping research.” - Dr. Quinn Lee, McGill, Canada

The introduction continues to include the following sentence highlighting examples of past work which relied of generating artificial movement and/or neural dat and which, by implication could have been done better (or at least accelerated and standardised) using our toolbox.

“Indeed, many past[13, 14, 15] and recent[16, 17, 18, 19, 6, 20, 21] models have relied on artificially generated movement trajectories and neural data.”

• Presentation: Some discussion of case studies in Introduction might address the above point on impact. It would be useful to have more discussion of how general the software is, and why the current feature set was chosen. For example, how well does RatInABox deal with environments of arbitrary shape? T-mazes? It might help illustrate the tool's generality to move some of the examples in supplementary figure to main text - or just summarize them in a main text figure/panel.

Thank you for this question. Since the initial submission of this manuscript RatInABox has been upgraded and environments have become substantially more “general”. Environments can now be of arbitrary shape (including T-mazes), boundaries can be curved, they can contain holes and can also contain objects (0-dimensional points which act as visual cues). A few examples are showcased in the updated figure 1 panel e.

To further illustrate the tools generality beyond the structure of the environment we continue to summarise the reinforcement learning example (Fig. 3e) and neural decoding example in section 3.1. In addition to this we have added three new panels into figure 3 highlighting new features which, we hope you will agree, make RatInABox significantly more powerful and general and satisfy your suggestion of clarifying utility and generality in the manuscript directly.

On the topic of generality, we wrote the manuscript in such a way as to demonstrate how the rich variety of ways RatInABox can be used without providing an exhaustive list of potential applications. For example, RatInABox can be used to study neural decoding and it can be used to study reinforcement learning but not because it was purpose built with these use-cases in mind. Rather because it contains a set of core tools designed to support spatial navigation and neural representations in general. For this reason we would rather keep the demonstrative examples as supplements and implement your suggestion of further raising attention to the large array of tutorials and demos provided on the GitHub repository by modifying the final paragraph of section 3.1 to read:

“Additional tutorials, not described here but available online, demonstrate how RatInABox can be used to model splitter cells, conjunctive grid cells, biologically plausible path integration, successor features, deep actor-critic RL, whisker cells and more. Despite including these examples we stress that they are not exhaustive. RatInABox provides the framework and primitive classes/functions from which highly advanced simulations such as these can be built.”

Reviewer #3 (Public Review):

George et al. present a convincing new Python toolbox that allows researchers to generate synthetic behavior and neural data specifically focusing on hippocampal functional cell types (place cells, grid cells, boundary vector cells, head direction cells). This is highly useful for theory-driven research where synthetic benchmarks should be used. Beyond just navigation, it can be highly useful for novel tool development that requires jointly modeling behavior and neural data. The code is well organized and written and it was easy for us to test.

We have a few constructive points that they might want to consider.

• Right now the code only supports X,Y movements, but Z is also critical and opens new questions in 3D coding of space (such as grid cells in bats, etc). Many animals effectively navigate in 2D, as a whole, but they certainly make a large number of 3D head movements, and modeling this will become increasingly important and the authors should consider how to support this.

Agents now have a dedicated head direction variable (before head direction was just assumed to be the normalised velocity vector). By default this just smoothes and normalises the velocity but, in theory, could be accessed and used to model more complex head direction dynamics. This is described in the updated methods section.

In general, we try to tread a careful line. For example we embrace certain aspects of physical and biological realism (e.g. modelling environments as continuous, or fitting motion to real behaviour) and avoid others (such as the biophysics/biochemisty of individual neurons, or the mechanical complexities of joint/muscle modelling). It is hard to decide where to draw but we have a few guiding principles:

1. RatInABox is most well suited for normative modelling and neuroAI-style probing questions at the level of behaviour and representations. We consciously avoid unnecessary complexities that do not directly contribute to these domains.

2. Compute: To best accelerate research we think the package should remain fast and lightweight. Certain features are ignored if computational cost outweighs their benefit.

3. Users: If, and as, users require complexities e.g. 3D head movements, we will consider adding them to the code base.

For now we believe proper 3D motion is out of scope for RatInABox. Calculating motion near walls is already surprisingly complex and to do this in 3D would be challenging. Furthermore all cell classes would need to be rewritten too. This would be a large undertaking probably requiring rewriting the package from scratch, or making a new package RatInABox3D (BatInABox?) altogether, something which we don’t intend to undertake right now. One option, if users really needed 3D trajectory data they could quite straightforwardly simulate a 2D Environment (X,Y) and a 1D Environment (Z) independently. With this method (X,Y) and (Z) motion would be entirely independent which is of unrealistic but, depending on the use case, may well be sufficient.

Alternatively, as you said that many agents effectively navigate in 2D but show complex 3D head and other body movements, RatInABox could interface with and feed data downstream to other softwares (for example Mujoco[11]) which specialise in joint/muscle modelling. This would be a very legitimate use-case for RatInABox.

We’ve flagged all of these assumptions and limitations in a new body of text added to the discussion:

“Our package is not the first to model neural data[37, 38, 39] or spatial behaviour[40, 41], yet it distinguishes itself by integrating these two aspects within a unified, lightweight framework. The modelling approach employed by RatInABox involves certain assumptions:

1. It does not engage in the detailed exploration of biophysical[37, 39] or biochemical[38] aspects of neural modelling, nor does it delve into the mechanical intricacies of joint and muscle modelling[40, 41]. While these elements are crucial in specific scenarios, they demand substantial computational resources and become less pertinent in studies focused on higher-level questions about behaviour and neural representations.

2. A focus of our package is modelling experimental paradigms commonly used to study spatially modulated neural activity and behaviour in rodents. Consequently, environments are currently restricted to being two-dimensional and planar, precluding the exploration of three-dimensional settings. However, in principle, these limitations can be relaxed in the future.

3. RatInABox avoids the oversimplifications commonly found in discrete modelling, predominant in reinforcement learning[22, 23], which we believe impede its relevance to neuroscience.

4. Currently, inputs from different sensory modalities, such as vision or olfaction, are not explicitly considered. Instead, sensory input is represented implicitly through efficient allocentric or egocentric representations. If necessary, one could use the RatInABox API in conjunction with a third-party computer graphics engine to circumvent this limitation.

5. Finally, focus has been given to generating synthetic data from steady-state systems. Hence, by default, agents and neurons do not explicitly include learning, plasticity or adaptation. Nevertheless we have shown that a minimal set of features such as parameterised function-approximator neurons and policy control enable a variety of experience-driven changes in behaviour the cell responses[42, 43] to be modelled within the framework.

• What about other environments that are not "Boxes" as in the name - can the environment only be a Box, what about a circular environment? Or Bat flight? This also has implications for the velocity of the agent, etc. What are the parameters for the motion model to simulate a bat, which likely has a higher velocity than a rat?

Thank you for this question. Since the initial submission of this manuscript RatInABox has been upgraded and environments have become substantially more “general”. Environments can now be of arbitrary shape (including circular), boundaries can be curved, they can contain holes and can also contain objects (0-dimensional points which act as visual cues). A few examples are showcased in the updated figure 1 panel e.

Whilst we don’t know the exact parameters for bat flight users could fairly straightforwardly figure these out themselves and set them using the motion parameters as shown in the table below. We would guess that bats have a higher average speed (speed_mean) and a longer decoherence time due to increased inertia (speed_coherence_time), so the following code might roughly simulate a bat flying around in a 10 x 10 m environment. Author response image 1 shows all Agent parameters which can be set to vary the random motion model.

Author response image 1.

• Semi-related, the name suggests limitations: why Rat? Why not Agent? (But its a personal choice)

We came up with the name “RatInABox” when we developed this software to study hippocampal representations of an artificial rat moving around a closed 2D world (a box). We also fitted the random motion model to open-field exploration data from rats. You’re right that it is not limited to rodents but for better or for worse it’s probably too late for a rebrand!

• A future extension (or now) could be the ability to interface with common trajectory estimation tools; for example, taking in the (X, Y, (Z), time) outputs of animal pose estimation tools (like DeepLabCut or such) would also allow experimentalists to generate neural synthetic data from other sources of real-behavior.

This is actually already possible via our “Agent.import_trajectory()” method. Users can pass an array of time stamps and an array of positions into the Agent class which will be loaded and smoothly interpolated along as shown here in Fig. 3a or demonstrated in these two new papers[9,10] who used RatInABox by loading in behavioural trajectories.

• What if a place cell is not encoding place but is influenced by reward or encodes a more abstract concept? Should a PlaceCell class inherit from an AbstractPlaceCell class, which could be used for encoding more conceptual spaces? How could their tool support this?

In fact PlaceCells already inherit from a more abstract class (Neurons) which contains basic infrastructure for initialisation, saving data, and plotting data etc. We prefer the solution that users can write their own cell classes which inherit from Neurons (or PlaceCells if they wish). Then, users need only write a new get_state() method which can be as simple or as complicated as they like. Here are two examples we’ve already made which can be found on the GitHub:

Author response image 2.

Phase precession: PhasePrecessingPlaceCells(PlaceCells)[12] inherit from PlaceCells and modulate their firing rate by multiplying it by a phase dependent factor causing them to “phase precess”.

Splitter cells: Perhaps users wish to model PlaceCells that are modulated by recent history of the Agent, for example which arm of a figure-8 maze it just came down. This is observed in hippocampal “splitter cell”. In this demo[1] SplitterCells(PlaceCells) inherit from PlaceCells and modulate their firing rate according to which arm was last travelled along.

• This a bit odd in the Discussion: "If there is a small contribution you would like to make, please open a pull request. If there is a larger contribution you are considering, please contact the corresponding author3" This should be left to the repo contribution guide, which ideally shows people how to contribute and your expectations (code formatting guide, how to use git, etc). Also this can be very off-putting to new contributors: what is small? What is big? we suggest use more inclusive language.

We’ve removed this line and left it to the GitHub repository to describe how contributions can be made.

• Could you expand on the run time for BoundaryVectorCells, namely, for how long of an exploration period? We found it was on the order of 1 min to simulate 30 min of exploration (which is of course fast, but mentioning relative times would be useful).

Absolutely. How long it takes to simulate BoundaryVectorCells will depend on the discretisation timestep and how many neurons you simulate. Assuming you used the default values (dt = 0.1, n = 10) then the motion model should dominate compute time. This is evident from our analysis in Figure 3f which shows that the update time for n = 100 BVCs is on par with the update time for the random motion model, therefore for only n = 10 BVCs, the motion model should dominate compute time.

So how long should this take? Fig. 3f shows the motion model takes ~10-3 s per update. One hour of simulation equals this will be 3600/dt = 36,000 updates, which would therefore take about 72,000*10-3 s = 36 seconds. So your estimate of 1 minute seems to be in the right ballpark and consistent with the data we show in the paper.

Interestingly this corroborates the results in a new inset panel where we calculated the total time for cell and motion model updates for a PlaceCell population of increasing size (from n = 10 to 1,000,000 cells). It shows that the motion model dominates compute time up to approximately n = 1000 PlaceCells (for BoundaryVectorCells it’s probably closer to n = 100) beyond which cell updates dominate and the time scales linearly.

These are useful and non-trivial insights as they tell us that the RatInABox neuron models are quite efficient relative to the RatInABox random motion model (something we hope to optimise further down the line). We’ve added the following sentence to the results:

“Our testing (Fig. 3f, inset) reveals that the combined time for updating the motion model and a population of PlaceCells scales sublinearly O(1) for small populations n > 1000 where updating the random motion model dominates compute time, and linearly for large populations n > 1000. PlaceCells, BoundaryVectorCells and the Agent motion model update times will be additionally affected by the number of walls/barriers in the Environment. 1D simulations are significantly quicker than 2D simulations due to the reduced computational load of the 1D geometry.”

And this sentence to section 2:

“RatInABox is fundamentally continuous in space and time. Position and velocity are never discretised but are instead stored as continuous values and used to determine cell activity online, as exploration occurs. This differs from other models which are either discrete (e.g. “gridworld” or Markov decision processes) or approximate continuous rate maps using a cached list of rates precalculated on a discretised grid of locations. Modelling time and space continuously more accurately reflects real-world physics, making simulations smooth and amenable to fast or dynamic neural processes which are not well accommodated by discretised motion simulators. Despite this, RatInABox is still fast; to simulate 100 PlaceCell for 10 minutes of random 2D motion (dt = 0.1 s) it takes about 2 seconds on a consumer grade CPU laptop (or 7 seconds for BoundaryVectorCells).”

• Regarding the Geometry and Boundary conditions, would supporting hyperbolic distance might be useful, given the interest in alternative geometry of representations (ie, https://www.nature.com/articles/s41593-022-01212-4)?

Whilst this would be very interesting it would likely represent quite a significant edit, requiring rewriting of almost all the geometry-handling code. We’re happy to consider changes like these according to (i) how simple they will be to implement, (ii) how disruptive they will be to the existing API, (iii) how many users would benefit from the change. If many users of the package request this we will consider ways to support it.

• In general, the set of default parameters might want to be included in the main text (vs in the supplement).

We also considered this but decided to leave them in the methods for now. The exact value of these parameters are subject to change in future versions of the software. Also, we’d prefer for the main text to provide a low-detail high-level description of the software and the methods to provide a place for keen readers to dive into the mathematical and coding specifics.

• It still says you can only simulate 4 velocity or head directions, which might be limiting.

Thanks for catching this. This constraint has been relaxed. Users can now simulate an arbitrary number of head direction cells with arbitrary tuning directions and tuning widths. The methods have been adjusted to reflect this (see section 6.3.4).

• The code license should be mentioned in the Methods.

We have added the following section to the methods:

6.6 License RatInABox is currently distributed under an MIT License, meaning users are permitted to use, copy, modify, merge publish, distribute, sublicense and sell copies of the software.

Author Response

LD Score regression (LDSC) is a software tool widely used in the field of genome-wide association studies (GWAS) for estimating heritabilities, genetic correlations, the extent of confounding, and biological enrichment. LDSC is for the most part not regarded as an accurate estimator of \emph{absolute} heritability (although useful for relative comparisons). It is relied on primarily for its other uses (e.g., estimating genetic correlations). The authors propose a new method called \texttt{i-LDSC}, extending the original LDSC in order to estimate a component of genetic variance in addition to the narrow-sense heritability---epistatic genetic variance, although not necessarily all of it. Epistasis in quantitative genetics refers to the component of genetic variance that cannot be captured by a linear model regressing total genetic values on single-SNP genotypes. \texttt{i-LDSC} seems aimed at estimating that part of the epistatic variance residing in statistical interactions between pairs of SNPs. To simplify, the basic model of \texttt{i-LDSC} for two SNPs $X_1$ and $X_2$ is

\begin{equation}\label{eq:twoX} Y = X_1 \beta_1 + X_2 \beta_2 + X_1 X_2 \theta + E, \end{equation}

and estimation of the epistatic variance associated with the product term proceeds through a variant of the original LD Score that measures the extent to which a SNP tags products of genotypes (rather than genotypes themselves). The authors conducted simulations to test their method and then applied it to a number of traits in the UK Biobank and Biobank Japan. They found that for all traits the additive genetic variance was larger than the epistatic, but for height the absolute size of the epistatic component was estimated to be non-negligible. An interpretation of the authors' results that perhaps cannot be ruled out, however, is that pairwise epistasis overall does not make a detectable contribution to the variance of quantitative traits.

We thank the reviewer for carefully reading of our manuscript and we appreciate the constructive comments. Our responses and edits to the specific major comments and minor issues are given below.

Major Comments

This paper has a lot of strong points, and I commend the authors for the effort and ingenuity expended in tackling the difficult problem of estimating epistatic (non-additive) genetic variance from GWAS summary statistics. The mere possibility of the estimated univariate regression coefficient containing a contribution from epistasis, as represented in the manuscript's Equation~3 and elsewhere, is intriguing in and of itself.

Is \texttt{i-LDSC} Estimating Epistasis?

Perhaps the issue that has given me the most pause is uncertainty over whether the paper's method is really estimating the non-additive genetic variance, as this has been traditionally defined in quantitative genetics with great consequences for the correlations between relatives and evolutionary theory (Fisher, 1930, 1941; Lynch & Walsh, 1998; Burger, 2000; Ewens, 2004).

Let us call the expected phenotypic value of a given multiple-SNP genotype the \emph{total genetic value}. If we apply least-squares regression to obtain the coefficients of the SNPs in a simple linear model predicting the total genetic values, then the partial regression coefficients are the \emph{average effects of gene substitution} and the variance in the predicted values resulting from the model is called the \emph{additive genetic variance}. (This is all theoretical and definitional, not empirical. We do not actually perform this regression.) The variance in the residuals---the differences between the total genetic values and the additive predicted values---is the \emph{non-additive genetic variance}. Notice that this is an orthogonal decomposition of the variance in total genetic values. Thus, in order for the variance in $\mathbf{W}\bm{\theta}$ to qualify as the non-additive genetic variance, it must be orthogonal to $\mathbf{X} \bm{\beta}$.

At first, I very much doubted whether this is generally true. And I was not reassured by the authors' reply to Reviewer~1 on this point, which did not seem to show any grasp of the issue at all. But to my surprise I discovered in elementary simulations of Equation~\ref{eq:twoX} above that for mean-centered $X_1$ and $X_2$, $(X_1 \beta_1 + X_2 \beta_2)$ is uncorrelated with $X_1 X_2 \theta$ for seemingly arbitrary correlation between $X_1$ and $X_2$. A partition of the outcome's variance between these two components is thus an orthogonal decomposition after all. Furthermore, the result seems general for any number of independent variables and their pairwise products. I am also encouraged by the report that standard and interaction LD Scores are ``lowly correlated' (line~179), meaning that the standard LDSC slope is scarcely affected by the inclusion of interaction LD Scores in the regression; this behavior is what we should expect from an orthogonal decomposition.

I have therefore come to the view that the additional variance component estimated by \texttt{i-LDSC} has a close correspondence with the epistatic (non-additive) genetic variance after all.

In order to make this point transparent to all readers, however, I think that the authors should put much more effort into placing their work into the traditional framework of the field. It was certainly not intuitive to multiple reviewers that $\mathbf{X}\bm{\beta}$ is orthogonal to $\mathbf{W}\bm{\theta}$. There are even contrary suggestions. For if $(\mathbf{X}\bm{\beta})^\intercal \mathbf{W} \bm{\theta} = \bm{\beta}^\intercal \mathbf{X}^\intercal \mathbf{W} \bm{\theta} $ is to equal zero, we know that we can't get there by $\mathbf{X}^\intercal \mathbf{W}$ equaling zero because then the method has nothing to go on (e.g., line~139). We thus have a quadratic form---each term being the weighted product of an average (additive) effect and an interaction coefficient---needing to cancel out to equal zero. I wonder if the authors can put forth a rigorous argument or compelling intuition for why this should be the case.

In the case of two polymorphic sites, quantitative genetics has traditionally partitioned the total genetic variance into the following orthogonal components:

\begin{itemize}

\item additive genetic variance, $\sigma^2_A$, the numerator of the narrow-sense heritability;

\item dominance genetic variance, $\sigma^2_D$;

\item additive-by-additive genetic variance, $\sigma^2_{AA}$;

\item additive-by-dominance genetic variance, $\sigma^2_{AD}$; and

\item dominance-by-dominance genetic variance, $\sigma^2_{DD}$.

\end{itemize}

See Lynch and Walsh (1998, pp. 88-92) for a thorough numerical example. This decomposition is not arbitrary or trivial, since each component has a distinct coefficient in the correlations between relatives. Is it possible for the authors to relate the variance associated with their $\mathbf{W}\bm{\theta}$ to this traditional decomposition? Besides justifying the work in this paper, the establishment of a relationship can have the possible practical benefit of allowing \texttt{i-LDSC} estimates of non-additive genetic variance to be checked against empirical correlations between relatives. For example, if we know from other methods that $\sigma^2_D$ is negligible but that \texttt{i-LDSC} returns a sizable $\sigma^2_{AA}$, we might predict that the parent-offspring correlation should be equal to the sibling correlation; a sizable $\sigma^2_D$ would make the sibling correlation higher. Admittedly, however, such an exercise can get rather complicated for the variance contributed by pairs of SNPs that are close together (Lynch & Walsh, 1998, pp. 146-152).

I would also like the authors to clarify whether LDSC consistently overestimates the narrow-sense heritability in the case that pairwise epistasis is present. The figures seem to show this. I have conflicting intuitions here. On the one hand, if GWAS summary statistics can be inflated by the tagging of epistasis, then it seems that LDSC should overestimate heritability (or at least this should be an upwardly biasing factor; other factors may lead the net bias to be different). On the other hand, if standard and interaction LD Scores are lowly correlated, then I feel that the inclusion of interaction LD Score in the regression should not strongly affect the coefficient of the standard LD Score. Relatedly, I find it rather curious that \texttt{i-LDSC} seems increasingly biased as the proportion of genetic variance that is non-additive goes up---but perhaps this is not too important, since such a high ratio of narrow-sense to broad-sense heritability is not realistic.

We thank the reviewer for taking the time to thoughtfully offer more context on how we might situate the i-LDSC framework within the greater context of traditional quantitative genetics. We now formalize the interaction component used in the i-LDSC model as an estimate of the phenotypic variance explained by additive-by-additive interactions between genetic variants (which we denote by 𝜎" to follow the conventional notation). In the newly revised Material and Methods, we also show how the i-LDSC model can be formulated to include dominance effects in a more general framework. Our updated derivations provide two key takeaways.

First, we assume that the additive and interaction effect sizes in the general model (𝜷,𝜽) are each normally distributed with variances proportional to their individual contributions to trait heritability: 𝛽& ∼ 𝒩(0, 𝜎"), 𝜃' ∼ 𝒩(0, 𝜎" ). This independence assumption implies that the additive and non- $ $$ additive components 𝑿𝜷 and 𝑾𝜽 are orthogonal where 𝔼[𝜷⊺𝑿⊺𝑾𝜽] = 𝔼[𝜷⊺]𝑿⊺𝑾𝔼[𝜽] = 𝟎. This is important because, as the reviewer points out, it means that there is a unique partitioning of genetic variance when studying a trait of interest. In the revised version of the manuscript, we show this derivation in the main text (see lines 129-143). We also extend this derivation in the Materials and Methods where we show the same result even after we include the presence of dominance effects in the generative model (see lines 415-417 and 438-457).

Second, we show that the genotype matrix 𝑿 and the matrix of genetic interactions 𝑾 are not linearly dependent because the additive-by-additive effects between two SNPs are encoded as the Hadamard product of two genotypic vectors in the form 𝒘! = 𝒙" ∘ 𝒙# (which is a nonlinear function of the genotypes). Linear dependence would have implied that one could find a transformation between a SNP and an interaction term in the form 𝒘! = 𝑐 × 𝒙" for some constant 𝑐. However, despite their linear independence, 𝑿 and 𝑾 are themselves not orthogonal and still have a nonzero correlation. This implies that the inner product between genotypes and their interactions is nonzero 𝑿⊺𝑾 ≠ 𝟎. To see this, we focus on a focal SNP 𝒙& and consider three different types of interactions:

• Scenario I: Interaction between a focal SNP with itself (𝒙" ∘ 𝒙").
• Scenario II: Interaction between a focal SNP with a different SNP (𝒙" ∘ 𝒙#).
• Scenario III: Interaction between a focal SNP with a pair of different SNPs (𝒙# ∘ 𝒙$).

In the Materials and Methods of the revised manuscript, we now provide derivations showing when would expect nonzero correlation between 𝑿 and 𝑾 which rely on the fact that: (1) we assume that genotypes have been mean-centered and scaled to have unit variance, and (2) under Hardy-Weinberg equilibrium, SNPs marginally follow a binomial distribution 𝒙& ∼ 𝐵𝑖𝑛(2, 𝑝) where 𝑝 represents the minor allele frequency (MAF) (Wray et al. 2007, Genome Res; Lippert et al. 2013, Sci Rep). These new additions are given in new lines 460-485).

Lastly, we agree with the reviewer that our results indicate that LDSC inflates estimates of SNP- based narrow-sense heritability. Our intuition for why this happens is largely consistent with the reviewer’s first point: since GWAS summary statistics can be inflated by the tagging of non- additive genetic variance, then it makes sense that LDSC should overestimate heritability. LDSC uses a univariate regression without the inclusion of cis-interaction scores. A simple consequence from “omitted variable bias” is likely happening where, since LDSC does not explicitly account for contributions from the tagged non-additive components which also contribute to the variance in the GWAS summary statistics, the estimate for the coefficient 𝜎" becomes slightly inflated.

How Much Epistasis Is \texttt{i-LDSC} Detecting?

I think the proper conclusion to be drawn from the authors' analyses is that statistically significant epistatic (non-additive) genetic variance was not detected. Specifically, I think that the analysis presented in Supplementary Table~S6 should be treated as a main analysis rather than a supplementary one, and the results here show no statistically significant epistasis. Let me explain.

Most serious researchers, I think, treat LDSC as an unreliable estimator of narrow-sense heritability; it typically returns estimates that are too low. Not even the original LDSC paper pressed strongly to use the method for estimating $h^2$ (Bulik-Sullivan et al., 2015). As a practical matter, when researchers are focused on estimating absolute heritability with high accuracy, they usually turn to GCTA/GREML (Evans et al., 2018; Wainschtein et al., 2022).

One reason for low estimates with LDSC is that if SNPs with higher LD Scores are less likely to be causal or to have large effect sizes, then the slope of univariate LDSC will not rise as much as it ``should' with increasing LD Score. This was a scenario actually simulated by the authors and displayed in their Supplementary Figure~S15. [Incidentally, the authors might have acknowledged earlier work in this vein. A simulation inducing a negative correlation between LD Scores and $\chi^2$ statistics was presented by Bulik-Sullivan et al. (2015, Supplementary Figure 7), and the potentially biasing effect of a correlation over SNPs between LD Scores and contributed genetic variance was a major theme of Lee et al. (2018).] A negative correlation between LD Score and contributed variance does seem to hold for a number of reasons, including the fact that regions of the genome with higher recombination rates tend to be more functional. In short, the authors did very well to carry out this simulation and to show in their Supplementary Figure~S15 that this flaw of LDSC in estimating narrow-sense heritability is also a flaw of \texttt{i-LDSC} in estimating broad-sense heritability. But they should have carried the investigation at least one step further, as I will explain below.

Another reason for LDSC being a downwardly biased estimator of heritability is that it is often applied to meta-analyses of different cohorts, where heterogeneity (and possibly major but undetected errors by individual cohorts) lead to attenuation of the overall heritability (de Vlaming et al., 2017).

The optimal case for using LDSC to estimate heritability, then, is incorporating the LD-related annotation introduced by Gazal et al. (2017) into a stratified-LDSC (s-LDSC) analysis of a single large cohort. This is analogous to the calculation of multiple GRMs defined by MAF and LD in the GCTA/GREML papers cited above. When this was done by Gazal et al. (2017, Supplementary Table 8b), the joint impact of the improvements was to increase the estimated narrow-sense heritability of height from 0.216 to 0.534.

All of this has at least a few ramifications for \texttt{i-LDSC}. First, the authors do not consider whether a relationship between their interaction LD Scores and interaction effect sizes might bias their estimates. (This would be on top of any biasing relationship between standard LD Scores and linear effect sizes, as displayed in Supplementary Figure~S15.) I find some kind of statistical relationship over the whole genome, induced perhaps by evolutionary forces, between \emph{cis}-acting epistasis and interaction LD Scores to be plausible, albeit without intuition regarding the sign of any resulting bias. The authors should investigate this issue or at least mention it as a matter for future study. Second, it might be that the authors are comparing the estimates of broad-sense heritability in Table~1 to the wrong estimates of narrow-sense heritability. Although the estimates did come from single large cohorts, they seem to have been obtained with simple univariate LDSC rather than s-LDSC. When the estimate of $h^2$ obtained with LDSC is too low, some will suspect that the additional variance detected by \texttt{i-LDSC} is simply additive genetic variance missed by the downward bias of LDSC. Consider that the authors' own Supplementary Table~S6 gives s-LDSC heritability estimates that are consistently higher than the LDSC estimates in Table~1. E.g., the estimated $h^2$ of height goes from 0.37 to 0.43. The latter figure cuts quite a bit into the estimated broad-sense heritability of 0.48 obtained with \texttt{i-LDSC}.

Here we come to a critical point. Lines 282--286 are not entirely clear, but I interpret them to mean that the manuscript's Equation~5 was expanded by stratifying $\ell$ into the components of s-LDSC and this was how the estimates in Supplementary Table~S6 were obtained. If that interpretation is correct, then the scenario of \texttt{i-LDSC} picking up missed additive genetic variance seems rather plausible. At the very least, the increases in broad-sense heritability reported in Supplementary Table~S6 are smaller in magnitude and \emph{not statistically significant}. Perhaps what this means is that the headline should be a \emph{negligible} contribution of pairwise epistasis revealed by this novel and ingenious method, analogous to what has been discovered with respect to dominance (Hivert et al., 2021; Pazokitoroudi et al., 2021; Okbay et al., 2022; Palmer et al., 2023).

This is an excellent question raised by the reviewer and, again, we really appreciate such a thoughtful and thorough response. First, we completely agree with the reviewer that the s-LDSC estimates previously included in the Supplementary Material should instead be discussed in the main text of the manuscript. In the revision, we have now moved the old Supplemental Table S6 to be the new Table 2. Second, we also agree that the conclusions about the magnitude of additive-by-additive effects should be based upon variance explained when using the cis- interaction score in addition to scores specific to different biological annotations when available, per s-LDSC.

However, we want to respectfully disagree that the results indicate a negligible contribution of additive-by-additive genetic variance to all the traits we analyzed (see Figure 4D). Although the additive-by-additive genetic variance component is not significant in any trait in the UK Biobank, there is little reason to expect that they would be given the inclusion of 97 other biological annotations from the s-LDSC model. Indeed, in the s-LDSC paper itself the authors look only for enrichment of heritability for a given annotation not a statistically significant test statistic. It also worth noting that jackknife approaches tend to be conservative and yield slightly larger standard errors for hypothesis testing. Taking all the great points that the reviewer mentioned into account, we believe that a moderate stance to the interpretation of our results is one that: (i) emphasizes the importance of using s-LDSC with the cis-interaction score to better assess the variance explained by additive-by-additive interaction effects and (ii) allows for the significance of the additive-by-additive component to not be the only factor when determining the importance of the role of non-additive effects in shaping trait architecture.

In the revision, we now write the following in lines 331-343:

Lastly, we performed an additional analysis in the UK Biobank where the cis-interaction scores are included as an annotation alongside 97 other functional categories in the stratified-LD score regression framework and its software s-LDSC (Materials and Methods). Here, s-LDSC heritability estimates still showed an increase with the interaction scores versus when the publicly available functional categories were analyzed alone, but albeit at a much smaller magnitude (Table 2). The contributions from the additive-by-additive component to the overall estimate of genetic variance ranged from 0.005 for MCHC (P = 0.373) to 0.055 for HDL (P = 0.575) (Figures 4C and 4D). Furthermore, in this analysis, the estimates of the additive-by-additive components were no longer statistically significant for any of the traits in the UK Biobank (Table 2). Despite this, these results highlight the ability of the i-LDSC framework to identify sources of “missing” phenotypic variance explained in heritability estimation. Importantly, moving forward, we suggest using the cis- interaction scores with additional annotations whenever they are available as it provides more conservative estimates of the role of additive-by-additive effects on trait architecture.

Lastly, in the Discussion, we now mention an area of future work would be to explore how the relationship between cis-interaction LD scores and interaction effect sizes might bias heritability estimates from i-LDSC (e.g., similar to the relationship explored standard LD scores and linear effect sizes in Figure 3 – figure supplement 8). See new lines 364-367.

Author Response

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

We agree with Reviewer #1 that it is not typical to include primary data in a review, but this seems to be a very unusual situation and it is not unprecedented. We seriously believe that it will significantly dilute the impact of the message if we were to separate this into two papers. We intended initially to do a comprehensive review of the αC-β4 motif as we think it is an extremely important element of secondary structure that has been rather overlooked in the protein kinase field. It is the site where the nucleotide and peptide/protein binding sites converge in the C:PKI complex and also in the RIα holoenzyme, which is also a pseudo-substrate inhibitor. This stable element is highly conserved in all protein kinases, and we think it is an extremely important allosteric site where the kinases differ. Thus, it is highly relevant for this set of Elife papers on kinase allostery. In parallel, we have developed the Local Spatial Pattern (LSP) alignment method for identifying Protein Residue Networks (PRNs) into a robust tool. When the Veglia team, our long-time collaborators, did their NMR analysis of the F100A mutant, which is in the αC-β4 loop, we thus decided to do the LSP analysis. The LSP results were so interesting and striking that we decided immediately to explore the motif further and to specifically compare the various crystal structures that we had solved in the past to see if indeed we had missed some changes. In addition to looking at the backbone, we decided to also look at the side chains and to compare the structures with the simulations. The results proved to be extremely informative and defined a multi-pronged approach that could be used to screen any disease mutation or alternatively as an Ala scan for any residue in any protein. I consider this to be one of the most important papers that I have published in many years. It describes a process for exploring the potential dynamic impact of any disease mutation or any point mutation. We emphasize repeatedly that the hypotheses generated from the computational screen will need to be validated experimentally, but our LSP analysis is a rapid and relatively inexpensive way to screen a set of mutations and predict which will have the greatest impact on dynamics. It is an especially powerful and robust way to identify allosteric sites as the LSP approach maps global changes of a single mutation across the entire protein. These mutants would then be prioritized for experimental follow-up. We are indeed now implementing this more comprehensive strategy in two ways. We are specifically exploring three disease mutations in the αC-β4 loop and, in parallel, are also doing a computational Ala scan of the entire loop (L95-L106); however, this is part of a separate and more comprehensive study that will take much longer. It will be the "Proof-of-Principle” of the hypotheses that we propose in our Elife paper. In addition to the LSP method, the MD simulations provide new and complementary insights into side chain dynamics in contrast to the static crystal structures. We will also begin to compare the αC-β4 loop in other kinases, specifically PKCβ2 and LRRK2, but once again this is part of a separate study and is clearly beyond the scope of this Elife paper. This focus on the αC-β4 loop is an excellent strategy that can be applied to any protein kinase. The LSP approach, however, can obviously also be applied to any protein or any motif, so it is potentially very powerful tool. We think that the impact and potential importance of this paper will be lost if it is split into two papers.

I went back to look at a recent review that we did for the Biochemical Journal on the PKA Cβ isoform, and there we also included some new primary data in the review. It was never questioned. We believe that our manuscript is so perfectly appropriate for this Elife series that is focus on allostery in kinases, and having our paper back-to-back with the Veglia NMR paper is especially important and relevant. We thus ask you will seriously consider keeping this as a single paper as part of this series on allostery.

---

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

Public Reviews:

Reviewer #1 (Public Review):

In this work Wu, J., et al., highlight the importance of a previously overlooked region on kinases: the αC-β4 loop. Using PKA as a model system, the authors extensively describe the conserved regulatory elements within a kinase and how the αC-β4 loop region integrates with these important regulatory elements. Previous biochemical work on a mutation within the αC-β4 loop region, F100A showed that this region is important for the synergistic high affinity binding of ATP and the pseudo substrate inhibitor PKI. In the current manuscript, the authors assess the importance of the αC-β4 loop region using computational methods such as Local Spatial Pattern Alignment (LSP) and MD simulations. LSP analysis of the F100A mutant showed decreased values for degree centrality and betweenness centrality for several key regulatory elements within the kinase which suggests a loss in stability/connectivity in the mutant protein as compared to the WT. Additionally, based on MD simulation data, the side chain of K105, another residue within the αC-β4 loop region had altered dynamics in the F100A mutant as compared to the WT protein. While these changes in the αC-β4 loop region seem to be consistent with the previous biochemical data, the results are preliminary and the manuscript can be strengthened (as the authors themselves acknowledge) with additional experiments. Specific comments/concerns are listed below.

1. MD simulations were carried out using a binary complex of the catalytic subunit of PKA and ATP/Mg and not the ternary complex of PKA, ATP/Mg and PKI. MD simulations carried out using the ternary complex instead of the binary complex would be more informative, especially on the role of the αC-β3 loop region in the synergistic binding of ATP/Mg and PKI.

Response 1. Thank you for your suggestion. We have included the data for the MD simulations of the ternary complex in the revised manuscript. This includes a new figure and was indeed informative (Figure 11). Text describing this simulation is also added on pages 15-17. All the changes in the revised manuscript are highlighted in red.

1. The LSP analysis shows a decrease in degree centrality for the αC-β4 loop region in the F100A mutant compared to the WT protein which suggests a gain in stability in this region for the F100A mutant (Fig. 8A). These results seem to be contradictory to the MD simulation data which shows the side chain dynamics of K105 destabilizes the αC-β4 loop region in the F100A mutant (Fig. 10B). It would be helpful if the authors could clarify this apparent discrepancy.

Response 2. In Figure 8A, the negative values of degree centrality for the αC-β4 loop region show that the value of DC is less in the WT compared to the mutant, suggesting that those regions are more stable in the mutant. This says that the mutation in the αC-β4 loop region both rigidifies the motif and alters the communication signaling networks between the two lobes.

The betweenness centrality plots (Figure 8B) also show how the connectivity between the two lobes is altered upon mutation. In the mutant the major connectors become V104 and I150 in the C-lobe, whereas connectivity was primarily governed by K72 (N-lobe) and D184 (C-lobe) in the wt C-subunit. Overall, the mutation causes rigidification of the αC-β4 loop and this leads to loss of allosteric communication between the two lobes.

The MD simulation results as shown in Figure 10B are not contradictory. This figure shows the overall dynamic profile of the protein, based on principal component analysis (PCA) using the parameter of the residual flexibility. It does not reflect a particular motif's stability or flexibility. Instead it shows that overall the protein upon mutation becomes more dynamic and can sample different conformational states, while, in contrast, the WT protein preferred a single global state of conformation. However, the LSP results showed that, compared to the other parts, the αC-β4 loop, especially V104 at the tip, becomes more stable following mutation, and this has an impact on the allosteric communication between the two lobes. We have added this information into the revised manuscript on page 14, also highlighted in red.

1. The foundation for the experiments carried out in this paper are based on previous NMR and computational data for the F100A mutant. However, the specific results and conclusions from these previous experiments are not clearly described.

Response 3. The NMR paper has been already accepted by eLIFE and here we are attaching the bioRxiv paper link, “https://www.biorxiv.org/content/10.1101/2023.09.12.557419v1.”

Reviewer #1 (Recommendations For The Authors):

In this work Wu, J., et al., draw attention to the αC-β4 loop, a previously neglected region within kinases. A comprehensive review on the important regulatory elements within the kinase along with how the αC-β4 loop (and the αE helix) integrates with these different regulatory elements is presented well. As the authors themselves acknowledge, the data presented here while promising is preliminary. Additional biochemical, NMR and computational experiments need to be carried out to assess the importance of F100, K105 and other residues in this region.

1. The authors indicate that previous computational studies predict a flip in the αC-β4 loop in the apo state. It would be helpful to have a figure showing the predicted flip as well as an explanation for the significance of this predicted flip.

Response 1. The NMR paper has been already accepted by eLIFE and here we are attaching the bioRxiv paper link, “https://www.biorxiv.org/content/10.1101/2023.09.12.557419v1.” The Figures 3 and 6 in that paper described the predicted flip in the αC-β4 loop in the apo state. We did not see a flip in any of our crystal structures, and the LSP analysis which is based on 200 ns simulations is not sufficient to see this major conformational change.

1. The authors cite previous NMR and biochemical experiments (reference 62), work that has just been submitted to eLife. Access to this work was difficult as this manuscript could not be found on the eLife website.

Response 2. The NMR paper has been already accepted by eLIFE and here we are attaching the bioRxiv paper link, “https://www.biorxiv.org/content/10.1101/2023.09.12.557419v1.”

Author Response

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

eLife assessment

Despite the importance of T follicular helper cells (Tfh cells) in vaccine-induced humoral responses, it is still unclear which type of Tfh cells (Tfh1, Tfh2, and Tfh17) is critical for generating protective humoral immunity. By using the rhesus macaques model (most similar to human), the authors have addressed this potentially important question and obtained suggestive data that Tfh1 is critical. Although being suggestive, the evidence for the importance of Tfh1 is incomplete.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Developing vaccination capable of inducing persistent antibody responses capable of broadly neutralizing HIV strains is of high importance. However, our ability to design vaccines to achieve this is limited by our relative lack of understanding of the role of T-follicular helper (Tfh) subtypes in the responses. In this report Verma et al investigate the effects of different prime and boost vaccination strategies to induce skewed Tfh responses and its relationship to antibody levels. They initially find that live-attenuated measles vaccine, known to be effective at inducing prolonged antibody responses has a significant minority of germinal center Tfh (GC-Tfh) with a Th1 phenotype (GC-Tfh1) and then explore whether a prime and boost vaccination strategy designed to induce GC-Tfh1 is effective in the context of anti-HIV vaccination. They conclude that a vaccine formulation referred to as MPLA before concluding that this is the case.

Clarification: MPLA serves as the adjuvant, and the vaccine formulation is characterized as a Th1 formulation based on the properties of the adjuvant.

Strengths:

While there is a lot of literature on Tfh subtypes in blood, how this relates to the germinal centers is not always clear. The strength of this paper is that they use a relevant model to allow some longitudinal insight into the detailed events of the germinal center Tfh (GC-Tfh) compartment across time and how this related to antibody production.

Weaknesses:

The authors focus strongly on the numbers of GC-Tfh1 as a proportion of memory cells and their comparison to GC-Tfh17. There seems to be little consideration of the large proportion of GC-Tfh which express neither CCR6 and CXCR3 and currently no clear reasoning for excluding the majority of GC-Tfh from most analysis. There seems to be an assumption that since the MPLA vaccine has a higher number of GC-Tfh1 that this explains the higher levels of antibodies. There is not sufficient information to make it clear if the primary difference in vaccine efficacy is due to a greater proportion of GC-Tfh1 or an overall increase in GC-Tfh of which the percentage of GC-Tfh1 is relatively fixed.

Response: We appreciate the reviewer's comment. Indeed, while there is substantial literature on Tfh subtypes in blood; the strength of our study lies in utilizing a relevant model to provide longitudinal insights into the dynamics of the germinal center Tfh (GC-Tfh) compartment over time and its relationship to antibody production. Regarding the concern about the comprehensive analysis of GC Tfh subsets, including GC-Tfh1, GC-Tfh17, and others not expressing CCR6 and/or CXCR3, we fully acknowledge its importance. To address this, we will conduct a detailed analysis of GC Tfh and GC Tfh1 frequencies, encompassing subsets without CCR6 and CXCR3 expression, to provide a more comprehensive view of the GC-Tfh population in our analysis.

Reviewer #2 (Public Review):

Summary:

Anil Verma et al. have performed prime-boost HIV vaccination to enhance HIV-1 Env antibodies in the rhesus macaque model. The authors used two different adjuvants, a cationic liposome-based adjuvant (CAF01) and a monophosphoryl lipid A (MPLA)+QS-21 adjuvant. They demonstrated that these two adjuvants promote different transcriptomes in the GC-TFH subsets. The MPLA+QS-21 adjuvant induces abundant GC TFH1 cells expressing CXCR3 at first priming, while the CAF01 adjuvant predominantly induced GC TFH1/17 cells co-expressing CXCR3 and CCR6. Both adjuvants initiate comparable Env antibody responses. However, MPLA+QS-21 shows more significant IgG1 antibodies binding to gp140 even after 30 weeks.

The enhancement of memory responses by MPLA+QS-21 consistently associates with the emergence of GC TFH1 cells that preferentially produce IFN-γ.

Strengths:

The strength of this manuscript is that all experiments have been done in the rhesus macaque model with great care. This manuscript beautifully indicated that MPLA+QS-21 would be a promising adjuvant to induce the memory B cell response in the HIV vaccine.

Weaknesses:

The authors did not provide clear evidence to indicate the functional relevance of GC TFH1 in IgG1 class-switch and B cell memory responses.

Response. We appreciate the recognition of our meticulous work in the rhesus macaque model and the potential of MPLA+QS-21 as an adjuvant for HIV vaccine-induced humoral immunity. We acknowledge the need to provide clearer evidence of the functional relevance of GC Tfh1 in IgG1 class-switching and B cell memory responses. We will attempt to address this concern in our revisions.

Recommendations for Authors:

Reviewer #1:

1. Is the proportion of GC-Tfh1 within GC-Tfh significantly increased in MPLA vs CAF01? The balance between Tfh1 and Tfh17 data is shown in 4C but appears quite a modest difference. Additionally, it excludes the majority of GC-Tfh since it only considers CCR6 and CXCR3 expressing cells.

Response. We have now included a comparison of the relative proportions of GC Tfh cells expressing CCR6 and CXCR3, as well as those lacking these markers. Our data now demonstrate an increased presence of Tfh1 within the GC-Tfh population when MPLA is employed at P1w2, as depicted in Figure 4D.

1. Is there any relationship between GC-Tfh17, 1/17 and non Th1/17 GC-Tfh and antibody levels? In Figure 5C only GC Tfh1 is examined making it impossible to judge if this is specific to GC-Tfh1 or a general relationship between higher total GC-Tfh and antibodies.

Response. In our revised description of the results, we have mentioned that GC Tfh frequencies correlated with antibody levels (r = 0.6, p < 0.05). However, it is important to note that this correlation was specific to the GC Tfh1 subset and was not observed with other subsets.

Other points:

1. The authors make a number of statements that rather exaggerate differences such as stating in the abstract that CAF01 induces Tfh1/17 while MPLA predominantly induces Tfh1. As shown in Figure 4C the majority of CCR6-CXCR3- GC-Tfh induced by CAF01 are GC-Tfh1 i.e. both formulations predominantly induce GC-Tfh1. Also, it is difficult to judge since the data is never provided but the predominant group of GC-Tfh appears to be CCR6-CXCR3- in both cases.

Response. We acknowledge the need for greater precision in our descriptions. In response, we have addressed this concern by providing the frequencies of CCR6-CXCR3- GC Tfh cells in Figure 4D. We have also included a comparison of the relative frequencies across the adjuvant groups in the Results section (Lines 331-338).

1. The authors use the term peripheral Tfh (pTfh), it may be better to use the more common term circulating Tfh (cTfh) to avoid confusion with T peripheral helper cells (Tph).

Response. We appreciate the reviewer's suggestion to use the more commonly accepted term "circulating Tfh (cTfh)" instead of "peripheral Tfh (pTfh)." We have incorporated this change into our manuscript to ensure clarity and avoid potential confusion with "peripheral helper cells (Tph).

1. Some further labelling of the pie chart in Figure 1G to at least specify larger groups such as Tfh2, Tfh17, Tfh1/17 would be helpful.

Response. We have incorporated the suggestion and identified cTfh2, cTfh17, and cTfh2/17 cells. We additionally now state in the legend that overlapping pie arcs correspond to specific polarized Tfh subsets denoted by arc color.

1. A gating example of the CXCR3, CCR6, CCR4 patterns in the GC Tfh would be helpful. "up to 25% of GC Tfh cells expressed CCR6" I think it is better to state the average here since 25% appears an outlier.

Response. We have now included a gating example of chemokine receptor expression, patterns in the GC Tfh. Additionally, we have revised the statement to mention the median (7%) of GC Tfh cells expressing CCR6 instead of specifying the upper limit.

1. Figure 1I, does this graph exclude triple negative cells? It's not clear from the figure legend but the numbers do not seem to add up with the graphical proportions shown in figure 1H.

Response. We have made the necessary clarification in both the results section, figure, and the figure legend to state that the Boolean analysis is based on cells expressing either CXCR3 or CCR6, thus explaining the exclusion of triple negative cells.

1. Figure 3C. Some label should be added to make clear which violins are from the CD95- and CD95+ groups. There may be too much data in this panel for p values to be legible. Either less graphs or more space may be needed.

Response. We have updated the Y axis labels in the figure to state that the violin plots show the differences in gene expression between CD95+ CD4 T cells and CD95- CD4 T cells (naive).

1. Figure 4B. Numbers attached to the gates (1, 17 etc) should be more clearly labeled Tfh1, Tfh17 etc since normally they might be expected to be gate percentages in this format. Gate percentages should also be added.

Response. We have clearly labeled the subsets as "Tfh1" and "Tfh17," making it easier for readers to interpret the figure. Additionally, we have included gate percentages in the flow plot. Furthermore, the percentages of GC Tfh subsets are now depicted in Figure 4D.

1. Overlarge and indistinct datapoint symbols are often a problem e.g. Figure 4G most of the CAF01 datapoints are merged into a single blob with no indication of where one point ends or begins. Supplementary figure 5E. Datapoint sizes are large to the extent that the lines are difficult to see. Lines indicating central tendency are often lost.

Response. We have reworked the graphs (including 4G, now 4I) to ensure clarity,

1. Generally greater care is needed with graph layout e.g. the B indicating figure 6B is on the graph of figure 6A.

Response. We have made the necessary adjustment to ensure that the letter "B" correctly corresponds to the graph in Figure 6B.

1. Figure 6J, the text seems to indicate "higher avidity with MPLA against autologous Env including V1V2 loops." However, the graph seems to indicate lower avidity for V1V2 loops? Response. We appreciate the careful observation. We have rectified this by updating the description in the results section to accurately reflect the graph, which shows higher avidity for V1V2 loops with CAF01.

2. Figure 6A. The authors state that significantly higher IgG1 was induced but Figure 6A seems to be the only graph lacking an indication of statistical significance.

Response. We have made the necessary adjustment to ensure that significance symbol is depicted in Figure 6A.

1. Brackets indicating significance are often unclear. e.g. in Figure 4B MPLA graph there are three groups and a single multipoint bracket with a single result making it unclear which groups are being compared.

Response. We have added clarification to the legend. It now states that the temporal comparisons in GC Tfh subsets for each vaccine group are made in relation to frequencies at baseline. This revision provides a clear reference point for the significance comparisons and ensures that readers can easily understand which groups are being compared.

Reviewer #2:

Overall, the manuscript is well-written and addresses an important issue. However, further investigation is warranted to understand how the MPLA+QS-21 induced GC TFH1 influenced on memory B cell response. This manuscript only showed the correlation between GC TFH1 and antibody responses. If the authors explain adjuvant preference in memory B cell responses, this manuscript could be more considerable for publication.

1. This reviewer recommends that the author provide more evidence to indicate the functional relevance of GC TFH1 in IgG1 class-switch and B cell memory responses. Some evidence supports that IFN-γ controls the antigen-specific IgG1 responses in humans, but it is still controversial. The author also suggests the involvement of IL-21, but this is also an open question even in the human system. This is also the case in the memory responses. There is no direct link between IFN-γ and memory B cell responses in the human system. The authors need more evidence of how GC TFH1 cell development has more advantages in IgG1 and memory responses than GC TFH1 /17 cells. I believe an antibody blockade of cytokines would be a possible strategy to prove these questions.

Response. We appreciate the reviewer's valuable suggestion to provide more evidence regarding the functional relevance of GC Tfh1 cells in IgG1 class-switch and B cell memory responses. It is indeed important to establish a direct link between GC Tfh1 cells and these responses, particularly in the context of cytokine skewing. The suggestion of antibody blockade studies to mechanistically link the modulation of the inflammatory milieu to Tfh differentiation and subsequent antibody functions is important. However, we must acknowledge that these studies are currently beyond the scope of our work. We have included this as a limitation in our study, recognizing the need for further studies to address these important questions.

1. In Fig.5, the authors use different scales to indicate the IgG antibody titer. A shows the log scale, while B shows the linear scale. Moreover, the differences are minimal, even though the authors indicated a significant difference. I am not sure this difference is meaningful.

Response. To clarify, we used a log scale in Figure 5A to demonstrate temporal changes over the course of vaccination. In Figure 5B, where we are comparing differences across vaccine regimens at week 30, a linear scale was deemed more appropriate, as it allows for a clear representation of the approximately two-fold difference observed. We fully acknowledge that to establish the biological significance of the observed difference, challenge studies will be essential.

Author Response

Reviewer #1 (Public Review):

This article proposes a new statistical approach to identify which of several experimenter-defined strategies best describes a biological agent's decisions when such strategies are not fully observable by choices made in a given trial. The statistical approach is described as Bayesian but can be understood instead as computing a smoothed running average (with decay) of the strategies' success at matching choices, with a winner-take-all inference across the rules. The article tests the validity of this statistical approach by applying it to both simulated agents and real data sets in mice and humans. It focuses on dynamically changing environments, where the strategy best describing a biological agent may change rapidly.

The paper asks an important question, and the analysis is well conducted; the paper is well-written and easy to follow. However, there are several concerns that limit the strength of the contribution. Major concerns include the framing of the method, considerations around the strategy space, limitations in how useful the technique may be, and missing details in analyses.

Reviewer #2 (Public Review):

In this study, the goal is to leverage the power of Bayesian inference to estimate online the probability that any given arbitrarily chosen strategy is being used by the decision-maker. By computing the trial-by-trial MAP and variance of the posterior distribution for each candidate strategy, the authors can not only see which strategy is primarily being used at every given time during the task and when strategy changes occur but also detect when the target rule of a learning task becomes the front-running strategy, i.e., when successful learning occurs.

Strengths:

1) The proposed approach adds to recent methods for capturing the dynamics of decision-making at finer temporal resolution (trials) (Roy et al., 2021; Ashwood et al., 2022) but it is novel and differs from these in that it is suited especially well for analyzing when learning occurs, or when a rule switches and learning must recommence, and it does not necessitate large numbers of trials.

2) The manuscript starts with a validation of the approach using synthetic data and then is applied to datasets of trial-based two-alternative forced choice tasks ranging from rodent to non-human primate to human, providing solid evidence of its utility.

3) Compared to classic procedures for identifying when an animal has learned a contingency which typically needs to be conservative in favor of better accuracy, this method retrieves signs of learning happening earlier (~30 trials earlier on average). This is achieved by identifying the moment (trial) when the posterior probability of the correct "target" rule surpasses the probability of all other strategies. Having greater temporal precision in detecting when learning happens may have a very significant impact on studies of the neural mechanisms of learning.

4) This approach seems amenable to testing many different strategies depending on the purpose of the analysis. In the manuscript, the authors test target versus non-target strategies (correct versus incorrect) and also in another version of the analysis, they test what they call "exploratory" strategies.

5) One of the main appeals of this method is its apparent computational simplicity. It necessitates only updating on every trial the parameters of a beta distribution (prior distribution for a given strategy) with the evidence that the behavior on trial was either consistent or inconsistent with the strategy. Two scalars, the mode of the posterior (MAP) and the inverse of the variance, are all that are required for identifying the decision criterion (highest MAP and if tied lowest variance) and the learning criterion (first trial where MAP for target strategy is higher than chance).

Weaknesses:

1) It seems like a limitation of this approach is that the candidate strategies to arbitrate between must be known ex-ante. It is not clear how this approach could be applied to uncover latent strategies that are not mixtures of the strategies selected.

2) Different strategies may be indistinguishable from each other and thus it may not be possible to distinguish between them. Similarly, the fact that two strategies seem to be competing for the highest MAP doesn't necessarily mean that those are correct strategies and perhaps interchangeable as the manuscript seems to suggest.

3) The decay parameter is a necessary component to make the strategy selection non-stationary and accommodate data sets where the rules are changing throughout the task. However, the choice of the decay parameter value bounds does not seem very principled. Having this parameter as a free-parameter adds a flexibility that seems to have significant effects on when the strategy switch is detected and how stable the detected switch is.

4) This method is a useful approach for arbitrating between strategies and describing the behavior with a temporal precision that may prove important for studies attempting to tie these precise events to changes in neural activity. However, it seems limited in its explanatory power. In its current form, this method does not provide a prediction of the probability to transition from one strategy to another. And, because the MAP of different strategies may be close at any given moment, it is hard to imagine using this approach to tease out the different "mental states" that represent each strategy being at play.

The reviewers’ detailed comments, not shared here, helped us considerably to improve the paper, and we thank the reviewers for their time here. We are unsure of the merits of sharing public reviews of a paper that has now changed considerably from the version that these reviews address. Nonetheless we shall address some key points of potential misunderstanding here.

“The statistical approach is described as Bayesian but can be understood instead as computing a smoothed running average (with decay) of the strategies' success at matching choices, with a winner-take-all inference across the rules.“

This is inaccurate. The algorithm performs sequential Bayesian updates on the evidence for and against the use of each strategy considered; for a given strategy i, its output at each trial is a fully parameterised posterior distribution over the probability of that strategy being used by the subject.

We are careful in the paper to separate the algorithm’s output from our further use of that output. To plot and analyse the output we often make use of the maximum a posteriori (MAP) estimate from each posterior. Other choices are of course possible, and we discuss them in the text.<br /> In one set of simulations we quantify the results using a decision rule that chooses the strategy with the highest MAP - this is presumably the “winner-takes-all inference” in the quoted text. We do not use this anywhere else in the paper, including the analyses of the 4 datasets, and so do not consider it as part of our method, but one possible use of the output of the algorithm.

“Major concerns include the framing of the method, considerations around the strategy space, limitations in how useful the technique may be, and missing details in analyses”

Our goal for this paper was to develop a computationally lightweight, trial-resolution, Bayesian approach to tracking the probability of user-specified strategies, so that we can capture the observer’s evidence for learning or for the features driving exploratory choice (e.g. whether subjects are responding to losses or wins; are they responding to cues or choice etc). The above quote reflects their detailed review comments, where we felt this reviewer wanted a solution to a different problem, that of a parameterised latent model of strategy use: while a perfectly valid research goal, this was not what we addressed here.

“1) It seems like a limitation of this approach is that the candidate strategies to arbitrate between must be known ex-ante. It is not clear how this approach could be applied to uncover latent strategies that are not mixtures of the strategies selected.”

The problem of knowing which strategies to analyse in advance only applies when running our algorithm in real-time. The fact that it could be run in real-time on modest computing hardware is to us one of its strengths, so we consider this a good problem to have.

As noted above, rather than determine latent strategies, our goal was to build an observer model that allows users to specify whatever strategy they wanted in order to answer their scientific question(s) of their data. For example, to define when a particular rule has been learnt; or to look for changes in response to particular features of the environment, such as a cue, or to a drug treatment or other intervention.

2) Different strategies may be indistinguishable from each other and thus it may not be possible to distinguish between them. Similarly, the fact that two strategies seem to be competing for the highest MAP doesn't necessarily mean that those are correct strategies and perhaps interchangeable as the manuscript seems to suggest.

As noted above, this is an observer model, and it is thus necessarily true that there are strategies for which the observer does not have sufficient evidence to distinguish. For example, a subject who continually chooses the rewarded left-hand lever will be doing both a strategy of “go left” and of “win-stay” in response to their choice. The inability to distinguish strategies is a property of the data, not of the algorithm. Also as noted above, we do not here consider the competition between strategies.

3) The decay parameter is a necessary component to make the strategy selection non-stationary and accommodate data sets where the rules are changing throughout the task. However, the choice of the decay parameter value bounds does not seem very principled. Having this parameter as a free-parameter adds a flexibility that seems to have significant effects on when the strategy switch is detected and how stable the detected switch is.

The revised manuscript draws together the existing simulations and analysis of the method to directly address this point, showing that there is a principled range of the decay parameter in which the algorithm should operate. The Discussion also points out that this is no different to a free parameter than any frequentist approach to strategy analysis, which must choose some time windows over which to compute the frequentist probability.

4) This method is a useful approach for arbitrating between strategies and describing the behavior with a temporal precision that may prove important for studies attempting to tie these precise events to changes in neural activity. However, it seems limited in its explanatory power. In its current form, this method does not provide a prediction of the probability to transition from one strategy to another. And, because the MAP of different strategies may be close at any given moment, it is hard to imagine using this approach to tease out the different "mental states" that represent each strategy being at play.

As noted above, this is an observer model and does not intend to infer mental states. The goal is to make accurate statements about observable behaviour. We agree that an interesting extension to this approach would be to model the transitions between strategies, and had already outlined this in the Discussion.

Author Response

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

REVIEWER 1:

Reviewer 1 stated: “The authors have provided strong evidence that high levels of auxin exposure perturb feeding behavior, survival rates, lipid metabolism, and gene expression patterns, providing a cautionary note for the field in using this technology. They also concluded that “overall, the experiments were suitably designed with appropriate sample size and data analysis methods.”

Reviewer 1 provided the following recommendations for improvement, which are addressed below:

Point 1: “Although authors showed that auxin causes gene expression changes including the possible alteration of Gal4 expression levels, no cell-type-specific data is provided. It would be informative to the Drosophila field if the authors could examine major Gal4 drivers in their expression levels, such as the ones used in studying metabolism and oogenesis.”

We agree with the reviewer that cell-type specific Gal4 expression should be thoroughly analyzed by scientists in the community wishing to use the current auxin-inducible gene expression system (AGES) in their studies; however, those analyses are beyond the scope of our manuscript. There are many tissues and cell types that are used to study metabolism and oogenesis (e.g., muscle, adipocytes, oenocytes, multiple cell types in the gut, multiple cell types in the ovary), and Gal4 expression patterns could be different depending on age, sex, and diet. It is therefore impossible for us to pinpoint one or two key tissues important for regulating lipid levels and would be a significant investment of time. We believe that each researcher should thoroughly check the Gal4 expression pattern for their specific tissue of interest under their normal standard or altered food conditions. As this reviewer pointed out, our current study provides a cautionary note for the field in using this technology. Nevertheless, we have provided a reference to a recent micropub (Hawley et al; PMID: 37396791) which describes neuronal Gal4 expression patterns comparing the AGES and temporal and regional gene expression targeting (TARGET) systems and updated the text in lines 539-544 of the revised manuscript.

Point 2: “Although the authors briefly mentioned aging research, feeding behavior, and lipid metabolism, RNA-seq data are provided only for short-term treatment (2 days). The ovary phenotype was examined with long-term treatment (15 days). It would be informative if the authors could also show other long-term treatment data.”

We respectfully point out to the reviewer that a 5-day auxin feeding assay was provided in Figure S4H, which reproduces the data provided for the 2-day auxin treatment. In addition, the original AGES paper (McClure et al, PMID: 35363137) provided adult survival data that extended to 80 days. In our updated manuscript, we have provided data for a 10-day auxin treatment that also addresses Point #4 below regarding whether the decrease in lipid levels upon auxin feeding is reversible.

Point 3: “The auxin used in this work is a more water-soluble version and at a high concentration (10 mM). In the C. elegans system, researchers are using a much lower concentration of auxin typically at 1 mM. Therefore, the discussion of their results in terms of potential impacts on other experimental systems should be done carefully. It would be helpful to know what impacts might be observed at a lower concentration of auxin. The recommendation would be that the authors add the 1 mM auxin data point to key elements of their analysis.”

The concentration of 10 mM auxin used in our study is the recommended dose to use in Drosophila (see McClure et al) and has been used in at least one additional study (Hawley et al). We also would like to point out that other systems (e.g., C. elegans and mice) have many differences in physiology and therefore the concentration of auxin used to elicit a response are likely to be different (e.g., 71.4 mM final concentration is the recommended concentration used in mice; Macdonald et al; PMID: 35736539). We have merely suggested that researchers using auxin for protein degradation should carefully check whether lipid levels (or other physiological processes of interest) are altered upon auxin feeding (or soaking) alone compared to a 0 mM auxin control. The text in lines 467-470 has been altered to reflect this. In addition, the specific recommended dose for Drosophila is highlighted and referenced in multiple places (i.e., methods and results and discussion) throughout the updated text.

Point 4: “Another related question is whether these detected changes are reversible or not after exposure to auxin at different concentrations. This would be informative for researchers to better design their temporally controlled experiments.”

We thank the reviewer for this suggestion and have provided the data in Figure S4I. Briefly, we found that after a 5-day treatment of auxin, removal of auxin for an additional 5 days does not recover lipid levels to those of control animals never exposed to auxin.

Point 5: “It would also be helpful to know whether spermatogenesis is affected or not.”

Although this would be an interesting developmental process to determine if affected by auxin exposure, we believe that these analyses are beyond the scope of the current manuscript.

Point 6: “A few other points include changing the nomenclature and validating some of the key genes shown in Figure 3 using quantitative RT-PCR experiments with the tissues where the affected genes are known to be expressed and functional.”

We thank the reviewer for this suggestion. We have provided qRT-PCR analysis using whole body samples and this data is now provided in the new Figure S8. We used whole-body samples for the qRT-PCR analysis because it would be impossible to pinpoint the specific tissue the differentially regulated genes are required for eliciting the response to auxin exposure. For example, according to Flybase (flybase.org) GstE3 transcripts are moderately to highly expressed in 15 of the 23 cell types annotated by the Fly Cell Atlas project (Li et al; PMID: 35239393).

REVIEWER 2:

Reviewer 2 stated: “The authors provide evidence of several Auxin effects. Experiments are suitably designed with appropriate sample size and data analysis methods.”

This reviewer expressed the following concerns, which are addressed below:

Point 1: “The provided information is limited and not very helpful for many applications. For example, although authors briefly mentioned aging research, feeding behavior, and lipid data, RNA seq data are provided only for short-term (48 hours) treatment. Especially, since ovary phenotype was examined with long-term treatment (15 days), authors should also show other data for long-term treatment as well.”

Please see our response to Point #2 of Reviewer 1 regarding long-term treatment experiments. Furthermore, although the ending timepoint for the ovarian analyses is 15 days, we also provide analysis at shorter time points (e.g., daily analysis for egg counts, 5 and 10 day timepoints for fixed sample analyses).

Point 2: “Although the authors show that Auxin causes a change in gene expression patterns and suggests the possible alteration of Gal4 expression levels, no cell-type-specific data is provided. It would be informative if the authors could examine the expression level of major Gal4 drivers. Authors should discuss how severe these changes are by comparing them with other treatments or conditions, such as starvation or mutant data (ideally, comparing with reported data or their own data if any?).”

Please see our response to Point #1 from Reviewer 1.

REVIEWER 3:

Reviewer 3 stated that they “found the study to be carefully done” and “this study will be of interest to researchers using the Drosophila system, especially those focusing on fatty acid metabolism or physiology.”

Reviewer 3 also had the following minor points, which are addressed below:

Point 1: “Auxin, actually 1-naphthaleneaceid acid here, which is a more water-soluble version of auxin (indole-3-acetic acid) is used at what I consider to be a high concentration-10 mM. The problem I have is that the authors are discussing their results in terms of potential impacts on other experimental systems. At least for C. elegans, I think this is not a reasonable extension of the current dataset. In the C. elegans system, researchers are using 1 mM auxin. The authors note that their RNA-seq results suggest a xenobiotic response. Could this apparent xenobiotic response be due to a metabolic byproduct following auxin administration at high concentrations? Figure S1A shows that there is quite a robust transcriptional response at 1 mM auxin. It would be helpful to know what impacts might be observed at this lower concentration in which the transcriptional induction could be used in the context of biologically meaningful experiments. The recommendation would be that the authors add the 1 mM auxin data point to key elements of their analysis.”

Regarding the comparisons to other model organisms, we refer to our response to Point #3 from Reviewer 1. We also point out that although there is a robust response to 1 mM auxin using the 3.1Lsp2-Gal4 driver, 1 mM is not sufficient for a robust response using additional driver lines in Drosophila (see Hawley et al). It is possible that the xenobiotic response is due to using the recommended dose of auxin (McClure et al).

However, given the fact that researchers are currently using the 10 mM dose for experiments in Drosophila, we believe that the 10 mM transcription dataset is the most relevant. Nevertheless, we do agree that researchers who choose to use lower concentrations of auxin in the future should carefully look at whether any transcriptional induction alters physiological processes of interest.

Point 2: “This reviewer was confused by the genetic nomenclature the authors use. The authors have chosen to use the designation 3.1Lsp2-Gal4 (3.1Lsp2-Gal4AID). I think this is potentially confusing because a reader might think that it is the Gal4 transcription factor that is the direct target of auxin- and TIR1-mediated protein degradation, as I initially did. Rather, it is the Gal80 repressor protein that is the direct target. The authors might consider a nomenclature that is more reflective of how this system works. It would also be helpful if the full genotypes of strains were included in each figure legend.”

We apologize for the nomenclature confusion in our original submission. We have changed our “AID” nomenclature throughout the manuscript to “AGES,” which is the nomenclature used in McClure et al. We respectfully note that the traditional nomenclature for using the temperature-sensitive Gal80 system is Gal80ts or adding the “ts” superscript to the Gal4 line used (e.g., 3.1Lsp2ts).

Point 3: “The RNA-seq dataset does not appear to be validated by RT-PCR experiments. The authors should consider validating some of the key genes shown in Figure 3 using quantitative RT-PCR experiments, potentially adding a 1 mM auxin data point.”

Please see our response to Point #6 to Reviewer 1.

REVIEWER 4:

Reviewer 4 stated: “Overall, the experiments were well-designed and carefully executed. The results were quantified with appropriate statistical analyses. The paper was also well-written and the results were presented logically.”

RECOMMENDATIONS FOR THE AUTHORS:

We have further addressed reviewer recommendations below. Thank you again, for your critique of our manuscript.

REVIEWER 2:

As I mentioned in my public review, long-term treatment data would be especially helpful. Examining changes in the expression level of major Gal4 lines is also informative.

Please see our responses to Points #1 and #2 to Reviewer 1 in the “Public Reviews” section. Although examination of Gal4 expression patterns is extremely important, we believe that these analyses should be carefully performed on a case-by-case basis in the future for labs who wish to continue to use this methodology.

REVIEWER 4:

I feel addressing #2 would be a great addition to the current version, while #1 and #3 could be addressed in future studies or by researchers who are interested in these processes.

Recommendation 1: “Both the metabolomics and transcriptome analyses were done using the whole animals, would it be more informative if these were done using specific tissue/organs such as the adult adipose tissue?”

Please see our response to Points #1 and #6 to Reviewer 1 in the “Public Reviews” section.

Recommendation 2: “Another related question is whether these detected changes are reversible or not after exposure to auxin? This would be informative for researchers to better design their temporally controlled experiments.”

We thank the reviewer for this suggestion and the analysis for this experiment is now provided in Figure S4I.

Recommendation 3: “Is spermatogenesis affected at all?”

We respectfully point out that many processes in spermatogenesis (as well as other biological processes) are affected by feeding (e.g., starvation) and would be extremely time consuming to carefully perform the analyses with the rigor required. We agree with Reviewer 4 and believe that this would be best to be performed on a case-by-case examination in the future.

Author Response

Our responses to the reviewers to go into the published pre-print. We thank the reviewers for their encouraging and thoughtful comments. These are good points that we would like to comment on as follows:

Reviewer 1:

Some important and interesting data are missing. For example, whether the gene therapy can extend the life span of these mutants? The overall in vivo voiding function is missing. AAV9/HSPE2 expression in the bladder wall is not shown.

A. Our study was not designed to determine whether gene therapy can improve life span of the Hpse2 mutant mice. We know that the mutant mice usually become ill after the first month of life and can die. However, we wanted to study the mice when they were generally well so that there would be no confounding effects on the bladder physiology caused by general ill health. Indeed, a recent study of Hpse2 inducible deletion in adult mice has shown evidence of exocrine pancreatic insufficiency (Kayal et al., PMID 37491420). We are currently exploring the status of the pancreas in our non-conditional juvenile Hpse2 mice, and whether gene transfer into the pancreas is possible.

B. We strongly agree that in vivo voiding studies will be important it the future, and suggest in vivo cystometry is the gold standard for this but is currently beyond the remit of this study.

C. It is correct that in this paper we have focussed on gene transduction into the pelvic ganglia, because the evidence is mounting that this is a neurogenic disease. Our ex vivo physiological studies show predominantly neurogenic defects that are corrected by the gene therapy. A detailed study of the bladder body is an interesting idea, in terms of possible transgene expression and detailed histology, and is something we will pursue in future studies.

Review 2:

Weaknesses include a lack of discussion of the basis for differences in carbachol sensitivity in Hpse2 mutant mice, limited discussion of bladder tissue morphology in Hpse2 mutant mice, some questions over the variability of the functional data, and a need for clarification on the presentation of statistical significance of functional data.

A. Yes, it is interesting that untreated male mutant mice have an increased bladder body contraction to carbachol compared with WT males. In a previous paper (Manak et al., 2020) we performed quantitative western blots for the M2 and M3 receptors and found levels were similar in mutants to the WTs, thus the increased sensitivity probably lies post-receptor.

B. A detailed study of the bladder body is an interesting idea, in terms of possible transgene expression and detailed histology, and is something we will pursue in future studies.

C. We have reported in our physiology graphs what we find. We do find some variability, particularly at lower frequencies, but our conclusions depend on analyses of the whole curve, which depend on multiple frequencies and show the expected overall pattern of frequency-dependent relaxation.

D. Thank you, the stats for Figure 8 will be corrected in the final version.

Reviewer 3:

Single-cell analysis of mutants versus control bladder, urethra including sphincter. This would be great also for the community.

A. Yes, in future we are very interested in using a single cell sequencing approach to look at the mutant, WT and rescued pelvic ganglia. In relation to this, there is a recent proof-of-principle paper pre-print in WT mouse pelvic ganglia, which suggests this may be feasible (Sivori et al., 2023).

Detailed tables showing data from each mouse examined.

B. In theory, it would be very interesting to correlate the strength of human gene transduction into the pelvic ganglia, with, for example, the effect on a physiological parameter. However, in general we used different sets of mice for these techniques so at the present we don’t have this information.

Use of measurements that are done in vivo (spot assay for example). This sounds relatively simple.

C. We strongly agree that in vivo voiding studies will be important it the future, and suggest in vivo cystometry is the gold standard for this but is currently beyond the remit of this study.

Assessment of viral integration in tissues besides the liver (could be done by QPCR).

D. This is an important point, and suggest the pancreas is a particularly interesting target for future studies. a recent study of Hpse2 inducible deletion in adult mice has shown evidence of exocrine pancreatic insufficiency (Kayal et al., PMID 37491420). We are currently exploring the status of the pancreas in our non-conditional juvenile Hpse2 mice, and whether gene transfer into the pancreas is possible.

Discuss subtypes of neurons that are present and targeted in the context of mutants and controls.

E. The make-up of the pelvic ganglia in Hpse2 mutant mice is a fascinating question. Future analysis using scRNA-Seq may be the most effective way to answer this question and is a molecular approach we are looking to pursue in the future.

Author Response

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

eLife assessment

This paper reports the development of SCA-seq, a new method derived from PORE-C for simultaneously measuring chromatin accessibility, genome 3D and CpG DNA methylation. Most of the conclusions are supported by convincing data. SCA-seq has the potential to become a useful tool to the scientific communities to interrogate genome structure-function relationships.

Public Reviews:

Reviewer #1 (Public Review):

In this work, Xie et al. developed SCA-seq, which is a multiOME mapping method that can obtain chromatin accessibility, methylation, and 3D genome information at the same time. SCA-seq first uses M.CviPI DNA methyltransferase to treat chromatin, then perform proximity ligation followed by long-read sequencing. This method is highly relevant to a few previously reported long read sequencing technologies. Specifically, NanoNome, SMAC-seq, and Fiber-seq have been reported to use m6A or GpC methyltransferase accessibility to map open chromatin, or open chromatin together with CpG methylation; Pore-C and MC-3C have been reported to use long read sequencing to map multiplex chromatin interactions, or together with CpG methylation. Therefore, as a combination of NanoNome/SMAC-seq/Fiber-seq and Pore-C/MC-3C, SCA-seq is one step forward. The authors tested SCA-seq in 293T cells and performed benchmark analyses testing the performance of SCA-seq in generating each data module (open chromatin and 3D genome). The QC metrics appear to be good and I am convinced that this is a valuable addition to the toolsets of multi-OMIC long-read sequencing mapping.

The revised manuscript addressed most of my questions except my concern about Fig. S9. This figure is about a theory that a chromatin region can become open due to interaction with other regions, and the author propose a mathematic model to compute such effects. I was concerned about the errors in the model of Fig. S9a, and I was also concerned about the lack of evidence or validation. In their responses, the authors admitted that they cannot provide biological evidence or validations but still chose to keep the figure and the text.

The revised Fig. S9a now uses a symmetric genome interaction matrix as I suggested. But Figure S9a still have a lot of problems. Firstly, the diagonal of the matrix in Fig. S9a still has many 0's, which I asked in my previous comments without an answer. The legend mentioned that the contacts were defined as 2, 0 or -2 but the revised Fig. S9a only shows 1,0, or -1 values. Furthermore, Fig. S9b,9c,9d all added a panel of CTCF+/- but there is no explanation in text or figure legend about these newly added panels. Given many unaddressed problems, I would still suggest deleting this figure.

In my opinion, this paper does not need Fig. S9 to support its major story. The model in this figure is independent of SCA-seq. I think it should be spinoff as an independent paper if the authors can provide more convincing analysis or experiments. I understand eLife lets authors to decide what to include in their paper. If the authors insist to include Fig. S9, I strongly suggest they should at least provide adequate explanation about all the figure panels. At this point, the Fig. S9 is not solid and clearly have many errors. The readers should ignore this part.

We appreciate the reviewer for raising these concerns regarding Fig. S9. After careful consideration, we have decided to address your concerns by deleting Fig. S9 and the corresponding text from the manuscript. We understand your point that the model presented in Fig. S9 is independent of SCA-seq and may require additional evidence and validation to be presented in a separate paper.

We agree that it is important to maintain the integrity and accuracy of the manuscript, and we appreciate your feedback in helping us make this decision.

Reviewer #2 (Public Review):

In this manuscript, Xie et al presented a new method derived from PORE-C, SCA-seq, for simultaneously measuring chromatin accessibility, genome 3D and CpG DNA methylation. SCA-seq provides a useful tool to the scientific communities to interrogate the genome structure-function relationship.

The revised manuscript has clarified almost of the concerns raised in the previous round of review, though I still have two minor concerns,

1. In fig 2a, there is no number presented in the Venn diagram (although the left panel indeed showed the numbers of the different categories, including the numbers in the right panel would be more straightforward).

We appreciate the reviewer for pointing out the need for clarification in the Venn diagram in Fig 2a. We have added the numbers to Venn diagram.

1. The authors clarified the discrepancy between sfig 7a and sfig 7g. However, the remaining question is, why is there a big difference in the percentage of the cardinality count of concatemers of the different groups between the chr7 and the whole genome?

We apologize for the confusion regarding the difference in the percentage of the cardinality count of concatemers between chr7 and the whole genome in figures S7a and S7g. The difference arises because the chr7 cardinality count only considers the intra-chromosome segments that are adjacent to each other on a SCA-seq concatemer, while the whole genome cardinality count includes both intra-chromosome and inter-chromosome segments.

In the case of a SCA-seq concatemer that contains both intra-chromosome junctions and inter-chromosome junctions, the whole genome cardinality count will be greater than the intra-chromosome cardinality count. This explains the difference in the percentages between chr7 and the whole genome in figures S7a and S7g.

To better clarify the definition of intra-chromosome cardinality, we have added an illustrative graph in figure S7a. In the updated figure S7a, the given exemplary SCA-seq concatemer has a whole genome cardinality of 4 and a chr7 intra-chromosome cardinality of 3.

Author Response

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

eLife assessment

This important study reports investigation of the dynamics of PKA at the single-cell level in in vitro and in epithelia in vivo. Using different fluorescent biosensors and optogenetic actuators, the authors dissect the signaling pathway responsible for PKA waves, finding that PKA activation is a consequence of PGE2 release, which in turn is triggered by calcium pulses, requiring high ERK activity. The evidence supporting the claims is solid. At this stage the work is still partly descriptive in nature, and additional measurements would increase the strength of mechanistic insights and physiological relevance.

We deeply appreciate Dr. Alejandro San Martín and Dr. Jonathan Cooper and the reviewers. Each comment is valuable and reasonable. We will revise our paper as much as possible.

We have described what we will do for the reviewer’s comments one by one in the below section.

Reviewer #1 (Recommendations For The Authors):

1. Even though the phenomenon of PGE2 signal propagation is elegantly demonstrated and well described, the whole paper is mostly of descriptive nature - the PGE2 signal is propagated via intercellular communication and requires Ca transients as well as MAPK activity, however function of these RSPAs in dense epithelium is not taken into consideration. What is the function of these RSPAs in cellular crowding? - Does it promote cell survival or initiate apoptosis? Does it feed into epithelial reorganization during cellular crowding? Still something else? The authors discuss possible roles of this phenomenon in cell competition context, but show no experimental or statistical efforts to answer this question. I believe some additional analysis or simple experiment would help to shed some light on the functional aspect of RSPAs and increase the importance of all the elegant demonstrations and precise experimental setups that the manuscript is rich of. Monolayer experiments using some perturbations that challenge the steady state of epithelial homeostasis - drug treatments/ serum deprivation/ osmotic stress/ combined with live cell imaging and statistical methods that take into account local cell density might provide important answers to these questions. The authors could consider following some of these ideas to improve the overall value of the manuscript.

We would like to thank the reviewer’s comment. Although we have intensively tried to identify the physiological relevance of RSPA, we could not detect the function at present.

In the case of MDCK, the treatment of NSAIDs, which cancels RSPA, did not affect its cell growth, ERK wave propagation during collective migration, migration velocity, cell survival, or apoptosis. In mouse epidermis, the frequency of RSPA was NOT affected by inflammation and collective cell migration, evoked by TPA treatment and wound, respectively.

Notably, RSPA also occurs in the normal epidermis, implying its relevance in homeostasis. However, at the current stage, we believe that the PGE2 dynamics and its regulation mechanism in the normal epidermis would be worth reporting to researchers in the field.

1. In the line 82-84 the authors claim: "We found that the pattern of cAMP concentration change is very similar to the activity change of PKA, indicating that a Gs protein-coupled receptor (GsPCR) mediates RSPA". In our opinion, this conclusion is not well-supported by the results. The authors should at least show that some measurements of the two patterns show correlation. Are the patterns of cAMP of the same size as the pattern of PKA? Do they have the same size depending on cell density? Do they occur at the same frequency as the PKA patterns, depending on the cell density? Do they have an all or nothing activation as PKA or their activation is shading with the distance from the source?

We have modified the text (line85)

“Although the increment of the FRET ratio was not so remarkable as that of Booster-PKA, Wwe found that the pattern of cAMP concentration change is very similar to the activity change of PKA, indicating that a Gs protein-coupled receptor (GsPCR) mediates RSPA. This discrepancy may be partially explained by the difference in the dynamic ranges for cAMP signaling in each FRET biosensor (Watabe2020). “

1. In general, the absolute radius of the waves is not a good measurement for single-cell biology studies, especially when comparing different densities or in vivo vs in vitro experiments. We suggest the authors add the measurement of the number of the cells involved in the waves (or the radius expressed in number of cells).

We appreciate the reviewer’s comment. We have analyzed our results to demonstrate the number of cells as in Fig2E, which would be easy for readers to understand.

1. In 6D, the authors should also show the single-cell trajectories to understand better the correlation between PKA and ERK peaks. Is the huger variability in ERK activity ratio dues to different peak time or different ERK activity levels in different cells? The authors should show both the variability in the time and intensity.

We have added a few representative results as Fig. S4.

1. In lines 130-132, the authors write, "This observation indicates that the amount of PGE2 secretion is predetermined and that there is a threshold of the cytoplasmic calcium concentration for the triggered PGE2 secretion". How could the author exclude that the amount of PGE2 is not regulated in its intensity as well? For sure, there is a threshold effect regarding calcium, but this doesn't mean that PGE2 secretion can be further regulated, e.g. by further increasing calcium concentration or by other mechanisms.

We agree with the reviewer’s comment. We have modified the text.

1. The manuscript shows that not all calcium transients are followed by RSPAs. Does the local cell density/crowding increase the probability of overlap between calcium transients and RSPAs?

We appreciate the reviewer’s comment. We have also hypothesized the model. However, we did not see the correlation that the reviewer pointed out. Currently, the increment of the RSPA frequency at high density is partially caused by the increment of calcium transients.

Reviewer #2 (Recommendations For The Authors):

1. The work is hardly conclusive as to the actual biological significance of the phenomenon. It would be interesting to know more under which physiological and pathological conditions PGE2 triggers such radial PKA activity changes. It is not well explained in which tissues and organs and under what conditions this type of cell-to-cell communication could be particularly important.

The greatest weakness of the study seems to be that the biological significance of the phenomenon is not clearly clarified. Although it can be deduced that PKA activation has many implications for cell signaling and metabolism, the work lacks the actual link to physiological or pathological significance.

We deeply appreciate the reviewer’s comment. Similar to the reseponse of reviewer#1, although we have intensively tried to identify the physiological relevance of RSPA, we could not detect the function.

On the other hand, we believe that the PGE2 dynamics and its regulation mechanism in the normal epidermis would be worth reporting to researchers in the field.

1. The authors do not explain further why in certain cells of the cell clusters Ca2+ signals occur spontaneously and thus trigger the phenomenon. What triggers these Ca2+ changes? And why could this be linked to certain cell functions and functional changes?

At this moment, we do not have a clear answer or model for the comment although the calcium transients have been reported in the epidermis (https://doi.org/10.1038/s41598-018-24899-7). Further studies are needed and we will pursue this issue as a next project.

1. What explains the radius and the time span of the radial signal continuation? To what extent are these factors also related to the degradation of PGE2? The work could be stronger if such questions and their answers would be experimentally integrated and discussed.

We agree with the reviewer’s comment. Although we have intensively studied that point, we have omitted the results because of its complications. In HeLa cells, but not MDCK cells, we demonstrate the meaning of the radius of RSPA (https://pubmed.ncbi.nlm.nih.gov/37813623/)

1. The authors could consider whether they could investigate the subcellular translocation of cPLA2 in correlation with cytosolic Ca2+ signals using GFP technology and high-resolution fluorescence microscopy with their cell model.

Actually, we tried to monitor the cPLA2 translocation using GFP-tagged cPLA2. However, the translocation of GFP-cPLA2 was detected, only when the cells were stimulated by calcium ionophore. At this point, we have concluded that the quantitative analysis of cPLA2 translocation would be difficult.  

Reviewer #3 (Recommendations For The Authors):

1. "The cell density in the basal layer is approximately 2x106 cells cm-2, which is markedly higher than that in MDCK cells (Fig. 2D). It is not clear whether this may be related to the lower frequency (~300 cm-2 h-1) and smaller radius of RSPA in the basal layer cells compared to MDCK cells (Fig. 2E)." Wasn't the relationship with cell density the opposite, higher density higher frequency? Isn't then this result contradicting the "cell density rule" that the authors argue is there in the in vitro system? The authors need to revise their interpretation of the data obtained.

We agree with the reviewer’s comment. Currently, we do not find the "cell density rule" in mouse epidermis. It would be difficult to identify common rules between mouse epidermis and MDCK cells. However, although it is descriptive, we believe it is worth comparing the MDCK results at this moment.

1. Similarly, the authors over conclude on the explanation of lack of change in the size of RSPA size when the change in fluorescence for the calcium reporter surpasses a threshold by saying that "This observation indicates that the amount of PGE2 secretion is predetermined and that there is a threshold of the cytoplasmic calcium concentration for the triggered PGE2 secretion." First, the study does not really measure directly PGE2 secretion. Hence, there is no way that they can argue that the level of PGE2 secreted is "predetermined". Instead, there could be an inhibitory mechanism that is triggered to limit further activation of PGE2 signaling/PKA in neighboring cells.

We agree with the reviewer’s comment. We have omitted the context.

1. To rule out a transcription-dependent mechanism in the apparent cell density-regulated sensitivity to PGE2, the authors need to inhibit transcription. We agree that our RNA-seq analysis would not 100% rule out the transcription-dependent mechanism. However, we believe that shutting down all transcription will show a severe off-target effect that indirectly affects the calcium transients and the PGE2-synthetase pathway. Therefore, our conclusion is limited.

4) EGF is reported to increase the frequency of RSPA but the change shown in Fig. 6F is not statistically significant, hence, EGF does not increase RSPA frequency in their experiments.

We have toned down the claim that EGF treatment increases the frequency (line172).

"Accordingly, the addition of EGF faintly increased the frequency of RSPA in our experiments, while the MEK and EGFR inhibitors almost completely abrogated RSPA (Fig. 6F), representing that ERK activation or basal ERK activity is essential for RSPA.“

1. The Discussion section is at times redundant with the results section. References to figures should be kept in the Results section.

We would like to argue in opposition to this comment. For readers, we believe that the reference to figures would be helpful and kind. However, if eLife recommends removing the reference from the Discussion section, we will follow the publication policy.

1. "Notably, the propagation of PKA activation, ~100 μm/min (Fig. 1H), is markedly faster than that of ERK activation, 2-4 μm/min (Hiratsuka et al., 2015)." The 2 kinase reporters are based on different molecular designs. Thus, it does not seem appropriate to compare the kinetics of both reporters as a proxy of the comparison of the kinetics of propagation of both kinases.

We think that we should discuss the comparison of the activity propagation between ERK and PKA. First, among many protein kinases, only ERK and PKA activities have been shown to spread in the epithelial cells. Second, both pathways are considered to be intercellular communication. Finally, crosstalk between these two pathways has been reported in several cells and organs.

1. In Figure 1E it is unclear what is significantly different from what. Statistical analysis should be added and reporting of the results should reflect the results from that analysis.

2. In Figure 3F and G the color coding is confusing. In F pink is radius and black is GCaMP6 and in G is RSPA+ and - cells. The authors should change the color to avoid ambiguity in the code.

We have amended the panels.

1. In Fig. 5C, how do they normalize per cell density if they are measuring radius of the response?

In Fig5C, we just measure the increment of FRET ratio in the view fields.

1. In Fig. 5D, what is the point of having a label for PTGER3 if data were not determined (ND)?

We have added what N.D. means.

“N.D. represents Not Detected.”

1. It is important to assess whether ERK activation depends of PGE2 signaling to better place ERK in the proposed signaling pathway. In fact, the authors argue that "ERK had a direct effect on the production of PGE2." But it could be that ERK is downstream PGE2 signaling instead.

It could be possible in other experimental conditions via EP1 and/or EP3 pathways. However, we never detected an effect of RSPA on ERK activity by analyzing our imaging system. In addition, treatment with NSAIDs or COX-2 depletion, which completely abolishes RSPA, did not affect ERK wave propagation. Thus, in our context, we concluded that ERK is not downstream of PGE2. This notion is also supported by the NGS results in Fig. 5D.

We have refrained from discussing the pathway of PGE2-dependent ERK activation because it would be redundant.

1. The authors need to explain better what they mean by "AND gate" if they want to reach a broad readership like that of eLife

We have modified the legend to explain the “AND gate” as much as possible (line639).

“Figure 7: Models for PGE2 secretion.

The frequency of calcium transients is cell density-dependent manner. While the ERK activation wave is there in both conditions. Because both calcium transient and ERK activation are required for RSPA, the probability for PGE2 secretion is regulated as “AND gate”. ”

1. In Fig. 5D, "The average intensity of the whole view field of mKate2 or mKOκ, at 20 to 30 min after the addition of PGE2, was applied to calculate the mKate2/mKOκ ratio." But this means that overlapping/densely plated cells in high density will show stronger changes in fluorescence. This should be done per cell not per field of view. It is obvious that the higher density will have more dense/brighter signal in a given field of view.

We are sorry for the confusion. The cell density does not affect the FRET ratio, although the brightness could be changed. A typical example is Fig1D. Thus, we are sure that our procedures represent the PKA activity in plated cells.

1. In Fig. 6B the authors need to explain how were the "randomly set positions" determined.

We have modified the legend section as below (line618).

“The ERK activities within 10 µm from the center of RSPA and within 10 µm from randomly set positions with a random number table generated by Python are plotted in the left panel. Each colored dot represents an average value of an independent experiment.”

1. Sentences 314-318 are repeated in 318-322.

We deeply appreciate the reviewer’s comment and have amended

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Here, Boor et al focus on the regulation of daf-7 transcription in the ASJ chemosensory neurons, which has previously been shown to be sensitive to a variety of external and internal signals. Interestingly, they find that soluble (but not volatile) signals released by food activate daf-7 expression in ASJ, but that this is counteracted by signals from the ASIC channels del-3 and del-7, previously shown to detect the ingestion of food in the pharynx. Importantly, the authors find that ASJ-derived daf-7 can promote exploration, suggesting a feedback loop that influences locomotor states to promote feeding behavior. They also implicate signals known to regulate exploratory behavior (the neuropeptide receptor PDFR-1 and the neuromodulator serotonin) in the regulation of daf-7 expression in ASJ. Additionally, they identify a novel role for a pathway previously implicated in C. elegans sensory behavior, HEN1/SCD-2, in the regulation of daf-7 in ASJ, suggesting that the SCD-2 homolog ALK may have a conserved role in feeding and metabolism.

Strengths:

The studies reported here, particularly the quantitation of gene expression and the careful behavioral analysis, are rigorously done and interpreted appropriately. The results suggest that, with respect to food, DAF-7 expression encodes a state of "unmet need" - the availability of nearby food to animals that are not currently eating. This is an interesting finding that reinforces and extends our understanding of the neurobiological significance of this important signaling pathway. The identification of a role for ASJ-derived daf-7 in motor behavior is a valuable advance, as is the finding that SCD-2 acts in the AIA interneurons to influence daf-7 expression in ASJ.

We appreciate the Reviewer 1’s thoughtful assessment of our work and inference that the expression of daf-7 encodes internal state corresponding to “unmet need.” Based on comments of Reviewer 1 and other reviewers, we have revised the title, abstract, and parts of the discussion to highlight not only the functional contribution of daf-7 expression in the ASJ neurons to behavioral state, but also the remarkable correlation between gene expression and internal state driving foraging behavior.

Weaknesses:

A limitation of the work is that some mechanistic relationships between the identified signaling pathways are not carefully examined, but this provides interesting opportunities for future work.

To enable the reader to begin to infer the relative contributions of the identified signaling pathways to the circuitry coupling distinct bacterial cues to foraging behavior, we have added data for the analysis of DAF-7 expression in the ASJ neurons in the tph-1 and pdfr-1 mutants in the complete absence of food. Our current leaning is that multiple pathways, including those we have begun to characterize here, may function in parallel to influence DAF-7 expression and internal state driving foraging behavior. Future work to explore this further is certainly of interest.

A minor weakness concerns the experiment in which daf-7 is conditionally deleted from ASJ. This is an ideal approach for probing the function of daf-7, but these experiments seem to be carried out in the well-fed, on-food condition in which control animals should express little or no daf-7 in ASJ. Thus, the experimental design does not allow an assessment of the role of daf-7 under conditions in which its expression is activated (e.g., in animals exposed to un-ingestible food).

The interpretation of genetic analysis in the complete absence of food is complicated by what we think are multiple parallel pathways that function to strongly promote roaming, as indicated in the prior work of Ben Arous et al. Our observation that the conditional deletion of daf-7 from the ASJ pair of neurons confers altered roaming behavior on a lawn of bacterial food supports physiological ongoing role for dynamic daf-7 expression from the ASJ neurons even in the presence of bacterial food that may contribute to the control of transitions between foraging states and the persistence of roaming and dwelling states.

To demonstrate the functional contribution of DAF-7 expression from the ASJ neuron pair during constitutive expression favoring roaming, we examined the roaming behavior of scd2(syb2455) animals that carry a gain-of-function mutation in scd-2 that promotes roaming and how the selective deletion of daf-7 from the ASJ neurons in the scd-2(syb2455) genetic background influences roaming behavior. This new experiment supports a model in which DAF-7 expression from the ASJ neurons contributes to the increased roaming behavior exhibited by scd-2(syb2455) animals. The new experiment is added as Figure 4I.

An additional minor issue concerns the interpretation of the scd-2 experiments. The authors' findings do support a role for scd-2 signaling in the activation of daf-7 expression by un-ingestible food, but the data also suggest that scd-2 signaling is not essential for this effect, as there is still an effect in scd-2 mutants (Figure 4B).

Considering that most of previous Figure 4B is redundant with previous Figure 4D, we removed previous Figure 4B. Our current Figure 4 has redesignated previous Figure 4D as 4B. We have also added qualification to the text to indicate that other pathways may modulate the daf-7 expression response to ingested food in parallel to SCD-2 signaling.

Reviewer #2 (Public Review):

Summary:

In this work, Boor and colleagues explored the role of microbial food cues in the regulation of neuroendocrine-controlled foraging behavior. Consistent with previous reports, the authors find that C. elegans foraging behavior is regulated by the neuroendocrine TGFβ ligand encoded by daf-7. In addition to its known role in the neuroendocrine/sensory ASI neurons, Boot and colleagues show that daf-7 expression is dynamically regulated in the ASJ sensory neurons by microbial food cues - and that this regulation is important for exploration/exploitation balance during foraging. They identify at least two independent pathways by which microbial cues regulate daf-7 expression in ASJ: a likely gustatory pathway that promotes daf-7 expression and an opposing interoceptive pathway, also likely chemosensory in nature but which requires microbial ingestion to inhibit daf-7 expression. Two neuroendocrine pathways known to regulate foraging (serotonin and PDF-1) appear to act at least in part via daf-7 induction. They further identify a novel role for the C. elegans ALK orthologue encoded by scd-2, which acts in interneurons to regulate daf-7 expression and foraging behavior. These results together imply that distinct cues from microbial food are used to regulate the balance between exploration and exploitation via conserved signaling pathways.

Strengths:

The findings that gustatory and interoceptive inputs into foraging behavior are separable and opposing are novel and interesting, which they have shown clearly in Figure 1. It is also clear from their results that removal of the interoceptive cue (via transfer to non-digestible food) results in rapid induction of daf-7::gfp in ASJ, and that ASJ plays an important role in the regulation of foraging behavior.

We thank Reviewer 2 for underscoring the modulation of neuroendocrine gene expression in the ASJ neuron pair by distinct gustatory and interoceptive inputs derived from bacterial food that we show in Figure 1.

The role of the hen-1/scd-2 pathway in mediating the effects of ingested food is also compelling and well-interpreted. The use of precise gain-of-function alleles further supports their conclusions. This implies that important elements of this food-sensing pathway may be conserved in mammals.

We thank Reviewer 2 for emphasizing the implications of our study on SCD-2/ALK as well as the generation and use of gain-of-function scd-2 alleles based on oncogenic mutations in ALK.

Weaknesses:

What is less clear to me from the work at this stage is how the gustatory input fits into this picture and to what extent can it be strongly concluded that the daf-7regulating pathways that they have identified (del-3/7, 5-HT, PDFR-1, scd-2) act via the interoceptive pathway as opposed to the gustatory pathway.

It follows from the work of the Flavell lab that del-3/7 likely acts via the interoceptive pathway in this context as well but this isn't shown directly - e.g. comparing the effects of aztreonam-treated bacteria and complete food removal to controls. The roles of 5-HT and PDFR-1 are even a bit less clear. Are the authors proposing that these are entirely parallel pathways? This could be explained in better detail.

We have added additional data regarding daf-7 expression from the ASJ neurons in the complete absence of food in the different mutant backgrounds noted by Reviewer 2. Data regarding daf-7 expression in the ASJ neurons under three distinct conditions—ingestible bacterial food, non-ingestible bacterial food, and the complete absence of food—enable the pairwise comparison of mutant data that allows for inference regarding the relative contributions of the genes to the interoceptive vs. gustatory pathways. In particular, effects on the interoceptive pathway can be inferred from the comparison of daf-7 expression on ingestible vs. non-ingestible food, whereas effects on the gustatory pathway can be inferred from the comparison of daf-7 expression on non-ingestible food vs. the absence of food (newly added).

These additional data are most informative for del-3; del-7 (Figure 1H), where the added data corroborate a role for these genes in the interoceptive pathway, consistent with the findings of the Flavell lab. Specifically, the observation that daf-7 expression levels are equivalent between wild-type and del-3;del-7 animals when there is no ingestible food (either no food or non-ingestible food conditions) suggest that DEL-3 and DEL-7 are functioning specifically to sense ingested food.

For pdfr-1, the analysis of the gain-of-function allele suggest that this pathway may have a greater relative effect on the gustatory pathway compared with the interoceptive pathway (Figure 3D). The robust upregulation seen in the pdfr-1(syb3826) animals between animals on ingestible and non-ingestible food, suggests that the interoceptive regulation is functional in these mutants, while the lack of upregulation between no-food and noningestible-food conditions suggests that the gustatory pathway is affected.

The observations with the 5-HT biosynthesis mutant are most consistent with serotonin signaling affecting daf-7 expression in the ASJ neurons through a mechanism that is parallel to the gustatory and interoceptive inputs into daf-7 expression in the ASJ neurons, as tph1(n4622) animals appear to have an elevated baseline expression of daf-7 in the ASJ neurons while retaining sensitivity to both gustatory and interoceptive food cues (Figure 3B).

The data with scd-2 are consistent with a role in the epistatic interoceptive pathway, considering the roughly equivalent levels of daf-7 expression in the ASJ neurons under all food conditions in scd-2(syb2455) animals (Figure 4B). However it is difficult to exclude the possibility that SCD-2 functions in both pathways or parallel to the gustatory and interoceptive inputs.

While we agree that our genetic analysis alone cannot distinguish between genes acting in parallel or directly in serial with the gustatory or interoceptive inputs, our data do establish that signaling through SCD-2, 5-HT or PDFR-1-dependent pathways can act on the same gene expression and signaling node (i.e. daf-7 expression in the ASJ neurons) to modulate the effects of bacterial food inputs on foraging behavior, with the effects on daf-7 expression in the ASJ neurons in scd-2, tph-1 and pdfr-1 mutants correlating with their effects on roaming and dwelling behaviors.

It would also be helpful to elaborate more on why the identified transcriptional positive feedback loop is predicted to extend roaming state duration - as opposed to some other mechanism of increasing roaming such as increased probability of roaming state initiation. This doesn't seem self-evident to me.

Given that animals can exist in only two states, the increased probability of roaming state initiation would present as shorter dwelling states, which we do not see for daf-7 mutants. As described in Flavell, et al., 2013, a decreased fraction of time roaming can be attributed to longer dwelling states, shorter roaming states, or both. Our positive feedback loop is predicted to extend roaming states because of the predicted effect of DAF-7 on stabilizing the roaming state.

Related to this point is the somewhat confusing conclusion that the effects of tph-1 and pdfr-1 mutations on daf-7 expression are due to changes in ingestion during roaming/dwelling. From my understanding (e.g. Cermak et al., 2020), pharyngeal pumping rate does not reliably decrease during roaming - so is it clear that there are in fact lower rates of ingestion during roaming in their experiments?

This is an interesting point. Despite consistent pumping rates, we still believe that roaming animals ingest less food than dwelling animals. For instance, dwelling animals are localized to areas with bacterial food, while roaming animals might traverse patches with no food where pumping does not result in food ingestion.

If so, why does increased roaming (via tph-1 mutation) result in further increases in daf-7 expression in animals fed aztreonam-treated food (Fig 3B)?

This is possibly because although roaming animals are eating less, when animals are on non-ingestible food, they’re not eating at all, resulting in further daf-7 upregulation.

Alternatively, there could be a direct signaling connection between the 5-HT/PDFR-1 pathways and daf-7 expression which could be acknowledged or explained.

Yes, this is certainly possible. We do not propose that all of the difference in daf-7 expression is due to changes in foraging behavior, but rather we are highlighting further instances of the correlation between daf-7 expression in the ASJ neurons and roaming. For instance, in the case of our tph-1 mutants, we see a relatively modest effect on daf-7 expression in the ASJ neurons but a large difference in the fraction of time roaming. This suggests that the magnitude of change in one (daf-7 expression in ASJ or roaming) does not predict the magnitude of the change in the other, but rather that they trend in the same direc<on.

Reviewer #3 (Public Review):

Summary:

In this interesting study, the authors examine the function of a C. elegans neuroendocrine TGF-beta ligand DAF-7 in regulating foraging movement in response to signals of food and ingestion. Building on their previous findings that demonstrate the critical role of daf-7 in a sensory neuron ASJ in behavioral response to pathogenic P. aeruginosa PA14 bacteria and different foraging behavior between hermaphrodite and male worms, the authors show, here, that ingestion of E. coli OP50, a common food for the worms, suppresses ASJ expression of daf-7 and secreted water-soluble cues of OP50 increases it. They further showed that the level of daf-7 expression in ASJ is positively associated with a higher level of roaming/exploration movement. Furthermore, the authors identify that a C. elegans ortholog of Anaplastic Lymphoma Kinase, scd-2, functions in an interneuron AIA to regulate ASJ expression of daf-7 in response to food ingestion and related cues. These findings place the DAF-7 TGF-beta ligand in the intersection of environmental food conditions, food intake, and foodsearching behavior to provide insights into how orchestrated neural functions and behaviors are generated under various internal and external conditions.

Strengths:

The study addresses an important question that appeals to a wide readership. The findings are demonstrated by generally strong results from carefully designed experiments.

We thank Reviewer 3 for the comments and interest in the work.

Weaknesses:

However, a few questions remain to provide a complete picture of the regulatory pathways and some analyses need to be strengthened. Specifically,

1. The authors show that diffusible cues of bacteria OP50 increase daf-7 expression in ASJ which is suppressed by ingestible food. Their results on del-3 and del-7 suggest that NSM neuron suppresses daf-7 ASJ expression. What sensory neurons respond to bacterial diffusible cues to increase daf-7 expression of ASJ? Since ASJ is able to respond to some bacterial metabolites, does it directly regulate daf-7 expression in response to diffusible cues of OP50 or does it depend on neurotransmission for the regulation? Some level of exploration in this question would provide more insights into the regulatory network of daf-7.

The focus of our study has been on the modulation of daf-7 expression in the ASJ neurons by distinct bacterial food cues and the downstream neuroendocrine circuitry that is influenced. The question of whether bacterial cues are directly sensed by the ASJ neurons remains unresolved by our study. However, we have previously demonstrated that the daf-7 expression in the ASJ neurons induced by P. aeruginosa metabolites is likely the result of direct detection by the ASJ neurons. We would also note (and have added to the manuscript) the observation of Zaslaver et al. (2015), in which increased calcium transients were observed in the ASJ neurons in response to the withdrawal of E. coli OP50 supernatant, which is consistent with our observations of the effect of a soluble bacterial food signal on daf-7 expression in the ASJ neurons.

1. The results including those in Figure 2 strongly support that daf-7 in ASJ is required for roaming. Meanwhile, authors also observe increased daf-7 expression in ASJ under several conditions, such as non-ingestible food. Does non-ingestible food induce more roaming?

Yes, this has been published by Ben Arous, et al., 2009. Figure 3C shows increased roaming on aztreonam-treated food. We have added specific mention of this in the text.

It would complete the regulatory loop by testing whether a higher (than wild type) level of daf-7 in ASJ could further increase roaming. The results in pdf-1 and scd-2 gain-of-function alleles support more ASJ leads to more roaming, but the effect of these gain-of-function alleles may not be ASJ-specific and it would be interesting to know whether ASJ-specific increase of daf-7 leads to a higher level of roaming. In my opinion, either outcome would be informative and strengthen our understanding of the critical function of daf-7 in ASJ demonstrated here.

We looked at roaming in animals with a ptrx-1::daf-7 cDNA transgene in a wild-type background and did not see changes in the fraction of time animals roam. However, multiple experimental factors could contribute to our inability to detect an effect, including relative promoter strength and context of other variables that alter daf-7 expression. Nevertheless, our data confirmed that ASJ neuron-specific expression of daf-7 cDNA can increase roaming in a daf-7 mutant background (Figure 2B).

We have also included an experiment (Figure 4I) looking at roaming in the scd-2(syb2455) gain-of-function animals in animals with daf-7 deleted from the ASJ neurons. These results suggest that part of the increased roaming seen in these scd-2(syb2455) animals is specifically due to increased daf-7 expression in the ASJ neurons.

1. The analyses in Figure 4 cannot fully support "We further observed that the magnitude of upregulation of daf-7 expression in the ASJ neurons when animals were moved from ingestible food to non-ingestible food was reduced in scd-2(syb2455) to levels only about one-fourth of those seen in wild-type animals (Figure 4D)...", because the authors tested and found the difference in daf-7 expression between ingestible and non-ingestible food conditions in both wild type and the mutant worms. The authors did not analyze whether the induction was different between wild type and mutant. Under the ingestible food condition, ASJ expression of daf-7 already looks different in scd-2(syb2455).

We appreciate the reviewer pointing out our lack of clarity in discussing our analysis of the data. The 4x difference represents the difference in fold change from ingested to noningested food in wild type and scd-2(syb2455) backgrounds. For wild-type animals, daf-7 expression in the ASJ neurons on non-ingestible food is 8.1-times higher on non-ingestible food than on ingestible food. In scd-2(syb2455) animals, this difference is 1.7 times. We have clarified this in the text.

1. The authors used unpaired two-tailed t-tests for all the statistical analyses, including when there are multiple groups of data and more than one treatment. In their previous study Meisel et al 2014, the authors used one-way ANOVA, followed by Dunnett's or Tukey's multiple comparison test when they analyzed daf-7 expression or lawn leaving in different mutants or under different bacterial conditions. It is not clear why a two-tailed t-test was used in similar analyses in this study

We have performed one-way ANOVAs for all comparisons included, and the results were largely consistent with what we found for t-tests. Ultimately, for our analysis we were most interested in pairwise comparisons and decided that t-tests would be most appropriate.

*Reviewer #1 (Recommendations For The Authors):

Line 170: For clarity, I suggest editing this to: "When animals are removed from edible food but are still exposed to soluble food signals, upregulation of daf-7..."

We have edited this in the text and appreciate the suggestion.

The authors report that pdfr-1(syb3826) was retrieved from "a screen done in parallel to this work." syb3826 is a Suny Biotech allele, suggesting that this screen may not have been done in the authors' lab but rather outsourced. Some additional details might be useful.

This S325F allele was originally recovered as qd385 in an EMS screen performed in our lab. syb3826 is an independently generated Suny Biotech allele we ordered to confirm that the S325F substitution in PDFR-1 was responsible for our phenotypes. This has been clarified in the text.

Line 210: Please provide a citation for the screen that identified hen-1(qd259).

This is the first time the allele is being published. The screen is included in two theses from our lab, Meisel 2016 and Park 2019.

Line 214: It would be useful here to also mention the previously identified role of scd2 in sensory integration.

Yes, we have added this to the text. Additionally, we have included a couple of sentences in the discussion about how previous studies that have found a role for SCD-2 in sensory integration may instead be detecting the role for SCD-2 in food sensing, as many of the assays used for sensory integration are also sensitive to nutritional status of the animals.

Line 271: Please provide a citation for the sex differences in food-leaving behavior (Lipton 2004 PMID 15329389 is the first careful characterization of this).<br /> We have added this to the text.

Author Response

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

Reviewer #2 (Recommendations For The Authors):

The evidence provided in this study reflects important discoveries on language lateralisation and most of the conclusions of this paper are supported by evidence. However, there are several areas regarding the characteristics of participants tested, hypotheses/predictions and the type of analysis, that need to be clarified and/or corrected.

1. There is a substantial disconnection between the introduction and the methods/results section.

One reason is because of lack of consistency. One example refers to the fact that, in the introduction, only IFC is mentioned. However, the analyses carried out to examine neural activity in different groups focused on IFC as well as other brain regions related to inhibitory control. However, these areas were not mentioned at all in the introduction. Second and related to the above, the rationale for conducting certain types of analyses is not specified. Some brain analyses focus on IFC only. Instead, other analyses focus on several areas.

Another weakness is that there is not sufficient detail regarding the hypotheses/predictions and the specific types of analyses chosen to test these hypotheses/predictions. For example, there is no mention of resting state fMRI data in the introduction, but later we discover that this type of data was collected and analyzed. Even a brief mention of the inclusion of resting state data in the introduction would be beneficial. Along the same lines, by reading the methods section we find out that VBM analyses were conducted. But it is unclear why. What was the purpose of this data analysis? This should be clarified briefly in the introduction and then in the methods section. It remains unclear why resting state results would be particularly informative for addressing the research question of this study. Task-related brain connectivity seems a more appropriate choice. Additionally, it is not explained what comparisons and outcomes would be informative/expected to distinguish between the two mentioned competing hypotheses. This should be made clear.

Another aspect that lacks clarity is the authors' predictions when investigating the relationship "between the lateralization of both functions and inter-hemispheric structural-functional connectivity, as well as with behavioural markers of certain clinical conditions that have been related with atypical lateralization". The hypotheses are completely omitted in this section.

Thank you for bringing this to our attention. We concur with Reviewer #2 that our introduction was somewhat lacking in detail and assumed too much prior knowledge on the part of the reader. This, together with a lack of a clear presentation of our tested hypotheses, made the introduction have a poor connection with both the results and discussion sections, which hindered the understanding of the paper.

As a result, we have made some additions to enhance the exposition of the following areas: (1) the causal and statistical hypotheses of lateralization (Lines 55-65); and (2) the hypotheses regarding subclinical markers of neurological disorders and the corpus callosum (Lines 90-104).

Furthermore, we have extensively revised the final paragraph of the introduction (Lines 105-121) to provide a clearer and more coherent linkage between the drivers presented during the introduction, our hypotheses, and the subsequent analyses.

1. It is important to provide more information on the language background of the participants. Were the participants in this study Catalan-Spanish bilinguals? If so, it is crucial for the authors to mention this.

Language background of the participants has been added to the corresponding section (Lines 138-145).

In fact, previous studies, including several publications from the authors themselves (Garbin et al., 2010; Rodríguez-Pujadas et al., 2013; Anderson et al., 2018), have shown that there are qualitative differences between bilinguals and monolinguals in the neural circuitry underlying executive control. Across all these studies, it was consistently reported that bilingual individuals, when engaged in non-linguistic inhibitory control tasks, recruited a broader network of left-brain regions associated with language control, including the left IFC, in comparison to monolingual individuals. If the participants in this study were indeed bilinguals, it raises concern if the aim of the study is to generalize the conclusions on lateralization effects beyond the bilingual population.

Rodríguez-Pujadas, A., Sanjuán, A., Ventura-Campos, N., Román, P., Martin, C., Barceló, F., … & Ávila, C. (2013). Bilinguals use language-control brain areas more than monolinguals to perform non-linguistic switching tasks. PLoS One, 8(9), e73028.

Garbin, G., Sanjuan, A., Forn, C., Bustamante, J. C., Rodríguez-Pujadas, A., Belloch, V., ... & Ávila, C. (2010). Bridging language and attention: Brain basis of the impact of bilingualism on cognitive control. NeuroImage, 53(4), 1272-1278.

Anderson, J. A., Chung-Fat-Yim, A., Bellana, B., Luk, G., & Bialystok, E. (2018). Language and cognitive control networks in bilinguals and monolinguals. Neuropsychologia, 117, 352-363.

Indeed, we have thoroughly reported that, when compared to monolinguals, bilinguals exhibit a significant implication of left brain regions during switching and inhibition tasks. So, this is a legitimate concern. Unfortunately, the society from which our participants were drawn is primarily bilingual, encompassing both active and passive bilinguals. The monolingual sample in those previous studies consisted of university students originating from predominantly monolingual regions of Spain. Given this context, it is unsurprising that the current study has a rather limited number of monolinguals (n=8), with only 2 displaying atypical language lateralization. Thus, we cannot provide a reliable answer to the role of bilingualism status in our data. Consequently, we have included a comment on this limitation on the discussion (Lines 504-512).

1. Regarding the methods section, I have the following specific queries. The first is about the control condition in the verb generation task. I find it puzzling that the 'task' and 'control' conditions differ in terms of the number of words uttered. Could the authors please provide further clarification on this?

Thank you for raising this question. Regarding the control condition, it is important to note that the design of this task drew inspiration from previously published verb generation tasks for fMRI (Benson et al., 1999; Fitzgerald et al., 1997) and PET (Petersen et al., 1988). In the fMRI tasks, a fixation cross served as the control condition, while the PET study used word repetition as the control. We acknowledged that a mere fixation cross might not adequately control for the movement and visual-related activations inherent in the verb generation task. Conversely, word repetition could potentially engage the default mode network due to the repetition of the same simple task, which might not be suitable for a control condition, and it could be overly linguistic because it involves a word. Consequently, we aimed to strike a balance by employing a control condition that consisted of reading letters. This approach allowed us to control for movement and vision factors without invoking semantics. Thus, after careful consideration, we ultimately opted on the reading of two letters to equate the response to the vocalization length of generating a verb.

Although we understand the concern of single vs. two vocalizations, it is worth emphasizing that this version of the verb generation task had undergone prior testing to assess its suitability for determining language lateralization in both healthy and clinical populations (Sanjuan et al., 2010). In fact, this task has been an integral component of our lab’s standard presurgical assessment protocol, which has been used for nearly two decades in individually evaluating language function in over 500 patients with central nervous system lesions.

Benson, R. R., Fitzgerald, D. B., Lesueur, L. L., Kennedy, D. N., Kwong, K. K., Buchbinder, B. R., Davis, T. L., Weisskoff, R. M., Talavage, T. M., Logan, W. J., Cosgrove, G. R., Belliveau, J. W., & Rosen, B. R. (1999). Language dominance determined by whole brain functional MRI in patients with brain lesions. Neurology, 4(52), 798–809.

Fitzgerald, D. B., Cosgrove, G. R., Ronner, S., Jiang, H., Buchbinder, B. R., Belliveau, J. W., Rosen, B. R., & Benson, R. R. (1997). Location of Language in the Cortex: A Comparison between Functional MR Imaging and Electrocortical Stimulation. AJNR Am J Neuroradiol, 18, 1529–1539.

Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, M., & Raichle, M. E. (1988). Positron emission tomographic studies of the cortical anatomy of single-word processing. Nature, 331(18), 585–589.

Sanjuán, A., Bustamante, J. C., Forn, C., Ventura-Campos, N., Barrós-Loscertales, A., Martínez, J. C., Villanueva, V., & Ávila, C. (2010). Comparison of two fMRI tasks for the evaluation of the expressive language function. Neuroradiology, 52(5), 407–415. https://doi.org/10.1007/s00234-010-0667-8

Second, it is mentioned that some participants were excluded from different tasks due to technical issues or time constraints. It is important to ensure that all the results can be attributed to the exact same sample of participants across all tasks.

We absolutely agree that excluding participants can be problematic when presenting the results of multiple sets of analyses. Therefore, we repeated all analyses while excluding the 7 participants that lacked resting-state data. All results remained virtually identical, with a few minor exceptions:

1) Region-wise analysis of the stop-signal task: Hemisphere × Group effect in the preSMA region is significant (uncorrected P = 0.019), but it does not survive Bonferroni correction (corrected P = 0.076)

2) Voxel-wise analysis of the stop-signal task: The Thalamus + STN and Caudate clusters are significant at the voxel level, but do not survive the cluster-based FWE correction. They do survive FDR correction, though.

3) Correlation between SPQ score and LI of the stop-signal task: This correlation weakens just behind statistical significance, with a P value of 0.053.

4) Correlation between reading variables and LIs of both tasks: Severe drops in P values are evident between both LIs and reading length accuracy (P = .111 and .133), as well as between verb generation LI and reading familiarity accuracy (P = .111). However, the association between the stop-signal LI and the reading length time is now significant (r = −.229, P = .042).

According to this, we have included this statement in the methods section: (Lines 218-220).“It is important to highlight that the exclusion of these seven participants across all analyses does not notably impact the overall results.“

It is unclear how the authors have estimated the RTs results from the practice trials. This requires more explanation. Also, why was the median used for the Go Reaction Time instead of the mean, when calculating the individual SSRT?

We adapted the procedure used by Xue et al. (2008), implementing their approach to calculate SSRT. This has been elaborated further (Lines 227-230), together with the use of practice trials (Lines 233-236).

Xue, G., Aron, A.R., and Poldrack, R.A. (2008). Common Neural Substrates for Inhibition of Spoken and Manual Responses. Cerebral Cortex 18, 1923–1932. 10.1093/CERCOR/BHM220.

On a final note, information about the different types of pre-processing and data analysis is all reported in the same paragraph. I think using subsections would increase the intelligibility of the section.

Thank you for this suggestion. We have added subsections in both the ‘image processing’ and ‘statistical analyses’ sections.

1. Data analysis and Interpretation of the results. It is unclear how the mean BOLD signal was extracted to conduct ROI analysis (Marsbar?).

Thank you for ponting this out. Indeed, we were not very accurate in the description of this procedure. We extracted the first eigenvariate via the VOI function within SPM12. This has been included in Lines 291-293.

I feel uneasy about the way results are corrected for multiple comparisons. For instance, it is mentioned that in the ROI analysis, all p-values were FDR-corrected for four comparisons, but it is unclear why. The correct procedure for supporting conclusions about the effect of specific brain would be to have 'brain region' (n=4) as another within-subject factor. Furthermore, the one-tailed correlation is appropriate but only when testing for the possibility of a relationship in one direction and completely disregarding the possibility of a relationship in the other direction. However, this does not seem to be the case here (see Introduction), so a two-tailed correlation would be more appropriate.

We agree with Reviewer #2 that presenting this analysis as a single MANOVA that includes a ‘Region’ factor is a more accurate approach. Consequently, we have made the aforementioned correction in the methods section (Lines 357-364) and the results section (Lines 395-406). The LI-LI one-tailed correlation was also changed to a two-tailed correlation in the methods section (Line 383), the results section (Line 417), and Figure 2 (Line 886).

I am quite confused about using the term interhemispheric connectivity to refer to the volume of the genu, body and splenium of the corpus callosum. In fact, the volumes of genu, body and splenium of the corpus callosum do not reflect a measure of how strongly RH and LH IFC are connected to each other.

We agree that using the term ‘interhemispheric connectivity’ when referring to callosal volume may be somewhat misleading. We have replaced every instance of this terminology throughout the paper.

Furthermore, it is unclear why in a set of analyses (ROI and whole brain analyses) the authors focus on brain responses in different ROIs but instead, in connectivity measures the focus is only on IFC.

Our initial rationale was to focus on regions that are prominently involved in language, particularly the IFC, for examining inter-hemispheric connectivity at rest.

However, upon more careful consideration, it is true that the preSMA is also implicated in the language network (Labache et al., 2018), and certain studies have reported an impact of STN stimulation on specific language skills (for a review, see Vos et al., 2021). Consequently, we have incorporated these two regions into the resting-state analysis, along with subsequent correlations with LIs (Table 1 and Lines 118, 321-322 & 449-452).

Labache, L., Joliot, M., Saracco, J., Jobard, G., Hesling, I., Zago, L., Mellet, E., Petit, L., Crivello, F., Mazoyer, B., & Tzourio-Mazoyer, N. (2018). A SENtence Supramodal Areas AtlaS (SENSAAS) based on multiple task-induced activation mapping and graph analysis of intrinsic connectivity in 144 healthy right-handers. Brain Structure and Function 2018 224:2, 224(2), 859–882. https://doi.org/10.1007/S00429-018-1810-2

Vos, S. H., Kessels, R. P. C., Vinke, R. S., Esselink, R. A. J., & Piai, V. (2021). The Effect of Deep Brain Stimulation of the Subthalamic Nucleus on Language Function in Parkinson’s Disease: A Systematic Review. Journal of Speech, Language, and Hearing Research, 64(7), 2794–2810. https://doi.org/10.1044/2021_JSLHR-20-00515

Minor corrections/comments:

It is unclear why in figure caption 1, the conjunction maps are mentioned even if formal conjunction analysis was not conducted.

This poor choosing of words has been replaced to ‘overlapping maps’.

Line 382. VHMC should be VMHC.

Fixed. Thank you.

Line 334. This sentence and especially its relationship with the results is not clear at all. What do you mean by 'This finding is consistent with previous reports showing that cognitive deficits appear only in specific cognitive domains'?

This has been clarified (Lines 521-525).

Author Response

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

Public Reviews:

Throughout the study, there is insufficient information about how experiments were performed and how often (imaging, pull-downs etc), how data was acquired, modified and analysed (especially imaging data, see below), how statistical analyses were done and what is presented in the figures (single planes or maximum intensity projections etc). This makes it difficult to evaluate the data and results.

We have incorporated additional experimental details to the Materials and Methods section: "Recent advancements in optical and camera technologies permit the acquisition of Z-stacks without perturbing Q cell division or overall animal development. Z-stack images were acquired over a range of -1.6 to +1.6 μm from the focal plane, at intervals of 0.8 μm. The field-of-view spanned 160 μm × 160 μm, and the laser power, as measured at the optical fiber, was approximately 1 mW. ImageJ software (http://rsbweb.nih.gov/ij/) was used to perform image analysis and measurement. Image stacks were z-projected using the average projection for quantification and using the maximum projection for visual display. "

The majority of our experimental procedures adhere to methodologies delineated in our prior publications and other scientific literature. We were pioneers in the development of fluorescence time-lapse live microscopy techniques for capturing Q cell migration and asymmetric division (Ou and Vale, Journal of Cell Biology, 2009; Ou et al., Science, 2010; Chai et al., Nature Protocols, 2012). Our innovative imaging protocol uncovered a novel mode of polarized, non-muscle myosin-II-dependent asymmetric cell division (Ou et al., Science, 2010). Subsequently, we unveiled another previously uncharacterized mechanism of asymmetric cell division dependent on polarized actin polymerization (Chai et al., Cell Discovery, 2022). In the present study, we have significantly refined our imaging and quantification protocols. Different from the single-focal-plane imaging employed in our earlier study by Ou et al. 2009, advancements in optical technologies and camera resolution now enable us to undertake time-lapse imaging across multiple focal planes and track signal differences between the anterior and posterior segments of dividing cells.

There is insufficient information about tools and reporters used. This is misleading and impacts the conclusions that can be made from the results presented. To give an example, in Figure 1D-F, the authors present data that HDA-1::GFP and LIN-53::mNeonGreen (both components of the nucleosome remodeling and deacetylation complex) but not the histone acetyltransferase MYS-1::GFP are 'asymmetrically segregated' during QR.a division. However, the authors do not mention that HDA-1::GFP and LIN-53::mNeonGreen are expressed at endogenous levels (they are CRISPR alleles) whereas MYS-1::GFP is overexpressed (integration of a multi-copy extrachromosomal array). The difference in 'segregation' could therefore be a consequence of different levels of expression rather than different modes of segregation ('asymmetric' versus 'symmetric').

Figure S2 shows overexpressed HDA-1, LIN-53 and CHD-3 are also asymmetrically segregated during ACD of QR.a, which indicates that different levels of expression do not affect the modes of segregation, at least for the NuRD subunits. In the main text, however, we presented the asymmetric segregation of HDA-1::GFP and LIN-53::mNeonGreen using their CRISPR KI alleles.

There is insufficient information about the phenotypes of the animals used (RNAi knock-downs of hda-1, lin-53 RNAi, pig-1 etc). Again this is misleading and impacts the conclusions that can be made. To give some examples,

1. In Figure 3A-G, control RNAi embryos are compared to hda-1 RNAi and lin-53 RNAi embryos. What the authors do not mention is that hda-1 RNAi and lin-53 RNAi embryos have severe developmental defects and essentially cannot be compared to control RNAi embryos. The differences between the embryos can be seen in Figure S7B where bright-field images of control RNAi, hda-1 RNAi and lin-53 RNAi embryos are shown. At the 350 min time point, a normal embryo is visible for the control, a 'ball of cells' embryo for hda-1 RNAi and an embryo that seems to have arrested at an earlier developmental stage (and therefore have much larger cells) for lin-53 RNAi. Because of these pleiotropic phenotypes, it is unclear whether differences seen for example in sAnxV::GFP positive cells (Figure 3A) are the result of a direct effect of hda-1(RNAi) on cell death or whether they are the result of global changes in development and cell fate induced by hda-1(RNAi). hda-1(RNAi) and lin-53(RNAi) embryos are also used for the data shown in Figures S6 and S7, raising the same concerns;

In the submitted manuscript, we mentioned that hda-1 RNAi and lin-53 RNAi caused embryonic lethality and that we could track some of the apoptotic events in hda-1 RNAi embryos arrested between the late gastrulation stage and bean stage. We agree with the reviewers that because of the pleiotropic phenotypes, we cannot distinguish whether sAnxV::GFP positive cells (Figure 3A) are the result of a direct effect of hda-1 (RNAi) on cell death or whether they are the result of global changes in development and cell fate induced by hda-1 (RNAi). We added the sentence to page 9 line 26: “Considering the pleiotropic phenotypes caused by loss of HDA-1, we cannot exclude the possibility that ectopic cell death might result from global changes in development, even though HDA-1 may directly contribute to the life-versus-death fate determination.”

1. The authors do not mention what the impact of Baf A1 treatment is on animals; however, the images provided in Figure 5E indicate that Baf A1 treatment causes pleiotropic effects in L1 larvae.

We have carefully checked the BafA1 treated animals, but have not been able to detect any visible defect in Baf A1 treated animals under a 25× dissection microscope at the given dosage and duration of treatment. We also searched for the published images or literature and did not find pleiotropic effects on the animal level at this dosage and duration; however, we agree with the reviewers that perturbation of pH homeostasis in lysosomes by BafA1 will certainly generate pleiotropic cellular defects. We discussed the issue below:

"Although BafA1-mediated disruption of lysosomal pH homeostasis is recognized to elicit a wide array of intracellular abnormalities, we found no evidence of such pleiotropic effects at the organismal level with the dosage and duration of treatment employed in this study."

There is a lack of adequate controls. Because of this, some of the data presented must be considered as preliminary. To give some examples:

1. Controls are lacking for the data shown in Figure 3D-G (i.e. genes other than egl-1). Since hda-1 RNAi has a pleiotropic effect and most likely affects H3K27 acetylation genome-wide, this is critical. Based on what is shown, it is unclear whether the results presented are specific to egl-1 or not;

In figure 3F, we added F23B12.1 and sru-43 as the controls of egl-1. We added “while the H3K27ac level of genes adjacent to egl-1 showed no significant changes” to Page 10 line 22 in the revised text.

1. The co-IP and mass spec data shown in Figure 4A, C and Figure S8 also lack a critical control, which is GFP only. Because of this, it is unclear whether subunits of the V-ATPase bind to HDA-1 or GFP. The co-IP and mass spec data forms the basis of Figures 5 and 6 as well as Figure S9. Data presented in these figures therefore has to be considered preliminary as well.

In the co-IP and mass spec shown in Figure 4A, we used ACT-4::GFP as the negative control, which can preclude V-ATPase subunits that bind to GFP. In Figure 4C, we used anti-V1A (V-ATPase V1 domain A subunit) antibody to confirm the interaction between V1A and HDA-1. In Figure S8B, we also used ACT-4::GFP as a control, showing other NuRD subunits bind to HDA-1 rather than GFP.

Inappropriate methods are used. For this reason, some of the data again must be considered preliminary. To give some examples:

1. In Figure 5A, B, the authors used super-ecliptic pHluorin to look at changes in pH in the daughter cells. However, the authors used quenching of super-ecliptic pHluorin fluorescence rather than a ratio-metric method to 'measure' changes in pH. Because of this, it is unclear whether the changes in fluorescence observed are due to changes in pH or changes in the amount of pHluorin protein. Figure 5A, B forms the basis for the experiments presented in the remaining parts of Figure 5 as well as in Figure 6 and Figure S9;

Bafilomycin A1 inhibits the activity of V-ATPase, presumably preventing the pumping of protons into the apoptotic daughter cell. It is more likely that the apoptotic daughter cell becomes less acidic and more neutral after the treatment of Baf1A, although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein. A ratio-metric method to measure changes in pH will be further used to distinguish the two possibilities.

We added “although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein.” to Page 12 line 12 in the revised text.

1. The authors' description of how some images were modified before quantitative analysis raises concerns. The figures of concern are particularly Figure 1 and Figure S4, where background subtraction with denoising and deconvolution was used. Background subtraction, with denoising and deconvolution is an image manipulation that enhances the contrast between background and what looks like foreground. Therefore, background subtraction should be applied primarily in experiments involving image segmentation not fluorescence intensity measurement. Not being provided any information by the authors about the kind of subtraction that was made, this processing could lead to an uneven subtraction across the image, which can easily lead to artefacts. Since the fluorescence intensity in the smaller daughter cell is lower, and thus closer to background, the algorithm the authors used may have misinterpreted the grey value information in the smaller daughter cell pixels. This could have led to an asymmetric subtraction of background in the two daughter cells, leading to a stronger subtraction in the smaller daughter cell. Ultimately, their processing could have artificially increased the intensity asymmetry between the two daughter cells in all their results.

As mentioned earlier, the imaging and quantification methods of this manuscript have been routinely used in our previous publications or studies from many other labs (Gräbnitz F, et al., Cell Rep. 2023; Herrero E, et al., Genetics. 2020; Roubinet C, et al., Curr Biol. 2021). Background subtraction is a standard procedure to quantify cellular fluorescence intensities. The fluorescence intensity of the slide background was measured from a region without worm bodies, of the same size as the region of interest. We have added how we measured the background to page 19 Line 24: “The fluorescence intensity of the slide background was measured from a region without worm bodies, of the same size as the region of interest.”

The imaging data is of low quality (for example Figures 1, 2, 5, 6; Figures S2, S3, S5, S6, S9). Since much of the study and the findings are based on imaging, this is a major concern. Critical parameters are not mentioned (number of sections in z-stack, size of the field-of-view, laser power used etc), which makes it difficult to understand what was done and what one is looking at.

Fluorescence images of neuroblast asymmetric cell division in developing C. elegans larvae has historically presented considerable challenges. Our recent methodological advancements have facilitated live imaging in this intricate system with improved resolution. In the revised manuscript, we have elucidated the specific z-stack parameters, field-of-view dimensions, and laser power settings employed: "Z-stack images were acquired over a range of -1.6 to +1.6 μm from the focal plane, at intervals of 0.8 μm. The field-of-view spaned 160 μm × 160 μm, and the laser power, as measured at the optical fiber, was approximately 1 mW."

To give some specific examples,

1. The images shown in Figure 2B are of very low quality with severe background from neighbouring cells. In addition, the outline of the cells (plasma membrane) or the nuclei of the daughter cells is unknown. Based on this it is not clear how the authors could have measured 'Fluorescence intensity ratio between sister nuclei' in an accurate and unbiased way (what is clear from these images is that there is an increase in HDA-1::GFP signal in ALL surviving daughters (asymmetric and symmetric divisions) post cytokinesis but not in the daughter cell that is about to die (asymmetric and unequal division));

We employed live-cell imaging in conjunction with automated cell lineage tracing algorithms (Du et al., Cell, 2014) to scrutinize NuRD asymmetry in embryos from the two- or four-cell stage up to the 350-cell stage. This sophisticated approach was initially pioneered by Dr. Zhirong Bao at Sloan Kettering and subsequently refined by Dr. Zhuo Du during Dr. Du's postdoctoral training in Dr. Bao's laboratory. This advanced imaging pipeline enables the scientific community to quantify cellular fluorescence intensity in an automated fashion, thereby substantially mitigating manual intervention and bias.

1. The images in Figure 6A and Figure S9A on VHA-17 segregation and its colocalization to ER and lysosome segregation during QR.a division are of very low quality and it is unclear to the reviewer how such images were used to obtain the quantitative data shown.

In some cases, there is a discrepancy between what is shown in figures and what the authors state in the text. To give some examples:

1. On page 7, the authors state "..., we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold at cytokinesis, respectively (Figure 1D-E)." Looking at the images for HDA-1 and LIN-53 in Figure 1D, the increase in the ratio mainly occurs between 4 min and 6 min, which is post cytokinesis and NOT prior to cytokinesis;

Thank the reviewer for pointing out this. The nuclear HDA-1 or LIN-53 asymmetry increased to 1.5 or 1.8-fold 6 min after the onset of anaphase, when QR.a just completes cytokinesis. Therefore, We change the sentence “we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold at cytokinesis, respectively (Figure 1D-E).” to “we found that nuclear HDA-1 or LIN-53 asymmetry gradually increased from 1.1-fold at the onset of anaphase to 1.5 or 1.8-fold upon the completion of cytokinesis, respectively (Figure 1D-E).”

However, nuclear HDA-1 or LIN-53 asymmetry initiates prior to cytokinesis. We started to see the nuclear HDA-1 or LIN-53 asymmetry (1.4 fold for HDA-1 and 1.2 fold for LIN-53 ) 2 min after the onset of anaphase (Figure 1D).

1. These images (Figure 1D) also show that there is an increase in the HDA-1 and LIN-53 signals in the larger daughter cells (QR.ap), which suggests that the increase in ratios (Figure 1E) is the result of increased HDA-1 and LIN-53 synthesis post cytokinesis. However, on top of page 8, the authors state "The total fluorescence of HDA-1, LIN-53 and MYS-1 remained constant during ACDs, suggesting that protein redistribution may establish NuRD asymmetry (Figure S4C)." In Figure S4C, the authors present straight lines for 'relative total fluorescence' for imaging (probably z-stacks) that was done every min over the course of 7 min. If there was no increase in material as the authors claim, they should have seen significant photobleaching over the course of the 7 min and therefore reduced level of 'relative total fluorescence' over time. How the data presented in Figure S4C was generated is therefore unclear. (Despite the fact that the authors claim that the asymmetry seen is not due to new synthesis in the larger daughter cell post cytokinesis, it would be more consistent with the first experiment presented in this study (Figure S1) that shows that there is more hda-1 mRNA in egl-1(-) cells compared to egl-1(+) cells);

Regarding the concern of photo-bleaching, we have meticulously calibrated our imaging system over the past several years. Rigorous controls, qualification, and analyses were scrupulously undertaken during the development of our fluorescence time-lapse imaging system for the investigation of Q cell dynamics, initially established by Dr. Guangshuo Ou in Ron Vale's laboratory—a renowned hub for avant-garde imaging techniques (Ou & Vale, Journal of Cell Biology, 2009; Ou et al., Science, 2010). Remarkably, no discernible photobleaching was observed even during two to three-hour imaging.

We agree that protein turnover, involving both degradation and synthesis, may occur. However, NuRD asymmetric distribution occurred within several minutes after metaphase and QR.a completes cytokinesis ~6min after the onset of anaphase, while GFP protein translation and maturation require ~ 30 min in Q neuroblast (Ou & Vale, Journal of Cell Biology, 2009). Even if hda-1::gfp mRNA is translated during cell division, the nascent GFP-tagged protein will mature long after the completion of cytokinesis. Consequently, we postulate that the influence of newly synthesized GFP-tagged protein during Q cell division is negligible for quantification purposes. It is plausible that the asymmetry in HAD-1 protein distribution is independent of hda-1 mRNA asymmetry.

1. On page 12, the authors state "..., in Baf A1-treated animals, QRaa inherited similar levels of HDA-1::GFP as its sister cell,...". However, looking at the image provided in Figure 5E (0 min), there seems to be a similar ratio of HDA-1::GFP between the daughter cells in DMSO and Baf A1-treated animals.

We have adjusted the images in Figure 5E to show the asymmetry in DMSO-treated control animals. We acknowledge variations among animals. Our quantifications from more than 10 animals show the HDA-1 asymmetry in DMSO-treated animals in Figure 5B.

Recommendations for the authors:

Conclusion 1

"Here, we demonstrate that the nucleosome remodeling and deacetylase (NuRD) complex is asymmetrically segregated into the surviving daughter cell rather than the apoptotic one during ACDs in Caenorhabditis elegans" (Abstract)

Results described on pages 6-9 ("NuRD asymmetric segregation during neuroblast ACDs" and "NuRD asymmetric segregation in embryonic cell lineages") and data shown in Figure S1, Figure 1, Figures S2, S3, S4, S5, Figure 2.

Conclusion 1 is not supported by the results as numerous concerns exist about the data in many of these figures (see above, major weaknesses). A more likely explanation for the authors' observations is that there is synthesis of NuRD post cytokinesis and that asymmetries in the amounts of NuRD observed in the two daughter cells is a consequence of their different cell sizes (QR.ap is 3x as large as QR.aa). This is supported by the finding that the loss of pig-1, which causes 'equal' division resulting in two daughter cells of similar sizes, abolishes the differences in NuRD seen between the daughter cells.

As discussed earlier, GFP protein translation and maturation require ~ 30 min in Q neuroblast (Ou & Vale, Journal of Cell Biology, 2009). Even if there is the synthesis of NuRD post cytokinesis, the nascent GFP-tagged protein will not mature within our imaging timeframe, Therefore, NuRD asymmetry is unlikely to be a result of the synthesis of NuRD post cytokinesis. In addition, We found that MYS-1::GFP was symmetrically segregated into the small apoptotic daughter cells and big surviving daughter cells, suggesting NuRD asymmetry may be irrelevant to cell size asymmetry.

Interestingly, despite the fact that the loss of pig-1 causes 100% of the divisions to be equal by size and symmetric with respect to NuRD amounts, it only causes about 30% of QR.aa cells to inappropriately survive. This demonstrates that there is a correlation between NuRD asymmetry and daughter cell size asymmetry but NOT between NuRD asymmetry and cell death. This also demonstrates that loss of 'NuRD asymmetry' and presence of NuRD in the daughter that should die is NOT sufficient to block its death.

Cordes et al. 2006 (DOI: 10.1242/dev.02447) reported that in pig-1 loss-of-function mutants, <40% of Q.p lineages produce extra neurons because Q.pp cells inappropriately survive. Noticeably, only 30% and 5% Q.p lineages produce extra neurons in ced-3 and egl-1 loss of function single mutant, respectively. pig-1 ced-3 double mutant or pig-1 egl-1 double mutants show a dramatically stronger phenotype than either single mutant, resulting in about 80% of Q.p lineages producing extra neurons. These results suggest that pig-1 functions in parallel to the EGL-1-CED-9-CED-4-CED-3 cell death pathway to promote Q cell apoptosis.

We agree with the reviewer that “loss of 'NuRD asymmetry' and presence of NuRD in the daughter that should die is NOT sufficient to block its death” in pig-1 loss-of-function mutants. However, these results do not rule out the correlation between NuRD asymmetry and cell death. In the pig-1 mutant, the concentration of NuRD in Q.pp might not be high enough to completely block the death pathway. Alternatively, NuRD may be one but not the only factor blocking the cell death pathway.

Lastly, it is imperative to underscore that cellular aberrations observed during early developmental stages frequently undergo compensatory correction during subsequent developmental stages or even initial aging stages. For example, in certain cell migration mutants exhibiting early migration defects, the initial penetrance exceeds 80%; however, the penetrance is mitigated to a mere 30% in adults. Such observations have been corroborated in our prior publications focusing on cell migration dynamics (Wang et al., PNAS, 2013; Zhu et al., Dev Cell, 2016). This appears to be a pervasive phenomenon, echoed by several laboratories specializing in neural development. Sengupta and Blacque’s labs has reported that early aging can ameliorate ciliary phenotypes in C. elegans mutants with compromised intraflagellar transport mechanisms. Accordingly, late developmental stages may act as a compensatory buffer for antecedent developmental abnormalities.

Conclusion 2

"The absence of NuRD triggers apoptosis via the EGL-1-CED-9-CED-4-CED-3 pathway, while an ectopic gain of NuRD enables apoptotic cells to survive." (Abstract) Results described on pages 8-10 ("Loss of the deacetylation activity of NuRD causes ectopic apoptosis" and "NuRD RNAi upregulates the egl-1 expression by increasing its H3K27 aceylation") and data shown in Figure S6, Figure 3, Figure S7 and data shown in Figure 5.

Because of the various concerns raised above (major weaknesses) about the data presented in Figure S6, Figure 3, Figure S7 (pleiotropic phenotypes of hda-1 and lin-53 RNAi animals, lack of controls etc), there is no evidence that NuRD has a specific and/or direct effect on egl-1 expression in cells programmed to die or that loss of NuRD causes ectopic egl-1-dependent cell death. The claim that "ectopic gain of NuRD enables apoptotic cells to survive." is based on Figure 5E, where the authors show that Baf A1 treatment causes symmetric NuRD segregation in 11/12 animals and that QR.aa survives in 11/12 animals. However, those data are unconvincing. As mentioned above (major weaknesses), from the low-quality images provided, it is not clear whether there is 'symmetric NuRD segregation' in Baf A1 treated animals, and the conditions of the animals are a concern because of pleiotropic effects of blocking V-ATPase. (I am not convinced I am actually looking at the same region of an L1 larvae in the three animals because the HDA-1::GFP signal seems inconsistent across them.) One process that is affected by a block of V-ATPase is engulfment. The fact that the authors observe that 130 min post-cytokinesis, QR.aa still persists in Baf A1 treated animals could therefore be the result of a delay in engulfment rather than a block in cell death. In addition, the claim that ectopic gain of NuRD enables apoptotic cells to survive contradicts their findings on loss of pig-1 described about ('Conclusion 1').

We acknowledge the limitations of our imaging system; however, as we pointed out earlier that we developed imaging methods and kept improving them. We have tried our best to obtain images from developing C. elegans larvae. On the other hand, we showed that hda-1 RNAi and lin-53 RNAi increase the expression of a subset of genes, including egl-1, either directly or indirectly (Fig. 3C). Figure 3B shows the ectopic cell death caused by loss of NuRD is dependent on EGL-1-CED-9-CED-4-CED-3 pathway. While we cannot exclude several other possibilities while addressing such a complex problem in such a challenging model system, these results provide some evidence supporting that our claim can be one of the possibilities.

Conclusion(s) 3

"We identified the vacuolar H+-adenosine triphosphatase (V-ATPase) complex as a crucial regulator of NuRD's asymmetric segregation. V-ATPase interacts with NuRD and is asymmetrically segregated into the surviving daughter cell. Inhibition of V-ATPase disrupts cytosolic pH asymmetry and NuRD asymmetry" (Abstract)

Results described on pages 10-13 ("V-ATPase regulates asymmetric segregation of NuRD during somatic ACDs") and data shown in Figures 4, 5, 6, Figures S8, S9.

As outlined above (major weaknesses), the evidence that HDA-1 interacts with the V-ATPase complex is preliminary (no GFP control), and I consider the imaging data showing that V-ATPase asymmetrically segregates very low quality and unconvincing (Figure 6). The data on pH changes are also preliminary as the experiment was not done the way it should have (quenching rather than ratiometric). Finally, there are concerns about the results that apparently demonstrate that inhibiting V-ATPase activity disrupts pH asymmetry and NuRD asymmetry (impact of Baf A1 treatment).

As discussed earlier, Bafilomycin A1 inhibits the activity of V-ATPase, presumably preventing the pumping of protons into apoptotic daughter cells. It is more likely that the apoptotic daughter cell becomes less acidic and more neutral after the treatment of Baf1A, although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein. A ratio-metric method to measure changes in pH will be further used to distinguish the two possibilities.

We added “although we cannot exclude the possibility that the changes in fluorescence could be due to changes in the amount of pHluorin protein.” to Page 12 line 12 in the revised text.

Conclusion 4

"We suggest that asymmetric segregation of V-ATPase may cause distinct acidification levels in the two daughter cells, enabling asymmetric epigenetic inheritance that specifies their respective life-versus-death fates." (Abstract) Discussion and model Figure 6C.

I consider the model premature and not based on any convincing data. In addition, the role of V-ATPase and acidification does not make sense. V-ATPase is involved in the acidification of the lysosomal system (lumen), and it is thought that cytosolic acidification in apoptotic cells is caused by lysosomal leakage. This is not consistent with the authors' model.

This manuscript lacks a section describing details of statistical analyses and the rationale for the chosen test, sample sizes, exclusion criteria, and replication details. Although the sample size is relatively smaller (less than 30), the authors used "unpaired t-test" for most of the tests. They should describe which type of t-test they used (parametric or non-parametric test). They also should provide replication details for non-statistical data set, for example Fig 3F and Fig 4C.

We used the Unpaired two-tailed parametric t-test. We have now added the information for statistic tests in the revised methods and figure legends.

Author Response

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

Thank you for the thoughtful consideration of our work, including both reviewers’ constructive comments. Our apologies for taking some extra time for this revision, but we wanted to adress comments thoroughly with new analyses, not to mention a PhD defense, parental leave and my teaching ultimately being the bottleneck for the team’s work!

Reviewer #1 (Public Review):

The authors use a combination of structural and MD simulation approaches to characterize phospholipid interactions with the pentameric ligand-gated ion channel, GLIC. By analyzing the MD simulation data using clusters of closed and open states derived previously, the authors also seek to compare lipid interactions between putative functional states. The ultimate goal of this work is to understand how lipids shape the structure and function of this channel.

The strengths of this article include the following:

1) The MD simulation data provide extensive sampling of lipid interactions in GLIC, and these interactions were characterized in putative closed and open states of the channel. The extensive sampling permits confident delineation of 5-6 phospholipid interaction sites per subunit. The agreement in phospholipid binding poses between structures and the all-atom MD simulations supports the utility of MD simulations to examine lipid interactions.

2) The study presents phospholipid binding sites/poses that agree with functionally-important lipid binding sites in other pLGICs, supporting the notion that these sites are conserved. For example, the authors identify interactions of POPC at an outer leaflet intersubunit site that is specific for the open state. This result is quite interesting as phospholipids or drugs that positively modulate other pLGICs are known to occupy this site. Also, the effect of mutating W217 in the inner leaflet intersubunit site suggests that this residue, which is highly conserved in pLGICs, is an important determinant of the strength of phospholipid interactions at this site. This residue has been shown to interact with phospholipids in other pLGICs and forms the binding site of potentiating neurosteroids in the GABA(A) receptor.

Weaknesses of this article include the following:

1) The authors describe in detail state-dependent lipid interactions from the MD simulations; however, the functional significance of these findings is unclear. GLIC function appears to be insensitive to lipids, although this understanding is based on experiments where GLIC proteoliposomes were fused to oocyte membranes, which may not be optimal to control the lipid environment. Without functional studies of GLIC in model membranes, the lipid dependence of GLIC function is not definitively known. Therefore, it is difficult to interpret the meaning of these state-dependent lipid interactions in GLIC.

2) It is unlikely that the bound phospholipids in the GLIC structures, which are co-purified from e. coli membranes, are POPC. Rather, these are most like PE or PG lipids. While it is difficult to accommodate mixed phospholipid membranes in all-atom MD simulations, the choice of POPC for this model, while practically convenient, seems suboptimal, especially since it is not known if PE or PG lipids modulate GLIC function. Nevertheless, it is striking that the overall binding poses of POPC from the simulations agree with those identified in the structures. It is possible that the identity of the phospholipid headgroup will have more of an impact on the strength of interactions with GLIC rather than the interaction poses (see next point).

3) The all-atom MD simulations provide limited insight into the strength of the POPC interactions at each site, which is important to interpret the significance of these interactions. It is unlikely that the system has equilibrated within the 1.7 microseconds of simulation for each replicate preventing a meaningful assessment of the lipid interaction times. Although the authors report exchange of up to 4 POPC interacting at certain residues in M4, this may not represent binding/unbinding events (depending on how binding/interaction is defined), since the 4 Å cutoff distance for lipid interactions is relatively small. This may instead be a result of small movements of POPC in and out of this cutoff. The ability to assess interaction times may have been strengthened if the authors performed a single extended replicate up to, for example, 10-20 microseconds instead of extending multiple replicates to 1.7 microseconds.

Reviewer #2 (Public Review):

The authors convincingly show multiple inner and outer leaflet non-protein (lipid) densities in a cryo-EM closed state structure of GLIC, a prokaryotic homologue of canonical pentameric ligand-gated ion channels, and observe lipids in similar sites during extensive simulations at both resting and activating pH. The simulations not only corroborate structural observations, but also suggest the existence of a state-dependent lipid intersubunit site only occupied in the open state. These important findings will be of considerable interest to the ion channel community and provide new hypotheses about lipid interactions in conjunction with channel gating.

Recommendations for the authors: please note that you control which, if any, revisions, to undertake

In particular, a discussion of whether the timescale of the simulations permit measurements of residence or interaction times of the lipids should be addressed.

Reviewer #1 (Recommendations for the authors):

Comment 1.1: The authors may consider expanding the discussion about the significance of state-dependent lipid interactions. On the one hand, they emphasize state-dependent interactions of POPC with closed and open states in the outer leaflet in the results. On the other hand, they state that GLIC is insensitive to its lipid environment. What is the significance of the state-dependent interactions of POPC in GLIC, if any? It is possible that GLIC agonist responses are sensitive to phospholipids (such as PE or PG found in e. coli)? The state-dependent differences in lipid interaction identified in this study support this possibility and suggest the need to better understand the effects of phospholipids on GLIC function.

Response 1.1: We agree with the reviewer that this is an interesting question and we have therefore extended the discussion with additional references on the functional effects on GLIC of various lipid membranes:

p. 11 (Discussion)

“Sampling was further simplified by performing simulations in a uniform POPC membrane. Prior experiments have been conducted to assess the sensitivity of GLIC in varying lipid environments (Labriola et al., 2013; Carswell et al., 2015; Menny et al., 2017), indicating that GLIC remains fully functional in pure POPC bilayers. In our cryo-EM experiments, the protein was recombinantly expressed from E. coli, which means that the experimental density would likely represent phosphatidylglycerol or phosphatidylethanolamine lipids. However, as the molecular identities of bound lipids could not be precisely determined, POPC lipids were built for straightforward comparison with simulation poses. While it appears that GLIC is capable of gating in a pure POPC bilayer, it remains plausible that its function could be influenced by different lipid species, especially due to the presence of multiple charged residues around the TMD/ECD interface which might interact differently with different lipid head groups. Further experiments would be needed to confirm whether the state dependence observed in simulations is also lipid-dependent. It is possible that certain types of lipids bind in one but not the other state, or that certain states are stabilized by a particular lipid type.”

Comment 1.2: It would be helpful to state in the discussion that the co-purified lipids from GLIC structures are likely PE or PG from e. coli membranes. Nevertheless, it is interesting that the phospholipid poses from the structures generally agree with those identified from the MD simulations using PC.

Response 1.2: Good point. We have clarified in the discussion that the native lipids in the cryo-EM structure are likely PG or PE lipids, as quoted in the preceding Response.

Comment 1.3: The authors describe a more deeply penetrating interaction of POPC in the outer intrasubunit cleft in the open state, but this is difficult to appreciate from the images in Fig. 4B, 4E or S3B. The same is true of the deep POPC interaction at the outer intersubunit site. It may be helpful to show these densities from a different perspective to appreciate the depth of these binding poses.

Response 1.3: We have added Figure 4 – figure supplement 1 to better show the depth of lipid binding poses, especially the ones in the outer leaflet intrasubunit cleft and at the inner intersubunit site, and cited the figure on p. 7 (Results).

Comment 1.4: The representation of the lipid densities in Fig. 4B is not easy to interpret. First, the meaning of resting versus activating conditions and closed versus open states can be easily missed for readers who are not familiar with the author's previous study. It may be helpful to describe this (i.e. how open and closed state clusters were generated from structures determined in resting and activating conditions) in greater detail in either the figure legend, results or methods. Second, the authors state that there are differences in lipid poses between the closed and open states but not resting and activating conditions. With the exception of the intersubunit density, this is difficult to appreciate from Fig. 4B. As stated in point #3, the difference, for example, in the complementary intrasubunit site may be better appreciated with an image from a different perspective.

Response 1.4: Acknowledged - the distinction between resting and activating conditions v.s. open and closed states can be confusing. We have tried to clarify these differences at the beginning of the results section, the methods section, and in the caption of Figure 4. Regarding differences in lipid poses between open and closed states, we agree it is difficult to appreciate from Figure 4, but here we refer the reader to Figure 4 – figure supplement 2 for an overlay between open and closed densities. Additionally, we now added Figure 1 – figure supplement 1 which provides lipid densities for all five subunits and overlays with the build cryo-EM lipids, possibly making differences easier to appreciate. Regarding images from different perspectives, we trust the new figure supplement described in Response 1.3 provides a better perspective.

p. 3 (Results)

“For computational quantification of lipid interactions and binding sites, we used molecular simulations of GLIC conducted under either resting or activating conditions (Bergh et al., 2021a). As described in Methods, resting conditions corresponded to neutral pH with most acidic residues deprotonated; activating conditions corresponded to acidic pH with several acidic residues protonated. Both open and closed conformations were present in both conditions, albeit with different probabilities.”

p. 8 (Figure 4)

“Overlaid densities for each state represent simulations conducted under resting (dark shades) or activating (light shades) conditions, which were largely superimposable within each state.”

p. 24 (Methods)

“We analyzed previously published MSMs of GLIC gating under both resting and activating conditions (Bergh et al., 2021a). Resting conditions corresponded to pH 7, at which GLIC is nonconductive in functional experiments, with all acidic residues modeled as deprotonated. Activating conditions corresponded to pH 4.6, at which GLIC is conductive and has been crystallized in an open state (Bocquet et al., 2009). These conditions were modeled by protonating a group of acidic residues (E26, E35, E67, E75, E82, D86, D88, E177, E243; H277 doubly protonated) as previously described (Nury et al., 2011).”

Comment 1.5: The new closed GLIC structure was obtained by merging multiple datasets. What were the conditions of the datasets used? Was it taken from samples in resting or also activating conditions?

Response 1.5: We have updated the Results, Discussion, and Methods to clarify this important point, in particular by merging datasets and rerunning the classification:

p. 3 (Results)

“In our cryo-EM work, a new GLIC reconstruction was generated by merging previously reported datasets collected at pH 7, 5, and 3 (Rovšnik et al., 2021). The predominant class from the merged data corresponded to an apparently closed channel at an overall resolution of 2.9 Å, the highest resolution yet reported for GLIC in this state (Figure 1 – figure supplement 2, Table 1).”

p. 11 (Discussion)

“Interestingly, the occupational densities varied remarkably little between resting and activating conditions (Figure 1 – figure supplement 1), indicating state- rather than pH- dependence in lipid interactions, also further justifying the approach of merging closed- state GLIC cryo-EM datasets collected at different pH conditions to resolve lipids.”

p. 14 (Methods)

“After overnight thrombin digestion, GLIC was isolated from its fusion partner by size exclusion in buffer B at pH 7, or in buffer B with citrate at pH 5 or 3 substituted for Tris. The purified protein was concentrated to 3–5 mg/mL by centrifugation. [...] Data from three different grids, at pH 7, 5, and 3, were merged and processed together.”

Comment 1.6: In Fig. 3D, do the spheres represent the double bond? If so, please state in the legend

Response 1.6: We have clarified in the legend of Figure 3D that the yellow spheres on the lipid tails represent a double bond.

Comment 1.7: In Fig. 3E, what is the scale of the color representation?

Response 1.7: We have clarified in the legend of Figure 3E that colors span 0 (white) to 137015 contacts (dark red).

Reviewer #2 (Recommendations For The Authors):

Comment 2.1: I'm not sure I fully understand how the final lipids were modeled (built). Fig. 1 caption suggests they may have been manually built? I understand that the idea was to place them in the overlap of simulation densities and structure densities, but can the authors please clarify if there were any quantifiable conditions that were employed during this process or if this was entirely manual placement in a pose that looked good? Regardless, it would be helpful to see an overlay of the built lipids with both the cryo and simulation densities (e.g., overly of Fig. 1F/H and G/H) to better visualize how the final built lipids compare.

Response 2.1: We thank the reviewer for pointing out unclarities regarding our methods. We have extended the methods section to clarify how the lipids were manually built in the cryo-EM structure. We have also added Figure 1 – figure supplement 1 showing overlays of the computational densities and built cryo-EM lipids.

p. 15 (Methods)

“Lipids were manually built in COOT by importing a canonical SMILES format of POPC (Kim et al., 2021) and adjusting it individually into the cryo-EM density in each of the sites associated with a single subunit, based in part on visual inspection of lipid densities from simulations, as described above. After building, 5-fold symmetry was applied to generate lipids at the same sites in the remaining four subunits.”

Comment 2.2: Regarding the state-dependent lipid entry to the outer leaflet intersubunit site associated with channel opening, if the authors could include a movie depicting this process that would be great. The current short explanation does not do this justice. Also, what were the dynamics of this process? Beyond the correlation between site occupancy and the pore being open, how did the timing of lipid entry/exit and pore opening/closing correlate?

Response 2.2: The point regarding the timing of state-dependent lipid binding at the subunit interface and pore opening is indeed an interesting one. We have added Figure 4 – figure supplement 3D showing that the state-dependent P250 lipid interaction precedes pore opening, as quantified by pore hydration levels, indicating a potential role in gating. The interaction between lipid binding and conformational change of the protein is also depicted in the newly added Figure 4 - video supplement 1, which we hope will be able to better communicate the conclusions regarding state-dependent interactions. We have also expanded the results and discussion to better explain these results:

p. 9 (Results)

“The lipid head made particularly close contacts with residue P250 on the M2-M3 loop, which undergoes substantial conformational change away from the pore upon channel opening, along with outer-leaflet regions of M1–M3 (Figure 4E, Figure 4—figure Supplement 3A,B,C, Figure 4—video 1). These conformational changes were accompanied by a flip of M1 residue F195, which blocked the site in the closed state but rotated inward to allow closer lipid interactions in the open state (Figure 4—figure Supplement 3C, Figure 4—video 1). Indeed, P250 was predominantly located within 3 Å of the nearest lipid atom in open- but not closed-state frames (Figure 4F). Despite being restricted to the open state, interactions with P250 were among the longest duration in all simulations (Figure 2C) and as these binding events preceded pore opening, it is plausible to infer a role for this state-dependent lipid interaction in the gating process (Figure 4 – figure supplement 3D).”

p. 12 (Discussion)

“The state-dependent binding event at this site preceded pore opening in MSMs, where lipid binding coincided with crossing a smaller energy barrier between closed and intermediate states, followed by pore opening at the main energy barrier between intermediate and open states (Figure 4 – figure supplement 3D). Further, since the P250- lipid interaction was characterized by relatively long residence times (Figure 2), it is possible this lipid interaction has a role to play in GLIC gating.”

Comment 2.3: Although the interaction times are helpful, I didn't get a great sense of how mobile the lipids are during the simulations. Can the authors discuss this a bit more. For example, are interaction times dominated by lipids that jiggle a bit away from a residue and then back again, vs how often are lipids exchanging with other lipids initially further away from the protein?

Response 2.3: We have now added various measures of lipid diffusion, both for initially interacting lipids and for bulk lipids, which are summarized in the new Figure 2 – figure supplement 1. We have further addressed the question of simulation timescales in Results, Discussion, and Methods. These numbers highlight that it is possible for lipids several nanometers away from the protein surface to exchange with lipids of the first lipid shell.

p. 3,6 (Results)

“Lateral lipid diffusion coefficients were estimated to 1.47 nm2/µs for bulk lipids and 0.68 nm2/µs for lipids of the first lipid shell (Figure 2 – figure supplement 1A), which is relatively slow compared to the timescales of each trajectory (1.7 µs). However, multiple residues throughout the M1, M3, and M4 helices exchanged contacts with 2-4 different lipid molecules in individual simulations (Figure 2C). Furthermore, 1.7-µs root mean square displacement of lipids originally in the first lipid shell was 2.15 nm, and 3.16 nm in the bulk bilayer, indicating such exchanges are not limited to nearby lipids (Figure 2 – figure supplement 1B). Thus, exchange events and diffusion estimates indicate that the duration of lipid contacts observed in this work can be at least partly attributed to interaction stabilities and not solely to sampling limitations.”

p. 11 (Discussion)

“Indeed, the unrestrained atomistic MD simulations studied here were not expected to capture the maximal duration of stable contacts, as indicated by some interaction times approaching the full 1.7-µs trajectory (Figure 2}). Nevertheless, simulations were of sufficient length to sample exchange of up to four lipids, particularly around the M4 helix. Calculation of lipid lateral diffusion coefficients resulted in average displacements at the end of simulations of 2.15 nm for lipids initially interacting with the protein surface, roughly corresponding to lipids diffusing out to the 4th lipid shell. Diffusion of bulk lipids was faster, allowing lipids originally 3.16 nm away from the protein surface to ingress the first lipid shell. This observation underscores the potential for lipid exchange events even among lipids initially distant from the protein surface. Of course, duration of exceptionally stable interactions, such as those involving T274 (Figure 2C), inevitably remain bounded by the length of our simulations. Still, diffusion metrics, supported by robust statistical analysis encompassing diverse starting conditions (500 trajectories), enable confident estimation of relative interaction times.“

p. 13 (Methods)

“Time-based measures of protein-lipid interactions, such as mean duration times and exchange of interactions, were calculated for the 100 x 1.7 µs-long simulations using prolintpy (Sejdiu and Tieleman, 2021) with a 4 Å interaction cutoff. Analysis of lateral lipid diffusion in individual simulations was carried out for two disjoint sets of lipids: the first lipid shell defined as lipids with any part within 4 Å of the protein surface (~90 lipids), and bulk lipids consisting of all other lipids (~280 lipids). Mean square displacements of each lipid set were calculated using GROMACS 2021.5 (Abraham et al., 2015b) with contributions from the protein center of mass removed. Diffusion coefficients for each set, DA, were calculated using the Einstein relation (Equation 1) by estimating the slope of the linear curve fit to the data.

where ri(t) is the coordinate of the center of mass of lipid i of set A at time t and DA is the self-diffusion coefficient.”

Comment 2.4: How symmetric or asymmetric are the cryo and simulation densities across subunits and was there subunit asymmetry in the final build lipids? I could not tell from any of the figures beyond the casual observation that they maybe look somewhat similar in Fig. 1?

Response 2.4: We thank the reviewer for this useful remark. We have clarified in the methods that the cryo-EM lipids were built in C5-symmetry, and thus the positions are symmetric. The computational densities were calculated independently for each subunit and are thus not necessarily symmetric. We have added Figure 1 – figure supplement 1 showing densities for all five subunits, also serving as an indication of convergence of the results.

p. 3 (Results) “Although the stochastic nature of simulations resulted in nonidentical lipid densities associated with the five GLIC subunits, patterns of lipid association were notably symmetric (Figure 1 – figure supplement 1).”

p. 14-15 (Methods)

“A smaller subset of particles was used to generate an initial model. All subsequent processing steps were done using 5-fold symmetry. […] A monomer of that model was fit to the reconstructed density and 5-fold symmetry was applied with PHENIX 1.19.2-4158 through NCS restraints detected from the reconstructed cryo-EM map, to generate a complete channel. […] After building, 5-fold symmetry was applied to generate lipids at the same sites in the remaining four subunits.”

Minor comments:

Comment 2.5: Fig. 1 is probably not easy to follow for the general reader and the caption is very brief. I suggest adding an additional explanation to the caption and/or additional annotations to the figure to help a general reader step through this.

Response 2.5: We have expanded the caption of Figure 1 and clarified the meanings of colors, labels, and annotations.

Comment 2.6: Fig. 1B - Caption is confusing. I would not call the state separation lines outlines as they are not closed loops. Also, I see red/orange and two shades of blue whereas the caption mentions orange and blue only. The caption should also explicitly say what the black lines are (other cluster separations).

Response 2.6: We have edited the caption to better describe colors, annotations, and the meaning of the data:

p. 4 (Figure 1)

“(B) Markov state models were used to cluster simulations conducted under resting (R) or activating (A) conditions into five states, including closed (left of the light or dark orange lines) and open (right of the light or dark blue lines). Black lines mark edges of other state clusters derived from MSM eigenvectors. Experimental structures are highlighted as white circles.”

Comment 2.7: Fig. 3F caption appears to conflict with data where interaction with W217A appears longer than W217. I think the authors want to suggest here that W217A reduces contact time with T274 as stated in the main text.

Response 2.7: We have clarified in this legend that “Mutation of residue W217, lining this pocket, reveals shortened interactions at the T274 binding site” (p. 6, Figure 3).

Comment 2.8: Ref 25 and 26 are the same.

Response 2.8: Apologies; this mistake has been corrected.

Author Response

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

Reviewer #1 (Public Review):

In this study, single neurons were recorded, using tetrodes, from the parahippocampal cortex of 5 rats navigating a double-Y maze (in which each arm of a Y-maze forks again). The goal was located at any one of the 4 branch terminations, and rats were given partial information in the form of a light cue that indicated whether the reward was on the right or left side of the maze. The second decision point was uncued and the rat had no way of knowing which of the two branches was correct, so this phase of the task was more akin to foraging. Following the outbound journey, with or without reward, the rat had to return (inbound journey) to the maze and start to begin again.

Neuronal activity was assessed for correlations with multiple navigation-relevant variables including location, head direction, speed, reward side, and goal location. The main finding is that a high proportion of neurons showed an increase in firing rate when the animal made a wrong turn at the first branch point (the one in which the correct decision was signalled). This increase, which the authors call rate remapping, persisted throughout the inbound journey as well. It was also found that head direction neurons (assessed by recording in an open field arena) in the same location in the room were more likely to show the rate change. The overall conclusion is that "during goal-directed navigation, parahippocampal neurons encode error information reflective of an animal's behavioral performance" or are "nodes in the transmission of behaviorally relevant variables during goal-directed navigation."

Overall I think this is a well-conducted study investigating an important class of neural representation: namely, the substrate for spatial orientation and navigation. The analyses are very sophisticated - possibly a little too much so, as the basic findings are relatively straightforward and the analyses take quite a bit of work to understand. A difficulty with the study is that it was exploratory (observational) rather than hypothesis-driven. Thus, the findings reveal correlations in the data but do not allow us to infer causal relationships.

We would like to clarify that this report consists of hypothesis-driven experiments, with post-hoc exploratory analyses. We have now made hypotheses more explicit in the text, and pointed out that follow-up analyses were to understand how these effects came to be. We thank the reviewer for pointing out that our hypotheses were not explicit in the introduction. We believe our results open the door for investigating the causal role of these regions in the propagation or generation of error signals during navigational behavior. Those types of experiments are however, outside the scope of the current work.

That said, the observation of increased firing in a subset of neurons following an erroneous choice is potentially interesting. However, the effect seems small. What were the actual firing rate values in Hz, and what was the effect size?

We thank the reviewer for the opportunity to clarify the effect size question. As there are multiple neurons in the analyses, differences in firing rate need necessarily to be normalized by a neuron's mean activity. For example, a difference of 1 spk/s is less meaningful when a neuron's base rate is 50 spk/s than when it is 10spks/s. Furthermore, our reports are for population level analyses, at which point comparing raw firing rate values and differences becomes more challenging. Nonetheless, we are including these raw metrics in two new supplemental figures (Figure 2 - figure supplement 4,5), where differences in individual neurons change can be up to 15 spks/s. Additionally, the patterns and statistical results observed in the main text are preserved, with outbound Right Cue minus Left Cue showing a left>stem>right (indicating error coding), and RW minus NRW showing negative values across all segments, indicating NRW>RW or higher activity following on inbound unrewarded trials. Statistics follow the corresponding main text results (Cue: segment LRT = 71.70; RW: segment LRT=45.80).

I also feel we are lacking information about the underlying behavior that accompanies these firing rate effects. The authors say "one possibility is that the head-direction signal in the parahippocampal region reflects a behavioral state related to the navigational choice or the lack of commitment to a particular navigational route" which is a good thought and raises the possibility that on error trials, rats are more uncertain and turn their heads more (vicarious trial and error) and thus sample the preferred firing direction more thoroughly. Another possibility is that they run more slowly, which is associated with a higher firing rate in these cells. I think we, therefore, need a better understanding of how behavior differed between error trials in terms of running speed, directional sampling, etc.

In terms of running speed, there was a small effect of mean running speed between correct and incorrect trials (across subjects LMEM: Cue correct>incorrect Z=2.3, p=0.02; RW Z=2.15, p=0.03). In most neurons, increases in speed will be accompanied by increases in firing rate. Thus, the differences in running speed cannot explain the observed results, as higher speed during correct trials would predict higher activity, which is the opposite of what we found.

A few good, convincing raw-data plots showing a remapping neuron on an error trial and a correct trial on the same arm would also be helpful (the spike plots were too tiny to get a good sense of this: fewer, larger ones would be more helpful).

Additional plots for individual units have been added, Figure 2 - figure supplement 3.

It would be useful to know at what point the elevated response returned to baseline, how - was it when the next trial began, and was the drop gradual (suggesting perhaps a more neurohumoral response) or sudden.

Due to the experimental design, this question cannot be addressed fully. Concretely, error trials incur a time-penalty in which the rats need to wait an additional 10 seconds before the next trial, while a new trial would start immediately when the animal nose-poked the home well after a correct trial. Nonetheless, the data on Reward provides insight into this question. The magnitude of the responses on left and right segments of the maze were larger than on the stem for Unrewarded (NRW) vs Rewarded (RW) trials on inbound trajectories, Fig. 4c. This suggests that while activity is still elevated post-incorrect throughout the maze, across units, this effect is smaller on the stem segment. Additionally, the analyses indicate that in the transition between outbound vs inbound trajectories (Figure 4 - figure supplement 3), activity patterns are sustained (incorrect>correct). Together, these results indicate that elevated "error-like" signal are slow in returning to baseline.  

Reviewer #2 (Public Review):

This work recorded neurons in the parahippocampal regions of the medial entorhinal cortex (MEC) and pre- and para-subiculum (PrS, PaS) during a visually guided navigation task on a 'tree maze'. They found that many of the neurons reflected in their firing the visual cue (or the associated correct behavioral choice of the animal) and also the absence of reward in inbound passes (with increased firing rate). Rate remapping explained best these firing rate changes in both conditions for those cells that exhibited place-related firing. This work used a novel task, and the increased firing rate at error trials in these regions is also novel. The limitation is that cells in these regions were analyzed together.

We acknowledge this limitation on our study, and we believe there might be interesting differences between these regions. Unfortunately, the post-mortem extraction of the recording implant micro-drive used for these experiments generated too much tissue damage for exact localization of the tetrodes. Nonetheless, given that the patterns were observed in all subjects, we are confident that at least the major findings of "error-like" signaling is present across the parahippocampal regions. At the same time, the distributions of functional cell types as defined in the open field are different across the PaS, PrS and MEC, leaving the possibility of a more nuanced error coding scheme by region.

Reviewer #3 (Public Review):

The authors set out to explore how neurons in the rodent parahippocampal area code for environmental and behavioral variables in a complex goal-directed task. The task required animals to learn the association between a cue and a spatial response and to use this information to guide behavior flexibly on a trial-by-trial basis. The authors then used a series of sophisticated analytical techniques to examine how neurons in this area encode spatial location, task-relevant cues, and correct vs. incorrect responding. While these questions have been addressed in studies of hippocampal place cells, these questions have not been addressed in these upstream parahippocampal areas.

Strengths:

1) The study presents data from ensembles of simultaneously recorded neurons in the parahippocampal region. The authors use a sophisticated method for ensuring they are not recording from the same neurons in multiple sessions and yet still report impressive sample sizes.

2) The use of the complex behavioral task guards against stereotyped behavior as rats need to continually pay attention to the relevant cue to guide behavior. The task is also quite difficult ensuring rats do not reach a ceiling level of performance which allows the authors to examine correct and incorrect trials and how spatial representations differ between them.

3) The authors take the unusual approach of not pre-processing the data to group neurons into categories based on the type of spatial information that they represent. This guards against preconceived assumptions as to how certain populations of neurons encode information.

4) The sophisticated analytical tools used throughout the manuscript allow the authors to examine spatial representations relative to a series of models of information processing.

5) The most interesting finding is that neurons in this region respond to situations where rewards are not received by increasing their firing rates. This error or mismatch signal is most commonly associated with regions of the basal ganglia and so this finding will be of particular interest to the field.

Weaknesses:

1) The histological verification of electrode position is poor and while this is acknowledged by the authors it does limit the ability to interpret these data. Recent advances have enabled researchers to look at very specific classes of neurons within traditionally defined anatomical regions and examine their interactions with well-defined targets in other parts of the brain. The lack of specificity here means that the authors have had to group MEC, PaS, and PrS into a functional group; the parahippocampus. Their primary aim is then to examine these neurons as a functional group. Given that we know that neurons in these areas differ in significant ways, there is not a strong argument for doing this.

See response to Reviewer 2.

2) The analytical/statistical tools used are very impressive but beyond the understanding of many readers. This limits the reader's ability to understand these data in reference to the rest of the literature. There are lots of places where this applies but I will describe one specific example. As noted above the authors use a complex method to examine whether neurons are recorded on multiple consecutive occasions. This is commendable as many studies in the field do not address this issue at all and it can have a major impact as analyses of multiple samples of the same neurons are often treated as if they were independent. However, there is no illustration of the outputs of this method. It would be good to see some examples of recordings that this method classifies as clearly different across days and those which are not. Some reference to previously used methods would also help the reader understand how this new method relates to those used previously.

We have added an additional Supplemental Figure (Figure 7 - figure supplement 1, that showcases the matching waveform approach. In the original manuscript, Fig. 7a provided an example output of the method.

3) The effects reported are often subtle, especially at the level of the single neuron. Examples in the figures do not support the interpretations from the population-level analysis very convincingly.

Additional plots for individual units have been added, Figure 2 - figure supplement 3. However, the effects, though small by unit, are consistent across neurons and subjects.

The authors largely achieve their aims with an interesting behavioral task that rats perform well but not too well. This allows them to examine memory on a trial-by-trial basis and have sufficient numbers of error trials to examine how spatial representations support memory-guided behavior. They report ensemble recordings from the parahippocampus which allows them to make conclusions about information processing within this region. This aim is relatively weak though given that this collection of areas would not usually be grouped together and treated as a single unitary area. They largely achieve their aim of examining the mechanisms underlying how these neurons code task-relevant factors such as spatial location, cue, and presence of reward. The mismatch or error-induced rate remapping will be a particularly interesting target for future research. It is also likely that the analytical tools used in this study could be used in future studies.

Reviewer #1 (Recommendations For The Authors):

1) Typo: "attempted to addresses these challenges"

We thank the reviewer for pointing out, this has been fixed.

2) "classified using tuning curve based metrics" - what does "tuning curve" mean in this context?

We have clarified this sentence in the main text.

3) "MEC neurons encode time-elapsed" should be "MEC neurons encode time elapsed" (no hyphen)

We thank the reviewer for pointing out, this has been fixed.

4) "a phenomenon referred to as 'global remapping'." - I dislike this term because it has two meanings in the literature. On the one hand, it is used to contrast with rate remapping: that is, it refers to a change in the location of place fields. On the other hand, it refers to the remapping of the whole population of cells at once, as contrasted with partial remapping. I suggest calling them location remapping (vs. rate) and complete remapping (vs. partial)

We agree that this is an overloaded term in the field. We have added 'location remapping' in the intro as a more specific term for global remapping.

5) " tasks with no trial-to-trial predictability or experimenter-controlled cues and goals in the same environment." - ambiguously worded as it isn't clear whether the "no" refers to one or both of what follows. Also needs a hyphen after experimenter.

We thank the reviewer for pointing out, this sentence has been reworded for clarity.

6) " neurons changed their firing activity as a function of cue identity" - this is confounded by behavior and reward contingency, both linked to cue identity, so the statement is slightly misleading.

We thank the reviewer for pointing this out, however, we disagree that this wording is misleading. Neurons changed their activity as a function cue identity and reward contingencies. Why neurons change their activity in such a manner is a different, more nuanced question, that we addressed through our analyses by converging on the "error" like signal that these signals seem to carry.

7) "remapping" - I am not fully comfortable with the use of this term in this context. It derives from the original reports of change in the firing location of place cells, and the proposal that these cells form a "map" with the change in activity reflecting recruitment of a new map. With observations of rate changes in some place cells, the new term "rate remapping" was introduced, and now the authors are using "rate remapping" to mean firing rate changes in non-spatial cells. The meaning is thus losing its value. "Re-coding" might be slightly better, although we can argue about whether "code" is much better than "map"

While we agree with the reviewer that "remapping" has been coerced into a grab-all term, these are the accepted semantics in the field. Thus, we are disinclined to change the language.

8) Figure 1 - it would be useful to indicate somehow that one of the decision points was cued and once free choice with the random outcome - it took me a while to work this out. Also, the choice of colors for the cues needs explaining - my understanding is that rats are very insensitive to these wavelengths. And what does Pse mean? I didn't understand those scatterplots at all.

The section "Tree-Maze behavior and electrophysiological recordings" under Results go into the details of the task. A reference and additional context for the selection of cues is now included in the "Behavioral Training" methods section. Rats possess dichromatic vision systems. Caption of Figure 1, 2 includes what Pse means, the performance of a subject for a given session. The scatter plots relate remapping to performance.

9) Also on Figure 1 - in the examples shown, it looks like the animals always checked the two end arms in the same order. Was this position habit typical?

We have added additional context in "Behavioral Training" methods section. Well trained rats do exhibit stereotyped behaviors (eg. going to one well then the other).

10) "...we hypothesized that the cue remapping score would be related to a subject's performance in the task." I am struggling to see why this doesn't follow trivially from the observation that remapping occurred on error trials.

We thank the reviewer for pointing out that this could use further clarity. We have added that the magnitude of remapping is what should relate to performance. To further clarify, remapping does not occur on error trials, remapping as operationally defined in this work, is the difference of spatial maps as a function of Cue identity or Reward contingency. Thus, as a difference metric, remapping occurs because there is a difference in activity between correct and incorrect trials. The magnitude of that difference need not relate to how the subject performed on the task.

11) "With this approach, found that incorrect coding units were more likely to overlap between cue and RW coding units than correct." Missing "we". I didn't understand this sentence - what does "overlap" mean?

We have added a sentence to further clarify this point.

12) "We found that incorrect>correct activity levels on outbound trajectories predicted incorrect>correct activity levels on inbound trajectories" - I don't understand how this can be the case given that many of these units were head direction tuned and therefore shouldn't even have been active in both directions.

As seen in Figure 7b, we were able to match 217 units across tasks. Of those, "Cluster 0" with 98 units showed strong head-direction coding. While "Cluster 0" units showed strong remapping effects, there were a lot of other units that could have contributed to the "incorrect>correct" across (out/in)-bound segments. Further, head-direction coding is defined in the Open-field environment, and there's no constraint on what these neurons could do on the Tree Maze task.

13). " Error or mismatch signals conform a fundamental computation" - should be "perform"

Wording slightly changed, but "conform" as in "act in accordance to" is what we intend here.

14) " provides it with the required stiffness and chemical resistivity"- what does "chemical resistivity" mean? To what chemicals?

This is mostly in reference to rat waste and cleaning products (alcohol). We changed the wording to durability for simplicity.

15) Supp Fig 1 shows that behavioral performance was very distinctly different for one of the animals. Was its neural data any different? What happens to the overall effect if this animal is removed from the analysis?

Unless otherwise stated, all analyses are performed through linear mixed effects with "subject" as a random effect. Thus, the effects of individual subjects are accounted for.

16) Histology - it would be useful to have a line drawing from the atlas alongside the micrographs to enable easier anatomical understanding.

There was variability in the medial lateral location of the tetrodes across animals and in the histological images provided and thus, we felt this would not be useful information as a single line drawing will not encompass/apply to all histology photos.

17) Supp. Fig. 5/6 I didn't understand what Left, Stem, and Right mean at the top. Also, the color keys are too tiny to be noticed

An additional sentence has been added to the caption to clarify that left, stem, right refer to what segment was selected via the ranking of scores.

Reviewer #2 (Recommendations For The Authors):

Was there a particular reason why cells in these regions were analyzed together? Can some of the results be tested for cells of a particular region, especially the MEC? One major limitation of these results is that it is unclear which regions it applies to, e.g., one cannot be certain that data shows here that MEC cells have these firing properties.

Damage due to the extraction of the recording tetrode bundle was extensive and we were not able to parcelate out individual regions. We have added additional details on this in the "Histology" section of the methods.

It is unclear how many cells in each region are included in each analysis. There is supplementary fig 3 but not sure how many of these met the criteria to be included in a certain analysis and it does not differentiate regions. Also, was any of the MUA included in the analyses?

Isolated single units were included in all analyses, but we did not differentiate from what region each unit came from. Analyses that include MUA are separate from the main findings, and are included in supplemental figures as reference.

Was the error trial defined as a trial when the animal did not make the right light-guided choice or did it also include cases in which the light-related arm choice was correct, but the animal first went to the unrewarded end arm? Nomenclature in the results is not explained well - what is an unrewarded trial or unrewarded trajectory or an error trial?

We have added a new paragraph in the methods under Behavioral Training that address trial nomenclature. This methods section is now referenced twice in the initial paragraphs of the results section.

Were any grid cells included in the data, especially could any cross-matched across the open field and the maze runs?

This was indeed a question of interest to us, however, the number of grid-cells recorded was not adequate for meaningful statistical inference. We further sought to avoid tuning curve based functional classifications of units.

In general, the results section is difficult to read, and its accessibility could be improved.

We thank the reviewer for this accessibility point. We hope that the small tweaks as a product of this revision will improve the readability of the manuscript. We tried to have major takeaways for each result, but the nature of the analyses necessarily make the text somewhat dense.

Minor:

One of the Figure 3f references should be Figure 3g and later, Figure 44 should be corrected.

We thank the reviewer for noting this, it has been fixed.

Reviewer #3 (Recommendations For The Authors):

There are a number of issues that I think could be addressed to improve the manuscript:

1) The figure could make it clearer where the LED panel is. Are the authors confident the rats see the cue on each trial?

We have added a new supplemental figure to address this question (Figure 1 - figure supplement 1). The new figures show the 3D geometry of the maze and the location of the Cue panel. The rats were able to see the cue, otherwise task performance would have remained at chance levels.

2) The same maze has been used in a series of studies of hippocampal place cells by Paul Dudchenko's group. They also went on to examine how these representations are affected in a very similar cued spatial response task. These studies should be acknowledged.

We thank the reviewer for pointing out this oversight. We have added the Ainge et al. citation ( https://doi.org/10.1523/JNEUROSCI.2011-07.2007) when first introducing the maze and in the methods.

3) In a number of supplementary figures, the authors present neurons that are selective for different properties such as segment, cue, reward, and direction. However, the number of spatially and cue-selective cells and the criteria by which cells are designated as selective are not reported. The analyses of spatial remapping and response to cues are done at the population level so I'm not sure how these cells are classified or selected for the figures.

The procedure for selection is included in the figure captions. Each unit is ranked based on the Uz score by segment as originally shown in Figures 2 and 4.

4) Related to this, the example cells on the figures do not clearly represent the effects presented. For example, given the title of Figure 2, I assume that the cells in 2B significantly remap. However, they don't look like they remap - the cells in the top row show rate remapping in one segment of the maze while the cells in the bottom do not show clear rate remapping responses. I suspect that traditional rate map-based analyses using maps based on consistently sized pixels rather than large segments would show only very modest changes in correlations or rates across these different types of trials. It is important to report the findings in this way as the authors interpret their data relative to the rate-remapping studies which have used these analyses. Readers who do not have the time or expertise to examine the methods in detail will conclude that the effects reported here are the same as previous rate remapping studies which the examples suggest is not the case.

Additional plots for individual units have been added to the supplement, Figure 2 - figure supplement 3. However, the effects, though small by unit, are consistent across neurons and subjects (Figure 2 - figure supplement 5).

5) Why is there a bias on the stem in 2C? This is of similar size to the effect on the right size and so deserves discussion.

The analysis in question is the across unit level bias in cue-coding by maze segment. The left segment shows elevated Right Cue coding, while the right segment shows elevated Left Cue coding. There was one reported statistical result, the main effect of segment in the Linear Mixed Effects model. We expand this result in the following two ways:

1. Individual statistical results by segment

a. Left Segment (Uz Coef. Estimate = 0.5, CI95%=[0.26, 0.75; p<1e-4])

b. Stem Segment (Uz Coef. Estimate = 0.22, CI95%=[-0.01, 0.47]; p=0.06)

c. Right Segment (Uz Coef. Estimate = -0.27, CI95%=[-0.51, -0.03], p=0.03)

1. Reporting the joint hypothesis test of left > stem > right by unit.

a. X2=90.45, p=2.28e-20

b. The comparison of left>stem by unit:

i. coefficient estimate = 0.28, CI95%=[0.11, 0.44], p=0.0008

Although the reviewer is correct in pointing out the effect size similarity, the appropriate statistical comparisons within and across units support the stated conclusions. In terms of systematic coding bias, there is a small bias across units (60% of units) and animals (4 out 5) for the Right Cue. Although interesting, this effect is orthogonal to the comparisons of interests (within unit differences). In order to highlight this point we have added the statistics of the joint hypothesis test of left>stem>right to the main manuscript.

Author Response

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

Response to Reviewer 1 Comments (Public Review):

Point 1: While the authors provided a large amount of data regarding the genes involved in the TOR pathway, it is highly descriptive and mostly confirmative data, as numerous papers have already shown that the TOR pathway plays essential roles in a myriad of biological processes in multiple fungi.

Response 1: Thank you for your comment. The target of rapamycin (TOR) signal pathway plays critical roles in various eukaryotic organisms. However, its specific role in controlling the development and virulence of opportunistic pathogenic fungi like A. flavus has remained unclear. Additionally, the underlying mechanism of the TOR pathway remains elusive in the A. flavus. As such, our study provides a useful contribution, as it is the first to comprehensively investigate the majority of genes in the conserved TOR signaling pathway in A. flavus.

Point 2: The authors seemed to perform a series of parallel studies in several genes involved in the TOR pathway in other fungi. However, their data are not properly interconnected to understand the TOR signaling pathway in this fungal pathogen. The authors frequently drew premature conclusions from basic phenotypic observations. For instance, based on their finding that sch9 mutant showed high calcium stress sensitivity, they concluded that Sch9 is the element of the calcineurin-CrzA pathway. Furthermore, based on their finding that the sch9 mutant show weak rapamycin sensitivity and increased Hog1 phosphorylation, they concluded that Sch9 is involved in TOR and HOG pathways. To make such conclusions, the authors should provide more detailed mechanistic data.

Response 2: Yes, we agree with the reviewer's comment. We have carefully reviewed the manuscript and made necessary revisions to eliminate arbitrary conclusions. For example, we have removed the statement that "Sch9 is the element of the calcineurin-CrzA pathway". Furthermore, we have rephrased our conclusions to better reflect our findings. "these results reflected that Sch9 regulates osmotic stress response via the HOG pathway in A. flavus"(Lines 279-280, page 13). We appreciate the reviewer's input, which has contributed to the clarity and accuracy of our work.

Point 3: In the section "Tor kinase plays important roles in A. flavus", some parts of their data are confusing. The authors said they identified a single Tor kinase ortholog, which is orthologous to S. cerevisiae Tor2. And then, they said failed to obtain a null mutant, but constructed a single copy deletion strain delta Tor1+/Tor2-. What does this mean? Does this mean A. flavus diploid strain? So is this heterozygous TOR/tor mutant? Otherwise, does the haploid A. flavus strain they used contain multiple copies of the TOR gene within its genome? What is the real name of A. flavus Tor kinase (Tor1 or Tor2?). "tor1+/tor2-" is the wrong genetic nomenclature. What is the identity of detalTor1+/Tor2-? Please provide detailed information on how all these mutants were generated. A similar issue was found in the analysis of TapA, which is speculated to be essential (what is the deltaTapA1+/TapA2-?). I couldn't find any detailed information even in Materials and Methods. The authors should provide southern blot data to validate all their mutants.

Response 3: Thank you for your comments. We acknowledge the confusion in our presentation and will ensure that accurate genetic nomenclature is used consistently throughout the paper.

In response to your queries, we have included a section in the Materials and Methods, titled "Detection of tor and tapA genes copy number in strains" (Lines 615-621, page 29), to provide details on how we determined the copy numbers of the tor and tapA genes in the strains. Our findings revealed that both the tor and tapA genes are present in double copies in our strains, which guided our decision to construct single-copy deletion strains using homologous recombination. We have verified these copy numbers using absolute quantification PCR (Table S1).

The use of the abbreviation '+/-' for the single copy knockout strains, such as tor+/- and tapA+/-, is consistent with common fungal literature practice. We apologize for any confusion caused by this nomenclature.

Although we did not employ southern blot data for validation, we conducted PCR and gene sequencing to confirm the mutants. We appreciate your comments to improve the clarity and accuracy of our manuscript.

Point 4: How were the FRB domain deletion mutants constructed? If the FKBP12-rapamycin binding (FRB) domain is specifically deleted in the Tor kinase allele, should it be insensitive and resistant to rapamycin? However, the authors showed that the FRB domain deleted TOR allele was indeed non-functional.

Response 4: We appreciate the reviewer's attention to the construction of the Fkbp12-rapamycin binding (FRB) domain deletion mutants and the discrepancy between the expected and observed results.

For the knockout of the FRB domain, we used the homologous recombination method, but because tor genes are double-copy genes, there are also double copies in the FRB domain. Despite our efforts, we encountered challenges in precisely determining the location of the other copy of the tor gene.

We speculate the common expectation that the deletion of the FRB domain should result in insensitivity and resistance to rapamycin, as it disrupts the binding site for Fkbp-rapamycin. However, we observed that the FRB domain-deleted mutant was more sensitive to rapamycin. This intriguing result suggests that there are additional factors or complexities involved in TOR signaling pathway regulation in A. flavus. We hypothesize that this result is related to the double copy of the tor gene. The reviewer's keen observation and comment have contributed to our efforts to better understand and explain this intriguing result.

Point 5: In Figure 4C, the authors should monitor Hog1 phosphorylation patterns under stressed conditions, such as NaCl treatment, and provide quantitative measurements. Similar issues were found in the western blot analysis of Slt2 (Fig. 8D).

Response 5: We agree with the reviewer that we should monitor Hog1 phosphorylation patterns under stressed conditions. In response to this valuable suggestion, we conducted additional experiments to examine Hog1 phosphorylation patterns under NaCl treatment for 30 minutes. The quantitative measurements of Hog1 phosphorylation levels under stress have been added to Figure 4E in the revised manuscript. Similarly, we have addressed the issue raised regarding Slt2 in Figure 8D.

Point 6: For all the deletion mutants generated in this study, the authors should generate complemented strains to validate their data.

Response 6: We appreciate the reviewer's suggestion to generate complemented strains for all the deletion mutants in our study to validate our data. However, due to the extensive number of genes involved in this research, it is hard to create complemented strains for each individual deletion mutant. As suggested by the reviewer, we have constructed complemented strains for several key deletion mutants, such as ΔsitA-C and Δppg1-C.

Response to Reviewer 1 Comments (Recommendations For The Authors):

Point 1: Overall, this manuscript was very poorly organized and not presented logically. It requires extensive English language editing.

Response 1: We appreciate the reviewer's feedback regarding the organization and language quality of our manuscript. To address these concerns, we have restructured the manuscript to improve its logical flow and coherence. We thank the reviewer for their constructive criticism, which has been instrumental in the manuscript's refinement.

Point 2: The authors did not present their figures in the order of description. For example, the authors suddenly described Figure 9A data in lines 128-130 in the middle of describing Figure 1. Furthermore, Figures 1D and 1F were described earlier than Figures 1B and 1C. In addition, Figure S2 was shown earlier than Figure S1. Please check this throughout the manuscript.

Response 2: We thank the reviewer for their insightful observation. We acknowledge the importance of a logical and coherent figure sequence for reader comprehension. After careful review, we have rearranged the text and images throughout the entire document to enhance the reading experience. The revised manuscript now presents figures in a consistent and logical order, following the sequence of descriptions. We believe this improvement will enhance the overall readability and comprehension of our research.

Point 3: The authors should follow the standard genetic nomenclature rules.

Response 3: Thank you for your suggestion. We have revised our manuscript to ensure that we are following the standard genetic nomenclature rules throughout. This includes the correct naming of genes, proteins, and mutations, as well as the use of appropriate italicization and formatting. We follow the rules: gene symbols are typically composed of three lowercase italicized letters, while protein symbols are not italicized, with an initial capital letter followed by lowercase letters.

Point 4: These are just a few examples. Besides the ones that I mentioned, I found numerous grammatically wrong or awkward sentences throughout the manuscript. So this manuscript requires extensive English proofreading.

Response 4: We apologize for the problem of our manuscript. We have asked an English native speaker to enhance the overall language quality and readability of the text. We believe that these improvements will significantly enhance the manuscript's overall quality and make it more accessible to a broader audience.

Response to Reviewer 2 Comments (Public Review):

Point 1: However, findings have not been deeply explored and conclusions mostly are based on parallel phenotypic observations. In addition, there are some concerns that exist surrounding the conclusions.

Response 1: We are grateful for the suggestion. We conduct additional experiments and analyses to delve more deeply into our findings and ensure a more robust basis for our conclusions.

Response to Reviewer 2 Comments (Recommendations For The Authors):

Point 1: Verification for mutants: a single copy deletion strain ΔTor1+/Tor2(containing one copy of the Tor gene), however, in the table of strain list, it seems like null mutants. There are no further verifications for relative genes' expression and no complementary strains.

A. Flavus ΔTor: Δku70; ΔniaD; ΔTor::pyrG

A. Flavus ΔTapA Δku70; ΔniaD; ΔTapA::pyrG

As described in pp208, "While we failed to obtained a null mutant, we constructed a single copy deletion strain ΔTor1+/Tor2- (containing one copy of the Tor gene) constructed by homologous recombination)"? But the authors think there was only one Tor kinase ortholog (AFLA_044350). It is hard to understand for this mutant What is the evidence to verify phenotypes of the ΔTor1+/Tor2- strain resulted from deletion of Tor2, no detail for how to make ΔTor1+/Tor2- strain.

Response 1: Thank you for your important comments and suggestion. We apologize for the confusion caused by genetic nomenclature. We make the necessary corrections in the table of strain lists to accurately reflect the genotypes of the strains (Table S3).

Multicopy variation of genes has not been explored in detail in fungi, especially in A. flavus, but is a commonly known phenomenon in mammalian genomes[1-2]. In yeast, the presence of two tor genes, tor1 and tor2, whereas in higher eukaryotes such as plants, animals, and filamentous fungi, there is only one tor gene[3-4]. The homology comparison results show that the genome of A. flavus contains only one tor gene. However, the tor gene in A. flavus exhibited varying copy numbers, as was confirmed by absolute quantification PCR at the genome level (Table S1).

In this study, we constructed a single copy deletion strain, tor+/-, through homologous recombination. This strain contains one copy of the tor gene. We provide a more detailed and explicit description of the methods used to detect of the genes copy number in strains (Lines 615-621, page 29). We thank the reviewer for pointing out these important issues.

Point 2: For a point mutant strain TORS1904L, they found that the sensitivity to rapamycin is consistent with the WT strain, it could not tell anything. It should be moved to Suppl.

Response 2: Thanks for your important comments. We acknowledge that these results may not provide significant insights. In response to this suggestion, we delete the data related to the TORS1904L point mutant strain and its sensitivity to rapamycin to ensure that the main manuscript focuses on the most pertinent and informative findings. Corresponding modifications have been made in the revised manuscript.

Point 3: For subtitle "Sch9 is correlate with the HOG and TOR pathways "What is the meaning for "correlate" similarly?

Response 3: Thank you for this comment. We apologize for the unclear wording. To enhance clarity, we revise the subtitle to more explicitly convey this conclusion, for example, "The Sch9 kinase is involved in aflatoxin biosynthesis and the HOG pathway". (Lines 242, page 12).

Point 4:for the ΔTapA 1+/TapA 2- strain (containing one copy of the TapA gene). It should have the complementary strain to verify the specific role of TapA. In FigS1B, ΔTOR and ΔTapA it could not tell TOR gene has been edited. Did you test mRNA of TOR gene?

Response 4: Thanks for your important comments. Due to the large number of genes involved, we did not perform a complementation experiment. However, we used PCR and sequencing to verify the editing of our gene. Additionally, we conducted copy number and mRNA analyses to verify its function. The transcriptional level of the tor gene in the tor+/- mutant was downregulated compared to the level in the wild-type strain (Fig. S6).

Response to Reviewer 3 Comments (Public Review):

Point 1: As for many results, I miss the re-complementation of the created mutants throughout the manuscript. This is standard praxis.

Response 1: Thanks for your suggestions. We acknowledge that re-complementation is a standard practice for validating the effects of gene deletions. However, due to the large number of genes involved in our study, we have performed supplementary experiments on a selection of them, such as ΔsitA-C and Δppg1-C. We are grateful to the reviewer for your understanding of this practical consideration.

Point 2: Fig. 1: cultures were grown for 48 h before measuring the transcript level. The authors show that brlA, abaA, and some sexual regulators are less expressed. In my opinion, this does not allow the conclusion that there is a direct control through rapamycin. Since the colonies grow very slowly in the presence of rapamycin, the authors should add rapamycin and follow gene expression after 15, 30, 60, 90 min. The figure legend needs to be more detailed. Which type of cultures were used, liquid, solid medium? Etc.

Response 2: We deeply appreciate the reviewer’s suggestion. Since we found that there were no significant differences in gene expression changes following shorter treatment times, we extended the treatment duration. We conduct additional experiments to examine the gene expression levels at longer time intervals (3, 6, and 9 h) after the addition of rapamycin (Figure 1H-1J). These time points allow us to capture the dynamic changes in gene expression in response to rapamycin more effectively. Additionally, we enhance the figure legend to provide a more comprehensive description that specifies the type of cultures used in the experiments.

Point 3: Why in chapter one Fig. 9 is already cited? Those data should then be included in Fig. 1 for the general phenotype.

Response 3: Thank you for the suggestion. We have reordered the figures in the updated version of the manuscript to ensure that the data for consistent and clarity.

Point 4: The authors wrote that radial growth and conidiation were gradually reduced with increasing rapamycin concentrations. This is not true. There is no gradient! However, it should be tested if there is a gradient if lower concentrations are used. The current data imply that there is a threshold concentration, so either there is 100 % growth or a reduction to 25 %. This looks strange.

Response 4: Thank you for underlining this deficiency. We agree that a threshold concentration versus a gradient is an important distinction that needs to be clarified. Our results show that the addition of excessive quantities of rapamycin does not increase the inhibition of A. flavus growth. As the concentration of the FK506 drug increases, there is a gradual decrease in the growth and cell production of A. flavus. This phenomenon could potentially be attributed to varying mechanisms of action exhibited by the drugs. Therefore, we have revised these confused sentences. ( Lines 120-121, Page 5)

Point 1: There are many wrong spellings:

Fig. 1. Before washed, before washing; RelaTEtive gene expERSion should read relative gene expression. Sclerotial should be sclerotia. See also Fig. 5 F, H, Fig. 6 E. 6D colon diameter should be colony diameter.

Fig. 4E. The expressED level... should read Expression level..... (also without article) Also in A, F, H.

Fig. 6C. TLC detection of WT.... The authors mean AF detection in extracts of WT..... AF was extracted and analyzed by TLC.....

Labelling of axes in one figure should be uniform.

Response 1: Thank you for your reminder. We apologize for the oversights, and we carefully address and correct all the mentioned spelling issues to ensure the accuracy and clarity of the manuscript.

Point 2: If the authors refer to the genes, I think they should be in small letters and italics, if it is the protein, the first letter should be capitalised tap1 (italics) and Tap1.

Response 2: We appreciate this suggestion. We have carefully checked the entire manuscript and revised follow the standard genetic nomenclature rules. We follow the naming conventions for microbial genes and proteins, where gene symbols are typically composed of three lowercase italicized letters, and protein symbols are not italicized, with an initial capital letter followed by lowercase letters.

Point 3: Very often articles are used where I would not use them.

Response 3: Thanks for your careful checks. We are sorry for our carelessness. Based on your comments, we have made the corrections to make the articles harmonized within the whole manuscript. We value the reviewer's feedback, which will contribute to the overall quality of our writing.

References:

[1] Handsaker R, Van Doren, V, Berman, J. et al. Large multiallelic copy number variations in humans. Nat Genet 47, 296–303 (2015).

[2] Wang Y, Wang S, Nie X. et al. Molecular and structural basis of nucleoside diphosphate kinase-mediated regulation of spore and sclerotia development in the fungus Aspergillus flavus. J Biol Chem. 2019 Aug 16;294(33):12415-12431.

[3] Kim DH, Sarbassov DD, Ali SM, et al. mTOR interacts with raptor to form a nutrient-sensitive complex that signals to the cell growth machinery. Cell. 2002; 110(2): 163-75.

[4] Fu L, Liu Y, Qin G, et al. The TOR-EIN2 axis mediates nuclear signalling to modulate plant growth. Nature. 2021; 591(7849): 288-292.

Author Response

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

Thank you for submitting your article "New genetic tools for mushroom body output neurons in Drosophila" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the assessment has been overseen by a Reviewing Editor and Albert Cardona as the Senior Editor.

eLife assessment:

This work advances on two Aso et al 2014 eLife papers to describe further resources valuable for the field. This paper adds more MBON split-Gal4s convincingly describing their anatomy, connectivity and function.

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript Rubin and Aso provide important new tools for the study of learning and memory in Drosophila. In flies, olfactory learning and memory occurs at the Mushroom Body (MB) and is communicated to the rest of the brain through Mushroom Body Output Neurons (MBONs). Previously, typical MBONs were thoroughly studied. Here, atypical MBONs that have dendritic input both within the MB lobes and in adjacent brain regions are studied. The authors describe new cell-type-specific GAL4 drivers for the majority of atypical MBONs (and other MBONs) and using an optogenetic activation screen they examined their ability to drive behaviors and learning.

The experiments in this manuscript were carefully performed and the results are clear. The tools provided in this manuscript are of great importance to the field.

Reviewer #2 (Public Review):

In this study, Aso and Rubin generated new split-GAL4 lines to label Drosophila mushroom body output neurons (MBONs) that previously lacked specific GAL4 drivers. The MBONs represent the output channels for the mushroom body (MB), a computational center in the fly brain. Prior research identified 21 types of typical MBONs whose dendrites exclusively innervate the MB and 14 types of atypical MBONs whose dendrites also innervate brain regions outside the MB. These MBONs transmit information from the MB to other brain areas and form recurrent connections to dopaminergic neurons whose axonal terminals innervate the MB. Investigating the functions of the MBONs is crucial to understanding how the MB processes information and regulates behavior. The authors previously established a collection of split-GAL4 lines for most of the typical MBONs and one atypical MBON. That split-GAL4 collection has been an invaluable tool for researchers studying the MB. This work extends their previous effort by generating additional driver lines labeling the MBON types not covered by the previous split-GAL4 collection. Using these new driver lines, the authors also activated the labeled MBONs using optogenetics and assessed their role in learning, locomotion, and valence coding. The expression patterns of the new split-GAL4 lines and the behavioral analysis presented in this manuscript are generally convincing. I believe that these new lines will be a valuable resource for the fly community.

Recommendations for the authors:

Minor additional suggestions:

1. Please ensure that the FlyLight links are provided for the new splitGal4s in the methods as well as results.

We added the requested link to the methods.

1. Correct a typo in 'ethyl lactate in the learning assays section of methods

corrected

Reviewer #1 (Recommendations For The Authors):

In the behavior assay, the authors use the same flies that were used for optogenetic olfactory conditioning and memory tests, to also examine the effects of activation in the absence of odors but with airflow. I think this may affect the interpretation of the results. If possible, it would be nice to show in the MBON types where a conditioning effect was found (i.e. MBON21, 29, 33) that performing the activation in the absence of odors but with airflow without previous conditioning yields the same results.

We share the reviewers concern that behavioral phenotypes during the later 10s LED sessions may be compromised by early optogenetic olfactory conditioning. Therefore, prior to running the experiment shown in Figure 2, we confirmed that the activation phenotypes of three positive control lines (MB011B and SS40755) could be observed after olfactory conditioning sessions. We added this data as Figure 2-figure supplement 2. For SS75200 and SS77383, a split-GAL4 driver for MBON33, we observed a loss of activation phenotype in the second trial of LED ON/OFF binary choice assay (Figure 3H). Therefore, we reran the 10s LED activation experiments without a previous optogenetic olfactory conditioning assay; these data are now also included in Figure 2-figure supplement 2.

Reviewer #2 (Recommendations For The Authors):

Below, I list some comments and suggestions which I hope could help the authors further improve their manuscript.

1. The authors identified 2 candidate lines for MBON28. It would be helpful if they could clarify how they determined whether a split-GAL4 correctly labels an MBON or is just a candidate line.

We have added in the methods section an explanation of the criteria used.

“The correspondence between the morphologies of EM skeletons and light microscopic images of GAL4 driver line expression patterns was used to assign GAL4 lines to particular cell types. This can be done with confidence when there are not multiple cell types with very similar morphology. However, in the case MBON28 we were not able to make a definitive assignment because of the similarity in the morphologies of MBON16, MBON17 and MBON28.”

1. The authors have previously shown that the expression pattern of a GAL4 driver is strongly influenced by the reporter used. The expression patterns of the split-GAL4 lines in this study are based on 20XUAS-Chrimson-mVenus trafficked (attp18), the expression strength of which may differ from other reporters or effectors. I suggest that the authors discuss this potential caveat in their manuscript. This will allow readers to be more cautious and check the expression patterns with their own reporters/effectors when using these new split-GAL4 lines.

We added the sentences below to address this concern.

“The expression patterns shown in this paper were obtained using an antibody against GFP which visualizes expression from 20xUAS-CsChrimson-mVenus in attP18. Directly visualizing the optogenetic effector is important since expression intensity, the number of labeled MBONs and off-targeted expression can differ when other UAS-reporter/effectors are used (for an example, see Figure 2—figure supplement 1 of Aso et al., 2014a).”

1. For the kinematic parameters in Fig. 2C, it is important to also show the baseline value of the parameters (i.e., the value before the light stimulation). For example, if a group of flies moves slower during the baseline period, their slower speed during the light-on period may not be due to MBON activation.

Figure 2 has been revised to include the z-scores for the 2s period just before turning on LED. The source data includes the parameter values used to calculate z-scores.

1. For Methods and Materials, the authors mostly refer to previous papers or websites for details. However, it would be helpful if they could include in this manuscript key information essential for repeating their experiments, such as the reporter/effector transgenes, empty-split controls, and antibodies and their working concentrations. It would also be helpful if they could provide the manufacturers and catalog numbers for the reagents used in this study.

We have added Appendix 1- Key Resource Table to list all the key reagents.

1. The original studies that identified the reward or punishment dopaminergic neurons mentioned in this manuscript should be cited.

We have added the following citations:

“Total number of synaptic connections from each MBON type to DANs and OANs. Based on the valence of memory when activation of DANs is used as unconditioned stimulus in olfactory conditioning (Aso et al., 2012, 2010; Aso and Rubin, 2016; Claridge-Chang et al., 2009; Huetteroth et al., 2015; Ichinose et al., 2015; Lin et al., 2014; Liu et al., 2012; Yamada et al., 2023; Yamagata et al., 2016, 2015)”

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

The proposed study provides an innovative framework for the identification of muscle synergies taking into account their task relevance. State-of-the-art techniques for extracting muscle interactions use unsupervised machine-learning algorithms applied to the envelopes of the electromyographic signals without taking into account the information related to the task being performed. In this work, the authors suggest including the task parameters in extracting muscle synergies using a network information framework previously proposed. This allows the identification of muscle interactions that are relevant, irrelevant, or redundant to the parameters of the task executed.

The proposed framework is a powerful tool to understand and identify muscle interactions for specific task parameters and it may be used to improve man-machine interfaces for the control of prostheses and robotic exoskeletons.

With respect to the network information framework recently published, this work added an important part to estimate the relevance of specific muscle interactions to the parameters of the task executed. However, the authors should better explain what is the added value of this contribution with respect to the previous one, also in terms of computational methods.

It is not clear how the well-known phenomenon of cross-talk during the recording of electromyographic muscle activity may affect the performance of the proposed technique and how it may bias the overall outcomes of the framework.

We thank reviewer 1 for their useful commentary on this manuscript.

Reviewer #2 (Public Review):

This paper is an attempt to extend or augment muscle synergy and motor primitive ideas with task measures. The authors idea is to use information metrics (mutual information, co-information) in 'synergy' constraint creation that includes task information directly. By using task related information and muscle information sources and then sparsification, the methods construct task relevant network communities among muscles, together with task redundant communities, and task irrelevant communities. This process of creating network communities may then constrain and help to guide subsequent synergy identification using the authors published sNM3F algorithm to detect spatial and temporal synergies.

The revised paper is much clearer and examples are helpful in various ways. However, figure 2 as presented does not convincingly show why task muscle mutual information helps in separating synergies, though it is helpful in defining the various network communities used in the toy example.

The impact of the information theoretic constraints developed as network communities on subsequent synergy separation are posited to be benign and to improve over other methods (e.g., NNMF). However, not fully addressed are the possible impacts of the methods on compositionality links with physiological bases, and the possibility remains of the methods sometimes instead leading to modules that represent more descriptive ML frameworks that may not support physiological work easily. Accordingly, there is a caveat. This is recognized and acknowledged by the authors in their rebuttal of the prior review. It will remain for other work to explore this issue, likely through testing on detailed high degree of freedom artificial neuromechanical models and tasks. This possible issue with the strategy here likely needs to be fully acknowledged in the paper.

The approach of the methods seeks to identify task relevant coordinative couplings. This is a meta problem for more classical synergy analyses. Classical analyses seek compositional elements stable across tasks. These elements may then be explored in causal experiments and generative simulations of coupling and control strategies. However, task-based understanding of synergy roles and functional uses is significant and is clearly likely to be aided by methods in this study.

Information based separation has been used in muscle synergy analyses using infomax ICA, which is information based at core. Though linear mixing of sources is assumed in ICA, minimized mutual information among source (synergy) drives is the basis of the separation and detects low variance synergy contributions (e.g., see Yang, Logan, Giszter, 2019). In the work in this paper, instead, mutual information approaches are used to cluster muscles and task features into network communities preceding the SNM3F algorithm use for separation, rather than using minimized information in separation. This contrast of an accretive or agglomerative mutual information strategy here used to cluster into networks, versus a minimizing mutual information source separation used in infomax ICA epitomizes a key difference in approach here.

Physiological causal testing of synergy ideas is neglected in the literature reviews in the paper. Although these are only in animal work (Hart and Giszter, 2010; Takei and Seki, 2017), the clear connection of muscle synergy analysis choices to physiology is important, and eventually these issues need to be better managed and understood in relation to the new methods proposed here, even if not in this paper.

Analyses of synergies using the methods the paper has proposed will likely be very much dependent on the number and quality of task variables included and how these are managed, and the impacts of these on the ensuing sparsification and network communities used prior to SNM3F. The authors acknowledge this in their response. This caveat should likely be made very explicit in the paper.

It would be useful in the future to explore the approach described with a range of simulated data to better understand the caveats, and optimizations for best practices in this approach.

A key component of the reviewers’ arguments here is their reductionist view of muscle synergies vs the emergentist view presented in our work here. In the reductionist lens, muscle groupings are the units (‘building blocks’) of coordinated movement and thus the space of intermuscular interactions is of particular interest for understanding movement construction. On the other hand, the emergentist view suggests that muscle groupings emerge from interactions between constituent parts (as quantified here using information theory, synergistic information is the information found when both activities are observed together). This is in line with recent work in the field showing modular control at the intramuscular level, exemplifying a scale-free phenomena. Nonetheless, we consider these approaches to muscle synergy research as complementary and beneficial for the field overall going forward.

Reviewer #3 (Public Review):

In this study, the authors developed and tested a novel framework for extracting muscle synergies. The approach aims at removing some limitations and constraints typical of previous approaches used in the field. In particular, the authors propose a mathematical formulation that removes constraints of linearity and couples the synergies to their motor outcome, supporting the concept of functional synergies and distinguishing the task-related performance related to each synergy. While some concepts behind this work were already introduced in recent work in the field, the methodology provided here encapsulates all these features in an original formulation providing a step forward with respect to the currently available algorithms. The authors also successfully demonstrated the applicability of their method to previously available datasets of multi-joint movements.

Preliminary results positively support the scientific soundness of the presented approach and its potential. The added values of the method should be documented more in future work to understand how the presented formulation relates to previous approaches and what novel insights can be achieved in practical scenarios and confirm/exploit the potential of the theoretical findings.

In their revision, the authors have implemented major revisions and improved their paper. The work was already of good quality and now it has improved further. The authors were able to successfully:

• improve the clarity of the writing (e.g.: better explaining the rationale and the aims of the paper);

• extend the clarification of some of the key novel concepts introduced in their work, like the redundant synergies;

• show a scenario in which their approach might be useful for increasing the understanding of motor control in patients with respect to traditional algorithms such as NMF. In particular, their example illustrates why considering the task space is a fundamental step forward when extracting muscle synergies, improving the practical and physiological interpretation of the results.

We thank reviewer 3 for their constructive commentary on this manuscript.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Figure 3 should report the distances between reaching points in panel A and the actual length distances of the walking paths in panel C.

The caption of fig.3 concerning the experimental setup of the datasets analysed has been updated with the following for dataset 1: “(A) Dataset 1 consisted of participants executing table-top point-to-point reaching movements (40cm distance from starting point P0) across four targets in forward (P1-P4) and backwards (P5-P8) directions at both fast and slow speeds (40 repetitions per task) [25]. The muscles recorded included the finger extensors (FE), brachioradialis (BR), biceps brachii (BI), medial-triceps (TM), lateral-triceps (TL), anterior deltoid (AD), posterior deltoid (PD), pectoralis major (PE), latissimus dorsi (LD) of the right, reaching arm.”. For dataset 3, to the best of the authors knowledge, this information was not given in the original paper.

Figure 4, what is the unit of the data shown?

The unit of bits is now mentioned in the toy example figure caption and in the caption of fig.5

Figure 4, the characteristics of the interactions are not fully clear, and the graphical representation should be improved.

We have made steps to improve the clarity of the figures presented.

For dataset 3, τ was the movement kinematics, but it is not specified how the task parameters were formulated. Did the authors use the data from all 32 kinematic markers, 4 IMUs, and force plates? If yes, it should be specified why all these signals were used. For sure, there will be signals included that are not relevant to the specific task. Did the authors select specific signals based on their relevance to the task (e.g., ankle kinematics)?

We have now clarified this in the text as follows: “For datasets 1 and 2, we determine the MI between vectors with respect to several discrete task parameters representing specific task attributes (e.g. reaching direction, speed etc.), while for dataset 3 we determined the task-relevant and -irrelevant muscles couplings in an unassuming way by quantifying them with respect to all available kinematic, dynamic and inertial motion unit (IMU) features.”

How did the authors endure that crosstalk did not affect their analysis, particularly between, e.g., finger extensors and brachioradialis and posterior deltoid and anterior deltoid (dataset 1)?

We have addressed this point in the previous round of reviews and made an explicit statement regarding cross-talk in the discussion section: “Although distinguishing task-irrelevant muscle couplings may capture artifacts such as EMG crosstalk, our results convey several physiological objectives of muscles including gross motor functions [66], the maintenance of internal joint mechanics and reciprocal inhibition of contralateral limbs [19,51].”

It would be informative to add some examples of not trivial/obvious task-related synergistic muscle combinations that have been extracted in the three datasets. Most of the examples reported in the manuscript are well-known biomechanically and quite intuitive, so they do not improve our understanding of synergistic muscle control in humans.

Our framework improves our understanding of synergistic motor control by enabling the formal quantification of synergistic muscle interactions, a capability not present among current approaches. Regarding the implications of this advance in terms of concrete examples, we have further clarified our examples presented in the results section, for example:

“Across datasets, many the muscle networks could be characterised by the transmission of complementary task information between functionally specialised muscle groups, many of which identified among the task-redundant representations (Fig.9-10 and Supp. Fig.2). The most obvious example of this is the S3 synergist muscle network of dataset 2 (Fig.11), which captures the complementary interaction between task-redundant submodules identified previously (S3 (Fig.9)).”

The description shows how our framework can extract the cross-module interactions that align with the higher-level objectives of the system, here the synergistic connectivity between the upper and lower body modules. Current approaches can only capture redundant and task-irrelevant interactions. Thus our framework provides additional insight into movement control.

The number of participations in dataset 2 is very limited and should be increased. We appreciate the reviewer's comment and would like to point out that for dataset 2 our aim was to increase the number of muscles (30), tasks (72) and trials for each task (30) which produced a very large dataset for each participant. This came at the expense of low number of participants, however all our statistical analyses here can be performed at the single-participant level. Furthermore, dataset 3 includes 25 participants and it enables us to demonstrate the reliability of the findings across participants.

Reviewer #2 (Recommendations For The Authors):

I believe it is important in the future to explore the approach proposed with a range of simulation data and neuromechanical models, to explore the issues I have raised and that you have acknowledged, though I agree it is likely out of scope for the paper here.

We agree with the reviewer that this would be valuable future work and indeed plan to do this in our future research.

The Github code for this paper should likely include the various data sets used in the paper and figures, appropriately anonymized, in order to allow the data to be explored and analyses replicated and package demonstrated to be exercised fully by a new user.

We thank the reviewer for this suggestion. Dataset3 is already available online at https://doi.org/10.1016/j.jbiomech.2021.110320. We will also make the other 2 datasets publicly available on our lab website very soon. Until then, as stated in the manuscript, we will make them available to anyone upon reasonable request.

Reviewer #3 (Recommendations For The Authors):

I have the following open points to suggest to the authors:

First, I recommend improving the quality of the figures: in the pdf version I downloaded, some writings are impossible to read.

We fully agree with the reviewer and note that in the pdf version of the paper, the figures are a lot worse than in the submitted word document submitted. Nevertheless, we will make further improvements on the figures as requested.

Even though the manuscript has improved, I still feel that some points were not addressed or were only partially addressed. In particular:

• The proposed comparison with NMF helps understanding why incorporating the task space is useful (and I fully agree with the authors about this point as the main reason to propose their contribution). However, the comparison does not help the reader to understand whether the synergies incorporating the task space are biased by the introduction of the task variables.

This question can be also reformulated as: are muscle synergies modified when task space variables are incorporated? Is the "weight" on task coefficients affecting the composition of muscle synergies? If so, the added interpretational power is achieved at the cost of losing the information regarding the neural substrate of synergies? I understand this point is not immediate to show, but it would increase the quality of the work.

• Reference to previous approaches that aimed at including task variables into synergy extraction are still missing in the paper. Even though it is not required to provide quantitative comparisons with other available approaches, there are at most 2-3 available algorithms in the literature (kinematics-EMG; force-EMG), that should not be neglected in this work. What did previous approaches achieve? What was improved with this approach? What was not improved?

Previous attempts of extracting synergies with non-linear approaches could also be described more.

In the latest version of the manuscript, we have referenced both the mixed NMF and autoencoders based algorithms. In both the introduction and discussion section of the manuscript, we also specify that our framework quantifies and decomposes muscle interactions in a novel way that cannot be done by other current approaches. In the results section we use examples from 3 different datasets to make this point clear, providing intuition on the use cases of our framework.

Author Response

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

Response to comments of editor/s:

• With regard to the comments on nonavailability of representative images/videos for Figures 1 A and B, in the revised manuscript we have added a representative video of GFP (-) and GFP (+) tracks in Supplemental video 1.

Response to comments of reviewer 2:

• With respect to the concern on figure 1, we have changed ‘% CD4+ T cell Migration’ to ‘% Proportion CD4+ T cell migration’ in Figures 1D & 1E in the revised manuscript. We also labelled the upper and lower panels of Figure 1I as ‘Untreated’ and ‘SDF1α’ respectively.

Response to comments of reviewer 1:

• With regard to the concern that ‘The transfection alone with siRNA may cause the lack of polarity’, we have added comparison of 2D migration MSD between control EGFP siRNA and Piezo1 siRNA-transfected CD4+ T cells as Supplementary Figure 1E.

• We have added new references as ref 42 and 43, with respect to PIEZO1 association with focal adhesions.

• With regard to the concerns around co-localization of Piezo1 and focal adhesions, we have added a representative image of Piezo1 and pFAK co-localization upon treatment of chemokine in revised Supplementary Fig. 3C. We have also used an additional focal adhesion marker, paxillin, to show that focal adhesion formation is not affected by Piezo1 KD (Revised Fig. 3E-3H). Upon comparing the mean pFAK and paxillin intensities, we observed no difference in Control and Piezo1 KD CD4+ T cells (Supplementary Figs. 3A, B).

• All the minor concerns and suggestions have been taken care of in the revised manuscript.

Author Response

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

Reviewer #1 (Public Review):

The manuscript is very-well written. Although the study is well-conducted the authors should be more convincing on how bacteria residing in tissues do not induce death. The association with IL-10 cytokine production appears weak and more experiments are needed to make it more robust

Reviewer #2 (Public Review):

Iske et al. provide experimental data that NAD+ lessens disease severity in bacterial sepsis without impacting on the host pathogen load. They show that in macrophages, NAD+ prevents Il1b secretion potentially mediated by Caspase11.

While the in vivo and in vitro data is interesting and hints towards a crucial role of NAD+ to promote metabolic adaptation in sepsis, the manuscript has shortcomings and would profit from several changes and additional experiments that support the claims.

Conceptually, the definition of sepsis is outdated. Sepsis is not SIRS, as in sepsis-2. Sepsis-3 defines sepsis as infection-associated organ dysfunction. This concept needs to be taken into account for the introduction and when describing the potential effects of NAD+ in sepsis. Also, LPS application cannot be considered a sepsis model, since it only recapitulates the consequence of TLR-4 activation. It is a model of endotoxemia. Also, the LPS data does not allow to draw conclusions about bacterial clearance (L135).

The authors state that protective effects by NAD were independent of the host pathogen load. This clearly indicates that NAD confers protection via enhancing a disease tolerance mechanism, potentially via reducing immunopathology. This aspect is not considered by the authors. The authors should incorporate the concept of disease tolerance in their work, cite the relevant literature on the topic and discuss it their findings in light of the published evidence for metabolic alteration sand adaptations in sepsis.

For the in vitro data, the manuscript would benefit from additional experiments using in vitro infection models.

In the merge manuscript, the authors provide two different versions of the figures. In one, bar plots are shown without individual data and in the other with scatter blots. All bar plots need to be provided as scatter plots showing individual values.

The authors should show further serology data for kidney and liver failure etc. as well as further cytokine data such as IL-6 and TNF to better characterize their models.

Careful revision of the entire manuscript, the figure legends and figures is required. The figure legend should not repeat the methods and materials section. The nomenclature for mouse protein and genes needs to be thoroughly revised.

L350. The authors write that they dissect the capacity of NAD+ to dampen auto- and alloimmunity. In this work, no data that supports this statement is shown and experiments with autoantigens or alloantigens are not performed.

L163 The authors describe pyroptosis but in the figure legend call it apoptosis. Specific markers for each cell death should be measured and determined which cell death mechanisms is involved.

Animal data comes from an infection model and LPS application. The RNAseq data is obtained from cells primed with Pam3CSK4 and subsequently subjected to LPS. It is unclear how the cell culture model reflects the animal model. As such the link between IFN signaling and the bacterial infection/LPS model are not convincing and need to be further elaborated.

Figure 5: It is unclear how many independent survival experiments were done, how many mice per group were used and whether the difference between groups was statistical significant. This information should be added.

Further experiments with primary cells from Il10 k.o. and Caspase11 k.o. animals should be provided that support the findings in macrophages.

Author Response:

Reviewer #1 (Public Review):

“The manuscript is very-well written. Although the study is well-conducted the authors should be more convincing on how bacteria residing in tissues do not induce death. The association with IL-10 cytokine production appears weak and more experiments are needed to make it more robust.”

Thank you very much for your thoughtful and constructive feedback on our manuscript. We appreciate your positive assessment of the writing quality and the acknowledgment of the wel-lconducted nature of the study.

In regard to the reviewer's comment that "The association with IL-10 cytokine production appears weak," we would like to provide a comprehensive response based on the findings and insights presented in our study (Fig 5). We would like to emphasize several key points to further elucidate this association:

The established knowledge underscores IL-10's capacity to hinder the activation and proliferation of macrophages, thereby safeguarding against an overly aggressive immune-inflammatory reaction (as referenced). In our earlier investigations, we demonstrated that NAD+ orchestrates a systemic generation of IL-10, which assumes a pivotal function in curtailing proinflammatory responses across various conditions, such as autoimmune diseases (as referenced), alloimmunity (as referenced), and bacterial infections (as referenced). In our latest research, we divulge that the introduction of NAD+ leads to an elevated occurrence of IL-10-producing CD4+ T cells, CD8+ T cells, and macrophages, although not dendritic cells (depicted in Figure 5B and C). Furthermore, our comprehensive analyses have substantiated that NAD+ administration thwarts pyroptosis by specifically targeting the non-canonical inflammasome pathway. Intriguingly, our in vitro outcomes suggest that the neutralization of the autocrine IL-10 signaling pathway through a neutralizing antibody and an IL-10 receptor antagonist partially reverses the NAD+-mediated blockage of pyroptosis. These in vitro results imply that NAD+ induces the production of IL-10 cytokines by macrophages, contributing to the suppression of pyroptosis. To corroborate our in vitro conclusions, we employed IL-10 knockout mice and wild-type mice, both treated with either NAD+ or a placebo solution. The wild-type mice treated with NAD+ displayed a survival rate exceeding 80%, whereas the IL-10 knockout mice exhibited a survival rate of "only" 40%. These in vivo findings align with our in vitro discoveries, underscoring the crucial role of NAD+mediated IL-10 cytokine production in impeding pyroptosis through NAD+ and shielding against septic shock. Drawing from our prior and current investigations, we respectfully disagree with the reviewer's characterization of our work as "weak."

Recommendations for the authors

‘’I suggest that animals subject to E. coli infection need to be followed-up for longer and sacrificed at a later time points. It is too difficult to believe that mice are surviving with full resting bacteria in tissues. Do results suggest a full shut-down of the mechanism? What was the level of infiltration of the tissues by neutrophils?’’

‘’I have difficulty to agree with the survival results of the IL-10(-/-) mice of Figure 5E. Can the authors provide the p-values and follow-up for longer? Why the WT and the IL-10(-/-) mice survive the same?’’

Thank you for your thoughtful and constructive comments on our manuscript. We appreciate your valuable insights, and we have carefully considered your suggestions.

We thank the reviewers for this comment. We have indeed followed-up for a longer period of time mice subjected to E. Coli infection and LPS (54mg/kg). Mice infected and treated with NAD+ survived for several months and recovered fully after 10 days. Mice survived for at least a year following infection. We have now included a sentence regarding the long-term survival in the results section of Figure 1 entitled “NAD+ protects mice against septic shock not via bacterial clearance but via inflammasome blockade”. Figure illustrating the level of infiltration of the tissues by neutrophils was added in supplementary data as supplementary figure 4.

In contrast, WT and IL-10-/- mice failed to withstand E. Coli or LPS (54mg/kg) administration when treated with a placebo solution. To our knowledge, our investigation represents the pioneering instance of successfully conferring protection against the lethal doses of E. Coli and LPS administered to animals. Considering the potent immunosuppressive nature of IL-10, our anticipation was that IL-10-/- mice would manifest an exacerbated inflammatory response subsequent to LPS administration, in contrast to WT mice. Our in vivo findings indeed corroborate this assumption, revealing that IL-10-/- mice succumbed more swiftly to LPS administration, displaying statistically significant disparities in survival rates compared to WT mice (p value of 0.0154). The pertinent p-value has been thoughtfully included in Figure 5E of our study.

Reviewer #2 (Public Review):

“Iske et al. provide experimental data that NAD+ lessens disease severity in bacterial sepsis without impacting on the host pathogen load. They show that in macrophages, NAD+ prevents Il1b secretion potentially mediated by Caspase11.

While the in vivo and in vitro data is interesting and hints towards a crucial role of NAD+ to promote metabolic adaptation in sepsis, the manuscript has shortcomings and would profit from several changes and additional experiments that support the claims.

Conceptually, the definition of sepsis is outdated. Sepsis is not SIRS, as in sepsis-2. Sepsis-3 defines sepsis as infection-associated organ dysfunction. This concept needs to be taken into account for the introduction and when describing the potential effects of NAD+ in sepsis. Also, LPS application cannot be considered a sepsis model, since it only recapitulates the consequence of TLR-4 activation. It is a model of endotoxemia. Also, the LPS data does not allow to draw conclusions about bacterial clearance (L135).

The authors state that protective effects by NAD were independent of the host pathogen load. This clearly indicates that NAD confers protection via enhancing a disease tolerance mechanism, potentially via reducing immunopathology. This aspect is not considered by the authors. The authors should incorporate the concept of disease tolerance in their work, cite the relevant literature on the topic and discuss it their findings in light of the published evidence for metabolic alteration sand adaptations in sepsis.

For the in vitro data, the manuscript would benefit from additional experiments using in vitro infection models.

In the merge manuscript, the authors provide two different versions of the figures. In one, bar plots are shown without individual data and in the other with scatter blots. All bar plots need to be provided as scatter plots showing individual values.

The authors should show further serology data for kidney and liver failure etc. as well as further cytokine data such as IL-6 and TNF to better characterize their models.

Careful revision of the entire manuscript, the figure legends and figures is required. The figure legend should not repeat the methods and materials section. The nomenclature for mouse protein and genes needs to be thoroughly revised.

L350. The authors write that they dissect the capacity of NAD+ to dampen auto- and alloimmunity. In this work, no data that supports this statement is shown and experiments with autoantigens or alloantigens are not performed.

L163 The authors describe pyroptosis but in the figure legend call it apoptosis. Specific markers for each cell death should be measured and determined which cell death mechanisms is involved.

Animal data comes from an infection model and LPS application. The RNAseq data is obtained from cells primed with Pam3CSK4 and subsequently subjected to LPS. It is unclear how the cell culture model reflects the animal model. As such the link between IFN signaling and the bacterial infection/LPS model are not convincing and need to be further elaborated.

Figure 5: It is unclear how many independent survival experiments were done, how many mice per group were used and whether the difference between groups was statistical significant. This information should be added.

Further experiments with primary cells from Il10 k.o. and Caspase11 k.o. animals should be provided that support the findings in macrophages.”

Thank you for taking the time to review our manuscript. We appreciate your insightful comments and valuable feedback regarding our study on the role protective role and underlying mechanisms of NAD+ in septic shock.

“While the in vivo and in vitro data is interesting and hints towards a crucial role of NAD+ to promote metabolic adaptation in sepsis, the manuscript has shortcomings and would profit from several changes and additional experiments that support the claims.”

We would like to point out that our current study does not underscore a metabolic adaptation in sepsis but more an immune regulation and a specific blockade of the non-canonical inflammasome signaling machinery.

“Conceptually, the definition of sepsis is outdated. Sepsis is not SIRS, as in sepsis-2. Sepsis-3 defines sepsis as infection-associated organ dysfunction. This concept needs to be taken into account for the introduction and when describing the potential effects of NAD+ in sepsis. Also, LPS application cannot be considered a sepsis model, since it only recapitulates the consequence of TLR-4 activation. It is a model of endotoxemia. Also, the LPS data does not allow to draw conclusions about bacterial clearance (L135).”

Our study uses highly lethal doses of E. Coli or LPS. These doses have been shown to result in multiple organ failure (1, 2). For many decades until now an un-numerable number of studies have used LPS as a model of sepsis (3, 4, 5). We have used LPS animal model based on a study published in 2013 by Kayagaki et al. (1), where the authors reported a novel TLR4-independent mechanism but mediated via activate caspase-11. We used the same animal model to demonstrate the specific role of NAD+ in targeting this TLR4-independent mechanism but mediated via activate caspase-11 and underscore NAD+’s mode of protection.

Moreover, we have not only used LPS but bacterial infection as well using E. Coli. We have also previously published an additional research article demonstrating the protective effect against Listeria Monocytogenes (6). The only model we currently did not use in our current study, is a cecal ligation puncture (CLP) model which is also another common animal model for sepsis.

Our conclusions regarding bacterial clearance are based not only on LPS results but also based on the bacterial load measurement and survival (Figure 1B&C) following E. Coli administration in different tissues (kidney and liver) and not LPS.

“The authors state that protective effects by NAD were independent of the host pathogen load. This clearly indicates that NAD confers protection via enhancing a disease tolerance mechanism, potentially via reducing immunopathology. This aspect is not considered by the authors. The authors should incorporate the concept of disease tolerance in their work, cite the relevant literature on the topic and discuss it their findings in light of the published evidence for metabolic alteration sand adaptations in sepsis.”

We respectfully disagree with the reviewer’s comment and do not believe that NAD+ enhances disease tolerance. We have supporting data indicating that NAD+ mediates protection via a specific blockade of the non-canonical inflammasome pathway, which prevents an over-zealous immune response that results in organ damage and multiple organ failure (MOF). Moreover, we demonstrate that not only NAD+ mediates protection via a specific blockade of the non-canonical inflammasome pathway but prevents septic shock induced death by an additional immunosuppression mediated by the systemic production of IL-10.

Both Caspase-11 and IL-10 pathways are crucial in NAD+ mediated protection against lethal doses of E. Coli and LPS administration. Figure 5A indicates that caspase-11-/- mice treated with PBS have a modest survival rate (~40% survival) when compared to the group of mice treated with NAD+ (>80% survival). These data indicate that NAD+ promotes survival via a caspase-11independent mechanism. Similarly, wild type mice subjected to NAD+ administration exhibited >80% survival, while NAD+ administration to IL-10-/- mice resulted only in a 40% survival rate. Based on these findings, we believe that NAD+ mediated protection against septic shock via a blockade of caspase-11 blockade and by IL-10 cytokine production that dampened the overzealous immune response rather than a disease tolerance.

“For the in vitro data, the manuscript would benefit from additional experiments using in vitro infection models.”

In the current study we have used two in vivo models using LPS and E. Coli a gram-negative bacterium. We have also previously reported the protective role of NAD+ in the context of Listeria Monocytogenes (6) a gram-positive bacterium. In the current study, our aim was to demonstrate the inhibitory role of NAD+ on the non-canonical pathway specifically. We believe that additional in vitro experiments for this study are out of scope.

“In the merge manuscript, the authors provide two different versions of the figures. In one, bar plots are shown without individual data and in the other with scatter blots. All bar plots need to be provided as scatter plots showing individual values.”

As requested by reviewer #2 all bar plots are now provided as scatter plots showing individual values.

“The authors should show further serology data for kidney and liver failure etc. as well as further cytokine data such as IL-6 and TNF to better characterize their models.”

We did not perform further serology analysis, but we did measure IL-6 and TNFα in mice treated with NAD+ or PBS. Mice treated with NAD+ had a reduced systemic level of both cytokines IL-6 and TNFα. We have now added the figures (Figure 1F). In addition, we performed a long-term survival, and all mice treated with NAD+ recovered fully after 10 days and survived over a year after infection. In addition, the mice that survived following NAD+ treatment died of old age.

“Careful revision of the entire manuscript, the figure legends and figures is required. The figure legend should not repeat the methods and materials section. The nomenclature for mouse protein and genes needs to be thoroughly revised.”

A Careful revision of the entire manuscript has been performed.

“L350. The authors write that they dissect the capacity of NAD+ to dampen auto- and alloimmunity. In this work, no data that supports this statement is shown and experiments with autoantigens or alloantigens are not performed.”

We thank the reviewer for this comment. We have now re-phrased our last sentence in the discussion and included references for our previous work. We have now stated:” We have previously reported that NAD+ administration can block auto- (7) and allo-immunity (8) via IL10 cytokine production. Here, we unveiled the capacity of NAD+ to protect against sepsisinduced death via a specific blockade of the non-canonical inflammasome pathway and a robust immunosuppression mediated by IL-10 cytokine production.

L163 The authors describe pyroptosis but in the figure legend call it apoptosis. Specific markers for each cell death should be measured and determined which cell death mechanisms is involved.

We thank the reviewer for this comment. We have focuses on pyoptosis-mediated cell death and not apoptosis. We have now replaced the term “apoptosis” by “pyroptosis-mediated to cell death”.

“Animal data comes from an infection model and LPS application. The RNAseq data is obtained from cells primed with Pam3CSK4 and subsequently subjected to LPS. It is unclear how the cell culture model reflects the animal model. As such the link between IFN signaling and the bacterial infection/LPS model are not convincing and need to be further elaborated.”

Our findings, depicted in Figure 3, pertain exclusively to in vitro investigations rather than in vivo examinations. Our research has demonstrated the selective inhibition of the non-canonical inflammasome pathway by NAD+, with a primary focus on unraveling the specific signaling pathway influenced by NAD+. Our in vitro outcomes indicate that the introduction of recombinant IFN-β counteracted the inhibitory effect of NAD+ on the non-canonical pathway. However, it's important to note that we have not evaluated the IFN-β pathway within our E. Coli and LPS in vivo models. Our primary intention was to exclusively decipher the roles of IFN-β and NAD+ in the context of inhibiting the non-canonical inflammasome, without extending our investigation to the broader in vivo scenarios.

“Figure 5: It is unclear how many independent survival experiments were done, how many mice per group were used and whether the difference between groups was statistical significant. This information should be added.”

We have now included the number of experiments, p values and number of animals used in Figure 5.

“Further experiments with primary cells from Il10 k.o. and Caspase11 k.o. animals should be provided that support the findings in macrophages.”

We concur with the reviewer's suggestion regarding the need for further experiments involving primary cells from IL-10-/- and Caspase-11-/- mice. However, we are uncertain about the potential contribution of these experiments in generating novel or supplementary findings to the existing study.

Recommendations For The Authors:

Besides the comments made in the public section, there are further issues that need to be considered by the authors.

“It is unclear what signifies „impressive, L106" or „dramatic, L257"”

“impressive” meant that we were surprised by the results since to the best of our knowledge prior this study there exists no report/study claiming such survival (>80%) following such high dose of E. Coli. In this aspect protective effects of NAD+ are unique. “dramatic” We (8) and others (9, 10) have previously used this term to describe a robust increase of cytokine production.

“L116. The authors describe „symptoms". It should be clarified what symptoms they observed and the data should be shown. If only temperature is available, then this should be said. It would be interesting to see effects of NAD+ on the glucose levels of the animals during sepsis.”

We thank the reviewer’s comment. We have measured only temperature. We believe that glucose level is beyond the scope of this study.

“L29. Sepsis is not restricted to bacterial and viral pathogens. Also fungi and protozoa can cause sepsis.”

We have now included fungi and protozoa.

“Suppl.Fig.1. A scale should be added.”

Scale has been added

“L822. Lethal dose of LPS would mean that this was lethal for all mice. However, the data suggests that NAD+ treated animals would not have died. This should be clarified.”

Here we meant lethal dose in absence of NAD+ treatment. Our study focuses on the protective role of NAD+ in a lethal context (bacterial and LPS).

“L823/824. The part of the sentence: ... IHC was performed staining for H&E.. is incomplete.”

We thank the reviewer’s comment. We have re-phrased our sentence.

“L804. IL-10 is not a pathway. This should be revised.”

We have replaced “pathway” by” mechanism”.

“The graphical abstract should be the last figure summarizing all findings.”

Figure 4 isn't the final illustration, as it doesn't encompass an overarching graphical summary of our discoveries. Instead, it exclusively highlights the findings related to NAD+'s impact on noncanonical inflammasome inhibition. Notably, this figure omits NAD+-mediated IL-10 cytokine generation and its crucial role in mitigating septic shock.

“The authors report that they used a dosage of 54mg/kg LPS (l.502). This is a rather unusual concentration. How was this determined?”

This was initially based on the first study reporting the role of casapase-11 in septic shock induced death published in 2013 by Kayagaki et al. (1). Many other have used this dosage for septic shock induced death animal model (11, 12, 13).

References:

1. Kayagaki N, et al. Noncanonical inflammasome activation by intracellular LPS independ ent of TLR4. Science 341, 1246‐1249 (2013).

2. Qin, X., Jiang, X., Jiang, X. et al. Micheliolide inhibits LPS-induced inflammatory response and protects mice from LPS challenge. Sci Rep 6, 23240 (2016).

3. Li Z, Qu W, Zhang D, Sun Y, Shang D. The antimicrobial peptide chensinin-1b alleviates the inflammatory response by targeting the TLR4/NF-κB signaling pathway and inhibits Pseudomonas aeruginosa infection and LPS-mediated sepsis. Biomed Pharmacother. 2023 Aug 1; 165:115227.

4. Ramani V, Madhusoodhanan R, Kosanke S, Awasthi S. A TLR4-interacting SPA4 peptide inhibits LPS-induced lung inflammation. Innate Immun. 2013 Dec;19(6):596610.

5. Zhang Y, Lu Y, Ma L, Cao X, Xiao J, Chen J, Jiao S, Gao Y, Liu C, Duan Z, Li D, He Y, Wei B, Wang H. Activation of vascular endothelial growth factor receptor-3 in macrophages restrains TLR4-NF-κB signaling and protects against endotoxin shock. Immunity. 2014 Apr 17;40(4):501-14.

6. Rodriguez Cetina Biefer H, Heinbokel T, Uehara H, Camacho V, Minami K, Nian Y, Koduru S, El Fatimy R, Ghiran I, Trachtenberg AJ, de la Fuente MA, Azuma H, Akbari O, Tullius SG, Vasudevan A, Elkhal A. Mast cells regulate CD4+ T-cell differentiation in the absence of antigen presentation. J Allergy Clin Immunol. 2018 Dec;142(6):18941908.e7.

7. Tullius SG, Biefer HR, Li S, Trachtenberg AJ, Edtinger K, Quante M, Krenzien F, Uehara H, Yang X, Kissick HT, Kuo WP, Ghiran I, de la Fuente MA, Arredouani MS, Camacho V, Tigges JC, Toxavidis V, El Fatimy R, Smith BD, Vasudevan A, ElKhal A. NAD+ protects against EAE by regulating CD4+ T-cell differentiation. Nat Commun. 2014 Oct 7;5:5101.

8. Elkhal A, et al. NAD(+) regulates Treg cell fate and promotes allograft survival via a systemic IL‐10 production that is CD4(+) CD25(+) Foxp3(+) T cells independent. Sci Rep 6, 22325 (2016).

9. Natalia Garcia-Becerra, Marco Ulises Aguila-Estrada, Luis Arturo Palafox-Mariscal, Georgina Hernandez-Flores, Adriana Aguilar-Lemarroy, Luis Felipe Jave-Suarez, FOXP3 Isoforms Expression in Cervical Cancer: Evidence about the Cancer-Related Properties of FOXP3Δ2Δ7 in Keratinocytes, Cancers, 15, 2, (347), (2023).

10. Estelle Bettelli, Maryam Dastrange, Mohamed Oukka. Foxp3 interacts with nuclear factor of activated T cells and NF-κB to repress cytokine gene expression and effector functions of T helper cells. Proceedings of the National Academy of Sciences. 2005.102; 14; 5138-5143.

11. Han Gyung Kim, Chaeyoung Lee, Ji Hye Yoon, Ji Hye Kim, Jae Youl Cho,BN82002 alleviated tissue damage of septic mice by reducing inflammatory response through inhibiting AKT2/NF-κB signaling pathway,Biomedicine & Pharmacotherapy,Volume 148,2022,112740.

12. Tao Q, Zhang Z-D, Qin Z, Liu X-W, Li S-H, Bai L-X, Ge W-B, Li J-Y and Yang Y-J (2022) Aspirin eugenol ester alleviates lipopolysaccharide-induced acute lung injury in rats while stabilizing serum metabolites levels. Front. Immunol. 13:939106.

13. Chen, N, Ou, Z, Zhang, W, Zhu, X, Li, P, Gong, J. Cathepsin B regulate non-canonical NLRP3 inflammasome pathway by modulating activation of caspase-11 in Kupffer cells. Cell Prolif. 2018; 51:e12487.

Author Response:

Reviewer #1:

1. This is a complex paper and would benefit from a schematic depicting the key findings.

This comment is appreciated. Unfortunately, due to time restraints, the authors were not able to graphically depict our findings.

1. The paper would benefit from additional supporting evidence. Would it be possible to measure fatty acid oxidation by metabolic tracing here, in IRG-deficient cells or in response to 4-OI? Although changes in protein level for Cpt1A are seen, this is correlated with fatty acid oxidation rather than direct demonstration. This may be challenging but would strengthen the manuscript.

This is a great comment. While we did not directly measure fatty acid flux in our manuscript, Weiss et al. Nature Metabolism 2023 did these studies in primary hepatocytes. They showed an increased palmitate incorporation into citrate.

1. The aspect concerning body temperature regulation is confusing. Would Itaconate not promote fatty acid oxidation to increase or maintain body temperature? Itaconate must therefore not be involved in the hypothermic response? Bringing UCP1 into the finding is confusing and needs to be better explained. Again a diagram would help, but enhanced BAT fatty acid oxidation and UCP1 expression appear linked here, with both being affected by Itaconate. This needs clarifying.

We appreciate this comment. The rationale is that if itaconate is stabilizing fatty acid oxidation, it would be necessary to fuel thermogenesis, a process dependent on fatty acid utilization. Our data support a role for itaconate in stabilizing body temperature following inflammation, potentially through enhanced fatty acid oxidation. This is evidenced by the hypothermic response to LPS in Acod1 KO mice. Furthermore, Mills et al. Nature 2018 show 4-OI injection boosts body temperature following LPS stimulation.

Reviewer #2:

Some conclusions involving the Irg1 knockout mice require important controls and clarifications to be fully convincing and some controls are missing.

We appreciate the needs for appropriate controls. Negative controls were omitted when baseline phenotypes were not observed. Due to time and resource limitations we were unable to repeat the experiments.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors explored correlations between taste features of botanical drugs used in ancient times and therapeutic uses, finding some potentially interesting associations between intensity and complexity of flavors and therapeutic potential, plus some more specific associations described in the discussion sections. I believe the results could be of potential benefit to the drug discovery community, especially for those scientists working in the field of natural products.

Strengths:

Owing to its eclectic and somehow heterodox nature, I believe the article might be of interest to a general audience. In fact, I have enjoyed reading it and my curiosity was raised by the extensive discussion.

The idea of revisiting a classical vademecum with new scientific perspectives is quite stimulating.

The authors have undertaken a significant amount of work, collecting 700 botanical drugs and exploring their taste and association with known uses via eleven trained panelists.

Weaknesses:

I have some methodological concerns. Was subjective bias within the panel of participants explored or minimized in any manner?

Yes, in all models we included ‘panellist’ as a random effect and therefore any biased perception by a single panellist across drugs or differences among panellists for an individual drug was accounted for. We now make this clearer in our methods.

Were the panelists exposed to the drugs blindly and on several occasions to assess the robustness of their perceptions?

The study was double blind, but blinding was not possible with the more well-known drugs (e.g., almonds, walnuts, thyme, mint). A random number generator was used to assign the drugs to the panellists, and according to the random distribution, some drugs were presented to the same panellist more than once. Robustness of panellists’ perception was not assessed specifically. We have added some text to the methods to clarify.

Judging from the total number of taste assessments recorded and from Supplementary Material, it seems that not every panelist tasted every drug. Why?

Because there were many drugs and panellists had time constraints. Overall, 3973 individual sensory trials were conducted, with an average of 361±153 trials per panellist and 5.7±1.3 trials per botanical drug.

It may be a good idea to explore the similarity in the assessments of the same botanical drug by different volunteers. If a given descriptor was reported by a single volunteer, was it used anyway for the statistical analysis or filtered out?

All responses were used as reported by the panellists, including potential ‘outliers’. As described above, the inclusion of ‘panellist’ as a random effect means that if one individual gave an unusual description of a particular drug in comparison to other individuals, this would be less impactful on any parameter estimates.

The idea of "versatility" is repeatedly used in the manuscript, but the authors do not clearly define what they call "versatile".

In line with suggestions made by reviewers, we have slightly adjusted the definition of therapeutic versatility and have now clearly defined the term on first use. Here, we define therapeutic versatility as the number of therapeutic ‘categories’ a drug is used for (the 25 broad categories are represented by shared iconography in Figure 1). Our revised results include analyses using this definition – which are qualitatively identical to our previous results which defined versatility using the 46 individual therapeutic uses.

The introduction should be expanded. There are plenty of studies and articles out there exploring the evolution of bitter taste receptors, and associating it with a hypothetical evolutionary advantage since bitter plants are more likely to be poisonous.

We agree. Bitter is arguably the most frequent chemosensory attribute of plants and botanical drugs perceived by humans. Our data shows that ‘poisons’ are not associated with bitterness but positively with ‘aromatic’, ‘sweet’ and ‘soapy’ – and negatively with ‘salty’ qualities.

We have added this paragraph to the introduction:

"The perception of taste and flavour (a combination of taste, smell and chemesthesis) here also referred to as chemosensation, has evolved to meet nutritional requirements and are particularly important in omnivores for seeking out nutrients and avoiding toxins (Rozin and Todd, 2016; Breslin, 2013; Glendinning, 2022). The rejection of bitter stimuli has generally been associated with the avoidance of toxins (Glendinning, 1994; Lindemann, 2001; Breslin, 2013) but to date no clear relationship between bitter compounds and toxicity at a nutritionally relevant dose could be established (Glendinning, 1994; Nissim et al., 2017). While bitter tasting metabolites occurring in fruits and vegetables have been linked with a lower risk for contracting cancer and cardiovascular diseases (Drewnoswski and Gomez-Carneros, 2000) the avoidance of pharmacologically active compounds is probably the reason why many medicines, including botanical drugs, taste bitter (Johns, 1990; Mennella et al., 2013)."

And expanded in the discussion:

"Though many bitter compounds are toxic, not all bitter plant metabolites are (Glendinning, 1994; Drewnoswski and Gomez-Carneros, 2000; e.g., iridoids, flavonoids, glucosinolates, bitter sugars). In part, this may be the outcome of an arms race between plant defence and herbivorous mammals’ bitter taste receptor sensitivities, resulting in the synthesis of metabolites capable of repelling herbivores and confounding the perception of potential nutrients by mimicking tastes of toxins. Here, poisons showed no association with bitter but positive associations with aromatic (px = 0.041), sweet (px = 0.022) and soapy (px = 0.025) as well as a negative association with salty (px = 0.046) qualities."

Since plant secondary metabolites are one of the most important sources of therapeutic drugs and one of their main functions is to protect plants from environmental dangers (e.g., animals), this evolutionary interplay should be at least briefly discussed in the introductory section.

This is now referred to in the introduction as well as in the discussion.

Since the authors visit some classical authors, Parecelsus' famous quote "All things are poison and nothing is without poison. Solely the dose determines that a thing is not a poison" may be relevant here. Also note that some authors have explored the relationship between taste receptors and pharmacological targets (e.g., Bioorg Med Chem Lett. 2012 Jun 15;22(12):4072-4).

We agree that pharmacologic action is determined by the dose. We now refer to the dose in the introduction: “…to date no clear relationship between bitter compounds and toxicity at a nutritionally relevant dose could be established (Glendinning, 1994; Nissim et al., 2017)”.

We are aware of the fact that several authors have explored the relationship between taste receptors as targets and their similarity with other targets. We use many examples from the literature to explain our data. Our analysis did, however, not highlight any association between sweet tastes and epilepsy (as reported in Bioorg Med Chem Lett. 2012 Jun 15;22(12):4072-4)). We are not able to explain all associations, and we acknowledge that there may be more associations between chemosensory receptors and therapeutic effects than those found and discussed here.

Reviewer #2 (Public Review):

Summary:

This is an unusual, but interesting approach to link the "taste" of plants and plant extracts to their therapeutic use in ancient Graeco-Roman culture. The authors used a panel of 11 trained tasters to test ~700 different medicinal plants and describe them in terms of 22 "taste" descriptors. They correlated these descriptors with the plant's medical use as reported in the De Materia Medica (DMM 1st Century, CE). Correcting for some of the plants' evolutionary phylogenetic relationships, the authors found that taste descriptors along with intensity measures were correlated with the "versatility" and/or specific therapeutic use of the medicine. For example, simple but intense tastes were correlated with the versatility of a medicine. Specific intense tastes were linked to versatility while others were not; intense bitter, starchy, musky, sweet, cooling, and soapy were associated with versatility, but sour and woody were negatively associated. Also, some specific tastes could be associated with specific uses - both positive and negative associations. Some of these findings make sense immediately, but others are somewhat surprising, and the authors propose some links between taste and medicinal use (both historical and modern use) in the discussion. The authors state that this study allows for a re-evaluation of pre-scientific knowledge, pointing toward a central role of taste in medicine.

Strengths:

The real strength of this study is the novelty of this approach - using modern-day tasters to evaluate ancient medicinal plants to understand the potential relationships between taste and therapeutic use, lending some support to the idea that the "taste" of a medicine is linked to its effectiveness as a treatment.

Weaknesses:

While I find this study very interesting and potentially insightful into the development and classification of certain botanical drugs for specific medicinal use, I would encourage the authors to revise the manuscript and the accompanying figures significantly to improve the reader's understanding of the methods, analyses, and findings. A more thorough discussion of the limitations of this particular study and this general type of approach would also be very important to include.

Figures were revised, one deleted (former Fig. 3), and another one put to the supplementary (former Fig. 4, now Figure supplement 1). We now acknowledge limitations in the final paragraph.

The metric of versatility seems somewhat arbitrary. It is not well explained why versatility is important and/or its relationship with taste complexity or intensity.

We have modified the definition of versatility in line with reviewers’ comments. We have provided a detailed explanation of this in our response to reviewer #1 but for ease of reference, we paste this again here:

Here, we define therapeutic versatility as the number of therapeutic ‘categories’ a drug is used for (the 25 broad categories are represented by shared iconography in Figure 1). Our revised results include analyses using this definition – which are qualitatively identical to our previous results which defined versatility using the 46 individual therapeutic uses.

The importance of versatility was not the focus but the impact of taste intensity and complexity on versatility. We hypothesize that associations between perceived complexity and intensity of chemosensory qualities with versatility of botanical drug use provides insights into the development of empirical pharmacological knowledge and therapeutic behaviour (now included in the introduction).

Similarly, the rationale for examining the relationships between individual therapeutic uses and taste intensity/complexity is not well explained, and given that a similar high intensity/low complexity relationship is common for most of the therapeutic uses, it restates the same concepts that were covered by the initial versatility comparison.

The examination of the relationships between individual therapeutic uses and taste intensity/complexity fine-tunes the overall analysis and shows that this concept is applicable in general. However, in general, the reviewer is correct, and this is not our main focus. We therefore shifted the analysis including the figure to the supplementary material and state in the discussion: “We also detected nuances in significance, and complete absence of significance across the relationships between individual therapeutic uses and complexity/intensity magnitudes for which we lack, however, more specific explanations (Figure supplement 1).

There are multiple issues with the figures - the use of icons is in many cases counterproductive and other representations are not clear or cause confusion (especially Figure 3).

We have excluded former Fig. 3. Otherwise, the use of iconography is to facilitate graphical representation and cross-referencing between figures without over-cluttering. We provide all text and numeric values in the supporting information if individual detail is required.

The phylogenetic information about the botanicals is missing. Also missing is any reference/discussion about how that analysis was able to disambiguate the confounding effects of shared uses and tastes of drugs from closely related species.

This is explained in the methods (sections: ‘Phylogenetic tree’ and ‘statistical procedure’). We highlight that all models showed high heritability which means that shared ancestry has a statistical influence on the model. The trees themselves are now represented in our modified Figure 2.

Reviewer #1 (Recommendations For The Authors):

Besides the points already covered in my public review, I believe it would be interesting to assess and discuss the differences between the category "food" (how many drugs were allocated there?) and the drugs used for therapeutic purposes. In this manner, the food category could serve as a retrospective negative control to test the authors' hypotheses. Does the food category include drugs of weak flavor? Does it include drugs of complex flavor?

All drugs in this database are associated with therapeutic uses. Only 96 are specifically mentioned to be also used as food while in total at least 152 are also used as food (many of the most obvious food drugs are not labelled as such in DMM). It is difficult to use the food category as a negative control (for testing whether food drugs have weaker tastes), because spices are included in the food category. If at all, only staples should be used for such an analysis. But this would be another study.

In the context of the present analyses, we do agree that there is interest and so we have therefore added a small section to our manuscript: The 96 botanical drugs specifically mentioned also for food (though there are more than 150 edible drugs in our dataset; Supplementary file 1) show positive associations with starchy (px = 0.005), nutty (px = 0.002) and salty (px = 0.001) and negative associations with bitter (px = 0.007), woody (px = 0.001) and stinging (px = 0.033) tastes and flavours.

Please replace "plant defence" with "plant defense".

Currently the whole MS is formatted BE. We are happy to revise on the basis of editorial policy.

Reviewer #2 (Recommendations For The Authors):

1. I would encourage replacing "taste" with "flavor" throughout the manuscript and in the title because this paper addresses "taste here defined as a combination of taste, odour and chemesthesis" which essentially is the definition of flavor, and should not be simplified to taste. Flavor is the more precise word, and there is no need to confuse readers by defining "taste" in this way when taste means just the gustatory aspect of flavor.

We now define flavour as a combination of taste, smell and chemesthesis and use ‘taste’ when referring to a specific taste quality. We use the term ‘chemosensory’ (perception, quality) and chemosensation for addressing the perception of both, taste and flavour qualities together. The abstract now reads: “The perception of taste and flavour (a combination of taste, smell and chemesthesis) here referred to as chemosensation, enables animals to find high-value foods and avoid toxins.”

We prefer to leave the title as it is in accordance with standard books (e.g., “Pharmacology of Taste” by Palmer and Servant) which address all kinds of chemosensory interactions and the fact that we’ve conducted a ‘tasting panel’ (and not a ‘flavour panel’), and because flavour as a concept is only used in English (and also there not consistently, with ‘taste’ being the preferred term used by English native speakers for describing perception where in a strict sense, ‘flavour’ would be the correct term, see Rozin P. "Taste-smell confusions" and the duality of the olfactory sense. Percept Psychophys. 1982 Apr;31(4):397-401)) and maybe also in French.

1. Methods - A much more detailed description of how the samples were prepared for the taste tests is needed. Were they sampled as a dry powder? No, they were sampled as dried pieces. We have added more information to our methods section to clarify.

Why is there such a big range in the amount provided (.1 to 2 g)? Because certain drugs are highly toxic (aconitum, opium) we could only provide a relatively small amount (that still permitted the perception of taste qualities). For practical reasons, half a walnut was dispensed. We have added more information to our methods section to clarify.

Also "Panelists were instructed to spit, rinse their mouth with drinking water and to take a break before tasting the next sample" This seems more likely that the samples were dissolved in a liquid if they were spitting and rinsing, but this is not clear. Also - take a break for how long between samples?

Panellists were instructed to chew the amount of sample necessary for taste perception, to annotate their perception, and to spit out residues of samples and finally rinse their mouth with drinking water. The breaks between tasting different samples depended on chemosensory persistence. We have added more information to our methods section to clarify.

How many samples were tested per day?

The number of tasted samples was different from panelist to panelist and depending on available time frames. On average each panellist tasted 17,2 drugs per hour using 10.5 sessions (18 sessions in total) lasting approximately two hours each. We have added more information to our methods section to clarify.

Did individual panelists get repeated samples?

Random distribution permitted that individual panellists were challenged also with repeated samples. We have added more information to our methods section to clarify.

1. Methods - Phylogenetic tree - Where is the output of this tree? It should be included in the figures and referred to in the results/discussion where the authors claim that they have been able to disambiguate phylogenetic closeness with taste and medicinal use.

We did not ‘build’ a phylogenetic tree, rather we modified an existing one. Therefore, the wording of that section in the methods has been adjusted for clarity. We refer to the tree in the results pertaining to phylogenetic relatedness by explicitly quantifying the extent of phylogenetic signal using the widely used heritability (h2) statistic. This means that shared ancestry has a statistical influence on the model. We have also added to our Figure 2 representations of the phylogenetic tree we used in our analysis, limited to the species for which we have data, also displaying the data (in this case, intensity and complexity) at the tips.

1. Taste intensity ratings should be better explained. Since the panelists are evaluating different amounts of samples (.1 to 2g) wouldn't the intensity of taste also depend on the amount of the substance?

The panelists were not told to introduce all the sample into their mouth but just enough to perceive the taste qualities clearly (explanation given in methods). E.g.: one black pepper corn is normally enough to perceive the taste and flavour of pepper while the same amount of hazelnut would be insufficient.

Or is this measure a relative value - "woodiness" vs "sourness" for example within the sample is strong/weak?

Chemosensation and sensory perception in general is always relative. (For instance, currently I can hear the birds singing outside. Was there music playing in my room I wouldn’t be able to hear them).

Because of this - are samples with strong tastes less likely to seem complex because the intensity of one stimulus masks the other?

Yes, we argue that drugs with strong tastes/flavours are less likely be perceived as being complex (fewer individual qualities perceived), arguably because strong stimuli overshadow weaker ones. We currently address this in the discussion and have made some modifications in line with the below comment.

This issue was presented briefly in the discussion when addressing the finding that samples with intense, but fewer tastes were more versatile, but this was highly confusing.

The authors presented both sides of the problem without referring to any of their own experiments to resolve the issue, or to highlight this as a potential limitation of the study at hand.

Yes, stronger tastes mask weaker tastes which addresses both sides of the problem.

We have modified the first paragraph of the discussion to make this clearer.

It now reads: "Unexpectedly, botanical drugs eliciting fewer but intense chemosensations were more versatile (Fig. 2). People often associate complexity with intensity, and taste complexity is popularly interpreted with a higher complexity of ingredients (Spence, and Wang, 2018). However, simple tastes can be associated with complex chemistry when intense tastes mask weaker tastes, or when tastants are blended (Breslin and Beauchamp, 1997; Green et al., 2010). For example, starchy flavours or sweet tastes can be sensed when bitter and astringent antifeedant compounds are present below a certain threshold while salts enhance overall flavour by suppressing the perception of bitter tastants (Breslin and Beauchamp, 1997; Johns, 1990). On the other hand, combinations of different tastants or olfactory stimuli do not necessarily result in increased perceived complexity (Spence and Wang, 2018; Weiss et al., 2012)."

It would be useful to understand the parameters a bit more - a data visualization of the relationships of intensity and complexity across all samples would be a welcome addition to Figure 2.

Shared ancestry has a statistical influence on the model. We have now also added to our Figure 2 representations of the phylogenetic tree we used in our analysis, limited to the species for which we have data, also displaying the data (in this case, intensity and complexity) at the tips.

1. "Therapeutic Versatility" is a measure of how many different therapeutic uses a given botanic is listed in the DMM. This is one of the primary comparisons of this study, but the authors do not provide much of a rationale for using this metric. Also, there are 46 therapeutic uses, but many are interrelated such as gastric, gynecology, muscle, neurological, respiratory, skin, and kidney. It is not clear in my reading of the methods if this was also treated in some type of "phylogeny" as well or not. I would assume a real therapeutic versatility metric should be higher for something used for cough, ulcers, gout, and menses rather than something that was used for 4 different, but skin-related complaints.

The reviewer is correct, and we appreciate this comment. We have modified the definition of versatility in line with the suggestions laid out here. We have provided a detailed explanation of this in our public responses but for ease of reference, we paste this again here:

Here, we define therapeutic versatility as the number of therapeutic ‘categories’ a drug is used for (the 25 broad categories are represented by shared iconography in Figure 1). Our revised results include analyses using this definition – which are qualitatively identical to our previous results which defined versatility using the 46 individual therapeutic uses.

We repeated our original ‘versatility’ analyses using the 25 broader categories rather than the 46 individual uses. The results remained largely the same.

1. Use of icons/pictorial representations in figures. Overall, the use of icons is not necessary - words could be used, and then readers would not need to keep going back and forth to the key in Figure 1 to identify the taste/use. I am very confused by Figure 3. How is the strength of taste shown in this figure? The use of the balance is a confusing representation since I don't associate strength/intensity with weight. Also there are specific tastes that are used more, and others that are used less (but the numbers of those are also more/less). I do not think this figure accomplishes the goal of relaying these findings.

Whilst we agree that iconography is not strictly necessary, we think it is a good way of graphically representing the results without over-crowding the figures or introducing text sizes too small to read in print. All values are provided in the supporting information if any individual detail is required.

We have decided on the basis of these comments to exclude former Fig. 3 and (Figure supplement 1). We hope that the removal of this figure and clearer signposting towards the text and numerical tables in the supplementary information alleviates the reviewer’s concerns.

1. Similarly, figure 4 is unclear. This could be better represented in a table with words and p values listed. But a larger issue is that this shows essentially the same overarching relationship across the therapeutic use cases - high intensity, low complexity. Only the pink kidney (other?) case differs from this pattern. In the discussion, several therapeutic uses are discussed that could need intense tasting medicine - but these are not related directly back to the relationships shown in Figure 4.

Yes, we agree with the reviewer and have now moved Fig. 4 to the supplementary (Figure supplement 1)

Author Response

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

Note to all Reviewers

We appreciate the reviewers’ comments and suggestions for improving the manuscript. Below is a summary of new data added and a brief description of the major new results. A detailed pointby-point response follows.

New data:

• Figure 1f

• Figure 2b, f, g

• Figure 4b

• Figure S7 • Figure S8

• Figure S9

Summary of major new results/edits:

• At the request of Reviewer #1 we have updated the name of the degradation tag to be more specific and we now call it the “LOVdeg” tag.

• We have added new controls demonstrating that light stimulation does not cause photobleaching or toxicity issues (Fig. S7).

• We now show that LOVdeg can function at various points in the growth cycle, demonstrating robust degradation (Fig. 1f, Fig. S8).

• We have included relevant controls for the AcrB-LOVdeg efflux pump results (Fig. 2f-g).

• We have included important benchmarking controls, such as an EL222-only control and SsrA tag control to provide a clearer view of how LOVdeg performance compares to other systems (Fig. S9, Fig. 4b).

Additional note:

• While repeating experiments during the revision process we found that the results for the combined action of EL222 and the LOVdeg tag were not as dramatic as in our original measurements, though the overall findings are consistent with our original results. Specifically, we still find that the combination of EL222 and the LOVdeg tag produces a lower signal than either on their own. We have updated these data in the revised manuscript (Fig. 4b).

Reviewer #1:

Public Review:

Specifically controlling the level of proteins in bacteria is an important tool for many aspects of microbiology, from basic research to protein production. While there are several established methods for regulating transcription or translation of proteins with light, optogenetic protein degradation has so far not been established in bacteria. In this paper, the authors present a degradation sequence, which they name "LOVtag", based on iLID, a modified version of the blue-light-responsive LOV2 domain of Avena sativa phototropin I (AsLOV2). The authors reasoned that by removing the three C-terminal amino acids of iLID, the modified protein ends in "-E-A-A", similar to the "-L-A-A" C-terminus of the widely used SsrA degradation tag. The authors further speculated that, given the light-induced unfolding of the C-terminal domain of iLID and similar proteins, the "-E-A-A" C-terminus would become more accessible and, in turn, the protein would be more efficiently degraded in blue light than in the dark.

Indeed, several tested proteins tagged with the "LOVtag" show clearly lower cellular levels in blue light than in the dark. While the system works efficiently with mCherry (10-20x lower levels upon illumination), the effect is rather modest (2-3x lower levels) in most other cases. Accordingly, the authors propose to use their system in combination with other light-controlled expression systems and provide data validating this approach. Unfortunately, despite the claim that the "LOVtag" should work faster than optogenetic systems controlling transcription or translation of protein, the degradation kinetics are not consistently shown; in the one case where this is done, the response time and overall efficiency are similar or slightly worse than for EL222, an optogenetic expression system.

The manuscript and the figures are generally very well-composed and follow a clear structure. The schematics nicely explain the underlying principles. However, limitations of the method in its main proposed area of use, protein production, should be highlighted more clearly, e.g., (i) the need to attach a C-terminal tag of considerable size to the protein of interest, (ii) the limited efficiency (slightly less efficient and slower than EL222, a light-dependent transcriptional control mechanism), and (iii) the incompletely understood prerequisites for its application. In addition, several important controls and measurements of the characteristics of the systems, such as the degradation kinetics, would need to be shown to allow a comparison of the system with established approaches. The current version also contains several minor mistakes in the figures.

We thank reviewer #1 for the feedback and suggestions to strengthen the manuscript. We have addressed these comments in the points that follow and now include important controls and benchmarks for our molecular tool.

Major points

1. The quite generic name "LOVtag" may be misleading, as there are many LOV-based tags for different purposes.

We appreciate that it would be beneficial to have a more specific name. We have updated the name to “LOVdeg” tag, which captures both the inclusion of LOV and the degradation function of the tag.

Updated throughout the manuscript and figures

1. Throughout the manuscript, the authors use "expression levels". As protein degradation is a post-expression mechanism, "protein levels" should be used instead.

We have transitioned to using “protein levels” at many points in the manuscript.

Updated throughout the manuscript

1. Degradation dynamics (time course experiments) should be shown. The only time this is done in the current version (in Fig. 4), degradation appears to be in the same range (even a bit slower) than for EL222, which does not support the claim that the "LOVtag" acts faster than other optogenetic systems controlling protein levels.

In the revised manuscript, time course data are now shown at multiple points. These include new data in Fig. 1f and Fig. S8 that demonstrate degradation at various stages of growth. Fig. S4 also shows the dynamics of degradation when comparing to the addition of exogenously expressed ClpA. We have added text in the results section to point the reader to these data. In addition, we have made minor modifications to the text in the Introduction to avoid making claims about speed comparisons. Fig. 1f, Fig. S8, Fig. S4

Results: Design and characterization of the AsLOV2-based degradation tag, Introduction

1. "Frequency" is used incorrectly for Fig. 3. A series of 5 seconds on, 5 seconds off corresponds to a frequency of 0.1 Hz (1 illumination round / 10 s), not of 0.5 Hz. What the authors indicate as "frequency" is the fraction of illumination time. However, the (correct) frequency should be given, as this is likely the more important factor.

We have changed how we calculate frequency to use the proposed definition of one pulse per time period. We updated the values in the text and in the figure. Fig. 3c

Results: Tuning frequency response of the LOVdeg tag

1. To properly evaluate the system, several additional controls are needed:

a. To test for photobleaching of mCherry by blue light illumination, untagged controls should be shown for the mCherry-based experiments. Fluorescence always seems to be lower upon illumination, except for the AsLOV2*(546) data, where it cannot be excluded that fluorescence readings are saturated. Relatedly, the raw data for OD and fluorescence should be included. Showing a Western blot against mCherry in at least one case would allow to separate the effects of photobleaching and degradation.

We appreciate the suggestion and have conducted these important controls. We now include new data demonstrating that light induction does not change fluorescence levels using an untagged mCherry control, nor does it significantly affect endpoint OD levels. Based on these results, we did not perform a Western blot because there were no effects to separate. Fig. S7

b. In Fig. 2b, light + IPTG should be shown to estimate the activity of the system at higher expression levels.

We have added these to the figure. Light + IPTG modestly increases expression compared to IPTG only, likely due to the saturating level of IPTG added, which achieves near full induction. Fig. 2b

c. In Fig. 4, EL222 alone should be shown to allow a comparison with the LOVtag. From the data presented, it looks like EL222 is both slightly faster and more efficient than the LOVtag.

We have added the EL222-only case for comparison with LOVdeg only and EL222 + LOVdeg. We note that Reviewer #3 raised a similar concern. Fig. 4b

d. The effect of the used light on bacterial viability under exponential and stationary conditions should be shown.

In this revision, we have added new data on light exposure at various points during exponential and stationary phase (Fig. 1f, Fig. S8). These OD data show that growth curves are similar for all cultures, regardless of the time light is applied during the growth phase. Additionally, we also now include ODs for the photobleaching experiments. These data also show that growth is not significantly altered under continuous light exposure. Figure 1f, Fig. S7b

1. The claim that "Post-translational control of protein function typically requires extensive protein engineering for each use case" is not correct. The authors should discuss alternative options, e.g. based on dimerization, more extensively and in a less biased manner.

We have toned down the language in this location and at other points in the manuscript. However, we maintain that other types of post-translational control, such as dimerization or LOV2 domain insertion, require more protein engineering than inserting a degradation tag. For example, we and others have directly demonstrated this in previous work (e.g. DOI: 10.1021/acssynbio.9b00395, 10.1101/2023.05.26.542511, 10.1038/s41467-023-38993-6), where numerous split site or insertion variants need to be screened and fine-tuned for successful light control. In contrast, a degradation mechanism has the potential to require less fine tuning to achieve a light response. We have included the above sources to clarify this point. Introduction, Results: Modularity of the LOVdeg tag

Minor points

1. In Suppl. Fig. 1, amino acid numbers seem to be off. Also, the alterations in iLID (compared to AsLOV2) that are not used in "LOVtag" appear to be missing and the iLID sequence incorrect, as a consequence.

Thank you for catching this. The number indices in Fig. S1 have been corrected. We also realized we were reporting the iLID(C530M) variant in our amino acid sequence and have reverted the 530M back to C. Fig. S1

1. Why is AsLOV2(543) more efficiently degraded than AsLOV2(543) (blue column in Fig. 1d) when the dark state should be stabilized in AsLOV2(543)?

We are not sure of the exact reason for the increased degradation response in the AsLOV2*(543) variant. It may be that the dark-state stabilizing mutations introduced also have more favorable interactions with degradation machinery, although this is highly speculative.

1. Why does the addition of EL222 reduce protein levels so strongly in the dark for CpFatB1* (Fig. 5)?

We believe this effect stems from the EL222 responsive promoter (PEL222). With LOVdeg only, CpFatB1* is expressed from an IPTG inducible promoter (PlacUV5) whereas EL222 responsive constructs necessitate a promoter switch containing an EL222 binding site. We have clarified this point and expanded our discussion of these results.

Results: Optogenetic control of octanoic acid production

1. Fig. 2f / S10 are difficult to interpret. Why does illumination only lead to a significant effect at 2.5 and 5 µg/ml and not at lower concentrations, where the degradation system would be expected to be most efficient?

We have expanded our discussion on these results to explain that this likely stems from basal protein levels of AcrB-LOVdeg in the light that can provide resistance at low antibiotic concentrations. We have also added new controls to this figure to show the chloramphenicol sensitivity of a ΔacrB strain and a ΔacrB strain with an IPTG-inducible version of acrB with no induction, demonstrating the lowest achievable chloramphenicol resistance from a standard inducible system.

Results: Modularity of the LOVdeg tag, Fig. 2f-g

1. Fig. 2f / S10 do not measure the MIC (which is a clearly defined value), but the sensitivity to Chloramphenicol.

We have changed the text to use the term chloramphenicol sensitivity instead of MIC. Results: Modularity of the LOVdeg tag

1. "***" in Fig. S1 should be explained.

We have removed the ‘***’ to avoid confusion. Fig. S1

1. The fold-change differences between light and dark, indicated in some selected cases, should be listed for all figures.

We have added fold-change values where appropriate. Fig 1d, Fig. 2b

Reviewer #2:

Public Review:

In this manuscript the authors present and characterize LOVtag, a modified version of the bluelight sensitive AsLOV2 protein, which functions as a light-inducible degron in Escherichia coli. Light has been shown to be a powerful inducer in biological systems as it is often orthogonal and can be controlled in both space and time. Many optogenetic systems target regulation of transcription, however in this manuscript the authors target protein degradation to control protein levels in bacteria. This is an important advance in bacteria, as inducible protein degradation systems in bacteria have lagged behind eukaryotic systems due to protein targeting in bacteria being primarily dependent on primary amino acid sequence and thus more difficult to engineer. In this manuscript, the authors exploit the fact that the J-alpha helix of AsLOV2, which unwinds into a disordered domain in response to blue light, contains an E-A-A amino acid sequence which is very similar to the C-terminal L-A-A sequence in the SsrA tag which is targeted by the unfoldases ClpA and ClpX. They truncate AsLOV2 to create AsLOV2(543) and combine this truncation with a mutation that stabilizes the dark state to generate AsLOV2*(543) which, when fused to the C-terminus of mCherry, confers light-induced degradation. The authors do not verify the mechanism of degradation due to LOVtag, but evidence from deletion mutants contained in the supplemental material hints that there is a ClpA dominated mechanism. They demonstrate modularity of this LOVtag by using it to degrade the LacI repressor, CRISPRa activation through degradation of MCP-SoxS, and the AcrB protein which is part of the AcrAB-TolC multidrug efflux pump. In all cases, measurement of the effect of the LOVtag is indirect as the authors measure reduction in LacI repression, reduction in CRISPRa activation, and drug resistance rather than directly measuring protein levels. Nevertheless the evidence is convincing, although seemingly less effective than in the case of mCherry degradation, although it is hard to compare due to the different endpoints being measured. The authors further modify LOVtag to contain a known photocycle mutation that slows its reversion time in the dark, so that LOVtag is more sensitive to short pulses of light which could be useful in low light conditions or for very light sensitive organisms. They also demonstrate that combining LOVtag with a blue-light transcriptional repression system (EL222) can decrease protein levels an additional 269-fold (relative to 15-fold with LOVtag alone). Finally, the authors apply LOVtag to a metabolic engineering task, namely reducing expression of octanoic acid by regulating the enzyme CpFatB1, an acyl-ACP thioesterase. The authors show that tagging CpFatB1 with LOVtag allows light induced reduction in octanoic acid titer over a 24 hour fermentation. In particular, by comparing control of CpFatB1 with EL222 transcriptional repression alone, LOVtag, or both the authors show that light-induced protein degradation is more effective than light-induced transcriptional repression. The authors suggest that this is because transcriptional repression is not effective when cells are at stationary phase (and thus there is no protein dilution due to cell division), however it is not clear from the available data that the cells were in stationary phase during light exposure. Overall, the authors have generated a modular, light-activated degron tag for use in Escherichia coli that is likely to be a useful tool in the synthetic biology and metabolic engineering toolkit.

We thank Reviewer #2 for the constructive feedback. In the updated manuscript, we now include data demonstrating degradation at different growth stages and address other points brought up in the review to improve understanding of the degradation tag.

Overall, the authors present a well written manuscript that characterizes an interesting and likely very useful tool for bacterial synthetic biology and metabolic engineering. I have a few suggestions that could improve the presentation of the material.

Major Comments:

• Could the authors clarify, perhaps through OD measurements, that the cultures in the octanoic acid experiment are actually in stationary phase during the relevant light induction. It isn't clear from the methods.

We have updated the Methods to clarify that the cells are entering stationary phase (OD600 = 0.6) when light is either kept on or turned off for production experiments. Production is continued for the following 24 hours. Note that we now show OD measurements in a separate set of experiments (Fig. 1f, Fig. S8).

Methods: Octanoic acid production experiment. Fig. 1f, Fig. S8

• Can the authors clarify why there is an overall decrease in protein in the clpX deletion? And is it this initial reduction that is the source of the change in fold in 1C? Similarly, for hslU is it because overall protein levels are higher with the tag? In general, I feel that the interpretation of Supplemental Figures S6-S10 could be moved in more detail to the main text, or at least the main takeaway points. But this is a personal preference, and not necessary to the major flow of the story which is about the utility of the LOVtag tool.

As shown in Fig. S5, expression of mCherry without any degradation tag is decreased in a clpX knockout strain compared to wild type. This difference may be the result of reduced cell health, and we now note this in the text. The strains shown in Fig. 1c are in wild type cells with normal expression, so this is not the source of the fold change. As for hslU, we agree it is interesting that expression seems to increase. However, the increase is modest and could stem from gene network regulation differences in that strain compared to wild type and may not be related to LOVdeg tag degradation. Each endogenous protease is involved in a wide range of functions within the cell, and it is unknown how global gene expression is impacted. We acknowledge the suggestion of moving the protease results to the main text, but we have ultimately elected to keep these data in the Supplementary Information to maintain the flow in the manuscript. However, we have added additional text pointing the reader to the Supplemental Text and include a brief summary of the findings in the main text.

Results: Design and characterization of the AsLOV2-based degradation tag

• What is the source of the poor repression in Figure 2D?

Presumably, this stems from low levels of the CRISPRa MCP-SoxS activator, even in the presence of light. We have added this point to the text.

Results: Modularity of the LOVdeg tag

• In general, it would be nice to have light-only controls for many of the experiments to validate that light is not affecting the indicated proteins or their function.

We thank the reviewer for this suggestion and note that Reviewer #1 raised a similar concern. We have now included light-only data for a strain containing IPTG-inducible mCherry without the LOVdeg tag (Fig. S7). These data show that light itself, at the levels used in this study, does not affect mCherry expression or cell growth. This strain serves as a direct control for data presented in Fig. 1 and Fig. 2b, as the systems are identical except for the addition of the LOVdeg tag onto either mCherry or the LacI repressor. Additionally, the control translates to other experiments since mCherry is used as a reporter for other systems in this study. Fig. S7

• It would be nice to directly measure the function of the tool at different phases of E. coli growth to show directly that protein degradation works at stationary phase, rather than the more indirect measurements used in the octanoic acid experiment.

We thank the reviewer for this suggestion, which significantly strengthens our results. We have added an experiment that tests the LOVdeg tag at different phases of growth (Fig. 1f, Fig. S8). In this experiment, cultures are growth from early exponential to stationary phase, and light is introduced at various points. Exposure windows of 4 hours, ranging from early exponential to stationary phase, all show functional light inducible degradation. Fig. 1f, Fig. S8.

Results: Design and characterization of the AsLOV2-based degradation tag

Minor Comments:

• It would be nice to make clear that the data in S6d and S7 is repeated, but with the HslUV data in S7.

We clarified this point in the caption of Fig. S4 (the former Fig. S7 in the original manuscript). Fig. S4 caption

• Why was 5s picked for the frequency response in Figure 3

We picked 5s because 1) it is a substantially shorter timescale than overall degradation dynamics seen for the LOVdeg tag, and 2) we found that shorter pulses could not be reliably achieved with the light stimulation hardware and software we used (Light Plate Apparatus with Iris software). To ensure high fidelity pulses, we opted for 5 second pulses that we empirically determined to be stable throughout long experiments. We have added text clarifying this. Results: Tuning frequency response of the LOVdeg tag

Reviewer #3:

Public Review:

The authors present the mechanism, validation, and modular application of LOVtag, a light-responsive protein degradation tag that is processed by the native degradosome of Escherichia coli. Upon exposure to blue light, the c-terminal alpha helix unfolds, essentially marking the protein for degradation. The authors demonstrate the engineered tag is modular across multiple complex regulatory systems, which shows its potential widespread use throughout the synthetic biology field. The step-by-step rational design of identifying the protein that was most dark stabilized as well as most light-responsive for degradation, was useful in terms of understanding the key components of this system. The most compelling data shows that the engineered LOVTag can be fused to multiple proteins and achieve light-based degradation, without affecting the original function of the fused protein; however, results are not benchmarked against similar degradation tagging and optogenetic control constructs. Creating fusion proteins that do not alter either of the original functions, is often difficult to achieve, and the novelty of this should be expanded upon to drive further impact.

We appreciate the feedback from Reviewer #3 to improve the manuscript. We have included important controls and benchmarking experiments to address the reviewer’s concerns, which are detailed in the points below.

Benchmarking:

The similarity between the L-A-A sequence of SsrA and the E-A-A sequence of LOVtag is one of the pieces of evidence that led the authors to their current protein design. The differences in degradation efficiency between the SsrA degradation tag and LOVtag are not shown, and benchmarking against SsrA would be a valuable way to demonstrate the utility of this construct relative to an established protein tagging tool.

We thank the reviewer for suggesting an experiment to benchmark performance. We have added new experimental data where a full length SsrA tag is added to a fusion protein of nearly identical size (mCherry-iLID), allowing us to directly compare performance to mCherryLOVdeg (Fig. S9). These results show that light inducible control with LOVdeg tag decreases protein expression levels to near those achieved with the native SsrA tag. Fig. S9.

Results: Design and characterization of the AsLOV2-based degradation tag

Additionally, there is a lack of an EL222-only control presented in Figure 4b and in the results section beginning with "Integrating the LOVtag with EL222...". Without benchmarking against this control the claim that "EL222 and the LOVtag work coherently to decrease expression" is unsubstantiated. No assumptions of synergy can be made.

We appreciate this comment and note that Reviewer #1 raised a similar concern. We have added data to Fig. 4b with an EL222-only control for comparison. Fig. 4b

The dramatic change in dark octanoic acid titer between the EL222, LOVtag and combined conditions are surprising, especially in comparison to the lack of change in the dark mCherry expression shown in Figure 4b. This data is the only to suggest that LOVtag may perform better than EL222. However, the inconsistencies in dark state regulation presented in the two experiments, and between conditions in this experiment bring the latter claim to question. A recommendation is that the authors either repeat this experiment, or comment on the observed discrepancy in dark state octanoic acid titers in their discussion.

First, a key difference between the data presented in Fig. 4 and Fig. 5 is that the production experiment is conducted over a long time period (24 hours) and the EL222/LOVdeg reporter experiment is conducted over 5 hours. Likely, performance differences between EL222 and the LOVdeg tag become more pronounced as protein accumulation occurs. Second, the LOVdeg only construct is expressed from a non-EL222 promoter which is able to achieve higher expression (see response to Reviewer #1, Minor point #3). Lastly, a convoluting factor is that the relationship between expression of CpFatB1 and octanoic acid production is not completely linear, and there are likely thresholds or expressions windows that result in similar endpoint titers. We agree a more detailed examination of how CpFatB1 changes over the course of the production period would be very interesting. However, this is beyond the scope of the present study, whose goal is to introduce and showcase the utility of the LOVdeg tag as a tool. We have added new discussion on this in the Results section to clarify some of these points. We have also repeated all experiments in Fig. 4 and consistently see the LOVdeg tag performing as well as or better than EL222. As noted in the remarks to all reviewers, these data have been updated in the revised manuscript.

Results: Optogenetic control of octanoic acid production. Fig. 4d

Based on the methodology presented, no change in the duration in light exposure was tested, even though this may be an important part of the system response. The on/off, for example in Figure 4b, is either all light or all dark, but they claim that their system is beneficial especially at stationary phase. The authors should consider showing the effects of shifting from dark to light at set intervals. (i.e. 1 hr dark then light, 2hr dark until light, etc.) This data would also aid in supporting the utility of this tag for controlling expression during different growth phases, where light may be used after the cells have reached a certain phase.

We have added new data showing the effect of light stimulation at different times in the growth cycle (see response to Reviewer #2, bullet point #5). These data demonstrate that the LOVdeg tag performs well at various points in the growth cycle. Fig. 1f, Fig. S8.

Results: Design and characterization of the AsLOV2-based degradation tag

Minor Revisions Figures:

• Figure 1:

• More clarity is needed in the naming conventions for this figure and in the body of the text. For example, a different convention than 546 and 543 should be used to refer to the full and truncated lengths of the tag. It would greatly aid understanding for this to be made more clear. The authors could simply continue to use "full" and "truncated" to refer to them. In addition, the term "stabilizing mutations" in 1c could be changed to read "dark state stabilizing mutations" to aid in clarity.

When describing the design of the LOVdeg tag, we opted towards a more technically accurate description over clarity in order to make our engineering process easily comparable to other LOV2 systems. As such, we kept the number-based nomenclature (543 or 546) to represent the domain within the phototropin 1 protein from Avena sativa (AsLOV2). The domain used in this study, and many other studies, are only amino acids 404-546, i.e. not the full sequence, thus saying simply ‘full’ or ‘truncated’ is not technically accurate. We believe the detailed nomenclature, which is limited to one section, is important to provide clarity on exactly what we used for protein engineering. In the revised version we introduce the nickname “LOVdeg” tag earlier and use it throughout the rest of the manuscript.

Results: Design and characterization of the AsLOV2-based degradation tag

• 1b It is not clear that this is the dark state stabilized structure in the figure, but is referred to as such only in the body of the text.

We have added text in the manuscript to clarify this is AsLOV2, not iLID, and have labeled it in the figure caption as well.

Results: Design and characterization of the AsLOV2-based degradation tag

• 1d. Fold change is reported in Figure 2d, and may be relevant to include those values in 1d as well.

Done. Fig. 1d

• 1e. It is not clear which tag is being used in this bar plot. Please specify that this is the dark state stabilized, truncated tag.

We have added a title to the plot and language to the caption, both of which clarify this point. Fig. 1e

• In addition, the microscopy images provided in supplemental material should be included in the first figure as it adds a compelling observation of LOVtag activity.

We are pleased to hear that the microscopy results are beneficial, however we elected to leave them in Supplementary to preserve the flow of the manuscript in the text surrounding Fig. 1.

• Figure 2:

• 2d. It is unclear what the 2.5x fold change is relative to (the baseline or the dark)

We have added a line in the figure to clarify the comparison being made. Fig. 2d

• 2f. More discussion can be added to describe what concentration of chloramphenicol is biologically/bioreactor relevant.

Our previous studies on the relationship between AcrAB expression and mutation rate (cited in the text) were carried out at a concentration within the range in which the LOVdeg tag is effective (5 μg/ml), suggesting this range to be relevant to tolerance and resistance.

• Figure 3:

• We recommend that this data and discussion are better suited for supplementary figures. The results shown here essentially recapitulate the same findings of Zoltowski et al., 2009. In addition, the paper describing this mutation should be cited in this figure caption in addition to the body of the text

Although these results are in line with previous findings, we believe this dataset is important for several reasons. First, the agreement with known mutations validates the unfolding-based mechanism for degradation control. Second, degradation that is contingent on unfolding of LOV2 offers a direct actuating mechanism of photocycle properties. Other systems, like that in Zoltowski et al., examine properties of purified proteins but lack the mechanism to translate its effect in live cells. This figure demonstrates how degradation can do so and lays the groundwork for degradation-based frequency processing circuits. Last, there are discrepancies between photocycle kinetics in situ, as reported by Li et al. (DOI: 10.1038/s41467-020-18816-8), and in cell-free studies such as in Zoltowski et al. The studies use different methods of measuring photocycle kinetics (in situ vs cell-free). This dataset substantiates relaxation times from Li et al. and suggests cell-free relaxation time constants are over estimated relative to our live cell results.

• Figure 4:

• There is a lack of an EL222-only control presented in Figure 4b. Without this data present, the claim that "EL222 and the LOVtag work coherently to decrease expression" is unsubstantiated. No assumptions of synergy can be made.

We have added EL222-only data to the figure; we note that Reviewer #1 made a similar request. Figure 4b

Manuscript

Results

• Design and characterization...

• Due to the extensive discussion of ClpX at the beginning of this section, more of the results on evaluating the binding partners and mechanism of LOVtag degradation should be presented in the main body of the manuscript and not in supplementary materials.

To maintain flow of the manuscript and focus on how the LOVdeg tag works as a synthetic biology tool, we have opted to keep this section in the Supplement Information, but have several lines in the text related to Fig. 1 that point the reader to this material. Results: Design and characterization of the AsLOV2-based degradation tag

• In the second paragraph of this section, the authors theorize that the C-terminal truncated E-AA sequence will "remain caged as part of the folded helix". How did the authors determine this? Was there any evidence to suggest that the truncated state would be any more responsive than the full length sequence? More data or rationale may need to be introduced to support the overall hypothesis presented in this paragraph.

We determined this by examining the crystal structure which shows that the E-A-A sequence is part of the folded helix. As seen in Fig. 1b, addition of amino acids after the EAAKEL sequence would not be part of the folded helix which ends prior to the terminal leucine. We added text to clarify our logic.

Results: Design and characterization of the AsLOV2-based degradation tag

• The similarity between the L-A-A sequence of SsrA and the E-A-A sequence of LOVtag is one of the pieces of evidence that brought the authors to their current protein design. The differences in degradation efficiency between the SsrA degradation tag and LOVtag are not clear, and benchmarking against SsrA would be a valuable way to demonstrate the utility of this construct relative to an established protein tagging tool.

We added an SsrA comparison to benchmark the system. Fig. S9

Results: Design and characterization of the AsLOV2-based degradation tag

• Tuning frequency and response...

• Overall the results presented in this section essentially recapitulate the effects that mutation presented in Zoltowski et. al., 2009 have on AsLOV2 dark state recovery and although this is a useful observation of LOVtag performance, a recommendation is to move this into a supplementary section.

See above response to Fig. 3 comment.

• Integrating the LOVtag with EL222...

• The claim is made in this section that LOVtag and EL222 work synergistically, however the experiments presented do not test repression due to EL222 activity alone. Without benchmarking against this control, the claim of synergy is not supported and we recommend that the authors perform this experiment again with the EL222-only control.

We have added this important control. Fig. 4b

Discussion

• The statement "the LOVtag can easily be integrated with existing optogenetic systems to enhance their function" is not substantiated without benchmarking LOVtag against an EL222- only control. As mentioned above this condition should be included in the experiments discussed in Figure 4 and in the section "Integrating the LOVtag with EL222.."

We added EL222-only regulation to benchmark the LOVdeg tag and LOVdeg + EL222 experiments. Fig. 4b

Experiments

Applications:

The application of this tag to the metabolic control of octanoic acid production could be more impactful. For instance, using the LOVtag with two different enzymes to change the composition of long/short chain fatty acids with light induction., Or possibly integrating the tag into a switch to activate production. However, the authors address that "decreasing titers is not the overall goal in metabolic engineering" in their discussion, and therefore the pursuit of this additional experiment is up to the authors' discretion.

We appreciate the suggestions for further applications of the LOVdeg tag. We envision that follow up studies will focus on the application of the LOVdeg tag in metabolic engineering. However, this will require significant development of production systems. We believe this to be out of the scope of this work, where the goal is to present the design and function of the LOVdeg tag as a tool.

Author Response

We are very thankful to the reviewers for a thorough review of our manuscript, and we are confident that we can address all identified weaknesses in the revised version. At the current point, we believe that it is important to mention the following:

1. The review by reviewer 1 contains factual errors. For example, the reviewer writes "There is much important information missing. For instance: how many animals were used per group and how was the breeding done?" Both animal numbers and the breeding scheme are described in detail in the manuscript.

2. Reviewer 3 criticizes our choice of animal ages used for the analysis of sperm DNA methylation aging. The reviewer suggests that the sperm of our younger group may contain spermatozoa from the 1st wave of spermatogenesis, while our older group cannot be considered chronologically old mice. We have experimental data that demonstrate that DNA regions that undergo methylation change with age have a linear association between methylation levels and age across the mouse lifespan (including ages used in our study). Thus, age-dependent changes in DNA methylation may be analyzed using any two ages as soon as they are different enough to detect the changes. We will include this experimental data in our resubmitted manuscript.

Author Response

Reviewer #1 (Public Review):

Question 1: The experiment that utilizes lactose or glucose supplementation to infer the importance of carbohydrate recognition by galectin-9 cannot be interpreted unequivocally owing to the growth-enhancing effect of lactose supplementation on Mtb during liquid culture in vitro.

Response: Thanks for this very constructive comment. We will repeat this experiment and lower the concentration of lactose in order to attenuate its effect on Mtb growth, thereby highlighting the reversed mycobacterial growth inhibition by galectin-9.

Question 2: Similar to the comment above, the apparent dose-independent effect of galectin-9 on Mtb growth in vitro is difficult to reconcile with the interpretation that galectin is functioning as claimed.

Response: We thank the reviewer for the correction. Indeed, as the reviewer pointed out, galectin-9 inhibits Mtb growth in dose-independent manner. We will correct the claim in the revised manuscript.

Question 3: The claimed differences in galectin-9 concentration in sera from tuberculin skin test (TST)-negative or TST-positive non-TB cases versus active TB patients are not immediately apparent from the data presented.

Response: We appreciate the reviewer’s concern. We will perform the detection of galectin-9 in sera in another independent cohort of active TB patients and healthy donors in China.

Question 4: Neither fluorescence microscopy nor electron microscopy analyses are supported by high-quality, interpretable images which, in the absence of supporting quantitative data, renders any claims of anti-AG mAb specificity (fluorescence microscopy) or putative mAb-mediated cell wall swelling (electron microscopy) highly speculative.

Response: We appreciate the reviewer’s concern. We will improve the procedure of the immunofluorescence assay to obtain high-quality and interpretable images with quantitative data. As for electron microscopy analyses, we will add a more precise label indicating cell wall in revised manuscript.

Question 5: Finally, the absence of any discussion of how anti-AG antibodies (similarly, galectin-9) gain access to the AG layer in the outer membrane of intact Mtb bacilli (which may additionally possess an extracellular capsule/coat) is a critical omission - situating these results in the context of current knowledge about Mtb cellular structure (especially the mycobacterial outer membrane) is essential for plausibility of the inferred galectin-9 and anti-AG mAb activities.

Response: Exactly, AG is hidden by mycolic acids in the outer layer of Mtb cell wall. As we have discussed in the Discussion part of previous manuscript (line286), we speculate that during Mtb replication, cell wall synthesis is active and AG becomes exposed, thereby facilitating its binding to galectin-9 or AG antibody and leading to Mtb growth arrest. It’s highly possible that galectin-9 or AG antibody targets replicating Mtb. We will describe this point more comprehensibly.

Reviewer #2 (Public Review):

Question 1: In light of other observations that cleaved galectin-9 levels in the plasma is a biomarker for severe infection (Padilla A et al Biomolecules 2021 and Iwasaki-Hozumi H et al. Biomoleucles 2021) it is difficult to reconcile the author's interpretation that the elevated gal-9 in Active TB patients (Figure 1E) contributes to the maintenance of latent infection in humans. The authors should consider incorporating these observations in the interpretation of their own results.

Response: Thank you for these very insightful comments. We observed elevated levels of galectin-9 in the serum of active TB patients, consistent with reports indicating that cleaved galectin-9 levels in the serum serve as a biomarker for severe infection (Iwasaki-Hozumi et al., 2021; Padilla et al., 2020). We interpret this to mean that elevated levels of galectin-9 in serum of active TB are an indicator of the host immune response to Mtb infection. However, the magnitude of elevated galectin-9 is insufficient to control Mtb infection thereby maintaining latent infection. This is comparable to other protective immune factors such as interferon gamma, which is considered protective and elevated in active TB, as well (El-Masry et al., 2007; Hasan et al., 2009).

Question 2: The anti-AG titers were measured only in individuals with active TB (Figure 3C), generally thought to be a less protective immunological state. The speculation that individuals with anti-AG titers have some protection is not founded. Further only 2 mAbs were tested to demonstrate restriction of Mtb in culture. It is possible that clones of different affinities for AG present within a patient's polyclonal AG-antibody responses may or may not display a direct growth restriction pressure on Mtb in culture. The authors should soften the claims about the presence of AG-titers in TB patients being indicative of protection.

Response: We appreciate the reviewer’s concern. As per the reviewer’s suggestion, we will soften the claim that anti-AG antibodies in the sera of TB patients indicate protection.

References El-Masry, S., Lotfy, M., Nasif, W.A., El-Kady, I.M., and Al-Badrawy, M. (2007). Elevated serum level of interleukin (IL)-18, interferon (IFN)-gamma and soluble Fas in patients with pulmonary complications in tuberculosis. Acta microbiologica et immunologica Hungarica 54, 65-77.

Hasan, Z., Jamil, B., Khan, J., Ali, R., Khan, M.A., Nasir, N., Yusuf, M.S., Jamil, S., Irfan, M., and Hussain, R. (2009). Relationship between circulating levels of IFN-gamma, IL-10, CXCL9 and CCL2 in pulmonary and extrapulmonary tuberculosis is dependent on disease severity. Scandinavian journal of immunology 69, 259-267.

Iwasaki-Hozumi, H., Chagan-Yasutan, H., Ashino, Y., and Hattori, T. (2021). Blood Levels of Galectin-9, an Immuno-Regulating Molecule, Reflect the Severity for the Acute and Chronic Infectious Diseases. Biomolecules 11.

Padilla, S.T., Niki, T., Furushima, D., Bai, G., Chagan-Yasutan, H., Telan, E.F., Tactacan-Abrenica, R.J., Maeda, Y., Solante, R., and Hattori, T. (2020). Plasma Levels of a Cleaved Form of Galectin-9 Are the Most Sensitive Biomarkers of Acquired Immune Deficiency Syndrome and Tuberculosis Coinfection. Biomolecules 10.

Author Response

We thank the reviewers for spending the time to read and provide reviews for our manuscript. The reviewers bring good points regarding the sample size, and the low exposure in the South Asian cohort owing to their unique cultural and social practices. We recognize these as limitations of the paper and will discuss these more extensively in the revised version. With respect to sample size, we are not attempting discovery but rather application of mDNA scores derived from external, large discovery samples. As such, though our sample sizes (n = 300–500) seem low for a typical EWAS, they are in a similar range as replication samples in other studies.

We would also like to take this opportunity to emphasize there is no possible overfitting as the score was tested in studies (FAMILY and START) independent of the discovery set (Joubert et al., 2016; n > 5,000) and the LASSO validation (CHILD; n = 352). In other words, the same participants used for LASSO validation were not used in testing. This is precisely to leverage the larger sample size from external studies to select more plausible CpGs as candidates to include in the model. In fact, the discovery sample size in Reese et al., (2017) was only n = 1,057 in comparison.

The validated score was then used for further testing in new datasets (FAMILY and START), where FAMILY achieved a more significant association than in the original validation sample (CHILD). At the same time, the mean squared error for the continuous smoking severity outcome (0 for no smoking, 1 for quit before pregnancy, 2 for quit during pregnancy, and 3 for current smoker) was 0.68 in CHILD and 1.43 in FAMILY, which indicate good fit; while the AUC for predicting current vs. non-smoker was 0.86 in CHILD and 0.9 in FAMILY. Taken together, these suggest the MRS constructed was not in violation of overfitting, or “failing to fit to additional data or predict future observations reliably”.

In terms of value, our derived score contained 11 CpGs that only overlapped 2 out of the 28 CpGs in the score that was derived in the reference provided (Reese, EHP, 2017, PMID 27323799), but they shared four genes that contributed the most weight to the score (MYO1G, CYP1A1, AHRR, and GFI1). In fact, using the 7 CpGs of the score derived in Reese that were present in all cohorts, we obtained slightly worse performance in CHILD (validation cohort; ANOVA p = 4.1E-5, AUC 0.74), and it was not associated with smoking history in FAMILY (testing cohort; p = 0.13). However, we do agree with the reviewer that including more CpGs will improve the performance, using 24/28 CpGs available in CHILD (HM450K), we obtained slightly better results (ANOVA p = 3.8E-7, AUC 0.94), but these were mostly due to the 14/24 CpGs that showed evidence of association with maternal smoking according to EWAS catalog. In conclusion, we believe our score captures the core genes with robust evidence of association and is more parsimonious for applying to external data, but it can also benefit from a larger sample size to capture CpGs that are moderately associated with maternal smoking.

Author Response

Reviewer #1 (Public Review):

Overall, the magnitude of the effect size due to FNDC5 deficiency in both male and female mice is rather modest. Looking at the data from a qualitative perspective, it is clear that knockout females still lose bone during lactation and on the low calcium diet (LCD). It is difficult to assess the physiologic consequence of the modest quantitative 'protection' seen in FNDC5 mutants since the mutants still show clear and robust effects of lactation and LCD on all parameters measured. Similarly, the magnitude of the 'increased' cortical bone loss in FNDC5 mutant males is also modest and perhaps could be related to the fact that these mice are starting with slightly more cortical bone. Since the authors do not provide a convincing molecular explanation for why FNDC5 deficiency causes these somewhat subtle changes, I would like to offer a suggestion for the authors to consider (below, point #2) which might de-emphasize the focus of the manuscript on FNDC5. If the authors chose not to follow this suggestion, the manuscript could be strengthened by addressing the consequences of the modest changes observed in WT versus FNDC5 KO mice.

We agree that the magnitude of the effect size due to FNDC5 deficiency is modest with regards to the quantitative cortical bone parameters. However, if one examines the changes in osteocyte lacunar size and the mechanical properties of these bones, the differences are greater. As shown in Figure 3 E, the lacunar area of the WT females on a low calcium diet increases by over 30% and the KO by less than 20%, while in the males it is approximately 38% in WT compared to 46% in KO mice. According to Sims and Buenzli (PMID: 25708054) a potential total loss of ~16,000 mm3 (16 mL) of bone occurs through lactation in the human skeleton. This was based on our measurements in lactation-induced murine osteocytic osteolysis (Qing et al PMID: 22308018). They used our 2D section of tibiae from lactating mice showing an increase in lacunar size from 38 to 46 um2. In that paper we also showed that canalicular width is increased with lactation. Therefore, this would suggest a dramatic decrease in intracortical porosity due to the osteocyte lacunocanalicular system in female KO on a low calcium diet compared to WT females and a dramatic increase in KO males compared to WT males. Also, PTH was higher in the serum of female WT compared to female KO mice on a low calcium diet, the opposite for males in order to maintain normal calcium levels (See Table 1). Based on this data, using the FNDC5 null animals, we would speculate that the product of FNDC5, irisin, is having a highly significant effect on the ultrastructure of bone in both males and females challenged with a low calcium diet.

2) The bone RNA-seq findings reported in Figures 4-6 are quite interesting. Although Youlten et al previously reported that the osteocyte transcriptome is sex-dependent, the work here certainly advances that notion to a considerable degree and likely will be of high interest to investigators studying skeletal biology and sexual dimorphism in general. To this end, one direction for the authors to consider might be to refocus their manuscript toward sexually-dimorphic gene expression patterns in osteocytes and the different effects of LCD on male versus female mice. This would allow the authors to better emphasize these major findings, and to then use FNDC5 deficiency as an illustrative example of how sexually-dimorphic osteocytic gene expression patterns might be affected by deletion of an osteocyte-acting endocrine factor. Ideally, the authors would confirm RNA-seq data comparing male versus female mice in osteocytes using in situ hybridization or immunostaining.

Thank you for this suggestion. We have compared the different effects of LCD on male versus female mice in our revised version and have added a figure containing this information.

3) Along the lines of point #2 (above), the presentation of the RNA-seq studies in Figures 4-6 is somewhat confusing in that the volcano plot titles seem to be reversed. For example, Figure 4A is titled "WT M: WT F", but the genes in the upper right quadrant appear to be up-regulated in female cortical bone RNA samples. Should this plot instead be titled "WT F: WT M"? If so, then all other volcano plots should be re-titled as well.

We have now insured that the plots are appropriately labeled.

4) Have the authors compared male versus female transcriptomes of LCD mice?

We have now compared the male vs female transcriptomes of LCD mice and added an additional figure.

5) It would be appreciated if the authors could provide additional serum parameters (if possible) to clarify incomplete data in both lactation and low-calcium diet models: RANKL/OPG ratio, Ctx, PTHrP, and 1,25-dihydroxyvitamin D levels.

It is not possible to quantitate each of these as the serum has been exhausted. We have checked the RANKL/OPG ratio in the RNA seq and qPCR data using osteocyte enriched bone chips and found no difference.

6) Lastly, the data that overexpressing irisin improved bone properties in Fig 2G was somewhat confusing. Based on Kim et al.'s (2018) work, irisin injection increased sclerostin gene expression and serum levels, thus reducing bone formation. Were sclerostin levels affected by irisin overexpression in this study? Was irisin's role in modulating sclerostin levels attenuated with additional calcium deficiency?

We have not observed any differences in the osteocyte Sost mRNA expression between WT and KO normal and low-calcium-diet male and female mice in our RNAseq and qPCR data. As such, we did not check the Sost levels for the 2G experiment.

Reviewer #2 (Public Review):

Summary:

The goal of this study was to examine the role of FNDC5 in the response of the murine skeleton to either lactation or a calcium-deficient diet. The authors find that female FNDC5 KO mice are somewhat protected from bone loss and osteocyte lacunar enlargement caused by either lactation or a calcium-deficient diet. In contrast, male FNDC5 KO mice lose more bone and have a greater enlargement of osteocyte lacunae than their wild-type controls. Based on these results, the authors conclude that in males irisin protects bone from calcium deficiency but that in females it promotes calcium removal from bone for lactation.

While some of the conclusions of this study are supported by the results, it is not clear that the modest effects of FNDC5 deletion have an impact on calcium homeostasis or milk production.

Specific comments:

1) The authors sometimes refer to FNDC5 and other times to irisin when describing causes for a particular outcome. Because irisin was not measured in any of the experiments, the authors should not conclude that lack of irisin is responsible. Along these lines, is there any evidence that either lactation or a calcium-deficient diet increases the production of irisin in mice?

The global FNDC5 KO mice used for our experiments do not produce or secrete irisin, therefore we have extrapolated that the observed effects are due to a lack of circulating irisin. However, this does not rule out that Fndc5 itself could have a function, but this would have to be most likely in muscle and not in the osteocyte as we do not detect significant levels of irisin in either primary osteoblasts nor primary osteocytes compared to muscle and C2C12 cells. As such, we concluded that the phenotypical differences we saw in our experiments are due to a lack of irisin. We now address the reviewer’s point in the discussion. The measurement of irisin in the circulation with lactation or with low calcium diet of normal mice has not been performed.

2) The results of the irisin-rescue experiment shown in figure 2G cannot be appropriately interpreted without normal diet controls. In addition, some evidence that the AAV8-irisin virus actually increased irisin levels in the mice would strengthen the conclusion.

We do not have the normal diet controls at this time. We have now added the quantitative data for tagged irisin in these mice showing highly significant expression

3) There is insufficient evidence to support the idea that the effect of FNDC5 on bone resorption and osteocytic osteolysis is important for the transfer of calcium from bone to milk. Previous studies by others have shown that bone resorption is not required to maintain milk or serum calcium when dietary calcium is sufficient but is critical if dietary calcium is low (Endo. 156:2762-73, 2015). To support the conclusions of the current study, it would be necessary to determine whether FNDC5 is required to maintain calcium levels when lactating mice lack sufficient dietary calcium.

We agree that it would be important to measure calcium levels in the milk to test the hypothesis that FNDC5 is important to maintain calcium levels in milk. However, as the calcium levels are normal in the serum, we are assuming they are normal in milk. This would require future experiments.

4) The amount of cortical bone loss due to lactation is very similar in both WT and FNDC5 KO mice. The results of the statistical analysis of the data presented in figure 1B are surprising given the very similar effect size of lactation. The key result from the 2-way ANOVA is whether there is an effect of genotype on the effect size of lactation (genotype-lactation interaction). The interaction terms were not provided. Similar concerns are noted for the results shown in figure 1G and H.

We agree, thanks. We will now add the interaction terms in the figure legends.

5) It is not clear what justifies the term 'primed' or 'activated' for resorption. Is there evidence that a certain level of TRAP expression lowers the threshold for osteocytic osteolysis in response to a stimulus?

The number of TRAP positive osteocytes in female KO mice are lower than in female WT. The number of TRAP positive osteocytes are lower in WT males compared to WT females. We propose that irisin plays a role in the number of TRAP positive osteocytes in normal, WT females by readying or preparing these cells to rapidly respond to low calcium. We will use the term ‘primed’ and will not use the term ‘activated’. We are open to any terminology or description as to why this is observed and what irisin could be doing to the osteocyte.

Reviewer #3 (Public Review):

Summary: Irisin has previously been demonstrated to be a muscle-secreted factor that affects skeletal homeostasis. Through the use of different experimental approaches, such as genetic knockout models, recombinant Irisin treatment, or different cell lines, the role of Irisin on skeletal homeostasis has been revealed to be more complex than previously thought and this warrants further examination of its role. Therefore, the current study sought to rigorously examine the effects of global Irisin knockout (KO) in male and female mouse bone. Authors demonstrated that in calcium-demanding settings, such as lactation or low-calcium diet, female Irisin KO mice lose less bone compared to wild-type (WT) female mice. Interestingly male Irisin KO mice exhibited worse skeletal deterioration compared to WT male mice when fed a low-calcium diet. When examined for transcriptomic profiles of osteocyte-enriched cortical bone, authors found that Irisin KO altered the expression of osteocytic osteolysis genes as well as steroid and fatty acid metabolism genes in males but not in females. These data support the authors' conclusion that Irisin regulates skeletal homeostasis in sex-dependent manner.

Strengths: The major strength of the study is the rigorous examination of the effects of Irisin deletion in the settings of skeletal maturity and increased calcium demands in female and male mice. Since many of the common musculoskeletal disorders are dependent on sex, examining both sexes in the preclinical setting is crucial. Had the investigators only examined females or males in this study, the conclusions from each sex would have contradicted each other regarding the role of Irisin on bone. Also, the approaches are thorough and comprehensive that assess the functional (mechanical testing), morphological (microCT, BSEM, and histology), and cellular (RNA-seq) properties of bone.

Weaknesses: One of the weaknesses of this study is a lack of detailed mechanistic analysis of why Irisin has a sex-dependent role on skeletal homeostasis. This absence is particularly notable in the osteocyte transcriptomic results where such data could have been used to further probe potential candidate pathways between LC females vs. LC males.

Our future studies will focus on understanding the molecular mechanism behind the sex-dependent effects of irisin. Our RNA seq data shows a significant difference in the lipid, steroid, and fat metabolism pathways between male and female mice, as well as between WT and KO mice. Future studies will focus on these pathways.

Another weakness is authors did not present data that convincingly demonstrate that Irisin secretion is altered in the skeletal muscle between female vs. male WT mice in response to calcium restriction. The supplement skeletal muscle data only present functional and electrophysiolgical outcomes. Since Itgav or Itgb5 were not different in any of the experimental groups, it is assumed that the changes in the level of Irisin is responsible for the phenotypes observed in WT mice. Assessing Irisin expression will further strengthen the conclusion based on observing skeletal changes that occur in Irisin KO male and female mice.

The problem is that the commercial assays for irisin are not dependable, and results can differ widely across and beyond the physiologic range of 1-10 ng/ml. In part this is due to the nature of the polyclonal antibodies used and the resultant cross reactivity with other proteins. It was shown in Islam et al, 2021 (Nature Metabolism) that the commercial ELISAs were completely unreliable in mice and the only reliable method of measuring circulating irisin is mass spectrometry.

Author Response

Reviewer #1 (Public Review):

Strengths:

1. In my assessment, the data sufficiently demonstrates that a modified version of Pertuzamab can bind both the wild-type and S310 mutant forms of ERBB2.

2. The engineering strategy employed is rational and effectively combines computational and experimental techniques.

3. Given the clinical activity of HER2-targeting ADCs, antibodies unaffected by ERBB2 mutations would be desired.

Weaknesses:

1. There is no data showing that the engineered antibody is equally specific as Pertuzamab i.e. that it does not bind to other (non-ERBB2) proteins.

Showing the specificity of the engineered antibodies is indeed important. We did not address it in the current ms, but it can be tested in the future.

1. There is no data showing that the engineered antibody has the desired pharmacokinetics/pharmacodynamics properties or efficacy in vivo.

In this ms we did not conduct in-vivo experiments. When moving forward, pharmacokinetics/pharmacodynamics properties and efficacy will be tested as well.

1. Computational approaches are only used to design a phage-screen library, but not used to prioritize mutations that are likely to improve binding (e.g. based on predicted impact on the stability of the interaction). A demonstration of how computational pre-screening or lead optimization can improve the time-intensive process would be a welcome advance.

Thank you for this important comment. In the present ms we indeed used a computational approach for prioritizing residues to be mutated, but we did not prioritize the mutations that are likely to improve binding. In the initial library design, we did prioritize the mutations. However, due to experimental approach limitations with codon’s selection for the library, we had decided to allow all possible residues in each position, knowing that the selection will remove non-binding variants.

Context:

The conflict of interest statement is inadequate. Most authors of the study (but not the first author) are employees of Biolojic, a company developing multi-specific antibodies, but the statements do not clarify whether the presented antibodies represent Biolojic IP, whether the company sponsored the research, and whether the company is further developing the specific antibodies presented.

The Conflict-of-Interest statement will be revised as such: The Biolojic Design authors are employees of Biolojic Design and have stock options in Biolojic Design. The company did not sponsor the research, does not hold IP for the presented antibodies, and is not further developing the presented antibodies.

Reviewer #2 (Public Review):

Strengths:

1. Deep computational analyses of large datasets of clinical data provide useful information about HER2 mutations and their potential relevance to antibody therapy resistance.

2. There is valuable information analyzing the residues within or near the interface between the antigen HER2 and the Pertuzumab antibody (heavy chain). The experimental antibody library screening obtained 90+ clones from 3.86×1011 sequences for further functional validation.

Weaknesses:

1. There is a lack of assessment for antibody variant functions in cancer cell phenotypes in vitro (proliferation, cell death, motility) or in vivo (tumor growth and animal survival). The only assay was the western blotting of phosphopho-HER3 in Figure 4. However, HER2 levels and phosphor-HER2 were not analyzed.

We indeed did not assess the engineered antibodies function in cancer cells. Regarding signaling assessment, previous works [1-3] also measured the signaling activation following HER2-HER3 dimerization by measuring pHER3, and we relied on them in this ms.

1. There is a misleading impression from the title of computational engineering of a therapeutic antibody and the statement in the abstract "we designed a multi-specific version of Pertuzumab that retains original function while also bindings these HER2 variants" for a few reasons:

a. The primary method used for variant antibody identification for HER2 mutant binding is rather traditional experimental screening based on yeast display instead of the computational design of a multi-specific version of Pertuzumab.

b. There is insufficient or lack of computational power in the antibody design or prioritization in choosing variant residues for the library construction of 3.86×1011 sequences. It seems random combinations from 6 residues out of 4 groups with 20 amino acid options.

c. The final version of the tri-binding variant is a combination of screened antibody clones instead of computation design from scratch.

d. There is incomplete experimental evidence about the therapeutic values of newly obtained antibody clones.

Thank you for this relevant comment. When addressing relevant residues to be mutated, the number of potential variants is enormous. The computational approach was aimed at identifying the most preferable residues, in which variation can improve binding and is not likely to harm important interactions. Although an initial smaller number of residues could be chosen, we decided to broaden our view and create a larger library, in the aim of combining the computational selection with an experimental selection. This indeed is not a computational design from scratch, but rather an intercourse between the computer and the lab, that yielded the presented results.

1. Figures can be improved with better labeling and organization. Some essential pieces of data such as Supplementary Figure 1B on HER2 mutations in S310 that abrogated its binding to Pertuzumab should be placed in the main figures.

Thank you for this comment, the relevant figures will be moved to the main text, and the labels will be revised.

1. It is recommended to provide a clear rationale or flowchart overview into the main Figure 1. Figure 2A can be combined with Figure 1 to the list of targeted residues.

Figures 1 and 2 will be divided differently, and the rationale will be detailed in the revised text.

1. The quality of Figures such as Figure 2B-C flow data needs to be improved.

This will be corrected in the revised text.

1. Diwanji, D., et al., Structures of the HER2-HER3-NRG1β complex reveal a dynamic dimer interface. Nature, 2021. 600(7888): p. 339-343.

2. Yamashita-Kashima, Y., et al., Mode of action of pertuzumab in combination with trastuzumab plus docetaxel therapy in a HER2-positive breast cancer xenograft model. Oncol Lett, 2017. 14(4): p. 4197-4205.

3. Kang, J.C., et al., Engineering multivalent antibodies to target heregulin-induced HER3 signaling in breast cancer cells. MAbs, 2014. 6(2): p. 340-53.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

This reviewer found the paper of very high interest, well supported, and well written. I have only a few suggestions to the authors for further improvement:

1. TRAIL mutants carrying individual mutations of basic residues R119, R122 and K125 were tested, but a TRAIL mutant lacking all three residues was not. This combined mutant protein would have allowed to test whether all heparin binding is abolished (e.g. that no other residues contribute to HS binding) and could have also been used as an independent control replacing heparin and heparinase treatment in binding/apoptosis studies. Given that the DR4/5 and heparin binding sites of TRAIL do not overlap, this form would be useful in determining the extent to which HS contributes to, or serves as a prerequisite for TRAIL binding to its receptor and cell death. Moreover, if bound to the receptor, this mutant TRAIL is expected to completely prevent HS-mediated receptor internalization. The added value of this experiment therefore is that it may provide an answer to the controversial debate on whether DR receptor internalization promotes or inhibits apoptosis.

In Fig. 5C, we provided data showing that the binding of R115A mutant of hTRAIL (equivalent to murine R199A mutant) to MB-453 cells was very similar to the binding of WT hTRAIL to heparin lyase treated cells. This finding suggests that nearly all HS-dependent binding to cell surface HS was abolished by mutating R115. Since a single mutant is sufficient, we felt there is little point in combining multiple mutations. We also used R115A mutant as an independent control replacing heparin and heparinase treatment in apoptosis assay in Fig. 7E. With regard to using the mutant in the internalization assay, we thank the reviewer for this excellent suggestion and will incorporate it into our future study as we intend to perform more in-depth investigation on the exact mechanism of internalization.

1. The domain data is interesting, but its physiological significance remains obscure and it also somewhat distracts from the main theme of the study. It may be removed from a revised manuscript.

We partially agree with the reviewer’s assessment, but we felt that this discovery is of sufficient novelty and should be made known to the whole community.

1. TUNEL data is shown as a picture in Figure 6, but quantification is lacking.

We have included the statistics of the TUNEL data in the final version as Fig. 6D.

1. Is the HS20 antibody a well-suited pan-anti-HS antibody? Why was this antibody used instead of heparinase digestion followed by the use of HS "stub" antibodies that were previously used as a reliable readout for overall sulfation?

The HS20 mAb has been very well characterized by Dr. Mitchell Ho group (Gao et al., 2016). We have also done side-by-side comparison of HS20 and the most commonly used anti-HS mAb 10E4 by immunostaining and FACS. In nearly all tissues and cells tested, HS20 gave better sensitivity and lower background (after heparin lyase treatment) compared to 10E4. The staining pattern of the two mAbs are usually identical, but the signal/noise ratio of HS20 is much better than 10E4. The HS ”stub” antibody can be useful in certain applications, but it is used mainly as an indicator of the distribution/abundance of HSPGs, rather than a readout of overall sulfation.

1. The discussion should be stripped from expressions such as interestingly, curiously, unexpectedly, certainly, undoubtedly and the like to improve readability. The manuscript should be checked for typos (for example surface plasma resonance line 473, was served line 481).

We thank the reviewer for the suggestions and many of these expressions were removed in the final version.

1. Last but not least: to test the physiological relevance of these findings, it would be of the highest interest to use a mouse model harboring a tumor cell line of choice and derived lines with impaired or increased HS expression, as outlined in my public comments, and to test tumor responsiveness to TRAIL treatment. If already planned, I wish you Good Luck with the experiments!

We thank the reviewer for this excellent suggestion and we have indeed planned to do exactly that!

Reviewer #2 (Recommendations For The Authors):

1. The authors showed in Fig.2 that 12mer HS forms complex with TRAIL homotrimer. Please clarify if 12mer HS binding leads to the formation of the TRAIL homotrimer or TRAIL can form homotrimer in the absence of HS binding. Do the TRAIL mutations that affect HS binding, such as R115A, also impact the homotrimer formation?

TRAIL automatically forms a homotrimer independent of HS. It is known that formation of the homotrimer critically depends on a zinc ion, which is located on the threefold axis of the trimer and is bound by cysteine 240. We have also verified that all TRAIL mutants remain homotrimeric by size exclusion chromatography.

1. Does 12mer HS also suppress TRAIL-mediated apoptosis in MDA-MB-453 cells?

We thank the reviewer for this question but felt performing this experiment will not add any more insight to the main conclusion. Most likely, the result will be similar to what we saw in Fig. 7D, where we found 12mer significantly inhibits TRAIL-induced apoptosis, but inhibits less efficiently compared to heparin.

1. The authors nicely showed the correlation between surface HS level and sensitivity to TRAIL-induced apoptosis in MM cell lines and implicated that such correlation could be related with the difference in the expression level of SDC1. This is an interesting point worth further validation. Does ectopic SDC1 expression in IM-9 cells lead to increase cell surface HS and sensitivity to TRAIL treatment? On the other hand, will depletion of SDC1 expression in U266 or RPMI8226 cells decrease their sensitivity to TRAIL treatment?

We agree that this would be an excellent experiment to try and have actually attempted to overexpress SDC1 in IM-9 cells. But we found IM-9 cells are very difficult to transfect and we only managed to convert a small percentage of SDC1 negative cells to positive cells. Also, the level of SDC1 expression on the SDC1-positive cells was not changed after overexpression. We have not tried depleting SDC1 expression in U266 and RPMI8226 cells because such an experiment might change the property of these cells in unexpected ways, which would make result interpretation impossible. A previous report has shown that knocking down SDC1 could enhance clustering of TRAIL receptors in H929 cells (Wu et al., J Immunol 2012;), which actually led to slightly increased apoptosis.

Author Response

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

eLife assessment

This important study extends insights on NAFLD and NASH regarding the role of plasma lactate levels using mice haplo-insufficient for the gene encoding lactate transporter MCT-1. While the evidence is largely convincing and the work significantly advances our understanding of the roles of distinct hepatic cell types in steatosis, a number of issues require attention and would best be solved by further experimentation.

RESPONSE: We agree with this assessment by eLife, and appreciate the reviewers’ view that the study is important and extends insights into liver disease.

Public Reviews:

Reviewer #1 (Public Review):

The authors put forth the hypothesis that hepatocyte and/or non-parenchymal liver MCT1 may be responsible for physiologic effects (lower body weight gain and less hepatic steatosis) in MCT1 global heterozygote mice. They generate multiple tools to test this hypothesis, which they combine with mouse diets that induce fatty liver, steatohepatitis and fibrosis. Novel findings include that deletion of hepatocyte MCT1 does not change liver lipid content, but increases liver fibrosis. Deletion of hepatic stellate cell (HSC) MCT1 does not substantially affect any liver parameter, but concomitant HSC MCT1 deletion does reverse fibrosis seen with hepatocyte MCT1 knockout or knockdown. In both models, plasma lactate levels do not change, suggesting that liver MCT1 does not substantially affect systemic lactate. In general, the data match the conclusions of the manuscript, and the studies are well-conducted and well-described. Further work would be necessary to dissect mechanism of fibrosis with hepatocyte MCT1, and whether this is due to changes in local lactate (as speculated by the authors) or another MCT1 substrate. This would be important to understand this novel potential cross-talk between hepatocytes and HSCs.

A parallel and perhaps more important advance is the generation of new methodology to target HSC in mice, using modified siRNA and by transduction of AAV9-Lrat-Cre. Both methods would reduce the need to cross floxed mice with the Lrat-Cre allele, saving time and resources. These tools were validated to an extent by the authors, but not sufficiently to ensure that there is no cross-reactivity with other liver cell types. For example, AAV9-LratCre-transduced MCT1 floxed mice show compelling HSC but not hepatocyte Mct1 knockdown, but other liver cell types should be assessed to ensure specificity. This is particularly important as overall liver Mct1 decreased by ~30% in AAV9-Lrat-Cre-transduced mice, which may exceed HSC content of these mice, especially when considering a 60-70% knockdown efficiency. This same issue also affects Chol-MCT1-siRNA, which the authors demonstrate to affect hepatocytes and HSC, but likely affects other cell types not tested. As this is a new and potentially valuable tool, it would be important to assess Mct1 expression across more non-parenchymal cells (i.e. endothelial, cholangiocytes, immune cells) to determine penetration and efficacy.

RESPONSE: We appreciate the reviewer’s view that the new methods we describe represent an important advance. To ensure the specificity of our novel AAV-Lrat-Cre construct, it would be fair to test its distribution among all possible hepatic cell types, including endothelial cells, cholangiocytes, and other immune cells, as suggested. Our efforts in this study were primarily focused on the major cell types thought to contribute to NASH, namely hepatocytes, Kupffer cells, and in particular hepatic stellate cells. The reasons for this focus were:

1) Our primary goal was to investigate the role of MCT1 in hepatic fibrogenesis. According to Manderacke et al. (2013, Nature Comm), hepatic stellate cells account for the dominant proportion (82-96%) of myofibroblast progenitors, which produce collagen fibers. While there may be interesting roles of MCT1 in those other cell types, to elucidate MCT1's role in fibrogenesis, focusing on the dominant fibrogenic cell type, hepatic stellate cells, was the most appropriate approach for this goal.

2) Considering the proportion of each hepatic cell type in the liver, hepatocytes constitute the majority (60-70%), followed by endothelial cells (15%), immune cells (10%), and stellate cells (5%), among others.

3) The AAV-Cre system is highly specific to its promoter, in this case, Lrat, which has been well established in multiple previous studies to exhibit high specificity for hepatic stellate cells in the liver. We will certainly conduct more comprehensive biodistribution studies in the future, as we believe that our AAV-Lrat-Cre system could be a valuable tool in this field.

Reviewer #2 (Public Review):

In this study, the authors seek to answer two main questions: 1) Whether interfering with lactate availability in hepatocytes through depletion of hepatocyte specific MCT-1 depletion would reduce steatosis, and 2) Whether MCT-1 in stellate cells promote fibrogenesis. While the first question is based on the observation that haploinsufficiency of MCT-1 makes mice resistant to steatosis, the rationale behind how MCT-1 could impact fibrogenesis in stellate cells is not clear. A more detailed discussion regarding how lactate availability would regulate two different processes in two different cell types would be helpful. The authors employ several mouse models and in vitro systems to show that MCT1 inhibition in hepatic stellate cells reduces the expression of COL-1. The significance of the findings is moderately impacted due to the following considerations:

RESPONSE: We have included additional in vitro data in order to provide a more comprehensive discussion of MCT1's potential role in regulating collagen production. Please refer to the new Figure 8, Supplementary Figure 6, and the results section (Potential Mechanism). Also note that our original hypothesis was that depleting MCT1 specifically in hepatocytes would protect mice with MCT1 haploinsufficiency from liver lactate overload and NAFLD. Furthermore, we postulated that this protection might prevent NASH progression since lipotoxicity-driven hepatocyte damage is a central factor in NASH pathogenesis. However, our findings did not support this hypothesis. We found only one brief article (2015, Z Gastroenterol et al., "Functional effects of monocarboxylate transporter 1 expression in activated hepatic stellate cells") that discussed the potential role of MCT1 depletion in hepatic stellate cells in regulating collagen production or fibrosis, as mentioned in their abstract. Unfortunately, the DOI for this article is not functional, and the data cannot be located. Moreover, when we attempted to replicate their results, we were unable to do so, leading us to report our own findings in the current paper.

a. Fibrosis in human NAFLD is a significant problem as a predictor of liver related mortality and is associated with type 1 and type 3 collagen. However, the reduction in COL1 in stellate cells did not amount to a reduction in liver fibrosis even in cell specific KO (in Fig 7E, there is no indication of whether Sirius red staining was different between HSC KO and control mice- the authors mention a downward trend in the text). The authors postulate that type 1 COL may not be the more predominant form of fibrosis in the model. This does not seem likely, since the same ob/ob mouse model was used to determine that fibrosis was enhanced with hepatocyte specific MCT-1 KO and decreased with Chol MCT-1KO. Measurements of different types of collagens in their model and the effect of MCT-1 on different types could be more informative. In particular, although collagens are the structural building blocks for hepatic fibrosis, fibrosis can also be controlled by matrix remodeling factors such as Timp1, Serpine 1, PAI-1 and Lox.

RESPONSE: We monitored the expression levels of matrix remodeling factors, such as Timp1 (Figure 5C, 5F). There was no change in expression upon Chol-MCT1-siRNA treatment, while a significant increase was observed upon GN-MCT1-siRNA treatment. This trend was similar to collagen expression in both cases. Regarding the different types of collagen, instead of measuring each individual type of collagen, we conducted Sirius red and trichrome staining, which enabled us to detect multiple types of collagen simultaneously (Figure 5G, Figure 7D).

b. The authors use multiple animal models including cell specific KO to conclude that stellate cell MCT-1 inhibition decreases COL-1. However, the mechanisms behind this reduced expression of COL-1 are not discussed or explored, making it descriptive.

RESPONSE: We agree that the mechanisms involved are not fully defined but have added new data (Figure 8, Supplement Figure 6) and text to discuss possibilities.

c. Different types of diets are used in this study which could impact lactate availability. Choline deficiency diets are reported to cause weight loss, and importantly have none of the metabolic features of human NASH. Therefore, their utility is doubtful, especially for this study which proposes to investigate if metabolic dysregulation and substrate availability could be a tool for therapy.

RESPONSE: Unfortunately, none of the rodent models used to study NASH completely replicate the condition in human patients, each having its own set of advantages and drawbacks. In line with the concern raised by reviewer #2, there has been a shift away from the use of severely detrimental methionine and choline-deficient diets in contemporary NASH research. Instead, diets that combine methionine and other amino acids with cholinedeficient diets, in conjunction with high-fat diets, have become more popular. The diet we employed in our study consists of high-fat diet combined with choline-deficient diets. We believe that our findings, which are consistent and established across two distinct NASH pathogenesis models and genetic backgrounds, lend additional robustness to our results.

d. Hepatocyte specific MCT-1 KO mice seem to have increased COL-1 production, despite no noticeable difference in hepatocyte steatosis. The reasons for this are not discussed. Fibrosis in NASH is thought to be from stellate cell activation secondary to signals from hepatocellular damage. There is no evidence that there was a difference in either of these parameters in the mouse models used.

RESPONSE: While lipotoxicity-driven liver damage remains a central aspect of NASH pathogenesis, the traditional two-hit theory has become less tractable, giving way to the multi-hit theory in the NASH field. The current debate revolves around whether steatosis is a decisive factor and requirement for NASH fibrogenesis. Our previous publication (Yenilmez et al., 2022, Mol Ther) demonstrated that nearly complete resolution of steatosis did not prevent other NASH features like inflammation and fibrosis, indicating the existence of multiple factors beyond steatosis in NASH pathogenesis. We believe that steatosis and fibrosis influence each other but can also develop independently.

e. The authors report that serum lactate levels did not rise after MCT-1 silencing, but the reasons behind this are unclear. There is insufficient data about lactate production and utilization in this model, which would be useful to interpret data regarding steatosis and fibrosis development. For example, does the MCT-1 KO prevent hepatocyte and stellate cell net import or export of lactate? What is the downstream metabolic consequence in terms of pyruvate, acetylCoA and the NAD/NADH levels. Does the KO have downstream effects on mitochondrial TCA cycling?

RESPONSE: Due to both biological and technical challenges (which are described in the new draft), conducting a comprehensive metabolomics study comparing hepatocyte MCT1 KO to hepatic stellate cell MCT1 KO was not feasible. It is important to note that MCT1 can also transport other substrates that are often overlooked, including pyruvate, short-chain fatty acids, and ketone bodies. Also, in addition to MCT1, there are at least two other functional isoforms of MCT: MCT2 and MCT4. Regrettably, due to these biological and technical complications, conducting a comprehensive metabolomic analysis is extremely complicated and difficult to interpret. Nevertheless, some insights are gained from a study involving MCT1 chaperone protein Basigin/CD147 knockout (KO) mice in a high-fat diet- induced hepatic steatosis model. Basigin acts as an auxiliary protein for MCT1, and its absence leads to improper localization and stabilization of MCT1, effectively simulating a state of MCT1 deficiency. In this context, hepatic lactate levels were reduced by half, and other metabolites such as pyruvate, citrate, α-ketoglutarate, fumarate, and malate were significantly decreased. While we must exercise caution when extrapolating these findings to our MCT1 study, they suggest that multiple metabolites, particularly pyruvate, may play a crucial role in the context of MCT1 deficiency.

f. MCT-1 protein expression is measured only in the in vitro assay. Similar quantitation through western blot is not shown in the animal models.

RESPONSE: We monitored MCT1 protein expression with either Western blot (Fig 2D, 2E (in vitro)) or immune-histology (Fig 4B, 4C (in vivo, ob/ob + GAN diet NASH model), Sup Fig 5F, 5G (in vivo, MCT1 f/f + CDHFD model)).

Reviewer #3 (Public Review):

A major finding of this work is that loss of monocarboxylate transporter 1 (MCT1), specifically in stellate cells, can decrease fibrosis in the liver. However, the underlying mechanism whereby MCT1 influences stellate cells is not addressed. It is unclear if upstream/downstream metabolic flux within different cell types leads to fibrotic outcomes. Ultimately, the paper opens more questions than it answers: why does decreasing MCT1 expression in hepatocytes exacerbate disease, while silencing MCT1 in fibroblasts seems to alleviate collagen deposition? Mechanistic studies in isolated hepatocytes and stellate cells could enhance the work further to show the disparate pathways that mediate these opposing effects. The work highlights the complexity of cellular behavior and metabolism within a disease environment but does little to mechanistically explain it.

RESPONSE: Described above to Reviewer #2

The observations presented are compelling and rigorous, but their impact is limited by the nearly complete lack of mechanistic insight presented in the manuscript. As also mentioned elsewhere, it is important to know whether lactate import or export (or the transport of another molecule-like ketone bodies, for example) is the decisive role of MCT1 for this phenotype. Beyond that, it would be interesting, albeit more difficult, to determine how that metabolic change leads to these fibrotic effects.

RESPONSE: Described above to Reviewer #2

Kuppfer cells are initially analyzed and targeted. These cells may play a major role in fibrotic response. It will be interesting to determine the effects of lactate metabolism in other cells within the microenvironment, like Kuppfer cells, to gain a complete understanding of how metabolism is altered during fibrotic change.

RESPONSE: To address the potential involvement of inflammatory cells, we added new data to the manuscript (Supplement Figure 4). Given the distinct hepatic cellular distribution of Chol-MCT1-siRNA and GN-MCT1-siRNA, the opposite fibrogenic phenotype observed may be attributed to MCT1’s role in non-hepatocyte cell types such as the inflammatory Kupffer cells and the fibrogenic hepatic stellate cells. To determine which hepatic cell type drives the opposite fibrotic phenotypes, we first hypothesized that GN-MCT1-siRNA activates M2 pro-fibrogenic macrophages more than Chol-MCT1-siRNA does. The representative M1/ M2 macrophage polarization gene markers were monitored in Kupffer cells. However, GN-MCT1-siRNA treatment caused comparable M1/M2 macrophage activation levels to Chol-MCT1-siRNA treatment (Supplement Figure 4A, 4B). These data suggest that the opposite fibrotic phenotypes caused by the different siRNA constructs are not due to M1/M2 macrophage polarization.

The timing of MCT1 depletion raises concern, as this is a largely prophylactic experiment, and it remains unclear if altering MCT1 would aid in the regression of established fibrosis. Given the proposal for translation to clinical practice, this will be an important question to answer.

RESPONSE: Agree these are important experiments for future evaluation.

Reviewer #1 (Recommendations For The Authors):

As above, in general, the conclusions match the data presented. The one exception is the authors discussion point that these data show the importance of lactate flux in fibrosis. As MCT1 has other substrates, it does not seem this is definitively due to lactate flux. It would be helpful to have additional experiments to clarify mechanism by which loss of hepatocyte MCT1 leads to increased fibrosis, while loss of HSC MCT1 reverses this finding. This may aid in concluding that altered fibrosis is in fact due to lactate flux in these cell types.

RESPONSE: Described above to Reviewer #2

In addition, it is unclear why the authors switched NASH models for the two tools generated (GAN diet for siRNA, CDHFD for AAV). Similarly, methodology to assess fibrosis switched between these two experiments - i.e. Sirius Red staining for siRNA-treated GAN diet-fed mice vs. Trichrome staining for AAV-transduced CDHFD-fed mice. These changes make it difficult to perform cross-comparisons of the data, to explain (for example), why GN-siRNA to Mct1 reduced body weight but AAV8-TBG-Cre did not. Similarly, GN-siRNA increased liver Col1a1 protein but AAV8-TBG-Cre did not. These differences could be explained by model system, or tool efficacy/off-target effects.

RESPONSE: We agree that different model systems can explain difference in results, but there is also an advantage of using different models and various methodologies as preclinical tests of consistency of data on NASH under different conditions. There are no perfect mouse models for human NASH.

• Phenotyping is also incomplete for the latter experiment, in particular amount of liver lipid content –

RESPONSE: We estimated lipid content by H&E (Fig 6E, F). In some experiments, we focused mostly on COL1 protein expression, as this rather than mRNA is the functional aspect of fibrosis.

Reviewer #2 (Recommendations For The Authors):

This study could benefit from standardization of the types of diet used across all animal models and a more comprehensive focus on the metabolic/substrate availability and utilization aspects of NAFLD and NASH affected in the mouse models with MCT-1 dependent lactate transport deficiency. Since hepatic fibrogenesis in NASH is impacted by signals following hepatocyte damage, the extent of cell death in these models could also be better characterized.

RESPONSE: Our ALT data provides indirect insight into hepatocyte damage. Our histology images did not reveal significant changes in cell morphology or integrity and there were no notable changes in caspase protein levels.

Other comments:

In Fig 4G, there is an increase in the number of lipid droplets with Chol- MCT-1 siRNA compared to GN-MCT1-sirRNA, suggesting that the stellate cell component might be responsible for this finding. The possible reasons for this are not discussed.

RESPONSE: The effects in Fig 4G were exceedingly small and there is no difference in total TG in these experiments, so it is hard to interpret these data and provide logical explanations.

In Fig 5A. A western Blot for aSMA and COL 1 is shown but the sample labeling is unclear i.e, do the lanes belong to different mice of the same condition? HFD mice vs Ctr mice?

RESPONSE: Both groups of ob/ob mice were fed a GAN diet. The graph in Fig 5 is a direct comparison between NTC-siRNA and MCT1-siRNA. To enhance clarity, this is indicated in the figure legends, and the data in Fig 5 is a continuation of the data presented in Fig 4

In Fig 5E, COL1 densitometry data should also be provided for non-silenced mice on HFD and Chow diet for appropriate comparison

RES\PONSE: Both groups of ob/ob mice were fed a GAN diet. The graph in Fig 5 represents a comparison between NTC-siRNA and MCT1-siRNA. It's important to note that, typically, ob/ob mice fed either a chow diet or a high-fat diet do not exhibit fibrogenic phenotypes within this time frame (3 weeks of dietary intervention).

There are many mis-statements throughout the text.Page 6 - "MCT1 silencing significantly inhibited Tgf1β-stimulated ACTA2 mRNA expression as well as collagen 1 protein production" but it is not stated that CO1A1 mRNA is unchanged in Fig 1C.

RESPONSE: We observed no change in CO1A1 mRNA levels (Fig 1C), so we focused on collagen 1 protein production (Fig 1B) on page 6. Given the consistent trend observed in Chol-MCT1-siRNA (Fig 5C), we proposed the possibility of MCT1's influence on collagen translation or protein turnover on page 11.

Page 7- ".......our Chol-MCT1-siRNA does not require transfection reagents as it is fully chemically modified". What does fully chemically modified mean and why does this mean in terms of transfection efficiency.

RESPONSE: One of the primary challenges in utilizing RNAi as a therapeutic approach has been the effective in vivo delivery strategy, particularly concerning stability and longevity against systemic nucleases. Recent developments in siRNA duplex chemical modification strategies, such as 2-Fluoro and 2-O-Methyl ribose substitutions, as well as phosphorothioate backbone replacements, have addressed these challenges (Please see Figure 3. In our current study, we employed 'chemically fully modified' siRNA, featuring several key modifications: (1) every single ribose is chemically modified to 2-F or 2-OMeribose, (2) phosphorothioate backbone replacement, (3) 5'-end of the antisense strand modification to (E)-Vinyl-phosphonate, and (4) 3'-end of the sense strand linkers such as Cholesterol or Tri-N-Acetyl-galactosamine. These chemical enhancements significantly improve transfection efficiency, longevity, and selectivity, setting it apart from traditional siRNA lacking such chemical modifications. A prior study from the Khvorova lab has demonstrated substantial efficiency differences between partially and fully modified siRNA in vivo.

Page 7- the results present for Fig 2 ignores Fig, 2C, if this is important it needs to be described if not, please delete.

RESPONSE: The dose-response potency results, crucial for identifying the most potent Chol-MCT1-siRNA compound, are depicted in Figure 2C. The wording "(Figure 2C)" has been inserted in the sentence as follows. “The silencing effect on Mct1 mRNA was monitored after 72 hours (Figure 2B). Several compounds elicited a silencing effect greater than 80% compared to the NTC-siRNA. The two most potent Chol-MCT1-siRNA, Chol- MCT1-2060 (IC50: 59.6nM, KD%: 87.2), and Chol-MCT1-3160 (IC50: 32.4nM, KD%: 87.7) (Figure 2C) were evaluated for their inhibitory effect on MCT1 protein levels (Figure 2D, 2E). Based on its IC50 value and silencing potency, Chol-MCT1-3160 construct was chosen for further studies in vivo (Table 2).”

Supplement Fig 1A-F should be analyzed by multiple comparisons not by paired t-tests.

RESPONSE: We performed t-tests for every comparison between two groups. However, for Sup Fig 1A-F, which involved a comparison among three different groups, we applied oneway ANOVA.

The x-axis in supplement Fig 2A and B are not labeled, and I assume are in weeks. The Fig 2B x-axis numbers also mis-labeled and should also be 0-3 and not 10-13.

RESPONSE: The x-axis is now appropriately labeled.

Page 10 - the description of supplement Fig 4A is not accurate. Srebf1 mRNA is unchanged by the GN-MCT1-siRNA treatment and Mlxipl mRNA is unchanged by Chol-MCT1-siRNA treatment. Is this total Mlxipl mRNA or can you distinguish between the alpha and beta variants.

RESPONSE: We adhered to NCBI nomenclature, where 'SREBP1' and 'ChREBP' represent proteins, not mRNA. The Mlxipl mRNA we tested pertains to total Mlxipl mRNA. Original draft shown below.

“To investigate the underlying mechanism by which lipid droplet morphological dynamics change, we monitored the effect of hepatic MCT1 depletion on DNL-related gene expression. Both GN-MCT1-siRNA and Chol-MCT1-siRNA strongly decreased the mRNA and protein levels related to representative DNL genes (Supplement Figure 4A-4D). Intriguingly, both modes of hepatic MCT1 depletion also inhibited expression of the upstream regulatory transcription factors SREBP1 and ChREBP.”

There are no molecular weight markers in supplement Fig 4C and D. Is the Srebp1c blot for the nuclear or precursor form?

RESPONSE: The Srebp1c blot presented represents the precursor form. I have edited the figure legend accordingly. It's worth noting that the cleaved form of Srebp1c either exhibited significantly lower expression compared to its precursor form or displayed comparable expression between the control group and the MCT1 depletion group.

Changes in mRNA and protein do not always reflect changes in activity (allosteric regulation). If you want to draw any conclusions about de novo lipogenesis you need to directly measure fatty acid synthesis rates from a carbohydrate precursor.

RESPONSE: We completely agree. Therefore, in the current study, we emphasized two key points: (1) hepatic MCT1 depletion affects the expression levels of representative DNL genes, and (2) however, this regulation was insufficient to resolve the steatosis phenotypes in our NASH model. We have added the text “while recognizing that the decreased expression of DNL genes does not necessarily indicate inhibited fatty acid synthesis rate” on page 15.

Reviewer #3 (Recommendations For The Authors):

Figure 1 - Are there changes to fibroblast phenotype with TGF-beta stimulation and are these changes reversed with MCT1 siRNA-mediated silencing, or is this purely an expression phenomenon?

RESPONSE: This study was designed to assess the preventative effect of MCT1 silencing on Tgf1β-induced fibrosis, rather than a reversal study. As detailed in the methods section, LX2 cells were initially cultured in DMEM/high glucose media with 2% FBS. The following day, we transfected the cells with either NTC-siRNA or MCT1-siRNA (IDT, cat 308915476) using Lipofectamine RNAi Max (ThermoFisher, cat 13778075) for 6 hours in serum-reduced Opti-MEM media (ThermoFisher, cat 31985062). Subsequently, the cells were maintained in serum-starved media, with or without 10ng/ml of recombinant human Tgf1β (R&D Systems, cat 240-B/CF), for 48 hours before harvesting.

Is lactate import/export itself responsible for this phenotype? It is presumed that MCT1 depletion alters import/export of lactate and subsequently modulates this phenotype, but this is never shown experimentally. Does lactate accumulate in these cells or in the medium in culture? The foundation of the paper rests on this hypothesis, so we believe that this is critical to establish. This is particularly relevant as MCT1 has been proposed to function primarily as a lactate importer, so the availability of medium lactate could be easily modulated to determine whether that mimics MCT1 loss.

RESPONSE: To address the underlying mechanism of MCT1/Lactate in stellate cells, we added a new figure to the manuscript (Figure 8). We had previously conducted an experiment to determine whether MCT1 depletion in LX2 cells in vitro influences extracellular lactate concentrations in DMEM/high glucose (25mM glucose) media supplemented with 1mM sodium pyruvate but without sodium lactate. Interestingly, we found no significant difference in extracellular glucose and lactate concentrations, which remained at 25mM and 5mM, respectively. These concentrations were comparable between groups, regardless of MCT1 loss. Additionally, we investigated the effects of MCT1 silencing in the presence of potent fibrogenic inducer TGF-β1. Intriguingly, MCT1 depletion effectively prevented TGF-β1-induced collagen production, irrespective of lactate (+/- pyruvate) supply in the media. LX2 cells with MCT1 depletion exhibited reduced collagen 1 production when lactate was solely generated by endogenous glycolysis (Figure 8F) and when exogenous lactate was supplied (Figure 8G).

Figure 2 - It is compelling that the Chol-MCT1-siRNA compounds are effective at targeting MCT1. However, is it clear how specific the siRNA target is? Are other MCT genes affected as well (if the siRNAs target areas of homology, for example)? Given that this siRNA strategy is used going forward and proposed as a therapeutic, it would be important to discuss and perhaps characterize off-target effects. A simple BLAST search for homology for the chosen siRNAs could help answer this question.

RESPONSE:

1) We designed the siRNA to specifically avoid any potential off-target effects on MCT1's 14 isoforms, and this approach aligns with the results obtained from the NCBI-BLAST analysis.

2) While there are 14 isoforms of MCTs, only the first four are functional. To assess the off-target effect of Chol-MCT1-siRNA on MCT2 and MCT4 (MCT3 was excluded due to its limited expression in retinal pigment epithelium), we conducted in vivo experiments in ob/ob mice, which demonstrated a highly selective MCT1 silencing effect. We have also included MCT1, MCT2, and MCT4 rt-qPCR data in the manuscript (Supplement Figure 2A, 2B).

3) We plan to further optimize and validate the human MCT1-targeting siRNA sequence for use in humanized mouse studies. It's important to note that the MCT1-siRNA used in this study was designed for mice.

Supplemental Figure 1 - brain would be one other highly metabolic tissue wherein it would be important to show lack of activity/accumulation.

RESPONSE: Undoubtedly, the brain is one of the most metabolically active tissues, playing a pivotal role in regulating signaling pathways and metabolism in other tissues. However, it poses a significant challenge in terms of targeting due to the presence of the blood-brain barrier (BBB). Overcoming BBB penetration remains one of the foremost challenges in the field of therapeutic siRNA delivery. For many therapeutic oligonucleotides, including Cholesterol-conjugated siRNAs, systemic administration alone is normally insufficient to achieve BBB penetration. Direct local injection or transient disruption of the BBB is normally required.

Figure 4 - The image shown for chol-MCT1-siRNA seems to show variation in lipid droplet size. Is this just this single image? The authors quantify smaller lipid droplets in this group, so the image may not be representative as there are many large droplets. Ultimately, additional mechanisms as to how alterations in lactate metabolism could mediate this phenotype are missing. This hypothesis also rests upon the assumption that MCT1 is modulating lactate, which is not shown experimentally, as discussed above.

RESPONSE: We changed the representative images (Fig 4B). We agree this aspect of the study is not resolved, and we have related text in the manuscript on this point: “neither GNMCT1-siRNA nor Chol-MCT1-siRNA decreased total hepatic TG levels (Figure 4H), although quantitative analysis of H&E images showed a small decrease in mean lipid droplet size and increased number of lipid droplets upon MCT1 silencing (Figure 4F, 4G). These data suggest the possibility that hepatic MCT1 depletion either 1) inhibits formation or fusion of lipid droplets, or 2) enhances lipolysis to diminish lipid droplet size.”

Figure 5 provides evidence that Chol-MCT1-siRNA expression decreases fibrosis but this is attributed to the effects on stellate cells. While GN-MCT1-siRNA and subsequent MCT1 silencing in hepatocytes has an opposite effect. The cell population that is not discussed, however, is the Kupffer cell. Could MCT1 silencing in this cell population be mediating part of the phenotype observed? How does MCT1 silencing affect Kupffer cell phenotype and activity?

This extends into Figure 6 where Kupffer cells are not given consideration in targeted experiments.

RESPONSE: Described above to Reviewer #3

Figure 6 and 7 use a different model to show that stellate cell depletion of MCT1, specifically, decreases collagen 1 protein levels in NASH, which reinforces the authors claims. Given the cell specificity of this experiment, it is more compelling data. It would be nice to show that Kupffer cell depletion of MCT1 does not have any affect (or perhaps show that it does.

RESPONSE: We agree, but Kupffer selective depletion is not possible to do with this siRNA technology. Please see the response above as our most recent attempt to address this question.

Figure 7 shows that even with decreased collagen deposition, there is no effect on liver stiffness or chronic liver injury as measure by ALT. This may suggest that the decreased level of fibrosis is either not significant to overall clinical outcome or that there are other fibroinflammatory mechanisms compensating for lack of COL1 deposition. Is there increased reticulin fibrosis when MCT1 is knocked down? This could be assessed with IHC or monitoring type 3 collogen (COL3A1).

RESPONSE: Reticulin fibrosis results from the excessive deposition of reticular fibers, primarily composed of type 3 collagen. However, based on our observation of trichrome staining in whole liver histology data (Fig 7D-E), which exhibited nearly identical trends to collagen type 1 expression (Fig 7A-C), it seems unlikely that type 3 collagen compensated for the decrease in type 1 collagen protein expression upon hepatic stellate cell MCT1 KO. We plan to perform detailed analysis of a more comprehensive list of ECM proteins including type 3 collagen in our humanized mouse model with engrafted human liver cells in future experiments.

Additional considerations:

It may be useful to know if inhibition of fibrosis affects survival/progression in these NASH models over a longer timeframe, although this may understandably be beyond the scope of the current work. The timing of MCT1 depletion is prophylactic and given the proposal to translate this research, it would be important to determine whether MCT1 inhibition reversed fibrosis, and if so, by what metabolic mechanism?

RESPONSE: We have observed that extending the duration of the NASH model increases the likelihood of hepatocarcinoma development. Exploring the aim to include survival and disease progression as well as reversal of fibrosis would be important in future experiments.

Summary of new Figures and Figures modified:

• Fig 1B: added "and" (significance) between the first and the third group, and the second and the last group.

• Fig 4B: replaced images with more representative ones as the mean lipid size was questioned by the reviewer.

• Fig 7D: made the images bigger (original images cropped and enlarged → 5X)

• Fig 8: newly created to explain the underlying pathway of lactate, and MCT1 regulating collagen production. Please find the results sections.

• Sup fig 2A, B: newly added to show our compounds’ selective silencing effect. - Sup Fig 2C-D: Added missing x-axis (moved from previous Figure 2A, 2B) - Sup Fig 2E-F: moved from sup Fig 3 not to have too many sup figures.

• Sup Fig 3C-D: showed both precursor and cleaved form of SREBP1 bands as requested (moved from previous sup Figure 4)

• Sup Fig 4: newly created, as questioned many times for the effect on Kupffer cells or other inflammatory cells.

• Sup Fig 6: newly created to explain the potential underlying mechanism of MCT1 depletion on collagen production.

• Sup Fig 7: moved from previous sup Fig 6.

• Sup Fig 8: moved from previous sup Fig 7.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

The authors have previously employed micrococcal nuclease tethered to various Mcm subunits to the cut DNA to which the Mcm2-7 double hexamers (DH) bind. Using this assay, they found that Mcm2-7 DH are located on many more sites in the S. cerevisiae genome than previously shown. They then demonstrated that these sites have characteristics consistent with origins of DNA replication, including the presence of ARS consensus sequences, the location of very inefficient sites of initiation of DNA replication in vivo, and for the most part are free of nucleosomes. They contain a G-C skew and they locate to intergenic regions of the genome. The authors suggest, consistent with published single molecule results, that there are many more potential origins in the S. cerevisiae genome than previously annotated, but also conclude that many of the newly discovered Mcm2-7 DH are very infrequently used as active origins of DNA replication.

The results are convincing and are consistent with prior observations. The analysis of the origin associated features is informative.

Specific Comments:

1. Page 8. The addition of an estimate of the most active origins using Southern blotting is fine for highly active origins, but how was Southern blotting used to calculate that 1-2% of cells in the eight cohort have an Mcm complex loaded.

We used a combination of Southern blotting and qPCR to measure licensing at the most active origins and then used our abundance curve to extrapolate these values to the less abundant cohorts. We expand on this point below, and we have changed the text to clarify this issue.

Reviewer #3 (Public Review):

By mapping the sites of the Mcm2-7 replicative helicase loading across the budding yeast genome using highresolution chromatin endogenous cleavage or ChEC, Bedalov and colleagues find that these markers for origins of DNA replication are much more broadly distributed than previously appreciated. Interestingly, this is consistent with early reconstituted biochemical studies that showed that the ACS was not essential for helicase loading in vitro (e.g. Remus et al., 2009, PMID: 19896182). To accomplish this, they combined the results of 12 independent assays to gain exceptionally deep coverage of Mcm2-7 binding sites. By comparing these sites to previous studies mapping ssDNA generated during replication initiation, they provide evidence that at least a fraction of the 1600 most robustly Mcm2-7-bound sequences act as origins. A weakness of the paper is that the group-based (as opposed to analyzing individual Mcm2-7 binding sites) nature of the analysis prevents the authors from concluding that all of the 1,600 sites mentioned in the title act as origins. The authors also show that the location of Mcm2-7 location after loading are highly similar in the top 500 binding sites, although the mobile nature of loaded Mcm2-7 double hexamers prevents any conclusions about the location of initial loading. Interestingly, by comparing subsets of the Mcm2-7 binding sites, they find that there is a propensity of at least a subset of these sites to be nucleosome depleted, to overlap with at least a partial match to the ACS sequence (found at all of the most well-characterized budding yeast origins), and a GC-skew centered around the site of Mcm loading. Each of these characteristics is related to previously characterized S. cerevisiae origins of replication.

Overall, this manuscript greatly broadens the number of sites that are capable of loading Mcm2-7 in budding yeast cells and shows that a subset of these additional sites act as replication origins. Although these studies show that the sequence specificity of S. cerevisiae replication origins still sets it apart from metazoan origins, the ability to license and initiate replication from sites with increasing sequence divergence suggests a previously unappreciated versatility.

Specific points:

1. The authors need to come up with a consistent name for loaded Mcms at an origin. In the manuscript they variously use 'MCM'(page 3), 'Mcm complexes' (page 4), 'MCM double hexamer' (page 6), and 'double-helicase' (page 8) to describe the Mcm2-7 complexes detected in their ChEC experiments. They should pick one name (Mcm2-7 double hexamer or MCM double hexamer would be the most accurate and clear) and stick with it throughout the manuscript.

We appreciate the criticism and agree that consistency is important for clarity, thus we tried using the term "Mcm2-7 double hexamer" in every instance in which we refer to Mcm loaded at an origin. However, upon reading the resulting manuscript, we felt that these changes hurt readability more than they helped with clarity, so we left the manuscript in its original form.

1. The authors state that "It is notable that, when Mcm is present, it is present predominantly as a single doublehexamer (right panel of Figure 3A), and that this remains true across the entire range of abundance shown in Figure 3A." This statement would be improved by prefacing it with "Based on the size of the protected regions" or some other clarifying statement that lets the reader know what they should be looking for in the data in 3A.


We thank the reviewer for the helpful suggestion. We have added the underlined words to the text to clarify this point.

It is notable that, when Mcm is present, it is present predominantly as a single doublehexamer (based on the size of the protected region in the right panel of Figure 3A), and that this remains true across the enAre range of abundance shown in Figure 3A.

1. The revised statements that "We have previously used Southern blotting to demonstrate that approximately 90% of the DNA at one of the most active known origins (ARS1103) is cut by Mcm-MNase (Foss et al., 2021), and to thereby infer that 90% of cells have a double- helicase loaded at this origin. Using this as a benchmark, we estimate that ~1-2% cells have an Mcm complex loaded at the Mcm binding sites in the eighth cohort (ranks 1401- 1600)." partially clarifies how the authors came to the 1-2% number, however, the calculation is still unclear. Based on Figure 1A, there are at least three logs (1,00 fold) difference in the number of CBMSs between the best origins (which is what they state the 90% comes from) to anywhere close to the 1400-1600 rank. Seems like the number should be at best 0.1% and probably less. Either way, the authors need to explain this calculation either in the text or in the text. This sort of number tends to get thrown around later and without a clear explanation readers cannot evaluate its credibility. 
<br /> We apologize for insufficiently clarifying how we arrived at our estimate of licensing. We believe that we have now remedied this, both by incorporating more measurements of licensing to improve our accuracy and by expanding the text to make our calculation unambiguous. We have added a supplemental figure showing the linear regression, based on 7 qPCR-based measurements of licensing, that we used to determine the median level of licensing of the first cohort of 200, and the altered text in the main text reads as follows:

We have previously used Southern blotting to demonstrate that approximately 90% of the DNA at one of the most active known origins (ARS1103) is cut by Mcm-MNase (Foss et al. 2021), and to thereby infer that 90% of cells have a double-helicase loaded at this origin. Combining this measurement with 6 additional measurements of licensing in cohort 1, we used linear regression (r2=0.7) to infer a median value of 69% for cohort 1. Because the median abundance in the 8th cohort is 1.5% of that in the first cohort, we estimate that CMBSs in the 8th cohort are typically licensed in 1% of cells in the population (69% x 0.015 = 1.0%).

1. The authors make the point in the introduction and discussion that recent single-molecule studies of replication origins indicate that as many as 20% of the origins identified are outside of known origins. This is very interesting but there seems to be a missed opportunity of comparing the location of these origins with the CBMSs. It would improve the manuscript to include some sort of comparison rather than using only the much older and less accurate ssDNA analysis.

Unfortunately, coverage and resolution with nanopore-based single-molecule precludes such an analysis.

1. The authors state at the end of the first paragraph on page 6 that the ChEC data is "very reproducible" which does seem to be the case but it is a little confusing for the knowledgeable reader since one would expect quite different results for an HU arrested strain versus a asynchronous or G1 arrested strain. This is hidden in the analysis in Figure S1 since 13 experiments are compared against one in each plot, however, if one x one comparisons were done there would certainly be substantial differences (or if there are not, there is a problem with the data - e.g. HU arrested cells should lack licensing at early firing origins).

It may appear counterintuiAve that one could obtain high r2 values when comparing G1 and HU-arrested samples. However, HU arrest was performed by transferring log phase cultures to 200 mM HU and harvesting cells after just 50 minutes. In this situation, most cells will be in G1 or very early S phase. Presumably, increasing times of incubation in HU would cause r2 values to decline.

1. On page 8 the authors state, "First, clear peaks of ssDNA extend down to the eighth cohort..." This seems to be stretching the data. There are clear peaks for the first five cohorts and then there is a notable change with any peak being much broader, extending over at least 10,000 bp. The authors should reconsider their statement here as it is not well supported by the data.

We have softened our language to the following: First, peaks of ssDNA signal, as judged by higher signal at the midpoints than the edges, extend down to the eighth cohort (brown line), which corresponds to CMBSs ranked 1401-1600.

1. There is one last missing reference. Wherever Eaton et al, 2010 is referenced Berbenetz, et al, 2010 (full ref below) should also be referenced as they come to very similar conclusions.

Berbenetz, N. M., Nislow, C. & Brown, G. W. Diversity of eukaryotic DNA replication origins revealed by genome-wide analysis of chromatin structure. PLoS Genet 6, (2010).

We have added this reference at all 4 instances in which we reference Eaton et al., 2010.

Recommendations for the authors:

Reviewer #3 (Recommendations For The Authors):

There are missing references in several places:

All references are included, and the references in point 3 have been split according to the reviewer's suggestion.

1. "For example, 15 of the 56 genes that contained a high abundance site have been implicated in meiosis and sporulation and are not expressed during vegetative growth (~5 out of 56 expected from random sampling), consistent with previous observations (Mori and Shirahige, 2007)." Should include Blitzblau et al., 2012 (PMC3355065) which showed that Mcm2-7 loading was impacted by differences in meiotic and mitotic transcription.

2. "In contrast to the low abundance sites, the most abundant 500 sites showed a preference for convergent over divergent transcription (left of vertical dotted line in Figure 4B), in agreement with a previous report (Li et al., 2014)." This preference was first pointed out in MacAlpine and Bell, 2005 (PMID: 15868424).

3. "This sequence is recognized by the Origin Recognition Complex (Orc), a 6-protein complex that loads MCM (Broach et al., 1983; Deshpande and Newlon, 1992; Eaton et al., 2010; Kearsey, 1984; Newlon and Theis, 1993; Singh and Krishnamachari, 2016; Srienc et al., 1985)." This list should include a reference to Bell and Stillman, 1992 (PMID: 1579162), which first described ORC and showed that it recognized the ACS. It would also be more helpful to the reviewer to distinguish the references that identified that ACS from those concerning ORC binding to it.

Author Response

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

On behalf of all the authors, I'd like to thank you for your insightful comments and valuable suggestions, which fully reflect your high level of scientific thinking and point the direction of our research and help us and other future researchers in the field to more comprehensively study and interpret the toxic effects of imidacloprid on honey bee larvae and its potential mechanisms, as well as the mechanisms of larval resistance and adaptations to imidacloprid. We have addressed each of the questions and revised the manuscript point-by-point in response to your comments. Below are detailed point-by-point responses to each question.

Public Review:

This study provides evidence of the ability of sublethal imidacloprid doses to affect growth and development of honeybee larva. While checking the effect of doses that do not impact survival or food intake, the authors found changes in the expression of genes related to energy metabolism, antioxidant response, and metabolism of xenobiotics. The authors also identified cell death in the alimentary canal, and disturbances in levels of ROS markers, molting hormones, weight and growth ratio. The study strengths come from exploring different aspects and impacts of imidacloprid exposure on honeybee juvenile stages and for that it demonstrates potential for assessing the risks posed by pesticides. The study weaknesses come from the lack of in depth investigation and an incomplete methodological design. For instance, many of the study conclusions are based on RT-qPCR, which show only a partial snapshot of gene expression, which was performed at a single time point and using whole larvae. There is no understanding of how different organs/tissues might respond to exposure and how they change over time. That creates a problem to understand the mechanisms of damage caused by the pesticide in the situation studied here. There is no investigation of what happens after pupation. The authors show that the doses tested have no impact on survival, food consumption and time to pupation, and the growth index drops from ~0.96 to ~0.92 in exposed larvae, raising the question of its biological significance. The origin of ROS are not investigated, nor do the authors investigate if the larvae recover from the damage observed in the gut after pupation. That is important as it could affect the adult workers' health. One of the study's central claims is that the reduced growth index is due to the extra energy used to overexpress P450s and antioxidant enzymes, but that is based on RT-qPCR only. Other options are not well explored and whether the gut damage could be causing nutrient absorption problems, or the oxidative stress could be impairing mitochondrial energy production is not investigated. These alternatives may also affect the growth index. The authors also state that the honeybee larvae has 7 instars, which is an incorrect as Apis mellifera have 5 larval instars. It is not clear from methods which precise stage of larval development was used for gut preparations. That information is important because prior to pupation larvae defecate and undergo shedding of gut lining. That could profoundly affect some of the results in case gut preparations for microscopy were made close to this stage. A more in-depth investigation and more complete methodological design that investigates the mechanisms of damage and whether the exposures tested could affect adult bees may demonstrate the damage of low insecticide doses to a vital pollinator insect species.

Recommendations for the authors:

This study presents a useful investigation on changes in gene expression by real time PCR and some of the physiological consequences of sublethal exposures to the neonicotinoid insecticide imidacloprid in honeybee larvae. It offers preliminary evidence of imidacloprid impacts on the development of bee larvae by interfering with molting and metabolism. Whereas the study provides evidence that small doses of imidacloprid affect larval growth rate, there is no investigation on whether that could affect the overall colony health, and some of the results open the possibility that the larvae may overcome some of the impacts of the exposure. As the authors state, the doses tested show no impact on larvae survival, food consumption or time to pupation. The investigation and methodological design lack in depth to explain the findings and provide incomplete evidence to support the authors conclusions. The study would benefit from a more thorough mechanistic characterization to better sustain the findings and demonstrate their biological relevance.

Response: I would like to express, on behalf of all the authors, our sincere appreciation for your insightful and insightful comments and suggestions, which significantly enhanced the quality of the manuscript. Your incisive insights point the way for future research in the field of bee biology on the mechanisms underlying imidacloprid-induced delays in larval development.

In this study, we investigated the effects of imidacloprid on honey bee larval development, including macro and micro changes and possible causes. This is the first of its kind in the field of honeybee biology research. However, we found that the underlying mechanism is extremely complex. The effects of toxic substances on animals and their interactions with larval development are complex and far-reaching. They include oxidative stress and damage; disruption of nutrient metabolic homeostasis; inhibition of detoxification and immunity; adverse effects on the nervous, circulatory, and digestive systems; inflammation, disease, and even organ failure; and subsequent effects on physiological activities such as development, reproduction, and behavior, and even death. These toxic effects interact in complex ways with the development of young animals, with some effects directly or indirectly affecting development while others do not.

Addressing this complex mechanistic issue based solely on the results of this study is a formidable challenge, which leads to some limitations of our study as pointed out by the reviewers. Although our study is not comprehensive enough in terms of mechanistic analysis and does not fully elucidate the mechanism, we believe it is an important and valuable first step in this area.

In the future, we will follow the reviewers' suggestions and deliberately redesign the experiments to focus on further research on the issues they raised. These include examining the effects of larval developmental delay on adult and colony health, investigating the post-pupal situation, identifying the source of ROS, and determining whether the larval gut damage observed after pupalization recovers.

In accordance with the reviewers' comments and suggestions, we have revised the manuscript to improve its rigor and scientific quality. We sincerely ask the reviewers to understand and accept this modification from us!

Next is our response to each of the questions and valuable suggestions provided by reviewers:

Recommendations For The Authors:

1. The authors found a reduction in growth index and body mass, but document no impact on survival, food consumption or time to pupariation. How much exactly is the reduction in growth index? It seems to be from ~0.96 to ~0.93. Is this biologically relevant? Would that be enough to impact the colony health?

Response: Thank you for your comments. In this study, we observed a gradual decrease in larval growth index from day 4, which stabilized by day 6. At the 4th, 5th and 6th instars, the growth index of the imidacloprid-treated groups were significantly lower than those of the control group by an average of 1.35%, 4.49% and 2.76%, respectively (Figure 1, source data 8). Statistical analysis confirmed the significance of the difference in these results. We have incorporated the above description into the red text on lines 148-152 of the Results section. Regarding the reviewer's inquiry on colony health, including imidacloprid-induced delayed larval development and some reduction in growth index and body weight with no effect on survival, food consumption, or time develop to pupation, because we do not currently have the technical capabilities to culture larvae to adulthood in laboratory incubators, this has resulted in a failure to further investigate the effects of imidacloprid-induced delayed larval development on adult colony health. However, this is a very important scientific question for future colony health. We will design experiments to address this issue in a follow-up study.

1. The authors find that P450s can help in detoxifying mechanisms to mitigate imidacloprid impacts. That however is a well-known fact. What is new about this claim?

Response: The point at which the ability to detoxify toxic substances is acquired during early development varies widely among animals. Although many studies have reported that the detoxification function of P450s helps mitigate the effects of imidacloprid in adult honey bees, there is no conclusive evidence as to whether or not honey bee larvae have acquired this ability at early stages of development. This ability is critical to the defense and health of honey bee larvae. Therefore, it is incumbent upon this study to clarify this issue, which is important in explaining the effects of imidacloprid on honey bee larvae.

1. Some references are cited incorrectly. The first and last name are swapped, for instance Charles et al.

Response: Thank you very much for pointing out this error, which we have corrected. Please see lines 92 and 889 in our revised version.

1. I still encounter important methodological flaws. The authors acknowledge my previous suggestions but only address a small fraction of them. The most relevant points regarding the understanding of the mechanisms behind the delayed growth rate remain unexplored. The expression levels of other nAChRs target of imidacloprid in honeybees were not investigated. The expression analyses are still based on a single time point and using whole larvae, which only superficially explore the problem and may lead to misinterpretations. I do not understand the authors claim that a technological breakthrough is required to address these issues, when performing more PCRs and doing dissections should cover the matter.

Response: Thank you very much for your important comment. You point out several unexplored issues related to understanding the mechanisms behind delayed growth rates. For example, The most relevant points regarding the understanding of the mechanisms behind the delayed growth rate remain unexplored. The expression levels of other nAChRs target of imidacloprid in honeybees were not investigated. The expression analyses are still based on a single time point and using whole larvae. Please allow me to explain. Honeybees (Apis mellifera) have nine different α-subunits, Amelα1-9, and two β-subunits, Amelβ1-2. Amelα5, Amelα7, and Amelα8 are expressed in MB Kenyon cells and AL neurons, and the Amelβ2 subunit is present in Kenyon cells. Amelα2, Amelα3, and Amelα7-2 are expressed in the optic lobes. The aim of this study was to investigate whether imidacloprid induces larval neurotoxicity. Based on the above information, we selected the two most representative nAChRs (Alph1 and Alph2) for analysis. The results showed that exposure to imidacloprid increased the expression of the Alph2 gene and inhibited AChE activity, indicating that imidacloprid is neurotoxic to larvae. This result answered our question of whether imidacloprid induces neurotoxicity in larvae. Therefore, we did not further analyze the expression levels of other nAChRs. We believe that this does not affect the understanding of the mechanism behind the delayed growth rate and that it is not necessarily necessary to analyze all 11 nAChRs to find an answer. We sincerely hope that the reviewers will understand and agree with this.

Furthermore, regarding the expression analysis based on a single time point and whole larvae. In this study, 72 h after imidacloprid exposure Fig. 1J, 5 days of age) was chosen for sampling because this is when imidacloprid has the greatest and most representative effect on larval development. Therefore, analyzing samples at this time point did not interfere with our exploration of the mechanisms by which imidacloprid causes larval developmental retardation. We used whole larvae rather than individual tissues for sample selection, which is a shortcoming for us. This was mainly due to technical challenges where we were unable to obtain pure single tissues through dissection. Nevertheless, we will make technical breakthroughs in the future so that we can sample and compare different tissues and developmental stages to obtain more comprehensive and accurate data. Thank you again for raising this important issue and for your valuable suggestions.

1. The authors could in many different ways explore what are the origin of ROS is. That is important to further develop their hypothesis on reduced energy levels.

Response: Thank you very much for your insightful comment and suggestion, it gives us great insight. Mitochondria are the main producers of ATP for cellular metabolism, accounting for approximately 90% of the total. However, mitochondria are also involved in the generation of reactive oxygen species (ROS). Excessive accumulation of ROS in mitochondria leads to oxidative stress, which in turn damages mitochondria and further increases ROS levels, creating a vicious cycle (Boovarahan and Kurian, 2018). In the present study, it was found that imidacloprid exposure led to increased ROS and MDA levels in larvae (Figure 5A and Figure 5-source data 14), indicating that imidacloprid induced severe oxidative stress and lipid damage, which may damage mitochondria and in turn affect mitochondrial ATP production, resulting in insufficient energy supply for larval development. This factor may also be an important explanation for the larval developmental delay caused by imidacloprid. We have included the above text in our revised manuscript. Please see the lines 432-442 in the revised manuscript.

1. If there is gut damage, is it restored in the adults? It is not clear from the methods which precise stage of larval development was used for gut preparations. That information is important because prior to pupation larvae defecate for the first time and undergo shedding of the gut lining. That could profoundly affect some of the results in case gut preparations for microscopy were made close to this stage. If no food residues are found in the gut of control larvae, does it mean that they are close to pupation? Could the apoptosis found in gut of exposed larvae be the natural shedding of gut lining prior to pupation? All these possibilities have to be discussed and authors should clarify the precise larval stage used in every assay.

Response: Thank you for your important comments. In this study, all samples used for the assay were larvae that had developed to 5-day-old after oral administration imidacloprid at 2-day-old. This is described in detail in the Materials and Methods. See lines 507, 517-521 in the revised manuscript. In general, 6-day-old bee larvae cease feeding and begin their first defecation at approximately 7-day-old. However, in our study, intestinal sections were prepared from 5-day-old larvae that had not fasted or defecated, when the intestinal mucosa was normal and not undergoing shedding. In this case, we found that imidacloprid caused damage to intestinal structures, apoptosis of intestinal cells, incomplete formation of the peritrophic membrane, and undigested food residues in the intestine. We believe that these results are objective and reliable.

1. Honeybee have 5 larval instars, not 7 (Figure 1). That creates confusion about which larval stage the authors used.

Response: Thank you very much for pointing out this editorial error, which we have corrected, please see Figure 1.

1. The Results section does not state the numbers by which parameters measures have changed, neither the values of significance. How much is the impact in growth index, body mass, gene fold change, etc?

Response: Thank you very much for pointing out this important problem. We have revised the Results section according to your suggestions. Please see the revised manuscript.

1. Mention figures in order (5c comes before 5b in the text)

Response: Thank you very much for the comment. We have revised according to your suggestions. Please see the lines 208-212 in the revised manuscript.

1. Paraquat is a herbicide not a pesticide

Response: Thank you for pointing out the loose wording. We have revised according to your suggestions. Please see the lines 316-319 in the revised manuscript.

1. What is the evidence that imidacloprid reduces growth index by inhibiting 20E? The authors provide real time data and discuss the data in terms of correlation. But correlation does not mean causation. Reduction in growth index could come from multitude of factors such as ROS affecting mitochondrial energy metabolism.

Response: We deeply appreciate your insightful comments and valuable suggestions. In this study, although we conducted an in-depth analysis of ecdysone regulation, which is crucial for insect larval development, and found some clues, as you pointed out, this is not the sole reason for larval developmental delay. In fact, animal growth and development are collectively regulated by numerous physiological, biochemical, and genetic factors. The the decline in the growth index may be due to other factors as you mentioned, such as oxidative stress impairing mitochondria, dysregulated neuro-endocrine axis caused by imidacloprid targeting neurons, poor nutrient absorption, impaired movement, etc, as animal growth and development are collectively regulated by numerous physiological, biochemical, and genetic factors. We have incorporated this understanding into the revised manuscript. Please see the lines 389-394 in the revised manuscript.

1. The authors state that "digestion and breakdown of nutrients is impaired by imidacloprid", the evidence discussed in the paragraph however supports only that imidacloprid impairs some of the genes involved in these processes.

Response: Thank you for your comments and valuable insights. In this paragraph, a lack of clarity and completeness in our writing may have led to the misconception that the evidence discussed only demonstrates the effects of imidacloprid on specific genes in these processes. In fact, our intent in this paragraph was to analyze and discuss the effects of imidacloprid on nutrient digestion and breakdown in larvae and to explore the causes of larval developmental delay. We demonstrated this using tissue sections, qRT-PCR and correlation analysis, which showed that the intestinal structure was disrupted and the expression of genes involved in nutrient digestion and catabolism was suppressed, resulting in defects in the catabolic utilization of food and consequently the presence of many food residues. In addition, there was a positive correlation between these genes and larval developmental delay. All this may be another important factor contributing to imidacloprid-induced larval developmental delay. We have revised and incorporate the above logic into the revised manuscript. Please see the lines 407-431 in the revised manuscript.

1. There is no evidence for the claim that overexpressing P450s and antioxidant enzymes cause a reduction in growth index. No transcriptome analysis was performed so it is unknown under the circumstances presented here how all the other P450s, antioxidant genes and overall gene profiles are responding. Surely, some genes will be repressed. Reduction in growth index could stem from, oxidative stress impairing mitochondria, dysregulated neuro-endocrine axis caused by imidacloprid targeting neurons, poor nutrient absorption, impaired movement, etc.

Response: Thank you for your comments and valuable insights. Indeed, as you have pointed out, drawing the conclusion that antioxidants and detoxification are significant contributors to larval developmental retardation solely based on correlation analysis is inherently flawed and lacks critical support, especially in the absence of P450 and antioxidant enzyme overexpression and comprehensive transcriptome analysis of other P450s, antioxidant genes, and the entire gene map. We have revised and included in the revised manuscript. Please see lines 461-467 in the red text in the revised manuscript. We have revised and incorporate the above logic into the revised manuscript. Please see the lines 407-431 in the revised manuscript.

1. How come the decreased ATP and glycogen levels have no effect on time to pupation? Extra time points for gene expression, measurements of gut damage, ATP levels, ROS, etc, are vital to answer how the exposed larvae eventually catch up with the unexposed group. Also, it is vital to understand whether these larval impacts translate to impacts on adults.

Response: We sincerely thank you for your insightful comments and suggestions! These important scientific issues you've raised are a good example of your high-level scientific thinking, and they will help us and other future researchers in the field to more comprehensively study and interpret the toxic effects of imidacloprid on honey bee larvae and their potential mechanisms, as well as the mechanisms of larval resistance and adaptation to imidacloprid. According to your comments, we will adapt our experiments and conduct more thorough research in the future to address the above issues.

1. I am confused about the author's definition of developmental rate; rate gives the notion of speed to achieve something. But the authors use developmental rate as a measure of viability (number of larvae that successfully pupated). There seems to be a significant decrease in their developmental rate plot (Fig 1i), but at the same time the authors show in Figure 1c (and mention throughout the manuscript) that there is no difference in probability of survival. This is quite confusing and the method section regarding these data is too concise and does little to help explain what the authors were trying to measure. The whole section on developmental traits would benefit of more details on how experiments were conducted and equipment used.

Response: Thank you so much for your valuable comments. Yes, as you can see, there appears to be a significant decrease in developmental rate but no difference in survival probability, which is an intriguing finding of this study. This finding suggests that the 377 ppb imidacloprid dose is not as harmful to the larvae as previously thought. Imidacloprid appeared to limit the larval ability to molt and develop only to a certain extent, but had no effect on the developmental process, let alone survival. It's worth investigating the underlying mechanism. As a result, we have included this question in the design of future studies. In addition, following your suggestion, we have revised the description of the material and methods in this section, including the experimental method in more detail. For more information, please see the revised manuscript, lines 530-541.

1. The authors should try to make it clear what percentage of exposed larvae become adults? I am confused because the plot called developmental rate might be trying to convey this message, but developmental rate and viability are very distinct traits. What is the difference, if any, in the time it takes for exposed larvae to become adults in comparison to non-exposed ones? Is there a difference in adult body weight? The answers to these last two questions are important to start understanding if the impacts of imidacloprid on larvae alimentation would still impact these same individuals once they become adults, i.e., would there be impacts for the colony and workers activity?

Response: Thank you very much for your insightful comments. Unfortunately, this is where the research falls short. Culturing larvae to adulthood in 24-well cell culture plates is a significant technical challenge that we have yet to overcome. As a result, the important questions you raise, such as what percentage of exposed larvae become adults? How does the time to adulthood differ (if at all) for exposed larvae versus non-exposed larvae? Is there a difference in adult weight? Do the effects of imidacloprid on larval feeding persist after these individuals reach adulthood? Does imidacloprid damage to larvae affect colony and adult activity? We do not have answers at this time. We are aware that answers to the above questions will help people better understand how serious the effects of imidacloprid environmental residues on honey bee larvae and adults, as well as bee colonies as a whole, are, and will draw sufficient attention to them. We intend to break through this technological bottleneck of culture larvae to adulthood in future studies and incorporate the above scientific questions into our next research design. Thank you again for your insightful comments! This gives us new research ideas.

Author Response

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

eLife assessment

This important paper builds on a method, previously conceptualized and validated, of genetic control for insect populations. The method, called pgSIT, uses integrated CRISPR-Cas9 based constructs to generate, in certain combinations of genotypes, mutations that cause both male sterility and female inviability. Release of such genotypes in sufficiently large numbers can lead to an inundation of a local insect population with sterile males and this can lead to localised population suppression, which represents an important method of control for problematic insect populations. The data are convincing and will be valuable to anyone working on vector control strategies.

Public Reviews:

Reviewer #1 (Public Review):

Precision guided sterile insect technology (pgSIT) is a means of mosquito vector control that aims to simultaneously kill females while generating sterile males for field release. These sterile males are expected to mate with 'wild' females resulting in very few eggs being laid or low hatching rates. Repeated releases are expected to result in the suppression of the mosquito population. This method avoids cumbersome sex-sorting while generating the sterile males. Importantly, until release, the two genetic elements that bring about female lethality and male sterility - the Cas9 and the gRNA carrying mosquitoes - are maintained as separate lines. They are crossed only prior to release, and therefore, this approach is considered to be more safe than gene drives.

The authors had made a version of this pgSIT in their 2021 paper where they targeted β-Tubulin 85D, which is only expressed in the male testes and its loss-of-function results in male sterility. In that pgSIT, they did not have female lethality, but generated flightless females by simultaneously targeted myosin heavy chain, which is expressed only in the female wings. Here the authors argue, that the survival of females is not ideal, and so modify their 2021 approach to achieve female lethality/sterility.

To do this, they target two genes - the female specific isoform of Dsx and intersex. They use multiple gRNAs against these genes and validate their ability to cause female lethality/sterility. Having verified that these do indeed affect female fertility, they combine gRNAs against Dsx and ix to generate female lethality/sterility and use β-Tubulin 85D to generate male sterility (previously validated). When these gRNA mosquitoes are crossed to Cas9 and the progeny crossed to WT (the set-up for pgSIT), they find that very few eggs are laid, larval death is high, and what emerges are males or intersex progeny that are sterile.

As this is the requirement for pgSIT, the authors then test if it is able to induce population suppression. To do this, they conduct cage trials and find that only when they use 20:1 or 40:1 ratio of pgSIT:WT cages, does the population crash in 4-5 generations. They model this pgSIT's ability to suppress a population in the wild. Unfortunately, I was not able to assess what parameters from their pgSIT were used in the model and therefore the predicted efficacy of their pgSIT, (though the range of 0-.1 is not great, given that the assessment is between 0-0.15).

We express our sincere appreciation for the valuable comments received. A wide range of ♀ viability and ♂ fertility values were explored in the model. The results determined that: “Achieving a ≥90% probability of elimination places slightly tighter restrictions on ♀ viability and ♂ fertility - a safe ballpark being ♀ viability and ♂ fertility both in the range 0-0.10, given a release scheme of ~26 releases of 250 pgSIT eggs per wild adult (Fig. 4B). These results suggest a target product profile for pgSIT to be ♀ viability and ♂ fertility both in the range 0-0.10.” A subsequent sentence has been added pointing out how the described pgSIT strain falls within this range: “The pgSIT strain described here falls well within these bounds, with ♀ viability of 0 and ♂ fertility of ~0.01.” The parameters of the described pgSIT strain are also listed throughout the paper and quoted here: “Cas9 in combination with gRNAdsx,ix,βTub induces either the lethality or transformation of pgSIT ♀’s into sterile unfit ⚥’s.” And: “Firstly, we determined that pgSIT males were not 100% sterile, with an estimated ~1% still producing some progeny.”

Finally, they also develop a SENSR with a rapid fluorescence read-out for detecting the transgene in the field. They show that this sensor is specific and sensitive, detecting low copy numbers of the transgene. This would be important for monitoring any release.

Overall, the data are clear and well presented. The manuscript is well written (albeit likely dense for the uninitiated!). I had concerns about the efficacy of generating the pgSIT animals - the overall number of eggs hatched from the gRNA (X) Cas9 cross appears to be low, therefore, very large numbers of parental animals would have to be reared and crossed to obtain enough sterile males for the SIT. In addition to this, I was concerned about the intersex progeny that can blood-feed. These could potentially contribute to the population and it would be useful to see the data that suggest that these numbers are low and the animals will not be competent in the field.

Reviewer #2 (Public Review):

This is a thorough and convincing body of work that represents an incremental but significant improvement on iterations of this method of CRISPR-based Sterile Insect Technique ('pgSIT'). In this version, compared to previous, the authors target more genes than previously, in order to induce both female inviability (targeting the genes intersex and doublesex, compared to fem-myo previously) and male sterility (targeting a beta-tubulin, as previously in the release generation. The characterization of the lines is extensive and this data will be useful to the field. However, what is lacking is some context as to how this formulation compares to the previous iteration. Mention is made of the possible advantage of removing most females, compared to just making them flightless (as previously) but there is no direct comparison, either experimental, or theoretical i.e. imputing the life history traits into a model. For me this is a weakness, yet easily addressed. In a similar vein, much is made in alluding to the 'safety concerns of gene drive' and how this is a more palatable half-way house, just because it has CRISPR component within it; it is not. It would be much more sensible, and more informative, to compare this pgSIT technology to RIDL (release of insects carrying a dominant lethal), which is essentially a transgene-based version of the Sterile Insect Technique, as is the work presented here.

We express our sincere appreciation for the valuable comments received. A wide range of ♀ viability and ♂ fertility values were explored in the model. Given the intricate nature of this study and taking into account the recommendations provided by multiple reviewers and the editor, we have eliminated superfluous comparisons among various methodologies.

The authors achieve impressive results and show that these strains, under a scenario of high levels of release ratios compared to WT, could achieve significant local suppression of mosquito populations. The sensitivity analysis that examines the effect of changing different biological or release parameters is well performed and very informative.

The authors are honest in acknowledging that there are still challenges in bringing this to field release, namely in developing sexing strains and optimizing release strategies - a question I have here is how to actually release eggs, and could variability in the efficiency of this aspect be modelled in the sensitivity analysis? It seems to me like this could be a challenge and inherently very variable.

We really appreciate comments. Several approaches are available to release eggs - either in pre-existing breeding sites in the field, or in artificial breeding sites (e.g., cups). We have added a sentence in the Discussion section to highlight that this is an area requiring further research: “Secondly, studies are required to determine the survival and mating competitiveness of released pgSIT males under field conditions, and to optimize their release protocol.” Regarding the efficiency of egg releases, the following sentence in the modeling results section has been added: “We assume released eggs have the same survival probability as wild-laid eggs; however if released eggs do have higher mortality, this would be equivalent to considering a smaller release.” As stated in the modeling results (and depicted in Figure 4 and Supplementary Figure 5): “Suppression outcomes were found to be most sensitive to release schedule parameters (number, size and interval of releases), ♂ fertility and ♀ viability.” It follows that suppression outcomes are equivalently sensitive to the efficiency of an egg release.

Reviewer #3 (Public Review):

Summary and Strengths:

The manuscript by Li et al. presents an elegant application of sterile insect technology (pgSIT) utilizing a CRISPR-Cas9 system to suppress mosquito vector populations. The pgSIT technique outlined in this paper employs a binary system where Cas9 and gRNA are conjoined in experimental crosses to yield sterile male mosquitoes. Employing a multiplexed strategy, the authors combine multiple gRNA to concurrently target various genes within a single locus. This approach successfully showcases the disruption of three distinct genes at different genomic positions, resulting in the creation of highly effective sterile mosquitoes for population control. The pioneering work of the Akbari lab has been instrumental in developing this technology, previously demonstrating its efficacy in Drosophila and Aedes aegypti. By targeting the female-specific splice isoform (exon-5) of doublesex in conjunction with intersex and β-tubulin, the researchers induce female lethality, leading to a predominance of sterile male mosquitoes. This innovation is particularly noteworthy as the deployment of sterile mosquitoes on a large scale typically requires substantial investment in sex sorting. However, this study circumvents this challenge through genetic manipulation.

Weaknesses:

One notable concern arising from this manuscript pertains to the absence of data regarding the potential off-target effects of the gRNA. Given the utilization of multiple gRNA, the risk of unintended mutations in non-target areas of the genome increases. With around 1% of males still capable of producing fertile offspring, understanding the frequency of unintended genome targeting becomes crucial. Such mutations could potentially become fixed within the natural population.

We express our sincere appreciation for the valuable comments received and fully agree with the reviewer regarding the importance of understanding the frequency of unintended genome targeting. However, the likelihood of off-target effects becoming fixed within the population is exceedingly low. To mitigate potential negative impacts, we employed CHOPCHOP V3.0.0 (https://chopchop.cbu.uib.no) for the selection of gRNAs, which will specifically tminimize the occurrence of genomic off-target cleavage events. Furthermore, our releasing process will be carried out in multiple rounds. In the event that an undesired mutant is introduced into the local population, the mutated gene will either be quickly eradicated through subsequent rounds of releases or be naturally eliminated through the process of natural selection over time.

The experiments are well-conceived, featuring suitable controls and repeated trials to yield statistically significant data. However, a primary issue with the manuscript lies in its data presentation. The authors' graphical representations are intricate and demand considerable attention to discern the nuances, especially due to the striking similarity between the symbols representing different genotypes. As it stands, the manuscript primarily caters to experts within the field, thereby warranting improvements in data visualization for broader comprehension.

We appreciate the comment. However, as this work is indeed complex and intricate and as there is limitations imposed by the publisher on data visualizations (i.e. number of figures in the main text, etc.) we have tried our best for presenting our data in full.

All three reviewers were appreciative of the work presented in this manuscript. There were some common concerns that we shared, that the authors could consider revising. They are listed below.

Essential revisions:

1. Formal comparison with the previous/other methods: The authors make many statements that compare this pgSIT with their previous method, gene drives, or with RIDL. We suggest that they focus their comparisons within the scope of data and avoid comparisons between RIDL, gene drive, and pgSIT that are based on perceptions of these methods. It would be useful if, for example, they could impute life history traits and demonstrate this pgSIT's efficacy over their previous versions.

We express our sincere appreciation for the valuable comments received. We have removed the unnecessary comparisons between different methods, please review the revised version.

1. Writing and presentation of figures: The authors should please take advantage of the eLife format and unpack each sentence/figure so that it's accessible to readers outside this field.

We appreciate your comment, and we have implemented some necessary changes based on your suggestions.

1. Data to support claims made in passing: There are many instances, such as detailed in the reviews (and the entire second paragraph in the discussion) that are not supported by data. The authors should either provide that data or not make these claims.

Thank you for the comment. We have removed these claims.

1. Off target effects: There is the formal possibility that off target effects that might get fixed in the population. Could the authors please address this in the discussion.

We appreciate the comment and fully agree with the reviewer regarding the importance of understanding the frequency of unintended genome targeting. However, the likelihood of off-target effects becoming fixed within the population is exceedingly low. We have address this in the discussion.

“Even though mutations could potentially become fixed within the natural population, the likelihood of off-target effects becoming fixed within the population is exceedingly low. To mitigate potential negative impacts, we employed CHOPCHOP V3.0.0 (https://chopchop.cbu.uib.no) for the selection of gRNAs, specifically to minimize the occurrence of genomic off-target cleavage events. Furthermore, our releasing process will be carried out in multiple rounds. Even in the event that an undesired mutant is introduced into the local population, it will either be completely eradicated through subsequent rounds of releases or be naturally eliminated through the process of natural selection over time.”

Aside from this, we ask that the authors please pay attention to the detailed reviews.

Reviewer #1 (Recommendations For The Authors):

The writing: Each sentence is packed with information and while this is fine for those immersed in the field, it might be dense for those who are not. There are a lot of nuances in such an approach and clearly laying it out for the reader is important. The authors should unpack some of these sentences to make their work more accessible.

Thank you for the comment. We have unpacked some of sentences, please review the revised version.

It will help to have a schematic linked to the introduction about how these mosquitoes are designed to be used. Which strains would be scaled up in the lab, which ones (and what stage) could be released, and in which animal/generation they expect sterility or lethality. This would be useful while interpreting the schematics of the genetic crosses in the rest of the figures (1B, 2B). Li et al 2021 has something to this effect. I say this particularly because in the text, 'pgSIT' is used to refer to both the lab stocks and the F1s.

We really appreciate the suggestion to incorporate a schematic into the introduction to clarify the intended use of these mosquitoes. Taking into account all the suggestions, we would like to keep textual descriptions and context provided within the manuscript, which, together with Figures 1B and 2B, illustrate our intentions. Nevertheless, we value your input and have taken other feedback into account to improve the overall quality of the content.

Because Figure 1A depicts all the gRNAs I thought that's what they were testing in the first results section. But the legends seems to suggest that the individual gRNAs have been tested. Such issues will be sorted with attention to the writing. It would also be nice to have Figure 2A here.

We apologize for any misunderstanding. Figure 1A displays two gRNA constructs: one for dsx (comprising 4 gRNAs) and another for ix (with 2 gRNAs). All of these gRNAs were tested in the initial results section. Subsequently, we engineered the final gRNA construct, denoted as gRNAdsx,ix,βTub, which combines the effective gRNAs described earlier (3 targeting dsx and 1 targeting ix, as illustrated in Supplementary Figure 2).

It wasn't clear to me how egg laying percentages were calculated or what it means.

We appreciate your comment. Female fecundity depends on the egg output (egg laying percentage) and the egg hatching rate, since insect female can lay unfertalized eggs that does not hatch. Egg laying percentages were calculated by dividing the numbers of laid eggs by a test female group by that of the control female group that laid the highest egg number. This procedure is called normalization and enable relative comparison of laid egg number.

How is hatching at times more than laying?

When a female group laid a small egg number but the high percentage of those eggs hatched.

Calling something 'intersex': The authors are assessing intersex by malformed genitalia, maxillary palps and ovaries. But the genitalia defects in Fig1D were not clear to me. Can the authors show better images? While the MP snd ovary phenotypes were clear, it would be nice to see these quantified - what proportion of the females have each/some/all of these phenotypes? It would be nice to see this quantified. (They have some of this in the supplementary table).

We express our gratitude for the comment received and acknowledge the issue regarding the clarity of the images. It is important to note that these photographs represent the highest level of clarity achieved thus far. We value your interest in the quantification of the observed phenotypes. However, due to certain constraints, we were unable to quantify the proportions for all the females, and we did not retain all the samples needed for this specific quantification.

It's interesting that 50% of the intersex don't blood-feed - is this because they do not have appropriately formed stylets? It would be important to quantify the number of hatch-able eggs. This is particularly important in the context of field application and should ideally be included in the mathematical modelling. In the discussion, the authors mention that they are not able to host-seek and a variety of other behaviours - these data should be presented as it would be important for assessing the efficacy of the pgSIT.

Thank you for the comment. We did not find the mutant stylets from these intersex mosquitoes. We agree with the reviewer that the number of hatchable eggs is particularly important in the context of field application. Indeed, the number of hatchable eggs is what was considered in the mathematical modeling. We did a blood feed assay (small cage and big cage) for host seeking behavior. Data were presented in Supplementary Table 5.

At the end of the first results section, the authors state, "Taken together, these findings reveal that ♀-specific lethality and/or ⚥..." But I don't see data that show female-specific lethality until Figure 2C.

Thank you for pointing out this. In order to describe our results clearly, we have deleted “♀-specific lethality and/or”

In the combined gRNA mosquito (the pgSIT), they find that the cross between the gRNA and Cas9 results in very few eggs being laid, high larval death, and what emerges are males. This suggests that it would be a poor pgSIT, right? You'd have to set up huge crosses to get enough males emerging in the wild to mate with WT females to bring about population suppression. Could the authors comment on this?

We appreciate the comment. Even in the presence of imperfections, such as reduced egg production resulting from the gRNA and Cas9 cross and the necessity of extensive mating to obtain an adequate number of males, population suppression is very promising with the pgSIT, both in terms of the potential to eliminate a mosquito population, or to suppress it to an extent that would largely interrupt disease transmission. It's worth noting that our current efforts serve as a validation of the system before its potential large-scale application, because we have demonstrated that removing females by disrupting sex determinate genes is possible with pgSIT, which can inform the development of such systems in other species in the future.

If I'm reading Figure 2C right, the authors have combined the results from two types of crosses in the last two plots: 1) the Cas9 (X) gRNA mosquitoes and 2) the progeny from these crossed to WTs. This is not ideal. I would suggest the authors unpack the text around this data and plot it separately.

We really appreciate the comment here, the panel 2C depicts the phenotypic data of the F1 progeny generated by the cross of the parents indicated below the X axis: egg-to-adult survival, larval death, sex ratios, and fertility. The fertility of F1 progeny is the major phenotypic feature for the project. To assess the fertility of the surviving F1 progeny, we had to cross the F1 females and males to WT males and females, respectively and assess the hatching rate of produced eggs before sacrificing emerged larvae and unhatched eggs. It's important to note that mosquito females can lay unfertilized eggs that fail to hatch.

The text around 2F needs to be more explanatory. There are lots of labels in the figure that are not referred to, making it difficult to follow the data.

We have gone through and expanded many of the figure legends and modified some figures to help make them more understandable.

The supplementary figure numbering is off.

We really appreciate the comment. The supplementary figure numbering have been fixed.

I cannot comment on Figure 4 as this is outside my expertise. However, I do feel that some attention to the writing might help make the approach more accessible to the invested advanced lay-person.

We appreciate the comment, and we re-wrote some of the sentences describing Figure 4.

Reviewer #2 (Recommendations For The Authors):

Line 49 'resistances' is a strange plural.

Corrected. Thank you so much!

the genitive, used with the sex symbols throughout, looks very weird e.eg line 60, 66 etc. Also the intersex symbol, on my copy at least, just prints as a square

These have been fixed in the revised version. Thank you so much!

Line 74 syntax (...: the spread of...") seems off

Corrected. Thank you for pointing out this.

Line 80-81 " to address some of the challenges with gene drives, pgSIT also leverages....." this is a straw man/red herring argument, and simply does not follow. It is this element that I raised above in the public review. See also line 84 'gene drive safety concerns'.

Thank you, we have re-wrote the paragraph.

Line 128 "the induced phenotypes were especially strong in intersex individuals" - this is a curious statement since, if intersex, they are by definition already showing a strongly induced phenotype

We apologize for the lack of clarity and have updated the text, we have deleted “the induced phenotypes were especially strong in intersex individuals”, to be more explicit, now stating “These gRNAdsx/+; Cas9/+ ⚥ exhibited multiple malformed morphological features, such as mutant maxillary palps, abnormal genitalia, and malformed ovaries”

The extent and completeness of the supplementary data is appreciated but there needs to be some statistical tests applied to back up statements like 'showed normal fertility' (line 138) or wind lengths 'were a bit larger'. None seem to have been applied.

We appreciate the comment. We've removed these sentences in the new version.

Supp Fig 4 - on left of panel C there is a small blue square at dsx locus that is unexplained. What is this?

Thank you for pointing this. It was a mistake, we have removed the small blue square from Sup Fig4.

Line 182 the reduction in flight activity in release genotype of pgSIT males - is it only those coming with the maternal source of Cas9 that are plotted (only pink dots)?

We appreciate the comment. pgSIT males, regardless of whether they originate from a maternal or paternal source of Cas9, exhibit a similar reduction in flight activity compared to wild-type (WT) males.

Figure 3A legend - I think there is a typo that says males were fed

Corrected. Thank you for pointing this out.

“♂’s” to “♀’s”

On the window of protection (WOP) plots (e.g. supp fig 12) what is the unit on Y-axis for WOP? It goes from 0-1, as if it were probability, but I was expecting some duration.

Thanks for the comment. The y-axis for WOP in Supp Fig 12 had been normalized unnecessarily. It has now been corrected to span from 0 to 5 years.

Fig 4B blue (line) on blue(shading) is impossible to decipher on my copy

Thank you for pointing this out. We have changed the colors of the traces (population dynamics), made the window of protection line thicker, and have made the shading less opaque to make the population dynamics in this figure clearer.

Line 250 and 252: supp Fig 13 (not 12)

Corrected. Thank you for pointing this out.

Line 279 "potentially a more widespread effect of sex determination genes than previously expected" - I simply don't see how this is so, or why there is the need to make such a claim. Dsx is known to underpin almost of somatic determination of sex-specific morphologies, in a range of insects.

We appreciate the comment. We have delete the sentence:

“Taken together, these observations indicate a potentially more widespread effect of sex determination genes than previously expected, though regardless.”

Line 320 "We would expect pgSIT to be regulated similarly to Oxitec's RIDL" because they are similar, which goes to my main point above about more appropriate context, and this warrants some direct attention to a comparison of the efficacy.

We appreciate the comment. We have delete these sentences:

“We would expect pgSIT to be regulated similarly to Oxitec's RIDL technology (Spinner et al., 2022), which has already been successfully deployed in numerous locations, including the United States.”

Was there a minimal performance advantage with strain #1 with the triple locus g-RNA suite, over the other two strains? Am just curious as to why one was chosen over the other

We appreciate the comment. There was no performance advantage with the strain #1 over the other two strains.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary

In this manuscript, Hagihara et al. characterized the relationship between the changes in lactate and pH and the behavioral phenotypes in different animal models of neuropsychiatric disorders at a large-scale level. The authors have previously reported that increased lactate levels and decreased pH are commonly observed in the brains of five genetic mouse models of schizophrenia (SZ), bipolar disorder (BD), and autism spectrum disorder (ASD). In this study, they expanded the detection range to 109 strains or conditions of animal models, covering neuropsychiatric disorders and neurodegenerative disorders. Through statistical analysis of the first 65 strains/conditions of animal models which were set as exploratory cohort, the authors found that most strains showed decreased pH and increased lactate levels in the brains. There was a significant negative correlation between pH and lactate levels both at the strain/condition level and the individual animal level. Besides, only working memory was negatively correlated with brain lactate levels. These results were successfully duplicated by studying the confirmative cohort, including 44 strains/conditions of animal models. In all strains/conditions, the lactate levels were not correlated with age, sex, or storage duration of brain samples.

Strengths

1. The manuscript is well-written and structured. In particular, the discussion is really nice, covering many potential mechanisms for the altered lactate levels in these disease models.

2. Tremendous efforts were made to recruit a huge number of various animal models, giving the conclusions sufficient power.

We are grateful to Reviewer #1 for the positive evaluation of our manuscript. As indicated in the responses that follow, we have taken all the comments and suggestions made by the reviewer into account in the revised version of our paper.

Weaknesses

1. The biggest concern of this study is the limited novelty. The point of "altered pH and/or lactate levels in the brains from human and rodent animals of neuropsychiatric disorders" has been reported by the same lab and other groups in many previous papers.

The previous study mentioned by the reviewer evaluated a small number of animal models of psychiatric disorders. The novelty of this study is underscored by two key findings: 1) the generality of changes in brain pH and lactate levels across a diverse range of disease models, and 2) the association of these phenomenon with specific behaviors. First, this large-scale animal model study revealed that alterations in brain pH/lactate levels can be found in approximately 30% of the animal models examined. This generality suggests a common basis in the neuropathophysiology of not only schizophrenia, bipolar disorder, and ASD, but also of Alzheimer’s disease (APP-J20 Tg mice), Down’s syndrome (Ts1Cje mice), Mowat–Wilson syndrome (Zeb2 KO mice), Dravet syndrome (Scn1a-A1783V KI mice), tuberous sclerosis complex (Tsc2 KO mice), Ehlers-Danlos syndrome (Tnxb KO mice), and comorbid depression in diabetes (streptozotocin-treated mice) and colitis (dextran sulfate sodium-treated mice). Secondly, this study demonstrated that these phenomenon in the brain are primarily associated with working memory impairment over depression- and anxiety-related behaviors. Importantly, developing these hypotheses in an exploratory cohort of animals and confirming them in an independent cohort within this study enhances the robustness and reliability of our hypotheses, which we believe are equally crucial as their novelty. Accordingly, we have revised the discussion section as follows (page 31, line 7):

Original text

"We performed a large-scale analysis of brain pH and lactate levels in 109 animal models of neuropsychiatric disorders, which revealed the diversity of brain energy metabolism among these animal models. Some strains of mice that were considered models of different diseases showed similar patterns of changes in pH and lactate levels. Specifically, the SZ/ID models (Ppp3r1 KO, Nrgn KO mice, and Hivep2 KO mice), BD/ID model (Camk2a KO mice), ASD model (Chd8 KO mice), depression models (mice exposed to social defeat stress, corticosterone-treated mice, and Sert KO mice), AD model (APP-J20 Tg mice), and DM model (Il18 KO and STZ-treated mice) commonly exhibited decreased brain pH and increased lactate levels."

Revised text

"We performed a large-scale analysis of brain pH and lactate levels in 109 animal models of neuropsychiatric disorders, which revealed the diversity of brain energy metabolism among these animal models. The key findings of this study are as follows: 1) the generality of changes in brain pH and lactate levels across a diverse range of disease models, and 2) the association of these phenomenon with specific behaviors. First, this large-scale animal model study revealed that alterations in brain pH/lactate levels can be found in approximately 30% of the animal models examined. This generality suggests a common basis in the neuropathophysiology of not only schizophrenia, bipolar disorder, and ASD, but also of Alzheimer’s disease (APP-J20 Tg mice), Down’s syndrome (Ts1Cje mice), Mowat–Wilson syndrome (Zeb2 KO mice), Dravet syndrome (Scn1a-A1783V KI mice), tuberous sclerosis complex (Tsc2 KO mice), Ehlers-Danlos syndrome (Tnxb KO mice), and comorbid depression in diabetes (streptozotocin-treated mice) and colitis (dextran sulfate sodium-treated mice). Secondly, this study demonstrated that these phenomenon in the brain are primarily associated with working memory impairment over depression- and anxiety-related behaviors. Importantly, developing these hypotheses in an exploratory cohort of animals and confirming them in an independent cohort within this study enhances the robustness and reliability of our hypotheses."

1. This study is mostly descriptive, lacking functional investigations. Although a larger cohort of animal models were studied which makes the conclusion more solid, limited conceptual advance is contributed to the relevant field, as we are still not clear about what the altered levels of pH and lactate mean for the pathogenesis of neuropsychiatric disorders.

We agree with the reviewer’s comment. To address this issue, it is necessary to comprehensively identify brain regions and cell types responsible for pH and lactate changes in each strain/condition of animals, as these may differ among them. Subsequently, based on such findings, we can then proceed with functional investigations that specifically target the identified brain regions/cell types. However, conducting such investigations would require a significant amount of time to complete, approximately 2–3 years, and is beyond the scope of this study. Therefore, we would like to conduct such studies in the future. We have mentioned this limitation by revising the discussion section of this study as follows (page 43, line 5):

Original text

"Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). The brain region-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Further studies are needed to address this issue by measuring microdissected brain samples and performing in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011) and MRS (Davidovic et al., 2011)."

Revised text:

"The major limitations of this study include the absence of analyses specific to brain regions or cell types and the lack of functional investigations. Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. It is known that certain molecular expression profiles and signaling pathways display brain region-specific alterations, and in some cases, even exhibit opposing changes in neuropsychiatric disease models (Hosp et al., 2017; Floriou-Servou et al. 2018; Reim et al., 2017). Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). Additionally, it has been reported that the basal intracellular pH differs between neurons and astrocytes (lower in astrocytes than in neurons), and their responsiveness to conditions simulating neural hyperexcitation and the metabolic acidosis in terms of intracellular pH also varies (Raimondo et al., 2016; Salameh et al., 2017). It would also be possible that the brain region/cell type-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Given the assumption that the brain regions and cell types responsible for pH and lactate changes vary across different strains/conditions, comprehensive studies are needed to thoroughly examine this issue for each animal model individually. This can be achieved through techniques such as evaluating microdissected brain samples, conducting in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011), and MRS (Davidovic et al., 2011). Subsequently, based on such findings, it is also necessary to conduct functional analyses for each model animal by manipulating pH or lactate levels in specific brain regions/cell types and evaluating behavioral phenotypes relevant to neuropsychiatric disorders."

1. The experiment procedure is also a concern. The brains from animal models were acutely collected without cardiac perfusion in this study, which suggests that resident blood may contaminate the brain samples. The lactate is enriched in the blood, making it a potential confounded factor to affect the lactate levels as well as pH in the brain samples.

We thank the reviewer for pointing this out. We have discussed this issue as follows (page 45, line 4):

We also note that there are several potential confounding factors in this study. The brain samples analyzed in this study contained cerebral blood. The cerebral blood volume is estimated to be approximately 20–50 μl/g in human and feline brains (Leenders et al., 1990; van Zijl et al., 1998). When we extrapolate these values to murine brains, it would imply that the proportion of blood contamination in the brain homogenates analyzed is 0.2–0.6%. Additionally, lactate concentrations in the blood are two to three times higher than those in the brains of mice (Béland-Millar et al., 2017). Therefore, even if there were differences in the amount of resident blood in the brains between control and experimental animals, the impact of such differences on the lactate measurements would likely be minimal.

1. The lactate and pH levels may also be affected by other confounded factors, such as circadian period, and locomotor activity before the mice were sacrificed. This should also be discussed in the paper.

Following the reviewer’s suggestion, we have discussed the matter as follows (page 45, line 12): Other confounding factors include circadian variation and locomotor activity before the brain sampling. Lactate levels are known to exhibit circadian rhythm in the rodent cortex, transitioning gradually from lower levels during the light period to higher levels during the dark period (Dash et al., 2012; Shram et al., 2002; Wallace et al., 2022). The variation in the times of sample collection during the day was basically kept minimized within each strain/condition of animals. However, the sample collection times were not explicitly matched across the different laboratories, which may contribute to variations in the baseline control levels of pH and lactate among different strains/conditions of animals (Table S3). In addition, motor activity and wake/sleep status immediately before brain sampling can also influence brain lactate levels (Neylor et al., 2012; Shram et al., 2002). These factors have the potential to act as confounding variables in the measurement of brain lactate and pH in animals.

1. Another concern is the animal models. Although previous studies have demonstrated that dysfunctions of these genes could cause related phenotypes for certain disorders, many of them are not acknowledged by the field as reliable disease models. Besides, gene deficiency could also cause many known or unknown unrelated phenotypes, which may contribute to the altered levels of lactate and pH, too. In this circumstance, the conclusion "pH and lactate levels are transdiagnostic endophenotype of neuropsychiatric disorders" is somewhat overstated.

We thank the reviewer for pointing this out. We should have taken this issue into consideration. Accordingly, we have discussed this issue as the limitation of this study in the discussion section as follows (page 34, line 14):

"While we analyzed 109 strains/conditions of animals, we included both those that are widely recognized as animal models for specific neuropsychiatric disorders and those that are not. For example, while interleukin 18 (Il18) KO mice and mitofusin 2 (hMfn2-D210V) Tg mice exhibited changes in pH and lactate levels, the evidence that these genes are associated with specific neuropsychiatric disorders is limited. However, these strains of mice exhibited behavioral abnormalities related to neuropsychiatric disorders, such as depressive-like behaviors and impaired working memory (Ishikawa et al., 2019, 2021; Yamanishi et al., 2019). Furthermore, these mice showed maturation abnormality in the hippocampal dentate gyrus and neuronal degeneration due to mitochondrial dysfunction, respectively, suggesting conceptual validity for utilization as animal models for neuropsychiatric and neurodegenerative disorders (Cunnane, et al., 2021; Burté et al., 2015; Hagihara et al., 2013, 2019). In contrast, mice with heterozygous KO of the synaptic Ras GTPase-activating protein 1 (syngap1), whose mutations have been identified in human patients with ID and ASD, showed an array of behavioral abnormalities relevant to the disorders (Komiyama et al., 2002; Nakajima et al., 2019), but did not show changes in brain pH or lactate levels. Therefore, while changes in brain pH and lactate levels could be transdiagnostic endophenotypes of neuropsychiatric disorders, they might occur depending on the subpopulation due to the distinct genetic and environmental causes or specific disease states in certain disorders."

Regarding the latter point suggested by the reviewer, we consider that alterations in brain pH and lactate levels occur, whether they are a direct and known consequence or indirect and unknown ones of genetic modifications. We have proposed that genetic modifications, along with environmental stimulations, may induce various changes, which subsequently converge toward specific endophenotypes in the brain, such as neuronal hyperexcitation, inflammation, and maturational abnormalities (Hagihara et al., 2013; Yamasaki et al., 2008). The findings of this study, demonstrating the commonality of alteration of brain pH and lactate levels, align with this concept, suggesting that these alterations could serve as brain endophenotypes in multiple neuropsychiatric disorders. We have revised the discussion section as follows (page 42, line 8):

Original text

"These findings suggest that the observed increase in lactate production and subsequent decrease in pH in whole-brain samples may be attributed to the hyperactivity of specific neural circuits in a subset of the examined animal models."

Revised text

"These findings suggest that neuronal hyperexcitation may be one of the common factors leading to increased lactate production and decreased pH in the brain. We consider that alterations in brain pH and lactate levels occur, whether they are a direct and known consequence or indirect and unknown ones of genetic modifications. We have proposed that genetic modifications, along with environmental stimulations, may induce various changes, which subsequently converge toward specific endophenotypes in the brain, such as neuronal hyperexcitation, inflammation, and maturational abnormalities (Hagihara et al., 2013; Yamasaki et al., 2008). The findings of this study, demonstrating the commonality of alterations in brain pH and lactate levels, align with this concept and suggest that these alterations could serve as brain endophenotypes in multiple neuropsychiatric disorders."

1. The negative correlationship between pH and lactate is rather convincing. However, how much the contribution of lactate to pH is not tested. In addition, regarding pH and lactate, which factor contributes most to the pathogenesis of neuropsychiatric disorders is also unclear. These questions may need to be addressed in the future study.

To estimate the degree of contribution of lactate to pH, we determined the contribution ratio using the regression coefficient within a linear regression model applied to a combined cohort. The results showed that 33.2% of changes in pH may be explained by changes in lactate level. We have added the following text in the Results section (page 28, line 7).

The contribution ratio of lactate to pH, calculated based on the regression coefficient in a linear regression model, was 33.2% at the individual level, suggesting a moderate level of contribution.

Regarding the latter suggestion, we would like to address the issue in the future study. Accordingly, we have added the following sentence in the discussion section (page 40, line 11):

Original text

"Further studies are needed to address these hypotheses by chronically inducing deficits in mitochondrial function to manipulate endogenous lactate levels in a brain region-specific manner and to analyze their effects on working memory."

Revised text

"Further studies are needed to address these hypotheses by chronically inducing deficits in mitochondrial function to manipulate endogenous lactate levels in a brain region-specific manner and to analyze their effects on working memory. It is also important to consider whether pH or lactate contributes more significantly to the observed behavioral abnormalities."

1. The authorship is open to question. Most authors listed in this paper may only provide mice strains or brain samples. Maybe it is better just to acknowledge them in the acknowledgments section.

In the light of the current circumstances, wherein there is no universally agreed definition of authorship (the Committee on Publication Ethics1), we acknowledge the reviewer’s concern. Collecting a comprehensive range of mouse strains and brain samples is a fundamental principle of this study. Maintaining mouse lines, breeding mice, genotyping, drug administration, and preparation of brain samples each require specialized expertise. Therefore, the scientific and technical contributions of individuals who only provided mouse strains or brain samples was also crucial for obtaining the data essential to this study. In accordance with the authorship guidelines outlined by the journal, which stipulate that “We recommend that all researchers who made substantial or important contributions to the design of a work, or the acquisition, analysis or interpretation of the data used in the paper, be included as authors.”, we would like to retain their authorship status. Furthermore, we ensured that all authors had read and approved the manuscript before submission, using Google Forms.

1. GUIDELINES ON GOOD PUBLICATION PRACTICE, Committee on Publication Ethics (COPE), https://publicationethics.org/files/u7141/1999pdf13.pdf

1. The last concern is about the significance of this study. Although the majority of strains showed increased lactate, some still showed decreased lactate levels in the brains. These results suggested that lactate or pH is an endophenotype for neuropsychiatric disorders, but it is hard to serve as a good diagnostic index as the change is not unidirectional in different disorders. In other words, the relationship between lactate level and neuropsychiatric disorders is not exclusive.

As pointed out by the reviewer, whether brain pH and lactate levels increase or decrease could vary among animal models. Such variation may represent subpopulations of patients or specific disease states. Considering both increases and decreases in changes in pH and lactate levels could be important to achieve that goal. Accordingly, we have revised the text as follows:

Added text (page 33, line 12)

"Detecting changes in brain pH and lactate levels, whether resulting in an increase or decrease due to their potential bidirectional alterations, using techniques such as MRS may help the diagnosis, subcategorization, and identification of specific disease states of these biologically heterogeneous and spectrum disorders, as has been shown for mitochondrial diseases (Lin et al., 2003)."

Added text (page 35, line 14)

"Therefore, while changes in brain pH and lactate levels could be transdiagnostic endophenotypes of neuropsychiatric disorders, they might occur depending on the subpopulation due to the distinct genetic and environmental causes or specific disease states in certain disorders."

Reviewer #2 (Public Review):

Hagihara et al. conducted a study investigating the correlation between decreased brain pH, increased brain lactate, and poor working memory. They found altered brain pH and lactate levels in animal models of neuropsychiatric and neurodegenerative disorders. Their study suggests that poor working memory performance may predict higher brain lactate levels.

However, the study has some significant limitations. One major concern is that the authors examined whole-brain pH and lactate levels, which might not fully represent the complexity of disease states. Different brain regions and cell types may have distinct protein and metabolite profiles, leading to diverse disease outcomes. For instance, certain brain regions like the hippocampus and nucleus accumbens exhibit opposite protein/signaling pathways in neuropsychiatric disease models.

We want to thank the reviewer for the valuable suggestions. To address this issue, it is necessary to comprehensively identify brain regions and cell types responsible for pH and lactate changes in each strain/condition of animals, as these may differ among them. Subsequently, based on such findings, we can then proceed with functional investigations that specifically target the identified brain regions/cell types. However, conducting such investigations would require a significant amount of time to complete, approximately 2–3 years, and is beyond the scope of this study. Therefore, we would like to conduct such studies in the future. We have mentioned this limitation by revising the discussion section of this study as follows (page 43, line 5):

Original text

"Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). The brain region-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Further studies are needed to address this issue by measuring microdissected brain samples and performing in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011) and MRS (Davidovic et al., 2011)."

Revised text

"The major limitations of this study include the absence of analyses specific to brain regions or cell types and the lack of functional investigations. Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. It is known that certain molecular expression profiles and signaling pathways display brain region-specific alterations, and in some cases, even exhibit opposing changes in neuropsychiatric disease models (Hosp et al., 2017; Floriou-Servou et al. 2018; Reim et al., 2017). Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). Additionally, it has been reported that the basal intracellular pH differs between neurons and astrocytes (lower in astrocytes than in neurons), and their responsiveness to conditions simulating neural hyperexcitation and the metabolic acidosis in terms of intracellular pH also varies (Raimondo et al., 2016; Salameh et al., 2017). It would also be possible that the brain region/cell type-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Given the assumption that the brain regions and cell types responsible for pH and lactate changes vary across different strains/conditions, comprehensive studies are needed to thoroughly examine this issue for each animal model individually. This can be achieved through techniques such as evaluating microdissected brain samples, conducting in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011), and MRS (Davidovic et al., 2011). Subsequently, based on such findings, it is also necessary to conduct functional analyses for each model animal by manipulating pH or lactate levels in specific brain regions/cell types and evaluating behavioral phenotypes relevant to neuropsychiatric disorders."

Moreover, the memory tests used in the study are specific to certain brain regions, but the authors did not measure lactate levels in those regions. Without making lactate measurements in brain-regions and cell types involved in these diseases, any conclusions regarding the role of lactate in CNS diseases is premature.

Regarding the point about “lactate measurements in brain-regions and cell types involved in these diseases,” please refer our responses provided above.

Additionally, evidence suggests that exogenous treatment with lactate has positive effects, such as antidepressant effects in multiple disease models (Carrard et al., 2018, Carrard et al., 2021, Karnib et al., 2019, Shaif et al., 2018). It also promotes learning, memory formation, neurogenesis, and synaptic plasticity (Suzuki et al., 2011, Yang et al., 2014, Weitian et al., 2015, Dong et al., 2017, El Hayek et al. 2019, Wang et al., 2019, Lu et al., 2019, Lev-Vachnish et a.l, 2019, Descalzi G et al., 2019, Herrera-López et al., 2020, Ikeda et al., 2021, Zhou et al., 2021,Roumes et al., 2021, Frame et al., 2023, Akter et al., 2023).

We thank the reviewer for pointing out many references regarding the effects of lactate that were not cited in our paper. We have since included these studies and discussed in more detail the effect of lactate at molecular, cellular, and behavioral levels (page 39, line 11).

Original text

"Moreover, increased lactate may have a positive or beneficial effect on memory function to compensate for its impairment, as lactate administration with an associated increase in brain lactate levels attenuates cognitive deficits in human patients (Bisri et al., 2016) and rodent models (Rice et al., 2002) of traumatic brain injury. In addition, lactate administration exerts antidepressant effects in a mouse model of depression (Carrard et al., 2016)."

Revised text

"Moreover, increased lactate may have a positive or beneficial effect on memory function to compensate for its impairment, as lactate administration with an associated increase in brain lactate levels attenuates cognitive deficits in human patients (Bisri et al., 2016) and rodent models (Rice et al., 2002) of traumatic brain injury. In addition, lactate administration exerts antidepressant effects in a mouse model of depression (Carrard et al., 2021, 2016; Karnib et al., 2019; Shaif et al., 2018). Lactate has also shown to promote learning and memory (Descalzi G et al., 2019; Dong et al., 2017; Hayek et al. 2019; Lu et al., 2019; Roumes et al., 2021; Suzuki et al., 2011), synaptic plasticity (Herrera-López et al., 2020; Yang et al., 2014; Zhou et al., 2021), adult hippocampal neurogenesis (Lev-Vachnish et al., 2019), and mitochondrial biogenesis and antioxidant defense (Akter et al., 2023), while its effects on adult hippocampal neurogenesis and learning and memory are controversial (Ikeda et al., 2021; Lev-Vachnish et al., 2019; Wang et al., 2019)."

In conclusion, the relevance of total brain pH and lactate levels as indicators of the observed correlations is controversial, and evidence points towards lactate having more positive rather than negative effects. It is important that the authors perform studies looking at brain-region-specific concentrations of lactate and that they modulate lactate levels (decrease) in animal models of disease to validate their conclusions. it is also important to consider the above-mentioned studies before concluding that "altered brain pH and lactate levels are rather involved in the underlying pathophysiology of some patients with neuropsychiatric disorders" and that "lactate can serve as a potential therapeutic target for neuropsychiatric disorders".

Regarding the points about positive effects of lactate, measurement of brain-region-specific lactate concentrations, and modulation of lactate levels, please refer to our responses provided earlier. The points raised by the reviewer are important and should be addressed in future studies.

Reviewer #2 (Recommendations For The Authors):

• Measure lactate in specific brain regions. The whole brain measurements are not relevant to the disease states.

We thank the reviewer for pointing this out. We totally agree with the reviewer’s comment and recognize that the lack of investigations in specific brain regions is one of the major limitations of this study. To address this issue, it is necessary to comprehensively identify brain regions and cell types responsible for pH and lactate changes in each strain/condition of animals, as these may differ among them. Subsequently, based on such findings, we can then proceed with functional investigations that specifically target the identified brain regions/cell types. However, conducting such investigations would require a significant amount of time to complete, approximately 2–3 years, and is beyond the scope of this study. Therefore, we would like to conduct such studies in the future. We have mentioned this limitation by revising the discussion section of this study as follows (page 43, line 5):

Original text

"Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). The brain region-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Further studies are needed to address this issue by measuring microdissected brain samples and performing in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011) and MRS (Davidovic et al., 2011)."

Revised text:

"The major limitations of this study include the absence of analyses specific to brain regions or cell types and the lack of functional investigations. Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. It is known that certain molecular expression profiles and signaling pathways display brain region-specific alterations, and in some cases, even exhibit opposing changes in neuropsychiatric disease models (Hosp et al., 2017; Floriou-Servou et al. 2018; Reim et al., 2017). Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). Additionally, it has been reported that the basal intracellular pH differs between neurons and astrocytes (lower in astrocytes than in neurons), and their responsiveness to conditions simulating neural hyperexcitation and the metabolic acidosis in terms of intracellular pH also varies (Raimondo et al., 2016; Salameh et al., 2017). It would also be possible that the brain region/cell type-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Given the assumption that the brain regions and cell types responsible for pH and lactate changes vary across different strains/conditions, comprehensive studies are needed to thoroughly examine this issue for each animal model individually. This can be achieved through techniques such as evaluating microdissected brain samples, conducting in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011), and MRS (Davidovic et al., 2011). Subsequently, based on such findings, it is also necessary to conduct functional analyses for each model animal by manipulating pH or lactate levels in specific brain regions/cell types and evaluating behavioral phenotypes relevant to neuropsychiatric disorders."

• Discuss in detail the studies that show the neuroprotective effects of lactate and reconcile these with the authors' conclusions.

As suggested by the reviewer, we have discussed in more detail the positive effect of lactate at molecular, cellular, and behavioral levels as below (page 39, line 11):

Original text

"Moreover, increased lactate may have a positive or beneficial effect on memory function to compensate for its impairment, as lactate administration with an associated increase in brain lactate levels attenuates cognitive deficits in human patients (Bisri et al., 2016) and rodent models (Rice et al., 2002) of traumatic brain injury. In addition, lactate administration exerts antidepressant effects in a mouse model of depression (Carrard et al., 2016)."

Revised text

"Moreover, increased lactate may have a positive or beneficial effect on memory function to compensate for its impairment, as lactate administration with an associated increase in brain lactate levels attenuates cognitive deficits in human patients (Bisri et al., 2016) and rodent models (Rice et al., 2002) of traumatic brain injury. In addition, lactate administration exerts antidepressant effects in a mouse model of depression (Carrard et al., 2021, 2016; Karnib et al., 2019; Shaif et al., 2018). Lactate has also shown to promote learning and memory (Descalzi G et al., 2019; Dong et al., 2017; Hayek et al. 2019; Lu et al., 2019; Roumes et al., 2021; Suzuki et al., 2011), synaptic plasticity (Herrera-López et al., 2020; Yang et al., 2014; Zhou et al., 2021), adult hippocampal neurogenesis (Lev-Vachnish et al., 2019), and mitochondrial biogenesis and antioxidant defense (Akter et al., 2023), while its effects on adult hippocampal neurogenesis and learning and memory are controversial (Ikeda et al., 2021; Lev-Vachnish et al., 2019; Wang et al., 2019)."

• Conduct experiments whereby you decrease/deplete/modulate lactate levels in animal models and show that there is amelioration of the symptoms.

Regarding this point, kindly refer to the responses we provided in the first comment from the reviewer. We have mentioned this limitation by revising the discussion section of this study as follows (page 43, line 5):

Original text

"Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). The brain region-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Further studies are needed to address this issue by measuring microdissected brain samples and performing in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011) and MRS (Davidovic et al., 2011)."

Revised text:

"The major limitations of this study include the absence of analyses specific to brain regions or cell types and the lack of functional investigations. Because we used whole brain samples to measure pH and lactate levels, we could not determine whether the observed changes in pH and/or lactate levels occurred ubiquitously throughout the brain or selectively in specific brain region(s) in each strain/condition of the models. It is known that certain molecular expression profiles and signaling pathways display brain region-specific alterations, and in some cases, even exhibit opposing changes in neuropsychiatric disease models (Hosp et al., 2017; Floriou-Servou et al. 2018; Reim et al., 2017). Indeed, brain region-specific increases in lactate levels were observed in human patients with ASD in an MRS study (Goh et al., 2014). Furthermore, while increased lactate levels were observed in whole-brain measurements in mice with chronic social defeat stress (Figure S7) (Hagihara et al., 2021a), decreased lactate levels were found in the dorsomedial prefrontal cortex (Yao et al., 2023). Additionally, it has been reported that the basal intracellular pH differs between neurons and astrocytes (lower in astrocytes than in neurons), and their responsiveness to conditions simulating neural hyperexcitation and the metabolic acidosis in terms of intracellular pH also varies (Raimondo et al., 2016; Salameh et al., 2017). It would also be possible that the brain region/cell type-specific changes may occur even in animal models in which undetectable changes were observed in the present study. This could be due to the masking of such changes in the analysis when using whole-brain samples. Given the assumption that the brain regions and cell types responsible for pH and lactate changes vary across different strains/conditions, comprehensive studies are needed to thoroughly examine this issue for each animal model individually. This can be achieved through techniques such as evaluating microdissected brain samples, conducting in vivo analyses using pH- or lactate-sensitive biosensor electrodes (Marunaka et al., 2014; Newman et al., 2011), and MRS (Davidovic et al., 2011). Subsequently, based on such findings, it is also necessary to conduct functional analyses for each model animal by manipulating pH or lactate levels in specific brain regions/cell types and evaluating behavioral phenotypes relevant to neuropsychiatric disorders."

Other corrections

Title page and Acknowledgements:

We have revised the affiliation information for the following co-authors: Drs. Anja Urbach8, Mohamed Darwish19, 20, Keizo Takao20, 22, Bong-Kiun Kaang53, 54, Michihiro Igarashi74, 75, Rie Ohashi87-89, and Nobuyuki Shiina87-89.

Page 56, line 12:

The term ‘The International Brain pH Consortium’ has been corrected to ‘The International Brain pH Project Consortium’.

Supplementary Table 1: Supplementary References:

1. Oota-Ishigaki A, Takao K, Yamada D, Sekiguchi M, Itoh M, Koshidata Y, et al. (2022): Prolonged contextual fear memory in AMPA receptor palmitoylation-deficient mice. Neuropsychopharmacology 47: 2150–2159.

We have updated the name of the mouse strain from “patDp” to “15q dup” throughout the manuscript.

We have made the following revisions to enhance readability.

Page 24, line 9: According to a simple correlation analysis, working memory measures (correct responses in the maze test) were significantly negatively correlated with brain lactate levels (r = -0.76, P = 1.93 × 10-5; Figure 1F).

Page 27, line 1:

Revised text

"We found that working memory measures (correct responses in the maze test) were the most frequently selected behavioral measures for constructing a successful prediction model (Figure 2E), which is consistent with the results of the exploratory study (Figure 1E)."

Figure 1 legend:

Revised text

"(F–H) Scatter plot showing correlations between actual brain lactate levels and measures of working memory (correct responses in the maze test) (F), the number of transitions in the light/dark transition test (G), and the percentage of immobility in the forced swim test (H)."

Figure 2 legend:

Revised text

"(F–H) Scatter plots showing correlations between actual brain lactate levels and working memory measures (correct responses in the maze test) (F), the acoustic startle response at 120 dB (G), and the time spent in dark room in the light/dark transition test (H)."

Page 30, line 2:

Original text

"The high to moderate-high pH/low to moderate-low lactate group included mouse models of ASD or developmental delay, such as Shank2 KO, Fmr1 KO, BTBR, Stxbp1 KO, Dyrk1 KO, Auts2 KO, and patDp mice (Table S1, Figure S7)."

Revised text

"The high pH/low lactate group and moderate-high pH/moderate-low lactate group included mouse models of ASD or developmental delay, such as Shank2 KO, Fmr1 KO, BTBR, Stxbp1 KO, Dyrk1 KO, Auts2 KO, and 15q dup mice (Table S1, Figure S7)."

Page 40, line 7:

Original text

"Moreover, increased lactate levels may also be involved in behavioral changes other than memory deficits such as anxiety."

Revised text

"Moreover, increased lactate levels may also be involved in behavioral changes other than memory deficits, such as anxiety."

Author Response

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

Reviewer #1 (Public Review):

The experimental design presented cannot clearly show that the effect of passive exposure was due to the specific exposure to task-relevant stimuli since there is no control group exposed to irrelevant stimuli.

We acknowledge the possibility that exposure to task-irrelevant stimuli could result in improvements in learning. Testing this possibility would be a worthwhile goal of future experiments, but it is outside the scope of our current study. We have been careful in our paper to only draw conclusions about the effects of exposure to task-relevant stimuli compared to no exposure. We have added a discussion of this point and relevant references to the literature in the Discussion section of our manuscript.

The conclusion that "passive exposure influences responses to sounds not used during training" (line 147) does not seem fully supported by the authors' analysis. The authors show that there is an increase in accuracy for intermediate sweep speeds despite the fact that this is the first time the animals encounter them in the active session. However, it seems impossible to exclude that this effect is not simply due to the increased accuracy of the extreme sounds that the animals had been trained on.

We have modified this sentence to emphasize that it refers to “intermediate” sounds. Regarding the reviewer’s concern, the conclusion is drawn from Figure 3, in which we show that mice exhibit an improvement on non-extreme stimuli after training on extreme stimuli. Panel 3D illustrates that the observed improvements are not just changes in psychometric performance driven by the extreme sounds. In the context of this result, the conclusion relates to generalization in performance on task-relevant stimuli that are closely related to the training stimuli. In our view, it was not entirely obvious a priori that this result would have to occur, since it is possible that performance could improve at the extremes without improving at the intermediate stimuli.

In the modelling section, the authors adjusted the hyper-parameters to maximize the difference between pure active and passive/active learning. This makes a comparison of learning rates between models somewhat confusing.

We apologize for the confusion. None of our conclusions are based on comparisons of learning speed between models, but perhaps this was not pointed out sufficiently clearly. The relevant comparisons between conditions for each specific model are made using the same hyperparameters. We have clarified this point in the modeling section of our manuscript.

The description of the sound does not state whether when reducing the slope of the sweeps the center or the onset frequency of the sounds is preserved.

Frequency modulated sounds of different FM slopes were generated such that the center frequency was always the same. This is now clarified in the updated version of the manuscript.

Reviewer #1 (Recommendations for the authors):

As mentioned, the specificity of the stimuli presented during the passive period is not explicitly addressed in either modelling or behaviour. For modelling, this could be quite straightforward to assess by manipulating the input stimuli during passive episodes. For the behaviour, this would require repeating the experiment with passive sessions during which unrelated sounds are presented (for example varying in frequency or intensity instead of frequency slope). I mainly include this suggestion to clarify my previous comment because this would require a huge amount of work.

We agree that varying the extent to which the presented passive stimuli are task-related to the task is an interesting point to study for future experiments. However, doing so for the experiments is outside the scope of the current study, and we believe exploring this only in the modeling part would add little value to the current study, because the outcome will highly depend on the details of the implementation.

Reviewer #2 (Public Review):

One limitation here is that the presented analysis is somewhat simplistic, does not include any detailed psychometric analysis (bias, lapse rates etc), and primarily focuses on learning speed.

In our preliminary analyses of trials that included extreme and intermediate stimuli after animals had learned the task (Figure 3), we investigated some metrics of the type that the reviewer suggests here. However, since such additional psychometric analyses were somewhat tangential to our main results (which are about learning speed and responses to sounds not included during training), we did not include these in our manuscript. In agreement with the reviewer’s concern, a main limitation of our study is that the available data does not allow for an analysis of psychometrics during the initial learning stages, since only the extreme stimuli were presented during the task.

Reviewer #2 (Recommendations for the authors):

The International Brain Lab has shown quite nicely that psychometric curves continue to improve (increased slope, decreased bias) across learning. This was not really discussed or presented in your data - is this observed during the S4 training portion?

We indeed saw improvements in the psychometric performance during stage S4, in particular for the active-only learners, as can be seen in Figure 3. We quantified these changes (now presented in the Results section), and added a discussion to the main text.

Why use a linear fit to extract the various quantities of interest? All of these quantities could be extracted from the raw behavioral data itself.

Because of the large variations in performance from day-to-day, a linear fit allowed us to extract a more reliable estimate of quantities like “Time to achieve 70%” and “Performance at 21 days” for each animal.

The analysis presented was focussed primarily on the fast learners. What about the slow learners? Are the ANN models able to recapitulate different aspects of their behavior?

We agree with the reviewer that the observation that the learners clustered into two groups calls for further investigation. In this study, we focused on the mice that learned more efficiently, because those allowed us to address our main research question about the influence of passive exposure. We believe, the slow learners could be modeled with ANNs that start with a less-easily discriminable input representation, which limits the performance that the trained network is ultimately able to achieve. This additional analysis is outside the scope of the current manuscript, but we hope to address these questions in the future.

Although I appreciate the thoroughness of the modeling, I was not entirely convinced by the narrative underlying models 1-5, since none of these models were able to successfully recapitulate your core findings. Would it not make more sense to focus primarily on the final model?

By starting with the simplest possible model that incorporates supervised and unsupervised learning, we were able to determine which ingredients were necessary to capture the behavioral data. We believe this could not have been clearly established by considering the final model alone.

Reviewer #3 (Public Review):

The first [major weakness] is that even Model 5 differs from their data. For example, the A+P (passive interleaved condition) learning curve in Figure 7 seems to be non-monotonic, and has some sort of complex eigenvalue in its decay to the steady state performance as trials increase. This wasn't present in their experimental data (Figure 2D), and implies a subtle but important difference. There also appear to be differences in how quickly the initial learning (during early trials) occurs for the A+P and A:P conditions. While both A+P and A:P conditions learn faster than A only in M5, A+P and A:P seem to learn in different ways, which isn't supported in their data.

The reviewer is correct that there are subtle differences between the two learning curves produced by Model 5. Due to expected variability in the experimental data, however, it is difficult to conclude whether such subtle distinctions also appear in the learning curves of the mice. Further, the slight overshoot of the learning curve that the reviewer mentions is not constrained by the experimental data due to different mice reaching asymptotic performance at different times, and many of them not having even reached asymptotic performance by the end of the training period.

However, even if there are minor discrepancies between the learning curves produced by the final version of the model and by the mice, we do not see this as being especially surprising or problematic. As in any model, there are a large number of potentially important features that are not included in any of our models–for example, realistic spectrotemporal neural responses, nonlinearity in neural activations, heterogeneity across mice, and many others. The aim of our modeling was to choose a space of possible models (which is inevitably restricted) and show which model version within that space best captures our experimental observations. Expanding the space of possible models that we considered to capture further nuances in the data will be a task for future work.

The second major weakness is that the authors also don't generate any predictions with M5. Can they test this model of learning somehow in follow-up behavioural experiments in mice? ... Without follow-up experiments to test their mechanism of why passive exposure helps in a schedule-independent way, the impact of this paper will be limited.

Although testing predictions from our models was beyond the scope of the current study, we do generate specific predictions with model M5 (in particular, about neural representations). Our model produces predictions about neural representations and the ways in which they evolve through learning, and we hope to test these predictions in future work.

I believe the authors need to place this work in the context of a large amount of existing literature on passive (unsupervised) and active (supervised) learning interactions. This field is broad both experimentally and computationally. For example, there is an entire sub-field of machine learning, called semi-supervised learning that is not mentioned at all in this work.

We thank the reviewer for pointing this out. The Discussion section of the updated manuscript now includes a discussion on how our results fit in with this literature.

Reviewer #3 (Recommendations for the authors):

All points made by the reviewer in their Recommendations For The Authors are associated with those presented in the Public Review and they are addressed in our response above.

Author Response

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

eLife assessment

This is a valuable study of Eph-Ephrin signaling mechanisms generating pathological changes in amyotropic lateral sclerosis. There are exciting findings bearing on the role of glial cells in this pathology. The study emerges with solid evidence for a novel astrocyte-mediated mechanism for disease propagation. It may help identify potential therapeutic targets.

Response to Editor’s decision letter: Drs. Huang and Zaidi: Thank you for considering this re-revision of our manuscript for potential publication in eLife. We have addressed the remaining comments of reviewer #2. We have included detailed response-to-reviewer comments below to address each of these remaining specific points from reviewer #2, and we have highlighted all the changes in the manuscript text (using a red font color) made in response to these comments. Based on the reviewers’ critiques, we feel our re-working of the manuscript has made for a greatly improved study.

Reviewer #1 (Recommendations For The Authors):

Reviewer comment: All questions/concerns have been addressed.

Response: We thank Reviewer #1 for the previous helpful comments that we used to improve our manuscript. As Reviewer #1 has no new comments, we have provided no additional responses to address this reviewer’s input. Instead, we only focus (in this new “Response to Reviewer Comments” document) on the remaining points from Reviewer #2 below.

Reviewer #2 (Recommendations For The Authors):

Overall, the authors have addressed most concerns raised in the prior review. A couple of very minor points remain, which would improve the clarity of the report.

Reviewer comment 1: The abstract has not been edited and still emphasizes that astrocyte-mediated upregulation in ephrinB2 signaling underlies pathogenicity in mutant SOD1-associated ALS. There is certainly sufficient evidence to suggest a large role for astrocytes, however, without a thorough investigation of other key cell types in the spinal cord, this cannot be concluded specifically. Especially given that a non-specific promoter (U6) was employed in the viral constructs.

Response: We apoplogize for this mistake. In response to the reviewer’s previous comment in the first round of review, we made changes throughout the manuscript to address this issue; however, we failed to do this in the Abstract. In this re-revised manucript, we now also make the necessary changes to the Abstract.

Reviewer comment 2: It is interesting to note that a non-specific promoter, U6, exhibited such large specificity to astrocytes in the cord as compared to neurons (Fig 2M). This is worth discussing briefly in the discussion and how this result compares to those in the literature.

Response: We have now added a brief discussion of this issue to the Discussion section, including describing our previous studies that used the Gfa2 promotor to achieve astrocyte-specific transduction when employing viral vectors in the rodent spinal cord.

Reviewer comment 3: I appreciate the authors including a supplemental figure on the expression of ephrinA4 receptors in the cervical ventral horn. Unfortunately, the quality of this image is very poor in conveying the receptor expression. The detailed discussion point on the expression of EphB receptors in the cervical ventral horn should be sufficient for readers to take into consideration.

Response: We have now removed this supplemental figure and keep only the text from the rerevised manuscript.

Reviewer comment 4: A few instances of motor neuron diameter being attributed to a 200μm2 size remain (e.g. pg 14).

Response: We have corrected this issue throughout the re-revised manuscript. The correct information is: somal diameter greater than 20 μm.

Reviewer comment 5: It is still a little unclear in the result text as to when assessment of lentiviral transduction was conducted following intraspinal injections.

Response: We have now added this detail about the time point of assessing transduction to both the Results section and the Materials/Methods section.

Reviewer comment 6: Some figures are missing markers of significance (e.g. Fig 2M).

Response: Below are our comments about significance markers for each graph in all figures.

Figure 1:

Panel E: We have now added asterisks for any statistically-significant comparisons. In addition, we provide the details of this statistical analysis in the text of the re-revised manuscript.

Figure 2:

Panel M: We have now added asterisks for statistical comparisons, as well as details in the text.

Panel N: The asterisk was already shown in the previous version of the figure.

Figure 3:

Panels B and G: The asterisks were already shown in the previous version of the figure.

Figure 4:

All panels: There are no significant differences; therefore, no asterisks are needed.

Figure 5:

Panel F and G: The asterisks were already shown in the previous version of the figure.

Panel H: The difference is not statistically-signficant.

Figure 6: No graphs are shown in this figure.

Reviewer comment 7: Since a wild type mouse control has not been included in the quantification of diaphragm NMJ innervation with and without ephrin knock-down, it would be useful to include a description or discussion on the phenotype of NMJ denervation exhibited in the SOD1G93A mouse model of ALS.

Response: We have now added description of diaphragm NMJ denervation that occurs in SOD1G93A mice, in particular at the age/time point of our NMJ analysis.

Author Response

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

eLife Assessment

This valuable manuscript investigates the roles of DKK3 in AD synapse integrity. Although previous work has identified the involvement of Wnt and DKK1 in synaptic physiology, this study provides compelling evidence that suppression of DKK3 rescues the changes in excitatory synapse numbers, as well as memory deficits in an established AD model mice. The authors provide both gain and loss of function data that support the main conclusion and advance our understanding of the mechanisms by which Wnt pathway mediates early synaptic dysfunction in AD models.

Public Reviews:

Reviewer #1 (Public Review):

In this study, Nuria Martin-Flores, Marina Podpolny and colleagues investigate the role of Dickkopf-3 (DKK3), a Wnt antagonist in synaptic dysfunction in Alzheimer's disease. Loss of synapses is a feature of Alzheimer's and other forms of dementia such as frontotemporal dementia and linked amyotrophic lateral sclerosis (FTD). The authors utilise a broad range of experimental approaches. They show that DKK3 levels are increased in Alzheimer's disease and that this occurs early in disease. This is an important finding since early disease changes are believed to be the most important. They also show increases in DKK3 in transgenic mouse models of Alzheimer's disease and that DKK3 knockdown restores synapse number and memory in one such model. Finally, they link these DKK3 increases to loss of excitatory synapses via the blockade of the Wnt pathway and subsequent activation of GSK3B; GSK3B is strongly linked to both Alzheimer's disease and FTD. The quality of the data is good and the conclusions well supported by these data. There are no major weaknesses. The findings support studies that target the Wnt pathway as a potential therapeutic for Alzheimer's disease.

Reviewer #2 (Public Review):

This manuscript by Martin-Flores et al., has examined the role of DKK3 in Alzheimer's disease, focusing on the regulation of synaptic numbers. By using human AD brain databases and tissue samples, the authors showed that DKK3 protein and mRNA levels are increased in the brains of AD patients. DKK3 is expressed in the excitatory neurons in WT mouse brains and accumulates at atrophic neurites around amyloid plaques in AD mouse brains. Interestingly, secretion of DKK3 appears to be regulated by NMDAR antagonist as well as chemical LTD. Through gain and loss of function studies, the authors showed that DKK3 regulates the number of excitatory as well as inhibitory synapses with distinct downstream pathways. Finally, the authors investigated the contribution of DKK3 to synaptic changes in AD and found that DKK3 loss of function rescues both the excitatory and inhibitory synaptic defects, resulting in the improvement of memory function in J20 mice.

Overall, the data is clearly presented and deals with novel roles of DKK3 in controlling excitatory and inhibitory synapses. The finding that shRNA expression of DKK3 in AD model mice rescues synaptic phenotypes and memory impairment is potentially interesting and may provide a new strategy for AD treatment.

We would like to thank the Editors and the Reviewers for their very insightful suggestions. We are delighted to receive very positive reviews of our manuscript. In response to the comments made by the reviewers, we have carried out an extensive revision of our manuscript. In the revised manuscript, we have addressed all the comments made by the reviewers.

Recommendations for the authors:

Reviewer #1:

My only comment regards the role of GSK3B activation in synaptic dysfunction and its targets. GSK3B is a Tau kinase but is also involved in IP3 receptor delivery of Ca2+ to mitochondria. This delivery is major regulator of mitochondrial ATP production and synaptic function is heavily dependent on ATP. Both Alzheimer's disease and FTD insults have been linked to GSK3B activation -see for e.g. Szabo EMBO R 2023, Gomez-Suaga Aging Cell 2022. It might be valuable to readers for the authors to speculate briefly on potential GSK3B synaptic targets in the Discussion.

We appreciate the reviewer for this suggestion. In the Discussion, we now included how GSK3β may contribute to synaptic dysfunction and loss in the context of increased DKK3 levels and in Alzheimer’s disease.

Reviewer #2:

1. In Fig 1B, the authors showed that soluble DKK3 levels were increased in Braak 1-3 patients, while no changes were observed in Braak 4-5. If the secretion of DKK3 is dependent on NMDAR activity, does this data imply that Braak 4-5 patients have reduced NMDAR activity in general, resulting in the reduced DKK3 release even with the increased mRNA levels? It would be interesting to test this hypothesis in a mouse AD model.

In Figure 1B, we analyzed the levels of soluble and insoluble DKK3 in the hippocampus of AD patients at different disease stages based on their Braak stages. As the reviewer indicated, soluble levels of DKK3 were increased in patients with Braak I-III but not at later stages. Importantly, DKK3 levels were also elevated in Braak IV-VI patients, but only in the insoluble fraction (Figure 1C), suggesting that DKK3 could accumulate within Aβ aggregates. Based on these findings, we cannot conclude that DKK3 release is reduced at later stages of the disease in patients.

To explore the underlying mechanisms regulating DKK3 levels, we used cultured hippocampal neurons and AD mouse brain slices. In mouse models, we have demonstrated that extracellular DKK3 levels (secreted DKK3 fraction) depends on NMDAR activation early in the disease progression (Figure 2E, F). Moreover, we also provide new data showing that antagonizing NMDAR partially blocks the increase of DKK3 extracellular levels induced by oligomeric Aβ (see response to question 4 of this reviewer and Figure S2G, H). It is well established that oligomeric Aβ promotes hyperexcitability through, in part, the aberrant activation of NMDAR (Li S et al., 2011, PMID: 21543591; Mucke L and Selkoe DJ et al., 2012, PMID: 22762015). In line with this, NMDAR blockers prevent Aβ-induced synapse loss and improve cognition in AD models (Hu NW et al., 2009, PMID: 19918059; Ye C et al., 2004, PMID: 15288443). In addition, an NMDAR antagonist is currently approved as a drug treatment for AD patients (Cumming J 2021, PMID: 33441154). Together, our findings in dissociated neurons, AD mouse brain and human samples indicate that soluble Aβ oligomers promote the release of DKK3 through NMDAR activation and suggest that this mechanism might also be occurring in the brain of AD patients.

1. Recent work (Yuan et al., 2022, Nature) has shown that dystrophic neurites/axonal spheroids found around Aβ deposits are filled with neuronal endolysosomes. Are DKK3 in ThioS positive amyloid plaques located in endolysosomes of these axonal spheroids? If so, does this data mean that DKK3 in Fig 2B-D represents the entrapped DKK3 protein population that fails to be secreted from dystrophic neurites?

The reviewer points an interesting question. Our results show that secretion of DKK3 is increased in two AD models before substantial plaque load. Later in the disease, DKK3 accumulates in dystrophic neurites (visualized as axonal spheroids) surrounding amyloid plaques. To address if DKK3 protein is located in vesicles of the endolysosomal pathway within axonal spheroids, we performed co-localization analyses of DKK3 and the endolysosomal marker LAMP1. We found that DKK3 colocalized with LAMP1 (Figure 2D) indicating the presence of DKK3 in axonal spheroids. These results indeed suggest that DKK3 is present in abnormally enlarged vesicles in dystrophic neurites around Aβ plaques. This could affect the axonal transport of DKK3. Given that proteins present in dystrophic neurites have been correlated with defects in bidirectional transport in the axon (Stokin GB et al., 2005, PMID: 15731448; Sadleir KR et al., 2016, PMID: 26993139), both DKK3 turnover and secretion could be affected.

1. Why does only LTD induce DKK3 release? Why not general activation of neuronal activity? It would be important to test the relationship between DKK3 secretion and neuronal activity with optogenetics and chemogenetics.

We tested whether neuronal activity triggered increased extracellular DKK3 levels by subjecting neurons to chemical long-term potentiation (cLTP) or long-term depression (cLTD). However, only cLTD increased extracellular DKK3, which we then confirmed in brain slices (Figure S3). This finding is not unexpected as it is well described that different patterns of activity can lead to different molecular outcomes. For example, high-frequency stimulation (HFS; an activity pattern that resembles LTP) and low-frequency stimulation (LFS; a different activity pattern resembling LTD) leads to opposing effects on surface levels of the Wnt receptor Frizzled-5 (Fz5) (Sahores M et al., 2010, PMID: 20530549). Furthermore, cLTP increases Fz5 s-acylation, an important post-translational modification that regulates the surface levels of Fz5, whereas cLTD decreases it (Teo S et al., 2023, PMID: 37557176). Another example is the BDNF receptor TrkB. Surface TrkB is increased by tetanic stimulation, which also induces LTP as HFS or cLTP, but not by LFS (Du J et al., 2000, PMID: 10995446). Our findings suggest that DKK3 might contribute to synaptic changes underlying cLTD. Future experiments using chemogenetics or optogenetics might elucidate the role of DKK3 in activity-induced synaptic changes.

1. Are Abeta oligomer treatment-dependent increases in DKK3 protein levels in the cellular lysate and the extracellular fraction also suppressed by APV?

Our results in AD mice indicate that increased DKK3 release is dependent on NMDAR activation. To investigate if amyloid-β oligomers (Aβo) increase DKK3 levels in the cell lysate and extracellular fractions through NMDAR, we blocked these receptors in hippocampal neurons using AP-V (Figure S2G, H). In these experiments, we use a lower concentration of Aβo (200nM of Aβ1-42) to avoid any potential cytotoxic effect. In line with our previous results using a higher concentration of Aβo, we observed that Aβo markedly increased DKK3 levels both in the cell lysate and in the extracellular fraction compared to the reverse Aβ42-1 control peptide. Kruskal-Wallis with Dunn’s test showed a trend to a reduced levels of DKK3 in the extracellular fraction when we compared neurons treated with Aβo and APV with those neurons treated with Aβ and vehicle (p = 0.0726). However, this reduced levels of DKK3 in the extracellular fraction reached statistical significance using a t-test (p = 0.0384). No differences were observed between the reverse control peptide and Aβo and APV conditions. These results suggest that blockade of the NMDAR partially occludes the ability of Aβo to increase DKK3 levels in the extracellular fraction.

1. Why does DKK3 shRNA only downregulate inhibitory synapses but not excitatory synapses in the WT brain slice? Does this mean that in the WT brain, other DKK proteins (without changes in their expression as shown in Fig S6) are sufficiently expressed and compensate for the roles of DKK3 in excitatory synapse integrity?

The reviewer points out an interesting result. In J20 mice, DKK3 knockdown affects both excitatory and inhibitory synapse density (Figure 6B, C). In Figure 3B, D, we show that in vivo downregulation of DKK3 leads to an increased number of inhibitory synapses without affecting excitatory ones in the brain of WT animals. These results indicate that in a healthy brain (WT), DKK3 is required for the maintenance of inhibitory synapses but not for excitatory synapses under our experimental conditions. Furthermore, DKK3 partially shares the mechanism of action with DKK1 as both DKK proteins promote excitatory synapse loss through the Wnt/GSK3β pathway (Figure 4A-C) (Marzo A et al., 2016, PMID: 27593374). Therefore, it is possible that endogenous DKK1 levels in the hippocampus could compensate for the reduced expression of DKK3 resulting in the lack of changes in excitatory synapse number when DKK3 is knockdown in WT animals.

1. Manipulating DKK3 in WT brains only affects Gephyrin but not VGAT, but in J20, both Gephyrin and VGAT seem to be affected by DKK3 shRNA (Fig 6). The authors need to provide the pre vs post synapse number in Fig 6 and discuss the potential differences.

We have now included the quantification of excitatory and inhibitory pre- and postsynaptic puncta for 4-months old (Figure S6B, C) and 9-months old (Figure S6D, E) WT and J20 mice. At 4-months old, the density of Homer1 puncta for excitatory synapses and both vGAT and Gephyrin for inhibitory synapses was increased and decreased respectively by knocking down DKK3 in the J20 mice. At 9-months, strong trends were observed in all the synaptic markers when downregulating DKK3, but significance was only reached for Homer1 puncta.

1. Where are the Wnt receptors expressed? Are they exclusively expressed in neurons? Can the authors exclude the potential involvement of glial cells in this process?

In neurons, Wnt receptors can be expressed in the synaptic terminals. For example, Wnt receptor Frizzled-5 is located at the presynaptic terminal and the dendritic shaft but not at spines (Sahores M et al., 2010, PMID: 20530549; McLeod F et al., 2018, PMID: 29694885), whereas Frizzled-7 is located at the dendritic shaft and spines (McLeod F et al., 2018, PMID: 29694885). In addition, the Wnt co-receptor LRP6 is present at both pre- and postsynaptic sites in excitatory synapses (Jones ME et al., 2023, PMID: 36638182). Kremen1, another receptor for Dkk proteins, is also highly expressed in the brain and our unpublished superresolution results show that this receptor is present in both pre- and postsynaptic sites of 53% of excitatory and 30% of inhibitory synapses. However, these receptors are not exclusively expressed in neurons and many of them are also highly expressed in astrocytes (Zhang Y et al., 2016, PMID: 25186741). Based on the literature and our findings, we cannot rule out the possibility that DKK3 may signal to other cell types such as astrocytes, which could also contribute to changes in synapse density. However, recombinant DKK3 induces structural and functional changes in excitatory and inhibitory synapses within 3-4h (Figure 3), suggesting that DKK3 acts on neurons leading to synaptic changes.

1. Does the shRNA treatment of DKK3 affect the size and number of amyloid plaques in the AD mice?

We thank the reviewer for raising this very important question. We have now evaluated the impact of DKK3 knockdown in Aβ pathology in the J20 mice. We did not observe differences in the Aβ coverage nor the averaged number and size of Aβ plaques when DKK3 was silenced in the CA3 (Figure S6F). Therefore, the changes we observe in excitatory and inhibitory synapse density around plaques after knocking down DKK3 are unlikely to be due to changes in Aβ plaques.

Author Response

eLife assessment

This study presents a valuable finding on the distinct subpopulation of adipocytes during brown-to-white conversion in perirenal adipose tissue (PRAT) at different ages. The evidence supporting the claims of the authors is convincing, although specific lineage tracing of this subpopulation of cells and mechanistic studies would expand the work. The work will be of interest to scientists working on adipose and kidney biology.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors performed single nucleus RNA-seq for perirenal adipose tissue (PRAT) at different ages. They concluded a distinct subpopulation of adipocytes arises through brown-to-white conversion and can convert to a thermogenic phenotype upon cold exposure.

Strengths:

PRAT adipose tissue has been reported as an adipose tissue that undergoes browning. This study confirms that brown-to-white and white-to-beige conversions also exist in PRAT, as previously reported in the subcutaneous adipose tissue.

We did not observe any white-to-beige conversion in PRAT under regular condition. The adipocyte population that arises from brown-to-white conversion (mPRAT-ad2) can respond to cold and restore their UCP1 expression. However, brown adipocytes that arise from the mPRAT-ad2 subpopulation after cold exposure have a distinct transcriptome to that of cold-induced beige adipocyte in iWAT (Figure S6K) and are more related to iBAT brown adipocytes (Figure 6E).

Weaknesses:

1. There is overall a disconnection between single nucleus RNA-seq data and the lineage chasing data. No specific markers of this population have been validated by staining.

We are not sure what “this population” refers to. We suspect it is the Ucp1-&Cidea+ mPRAT-ad2 adipocyte subpopulation. If so, we did not identify specific markers for these adipocytes as shown in Figure 1H and statement in the Discussion. mPRAT-ad2 is negative for Ucp1 and Cyp2e1, which are markers for mPRAT-ad1 and mPRAT-ad3&4, respectively. Therefore, we plan to stain the mPRAT with Ucp1, Cyp2e1 and Perilipin (a pan adipocyte marker) antibodies. Cells that are Perilipin+&Ucp1-&Cyp2e1- will represent the mPRAT-ad2 subpopulation.

1. It would be nice to provide more evidence to support the conclusion shown in lines 243 to 245 "These results indicated that new BAs induced by cold exposure were mainly derived from UCP1- adipocytes rather than de novo ASPC differentiation in puPRAT". Pdgfra-negative progenitor cells may also contribute to these new beige adipocytes.

Our sequencing data and many previous studies (Angueira et al., 2021; Burl et al., 2022; Dong et al., 2022) have shown that Pdgfra is a marker for all ASPCs. We will also check adipocyte labelling pattern of mPRAT in the PdgfraCre;Ai14 mice. If all adipocytes are Tomato+, it suggests that adipocytes in mPRAT are all derived from Pdgfra-expressing cells. Also, the cold-induced adipocytes in mPRAT resemble more to the brown adipocytes of iBAT than the beige adipocytes of iWAT (Figure 6E and S6K).

Angueira, A.R., Sakers, A.P., Holman, C.D., Cheng, L., Arbocco, M.N., Shamsi, F., Lynes, M.D., Shrestha, R., Okada, C., Batmanov, K., et al. (2021). Defining the lineage of thermogenic perivascular adipose tissue. Nat Metab 3, 469-484. 10.1038/s42255-021-00380-0.

Burl, R.B., Rondini, E.A., Wei, H., Pique-Regi, R., and Granneman, J.G. (2022). Deconstructing cold-induced brown adipocyte neogenesis in mice. Elife 11. 10.7554/eLife.80167.

Dong, H., Sun, W., Shen, Y., Balaz, M., Balazova, L., Ding, L., Loffler, M., Hamilton, B., Kloting, N., Bluher, M., et al. (2022). Identification of a regulatory pathway inhibiting adipogenesis via RSPO2. Nat Metab 4, 90-105. 10.1038/s42255-021-00509-1.

1. The UCP1Cre-ERT2; Ai14 system should be validated by showing Tomato and UCP1 co-staining right after the Tamoxifen treatment.

We will inject Ucp1CreERT2;Ai14 mice at 1- and 6-month-old of age with tamoxifen and collect one day after the last injection to check the overlap between the Tomato signal and UCP1 immunofluorescent staining.

Reviewer #2 (Public Review):

Summary:

In the present manuscript, Zhang et al utilize single-nuclei RNA-Seq to investigate the heterogeneity of perirenal adipose tissue. The perirenal depot is interesting because it contains both brown and white adipocytes, a subset of which undergo functional "whitening" during early development. While adipocyte thermogenic transdifferentiation has been previously reported, there remain many unanswered questions regarding this phenomenon and the mechanisms by which it is regulated.

Strengths:

The combination of UCP1-lineage tracing with the single nuclei analysis allowed the authors to identify four populations of adipocytes with differing thermogenic potential, including a "whitened" adipocyte (mPRAT-ad2) that retains the capacity to rapidly revert to a brown phenotype upon cold exposure. They also identify two populations of white adipocytes that do not undergo browning with acute cold exposure.

Anatomically distinct adipose depots display interesting functional differences, and this work contributes to our understanding of one of the few brown depots present in humans.

Weaknesses:

The most interesting aspect of this work is the identification of a highly plastic mature adipocyte population with the capacity to switch between a white and brown phenotype. The authors attempt to identify the transcriptional signature of this ad2 subpopulation, however, the limited sequencing depth of single nuclei somewhat lessens the impact of these findings. Furthermore, the lack of any form of mechanistic investigation into the regulation of mPRAT whitening limits the utility of this manuscript. However, the combination of well-executed lineage tracing with comprehensive cross-depot single-nuclei presented in this manuscript could still serve as a useful reference for the field.

The sequencing depth of our data is comparable, if not better than previously published snRNA-seq studies on adipose tissue (Burl et al., 2022; Sarvari et al., 2021; Sun et al., 2020). Therefore, the depth of our data has reached the limit of the 3’ sequencing methods. Unfortunately, due to size limitation of the adipocytes, it is also not feasible to sort them for Smart-seq.

Burl, R.B., Rondini, E.A., Wei, H., Pique-Regi, R., and Granneman, J.G. (2022). Deconstructing cold-induced brown adipocyte neogenesis in mice. Elife 11. 10.7554/eLife.80167.

Sarvari, A.K., Van Hauwaert, E.L., Markussen, L.K., Gammelmark, E., Marcher, A.B., Ebbesen, M.F., Nielsen, R., Brewer, J.R., Madsen, J.G.S., and Mandrup, S. (2021). Plasticity of Epididymal Adipose Tissue in Response to Diet-Induced Obesity at Single-Nucleus Resolution. Cell Metab 33, 437-453 e435. 10.1016/j.cmet.2020.12.004.

Sun, W., Dong, H., Balaz, M., Slyper, M., Drokhlyansky, E., Colleluori, G., Giordano, A., Kovanicova, Z., Stefanicka, P., Balazova, L., et al. (2020). snRNA-seq reveals a subpopulation of adipocytes that regulates thermogenesis. Nature 587, 98-102. 10.1038/s41586-020-2856-x.

Author Response

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

Public Reviews:

The study could also valuably explore what kinds of genes experienced what forms of expression evolution. A brief description of GO terms frequently represented in genes which showed strong patterns of expression evolution might be suggestive of which selective pressures led to the changes in expression in the C. bursa-pastoris lineage, and to what extent they related to adaptation to polyploidization (e.g. cell-cycle regulators), compensating for the initial pollen and seed inviability or adapting to selfing (endosperm- or pollen-specific genes), or adaptation to abiotic conditions. ”

We did not include a gene ontology (GO) analysis in the first place as we did not have a clear expectation on the GO terms that would be enriched in the genes that are differentially expressed between resynthesized and natural allotetraploids. Even if we only consider adaptive changes, the modifications could occur in various aspects, such as stabilizing meiosis, adapting to the new cell size, reducing hybrid incompatibility and adapting to self-fertilization. And each of these modifications involves numerous biological processes and molecular functions. As we could make post-hoc stories for too many GO terms, extrapolating at this stage have limited implications and could be misleading.

Nonetheless, we are not the only study that compared newly resynthesized and established allopolyploids. GO terms that were repeatedly revealed by this type of exploratory analysis may give a hint for future studies. For this reason, now we have reported the results of a simple GO analysis.

Recommendations for the authors: please note that you control which, if any, revisions, to undertake

The majority of concerns from reviewers and the reviewing editor are in regards to the presentation of the manuscript; that the framing of the manuscript does not help the general reader understand how this work advances our knowledge of allopolyploid evolution in the broad sense. The manuscript may be challenging to read for those who aren't familiar with the study system or the genetic basis of polyploidy/gene expression regulation. Further, it is difficult to understand from the introduction how this work is novel compared to the recently published work from Duan et al and compared to other systems. Because eLife is a journal that caters to a broad readership, re-writing the introduction to bring home the novelty for the reader will be key.

Additionally, the writing is quite technical and contains many short-hands and acronyms that can be difficult to keep straight. Revising the full text for clarity (and additionally not using acronyms) would help highlight the findings for a larger audience.

Reviewer #1 (Recommendations For The Authors):

Most of my suggestions on this interesting and well-written study are minor changes to clarify the writing and the statistical approaches.

The use of abbreviations throughout for both transcriptional phenomena and lines is logical because of word limits, but for me as a reader, it really added to the cognitive burden. Even though writing out "homoeolog expression bias" or "hybridization-first" every time would add length, I would find it easier to follow and suspect others would too.

Thank you for this suggestion. Indeed, using less uncommon acronyms or short-hands should increase the readability of the text for broader audience. Now in most places, we refer to “Sd/Sh” and “Cbp” as “resynthesized allotetraploids” and “natural allotetraploids”, respectively. We have also replaced the most occurrences of the acronyms for transcriptional phenomena (ELD, HEB and TRE) with full phrases, unless there are extra attributes before them (such as “Cg-/Co-ELD” and “relic/Cbp-specific ELD”).

It would be helpful to include complete sample sizes to either a slightly modified Figure 1 or the beginning of the methods, just to reduce mental arithmetic ("Each of the five groups was represented by six "lines", and each line had six individuals" so there were 180 total plants, of which 167 were phenotyped - presumably the other 13 died? - and 30 were sequenced).

The number 167 only applied to floral morphorlogical traits (“Floral morphological traits were measured for all five groups on 167 plants…”), but the exact total sample size for other traits differed. Now the total sample sizes of other traits have also been added to beginning of the second paragraph of the methods.

For this study 180 seedings have been transplanted from Petri dishes to soil, but 8 seedlings died right after transplanting, seemingly caused by mechanical damage and insufficient moistening. Later phenotyping (2020.02-2020.05) was also disrupted by the COVID-19 pandemic, and some individuals were not measured as we missed the right life stages. Specifically, 5 individuals were missing for floral morphological traits (sepal width, sepal length, petal width, petal length, pistil width, pistil length, and stamen length), 30 for pollen traits, 1 for stem length, and 2 for flowering time. As for seed traits, we only measured individuals with more than ten fruits, so apart from the reasons mentioned above, individuals that were self-incompatible and had insufficient hand-pollination were also excluded. We spotted another mistake during the revision: two individuals with floral morphological measurements had no positional information (tray ID). These measurements were likely mis-sampled or mislabeled, and were therefore excluded from analysis. We assumed most of these missing values resulted from random technical mistakes and were not directly related to the measured traits.

In general, the methods did a thorough job of describing the genomics approaches but could have used more detail for the plant growth (were plants randomized in the growth chamber, can you rule out block/position effects) and basic statistics (what statistical software was used to perform which tests comparing groups in each section, after the categories were identified).

When describing the methods, mention whether the plants; this should be straightforward as a linear model with position as a covariate.

Data used in the present study and a previously published work (Duan et al., 2023) were different subsets of a single experiment. For this reason, we spent fewer words in describing shared methods in this manuscript but tried to summarize some methods that were essential for understanding the current paper. But as you have pointed out, we did miss many important details that should have been kept. Now we have added some description and a table (Supplementary file 1) in the “Plant material” section for explaining randomization, and added more information of the software used for performing statistic tests in the “Phenotyping” section.

Although we did not mention in the present manuscript, we used a randomized block design for the experiment (Author response image 1).

Author response image 1.

Plant positions inside the growth chamber.

Plants used in the present study and Duan et al. (2023) were different subsets of a single experiment. The entire experiment had eight plant groups, including the five plant groups used in the present study (diploid C. orientalis (Co2), diploid C. grandiflora (Cg2), “whole-genome-duplication-first” (Sd) and “hybridization-first”(Sh) resynthesized allotetraploids, and natural allotetraploids, C. bursa pastoris (Cbp), as well as three plant groups that were only used in Duan et al. (2023; tetraploid C. orientalis (Co4), tetraploid C. grandiflora (Cg4) and diploid hybrids (F)). Each of the eight plant groups had six lines and each line represented by six plants, resulting in 288 plants (8 groups x 6 lines x 6 individuals = 288 plants). The 288 plants were grown in 36 trays placed on six shelves inside the same growth chamber. Each tray had exactly one plant from each of the eight groups, and the position of the eight plants within each tray (A-H) were randomized with random.shuffle() method in Python (Supplementary file 1). The position of the 36 trays inside the growth room (1-36) was also random and the positions of all trays were shuffled once again 28 days after germination (randomized with RAND() and sorting in Microsoft Excel Spreadsheet). (a) Plant distribution; (b) An example of one tray; (c) A view inside the growth chamber, showing the six benches.

With the randomized block design and one round of shuffling, positional effect is very unlikely to bias the comparison among the five plant groups. The main risk of not adding positions to the statistical model is increasing error variance and decreasing the statistical power for detecting group effect. As we had already observed significant among-group variation in all phenotypic traits (p-value <2.2e-16 for group effect in most tests), further increasing statistical power is not our primary concern. In addition, during the experiment we did not notice obvious difference in plant growth related to positions. Although we could have added more variables to account for potential positional effects (tray ID, shelf ID, positions in a tray etc.), adding variables with little effect may reduce statistical power due to the loss of degree of freedom.

Due to one round of random shuffling, positions cannot be easily added as a single continuous variable. Now we have redone all the statistical tests on phenotypic traits and included tray ID as a categorical factor (Figure 2-Source Data 1). In general, the results were similar to the models without tray ID. The F-values of group effect was only slightly changed, and p-values were almost unchanged in most cases (still < 2.2e-16). The tray effect (df=35) was not significant in most tests and was only significant in petal length (p-value=0.0111), sepal length (p-value=0.0242) and the number of seeds in ten fruits (p-value=0.0367). As expected, positions (tray ID) had limited effect on phenotypic traits.

Figure 2 - I assume the numbers at the top indicate sample sizes but perhaps add this to the figure caption.

Statistical power depends on both the total sample size and the sample size of each group, especially the group with the fewest observations. We lost different number of measurements in each phenotypic trait, and for pollen traits we did have a notable loss, so we chose to show sample sizes above each group to increase transparency. Since we had five different sets of sample sizes (for floral morphological traits, stem length, days to flowering, pollen traits and seed traits, respectively), it would be cumbersome to introduce all 25 numbers in figure caption and could be hard for readers to match the sample sizes with results. For this reason, we would like to keep the sample sizes in the figure, and now we have modified the legend to clarify that the numbers above groups are sample sizes.

’The trend has been observed in a wide range of organisms, including ...’ - perhaps group Brassica and Raphanobrassica into one clause in the sentence, since separating them out undermines the diversity somewhat.

Indeed, it is very strange to put “cotton” between two representatives from Brassicaceae. Now the sentence is changed to “… including Brassica (Wu et al., 2018; Li et al., 2020; Wei et al., 2021) and Raphanobrassica (Ye et al., 2016), cotton (Yoo et al., 2013)…”

The diagrams under the graph in Figure 4B are particularly helpful for understanding the expression patterns under consideration! I appreciated them a lot!

Thank you for the comment. We also feel the direction of expression level dominance is convoluted and hard to remember, so we adopted the convention of showing the directions with diagrams.

Reviewer #2 (Recommendations For The Authors):

The science is very interesting and thorough, so my comments are mostly meant to improve the clarity of the manuscript text:

• I found it challenging to remember the acronyms for the different gene expression phenomena and had to consistently cross-reference different parts of the manuscript to remind myself. I think using the full phrase once or twice at the start of a paragraph to remind readers what the acronym stands for could improve readability.

Thank you for this reasonable suggestion. Now we have replaced the most occurrence of acronyms with the full phrases.

• There are some technical terms, such as "homoeologous synapsis" and "disomic inheritance", which I think are under-defined in the current text.

Indeed these terms were not well-defined before using in the manuscript. Now we have added a brief explanation for each term.

• Under the joint action of these forces, allopolyploid subgenomes are further coordinated and degenerated, and subgenomes are often biasedly fractionated" This sentence has some unclear terminology. Does "coordinated" mean co-adapted, co-inherited, or something else? Is "biasedly fractionated" referring to biased inheritance or evolution of one of the parental subgenomes?

We apologize for not using accurate terms. With “coordinated” we emphasized the evolution of both homoeologs depends on the selection on total expression of both homoeologs, and on both relative and absolute dosages, which may have shifted away from optima after allopolyploidization. “Co-evolved” or “co-adapted” might be a better word.

But the term "biasedly fractionation" has been commonly used for referring to the phenomenon that genes from one subgenome of polyploids are preferentially retained during diploidization (Woodhouse et al., 2014; Wendel, 2015). Instead of inventing a new term, we prefer to keep the same term for consistency, so readers could link our findings with numerous studies in this field. Now the sentence is changed to “Under the joint action of these forces, allopolyploid subgenomes are further co-adapted and degenerated, and subgenomes are often biasedly retained, termed biased fractionation”.

• There are a series of paragraphs in the results, starting with "Resynthesized allotetraploids and the natural Cbp had distinct floral morphologies", which consistently reference Figure 1 where they should be referencing Figure 2.

Thank you for spotting this mistake! Now the numbers have been corrected.

• ‘The number of pollen grains per flower decreased in natural Cbp’ this wording implies it's the effect of some experimental treatment on Cbp, rather than just measured natural variation.

Yes, it is not scientifically precise to say this in the Results section, especially when describing details of results. We meant that assuming resynthesized allopolyploids are good approximation of the initial state of natural allotetraploid C. bursa-pastoris, our results indicate that the number of pollen grains had decreased in natural C. bursa-pastoris. But this is an implication, rather than an observation, so the sentence is better rewritten as “Natural allotetraploids had less pollen grains per flower.”

• ‘The percentage of genes showing complete ELD was altogether limited but doubled between resynthesized allotetraploid groups and natural allotetraploids’ for clarity, I would suggest revising this to something like "doubled in natural allotetraploids relative to resynthesized allotetraploids

Thank you for the suggestion. The sentence has been revised as suggested.

• I'm not sure I understand what the difference is between expression-level dominance and homeolog expression bias. It seems to me like the former falls under the umbrella of the latter.

Expression-level dominance and homeolog expression bias are easily confused, but they are conceptually independent. One gene could have expression-level dominance without any homeolog expression bias, or strong homeolog expression bias without any expression-level dominance. The concepts were well explained in Grover et al., (2012) with nice figures.

Expression level dominance compares the total expression level of both homoeologs in allopolyploids with the expression of the same gene in parental species, and judges whether the total expression level in allopolyploids is only similar to one of the parental species. The contributions from different homoeologs are not distinguished.

While homoeolog expression bias compares the relative expression level of each homoeologs in allopolyploids, with no implication on the total expression of both homoeologs.

Let the expression level of one gene in parental species X and Y be e(X) and e(Y), respectively. And let the expression level of x homoeolog (from species X) and y homoeolog (from species Y) in allopolyploids be e(x) and e(y), respectively.

Then a (complete) expression level dominance toward species X means: e(x)+e(y)=e(X) and e(x)+e(y)≠e(Y);

While a homoeolog expression bias toward species X means: e(x) > e(y), or e(x)/e(y) > e(X)/e(Y), depending on the definition of studies.

Both expression-level dominance and homeolog expression bias have been widely studied in allopolyploids (Combes et al., 2013; Li et al., 2014; Yoo et al., 2014; Hu & Wendel, 2019). As the two phenomena could be in opposite directions, and may be caused by different mechanisms, we think adopting the definitions in Grover et al., (2012) and distinguishing the two concepts would facilitate communication.

• Is it possible to split up the results in Figure 7 to show which of the two homeologs was lost (i.e. orientalis vs. grandiflora)? Or at least clarify in the legend that these scenarios are pooled together in the figure?

Maybe using acronyms without explanation made the figure titles hard to understand, but in the original Figure 7 the loss of two homoeologs were shown separately. Figure 7a,c showed the loss of C. orientalis-homoeolog (“co-expession loss”), and Figure 7b,d showed the loss of C. grandiflora-homoeolog (“cg-expession loss”). Now the legends have been modified to explain the Figure.

• The paragraph starting with "The extant diploid species" is too long, should probably be split into two paragraphs and edited for clarity.

The whole paragraph was used to explain why the resynthesized allotetraploids could be a realistic approximation of the early stage of C. bursa-pastoris with two arguments:

1) The further divergence between C. grandiflora and C. orientalis after the formation of C. bursa-pastoris should be small compared to the total divergence between the two parental species; 2) The mating systems of real parental populations were most likely the same as today. Now the two arguments were separated as two paragraphs, and the second paragraph has been shortened.

• On the other hand, the number of seeds per fruit" implies this is evidence for an alternative hypothesis, when I think it's really just more support for the same idea.

“On the other hand” was used to contrast the reduced number of pollen grains and the increased number of seeds in natural allotetraploids. As both changes are typical selfing syndrome, indeed the two support the same idea. We replaced the “On the other hand” with “Moreover”.

• ‘has become self-compatible before the formation" "has become" should be "became".

The tense of the word has been changed.

• If natural C. bursa-pastoris indeed originated from the hybridization between C. grandiflora-like outcrossing plants and C. orientalis-like self-fertilizing plants, the selfing syndrome in C. bursa-pastoris does not reflect the instant dominance effect of the C. orientalis alleles, but evolved afterward.’ This sentence should be closer to the end of the paragraph, after the main morphological results are summarized.

Thank you for the suggestion. The paragraph is indeed more coherent after moving the conclusion sentence.

References

Combes, M.C., Dereeper, A., Severac, D., Bertrand, B. & Lashermes, P. (2013) Contribution of subgenomes to the transcriptome and their intertwined regulation in the allopolyploid Coffea arabica grown at contrasted temperatures. New Phytologist, 200, 251–260.

Grover, C.E., Gallagher, J.P., Szadkowski, E.P., Yoo, M.J., Flagel, L.E. & Wendel, J.F. (2012) Homoeolog expression bias and expression level dominance in allopolyploids. New Phytologist, 196, 966–971.

Hu, G. & Wendel, J.F. (2019) Cis – trans controls and regulatory novelty accompanying allopolyploidization. New Phytologist, 221, 1691–1700.

Li, A., Liu, D., Wu, J., Zhao, X., Hao, M., Geng, S., et al. (2014) mRNA and Small RNA Transcriptomes Reveal Insights into Dynamic Homoeolog Regulation of Allopolyploid Heterosis in

Nascent Hexaploid Wheat. The Plant Cell, 26, 1878–1900. Wendel, J.F. (2015) The wondrous cycles of polyploidy in plants. American Journal of Botany, 102, 1753–1756.

Woodhouse, M.R., Cheng, F., Pires, J.C., Lisch, D., Freeling, M. & Wang, X. (2014) Origin, inheritance, and gene regulatory consequences of genome dominance in polyploids. Proceedings of the National Academy of Sciences of the United States of America, 111, 5283–5288.

Yoo, M.J., Liu, X., Pires, J.C., Soltis, P.S. & Soltis, D.E. (2014) Nonadditive Gene Expression in Polyploids. https://doi.org/10.1146/annurev-genet-120213-092159, 48, 485–517.

Author Response

Public Reviews:

Roget et al. build on their previous work developing a simple theoretical model to examine whether ageing can be under natural selection, challenging the mainstream view that ageing is merely a byproduct of other biological and evolutionary processes. The authors propose an agent-based model to evaluate the adaptive dynamics of a haploid asexual population with two independent traits: fertility timespan and mortality onset. Through computational simulations, their model demonstrates that ageing can give populations an evolutionary advantage. Notably, this observation arises from the model without invoking any explicit energy tradeoffs, commonly used to explain this relationship.

The model’s results are based on both numerical simulations and formal mathematical analysis.

Additionally, the theoretical model developed here indicates that mortality onset is generally selected to start before the loss of fertility, irrespective of the initial values in the population. The selected relationship between the fertility timespan and mortality onset depends on the strength of fertility and mortality effects, with larger effects resulting in the loss of fertility and mortality onset being closer together. By allowing for a trans-generational effect on ageing in the model, the authors show that this can be advantageous as well, lowering the risk of collapse in the population despite an apparent fitness disadvantage in individuals. Upon closer examination, the authors reveal that this unexpected outcome is a consequence of the trans-generational effect on ageing increasing the evolvability of the population (i.e., allowing a more effective exploration of the parameter landscape), reaching the optimum state faster.

The simplicity of the proposed theoretical model represents both the major strength and weakness of this work. On one hand, with an original and rigorous methodology, the logic of their conclusions can be easily grasped and generalised, yielding surprising results. Using just a handful of parameters and relying on direct competition simulations, the model qualitatively recapitulates the negative correlation between lifespan and fertility without requiring energy tradeoffs. This alone makes this work an important milestone for the rapidly growing field of adaptive dynamics, opening many new avenues of research, both theoretically and empirically.

We thank the reviewers and editor for highlighting the importance of the work presented here.

On the other hand, the simplicity of the model also makes its relationship with living organisms difficult to gauge, leaving open questions about how much the model represents the reality of actual evolution in a natural context.

We presented both in results and discussion how the mathematical trade-offs between fertility and survival time give rise to (xb, xd) configuration representative of existing aging modes.

In particular, a more explicit discussion of how the specifics of the model can impact the results and their interpretation is needed. For example, the lack of mechanistic details on the trans-generational effect on ageing makes the results difficult to interpret.

We discussed the role of the transgenerational Lansing effect played to its function, there is no need for a particular mechanism beyond that function of transgenerational negative effect. We reinforce this in the discussion by adding the following sentence “Regarding the nature of the transgenerational effect, our model is agnostic and the mere transmission of any negative effect would be sufficient to exert the function. “

Even if analytical results are obtained, most of the observations appear derived from simulations as they are currently presented. Also, the choice of parameters for the simulations shown in the paper and how they relate to our biological knowledge are not fully addressed by the authors.

The long time limit of the system with and without the Lansing effect is based on analytical results later confirmed using numerical simulations. The choice of parameters is explained in the introduction as being the minimum ones for defining a living organism. As for the parameters’ values, our numerical analysis gives a solution for any ib, id, xb and xd on R+, making the choice of initial value a mere random decision.

Finally, the conclusions of evolvability are insufficiently supported, as the authors do not show if the wider genotypic variability in populations with the ageing trans-generational effect is, in fact, selected.

We do not show nor claim that evolvability per se is selected for but that the apparent advantage given by this transgenerational effect seems to be mediated by an increased genotypic/phenotypic variability conferred to the lineage that we interpreted as evolvability.

Author Response

Reviewer #1 (Public Review):

De Seze et al. investigated the role of guanine exchange factors (GEFs) in controlling cell protrusion and retraction. In order to causally link protein activities to the switch between the opposing cell phenotypes, they employed optogenetic versions of GEFs which can be recruited to the plasma membrane upon light exposure and activate their downstream effectors. Particularly the RhoGEF PRG could elicit both protruding and retracting phenotypes. Interestingly, the phenotype depended on the basal expression level of the optoPRG. By assessing the activity of RhoA and Cdc42, the downstream effectors of PRG, the mechanism of this switch was elucidated: at low PRG levels, RhoA is predominantly activated and leads to cell retraction, whereas at high PRG levels, both RhoA and Cdc42 are activated but PRG also sequesters the active RhoA, therefore Cdc42 dominates and triggers cell protrusion. Finally, they create a minimal model that captures the key dynamics of this protein interaction network and the switch in cell behavior.

We thank reviewer #1 for this assessment of our work.

The conclusions of this study are strongly supported by data. Perhaps the manuscript could include some further discussion to for example address the low number of cells (3 out of 90) that can be switched between protrusion and retraction by varying the frequency of the light pulses to activate opto-PRG.

The low number of cells being able to switch can be explained by two different reasons:

1) first, we were looking for clear inversions of the phenotype, where we could see clear ruffles in the case of the protrusion, and clear retractions in the other case. Thus, we discarded cells that would show in-between phenotypes, because we had no quantitative parameter to compare how protrusive or retractile they were. This reduced the number of switching cells

2) second, we had a limitation due to the dynamic of the optogenetic dimer used here. Indeed, the control of the frequency was limited by the dynamic of unbinding of the optogenetic dimer. This dynamic of recruitment (~20s) is comparable to the dynamics of the deactivation of RhoA and Cdc42. Thus, the differences in frequency are smoothed and we could not vary enough the frequency to increase the number of switches. Thanks to the model, we can predict that decreasing the unbinding rate of the optogenetic tool should allow us to increase the number of switching cells.

We will add further discussion of this aspect to the manuscript.

Also, the authors could further describe their "Cell finder" software solution that allows the identification of positive cells at low cell density, as this approach will be of interest for a wide range of applications.

There is a detailed explanation of the ‘Cell finder’ in the method sections. It is also available on github at https://github.com/jdeseze/cellfinder and currently in development to be more user-friendly and properly commented.

Reviewer #2 (Public Review):

Summary:

This manuscript builds from the interesting observation that local recruitment of the DHPH domain of the RhoGEF PRG can induce local retraction, protrusion, or neither. The authors convincingly show that these differential responses are tied to the level of expression of the PRG transgene. This response depends on the Rho-binding activity of the recruited PH domain and is associated with and requires (co?)-activation of Cdc42. This begs the question of why this switch in response occurs. They use a computational model to predict that the timing of protein recruitment can dictate the output of the response in cells expressing intermediate levels and found that, "While the majority of cells showed mixed phenotypes irrespectively of the activation pattern, in few cells (3 out of 90) we were able to alternate the phenotype between retraction and protrusion several times at different places of the cell by changing the frequency while keeping the same total integrated intensity (Figure 6F and Supp Movie)."

Strengths:

The experiments are well-performed and nicely documented. However, the molecular mechanism underlying the shift in response is not clear (or at least clearly described). In addition, it is not clear that a prediction that is observed in ~3% of cells should be interpreted as confirming a model, though the fit to the data in 6B is impressive.

Overall, the main general biological significance of this work is that RhoGEF can have "off target effects". This finding is significant in that an orthologous GEF is widely used in optogenetic experiments in drosophila. It's possible that these findings may likewise involve phenotypes that reflect the (co-)activation of other Rho family GTPases.

We thank reviewer #2 for having assessed our work. Indeed, the main finding of this work is the change in the GEF function upon its change in concentration, which could be explained with a simple model supported by quantitative data. We think that the mechanism of the switch is quite clear, supported by the data showing the double effect of the PH domain and the activation of Cdc42. The few cells that are able to switch phenotype have to be seen as an honest data confirming that 1) concentration is indeed the main determinant of the protein’s function, and the switch is hard to obtain (which is also predicted by the model) 2) the two underlying networks are being activated at different timescales, which leaves some space for differential activation in the same cell. We are here limited by the dynamic of the optogenetic tool, as explained in the response to reviewer #1, and the intrinsic cell-to-cell variability.

Regarding the interpretation of our results as RhoGEF “off target effects”, we think that it might be too reductive. As said in the discussion, we proposed that the dual role of the RhoGEF could have physiological implications on the induction of front protrusions and rear retractions. While we do not demonstrate it here, it opens the door for further investigation.

Weaknesses:

The manuscript makes a number of untested assumptions and the underlying mechanism for this phenotypic shift is not clearly defined.

We may not have been clear in our manuscript, but we think that the underlying mechanism for this phenotypic shift is clearly explained and backed up by the data and the literature. It relies on 1) the ability of PRG to activate both RhoA and Cdc42 and 2) the ability of the PH domain to directly bind to active RhoA (which is, as shown in the manuscript, necessary but not sufficient for protrusions to happen). The model succeeds in reproducing the data of RhoA with only one free parameter and two independently fitted ones. The fact that activation of RhoA and Cdc42 lead to retraction and protrusion respectively is known since a long time. Thus, we think that the switch is clearly and quantitatively explained.

This manuscript is missing a direct phenotypic comparison of control cells to complement that of cells expressing RhoGEF2-DHPH at "low levels" (the cells that would respond to optogenetic stimulation by retracting); and cells expressing RhoGEF2-DHPH at "high levels" (the cells that would respond to optogenetic stimulation by protruding). In other words, the authors should examine cell area, the distribution of actin and myosin, etc in all three groups of cells (akin to the time zero data from figures 3 and 5, with a negative control). For example, does the basal expression meaningfully affect the PRG low-expressing cells before activation e.g. ectopic stress fibers? This need not be an optogenetic experiment, the authors could express RhoGEF2DHPH without SspB (as in Fig 4G).

We thank reviewer #2 for this suggestion. PRG-DHPH is known to affect the phenotype of the cell as shown in Valon et al., 2017. Thus, we really focused on the change implied by the change in optoPRG expression, to understand the phenotype difference. However, we agree that this could be an interesting data to add and will do the experiments for the revised version of the manuscript.

Relatedly, the authors seem to assume ("recruitment of the same DH-PH domain of PRG at the membrane, in the same cell line, which means in the same biochemical environment." supplement) that the only difference between the high and low expressors are the level of expression. Given the chronic overexpression and the fact that the capacity for this phenotypic shift is not recruitment-dependent, this is not necessarily a safe assumption. The expression of this GEF could well induce e.g. gene expression changes.

We agree with reviewer #2 that there could be changes in gene expression. In the next point of this supplementary note, we had specified it, by saying « that overexpression has an influence on cell state, defined as protein basal activity or concentration before activation. » We are sorry if it was not clear and will change this sentence for the new version.

One of the interests of the model is that it does not require any change in absolute concentrations, beside the GEF. The model is thought to be minimal and fits well and explains the data with very few parameters. We don’t show that there is no change in concentration but we show that it is not required to invoke it.

We will add in the revised version of the manuscript a paragraph discussing this question.

The third paragraph of the introduction, which begins with the sentence, "Yet, a large body of works on the regulation of GTPases has revealed a much more complex picture with numerous crosstalks and feedbacks allowing the fine spatiotemporal patterning of GTPase activities" is potentially confusing to readers. This paragraph suggests that an individual GTPase may have different functions whereas the evidence in this manuscript demonstrates, instead, that a particular GEF can have multiple activities because it can differentially activate two different GTPases depending on expression levels. It does not show that a particular GTPase has two distinct activities. The notion that a particular GEF can impact multiple GTPases is not particularly novel, though it is novel (to my knowledge) that the different activities depend on expression levels.

We thank the reviewer for this remark and didn’t intended to confuse the readers. Indeed, we think that this manuscript confirms the canonical view on the GTPases (as most optogenetic experiments did in the past years). We show here that it is more complicated at the level of the GEF. We agree that this is not particularly novel. However, to our knowledge, there is no example of such clear phenotypic control, explained solely by the change in concentration.

We think that the last paragraph of the introduction is quite clear in the fact that it is the GEF itself that switches its function, and not the Rho-GTPases, but we will reconsider the phrasing of this paragraph for the revised version.

Concerning the overall model summarizing the authors' observations, they "hypothesized that the activity of RhoA was in competition with the activity of Cdc42"; "At low concentration of the GEF, both RhoA and Cdc42 are activated by optogenetic recruitment of optoPRG, but RhoA takes over. At high GEF concentration, recruitment of optoPRG lead to both activation of Cdc42 and inhibition of already present activated RhoA, which pushes the balance towards Cdc42."

These descriptions are not precise. What is the nature of the competition between RhoA and Cdc42? Is this competition for activation by the GEFs? Is it a competition between the phenotypic output resulting from the effectors of the GEFs? Is it competition from the optogenetic probe and Rho effectors and the Rho biosensors? In all likelihood, all of these effects are involved, but the authors should more precisely explain the underlying nature of this phenotypic switch. Some of these points are clarified in the supplement, but should also be explicit in the main text.

We are going to precise these descriptions for the revised version of the manuscript. The competition between RhoA and Cdc42 was thought as a competition between retraction due to the protein network triggered by RhoA (through ROCK-Myosin and mDia-bundled actin) and the protrusion triggered by Cdc42 (through PAK-Rac-ARP2/3-branched Actin). We will make it explicit in the main text.

Author Response

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

eLife assessment

The findings of this study are valuable as they provide new insights into the role of acetylcholine in modulating sensory processing in the auditory cortex. This paper reports a systematic measurement of cell activity in the auditory cortex before and after applying ACh during an oddball and cascade sequence of auditory stimuli in anesthetized rats. The results presented are solid given the rigorous experimental design and statistical analysis. The conclusions are provocative and will interest researchers in auditory neuroscience and neuromodulation, as well as clinicians and individuals with auditory processing disorders. However, the findings support multiple interpretations, beyond that offered by the authors.

Our reply: First and foremost, we would like to thank the editors and reviewers for their constructive criticisms, as well as their thoughtful and thorough evaluations of our manuscript. We greatly appreciate their assessment about the novelty and general significance in our study and have revised the manuscript according to their recommendations. In the following we include detailed responses and revisions based on the reviewer’s recommendations.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study examined the impact of exogenous microapplication of acetylcholine (Ach) on metrics of novelty detection in the anesthetized rat auditory cortex. The authors found that the majority of units showed some degree of modulation of novelty detection, with roughly similar numbers showing enhanced novelty detection, suppressed novelty detection, or no change. Enhanced novelty responses were driven by increases in repetition suppression. Suppressed novelty responses were driven by deviance suppression. There were no compelling differences seen between auditory cortical subfields or layers, though there was heterogeneity in the Ach effects within subfields. Overall, these findings are important because they suggest that fluctuations in cortical Ach, which are known to occur during changes in arousal or attentional states, will likely influence the capacity of individual auditory cortical neurons to respond to novel stimuli.

Strengths:

The work addresses an important problem in auditory neuroscience. The main strengths of the study are that the work was systematically done with appropriate controls (cascaded stimuli) and utilizes a classical approach that ensures that drug application is isolated to the micro-environment of the recorded neuron. In addition, the authors do not isolate their study to only the primary auditory cortex, but examine the impact of Ach across all known auditory cortical subfields.

Our reply: Thank you very much for these supportive comments and the appreciation of our work.

Weaknesses:

1. As acknowledged by the authors, this study explicitly examines a phenomenon of high relevance to active listening but is done in anesthetized animals, limiting its applicability to the waking state.

Our reply: We agree; and indeed, this weakness was already recognized in the original manuscript but is now emphasized in the discussion.

1. The authors do not make any attempt to determine, by spike shape/duration, if their units are excitatory or inhibitory, which may explain some of the variance of the data.

Our reply: This is a very interesting question, and in fact, we have previously estimated whether neurons are excitatory or inhibitory based on the spike shape (Pérez-Gonzalez et al., 2021). Originally, we sought to implement a similar analysis here and tried to estimate if the recorded units were excitatory or inhibitory based on the spike shapes. But when we tried to perform this analysis, we found that in many cases the recordings had captured occasional spikes from other neurons. This caveat had introduced alterations in the average spike shape, and thus precluded an accurate categorization. Therefore, we decided to discard this analysis for the sake of correctness. This weakness is further commented on in the discussion.

1. The application of exogenous Ach, potentially in supra-physiological amounts, makes this study hard to extrapolate to a behaving animal. A more compelling design would be to block Ach, particularly at particular receptor types, to determine the effect of endogenous Ach.

Our reply: We agree again with the reviewer; this weakness was already acknowledged, but this is now further highlighted in discussion where we comment that future studies should analyze the effect of muscarinic- and nicotinic- receptors and blockade them to potentially observe more physiologically-comparable effects. Moreover, this issue is also related to a comment raised by reviewer#2 on a possible ‘dose-response relationship’ issue.

Reviewer #2 (Public Review):

Summary:

In this study, the authors investigate the effect of ACh on neuronal responses in the auditory cortex of anesthetized rats during an auditory oddball task. The paradigm consisted of two pure tones (selected from the frequency responses at each recording site) presented in a pseudo-random sequence. One tone was presented frequently (the "standard" tone) and the other infrequently (the "deviant" tone). The authors found that ACh enhances the detection of unexpected stimuli in the auditory environment by increasing or decreasing the neuronal responses to deviant and standard tones.

Strengths:

The study includes the use of appropriate and validated methodology in line with the current state-of-the-art, rigorous statistical analysis, and the demonstration of the effects of acetylcholine on auditory processing.

Our reply: Thank you very much for these supportive comments and the appreciation of our work.

Weaknesses:

The study was conducted in anesthetized rats, and further research is needed to determine the behavioral relevance of these findings.

Our reply: We agree; and indeed, this weakness was already recognized but is now emphasized in discussion.

Reviewer #1 (Recommendations For The Authors):

As outlined above, breaking out the units into those that are putative excitatory or inhibitory cells would be helpful, if possible. Other critiques are minor:

1. "Acetylcholine", "ACh" and "Ach" are used throughout the manuscript. Please define the chosen abbreviation at first use, and be consistent.

2. Line 116, remove comma after "ACh".

3. Line 123, I would add "in the rat at the end of the first sentence since the species was not mentioned up to this point.

4. Fig 2 - it would be useful in the Figure (not just in the text) to label red as being the deviant tone and blue as being the standard.

5. In many Figures (e.g., Fig 5), the term "effect" is found in the legend rather than "ACh". It would seem more intuitive to label these as "ACh".

6. The AUC and MI interpretations are not clear. Both are metrics that quantify similarity but the authors state that when these values decrease the neurons are less able to discriminate between them (i.e., they are more similar). Some clarifying text would be useful.

7. L276 - should "SI increase" be "SI decrease"?

8. L285 - would replace "solely" with "primarily".

9. Fig 7 - the authors may consider indicating with a label what the difference is between A and C compared to B and D.

10. L634 - why were only females used?

11. L646 - "bran" should be "brain".

12. L649 - "homoeothermic" should be "homeothermic".

13. L661 - "allowed to generate" should be "allowed the generation of".

14. L670 - no need for both "about" and "approximately".

15. L681 - please state what the search stimuli were.

16. L688 - should be "closed-field".

17. L754 - add a hyphen to "time-consuming".

Our reply: Thanks so much for the detailed proofreading of the manuscript and suggestions. All them have been clarified or implemented and corrected in the text.

Reviewer #2 (Recommendations For The Authors):

The authors could investigate the effects of different doses of ACh on auditory processing to determine if there is a dose-response relationship.

Our reply: We agree that this is an interesting question also relate to a matter raised by Reviewer#1 that could be linked to the issue of ‘exogenous Ach’.

The study only investigated the effects of ACh on neuronal responses during an auditory oddball task. It would be interesting to investigate the effects of ACh on other aspects of auditory processing, such as sound localization or the discrimination of tones.

Our reply: We agree that, while these aspects of auditory processing are very fascinating, they were outside the scope of the study, and not directly related to predictive coding and precision, so each one of these characteristics would be a full, future project in itself.

The authors could provide more context on the significance of their findings for individuals with auditory processing disorders.

Our reply: Thanks for the suggestion. It remains unclear how abnormal brainstem and cortical processing associated with auditory processing disorders arises (Moore, 2006, 2012). While we are not aware of any known direct connection between auditory processing disorders and acetylcholine, individuals with auditory processing disorders do have difficulties with auditory selective attention, so perhaps one could speculate that ACh, by modulating SSA/prediction error, could have some impact on encoding salient events, and if disrupted could lead to problems with selective attention. Moore (2012) speculated that auditory processing disorders may arise from unbalanced processing in bottom-up and top-down contributions.

Since ACh has been implicated in some neurogenerative diseases and neurodevelopmental disorders, we have also added in the Discussion dialogue about a possible relationship between the modulatory effect of ACh on predictive coding (which involves bottom-up and top-down contributions) and auditory processing disorders. We also cite the recent work by Felix and colleagues (2019) which is the only study we have found on the effects of ACh on auditory processing disorders where they analyzed altered temporal processing at the level of the brainstem in α7-subunit of the nicotinic acetylcholine receptor (α7-nAChR)-deficient mice. After studying α7-nAChR knockout mice of both sexes and wild-type colony controls, they concluded that the malfunction of the CHRNA7 gene that encodes the α7-nAChR may contribute to degraded spike timing in the midbrain, which may underlie the observed timing delay in the ABR signals. These authors propose that their findings are consistent with a role for the α7-nAChR in types of neurodevelopmental and auditory processing disorders. There is also evidence on cholinergic system disfunction being related to the pathophysiology of Alzheimer’s disease (Pérez-González et al., 2022). For instance, disfunction of the synapses of cholinergic neurons in the hippocampus and nucleus basalis of Meynert, as well as decreased choline acetyltransferase activity, is associated to memory disorders in Alzheimer’s disease (Hampel et al., 2018). Also, A Alzheimer’s disease D patients show reduced amounts of the vesicular ACh transporter in some brain areas (Aghourian et al., 2017). Finally, cholinesterase inhibitors seem to have some favorable effect in the treatment of Alzheimer’s disease patients (Sharma, 2019).

Aghourian M, Legault-Denis C, Soucy J-P, Rosa-Neto P, Gauthier S, Kostikov A, et al. 2017. Quantification of brain cholinergic denervation in Alzheimer’s disease using PET imaging with [18F]-FEOBV. Mol. Psychiatry 22:1531–1538. doi: 10.1038/mp.2017.183

Felix RA 2nd, Chavez VA, Novicio DM, Morley BJ, Portfors CV. 2019. Nicotinic acetylcholine receptor subunit α7-knockout mice exhibit degraded auditory temporal processing. J Neurophysiol. 122(2):451-465. doi: 10.1152/jn.00170.2019.

Hampel H, Mesulam M-M, Cuello AC, Khachaturian AS, Vergallo A, Farlow MR, et al. 2018. Revisiting the Cholinergic Hypothesis in Alzheimer’s Disease: emerging Evidence from Translational and Clinical Research. J. Prev. Alzheimers Dis. 6:1–14. doi:10.14283/jpad.2018.43

Moore DR. 2006. Auditory processing disorder (APD)-potential contribution of mouse research. Brain Res. 1091:200–206.

Moore DR. 2012. Listening difficulties in children: bottom-up and top-down contributions. J Commun Disord. ;45:411–418.

Pérez-González D, Parras GG, Morado-Díaz CJ, Aedo-Sánchez C, Carbajal GV, Malmierca MS. 2021. Deviance detection in physiologically identified cell types in the rat auditory cortex. Hear Res. 2021 Jan;399:107997. doi: 10.1016/j.heares.2020.107997.

Pérez-González D, Schreiner TG, Llano DA and Malmierca MS. 2022. Alzheimer’s Disease, Hearing Loss, and Deviance Detection. Front. Neurosci. 16:879480. doi: 10.3389/fnins.2022.879480

Sharma K. 2019. Cholinesterase inhibitors as Alzheimer’s therapeutics. Mol. Med. Rep. 20:1479–1487. doi:10.3892/mmr.2019.1 0374

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript by Heyndrickx et al describes protein crystal formation and function that bears similarity to Charcot-Leyden crystals made of galectin 10, found in humans under similar conditions. Therefore, the authors set out to investigate CLP crystal formation and their immunological effects in the lung. The authors reveal the crystal structure of both Ym1 and Ym2 and show that Ym1 crystals trigger innate immunity, activated dendritic cells in the lymph node, enhancing antigen uptake and migration to the lung, ultimately leading to induction of type 2 immunity.

Strengths:

We know a lot about expression levels of CLPs in various settings in the mouse but still know very little about the functions of these proteins, especially in light of their ability to form crystal structures. As such data presented in this paper is a major advance to the field.

Resolving the crystal structure of Ym2 and the comparison between native and recombinant CLP crystals is a strength of this manuscript that will be a very powerful tool for further evaluation and understanding of receptor, binding partner studies including the ability to aid mutant protein generation.

The ability to recombinantly generate CLP crystals and study their function in vivo and ex vivo has provided a robust dataset whereby CLPs can activate innate immune responses, aid activation and trafficking of antigen presenting cells from the lymph node to the lung and further enhances type 2 immunity. By demonstrating these effects the authors directly address the aims for the study. A key point of this study is the generation of a model in which crystal formation/function an important feature of human eosinophilic diseases, can be studied utilising mouse models. Excitingly, using crystal structures combined with understanding the biochemistry of these proteins will provide a potential avenue whereby inhibitors could be used to dissolve or prevent crystal formation in vivo.

The data presented flows logically and formulates a well constructed overall picture of exactly what CLP crystals could be doing in an inflammatory setting in vivo. This leaves open a clear and exciting future avenue (currently beyond the scope of this work) for determining whether targeting crystal formation in vivo could limit pathology.

Weaknesses:

Although resolving the crystal structure of Ym2 in particular is a strength of the authors work, the weaknesses are that further work or even discussion of Ym2 versus Ym1 has not been directly demonstrated. The authors suggest Ym2 crystals will likely function the same as Ym1, but there is insufficient discussion (or data) beyond sequence similarity as to why this is the case. If Ym1 and Ym2 crystals function the same way, from an evolutionary point, why do mice express two very similar proteins that are expressed under similar conditions that can both crystalise and as the authors suggest act in a similar way. Some discussion around these points would add further value.

We agree with reviewer. We have further elaborated the discussion section including these points, stating clearly that more research needs to be done using Ym2 crystals before we can draw parallels in vivo.

Additionally, the crystal structure for Ym1 has been previously resolved (Tsai et al 2004, PMID 15522777) and it is unclear whether the data from the authors represents an advance in the 3D structure from what is previously known.

The crystal structure of Ym1 has indeed been previously solved, and we refer to that paper. In addition, we also provide the crystal structure of in vitro grown Ym1, ashowing biosimilarity. This, for the field of crystallography is a major finding, since it validates the concept that crystal structures generated in vitro can reflect in vivo grown structures. Moreover, the in vivo crystallization of Ym2 was unknown prior to this work, and is now clear as revealed by the ex vivo X-ray crystallography. The strength of our story is that we can now compare Ym1 and Ym2 crystals structures in detail.

Whilst also generating a model to understand Charcot-Leyden crystals (CLCs), the authors fail to discuss whether crystal shape may be an important feature of crystal function. CLCs are typically needle like, and previous publications have shown using histology and TEM that Ym1 crystals are also needle like. However, the crystals presented in this paper show only formation of plate like structures. It is unclear whether these differences represent different methodologies (ie histology is 2D slides), or differences in CLP crystals that are intracellular versus extracellular. These findings highlight a key question over whether crystal shape could be important for function and has not been addressed by the authors.

In contrast to the bipyramidal, needle-like CLC crystals formed by human galectin-10 protein (hexagonal space group P6522), the in vivo grown Ym1 and Ym2 crystals we were able to isolate for X-ray diffraction experiments had a plate-like morphology with identical crystallographic parameters as recombinant Ym1/Ym2 crystals (space group P21). We note that depending on the viewing orientation of the thin plate-like Ym1 crystals, they may appear needle-like in histology and TEM images. In addition, we can fully not exclude that both Ym1 or Ym2 may crystallize in vivo in different space groups (which could result in different crystal morphologies for Ym1/Ym2) but we have no data to support this. It is finally also a possibility that plate like structures can break up in vivo along a long axis as a result of mechanical forces, and end up as rod-or needle like shapes.

Ym1/Ym2 crystals are often observed in conditions where strong eosinophilic inflammation is present. However, soluble Ym1 delivery in naïve mice shows crystal formation in the absence of a strong immune response. There is no clear discussion as to the conditions in which crystal formation occurs in vivo and how results presented in the paper in terms of priming or exacerbating an immune response align with what is known about situations where Ym1 and Ym2 crystals have been observed.

Although Ym1 and Ym2 crystals are often observed in mice at sites of eosinophilic inflammation, they are not made by eosinophils, but mainly by macrophages and epithelial cells, respectively. In vitro, protein crystallization typically starts from supersaturated solutions that support crystal nucleation. Several factors such as temperature and pH can affect the solubility of Ym1 and Ym2 in vivo and thus affect the nucleation and crystallization process. For Ym1 and Ym2 we noticed in vitro that a small drop in pH facilitates the crystallization process. Although the physiological pH is 7.4, during inflammation, there is a drop in pH. This drop in pH is the result of the infiltration and activation of inflammatory cells in the tissue, which leads to an increased energy and oxygen demand, accelerated glucose consumption via glycolysis and thus increased lactic acid secretion. In addition, we cannot exclude that in vivo, the nucleation process for Ym1/Ym2 is facilitated by interaction with ligands in the extracellular space (e.g. polysaccharide ligands or other – yet to be identified – specific ligands to Ym1/Ym2).

Reviewer #2 (Public Review):

Summary:

This interesting study addresses the ability of Ym1 protein crystals to promote pulmonary type 2 inflammation in vivo, in mice.

Strengths:

The data are extremely high quality, clearly presented, significantly extending previous work from this group on the type 2 immunogenicity of protein crystals.

Weaknesses:

There are no major weaknesses in this study. It would be interesting to see if Ym2 crystals behave similarly to Ym1 crystals in vivo. Some additional text in the Introduction and Discussion would enrich those sections.

We agree that this would be interesting to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Suggestions for improved experiments and to strengthen findings:

I think additional data on the ability of Ym2 crystals to induce an immune response would be advantageous. I'm not by any means suggesting the authors repeat all the experiments with Ym2 crystals, but even just the ability to show that Ym2 could promote type 2 immunity in the acute OVA model, would help to strengthen the argument that these crystals in general function in a similar way. Alternatively, a discussion on whether these protein crystals may function in different scenarios/tissues or conditions could help in light of additional data

We agree that this is an interesting point to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Measuring IL-33 in lung tissue is difficult to interpret as cells will express intracellular IL-33 that is not active and may explain why the results in Fig 2D are not overly convincing. It could just be that Ym1 crystals are changing the number of cells expressing IL-33 (e.g macrophages, or type 2 pneumocytes) Did the authors also measure active IL-33 release in the BAL fluid which may give a better indication of Ym1's ability to activate DAMPs?

We also measured active IL-33 release in the BAL fluid, but due to the limited sample availability we could only measure this in one of the two repeat experiments, resulting in non-significant results for the BAL fluid. However, certainly for the 6h timepoint we saw a similar trend in the BAL fluid as in the lung tissue, meaning higher levels of IL-33 in the Ym1 crystal group compared to the PBS and soluble Ym1 group.

Crystals in Fig 2F staining with Ym1 appear to be brighter in the soluble Ym1 group. Is this related to increased packing of Ym1 in the crystals formed in vivo as opposed to those formed in vitro? Aside from reduced amount of crystals that form when you give soluble Ym1, could the type of crystal also be influencing the ability of soluble Ym1 crystals to generate an immune response?

Our X-ray diffraction experiments show that the packing of Ym1 is identical for in vivo and in vitro grown crystals. Possibly the apparent difference in brightness is caused by stochastic staining by the antibody. In this regard we note that the crystals formed from soluble Ym1 after 24h also can appear as less bright in a similar fashion as recombinant Ym1 crystals.

Overall, the data and writing of the manuscript is presented to a very high standard

A few minor points:

• Fig 2F - a little unsure what the number in the left top corner of the images represented.

These numbers represent the picture numbers generated by the software, but as they don’t have any added value for the story, we removed these numbers from the images.

• Not clear why two different expression vectors were used - one for Ym1 and one for Ym2?

Because we observed that recombinant Ym2 is more poorly secreted in the mammalian cell culture supernatant as compared to recombinant Ym1, we produced Ym2 with an N-terminal hexahistidine-tag followed by a Tobacco Etch Virus (TEV)-protease cleavage site to facilitate its purification.

Reviewer #2 (Recommendations For The Authors):

The authors briefly outline in their Introduction potential Sources of Ym1/2 in vivo, highlighting monocytes, M2 macrophages, alveolar macrophages, neutrophils and epithelial cells. Do DCs also make detectable/meaningful amounts of Ym1/2 in vivo, particularly in type 2 settings?

In the introduction we only highlighted the main cellular sources of Ym1 and Ym2, but there is literature available stating/showing that Ym1/2 is not only expressed by macrophages, neutrophils, monocytes and epithelial cells, but can also be induced in DCs and mast cells. We added the word ‘mainly’ to this sentence in the introduction, to make clear that macrophages, neutrophils and monocytes are not the only sources of Ym1.

Given the nicely demonstrated similarity of recombinant Ym1 and Ym2 crystals, I think it is important for the authors to include at least initial data on the outcome of recombinant Ym2 crystal admin to mice, in comparison to their Ym1 data.

We agree that this is an interesting point to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Given the generation of crystals following in vivo administration of soluble Ym1, albeit at a lower level than when crystals were administered, it would be interesting to see if increased concentrations of soluble material show a dose dependent increase in lung inflammation readouts.

We agree that this would be an interesting point to investigate. Alongside this we could also titrate down the crystal dose, to see if there is a dose dependent decrease in lung inflammation readouts. However, at this time, we choose to not investigate this further.

I couldn't easily follow the authors' Discussion about potential ability of anti Ym-1/2 Abs to dissolve Ym1/2 crystals (similar to what they have demonstrated for Abs vs Gal10 crystals). Have they addressed this possibility experimentally? If so, addition of such data to the manuscript would be extremely interesting, given the obvious potential Ym1/2 crystal dissolving Abs for investigation of the role of these in a range of different murine models of type 2 inflammation.

We agree that the phrasing of this part of the discussion can be unclear/confusing. We rephrased this part to make it clearer. However, we did not address the possibility of Ym1/2 crystal dissolving antibodies experimentally.

In the Results section, the authors briefly comment on the pro-type 2 nature of Ym1 crystals in relation to their previous work with uric acid and Gal10 crystals, proposing that the pulmonary type 2 response may be a 'generic response to crystals of different chemical composition'. The Discussion would be enriched by deeper exploration of this comment.

We have further elaborated the discussion section including this point.

Author Response

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

We thank the reviewers for their thorough reading of the manuscript and insightful comments. We have responded to both the “public review” and the “recommendations” and feel that the manuscript is now significantly strengthened.

Public Review comments

Reviewer #1:

Weaknesses:

1. The abstract does not discuss the reduction of E-gel consumption that occurs after multiple days of exposure to the THC formulation, but rather implies that a new model for chronic oral self-administration has been developed. Given that only two days of consumption was assessed, it is not clear if the model will be useful to determine THC effects beyond the acute measures presented here. The abstract should clarify that there was evidence of reduced consumption/aversive effects with repeated exposures.

Thank you for your observation. We have added language to address this in the manuscript and the abstract. The model developed in the manuscript is an acute exposure model, with the intention of further chronic exposure adaptations to be developed separately (page 2, line 29).

1. In the results section, the authors sometimes describe effects in terms of the concentration of gel as opposed to the dose consumed in mg/kg, which can make interpretation difficult. For example, the text describing Figure 1i states that significant effects on body temperature were achieved at 4 mg CTR-gel and 5 mg THC-gel, but were essentially equivalent doses consumed? It would be helpful to describe what average dose of THC produced effects given that consumption varied within each group of mice assigned to a particular concentration.

We thank the reviewer for this comment and have edited our text to clarify our results. For example, this point is further emphasized by the correlation of the data in Figure1l-n showing the relationship between individual consumption and behavioral readouts (page 11, line 225-226).

1. The description of the PK data in Figure 3 did not specify if sex differences were examined. Prior studies have found that males and females can exhibit stark differences in brain and plasma levels of THC and metabolites, even when behavioral effects are similar. However, this does depend on species, route, timing of tissue collection. It would be helpful to describe the PK profile of males and females separately.

We did compare sex dependent effects and found no significant effects after THC E-gel consumption. We’ve added additional language to address this point in the discussion (Supplementary tables T1 and T2).

1. In Figure 5, it is unclear how the predicted i.p. THC dose could be 30 mg/kg when 30 mg/kg was not tested by the i.p. route according to the figure, and if it had been it would have likely been almost zero acoustic startle, not the increased startle that was observed in the 2 hr gel group. It seems more likely that it would be equivalent to 3 mg/kg i.p. Could there be an error in the modeling, or was it based on the model used for the triad effects? This should be clarified.

We apologize for the confusion created by that data, and it has now been updated for clarity. The original ~30mg/kg was not a predicted dose consumed, but rather an expected dose consumed based on individual male v. female consumption data in Supplemental Figure S1b. For clarity on the figure, we’ve instead placed dashed lines that draw attention only to the predicted startle response expected from our THC-E-gel model. We have also updated the text which hopefully makes this clearer.

Reviewer #2:

Weaknesses:

Certainly, more THC mediated behavioral outcomes could have been tested, but the work presents a proof-of-concept study to investigate acute THC treatment.

It would have been interesting if this application form is also possible for chronic treatment regimen

We agree that a chronic treatment regimen and additional behavioral outcomes is the next, most exciting step for expanding this oral THC-E-gel consumption model, and something we are actively pursuing.

Reviewer #3:

Weaknesses:

The main weaknesses of the manuscript revolve around clarification of the Methods section. All of these weaknesses are described in the "Recommendations to authors" section. Revising the manuscript would account for many of these weaknesses.

Thank you for carefully reading through our methodology. We have made edits according to everything brought up in the recommendation section of reviewer comments.

Recommendations for Authors

Reviewer #1:

Minor edits to the text:

Abstract: "intraperitoneal contingent" should be "intraperitoneal noncontingent".

Line 221, this sentence needs editing for clarity.

Lines 249-250, incomplete sentence.

Line 284, the word "activity" is missing from "locomotor between mice".

Lines 299-301, incomplete sentence.

Thank you for finding these mistakes. All these recommendations have been incorporated into the final publication.

Reviewer #2:

1. The typical THC tetrad includes catalepsy. Why was this behavioral outcome not monitored?

We felt that locomotion, analgesia, and body temperature were robust behavioral readouts for monitoring cannabimimetic responses and that acoustic startle served as an additional, novel means of understanding THC-E-gel effects.

1. Please specify the exact substrain of C57BL/6 (i.e., J or N or some other)

C57BL/6J mice were used for the publication. This clarification has been made in the methods section.

1. Figure S3 is not mentioned in the result part, but only in the discussion.

Figure S3 is now referenced in the main body of the Results section.

1. It might be interesting to follow up the issue that the individual THC consumption is considerable, as depicted in Fig. 1e (at high dose). This will presumably also lead to different behavioral responses. Or is there individual metabolism, also difference male vs. female?

Thank you for the suggestion. We agree that the distribution of THC doses consumed (calculation based on weight) would be worth further investigating and have now included language about this (page 20, line 436). Please note that we did not find a sex difference (Supplemental Figure S1b), but it would be exciting to discover some biologically relevant cause such as individual absorption or metabolism

Reviewer #3:

Major

1. Methods: Were the observers of experiments blinded to animal treatment? Why or why not?

Multiple investigators performed the behavioral measurements and were not blinded to mouse treatments, but the dose consumed by each mouse remained blind. Thus, because animals consumed THC gelatin of their own volition while having ad libitum access, we performed the correlational analysis presented in Figure 1 l-n.

1. Methods: The authors could consider relating their study design to the ARRIVE guidelines and providing a statement as to whether their study adheres to these guidelines. Related to this, were mice provided with any environmental enrichment during the study?

We followed the ARRIVE guidelines with exception to investigator blinding (described above). Please note that mice were not provided with additional environmental enrichment during the study, a point that we specified in our methods (page 5, line 91).

1. Methods / Results: In the Methods it is stated that the triad of cannabimimetic behaviors was measured 1 h post-injection or immediately after gelatin exposure. Why were these timepoints chosen? Perhaps this wording should be revised because measurements of cannabimimetic effects were taken several times after drug exposure. Peak i.p. drug may occur earlier than 1 h whereas peak oral drug effect is likely to occur over a longer time period (i.e., not immediately after) due to delays of absorption and first pass metabolism. Is it possible that the authors have underestimated oral drug effects by selecting these timepoints? Please discuss.

We observed a reduction in locomotion activity starting 1 h following the beginning of exposure to the gelatin (Figure 2), suggesting initial cannabimimetic changes. Based on this observable response we chose to measure all cannabimimetic behaviors immediately following gelatin exposure. The exposure timeline for i.p. injection (1 h post-injection) was selected based on a standard published protocol (Metna-Laurent et al, 2017).

a. Pharmacodynamics: Related to this and because the aim of this paper is to establish a rodent oral dose model, could the authors discuss the need for better characterization of the time course of drug effects? For example, how might anti-nociception or locomotor activity vary following THC E-gel consumption? This is somewhat addressed in the locomotion time course in Figure 2G but could be elaborated on or discussed in more detail.

We agree that future studies should include additional time points measuring behavioral changes. This important point is now emphasized in the discussion (page 21, line 455).

b. Pharmacokinetics: Related to this point above, have the authors considered collecting blood or tissue samples from their i.p.-injected animals to assess drug pharmacokinetics as they relate to drug effect and as compared to oral THC consumption? I am not suggesting the authors conduct a completely new study for this manuscript; however, this could be raised as a future study and/or as a weakness of the current study.

We did not measure blood and tissue concentrations after i.p. administration due to the number of studies reporting these values by our co-author, Dr. Daniele Piomelli, that established these pharmacokinetic measures. Thus, we chose to reference these studies. Please note that repeating such measurements would be labor intensive, unnecessary use federal NIH resources and animals, while being very redundant to the existing literature.

c. Minor, but related to these points: In the results, page 14 line 299: the first sentence of this paragraph is confusing as written. The Reviewer recognizes that the authors are relating the pharmacokinetic work to previously published findings, but still thinks that measuring and comparing THC levels from their cohort of i.p.-injected animals would have benefitted the present study.

Thank you, this edit has been made in the manuscript.

1. Methods, Histology: The methods as described do not contain sufficient detail regarding THC and THC metabolite quantification. In addition, it is not clear from this section what Histology was performed and how (no histology results appear in the manuscript). Please add more detail to this section of the Methods.

We apologize for this typo and have corrected it in the methods section of the manuscript.

1. Methods / Results: The statistics section requires additional detail regarding the rationale for tests being performed on different datasets. In addition, a description of the curve fitting used for data in figures 1H-J, 4B-D, and S4 would be helpful to the reader.

Thank you, we have updated and provided more information regarding the curve fitting that was used in the methods and results section for the respective figure panels (page 9, line 183-184).

Minor

1. Throughout: The use of the phrase "high" dose is somewhat arbitrary and not defined relative to other doses of the THC formulation throughout the manuscript. The Reviewer suggests simply stating that THC was used, specifying the dose, or justifying in the Abstract and/or Introduction the classification of "high" based on relevant literature.

Thank you for the observation. We have removed this ambiguity by specifically mentioning the dose that was consumed (e.g., abstract page 2, line 20).

1. Abstract: define "CB1" in the abstract. Although this is a common abbreviation within the field, its use should be defined.

We have added this definition in the abstract for clarification.

1. Figure 2: why are the consumption panels B, C, and D given separate labels but the locomotor data are all labeled together as panel G?

Thank you for the observation, we have adjusted the labeling, so it is equal for both sets of panels.

Author Response

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

Thank you very much for forwarding these two important reviews on our paper. Please find hereby our point-by-point responses addressing the ideas, arguments and points of concern raised by the reviewers. We provide explanation of how these points have been incorporated in the paper.

We feel the review process has been a useful exercise and that the paper has greatly benefited in terms of clarity and accessibility. It is our hope that our findings may ignite renewed interest on unexplored and “unexpected” aspects of great ape vocal communication, inspire novel research, and invite bold new advances on the long-standing puzzle of language origins and evolution. In several relevant sections, we have also sought to explicitly address the point of doubt raised in eLife’s editorial assessment, published alongside the reviewed preprint of our paper. The editorial assessment stated that “…However the evidence provided to support the major claims of the paper is currently incomplete. Specifically, it is not yet clear how the rhythmic structuring found in these long calls is more similar to human language recursion per se rather than isochrony as a broader, more common phenomenon.” To directly clarify this point, we provide now various examples of how recursion is distinct from repetition, using everyday objects for an intuitive understanding (e.g., lines 43-51). We have also expanded the discussion to better contextualise and clarify the implications of our findings on language evolution theory. We hope this will help addressing the implicit request for clarification in the previous editorial assessment.

Thank you very much for your kind and dedicated attention in the processing of our study.

Public Reviews:

Reviewer #1 (Public Review):

This study investigates the structuring of long calls in orangutans. The authors demonstrate long calls are structured around full pulses, repeated following a regular tempo (isochronic rhythm). These full pulses are themselves structured around different sub-pulses, themselves repeated following an isochronic rhythm. The authors argue this patterning is evidence for self-embedded, recursive structuring in orangutang long calls.

The analyses conducted are robust and compelling and they support the rhythmicity the authors argue is present in the long calls. Furthermore, the authors went above and beyond and confirmed acoustically the sub-categories identified were accurate.

We thank the reviewer for this important support regarding our methods and findings.

However, I believe the manuscript would benefit from a formal analysis of the specific recursive patterning occurring in the long call. Indeed, as of now, it is difficult for the reader to identify what the authors argue to be recursion and distinguish it from simple repetitions of motifs, which is essential.

We agree with the reviewer that the distinction between repetition and recursion is very important for the adequate interpretation of our findings. Following the reviewer’s point (and the Editorial Assessement), we have now rephrased several passages in the initial paragraph of the paper for added clarity, where recursion is introduced and explained. We now also provide various new examples of recursion in everyday life and popular culture to better illustrate in an easy and accessible way the fundamental nature of recursion. We then use two of these common examples (computer folders and Russian dolls) to specifically distinguish repetition from recursion.

Although the authors already discuss briefly why linear patterning is unlikely, the reader would benefit from expanding on this discussion section and clarifying the argument here (a lay terminology might help).

Corrected accordingly.

I believe an illustration here might help. In the same logic, I believe a tree similar to the trees used in linguistics to illustrate hierarchical structuring would help the reader understand the recursive patterning in place here. This would also help get the "big picture", as Fig 1A is depicting a frustratingly small portion of the long call.

We completely understand the reviewer’s concern here. As proposed by the reviewer, and in addition to changes in the Introduction (see above) and Discussion (see below), we have now added a new figure in the Discussion to help the reader get the “big picture” of our findings.

We have also made revisions throughout the Introduction and Discussion to simplify the text, clarify our exposition and facilitate the reader better and intuitively understand the nature and relevance of our results.

Notwithstanding these comments, this paper would provide crucial evidence for recursion in the vocal production of a non-human ape species. The implication it would have would represent a key shift in the field of language evolution. The study is very elegant and well-constructed. The paper is extremely well written, and the point of view adopted is original, well-argued and compelling.

We are humbled by the reviewer’s words, and we thank the reviewer for attributing these qualities to our paper. This feedback reassures us of the disruptive potential that these and similar future findings may have on our understanding of language evolution.

Reviewer #2 (Public Review):

I am not qualified to judge the narrow claim that certain units of the long calls are isochronous at various levels of the pulse hierarchy. I will assume that the modelling was done properly. I can however say that the broad claims that (i) this constitutes evidence for recursion in non-human primates, (ii) this sheds light on the evolution of recursion and/or language in humans are, when not made trivially true by a semantic shift, unsupported by the narrow claims. In addition, this paper contains errors in the interpretation of previous literature.

We report the first confirmed case of “vocal sequences within vocal sequences” in a wild nonhuman primate, namely a great ape. The currently prevailing models of language evolution often rest on the (purely theorical) premise that such structures do not exist in any animal bar humans. We find the discovery of such structures in a wild great ape exciting, remarkable, and promising. We regret that the reviewer does not share this sentiment with us. We feel that the statement that these findings are trivial and narrow is unfounded.

In order to clarify and better communicate the significance of our findings, we now explain in more detail in the Introduction and Discussion how the discovery of nested isochrony in wild orangutans promises to stimulate new series of studies in nature and captivity. Our findings dovetail nicely with previous captive studies that have shown that animals can learn how to recognise recursive patterns and invite new research efforts for the investigation of recursive abilities in the wild and in the absence of human priming and in nonhuman primates.

The main difficulty when making claims about recursion is to understand precisely what is meant by "recursion" (arguably a broader problem with the literature that the authors engage with). The authors offer some characterization of the concept which is vague enough that it can include anything from "celestial and planetary movement to the splitting of tree branches and river deltas, and the morphology of bacteria colonies". With this appropriately broad understanding, the authors are able to show "recursion" in orangutans' long calls. But they are, in fact, able to find it everywhere.

The reviewer is correct in highlighting that recursion is ubiquitous in nature and this is something that we explicitly state in the paper. This only makes it the more surprising that, when it comes to vocal combinatorics, recursion has only been described in human language and music, but in no other animals. If studies providing such evidence are known to reviewer, we kindly request their corresponding references.

In the new revised version, we have paid attention to this aspect raised by the reviewer, and we have sought to disambiguate that our observations pertain to temporal recursion. This clarification will hopefully allow a better understanding of our results.

The sound of a plucked guitar string, which is a sum of self-similar periodic patterns, count as recursive under their definition as well.

The example pointed out here by reviewer is factually correct; sound harmonics represent a recursive pattern of a fundamental frequency. (In fact, we explain this phenomenon in the Discussion.) The reviewer’s comment seems to offer an analogy to oscillatory phenomena in the physiology of the vocal folds, and so, it is misplaced with regards to our present study, which focused vocal sequences. Admittedly, this misinterpretation may have been implicitly caused by our wording and we apologise for this. We now refer to “vocal combinatorics” instead of “vocal production” throughout the paper to avoid the reader considering that our findings pertain to the physiology of the vocal folds.

One can only pick one's definition of recursion, within the context of the question of interest: evolution of language in humans. One must try to name a property which is somewhat specific to human language, and not a ubiquitous feature of the universe we live in, like self-similarity. Only after having carved out a sufficiently distinctive feature of human language, can we start the work of trying to find it in a related species and tracing its evolutionary history. When linguists speak of recursion, they speak of in principle unbounded nested structure (as in e.g., "the doctor's mother's mother's mother's mother ..."). The author seems to acknowledge this in the first line of the introduction: "the capacity to iterate a signal within a self-similar signal" (emphasis added). In formal language theory, which provides a formal and precise definition of one notion of recursivity appropriate for human language, unbounded iteration makes a critical difference: bounded "nested structures" are regular (can be parsed and generated using finite-state machines), unbounded ones are (often) context-free (require more sophisticated automaton). The hierarchy of pulses and sub-pulses only has a fixed amount of layers, moreover the same in all productions; it does not "iterate".

The reviewer explains here how recursion, in its fully fledged form in modern language(s), is defined by linguistics. We fully agree and do not contest such descriptions and definitions in any way. These descriptions and definitions aim to describe how recursion operates today, not how it evolved. Nor do these descriptions and definitions generate data-driven, testable predictions about precursors or proto-states of recursion as used by modern language-able humans. This is scientifically problematic and heuristically unsatisfying regarding the open question of language evolution.

Following human-specific definitions for recursion, as proposed by the reviewer, cannot per se be used to undertake a comparative approach to evolution because they leave nothing to compare recursion with in other (wild) species. Using human-specific definitions unavoidably leads to black-and-white notions that language is always absolutely present in humans and always absolutely absent in other animals, regardless of their degree of relatedness to humans. It is unpreventable that these descriptions flout foundational principles of evolution, such as descent with modification and shared ancestry.

This conceptual problem is not new. Less than a century ago, it was believed that humans were the only tool-user (thousands of examples are known today in nonhuman animals, including fish and invertebrates), and later, that humans were the only cultural animal (today it is known that migrating caribou and fruit flies can establish traditions based on social learning). We must follow in the footsteps of those who have helped redefine human nature in the past. As famously stated by Louis Leakey when presented with evidence for chimpanzee tool-use collected by Jane Goodall, “Now we must redefine tool, redefine man, or accept chimpanzees as human”. Therefore, as a matter of course, we must redefine recursion, embracing empirically (other than purely theoretically) definitions that allow recursion to take on forms and functions different from that of modern language-able humans.

Another point is that the authors don't show that the constraints that govern the shape of orangutans long calls are due to cognitive processes.

The reviewer is indeed correct. This does not, however, refute our findings. We do not directly show that cognitive processes govern recursion in orangutan long calls. Instead, we show that the observed patterns cannot be explained by simple bodily or motoric processes, excluding therefore low-level explanations. With more than 50 years of accumulated field experience in primatology, this was the only possible way that our team found to go about conducting research and analyses on natural behaviour, in the wild, with a critically endangered primate. We would be very interested in learning from the reviewer what ethical and non-invasive methods, specific locations in the wild, and type of behavioural or socio-ecological data could be otherwise viably used to demonstrate what the reviewer requests. If other scientists believe that the patterns observed in wild orangutan long calls – three independent, but simultaneously-occurring recursive motifs – can be generated based on low-level physiological mechanisms alone, the burden of proof resides with them.

Any oscillating system will, by definition, exhibit isochrony.

We disagree with this statement. The example provided above by the reviewer him/her-self disproves the statement: a guitar string when struck is an oscillating system but it is not isochronic nor is it combinatorial. Isochrony cannot be established with single events, only with event sequences (in practice, ideally >3).

For instance, human trills produce isochronouns or near isochronous pulses. No cognitive process is needed to explain this; this is merely the physics of the articulators. Do we know that the rhythm of the pulses and sub-pulses in orangutans is dictated by cognition as opposed to the physics of the articulators?

The reviewer seems to misinterpret our results here. Our focus is on vocal combinatorics, not vocal fold oscillation (see previous response). We have now reworded all instances where the text could be unclear.

Even granting the authors' unjustified conclusion that wild orangutans have "recursive" structures and that these are the result of cognition, the conclusions drawn by the authors are too often fantastic leaps of induction. Here is a cherry-picked list of some of the far-fetched conclusions: - "our findings indicate that ancient vocal patterns organized across nested structural strata were likely present in ancestral hominids". Does finding "vocal patterns organized across nested structural strata" in wild orangutans suggest that the same were present in ancestral hominids?

Following the reviewer’s comment, we have now rephrased and toned down this passage, stating that such structures “may have been present” in ancestral hominids. We are grateful to the reviewer for this comment.

• "given that isochrony universally governs music and that recursion is a feature of music, findings (sic.) suggest a possible evolutionary link between great ape loud calls and vocal music". Isochrony is also a feature of the noise produced by cicadas. Does this suggest an evolutionary link between vocal music and the noise of cicadas?

We apologise, but it is unclear what the reviewer is exactly suggesting or proposing here. It seems as though it is believed that cicadas are as phylogenetically related to humans as great apes are. Our last common ancestor with great apes diverged about 10mya, but with cicadas 600mya. The last common ancestor with great apes was a great ape (or hominid). The human-cicada last common ancestor would have looked like a worm (it is probable it would already have a nervous bulge at the head, or “brain”). In order to avoid similar misinterpretations, we have now clarified in several instances that our study and interpretation of results are based on shared ancestry within the Hominid family.

It seems that the reviewer may be also misinterpreting our findings. We do not simply report isochrony in a wild great ape (multiple references for isochronous calls in primate are provided in the Discussion). We report isochrony within isochrony in three non-exclusive rhythmic arrangements. In case the reviewer knows of a study on cicadas, or any non-human species, showing recursive sound combinatorics of this nature, we kindly request the citation. We can only hope that such new cases may be gradually unveiled in wild animals to help propel our general understanding of possible ways of how insipient recursive vocal combinatorics in ancient hominids could have given rise to recursion as used today by language-able modern humans.

Finally, some passages also reveal quite glaring misunderstandings of the cited literature. For instance:

• "Therefore, the search for recursion can be made in the absence of meaning-base operations, such as Merge, and more generally, semantics and syntax". It is precisely Chomsky's (disputable) opinion that the main operation that govern syntax, Merge, has nothing to do with semantics. The latter is dealt within a putative conceptual-intentional performance system (in Chomsky's terminology), which is governed by different operations.

Following the reviewer’s comment, we have now removed “meaning-base operations, such as Merge, and more generally” from the target sentence in order to avoid confusion. Thank you.

• "Namely, experimental stimuli have consisted of artificial recursive signal sequences organized along a single temporal scale (though not structurally linear), similarly with how Merge and syntax operate". The minimalist view advocated by Chomsky assumes that mapping a hierarchichal structure to a linear order (a process called linearizarion) is part of the articulatory-perceptual system. This system is likewise not governed by Merge and is not part of "syntax" as conceived by the Chomskyan minimalists.

Following the reviewer’s comment, we have not omitted the target sentence for added clarity.

Reviewer #1 (Recommendations For The Authors):

L55-67: I feel there is a step missing in the logic of the argumentation here. The studies cited by the authors here are mostly about syntactic-like structuring but not recursion. Hence when the authors mention in the next sentence that these studies investigate the perception of recursive signalling, it seems incorrect. I agree with the logic, but the references do not seem appropriate. I would further suggest that if there are no other references, that would make the introduction of the study here even easier: there is very little work investigating this capacity in non-human animals, let alone on a production perspective, therefore, the study conducted here is paramount and fills this important gap in the literature.

We are grateful to Reviewer #1 for these comments, and we are honoured to hear that our findings are filling a literature gap. We have now carefully revised the manuscript, hopefully, streamlining our line of reasoning and improving the paper’s overall readability. We agree that there is very little work investigating the spontaneous “production” of recursion in nonhuman animals. We decided to better detail the logic of our paper by clarifying the difference between recursion and repetition and clarifying that the motifs that we identify in wild orangutan represent a case of "temporal recursion".

L59: Johan J should be removed (same in discussion).

Removed, thanks.

L60: For example is repeated twice, here and L55.

We have rephrased this part of the manuscript, thanks.

L72-73: If we consider the Watson et al., 2020 study an example of recursive perception (which I do not think is true), this was conducted using a passive design - i.e. with no active training.

We have rephrased this part of the manuscript, thanks.

L240-241: Again, non-adjacent dependency processing does not equal recursion.

We agree that non-adjacent dependency processing does not equal recursion. We have now clarified this section accordingly.

L269: one of the most.

Corrected, thanks.

L296: add space after settings.

Corrected, thanks.

Reviewer #2 (Recommendations For The Authors):

In addition to the public portion of the review, I advise the authors' to substantially alter their style of writing. The language used is not accurate and the intended meaning is often not clear. This makes it hard for any reader to follow the authors' reasoning fully. Below I list only a few of the egregious examples but the examples abound:

• "this hints at a neuro-cognitive or neuro-computational transformation in the human brain" what meaning do the author assign to "neuro-cognitive" and "neuro-computational" ? what difference do they place between the two (so that they would be disjoined.) ? What "transformation" are we talking about ? From what to what ?

• " However, recursive signal structures can also unfold in other manners, such as across nested temporal scales and in the absence of semantics (Fitch, 2017a), as in music." what is meant here by nested temporal scales ?

• "The simultaneous occurrence of non-exclusive recursive patterns excludes the likelihood that orangutans concatenate long calls and their subunits in linear structure without any recursive processes": isn't there a more straightforward way to say "excludes the likelihood"? What is meant by "non-exclusive recursive patterns"?

It seems that Reviewer #2 does not share our writing style. Nonetheless, we have tried to meet the reviewer halfway, clarifying throughout the new revised version our definitions, our line of argument, our motivations, our results, the context of our findings in what is known about recursion in animals, and the implication of our discovery for language evolution theory.

Author Response

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

We agree with the reviewer that the statistics are buried in a dense excel file without a read-me page. We will address this by making a summary excel page for p-values during the production process.

---

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

eLife assessment

This important study uses genomically-engineered glypican alleles to demonstrate convincingly that Dally (not Dally-like protein [Dlp]) is the key contributor to formation of the Dpp/BMP morphogen gradient in the wing disc of Drosophila. The authors provide solid genetic evidence that, surprisingly, the core domain of Dally appears to suffice to trap Dpp at the cell surface. They conclude with a model according to which Dally modulates the range of Dpp signaling by interfering with Dpp's internalization by the Dpp receptor Thickveins.

Public Reviews:

Reviewer #1 (Public Review):

How morphogens spread within tissues remains an important question in developmental biology. Here the authors revisit the role of glypicans in the formation of the Dpp gradient in wing imaginal discs of Drosophila. They first use sophisticated genome engineering to demonstrate that the two glypicans of Drosophila are not equivalent despite being redundant for viability. They show that Dally is the relevant glypican for Dpp gradient formation. They then provide genetic evidence that, surprisingly, the core domain of Dally suffices to trap Dpp at the cell surface (suggesting a minor role for GAGs). They conclude with a model that Dally modulates the range of Dpp signaling by interfering with Dpp's degradation by Tkv. These are important conclusions, but more independent (biochemical/cell biological) evidence is needed.

As indicated above, the genetic evidence for the predominant role of Dally in Dpp protein/signalling gradient formation is strong. In passing, the authors could discuss why overexpressed Dlp has a negative effect on signaling, especially in the anterior compartment. The authors then move on to determine the role of GAG (=HS) chains of Dally. They find that in an overexpression assay, Dally lacking GAGs traps Dpp at the cell surface and, counterintuitively, suppresses signaling (fig 4 C, F). Both findings are unexpected and therefore require further validation and clarification, as outlined in a and b below.

a. In loss of function experiments (dallyDeltaHS replacing endogenous dally), Dpp protein is markedly reduced (fig 4R), as much as in the KO (panel Q), suggesting that GAG chains do contribute to trapping Dpp at the cell surface. This is all the more significant that, according to the overexpression essays, DallyDeltaHS seems more stable than WT Dally (by the way, this difference should also be assessed in the knock-ins, which is possible since they are YFP-tagged). The authors acknowledge that HS chains of Dally are critical for Dpp distribution (and signaling) under physiological conditions. If this is true, one can wonder why overexpressed dally core 'binds' Dpp and whether this is a physiologically relevant activity.

According to the overexpression assay, DallyDeltaHS seems more stable than WT Dally (Fig. 4B’, E’, 5A’, B’). As the reviewer suggested, we addressed the difference using the two knock-in alleles and found that DallyDeltaHS is more stable than WT Dally (Fig.4 L, M inset), further emphasizing the insufficient role of core protein of Dally for extracellular Dpp distribution.

In summary, we showed that, although Dally interacts with Dpp mainly through its core protein from the overexpression assay (Fig. 4E, I), HS chains are essential for extracellular Dpp distribution (Fig. 4R). Thus, the core protein of Dally alone is not sufficient for extracellular Dpp distribution under physiological conditions. These results raise a question about whether the interaction of core protein of Dally with Dpp is physiologically relevant. Since the increase of HS upon dally expression but not upon dlp expression resulted in the accumulation of extracellular Dpp (Fig. 2) and this accumulation was mainly through the core protein of Dally (Fig. 4E, I), we speculate that the interaction of the core protein of Dally with Dpp gives ligand specificity to Dally under physiological conditions.

To understand the importance of the interaction of core protein of Dally with Dpp under physiological conditions, it is important to identify a region responsible for the interaction. Our preliminary results overexpressing a dally mutant lacking the majority of core protein (but keeping the HS modified region intact) showed that HS chains modification was also lost. Although this is consistent with our results that enzymes adding HS chains also interact with the core protein of Dally (Fig. 4D), the dally mutant allele lacking the core protein would hamper us from distinguishing the role of core protein of Dally from HS chains.

Nevertheless, we can infer the importance of the interaction of core protein of Dally with Dpp using dally[3xHA-dlp, attP] allele, where dlp is expressed in dally expressing cells. Since Dally-like is modified by HS chains but does not interact with Dpp (Fig. 2, 4), dally[3xHA-dlp, attP] allele mimics a dally allele where HS chains are properly added but interaction of core protein with Dpp is lost. As we showed in Fig.3O, S, the allele could not rescue dallyKO phenotypes, consistent with the idea that interaction of core protein of Dally with Dpp is essential for Dpp distribution and signaling and HS chain alone is not sufficient for Dpp distribution.

b. Although the authors' inference that dallycore (at least if overexpressed) can bind Dpp. This assertion needs independent validation by a biochemical assay, ideally with surface plasmon resonance or similar so that an affinity can be estimated. I understand that this will require a method that is outside the authors' core expertise but there is no reason why they could not approach a collaborator for such a common technique. In vitro binding data is, in my view, essential.

We agree with the reviewer that a biochemical assay such as SPR helps us characterize the interaction of core protein of Dally and Dpp (if the interaction is direct), although the biochemical assay also would not demonstrate the interaction under the physiological conditions.

However, SPR has never been applied in the case of Dpp, probably because purifying functional refolded Dpp dimer from bacteria has previously been found to be stable only in low pH and be precipitated in normal pH buffer (Groppe J, et al., 1998)(Matsuda et al., 2021). As the reviewer suggests, collaborating with experts is an important step in the future.

Nevertheless, SPR was applied for the interaction between BMP4 and Dally (Kirkpatrick et al., 2006), probably because BMP4 is more stable in the normal buffer. Although the binding affinity was not calculated, SPR showed that BMP4 directly binds to Dally and this interaction was only partially inhibited by molar excess of exogenous HS, suggesting that BMP4 can interact with core protein of Dally as well as its HS chains. In addition, the same study applied Co-IP experiments using lysis of S2 cells and showed that Dpp and core protein of Dally are co-immunoprecipitated, although it does not demonstrate if the interaction is direct.

In a subsequent set of experiments, the authors assess the activity of a form of Dpp that is expected not to bind GAGs (DppDeltaN). Overexpression assays show that this protein is trapped by DallyWT but not dallyDeltaHS. This is a good first step validation of the deltaN mutation, although, as before, an invitro binding assay would be preferable.

Our overexpression assays actually showed that DppDeltaN is trapped by DallyWT and by dallyDeltaHS at similar levels (Fig. 5C), indicating that interaction of DppDeltaN and HS chains of Dally is largely lost but DppDeltaN can still interact with core protein of Dally.

We thank the reviewer for the suggesting the in vitro experiment. Although we decided not to develop biophysical experiments such as SPR for Dpp in this study due to the reasons discussed above, we would like to point out that our result is consistent with a previous Co-IP experiment using S2 cells showing that DppDeltaN loses interaction with heparin (Akiyama2008).

However, in contrast to our results, the same study also proposed by Co-IP experiments using S2 cells that DppDeltaN loses interaction with Dally (Akiyama2008). Although it is hard to conclude since western blotting was too saturated without loading controls and normalization (Fig. 1C in Akiyama 2008), and negative in vitro experiments do not necessarily demonstrate the lack of interaction in vivo. One explanation why the interaction was missed in the previous study is that some factors required for the interaction of DppDeltaN with core protein of Dally are missing in S2 cells. In this case, in vivo interaction assay we used in this study has an advantage to robustly detect the interaction.

Nevertheless, the authors show that DppDeltaN is surprisingly active in a knock-in strain. At face value (assuming that DeltaN fully abrogates binding to GAGs), this suggests that interaction of Dpp with the GAG chains of Dally is not required for signaling activity. This leads to authors to suggest (as shown in their final model) that GAG chains could be involved in mediating the interactions of Dally with Tkv (and not with Dpp. This is an interesting idea, which would need to be reconciled with the observation that the distribution of Dpp is affected in dallyDeltaHS knock-ins (item a above). It would also be strengthened by biochemical data (although more technically challenging than the experiments suggested above). In an attempt to determine the role of Dally (GAGs in particular) in the signaling gradient, the paper next addresses its relation to Tkv. They first show that reducing Tkv leads to Dpp accumulation at the cell surface, a clear indication that Tkv normally contributes to the degradation of Dpp. From this they suggest that Tkv could be required for Dpp internalisation although this is not shown directly. The authors then show that a Dpp gradient still forms upon double knockdown (Dally and Tkv). This intriguing observation shows that Dally is not strictly required for the spread of Dpp, an important conclusion that is compatible with early work by Lander suggesting that Dpp spreads by free diffusion. These result show that Dally is required for gradient formation only when Tkv is present. They suggest therefore that Dally prevents Tkv-mediated internalisation of Dpp. Although this is a reasonable inference, internalisation assays (e.g. with anti-Ollas or anti-HA Ab) would strengthen the authors' conclusions especially because they contradict a recent paper from the Gonzalez-Gaitan lab.

Thanks for suggesting the internalization assay. As we discussed in the discussion, our results suggest that extracellular Dpp distribution is severely reduced in dally mutants due to Tkv mediated internalization of Dpp (Fig. 6). Thus, extracellular Dpp available for labelling with nanobody is severely reduced in dally mutants, which can explain the reduced internalization of Dpp in dally mutants in the internalization assay. Therefore, we think that the nanobody internalization assay would not distinguish the two contradicting possibilities.

The paper ends with a model suggesting that HS chains have a dual function of suppressing Tkv internalisation and stimulating signaling. This constitutes a novel view of a glypican's mode of action and possibly an important contribution of this paper. As indicated above, further experiments could considerably strengthen the conclusion. Speculation on how the authors imagine that GAG chains have these activities would also be warranted.

Thank you very much!

Reviewer #2 (Public Review):

The authors are trying to distinguish between four models of the role of glypicans (HSPGs) on the Dpp/BMP gradient in the Drosophila wing, schematized in Fig. 1: (1) "Restricted diffusion" (HSPGs transport Dpp via repetitive interaction of HS chains with Dpp); (2) "Hindered diffusion" (HSPGs hinder Dpp spreading via reversible interaction of HS chains with Dpp); (3) "Stabilization" (HSPGs stabilize Dpp on the cell surface via reversible interaction of HS chains with Dpp that antagonizes Tkv-mediated Dpp internalization); and (4) "Recycling" (HSPGs internalize and recycle Dpp).

To distinguish between these models, the authors generate new alleles for the glypicans Dally and Dally-like protein (Dlp) and for Dpp: a Dally knock-out allele, a Dally YFP-tagged allele, a Dally knock-out allele with 3HA-Dlp, a Dlp knock-out allele, a Dlp allele containing 3-HA tags, and a Dpp lacking the HS-interacting domain. Additionally, they use an OLLAS-tag Dpp (OLLAS being an epitope tag against which extremely high affinity antibodies exist). They examine OLLAS-Dpp or HA-Dpp distribution, phospho-Mad staining, adult wing size.

They find that over-expressed Dally - but not Dlp - expands Dpp distribution in the larval wing disc. They find that the Dally[KO] allele behaves like a Dally strong hypomorph Dally[MH32]. The Dally[KO] - but not the Dlp[KO] - caused reduced pMad in both anterior and posterior domains and reduced adult wing size (particularly in the Anterior-Posterior axis). These defects can be substantially corrected by supplying an endogenously tagged YFP-tagged Dally. By contrast, they were not rescued when a 3xHA Dlp was inserted in the Dally locus. These results support their conclusion that Dpp interacts with Dally but not Dlp.

They next wanted to determine the relative contributions of the Dally core or the HS chains to the Dpp distribution. To test this, they over-expressed UAS-Dally or UAS-Dally[deltaHS] (lacking the HS chains) in the dorsal wing. Dally[deltaHS] over-expression increased the distribution of OLLAS-Dpp but caused a reduction in pMad. Then they write that after they normalize for expression levels, they find that Dally[deltaHS] only mildly reduces pMad and this result indicates a major contribution of the Dally core protein to Dpp stability.

Thanks for the comments. We actually showed that compared with Dally overexpression, Dally[deltaHS] overexpression only mildly reduces extracellular Dpp accumulation (Fig. 4I). This indicates a major contribution of the Dally core protein to interaction with Dpp, although the interaction is not sufficient to sustain extracellular Dpp distribution and signaling gradient.

The "normalization" is a key part of this model and is not mentioned how the normalization was done. When they do the critical experiment, making the Dally[deltaHS] allele, they find that loss of the HS chains is nearly as severe as total loss of Dally (i.e., Dally[KO]). Additionally, experimental approaches are needed here to prove the role of the Dally core.

Since the expression level of Dally[deltaHS] is higher than Dally when overexpressed, we normalized extracellular Dpp distribution (a-Ollas staining) against GFP fluorescent signal (Dally or Dally[deltaHS]). To do this, we first extracted both signal along the A-P axis from the same ROI in the previous version. The ratio was calculated by dividing the intensity of a-Ollas staining with the intensity of GFP fluorescent signal at a given position x. The average profile from each normalized profile was generated and plotted using the script described in the method (wingdisc_comparison.py) as other pMad or extracellular staining profiles.

Although this analysis provides normalized extracellular Dpp accumulation at different positions along the A-P axis, we are more interested in the total amount of Dpp or DppDeltaN accumulation upon Dally or dallyDeltaHS expression. Therefore, in the revised ms, we decided to normalize total amount of extracellular Dpp against the level of Dally or Dally[deltaHS] by dividing total signal intensity of extracellular Dpp staining (ExOllas staining) by total GFP fluorescent signal (Dally or Dally[deltaHS]) around the Dpp producing cells in each wing disc. Statistical analysis showed that accumulation of extracellular Dpp is only slightly reduced without HS chains (Fig.4I), indicating that Dally interacts with Dpp mainly through its core protein.

We agree with the reviewer that additional experimental approaches are needed to address the role of the core protein of Dally. As we discussed in the response to the reviewer1, to understand the importance of the interaction of core protein of Dally with Dpp, it is important to identify a region responsible for the interaction. Our preliminary results overexpressing a dally mutant lacking the majority of core protein (but keeping the HS modified region intact) showed that HS chains modification was also lost. Although this is consistent with our results that enzymes adding HS chains also interact with the core protein of Dally (Fig. 4D), the dally mutant allele lacking the core protein would hamper us from distinguishing the role of the core protein of Dally from HS chains.

Nevertheless, we can infer the importance of the interaction of core protein of Dally with Dpp using dally[3xHA-dlp, attP] allele, where dlp is expressed in dally expressing cells. Since Dally-like is modified by HS chains but does not interact with Dpp (Fig. 2, 4), dally[3xHA-dlp, attP] allele mimics a dally allele where HS chains are properly added but interaction of core protein with Dpp is lost. As we showed in Fig.3O, S, the allele could not rescue dallyKO phenotypes, consistent with the idea that interaction of core protein of Dally with Dpp is essential for Dpp distribution and signaling.

Prior work has shown that a stretch of 7 amino acids in the Dpp N-terminal domain is required to interact with heparin but not with Dpp receptors (Akiyama, 2008). The authors generated an HA-tagged Dpp allele lacking these residues (HA-dpp[deltaN]). It is an embryonic lethal allele, but they can get some animals to survive to larval stages if they also supply a transgene called “JAX” containing dpp regulatory sequences. In the JAX; HA-dpp[deltaN] mutant background, they find that the distribution and signaling of this Dpp molecule is largely normal. While over-expressed Dally can increase the distribution of HA-dpp[deltaN], over-expression of Dally[deltaHS] cannot. These latter results support the model that the HS chains in Dally are required for Dpp function but not because of a direct interaction with Dpp.

Our overexpression assays actually showed that both Dally and Dally[deltaHS] can accumulate Dpp upon overexpression and the accumulation of Dpp is comparable after normalization (Fig. 5C), consistent with the idea that interaction of DppdeltaN and HS chains are largely lost. As the reviewer pointed out, these results support the model that the HS chains in Dally are required for Dpp function but not because of a direct interaction with Dpp.

In the last part of the results, they attempt to determine if the Dpp receptor Thickveins (Tkv) is required for Dally-HS chains interaction. The 2008 (Akiyama) model posits that Tkv activates pMad downstream of Dpp and also internalizes and degrades Dpp. A 2022 (Romanova-Michaelides) model proposes that Dally (not Tkv) internalizes Dpp.

To distinguish between these models, the authors deplete Tkv from the dorsal compartment of the wing disc and found that extracellular Dpp increased and expanded in that domain. These results support the model that Tkv is required to internalize Dpp.

They then tested the model that Dally antagonizes Tkv-mediated Dpp internalization by determining whether the defective extracellular Dpp distribution in Dally[KO] mutants could be rescued by depleting Tkv. Extracellular Dpp did increase in the D vs V compartment, potentially providing some support for their model. However, there are no statistics performed, which is needed for full confidence in the results. The lack of statistics is particularly problematic (1) when they state that extracellular Dpp does not rise in ap>tkv RNAi vs ap>tkv RNAi, dally[KO] wing discs (Fig. 6E) or (2) when they state that extracellular Dpp gradient expanded in the dorsal compartment when tkv was dorsally depleted in dally[deltaHS] mutants (Fig. 6I). These last two experiments are important for their model but the differences are assessed only visually. In fact, extracellular Dpp in ap>tkv RNAi, dally[KO] (Fig. 6B) appears to be lower than extracellular Dpp in ap>tkv RNAi (Fig. 6A) and the histogram of Dpp in ap>tkv RNAi, dally[KO] is actually a bit lower than Dpp in ap>tkv RNAi, But the author claim that there is no difference between the two. Their conclusion would be strengthened by statistical analyses of the two lines.

We provided statistics for all the quantifications for pMad and extracellular Dpp distribution as supplementary data. In the previous version, we argued that extracellular Dpp level in ap>tkvRNAi, dallyKO (Fig.6B) does not increase compared with that in ap>tkvRNAi (Fig.6A). Statistical analysis (t-test) showed that the extracellular Dpp level in Fig. 6B is similar to or lower than that in Fig. 6A (Fig. 6E), confirming our conclusion. Statistical analysis (t-test) also confirmed that extracellular Dpp distribution expanded when tkv was knocked down in dallyHS mutants (Fig. 6I).

Strengths:

1. New genomically-engineered alleles

A considerable strength of the study is the generation and characterization of new Dally, Dlp and Dpp alleles. These reagents will be of great use to the field.

Thanks. We hope that these resources are indeed useful to the field.

1. Surveying multiple phenotypes

The authors survey numerous parameters (Dpp distribution, Dpp signaling (pMad) and adult wing phenotypes) which provides many points of analysis.

Thanks!

Weaknesses:

1. Confusing discussion regarding the Dally core vs HS in Dpp stability. They don't provide any measurements or information on how they "normalize" for the level of Dally vs Dally[deltaHS]? This is important part of their model that currently is not supported by any measurements.

We explained how we normalized in the above section and updated the method section in the revised ms.

1. Lacking quantifications and statistical analyses:

a. Why are statistical significance for histograms (pMad and Dpp distribution) not supplied? These histograms provide the key results supporting the authors' conclusions but no statistical tests/results are presented. This is a pervasive shortcoming in the current study.

Thanks. We provided t-test analyses together with the raw data as supplementary data.

b. dpp[deltaN] with JAX transgene - it would strengthen the study to supply quantitative data on the percent survival/lethal stage of dpp[deltaN] mutants with or without the JAK transgene

In this study, we are interested in the role of dpp[deltaN] during the wing disc development. Therefore, we decided not to perform the detailed analysis on the percent survival/lethal stage of dpp[deltaN] mutants with or without the JAX transgene in the current study. Nevertheless, the fact that dpp[deltaN] allele is maintained with a balanced stock and JAX;dpp[deltaN] allele can be maintained as homozygous stock indicates that the lethality of dpp[deltaN] allele comes from the early stages. Indeed, our preliminary results showed that pMad signal is severely lost in the dpp[deltaN] embryo without JAX (data not shown), indicating that the allele is lethal at early embryonic stages.

c. The graphs on wing size etc should start at zero.

Thanks. We corrected this in the current ms.

d. The sizes of histograms and graphs in each figure should be increased so that the reader can properly assess them. Currently, they are very small.

Thanks. We changed the sizes in the current ms.

The authors' model is that Dally (not Dlp) is required for Dpp distribution and signaling but that this is not due to a direct interaction with Dpp. Rather, they posit that Dally-HS antagonize Tkv-mediated Dpp internalization. Currently the results of the experiments could be considered consistent with their model, but as noted above, the lack of statistical analyses of some parameters is a weakness.

Thanks. We now performed and provided the statistical analyses in the revised ms.

One problematic part of their result for me is the role of the Dally core protein (Fig. 7B). There is a mis-match between the over-expression results and Dally allele lacking HS (but containing the core). Finally, their results support the idea that one or more as-yet unidentified proteins interact with Dally-HS chains to control Dpp distribution and signaling in the wing disc.

Our results simply suggest that Dpp can interact with Dally mainly through core protein but this interaction is not sufficient to sustain extracellular Dpp gradient formation under physiological conditions (dallyDeltaHS) (Fig. 4Q). We find that the mis-match is not problematic if the role of Dally is not simply mediated through interaction with Dpp. We speculate that interaction of Dpp and core protein of Dally is transient and not sufficient to sustain the Dpp gradient without HS chains of Dally stabilizing extracellular Dpp distribution by blocking Tkv-mediated Dpp internalization.

There is much debate and controversy in the Dpp morphogen field. The generation of new, high quality alleles in this study will be useful to Drosophila community, and the results of this study support the concept that Tkv but not Dally regulate Dpp internalization. Thus the work could be impactful and fuel new debates among morphogen researchers.

Thanks.

The manuscript is currently written in a manner that really is only accessible to researchers who work on the Dpp gradient. It would be very helpful for the authors to re-write the manuscript and carefully explain in each section of the results (1) the exact question that will be asked, (2) the prior work on the topic, (3) the precise experiment that will be done, and (4) the predicted results. This would make the study more accessible to developmental biologists outside of the morphogen gradient and Drosophila communities.

Thanks. We modified texts and changed the order of Fig.5. We hope that the changes make this study more accessible to developmental biologists outside of the field.

Author Response

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

We thank the reviewers for their feedback. Our response and a summary of the changes made to the manuscript are shown below. In addition to the changes made in response to the reviewer’s comments, we made the following changes to improve the manuscript:

• We updated figures 8 and 9 using data with improved preprocessing and source reconstruction. We now also include graphical network plots. This helps in the cross method (figure 8 vs 9) and cross dataset (figure 9 vs 10) comparison.

• We added funding acknowledgments and a credit author statement.

Reviewer #1 (Public Review):

Summary:

These types of analyses use many underlying assumptions about the data, which are not easy to verify. Hence, one way to test how the algorithm is performing in a task is to study its performance on synthetic data in which the properties of the variable of interest can be apriori fixed. For example, for burst detection, synthetic data can be generated by injected bursts of known durations, and checking if the algorithm is able to pick it up. Burst detection is difficult in the spectral domain since direct spectral estimators have high variance (see Subhash Chandran et al., 2018, J Neurophysiol). Therefore, detected burst lengths are typically much lower than injected burst lengths (see Figure 3). This problem can be solved by doing burst estimation in the time domain itself, for example, using Matching Pursuit (MP). I think the approach presented in this paper would also work since this model is also trained on data in the time domain. Indeed, the synthetic data can be made more "challenging" by injecting multiple oscillatory bursts that are overlapping in time, for which a greedy approach like MP may fail. It would be very interesting to test whether this method can "keep up" as the data is made more challenging. While showing results from brain signals directly (e.g., Figure 7) is nice, it will be even more impactful if it is backed up with results obtained from synthetic data with known properties.

We completely agree with the reviewer that testing the methods using synthetic data is an important part of validating such an approach. Each of the original papers that apply these methods to a particular application do this. The focus of this manuscript is to present a toolbox for applying these methods rather than to introduce/validate the methods themselves. For a detailed validation of the methods, the reader should see the citations. For example, the following paper introduces the HMM as a method for oscillatory burst detection:

• A.J. Quinn, et al. “Unpacking transient event dynamics in electrophysiological power spectra”. Brain topography 32.6 (2019): 1020-1034. See figures 2 and 3 for an evaluation of the HMM’s performance in detecting single-channel bursts using synthetic data.

We have added text to paragraph 2 in section 2.5 to clarify this burst detection method has been validated using simulated data and added references.

I was wondering about what kind of "synthetic data" could be used for the results shown in Figure 8-12 but could not come up with a good answer. Perhaps data in which different sensory systems are activated (visual versus auditory) or sensory versus movement epochs are compared to see if the activation maps change as expected. We see similarities between states across multiple runs (reproducibility analysis) and across tasks (e.g. Figure 8 vs 9) and even methods (Figure 8 vs 10), which is great. However, we should also expect the emergence of new modes specific to sensory activation (say auditory cortex for an auditory task). This will allow us to independently check the performance of this method.

The following papers study the performance of the HMM and DyNeMo in detecting networks using synthetic data:

• D. Vidaurre, et al. “Spectrally resolved fast transient brain states in electrophysiological data”. Neuroimage 126 (2016): 81-95. See figure 3 in this paper for an evaluation of the HMM’s performance in detecting oscillatory networks using simulation data.

• C. Gohil, et al. “Mixtures of large-scale dynamic functional brain network modes”. Neuroimage 263 (2022): 119595. See figures 4 and 5 for an evaluation of DyNeMo performance in detecting overlapping networks and long-range temporal structure in the data.

We have added text to paragraph 2 in section 2.5 to clarify these methods have been well tested on simulated data and added references.

The authors should explain the reproducibility results (variational free energy and best run analysis) in the Results section itself, to better orient the reader on what to look for.

Considering the second reviewer’s comments, we moved the reproducibility results to the supplementary information (SI). This means the reproducibility results are no longer part of the main figures/text. However, we have added some text to help the reader understand what aspects indicate the results are reproducible in section 2 of the SI.

Page 15: the comparison across subjects is interesting, but it is not clear why sensory-motor areas show a difference and the mean lifetime of the visual network decreases. Can you please explain this better? The promised discussion in section 3.5 can be expanded as well.

It is well known that the frequency and amplitude of neuronal oscillations changes with age. E.g. see the following review: Ishii, Ryouhei, et al. "Healthy and pathological brain aging: from the perspective of oscillations, functional connectivity, and signal complexity." Neuropsychobiology 75.4 (2018): 151-161. We observe older people have more beta activity and less alpha activity. These changes are seen in time-averaged calculations, i.e. the amplitude of oscillations are calculated using the entire time series for each subject.

The dynamic analysis presented in the paper provides further insight into how changes in the time-averaged quantities can occur through changes in the dynamics of frequency-specific networks. The sensorimotor network, which is a network with high beta activity, has a higher fractional occupancy. This indicates the change we observe in time-average beta power may be due to a longer amount of time spent in the sensorimotor network. The visual network, which is a network with high alpha activity, shows reduced lifetimes, which can explain the reduced time-averaged alpha activity seen with ageing.

We hope the improved text in the last paragraph of section 3.5 clarifies this. It should also be taken into account that the focus of this manuscript is the tools rather than an in-depth analysis of ageing. We use the age effect as an example of the potential analysis this toolbox enables.

Reviewer #2 (Public Review):

Summary:

The authors have developed a comprehensive set of tools to describe dynamics within a single time-series or across multiple time-series. The motivation is to better understand interacting networks within the human brain. The time-series used here are from direct estimates of the brain's electrical activity; however, the tools have been used with other metrics of brain function and would be applicable to many other fields.

Strengths:

The methods described are principled, and based on generative probabilistic models.

This makes them compact descriptors of the complex time-frequency data.

Few initial assumptions are necessary in order to reveal this compact description.

The methods are well described and demonstrated within multiple peer-reviewed articles.

This toolbox will be a great asset to the brain imaging community.

Weaknesses:

The only question I had was how to objectively/quantitatively compare different network models. This is possibly easily addressed by the authors.

We thank the reviewer for his/her comments. We address the weaknesses in our response in the “Recommendations For The Authors” section.

Reviewer #1 (Recommendations For The Authors):

Figure 2 legend: Please add the acronym for LCMV also.

We have now done this.

Section 2.5.1 page 8: the pipeline is shown in Figure 4, not 3.

This has been fixed.

Reviewer #2 (Recommendations For The Authors):

This is a great paper outlining a resource that can be applied to many different fields. I have relatively minor comments apart from one.

How does one quantitatively compare network descriptors (from DyNeMo and TDE-HMM for example)? At the moment the word 'cleaner' (P17) is used, but is there any non-subjective way? (eg Free energy/ cross validation etc). At the moment it is useful that one method gives a larger effect size (in a comparison between groups).. but could the authors say something about the use of these methods as more/less faithful descriptors of the data? Or in other words, do all methods generate datasets (from the latent space) that can be quantitatively compared with the original data?

In principle, the variational free energy could be used to compare models. However, because we use an approximate variational free energy (an exact measure is not attainable) for DyNeMo and an exact free energy for the HMM, it is possible that any differences we see in the variational free energy between the HMM and Dynemo are caused by the errors in its approximation. This makes it unreliable for comparing across models. That said, we can still use the variational free energy to compare within models. Indeed, we use the variational free energy for quantitative model comparisons when we select the best run to analyse from a set of 10.

One viable approach for comparing models is to assess their performance on downstream tasks. In this manuscript, examples of downstream tasks are the evoked network response and the young vs old group difference. We argue a better performance in the downstream task indicates a more useful model within that context. This performance is a quantitative measure. Note, there is no straightforward answer to which is the best model. It is likely different models will be useful for different downstream tasks.

In terms of which model provides a more faithful description of the data. The more flexible generative model for DyNeMo means it will generate more realistic data. However, this doesn’t necessarily mean it’s the best model (for a particular downstream task). Both the HMM and DyNeMo provide complementary descriptions that can be useful.

We have clarified the above in paragraph 5 of section 4.

Other comments:

• Footnote 6 - training on concatenated group data seems to be important. It could be more useful in the main manuscript where the limitations of this could be discussed.

By concatenating the data across subjects, we learn a group-level model. By doing this, we pool information across all subjects to estimate the networks. This can lead to more robust estimates. We have moved this footnote to the main text in paragraph 1 of section 2.5 and added further information.

• In the TDE burst detection section- please expand on why/how a specific number of states was chosen.

As with the HMM dynamic network analysis, the number of states must be pre-specified. For burst detection, we are often interested in an on/off type segmentation, which can be achieved with a 2 state HMM. However, if there are multiple burst types, these will all be combined into a single ‘on’ state. Therefore, we might want to increase the number of states to model multiple burst types. 3 was chosen as a trade-off to stay close to the on/off description but allow the model to learn more than 1 burst type. We have added text discussing this in paragraph 4 of section 4.

• Normally the value of free energy is just a function of the data - and only relative magnitude is important. I think figures (eg 7c) would be clearer if the offset could be removed.

We agree only the relative magnitude is important. We added text clarifying this in section 2 of the SI. We think it would still be worthwhile to include the offset so that future users can be sure they have correctly trained a model and calculated the free energy.

• Related to the above- there are large differences in model evidence shown between sets. Yet all sets are the same data, and all parameter estimates are more or less the same. Could the authors account for this please (i.e. is there some other parameter that differentiates the best model in one set from the other sets, or is the free energy estimate a bit variable).

We would like to clarify only the model parameters for the best run are shown in the group-level analysis. This is the run with the lowest variational free energy, which is highlighted in red. We have now clarified this in the caption of each figure. The difference in free energy for the best runs (across sets) is relatively small compared to the variation across runs within a set. If we were to plot the model parameters for each of the 10 runs in a set, we would see more variability. We have now clarified this in section 2 of the SI.

Also note, the group analysis usually involves taking an average. Small differences in the variational free energy could reflect small differences in subject-specific model parameters, which are then averaged out, giving virtually identical group effects.

• And related once again, if the data are always the same, I wonder if the free-energy plots and identical parameter estimates could be removed to free up space in figures?

The reproducibility results have now been moved to the supplementary information (SI).

• When citing p-values please specify how they are corrected (and over what please eg over states, nodes, etc?). This would be useful didactically as I imagine most users will follow the format of the presentation in this paper.

We now include in the caption further details of how the permutation significance testing was done.

• Not sure of the value of tiny power maps in 9C. Would consider making it larger or removing it?

The scale of these power maps is identical to part (A.I). We have moved the reproducibility analysis to the SI, enlarged the figure and added colour bars. We hope the values are now legible.

• Figure 3. I think the embedding in the caption doesn't match the figure (+-5 vs +-7 lags). Would be useful to add in the units of covariance (cii).

The number of embeddings in the caption has been fixed. Regarding the units for the covariances, as this is simulated data there aren’t really any units. Note, there is already a colour bar to indicate the values of each element.

• Minimize variational free energy - it may be confusing for some readers that other groups maximize the negative free energy. Maybe a footnote?

We thank the reviewer for their suggestion. We have added a footnote (1).

• Final question- and related to the Magnetoencephalography (MEG) data presented. These data are projected into source space using a beamformer algorithm (with its own implicit assumptions and vulnerabilities). Would be interested in the authors' opinion on what is standing between this work and a complete generative model of the MEG data - i.e. starting with cortical electrical current sources with interactions modeled and a dynamic environmental noise model (i.e. packing all assumptions into one model)?

In principle, there is nothing preventing us from including the forward model in the generative model and training on sensor level MEG data. This would be a generative model starting from the dipoles inside the brain to the MEG sensors. This is under active research. If the reviewer is referring to a biophysical model for brain activity, the main barrier for this is the inference of model parameters. However, note that the new inference framework presented in the DyNeMo paper (Gohil, et al. 2022) actually makes this more feasible. Given the scope of this manuscript is to present a toolbox for studying dynamics with existing methods, we leave this topic as future work.