Author response:

Thank you for your thoughtful and constructive feedback on our manuscript. We greatly appreciate your recognition of the strengths in our work, particularly regarding the genetic evidence demonstrating the role of beta-adrenergic receptors in cavernous malformation (CCM) development and the therapeutic potential of beta-blocker drugs.

We acknowledge your concerns and have addressing them to improve the clarity and rigor of our study. Specifically:

For Reviewer 1:

(1) Figure Annotation: We will enhance the annotation of the figure panels to provide clearer and more detailed descriptions, ensuring that each panel is easily interpretable and contributes effectively to the overall narrative of the study.

(2) Baseline Control for Klf2 Expression: We will include this control (klf2 expression in control Morpholino-injected embryos) in our revised figures and text to provide a more complete context for the changes in klf2 expression observed in response to genetic loss of adrb1 in ccm2 morphants and ccm2-CRISPR embryos.

For Reviewer 2:

(3) Sample Sizes in Figures 1, 2, and 3: We agree that reporting sample sizes is crucial for the transparency and reproducibility of our findings. For Figures 1B and D, as well as Figures 2G and 3B, we will update the figure legends to include the number of embryos and adult fish in which lesions were scored.

(4) Figure 4 Concerns: Use of adrb1 Morphants: We will note that the adrb1 morphant shows similar hemodynamic changes to the adrb1 mutant and that the morphants did not prevent the mosaic klf2 expression in ccm2 CRISPR embryos. Thus supporting our conclusion that protection from CVP cavernomas in the adrb1 mutant are not due to altered hemodynamics blocking mosaic klf2 expression.

Author response:

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

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

In this manuscript, Day et al. present a high-throughput version of expansion microscopy to increase the throughput of this well-established super-resolution imaging technique. Through technical innovations in liquid handling with custom-fabricated tools and modifications to how the expandable hydrogels are polymerized, the authors show robust ~4-fold expansion of cultured cells in 96-well plates. They go on to show that HiExM can be used for applications such as drug screens by testing the effect of doxorubicin on human cardiomyocytes. Interestingly, the effects of this drug on changing DNA organization were only detectable by ExM, demonstrating the utility of HiExM for such studies. 

Overall, this is a very well-written manuscript presenting an important technical advance that overcomes a major limitation of ExM - throughput. As a method, HiExM appears extremely useful, and the data generally support the conclusions. 

Strengths: 

Hi-ExM overcomes a major limitation of ExM by increasing the throughput and reducing the need for manual handling of gels. The authors do an excellent job of explaining each variation introduced to HiExM to make this work and thoroughly characterize the impressive expansion isotropy. The dox experiments are generally well-controlled and the comparison to an alternative stressor (H2O2) significantly strengthens the conclusions. 

Weaknesses: 

(1) Based on the exceedingly small volume of solution used to form the hydrogel in the well, there may be many unexpanded cells in the well and possibly underneath the expanded hydrogel at the end of this. How would this affect the image acquisition, analysis, and interpretation of HiExM data? 

The hydrogel footprint covers approximately 5% of the surface within an individual well and only cells within this area are embedded in the polymerized hydrogel for subsequent processing steps. Cells that are outside of this footprint are not incorporated into the gel because these cells are digested by Proteinase K and washed away by the excess water exchange in the gel swelling step. Note that different cell types may require higher or lower concentrations of Proteinase K to adequately digest cells for expansion while maintaining fluorescence signal. Given the compatibility of HiExM with 96-well plates, this titration can be performed rapidly in a single experiment. Although cells outside of the hydrogel footprint are removed prior to imaging, we do occasionally observe Hoechst signal that appears to be underneath the gels. We believe this signal is likely from excess DNA from digested cells that was not fully washed out in the gel swelling step. This signal is both spatially and morphologically distinct from the nuclear signal of intact cells and it does not affect image acquisition, analysis, or data interpretation. 

(2) It is unclear why the expansion factor is so variable between plates (e.g., Figure 2H). This should be discussed in more detail. 

The variability in expansion factor across plates can likely be attributed to the small volume of gel solution (~250 nL) required for expansion within 96 well plates. Small variations in gel volume could impact gel polymerization compared to standard ExM gels. For example, gels in HiExM are more sensitive to evaporation because of the ~1000x reduced volume compared to standard expansion gel preparations, resulting in an increased air-liquid-interface. Evaporation in HiExM gels would increase monomer and cross linker concentrations, leading to variation in expansion factor across plates. We note that expansion factor is robust within well plates and that variance is slightly increased between plates. These considerations are discussed in the revised manuscript.

(3) The authors claim that CF dyes are more resistant to bleaching than other dyes. However, in Figure. S3, it appears that half of the CF dyes tested still show bleaching, and no data is shown supporting the claim that Alexa dyes bleach. It would be helpful to include data supporting the claim that Alexa dyes bleach more than CF dyes and the claim that CF dyes in general are resistant to bleaching should be modified to more accurately reflect the data shown. 

We did not show data using Alexa dyes because these fluorophores are highly sensitive to photobleaching using Irgacure and thus we could not obtain images. In contrast, some CF dyes are more robust to bleaching in HiExM including CF488A, CF568, and CF633 dyes.  We have recently adapted our protocol to PhotoExM chemistry which is compatible with a wider range of fluorophores as described by Günay et al. (2023) and as shown in Fig. S16.

(4) Related to the above point, it appears that Figure S11 may be missing the figure legend. This makes it hard to understand how HiExM can use other photo-inducible polymerization methods and dyes other than CF dyes.

We revised the legend for revised Fig. S11 (now Fig. S16) as follows: Example of a cell expanded in HiExM using Photo-ExM gel chemistry. Photo-ExM does not require an anoxic environment for gel deposition and polymerization, improving ease of use of HiExM. Mitochondria were stained with an Alexa 647 conjugated secondary antibody, demonstrating that HiExM is compatible with additional fluorophores when combined with Photo-ExM.

(5) The use of automated high-content imaging is impressive. However, it is unclear to me how the increased search space across the extended planar area and focal depths in expanded samples is overcome. It would be helpful to explain this automated imaging strategy in more detail. 

We imaged plates on the Opera Phenix using the PreciScan Acquisition Software in Harmony. In brief, each well is imaged at 5x magnification in the Hoechst channel to capture the full well at low resolution. Hoechst is used for this step given its signal brightness, ubiquity across established staining protocols, and spectral independence from most fluorophores commonly conjugated to secondary antibodies. Using this information, the microscope detects regions of interest (nuclei) based on criteria including size, brightness, circularity, etc. Finally, the positional information for each region is stored, and the microscope automatically images those regions at 63x magnification. The working distance for the objective used in this study is 600 µm which is sufficient to capture the entirety of expanded cells in the Z direction. This strategy minimizes offtarget imaging and allows robust image acquisition even in cultures with lower seeding density. A detailed description of the automated imaging strategy is included in the methods section of the revised manuscript.

(6) The general method of imaging pre- and post-expansion is not entirely clear to me. For example, on page 5 the authors state that pre-expansion imaging was done at the center of each gel. Is pre-expansion imaging done after the initial gel polymerization? If so, this would assume that the gelation itself has no effect on cell size and shape if these gelled but not yet expanded cells are used as the reference for calculating expansion factor and isotropy. 

Pre-expansion imaging is performed after staining is complete, but prior to the application of AcX, which is the first step of the HiExM protocol. Following staining and imaging, plates can be sealed with parafilm and stored at 4˚C for up to a week prior to starting the expansion protocol. We typically image 61 fields of view at the center of the well plate (where the gel will be deposited) to obtain sufficient pre-expansion images as shown in Figure 2b (left). After preexpansion imaging, we perform the HiExM protocol followed by image acquisition. We then tile all the images, as shown in Figure 2b, and compare tiled images from the same well pre- and post-expansion to manually identify the same cells. Comparisons of the pre- and postexpansion images of the same cell are used to calculate expansion factor and isotropy measurements as described. A detailed description of this process is included in the revised manuscript.

(7) In the dox experiments, are only 4 expanded nuclei analyzed? It is unclear in the Figure 3 legend what the replicates are because for the unexpanded cells, it says the number of nuclei but for expanded it only says n=4. If only 4 nuclei are analyzed, this does not play to the strengths of HiExM by having high throughput.

We performed the doxorubicin titration assay across four different well plates (n=4). For each condition, the total number of expanded nuclei measured was 118, 111, 110, 113, and 77 for DMSO, 1nM, 10nM, 100nM, and 1µM, respectively. For SEM calculations, we included the number of independent experiments to avoid underestimating error. We revised the Fig. 3 legend to include these experimental details.

(8) I am not sure if the analysis of dox-treated cells is accurate for the overall phenotype because only a single slice at the midplane is analyzed. It would be helpful to show, at least in one or two example cases, that this trend of changing edge intensity occurs across the whole 3D nucleus.  

For this analysis, the result is heavily dependent on the angle at which the edge of the nucleus intersects the image plane in the orthogonal view. For this reason, we opted to only use the optimal image plane for each nucleus. We repeated our analysis on an image using multiple optical sections to demonstrate this point. These new data are included as Fig. S11 of the revised manuscript.

(9) It would be helpful to provide an actual benchmark of imaging speed or throughput to support the claims on page 8 that HiExM can be combined with autonomous imaging to capture thousands of cells a day. What is the highest throughput you have achieved so far?  

The parameters that dictate imaging speed in HiExM include exposure time, z-stack height, and number of fluorophore channels. Depending on the signal intensity for a given channel, exposure times vary from 200ms to 1000ms. For z-stack height, we found that imaging 65 sections with 1µm spacing allowed for robust identification of each region of interest in the 5x pre-scan. As an example, collecting images for a full well plate (e.g., 20 images per well with 4 channels) requires approximately 24 hours of autonomous image acquisition using the Opera Phenix. Depending on cell size, this process yields imaging data for 1200 cells (1 cell per field of view) to 6000 cells (5 cells per field of view). Different autonomous imagers as well as improving staining techniques that increase signal:noise can be expected to significantly decrease the exposure time as it will reduce the number of z-stacks needed for each region.

Reviewer #2 (Public Review): 

Summary: 

In the present work, the authors present an engineering solution to sample preparation in 96well plates for high-throughput super-resolution microscopy via Expansion Microscopy. This is not a trivial problem, as the well cannot be filled with the gel, which would prohibit the expansion of the gel. A device was engineered that can spot a small droplet of hydrogel solution and keep it in place as it polymerizes. It occupies only a small portion of space at the center of each well, the gel can expand into all directions, and imaging and staining can proceed by liquid handling robots and an automated microscope. 

Strengths: 

In contrast to Reference 8, the authors' system is compatible with standard 96 well imaging plates for high-throughput automated microscopy and automated liquid handling for most parts of the protocol. They thus provide a clear path towards high-throughput ExM and highthroughput super-resolution microscopy, which is a timely and important goal. 

Weaknesses: 

The assay they chose to demonstrate what high-throughput ExM could be useful for, is not very convincing. But for this reviewer that is not important. 

We believe the data provide an example of the utility of HiExM that would benefit experiments that require many samples (e.g., conditions, replicates, timepoints, etc.) by enabling easier sample processing and autonomous acquisition of thousands of nanoscale images in parallel. The ability to generate large data sets also enables quantitative analysis of images with appropriate statistical power. The intention of this work is to provide a proof-of-concept example of the robustness, accessibility, and experimental design flexibility of HiExM.

Reviewer #3 (Public Review):

Summary: 

Day et al. introduced high-throughput expansion microscopy (HiExM), a method facilitating the simultaneous adaptation of expansion microscopy for cells cultured in a 96-well plate format. The distinctive features of this method include 1) the use of a specialized device for delivering a minimal amount (~230 nL) of gel solution to each well of a conventional 96-well plate, and 2) the application of the photochemical initiator, Irgacure 2959, to successfully form and expand the toroidal gel within each well.  

Strengths: 

This configuration eliminates the need for transferring gels to other dishes or wells, thereby enhancing the throughput and reproducibility of parallel expansion microscopy. This methodological uniqueness indicates the applicability of HiExM in detecting subtle cellular changes on a large scale. 

Weaknesses: 

To demonstrate the potential utility of HiExM in cell phenotyping, drug studies, and toxicology investigations, the authors treated hiPS-derived cardiomyocytes with a low dose of doxycycline (dox) and quantitatively assessed changes in nuclear morphology. However, this reviewer is not fully convinced of the validity of this specific application. Furthermore, some data about the effect of expansion require reconsideration. 

The application we chose was intended as a methods proof-of-concept that could enable future deep biological investigations using HiExM. We believe the data provide an example of the utility of HiExM for collecting thousands of nanoscale images that would benefit experiments that require many samples (e.g., conditions, replicates, timepoints, etc.). The ability to generate large data sets also enables quantitative analysis of images with appropriate statistical power. The intention of this experiment was to provide a proof-of-concept example of the robustness, accessibility, and experimental design flexibility of HiExM. 

The variability in expansion factor across plates can likely be attributed to the small volume (~250 nL) deposited by the device posts. Small variations in gel volume could impact gel polymerization compared to standard ExM gels. For example, HiExM gels are more sensitive to evaporation due to an increased air-liquid-interface because they are ~1000x smaller than standard expansion gel preparations. Evaporation in HiExM gels likely increases monomer and cross linker concentrations, leading to variation in expansion factor across plates. We note that expansion factor is robust within well plates and that the expansion factor can be more variable between plates, likely due to differences in gel volumes and evaporation. Future iterations of the platform are expected to control for these environmental conditions. These differences are discussed in the revised manuscript.

Recommendations for the authors:.

Reviewer #1 (Recommendations For The Authors):

(1) Please include a scale bar in Figure 3a.

A scale bar has been added to Figure 3a.

(2) Please show the data related to nuclear volume after dox treatment.

We have added a supplementary figure (Fig. S10) showing nuclear volume and sphericity for post-expansion nuclei as well as nuclear area and circularity for pre-expansion nuclei.

(3) I think it would be extremely helpful for the method as a whole if analysis code and files for device fabrication were made publicly available rather than upon request.

The analysis code has been included in the supplementary files as CM_Hoechst_Analysis_for publication.ipynb. Device design files are also available at the supplementary files link as hiExM_device.SLDPRT (96-well plate device) and MultiExM_24_July28_2022.SLDPRT (24-well plate device).

(4) Some details are missing from the methods, such as the concentration of AcX used for HiExM, the concentration of antibodies, etc. Related, how long does the photopolymerization take? Just the 60 seconds that the UVA light is on?

Additional protocol details are included in the methods section of the revised manuscript. The photopolymerization does only take 60 seconds.

Reviewer #2 (Recommendations For The Authors):

(1) The first three references are chosen a little strangely here. I suggest citing STED, SIM, and PALM/STORM from the original manuscripts here. Also, EM is technically not a super-resolution technique as it is within the resolution of electron beams. This reviewer would stay with light microscopy methods when discussing "super-resolution".

We removed the reference to EM and added citations to the original publications for SIM, STED, and STORM.

(2) The sentence after citation 4 is a little off in its meaning.

We have edited the sentence to improve clarity.

(3) It is highly useful and great that the authors include the observations on the effect of photopolymerization with Irgacure 2959 on dyes.

(4) In the discussion, the authors could mention new high NA silicone oil objectives that may further optimise the resolution in their scheme.

We added a sentence in the discussion to reflect this important point.

(5) The files for the manufacture of the HiExM devices must be in the supplementary data rather than available on request.

The Solidworks designs for the 96 and 24 well plate devices are included in the supplementary files as hiExM_device.SLDPRT and MultiExM_24_July28_2022.SLDPRT, respectively.

(6) It would be useful if the authors could discuss their thoughts on the high throughput processing of expansion factors in the data analysis routine.

We added details to the methods section describing how images are processed and analyzed.

Reviewer #3 (Recommendations For The Authors):

Major:

(1) In the experiments depicted in Figure 3, the authors attempted cellular phenotyping using hiPCS-derived cardiomyocytes treated with doxorubicin (dox). They addressed that the relative intensity of Hoechst at the nuclear periphery increased solely in post-expansion images, although this trend is not clearly evidenced in the provided data (e.g., DMSO control vs. 1 nM dox, Figure 3b). Moreover, this observed phenomenon lacks clear biological significance and may not be suitable as a demonstration for proof-of-concept (POC) acquisition. It is crucial to delineate the biological processes linked with the specific enhancement of DNA binding dye signals in the nuclear periphery and how to rule out the possibility of heterogeneous redistribution of nuclear components rather than enhancing resolution. For instance, if this change can be associated with a biological process such as DNA damage, quantitative detection of the accumulated proteins related to DNA repair, or the specific histone marks, may be more suitable and less susceptible to heterogeneous expansion factors. Additionally, the authors noted the absence of significant changes in nuclear volume, yet the corresponding data was not presented. Moreover, the application insufficiently demonstrated the HiExM's scalable feature employing various well plates. If only acquiring images of dozens of nuclei (Figure 3 legend, p15), a single well per condition would suffice. Therefore, it is necessary to elucidate why this application necessitates a 96-well format for demonstration purposes. The potential experimental design should also incorporate the requirement for well-to-well replication and the acquisition of features at the individual well level, rather than at the single-cell level. Also, related to Figure S10, whether outer gradient slope, but not inner gradient slope, is linked to apoptosis (Page 8, Line 2-4) remains unclear in the H2O2-treated cells.

We believe the data provide an example of the utility of HiExM that would benefit experiments that require many samples (e.g., conditions, replicates, timepoints, etc.) by enabling easier sample processing and autonomous acquisition of thousands of nanoscale images in parallel. The ability to generate large data sets also enables quantitative analysis of images with appropriate statistical power. The intention of this work is to provide a proof-of-concept example of the robustness, accessibility, and experimental design flexibility of the HiExM method. As discussed in the manuscript, dox treatment is associated with DNA damage, cellular stress, and apoptosis, and commonly observed at high dox concentrations (>200 nM) in in vitro studies using conventional microscopy. Our data suggest that cardiomyocytes exhibit sensitivity to lower concentrations of dox than previously anticipated. Although direct evidence specifically linking dox to increased DNA condensation at the nuclear periphery is limited, the known proapoptotic effects of dox strongly suggest that our observations correlate with these changes. We have now included the data analysis on nuclear morphology in revised Fig. S10. We agree that deeper biological interpretation of the observed changes in Hoechst signal upon dox treatment (or other cellular stressors such as H2O2) using HiExM and whether these changes are correlated with DNA damage or other cellular alterations remains an exciting future direction to develop a more sensitive platform for assessing drug responses.

For expanded samples, we performed the doxorubicin titration assay across four different well plates (n=4). For each condition, the total number of nuclei measured was 118, 111, 110, 113, and 77 for DMSO, 1nM, 10nM, 100nM, and 1µM, respectively. We apologize for the confusion with respect to the number of replicates and cells analyzed. For SEM calculations, we used the number of independent experiments to avoid underestimating error. 

(2) In Figure 2b, do the orange arrows indicate the same cell with a unique shape in both the pre- and post-expansion images? Additionally, in Figure 3b, why do the pre- and post-expansion nuclei exhibit such different global shapes? Considering that the gel may freely rotate within the well during expansion, it raises doubts about whether one can identify cells with consistent shapes in both the pre- and post-expansion images. Furthermore, this reviewer observed a similar issue regarding reproducibility among different well plates, as shown in Figure 2h. The panel illustrates that different plates yielded distinct populations of gel sizes. The expansion factors provided in the figure legend (page 13) ranged from 3.5x to 5.1x across gels, indicating a relatively large variation in expansion size. What is the reason behind these variations, and how can they be minimized? These variations could become critical when considering large-scale screening across multiple plates.

The orange arrow is intended to indicate the same cell with a unique shape in both the pre- and post-expansion images, albeit at a different orientation given that the gel is not fixed within the well. We agree that improved methods to identify the same cells pre- and post-expansion could facilitate error measurements. We have referenced recent methods that could be combined with HiExM to automate and improve error and distortion detection to the discussion of the revised manuscript. 

Fig. 2 illustrates the ability of HiExM to achieve reproducible gel formation with minimal error within gels, wells, and across plates, measurements consistent with proExM. While uniform within gels, the expansion factor is somewhat variable between gels and plates. We attribute these differences primarily to the small size of the gels, making them vulnerable to the effects of evaporation between experiments. We note this variability should be taken into consideration for studies where absolute length measurements between plates are important for biological interpretation. Future iterations of the platform that allow precise delivery of gel volumes and that minimizes environmental exposure are expected to improve the expansion factor reproducibility across plates to further enable the use of HiExM as a tool for high-throughput nanoscale imaging.

Minor:

(1) Considering the signal loss due to photobleaching and fluorophore dilution during expansion, protein imaging may occasionally lack the sensitivity required to detect subtle morphological changes in cellular machinery. This potential limitation should be addressed or discussed in the text.

A sentence reflecting this point has been added to the manuscript.

(2) On page 15, the figure legend for panel d states, "Heatmaps of nuclei in b showing..." However, it appears that the panel referred to in this sentence corresponds to panel c.

The typo has been fixed.

(3) The type of glass 96-well plate utilized in this study should be specified, as the quality of the product could impact the expansion results.

The supplier and product number of the well plate used in our study has been added to the methods section.

(4) In Figure S3, the raw pixel values of CF305 dye are exceptionally low. Is there a specific reason for the very low signals observed when using this dye?

CF® 350 (305 was a typo) does not excite well at 405 nm, which is the excitation wavelength for the channel we used.

Author response:

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

Reviewer #1 (Public Review):

In their manuscript, Gan and colleagues identified a functional critical residue, Tyr404, which when mutated to W or A results in GOF and LOF of TRPML1 activity, respectively. In addition, the authors provide a high-resolution structure of TRPML1 with PI(4,5)P2 inhibitor. This high-resolution structure also revealed a bound phospholipid likely sphingomyelin at the agonist/antagonist site, providing a plausible explanation for sphingomyelin inhibition of TRPML1.

This is an interesting study, revealing valuable additional information on TRPML1 gating mechanisms including effects on endogenous phospholipids on channel activity. The provided data are convincing. Some major open questions remain. The work will be of interest to a wide audience including industry researchers occupied with TRPML1 exploration as a drug target.

We appreciate reviewer #1’s positive comments and the specific points raised by this reviewer are addressed in our response to Recommendations For The Authors

Reviewer #2 (Public Review):

The transient receptor potential mucolipin 1 (TRPML1) functions as a lysosomal organelle ion channel whose variants are associated with lysosomal storage disorder mucolipidosis type IV. Understanding sites that allosterically control the TRPML1 channel function may provide new molecular moieties to target with prototypic drugs.

Gan et al provide the first high-resolution cryo-EM structures of the TRPML1 channel (Y404W) in the open state without any activating ligands. This new structure demonstrates how a mutation at a site some distance away from the pore can influence the channel's conducting state. However, the authors do not provide a structural analysis of the Y404W pore which would validate their open-state claims. Nonetheless, Gan et al provide compelling electrophysiology evidence which supports the proposed Y404W gain of function effect. The authors propose an allosteric mechanism with the following molecular details- the Y404 to W sidechain substitution provides extra van der Waals contacts within the pocket surrounded by helices of the VSD-like domain and causes S4 bending which in turn opens to the pore through the S4-S5 linker. Conversely, the author functionally demonstrates that an alanine mutation at this site causes a loss of function. Although the authors do not provide a structure of the Y404A mutation, they propose that the alanine substitution disrupts the sidechain packing and likely destabilizes the open conformation. TRPM1 channels are regulated by PIP2 species, which is related to their cell function. In the membrane of lysosomes, PI(3,5)P2 activates the channel, whereas PI(4,5)P2 found in the plasma membrane has inhibitory effects. To understand its lipid regulation, the authors solved a cryo-EM structure of TRPM1 bound to PI(4,5)P2 in its presumed closed state. Again, while the provided functional evidence suggests that PI(4,5)P2 occupancy inhibits TRPML1 current, the authors do not provide analysis of the pore which would support their closed state assertion. Within this same structure, the authors observe a density that may be attributed to sphingomyelin (or possibly phosphocholine). Using electrophysiology of WT and the Y404W channels, the authors report sphingomyelins antagonist effect on TRPML1 currents under low luminal (external) pH. Taken together, the results described in Gan et al provide compelling evidence for a gating (open, closed) mechanism of the TRPML1 pore which can be allosterically regulated by altered packing and lipid interactions within the VSDL.

We appreciate reviewer #2’s positive comments and constructive suggestions. We functionally demonstrated that the Y404A mutant is more stable in the closed state. We did not pursue the structure of this mutant as we expect its structure will be the same as the apo closed TRPML1. To verify the open conformation of the Y404W mutant and the closed conformation of PI(4,5)P2 –bound TRPML1, we analyzed the pore radii of our structures in the revision as suggested by the reviewer and compared them with open and closed pores from previously determined TRPML1. Some specific points raised by this reviewer are addressed in our response to Recommendations For The Authors

Reviewer #1 (Recommendations For The Authors):

(1) Mutations in TRPML1 cause Mucolipidosis type IV. One patient mutation reported earlier (Chen et al., 2014 https://pubmed.ncbi.nlm.nih.gov/25119295/) to be a LOF mutation is R403C. This mutation resides just next to the here-identified Y404 position which can be converted in either LOF or GOF. Another patient mutation, F408del (also reported previously: (Chen et al., 2014 https://pubmed.ncbi.nlm.nih.gov/25119295/)) results in a mild activity reduction, in particular of the PI(3,5)P2 effect. Can the authors please discuss their findings in the context of the reported literature on these patient mutations and provide explanations as to why this part of the TRPML1 protein seemingly is such a hotspot for mutations affecting channel activity and how they explain this based on their structural evidence? What characteristics would be required for a small molecule agonist of TRPML1 in order to elicit larger activation in these patient LOF mutations if possible?

We thank the reviewer for highlighting these mutations identified in human patients. R403 appears to play two key roles. Firstly, its side chain participates in stabilizing Y404 in the open state. Secondly, as demonstrated in our previous study on TRPML1 (PMID: 35131932), the R403 side chain points towards the PI(3,5)P2 binding pocket, where it forms a critical salt bridge with the C3 phosphate group in the open state. Therefore, R403C mutation likely abolishes PI(3,5)P2 activation and also destabilizes the open state, resulting in the loss of function of the channel. We have expanded our discussion on this mutation in the revision. F408 is positioned at the junction between S4 and the S4-S5 linker. Its deletion mutation could change the stability or the folding of the protein. It is difficult to speculate the exact cause of the F408Δ LOF based on the TRPML1 structure. We don’t feel the effect of this mutation is relevant to the findings of this study.

(2) The authors used ML-SA1 only as a basis for their claims. Could they possibly provide some key data also on alternative small molecule agonists such as SF-51 and/or MK6-83?

We thank the reviewer for this suggestion. The TRPML1 agonists such as  ML-SA1 (derived from SF-51) and  MK6-83 have been well characterized in previous studies. In our study of sphingomyelin effect on TRPML1 activity, we used SF-51 to activate the channel (Figure 4b). The goal of our study is to demonstrate that agonist and antagonist can still allosterically regulate the LOF Y404A and GOF Y404W mutant channels, respectively, and their competition with sphingomyelin. We chose ML-SA1 in our experiment simply because it has been a commonly used TRPML1 agonist and its binding has been structurally defined, allowing us to compare various TRPML1 structures with different ligands. We don’t feel the use of other agonists would add extra information to our findings.  

(3) Sphingomyelin effects on TRPML1 have been confirmed by other groups as well (see e.g. Prat Castro et al., 2022 https://pubmed.ncbi.nlm.nih.gov/36139381/ Fig.3. Interestingly TPC2 seems unaffected by sphingomyelin albeit it is also activated by PI(3,5)P2. Can the authors provide possibly some modeling and/or cryoEM data on TPC2 with sphingomyelin to potentially explain why TPC2 is seemingly unaffected by sphingomyelin?

We appreciate the reviewer for providing additional evidence of sphingomyelin's effects on TRPML1 and have included the reference in revision. The binding site and activation mechanism of PI(3,5)P2 are different between TPC2 and TRPML1. It is beyond the scope of this study and also too speculative to model sphingomyelin binding (if any) in TPC2 to explain its lack of effect on TPC2 activity.

Reviewer #2 (Recommendations For The Authors):

The findings from Gan et al provide structural insights into the allosteric regulation of TRPML1 channel gating. The authors have provided compelling and hard-won cryo-EM structural evidence of the channel regulation by PI4,5P2 and at sites that pack the gating pore of the VSDL (S4). However, as noted in the public review, the analysis of the cryo-EM structures that would support claims of open and closed channel states is woefully lacking. Additional information related to the functional results is required to evaluate the activation and inhibition kinetic effect of lipids and pharmacological agents used to support their allosteric mechanism of TRPML1 gating.

Major concerns:

(a) At the very least, the pore domains of the new channels (PDB 9CBZ and 9CC2) should be analyzed using the HOLE (or other) programs to estimate the distances along the ion-conducting pathway - are the structures wide enough to support the passage of hydrated or partially hydrated cations? Additional figure panels should provide this comparative analysis.

We thank the reviewer for this valuable suggestion. We have added a figure (Figure Supplement 3b) of pore domain radius analysis using the HOLE program in the revision and have also included the radius comparison with previous determined open and closed TRPML1 structures.  

(b) At the very least, all current traces (Figures 1C, 1D, 1E, 4B, 4C) should be accompanied by time course plots of current amplitudes. It is impossible to evaluate the authors' claims of lipid and drug effects on TRPML1 channels without this information.

The corresponding time course plots of current amplitudes have now been included in the revision as Figure Supplement 1 and 7.

(c) Regarding the gain of function Y404W mutation structure, the authors' allosteric mechanistic hypothesis centers on side chain packing details within the VSD-like domain S4 (which in turn opens the pore through the S4-S5 linker). However, the local resolution within the structures at this site is not described. To assess the veracity of these claims, at the very least, authors should provide electron density maps of this region, either in Figure 2 or in Figure Supplement 4.

We have included the electron density map and local resolution information surrounding the W404 residue from the Y404W GOF mutant structure in the revision as  Figure Supplement 3c.

Minor concerns:

(d) Additional evidence related to the identity of the pore domain-associated lipid density (PC or sphingomyelin), and its channel regulation would improve the manuscript. The authors examine sphingomyelin, but what is the functional impact of PC on TRPML1 currents? While this is suggested, it is at the authors' discretion whether or not to carry out this analysis.

We thank the reviewer for raising this question. Adding extra PC has no effect on TRPML1 activity. This is expected since PC is the major lipid component of the membrane.

(e) The manuscript is well written. However, a few errors were noted while reviewing this draft.

i. Line 138, (Figure F&G).

ii. Line 66, "signaling transduction".

These errors have now been corrected.

Author response:

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

Reviewer #1 (Public Review):

Summary:

This manuscript by Meissner and colleagues described a novel take on a classic social cognition paradigm developed for marmosets. The classic pull task is a powerful paradigm that has been used for many years across numerous species, but its analog approach has several key limitations. As such, it has not been feasible to adopt the task for neuroscience experiments. Here the authors capture the spirit of the classic task but provide several fundamental innovations that modernize the paradigm - technically and conceptually. By developing the paradigm for marmosets, the authors leverage the many advantages of this primate model for studies of social brain functions and their particular amenability to freely-moving naturalistic approaches.

Strengths:

The current manuscript describes one of the most exciting paradigms in primate social cognition to be developed in many years. By allowing for freely-moving marmosets to engage in high numbers of trials, while precisely quantifying their visual behavior (e.g. gaze) and recording neural activity this paradigm has the potential to usher in a new wave of research on the cognitive and neural mechanisms underlying primate social cognition and decision-making. This paradigm is an elegant illustration of how naturalistic questions can be adapted to more rigorous experimental paradigms. Overall, I thought the manuscript was well written and provided sufficient details for others to adopt this paradigm. I did have a handful of questions and requests about topics and information that could help to further accelerate its adoption across the field.

Weaknesses:

LN 107 - Otters have also been successful at the classic pull task (https://link.springer.com/article/10.1007/s10071-017-1126-2)

We have added this reference to the manuscript.

LN 151 - Can you provide a more precise quantification of timing accuracy than the 'sub-second level'. This helps determine synchronization with other devices.

We have included more precise timing details, noting that data is stored at the millisecond level.

Using this paradigm, the marmosets achieved more trials than in the conventional task (146 vs 10). While this is impressive, given that only ~50 are successful Mutual Cooperation trials it does present some challenges for potential neurophysiology experiments and particular cognitive questions. The marmosets are only performing the task for 20 minutes, presumably because they become sated and are no longer motivated. This seems a limitation of the task and is something worth discussing in the manuscript. Did the authors try other food rewards, reduce the amount of reward, food/water restrict the animals for more than the stated 1-3 hours? How might this paradigm be incorporated into in-cage approaches that have been successful in marmosets? Any details on this would help guide others seeking to extend the number of trials performed each day.

We have added a discussion addressing the use of liquid rewards, minimal food and water restriction, and the potential for further optimization to increase task engagement and trial numbers. This is now reflected in the revised manuscript.

Can you provide more details on the DLC/Anipose procedure? How were the cameras synchronized? What percentage of trials needed to be annotated before the model could be generalized? Did each monkey require its own model, or was a single one applied to all animals?

We have added more detailed information on the DLC and Anipose tracking which can be found in the Multi-animal 3D tracking section under Materials & Methods.

Will the schematics and more instructions on building this system be made publicly available? A number of the components listed in Table 1 are custom-designed. Although it is stated that CAD files will be made available upon request, sharing a link to these files in an accessible folder would significantly add to the potential impact of this paradigm by making it easier for others to adopt.

We have made the SolidWorks CAD files publicly available. They can now be found in the Github repository alongside the apparatus and task code.

In the Discussion, it would be helpful to have some discussion of how this paradigm might be used more broadly. The classic pulling paradigm typically allows one to ask a specific question about social cognition, but this task has the potential to be more widely applied to other social decision-making questions. For example, how might this task be adopted to ask some of the game-theory-type approaches common in this literature? Given the authors' expertise in this area, this discussion could serve to provide a roadmap for the broader field to adopt.

Although this paradigm was developed specifically for marmosets, it seems to me that it could readily be adopted in other species with some modifications. Could the authors speak to this and their thoughts on what may need to be changed to be used in other species? This is particularly important because one of the advantages of the classic paradigm is that it has been used in so many species, providing the opportunity to compare how different species approach the same challenge. For example, though both chimps and bonobos are successful, their differences are notably illuminating about the nuances of their respective social cognitive faculties.

We have expanded the discussion for the broader applications of this apparatus both for other decision-making research questions as well as its adaptability for use in other species.

Reviewer #2 (Public Review):

Summary:

This important work by Meisner et al., developed an automated apparatus (MarmoAPP) to collect a wide array of behavioral data (lever pulling, gaze direction, vocalizations) in marmoset monkeys, with the goal of modernizing collection of behavioral data to coincide with the investigation of neurological mechanisms governing behavioral decision making in an important primate neuroscience model. The authors show a variety of "proof-of-principle" concepts that this apparatus can collect a wide range of behavioral data, with higher behavioral resolution than traditional methods. For example, the authors highlight that typical behavioral experiments on primate cooperation provide around 10 trials per session, while using their approach the authors were able to collect over 100 trials per 20-minute session with the MarmoAAP.

Overall the authors argue that this approach has a few notable advantages:<br /> (1) it enhances behavioral output which is important for measuring small or nuanced effects/changes in behavior;<br /> (2) allows for more advanced analyses given the higher number of trials per session;<br /> (3) significantly reduces the human labor of manually coding behavioral outcomes and experimenter interventions such as reloading apparatuses for food or position;<br /> (4) allows for more flexibility and experimental rigor in measuring behavior and neural activity simultaneously.

Strengths:

The paper is well-written and the MarmoAPP appears to be highly successful at integrating behavioral data across many important contexts (cooperation, gaze, vocalizations), with the ability to measure significantly many more behavioral contexts (many of which the authors make suggestions for).

The authors provide substantive information about the design of the apparatus, how the apparatus can be obtained via a long list of information Apparatus parts and information, and provide data outcomes from a wide number of behavioral and neurological outcomes. The significance of the findings is important for the field of social neuroscience and the strength of evidence is solid in terms of the ability of the apparatus to perform as described, at least in marmoset monkeys. The advantage of collecting neural and freely-behaving behavioral data concurrently is a significant advantage.

Weaknesses:

While this paper has many significant strengths, there are a few notable weaknesses in that many of the advantages are not explicitly demonstrated within the evidence presented in the paper. There are data reported (as shown in Figures 2 and 3), but in many cases, it is unclear if the data is referenced in other published work, as the data analysis is not described and/or self-contained within the manuscript, which it should be for readers to understand the nature of the data shown in Figures 2 and 3.

(1) There is no data in the paper or reference demonstrating training performance in the marmosets. For example, how many sessions are required to reach a pre-determined criterion of acceptable demonstration of task competence? The authors reference reliably performing the self-reward task, but this was not objectively stated in terms of what level of reliability was used. Moreover, in the Mutual Cooperation paradigm, while there is data reported on performance between self-reward vs mutual cooperation tasks, it is unclear how the authors measured individual understanding of mutual cooperation in this paradigm (cooperation performance in the mutual cooperation paradigm in the presence or absence of a partner; and how, if at all, this performance varied across social context). What positive or negative control is used to discern gained advantages between deliberate cooperation vs two individuals succeeding at self-reward simultaneously?

Thank you for your comment. This Tools & Resources paper is focused solely on the development of the apparatus and methods. Future publications will provide more details on training performance, learning behaviors, and include appropriate controls to distinguish deliberate cooperation from simultaneous success in self-reward tasks.

(2) One of the notable strengths of this approach argued by the authors is the improved ability to utilize trials for data analysis, but this is not presented or supported in the manuscript. For example, the paper would be improved by explicitly showing a significant improvement in the analytical outcome associated with a comparison of cooperation performance in the context of ~150 trials using MarmoAAP vs 10-12 trials using conventional behavioral approaches beyond the general principle of sample size. The authors highlight the dissection of intricacies of behavioral dynamics, but more could be demonstrated to specifically show these intricacies compared to conventional approaches. Given the cost and expertise required to build and operate the MarmoAAP, it is critical to provide an important advantage gained on this front. The addition of data analysis and explicit description(s) of other analytical advantages would likely strengthen this paper and the advantages of MarmoAAP over other behavioral techniques.

Thank you for the suggestion. While this manuscript focuses on the apparatus and methods, the increase in trial numbers itself provides clear advantages, including greater statistical power and more robust analyses of behavioral dynamics. Future publications will offer more in-depth analyses comparing the performance and cooperation behavior observed with MarmoAAP, further demonstrating these analytical benefits.

Reviewer #3 (Public Review):

Summary:

The authors set out to devise a system for the neural and behavioral study of socially cooperative behaviors in nonhuman primates (common marmosets). They describe instrumentation to allow for a "cooperative pulling" paradigm, the training process, and how both behavioral and neural data can be collected and analyzed. This is a valuable approach to an important topic, as the marmoset stands as a great platform to study primate social cognition. Given that the goals of such a methods paper are to (a) describe the approach and instrumentation, (b) show the feasibility of use, and (c) quantitatively compare to related approaches, the work is easily able to meet those criteria. My specific feedback on both strengths and weaknesses is therefore relatively limited in scope and depth.

Strengths:

The device is well-described, and the authors should be commended for their efforts in both designing this system but also in "writing it up" so that others can benefit from their R&D.

The device appears to generate more repetitions of key behavior than other approaches used in prior work (with other species).

The device allows for quantitative control and adjustment to control behavior.

The approach also supports the integration of markerless behavioral analysis as well as neurophysiological data.

Weaknesses:

A few ambiguities in the descriptions are flagged below in the "Recommendations for authors".

The system is well-suited to marmosets, but it is less clear whether it could be generalized for use in other species (in which similar behaviors have been studied with far less elegant approaches). If the system could impact work in other species, the scope of impact would be significantly increased, and would also allow for more direct cross-species comparisons. Regardless, the future work that this system will allow in the marmoset will itself be novel, unique, and likely to support major insights into primate social cognition.

Thank you for this feedback. We have expanded the discussion to include how the apparatus could be adapted for use in other species, highlighting the potential modifications required, such as adjusting the size and strength of the servo motor and components. These changes would enable broader applications and facilitate cross-species comparisons.

Author response:

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

Reviewer #1 (Public Review):

Summary:

Understanding large-scale neural activity remains a formidable challenge in neuroscience. While several methods have been proposed to discover the assemblies from such large-scale recordings, most previous studies do not explicitly model the temporal dynamics. This study is an attempt to uncover the temporal dynamics of assemblies using a tool that has been established in other domains.

The authors previously introduced the compositional Restricted Boltzmann Machine (cRBM) to identify neuron assemblies in zebrafish brain activity. Building upon this, they now employ the Recurrent Temporal Restricted Boltzmann Machine (RTRBM) to elucidate the temporal dynamics within these assemblies. By introducing recurrent connections between hidden units, RTRBM could retrieve neural assemblies and their temporal dynamics from simulated and zebrafish brain data.

Strengths:

The RTRBM has been previously used in other domains. Training in the model has been already established. This study is an application of such a model to neuroscience. Overall, the paper is well-structured and the methodology is robust, the analysis is solid to support the authors' claim.

Weaknesses:

The overall degree of advance is very limited. The performance improvement by RTRBM compared to their cRBM is marginal, and insights into assembly dynamics are limited.

(1) The biological insights from this method are constrained. Though the aim is to unravel neural ensemble dynamics, the paper lacks in-depth discussion on how this method enhances our understanding of zebrafish neural dynamics. For example, the dynamics of assemblies can be analyzed using various tools such as dimensionality reduction methods once we have identified them using cRBM. What information can we gain by knowing the effective recurrent connection between them? It would be more convincing to show this in real data.

See below in the recommendations section.

(2) Despite the increased complexity of RTRBM over cRBM, performance improvement is minimal. Accuracy enhancements, less than 1% in synthetic and zebrafish data, are underwhelming (Figure 2G and Figure 4B). Predictive performance evaluation on real neural activity would enhance model assessment. Including predicted and measured neural activity traces could aid readers in evaluating model efficacy.

See below in the recommendations section.

Recommendations:

(1) The biological insights from this method are constrained. Though the aim is to unravel neural ensemble dynamics, the paper lacks in-depth discussion on how this method enhances our understanding of zebrafish neural dynamics. For example, the dynamics of assemblies can be analyzed using various tools such as dimensionality reduction methods once we have identified them using cRBM. What information can we gain by knowing the effective recurrent connection between them? It would be more convincing to show this in real data.

We agree with the reviewer that our analysis does not explore the data far enough to reach the level of new biological insights. For practical reasons unrelated to the science, we cannot further explore the data in this direction at this point, however, funding permitting, we will pick up this question at a later stage. The only change we have made to the corresponding figure at the current stage was to adapt the thresholds, which better emphasizes the locality of the resulting clusters.

(2) Despite the increased complexity of RTRBM over cRBM, performance improvement is minimal. Accuracy enhancements, less than 1% in synthetic and zebrafish data, are underwhelming (Figure 2G and Figure 4B). Predictive performance evaluation on real neural activity would enhance model assessment. Including predicted and measured neural activity traces could aid readers in evaluating model efficacy.

We thank the reviewer kindly for the comments on the performance comparison between the two models. We would like to highlight that the small range of accuracy values for the predictive performance is due to both the sparsity and stochasticity of the simulated data, and is not reflective of the actual percentage in performance improvement. To this end, we have opted to use a rescaled metric that we call the normalised Mean Squared Error (nMSE), where the MSE is equal to 1 minus the accuracy, as the visible units take on binary values. This metric is also more in line with the normalised Log-Likelihood (nLLH) metric used in the cRBM paper in terms of interpretability. The figure shows that the RTRBM can significantly predict the state of the visible units in subsequent time-steps, whereas the cRBM captures the correct time-independent statistics but has no predictive power over time.

We also thank the reviewer for pointing out that there is no predictive performance evaluation on the neural data. This has been chosen to be omitted for two reasons. First, it is clear from Fig. 2 that the (c)RBM has no temporal dependencies, meaning that the predictive performance is determined mostly by the average activity of the visible units. If this corresponds well with the actual mean activity per neuron, the nMSE will be around 0. This correspondence is already evaluated in the first panel of 3F. Second, as this is real data, we can not make an estimate of a lower bound on the MSE that is due to neural noise. Because of this, the scale of the predictive performance score will be arbitrary, making it difficult to quantitatively assess the difference in performance between both models.

(3) The interpretation of the hidden real variable $r_t$ lacks clarity. Initially interpreted as the expectation of $\mathbf{h}_t$, its interpretation in Eq (8) appears different. Clarification on this link is warranted.

We thank the reviewer kindly for the suggested clarification. However, we think the link between both values should already be sufficiently clear from the text in lines 469-470:

“Importantly, instead of using binary hidden unit states 𝐡[𝑡−1], sampled from the expected real valued hidden states 𝐫[𝑡−1], the RTRBM propagates these real-valued hidden unit states directly.”

In other words, both indeed are the same, one could sample a binary-valued 𝐡[𝑡-1] from the real-valued 𝐫[𝑡-1] through e.g. a Bernoulli distribution, where 𝐫[𝑡-1] would thus indeed act as an expectation over 𝐡[𝑡−1]. However, the RTRBM formulation keeps the real-valued 𝐫[𝑡-1] to propagate the hidden-unit states to the next time-step. The motivation for this choice is further discussed in the original RTRBM paper (Sutskever et al. 2008).

(4) In Figure 3 panel F, the discrepancy in x-axis scales between upper and lower panels requires clarification. Explanation regarding the difference and interpretation guidelines would enhance understanding.

Thank you for pointing out the discrepancy in x-axis scales between the upper and lower panels of Figure 3F. The reason why these scales are different is that the activation functions in the two models differ in their range, and showing them on the same scale would not do justice to this difference. But we agree that this could be unclear for readers. Therefore we added an additional clarification for this discrepancy in line 215:

“While a direct comparison of the hidden unit activations between the cRBM and the RTRBM is hindered by the inherent discrepancy in their activation functions (unbounded and bounded, respectively), the analysis of time-shifted moments reveals a stronger correlation for the RTRBM hidden units ($r_s = 0.92$, $p<\epsilon$) compared to the cRBM ($r_s = 0.88$, $p<\epsilon$)”

(5) Assessing model performance at various down-sampling rates in zebrafish data analysis would provide insights into model robustness.

We agree that we would have liked to assess this point in real data, to verify that this holds as well in the case of the zebrafish whole-brain data. The main reason why we did not choose to do this in this case is that we would only be able to further downsample the data. Current whole brain data sets are collected at a few Hz (here 4 Hz, only 2 Hz in other datasets), which we consider to be likely slower than the actual interaction speed in neural systems, which is on the order of milliseconds between neurons, and on the order of ~100 ms (~10 Hz) between assemblies. Therefore reducing the rate further, we expect to only see a reduction in quality, which we considered less interesting than finding an optimum. Higher rates of imaging in light-sheet imaging are only achievable currently by imaging only single planes (which defies the goal of whole brain recordings), but may be possible in the future when the limiting factors (focal plane stepping and imaging) are addressed. For completeness, we have now performed the downstepping for the experimental data, which showed the expected decrease in performance. The results have been integrated into Figure 4.

Reviewer #2 (Public Review):

Summary:

In this work, the authors propose an extension to some of the last author's previous work, where a compositional restricted Boltzmann machine was considered as a generative model of neuron-assembly interaction. They augment this model by recurrent connections between the Boltzmann machine's hidden units, which allow them to explicitly account for temporal dynamics of the assembly activity. Since their model formulation does not allow the training towards a compositional phase (as in the previous model), they employ a transfer learning approach according to which they initialise their model with a weight matrix that was pre-trained using the earlier model so as to essentially start the actually training in a compositional phase. Finally, they test this model on synthetic and actual data of whole-brain light-sheet-microscopy recordings of spontaneous activity from the brain of larval zebrafish.

Strengths:

This work introduces a new model for neural assembly activity. Importantly, being able to capture temporal assembly dynamics is an interesting feature that goes beyond many existing models. While this work clearly focuses on the method (or the model) itself, it opens up an avenue for experimental research where it will be interesting to see if one can obtain any biologically meaningful insights considering these temporal dynamics when one is able to, for instance, relate them to development or behaviour.

Weaknesses:

For most of the work, the authors present their RTRBM model as an improvement over the earlier cRBM model. Yet, when considering synthetic data, they actually seem to compare with a "standard" RBM model. This seems odd considering the overall narrative, and it is not clear why they chose to do that. Also, in that case, was the RTRBM model initialised with the cRBM weight matrix?

Thank you for raising the important point regarding the RTRBM comparison in the synthetic data section. Initially, we aimed to compare the performance of the cRBM with the cRTRBM. However, we encountered significant challenges in getting the RTRBM to reach the compositional phase. To ensure a fair and robust comparison, we opted to compare the RBM with the RTRBM.

A few claims made throughout the work are slightly too enthusiastic and not really supported by the data shown. For instance, when the authors refer to the clusters shown in Figure 3D as "spatially localized", this seems like a stretch, specifically in view of clusters 1, 3, and 4.

Thanks for pointing out this inaccuracy. When going back to the data/analyses to address the question about locality, we stumbled upon a minor bug in the implementation of the proportional thresholding, causing the threshold to be too low and therefore too many neurons to be considered.

Fixing this bug reduces the number of neurons, thereby better showing the local structure of the clusters. Furthermore, if one would lower the threshold within the hierarchical clustering, smaller, and more localized, clusters would appear. We deliberately chose to keep this threshold high to not overwhelm the reader with the number of identified clusters. We hope the reviewer agrees with these changes and that the spatial structure in the clusters presented are indeed rather localized.

Moreover, when they describe the predictive performance of their model as "close to optimal" when the down-sampling factor coincided with the interaction time scale, it seems a bit exaggerated given that it was more or less as close to the upper bound as it was to the lower bound.

We thank the reviewer for catching this error. Indeed, the best performing model does not lay very close to the estimated performance of an optimal model. The text has been updated to reflect this.

When discussing the data statistics, the authors quote correlation values in the main text. However, these do not match the correlation values in the figure to which they seem to belong. Now, it seems that in the main text, they consider the Pearson correlation, whereas in the corresponding figure, it is the Spearman correlation. This is very confusing, and it is not really clear as to why the authors chose to do so.

Thank you for identifying the discrepancy between the correlation values mentioned in the text and those presented in the figure. We updated the manuscript to match the correlation coefficient values in the figure with the correct values denoted in the text.

Finally, when discussing the fact that the RTRBM model outperforms the cRBM model, the authors state it does so for different moments and in different numbers of cases (fish). It would be very interesting to know whether these are the same fish or always different fish.

Thank you for pointing this out. Keeping track of the same fish across the different metrics makes sense. We updated the figure to include a color code for each individual fish. As it turns out each time the same fish are significantly better performing.

Recommendations:

Figure 1: While the schematic in A and D only shows 11 visible units ("neurons"), the weight matrices and the activity rasters in B and C and E and F suggest that there should be, in fact, 12 visible units. While not essential, I think it would be nice if these numbers would match up.

Thank you for pointing out the inconsistency in the number of visible units depicted in Figure 1. We agree that this could have been confusing for readers. The figure has been updated accordingly. As you suggested, the schematic representation now accurately reflects the presence of 12 visible units in both the RBM and RTRBM models.

Figure 3: Panel G is not referenced in the main text. Yet, I believe it should be somewhere in lines 225ff.

Thank you for mentioning this. We added in line 233 a reference to figure 3 panel G to refer to the performance of the cRBM and RTRBM on the different fish.

Line 637ff: The authors consider moments <v\_i h\_μ> and <v\_i h\_j>, and from the context, it seems they are not the same. However, it is not clear as to why because, judging from the notation, they should be the same.

The second-order statistic <v\_i h\_j> on line 639 was indeed already mentioned and denoted as <v\_i h\_μ> on line 638. It has now been removed accordingly in the updated manuscript.

I found the usage of U^ and U throughout the manuscript a bit confusing. As far as I understand, U^ is a learned representation of U. However, maybe the authors could make the distinction clearer.

We understand the usage of Û and U throughout the text may be confusing for the reader. However, we would like to notify the reviewer that the distinction between these two variables is explained in line 142: “in addition to providing a close estimate (̂Û) to the true assembly connectivity matrix U”. However, for added clarification to the reader, we added additional mentions of the estimated nature of Û throughout the text in the updated manuscript.

Equation 3: It would be great if the authors could provide some more explanation of how they arrived at the identities.

These identities have previously been widely described in literature. For this reason, we decided not to include their derivation in our manuscript. However, for completeness, we kindly refer to:

Goodfellow, I., Bengio, Y., & Courville, A. (2016). Chapter 20: Deep generative models [In Deep Learning]. MIT Press. https://www.deeplearningbook.org/contents/generative_models.html

Typos:

-  L. 196: "connectiivty" -> "connectivity"

-  L. 197: Does it mean to say "very strong stronger"?

-  L. 339: The reference to Dunn et al. (2016) should appear in parentheses.

-  L. 504f: The colon should probably be followed by a full sentence.

-  Eq. 2: In the first line, the potential V still appears, which should probably be changed to show the concrete form (-b * h) as in the second line.

-  L. 351: Is there maybe a comma missing after "cRBM"?

-  L. 271: Instead of "correlation", shouldn't it rather be "similarity"? - L. 218: "Figure 3D" -> "Figure 3F"

We thank the reviewer for pointing out these typos, which have all (except one) been fixed in the text. We do emphasize the potential V to show that there are alternative hidden unit potentials that can be chosen. For instance, the cRBM utilizes dReLu hidden unit potentials.

Reviewer #3 (Public Review):

With ever-growing datasets, it becomes more challenging to extract useful information from such a large amount of data. For that, developing better dimensionality reduction/clustering methods can be very important to make sense of analyzed data. This is especially true for neuroscience where new experimental advances allow the recording of an unprecedented number of neurons. Here the authors make a step to help with neuronal analyses by proposing a new method to identify groups of neurons with similar activity dynamics. I did not notice any obvious problems with data analyses here, however, the presented manuscript has a few weaknesses:

(1) Because this manuscript is written as an extension of previous work by the same authors (van der Plas et al., eLife, 2023), thus to fully understand this paper it is required to read first the previous paper, as authors often refer to their previous work for details. Similarly, to understand the functional significance of identified here neuronal assemblies, it is needed to go to look at the previous paper.

We agree that the present Research Advance has been written in a way that builds on our previous publication. It was our impression that this was the intention of the Research Advance format, as spelled out in its announcement "eLife has introduced an innovative new type of article – the Research Advance – that invites the authors of any eLife paper to present significant additions to their original research". In the previous formatting guidelines from eLife this was more evident with a strong limitation on the number of figures and words, however, also for the present, more liberal guidelines, place an emphasis on the relation to the previous article. We have nonetheless tried in several places to fill in details that might simplify the reading experience.

(2) The problem of discovering clusters in data with temporal dynamics is not unique to neuroscience. Therefore, the authors should also discuss other previously proposed methods and how they compare to the presented here RTRBM method. Similarly, there are other methods using neural networks for discovering clusters (assemblies) (e.g. t-SNE: van der Maaten & Hinton 2008, Hippocluster: Chalmers et al. 2023, etc), which should be discussed to give better background information for the readers.

The clustering methods suggested by the reviewer do not include modeling any time dependence, which is the crucial advance presented here by the introduction of the RTRBM, in extending the (c)RBM. In our previous publication on the cRBM (an der Plas et al., eLife, 2023), this comparison was part of the discussion, although it focussed on a different set of methods. While clustering methods like t-SNE, UMAP and others certainly have their value in scientific analysis, we think it might be misleading the reader to think that they achieve the same task as an RTRBM, which adds the crucial dimension of temporal dependence.

(3) The above point to better describe other methods is especially important because the performance of the presented here method is not that much better than previous work. For example, RTRBM outperforms the cRBM only on ~4 out of 8 fish datasets. Moreover, as the authors nicely described in the Limitations section this method currently can only work on a single time scale and clusters have to be estimated first with the previous cRBM method. Thus, having an overview of other methods which could be used for similar analyses would be helpful.

We think that the perception that the RTRBM performs only slightly better is based on a misinterpretation of the performance measure, which we have tried to address (see comments above) in this rebuttal and the manuscript. In addition we would like to emphasize that the structural estimation (which is still modified by the RTRBM, only seeded by the cRBMs output), as shown in the simulated data, makes improved structural estimates, which is important, even in cases where the performance is comparable (which can be the case if the RBM absorbs temporal dependencies of assemblies into modified structure of assemblies). We have clarified this now in the discussion.

Recommendations:

(1) Line 181: it is not explained how a reconstruction error is defined.

Dear reviewer, thanks for pointing this out. A definition of the (mean square) reconstruction error is added in this line.

(2) How was the number of hidden neurons chosen and how does it affect performance?

Thank you for pointing this out. Due to the fact that we use transfer learning, the number of hidden units used for the RTRBM is given by the number of hidden units used for training the cRBM. In further research, when the RTRBM operates in the compositional phase, we can exploit a grid search over a set of hyper parameters to determine the optimal set of hidden units and other parameters.

Author response:

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

Reviewing Editor (Recommendations For The Authors):

The revised manuscripts and rebuttal sufficiently answered all the questions raised by three Reviewers. Overall, the manuscript is well written and the results are clear based on a straightforward experiment in a pursuit of comparing dimeric PTH analog to 1-34 PTH analog, which has established clinical efficacy. The study's results are valuable as it utilized large animal models, specifically examined the local bone integration effects, and demonstrated the comparable therapeutic efficacy of the new PTH analog to 1-34 PTH. However, the data did not convincingly show how the dimeric PTH analog overcomes the limitations of 1-34 PTH. I suggest that the discussion should focus more on the differences between the two analogs.

We sincerely appreciate your thorough review and valuable feedback. We have carefully considered your comments and would like to address them as follows:

“Regarding the results on the effect of dimeric R25CPTH(1-34) in the OVX mouse model (Noh et al., 2024), bone formation markers were increased in the dimeric R25CPTH(1-34) group compared to the rhPTH (1-34) group. Additionally, bone resorption markers were decreased in the rhPTH (1-34) group compared to the control group. However, no significant differences were observed in the dimeric R25CPTH(1-34) group. This suggests that the mechanism of action of the dimeric peptide differs from that of the wildtype peptide. Furthermore, based on unpublished data comparing mRNA expression in bone and kidney tissues between the dimeric R25CPTH(1-34) and rhPTH (1-34) treated groups, we strongly believe that dimeric R25CPTH(1-34) exhibits distinct biological activity from rhPTH (1-34). These differences may arise from variations in PTH receptor binding, involvement of different G protein subtypes, or downstream intracellular signaling pathways.

The distinct effects of dimeric R25CPTH(1-34) and rhPTH (1-34) on osteoblasts and osteoclasts could indicate that while remodeling-based osteogenesis has a limited clinical use period, the dimeric form might promote sustained bone formation and increased bone density over a longer duration. Given that patients with this mutation, who have been exposed to the mutant dimer throughout their lives, exhibit high bone density, this suggests significant potential for dimeric R25CPTH(1-34) as a novel therapeutic option alongside wildtype PTH.” (Discussion section 2nd paragraph)

A few minor points I 'd like to point out. This line number is based on a Word file.

Line 146-148 - However, both were insufficient compared to the control group and did not illustrate any bone filling. The measured bone-implant contact ratio was 18.32 {plus minus} 16.19% for the control group, 48.13 {plus minus} 29.81% for the group, and 39.53 {plus minus} 26.17% (P < 0.05).

- Does it mean that bone generation of both treatment group is inferior to the control group? please specify which groups the values are belong to and between which groups P-value compare.

Thank you very much for your suggestion to improve the manuscript. We have recognized the previous omission and have revised the sentence clearly as follows.

"The measured bone–implant contact ratio was 18.32 ± 16.19% for the control group, 48.13 ± 29.81% for the rhPTH(1-34) group, and 39.53 ± 26.17% for the dimeric R25CPTH(1-34) group, illustrating the significant improvement in osseointegration. (P < 0.05 for the control group compared to both PTH groups; however, the difference between the PTH groups was not significant.)"

Line 157 - incompleteness over the same period. The rhPTH(1-34) group exhibited a mature trabecularcfghnc

- Please correct misspellings.

As the reviewer mentioned, I have corrected "trabecularcfghnc" to "trabecular." Thank you.

Line 165-168 and Figure 4 M-N - Both the rhPTH(1-34) and dimeric R25CPTH(1-34) groups showed a significantly higher number of TRAP+ cells at both bone defects, with and without a xenograft, compared to the control group (Figure 4M,N). (P < 0.05) In addition, the number of TRAP+ cells in the dimeric R25CPTH(1-34)group was significantly higher than in the vehicle, yet lower than in the rhPTH(1-34) group (Figure 4M,N).)

- I believe the heading of figure 4M-N should be changed to with or without xenograft. And maybe you want to explain the significant difference of TRAP positive cells between two groups (with vs. without xenograft). Minor point: was - were

We totally agree with reviewer’s comment. We changed figure 4. Also, based on the revised figure, the figure legends for figure 4 were also revised as follows. “The number of TRAP-positive cells in the mandible with and without xenograft in the rhPTH(1-34) and dimeric R25CPTH(1-34)-treated beagle groups.” Following the reviewer's comments, the be verb in the sentences in the results section was changed from ‘was’ to ‘were’. “The capability of rhPTH(1-34) and dimeric R25CPTH(1-34) in bone remodeling were evaluated by tartrate-resistant acid phosphatase (TRAP) immunohistochemical staining.”

Line 182-186 - This study investigated the therapeutic effects of rhPTH(1-34) and dimeric R25CPTH(1-34) on bone regeneration and osseointegration in a large animal model with postmenopausal osteoporosis. rhPTH(1-34) and dimeric R25CPTH(1-34) have shown significant clinical efficacy, and although there have been a few studies investigating their effects on bone regeneration in rodents (Garcia et al., 2013), the authors in this study aimed to investigate the effects using a large animal model that more accurately mimics osteoporotic humans (Cortet, 2011).

- Please split the sentences for better clarity. In last sentence, I'm unsure what Cortet 2011 citation here is for. The statement should be written in the first person not the third person.

We appreciate your attention to detail, which has helped improve the clarity and accuracy of this manuscript. As per the reviewer's suggestion, I have reordered and changed the references to fit the content and revised the sentences to the first person.

“rhPTH(1-34) and dimeric R25CPTH(1-34) have shown significant clinical efficacy. Although there have been a few studies investigating their effects on bone regeneration in rodents (Garcia et al., 2013), we aimed to investigate these effects using a large animal model. We chose this model because it more accurately mimics osteoporotic humans (Jee and Yao, 2001).”

Line 196-197 - Furthermore, by demonstrating that dimeric R25CPTH(1-34) exhibits a distinct pharmacological profile different from rhPTH(1-34) but still provides a clear anabolic effect in the localized jaw region, the authors have shown that it may possess different potential therapeutic indications from rhPTH(1-34).

- This study does not include any pharmacological data. (Please cite reference). Again, I would suggest writing it in the first person. It sounds like you are reviewing someone else's work

Thank you for your insightful comments. We acknowledge that our study did not include pharmacological data. We have changed the sentence to clarify that the pharmacological profile information is derived from previous studies. A suitable citation was included to substantiate this assertion. As suggested, we have revised the statement in the first person to more accurately represent our own research and discoveries.

“Furthermore, we have shown that dimeric R25CPTH(1-34) has a distinct anabolic effect in the localized mandible region, which is comparable to that of rhPTH(1-34). Our findings indicate that dimeric R25CPTH(1-34) may have distinct potential therapeutic indications, as demonstrated by prior pharmacological studies  (Bae et al., 2016), which demonstrated that it possesses a distinct pharmacological profile from rhPTH(1-34).”

Line 201 - One of the potential clinical advantages of dimeric R25CPTH(1-34) is its partial agonistic effect in pharmacodynamics.

- it needs reference

Thank you for your insightful advice. As reviewer’s suggestion, we have included references as follows.

“Additionally, the potency of cAMP production in cells was lower for dimeric R25CPTH compared to monomeric R25CPTH, consistent with its lower PTH1R-binding affinity (Noh et al., 2024).”

Line 206-207 - Also, the effects of dimer were prominent, as we mentioned better bone formation than the control group

- But not compared with monomeric 1-34 PTH

We have revised the statement to more accurately reflect our findings.

“Also, the impact of dimeric R25CPTH(1-34) was notable, as we observed a noticeable improvement in bone formation when compared to the control group. However, these effects were not as strong as those of rhPTH(1-34). Both PTH analogs demonstrated enhanced anabolic effects around the titanium implants, promoting bone regeneration and remodeling.”

Line 224 - The authors have attributed this phenomenon to the unique anatomical characteristics observed in the jawbone.

- I would suggest writing it in the first person

We totally understood the reviewer’s comment. We have corrected the sentences as follows.

“The anabolic effects of both PTH analogs in this specific region may have been enhanced by the unique anatomical characteristics of the mandible, which we attribute to these improvements.”

Line 236 - The authors have attributed this phenomenon to the unique anatomical characteristics observed in the jawbone.

- This is outdated as the label of two year limit of Forteo use was lifted by FDA in 2021

Thank you for your valuable comments regarding the FDA’s decision to lift the two-year limit on Forteo (teriparatide) use in 2021. We have revised sentences to reflect this recent information in FDA guidelines as follows.

“Despite the FDA's decision to remove the two-year treatment limit in 2021, which opens possibilities for broader clinical applications, there are still numerous challenges that need to be addressed. There are ongoing concerns about the potential long-term effects of extended use, including accelerated bone remodeling, possible hypercalcemic conditions, and heightened bone resorption”

Line 380-382 - bone volume (TV; mm3), trabecular number (Tb.N; 1/mm), trabecular thickness (Tb. Th; um), trabecular separation (Tb.sp; µm).

- minor points- please superscript mm3, and change u -> µ

We appreciate reviewer’s detailed comments. We have corrected the part about unit display in figure legend.

Line 405-406 - following treatment with dimeric dimeric R25CPTH(1-34)

- please remove redundancy.

We removed dimeric duplication in the figure legend for figure 5 as follows.

“Figure 5. Measurement of biochemical Marker Dynamics in serum. The serum levels of calcium, phosphorus, P1NP, and CTX across three time points (T0, T1, T2) following treatment with dimeric R25CPTH(1-34), rhPTH(1-34) and control.”

Line 409-410 - CTX levels, associated with bone resorption, show no significant differences between groups.

- there is a missing figure identification. please specify relevant figure - I guess (E)

We appreciate the reviewer's insightful comment regarding the missing figure identification in the sentence about CTX levels. After reviewing Figure 5, we have specified the relevant figure panel as follows:

“Figure 5. (A) The study timeline. (B-C) Calcium and phosphorus levels show an upward trend in response to both PTH treatments compared to control, indicating enhanced bone mineralization. (D) P1NP levels, indicative of bone formation, remain relatively stable across time and treatments. (E) CTX levels, associated with bone resorption, show no significant differences between groups.”

Author response:

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

eLife Assessment 

This study is a detailed investigation of how chromatin structure influences replication origin function in yeast ribosomal DNA, with focus on the role of the histone deacetylase Sir2 and the chromatin remodeler Fun30. Convincing evidence shows that Sir2 does not affect origin licensing but rather affects local transcription and nucleosome positioning which correlates with increased origin firing. However, the evidence remains incomplete as the methods employed do not rigorously establish a key aspect of the mechanism, fully address some alternative models, or sufficiently relate to prior results. Overall, this is a valuable advance for the field that could be improved to establish a more robust paradigm. 

We have added extensive new results to the manuscript that, we believe, address all three criticisms above, namely that the methods employed do not (1) rigorously establish a key aspect of the mechanism; (2) fully address some alternative models; or (3) sufficiently relate to prior results.

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

This paper presents a mechanistic study of rDNA origin regulation in yeast by SIR2. Each of the ~180 tandemly repeated rDNA gene copies contains a potential replication origin. Earlyefficient initiation of these origins is suppressed by Sir2, reducing competition with origins distributed throughout the genome for rate-limiting initiation factors. Previous studies by these authors showed that SIR2 deletion advances replication timing of rDNA origins by a complex mechanism of transcriptional de-repression of a local PolII promoter causing licensed origin proteins (MCMcomplexes) to re-localize (slide along the DNA) to a different (and altered) chromatin environment. In this study, they identify a chromatin remodeler, FUN30, that suppresses the sir2∆ effect, and remarkably, results in a contraction of the rDNA to about onequarter it's normal length/number of repeats, implicating replication defects of the rDNA. Through examination of replication timing, MCM occupancy and nucleosome occupancy on the chromatin in sir2, fun30, and double mutants, they propose a model where nucleosome position relative to the licensed origin (MCM complexes) intrinsically determines origin timing/efficiency. While their interpretations of the data are largely reasonable and can be interpreted to support their model, a key weakness is the connection between Mcm ChEC signal disappearance and origin firing.  

Criticism: The reviewer expressed concern about the connection between Mcm ChEC signal disappearance and origin firing.

To further support our claim that the disappearance of the MCM signal in our ChEC datasets reflects origin firing, we now present additional data using the well-established method of MCM Chromatin IP (ChIP).

(1) New Supporting Evidence:  ChIP at genome-wide origins. In Figure 5 figure supplement 2, we demonstrate that the Mcm2 ChIP signal in cells released into hydroxyurea (HU) is significantly reduced at early origins compared to late origins, which mirrors the pattern observed with the MCM2 ChEC signal. This reduction in the ChIP signal at early origins supports the interpretation that the MCM signal disappearance is associated with origin firing.

(2) New supporting based evidence:  ChIP at rDNA Origins. Our ChIP analysis also shows that the disappearance of the MCM signal at rDNA origins in sir2Δ cells released into HU is accompanied by signal accumulation at the replication fork barrier (RFB), indicative of stalled replication forks at this location (Figure 5 figure supplement 3). This pattern is consistent with the initiation of replication at these origins and fork stalling at the RFB.

(3) New supporting evidence:  2D gels with quantification. Furthermore, additional 2D gel electrophoresis results provide ample independent evidence of rDNA origin firing in HU in sir2Δ mutants and suppression of origin firing in sir2 fun30 cells. These new data include 1) quantification of 2D gels in Figure 4D and 2) new 2D gels presented in Figure 4C as described below in greater detail. Collectively, these results demonstrate that rDNA origins fire prematurely in HU in sir2 cells and that firing is suppressed by FUN30 deletion. These additional data reinforce our model and support the association between MCM signal disappearance and replication initiation.

While the cyclical chromatin association-dissociation of MCM proteins with potential origin sequences may be generally interpreted as licensing followed by firing, dissociation may also result from passive replication and as shown here, displacement by transcription and/or chromatin remodeling.

The reviewer raised a concern that the cyclical chromatin association-dissociation of MCM proteins could be interpreted as licensing followed by firing, but might also result from passive replication or displacement by transcription and chromatin remodeling.

Addressing Alternative Explanations:

(1) Selective Disappearance of MCM Complexes: While transcription and passive replication can indeed cause the MCM-ChEC signal to disappear, these processes cannot selectively cause the disappearance of the displaced MCM complex without also affecting the non-displaced MCM complex. Specifically, RNA polymerase transcribing C-pro would first need to dislodge the normally positioned MCM complex before reaching the displaced complex, which is not observed in our data.

(2) Role of FUN30 Deletion:  FUN30 deletion results in increased C-pro transcription and reduced disappearance of the displaced MCM complex. This observation supports our model, as transcription alone would not selectively affect the displaced MCM complex while leaving the normally positioned MCM complex unaffected.

(3) Licensing Restrictions: It is crucial to note that continuous replenishment of displaced MCMs with newly loaded MCMs is not possible in our experimental conditions, as the cells are in S phase and licensing is restricted to G1. This temporal restriction further supports our interpretation that the disappearance of the MCM signal reflects origin firing rather than alternative processes.

In summary, while alternative explanations such as transcription and passive replication could potentially account for MCM signal disappearance, our data indicate that these processes cannot selectively affect the displaced MCM complex without impacting the non-displaced complex. The selective disappearance observed in our experiments, along with the effects of FUN30 deletion and the temporal constraints on MCM loading, strongly support our interpretation that the disappearance of the MCM signal reflects origin firing.

Moreover, linking its disappearance from chromatin in the ChEC method with such precise resolution needs to be validated against an independent method to determine the initiation site(s). Differences in rDNA copy number and relative transcription levels also are not directly accounted for, obscuring a clearer interpretation of the results. 

The reviewer raised concerns about the need to validate the disappearance of MCM from chromatin observed using the ChEC method against an independent method to determine initiation sites. Additionally, they pointed out that differences in rDNA copy number and relative transcription levels are not directly accounted for, which may obscure the interpretation of the results.

(1) Reduced rDNA Copy Number promotes Early Replication: Copy number reduction of the magnitude caused by deletion of both SIR2 and FUN30 is not expected to suppress early rDNA replication in sir2, but rather to exacerbate it. Specifically, deletion of SIR2 and FUN30 causes the rDNA to shrink to approximately 35 copies. Kwan et al., 2023 (PMID: 36842087) have shown that a reduction in rDNA copy number to 35 copies results in a dramatic acceleration of rDNA replication in a SIR2+ strain. Therefore, the effect of rDNA size on replication timing reinforces our conclusion that deletion of FUN30 suppresses rDNA replication.

(2) New 2D Gels in sir2 and sir2 fun30 strains with equal number of rDNA repeats: To directly address the concern regarding differences in the number of rDNA repeats, we have included new 2D gel analyses in the revised manuscript. By using a fob1

background, we were able to equalize the repeat number between the sir2 and sir2 fun30 strains (Figure 4E). The 2D gels conclusively show that the suppression of rDNA origin firing upon FUN30 deletion is independent of both rDNA size and FOB1.

Nevertheless, this paper makes a valuable advance with the finding of Fun30 involvement, which substantially reduces rDNA repeat number in sir2∆ background. The model they develop is compelling and I am inclined to agree, but I think the evidence on this specific point is purely correlative and a better method is needed to address the initiation site question. The authors deserve credit for their efforts to elucidate our obscure understanding of the intricacies of chromatin regulation. At a minimum, I suggest their conclusions on these points of concern should be softened and caveats discussed. Statistical analysis is lacking for some claims. 

Strengths are the identification of FUN30 as suppressor, examination of specific mutants of FUN30 to distinguish likely functional involvement. Use of multiple methods to analyze replication and protein occupancies on chromatin. Development of a coherent model. 

Weaknesses are failure to address copy number as a variable; insufficient validation of ChEC method relationship to exact initiation locus; lack of statistical analysis in some cases. 

With regard to "insufficient validation of ChEC method relationship to exact initiation locus":  The two potential initiation sites that one would monitor (non-displaced and displaced) are separated by less than 150 base pairs, and other techniques simply do not have the resolution necessary to distinguish such differences. Indeed, our new ChIP results presented in Figure 5 figure supplement 3 clearly demonstrate that while the resolution of ChIP is adequate to detect the reduction of MCM signal at the replication initiation site and its relocation to the RFB ( ~2 kb away), it lacks the resolution required to differentiate closely spaced MCM complexes.

Furthermore, as we suggest in the manuscript, our results are consistent with a model in which it is only the displaced MCM complex that is activated, whether in sir2 or WT.  If no genotypedependent difference in initiation sites is even expected, it would be hard to interpret even the most precise replication-based assays.  

We appreciate the reviewer pointing out that some statistical analyses were lacking: we have added statistical analysis for 2D gels (Figures 4D and 4E),  EdU incorporation experiments in Figure 4F and disappearance of MCM ChEC and ChIP signal upon release of cells into HU (Figure 5 supplement 1 and Supplement 2).  

Additional background and discussion for public review: 

This paper broadly addresses the mechanism(s) that regulate replication origin firing in different chromatin contexts. The rDNA origin is present in each of ~180 tandem repeats of the rDNA sequence, representing a high potential origin density per length of DNA (9.1kb repeat unit). However, the average origin efficiency of rDNA origins is relatively low (~20% in wild-type cells), which reduces the replication load on the overall genome by reducing competition with origins throughout the genome for limiting replication initiation factors. Deletion of histone deacetylase SIR2, which silences PolII transcription within the rDNA, results in increased early activation or the rDNA origins (and reduced rate of overall genome replication). Previous work by the authors showed that MCM complexes loaded onto the rDNA origins (origin licensing) were laterally displaced (sliding) along the rDNA, away from a well-positioned nucleosome on one side. The authors' major hypothesis throughout this work is that the new MCM location(s) are intrinsically more efficient configurations for origin firing. The authors identify a chromatin remodeling enzyme, FUN30, whose deletion appears to suppress the earlier activation of rDNA origins in sir2∆ cells. Indeed, it appears that the reduction of rDNA origin activity in sir2∆ fun30∆ cells is severe enough to results in a substantial reduction in the rDNA array repeat length (number of repeats); the reduced rDNA length presumably facilitates it's more stable replication and maintenance. 

Analysis of replication by 2D gels is marginally convincing, using 2D gels for this purpose is very challenging and tricky to quantify. 

We address this criticism by carefuly quantifying 2 D gel results using single rARS signal for normalizing bubble arc as discussed below.

The more quantitative analysis by EdU incorporation is more convincing of the suppression of the earlier replication caused by SIR2 deletion. 

We have also added quantification of EdU results to strengthen our arguments.  

To address the mechanism of suppression, they analyze MCM positioning using ChEC, which in G1 cells shows partial displacement of MCM from normal position A to positions B and C in sir2∆ cells and similar but more complete displacement away from A to positions B and C in sir2fun30 cells. During S-phase in the presence of hydroxyurea, which slows replication progression considerably (and blocks later origin firing) MCM signals redistribute, which is interpreted to represent origin firing and bidirectional movement of MCMs (only one direction is shown), some of which accumulate near the replication fork barrier, consistent with their interpretation. They observe that MCMs displaced (in G1) to sites B or C in sir2∆ cells, disappear more rapidly during S-phase, whereas the similar dynamic is not observed in sir2∆fun30∆. This is the main basis for their conclusion that the B and C sites are more permissive than A. While this may be the simplest interpretation, there are limitations with this assay that undermine a rigorous conclusion (additional points below). The main problem is that we know the MCM complexes are mobile so disappearance may reflect displacement by other means including transcription which is high is the sir2∆ background. Indeed, the double mutant has greater level of transcription per repeat unit which might explain more displaced from A in G1. Thus, displacement might not always represent origin firing. Because the sir2 background profoundly changes transcription, and the double mutant has a much smaller array length associated with higher transcription, how can we rule out greater accessibility at site A, for example in sir2∆, leading to more firing, which is suppressed in sir2 fun30 due to greater MCM displacement away from A? 

I think the critical missing data to solidly support their conclusions is a definitive determination of the site(s) of initiation using a more direct method, such as strand specific sequencing of EdU or nascent strand analysis. More direct comparisons of the strains with lower copy number to rule out this facet. As discussed in detail below, copy number reduction is known to suppress at least part of the sir2∆ effect so this looms over the interpretations. I think they are probably correct in their overall model based on the simplest interpretation of the data but I think it remains to be rigorously established. I think they should soften their conclusions in this respect. 

Please see discussion below about these issues.

Reviewer #2 (Public Review): 

Summary: 

In this manuscript, the authors follow up on their previous work showing that in the absence of the Sir2 deacetylase the MCM replicative helicase at the rDNA spacer region is repositioned to a region of low nucleosome occupancy. Here they show that the repositioned displaced MCMs have increased firing propensity relative to non-displaced MCMs. In addition, they show that activation of the repositioned MCMs and low nucleosome occupancy in the adjacent region depend on the chromatin remodeling activity of Fun30. 

Strengths: 

The paper provides new information on the role of a conserved chromatin remodeling protein in the regulation of origin firing and in addition provides evidence that not all loaded MCMs fire and that origin firing is regulated at a step downstream of MCM loading. 

Weaknesses: 

The relationship between the author's results and prior work on the role of Sir2 (and Fob1) in regulation of rDNA recombination and copy number maintenance is not explored, making it difficult to place the results in a broader context. Sir2 has previously been shown to be recruited by Fob1, which is also required for DSB formation and recombination-mediated changes in rDNA copy number. Are the changes that the authors observe specifically in fun30 sir2 cells related to this pathway? Is Fob1 required for the reduced rDNA copy number in fun30 sir2 double mutant cells? 

We have conducted additional studies in the fob1 background to address how FOB1 and the replication fork barrier (RFB) influence the kinetics of rDNA size reduction upon FUN30 deletion (Figure 2 - figure supplement 2), rDNA replication timing (Figure 2 - figure supplement 3), and rDNA origin firing using 2D gels (Figure 4C).

Strains lacking SIR2 exhibit unstable rDNA size, and FOB1 deletion stabilizes rDNA size in a sir2 background (and otherwise). Similarly, we found that FOB1 deletion influences the kinetics of rDNA size reduction in sir2 fun30 cells. Specifically, we were able to generate a fob1 sir2 fun30 strain with more than 150 copies. Nonetheless, and consistent with our model, this strain still exhibited delayed rDNA replication timing (Figure 2 - figure supplement 3), and its rDNA still shrank upon continuous culture (Figure 2 figure supplement 2). These results demonstrate that, although FOB1 affects the kinetics of rDNA size reduction in sir2 fun30 strains, the reduced rDNA array size or delayed replication timing upon FUN30 deletion size does not depend on FOB1.

The use of the fob1 background allowed us to compare the activation of rDNA origins in sir2 and sir2 fun30 strains with equally short rDNA sizes. 2D gels demonstrate robust and reproducible suppression of rDNA origin activity upon deletion of FUN30 in sir2 fob1 strains with 35 rDNA copies (Figure 4C). These results indicate that the main effect we are interested in—FUN30-induced reduction in origin firing—is independent of both FOB1 and rDNA size.

Our additional studies conclusively show that the FUN30-induced reduction in rDNA origin firing is independent of both FOB1 and rDNA size. These findings provide important insights into the mechanisms regulating rDNA copy number maintenance, placing our results within the broader context of existing knowledge on Sir2 and Fob1 functions.

Reviewer #3 (Public Review): 

Summary: 

Heterochromatin is characterized by low transcription activity and late replication timing, both dependent on the NAD-dependent protein deacetylase Sir2, the founding member of the sirtuins. This manuscript addresses the mechanism by which Sir2 delays replication timing at the rDNA in budding yeast. Previous work from the same laboratory (Foss et al. PLoS Genetics 15, e1008138) showed that Sir2 represses transcription-dependent displacement of the Mcm helicase in the rDNA. In this manuscript, the authors show convincingly that the repositioned Mcms fire earlier and that this early firing partly depends on the ATPase activity of the nucleosome remodeler Fun30. Using read-depth analysis of sorted G1/S cells, fun30 was the only chromatin remodeler mutant that somewhat delayed replication timing in sir2 mutants, while nhp10, chd1, isw1, htl1, swr1, isw2, and irc3 had not effect. The conclusion was corroborated with orthogonal assays including two-dimensional gel electrophoresis and analysis of EdU incorporation at early origins. Using an insightful analysis with an Mcm-MNase fusion (Mcm-ChEC), the authors show that the repositioned Mcms in sir2 mutants fire earlier than the Mcm at the normal position in wild type. This early firing at the repositioned Mcms is partially suppressed by Fun30. In addition, the authors show Fun30 affects nucleosome occupancy at the sites of the repositioned Mcm, providing a plausible mechanism for the effect of Fun30 on Mcm firing at that position. However, the results from the MNAse-seq and ChEC-seq assays are not fully congruent for the fun30 single mutant. Overall, the results support the conclusions providing a much better mechanistic understanding how Sir2 affects replication timing at rDNA, 

The observation that the MNase-seq plot in fun30 mutant shows a large signal at the +3 nucleosome and somewhat smaller at position +2, while the ChEC-seq plot exhibits negligible signals, is indeed an important point of consideration. This discrepancy arises because most of the MCM in fun30 mutant remains at its original site where it abuts +1 nucleosome. As a result, the MCM-MNase fusion protein fails to reach and “light up” the +3 nucleosome, which is, nonetheless, well-visualized with exogenous MNase.  The paucity of displaced MCMs, which is responsible for cutting +2 nucleosome, explains the discrepancy in the +2 nucleosome signal between exogenous MNase and CheC datasets in the fun30 mutant.  

Despite this apparent discrepancy, the overall results support our conclusions and provide a much better mechanistic understanding of how Sir2 affects replication timing at rDNA. The MNaseseq data reflect nucleosome positioning and chromatin structure, while the ChEC-seq data specifically highlights the locations where MCM is bound and active.  

Strengths 

(1) The data clearly show that the repositioned Mcm helicase fires earlier than the Mcm in the wild type position. 

(2) The study identifies a specific role for Fun30 in replication timing and an effect on nucleosome occupancy around the newly positioned Mcm helicase in sir2 cells. 

Weaknesses 

(1) It is unclear which strains were used in each experiment. 

(2) The relevance of the fun30 phospho-site mutant (S20AS28A) is unclear. 

We appreciate the reviewer pointing out places in which our manuscript omitted key pieces of information (items 1 and 3), we have included the strain numbers in our revision.  With regard to point 2, we had written:  

Fun30 is also known to play a role in the DNA damage response; specifically, phosphorylation of Fun30 on S20 and S28 by CDK1 targets Fun30 to sites of DNA damage, where it promotes DNA resection (Chen et al. 2016; Bantele et al. 2017). To determine whether the replication phenotype that we observed might be a consequence of Fun30's role in the DNA damage response, we tested non-phosphorylatable mutants for the ability to suppress early replication of the rDNA in sir2; these mutations had no effect on the replication phenotype (Figure 2B), arguing against a primary role for Fun30 in DNA damage repair that somehow manifests itself in replication. 

(3) For some experiments (Figs. 3, 4, 6) it is unclear whether the data are reproducible and the differences significant. Information about the number of independent experiments and quantitation is lacking. This affects the interpretation, as fun30 seems to affect the +3 nucleosome much more than let on in the description. 

We have provided replicas and quantitation for the results in these figures.

(Replica ChEC Southern blot with quantification (Figure 3 figure supplement 1), quantification and replicas for 2D gels in Figure 4 and replicas for nucleosome occupancy (Figure 6 supplement 1).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

Fig. 3-Examination of MCM occupancy at the rDNA ARS region using a variation of ChEC.

Presumably these are these G1-arrested cells but does not seem to be stated. Please confirm. 

The 2D gels results are not very convincing of their conclusions. We are asked to compare bubble to fork arcs at 30 minutes, but this is not feasible. It is the author's job to quantify the data from multiple replicates, but none is given. After much careful examination, comparing the relative intensities of ascending bubble and Y-arcs, I think I can accept that 4A shows highest early efficiency for sir2 over WT and fun30, which are similar to each other, and lowest for sir2 fun30, at 60 and 90 min. 

In the revision we provide a careful quantification of the 2D gels in Figure 4. For assessing rDNA origin activity, we normalized the bubble arc during the HU time course to a single rARS signal, that appears as large 24.4kb Nhe1I fragment originating from the  rightmost rDNA repeat (see Figures 4A and 4B). The description of the quantification in the text is provided below. 

“Prior to separation on 2D gels, DNA was digested with NheI, which releases a 4.7 kb rARScontaining linear DNA fragment at the internal rDNA repeats (1N) and a much larger, 24.5 kb single-rARS-containing fragment originating from the rightmost repeat. In 2D gels, active origins generate replication bubble arc signals, whereas passive replication of an origin appears as a y-arc. Having a signal emanating from a single ARS-containing fragment simplifies the comparison of rDNA origin activity in strains with different numbers of rDNA repeats, such as in sir2 vs sir2 fun30 mutants. Origin activity is expressed as a ratio of the bubble to the single-ARS signal, effectively measuring the number of active rDNA origins per cell at a given time point. 

As seen previously (Foss et al. 2019), deletion of SIR2 increased the number of activated rDNA origins, while deletion of FUN30 suppressed this effect. When analyzed in aggregate at 20, 30, 60 and 90 minutes following release into HU, the average number of activated rDNA origin activity in sir2 mutant was increased 6.3-fold compared to those in WT (5.0±2.3 in sir2 vs 0.8±0.4 in wt, p<0.05 by 2 tailed t-test), and the increased number was reduced upon FUN30 deletion (1.3±0.7 in sir2 fun30, p<0.05 by 2 tailed t-test vs sir2, NS for comparison to WT).”

However, for part 4B, they state (p. 11) that deletion of FUN30 in a SIR2 background had no perceptible effect (on ARS305) but I think the data appear otherwise: the FUN30 cells show more Y-arc than WT.

We now provide the assessment of ARS305 activity in HU cells as a ratio of bubble-arc to 1N signal. The reviewer is right that FUN30 has a more robust bubble arc signal compared to WT.

However, after normalization to 1N this difference did not appear significant (3.7 vs 5.1). Overall the analysis of activity or ARS305 origins demonstrates a reciprocity with the activity of rDNA origins in each of the four genotypes.  Furthermore, this observation is confirmed in our EdU-based analysis of 111 genomic origins, with statistical analysis showing a very high level of significance (see below).  

Ultimately, analysis of unsynchronized cells would give unambiguous results about origin efficiency. In this regard I note that analysis of rDNA origin firing by 2D gels with HU versus asynchronous gives different results in WT versus sir2∆, with no difference in unsynchronized cells (He et al. 2022). It would be interesting to test the strains here unsynchronized, though copy number size would still be a variable to address.

Origin activity in log cultures is typically assessed by comparing replication initiation within an origin, presenting as a bubble arc, to passively replicated DNA (Y-arc). However, such an analysis at tandemly arrayed origins, such as rDNA, is not feasible, as both active and passive replication are the result of activation of the same origins. This explains the lack of difference between WT and sir2 cells previously reported (He et al. 2022), which we have also observed. Differences in activation of rDNA origins in WT vs sir2 cells is clearly reflected in HU experiments, as was the case in the earlier report (He et al. 2022). 

To address the issue of differences in copy number between sir2 and sir2 fun30 cells we have now done experiments in a fob1 background where we can equalize the copy number among the two genotypes. These 2D gels are presented in Figure 4C. We address this issue in the revised manuscript as follows:

“The overall impact of FUN30 deletion on rDNA origin activity in a sir2 background is expected to be a composite of two opposing effects: a suppression of rDNA origin activation and increased rDNA origin activation due to reduced rDNA size (Kwan et al. 2023). To evaluate the effect FUN30 on rDNA origin activation independently of rDNA size, we generated an isogenic set of strains in a fob1 background, all of which contain 35 copies of the rDNA repeat.  (Deletion of FOB1 is necessary to stabilize rDNA copy number.)  Comparing rDNA origin activity in sir2 versus sir2 fun30 genotypes, we observed a robust and reproducible reduction in rDNA origin activity upon FUN30 deletion. This finding confirms that the FUN30 suppresses rDNA origin firing in sir2 background independently of both rDNA size and FOB1 status.”

-EdU analysis is more convincing regarding relative effects on genome versus rDNA, however, again, the effect of reduced rDNA array size in the sir2 fun30 cells may also be the proximal cause of the reduced effect on genome (early origins) replication rather than a direct effect on origin efficiency. No statistic provided to support that fun30 suppresses sir2 for rDNA activity. 

This comment raises three distinct, but related, issues: 

First, the reviewer is asking whether the reduced rDNA size, of the magnitude we observed in sir2 fun30 cells, could by itself be responsible for increased origin activity elsewhere in the genome, just because there is less rDNA that needs to be replicated. As noted earlier (Kwan et al. 2023), Kwan et al. examined the effect of rDNA size reduction and observed: 1) marked increased in rDNA origin activity and 2) reciprocal reduction in origin activity elsewhere in the genome. This counterintuitive finding suggests that a smaller rDNA size exerts more competition for limited replication resources compared to a larger rDNA size. In light of this, our findings with FUN30 deletion become even more compelling. The suppression of rDNA firing upon FUN30 deletion is so significant that it overrides the expected effects of rDNA size reduction.

Second, the reviewer points out our lack of statistical analysis to support our contention that fun30 suppresses sir2 with regard to rDNA origin activity. We have now addressed this issue as well, by quantifying 2D gel signals, as described above in the text that begins with "Prior to separation on 2D gels, DNA was digested with NheI ...". 

Third, we have now provided a statistical analysis to support our conclusion that EdU-based analysis of activity of 111 early origins shows suppression upon deletion of SIR2 that is largely reversed by additional deletion of FUN30. 

"Deletion of FUN30 in a sir2 background partially restored EdU incorporation at early origins, concomitant with reduced EdU incorporation at rDNA origins. In particular, the median value of log10 of read depths at 111 early origins, as the data are shown in Figure 4F, dropped from 6.5 for wild type to 6.2 for sir2 but then returned almost to wild type levels (6.4) in sir2 fun30.  The p value obtained by Student's t test, comparing the drop in 111 origins from wild type to sir2 with that from wild type to sir2 fun30 was highly significant (<< 10-16)  In contrast, FUN30 deletion in the WT background did not reduce EdU incorporation at genomic origins (median 6.6). These findings highlight that FUN30 deletion-induced suppression of rDNA origins in sir2 is accompanied by the activation of genomic origins."

Use loss of Mcm-ChEC signal as proxy for origin firing. Reasonably convincing that decrease correlates with origin firing on a one-to-one basis (Fig. 5B), though no statistic given. 

We provide the statistical analysis in Figure 5-figure supplement 1.

However, there is no demonstration of ability to observe this correlation with fine resolution as needed for the claims here. It seems equally possible that sir2 deletion causes more firing by repositioning MCMs to a better location or that the prior location, which still contains substantial MCM, becomes more permissive. The MCM signal appears to be mobile, so perhaps the role of FUN30 is to prevent to mobility of MCM away from the original site in WT cells; note that significantly less Mcm signal is at the original position in sir2 fun30. No accumulation of MCM occurs near the RFB in WT (and fun30) cells. I understand that origin firing is lower in WT but raises concerns about sensitivity and dynamic range of this assay and that MCM positions may reflect transcription versus replication. 

Please see the section above labeled "Addressing Alternative Explanations".  

Is Fig 6A Y-axis correctly labeled? I understand this figure to represent MNase-seq reads; is there any Mcm2-ChEC-seq in part A? 

We have corrected the labeling. 6A represent MNase-seq reads. Thank you for pointing this out.

I understand part B to represent nucleosome-sized fragments released by Mcm2-ChEC interpreted to be nucleosomes. But could they be large fragments potentially containing adjacent MCM-double hexamers?  

Our representation of ChEC-seq data in Figure 1 supplement 1, where we can see the entire spectrum of fragment sizes, demonstrates two distinct populations of fragments: nucleosome size and MCM-size fragments.

Reviewer #2 (Recommendations For The Authors): 

Suggestions for the authors to consider: 

(1) The authors make a good case for the importance of replication balance between rDNA and euchromatin in ensuring that the genome is replicated in a timely fashion. This seems to be clearly regulated by Sir2. However, Sir2 also affects rDNA copy number and suppresses unequal cross over events, which are stimulates by Fob1. Does Fun30 suppress Fob1-dependent recombination events in sir2D cells? 

It is unclear why FUN30 only affects rDNA repeat copy number in sir2 cells. Why doesn't Fun30 reduce copy number in wild-type cells? 

Deletion of SIR2 causes rightward repositioning of MCMs to a position where they are more prone to fire, as shown by our HU ChEC datasets in which we show that the repositioned MCMs are more prone to activation than the non-repositioned ones. FUN30 deletion suppresses activation of these, activation-prone repositioned MCMs, as shown by HU ChEC. This suppression of rDNA origin activation in sir2 cells causes rDNA to shrink. In fun30 single mutants, due to the paucity of non-repositioned MCMs, we do not observe significant suppression of rDNA origin firing, and consequently, there is no reduction in rDNA size in fun30 cells.

(2) The authors use Mcm-MNase to map the location of the MCM helicase. Can these results be confirmed using the more standard and direct ChIP assay to examine changes in MCM localization

We carried out suggested MCM ChIP experiments and present these results in Figure 5 supplement 2 and supplement 3. These ChIP data demonstrate that: 

(1) MCM signal disappears preferentially at early origins compared to late origins, as seen in our ChEC results.

(2) The disappearance of ChEC signal at rDNA origins in sir2 mutant is accompanied by the signal accumulation at the RFB, consistent with fork stalling at the RFB mirroring the results we obtained by ChEC. While these results indicate that that ChIP has adequate resolution to detect MCM repositioning at 2 kb, scale, its resolution was insufficient for fine scale discrimination of repositioned and non-repositioned MCMs.

In this regard, the specific role of Fun30 in regulation of MCM firing at rDNA is interesting. 

Does Fun30 localize to the ARS region of rDNA? How is Fun30 specifically recruited to rDNA?  

We carried out ChIP for Fun30 and observed, similarly to previous reports (Durand-Dubief et al. 2012), a wide distribution of Fun30 throughout the genome and at rDNA. We have elected not to include these results in the current manuscript.

(3) The 2D gels in Figure 4 are difficult to interpret. The bubble to arc ratios in fun30D seem different from both wild-type and sir2D. It may be helpful to the reader to quantify the bubble to arc ratios. fun30D also seems to be affecting ARS305 by itself.

We provide quantification of 2 D gels in Figure 4.

(4) Figure 5. 

(4.1) For examining origin firing based on the disappearance of the Mcm-MNase reads, is HU arrest necessary? HU may be causing indirect effects due to replication fork stalling. In principle, the authors should be able to perform this analysis without HU, since their cells are released from synchronized arrest in G1 (and at least for the first cell cycle should proceed synchronously on to S phase). In addition, validation of Mcm-ChEC results using ChIP for one of the subunits of the MCM complex would increase confidence in the results. 

The HU arrest allows us to examine early events in DNA replication at much finer spatial and temporal resolution than it would be possible without it.

We have now used Mcm2 ChIP to confirm that the signal disappears at the MCM loading site in HU in sir2 cells as discussed above (Figure 5 figure supplement 3). However, the resolution is inadequate to discriminate non-repositioned vs repositioned MCMs.

(4.2) The non-displaced Mcm-ChEC signal in sir2D seems like it's decreasing more than in wildtype cells. Explain. It would be helpful to quantify these results by integrating the area under each peek (or based on read numbers). It looks like one of the displaced Mcm signals (the one more distal from the non-displaced) is changing at a similar rate to the non-displaced.  

Integrating the area under each Mcm-ChEC peak or using read numbers is superfluous for the following reasons:  (1) The rectangular appearance of the peaks in Figure 5 clearly reflects signal intensity, making additional numerical integration redundant. (2) The visual differences between wild-type and sir2D cells are distinct and sufficient for drawing conclusions without further quantification.  (3) Keeping the analysis straightforward avoids unnecessary complexity and maintains clarity.

(4.3) Can the authors explain why fun30D seems to be suppressing only one of the 2 displaced Mcms from firing? 

We speculate that the local environment is more conductive for firing one of two displaced MCMs, but we do not understand why.

(5) Figure 6. Why would the deletion of SIR2, a silencing factor, results in increased nucleosome occupancy at rDNA? 

If we understand correctly, the reviewer is referring to a small increase in +2 and +3 signal in sir2 compared to the WT. In WT G1 cells, there is a single MCM between +1 and +3 nucleosome. This space cannot accommodate a +2 nucleosome in G1 cells because MCM is loaded at that position in most cells (in G2 cells however, this space is occupied by a nucleosome (Foss et al., 2019). MCM repositioning in sir2 mutant would displace MCM from this location making it possible for this space to be now occupied by a nucleosome.

The changes in nuc density seem modest. Also, nucleosome density is similarly increased in sir2D and fun30D cells, but sir2 has a dramatic effect on origin firing but fun30D does not. Explain. 

We believe that the FUN30 status makes most of the difference for firing of displaced MCMs.

Since there are few displaced MCMs in SIR2 cells, there is not large impact on origin firing. Furthermore, the rDNA already fires late in WT cells, so our ability to detect further delay upon  FUN30 deletion could be more difficult.

(6) Discussion. At rDNA Sir2 may simply act by deacetylating nucleosomes and decreasing their mobility. This is unrelated to compaction which is usually only invoked regarding the activities of the full SIR complex (Sir2/3/4) at telomeres and the mating type locus. The arguments regarding polymerase size, compaction etc may not be relevant to the main point since although the budding yeast Sir2 participates in heterochromatin formation at the mating type loci and telomeres, at rDNA it may act locally near its recruitment site at the RFB. 

This is a valid point. We have added this sentence in the discussion to highlight the differences between silencing at rDNA and those at the silent mating loci and telomeres that SIR-complex dependent.

“Steric arguments such as these are even less compelling when made for rDNA than for the silent mating type loci and telomeres, because chromatin compaction has been studied mostly in the context of the complete Sir complex (Sir1-4). In contrast, Sir1, 3, and 4 are not present at the rDNA.”

Minor 

It would be interesting to see if deletion of any histone acetyltranferases acts in a similar way to Fun30 to reduce rDNA copy number in sir2D cells. 

Thank you for this suggestion.

Reviewer #3 (Recommendations For The Authors): 

(1) The design of Figure 3 could be improved. A scheme could help understand the assay without flipping back to Figure 1. The numbers below the gel bands need definition. 

We have included the scheme describing the restriction and MCM-MNase cut sites and the location of the probe for the Southern blot.

(2) The design of Figure 4 could be improved by adding a scheme to help interpret the 2d gel picture. The figure also lacks quantitation. Are the results reproducible and the differences significant? 

We have added the scheme, quantification and statistics in Figure 4.

(3) Please list in each figure legend the exact strains from Table S1 which were used. 

We have included the strain numbers in the Figure legend.

Durand-Dubief M, Will WR, Petrini E, Theodorou D, Harris RR, Crawford MR, Paszkiewicz K, Krueger F, Correra RM, Vetter AT et al. 2012. SWI/SNF-like chromatin remodeling factor Fun30 supports point centromere function in S. cerevisiae. PLoS Genet 8: e1002974.

Foss EJ, Gatbonton-Schwager T, Thiesen AH, Taylor E, Soriano R, Lao U, MacAlpine DM, Bedalov A. 2019. Sir2 suppresses transcription-mediated displacement of Mcm2-7 replicative helicases at the ribosomal DNA repeats. PLoS Genet 15: e1008138.

He Y, Petrie MV, Zhang H, Peace JM, Aparicio OM. 2022. Rpd3 regulates single-copy origins independently of the rDNA array by opposing Fkh1-mediated origin stimulation. Proc Natl Acad Sci U S A 119: e2212134119.

Kwan EX, Alvino GM, Lynch KL, Levan PF, Amemiya HM, Wang XS, Johnson SA, Sanchez JC, Miller MA, Croy M et al. 2023. Ribosomal DNA replication time coordinates completion of genome replication and anaphase in yeast. Cell Rep 42: 112161.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

This manuscript describes soluble Uric Acid (sUA) as an endogenous inhibitor of CD38, affecting CD38 activity and NAD+ levels both in vitro and in vivo. Importantly, the inhibition constants calculated support the claim that sUA inhibits CD38 under physiological conditions. These findings are of extreme importance to understanding the regulation of an enzyme that has been shown to be the main NAD+/NMN-degrading enzyme in mammals, which impacts several metabolic processes and has major implications for understanding aging diseases. The manuscript is well written, the figures are self-explanatory, and in the experiments presented, the data is very solid. The authors discuss the main limitations of the study, especially in regard to the in vivo results. As a whole, I believe that this is a very interesting manuscript that will be appreciated by the scientific community and that opens a lot of new questions in the field of metabolism and aging. I found some issues that I believe constitute a weakness in the manuscript, and although they do not require new experiments, they may be considered by the authors for discussion in the final version of the manuscript.

We greatly appreciate the reviewer’s thoughtful comments and favorable review of our work.

The authors acknowledge the existence of several previous papers involving pharmacological inhibition of CD38 and their impact on several models of metabolism and aging. However, they only cite reviews. Given the focus of the manuscript, I believe that the seminal original papers should be cited.

Yes, we agreed with the reviewer. Two representative papers regarding the pioneering findings [Ref 1, 2] of pharmacological inhibition of CD38 were cited in the discussion of current manuscript.

(1) Tarragó, M. G., et al. (2018). A Potent and Specific CD38 Inhibitor Ameliorates Age-Related Metabolic Dysfunction by Reversing Tissue NAD+ Decline. Cell Metab 27(5): 1081-1095.e1010.

(2) Escande, C., et al. (2013). Flavonoid apigenin is an inhibitor of the NAD+ ase CD38: implications for cellular NAD+ metabolism, protein acetylation, and treatment of metabolic syndrome. Diabetes 62(4): 1084-1093.

Related to the previous comment, the authors show that they have identified the functional group on sUA that inhibits CD38, 1,3-dihydroimidazol-2-one. How does this group relate with previous structures that were shown to inhibit CD38 and do not have this chemical structure? Is sUA inhibiting CD38 in a different site? A crystallographic structure of CD38-78c is available in PDB that could be used to study or model these interactions.

Currently, there are several kinds of CD38 inhibitors, including NAD+/NMN analogs, flavonoids, 4-quinolines, etc. [Ref 1], but they do not have 1,3-dihydroimidazol-2-one or similar groups. We also noticed that sUA and its analogs have no remarkable structural similarity with these inhibitors. We have ever tried to identify the binding sites of sUA on CD38 by NMR. Since our NMR method required a large sample size, we had to prepare recombinant human CD38 using a cell-free protein synthesis system. However, the obtained CD38 protein showed a lower Vmax than commercial recombinant CD38 expressed in HEK293 cells, raising a concern of spatial conformation deference in the synthesized CD38. Thus, we were unable to get convinced data to confirm if sUA has different binding sites. Given the difference in structural feature and inhibition type, we did not use the PDB data regarding 78c-CD38 interaction for analysis in this study.

(1) Chini, E. N., et al. (2018). The Pharmacology of CD38/NADase: An Emerging Target in Cancer and Diseases of Aging. Trends Pharmacol Sci 39(4): 424-436.

Although the mouse model used to manipulate sUA levels is not ideal, the authors discuss its limitations, and importantly, they have CD38 KO mice as control. However, all the experiments were performed in very young mice, where CD38 expression is low in most tissues (10.1016/j.cmet.2016.05.006). This point should be mentioned in the discussion and maybe put in the context of variations of sUA levels during aging.

We appreciate the reviewer’s kind suggestions. Yes, CD38 expression in young mice is relatively low and we used young mice in this study; thus, aged mice would be promising to furthest evaluate the interaction between CD38 and sUA. Regarding the changes in sUA levels during aging, previous reports indicate that sUA levels seem to increase with age in mice and humans [Ref 1, 2]. We speculate that this increase is a physiologically compensatory response to aging in organisms. Accordingly, we added more details in the discussion (second paragraph).

(1) Iwama, M., et al. (2012). Uric acid levels in tissues and plasma of mice during aging. Biol Pharm Bull 35(8): 1367-1370.

(2) Kuzuya, M., et al. (2002). Effect of aging on serum uric acid levels: longitudinal changes in a large Japanese population group. J Gerontol A Biol Sci Med Sci 57(10): M660-664.

Reviewer #2 (Public Review):

Summary:

This is an interesting work where Wen et al. aimed to shed light on the mechanisms driving the protective role of soluble uric acid (sUA) toward avoiding excessive inflammation. They present biochemical data to support that sUA inhibits the enzymatic activity of CD38 (Figures 1 and 2). In a mouse model of acute response to sUA and using mice deficient in CD38, they find evidence that sUA increases the plasma levels of nicotinamide nucleotides (NAD+ and NMN) (Figure 3) and that sUA reduces the plasma levels of inflammasome-driven cytokines IL-1b and IL-18 in response to endotoxin, both dependent on CD38 (Figure 4). Their work is an important advance in the understanding of the physiological role of sUA, with mechanistic insight that can have important clinical implications.

Strengths:

The authors present evidence from different approaches to support that sUA inhibits CD38, impacts NAD+ levels, and regulates inflammatory responses through CD38.

We deeply thank the reviewer for the thoughtful comments and appreciation of our findings.

Weaknesses:

The authors investigate macrophages as the cells impacted by sUA to promote immunoregulation, proposing that inflammasome inhibition occurs through NAD+ accumulation and sirtuin activity due to sUA inhibition of CD38. Unfortunately, the study still lacks data to support this model, as they could not replicate their in vivo findings using murine bone marrow-derived macrophages, a standard model to assess inflammasome activation. Without an alternative approach, the study lacks data to establish in vitro that sUA inhibition of CD38 reduces inflammasome activation in macrophages - consequently, they cannot determine yet if both NAD+ accumulation and sirtuin activity in macrophages is a mechanism leading to sUA role in vivo.

We deeply thank the reviewer for pointing out this weakness in our work. In fact, we tried to prepare stable CD38 KD/KO THP-1 cells in the middle of 2021; however, we faced some technical problems due to the limitations of instruments. Thus, we used CD38 KO mice to prepare CD38 KO BMDMs, as shown in the first version of manuscript, we failed to replicate the results in BMDMs because of the low uptake of sUA. To address the reviewer’s concern regarding the lack of an in vitro link between CD38 and sUA immunosuppression, we used 78c, a highly specific and potent inhibitor of CD38, to block CD38 in primed THP-1 cells. Then we evaluated the effect of sUA pre-incubation on MSU crystal-induced IL-1β release in primed THP-1 cells (vehicle and CD38 blockade). The added results in Figure 4-figure supplement 2B and 2C indicated that CD38 blockade largely impaired the immunosuppressive effect of sUA without reducing sUA uptake. In addition, we found that sUA at physiological levels boosted NAD+ levels in THP-1 cells (Figure 3-figure supplement 1B) without affecting the activities of other key enzymes involved in NAD+ synthesis and degradation, including NAMPT and PARP (Figure 3-figure supplement 2). All these results supported that CD38 is a key mediator for sUA at physiological levels to regulate inflammasome activation in vitro.

Reviewer #3 (Public Review):

Summary:

In the present manuscript, the authors propose that soluble Uric acid (sUA) is an enzymatic inhibitor of the NADase CD38 and that it controls levels of NAD modulating inflammatory response. Although interesting the studies are at this stage preliminary and validation is needed.

Strengths:

The study characterizes the potential relevance of sUA in NAD metabolism.

We greatly appreciate the reviewer for the thoughtful comments and valuable suggestions.

Weaknesses:

(1) A full characterization of the effect of sUA in other NAD-consuming and synthesizing enzymes is needed to validate the statement that the mechanism of regulation of NAD by sUA is mediated by CD38, The CD38 KO may not serve as the ideal control since it may saturate NAD levels already. Analysis of multiple tissues is needed.

Yes, it is necessary to confirm if sUA affects other NAD+-consuming and synthesizing enzymes. To address the concern and to provide additional validation, we tested the direct effects of sUA and other purine derivates on the activities of another two key enzymes involved in the metabolic network of NAD+, including PARP (NAD+-consuming enzyme) and NAMPT (NAD+-synthesizing enzyme). The added results in Figure 3-figure supplement 2 showed that sUA has no effect on PARP and NAMPT activity, suggesting that CD38 is a main target for sUA in regulating NAD+ availability. In addition, we also confirmed both PARP and NAMPT were not affected by purine metabolism under physiological conditions. Although hypoxanthine and xanthine, at 500 μM (supraphysiological levels), slightly inhibited PARP activity, it has no physiological significance due to their low physiological concentrations (generally below 20 μM). Further evaluation of these inhibitory effects under pathological conditions would be of interest but were beyond the focus of this study.

Given that tissue sUA uptake is saturated under physiological conditions (tissue sUA did not increase in our models, Figure 3-figure supplement 5A and 5B), CD38 and other potential targets in tissues may be not affected by sUA in our models. We used CD38 KO mice to confirm if sUA interacts with other targets to regulate NAD+ degradation and inflammatory responses. A previous study [Ref 1] revealed that inhibition of other enzymes involved in NAD+ metabolism, such as PARP, resulted in a significant increase of NAD+ availability in CD38 KO mice, which indicates that CD38 KO mice can be used to exclude the potential interaction between sUA and other targets. In fact, we did not observe significant effects of sUA in CD38 KO mice. More importantly, we added the additional validation regarding PARP and NAMPT activity according to the reviewer’s kind suggestion, which further confirmed that CD38 is the main target for sUA in our models.

(1) Tarragó, M. G., et al. (2018). A Potent and Specific CD38 Inhibitor Ameliorates Age-Related Metabolic Dysfunction by Reversing Tissue NAD+ Decline. Cell Metab 27(5): 1081-1095.e1010.

(2) The physiological role of sUA as an endogenous inhibitor of CD38 needs stronger validation (sUA deficient model?).

We thank the reviewer’s insightful suggestions. Yes, sUA depletion model is ideal for further validation, as we discussed in the limitations of this study. Given that introduction of exogenous recombinant uricase (immunometabolism may be affected) to deplete sUA is not ideal for the evaluation under physiological conditions, uricase-transgenic mice would be a promising model. However, now we have no uricase-transgenic mice, and we are unable to prepare CD38 KO/uricase-transgenic mice for additional validation within a reasonable time. In the first version of manuscript, therefore, we used an sUA-release model in sUA-supplementation mice as a further validation in Figure 3E.

(3) Flux studies would also be necessary to make the conclusion stronger.

Answer: We highly appreciate the reviewer’s suggestion regarding metabolic flux analysis. Yes, flux analysis using specifically designed isotope-labeled NAD+ is an ideal validation in mice, as it can track any sUA-induced changes in NAD+ metabolism. However, we are unable to synthesize or obtain suitable isotope-labeled substrates for in vivo validation due to the technical limitations and financial burdens.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

The manuscript is very solid and very well-presented and discussed. In my opinion, the only weakness in the writing is that the message about finding an endogenous regulator of CD38 activity and NAD levels gets blurred by the temptation of jumping into the potential of developing new pharmacological CD38 inhibitors based on sUA structure. My recommendation would be to focus on delivering a clear message about sUA as a physiological inhibitor of CD38, and the possible implications for understanding the onset and evolution of metabolic diseases and aging. Maybe leave the potential of developing novel sUA-based CD38 inhibitors for a final comment. I understand this last point is very attractive, but there are very potent pharmacological CD38 inhibitors already available with promising results.

We greatly appreciate the reviewer’s valuable suggestions. Yes, there are some promising CD38 inhibitors with nanomolar Ki such as 78c. To clearly focus on sUA as a physiological inhibitor of CD38, we simplified the description in the manuscript and just keep the discussion of functional group. Since the data regarding the development of CD38 inhibitors in our manuscript remain limited, we did not put it in a separate part. We believe the simplified information is still sufficient for medicinal chemists who are interested in the development of CD38 inhibitors.

Reviewer #2 (Recommendations For The Authors):

Major comments:

(1) The authors present several pieces of data to explain why there is no impact of sUA in inflammasome-mediated cytokine secretion, which is convincing and supports that BMDM will not be a model of choice to establish macrophages as the target of sUA. However, their data with THP-1 cells is compelling and could be further explored using shRNA, or CRISPR approaches to deplete CD38, thus establishing a mechanistic link in vitro.

We greatly appreciate the reviewer’s thoughtful comments. We added more results as mentioned before (please see our response to public review).

(2) There is a severe lack of linearity in how the figures are presented in the main text, making it difficult to read through the manuscript. The figures should be presented in the order they appear in the panels.

We thank the reviewer for pointing out this issue. We improved the figure assembly according to eLife guidelines.

Minor comments:

(1) The authors do not appropriately address the finding that CD38 impacts the secretion of IL-1b in BMDM (Figure S6B) and in vivo (Figure 4), possibly independently of sUA.

Yes, the regulatory effect of CD38 on cytokine release seems independent of sUA. In fact, we used CD38 KO BMDMs to validate the role of CD38 in the inflammasome activation. We showed that sUA levels are comparable between WT and KO mice, suggesting that CD38 KO does not affect the baselines and boosted levels of sUA in our models. In this situation, we were able to evaluate the immunosuppressive effects of sUA at the same physiological levels in WT and CD38 KO mice, thus providing evidence to support that sUA at physiological levels limits excessive inflammation via CD38.

(2) Figure 3F: the legend on the x-axis lacks the indication of which groups were treated with recombinant hCD38.

We appreciate the reviewer’s comments. We improved Figure 3 by adding more information.

(3) While the results on panels 3 and 4 provide robust evidence that sUA is anti-inflammatory through CD38, the title of the figures extrapolates their findings (i.e., no data shows CD38-sUA, either in vitro or in vivo).

We appreciate the reviewer’s kind suggestion. We provided data to support the direct interaction between CD38 and sUA in Figure 3F; we admitted that we did not show the data regarding the direct interaction in mice in Figure 4. To help readers easily track the results, however, we used a conclusion-like title.

(4) The introduction could briefly mention NMN and CD38 activity as an ecto-enzyme to facilitate the understanding of their findings by a general audience, especially the dosing of NMN and their data on BMDM.

We added more description regarding CD38 and NMN in the introduction part. Once again, we deeply thank the reviewer for the valuable suggestions.

Reviewer #3 (Recommendations For The Authors):

(1) A full characterization of the effect of sUA in other NAD-consuming and synthesizing enzymes is needed to validate the statement that the mechanism of regulation of NAD by sUA is mediated by CD38, The CD38 KO may not serve as the ideal control since it may saturate NAD levels already. Analysis of multiple tissues is needed.

We greatly appreciate the reviewer’s valuable suggestions. We added more results to validate CD38 as the main target of sUA in our models. Please see our response to public review.

(2) The physiological role of sUA as an endogenous inhibitor of CD38 needs stronger validation (sUA deficient model?).

We greatly appreciate the reviewer’s valuable suggestions. Please see our response to public review.

(3) Flux studies would also be necessary to make the conclusion stronger.

We greatly appreciate the reviewer’s valuable suggestions. Please see our response to public review.

Author response:

Reviewer #1 (Public Review):

Summary:

In previous work, the authors described necrosis-induced apoptosis (NiA) as a consequence of induced necrosis. Specifically, experimentally induced necrosis in the distal pouch of larval wing imaginal discs triggers NiA in the lateral pouch. In this manuscript, the authors confirmed this observation and found that while necrosis can kill all areas of the disc, NiA is limited to the pouch and to some extent to the notum, but is excluded from the hinge region. Interestingly and unexpectedly, signaling by the Jak/Stat and Wg pathways inhibits NiA. Further characterization of NiA by the authors reveals that NiA also triggers regenerative proliferation which can last up to 64 hours following necrosis induction. This regenerative response to necrosis is significantly stronger compared to discs ablated by apoptosis. Furthermore, the regenerative proliferation induced by necrosis is dependent on the apoptotic pathway because RNAi targeting the RHG genes is sufficient to block proliferation. However, NiA does not promote proliferation through the previously described apoptosis-induced proliferation (AiP) pathway, although cells at the wound edge undergo AiP. Further examination of the caspase levels in NiA cells allowed the authors to group these cells into two clusters: some cells (NiA) undergo apoptosis and are removed, while others referred to as Necrosis-induced Caspase Positive (NiCP) cells survive despite caspase activity. It is the NiCP cells that repair cellular damage including DNA damage and that promote regenerative proliferation. Caspase sensors demonstrate that both groups of cells have initiator caspase activity, while only the NiA cells contain effector caspase activity. Under certain conditions, the authors were also able to visualize effector caspase activity in NiCP cells, but the level was low, likely below the threshold for apoptosis. Finally, the authors found that loss of the initiator caspase Dronc blocks regenerative proliferation, while inhibiting effector caspases by expression of p35 does not, suggesting that Dronc can induce regenerative proliferation following necrosis in a non- apoptotic manner. This last finding is very interesting as it implies that Dronc can induce proliferation in at least two ways in addition to its requirement in AiP.

Strengths:

This is a very interesting manuscript. The authors demonstrate that epithelial tissue that contains a significant number of necrotic cells is able to regenerate. This regenerative response is dependent on the apoptotic pathway which is induced at a distance from the necrotic cells. Although regenerative proliferation following necrosis requires the initiator caspase Dronc, Dronc does not induce a classical AiP response for this type of regenerative response. In future work, it will be very interesting to dissect this regenerative response pathway genetically.

Weaknesses:

No weaknesses were identified.

We thank the reviewer for their positive evaluation and kind words.

Reviewer #2 (Public Review):

Summary / Strengths:

In this manuscript, Klemm et al., build on past published findings (Klemm et al., 2021) to characterize caspase activation in distal cells following necrotic tissue damage within the Drosophila wing imaginal disc. Previously in Klemm et al., 2021, the authors describe necrosis-induced-apoptosis (NiA) following the development of a genetic system to study necrosis that is caused by the expression of a constitutive active GluR1 (Glutamate/Ca2+ channel), and they discovered that the appearance of NiA cells were important for promoting regeneration.

In this manuscript, the authors aim to investigate how tissues regenerate following necrotic cell death. They find that the cells of the wing pouch are more likely to have non-autonomous caspase activation than other regions within the wing imaginal disc (hinge and notum),two signaling pathways that are known to be upregulated during regeneration, Wnt (wingless) and JAK/Stat signaling, act to prevent additional NiA in pouch cells, and may explain the region specificity, the presence of NiA cells promotes regenerative proliferation in late stages of regeneration, not all caspase-positive cells are cleared from the epithelium (these cells are then referred to as Necrosis-induced Caspase Positive (NiCP) cells), these NiCP cells continue to live and promote proliferation in adjacent cells, the caspase Dronc is important for creating NiA/NiCP cells and for these cells to promote proliferation. Animals heterozygous for a Dronc null allele show a decrease in regeneration following necrotic tissue damage.

The study has the potential to be broadly interesting due to the insights into how tissues differentially respond to necrosis as compared to apoptosis to promote regeneration.

Weaknesses:

However, here are some of my current concerns for the manuscript in its current version:

The presence of cells with activated caspase that don't die (NiCP cells) is an interesting biological phenomenon but is not described until Figure 5. How does the existence of NiCP cells impact the earlier findings presented? Is late proliferation due to NiA, NiCP, or both? Does Wg and JAK/STAT signaling act to prevent the formation of both NiA and NiCP cells or only NiA cells? Moreover, the authors are able to specifically manipulate the wound edge (WE) and lateral pouch cells (LP), but don't show how these manipulations within these distinct populations impact regeneration. The authors provide evidence that driving UAS-mir(RHG) throughout the pouch, in the LP or the WE all decrease the amount of NiA/NiCP in Figure 3G-O, but no data on final regenerative outcomes for these manipulations is presented (such as those presented for Dronc-/+ in Fig 7M). The manuscript would be greatly enhanced by quantification of more of the findings, especially in describing if the specific manipulations that impacted NiA /NiCP cells disrupt end-point regeneration phenotypes.

We thank the reviewer for their assessment and helpful suggestions to improve the manuscript. Regarding the presence of NiA and NiCP cells, and the proportion of each within a regenerating wing disc, unfortunately we are currently limited in our ability to distinguish each type of cell using available tools. This applies to both visualizing these cells via anti-cDcp-1 staining or the activity of GC3Ai, DBS-GFP and CasExpress, and detecting their function via their influence on proliferation. As such, although the identification of NiCP does not change any of the conclusions prior to Figure 5 in which NiCP are described, we are currently unable to comment on the contribution of NiA versus NiCP to late proliferation, or whether they are differently affected by Wg and JAK/STAT signaling. This issue is touched on in the discussion, but we will expand upon our commentary to better highlight these issues.

With respect to the reviewer’s suggestion to include evidence on whether blocking NiA/NiCP influences final regenerative outcomes, these data were published in our first paper on this work (Klemm et al., 2021, PMID: 34740246), which we will gladly reiterate in this work.

Finally, we agree that further quantification of our findings will strengthen the work, which is also suggested by Reviewer 3, and plan to add it where possible in a revised manuscript.

How fast does apoptosis take within the wing disc epithelium? How many of the caspase(+) cells are present for the whole 48 hours of regeneration? Are new cells also induced to activate caspase during this time window? The author presented a number of interesting experiments characterizing the NiCP cells. For the caspase sensor GC3Ai experiments in Figure 5, is there a way to differentiate between cells that have maintained fluorescent CG3Ai from cells that have newly activated caspase? What is the timeline for when NiA and NiCP are specified? In addition, what fraction of NiCP cells contribute to the regenerated epithelium? Additional information about the temporal dynamics of NiA and NiCP specification/commitment would be greatly appreciated.

Regarding the timing of apoptosis, Schott et al., 2017 (PMID:28870988) demonstrated that apoptotic GC3Ai-labeled cells in imaginal discs are extruded within 1 hr of labeling, the kinetics of which agree with previously published work on the temporal dynamics of apoptotic cell extrusion by Monier et al., 2015 (PMID:25607361). This is much faster than the continued labeling that we observe up to 64 hr post necrosis. We will include this information alongside a quantification of the percent of the wing pouch with GC3Ai-positive cells over time to better address whether the GC3Ai signal is maintained over time or if newly activated caspases account for the signal in late regenerating discs. We plan to include PH3 staining to distinguish between cells that have activated GC3Ai and are proliferating versus new caspase activity. Additionally, we plan to include new experimental evidence to evaluate the timing of GC3Ai-labelled apoptotic cell loss in our system.

The question of when NiA/NiCP are specified is difficult to address due to the issue of not being able to easily distinguish between these cell types. We previously attempted to answer this particular question, and also to determine what fraction of these cells contribute to the regenerated epithelium, using caspase-based lineage tracing with CasExpress. However, as shown in the paper, we are unable to label NiA/NiCP with CasExpress, either due to the lack of caspase activity level or subcellular localization. We are currently attempting to combine other caspase reporters with lineage tracing tools and examine late-stage wing discs to address these questions.

The notum also does not express developmental JAK/STAT, yet little NiA was observed within the notum. Do the authors have any additional insights into the differential response between the pouch and notum? What makes the pouch unique? Are NiA/NiCP cells created within other imaginal discs and other tissues? Are they similarly important for regenerative responses in other contexts?

As noted by Martin et al., 2017 (PMID:28935711), Martin & Morata, 2018 (PMID:29938762), and our own observations in Harris et al., 2016 (PMID:26840050), the notum does not respond to damage in a way that leads to regeneration, while the pouch does. As NiA/NiCP are a pro-regenerative response, we speculate that this intrinsic difference in regenerative capacity that is potentially caused by a different proliferative and genetic response to injury may account for the disparity in NiA/NiCP occurrence in the pouch vs the notum. A difference in the presence of the (currently unidentified) DAMPs or PRRs in notum vs pouch cells may also be responsible. Alternatively, since the hinge tissue is also refractory to NiA/NiCP due to the presence of genetic factors such as Wg and JAK/STAT, there may be an analogous pathway present in notum cells that acts to protect against the induction of pro-apoptotic factors. Indeed, caspase 3 activation does not seem to occur upon ablation of the notum (Bergantinos et al. 2010, PMID:20215351). We plan to add these points to the discussion.

Regarding the existence of NiA/NiCP in other contexts, we have additional data stemming from our clonal patch experiments (Figure S1) that demonstrates this phenomenon occurs in other imaginal discs, which we plan to include in the revised manuscript.

Reviewer #3 (Public Review):

The manuscript "Regeneration following tissue necrosis is mediated by non- apoptotic caspase activity" by Klemm et al. is an exploration of what happens to a group of cells that experience caspase activation after necrosis occurs some distance away from the cells of interest. These experiments have been conducted in the Drosophila wing imaginal disc, which has been used extensively to study the response of a developing epithelium to damage and stress. The authors revise and refine their earlier discovery of apoptosis initiated by necrosis, here showing that many of those presumed apoptotic cells do not complete apoptosis. Thus, the most interesting aspect of the paper is the characterization of a group of cells that experience mild caspase activation in response to an unknown signal, followed by some effector caspase activation and DNA damage, but that then recover from the DNA damage, avoid apoptosis, and proliferate instead. Many questions remain unanswered, including the signal that stimulates the mild caspase activation, and the mechanism through which this activation stimulates enhanced proliferation.

The authors should consider answering additional questions, clarifying some points, and making some minor corrections:

Major concerns affecting the interpretation of experimental results:

Expression of STAT92E RNAi had no apparent effect on the ability of hinge cells to undergo NiA, leading the authors to conclude that other protective signals must exist. However, the authors have not shown that this STAT92E RNAi is capable of eliminating JAK/STAT signaling in the hinge under these experimental conditions. Using a reporter for JAK/STAT signaling, such as the STAT-GFP, as a readout would confirm the reduction or elimination of signaling. This confirmation would be necessary to support the negative result as presented.

We thank the reviewer for their assessment and helpful suggestions to improve the manuscript. Although the knockdown of Stat92E using this RNAi line has been shown to produce phenotypes associated with loss of JAK/STAT signaling in previous papers (Monahan and Starz-Gaiano, 2014, 2016 PMID:26277564, 26993259), we agree it would be useful to demonstrate this in our hands and therefore plan to include these data.

Similarly, the authors should confirm that the Zfh2 RNAi is reducing or eliminating Zfh2 levels in the hinge under these experimental conditions, before concluding that Zfh2 does not play a role in stopping hinge cells from undergoing NiA.

We attempted to demonstrate the loss of Zfh2 using this RNAi line, but as noted by the reviewer the antibody staining appears mostly unchanged. A reduction in Zfh2 protein levels by this RNAi has previously been demonstrated in cells of the gut (Rojas Villa et al., 2019, PMID: 31841513), suggesting that the persistent Zfh2 staining we see could be due to perdurance of the Zfh2 protein, high levels of expression or high sensitivity of the Zfh2 antibody (or a combination of these factors). We plan to repeat the experiment using a longer knockdown duration prior to ablation to show a change in Zfh2, and/or test alternative RNAi lines. In the absence of these data, we will alter our conclusions to state that Zfh2 cannot be ruled out as playing a role in preventing NiA/NiCP formation in the hinge.

EdU incorporation was quantified by measuring the fluorescence intensity of the pouch and normalizing it to the fluorescence intensity of the whole disc. However, the images show that EdU fluorescence intensity of other regions of the disc, especially the notum, varied substantially when comparing the different genetic backgrounds (for example, note the substantially reduced EdU in the notum of Figure 3 B' and B'). Indeed, it has been shown that tissue damage can lead to suppression of proliferation in the notum and elsewhere in the disc, unless the signaling that induces the suppression is altered. Therefore, the normalization may be skewing the results because the notum EdU is not consistent across samples, possibly because the damage-induced suppression of proliferation in the notum is different across the different genetic backgrounds.

We agree with the reviewer that the use of EdU cannot distinguish between an increase in proliferation in the pouch versus a decrease in proliferation of the notum (or a combination of the two), since EdU incorporation by its nature is a relative rather than absolute measure of proliferation. However, we believe that the important finding is that a localized change in proliferation is observed late in necrosis, which is dependent on NiA/NiCP since blocking the formation of these cells prevents this change. While it is possible that this observed change is caused by a reduction in proliferation of the notum, with little or even no alteration in the pouch, this would imply that NiA/NiCP act to non-autonomously limit the proliferation of cells far from where they appear in the pouch, rather than causing localized proliferation in the immediately surrounding tissue that is representative of a blastema. Although we cannot rule this possibility out, our use of a different marker for proliferation in this work (fluorescent E2F) and a more objective proliferation marker, PH3, (Klemm et al., 2021, PMID: 34740246) agree with our observations made using EdU and suggest the formation of a localized blastema in the pouch. To attempt to address this issue, due to the variability of EdU staining between samples, we aim to quantify changes in EdU that are normalized to undamaged discs stained and mounted in the same sample, thus allowing a more objective background level of proliferation to be used for comparison.

The authors expressed p35 to attempt to generate "undead cells". They take an absence of mitogen secretion or increased proliferation as evidence that undead cells were not generated. However, there could be undead cells that do not stimulate proliferation non-autonomously, which could be detected by the persistence of caspase activity in cells that do not complete apoptosis. Indeed, expressing p35 and observing sustained effector caspase activation could help answer the later question of what percentage of this cell population would otherwise complete apoptosis (NiA, rescued by p35) vs reverse course and proliferate (NiCP, unaffected by p35).

While it is very possible that expression of P35 in NiA/NiCP could induce a previously uncharacterized type of undead cell that persists but does not secrete known AiP-related factors, the way in which P35 blocks activity of effector caspases (Drice and Dcp-1) precludes our ability to reliably detect and assay NiA/NiCP over time: P35 inactivates caspases by binding to their catalytic site, which causes cDcp-1 labeling to become weak and diffuse (Klemm et al 2021, PMID: 34740246), likely because the detectable epitope is in the catalytic site (Florentin & Arama, 2012. PMID: 22351928). Similarly, the GC3Ai reporter acts as a substrate for caspases and must be cleaved for fluorescence to occur (Zhang et al., 2013 PMID: 23857461). Thus, co-expressing P35 with GC3Ai actually reduces the number of NiA/NiCP cells labeled by GC3Ai and weakens cDcp-1 staining, preventing us from assaying their persistence or association with proliferative markers.

It is unclear if the authors' model is that the NiCP cells lead to autonomous or non-autonomous cell proliferation, or both. Could the lineage-tracing experiments and/or the experiments marking mitosis relative to caspase activity answer this question?

While we see GC3Ai-labeled NiA/NiCP in the same area of the pouch with high levels of proliferation (PH3), we observe a mixture of GC3Ai cells that overlapped the PH3 cells and GC3Ai cells that were adjacent to PH3(+) cells. Thus, we are unable to conclusively say whether proliferation is induced autonomously or non-autonomously. We have attempted to answer this question with lineage tracing, however NiA/NiCP are not labeled by the CasExpress tool, and we were unable to define a relationship between NiA/NiCP and proliferation through lineage tracing. However, we add further explanation of our findings to better clarify the proposed model of NiA/NiCP-induced proliferation.

Many of the conclusions rely on single images. Quantification of many samples should be included wherever possible.

As suggested by Reviewers 2 and 3 we plan to strengthen our findings by adding quantification of phenotypes where possible, in particular in Figure 2 as mentioned in the “Recommendations for the authors”.

Why does the reduction of Dronc appear to affect regenerative growth in females but not males?

We note that the effect on regenerative growth does appear to be present in males, but that the effect is not significant. We suspect that the lower n for this experiment is the cause, and are addressing this by repeating the experiment to increase the n.

Author response:

The following is the authors’ response to the previous reviews

Reviewer 1 (Public Review):

Multiple sclerosis (MS) is a debilitating autoimmune disease that causes loss of myelin in neurons of the central nervous system. MS is characterized by the presence of inflammatory immune cells in several brain regions as well as the brain barriers (meninges). This study aims to understand the local immune hallmarks in regions of the brain parenchyma that are adjacent to the leptomeninges in a mouse model of MS. The leptomeninges are known to be a foci of inflammation in MS and perhaps "bleed" inflammatory cells and molecules to adjacent brain parenchyma regions. To do so, they use novel technology called spatial transcriptomics so that the spatial relationships between the two regions remain intact. The study identifies canonical inflammatory genes and gene sets such as complement and B cells enriched in the parenchyma in close proximity to the leptomeninges in the mouse model of MS but not control. The manuscript is very well written and easy to follow. The results will become a useful resource to others working in the field and can be followed by time series experiments where the same technology can be applied to the diAerent stages of the disease.

Comments on revised version:

I agree that the authors successfully addressed most of my comments/critiques. However, the fact that the control mice were not injected with CFA and pertussis toxin is somewhat concerning, because it will be hard to interpret the cause of the transcriptomic readouts described in this study. Some of the described eAects might be due to CFA or pertussis (which was used in the EAE but not the "naive" group), and not necessarily to the relapsing-remitting EAE immune features recapitulated in this mouse model. Moreover, this caveat associated with the "naive" control group is not being clearly stated throughout the manuscript and might go unnoticed to readers.

The authors should clearly state, in the methods section (in the section "Induction of SJL EAE"), that the naive control group was not injected with CFA or pertussis toxin.

Additionally, this potential confounder, of not using a control group injected with the same CFA and pertussis toxin regimen of the EAE group, should be mentioned in paragraph two of the discussion alongside the other limitations of the study already highlighted by the authors (or in another section of the discussion).

We thank the reviewer for highlighting this point. Our choice of healthy/naïve, rather than CFA only, controls was intentional, given our desire to sensitively measure genes changing during neuroinflammation. Ultimately, however, we believe the choice of control group had little effect on our conclusions. We would like to note that SJL-EAE does not require pertussis toxin, so the only difference between naïve and CFA only groups is a single injection of CFA 11 weeks prior to experiment endpoint. We have performed additional IHC imaging of naïve and CFA only groups, finding no difference in glial reactivity by MFI measurement of GFAP, IBA1, or CD68 (updated Supplementary Figure 1C–E).

We have also added sections to the Results and Discussion section to clearly address this point. In the Results: “Since naïve animals were used as controls, we confirmed that CFA alone does not produce lasting glial reactivity or LMI formation. Groups of animals were given CFA only or left naïve. Neither group developed neurologic signs, and after 11 weeks the brains were processed for IHC analysis. There was no evidence of LMI development, and no difference in glial reactivity as measured by GFAP, IBA1, or CD68 intensity (Supplemental Figure 1C–E).” In the Discussion: “Another important consideration in these experiments is our choice of naïve, rather than CFA only, controls. While often used as the control in EAE studies focused on mechanisms of autoimmunity, CFA only can independently induce systemic inflammation. Since this study seeks to describe transcriptomic changes in neuroinflammation more broadly, we chose to use a healthy comparison group to maximize our ability to find genes enriched in neuroinflammation. Ultimately, however, the choice of naïve or CFA only controls is unlikely to have affected our conclusions. SJL-EAE, unlike the more common C57Bl6-EAE, does not require pertussis toxin during the induction. The only difference between naïve and CFA only controls is the subcutaneous CFA delivered at time of immunization (11 weeks prior to experiment endpoint). Indeed, when we compared CFA only and healthy animals at 11 weeks there was no difference in glial reactivity by GFAP, IBA1, or CD68 MFI. There was also no evidence of neurologic symptoms or LMI development in CFA only controls.”

Reviewer 2 (Public Review):

Accumulating data suggests that the presence of immune cell infiltrates in the meninges of the multiple sclerosis brain contributes to the tissue damage in the underlying cortical grey matter by the release of inflammatory and cytotoxic factors that diAuse into the brain parenchyma. However, little is known about the identity and direct and indirect eAects of these mediators at a molecular level. This study addresses the vital link between an adaptive immune response in the CSF space and the molecular mechanisms of tissue damage that drive clinical progression. In this short report the authors use a spatial transcriptomics approach using Visium Gene Expression technology from 10x Genomics, to identify gene expression signatures in the meninges and the underlying brain parenchyma, and their interrelationship, in the PLP-induced EAE model of MS in the SJL mouse. MRI imaging using a high field strength (11.7T) scanner was used to identify areas of meningeal infiltration for further study. They report, as might be expected, the upregulation of genes associated with the complement cascade, immune cell infiltration, antigen presentation, and astrocyte activation. Pathway analysis revealed the presence of TNF, JAK-STAT and NFkB signaling, amongst others, close to sites of meningeal inflammation in the EAE animals, although the spatial resolution is insuAicient to indicate whether this is in the meninges, grey matter, or both.

UMAP clustering illuminated a major distinct cluster of upregulated genes in the meninges and smaller clusters associated with the grey matter parenchyma underlying the infiltrates. The meningeal cluster contained genes associated with immune cell functions and interactions, cytokine production, and action. The parenchymal clusters included genes and pathways related to glial activation, but also adaptive/B-cell mediated immunity and antigen presentation. This again suggests a technical inability to resolve fully between the compartments as immune cells do not penetrate the pial surface in this model or in MS. Finally, a trajectory analysis based on distance from the meningeal gene cluster successfully demonstrated descending and ascending gradients of gene expression, in particular a decline in pathway enrichment for immune processes with distance from the meninges.

Comments on revised version:

The authors have addressed all of my comments regarding the lack of spatial resolution between the grey matter and the overlying meninges and also concerning the diAiculties in extrapolating from this mouse model to MS itself.

I am however very concerned about the lack of the correct control group. Immunization of rodents with complete freunds adjuvant and pertussis alone gives rise to widespread microglial activation, some immune cell infiltration and also structural changes to axons, particularly at nodes of Ranvier (https://doi.org/10.1097/NEN.0b013e3181f3a5b1). This will inevitably make it diAicult to interpret the transcriptomics results, depending on whether these changes are reversible or not and the time frame of the reversal. In the C57Bl6 EAE models adjuvant induced microglial activation becomes chronic, whereas the axonal changes do reverse by 10 weeks. Whether this is the same in SJL EAE model is not clear.

We thank the reviewer for bringing up this concern regarding control group, which we discussed above in point 1.1. To specifically address reviewer 2’s point regarding microglial activation, we performed IHC analysis comparing naïve and CFA only groups of SJL animals. We found no substantial diAerence in astrocyte or microglial activation in these animals after 11 weeks, as measured by GFAP, IBA1, and CD68. This new data appears in updated Supplementary Figure 1C–D.

Recommendations for the authors:

Both reviewers agree that the revised version has improved and some of their major concerns were adequately addressed. However, both reviewers also agree that critical experimental controls are missing, including the FCA and pertussis toxin injected mice which likely show some degree of inflammation in their brain and are needed to compare your experimental MS group and interpret the transcriptomics data.

We appreciate both reviewers’ important comments on the control group used in this study. In this revised manuscript we have described our rationale for choosing naïve controls, rather than CFA only, and believe they are the most appropriate comparison group. Additionally, we believe that both CFA only and naïve will have similar degrees of baseline neuroinflammation at the 11- week time point. We apologize for not clarifying before, but pertussis toxin is not used in the SJL-EAE, and therefore the “CFA only” control is much milder in SJL-EAE compared to C57Bl6-EAE. Given that many signs of inflammation resolve by 10 weeks in CFA only with pertussis controls (https://academic.oup.com/jnen/article/69/10/1017/2917071; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10902151/),) CFA only without pertussis controls are unlikely to have any substantial remaining neuroinflammation at 11 weeks. To test this, we performed an additional experiment directly comparing naïve and CFA only without pertussis.

These groups showed similar degrees of glial reactivity.

Given the costs of repeating a spatial transcriptomic experiment and inevitable batch effects should we add a group at this point, we have chosen to not as a CFA only control condition to our transcriptomics analysis. However, we believe our added text clarifying the rationale behind control choice and added immunofluorescence data gives readers the appropriate context to accurately interpret our results.

Author response:

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

We would like to thank the reviewers for their positive and constructive comments on the manuscript.

We committed in our original rebuttal letter to implement the following revisions to both DGRPool and the corresponding manuscript to address the reviewers’ comments:

(1) We agree with reviewer #1 that normalizing the data could potentially improve the GWAS results. Thus, for computing the GWAS results, we are now using these two additional options in PLINK2: “--quantile-normalize --variance-standardize”. We assessed the impact of these options on the overall results, which revealed only minor improvements of the results, globally being a bit more stringent. In this direction, we also now filter the top results with a nominal p-value of 0.001 instead of 0.01, also because it provided better results for the new gene set enrichment step.

(2) We added a KRUSKAL test next to the ANOVA test for assessing the links between the phenotypes and the 6 known covariates, as well as a Shapiro-Wilk test of normality.

(3) We agree with both reviewers that gene expression information is of interest. As mentioned before, adding gene expression data to the portal would have required extensive work, beyond the current scope of this paper, which primarily focuses on phenotypes and genotype-phenotype associations. Nonetheless, we included more gene-level outlinks to Flybase. Additionally, we now link variants and genes to Flybase's online genome browser, JBrowse. By following the reviewers' suggestions, we aim to guide DGRPool users to potentially informative genes.

(4) Consistent with the latter point, and in agreement with reviewer #2, we acknowledge that additional tools could enhance DGRPool's functionality and facilitate meta- analyses for users. Therefore, we developed a gene-centric tool that now allows users to query the database based on gene names. Moreover, we integrated ortholog databases into the GWAS results. This feature will enable users to extend Drosophila gene associations to other species if necessary.

(5) We amended the manuscript to describe all the new tools and features that were developed and implemented. In short, the new features include a new gene-centric page with diverse links (Phenotypes, Genome Browser JBrowse, Orthologs …), a variant-centric page (variant details, and PheWAS), an API for programmatic access to the database, and other statistical outputs and filtering options.

We will detail these advances in the point-by-point response below and in the revised manuscript.

Reviewer #1 (Public Review):

This is a technically sound paper focused on a useful resource around the DRGP phenotypes which the authors have curated, pooled, and provided a user-friendly website. This is aimed to be a crowd-sourced resource for this in the future.

The authors should make sure they coordinate as well as possible with the NC datasets and community and broader fly community. It looks reasonable to me but I am not from that community.

We thank the reviewer for the positive comments. We will leverage our connections to the fly and DGRP communities to make the resource as valuable as possible. DGRPool in fact already reflects the input of many potential users and was also inspired by key tools on the DGRP2 website. Furthermore, it also rationalizes why we are bridging our results with other resources, such as linking out to Flybase, which is the main resource for the Drosophila community at large.

I have only one major concern which in a more traditional review setting I would be flagging to the editor to insist the authors did on resubmission. I also have some scene setting and coordination suggestions and some minor textual / analysis considerations.

The major concern is that the authors do not comment on the distribution of the phenotypes; it is assumed it is a continuous metric and well-behaved - broad gaussian. This is likely to be more true of means and medians per line than individual measurements, but not guaranteed, and there could easily be categorical data in the future. The application of ANOVA tests (of the "covariates") is for example fragile for this.

The simplest recommendation is in the interface to ensure there is an inverse normalisation (rank and then project on a gaussian) function, and also to comment on this for the existing phenotypes in the analysis (presumably the authors are happy). An alternative is to offer a kruskal test (almost the same thing) on covariates, but note PLINK will also work most robustly on a normalised dataset.

We thank the reviewer for raising this interesting point. Indeed, we did not comment on the distribution of individual phenotypes due to the underlying variability from one phenotype to another, as suggested by the reviewer. Some distributions appear normal, while others are clearly not normally distributed. This information is 'visible' to users by clicking on any phenotype; DGRPool automatically displays its global distribution if the values are continuous/quantitative. Now, we also provide a Shapiro-Wilk test to assess the normality of the distribution.

We acknowledge the reviewer's concerns regarding the use of ANOVA tests. However, we want to point out that the ANOVA test is solely conducted to assess whether any of the well- established inversions or symbiont infection status (that, for simplification, we call “covariates” or “known covariates”) are associated with the phenotype of interest. This is merely informational, to help the user understand if their phenotype of interest is associated with a known covariate. But all of these known covariates are put in the model in any case, so PLINK2 will automatically correct for them, whatever is the output of the ANOVA test.

Still, we amended the manuscript to better explain this, and we added a Kruskal-Wallis test (in addition to the ANOVA test) in the results, so the users can have a better overview of potentially associated known covariates. We added this text on p. 10 of the revised manuscript:

“The tool further runs a gene set enrichment analysis of the results filtered at p<0.001 to enrich the associated genes to gene ontology terms, and Flybase phenotypes. We also provide an ANOVA and a Kruskal-Wallis test between the phenotype and the six known covariates to uncover potential confounder effects (prior correction), which is displayed as a “warning” table to inform the user about potential associations of the phenotype and any of the six known covariates. It is important to note that these ANOVA and Kruskal tests are conducted for informational purposes only, to assess potential associations between well-established inversions or symbiont infection status and the phenotype of interest. However, all known covariates are included in the model regardless, and PLINK2 will automatically correct for them, irrespective of the results from the ANOVA or Kruskal tests. “

We also acknowledge in the manuscript (Methods section) that the Kruskal-Wallis test is used for a single factor (independent variables) at a time. This is unlike the ANOVA test that we initially performed, which was handling multiple factors simultaneously (given that it was performed in a multifactorial design). For a more direct comparison with our ANOVA model, we ran separate Kruskal-Wallis tests for each factor, but then we acknowledged its potential limitations compared to our multifactorial ANOVA, since each of these tests treats the factor in question as the only source of variation, not considering other factors. But since the test is not intended for interactions or combined effects of these factors, we deem it to be sufficient.

Nevertheless, we concur with the reviewer that normalizing the data could potentially enhance GWAS results. Consequently, we have rerun the GWAS analyses using the PLINK2 --quantile- normalize and --variance-standardize options. We have updated all results on the website and also updated the plots in the manuscript, accordingly.

Minor points:

On the introduction, I think the authors would find the extensive set of human GWAS/PheWAS resources useful; widespread examples include the GWAS Catalog, Open Targets PheWAS, MR-base, and the FinnGen portal. The GWAS Catalog also has summary statistics submission guidelines, and I think where possible meta-data harmonisation should be similar (not a big thing). Of course, DRGP has a very different structure (line and individuals) and of course, raw data can be freely shown, so this is not a one-to-one mapping.

Thank you for the suggestion. We cited these resources in the Introduction.

“This aligns with the harmonization effort undertaken by other human GWAS/PheWAS resources, such as the GWAS Catalog, Open Targets PheWAS, MR-base, and the FinnGen portal, which provide extensive examples of effective data use and accessibility. Although the structure of DGRPool differs from these human databases, we acknowledge the importance of similar meta-data harmonization guidelines. Inspired by the GWAS Catalog's summary statistics submission guidelines, we propose submission guidelines for DGRP phenotyping data in this paper. “

For some authors coming from a human genetics background, they will be interpreting correlations of phenotypes more in the genetic variant space (eg LD score regression), rather than a more straightforward correlation between DRGP lines of different individuals. I would encourage explaining this difference somewhere.

We understand that this is a potential issue and we made the distinction clearer in the manuscript to avoid any confusion. We added this text on p.7, at the beginning of the correlation results section:

“Of note, by “phenotype correlations”, we mean direct phenotype-phenotype correlations, i.e. a straightforward Spearman’s correlation of two phenotypes between common DRGP lines, and we repeated this process for each pair of phenotypes. “

This leads to an interesting point that the inbred nature of the DRGP allows for both traditional genetic approaches and leveraging the inbred replication; there is something about looking at phenotype correlations through both these lenses, but this is for another paper I suspect that this harmonised pool of data can help.

We agree with the reviewer and hope that more meta-analyses will be made possible by leveraging the harmonized data that are made available through DGRPool.

I was surprised the authors did not crunch the number of transcript/gene expression phenotypes and have them in. Is this because this was better done in other datasets? Or too big and annoying on normalisation? I'd explain the rationale to leave these out.

This is a very good point and is in fact something that we initially wanted to do. However, to render the analysis fair and robust, it would require processing all datasets in the same way. This implies cataloging all existing datasets and processing them through the same pipeline. In addition, it would require adding a “cell type” or “tissue” layer, because gene expression data from whole flies is obviously not directly comparable to gene expression data from specific tissues or even specific conditions. This would be key information as phenotypes are often tissue-dependent. Consequently, and as implied by the reviewer, we deemed this too big of a challenge beyond the scope of the current paper. Nevertheless, we plan to continue investigating this avenue in a potential follow-up paper.

We still added a gene-centric tool to be able to query the GWAS results by gene. We also added orthologs and Flybase gene-phenotype information, both in this new gene-centric tool and also in all GWAS results.

I think 25% FDR is dangerously close to "random chance of being wrong". I'd just redo this section at a higher FDR, even if it makes the results less 'exciting'. This is not the point of the paper anyway.

We agree with the reviewer that this threshold implies a higher risk of false positive results. However, this is not an uncommonly used threshold (Li et al., PLoS biology, 2008; Bevers et al., Nature Metabolism, 2019; Hwangbo et al, Elife, 2023), and one that seems robust enough in our analysis since similar phenotypes are significant in different studies at different FDR thresholds.

Nevertheless, we revisited these results with a stronger threshold of 5% FDR in the main Figure 3C. Most of the conclusions were maintained, except for the relation between longevity and “food intake”, as well as “sleep duration”. We modified the manuscript accordingly, notably removing these points from the abstract, and tuning down the results section. We kept the 25% FDR results as supplemental information.

I didn't buy the extreme line piece as being informative. Something has to be on the top and bottom of the ranks; the phenotypes are an opportunity for collection and probably have known (as you show) and cryptic correlations. I think you don't need this section at all for the paper and worry it gives an idea of "super normals" or "true wild types" which ... I just don't think is helpful.

We appreciate the reviewer’s feedback on the section regarding extreme DGRP lines and understand the concern about potential implications of “super normals” or “true wild types.” This section aimed to explore whether specific DGRP lines consistently rank in the extremes of phenotypic measures, particularly those tied to viability-related traits. Our hypothesis was that if particular lines consistently appear at the top or bottom, this might suggest some inherent bias or inbreeding-related weakness that could influence genetic association studies.

However, as per the analyses presented, we did not discover support for this phenomenon. Importantly, the observed mild correlation in extremeness across sexes, while not profound, further suggested that this phenomenon is not a consistent population-wide feature.

Nevertheless, we consider that this message is still important to convey. In response to the reviewer's feedback, we have provided a clearer conclusion of this paper section by adding the following paragraph:

“In conclusion, this analysis showed that while certain lines exhibit lower longevity or outlier behavior for specific traits, we found no evidence of a general pattern of extremeness across all traits. Therefore, the data do not support the idea of 'super normals' or any other inherently biased lines that could significantly affect genetic studies. “

I'd say "well-established inversion genotypes and symbiot levels" rather than generic covariates. Covariates could mean anything. You have specific "covariates" which might actually be the causal thing.

We thank the author for the suggestion. We agree and modified the manuscript accordingly.

I wouldn't use the adjective tedious about curation. It's a bit of a value judgement and probably places the role of curation in the wrong way. Time-consuming due to lack of standards and best practice?

We thank the author for the suggestion. We agree and modified the manuscript accordingly, replacing the occurrences by “thorough” and “rigorous” which correspond better to the initial intended meaning.

Reviewer #2 (Public Review):

Summary:

In the present study, Gardeux et al provide a web-based tool for curated association mapping results from DRP studies. The tool lets users view association results for phenotypes and compare mean phenotype ~ phenotype correlations between studies. In the manuscript, the authors provide several example utilities associated with this new resource, including pan-study summary statistics for sex, traits, and loci. They highlight cross-trait correlations by comparing studies focused on longevity with phenotypes such as oxphos and activity.

Strengths:

-Considerable efforts were dedicated toward curating the many DRG studies provided.

-Available tools to query large DRP studies are sparse and so new tools present appeal

Weaknesses:

The creation of a tool to query these studies for a more detailed understanding of physiologic outcomes seems underdeveloped. These could be improved by enabling usages such as more comprehensive queries of meta-analyses, molecular information to investigate given genes or pathways, and links to other information such as in mouse rat or human associations.

We appreciate the reviewer's kind comments.

Regarding the tools, we concur with the reviewer that incorporating additional tools could enhance DGRPool and facilitate users in conducting meta-analyses. Therefore, we developed two new tools: a gene-centric tool that enables users to query the database based on gene names, and a variant-centric tool mostly for studying the impact of specific genomic loci on phenotypes. Additionally, in all GWAS results, we added links to ortholog databases, thereby allowing users to extend fly gene associations to other species, if required.

Furthermore, we added links to the Flybase database, for variants, phenotypes, and genes that are already present in Flybase. We also link out to a 'genome browser-like' view (Flybase’s JBrowse tool) of the GWAS results centered around the affected variants/genes.

Finally, we now also perform a gene-set enrichment analysis for each GWAS result, both in the Flybase gene-phenotype database and the Gene Ontology (GO) database.

Reviewer #2 (Recommendations For The Authors):

(1) The authors discuss how current available DRG databases are basically data-dump sites and there is a need for integrative queries. Clearly, they spent (and are spending) considerable efforts into curating associations from available studies so the current resource seems to contain several areas of missed opportunities. The most clear addition would be to integrate gene-level queries. For example which genes underlie associations to given traits, what other traits map to a specific gene, or multiple genes which map to traits. This absence of integration is somewhat surprising given the lab's previous analyses of eQTL data in DRPs (https://doi.org/10.1371/journal.pgen.1003055 ) and readily available additional data (ex. 10.1101/gr.257592.119 ,flybase) simple intersections between these at the locus level would provide much deeper molecular support for searching this database.

The point raised by the reviewer concerning eQTL / transcriptomic data is in fact similar to the one raised by reviewer #1. We strongly agree with both reviewers that incorporating eQTL results in the tool would be very valuable, and this is in fact something that we initially wanted to do. However, to render the analysis fair and robust, it would require re-processing multiple public datasets in the same way. This would imply cataloging all existing datasets and processing them through the same pipeline. In addition, it would require adding a “cell type” or “tissue” layer, because gene expression data from whole flies is obviously not directly comparable to gene expression data from specific tissues or even specific conditions. This would be key information as phenotypes are often tissue-dependent. Consequently, we deemed implementing all these layers too big of a challenge beyond the scope of the current paper, but we plan to continue investigating this avenue in a potential follow-up paper.

As mentioned before, we still integrated gene-level queries in a new tool, querying genes in the context of GWAS results. We acknowledge that this is not directly related to gene expression, and thus not implicating eQTL datasets (at least for now), but we think that it is for now a good alternative, reinforcing the interpretation of the GWAS results.

Since this point was raised by both reviewers, we added a discussion about this in the manuscript.

“We recognize certain limitations of the current web tool, particularly the lack of eQTL or gene expression data integration. Properly integrating DGRP GWAS results with gene expression data in a fair and robust manner would require uniform processing of multiple public datasets, necessitating the cataloging and standardization of all available datasets through a consistent pipeline. Moreover, incorporating a “cell type” or “tissue” layer would be essential, as gene expression data from whole flies is not directly comparable to data from specific tissues or even specific conditions. Since phenotypes are often tissue-dependent, this information is vital. However, implementing these layers presented too big of a challenge and was beyond the scope of this paper. “

(2) Another area that would help to improve is to provide either a subset or the ability to perform a meta-analysis of the studies proposed to see where phenotype intersections occur, as opposed to examining their correlation structure. For any given trait the PLINK data or association results seem already generated so running together and making them available seems fairly straightforward. This can be done in several ways to highlight the utility (for example w/wo specific covariates from Huang et al., 2014 and/or comparing associations that occur similarly or differently between sexes).

We are not 100% sure what the reviewer refers to when mentioning “phenotype intersection”, but we interpreted it as a “PheWAS capability”. Currently, in DGRPool, for every variant, there is a PheWAS option, which scans all phenotypes across all studies to see if several phenotypes are impacted by this same variant.

We tried to make this tool more visible, both in the GWAS section of the website, but also in the “Check your phenotype” tool, when users are uploading their own data to perform a GWAS. We have also created a “Variants” page, accessible from the top menu, where users can view particular variants and explore the list of phenotypes they are significantly associated with.

From both result pages, users can download the data table as .tsv files.

(3) As pointed out by the authors, an advantage of DRGs is the ease of testing on homozygous backgrounds. For each phenotype queried (or groups of related phenotypes would be of interest too), I imagine that subsetting strains by the response would help to prioritize lines used for follow-up studies. For example, resistant or sensitive lines to a given trait. This is already done in Fig 4C and 4E but should be an available analysis for all traits.

For all quantitative phenotypes, we show the global distribution by sex, followed by the sorted distribution by DGRP line. Since the data can be directly downloaded from the corresponding plots, resistant and sensitive lines can then be readily identified for all phenotypes.

(4) To researchers beyond the DRP community, one feature to consider would be seeing which other associations are conserved across species. While doing this at the phenotype level might be tricky to rename, assigning gene-level associations would make this streamlined. For example, a user could query longevity, subset by candidate gene associations then examine outputs  for  what  is  associated  with  orthologue  genes  in  humans (ex. https://www.ebi.ac.uk/gwas/docs/file-downloads) or other reference panels such as mice and rats.

In all GWAS results, and in the gene-centric tool, we have added links to ortholog databases. In short, when clicking on a variant, users can see which gene is potentially impacted by this variant (gene-level variant annotation). When clicking on these genes, the user can then open the corresponding, detailed gene page.

To address the reviewer’s comment, in the gene page, we have added two orthologous databases (Flybase and OrthoDB), which enables cross-species association analyses.

(5) Related to enabling a meta-data analysis, it would be helpful to let users download all PLINK or DGRP tables in one query. This would help others to query all data simultaneously.

We would like to kindly point out that all phenotyping data can already be downloaded from the front page, which includes the phenotypes, the DGRP lines and the studies’ data and metadata. However, we did not provide the global GWAS results through a single file, because the data is too large. Instead, we provide each GWAS dataset via a unique file, available per phenotype, on the corresponding GWAS result page of this phenotype. This file is filtered for p<0.001, and contains GWAS results (PLINK beta, p and FDR) as well as gene and regulatory annotations.

(6) Following analysis of association data an interesting feature would be to enable users to subset strains for putative LOF variants at a given significant locus. This is commonly done for mouse strains (ex. via MGI).

The GWAS result table available for each phenotype can be filtered for any variant of interest. We added the capability to filter by variant impact; LOF variants being usually referred to as HIGH impact variants.

(7) Viewing the locus underlying annotation can also provide helpful information. For example, several nice fly track views are shown in 10.1534/g3.115.018929, which would help users to interpret molecular mechanisms.

We now link the GWAS results out to Flybase’s JBrowse genome browser.

Author response:

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

We are grateful to all three reviewers and editors for their critical comments and suggestions.

Reviewer #2 (Recommendations For The Authors):

The authors responded satisfactorily to all my comments and suggestions.

We thank the reviewer for his time and feedback.

Reviewer #3 (Recommendations For The Authors):

Comments for authors:

The authors have addressed most of the reviewer's concerns. Although no additional data were included to strengthen the manuscript, they have clarified some relevant points, and the manuscript has been updated accordingly. In my view, the current manuscript is well-written and mostly straightforward.

We thank the reviewer for his time and suggestions. Addressing them have improved the quality of our manuscript.

After a second revision, I just have a few minor comments (mostly editorial) that should be easy to address.

(1) Page 16: "The dominant presence of the GRIK1-1 gene was also reported in retinal Off bipolar cells..." Please include reference(s).

We have now cited the following reference:

Lindstrom, S.H., Ryan, D.G., Shi, J., DeVries, S.H., 2014. Kainate receptor subunit diversity underlying response diversity in retinal Off bipolar cells. J. Physiol. 592, 1457–1477. https://doi.org/10.1113/jphysiol.2013.265033

(2) Page 18: "Based on our functional assays, the splice seems to affect the interaction between the receptor and auxiliary proteins". Please remove or tone down this statement; the current data do not support this claim.

We have revised the sentence as following: “Based on our functional assays, the splice may possibly affect the interaction between the receptor and auxiliary proteins.”

(3) Page 24: "cultures ... at 0.5 µg/mL were transfected". In the current context, it is not clear what you mean with 0.5 µg/mL. Please check and correct.

Thanks for pointing out this error. We have corrected it.

(4) Page 30. He et al. reference is repeated.

Thanks. We have fixed it now.

(5) Figure 3, Panel C: Please incorporate the EC50 value for the red trace into the figure; it appears to be a different data set and, consequently, a different fitting compared with Figure 2C.

The GluK1-1a data set (red trace) is identical to that in Figure 2c, though it may appear different due to the scale of the X and Y axis. As suggested, we have now included the EC50 value for this data set in Figure 3, panel C.

(6) Figure legend 4: Please check two minor issues here:

(a) "Bar graphs... with or without Neto1 protein..." This statement is apparently wrong; Figure 4 does not show the effect of Neto1.

(b) "The wild type GluK1 splice variant data is the same as from Figure 1.." I think the authors mean Figure 2A instead of Fig. 1. Please check.

Thanks for pointing out the error. We have fixed the same in the revised manuscript.

(7) Please check and correct spelling/wording issues in the text. Here are some examples:

(a) Page 9 " Figure 3G - I, Table2.." (There is no Panel I). 

Fixed.

(b) Page 16 "... and is involved in various pathophysiology..." 

We have revised the sentence as “… and is involved in various pathophysiological conditions”

(c) Page 19 "The constructs used for this study were HEK293 WT mammalian cells were seeded on..." 

Fixed. Thanks.

(d) Page 23 "The immunoblots were probed..." Please check the whole paragraph and correct the issues.

Fixed. Thanks.

(e) Page 27 "initially, 1,97,908 particles were picked". Check the value; the same issue occurs in Fig.6 table supplement 1. 

Thanks. We have now modified the sentence to clarify that for  GluK1-1aEM ND-SYM, initially, 1,97,908 particles were picked and subjected to multiple rounds of clean-up using 2D and 3D classification. Finally,  24,531 particles were used for the final 3D reconstruction and refinement.

(f) Legend Figure 2: Remove "(F)" from the legend. 

Thanks. Fixed.

(g) Legend Figure 2-Sup.1: Check/correct spelling issues. 

Thanks. Fixed.

(h) Figure 5-figure supplement 1: There is a mistake in panel B: "GFP" label is shown for Gluk1 and Neto2, but the authors mention that the pull-down was done with Anti-His antibodies. Please correct.

Thanks. The pull-down experiments were done with anti-His for both the blots presented in panels A and B as mentioned in both the figures (right side panels of both A and B). However, for the GluK1 and Neto2 pull downs (panel B), the blots were probed with anti-GFP antibody which would detect both the receptor (as the receptor has both GFP-His8) and Neto2-GFP at their respective sizes. This has been indicated in the figure panel B.

(8) Related to the point-by-point document:

Major concern 2: Interpreting the effect of mutants on the regulation by Neto proteins requires knowing how the mutant is affecting the channel properties without Neto. In my view, if the data showing the K368/375/379/382H376-E mutant without Neto is missing (in this case due to low current amplitude), then, the pink bars in Fig. 5 should be removed from the figure. 

We thank the reviewer for raising this interesting point and agree that it would be valuable to characterize the channel properties of all the mutants individually. However, as mentioned earlier, the functions of some mutant receptors are only rescued, or reliable, measurable currents are detected, when they are co-expressed with Neto proteins. We still believe that comparing wild-type and mutant receptors co-expressed with Neto proteins provides important insights, and therefore, we would like to retain the K368/375/379/382H376-E mutant data in the figure.

Major concern 4: Figure 6-figure Supplement 8 is not mentioned in the manuscript. It would help to include a proper description in the Results section similar to the answer included in the point-by-point document.

Figure6-figure Supplement 8 has already been cited on page 15. We have also cited Figure6-figure Supplement 9 on the same page and have added following sentences in the text:

“A superimposition of GluK1-1aEM (detergent-solubilized or reconstituted in nanodiscs) and GluK1-2a (PDB:7LVT) showed an overall conservation of the structures in the desensitized state. No significant movements were observed at both the ATD and LBD layers of GluK1-1a with respect to GluK1-2a (Figure 6; Figure 6-figure supplement 9).”

Major concern 5: The ramp/recovery protocol was not included properly in the manuscript; please include the time of the ramp pulse and the time used for the recovery period.

Elaborated ramp and recovery protocols are included in the methods section. The time used for the recovery period was variable and was tuned as per the recovery kinetics. All the figures were representative traces are shown include the scale bar showing the time period of agonist application.

Minor concern 1: The proposed change was not included in the manuscript; check page 7.

Thanks for highlighting this error. We have now changed it in the revised manuscript.

Minor concern 10: The manuscript was not corrected as indicated. Please check.

Thanks. We have now modified the sentence as following: “…..a reduction was observed for K375/379/382H376-E receptors (1.17 ± 0.28 P=0.3733) compared to wild-type although differences do not reach statistical significance”

Minor concern 14: The figure was not corrected as indicated. Please check.

Thanks for highlighting this error. We have now changed it in the revised manuscript.

Minor concern 19: I suggest including this briefly in the Discussion section.

Thanks for the suggestion. We have included the following sentence in the discussion:

“The differences in observations could be due to variations in experimental conditions, such as the constructs and recording conditions used.”

Author response:

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

Reviewer #1 (Public Review):

Weaknesses:

Given that all mutants tested showed the same degree of activation by PEG400, it seemed possible that PEG400 might be an allosteric activator of WNK1/3 through direct binding interactions. Perhaps PEG400 eliminates CWN1/2 waters by inducing conformational changes so that water loss is an effect not a cause of activation. To address this it would be helpful to comment on whether new electron densities appeared in the X-ray structure of WNK1/SA/PEG400 that might reflect PEG400 interactions with chains A or B.

We re-evaluated the WNK1/SA/PEG400 electron density looking for non-protein densities larger than water. No new densities were found. However, we do observe a PEG400-destabilizing effect using differential scanning fluorimetry, and have included this data into Figure 2. We conclude that the effects on the water structure and destabilization are due to demands on solvent.

We have included in the second paragraph of the introduction references to primary literature that advance similar arguments to explain osmolyte induced effects on activity.

Specifically, Colombo MF, Rau DC, Parsegian VA (1992) Protein solvation in allosteric regulation: a water effect on hemoglobin. Science 256: 655-659 and LiCata VJ, Allewell NM (1997) Functionally linked hydration changes in Escherichia coli aspartate transcarbamylase and its catalytic subunit. Biochemistry 36: 10161—10167. 

It would also be helpful to discuss any experiments that might have been done in previous work to examine the direct binding of glycerol and other osmolytes to WNKs.

We did not observe PEG400 in WNK1/SA/PEG400 despite effects on the space group and subunit packing. On the other hand, glycerol was observed in WNK1/SA, which was cryoprotected in glycerol (PDB file 6CN9). We have highlighted these differences in the second section of the results. A thorough analysis on the effects of various osmolytes on WNK structure, stability, and activity is a potential future direction.

The study would benefit from a deeper discussion about how to reconcile the different effects of mutations. For example, wouldn't most or all of the mutations be expected to disrupt the water network, and relieve the proposed autoinhibition? This seemed especially true for some of the residues, like Y420(Y346), D353(D279), and K310(K236), which based on Fig 3 appeared to interact with waters that were removed by PEG400.

The manuscript has been updated with new data and better discussion of this point. Given the inconsistencies on the effects of mutation in static light scattering (SLS), we addressed the possibility that the reducing agent was not constant across experiments. In a repeated study, including reducing agent (1 mM TCEP), we obtained results on mutant mass more similar to wild-type than in the original experiment. An exception was that two of the mutants were much more monomeric than wild-type. It follows that the network CWN1 stabilizes the inactive dimer. The reduced activity of some of the mutants probably reflects the position of CWN1 and the AL-CL Cluster in the active site, such that mutants can affect substrate binding or catalysis. This is now better discussed both in the data and discussion sections.

Mutants have a tendency to have complex effects on activity and structure. It was satisfying to find any activating mutants. We point out that we have been careful to present all of our data including mutants that are not easily explained by our models.

Alternatively, perhaps the waters in CWN2 are more important for maintaining the autoinhibited structure. This possibility would be useful to discuss, and perhaps comment on what may be known about the energetic contributions of bound water towards stabilizing dimers.

This research focused on the most salient unique feature of WNK1- CWN1. We also identified CWN2. Mutational analysis of CWN2 can’t be done without disrupting the dimer interface, greatly complicating data interpretation.

It would also be useful to comment on why aggregation of E319Q/A (E314) shouldn't inhibit kinase activity instead of activating it.

On recollection of the SLS data in the presence of reducing agent, we saw reduced aggregation. WNK3/D279N and WNK3/E314Q were more monomeric, especially at the higher protein concentration used. WNK3/E314Q is one of the more active mutants.

The X-ray work was done entirely with WNK1 while the mutational work was done entirely with WNK3. Therefore, a simple explanation for the disconnect between structure and mutations might be that WNK1 and WNK3 differ enough that predictions from the structure of one are not applicable to mutations of the other. It would be helpful to describe past work comparing the structure and regulation of WNK1 and WNK3 that support the assumption of their interchangeability.

We have responded directly to this concern. We introduced our most interesting amino acid replacement WNK3/E314A into WNK1, making WNK1/E388A. Similar trends in chloride inhibition and mutational activation were observed in WNK1 as in WNK3. This supports the assumption of interchangeability of WNK1 and WNK3 we invoked for practical reasons.  As expected, the overall activity of WNK1 is lower than WNK3. Overall, the lower activity limited data collection. However, the lower activity did allow us to fit the chloride inhibition data to a kinetic model for WNK1.  Panels on WNK1 activity, mutation, and chloride inhibition were added to Figure 5 and to Supplemental data (Table S6).

Reviewer #2 (Public Review):

Strengths:

The most interesting result presented here is that P1 crystals of WNK1 convert to P21 in the presence of PEG400 and still diffract (rather than being destroyed as the crystal contacts change, as one would expect). All of the assays for activity and osmolyte sensing are carried out well.

Thank you. We have emphasized this point in the Results section with the word “remarkably”

Weaknesses:

The rationale for using WNK3 for the mutagenesis study is that it is more sensitive to osmotic pressure than WNK1. I think that WNK1 would have been a better platform because of the direct correlation to the structural work leading to the hypothesis being tested. All of the crystallographic work is WNK1; it is not logical to jump to WNK3 without other practical considerations.

This point is addressed in the last comment to Reviewer 1. We added autophosphorylation assay data on our most interesting mutant (WNK3/E314A) in WNK1 (WNK1/E388A). Conversely, we have crystallographic data on uWNK3 (on uWNK3/E314A collected to 3.3Å). These new data justify the assumption of interchangeability of results obtained for uWNK1 and uWNK3.

Osmolyte sensing was tested by measuring ATP consumption as a function of PEG400 (Figure 6). Data for the subset of mutants analyzed by this assay showed increasing activity. It is not clear why the same collection of mutant proteins analyzed in the experiments of Figure 5 was not also measured for osmolyte sensing in Figure 6.

These data are now more complete, having been now collected for all of the WNK3 mutants (now Figure 7).

The last set of data presented uses light scattering to test whether the WNK3 mutant proteins exhibit quaternary structural changes consistent with the monomer/dimer hypothesis. If they did, one would expect a higher degree of monomer for those that are activated by mutation, and a lower amount of monomer (like wt) for those that are not. Instead, one of the mutant proteins that showed the most chloride inhibition (Y346F) had a quaternary structure similar to the wt protein, and others have similar monomer/dimer mixtures but distinct chloride inhibition profiles (K307A and M301A). I don't see how the light scattering data contribute to this story other than to refute the hypothesis by showing a lack of correlation between quaternary structure, water binding, and activity. This is another reason why the disconnect between WNK1 and WNK3 could be a problem. All of the detailed structural work with WNK1 must be assumed with WNK3; perhaps the light scattering data are contradicting this assumption?

As noted above, on recollection of the SLS data in the presence of reducing agent, we saw reduced aggregation and more consistency with our model. Thus, we now feel it is a useful contribution to the manuscript. The table in Supplemental data has been updated.

Reviewer #1 (Recommendations For The Authors):

Fig 3D in the PDF manuscript seemed distorted - waters were cut off. Also Fig 2D would benefit from showing the whole molecule, instead of cutting off the top and bottom of the kinase domain.<br /> We suspect this is a data transfer problem, since we don’t see these truncations.

Both Figure 2 and 3 have been changed, addressing these concerns and adding new differential scanning fluorimetry data as discussed in reply to Reviewer 1. Figure 2 was simplified by eliminating Figures 2A-2C, and replacing them with a new Figure 2B, the superposition of WNK1/SA/PEG400 (PDB 9D3F), WNK1/SA (PDB 6CN9).  

In Figure 3, we added a panel highlighting the volume change around CWN1 in presence of PEG400 (Figure 3C). Hopefully, inappropriate cropping has been eliminated.

Line 162: Y314F should be Y346F.

This has been corrected. Thank you.

Lines 211-213 - these two sentences do not seem to logically go together: "Two hyper-active mutants were discovered, WNK3/E314A, and WNK3/E314Q. These mutants are straightforward to interpret based on our model: the mutated residues support and stabilize inactive dimeric WNK."

An extensive rewrite has been conducted to address the difference in activity between the higher activity mutants versus less active mutants, now discussed in two paragraphs, and two Figures, Figure 5 and 6. The SLS data, recollected with more reducing agent, has given more consistent results (Supplemental), making the discussion more straightforward (discussed above).

Reviewer #2 (Recommendations For The Authors)

I think WNK1 would be a better platform for mutagenesis than WNK3. Or minimally the authors should better justify the switch to WNK3 from WNK1. Analyze the same set of mutants in Figure 5 into Figure 6.

Again, we have added assay data on uWNK1/E388A, and structural data on uWNK3/E314A.

I would analyze the same set of mutants in Figures 5 and 6.

We have analyzed all of the WNK3 mutants in the ADP-Glo assays (Figure 7).

Will the P21 crystal form grow independently in PEG400?

Attempts to crystallize WNK1/SA or WNK3/SA or other constructs in PEG400 have been unsuccessful.

I would also add some context about the role of water in allosteric mechanisms. I know there is a long history in hemoglobin in which specific waters have been associated with the T and R states such as that by Marcio Colombo. There is a relatively recent article in J. Phys Chem. that would provide good context. Leitner et al., J. Chem. Phys. 152, 240901 (2020)

Thank you. Good call.

Author response:

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

eLife assessment

This fundamental study uses a creative experimental system to directly test Ohno's hypothesis, which describes how and why new genes might evolve by duplication of existing ones. In agreement with existing criticism of Ohno's original idea, the authors present compelling evidence that having two gene copies does not speed up the evolution of a new function as posited by Ohno, but instead leads to the rapid inactivation of one of the copies through the accumulation of mostly deleterious mutations. These findings will be of broad interest to evolutionary biologists and geneticists.

We thank the editors and the reviewers for their positive feedback concerning our experimental system and for the constructive feedback on how to further improve the manuscript. We have now addressed the reviewer’s comments in a revised version.

Reviewer #1 (Public Review):

Overview:

The authors construct a pair of E. coli populations that differ by a single gene duplication in a selectable fluorescent protein. They then evolve the two populations under differing selective regimes to assess whether the end result of the selective process is a "better" phenotype when starting with duplicated copies. Importantly, their starting duplicated population is structured to avoid the duplication- amplification process often seen in bacterial artificial evolution experiments. They find that while duplication increases robustness and speed of adaptation, it does not result in more highly adapted final states, in contrast to Ohno's hypothesis.

Major comments:

This is an excellent study with a very elegant experimental setup that allows a precise examination of the role of duplication in functional evolution, exclusive of other potential mechanisms. My main concern  is  to  clarify  some  of  the  arguments  relating  to  Ohno's  hypothesis.

I think my main confusion on first reading the manuscript was in the precise definition of Ohno's hypothesis. I think this confusion was mine and not the authors, but it is likely common and could be addressed.

Most evolutionary biologists think of gene duplication as making neofunctionalization "easier" by providing functional redundancy and a larger mutational target, such that the evolutionary process of neofunctionalization is faster (as the authors observed). In this framework, the final evolved state might not differ when selection is applied to duplicated copies or a single-copy gene. Ohno's hypothesis, by contrast, argues that there generally exist adaptive conflicts between the ancestral function and the "desired" novel function, such that strong selection on a single-copy gene cannot produce the evolutionary optima that selection on two copies would. This idea is hinted at in the quotation from Ohno in paragraph 2 of the introduction. However, the sentences that follow I don't think reinforce this concept well enough and lead to some confusion.

With that definition in mind, I agree with the authors' conclusion that these data do not support Ohno's hypothesis. My quibble would be that what is actually shown here is that adaptive conflict in function is not universal: there are cases where a single gene can be optimized for multiple functions just as well as duplicated copies. I do not think the authors have, however, refuted the possibility that such adaptive conflicts are nonetheless a significant barrier to evolutionary innovation in the absence of gene duplication generally. Perhaps just a sentence or two to this effect might be appropriate.

We fully agree with the reviewer that trade-offs might play an important role in the evolution of single copy and of duplicated genes, depending on the gene and on the selection regime. And while trade-offs are not likely to play a big role in the selection regime we discuss in detail in the main text (evolution towards more green), they probably are important for at least one our selection regimes. In fact, we so state in the following passage of the discussion. In addition, we have now added a sentence that acknowledges the importance of trade-offs for evolution in the absence of gene duplication:

“A single gene encoding such a protein suffers from an adaptive conflict between the two activities. Gene duplication may provide an escape from this adaptive conflict, because each duplicate may specialize on one activity14, 15. For coGFP, a trade-off likely exists for fluorescence in these two colors, because improvement of green fluorescence entails a loss of blue fluorescence during evolution (Figure S8 and Figure S16). We therefore expected that during selection for both green and blue fluorescence, one cogfp copy in double-copy populations would “specialize” on green fluorescence whereas the other copy would specialize on blue fluorescence. However, when we analyzed individual population members with two active gene copies we could not find any such specialization (Figure S21). Moreover, the identified key mutations at positions 147 and 162 have a very low frequency (<1%) in these populations (Figure S15). Future experiments with different selection strategies might reveal the reasons for this observation and the conditions under which such a specialization can occur.“

I also think the authors need to clarify their approach to normalizing fluorescence between the two populations to control for the higher relative protein expression of the population with a duplicated gene. Since each population was independently selected with the highest fluorescing 60% (or less) of the cells selected, I think this normalization is appropriate. Of course, if the two populations were to compete against each other, this dosage advantage of the duplicates would itself be a selective benefit. Even as it is, the dosage advantage should be a source of purifying selection on the duplication, and perhaps this should be noted.

The reviewer is correct. To be able to follow the evolutionary trajectories of the different constructs, the populations were treated separately. The gates were adjusted for each library separately to select for the top 60, 1 or 0.01% of cells and the gates for the double-copy populations were set to slightly higher fluorescence, reflected in the higher fluorescence of these populations in Figure 3A. Indeed, if individuals in these populations were to compete against each other, the double-copy populations would have a benefit due to the dosage advantage. However, as we already pointed out in the manuscript, we did not see any additional advantage beyond the increased gene dosage provided by the second copy (Figure 3B). To discuss this issue in more detail, we have now added the following text to the discussion:

“It is worth noting that we evolved each of our single- and double-copy populations separately and in parallel to follow their individual evolutionary trajectories. In a natural population, individuals with one or two copies might occur in the same population and compete against each other. In this situation any dosage advantage of a duplicate gene would itself entail selective benefit. Our approach allowed us to find out if gene duplication facilitates phenotypic evolution beyond any such gene dosage effect. At least for the specific genes, selection pressures, and mutation rates we used, the data suggest that it does not.”

Finally, I am slightly curious about the nature of the adaptations that are evolving. The authors primarily discuss a few amino-acid changing mutations that seem to fix early in the experiment. Looking at Figure 3, it however, appears that the populations are still evolving late in the experiment, and so presumably other changes are occurring later on. Do the authors believe that perhaps expression changes to increase protein levels are driving these later changes?

Figure S15 shows that some mutations are indeed still increasing in frequency during late evolutionary rounds, in particular S2L, V141L and V205L. We have measured the emission spectra of these mutants (Figure S16), and these mutations increase fluorescence both in green and blue. It is therefore likely that these mutations, similar to L98M, increase protein expression, solubility, or thermal stability, as suggested by the reviewer. We now clarify this matter in a new passage of the results:

“Like L98M, the additional mutations S2I, V141I and V25L also occurred in all selection regimes, but they reached lower frequencies than L98M during the 5 generations of the experiment. We hypothesized that mutations observed in all selection regimes do not derive their benefit from increasing the intensity of any one fluorescent color. Instead, they may increase protein expression, solubility, or thermal stability.”

Reviewer #2 (Public Review):

Summary:

Drawing from tools of synthetic biology, Mihajlovic et al. use a cleverly designed experimental system to dissect Ohno's hypothesis, which describes the evolution of functional novelty on the gene-level through the process of duplication & divergence.

Ohno's original idea posits that the redundancy gained from having two copies of the same gene allows one of them to freely evolve a new function. To directly test this, the authors make use of a fluorescent protein with two emission maxima, which allows them to apply different selection regimes (e.g. selection for green AND blue, or, for green NOT blue). To achieve this feat without being distracted by more complex evolutionary dynamics caused by the frequent recombination between duplicates, the authors employ a well-controlled synthetic system to prevent recombination: Duplicates are placed on a plasmid as indirect repeats in a recombination-deficient strain of E.coli. The authors implement their directed evolution approach through in vitro mutagenesis and selection using fluorescent-activated cell sorting. Their in-depth analysis of evolved mutants in single-copy versus double-copy genotypes provides clear evidence for Ohno's postulate that redundant copies experience relaxed purifying selection. In contrast to Ohno's original postulate, however, the authors go on to show that this does not in fact lead to more rapid phenotypic evolution, but rather, the rapid inactivation of one of the copies.

Strengths:

This paper contributes with great experimental detail to an area where the literature predominantly leans on genomics data. Through the use of a carefully designed, well-controlled synthetic system the authors are able to directly determine the phenotype & genotype of all individuals in their evolving populations and compare differences between genotypes with a single or double copy of coGFP. With it they find clear evidence for what critics of Ohno's original model have termed "Ohno's dilemma", the rapid non- functionalization by predominantly deleterious mutations.

Including an expressed but non-functional coGFP in (phenotypically) single copy genotypes provides an especially thoughtful control that allows determining a baseline dN/dS ratio in the absence of selection. All in all the study is an exciting example of how the clever use of synthetic biology can lead to new insights.

Weaknesses:

The major weakness of the study is tied to its biggest strength (as often in experimental biology there is a trade-off between 'resolution' and 'realism').

The paper ignores an important component of the evolutionary process in favour of an in-depth characterization of how two vs one copy evolve. Specifically, by employing a recombination-deficient strain and constructing their duplicates as inverted repeats their experimental design completely abolishes recombination between the two copies.

This is problematic for two reasons:

i)  In nature, new duplicates do not arise as inverted, but rather as direct (tandem) repeats and - as the authors correctly point out - these are very unstable, due to the fact that repeated DNA is prone to recA- dependent homologous recombination (which arise orders of magnitude more frequently than point mutations).

ii)  This instability often leads to further amplification of the duplicates under dosage selection both in the lab and in the wild (e.g. Andersson & Hughes, Annu. Rev. Genet. 2009), and would presumably also be an outcome under the current experimental set-up if it was not prevented from happening?

So in sum, recombination between duplicate genes is not merely a nuisance in experiments, but occurring at extremely high frequencies in nature (such that the authors needed to devise a clever engineering solution to abolish it), and is often observed in evolving populations, be it in the laboratory or the wild.

The manuscript sells controlling of copy number as a strength. And clearly, without it, the same insights could not be gained. However, if the basis for the very process of what Ohno's model describes is prevented from happening for the process to be studied, then, for reasons of clarity and context this needs pointing out, especially, to readers less familiar with the principles of molecular evolution.

Connected to this, there are several places in the introduction and the discussion where I feel that the existing literature, in particular models put forward since Ohno that invoke dosage selection (such as IAD) end up being slightly misrepresented.

My point is best exemplified in line 1 of Discussion: "To test Ohno's hypothesis and to distinguish its predictions from those of competing hypotheses, it is necessary to maintain a constant and stable copy number of duplicated genes during experimental evolution."

We understand the reviewer’s position and fully agree that we needed to clarify better what our experiments aimed to achieve. To this end, we rewrote the beginning of the discussion to read:

“Our aim was to study whether gene duplication can affect mutational robustness and phenotypic evolution beyond any effect of increased gene dosage provided by multiple gene copies. To this end, we needed to maintain a constant and stable copy number of duplicated genes during experimental evolution.”

I think this statement is simply not true and might be misleading. To take the exaggerated position of a devil's advocate, the goal of evolutionary biology should be to find out how evolution actually proceeds in nature most of the time, rather than creating laboratory systems that manage to recapitulate influential ideas.

On this point, we respectfully disagree. To ask questions like ours, laboratory experiments that are highly controlled albeit possibly “unnatural” can be essential. And we would argue that our experiments do not merely aim to “recapitulate” an influential idea but to validate it and potentially refute it, as we did for our study system. Validating theory is an essential aspect of experimental science. Textbooks in biology and beyond are rife with examples.

While fixing copy number may be a necessary step to understand how one copy evolves if a second one is present, it seems that if Ohno's hypothesis only works out in recA-deficient bacterial strains and on engineered inverted repeats, that Ohno might have missed one crucial aspect of how paralogs evolve. The mentioned competing hypotheses have been put forward to (a) address Ohno's dilemma (which the present study beautifully demonstrates exists under their experimental conditions) and (b) to reflect a commonly observed evolutionary process in bacteria (dosage gain in response to selection, e.g. a classic way of gaining antibiotic resistance). Fixing the copy number allowed the authors to show which predictions of Ohno's model hold up and which don't (under these specific conditions). But they do so without even preventing the processes described by alternative models from happening, so the experimental system is hardly appropriate to distinguish between Ohno & alternatives. Therefore, I think it could be made clearer that the experimental system is great to look at certain aspects Ohno's hypothesis in  detail, but  it  can  only inform  us about  a  universe  without  recombination.

(1)  Citing the works by ref 8, 26, 27 to merely state that "in some copies were gained and some were lost (ref 6, ref 25)" makes it seem as if fixing at 2 copies is some sort of sensible average. Yet ref 6 (Dhar et al.) specifically states that dosage is the most important response. Moreover, in this study gene copies are lost, but plasmid copies are gained instead. In Holloway et al. 2007 (ref 25), the 2 copies resided on different plasmids, so entirely different underlying molecular genetics might be at work (high cost of plasmid maintenance, and competitive binding on both proteins onto the respective (off)-target, where either way selection favored a single copy, so a different situation altogether). In both cited studies, fixing the copy would have prohibited learning something about the process of duplication & divergence.

Hence this statement seems to distract the readers from the main message, which seems that preventing recombination experimentally allows to follow the divergence of each copy and studying a response that does not involve dosage-increase.

(2)   "These studies highlighted the importance of gene duplication in providing fast adaptation under changing environmental conditions but they focused on the importance of gene dosage." I think this constructs a false dichotomy. Instead, these studies pointed out that dosage (and with it, selection for dosage)  is  an  important  part  of  the  equation  that  might  have  been  missed  by  Ohno.

Your points are well taken. To clarify the insights from previous experiments and the aims of our experiments we rewrote this passage in question as follows.

“These studies underline the importance of gene duplication in providing fast adaptation under changing environmental conditions. In some studies one copy was lost6, 25, while in others, additional copies were gained8, 26, 27. Together these studies highlight that gene dosage and selection for dosage can play an important role during the evolution of duplicated genes6, 8, 25-28.

These studies also raise the question whether gene duplication can provide an advantage beyond its effects on gene dosage. To find out it is necessary to study the evolution of gene duplicates while keeping the copy number of the duplicated gene exactly at two. This is challenging because gene duplication causes recombinational instability and high variability in copy number. No previous experimental studies were designed to control copy number. Here, we present an experimental system that allowed us to keep the copy number fixed at one or two genes, and to follow the evolution of each gene copy in the absence of any dosage increase.”

(3)  "Such models are also easier to test experimentally, because they do not require precise control of gene copy number. The necessary tests can even benefit from massive gene amplifications8. Although Ohno's hypothesis is more difficult to test experimentally (...)" - again, I feel the wording is slightly misleading. The point is not that IAD is easier to test and Ohno's model is harder to test in laboratory experiments, rather, experiments (and some more limited observations of naturally evolving populations) seem to suggest that in reality evolution proceeds (more often?) according to IAD rather than Ohno's neofunctionalization hypothesis. However, as the authors point out, it will be exciting to see their clever experimental system used to test other genes and conditions to get a more comprehensive understanding of what gene- and selection- parameter values would overcome Ohno's dilemma.

We agree and in response rewrote the paragraph in question to read:

“The challenge that a duplicated gene copy must remain free of frequent deleterious mutations long enough to acquire beneficial mutations that provide a new selectable phenotype is known as Ohno’s dilemma13. Our experiments confirm that this challenge is highly relevant for post-duplication evolution. Other models such as the innovation-amplification-divergence (IAD) model8, 13 postulate that this dilemma can be resolved through an increase in gene dosage that allows latent pre-duplication phenotypes to come under the influence of selection. To distinguish between the effects of gene dosage and other benefits of gene duplication, we prevented recombination and gene amplification to prevent copy number increases beyond two copies. We are aware that our experimental design does not reflect how evolution may occur in the wild. However, this design allowed us to study evolutionary forces separately that are otherwise difficult to disentangle. “

Finally, we also made two changes in the abstract (highlighted in red) to take your feedback into account.

Reviewer #2 (Recommendations For The Authors):

The paper is very well written, with a lot of emphasis put on explaining every step and every finding. It was a joy to read.

Thanks!

Full stop missing in line 5 of abstract.

Corrected.

Author response:

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

Reviewer #2 (Recommendations for the authors):

Comments to the authors:

R1. The authors show a similar reduction in ECAR as a measure of glycolytic inhibition upon treating H37Rv-infected unprimed MDMs with 5 mM 2-DG at 1 h and 24 h. However, the data pertaining to the extent of glycolytic inhibition upon 2-DG treatment in IFN-γ or IL-4 primed AMs or MDMs is not included.

We acknowledge that we have not checked the ECAR of every dataset herein treated with 2DG. However, we have provided evidence that 2DG reduced ECAR in the control datasets, and moreover, 2DG is functionally affecting the cells (e.g. the presence of 2DG altered cytokine production in both AM and MDM, even in the presence of IFN-γ or IL-4).

R2. The authors have replotted the same data as percent change and fold difference with different normalizing samples. While they have corrected one of the highlighted discrepancies in the data plotting of Fig. 1A and 1C, similar discrepancies are still evident in many other instances. Based on my understanding of the data and normalization methodology stated by the authors in response to comment (#5) by reviewer 1, the authors are plotting fold changes across all samples with respect to unstimulated and unprimed macrophages, whereas percent changes are plotted for stimulated (LPS or dead H37Rv) samples with respect to baseline measurements for each unstimulated sample under differently primed macrophages. I believe the slope of lines connecting unstimulated and LPS stimulates/H37Rv infected upon percent increase or decrease (from the baseline of unstimulated samples) will still maintain their trend in fold changes (relative to unstimulated and unprimed macrophages) irrespective of changes in absolute values. For instance, in Fig. 1F, there are at least 3 samples that show an increase in fold change in OCR upon H37Rv infection in IFN-γ primed MDMs. However, Fig. 1H, plotted from the same data, shows a decrease in OCR across all IFN-γ primed MDMs upon H37Rv infection. The authors have also highlighted that this decrease in OCR upon H37Rv infection in IFN- γ primed MDMs is highly significant (P < 0.01). The same data is again plotted as a bar plot in Fig. 1J as fold change relative to unstimulated and unprimed macrophages (mislabeled as percent change to unstimulated), showing no difference upon H37Rv infection of IFN-γ primed MDMs.

We have amended the axis in Figure 1 and Supplemental Figure 1 to more accurately describe the two different forms of analysis. We have fixed the errors outlined. We have also amended the methods in the text to clarify the two analyses carried out on the metabolic data. Lines 113-122 as follows:

“Fold change in ECAR and OCR was calculated compared to unstimulated unprimed controls at 150 minutes, where unstimulated unprimed macrophages were set to 1. This allows for analysis of the effects of both priming and subsequent stimulation for and accounts for the variation in the raw ECAR and OCR reading between runs thereby making each donor its own control.

Percent change in ECAR and OCR was also calculated to equalise groups to the same point prior to stimulation. Each condition was compared to its own respective primed or unprimed baseline at 30 minutes and this was set to 100%, prior to stimulation, this was carried to examine the capacity of cells to increase metabolic parameters in response to stimulation. Post stimulation percent change data was then extracted and analysed at 150 minutes. This controls for the priming effect and enables the analysis of metabolic capacity in each dataset.”

For figure 1J, the data is replotted from fold change datasets (not percentage change where the decrease in OCR is significant). The axis label has been revised for accuracy.

Author response:

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

This study investigates the role of Caspar (Casp), an orthologue of human Fas- associated factor-1, in regulating the number of primordial germ cells that form during Drosophila embryogenesis. The findings are important in that they reveal an additional pathway involved in germ cell specification and maintenance. The evidence supporting the conclusions is solid, as the authors identify Casp and its binding partner Transitional endoplasmic reticulum 94 (TER94) as factors that influence germ cell numbers. Minor changes to the title, text, and experimental design are recommended.

We thank the Editors and Reviewers for their overall positive and thoughtful feedback. Based on these comments, we have revised our manuscript. The changes in the manuscript have been highlighted in ‘blue’ font for easy visualisation.

Reviewer #1 (Public Review):

Summary:

The authors were seeking to define the roles of the Drosophila caspar gene in embryonic development and primordial germ cell (PGC) formation. They demonstrate that PGC number, and the distribution of the germ cell determinant Oskar, change as a result of changes in caspar expression; reduction of caspar reduces PGC number and the domain of Oskar protein expression, while overexpression of caspar does the reverse. They also observe defects in syncytial nuclear divisions in embryos produced from caspar mutant mothers. Previous work from the same group demonstrated that Caspar protein interacts with two partners, TER94 and Vap33. In this paper, they show that maternal knockdown of TER94 results in embryonic lethality and some overlap of phenotypes with reduction of caspar, supporting the idea may work together in their developmental roles. The authors propose models for how Caspar might carry out its developmental functions. The most specific of these is that Caspar and its partners might regulate oskar mRNA stability by recruiting ubiquitin to the translational regulator Smaug.

Strengths:

The work identifies a new factor that is involved in PGC specification and points toward an additional pathway that may be involved in establishing and maintaining an appropriate distribution of Oskar at the posterior pole of the embryo. It also ties together earlier observations about the presence of TER94 in the pole plasm that have not heretofore been linked to a function.

Weaknesses:

(1)  A PiggyBac insertion allele casp[c04227] is used throughout the paper and referred to as a loss-of-function allele (casp[lof]). However, this allele does not appear to act strictly as a loss-of-function. Figure 1E shows that some residual Casp protein is present in early embryos produced by casp[lof]/Df females, and this protein is presumably functional as the PiggyBac insertion does not affect the coding region. Also, Figures 1B and 1C show that the phenotypes of casp[lof] homozygotes and casp[lof]/Df are not the same; surprisingly, the homozygous phenotypes are more severe. These observations are unexplained and inconsistent with the insertion being simply a loss-of-function allele. Might there be a second-site mutation in casp[c04227]?

The term loss-of-function (lof) is used rather than null or amorph. casplof is a strong hypomorph, with residual (and functional) protein estimated in the range 5-10% when compared to the wild type. The caspc04227 was procured from BDSC, and based on the decrease in lethality of the casplof/casp(Df) compared to casplof, we assume that second site hits in the casplof line are the reason for the enhanced lethality. For this very reason, we have used casplof/ casp(Df) for all subsequent experiments. We also conducted rescue experiments wherever possible to confirm the specificity with caspWT and various deletion variants of casp.

(2)  TER94 knockdown phenotypes have been previously published (Zhang et al 2018 PMID 30012668), and their effects on embryonic viability and syncytial mitotic divisions were described there. This paper is inappropriately not cited, and the data in Figure 4 should be presented in the context of what has been published before.

We apologize for the oversight. Indeed, Zhang et al. (2018) highlighted TER94 as one of the loci uncovered in their screen and some of the relevant phenotypes are described there. We have referred to their findings at the appropriate junctures as suggested (pg 11, pg13, pg 15).

(3)  The peptide counts in the mass spectrometry experiment aimed at finding protein partners for Casp are extremely low, except for Casp itself and TER94. Peptide counts of 1-2 seem to me to be of questionable significance.

Peptide counts are indeed low, but the fact that they are enriched at all, in comparison to controls, considering that we are using whole embryo lysates rather than isolated PGC lysates, suggests interaction with Casp could be biologically/ functionally meaningful. The data is restricted to the supplementary material and is not analyzed in isolation; we have combined data from multiple mass spectrometry experiments by other researchers to link Casp to pole plasm components.

(4)  The pole bud phenotypes from TER94 knockdown and casp mutant shown in Fig 5 appear to be quite different. These differences are unexplained and seem inconsistent with the model proposed that the two proteins work in a common pathway. Whole embryos should also be shown, as the TER94 KD phenotype could result from a more general dysmorphism.

We agree that TER94 KD is a stronger phenotype, with TER94 having essential cell division and patterning roles. In fact, the TER94 RNAi embryos, unlike casplof, stall in terms of their developmental program before Stage 4. This has been noted in the earlier study (Zhang et al., 2018). As a result, we focused on pole bud stage embryos that were rare - but present in the collections. We report that PGC from very early TER94 RNAi embryos have fewer pole buds.

The rationale behind the presumption that these two proteins may work in a common pathway is clear-cut. We have validated the physical interaction using protein lysates from two developmental time points. Satisfyingly, an affinity purification using antibodies against TER94 or Casp invariably enriches the other protein as the primary interacting partner. Our model integrates data from mammalian and fly systems to support the idea that there must be an overlap between TER94/Casp function, with these two proteins working together to engineer the degradation of ubiquitinated Smaug. Future experiments are necessary to confirm and extend this claim.

(5)  Figure 6 is not quantitative, lacking even a second control staining to check for intensity variation artifacts. Therefore, it shows that the distribution of Oskar protein changes in the various genotypes, but not convincingly that the level of Oskar changes as the paper claims.

We appreciate that oskar RNA localization is also somewhat altered due to change in casp levels. We have acknowledged the variability in the various phenotypes, and as such, it is unsurprising that it has also reflected in the Oskar levels. However, it is evident that a statistically significant number of mutant embryos show a decrease in Oskar levels.

(6)  The error bars are huge in the graphs in Figure 7H, I, and J, leading me to question whether these changes are statistically significant. Calculations of statistical significance are missing from these graphs and need to be added.

The data in the Western blots represents the whole embryo, as the lysates used are from embryos 0-1, 1-2, 2-3 hrs. We have averaged and plotted data from 5 Western blots. The changes are not statistically significant. Even without the statistical significance, the data for Fig. 7I led us to examine Smaug in the pole cells, rather than in the whole embryo. The pole cell data (Fig8-D3) is striking and led to the conclusion – that Smaug protein perdures in the pole cells during the stages of syncytial/cellular blastoderm.

(7)  There are many instances of fuzzy and confusing language when describing casp phenotypes. For example, on lines 211-212, it is stated that 'casp[lof] adults are only partially homozygous viable as ~70% embryos laid by the homozygous mutant females failed to hatch into larvae'. Isn't this more accurately described as 'casp[c04227] is a maternal-effect lethal allele with incomplete penetrance'? Another example is on line 1165, what exactly is a 'semi-vital function'?

We thank the reviewer for reading the manuscript in detail. We have tried to pay attention to reduce the ambiguity and fixed the text accordingly (pg 7, line 214; pg 33, line 1169, word semi-vital is deleted).

Reviewer #2 (Public Review):

Summary:

This study investigated the role of the Caspar (Casp) gene, a Drosophila homolog of human Fas-associated factor-1. It revealed that maternal loss of Casp led to centrosomal and cytoskeletal abnormalities during nuclear cycles in Drosophila early embryogenesis, resulting in defective gastrulation. Moreover, Casp regulates PGC numbers, likely by regulating the levels of Smaug and then Oskar. They demonstrate that Casp protein levels are linearly correlated to the PGC number. The partner protein TER94, an ER protein, shows similar but slightly distinct phenotypes. Based on the deletion mutant analysis, TER94 seems functionally relevant for the observed Casp phenotype. Additionally, it is likely involved in regulating protein degradation during PGC specification.

Strengths:

The paper reveals an unexpected function of the maternally produced Casp gene, previously implicated in immune response regulation and NF-kB signaling inhibition, in nuclear division and PGC formation in early fly embryos. Experiments are properly conducted and strongly support the conclusion. The rescue experiment using deletion mutant form is particularly informative as it suggests the requirement of each domain function.

Weaknesses:

Functional relationships among molecules shown here (and other genes known to regulate these processes) are still unclear.

We completely agree with this assessment. In our view this is an interesting albeit initial report. We also appreciate that understanding the mechanistic underpinnings of these results will be critical. We have ensured that our present claims are backed up by data, however, are fully sensitive to the fact that newer observations will refine or even alter these claims. We are continuing to work on the problem and will hopefully make further inroads in mechanism in the coming years.

Reviewer #3 (Public Review):

Summary:

Das et al. discovered a maternal role for Caspar (Casp), the Drosophila orthologue of human Fas-associated factor-1 (FAF1), in embryonic development and germ cell formation. They find that Casp interacts with Transitional endoplasmic reticulum 94 (TER94). Loss of Casp or TER94 leads to partial embryonic lethality, correlated with aberrant centrosome behavior and cytoskeletal abnormalities. This suggests that Casp, along with TER94, promotes embryonic development through a still unidentified mechanism. They also find that Casp regulates germ cell number by controlling a key determinant of germ cell formation, Oskar, through its negative regulator, Smaug.

Strengths:

Overall, the experiments are well-conducted, and the conclusions of this paper are mostly well-supported by data.

Weaknesses:

Some additional controls could be included, and the language could be clarified for accuracy.

Reviewer #1 (Recommendations For The Authors):

(1)  The paper is inconsistent in using standard Drosophila nomenclature. Often the name of the mammalian counterpart is used instead. This needs to be cleaned up as it is very confusing to the reader.

The names of the mammalian counterpart are explicitly used, when we intended, to underscore the parallels between mammalian vs Drosophila function, specifically in the context of the major players in this study, TER94 vs VCP; Caspar vs FAF1. Since we do not have direct biochemical data indicating that TER94/Casp degrades Smaug, we use published mammalian literature to draw parallels. At no point have we swapped terminology casually.

(2)  The Discussion is far too long and in my view extends too far beyond the experimental data in the paper. As a start for editing, its first two paragraphs (lines 1138-1164) include mostly general statements and could be greatly reduced or eliminated.

Our aim was to emphasize the repurposing of factors between early development and later/adult stages for different functional contexts. Our laboratory (Ratnaparkhi) works on Casp in terms of its roles in NF-kappa B signalling. We serendipitously stumbled on the embryonic lethality while characterizing the casplof allele, which, later, led us to examine the function of Casp during embryonic germ cell development.

(3)  The Introduction is weak in its description of the developmental function of Toll and Dorsal. This could be summarized in a sentence or two.

As suggested, a few sentences that highlight the developmental function of Toll/Dorsal signalling have been added to the text (pg 3, line 90-92).

(4)  Even if correctly cited, it is not appropriate to simply reproduce an image from a public database, as was done in Figure S1C. This should be removed.

Figure S1C has been deleted.

(5)  The Materials and Methods section should be moved to after the Discussion so it does not interrupt the flow of the Results.

The Section has been moved as suggested.

Reviewer #2 (Recommendations For The Authors):

For general readers, more detailed information about the PGC specification will be helpful in the Introduction or Results section.

PGC specification is introduced in the text as the story transits from global embryonic effects of casp knockdown to specific effects on PGCs. A few additional sentences have been added to bolster the text (pg 11, first paragraph).

The Methods section talks about live imaging, but I could not find the experiments in the figures. Are the data available for asynchronous nuclear divisions in the live imaging?

The live imaging relates to DIC movies that are part of Suppl. Fig 2A. The movies are embedded in an MS PowerPoint slide, which has been uploaded as a PowerPoint (and not a PDF).

To ensure that the mutant changes the Osk translation rate, showing the Osk RNA level may be helpful.

oskar RNA localization is quite distinct as compared to Oskar and Vasa protein. It has been shown that oskar RNA is localized to the founder granules and is, in fact, excluded from the germ granules that contain Vasa, Oskar and nos RNA etc. Gavis lab recently reported (Eichler et al., 2020) that ectopic localization of osk RNA in the germ granules is toxic to pole cells. Thus, it will be of interest to analyze whether and how oskar RNA is localized in casp embryos.

More discussion about the difference between Casp and ter94 phenotypes and potential reasons would be informative.

TER94 appears to be an essential maternal gene. Hypomorphic knockdown of TER94 using RNAi is sufficient to induce early embryonic lethality. In fact, Zhang et. al., 2018 et al., using stronger/earlier maternal drivers highlighted the lethality and somatic cell division defects caused due to the severe loss of TER94. The UBX domain is present in multiple proteins, in addition to Casp. TER94 possible plays a vital role in protein degradation of critical cell cycle proteins, such as cyclins that need to be degraded for efficient genomic duplications in the 10’ nuclear division cycles that predominate the first few hours of embryogenesis.

N=3 (Fig1 legend) and N =15 (Fig2). What are those numbers?

N=3 indicated the number of repeats of the western blot. This reference has been deleted. N=15, represents the number of embryos imaged for data in panels G and H.

Reviewer #3 (Recommendations For The Authors):

Major Suggestion:

(1)  Oskar (Osk) mRNA Localization: Does Osk mRNA localization change upon overexpression or LOF of Casp? Since TER94 has been implicated in Osk mRNA localization (Ruden et al., 2000), this would be a good control to include.

As mentioned earlier, in the response to editors, data presented in our manuscript indicates that Caspar is unique in its ability to regulate both Oskar levels and centrosome dynamics. As the reviewer pointed out, we are in the process of analysing the possible localization defects in oskar mRNA in the embryos. Since the preliminary data are promising, we are pursuing this carefully to better understand the involvement of Caspar. We are focusing on the ability of Caspar to regulate early nuclear divisions prior to pole cell formation. It is possible that in casp mutant embryos the nuclei/centrosomes that enter the pole plasm are already defective and thus can influence release of the pole plasm components. This needs to be examined carefully, and we are conducting these experiments.

(2)  Western Blot for Osk Protein: It would also be beneficial to perform a western blot for Osk protein to demonstrate that it is indeed increased upon Casp overexpression.

This is a good suggestion. However, Oskar antibodies are not readily available, and we have a very limited supply which have been used for embryo staining experiments. We considered these more useful as in addition to the absolute levels, staining experiment can reveal localization pattern. It was thus possible to correlate Oskar function with the pole cell counts in respective genetic backgrounds.

(3)  Title Clarification: The title states, "Caspar determines primordial germ cell identity in Drosophila melanogaster." The current experiments do not show that Casp determines germ cell identity. It would be more accurate to conclude that Casp regulates germ cell numbers.

Please refer to the introductory paragraphs where we explain our views in this regard. We have modified our title to “Caspar specifies primordial germ cell count and identity in Drosophila melanogaster."

Minor Suggestions:

(1)  Line 69: Delete the use of "recent" for papers published in 2001 and 2007. These papers are around 20 years old.

The word has been deleted.

(2)  Paragraph from Line 110: Consider splitting this paragraph into two for better readability and clarity.

Paragraph has been split into two; this has improved readability.

(3)  Line 266: Check and correct the formatting issues in this line.

Edited, based on suggestion. A line break was added after the title.

(4)  Line 328: Adding references to earlier studies here will be useful for providing context and supporting information.

References that introduce Centrosomes and their roles as organizing centres have been added in line 336.

(5)  Line 564: It is best to avoid using the word "master." Please consider using other terms such as "key" or "principal."

Edited, based on suggestion.

(6)  Citations: The authors should also cite Cinalli et al., 2013 for the Gcl reference to ensure comprehensive citation of relevant literature.

Thank you for the suggestion. The reference has been added on pages 16 and 29.

(7)  Overall Length: The paper is quite long. If it can be shortened, it will be easier to read. Consider condensing sections where possible without losing essential information.

The paper is indeed longer than average, but the choice of eLife as the home for this study was, in part, determined by the platform's flexibility regarding length/ word count. It seemed worthwhile to elaborate the text in places to accentuate the novelty of the findings.

These additions and adjustments would help to further substantiate the claims and improve the clarity of the paper.

We hope that the claims made in our manuscript are substantiated by the data that are presented. Wherever possible, we have tried to modify the text suitably to improve clarity.

Author response:

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

Reviewer 1 (Public Review):

Summary: Wilmes and colleagues present a computational model of a cortical circuit for predictive processing which tackles the issue of how to learn predictions when different levels of uncertainty are present for the predicted sensory stimulus. When a predicted sensory outcome is highly variable, deviations from the average expected stimulus should evoke prediction errors that have less impact on updating the prediction of the mean stimulus. In the presented model, layer 2/3 pyramidal neurons represent either positive or negative prediction errors, SST neurons mediate the subtractive comparison between prediction and sensory input, and PV neurons represent the expected variance of sensory outcomes. PVs therefore can control the learning rate by divisively inhibiting prediction error neurons such that they are activated less, and exert less influence on updating predictions, under conditions of high uncertainty.

Strengths: The presented model is a very nice solution to altering the learning rate in a modality and context-specific way according to expected uncertainty and, importantly, the model makes clear, experimentally testable predictions for interneuron and pyramidal neuron activity. This is therefore an important piece of modelling work for those working on cortical and/or predictive processing and learning. The model is largely well-grounded in what we know of the cortical circuit.

Weaknesses: Currently, the model has not been challenged with experimental data, presumably because data from an ad- equate paradigm is not yet available. I therefore only have minor comments regarding the biological plausibility of the model:

Beyond the fact that some papers show SSTs mediate subtractive inhibition and PVs mediate divisive inhibition, the selection of interneuron types for the different roles could be argued further, given existing knowledge of their properties. For instance, is a high PV baseline firing rate, or broad sensory tuning that is often interpreted as a ’pooling’ of pyramidal inputs, compatible with or predicted by the model?

Thank you for this nice suggestion. We added a section to the discussion expanding on this: “The model predicts that the divisive interneuron type, which we here suggest to be the PVs, receive a representation of the stimulus as an input. PVs could be pooling the inputs from stimulus-responsive layer 2/3 neurons to estimate uncertainty. The more the stimulus varies, the larger the variability of the pyramidal neuron responses and, hence, the variability of the PV activity. The broader sensory tuning of PVs (Cottam et al. 2013) is in line with the model insofar as uncertainty modulation could be more general than the specific feature, which is more likely for low-level features processed in primary sensory cortices. PVs were shown to connect more to pyramidal cells with similar feature-tuning (Znamenskyiy et al. 2024); this would be in line with the model, as uncertainty modulation should be feature-related. In our model, some SSTs deliver the prediction to the positive prediction error neurons. SSTs are already known to be involved in spatial prediction, as they underlie the effect of surround suppression (Adesnik et al. 2012), in which SSTs suppress the local activity dependent on a predictive surround.”

On a related note, SSTs are thought to primarily target the apical dendrite, while PVs mediate perisomatic inhibition, so the different roles of the interneurons in the model make sense, particularly for negative PE neurons, where a top-down excitatory predicted mean is first subtractively compared with the sensory input, s, prior to division by the variance. However, sensory input is typically thought of as arising ’bottom-up’, via layer 4, so the model may match the circuit anatomy less in the case of positive PE neurons, where the diagram shows ’s’ arising in a top-down manner. Do the authors have a justification for this choice?

We agree that ‘s’ is a bottom-up input and should have been more clear about that we do not consider ‘s’ to be a top-down input like the prediction. We hence adjusted the figure correspondingly and added a few clarifying sentences to the manuscript. The reviewer, however, raises an important point, which is not talked about enough. Namely, that if the bottom-up input ‘s’ comes from L4, how can it be compared in a subtractive manner with the top-down prediction arriving in the superficial layers? In Attinger et al. it was shown that the visual stimulus had subtractive effects on SST neurons. The axonal fibers delivering the stimulus information are hence likely to arrive in the vicinity of the apical dendrites, where SSTs target pyramidal cells. Hence, those axons delivering stimulus information could also target the apical dendrites of pyramidal cells. As the reviewer probably had in mind, L4 input tends to arrive in the somatic layer. However, there are also stimulus-responsive cells in layer 2/3, such that the stimulus information does not need to come directly from L4, it could be relayed via those stimulus-responsive layer 2/3 cells. It has been shown that L2/3→L3 axons are mostly located in the upper basal dendrites and the apical oblique dendrites, above the input from L4 (Petreanu et al. The subcellular organization of neocortical excitatory connections). Hence, stimulus information could arrive on the apical dendrites, and be subtractively modulated by SSTs. We would also like to note that the model does not take into account the precise dendritic location of the inputs. The model only assumes that the difference between stimulus and prediction is calculated before the divisive modulation by the variance.

In cortical circuits, assuming a 2:8 ratio of inhibitory to excitatory neurons, there are at least 10 pyramidal neurons to each SST and PV neuron. Pyramidal neurons are also typically much more selective about the type of sensory stimuli they respond to compared to these interneuron classes (e.g., Kerlin et al., 2012, Neuron). A nice feature of the proposed model is that the same interneurons can provide predictions of the mean and variance of the stimulus in a predictor-dependent manner. However, in a scenario where you have two types of sensory stimulus to predict (e.g., two different whiskers stimulated), with pyramidal neurons selective for prediction errors in one or the other, what does the model predict? Would you need specific SST and PV circuits for each type of predicted stimulus?

If we understand correctly, this would be a scenario in which the same context (e.g., sound) is predicting two types of sensory stimulus. In that case, one may need specific SST and PV circuits for the different error neurons selective for prediction errors in these stimuli, depending on how different the predictions are for the two stimuli as we elaborate in the following. The reviewer is raising an important point here and that is why we added a section to the discussion elaborating on it.

We think that there is a reason why interneurons are less selective than pyramidal cells and that this is also a feature in prediction error circuits. Similarly-tuned cells are more connected to each other, because they tend to be activated together as the stimuli they encode tend to be present in the environment together. Also, error neurons selective to nearby whiskers are more likely to receive similar stimulus information, and hence similar predictions. Hence, because nearby whiskers are more likely to be deflected similarly, a circuit structure may have developed during development such that neurons selective for prediction errors of nearby whiskers, may receive inputs from the same inhibitory interneurons. In that case, the same SST and PV cells could innervate those different neurons. If, however, the sensory stimuli to be predicted are very different, such that their representations are likely to be located far away from each other, then it also makes sense that the predictions for those stimuli are more diverse, and hence the error neurons selective to these are unlikely to be innervated by the same interneurons.

We added a shorter version of this to the discussion: “The lower selectivity of interneurons in comparison to pyramidal cells could be a feature in prediction error circuits. Error neurons selective to similar stimuli are more likely to receive similar stimulus information, and hence similar predictions. Therefore, a circuit structure may have developed such that prediction error neurons with similar selectivity may receive inputs from the same inhibitory interneurons.”

Reviewer 2 (Public Review):

Summary: This computational modeling study addresses the observation that variable observations are interpreted differently depending on how much uncertainty an agent expects from its environment. That is, the same mismatch between a stimulus and an expected stimulus would be less significant, and specifically would represent a smaller prediction error, in an environment with a high degree of variability than in one where observations have historically been similar to each other. The authors show that if two different classes of inhibitory interneurons, the PV and SST cells, (1) encode different aspects of a stimulus distribution and (2) act in different (divisive vs. subtractive) ways, and if (3) synaptic weights evolve in a way that causes the impact of certain inputs to balance the firing rates of the targets of those inputs, then pyramidal neurons in layer 2/3 of canonical cortical circuits can indeed encode uncertainty-modulated prediction errors. To achieve this result, SST neurons learn to represent the mean of a stimulus distribution and PV neurons its variance.

The impact of uncertainty on prediction errors is an understudied topic, and this study provides an intriguing and elegant new framework for how this impact could be achieved and what effects it could produce. The ideas here differ from past proposals about how neuronal firing represents uncertainty. The developed theory is accompanied by several predictions for future experimental testing, including the existence of different forms of coding by different subclasses of PV interneurons, which target different sets of SST interneurons (as well as pyramidal cells). The authors are able to point to some experimental observations that are at least consistent with their computational results. The simulations shown demonstrate that if we accept its assumptions, then the authors’ theory works very well: SSTs learn to represent the mean of a stimulus distribution, PVs learn to estimate its variance, firing rates of other model neurons scale as they should, and the level of un- certainty automatically tunes the learning rate, so that variable observations are less impactful in a high uncertainty setting.

Strengths: The ideas in this work are novel and elegant, and they are instantiated in a progression of simulations that demonstrate the behavior of the circuit. The framework used by the authors is biologically plausible and matches some known biological data. The results attained, as well as the assumptions that go into the theory, provide several predictions for future experimental testing.

Weaknesses: Overall, I found this manuscript to be frustrating to read and to try to understand in detail, especially the Results section from the UPE/Figure 4 part to the end and parts of the Methods section. I don’t think the main ideas are so complicated, and it should be possible to provide a much clearer presentation.

For me, one source of confusion is the comparison across Figure 1EF, Figure 2A, Figure 3A, Figure 4AB, and Figure 5A. All of these are meant to be schematics of the same circuit (although with an extra neuron in Figure 5), yet other than Figures 1EF and 4AB, no two are the same! There should be a clear, consistent schematic used, with identical labeling of input sources, neuron types, etc. across all of these panels.

We changed all figures to make them more consistent and pointed out that we consider subparts of the circuit.

The flow of the Results section overall is clear until the “Calculation of the UPE in Layer 2/3 error neurons” and Figure 4, where I find that things become significantly more confusing. The mention of NMDA and calcium spikes comes out of the blue, and it’s not clear to me how this fits into the authors’ theory. Moreover: Why would this property of pyramidal cells cause the PV firing rate to increase as stated? The authors refer to one set of weights (from SSTs to UPE) needing to match two targets (weights from s to UPE and weights from mean representation to UPE); how can one set of weights match two targets? Why do the authors mention “out-of-distribution detection’ here when that property is not explored later in the paper? (see also below for other comments on Figure 4)

We agree that the introduction of NMDA and calcium spikes was too short and understand that it was confusing. We therefore modified and expanded the section. To answer the two specific questions: First, Why would this property of pyramidal cells cause the PV firing rate to increase as stated? This property of pyramidal cells does not cause the PV firing rate to increase. When for example in positive error neurons, the mean input increases, then the PVs receive higher stimulus input on average, which is not compensated by the inhibitory prediction (which is still at the old mean), such that the PV firing rate increases. Due to the nonlinear integration in PVs, the firing rate can increase a lot and inhibit the error neurons strongly. If the error neurons integrate the difference nonlinearly, they compensate for the increased inhibition by PVs. In Figure 5, we show that a circuit in which error neurons exhibit a dendritic nonlinearity matches an idealised circuit in which the PVs perfectly represent the variance. We modified the text to clarify this.

Second, how can one set of weights match two targets? In our model, one set of weights does not need to match two targets. We apologise that this was written in such a confusing way. In positive error neurons, the inhibitory weights from the SSTs need to match the excitatory weights from the stimulus, and in negative error neurons, the inhibitory weights from the SSTs need to match the excitatory weights from the prediction. The weights in positive and negative circuits do not need to be the same. So, on a particular error neuron, the inhibition needs to match the excitation to maintain EI balance. Given experimental evidence for EI balance and heterosynaptic plasticity, we think that this constraint is biologically achievable. The inhibitory and excitatory synapses that need to match are targeting the same postsynaptic neuron and could hence have access to their postsynaptic effect. We modified the text to be more clear. Finally, we omitted the mentioning of out-of-distribution detection, see our reply below.

Coming back to one of the points in the previous paragraph: How realistic is this exact matching of weights, as well as the weight matching that the theory requires in terms of the weights from the SSTs to the PVs and the weights from the stimuli to the PVs? This point should receive significant elaboration in the discussion, with biological evidence provided. I would not advocate for the authors’ uncertainty prediction theory, despite its elegant aspects, without some evidence that this weight matching occurs in the brain. Also, the authors point out on page 3 that unlike their theory, “...SSTs can also have divisive effects, and PVs can have subtractive effects, dependent on circuit and postsynaptic properties”. This should be revisited in the Discussion, and the authors should explain why these effects are not problematic for their theory. In a similar vein, this work assumes the existence of two different populations of SST neurons with distinct UPE (pyramidal) targets. The Discussion doesn’t say much about any evidence for this assumption, which should be more thoroughly discussed and justified.

These are very important points, we agree that the biological plausibility of the model’s predictions should be discussed and hence expanded the discussion with three new paragraphs:

To enable the comparison between predictions and sensory information via subtractive inhibition, we pointed out that the weights of those inputs on the postsynaptic neuron need to match. This essentially means that there needs to be a balance of excitatory and inhibitory inputs. Such an EI balance has been observed experimentally (Tan and Wehr, 2009). And it has previously been suggested that error responses are the result of breaking this EI balance (Hertäg und Sprekeler, 2020, Barry and Gerstner, 2024). Heterosynaptic plasticity is a possible mechanism to achieve EI balance (Field et al. 2020). For example, spike pairing in pre- and postsynaptic neurons induces long-term potentiation at co-activated excitatory and inhibitory synapses with the degree of inhibitory potentiation depending on the evoked excitation (D’amour and Froemke, 2015), which can normalise EI balance (Field et al. 2020).

In the model we propose, SSTs should be subtractive and PVs divisive. However, SSTs can also be divisive, and PVs subtractive dependent on circuit and postsynaptic properties (Seybold et al. 2015, Lee et al. 2012, Dorsett et al. 2021). This does not necessarily contradict our model, as circuits in which SSTs are divisive and PVs subtractive could implement a different function, as not all pyramidal cells are error neurons. Hence, our model suggests that error neurons which can calculate UPEs should have similar physiological properties to the layer 2/3 cells observed in the study by Wilson et al. 2012.

Our model further posits the existence of two distinct subtypes of SSTs in positive and negative error circuits. Indeed, there are many different subtypes of SSTs. SST is expressed by a large population of interneurons, which can be further subdivided. There is e.g. a type called SST44, which was shown to specifically respond when the animal corrects a movement (Green et al. 2023). Our proposal is hence aligned with the observation of functionally specialised subtypes of SSTs.

Finally, I think this is a paper that would have been clearer if the equations had been interspersed within the results. Within the given format, I think the authors should include many more references to the Methods section, with specific equation numbers, where they are relevant throughout the Results section. The lack of clarity is certainly made worse by the current state of the Methods section, where there is far too much repetition and poor ordering of material throughout.

We implemented the reviewer’s detailed and helpful suggestions on how to improve the ordering and other aspects of the methods section and now either intersperse the equations within the results or refer to the relevant equation number from the Methods section within the Results section.

Reviewer 3 (Public Review):

Summary: The authors proposed a normative principle for how the brain’s internal estimate of an observed sensory variable should be updated during each individual observation. In particular, they propose that the update size should be inversely proportional to the variance of the variable. They then proposed a microcircuit model of how such an update can be implemented, in particularly incorporating two types of interneurons and their subtractive and divisive inhibition onto pyramidal neurons. One type should represent the estimated mean while another represents the estimated variance. The authors used simulations to show that the model works as expected.

Strengths: The paper addresses two important issues: how uncertainty is represented and used in the brain, and the role of inhibitory neurons in neural computation. The proposed circuit and learning rules are simple enough to be plausible. They also work well for the designated purposes. The paper is also well-written and easy to follow.

Weaknesses: I have concerns with two aspects of this work.

(1) The optimality analysis leading to Eq (1) appears simplistic. The learning setting the authors describe (estimating the mean of a stationary Gaussian variable from a stream of observations) is a very basic problem in online learning/streaming algorithm literature. In this setting, the real “optimal” estimate is simply the arithmetic average of all samples seen so far. This can be implemented in an online manner with µˆt = µˆt−1 +(st −µˆt−1)/t. This is optimal in the sense that the estimator is always the maximum likelihood estimator given the samples seen up to time t. On the other hand, doing gradient descent only converges towards the MLE estimator after a large number of updates. Another critique is that while Eq (1) assumes an estimator of the mean (mˆu), it assumes that the variance is already known. However, in the actual model, the variance also needs to be estimated, and a more sophisticated analysis thus needs to take into account the uncertainty of the variance estimate and so on. Finally, the idea that the update should be inverse to the variance is connected to the well-established idea in neuroscience that more evidence should be integrated over when uncertainty is high. For example, in models of two-alternative forced choices it is known to be optimal to have a longer reaction time when the evidence is noisier.

We agree with the reviewer that the simple example we gave was not ideal, as it could have been solved much more elegantly without gradient descent. And the reviewer correctly pointed out that our solution was not even optimal. We now present a better example in Figure 7, where the mean of the Gaussian variable is not stationary. Indeed, we did not intend to assume that the Gaussian variable is stationary, as we had in mind that the environment can change and hence also the Gaussian variable. If the mean is constant over time, it is indeed optimal to use the arithmetic mean. However, if the mean changes after many samples, then the maximum likelihood estimator model would be very slow to adapt to the new mean, because t is large and each new stimulus only has a small impact on the estimate. If the mean changes, uncertainty modulation may be useful: if the variance was small before, and the mean changes, then the resulting big error will influence the change in the estimate much more, such that we can more quickly learn the new mean. A combination of the two mechanisms would probably be ideal. We use gradient descent here, because not all optimisation problems the brain needs to solve are that simple. The problem with converging only after a large number of updates is a general problem of the algorithm. Here, we propose how the brain could estimate uncertainty to achieve the uncertainty-modulation observed in inference and learning tasks observed in behavioural studies. To give a more complex example, we present in a new Figure 8 how a hierarchy of UPE circuits can be used for uncertainty-based integration of prior and sensory information, similar to Bayes-optimal integration.

Yes, indeed, there is well-known behavioural evidence, we would like to thank the reviewer for pointing out this connection to two-alternative forced choice tasks. We now cite this work. Our contribution is not on the already established computational or algorithmic level, but the proposal of a neural implementation of how uncertainty could modulate learning. The variance indeed needs to be estimated for optimal mean updating. That means that in the beginning, there will be non-optimal updating until the variance is learned. However, once the variance is learned, mean-updating can use the learned variance. There may be few variance contexts but many means to be learned, such that variance contexts can be reused. In any case, this is a problem on the algorithmic level, and not so much on the implementational level we are concerned with.

(2) While the incorporation of different inhibitory cell types into the model is appreciated, it appears to me that the computation performed by the circuit is not novel. Essentially the model implements a running average of the mean and a running average of the variance, and gates updates to the mean with the inverse variance estimate. I am not sure about how much new insight the proposed model adds to our understanding of cortical microcircuits.

We here suggest an implementation for how uncertainty could modulate learning via influencing prediction error com- putation. Our model can explain how humans could estimate uncertainty and weight prior versus sensory information accordingly. The focus of our work was not to design a better algorithm for mean and variance estimation, but rather to investigate how specialised prediction error circuits in the brain can implement these operations to provide new experimental hypotheses and predictions.

Reviewer 1 (Recommendations For The Authors):

Clarity and conciseness are a strength of this manuscript, but a more comprehensive explanation could improve the reader’s understanding in some instances. This includes the NMDA-based nonlinearity of pyramidal neuron activation - I am a little unclear exactly what problem this solves and how (alongside the significance of 5D and E).

We agree that the introduction of the NMDA-based nonlinearity was too short and understand that it was confusing. We therefore modified and expanded the section, where we introduce the dendritic nonlinearity of the error neurons.

Page 5: I think there is a ’positive’ and ’negative’ missing from the following sentence: ’the weights from the SSTs to the UPE neurons need to match the weights from the stimulus s to the UPE neuron and from the mean representation to the UPE neuron, respectively.’

Thanks for pointing that out! We changed the sentence to be more clear to the following: “To ensure a comparison between the stimulus and the prediction, the inhibition from the SSTs needs to match the excitation it is compared to in the UPE neurons: In the positive PE circuit, the weights from the SSTs representing the prediction to the UPE neurons need to match the weights from the stimulus s to the UPE neurons. In the negative PE circuit, the weights from SSTs representing the stimulus to the negative UPE neurons need to match the weights from the mean representation to the UPE neurons, respectively.”

Reviewer 2 (Recommendations For The Authors):

Related to the first point above: I don’t feel that the authors adequately explained what the “s” and “a” information (e.g., in Figures 2A, 3A) represent, where they are coming from, what neurons they impact and in what way (and I believe Fig. 3A is missing one “a” label). I think they should elaborate more fully on these key, foundational details for their theory. To me, the idea of starting from the PV, SST, and pyramidal circuit, and then suddenly introducing the extra R neuron in Figure 5, just adds confusion. If the R neuron is meant to be the source, in practice, of certain inputs to some of the other cell types, then I think that should be included in the circuit from the start. Perhaps a good idea would be to start with two schematics, one in the form of Figure 5A (but with additional labeling for PV, SST) and one like Figure 1EF (but with auditory inputs as well), with a clear indication that the latter is meant to represent a preliminary, reduced form of the former that will be used in some initial tests of the performance of the PV, SST, UPE part of the circuit. Related to the Methods, I also can give a list of some specific complaints (in latex):

(1) φ, φP V are used in equations (10), (11), so they should be defined there, not many equations later.

Thank you, we changed that.

(2) β, 1 − β appear without justification or explanation in (11). That is finally defined and derived several pages later.

Thank you, we now define it right at the beginning.

(3) Equations (10)-(12) should be immediately followed by information about plasticity, rather than deferring that.

That’s a great idea. We changed it. Now the synaptic dynamics are explained together with the firing rate dynamics.

(4) After the rate equations (10)-(12) and weight change equations (23)-(25) are presented, the same equations are simply repeated in the “Explanation of the synaptic dynamics” subsection.

We agree that this was suboptimal. We moved the explanation of the synaptic dynamics up and removed the repetition.

(5) In the circuit model (13)-(19), it’s not clear why rR shows up in the SST+ and PV− equations vs. rs in PV+ and SST−. Moreover, rs is not even defined! Also, I don’t see why wP V +,R shows up in the equation for rP V − .

We added more explanation to the Methods section as to why the neurons receive these inputs and renamed rs to s, which is defined. The “+” in wP V +,R was a typo. Thank you for spotting that.

(6) The authors should only number those equations that they will reference by number. Even more importantly, there are many numbers such as (20), (26), (32), (39) that are just floating there without referring to an equation at all.

Thank you for spotting that. We corrected this.

(7) The authors fail to specify what is ra in Figure 8. Moreover, it seems strange to me that wP V,a approaches σ rather than wP V,ara approaching σ, since φP V is a function of wP V,ara.

You are right, wP V,ara should approach σ, but since ra is either 1 or 0 to indicate the presence of absence of the cue, and only wP V,a is plastic and changing„ wP V,a approaches σ.

(8) I don’t understand the rationale for the authors to introduce equation. (30) when they already had plasticity equations earlier. What is the relation of (30), (31) to (24)?

It is the same equation. In 30 we introduce simpler symbols for a better overview of the equations. 31 is equal to 30, with rP V replaced by it’s steady state.

(9) η is omitted from (33) - it won’t affect the final result but should be there.

We fixed this.

I have many additional specific comments and suggestions, some related to errors that really should have been caught before manuscript submission. I will present these based on the order in which they arise in the manuscript.

(1) In the abstract, the mention of layer 2/3 comes out of nowhere. Why this layer specifically? Is this meant to be an abstract/general cortical circuit model or to relate to a specific brain area? (Also watch for several minor grammatical issues in the abstract and later.)

Thank you for pointing this out. We now mention that the observed error neurons can be found in layer 2/3 of diverse brain areas. It is meant to be a general cortical circuit model independent of brain area.

(2) In par. 2 of the introduction, I find sentences 3-4 to be confusing and vague. Please rewrite what is meant more directly and clearly.

We tried to improve those sentences.

(3) Results subtitle 1: “suggests” → “suggest”

Thank you.

(4) Be careful to use math font whenever variables, such as a and N, are referenced (e.g., use of a instead of a bottom pg. 2).

We agree and checked the entire manuscript.

(5) Ref. to Fig. 1B bottom pg. 2 should be Fig. 1CD. The panel order in the figure should then be changed to match how it is referenced.

We fixed it and matched the ordering of the text with the ordering of the figure.

(6) Fig. 2C and 3E captions mention std but this is not shown in the figures - should be added.

It is there, it is just very small.

(7) Please clarify the relation of Figure 2C to 2F, and Figure 3F to 3H.

We colour-coded the points in 2F that correspond to the bars in 2C. We did the same for 3F and 3H.

(8) Figures 3E,3F appear to be identical except for the y-axis label and inclusion of std in 3F. Either more explanation is needed of how these relate or one should be cut.

The difference is that 3E shows the activity of PVs based on only the sound cue in the absence of a whisker stimulus. And 3F shows the activity of PVs based on both the sound cue and whisker stimuli. We state this more clearly now.

(9) Bottom of pg. 4: clarify that a quadratic φP V is a model assumption, not derived from results in the figure.

We added that we assume this.

(10) When k is referenced in the caption of Figure 4, the reader has no idea what it is. More substantially, most panels of Figure 4 are not referenced in the paper. I don’t understand what point the authors are trying to make here with much of this figure. Indeed, since the claim is that the uncertainy prediction should be based on division by σ2, why aren’t the numerical values for UPE rates much larger, since σ gets so small? The authors also fail to give enough details about the simulations done to obtain these plots; presumably these are after some sort of (unspecified) convergence, and in response to some sort of (unspecified) stimulus? Coming back to k, I don’t understand why k > 2 is used in addition to k = 2. The text mentions – even italicizes – “out-of-distribution dectection’, but this is never mentioned elsewhere in the paper and seems to be outside the true scope of the work (and not demonstrated in Figure 4). Sticking with k = 2 would also allow authors to simply use (·)k below (10), rather than the awkward positive part function that they have used now.

We now introduce the equation for the error neurons in Eq. 3 within the text, such that k is introduced before the caption. It also explains why the numerical values do not become much larger. Divisive inhibition, unlike mathematical division, cannot lead to multiplication in neurons. To ensure this, we add 1 to the denominator.

We show the error neuron responses to stimuli deviating from the learned mean after learning the mean and variance. The deviation is indicated either on the x-axis or in the legend depending on the plot. We now more explicitly state that these plots are obtained after learning the mean and the variance.

We removed the mentioning of the “out-of-distribution detection” as a detailed treatment would indeed be outside of the scope.

(11) Page 5, please clarify what is meant by “weights from the sound...”. You have introduced mathematical notation - use it so that you can be precise.

We added the mathematical notation, thank you!

(12) Figure 5D: legend has 5 entries but the figure panel only plots 4 quantities.

The SST firing rate was below the R firing rate. We hence omitted the SST firing rate and its legend.

(13) Figure 5: I don’t understand what point is being made about NMDA spikes. The text for Figure 5 refers to NMDA spikes in Figure 4, but nothing was said about NMDA spikes in the text for Figure 4 nor shown in Figure 4 itself.

We were referring to the nonlinearity in the activation function of UPEs in Figure 4. We changed the text to clarify this point.

(14) Figure 6: It is too difficult to distinguish the black and purple curves even on a large monitor. Also, the authors fail to define what they mean by “MM” and also do not define the quantities Y+ and Y− that they show. Another confusing aspect is that the model has PV+ and PV− neurons, so why doesn’t the figure?

Thank you for the comment. We changed the colour for better visibility, replaced the Upsilons with UPE (we changed the notation at some point and forgot to change it in the figure), and defined MM, which is the mismatch stimulus that causes error activity. We did not distinguish between PV+ and PV− in the plot as their activity is the same on average. We plotted the activity of the PV+. We now mention that we show the activity of PV+ as the representative.

(15) Also Figure 6: The authors do not make it clear in the text whether these are simulation results or cartoons. If the latter, please replace this with actual simulation results.

They are actual simulation results. We clarified this in the text.

(16) This work assumes the existence of two different populations of SST neurons with distinct UPE (pyramidal) targets. The Discussion doesn’t say much about any evidence for this assumption, which should be more thoroughly discussed and justified.

We now discuss this in more detail in the discussion as mentioned in our response to the public review.

(17) Par. 2 of the discussion refers to “Bayesian” and “Bayes-optimal” several times. Nothing was said earlier in the paper about a Bayesian framework for these results and it’s not clear what the authors mean by referring to Bayes here. This paragraph needs editing so that it clearly relates to the material of the results section and its implications.

We added an additional results section (the last section with Figure 8) on integrating prior and sensory information based on their uncertainties, which is also the case for Bayes-optimal integration, and show that our model can reproduce the central tendency effect, which is a hallmark of Bayes-optimal behaviour.

Reviewer 3 (Recommendations For The Authors):

See public review. I think the gradient-descent type of update the authors do in Equation (1) could be more useful in a more complicated learning scenario where the MLE has no closed form and has to be computed with gradient-based algorithms.

We responded in detail to your points in our point-by-point response to the public review.

Author response:

Joint Public Review:

Summary:

This study presents a strategy to efficiently isolate PcrV-specific BCRs from human donors with cystic fibrosis who have/had Pseudomonas aeruginosa (PA) infection. Isolation of mAbs that provide protection against PA may be a key to developing a new strategy to treat PA infection as the PA has intrinsic and acquired resistance to most antibiotic drug classes. Hale et al. developed fluorescently labeled antigen-hook and isolated mAbs with anti-PA activity. Overall, the authors' conclusion is supported by solid data analysis presented in the paper. Four of five recombinantly expressed PcrV-specific mAbs exhibited anti-PA activity in a murine pneumonia challenge model as potent as the V2L2MD mAb (equivalent to gremubamab). However, therapeutic potency for these isolated mAbs is uncertain as the gremubamab has failed in Phase 2 trials. Clarification of this point would greatly benefit this paper.

Strengths:

(1) High efficiency of isolating antigen-specific BCRs using an antigenic hook.

(2) The authors' conclusion is supported by data.

Weaknesses:

Although the authors state that the goal of this study was to generate novel protective mAbs for therapeutic use (P12; Para. 2), it is unclear whether PcrV-specific mAbs isolated in this study have therapeutic potential better than the gremubamab, which has failed in Phase 2 trials. Four of five PcrV-specific mAbs isolated in this study reduced bacterial burdens in mice as potent as, but not superior to, gremubamab-equivalent mAb. Clarification of this concern by revising the text or providing experimental results that show better potential than gremubamab would greatly benefit this paper.

The authors thank the reviewer for their thoughtful positive assessment. As noted by the reviewer, the studies described here, which were performed in mice, show that our MBC-derived mAbs are as effective as V2L2MD, a mAb that is one component of the gremubamab bi-specific. However, key theoretical strengths of MBC-derived mAbs (reduced immunogenicity, full participation in effector functions) are not easily tested in mice. We have clarified and expanded our discussion of these points in our revised manuscript, particularly in the Discussion paragraph 4.

Author response:

The following is the authors’ response to the previous reviews

We greatly appreciate all the reviewers’ constructive comments on our previously revised manuscript. In the current revision, we added several experimental data for answering the reviewers’ comments. Below we describe our point-by-point responses to their comments:

Reviewer #1 (Public Review):

Unaddressed and additional concerns (re-submission)

In this revised version of the manuscript, the authors have made important modifications in the text, inserted new references, and incorporated additional quantifications of cFos immunolabeling in three brain regions, as recommended by the reviewers. While these modifications have significantly improved the quality of the manuscript, other critical concerns raised during the initial submission of the

manuscript (Major concerns 1, 2, and 4; some of them also raised by the other reviewers) were not properly addressed by the authors. On several occasions, the authors recognize the importance of clarifying the points for the correct interpretation of the results but opt for leaving the open questions to be addressed during future studies. Therefore, the authors might consider adding a new section at the end of the manuscript to include all the caveats and future directions.

In the current revision, in order to answer the reviewer #1’s original concerns 1, 2, and 4, we added several experimental data.

Original major concerns 1) and 2): Regarding whether mice are detecting qualitative or quantitative differences between fresh and old cat saliva.

To address these concerns, as shown in new Figure 1I and J, we measured volumes of saliva contained in in individual swabs and total protein concentrations at the time of behavior tests: Fresh (15 minutes after collection) and Old (4 hours after collection). The saliva volumes at the time of behavioral testing were indistinguishable between fresh and old samples (Figure 1I). In addition, the concentrations of total proteins in both fresh and old saliva were also indiscernible (Figure 1J). Furthermore, we also examined the difference of the amount of Fel d 4 protein, one of the most abundant proteins in cat saliva, between fresh and old saliva by conducting western blotting analyses. As shown in new Supplemental Figure 2, the amount of Fel d 4 was nearly equivalent between fresh and old saliva. Indeed, our analyses using recombinant Fel d 4 protein showed that Fel d 4 does not induce freezing behavior (Supplemental Figure 5). Based on these findings, we believe that the difference between fresh and old cat saliva lies in specific components rather than the total or major saliva content. One possible explanation for this difference is the time-dependent reduction of specific freezing-inducing components in old saliva.

To investigate such a possibility, we also examined mouse behavior directed toward swabs containing diluted fresh cat saliva. Indeed, exposure to diluted fresh saliva resulted in a shorter duration of freezing behavior. Fresh saliva diluted to 70% induced freezing behavior for a duration equivalent to that of undiluted fresh saliva, while freezing behavior in response to 50% and 30% fresh saliva was significantly reduced to the same duration as that observed with old saliva (Figure 1K). The duration of direct interaction with swabs containing 70% and 50–30% fresh saliva also exhibited a similar trend to that observed with fresh and old saliva swabs, respectively (Figure 1L).

These new results provide compelling evidence that the differential freezing response of mice to fresh versus old cat saliva is not attributed to quantitative differences, such as total volume, total protein concentration, or the amount of major proteins like Fel d 4. However, when fresh saliva was diluted, we observed a corresponding reduction in freezing behavior, suggesting that specific components within the saliva—those responsible for inducing freezing—may decrease over time.

Our findings indicate that while the overall content of saliva remains consistent over time, specific freezing-inducing components seem to degrade or reduce at a different rate than other components, which alters the composition of saliva over time. The speed of reduction of these freezing-inducing components appears to be different from more stable proteins such as Fel d 4. As a result, the composition of saliva changes over time, leading to a qualitative difference between fresh and old saliva that mice can detect. This ability to discern such subtle chemical changes likely reflects an adaptive sensory mechanism, allowing mice to respond to predator cues to induce optimal defensive behavior in a certain context. Identifying the specific freezing-inducing components through traditional purification processes, such as high-performance liquid chromatography followed by behavioral examination (Haga-Yamanaka et al., 2014; Kimoto et al., 2005), is crucial for a deeper understanding of the mechanisms underlying the observed behavior. Our research team is actively working to isolate these molecules, and we hope to report our findings in future studies.

(4) The interpretation that fresh and old saliva activates different subpopulations of neurons in the VMH based on the observation that cFos positively correlates with freezing responses only with the fresh saliva lacks empirical evidence. To address this question, the authors should use two neuronal activity markers to track the response of the same population of VHM cells within the same animals during exposure to fresh vs. old saliva.

To address this issue, as shown in the new Figure 7, we performed a double exposure experiment using Fos2A-iCreERT2; Ai9 (TRAP2) mice (Allen et al., 2017; DeNardo et al., 2019). In this experiment, mice were exposed to the first stimulus under the treatment of 4-hydroxytamoxifen (4-OHT). One week after the initial exposure, the same mice were subjected to a second stimulus exposure for one hour. Through this paradigm, neurons activated by the first stimulus were visualized by tdTomato, while ones activated by the second stimulus were detected as cFos-IR (Figure 7A). Quantification of tdTomato and cFos-IR double-positive cells among tdTomato-labeled cells revealed that 43% (mean per animal: 61 / 143) of cells activated by fresh saliva during the first exposure were also activated by fresh saliva during the second exposure, whereas only 16% (17 / 106) of cells activated by old saliva during the first exposure were activated by fresh saliva during the second exposure (p = 7.5e-6, Chi-squared test). The difference in the fraction of overlapping cells between fresh and old saliva exposures was found significant when we compared the two groups of animals (Figure 7D, p = 0.0035, permutation test). Additionally, quantification of tdTomato and cFos-IR double-positive cells among cFos-IR cells indicated that over 27% (61 / 226) of cells activated by fresh saliva during the second exposure were previously activated by fresh saliva, whereas only 15% (17 / 112) of cells activated by fresh saliva during the second exposure were previously activated by old saliva (p = 0.015, Chi-squared test). The difference in the fraction of overlapping cells between fresh and old saliva exposures was also significant in this analysis (Figure 7E,p = 0.0060, permutation test). Together, these results demonstrate that fresh and old cat saliva activate largely different populations of neurons within the VMH. These new results were described on page 11 line 18 – page 12 line 8.

In addition to these unaddressed concerns, some new issues have emerged in the new version of the manuscript. For example, the following paragraph introduced in the discussion section is not supported by the experimental findings.

"We assume that such differential activations of the mitral cells between fresh and old saliva result in the differential activation of targeting neural substrates, possibly MeApv, which results in differential activation of VMH neurons (Figure 7)."

Although the authors did not observe statistical differences in cFos expression in the pvMeA among groups, they claim that the differences in cFos expression in the VMH between fresh vs. old saliva are mediated by differential activation of upstream neurons in the MeApv. The lack of statistical differences may be caused by the reduced number of subjects in each group, as recognized in the text by the

authors.

We appreciate the reviewer's thoughtful comment. We agree that the paragraph in the comment, which presented a working hypothesis regarding differential activations of mitral cells and the MeApv between fresh and old saliva exposures, was speculative and not fully supported by our experimental findings. To address this, we have removed the assumptions related to the differential responses of mitral cells and the MeApv from the discussion and have updated the figure accordingly (now presented as new Figure 8).

Moreover, the authors propose that in addition to fel d 4, multiple molecules present in the cat saliva can be inducing distinct defensive responses in the animals, but they do not provide any reference to support their claim.

We thank the reviewer for highlighting this point. Our claim regarding the presence of other molecules in cat saliva inducing freezing defensive responses is based on our observation, as shown in the new Supplemental Figure 5, that recombinant Fel d 4 protein alone does not induce freezing behavior. This suggests the existence of other unidentified components in cat saliva that may contribute to freezing behavior. As we agree that identifying these specific freezing-inducing components is important for a more comprehensive understanding of the underlying mechanisms, our research team is actively working to isolate these molecules, and we hope to report our findings in future studies.

Reviewer #2 (Public Review):

The findings are relatively preliminary. The identities of the receptor and the ligand in the cat saliva that induces the behavior remain unclear. The identity of VMH cells that are activated by the cat saliva remains unclear. There is a lack of targeted functional manipulation to demonstrate the role of V2R-A4 or VMH cells in the behavioral response to the cat saliva.

We thank the reviewer’s important insight on the need for further investigation into the molecular and neural mechanisms underlying the behavioral response to cat saliva. We recognize the importance of conducting studies involving V2R-A4 receptor knockouts and targeted functional manipulations within the VMH using neural circuit perturbation approaches.

However, the V2R-A4 subfamily consists of 25 Vmn2r genes, most of which are closely grouped together, forming a V2R-A4 gene cluster within a 2.5-megabase chromosomal region. As we described in our recent review article (Rocha et al., 2024), the Vmn2r genes within the V2R-A4 subfamily display a high degree of homology, with nucleotide and amino acid identities among the several Vmn2rs surpassing 97-99%, suggesting possible redundancy among these receptor genes. This is in stark contrast to the diversity typically observed within other V2R subfamilies. Consequently, knockout strategies targeting a single receptor gene, which have been successful for other vomeronasal receptors, may not be effective for V2R-A4 receptor genes. The most appropriate strategy for examining the necessity of V2R-A4 receptors would be knocking out the entire V2R-A4 gene cluster, spanning a 2.5-megabase chromosomal region. Due to the technical challenges involved, addressing this issue is not feasible in the foreseeable future. Moreover, in our current study, we aimed to establish the foundational relationship between predator cues in cat saliva and defensive behaviors. We view our findings as an important first step that sets the stage for these more targeted and mechanistic studies involving the neural circuit perturbation experiments, such as optogenetics and Designer Receptors Exclusively Activated by Designer Drugs (DREADDs), in the next step.

Reviewer #3 (Public Review):

Weaknesses:

(1)  It is unclear if fresh and old saliva indeed alter the perceived imminence of predation, as claimed by the authors. Prior work indicates that lower imminence induces anxiety-related actions, such as re- organization of meal patterns and avoidance of open spaces, while slightly higher imminence produces freezing. Here, the authors show that fresh and old predator saliva only provoke different amounts of freezing, rather than changing the topography of defensive behaviors, as explained above. Another prediction of predatory imminence theory would be that lower imminence induced by old saliva should produce stronger cortical activation, while fresh saliva would activate amygdala, if these stimuli indeed correspond to significantly different levels of predation imminence.

We appreciate the reviewer’s insightful comments regarding the perceived imminence of predation and the behavioral responses to fresh and old saliva. Our study specifically focused on comparing the defensive behaviors of mice in response to 15-minute-old and 4-hour-old cat saliva, particularly within the context of freezing behavior in their home cages. We chose these specific time points to capture the potential variation in behavioral intensity rather than the full spectrum of defensive behaviors. While a more comprehensive analysis—including varying time points, different types of defensive behaviors, and broader neural activation patterns (e.g., cortical versus amygdala activation)—might provide further insights into predation imminence theory, these aspects were beyond the scope of our current study. Future research could certainly address these points by examining behavioral and neural responses across additional saliva aging intervals and in varied behavioral contexts. Such studies would complement and extend the findings presented here, further elucidating the relationship between predator cue characteristics and defensive behaviors.

(2)  It is known that predator odors activate and require AOB, VNO and VMH, thus replications of these findings are not novel, decreasing the impact of this work.

As the reviewer mentioned, the activation of the AOB, VNO, and VMH by predator odors has been established in prior studies. However, our study provides new insights by demonstrating that defensive freezing behavior in response to predator odors is mediated through the vomeronasal organ (VNO) sensory circuit, which has not been previously shown. The novelty of our work lies in two key findings: 1) the introduction of a new behavioral paradigm that assesses freezing responses to predator cues based on the freshness of chemosensory signals in cat saliva, and 2) the demonstration that the vomeronasal sensory circuit mediates defensive freezing behavior in response to cat saliva.

Additionally, our results show that cat saliva of different freshness levels differentially activates VNO sensory neurons that express the same subfamily of sensory receptors. This differential activation subsequently modulates the downstream neural circuits, leading to varied freezing behavioral outcomes. We believe these findings provide a novel conceptual advance over previous studies by elucidating a more detailed mechanism of how predator-derived cues influence defensive behaviors through the accessory olfactory system.

(3)  There is a lack of standard circuit dissection methods, such as characterizing the behavioral effects of increasing and decreasing neural activity of relevant cell bodies and axonal projections, significantly decreasing the mechanistic insights generated by this work

We thank the reviewer for this valuable comment. Investigating the behavioral effects of manipulating specific cell types and axonal projections, as well as characterizing circuit connectivity, is essential for a more comprehensive understanding of the underlying neural circuits. These approaches, such as modulating neural activity in defined cell populations and dissecting circuit pathways, using optogenetics, DREADD, etc., would provide deeper mechanistic insights. In our current study, however, we aimed to establish the foundational relationship between predator cues in cat saliva and defensive behaviors. We view our findings as an important first step that sets the stage for these more targeted and mechanistic studies in the future.

(4)  The correlation shown in Figure 5c may be spurious. It appears that the correlation is primarily driven by a single point (the green square point near the bottom left corner). All correlations should be calculated using Spearman correlation, which is non-parametric and less likely to show a large correlation due to a small number of outliers. Regardless of the correlation method used, there are too few points in Figure 5c to establish a reliable correlation. Please add more points to 5c.

We appreciate the reviewer’s suggestion regarding the correlation analysis in Figure 5E. We assessed the normality of our data using both the Shapiro-Wilk and Kolmogorov-Smirnov tests, which confirmed that the dataset is parametric, justifying the use of a parametric correlation method in this context. However, we acknowledge the concern about the limited number of data points and the influence of potential outliers on the observed correlation. Increasing the sample size might provide a more robust assessment of correlation patterns and reduce the potential impact of any single data point. While this would be an important direction for future research, such as with larger sample sizes, it is beyond the scope of the current study.

(5)  Please cite recent relevant papers showing VMH activity induced by predators, such as https://pubmed.ncbi.nlm.nih.gov/33115925/ and https://pubmed.ncbi.nlm.nih.gov/36788059/

We thank the reviewer’s suggestion to cite these important papers. https://pubmed.ncbi.nlm.nih.gov/33115925/ (Esteban Masferrer et al., 2020) and https://pubmed.ncbi.nlm.nih.gov/36788059/ (Tobias et al., 2023) are now cited at page 16 line 10 in Discussion under “Differential activation of VMH neurons potentially underlying distinct intensities of freezing behavior.”

(6)  Add complete statistical information in the figure legends of all figures, which should include n, name of test used and exact p values.

We included statistical analysis results in figure legends; for Figure 6B, we provided statistical analysis results in Supplemental Table 1.

(7)  Some of the findings are disconnected from the story. For example, the authors show V2R-A4- expressing cells are activated by predator odors. Are these cells more likely to be connected to the rest of the predatory defense circuit than other VNO cells?

Yes, our hypothesis posits that V2R-A4-expressing VNO sensory neurons serve as receptor neurons for predator cues present in cat saliva. Additionally, we assume that these specific sensory neurons have stronger anatomical connections with the defensive circuit compared to VNO sensory neurons expressing other receptor subfamilies. In our modified Discussion section, we discussed this point under “V2R-A4 subfamily as the receptor for predator cues in cat saliva.”

(8)  Please paste all figure legends directly below their corresponding figure to make the manuscript easier to read

We have added figure legends directly below their corresponding figures.

(9)  Were there other behavioral differences induced by fresh compared to old saliva? Do they provoke differences in stretch-attend risk evaluation postures, number of approaches, average distance to odor stimulus, velocity of movements towards and away the odor stimulus, etc?

We appreciate the reviewer's valuable comments. We have now incorporated an analysis of stretch-sniff risk assessment behavior, presented in new Figure 1F (graph) and Supplemental Figure 1B (raster plot). Mice exhibited stretch-sniff risk assessment behavior, which remained consistent across control, fresh saliva, and old saliva swabs. Additionally, we have also included a raster plot for direct investigation, previously noted as ‘interaction’ in the original manuscript (Supplemental Figure 1C). Mice exposed to a swab containing either fresh or old saliva significantly avoided directly investigating the swab. In contrast, mice exposed to a clean control swab spent a significant amount of time directly investigating the swab, engaging in behaviors such as sniffing and chewing (Figure 1G). A comparison of temporal behavioral patterns revealed a slightly higher frequency of direct investigation behavior toward old saliva compared to fresh saliva at the beginning of the exposure period (Supplemental Figure 1C).

Reviewer #3 (Recommendations For The Authors):

The authors have partially addressed several important points raised in the prior review, increasing the strength of the manuscript. However, 2 key questions already raised previously, were not addressed:

(1)  Is old saliva qualitatively different from new saliva, or is it the same as a smaller amount of new saliva? As Reviewer 1 wrote: "An important point that the authors should clarify in this study is whether mice are detecting qualitative or quantitative differences between fresh and old cat saliva."

Since one of the author's main points is that fresh and old saliva elicit different perceived threat imminences, it is crucial to show that these two stimuli are somehow qualitatively different.

One way to investigate this could be to show that animals perform different behaviors when exposed to smaller among of new saliva vs old saliva, or that the cfos activation patterns are different in these two conditions.

The answers to these concerns are provided in the Public review Comment from Reviewer #1.

(2)  The other key question is if different VMH populations are activated by new vs old saliva.

The answer to this concern is provided in the Public Review comment from Reviewer #1.

Lastly, although the new analysis and text changes improved the manuscript, many issues raised were addressed with some variation of 'future studies will be done', or 'we concur with the Reviewer'. However, the extra experiments required to answer these questions were not done. For this reason, even though the authors have numerous exciting pieces of data, overall the work is still incomplete. I highlight below some examples in which the authors agree with the Reviewer, but do not answer the question with the new work that would be required, or propose to do the work in future studies.

In this revised manuscript, we have conducted several additional experiments to address key concerns raised by the reviewers that are directly relevant to our claims. Specifically, we have examined: 1) whether qualitative or quantitative differences between fresh and old cat saliva are detected by mice to modulate behavior (NEW Figure 1I, J, K, and L, and NEW Supplemental Figure 2); 2) the involvement of Fel d 4 in freezing behavior (NEW Supplemental Figure 5); and 3) whether different VMH populations are activated by fresh versus old saliva (NEW Figure 7). However, some concerns raised by the reviewers fall outside the scope of the current manuscript. These include: 1) identifying the specific components that induce freezing, 2) examining the necessity of V2R-A4 receptors, 3) conducting neural circuit perturbations, and 4) performing a comprehensive analysis—including varying time points, different types of defensive behaviors, and broader neural activation patterns (e.g., cortical versus amygdala activation)—of the mouse’s defensive response to different levels of predator threat imminence. As these aspects are beyond the focus of our current manuscript, we have noted in the Public Review comments.

References:

Allen WE, DeNardo LA, Chen MZ, Liu CD, Loh KM, Fenno LE, Ramakrishnan C, Deisseroth K, Luo L. 2017. Thirst-associated preoptic neurons encode an aversive motivational drive. Science 357:1149– 1155.

DeNardo LA, Liu CD, Allen WE, Adams EL, Friedmann D, Fu L, Guenthner CJ, Tessier-Lavigne M, Luo L. 2019. Temporal evolution of cortical ensembles promoting remote memory retrieval. Nat Neurosci 22:460–469.

Haga-Yamanaka S, Ma L, He J, Qiu Q, Lavis LD, Looger LL, Yu CR. 2014. Integrated action of pheromone signals in promoting courtship behavior in male mice. Elife 3:e03025.

Kimoto H, Haga S, Sato K, Touhara K. 2005. Sex-specific peptides from exocrine glands stimulate mouse vomeronasal sensory neurons. Nature 437:898–901.

Rocha A, Nguyen QAT, Haga-Yamanaka S. 2024. Type 2 vomeronasal receptor-A4 subfamily: Potential predator sensors in mice. Genesis 62:e23597.

Author response:

Reviewer #1 (Public review):

This manuscript from Schwintek and coworkers describes a system in which gas flow across a small channel (10^-4-10^-3 m scale) enables the accumulation of reactants and convective flow. The authors go on to show that this can be used to perform PCR as a model of prebiotic replication.

Strengths:

The manuscript nicely extends the authors' prior work in thermophoresis and convection to gas flows. The demonstration of nucleic acid replication is an exciting one, and an enzyme-catalyzed proof-of-concept is a great first step towards a novel geochemical scenario for prebiotic replication reactions and other prebiotic chemistry.

The manuscript nicely combines theory and experiment, which generally agree well with one another, and it convincingly shows that accumulation can be achieved with gas flows and that it can also be utilized in the same system for what one hopes is a precursor to a model prebiotic reaction. This continues efforts from Braun and Mast over the last 10-15 years extending a phenomenon that was appreciated by physicists and perhaps underappreciated in prebiotic chemistry to increasingly chemically relevant systems and, here, a pilot experiment with a simple biochemical system as a prebiotic model.

I think this is exciting work and will be of broad interest to the prebiotic chemistry community.

Weaknesses:

The manuscript states: "The micro scale gas-water evaporation interface consisted of a 1.5 mm wide and 250 µm thick channel that carried an upward pure water flow of 4 nl/s ≈ 10 µm/s perpendicular to an air flow of about 250 ml/min ≈ 10 m/s." This was a bit confusing on first read because Figure 2 appears to show a larger channel - based on the scale bar, it appears to be about 2 mm across on the short axis and 5 mm across on the long axis. From reading the methods, one understands the thickness is associated with the Teflon, but the 1.5 mm dimension is still a bit confusing (and what is the dimension in the long axis?) It is a little hard to tell which portion (perhaps all?) of the image is the channel. This is because discontinuities are present on the left and right sides of the experimental panels (consistent with the image showing material beyond the channel), but not the simulated panels. Based on the authors' description of the apparatus (sapphire/CNC machined Teflon/sapphire) it sounds like the geometry is well-known to them. Clarifying what is going on here (and perhaps supplying the source images for the machined Teflon) would be helpful.

We understand. We will update the figures to better show dimensions of the experimental chamber. We will also add a more complete Figure in the supplementary information. Part of the complexity of the chamber however stems from the fact that the same chamber design has also been used to create defined temperature gradients which are not necessary and thus the chamber is much more complex than necessary.

The data shown in Figure 2d nicely shows nonrandom residuals (for experimental values vs. simulated) that are most pronounced at t~12 m and t~40-60m. It seems like this is (1) because some symmetry-breaking occurs that isn't accounted for by the model, and perhaps (2) because of the fact that these data are n=1. I think discussing what's going on with (1) would greatly improve the paper, and performing additional replicates to address (2) would be very informative and enhance the paper. Perhaps the negative and positive residuals would change sign in some, but not all, additional replicates?

To address this, we will show two more replicates of the experiment and include them in Figure 2.

We are seeing two effects when we compare fluorescence measurements of the experiments.

Firstly, degassing of water causes the formation of air-bubbles, which are then transported upwards to the interface, disrupting fluorescence measurements. This, however, mostly occurs in experiments with elevated temperatures for PCR reactions, such as displayed in Figure 4.

Secondly, due to the high surface tension of water, the interface is quite flexible. As the inflow and evaporation work to balance each other, the shape of the interface adjusts, leading to alterations in the circular flow fields below.

Thus the conditions, while overall being in steady state, show some fluctuations. The strong dependence on interface shape is also seen in the simulation. However, modeling a dynamic interface shape is not so easy to accomplish, so we had to stick to one geometry setting. Again here, the added movies of two more experiments should clarify this issue.

The authors will most likely be familiar with the work of Victor Ugaz and colleagues, in which they demonstrated Rayleigh-Bénard-driven PCR in convection cells (10.1126/science.298.5594.793, 10.1002/anie.200700306). Not including some discussion of this work is an unfortunate oversight, and addressing it would significantly improve the manuscript and provide some valuable context to readers. Something of particular interest would be their observation that wide circular cells gave chaotic temperature profiles relative to narrow ones and that these improved PCR amplification (10.1002/anie.201004217). I think contextualizing the results shown here in light of this paper would be helpful.

Thanks for pointing this out and reminding us. We apologize. We agree that the chaotic trajectories within Rayleigh-Bénard convection cells lead to temperature oscillations similar to the salt variations in our gas-flux system. Although the convection-driven PCR in Rayleigh-Bénard is not isothermal like our system, it provides a useful point of comparison and context for understanding environments that can support full replication cycles. We will add a section comparing approaches and giving some comparison into the history of convective PCR and how these relate to the new isothermal implementation.

Again, it appears n=1 is shown for Figure 4a-c - the source of the title claim of the paper - and showing some replicates and perhaps discussing them in the context of prior work would enhance the manuscript.

We appreciate the reviewer for bringing this to our attention. We will now include the two additional repeats for the data shown in Figure 4c, while the repeats of the PAGE measurements are already displayed in Supplementary Fig. IX.2. Initially, we chose not to show the repeats in Figure 4c due to the dynamic and variable nature of the system. These variations are primarily caused by differences at the water-air interface, attributed to the high surface tension of water. Additionally, the stochastic formation of air bubbles in the inflow—despite our best efforts to avoid them—led to fluctuations in the fluorescence measurements across experiments. These bubbles cause a significant drop in fluorescence in a region of interest (ROI) until the area is refilled with the sample.

Unlike our RNA-focused experiments, PCR requires high temperatures and degassing a PCR master mix effectively is challenging in this context. While we believe our chamber design is sufficiently gas-tight to prevent air from diffusing in, the high surface-to-volume ratio in microfluidics makes degassing highly effective, particularly at elevated temperatures. We anticipate that switching to RNA experiments at lower temperatures will mitigate this issue, which is also relevant in a prebiotic context.

The reviewer’s comments are valid and prompt us to fully display these aspects of the system. We will now include these repeats in Figure 4c to give readers a deeper understanding of the experiment's dynamics. Additionally, we will provide videos of all three repeats, allowing readers to better grasp the nature of the fluctuations in SYBR Green fluorescence depicted in Figure 4c.

I think some caution is warranted in interpreting the PCR results because a primer-dimer would be of essentially the same length as the product. It appears as though the experiment has worked as described, but it's very difficult to be certain of this given this limitation. Doing the PCR with a significantly longer amplicon would be ideal, or alternately discussing this possible limitation would be helpful to the readers in managing expectations.

This is a good point and should be discussed more in the manuscript. Our gel electrophoresis is capable of distinguishing between replicate and primer dimers. We know this since we were optimizing the primers and template sequences to minimize primer dimers, making it distinguishable from the desired 61mer product. That said, all of the experiments performed without a template strand added did not show any band in the vicinity of the product band after 4h of reaction, in contrast to the experiments with template, presenting a strong argument against the presence of primer dimers.

Reviewer #2 (Public review):

Schwintek et al. investigated whether a geological setting of a rock pore with water inflow on one end and gas passing over the opening of the pore on the other end could create a non-equilibrium system that sustains nucleic acid reactions under mild conditions. The evaporation of water as the gas passes over it concentrates the solutes at the boundary of evaporation, while the gas flux induces momentum transfer that creates currents in the water that push the concentrated molecules back into the bulk solution. This leads to the creation of steady-state regions of differential salt and macromolecule concentrations that can be used to manipulate nucleic acids. First, the authors showed that fluorescent bead behavior in this system closely matched their fluid dynamic simulations. With that validation in hand, the authors next showed that fluorescently labeled DNA behaved according to their theory as well. Using these insights, the authors performed a FRET experiment that clearly demonstrated the hybridization of two DNA strands as they passed through the high Mg++ concentration zone, and, conversely, the dissociation of the strands as they passed through the low Mg++ concentration zone. This isothermal hybridization and dissociation of DNA strands allowed the authors to perform an isothermal DNA amplification using a DNA polymerase enzyme. Crucially, the isothermal DNA amplification required the presence of the gas flux and could not be recapitulated using a system that was at equilibrium. These experiments advance our understanding of the geological settings that could support nucleic acid reactions that were key to the origin of life.

The presented data compellingly supports the conclusions made by the authors. To increase the relevance of the work for the origin of life field, the following experiments are suggested:

(1) While the central premise of this work is that RNA degradation presents a risk for strand separation strategies relying on elevated temperatures, all of the work is performed using DNA as the nucleic acid model. I understand the convenience of using DNA, especially in the latter replication experiment, but I think that at least the FRET experiments could be performed using RNA instead of DNA.

We understand the request only partially. The modification brought about by the two dye molecules in the FRET probe to be able to probe salt concentrations by melting is of course much larger than the change of the backbone from RNA to DNA. This was the reason why we rather used the much more stable DNA construct which is also manufactured at a lower cost and in much higher purity also with the modifications. But we think the melting temperature characteristics of RNA and DNA in this range is enough known that we can use DNA instead of RNA for probing the salt concentration in our flow cycling.

Only at extreme conditions of pH and salt, RNA degradation through transesterification, especially under alkaline conditions is at least several orders of magnitude faster than spontaneous degradative mechanisms acting upon DNA [Li, Y., & Breaker, R. R. (1999). Kinetics of RNA degradation by specific base catalysis of transesterification involving the 2 ‘-hydroxyl group. Journal of the American Chemical Society, 121(23), 5364-5372.]. The work presented in this article is however focussed on hybridization dynamics of nucleic acids. Here, RNA and DNA share similar properties regarding the formation of double strands and their respective melting temperatures. While RNA has been shown to form more stable duplex structures exhibiting higher melting temperatures compared to DNA [Dimitrov, R. A., & Zuker, M. (2004). Prediction of hybridization and melting for double-stranded nucleic acids. Biophysical Journal, 87(1), 215-226.], the general impact of changes in salt, temperature and pH [Mariani, A., Bonfio, C., Johnson, C. M., & Sutherland, J. D. (2018). pH-Driven RNA strand separation under prebiotically plausible conditions. Biochemistry, 57(45), 6382-6386.] on respective melting temperatures follows the same trend for both nucleic acid types. Also the diffusive properties of RNA and DNA are very similar [Baaske, P., Weinert, F. M., Duhr, S., Lemke, K. H., Russell, M. J., & Braun, D. (2007). Extreme accumulation of nucleotides in simulated hydrothermal pore systems. Proceedings of the National Academy of Sciences, 104(22), 9346-9351.].

Since this work is a proof of principle for the discussed environment being able to host nucleic acid replication, we aimed to avoid second order effects such as degradation by hydrolysis by using DNA as a proxy polymer. This enabled us to focus on the physical effects of the environment on local salt and nucleic acid concentration. The experiments performed with FRET are used to visualize local salt concentration changes and their impact on the melting temperature of dissolved nucleic acids.  While performing these experiments with RNA would without doubt cover a broader application within the field of origin of life, we aimed at a step-by-step / proof of principle approach, especially since the environmental phenomena studied here have not been previously investigated in the OOL context. Incorporating RNA-related complexity into this system should however be addressed in future studies. This will likely require modifications to the experimental boundary conditions, such as adjusting pH, temperature, and salt concentration, to account for the greater duplex stability of RNA. For instance, lowering the pH would reduce the RNA melting temperature [Ianeselli, A., Atienza, M., Kudella, P. W., Gerland, U., Mast, C. B., & Braun, D. (2022). Water cycles in a Hadean CO2 atmosphere drive the evolution of long DNA. Nature Physics, 18(5), 579-585.].

(2) Additionally, showing that RNA does not degrade under the conditions employed by the authors (I am particularly worried about the high Mg++ zones created by the flux) would further strengthen the already very strong and compelling work.

Based on literature values for hydrolysis rates of RNA [Li, Y., & Breaker, R. R. (1999). Kinetics of RNA degradation by specific base catalysis of transesterification involving the 2 ‘-hydroxyl group. Journal of the American Chemical Society, 121(23), 5364-5372.], we estimate RNA to have a halflife of multiple months under the deployed conditions in the FRET experiment (High concentration zones contain <1mM of Mg2+). Additionally, dsRNA is multiple orders of magnitude more stable than ssRNA with regards to degradation through hydrolysis [Zhang, K., Hodge, J., Chatterjee, A., Moon, T. S., & Parker, K. M. (2021). Duplex structure of double-stranded RNA provides stability against hydrolysis relative to single-stranded RNA. Environmental Science & Technology, 55(12), 8045-8053.], improving RNA stability especially in zones of high FRET signal. Furthermore, at the neutral pH deployed in this work, RNA does not readily degrade. In previous work from our lab [Salditt, A., Karr, L., Salibi, E., Le Vay, K., Braun, D., & Mutschler, H. (2023). Ribozyme-mediated RNA synthesis and replication in a model Hadean microenvironment. Nature Communications, 14(1), 1495.], we showed that the lifetime of RNA under conditions reaching 40mM Mg2+ at the air-water interface at 45°C was sufficient to support ribozymatically mediated ligation reactions in experiments lasting multiple hours.

With that in mind, gaining insight into the median Mg2+ concentration across multiple averaged nucleic acid trajectories in our system (see Fig. 3c&d) and numerically convoluting this with hydrolysis dynamics from literature would be highly valuable. We anticipate that longer residence times in trajectories distant from the interface will improve RNA stability compared to a system with uniformly high Mg2+ concentrations.

(3) Finally, I am curious whether the authors have considered designing a simulation or experiment that uses the imidazole- or 2′,3′-cyclic phosphate-activated ribonucleotides. For instance, a fully paired RNA duplex and a fluorescently-labeled primer could be incubated in the presence of activated ribonucleotides +/- flux and subsequently analyzed by gel electrophoresis to determine how much primer extension has occurred. The reason for this suggestion is that, due to the slow kinetics of chemical primer extension, the reannealing of the fully complementary strands as they pass through the high Mg++ zone, which is required for primer extension, may outcompete the primer extension reaction. In the case of the DNA polymerase, the enzymatic catalysis likely outcompetes the reannealing, but this may not recapitulate the uncatalyzed chemical reaction.

This is certainly on our to-do list. Our current focus is on templated ligation rather than templated polymerization and we are working hard to implement RNA-only enzyme-free ligation chain reaction, based on more optimized parameters for the templated ligation from 2’3’-cyclic phosphate activation that was just published [High-Fidelity RNA Copying via 2′,3′-Cyclic Phosphate Ligation, Adriana C. Serrão, Sreekar Wunnava, Avinash V. Dass, Lennard Ufer, Philipp Schwintek, Christof B. Mast, and Dieter Braun, JACS doi.org/10.1021/jacs.3c10813 (2024)]. But we first would try this at an air-water interface which was shown to work with RNA in a temperature gradient [Ribozyme-mediated RNA synthesis and replication in a model Hadean microenvironment, Annalena Salditt, Leonie Karr, Elia Salibi, Kristian Le Vay, Dieter Braun & Hannes Mutschler, Nature Communications doi.org/10.1038/s41467-023-37206-4 (2023)] before making the jump to the isothermal setting we describe here. So we can understand the question, but it was good practice also in the past to first get to know the setting with PCR, then jump to RNA.

Reviewer #2 (Recommendations for the authors):

(1) Could the authors comment on the likelihood of the geological environments where the water inflow velocity equals the evaporation velocity?

This is an important point to mention in the manuscript, thank you for pointing that out. To produce a defined experiment, we were pushing the water out with a syringe pump, but regulated in a way that the evaporation was matching our flow rate. We imagine that a real system will self-regulate the inflow of the water column on the one hand side by a more complex geometry of the gas flow, matching the evaporation with the reflow of water automatically. The interface would either recede or move closer to the gas flux, depending on whether the inflow exceeds or falls short of the evaporation rate. As the interface moves closer, evaporation speeds up, while moving away slows it down. This dynamic process stabilizes the system, with surface tension ultimately fixing the interface in place.

We have seen a bit of this dynamic already in the experiments, could however so far not yet find a good geometry within our 2-dimensional constant thickness geometry to make it work for a longer time. Very likely having a 3-dimensional reservoir of water with less frictional forces would be able to do this, but this would require a full redesign of a multi-thickness microfluidics. The more we think about it, the more we envisage to make the next implementation of the experiment with a real porous volcanic rock inside a humidity chamber that simulates a full 6h prebiotic day. But then we would lose the whole reproducibility of the experiment, but likely gain a way that recondensation of water by dew in a cold morning is refilling the water reservoirs in the rocks again. Sorry that I am regressing towards experiments in the future.

(2) Could the authors speculate on using gases other than ambient air to provide the flux and possibly even chemical energy? For example, using carbonyl sulfide or vaporized methyl isocyanide could drive amino acid and nucleotide activation, respectively, at the gas-water interface.

This is an interesting prospect for future work with this system. We thought also about introducing ammonia for pH control and possible reactions. We were amazed in the past that having CO2 instead of air had a profound impact on the replication and the strand separation [Water cycles in a Hadean CO2 atmosphere drive the evolution of long DNA, Alan Ianeselli, Miguel Atienza, Patrick Kudella, Ulrich Gerland, Christof Mast & Dieter Braun, Nature Physics doi.org/10.1038/s41567-022-01516-z (2022)]. So going more in this direction absolutely makes sense and as it acts mostly on the length-selectively accumulated molecules at the interface, only the selected molecules will be affected, which adds to the selection pressure of early evolutionary scenarios.

Of course, in the manuscript, we use ambient air as a proxy for any gas, focusing primarily on the energy introduced through momentum transfer and evaporation. We speculate that soluble gasses could establish chemical gradients, such as pH or redox potential, from the bulk solution to the interface, similar to the Mg2+ accumulation shown in Figure 3c. The nature of these gradients would depend on each gas's solubility and diffusivity. We have already observed such effects in thermal gradients [Keil, L. M., Möller, F. M., Kieß, M., Kudella, P. W., & Mast, C. B. (2017). Proton gradients and pH oscillations emerge from heat flow at the microscale. Nature communications, 8(1), 1897.] and finding similar behavior in an isothermal environment would be a significant discovery.

(3) Line 162: Instead of "risk," I suggest using "rate".

Oh well - thanks for pointing this out! Will be changed.

(4) Using FRET of a DNA duplex as an indicator of salt concentration is a decent proxy, but a more direct measurement of salt concentration would provide further merit to the explicit statement that it is the salt concentration that is changing in the system and not another hidden parameter.

Directly observing salt concentration using microscopy is a difficult task. While there are dyes that change their fluorescence depending on the local Na+ or Mg2+ concentration, they are not operating differentially, i.e. by making a ratio between two color channels. Only then we are not running into artifacts from the dye molecules being accumulated by the non-equilibrium settings. We were able to do this for pH in the past, but did not find comparable optical salt sensors. This is the reason we ended up with a FRET pair, with the advantage that we actually probe the strand separation that we are interested in anyhow. Using such a dye in future work would however without a doubt enhance the understanding of not only this system, but also our thermal gradient environments.

(5) Figure 3a: Could the authors add information on "Dried DNA" to the caption? I am assuming this is the DNA that dried off on the sides of the vessel but cannot be sure.

Thanks to the reviewer for pointing this out. This is correct and we will describe this better in the revised manuscript.

(6) Figure 4b and c: How reproducible is this data? Have the authors performed this reaction multiple independent times? If so, this data should be added to the manuscript.

The data from the gel electrophoresis was performed in triplicates and is shown in full in supplementary information. The data in c is hard to reproduce, as the interface is not static and thus ROI measurements are difficult to perform as an average of repeats. Including the data from the independent repeats will however give the reader insight into some of the experimental difficulties, such as air bubbles, which form from degassing as the liquid heats up, that travel upwards to the interface, disrupting the ongoing fluorescence measurements.

(7) Line 256: "shielding from harmful UV" statement only applies to RNA oligomers as UV light may actually be beneficial for earlier steps during ribonucleoside synthesis. I suggest rephrasing to "shielding nucleic acid oligomers from UV damage.".

Will be adjusted as mentioned.

(8) The final paragraph in the Results and Discussion section would flow better if placed in the Conclusion section.

This is a good point and we will merge results and discussion closer together.

(9) Line 262, "...of early Life" is slightly overstating the conclusions of the study. I suggest rephrasing to "...of nucleic acids that could have supported early life."

This is a fair comment. We thank the reviewer for his detailed analysis of the manuscript!

(10) In references, some of the journal names are in sentence case while others are in title case (see references 23 and 26 for example).

Thanks - this will be fixed.

Author response:

Reviewer #1 (Public review):

Summary:

Lejeune et al. demonstrated sex-dependent differences in the susceptibility to MRSA infection. The authors demonstrated the role of the microbiota and sex hormones as potential determinants of susceptibility. Moreover, the authors showed that Th17 cells and neutrophils contribute to sex hormone-dependent protection in female mice.

Strengths:

The role of microbiota was examined in various models (gnotobiotic, co-housing, microbiota transplantation). The identification of responsible immune cells was achieved using several genetic knockouts and cell-specific depletion models. The involvement of sex hormones was clarified using ovariectomy and the FCG model.

Weaknesses:

The mechanisms by which specific microbiota confer female-specific protection remain unclear.

We thank the reviewer for highlighting the strength of the manuscript including the models and techniques we employ. We agree that the relationship between the microbiota and sex-dependent protection is less developed compared with other aspects of the study. In preparation of a revised manuscript, we intend on performing a more thorough comparison of male vs. female microbiota, along with quantification of sex hormones and downstream Th17 function (neutrophil recruitment and activation).

Reviewer #2 (Public review):

Overall, the paper nicely adds to the growing body of literature investigating how biological sex impacts the immune system and the burden of infectious disease. The conclusions are mostly supported by the data although there are some aspects of the data that could be better addressed and clarified.

We thank the reviewer for appreciating our contribution. We intend on performing experiments to fill-in gaps and text revisions to increase clarity and acknowledge limitations.

(1) There is something of a disconnect between the initial microbiome data and the later data that analyzes sex hormones and chromosomes. While there are clearly differences in microbial species across the two sites (NYU and JAX) how these bacterial species might directly interact with immune cells to induce female-specific responses is left unexplored. At the very least it would help to try and link these two distinct pieces of data to try and inform the reader how the microbiome is regulating the sex-specific response. Indeed, the reader is left with no clear exploration of the microbiota's role in the persistence of the infection and thus is left wanting.

We agree. This comment is similar to Reviewer #1’s feedback. As mentioned above, we anticipate clarifying the association between sex differences and the microbiota. We will attempt to investigate specific bacteria, although some aspects of microbiota characterization may be outside the timeframe of the revision.

(2) While the authors make a reasonable case that Th17 T cells are important for controlling infection (using RORgt knockout mice that cannot produce Th17 cells), it is not clear how these cells even arise during infection since the authors make most of the observations 2 days post-infection which is longer before a normal adaptive immune response would be expected to arise. The authors acknowledge this, but their explanation is incomplete. The increase in Th17 cells they observe is predicated on mitogenic stimulation, so they are not specific (at least in this study) for MRSA. It would be helpful to see a specific restimulation of these cells with MRSA antigens to determine if there are pre-existing, cross-reactive Th17 cells specific for MRSA and microbiota species which could then link these two as mentioned above.

We acknowledge that this is a major limitation of our study. Although an experiment demonstrating pre-existing, cross-reactive T cells would help support our conclusion, aspects of MRSA biology may make the results of this experiment difficult to interpret. We have consulted with an expert on MRSA virulence factors, co-lead author Dr. Victor Torres, about the feasibility of this experiment. MRSA possess superantigens, such as Staphylococcal enterotoxin B, which bind directly to specific Vβ regions of T-cell receptors (TCR) and major histocompatibility complex (MHC) class II on antigen-presenting cells, resulting in hyperactivation of T lymphocytes and monocytes/macrophages. Additionally, other MRSA virulence factors, such as α-hemolysin and LukED, can induce cell death of lymphocytes. MRSA’s enterotoxins are heat stable, so heat-inactivation of the bacterium may not help in this matter.  For these reasons, restimulation of lymphocytes with MRSA antigens may be difficult to interpret. We humbly suggest that addressing this aspect of the mechanism is outside the scope of this manuscript.

A study by Shao et al. provides an example of a host commensal species inducing Th17 cells with cross-reactivity against MRSA. Upon intestinal colonization, the intestinal fungus Candida albicans influences T cell polarization towards a Th17 phenotype in the spleen and peripheral lymph nodes which provided protection to the host against systemic candidemia. Interestingly, this induction of protective Th17 cells, increased IL-17 and responsiveness in circulating Ly6G+ neutrophils also protected mice from intravenous infection with MRSA, indicating that T cell activation and polarization by intestinal C. albicans leads to non-specific protective responses against extracellular pathogens.

Shao TY, Ang WXG, Jiang TT, Huang FS, Andersen H, Kinder JM, Pham G, Burg AR, Ruff B, Gonzalez T, Khurana Hershey GK, Haslam DB, Way SS. Commensal Candida albicans Positively Calibrates Systemic Th17 Immunological Responses. Cell Host & Microbe. 2019 Mar 13;25(3):404-417.e6. doi: 10.1016/j.chom.2019.02.004. PMID: 30870622; PMCID: PMC6419754.

Reviewer #3 (Public review):

Strengths:

A strength of the work is the rigorous experimental design. Appropriate controls were executed and, in most cases, multiple approaches were conducted to strengthen the authors' conclusions. The conclusions are supported by the data.

The following suggestions are offered to improve an already strong piece of scholarship.

Weaknesses:

The correlation between female sex hormones and the elimination of S. aureus from the gut could be further validated by quantifying sex hormones produced in the four core genotype mice in response to colonization. Additionally, and this may not be feasible, but according to the proposed model administering female sex hormones to male mice should decrease colonization. Finally, knowing whether the quantity of IL-17a CD4+ cells change in the OVX mice has the potential to discern whether abundance/migration of the cells or their activation is promoted by female sex hormones.

In the Discussion, the authors highlight previous work establishing a link between immune cells and sex hormone receptors, but whether the estrogen (and progesterone) receptor is differentially expressed in response to S. aureus colonization could be assessed in the RNAseq dataset. Differential expression of known X and Y chromosome-linked genes were discussed but specific sex hormones or sex hormone receptors, like the estrogen receptor, were not. This potential result could be highlighted.

We appreciate the comment on the scholarship and thank the Reviewer for the insightful suggestions to improve this manuscript. We intend on measuring hormone levels and performing the recommended (or similar) experiments based on availability of reagents and mice during the revision period. We also apologize for not including references that address some of the Reviewer’s questions. Other research groups have compared the levels of hormones between XX and XY males and females in the four core genotypes model and have found similar levels of circulating testosterone in adult XX and XY males. No difference was found in circulating estradiol levels in XX vs XY- females when tested at 4-6 or 7-9 months of age.

Karen M. Palaszynski, Deborah L. Smith, Shana Kamrava, Paul S. Burgoyne, Arthur P. Arnold, Rhonda R. Voskuhl, A Yin-Yang Effect between Sex Chromosome Complement and Sex Hormones on the Immune Response. Endocrinology, Volume 146, Issue 8, 1 August 2005, Pages 3280–3285, https://doi.org/10.1210/en.2005-0284

Sasidhar MV, Itoh N, Gold SM, Lawson GW, Voskuhl RR. The XX sex chromosome complement in mice is associated with increased spontaneous lupus compared with XY. Ann Rheum Dis. 2012 Aug;71(8):1418-22. doi: 10.1136/annrheumdis-2011-201246. Epub 2012 May 12. PMID: 22580585; PMCID: PMC4452281.

Examination of the levels of estrogen, progesterone, and androgen receptors in our cecal-colonic lamina propria RNA-seq dataset is an excellent idea. We will add these analyses to the revised manuscript. We are planning additional experiments to better understand the contributions of hormones or their receptors and anticipate including such data in either a response letter or revised manuscript.

Author response:

Reviewer #1 (Public Review):

Strengths:

Overall there are some very interesting results that make an important contribution to the field. Notably, the results seem to point to differential recruitment of the PL-DMS pathway in goal-tracking vs sign-tracking behaviors.

Thank you.

Weaknesses:

There is a lot of missing information and data that should be reported/presented to allow a complete understanding of the findings and what was done. The writing of the manuscript was mostly quite clear, however, there are some specific leaps in logic that require more elaboration, and the focus at the start and end on cholinergic neurons and Parkinson's disease are, at the moment, confusing and require more justification.

In the revised paper, we provide additional information in support of results and clarify procedures and findings. Furthermore, we expand the discussion of the proposed interpretational framework that suggests that the contrasts between the cortical-striatal processing of movement cues in sign- versus goal trackers are related to previously established, parallel contrasts in the cortical cholinergic detection of attention-demanding cues.

Reviewer #2 (Public review):

Strengths:

The power of the sign- and goal-tracking model to account for neurobiological and behavioral variability is critically important to the field's understanding of the heterogeneity of the brain in health and disease. The approach and methodology are sound in their contribution to this important effort.

The authors establish behavioral differences, measure a neurobiological correlate of relevance, and then manipulate that correlate in a broader circuitry and show a causal role in behavior that is consistent with neurobiological measurements and phenotypic differences.

Sophisticated analyses provide a compelling description of the authors' observations.

Thank you.

Weaknesses:

It is challenging to assess what is considered the "n" in each analysis (trial, session, rat, trace (averaged across a session or single trial)). Representative glutamate traces (n = 5 traces (out of hundreds of recorded traces)) are used to illustrate a central finding, while more conventional trial-averaged population activity traces are not presented or analyzed. The latter would provide much-needed support for the reported findings and conclusions. Digging deeper into the methods, results, and figure legends, provides some answers to the reader, but much can be done to clarify what each data point represents and, in particular, how each rat contributes to a reported finding (ie. single trial-averaged trace per session for multiple sessions, or dozens of single traces across multiple sessions).

Representative traces should in theory be consistent with population averages within phenotype, and if not, discussion of such inconsistencies would enrich the conclusions drawn from the study. In particular, population traces of the phasic cue response in GT may resemble the representative peak examples, while smaller irregular peaks of ST may be missed in a population average (averaged prolonged elevation) and could serve as a rationale for more sophisticated analyses of peak probability presented subsequently.

Figures 5c-f depict individual data from all rats and trials. For all major analyses, the revised manuscript consolidates information about the number of rats per phenotype and sex, and the number of trials contributed by individual rats, in the result section.

As detailed in the section on statistical methods, and as mentioned by the reviewer under Strengths, we used advanced statistical methods to assure that data from individual animals contribute equally to the overall result, and to minimize the possibility that an inordinate number of trials obtained from just one or a couple of rats biased the overall analysis.

As the reviewer correctly pointed out, we have chosen not to show trial- or subject-averaged traces to illustrate glutamate release dynamics across trials. The present analyses focus on peak glutamate concentrations, the number of peaks, and the timing of peaks relative to a task cue or a behavioral event. Within a response bin, such as the 2-s period following turn cues, glutamate peaks – as defined in Methods - occur at variable times relative to cue onset.  Averaging traces over a population of rats or trials would “wash-out” the phenotype- and task event-dependent patterns of glutamate peaks, yielding, for example, a single, nearly 2-s long plateau for cue-locked glutamate recordings from STs (Figure 5b). Thus, subject- or trial-averaged traces would not illustrate the major findings described in this paper and would rather be uninformative. As already mentioned, individual data from all subjects and trials are shown in Figs 5c-f.

Reviewer #3 (Public review):

Strengths:

Overall these studies are interesting and are of general relevance to a number of research questions in neurology and psychiatry. The assessment of the intersection of individual differences in cue-related learning strategies with movement-related questions - in this case, cued turning behavior - is an interesting and understudied question. The link between this work and growing notions of corticostriatal control of action selection makes it timely.

Thank you.

Weaknesses:

The clarity of the manuscript could be improved in several places, including in the graphical visualization of data. It is sometimes difficult to interpret the glutamate results, as presented, in the context of specific behavior, for example.

We appreciate the reviewer’s concerns about the complexity of some of the graphics, particularly the results from the arguably innovative analysis illustrated in Figure 6. Figure 6 illustrates that the likelihood of a cued turn can be predicted based on single and combined glutamate peak characteristics. The revised legend for this figure provides additional information and examples to ease the readers’ access to this figure.

Author response:

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

Reviewer #1 (Public Review):

Summary:

The authors found that the loss of cell-ECM adhesion leads to the formation of giant monocular vacuoles in mammary epithelial cells. This process takes place in a macropinocytosis-like process and involves PI3 kinase. They further identified dynamin and septin as essential machinery for this process. Interestingly, this process is reversible and appears to protect cells from cell death.

Strengths:

The data are clean and convincing to support the conclusions. The analysis is comprehensive, using multiple approaches such as SIM and TEM. The discussion on lactation is plausible and interesting.

We thank the reviewer for the summary of our study and the positive comment.

Weaknesses:

As the first paper describing this phenomenon, it is adequate. However, the elucidation of the molecular mechanisms is not as exciting as it does not describe anything new. It is hoped that novel mechanisms will be elucidated in the future. In particular, the molecules involved in the reversing process could be quite interesting.

We agree with the reviewer’s comments and believe that investigating the molecular mechanisms involved in reversing GUVac formation, as illustrated in Figure 5J, would be valuable for future research.

Additionally, the relationship to conventional endocytic compartments, such as early and late endosomes, is not analyzed.

We thank the reviewer for the valuable comment. To determine whether GUVac displays markers of other endomembrane systems, we analyzed several markers, including EEA1, Rab5, LC3B, LAMP1, and Transferrin receptor (TfR). At early time points (1 h), we observed several large vesicles that had taken up 70kDa Dextran and exhibited EEA1 or Rab5, markers of early endosomes. By 6 hours, some of these large vesicles showed lysotracker positivity, indicating a transition from early to late endosomal fate, similar to the maturation process of conventional macropinocytic vesicles (see new Figure 1-figure supplement 2A). However, once the vesicles fused, grew, and became GUVac, these markers did not consistently correspond with the GUVac membrane but were instead unevenly distributed around it (new Figure 1-figure supplement 2B, C). This made it difficult to determine whether they were localized to separate organelles or part of the GUVac membrane. Interestingly, we found that the Transferrin receptor (TfR), which also marks a general membrane population involved in the endocytic pathway (such as PM invagination), was evenly distributed within the GUVac membrane (new Figure 1-figure supplement 2B, D). Therefore, GUVac appears to possess heterogeneous characteristics of the endocytic membrane, mainly with the TfR marker (likely due to PM invagination) and some partial endomembrane system markers. However, further analysis would be required to confirm this.

Reviewer #2 (Public Review):

Summary:

The manuscript "Formation of a giant unilocular vacuole via macropinocytosis-like process confers anoikis resistance" describes an interesting observation and provides initial steps towards understanding the underlying molecular mechanism.

The manuscript describes that the majority of non-tumorigenic mammary gland epithelial cells (MCF-10A) in suspension initiate entosis. A smaller fraction of cells forms a single giant unilocular vacuole (hereafter referred to as a GUVac). GUVac appeared to be empty and did not contain invading (entotic) cells. The formation of GUVac could be promoted by disrupting actin polymerisation with LatB and CytoD. The formation of GUVacs correlated with resistance to anoikis. GUVac formation was detected in several other epithelial cells from secretory tissues.

The authors then use electron microscopy and super-resolution imaging to describe the biogenesis of GUVac. They find that GUVac formation is initiated by a micropinocytosis-like phenomenon (that is independent of actin polymerisation). This process leads to the formation of large plasma membrane invaginations, that pinch off from the PM to form larger vesicles that fuse with each other into GUVacs.

Inhibition of actin polymerisation in suspended MCF-10a leads to the recruitment of Septin 6 to the PM via its amphipathic helix. Treatment with FCF (a septin polymerisation inhibitor) blocked GUVac biogenesis, as did pharmacological inhibition of dynamin-mediated membrane fission. The fusion of these vesicles in GUVacs required (perhaps not surprisingly) PI3P.

Strengths:

The authors have made an interesting and potentially important observation. They describe the formation of an endo-lysosomal organelle (a giant unilocular vacuole - GUVac) in suspended epithelial cells and correlate the formation of GUVacs with resistance to aniokis.

We thank the reviewer for the summary of our study and the positive comment.

Weaknesses:

My major concern is the experimental strategy that is used throughout the paper to induce and study the formation GUVac. Almost every experiment is conducted in suspended cells that were treated with actin depolymerising drugs (e.g. LatB) and thus almost all key conclusions are based on the results of these experiments. I only have a few suggestions that would improve these experiments or change their outcome and interpretation. Yet, I believe it is essential to identify the endogenous pathway leading to the actin depolymerisation that drives the formation of GUVacs in detached epithelial cells (or alternatively to figure out how it is suppressed in most detached cells). A first step in that direction would be to investigate the polymerization status of actin in MCF-10a cells that 'spontaneously' form GUVacs and to test if these cells also become resistant to anoikis.

We thank the reviewer for the valuable comments and fully acknowledge the limitations of our approach. Many detached cells likely tend to contact each other for cell aggregations to suppress GUVac formation. However, it is unclear whether cells that spontaneously form GUVac in suspension have a weakened F-actin structure, which would be valuable to investigate in future studies.

Also, it would be great (and I believe reasonably easy) to better characterise molecular markers of GUVacs (LAMP's, Rab's, Cathepsins, etc....) to discriminate them from other endosomal organelles

In response to a similar comment from Reviewer 1, we analyzed markers of other endocytic compartments, including EEA1, Rab5, Transferrin receptor (TfR), LC3B, and LAMP1. At early time points (1 h), we observed several large vesicles that had taken up 70kDa Dextran and exhibited EEA1 or Rab5, markers of early endosomes. By 6 hours, some of these large vesicles showed lysotracker positivity, indicating a transition from early to late endosomal fate, similar to the maturation process of conventional macropinocytic vesicles (see new Figure 1-figure supplement 2A). However, once the vesicles fused, grew, and became GUVac, these markers did not consistently correspond with the GUVac membrane but were instead unevenly distributed around it (new Figure 1-figure supplement 2B, C). This made it difficult to determine whether they were localized to separate organelles or part of the GUVac membrane. Interestingly, we found that the Transferrin receptor (TfR), which also marks a general membrane population involved in the endocytic pathway (such as PM invagination), was evenly distributed within the GUVac membrane (new Figure 1-figure supplement 2B, D). Therefore, GUVac appears to possess heterogeneous characteristics of the endocytic membrane, mainly with the TfR marker (likely due to PM invagination) and some partial endomembrane system markers. However, further analysis would be required to confirm this.

Reviewer #3 (Public Review):

Summary:

Loss of cell attachment to extracellular matrix (ECM) triggers aniokis (a type of programmed cell death), and resistance to aniokis plays a role in cancer development. However, mechanisms underlying anoikis resistance, and the precise role of F-actin, are not fully known.

Here the authors describe the formation of a new organelle, giant unilocular vacuole (GUVac), in cells whose F-actin is disrupted during loss of matrix attachment. GUVac formation (diameter >500 nm) resulted from a previously unrecognised macropinocytosis-like process, characterized by inwardly curved micron-sized plasma membrane invaginations, dependent on F-actin depolymerization, septin recruitment, and PI(3)P. Finally, the authors show GUVac formation after loss of matrix attachment promotes resistance to anoikis.

From these results, the authors conclude that GUVac formation promotes cell survival in environments where F-actin is disrupted and conditions of cell stress.

Strengths:

The manuscript is clear and well-written, figures are all presented at a very high level.

A variety of cutting-edge cell biology techniques (eg time-lapse imaging, EM, super-resolution microscopy) are used to study the role of the cytoskeleton in GUVac formation. It is discovered that: (i) a macropinocytosis-like process dependent on F-actin depolymerisation, SEPT6 recruitment, and PI(3)P contributes to GUVac formation, and (ii) GUVac formation is associated with resistance to cell death.

We thank the reviewer for the concise summary of our study and positive comments.

Weaknesses:

The manuscript is highly reliant on the use of drugs, or combinations of drugs, for long periods of time (6hr, 18hr..). Wherever possible the authors should test conclusions drawn from experiments involving drugs also using other canonical cell biology approaches (eg siRNA, Crispr). Although suggestive as a first approach, it is not reliable to draw conclusions from experiments where only drug combinations are being advanced (eg LatB + FCF).

We thank the reviewer for the comment and suggestion. As suggested, we employed siRNAs targeting Septin2 and Septin9 in cells treated with LatB as an alternative to the drug combination approach. This genetic approach, combined with chemical treatment, led to a consistent reduction in GUVac formation, similar to the results observed with LatB+FCF treatment (see new Figure 3D-WB and graph).

F-actin is well known to play a wide variety of roles in cell death and other canonical cell death pathways (PMID: 26292640). The authors show using pharmacological inhibition that F-actin is key for GUVac formation. However, especially when testing for physiological relevance, how can these other roles for F-actin be ruled out?

In Figure 5, we investigate the physiological relevance of GUVac, highlighting its role in suppressing apoptosis and enhancing anoikis resistance. As the reviewer correctly noted, F-actin inhibition is known to reduce apoptotic signaling (PMID: 16072039). However, we observed that anoikis resistance is lost when GUVac is suppressed through knockout of either PI3KC2alpha or VPS34 in cells with F-actin disrupted by LatB (Figure 5I). This suggests that GUVac plays a role in suppressing apoptosis independently of F-actin depolymerization-induced apoptosis resistance.

To test the role of septins in GUVac formation only recruitment studies and no direct functional work is performed. A drug forchlofeneuron (FCF) is used, but this is well known to have off-target effects (PMID: 27473917).

We thank the reviewer for the valuable comments. To eliminate potential off-target effects of FCF, as described above, we employed siRNA targeting Septin 2 and Septin 9 and observed similar results (see new Figure 3D).

Cells that possess GUVac are resistant to aniokis, but how are these cells resistant? This report is focused on mechanisms underlying GUVac formation and does not directly test for mechanisms underlying aniokis resistance.

We fully agree with the reviewer’s comments and recognize the importance of uncovering the mechanism behind GUVac-mediated anoikis resistance for future research. It will likely be essential to investigate how prosurvival signaling pathways are activated, like the PI3K-AKT signaling (as shown in Figure 5-Supplement 1) or the YAP/TAZ pathway.

Reviewer #1 (Recommendations For The Authors):

Figure 4 Supplemental 1. What are the faint bands in clones 23, 26, and 29? Are they cross-reacting bands? Or Vps34?

We apologize if the data in our original manuscript were misleading. To clarify the specificity of the VPS34 antibody in the Western blot analysis of VPS34 KO clones, we compared these samples with those from siRNA-mediated VPS34-depleted cells (see new Figure 4-Supplement 1E, which replaces the original Figure). Consistent with the known size of VPS34 at approximately 100 kDa, we observed a clear disappearance of the VPS34 band at around 100 kDa in the sgVPS34 clones, which was comparable to the size observed in siRNA-treated cells.

Reviewer #2 (Recommendations For The Authors):

Figure 2B: Only 4 cells were counted. Please comment.

At the outset of this study, we faced technical difficulties in preparing TEM samples, which limited the number of samples included in Figure 2B. However, subsequent experiments that combined TEM with super-resolution microscopy, as shown in Figure 4D-F, produced similar data on plasma membrane invagination, as depicted in Figure 2B, which is the initial step in the formation of GUVac.

Figure 2C: do cells shrink after treatment with EIPA or LatB? Please comment.

We apologize if the data presented in our original manuscript were misleading. Control cells treated with DMSO display multiple cell-in-cell structures (known as 'entosis'), which typically results in a larger overall cell size compared to EIPA or LatB-treated non-entotic single cells. This might have created the impression that cells shrink relative to the control under EIPA or LatB treatment. We hope this explanation has answered the reviewer’s question.

Figure 3A: The changes in the localization of mCherry-Spetin6 appear to be very dramatic. Are these results properly reflected by the quantification in Figure 3B? Is indeed the entire mCherry-Spetin6 pool recruited to the plasma membrane? Wouldn't that imply that all other septin6-regulated processes are blocked?

Again, we apologize if the data presented in our original manuscript caused any confusion. In Figure 3B, we quantified only the number of filament-like Septin6 structures predominantly observed in LatB-treated cells, rather than measuring changes in the relative fluorescence intensity of Septin6 between the plasma membrane and the cytosol. Although we could not estimate the proportion of total Septin6 recruited to the plasma membrane from the cytosol based solely on Figure 3A-B, conducting plasma membrane fractionation experiments with endogenous Septin6, followed by Western blot analysis, would be valuable for addressing this issue in future studies.

Figure 3D: Please also provide data for the 6h time-point (as in all other experiments).

We apologize for omitting the 6-hour time point, which may have caused confusion. The new Figure 3E (previously Figure 3D) shows that recruitment of wild-type Septin6, but not the amphipathic helix (AH) deletion mutant, occurs at a 6-hour time point.

Figure 3E: Molecular weight for western blot is missing.

We thank the reviewer for pointing this out and have revised the figure accordingly.

Line 188 - Title of subchapter could include dynamin.

We appreciate the reviewer’s helpful suggestion and have updated the revised manuscript to reflect this. The phrase "Recruitment of Septin to the Fluctuating Plasma Membrane Drives Macropinocytosis-like Process" has been revised to "Septin and Dynamin Drive Macropinocytosis-like Process".

Line 450 - please describe how the genotyping of MCF10a gene-engineered cells was performed.

We confirmed the knockout of MCF10A cell lines by Western blot analysis using specific antibodies against VPS34 and PI3KC2α, rather than through genotyping.

Reviewer #3 (Recommendations For The Authors):

(1) The manuscript is highly reliant on the use of drugs, or combinations of drugs, for long periods of time (6hr, 18hr..). Wherever possible authors should test conclusions drawn from experiments involving drugs also using other canonical cell biology approaches (eg siRNA, Crispr). Although suggestive as a first approach, it is not reliable to draw conclusions from experiments where only drug combinations are being advanced (eg LatB + FCF).

We thank the reviewer for the comment. As suggested, we employed siRNAs targeting Septin2 and Septin9 in cells treated with LatB as an alternative to the drug combination approach. This genetic approach, combined with chemical treatment, led to a consistent reduction in GUVac formation, similar to the results observed with LatB+FCF treatment (see new Figure 3D-WB and graph).

(2) SEPT6 is recruited at an inwardly curved plasma membrane. Can the authors better describe what type of structure is being recruited/quantified (filaments, collar-like structures, etc)?

We apologize if the data presented was unclear. As outlined in the Methods section in the original manuscript, we detected puncta-like Septin6 structures using the Find Maxima tool in ImageJ, which could include both filamentous and collar-like structures that were less apparent in the DMSO control. We have added additional explanations in the revised manuscript in the legend of Figure 3B to clarify the recruitment of Septin6.

Previous work has shown that octameric septin complexes are linking actin to the plasma membrane (PMID: 36562751). Tests for the recruitment/function of other key septins such as SEPT7 and SEPT9 to support conclusions.

As previously mentioned, to further explore the role of other septin family members in GUVac formation, we tested the roles of Septin9 and Septin2 using siRNAs and found that they are essential for this process (see new Figure 3D). Unfortunately, we were unable to assess the localization of Septin2 and Septin9 due to the lack of suitable antibodies for detecting endogenous proteins by immunofluorescence.

(3) SEPT6 recruitment is impaired when cells are treated with FCF. FCF is well known to have off-target effects (PMID: 25217460, PMID: 27473917). siRNA for SEPT2, SEPT7 and/or SEPT9 can be used to test phenotypes obtained using FCF.

We thank the reviewer for the comment. As also mentioned above, to eliminate potential off-target effects of FCF, we used siRNA to target Septin2 and Septin9, and obtained similar results (see new Figure 3D).

(4) SEPT6 is recruited to the fluctuating cell membrane via the amphipathic helix (AH) domain (Figure 3D). Are these only representative images? It is not clear what readers should be looking at - can the authors provide arrows to highlight what is the difference +/- AH? Can something be quantified?

We thank the reviewer for the suggestion and have added arrows from the inset of the merge pannel Figure 3E, along with line profile analysis, to emphasize the failure of the AH deletion mutant of Septin6 to recruit to the plasma membrane.

Throughout Figure 3, why use LatB treatment at different times?

We apologize if this was not clearly addressed in our original manuscript. Throughout the study, we primarily used an 18-hour LatB treatment to evaluate GUVac formation, as this longer period allows for gradual vesicle fusion. In contrast, we utilized 6-hour treatments to demonstrate that Septin6 recruitment and subsequent plasma membrane invagination occur at earlier time points, as evidenced by the data in Figure 2G (super-resolution live imaging) and Figure 4D (electron microscopy analysis). This clarification has been incorporated into the revised manuscript.

(5) F-actin is well known to play a wide variety of roles in cell death and other canonical cell death pathways (PMID: 26292640). The authors show using pharmacological inhibition that F-actin is key for GUVac formation. However, especially when testing for physiological relevance, how can these other roles for F-actin be ruled out?

In Figure 5, we investigate the physiological relevance of GUVac, highlighting its role in suppressing apoptosis and enhancing anoikis resistance. As the reviewer correctly noted, F-actin inhibition is known to reduce apoptotic signaling (PMID: 16072039). However, when GUVac is suppressed through knockout of either PI3KC2alpha or VPS34 in cells with F-actin disrupted by LatB, anoikis resistance is lost (see Figure 5H, I). This suggests that GUVac plays a role in suppressing apoptosis independently of F-actin depolymerization-induced apoptosis resistance.

(6) Cells that possess GUVac are resistant to aniokis, but how are these cells resistant? This report is focused on mechanisms underlying GUVac formation and does not directly test for mechanisms underlying aniokis resistance.

We fully agree with the reviewer’s comments and recognize the importance of uncovering the mechanism behind GUVac-mediated anoikis resistance for future research. It will likely be essential to investigate how prosurvival signaling pathways are activated, like the PI3K-AKT signaling (as shown in Figure 5-Supplement 1) or the YAP/TAZ pathway.

(7) In the Discussion, there is a lot of text on involution and speculative relevance of GUVac formation. I would focus the Discussion more on the clear results discovered here.

We thank the reviewer’s feedback and have revised the discussion to reduce its length concerning involution.

(8) Figure 5. GUVac formation promotes cell survival in altered actin and matrix environments. In Figure 5J, it will not be clear to readers outside the field what is being shown here.

We appreciate the reviewer’s suggestion and have added two distinct dotted lines around the vacuole and cell area in the revised figure to emphasize the gradual reduction in its size over time.

Author response:

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

Reviewer #1 (Public Review):

Summary:

SUMO proteins are processed and then conjugated to other proteins via a C-terminal di-glycine motif. In contrast, the N-terminus of some SUMO proteins (SUMO2/3) contains lysine residues that are important for the formation of SUMO chains. Using NMR studies, the N-terminus of SUMO was previously reported to be flexible (Bayer et al., 1998). The authors are investigating the role of the flexible (referred to as intrinsically disordered) N-terminus of several SUMO proteins. They report their findings and modeling data that this intrinsically disordered N-terminus of SUMO1 (and the C. elegans Smo1) regulates the interaction of SUMO with SUMO interacting motifs (SIMs).

Strengths:

Among the strongest experimental data suggesting that the N-terminus plays an inhibitory function are their observations that

(1) SUMO1∆N19 binds more efficiently to SIM-containing Usp25, Tdp2, and RanBp2,<br /> (2) SUMO1∆N19 shows improved sumoylation of Usp25,<br /> (3) changing negatively-charged residues, ED11,12KK in the SUMO1 N-terminus increased the interaction and sumoylation with/of USP25.

The paper is very well-organized, clearly written, and the experimental data are of high quality. There is good evidence that the N-terminus of SUMO1 plays a role in regulating its binding and conjugation to SIM-containing proteins. Therefore, the authors are presenting a new twist in the ever-evolving saga of SUMO, SIMs, and sumoylation.

Weaknesses:

Much has been learned about SUMO through structure-function analyses and this study is another excellent example. I would like to suggest that the authors take some extra time to place their findings into the context of previous SUMO structure-function analyses. Furthermore, it would be fitting to place their finding of a potential role of N-terminally truncated Smo1 into the context of the many prior findings that have been made with regard to the C. elegans SUMO field. Finally, regarding their data modeling/simulation, there are questions regarding the data comparisons and whether manipulations of the N-terminus also have an effect on the 70/80 region of the core.

We thank the reviewer for insightful and constructive comments to improve our manuscript. We have now placed our findings in the context of previous structure-function analyses at several occasions, details of which can be found in our replies to the detailed comments.

We are also placing the C. elegans data into context of previously published findings on the various functions of SMO-1 in controlling development and maintaining genomic stability (lines 510ff). Finally, we addressed all questions and suggestions regarding comparison of MD simulation and NMR data, and addressed the question whether mutations in the N-terminus affected the 70/80 region. We have now clarified in the manuscript that the sum of MD and NMR data does not allow a clear-cut conclusion on the 70/80 interactions. 

Reviewer #2 (Public Review):

Summary:

This very interesting study originated from a serendipitous observation that the deletion of the disordered N-terminal tail of human SUMO1 enhances its binding to its interaction partners. This suggested that the N terminus of SUMO1 might be an intrinsic competitive inhibitor of SUMO-interacting motif (SIM) binding to SUMO1. Subsequent experiments support this mechanism, showing that in humans it is specific to SUMO1 and does not extend to SUMO2 or SUMO3 (except, perhaps, when the N terminus of SUMO2 becomes phosphorylated, as the authors intriguingly suggest - and partially demonstrate). The auto-inhibition of SUMO1 via its N-terminal tail apparently explains the lower binding of SUMO1 compared to SUMO2 to some SIMs and lower SIM-dependent SUMOylation of some substrates with SUMO1 compared to SUMO2, thus adding an important element to the puzzle of SUMO paralogue preference. In line with this explanation, N-terminally truncated SUMO1 was equally efficient to SUMO2 in the studied cases. The inhibitory role of SUMO1's N terminus appears conserved in other species including S. cerevisiae and C. elegans, both of which contain only one SUMO. The study also elucidates the molecular mechanism by which the disordered N-terminal region of SUMO1 can exert this auto-inhibitory effect. This appears to depend on the transient, very highly dynamic physical interaction between the N terminus and the surroundings of the SIM-binding groove based mostly on electrostatic interactions between acidic residues in the N terminus and basic residues around the groove.

Strengths:

A key strength of this study is the interplay of different techniques, including biochemical experiments, NMR, molecular dynamics simulations, and, at the end, in vivo experiments. The experiments performed with these different techniques inform each other in a productive way and strengthen each others' conclusions. A further strength is the detailed and clear text, which patiently introduces, describes, and discusses the study. Finally, in terms of the message, the study has a clear, mechanistic message of fundamental importance for various aspects of the SUMO field, and also more generally for protein biochemists interested in the functional importance of intrinsically disordered regions.

Weaknesses:

Some of the authors' conclusions are similar to those from a recent study by Lussier-Price et al. (NAR, 2022), the two studies likely representing independent inquiries into a similar topic. I don't see it as a weakness by itself (on the contrary), but it seems like a lost opportunity not to discuss at more length the congruence between these two studies in the discussion (Lussier-Price is only very briefly cited). Another point that can be raised concerns the wording of conclusions from molecular dynamics. The use of molecular dynamics simulations in this study has been rigorous and fruitful - indeed, it can be a model for such studies. Nonetheless, parameters derived from molecular dynamics simulations, including kon and koff values, could be more clearly described as coming from simulations and not experiments. Lastly, some of the conclusions - such as enhanced binding to SIM-containing proteins upon N-terminal deletion - could be additionally addressed with a biophysical technique (e.g. ITC) that is more quantitative than gel-based pull-down assays - but I don't think it is a must.

Thank you very much for pointing towards the study of Lussier-Price. We now point out congruent findings in more detail in the discussion.

We also thank the reviewer for the advice to present and discuss the MD findings more clearly, and more explicitly specify which parameters were obtained from MD. We have made changes throughout the Results and Discussion sections.

We agree that it would be a nice addition to use ITC measurements as a more quantitative method to assess differences in binding affinities upon deletion of the SUMO N-terminus. We had tried to measure affinities between SUMO and SIM-containing binding partners by ITC but in our hand, this failed. In the study of Lussier-Price et al., the authors were able to measure differences in SIM binding upon deleting the N-terminus but only when they used phosphorylated SIM peptides. Follow-up studies, e.g., on the effect of SUMO’s N-terminal modifications should certainly include more quantitative measurement such as ITCs, however these studies will have to be picked up by others. The main PI Frauke Melchior and most contributing authors moved on to new challenges.

Reviewing Editor (Recommendations For The Authors):

Both reviewers agreed that your manuscript presents novel results and the key findings including the self-inhibitory role of the N-terminal tail of SUMO proteins in their interaction with SIM are overall well supported by the data. The reviewers also provided constructive suggestions. They pointed out that some simulation results are not clear, which could be strengthened by control analysis and by toning down the related descriptions. In addition, Reviewer 2 suggested that the conclusions from the current biochemical and simulation studies could be further reinforced by more quantitative binding measurements. We hope that these points can be addressed in the revision.

We thank both reviewers for their insightful and constructive comments and the appreciative tone. In our replies above and below we address most of the raised concerns.

We strongly recommend the change of the current title. eLife advises that the authors avoid unfamiliar abbreviations or acronyms, or spell out in full or provide a brief explanation for any acronyms in the title.

We changed the title to “The intrinsically disordered N-terminus of SUMO1 is an intramolecular inhibitor of SUMO1 interactions” to avoid acronyms in the title.

Reviewer #1 (Recommendations For The Authors):

Major:

Lines 190-262: The authors use NMR experiments and all-atom molecular dynamics (MD) simulations. They state that this approach reveals a highly dynamic interaction of the SUMO1 N-terminus with the core and that the SIM binding groove and the 70/80 region are temporarily occupied by the SUMO1 N-terminus (Fig. 3C). After comparing SUMO1, Smt3, SUMO2, and Smo1 by this approach they state that the most striking differences exist for the interaction with the SIM-binding groove, while interactions with the 70/80 region are rather comparable.

The authors then compare the average binding time data of Figure 3C, D, E, F in Figure 3G.

It is not clear which data points are included in the bar graphs of Figure 3G and how the individual data points (there are maybe 8 shown in each bar) correspond to the data shown in 3C, D, E, and F or if they are iterations (n?) of the modeled data. This should be clarified. Also, for comparison, the authors should also graph the average data of the 70/80 region.

We clarified the data shown in Figure 3G as well as 3C-F, and how It relates to each other. Indeed, Figure 3G shows 8 data points for 8 trajectories, and their average. Figure 3C-F are based on the same 8 trajectories, in this case broken down per residue of the protein. The average data of the 70/80 region does not show any significant differences across the proteins, as already pretty well visible from panels 3C-F.

Line 322: More concerning, in Figure 5, the authors model how a ED11,12KK mutations disrupt the interaction between the N-terminus and the SIM-binding groove and state that this mutation leaves interactions with the 70/80 region largely untouched. Again, it is not clear which data points are included in the bar graph 5D and 5G and how many iterations. Furthermore, data of 5B, C (SUMO1) and 5 E, F (smo1) do show clear differences between the WT and mutants affecting both the SIM binding groove and the 70/80 region. The double mutation clearly seems to affect the 70/80 region when comparing 5B, C (SUMO1) and 5 E, F (smo1), but this result is not mentioned. Indeed, the authors state that the double mutants leave the interactions with the 70/80 region largely untouched, but this is not borne out by the data presented.

We improved the clarity of the legend of Figure 5 as suggested. We also thank the reviewer for the comment on the changes in the 70/80 region, to which we point the reader explicitly now in the corresponding Results section. We, however, refrain from drawing conclusions from the MD in this case, as this change is not supported by the NMR measurements (Fig 5a). Charge-charge interactions in the charge-rich double mutants might be overstabilized in the MD simulations, a problem known for the canonical force fields used here, albeit tailoring it for IDPs. We now cite a corresponding reference. Another potential explanation for that the CMPs do not take this change up upon mutation could be a pronounced fuzziness in this region, which however, in turn, is not apparent from the simulations. We would therefore not overinterpret these differences in the 70/80 region. Our key conclusion is the loss of interactions with the SIM-binding groove – and thus of cis-inhibition – by mutations, which is supported by both, MD and NMR.  

341: In their N-termini substitution experiments, the authors show that the SUMO1 core that carries the SUMO2 N-terminus (S2N-S1C) binds USP25 more efficiently than wt SUMO1. However, the SUMO1 core that carries the SUMO2 N-terminus is also reduced in its interaction with Usp25. This is concerning as the SUMO2 N-terminus was not predicted to interfere with SIM binding.

We were excited to see that the inhibitory potential could be partially transplanted by swapping the N-termini of SUMO1 and SUMO2 demonstrating that some important determinants are contained within the N-terminal tail of SUMO proteins. However, the observed effects were partial indicating that also other determinants contribute and that we do not yet understand all aspects. Obviously, the SUMO1 and SUMO2 cores are similar (also in the area comprising the SIM binding groove) but not identical, and as the inhibition arises from dynamic interactions of the N-terminus with the SIM binding area, differences in the SUMO cores and in residues flanking SUMO’s N-terminus are likely to influence the inhibitory potential as well.

Blue bars in 3G, 5D, and 6A look surprisingly similar down to the individual data points - does that mean that the same SUMO1 WT data was recycled for these different experiments? This is concerning to me.

The data displayed in the figures listed above are derived from in silico simulations and indeed display the same data set for the case of SUMO1 WT repeatedly, as we also state in the figure legends (we had done so for 5D “(identical to Fig. 3C)”, and now added the same comment to 6A, thanks for pointing this out). We show the SUMO1 WT data again to facilitate comparing the different SUMO variants in MD simulations.

Line 352 and 496: The authors used phosphomimetic mutants to assess the effect of SUMO2 N-term phosphorylation on interaction with Usp25. The data suggest a mild phenotype (6G) which is borne out by the quantization in 6H. In contrast, the effect of an array of modifications for SUMO1 (Figures 6A - C) was solely analyzed by MD simulation. If possible, this data should be confirmed, at least by using a phosphomimetic at the Ser9 position of SUMO1. Alternatively, a caveat explaining the need to confirm these predictions by actual experiments should be added to the text.

Already now we state in “Limitations of the study” that “While our MD simulations and in vitro studies with selected mutants point in this direction, we have not been able to generate quantitatively acetylated and/or phosphorylated SUMO variants to test this hypothesis.”

We agree that the hypothesis needs experimental validation. Phosphomimetic amino acids can be a useful tool in some cases but fail to mimic a phosphor group in other cases. In the past we had tested whether replacing Ser9 by a potentially phospho-mimicking amino acid (Glu) would further diminish binding of SIM-containing proteins compared to already strongly reduced binding to wt SUMO1 but the effect was too mild to yield a significant difference, at least in our assay. Whether this is due to a lack of Glu in mimicking phosphorylation of Ser9, due to limited sensitivity of our pulldown assay combined with the challenge to detect inhibition compared to an already inhibited state, or a failure in our hypothesis we were not able to clarify so far. We therefore now also added a sentence to the paragraph introducing phosphoSer9 MD simulations (now line 367) stating that this hypothesis needs to be tested experimentally.

Minor:

Line 110: the authors should include references for their summary statement that "A defining feature of SUMO proteins is the intrinsically disordered N-terminus, whose function is only partly understood." Also cite in line 119.

Thank you, we now included some references.

Line 75: Please indicate early on that the N-terminus of some SUMO proteins contains lysines for the formation of SUMO chains. Please list them.

We now list, which of the SUMO proteins used in this study contain lysine residues in their N-termini.

Line 113: Please cite studies that elucidated the sumoylation of lysines in the N-terminus of SUMO2/3 proteins.

Thank you, we now included some references.

Line 153: The authors should include additional references on Smt3 structure function analyses to provide better context. One important detail, for example, is the important finding that Yeast SUMO (Smt3) deletion can be complemented by hsSUMO1 but not hsSUMO2 and hsSUMO3. Additionally, in yeast the entire Smt3 N-terminus can be deleted without detectable effects on growth, underscoring the enigmatic role of the N-terminus (Newman et al., 2017). Caveat also applies to line 266.

Thank you, we now included some additional information and references around line 153 and below.

164: The hypothesis that the SUMO1 N-terminus interferes with SIM binding groove ignores the previous observation that deletion of the SUMO2 N-terminus does not have an effect on binding (in vitro). While this is addressed later, the authors should clarify this e.g. by stating "a unique feature of the SUMO1 N-terminus".
>

We now explicitly mention that this feature appears to be unique to SUMO1.

374 and 499: The authors should discuss the caveat that the deletion of the N-terminus of Smt3 does not have a phenotype in yeast in vivo (Newman et al., 2017).

We now discuss that Smt3’s N-terminus can be deleted without detectable phenotype, both in the results as well as in “Limitations of the study”.

Line 367: I feel this is overstated and I do not see any evidence that post translation modifications of the SUMO core plays a role. Therefore, I suggest: Our data and modeling are consistent with an interpretation that the N-termini of human and C. elegans SUMO1 proteins are inhibitory and that other SUMO N-termini may acquire such a function upon posttranslational modification of the N-terminus.

We agree that this is pure speculation and therefore restrict our hypothesis to modifications of the N-terminus.

Line 374 ff: Since Smo-∆N12 increases sumoylation (Fig. 2I), it is likely that the in vivo defect is due to over-sumoylation in C. elegans. The authors should discuss this possibility and quote appropriate literature e.g.: Rytinki et al., Overexpression of SUMO perturbs the growth and development of Caenorhabditis elegans. Cell Mol Life Sci. 2011 Oct;68(19):3219-32. PMID: 21253676.

In our study, we employ in vitro SUMOylation as a means to assess the SIM binding capability in an in-solution assay. For this, we use USP25 as a specific substrate known to depend on a SIM for its SUMOylation. We cannot exclude that some specific substrates depending on this same mechanism for their modification may be upregulated in modification also in the Smo-1∆N12 worms. In vivo however, the majority of SUMO substrates is not subject to SIM-dependent SUMOylation. We now added a control experiment showing that we neither observe significantly increased SUMO levels nor upregulated steady state levels of SUMOylation in these worms (Supplemental figure 8).

The phenotypes shown in the paper by Rytinki et al. do not resemble the smo-1∆N12 mutants. Rather, we observed a specific defect in the meiotic germ cells at the pachytene stage causing increased apoptosis Moreover, we show by western blot analysis that there is no global over-sumoylation occurring in smo-1∆N12 mutants (Fig. s8). Together, our data point to a germline-specific function of the SMO-1 N-terminus in maintaining genome stability (lines 510ff).

Reviewer #2 (Recommendations For The Authors):

Page2 - "Small Ubiquitin-related modifiers of the SUMO family regulate thousands of proteins in eukaryotic cells" - The authors could consider a more precise statement, e.g. that SUMO modifiers have been detected on thousands of proteins and their regulatory effect on many proteins have been demonstrated.

To be a bit more precise, the sentence now reads: “Ubiquitin-related proteins of the SUMO family are reversibly attached to thousands of proteins”. The summary has a word limit, hence we did not expand further at this place.

Page 4 - "Both events require SUMO-binding motifs (reviewed, e.g. in 7 ." - The end bracket is missing. Also, isn't it too strong a statement that paralogue specificity always requires a SIM? I don't know all the literature sufficiently well, but the authors could double-check if it is correct to say that paralogue-specific SUMOylation always depends on a SIM.

Thank you, we added the missing bracket. We agree that it would not be correct to say that paralogue-specificity always depends on a SIM. One alternative example is Dpp9, which shows a clear preference for SUMO1 without owning a SIM. Instead, Dpp9 harbors an alternative SUMO-binding motif, the E67-interacting loop, with a strong paralogue-preference (Pilla et al., 2012). We never intended to imply that a SIM is required for paralogue preference and we also rather generically wrote “SUMO binding motif” instead of “SIM”. However, in the subsequent paragraph about SUMO binding motifs we only go into details of SIMs as one of three classes of SUMO binding motifs not even mentioning the alternative classes. To make this more obvious, we now list the two other known classes of SUMO binding motifs hoping that it will shed the correct light onto our previous statement about paralogue preference.

Page 4 - In the nice discussion of different types of SIMs, the authors could consider mentioning also the special case of TDP2, which is used later by them as a model binding protein. This could provide an occasion to explain what the unusual "split SIM", mentioned on page 6, but not discussed, is, and what its relation to a normal SIM is. Also, it can perhaps be mentioned that TDP2 contacts SUMO2 not only through the two hydrophobic elements contiguous in space that mimic a SIM but also through a slightly larger interface around these regions on the surface of a folded domain.

Thank you for pointing this out. In the introduction, we extended our section on SUMO binding and now also included TDP2’s “split SIM”.

Page 11-12 - In the section "Interaction between SUMO's disordered N-termini and the SIM binding groove is highly dynamic" (and corresponding figures), it should be stated that the discussed kinetic parameters are derived from molecular dynamics simulations and not experimental measurements. It was not very clear to me. This also applies to this sentence on page 17: "First, we observed a very fast (ns) rate of the binding/unbinding process", which in its current form suggests direct observation rather than simulation.

We thank the reviewer for pointing this out, and in fact, Rev #1 made the same comment. We specified now clearly that the rates were calculated from MD simulations, in the Results and Discussion sections (on page 11-12 and 18 (previously 17)).

Page 16 - The authors could briefly mention that this relatively long disordered N-terminal tail is a specific feature of SUMO proteins that distinguishes them from ubiquitin. I guess it is obvious to people from the SUMO field, but I don't think it is explicitly stated anywhere in the text and it could be interesting for readers who are less familiar with SUMO/ubiquitin differences.

Thank you, we added a short half-sentence pointing out this difference.

Page 17 - "The N-terminal region remains fully disordered in the bound state and is thus a classic example of intrinsic disorder irrespective of the binding state." - it could be added to this sentence that this is suggested by molecular dynamics simulations and not directly observed.

We added the information that this finding is based on the MD simulations.

Page 18 - "(e.g., 41,53 or flanking the SIM binding groove24,42" - the end bracket is missing.

Thanks, we added it.

Page 19 - "Our analysis in C. elegans (Fig. 7) suggests that this N-terminal function is particularly important in DNA damage response, a pathway that is strongly dependent on the SUMO system." - this brief description of the in vivo data seems to overgeneralise them a little bit. Perhaps one can describe what was observed with slightly more nuance.

See changes on p.19, lines 510ff.

Author response

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

We thank the editors and reviewers for their thoughtful comments on our manuscript. We greatly appreciated the suggestions and recommendations that helped us to improve the study. With adaptations, and inclusion of novel data and analyses, we have addressed all points raised, and hope that by these improvements the study further meets the standards for eLife. 

Reviewer #1 (Recommendations For The Authors):

Minor text edits should be made.

(1.1) As a recent study from the Wong lab also showed sebaceous gland regeneration following complete ablation (Veniaminova et al., 2023), this finding should be mentioned in the text, and the abstract ("Most strikingly...") should be toned down.

We thank the reviewer for the positive feedback, and for highlighting this part of the study from the Wong lab. Although we cited this study study in a different context, we had not discussed the sebaceous gland regeneration finding. We have now added this to the discussion section of the manuscript.

(1.2) Introduction: In lines 31-33 discussing the connection of sebaceous glands with skin disorders, the 5 references cited seem to replicate the citations from a similar sentence in Veniaminova et al., 2019. The authors should vary their citations, as there are likely other publications that can be cited here.

Additional references have been added.

Reviewer #2 (Recommendations For The Authors):

The manuscript is well written and the data are well presented in the figures.

We thank the reviewer for the positive feedback.

(2.1) Here are some points that could be taken into consideration to improve the manuscript:

- Row 75 "the primary" regulator could be changed to "a crucial".

We appreciate this suggestion and have made the text edit.

- Row 86 could be added: ...is the dominant ligand of the Notch signalling.

We have made the text edit as suggested.

(2.2) Row 107-109 from the quantification of Figure 1G and Figure 2 it seems that only the aJ2 treatment has an SG phenotype. Why aJ1 doesn't have any effect? (same is true in other figures). If the data on aJ1 are maintained in the manuscript, this should be argued in the discussion section.

The reviewer is correct in noting that the aJ1 treatment does not cause the phenotype, and this is indeed one of the key findings of the study. This is maintained throughout the manuscript. We have also cited references showing that embryonic and adult deletions of Jag1 do not cause any sebaceous gland defects. All these data argue that Jag1 is not the relevant Notch signaling ligand in sebocyte differentiation. We have further clarified this in the manuscript.

(2.3) Related to Figure 3G. As the Lrig1 stem cells can go towards both the sebocyte differentiation, or the sebaceous duct differentiation, it would be interesting to evaluate if the differentiation impairment caused by the antibody treatment affects in a similar manner (or not) the sebaceous duct differentiation. This could be tested through immunofluorescence, selecting markers of sebaceous duct.

We thank the reviewer for this thoughtful question. We are unable to find any unique markers of the sebaceous ducts (that are not expressed in other parts of the sebaceous gland, especially sebocytes) in the literature, thus, any analysis of markers would be confounded by its change of expression due to the loss of sebocytes.

However, we have evaluated the histology using bursting sebocytes releasing sebum as a proxy of a functional sebaceous duct. We have not found any significant differences between treatments using this metric (Fig. S1).

(2.4) As the word "therapeutic" is often underlined in the manuscript, maybe a few sentences on the transnational aspects of the results could be added to the discussion.

We thank the reviewer for highlighting this point. We have added this to the discussion.

(2.5) Figure 3 suggests that Jag2 is produced by basal sebocytes and used by these cells to induce sebocyte differentiation. I'm wondering if in an in vitro cell system (with a mixture of marked Jag2-expressing cells and marked Jag2-negative cells), it would be possible to understand if this mechanism of differentiation is a cell-autonomous mechanism or a mechanism based on cell competition (for instance, it would be possible that the progenitors compete for their niche on the basal layer by pushing neighbouring basal cells to differentiate presenting them Jag2).

We thank the reviewer for the insightful suggestion. The mechanistic underpinning of how Notch signaling induces sebocyte differentiation is still unclear, and we find the reviewer’s suggestion very interesting. However, establishing an in vitro model that captures the aspects mentioned, would require a lot of optimization and validation. To help rapid dissemination of our findings we elected to keep this out of the manuscript, but we will certainly consider it for future studies.

Reviewer #3 (Recommendations For The Authors):

(3.1) The authors focussed on mouse back skin sebaceous glands to analyse the phenotype. Are the effects also reproducible in the sebaceous glands of the mouse ears and tail epidermis? If so, the data should be strengthened by quantifying the phenotype using tail epidermal whole mounts (Braun et al., 2003; Development, PMID: 12954714), ideally by co-staining sebaceous glands for differentiation markers (e.g. FASN, Adipophilin) or lipid deposits (e.g., Oil red O). Also, the authors need to clarify how many sebaceous glands were scored per mouse. If not, please provide a rationale explaining the location restriction.

We thank the reviewer for pointing this out. Indeed, we have only incorporated data from the telogen dorsal skin of the animals. We have now more accurately reflected this in the revised manuscript. Additionally, we have added the number of sebaceous glands quantified in each figure per the reviewer’s suggestion.

Since the stage of hair growth cycle can affect the sebaceous glands, we chose the resting (telogen) phase of the hair cycle to reliably study the sebaceous glands. At 8 weeks of age, hair follicles have uniformly entered the telogen phase. As subsequent re-entry into the anagen phase is asynchronous in the adult skin, the color of the dorsal skin of C57BL/6 mice can be used to determine whether the hair follicles are in the telogen phase or not. These reasons led us to choose this location, allowing us to study only telogen phase hair follicles.

We also point out that previously reported data (Estrach et al., 2006) did not show differences between dorsal and tail skin, so we assume the mechanisms must largely be conserved. However, as the reviewer rightfully points out, we cannot be sure and have, therefore, indicated the dorsal location throughout the manuscript.

(3.2) The micrographs in Figure 2 suggest that expression of both Jagged2 and Notch1 (intercellular domain) is not restricted to the sebaceous glands, as both molecules appear to be detected also in the isthmus and lower hair follicle. Of note, the online tool provided by the Kasper and Linnarsson labs (http://linnarssonlab.org/epidermis/) shows that both molecules are more widely expressed in mouse back skin. Please provide some analysis of the overall expression of these molecules in mouse skin. In line, is the observed effect of using the antagonising antibodies restricted to the sebaceous glands? Please provide additional data on proliferation and differentiation in the interfollicular epidermis, hair follicle cycling, and other skin compartments. For instance, the data published in the cited paper by Lafkas et al. (2005) suggest a thickening of the dermal adipocyte layer upon Jagged2 inhibition using monoclonal therapeutic antibodies.

The reviewer is correct in noting that expression of both Jag2 and Notch1 is not restricted to the sebaceous gland. The Notch signaling pathway is a well-known regulator for epidermal differentiation, and members of the pathway are expressed in various locations of the skin, including the interfollicular epidermis and the hair follicle. The expression and function of Notch signaling in these locations has been reviewed in (Hsu et al., 2014; Nowell and Radtke, 2013; Watt et al., 2008). We have also added zoomed out images showing expression of Jag2 and Notch1 in the skin (Figure S2e,f).

The effect of the antagonizing antibodies is not restricted to sebaceous glands, as we already noted in our discussion section: “While injections of the Notch blocking antibodies are systemic, we only observed a reduction in the number of Notch-active cells in the IFE, but not a complete loss.” The functional impact of the antibodies is likely beyond the sebaceous gland, as the reviewer points out, but understanding the full effect in other compartments, we consider beyond the scope of the current study.

In our previous study (Lafkas et al., 2015), the skin was examined at different animal ages/gender and using different antibody dosing regimens, which is the likely explanation for the differences observed. We have now quantified the width of the adipocyte layer and the IFE and show that there are no significant differences between treatments (Figure S1g-j). This together with the histology suggest that there are no significant differences in the differentiation and proliferation of these compartments.

(3.3) Since Jagged1 is a Wnt/beta-catenin target gene that is essential for (ectopic) hair follicle formation and differentiation (Estrach et al., 2006, Development, PMID: 17035290) and the sebaceous gland is widely considered as an epidermal compartment with absent/low Wnt/beta-catenin pathway activity during normal homeostasis (Lim & Nusse, 2013, Cold Spring Habor Perspectives in Biology, PMID: 23209129), how is the expression of Notch1 and Jagged2 regulated upstream in sebocyte progenitors? It would be important to bring some more mechanistic insights into the upstream regulation of Notch activity. In line with comment 2, how are the compartment-specific effects molecularly regulated if the effects are not restricted to the sebaceous glands?

The reviewer is correct in noting that the Wnt pathway does not seem to be a likely candidate for driving sebocyte differentiation through Notch signaling. Indeed, Wnt inhibition is required for sebocyte differentiation (Merrill et al., 2001; Niemann et al., 2002), and the Jag2 promoter region also does not contain TCF binding sites (Katoh and Katoh, 2006).

We speculate that Myc might regulate Notch signaling in the sebaceous gland. It is expressed in the sebaceous gland basal stem cells and has been reported to positively regulate sebocyte differentiation (Cottle et al., 2013). In addition, studies have shown that Jag2 is a Myc target gene (Fiaschetti et al., 2014; Yustein et al., 2010). However, evaluating which upstream pathway potentially regulates Notch signaling, and resolving the regulatory network of sebocyte differentiation beyond the direct Notch ligands and receptors would require extensive in vivo modeling using KO and transgenic animals, which we consider to be beyond the scope of the current manuscript.

References

Cottle DL, Kretzschmar K, Schweiger PJ, Quist SR, Gollnick HP, Natsuga K, Aoyagi S, Watt FM. 2013. c-MYC-Induced Sebaceous Gland Differentiation Is Controlled by an Androgen Receptor/p53 Axis. Cell Rep 3:427–441. doi:10.1016/j.celrep.2013.01.013

Estrach S, Ambler CA, Celso CLL, Hozumi K, Watt FM. 2006. Jagged 1 is a β-catenin target gene required for ectopic hair follicle formation in adult epidermis. Development 133:4427–4438. doi:10.1242/dev.02644

Fiaschetti G, Schroeder C, Castelletti D, Arcaro A, Westermann F, Baumgartner M, Shalaby T, Grotzer MA. 2014. NOTCH ligands JAG1 and JAG2 as critical pro-survival factors in childhood medulloblastoma. Acta Neuropathol Commun 2:39. doi:10.1186/2051-5960-2-39

Hsu Y-C, Li L, Fuchs E. 2014. Emerging interactions between skin stem cells and their niches. Nat Med 20:847–856. doi:10.1038/nm.3643

Katoh Masuko, Katoh Masaru. 2006. Notch ligand, JAG1, is evolutionarily conserved target of canonical WNT signaling pathway in progenitor cells. Int J Mol Med. doi:10.3892/ijmm.17.4.681

Lafkas D, Shelton A, Chiu C, Boenig G de L, Chen Y, Stawicki SS, Siltanen C, Reichelt M, Zhou M, Wu X, Eastham-Anderson J, Moore H, Roose-Girma M, Chinn Y, Hang JQ, Warming S, Egen J, Lee WP, Austin C, Wu Y, Payandeh J, Lowe JB, Siebel CW. 2015. Therapeutic antibodies reveal Notch control of transdifferentiation in the adult lung. Nature 528:127–131. doi:10.1038/nature15715

Merrill BJ, Gat U, DasGupta R, Fuchs E. 2001. Tcf3 and Lef1 regulate lineage differentiation of multipotent stem cells in skin. Genes Dev 15:1688–1705. doi:10.1101/gad.891401

Niemann C, Owens DM, Hülsken J, Birchmeier W, Watt FM. 2002. Expression of ΔNLef1 in mouse epidermis results in differentiation of hair follicles into squamous epidermal cysts and formation of skin tumours. Development 129:95–109. doi:10.1242/dev.129.1.95

Nowell C, Radtke F. 2013. Cutaneous Notch Signaling in Health and Disease. Cold Spring Harb Perspect Med 3:a017772. doi:10.1101/cshperspect.a017772

Watt FM, Estrach S, Ambler CA. 2008. Epidermal Notch signalling: differentiation, cancer and adhesion. Curr Opin Cell Biol 20:171–179. doi:10.1016/j.ceb.2008.01.010

Yustein JT, Liu Y-C, Gao P, Jie C, Le A, Vuica-Ross M, Chng WJ, Eberhart CG, Bergsagel PL, Dang CV. 2010. Induction of ectopic Myc target gene JAG2 augments hypoxic growth and tumorigenesis in a human B-cell model. Proc Natl Acad Sci 107:3534–3539. doi:10.1073/pnas.0901230107

Author response:

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

eLife assessment

This important research uses an elegant combination of protein-protein biochemistry, genetics, and microscopy to demonstrate that the novel bacterial protein FipA is required for polar flagella synthesis and binds to FlhF in multiple bacterial species. This manuscript is convincing, providing evidence for the early stages of flagellar synthesis at a cell pole; however, the protein biochemistry is incomplete and would benefit from additional rigorous experiments. This paper could be of significant interest to microbiologists studying bacterial motility, appendages, and cellular biology.

We are very grateful for the very positive and helpful evaluation.

Joint Public Review:

Bacteria exhibit species-specific numbers and localization patterns of flagella. How specificity in number and pattern is achieved in Gamma-proteobacteria needs to be better understood but often depends on a soluble GTPase called FlhF. Here, the authors take an unbiased protein-pulldown approach with FlhF, resulting in identifying the protein FipA in V. parahaemolyticus. They convincingly demonstrate that FipA interacts genetically and biochemically with previously known spatial regulators HubP and FlhF. FipA is a membrane protein with a cytoplasmic DUF2802; it co-localizes to the flagellated pole with HubP and FlhF. The DUF2802 mediates the interaction between FipA and FlhF, and this interaction is required for FipA function. Altogether, the authors show that FipA likely facilitates the recruitment of FlhF to the membrane at the cell pole together with the known recruitment factor HupB. This finding is crucial in understanding the mechanism of polar localization. The authors show that FipA co-occurs with FlhF in the genomes of bacteria with polarly-localized flagella and study the role of FipA in three of these organisms: V. parahaemolyticus, S. purtefaciens, and P. putida. In each case, they show that FipA contributes to FlhF polar localization, flagellar assembly, flagellar patterning, and motility, though the details differ among the species. By comparing the role of FipA in polar flagellum assembly in three different species, they discover that, while FipA is required in all three systems, evolution has brought different nuances that open avenues for further discoveries.

Strengths:

The discovery of a novel factor for polar flagellum development. The solid nature and flow of the experimental work.

The authors perform a comprehensive analysis of FipA, including phenotyping of mutants, protein localization, localization dependence, and domains of FipA necessary for each. Moreover, they perform a time-series analysis indicating that FipA localizes to the cell pole likely before, or at least coincident with, flagellar assembly. They also show that the role of FipA appears to differ between organisms in detail, but the overarching idea that it is a flagellar assembly/localization factor remains convincing.

The work is well-executed, relying on bacterial genetics, cell biology, and protein interaction studies. The analysis is deep, beginning with discovering a new and conserved factor, then the molecular dissection of the protein, and finally, probing localization and interaction determinants. Finally, the authors show that these determinants are important for function; they perform these studies in parallel in three model systems.

Weaknesses:

The comparative analysis in the different organisms was on balance, a weakness. Mixing the data for the organisms together made the text difficult to read and took away key points from the results. The individual details crowded out the model in its current form. Indeed, because some of the phenotypes and localization dependencies differ between model systems, the comparison is challenging to the reader. The authors could more clearly state what these differences mean, why they arise, and (in the discussion) how they might relate to the organism's lifestyle.

More experiments would be needed to fully analyze the effects of interacting proteins on individual protein stability; this absence slightly detracted from the conclusions.

We have tried our best to improve the manuscript according to the insightful suggestions of the reviewers. Please find our answers to the raised issues below.

Reviewer #1 (Recommendations For The Authors):

We are very grateful to this reviewer for the very positive evaluation and the great suggestions to improve the manuscript.

I think there is value to the comparative analysis but how to present it in such a way that the key similarities and differences stand out is the challenge. Perhaps a table that compares the three datasets is sufficient. Or tell the story of V. parahaemolyticus first to establish the model, followed by comparative analysis of the other two organisms highlighting differences and relegating similarities to supplemental?

We agree that the our previous presentation of our comparative analysis made it very hard to follow the major findings and the general role(s) of FipA, and we are very grateful for the suggestions on how to improve this. We have decided to change the presentation as the reviewer recommended. We used V. parahaemolyticus as a ‚lead model‘ to describe the role of FipA, and we then compared the major findings to the other two species. We hope that the story is now easier to follow.

This is not something that needs to be addressed in the text but I wanted to bring the protein SwrB to the authors' attention which may further expand FipA relevance. Bacillus subtilis uses FlhFG to somehow pattern flagella in a peritrichous arrangement and there are a number of striking similarities, in my opinion, between FipA and SwrB. The two proteins have very similar domain architecture/topology, both proteins promote flagellar assembly, and the genetic neighborhood/operon organization is uncannily similar. There are other more minor similarities dependent on the organism in this paper.

Phillips, Kearns. 2021. Molecular and cell biological analysis of SwrB in Bacillus subtilis. J Bacteriol 203:e0022721

Phillips, Kearns. 2015. Functional activation of the flagellar type III secretion export apparatus. PLoS Genet 11:e1005443.

We thank this reviewer for pointing out these intriguing similarities. For this study we have decided to exclusively concentrate on polarly flagellated bacteria. FlhF und FlhG are also present in B. subtilis where they play a role in organizing flagellation, but we feel that this would be out of scope for this manuscript.

Reviewer #2 (Recommendations For The Authors):

We would like to thank this reviewer for the very positive evaluation and for pointing out several issues to strengthen the story.

Figure 3A data are problematic since everything is too small to visualize. Since these are functional GFP fusions (or mCherry for 2E data), why are they not presented in color?

Again - why are color figures not used to help the reader in Fig 4A and 5F & 5G to confirm what is asserted?

Again, it is difficult to see the images presented. It is asserted that FipA is recruited to the cell pole after cell division and before flagellum assembly, but one has to take their word for it.

We fully agree that in some case the localization pattern is hard to see on the micrographs presented. We have, therefore, provided enlarged micrographs in the supplemental part which allow to better see the fluorescent foci within the cells. With respect to presentations in color – we found that this did not improve the visibility of localizations and therefore have decided to use the grayscale images.

Here, what is missing are turnover assays. Do FipA, FlhF, and HubP all co-localize as complex or is the absence of one leading to the protein turnover of other partners? I think this needs to be sorted out before final conclusions can be made.

Thanks for pointing out this important point. We have now provided western analysis which demonstrate that FipA and FlhF are produced and stable in the absence of the other partners (see Supplemental Figure 5). Stability of HubP as a general polar marker not only required for flagellation was not determined.

Minor comments:

Line 58: change "around" to "in timing with"

Line 79: what "signal" is transferred from the C-ring to the MS-ring. Are they not fully connected such that rotation is the entire structure - C-ring-MS-ring-Rod-Hook-Filament. Is it not the change in the relationship to the stator complex where the signal is transferred?

Line 85: change "counting" to "control of flagellar numbers per cell"

Line 110: change "is (co-)responsible for recruiting" to "facilitates recruitment of"

Thanks for pointing this out. We have adjusted the wording according to the reviewer’s suggestions.

Given that motility phenotypes vary on individual plates (volumes and dryness vary), why in Figure 2C are the motility assays for fipA and flhF mutants of P. putida done on different plates?

For better visualisation, we have rearranged the spreading halos for the figure. All strain spreading comparisons on soft agar were always conducted on the same plate due to the reasons this reviewer mentioned.

Reviewer #3 (Recommendations For The Authors):

We thank this reviewer for the very positive evalution and the great suggestions.

One possibility is to describe first all the results relating to FipA in Vibrio and then add the result sections at the end to illustrate the differences between Vibrio and Shewanella, and then Vibrio and Pseudomonas. This may make it easier to follow for the reader.

We agree that the our previous presentation of our comparative analysis made it very hard to follow the major findings and the general role(s) of FipA, and we are very grateful for the suggestions on how to improve this. We have decided to change the presentation as the reviewer recommended. We used V. parahaemolyticus as a ‚lead model‘ to describe the role of FipA, and we then compared the major findings to the other two species. We hope that the story is now easier to follow.

I would have liked to see some TEM analysis of flagella in fipA/hubP double mutants strains and was also wondering if FipA/FlhF/HubP colocalization had been studied in E. coli when all proteins are expressed together, at least with two bearing fluorescent tags.

Thanks for these great suggestions. In this study, we have concentrated on the localization of FlhF by FipA and HubP. HubP has multiple functions in the cell and may also affect flagellar synthesis to some extent in a species-specific fashion. Therefore, any findings would have to be discussed very carefully, so we have decided to leave that out for the time being.

With respect to the FipA/HubP/FlhF production in a heterologous host such as E. coli, this has been partly done (without FipA) in a second parallel story (see reference to Dornes et al (2024) in this manuscript). Rebuilding larger parts of the system in a heterologous host is currently done in an independent study. Therefore, we have decided not to include this already here.

From the Reviewing Editor:

We are grateful for handling the fair reviewing process, for the positive evaluation and the helpful hints.

The microscopy was inconsistent (DIC versus phase) for unclear reasons. Did using different microscopes impact the ability to acquire low-intensity fluorescence signals? Please add a sentence in the Methods section to clarify.

We are sorry for this inconsistency. As the imaging was carried out by different labs (to some part before the projects were joined), the corresponding preferred microscopy settings were used. We have added an explaining sentence to the Methods section.

Also, some subcellular fluorescence localizations were not visible in the selected images (e.g., Figures 3 and 5). The reader had to rely on the authors' statements and analyses. The conclusions could be more robust with fluorescence measurements across the cell body for a subset of cells. The authors could provide this data analysis in the Supplemental; this measurement would more clearly show an accumulation of fluorescence at the cell pole, particularly in low-intensity images.

We fully agree that in some case the localization pattern is hard to see on the micrographs presented. Unfortunately, often the signal is not sufficiently strong to provied proper demographs. We have, therefore, provided enlarged micrographs in the supplemental part, which allow to better see the fluorescent foci within the cells.

Author response:

We sincerely thank the reviewers for their thoughtful, critical, and constructive comments, which will help us in further exploring the mechanisms by which LDH regulates glycolysis, the tricarboxylic acid cycle, and oxidative phosphorylation future studies. The following is our responses to the reviewers' comments.

Reviewer #1 (Public Review):

Summary:

Zeng et al. have investigated the impact of inhibiting lactate dehydrogenase (LDH) on glycolysis and the tricarboxylic acid cycle. LDH is the terminal enzyme of aerobic glycolysis or fermentation that converts pyruvate and NADH to lactate and NAD+ and is essential for the fermentation pathway as it recycles NAD+ needed by upstream glyceraldehyde-3-phosphate dehydrogenase. As the authors point out in the introduction, multiple published reports have shown that inhibition of LDH in cancer cells typically leads to a switch from fermentative ATP production to respiratory ATP production (i.e., glucose uptake and lactate secretion are decreased, and oxygen consumption is increased). The presumed logic of this metabolic rearrangement is that when glycolytic ATP production is inhibited due to LDH inhibition, the cell switches to producing more ATP using respiration. This observation is similar to the well-established Crabtree and Pasteur effects, where cells switch between fermentation and respiration due to the availability of glucose and oxygen. Unexpectedly, the authors observed that inhibition of LDH led to inhibition of respiration and not activation as previously observed. The authors perform rigorous measurements of glycolysis and TCA cycle activity, demonstrating that under their experimental conditions, respiration is indeed inhibited. Given the large body of work reporting the opposite result, it is difficult to reconcile the reasons for the discrepancy. In this reviewer's opinion, a reason for the discrepancy may be that the authors performed their measurements 6 hours after inhibiting LDH. Six hours is a very long time for assessing the direct impact of a perturbation on metabolic pathway activity, which is regulated on a timescale of seconds to minutes. The observed effects are likely the result of a combination of many downstream responses that happen within 6 hours of inhibiting LDH that causes a large decrease in ATP production, inhibition of cell proliferation, and likely a range of stress responses, including gene expression changes.

Strengths:

The regulation of metabolic pathways is incompletely understood, and more research is needed, such as the one conducted here. The authors performed an impressive set of measurements of metabolite levels in response to inhibition of LDH using a combination of rigorous approaches.

Weaknesses:

Glycolysis, TCA cycle, and respiration are regulated on a timescale of seconds to minutes. The main weakness of this study is the long drug treatment time of 6 hours, which was chosen for all the experiments. In this reviewer's opinion, if the goal was to investigate the direct impact of LDH inhibition on glycolysis and the TCA cycle, most of the experiments should have been performed immediately after or within minutes of LDH inhibition. After 6 hours of inhibiting LDH and ATP production, cells undergo a whole range of responses, and most of the observed effects are likely indirect due to the many downstream effects of LDH and ATP production inhibition, such as decreased cell proliferation, decreased energy demand, activation of stress response pathways, etc.

We appreciate the reviewer’s critical comments. The main argument is whether the inhibition of LDH induces a temporal perturbation in glycolysis, the TCA cycle, and OXPHOS, or if it leads to a shift to a new steady state. We argue that this shift represents a transition between two steady states; specifically, GNE-140 treatment drives metabolism from one steady state to another.

Before conducting the experiment, we performed a time course experiment, measuring glucose consumption and lactate production in cells treated with GNE-140. The results demonstrated a very good linearity, indicating that the glycolytic rate remained constant—thus confirming that glycolysis was at steady state. Given the tight coupling between glycolysis, the TCA cycle, and OXPHOS, we infer that the TCA cycle and OXPHOS were also at steady state. However, this ‘infer’ requires further confirmation.

Multiple published reports have shown that LDH inhibition in cancer cells causes a shift from fermentative ATP production to respiratory ATP production. This notion persists because it is often compared to the well-established Crabtree and Pasteur effects, where cells toggle between fermentation and respiration based on glucose and oxygen availability. However, in the Pasteur or Crabtree effects, the deprivation of oxygen—the terminal electron acceptor—drives the switch, which is fundamentally different from LDH inhibition.

Reviewer #2 (Public Review):

Summary:

Zeng et al. investigated the role of LDH in determining the metabolic fate of pyruvate in HeLa and 4T1 cells. To do this, three broad perturbations were applied: knockout of two LDH isoforms (LDH-A and LDH-B), titration with a non-competitive LDH inhibitor (GNE-140), and exposure to either normoxic (21% O2) or hypoxic (1% O2) conditions. They show that knockout of either LDH isoform alone, though reducing both protein level and enzyme activity, has virtually no effect on either the incorporation of a stable 13C-label from a 13C6-glucose into any glycolytic or TCA cycle intermediate, nor on the measured intracellular concentrations of any glycolytic intermediate (Figure 2). The only apparent exception to this was the NADH/NAD+ ratio, measured as the ratio of F420/F480 emitted from a fluorescent tag (SoNar).

The addition of a chemical inhibitor, on the other hand, did lead to changes in glycolytic flux, the concentrations of glycolytic intermediates, and in the NADH/NAD+ ratio (Figure 3). Notably, this was most evident in the LDH-B-knockout, in agreement with the increased sensitivity of LDH-A to GNE-140 (Figure 2). In the LDH-B-knockout, increasing concentrations of GNE-140 increased the NADH/NAD+ ratio, reduced glucose uptake, and lactate production, and led to an accumulation of glycolytic intermediates immediately upstream of GAPDH (GA3P, DHAP, and FBP) and a decrease in the product of GAPDH (3PG). They continue to show that this effect is even stronger in cells exposed to hypoxic conditions (Figure 4). They propose that a shift to thermodynamic unfavourability, initiated by an increased NADH/NAD+ ratio inhibiting GAPDH explains the cascade, calculating ΔG values that become progressively more endergonic at increasing inhibitor concentrations.

Then - in two separate experiments - the authors track the incorporation of 13C into the intermediates of the TCA cycle from a 13C6-glucose and a 13C5-glutamine. They use the proportion of labelled intermediates as a proxy for how much pyruvate enters the TCA cycle (Figure 5). They conclude that the inhibition of LDH decreases fermentation, but also the TCA cycle and OXPHOS flux - and hence the flux of pyruvate to all of those pathways. Finally, they characterise the production of ATP from respiratory or fermentative routes, the concentration of a number of cofactors (ATP, ADP, AMP, NAD(P)H, NAD(P)+, and GSH/GSSG), the cell count, and cell viability under four conditions: with and without the highest inhibitor concentration, and at norm- and hypoxia. From this, they conclude that the inhibition of LDH inhibits the glycolysis, the TCA cycle, and OXPHOS simultaneously (Figure 7).

Strengths:

The authors present an impressively detailed set of measurements under a variety of conditions. It is clear that a huge effort was made to characterise the steady-state properties (metabolite concentrations, fluxes) as well as the partitioning of pyruvate between fermentation as opposed to the TCA cycle and OXPHOS.

A couple of intermediary conclusions are well supported, with the hypothesis underlying the next measurement clearly following. For instance, the authors refer to literature reports that LDH activity is highly redundant in cancer cells (lines 108 - 144). They prove this point convincingly in Figure 1, showing that both the A- and B-isoforms of LDH can be knocked out without any noticeable changes in specific glucose consumption or lactate production flux, or, for that matter, in the rate at which any of the pathway intermediates are produced. Pyruvate incorporation into the TCA cycle and the oxygen consumption rate are also shown to be unaffected.

They checked the specificity of the inhibitor and found good agreement between the inhibitory capacity of GNE-140 on the two isoforms of LDH and the glycolytic flux (lines 229 - 243). The authors also provide a logical interpretation of the first couple of consequences following LDH inhibition: an increased NADH/NAD+ ratio leading to the inhibition of GAPDH, causing upstream accumulations and downstream metabolite decreases (lines 348 - 355).

Weaknesses:

Despite the inarguable comprehensiveness of the data set, a number of conceptual shortcomings afflict the manuscript. First and foremost, reasoning is often not pursued to a logical conclusion. For instance, the accumulation of intermediates upstream of GAPDH is proffered as an explanation for the decreased flux through glycolysis. However, in Figure 3C it is clear that there is no accumulation of the intermediates upstream of PFK. It is unclear, therefore, how this traffic jam is propagated back to a decrease in glucose uptake. A possible explanation might lie with hexokinase and the decrease in ATP (and constant ADP) demonstrated in Figure 6B, but this link is not made.

We appreciate the reviewer's critical comment. In Figure 3C, there is no accumulation of F6P or G6P, which are upstream of PFK1. This is because the PFK1-catalyzed reaction sets a significant thermodynamic barrier. Even with treatment using 30 μM GNE-140, the ∆GPFK1 (Gibbs free energy of the PFK1-catalyzed reaction) remains -9.455 kJ/mol (Figure 3D), indicating that the reaction is still far from thermodynamic equilibrium, thereby preventing the accumulation of F6P and G6P.

We agree with the reviewer that hexokinase inhibition may play a role, this requires further investigation.

The obvious link between the NADH/NAD+ ratio and pyruvate dehydrogenase (PDH) is also never addressed, a mechanism that might explain how the pyruvate incorporation into the TCA cycle is impaired by the inhibition of LDH (the observation with which they start their discussion, lines 511 - 514).

We agree with the reviewer’s comment. In this study, we did not explore how the inhibition of LDH affects pyruvate incorporation into the TCA cycle. As this mechanism was not investigated, we have titled the study: "Elucidating the Kinetic and Thermodynamic Insights into the Regulation of Glycolysis by Lactate Dehydrogenase and Its Impact on the Tricarboxylic Acid Cycle and Oxidative Phosphorylation in Cancer Cells."

It was furthermore puzzling how the ΔG, calculated with intracellular metabolite concentrations (Figures 3 and 4) could be endergonic (positive) for PGAM at all conditions (also normoxic and without inhibitor). This would mean that under the conditions assayed, glycolysis would never flow completely forward. How any lactate or pyruvate is produced from glucose, is then unexplained.

This issue also concerned me during the study. However, given the high reproducibility of the data, we consider it is true, but requires explanation.

The PGAM-catalyzed reaction is tightly linked to both upstream and downstream reactions in the glycolytic pathway. In glycolysis, three key reactions catalyzed by HK2, PFK1, and PK are highly exergonic, providing the driving force for the conversion of glucose to pyruvate. The other reactions, including the one catalyzed by PGAM, operate near thermodynamic equilibrium and primarily serve to equilibrate glycolytic intermediates rather than control the overall direction of glycolysis, as previously described by us (J Biol Chem. 2024 Aug 8;300(9):107648).

The endergonic nature of the PGAM-catalyzed reaction does not prevent it from proceeding in the forward direction. Instead, the directionality of the pathway is dictated by the exergonic reaction of PFK1 upstream, which pushes the flux forward, and by PK downstream, which pulls the flux through the pathway. The combined effects of PFK1 and PK may account for the observed endergonic state of the PGAM reaction.

However, if the PGAM-catalyzed reaction were isolated from the glycolytic pathway, it would tend toward equilibrium and never surpass it, as there would be no driving force to move the reaction forward.

Finally, the interpretation of the label incorporation data is rather unconvincing. The authors observe an increasing labelled fraction of TCA cycle intermediates as a function of increasing inhibitor concentration. Strangely, they conclude that less labelled pyruvate enters the TCA cycle while simultaneously less labelled intermediates exit the TCA cycle pool, leading to increased labelling of this pool. The reasoning that they present for this (decreased m2 fraction as a function of DHE-140 concentration) is by no means a consistent or striking feature of their titration data and comes across as rather unconvincing. Yet they treat this anomaly as resolved in the discussion that follows.

GNE-140 treatment increased the labeling of TCA cycle intermediates by [13C6]glucose but decreased the OXPHOS rate, we consider the conflicting results as an 'anomaly' that warrants further explanation. To address this, we analyzed the labeling pattern of TCA cycle intermediates using both [13C6]glucose and  [13C5]glutamine. Tracing the incorporation of glucose- and glutamine-derived carbons into the TCA cycle suggests that LDH inhibition leads to a reduced flux of glucose-derived acetyl-CoA into the TCA cycle, coupled with a decreased flux of glutamine-derived α-KG, and a reduction in the efflux of intermediates from the cycle. These results align with theoretical predictions. Under any condition, the reactions that distribute TCA cycle intermediates to other pathways must be balanced by those that replenish them. In the GNE-140 treatment group, the entry of glutamine-derived carbon into the TCA cycle was reduced, implying that glucose-derived carbon (as acetyl-CoA) entering the TCA cycle must also be reduced, or vice versa.

This step-by-step investigation is detailed under the subheading "The Effect of LDHB KO and GNE-140 on the Contribution of Glucose Carbon to the TCA Cycle and OXPHOS" in the Results section in the manuscript.

In the Discussion, we emphasize that caution should be exercised when interpreting isotope tracing data. In this study, treatment of cells with GNE-140 led to an increase labeling percentage of TCAC intermediates by [13C6]glucose (Figure 5A-E). However, this does not necessarily imply an increase in glucose carbon flux into TCAC; rather, it indicates a reduction in both the flux of glucose carbon into TCAC and the flux of intermediates leaving TCAC. When interpreting the data, multiple factors must be considered, including the carbon-13 labeling pattern of the intermediates (m1, m2, m3, ---) (Figure 5G-K), replenishment of intermediates by glutamine (Figure 5M-V), and mitochondrial oxygen consumption rate (Figure 5W). All these factors should be taken into account to derive a proper interpretation of the data. 

Reviewer #3 (Public Review):

Hu et al in their manuscript attempt to interrogate the interplay between glycolysis, TCA activity, and OXPHOS using LDHA/B knockouts as well as LDH-specific inhibitors. Before I discuss the specifics, I have a few issues with the overall manuscript. First of all, based on numerous previous studies it is well established that glycolysis inhibition or forcing pyruvate into the TCA cycle (studies with PDKs inhibitors) leads to upregulation of TCA cycle activity, and OXPHOS, activation of glutaminolysis, etc (in this work authors claim that lowered glycolysis leads to lower levels of TCA activity/OXPHOS). The authors in the current work completely ignore recent studies that suggest that lactate itself is an important signaling metabolite that can modulate metabolism (actual mechanistic insights were recently presented by at least two groups (Thompson, Chouchani labs). In addition, extensive effort was dedicated to understanding the crosstalk between glycolysis/TCA cycle/OXPHOS using metabolic models (Titov, Rabinowitz labs). I have several comments on how experiments were performed. In the Methods section, it is stated that both HeLa and 4T1 cells were grown in RPMI-1640 medium with regular serum - but under these conditions, pyruvate is certainly present in the medium - this can easily complicate/invalidate some findings presented in this manuscript. In LDH enzymatic assays as described with cell homogenates controls were not explained or presented (a lot of enzymes in the homogenate can react with NADH!). One of the major issues I have is that glycolytic intermediates were measured in multiple enzyme-coupled assays. Although one might think it is a good approach to have quantitative numbers for each metabolite, the way it was done is that cell homogenates (potentially with still traces of activity of multiple glycolytic enzymes) were incubated with various combinations of the SAME enzymes and substrates they were supposed to measure as a part of the enzyme-based cycling reaction. I would prefer to see a comparison between numbers obtained in enzyme-based assays with GC-MS/LC-MS experiments (using calibration curves for respective metabolites, of course). Correct measurements of these metabolites are crucial especially when thermodynamic parameters for respective reactions are calculated. Concentrations of multiple graphs (Figure 1g etc.) are in "mM", I do not think that this is correct.

While the roles of lactate as a signaling metabolite and metabolic models are important areas of research, our work focuses on different aspects.

It is true that cell homogenates contain many enzymes that use NAD as a hydride acceptor or NADH as a hydride donor. However, in our assay system, the substrates are pyruvate and NADH, meaning only enzymes that catalyze the conversion of pyruvate + NADH to NAD + lactate can utilize NADH. Other enzymes do not interfere with this reaction. Although some enzymes may also catalyze this reaction, their catalytic efficiency is markedly lower than that of LDH, ensuring the validity of this assay.

Similarly, the assays for glycolytic intermediates are validated by the substrate specificity.

We have developed an LC-MS methodology for some glycolytic intermediates, but the accuracy of quantification remains unsatisfactory due to inherent limitations of this methodology.

Author response:

Reviewer #1 (Public review):

Summary:

Lodhiya et al. demonstrate that antibiotics with distinct mechanisms of action, norfloxacin, and streptomycin, cause similar metabolic dysfunction in the model organism Mycobacterium smegmatis. This includes enhanced flux through the TCA cycle and respiration as well as a build-up of reactive oxygen species (ROS) and ATP. Genetic and/or pharmacologic depression of ROS or ATP levels protect M. smegmatis from norfloxacin and streptomycin killing. Because ATP depression is protective, but in some cases does not depress ROS, the authors surmise that excessive ATP is the primary mechanism by which norfloxacin and streptomycin kill M. smegmatis. In general, the experiments are carefully executed; alternative hypotheses are discussed and considered; the data are contextualized within the existing literature. Clarification of the effect of 1) ROS depression on ATP levels and 2) ADP vs. ATP on divalent metal chelation would strengthen the paper, as would discussion of points of difference with the existing literature. The authors might also consider removing Figures 9 and 10A-B as they distract from the main point of the paper and appear to be the beginning of a new story rather than the end of the current one. Finally, statistics need some attention.

Strengths:

The authors tackle a problem that is both biologically interesting and medically impactful, namely, the mechanism of antibiotic-induced cell death.

Experiments are carefully executed, for example, numerous dose- and time-dependency studies; multiple, orthogonal readouts for ROS; and several methods for pharmacological and genetic depletion of ATP.

There has been a lot of excitement and controversy in the field, and the authors do a nice job of situating their work in this larger context.

Inherent limitations to some of their approaches are acknowledged and discussed e.g., normalizing ATP levels to viable counts of bacteria.

We sincerely thanks appreciate the reviewer’s encouraging feedback.

Weaknesses:

The authors have shown that treatments that depress ATP do not necessarily repress ROS, and therefore conclude that ATP is the primary cause of norfloxacin and streptomycin lethality for M. smegmatis. Indeed, this is the most impactful claim of the paper. However, GSH and dipyridyl beautifully rescue viability. Do these and other ROS-repressing treatments impact ATP levels? If not, the authors should consider a more nuanced model and revise the title, abstract, and text accordingly.

We thank the reviewer for asking this question. In the revised version of the manuscript, we will include data on the impact of the antioxidant GSH on ATP levels.

Does ADP chelate divalent metal ions to the same extent as ATP? If so, it is difficult to understand how conversion of ADP to ATP by ATP synthase would alter metal sequestration without concomitant burst in ADP levels.

We sincerely thank the reviewer for raising this insightful question. Indeed, ADP and AMP can also form complexes with divalent metal ions; however, these complexes tend to be less stable. According to the existing literature, ATP-metal ion complexes exhibit a higher formation constant compared to ADP or AMP complexes. This has been attributed to the polyphosphate chain of ATP, which acts as an active site, forming a highly stable tridentate structure (Khan et al., 1962; Distefano et al., 1953). An antibiotic-induced increase in ATP levels, irrespective of any changes in ADP levels, could still result in the formation of more stable complexes with metal ions, potentially leading to metal ion depletion. Although recent studies indicate that antibiotic treatment stimulates purine biosynthesis (Lobritz MA et al., 2022; Yang JH et al., 2019), thereby imposing energy demands and enhancing ATP production, the possibility of a corresponding increase in total purine nucleotide levels (ADP+ATP) exist (is mentioned in discussion section). However, this hypothesis requires further investigation.

Khan MMT, Martell AE. Metal Chelates of Adenosine Triphosphate. Journal of Physical Chemistry (US). 1962 Jan 1;Vol: 66(1):10–5

Distefano v, Neuman wf. Calcium complexes of adenosinetriphosphate and adenosinediphosphate and their significance in calcification in vitro. Journal of Biological Chemistry. 1953 Feb 1;200(2):759–63

Lobritz MA, Andrews IW, Braff D, Porter CBM, Gutierrez A, Furuta Y, et al. Increased energy demand from anabolic-catabolic processes drives β-lactam antibiotic lethality. Cell Chem Biol [Internet]. 2022 Feb 17.

Yang JH, Wright SN, Hamblin M, McCloskey D, Alcantar MA, Schrübbers L, et al. A White-Box Machine Learning Approach for Revealing Antibiotic Mechanisms of Action. Cell [Internet]. 2019 May 30

Some of the results in the paper diverge from what has been previously reported by some of the referenced literature. These discrepancies should be clarified.

We apologize for any confusion, but we are uncertain about the specific discrepancies the reviewer is referring. In the discussion section, we have addressed and analysed our results within the broader context of the existing literature, regardless of whether our findings align with or differ from previous studies.

Reviewer #2 (Public review):

Summary:

The authors are trying to test the hypothesis that ATP bursts are the predominant driver of antibiotic lethality of Mycobacteria.

Strengths:

This reviewer has not identified any significant strengths of the paper in its current form.

Weaknesses:

A major weakness is that M. smegmatis has a doubling time of three hours and the authors are trying to conclude that their data would reflect the physiology of M. tuberculosis which has a doubling time of 24 hours. Moreover, the authors try to compare OD measurements with CFU counts and thus observe great variabilities.

If the authors had evidence to support the conclusion that ATP burst is the predominant driver of antibiotic lethality in mycobacteria then this paper would be highly significant. However, with the way the paper is written, it is impossible to make this conclusion.

We have identified this new mechanism of antibiotic action in Mycobacterium smegmatis and have also mentioned that whether and how much of this mechanism is true in other organism needs to be tested as argued extensively in the discussion section of the manuscript.

We have always drawn inferences from the CFU counts as the OD600nm is never a reliable method as reported in all of our experiments.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors use microscopy experiments to track the gliding motion of filaments of the cyanobacteria Fluctiforma draycotensis. They find that filament motion consists of back-and-forth trajectories along a "track", interspersed with reversals of movement direction, with no clear dependence between filament speed and length. It is also observed that longer filaments can buckle and form plectonemes. A computational model is used to rationalize these findings.

We thank the reviewer for this accurate summary of the presented work.

Strengths:

Much work in this field focuses on molecular mechanisms of motility; by tracking filament dynamics this work helps to connect molecular mechanisms to environmentally and industrially relevant ecological behavior such as aggregate formation.

The observation that filaments move on tracks is interesting and potentially ecologically significant.

The observation of rotating membrane-bound protein complexes and tubular arrangement of slime around the filament provides important clues to the mechanism of motion.

The observation that long filaments buckle has the potential to shed light on the nature of mechanical forces in the filaments, e.g. through the study of the length dependence of buckling.

We thank the reviewer for listing these positive aspects of the presented work.

Weaknesses:

The manuscript makes the interesting statement that the distribution of speed vs filament length is uniform, which would constrain the possibilities for mechanical coupling between the filaments. However, Figure 1C does not show a uniform distribution but rather an apparent lack of correlation between speed and filament length, while Figure S3 shows a dependence that is clearly increasing with filament length. Also, although it is claimed that the computational model reproduces the key features of the experiments, no data is shown for the dependence of speed on filament length in the computational model. The statement that is made about the model "all or most cells contribute to propulsive force generation, as seen from a uniform distribution of mean speed across different filament lengths", seems to be contradictory, since if each cell contributes to the force one might expect that speed would increase with filament length.

We agree that the data shows in general a lack of correlation, rather than strictly being uniform. In the revised manuscript, we intend to collect more data from observations on glass to better understand the relation between filament length and speed. 

In considering longer filaments, one also needs to consider the increased drag created by each additional cell - in other words, overall friction will either increase or be constant as filament length increases. Therefore, if only one cell (or few cells) are generating motility forces, then adding more cells in longer filaments would decrease speed.

Since the current data does not show any decrease in speed with increasing filament length, we stand by the argument that the data supports that all (or most) cells in a filament are involved in force generation for motility. We would revise the manuscript to make this point - and our arguments about assuming multiple / most cells in a filament contributing to motility - clear.

The computational model misses perhaps the most interesting aspect of the experimental results which is the coupling between rotation, slime generation, and motion. While the dependence of synchronization and reversal efficiency on internal model parameters are explored (Figure 2D), these model parameters cannot be connected with biological reality. The model predictions seem somewhat simplistic: that less coupling leads to more erratic reversal and that the number of reversals matches the expected number (which appears to be simply consistent with a filament moving backwards and forwards on a track at constant speed).

We agree that the coupling between rotation, slime generation and motion is interesting and important when studying the specific mechanism leading to filament motion. However, we believe it even more fundamental to consider the intercellular coordination that is needed to realise this motion. Individual filaments are a collection of independent cells. This raises the question of how they can coordinate their thrust generation in such a way that the whole filament can both move and reverse direction of motion as a single unit. With the presented model, we want to start addressing precisely this point.

The model allows us to qualitatively understand the relation between coupling strength and reversals (erratic vs. coordinated motion of the filament). It also provides a hint about the possibility of de-coordination, which we then look for and identify in longer filaments.

While the model results seem obvious in hindsight, the analysis of the model allows phrasing the question of cell-to-cell coordination, which has not been brought up previously when considering the inherently multi-cell process of filament motility.

Filament buckling is not analysed in quantitative detail, which seems to be a missed opportunity to connect with the computational model, eg by predicting the length dependence of buckling.

Please note that Figure S10 provides an analysis of filament length and number of buckling instances observed. This suggests that buckling happens only in filaments above a certain length.

We do agree that further analyses of buckling - both experimentally and through modelling would be interesting.  This study, however,  focussed on cell-to-cell coupling / coordination during filament motility. We have identified the possibility of de-coordination through the use of a simple 1D model of motion, and found evidence of such de-coordination in experiments. Notice that the buckling we report does not depend on the filament hitting an external object. It is a direct result of a filament activity which, in this context, serves as evidence of cellular de-coordination.

Now that we have observed buckling and plectoneme formation, these processes need to be analysed with additional experiments and modelling. The appropriate model for this process needs to be 3D, and should ideally include torques arising from filament rotation. Experimentally, we need to identify means of influencing filament length and motion and see if we can measure buckling frequency and position across different filament lengths. These works are ongoing and will have to be summarised in a separate, future publication.

Reviewer #2 (Public review):

Summary:

The authors combined time-lapse microscopy with biophysical modeling to study the mechanisms and timescales of gliding and reversals in filamentous cyanobacterium Fluctiforma draycotensis. They observed the highly coordinated behavior of protein complexes moving in a helical fashion on cells' surfaces and along individual filaments as well as their de-coordination, which induces buckling in long filaments.

We thank the reviewer for this accurate summary of the presented work.

Strengths:

The authors provided concrete experimental evidence of cellular coordination and de-coordination of motility between cells along individual filaments. The evidence is comprised of individual trajectories of filaments that glide and reverse on surfaces as well as the helical trajectories of membrane-bound protein complexes that move on individual filaments and are implicated in generating propulsive forces.

We thank the reviewer for listing these positive aspects of the presented work.

Limitations:

The biophysical model is one-dimensional and thus does not capture the buckling observed in long filaments. I expect that the buckling contains useful information since it reflects the competition between bending rigidity, the speed at which cell synchronization occurs, and the strength of the propulsion forces.

Cell-to-cell coordination is a more fundamental phenomenon than the buckling and twisting of longer filaments, in that the latter is a consequence of limits of the former. In this sense, we are focussing here on something that we think is the necessary first step to understand filament gliding. The 3D motion of filaments (bending, plectoneme formation) is fascinating and can have important consequences for collective behaviour and macroscopic structure formation. As a consequence of cellular coupling, however, it is beyond the scope of the present paper.

Please also see our response above. We believe that the detailed analysis of buckling and plectoneme formation requires (and merits) dedicated experiments and modelling which go beyond the focus of the current study (on cellular coordination) and will constitute a separate analysis that stands on its own. We are currently working in that direction.

Future directions:

The study highlights the need to identify molecular and mechanical signaling pathways of cellular coordination. In analogy to the many works on the mechanisms and functions of multi-ciliary coordination, elucidating coordination in cyanobacteria may reveal a variety of dynamic strategies in different filamentous cyanobacteria.

We thank the reviewer for highlighting this point again and seeing the value in combining molecular and dynamical approaches.

Reviewer #3 (Public review):

Summary:

The authors present new observations related to the gliding motility of the multicellular filamentous cyanobacteria Fluctiforma draycotensis. The bacteria move forward by rotating their about their long axis, which causes points on the cell surface to move along helical paths. As filaments glide forward they form visible tracks. Filaments preferentially move within the tracks. The authors devise a simple model in which each cell in a filament exerts a force that either pushes forward or backwards. Mechanical interactions between cells cause neighboring cells to align the forces they exert. The model qualitatively reproduces the tendency of filaments to move in a concerted direction and reverse at the end of tracks.

We thank the reviewer for this accurate summary of the presented work.

Strengths:

The observations of the helical motion of the filament are compelling.

The biophysical model used to describe cell-cell coordination of locomotion is clear and reasonable. The qualitative consistency between theory and observation suggests that this model captures some essential qualities of the true system.

The authors suggest that molecular studies should be directly coupled to the analysis and modeling of motion. I agree.

We thank the reviewer for listing these positive aspects of the presented work and highlighting the need for combining molecular and biophysical approaches.

Weaknesses:

There is very little quantitative comparison between theory and experiment. It seems plausible that mechanisms other than mechano-sensing could lead to equations similar to those in the proposed model. As there is no comparison of model parameters to measurements or similar experiments, it is not certain that the mechanisms proposed here are an accurate description of reality. Rather the model appears to be a promising hypothesis.

We agree with the referee that the model we put forward is one of several possible. We note, however, that the assumption of mechanosensing by each cell - as done in this model - results in capturing both the alignment of cells within a filament (with some flexibility) and reversal dynamics. We have explored an even more minimal 1D model, where the cell’s direction of force generation is treated as an Ising-like spin and coupled between nearest neighbours (without assuming any specific physico-chemical basis). We found that this model was not fully able to capture both phenomena. In that model, we found that alignment required high levels of coupling (which is hard to justify except for mechanical coupling) and reversals were not readily explainable (and required additional assumptions). These points led us to the current, mechanically motivated model.

The parameterisation of the current model would require measuring cellular forces. To this end, a recent study has attempted to measure some of the physical parameters in a different filamentous cyanobacteria [1] and in our revision we will re-evaluate model parameters and dynamics in light of that study. We will also attempt to directly verify the presence of mechano-sensing by obstructing the movement of filaments.

Author response:

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

eLife assessment

The authors present valuable findings on how to determine the genetic architecture of extreme phenotype values by using data on sibling pairs. While the authors' derivations of the method are correct, the scenarios considered are incomplete, making it difficult to have confidence in the interpretation of the results as demonstrating the influence of de-novo or Mendelian (rare, penetrant-variant) architectures. The method nevertheless shows promise and will be of interest to researchers studying complex trait genetics. 

A.1: We have now expanded our consideration of the scenarios and we have ensured that we do not over-interpret our results as being due to de novo or Mendelian architectures. Instead, we make clear that our statistical tests are powered to identify these architectures but that there are other potential causes of significant results (e.g. measurement error or uncontrolled environmental factors from heavy-tailed distributions), making follow-up validation studies necessary before underlying architectures can be confirmed. We consider this to be typical of observational research, in which significant results may indicate causal effects unless uncontrolled confounding factors explain the observed associations, requiring experimental/trial follow-up for validation. We believe that our tests are useful for providing initial inference, and that in some settings – e.g. prioritising samples for sequencing to identify rare variants – could be useful as an initial screening step to increase the efficacy of a planned analysis or study.

Additionally, we have now developed “SibArc”, an openly available software tool that takes input sibling trait data and estimates conditional sibling heritability across the trait distribution. Then - based on our theoretical framework developed and described in the paper - for each tail of the trait distribution, estimates effect sizes and generates P-values corresponding to our de novo and Mendelian tests, and performs a Kolmogorov-Smirnov test to identify general departures from our null model. Furthermore, SibArc also provides additional functionality for users under preliminary beta form, for example, running an iterative optimisation routine to infer approximate relative degrees of polygenic, de novo, and Mendelian architectures prevailing in each trait tail. We have made this software tool, Quick Start tutorial, and sample data available online at Github and are hosting these on a dedicated website: www.sibarc.net.

Reviewer #1 (Public Review):

This is a clever and well-done paper that should be published. The authors sought to craft a method, applicable to biobank-scale data but without necessarily using genotyping or sequencing, to detect the presence of de novo mutations and rare variants that stand out from the polygenic background of a given trait. Their method depends essentially on sibling pairs where one sibling is in an extreme tail of the phenotypic distribution and whether the other sibling's regression to the mean shows a systematic deviation from what is expected under a simple polygenic architecture. 

Their method is successful in that it builds on a compelling intuition, rests on a rigorous derivation, and seems to show reasonable statistical power in the UK Biobank. (More biobanks of this size will probably become available in the near future.)  It is somewhat unsuccessful in that rejection of the null hypothesis does not necessarily point to the favored hypothesis of de novo or rare variants. The authors discuss the alternative possibility of rare environmental events of large effect. Maybe attention should be drawn to this in the abstract or the introduction of the paper. Nevertheless, since either of these possibilities is interesting, the method remains valuable. 

A.2: We agree with the reviewer that we should have made it clearer that - while our statistical tests are powered to identify de novo and Mendelian architectures – significant findings from our tests could also be explained by rare environmental events of large effect (specifically by uncontrolled environmental factors with heavy-tailed distributions). We have now made this clear throughout the manuscript (see A.1).

Moreover, we agree with the reviewer that whether the cause of deviations from expectations are due to de novo or rare variants, or environmental factors, either possibility is interesting. For example, in either scenario, our results can highlight inaccuracy in PRS prediction of extreme trait values for certain traits, and also provides a relative measure across different traits of large effects impacting on the trait tails, irrespective of whether genetic or environmental. We now place more emphasis on this point throughout the manuscript.

Reviewer #2 (Public Review):

Souaiaia et al. attempt to use sibling phenotype data to infer aspects of genetic architecture affecting the extremes of the trait distribution. They do this by considering deviations from the expected joint distribution of siblings' phenotypes under the standard additive genetic model, which forms their null model. They ascribe excess similarity compared to the null as due to rare variants shared between siblings (which they term 'Mendelian') and excess dissimilarity as due to de-novo variants. While this is a nice idea, there can be many explanations for rejection of their null model, which clouds interpretation of Souaiaia et al.'s empirical results.

A.3: We agree with the reviewer that we should have made clearer that there are other explanations for significant results from our tests and we have now fully addressed this point – (see A.1, A.2, A.4, A.5 for more detail).  In addition, we now elaborate on exactly what our null hypothesis is: which is not only that the expected joint distribution of siblings’ phenotypes is governed by the standard additive genetic model, but that environmental effects are either controlled for or else their combined effect is approximately Gaussian. Furthermore, by selecting only those traits whose raw trait distribution most closely corresponds to a Gaussian distribution from the UK Biobank, we increase the probability that significant results from our tests are due to rare variants (shared or unshared among siblings).

The authors present their method as detecting aspects of genetic architecture affecting the extremes of the trait distribution. However, I think it would be better to characterize the method as detecting whether siblings are more or less likely to be aggregated in the extremes of the phenotype distribution than would be predicted under a common variant, additive genetic model.

A.4: As discussed above we should have stated more clearly that significant results could be due to non-genetic factors, we have now addressed this.

However, we do not think that it would be appropriate to characterise our tests as merely corresponding to over and under aggregation of siblings in the tails. Firstly, environmental factors should be controlled for as part of our testing, increasing the probability that significant results are due to genetic, and not environmental factors. Secondly, tests for identifying broad over and under aggregation of siblings in the tails should be designed differently and, accordingly, the tests that we have developed here would not be optimal to detect over/under aggregation of siblings in trait tails. Our test for inference of de novo variants, for example, exploits the fact that de novo alleles of large effect result in one sibling being extreme and all others being drawn from the background distribution, so that the mean of other siblings is relatively low – not merely that other siblings are less likely to be found in the tail. For more discussion on this issue in relation to one of reviewer 1’s points, see A.9.

Exactly how the rareness and penetrance of a genetic variant influence the conditional sibling phenotype distribution at the extremes is not made clear. The contrast between de-novo and 'Mendelian' architectures is somewhat odd since these are highly related phenomena: a 'Mendelian' architecture could be due to a de-novo variant of the previous generation. The fact that these two phenomena are surmised to give opposing signatures in the authors' statistical tests seems suboptimal to me: would it not be better to specify a parameter that characterizes the degree or sharing between siblings of rare factors of large effect? This could be related to the mixture components in the bimodal distribution displayed in Fig 1. In fact, won't the extremes of all phenotypes be influenced by all three types of variants (common, rare, de-novo) to greater or lesser degree? By framing the problem as a hypothesis testing problem, I think the authors are obscuring the fact that the extremes of real phenotypes likely reflect a mixture of causes: common, de-novo, and rare variants (and shared and non-shared environmental factors). 

A.5: We absolutely recognise that there will typically be a complex and continuous mix of genetic architectures underlying complex traits in their tails, dictated by the 2-dimensional relationship between allele frequency and effect size. We did consider developing a fully Bayesian statistical framework to model this, but soon realised that doing this properly would require a substantial amount of model development, accounting for multiple factors in ways that would require a great deal of further investigation; for example, performing a range of complex simulations to investigate the effects of different selective pressures over time, different patterns of assortative mating, and effect size generating distributions. We are in the process of applying for funding for a multi-year project that will perform exactly these investigations as a step towards developing more sophisticated models of inference. In the meantime, we do believe that the simpler hypothesis-testing framework that we have developed here does have important value. Assuming that environmental factors are accounted for, or that any that are not accounted for have combined Gaussian effects, then our tests will indeed infer enrichments of de novo and ‘Mendelian’ rare alleles of large effect in the tails of complex traits. Results from these tests can also be compared within and across traits to compare the relative degree of such enrichments among traits. For some traits we observe significant results from both tests, and for other traits we observe highly significant results from one of our tests but not the other. Thus, while our tests do not provide a complete picture about the genetic architecture in the tails of complex traits, they do offer some intriguing initial insights into tail architecture, important given the enrichment of disease in trait tails.

To better enable interpretation of the results of this method, a more comprehensive set of simulations is needed. Factors that may influence the conditional distribution of siblings' phenotypes beyond those considered include: non-normal distribution, assortative mating, shared environment, interactions between genetic and shared environmental factors, and genetic interactions. 

A.6: As described above (see A.5) we do agree that a more comprehensive set of simulations is exactly what is needed to further extend this work. However, we believe that the tests that we have developed so far, which make some simplifying assumptions that we think would often hold in practice, is a useful start to what is an entirely novel approach to inferring genetic architecture from family trait-only (non-genetic) data. Our work could already be useful for method developers who may wish to extend our approach in ways that we may not think of. It could also be useful for applied scientists focusing on specific traits who will be able to gain initial, inference-level, insights by applying our tests to their data, while the results of applying our tests may even guide study design of rare variant mapping studies.

In summary, I think this is a promising method that is revealing something interesting about extreme values of phenotypes. Determining exactly what is being revealed is going to take a lot more work, however. 

A.7: We thank the reviewer for highlighting the promise in our approach and agree that it is revealing something interesting about complex traits. We also agree that it is going to take a lot more work to reveal exactly what that is for different traits, which we plan to work on ourselves and hope that this paper will help other interested scientists to follow-up on and extend as well.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

R.1.1: Why these particular traits (body fat, mean corpuscular haemoglobin, neuroticism, heel bone mineral density, monocyte count, sitting height)? 

A.8: Traits were initially selected to cover a variety of traits (anthropometric, metabolic, personality..) and to illustrate different examples of tail architecture. However, in response to a point from reviewer 2 (see A.17), we have now overhauled our quality control of traits to ensure that only traits closely matching Gaussian distributions are included. In total, 18 traits were selected, with detailed results presented in Appendix 4 and results corresponding to 6 of the traits presented in the main text (Figure 6) to show examples of different types of tail architecture.

R.1.2: Why are there separate tests for de novo and Mendelian architectures? It seems that one could use either of the derived tests for both purposes, simply by switching to a two-sided test for each tail. My guess is that the score test of whether alpha is zero would be the more statistically powerful test. 

A.9: The score test of whether alpha is zero has limited power to detect Mendelian architectures. This is because under Mendelian effects, half the siblings in a family have trait values reflecting the background distribution, such that the mean of sibling trait values is not so different from the polygenic expectation (i.e. alpha close to 0). The Mendelian score test that we developed is substantially more powerful because it evaluates co-occurrence of siblings in the tails, which is far higher under Mendelian architecture in the tail than compared to polygenic architecture.

However, in order test for general departures from our null model, including those of non-Gaussian environmental factors, we now include results from performing a Kolmogorov-Smirnoff test of difference from the expected distribution, and also provide this test as an option in our ‘SibArc’ software tool.

R.1.3: This method assumes that assortative mating is absent. I worry that sitting height might not be a good trait to analyze, since there is some assortative mating (~0.3) for height (e.g., Yengo et al., 2018). Perhaps this trait should not be included among those that are analyzed in this paper. Then again, it is possible that there is less assortative mating for sitting height than total height (i.e., leg length) (Jensen & Sinha, 1993). 

A.10:  It is true that our method assumes random mating. We note that while  assortative mating increases sibling similarity relative to expectation, if it is stable across the trait distribution it will also bias heritability estimation upward which is likely it’s potential impact in our framework.  However, if assortative mating is more prevalent in the tails of the distribution, it can result in excess kurtosis – an impact that can increase false positive Mendelian tests and false negative de novo tests.  Given that the trait distribution for Sitting Height has only moderate excess Kurtosis (~0.4, see Fig 9, Appendix 4) and we inferred de novo architecture only for this trait, we feel that including it in the paper is appropriate. 

R.1.4: I wonder if it's possible to discuss the impact of non-additive genetic variance on the method. How does this affect the estimation of heritability, which calibrates the expectation for regression to the mean? Can non-additive genetic deviations explain a rejection of the null hypothesis of simple polygenicity? 

A.11: Yes, the heritability estimation, which calibrates expectation for regression to the mean, assumes additivity of effects, as do the most popular estimators of heritability from GWAS data in the field: GCTA-GREML, LD Score regression and LDAK. Accordingly, non-additive genetic effects could result in rejection of the null hypothesis. We have highlighted this point in the Discussion. However, we also point out that current evidence suggests that the contribution of non-additive genetic effects to complex trait variation is relatively small (Hivert 2021) and that non-additive genetic effects that have a similar impact across the trait distribution should not be a problem for our approach (only those that have an increasing effect towards the tails would be).

R.1.5: p.5: Maybe a more realistic way to simulate a genetic architecture is to draw the MAF from the distribution [MAF(1 - MAF)]^{-1} and then an effect of the minor allele from some mound-shaped distribution (e.g., mixture of normals). The absolute or squared effect of the minor allele should increases as the MAF decreases, and there have been some papers trying to estimate this relationship (e.g., Zeng et al., 2021). Maybe make the number of causal SNPs 10,000. I don't rate this as an urgent suggestion because my sense is that the method should be robust, making adequate even a fairly minimal simulation confirming its accuracy. 

A.11: In separate work, we have performed a comprehensive simulation study using the forward-in-time population genetic simulator SLIM-3 (Haller and Messer, 2019), which generates genetic effects according to Gaussian and Gamma distributions and models different selective pressures on complex traits. We plan to publish this work shortly and also extend the simulations to family data, from which we will be able to test the performance of our methods here under a range of different scenarios of genetic variation generation, including a variety of relationships between allele frequency and effect sizes. We agree with the reviewer that at this point, however, our minimal simulation should be sufficient to confirm our tests’ general robustness and so we will perform further testing once we have extended our more sophisticated simulation study.

R.1.6: p.6: Step D seems to leave out a normalization of G to have unit variance. Also, the last part should say "the square of the correlation between the genetic liability and the trait is equal to the heritability." 

A.12: Corrected – we thank the reviewer for spotting this.

R.1.7: Figure 5: The power being adequate if roughly 1 of a 1000 index siblings with an extreme trait value owes their values to de novo mutations makes me think that there should be a discussion of the prior probability. The average person carries about 80 de novo mutations. How many of these are likely to affect, e.g., height? Zeng et al. (2021) gave estimates of mutational targets. Given that a mutation affects height, will its likely effect size be large enough to be detected with the method? Kemper et al. (2012) discussed this point in a perhaps useful way. 

A.13: We find the work investigating mutational target sizes and generating effect sizes of different mutations (de novo or rare) to be extremely interesting and critical for understanding the causes of observed genetic variation. However, we think that this work is insufficiently progressed at this point to build on directly here for making more nuanced interpretation of our results. We are, however, exploring the impact of mutational target sizes, effect size distributions and selection effects, on the genetic architecture of complex traits via population genetic simulations (see A.11), and so we hope to be able to provide more in-depth interpretation of our results in the future.

R.1.8: Figure 6: The number in the tables for Mendelian architecture are presumably observed and expected counts. But what about the numbers for de novo architecture? Those don't look like counts. Maybe they are conditional expectations of standardized trait values. Whatever the case may be, the caption should provide an explanation. 

A.14: The observed and expected values for the de novo statistical test represent the expected and observed mean standardized trait values for siblings of individuals in the bottom and top 1% of the distribution. We have now made this clear in our updated figure.

R.1.9: p. 16: Element (2,1) in the precision matrix after Equation 15 is missing a negative sign. 

A.15: Corrected – we thank the reviewer for spotting this.

R.1.10: p. 20: Shouldn't Equation 20 place an exponent of n on the factor outside of the exponential? 

A.16: Corrected – we thank the reviewer for spotting this.

Reviewer #2 (Recommendations For The Authors):

R.2.1: The first concern that I have is that their statistical tests rely heavily on an assumption of bivariate normal distribution for sibling pair's phenotypes. Real phenotypes do not have such a distribution in general. The authors rely upon an inverse-normal transform when applying their method to real data. While the inverse-normal transform will ensure that the siblings' phenotypes have a marginal normal distribution, such a transform does not ensure that the joint distribution is bivariate normal. The authors should examine their procedure for simulated phenotypes with a non-normal distribution to see if their statistical tests remain properly calibrated. Related to this, I am concerned about applying an inverse normal transform to the neuroticism phenotype that contains only 13 unique values in UKB. How does the transform deal with tied values? Can we sensibly talk about extreme trait values for such a set of observations? 

A.17: The reviewer is correct that a bivariate normal distribution for sibling pairs’ trait values does not necessarily hold, and only does so if the assumptions of our null model are met (polygenic effects, Gaussian environmental effects, random mating..). We have now more clearly described the assumptions of our null model, and to increase the matching of our selected traits to those assumptions we have expanded our analyses and now present results on traits that are close to Gaussian. As part of this more strict quality control, only traits with more than 50 unique values are included, meaning that neuroticism is excluded in our final analysis. We also now note that performing an inverse normal transformation on the traits only increases the robustness of the tests to some of our modelling assumptions. In future work we plan to investigate how best to model the conditional sibling distribution under a variety of non-Gaussian environmental effects and different non-random patterns of mating.

R.2.2: The joint sibling phenotype distribution (Equation 4) can be derived by applying the formula for the conditional distribution of a multivariate Gaussian to the standard additive genetic model. The authors' derivation is unnecessarily complex. Furthermore, many of the formulae have been used in Shai Carmi's work on embryo screening, but this work is not cited. 

A.18: We now state in the text that the conditional sibling distribution can also be derived from the joint trait distribution of related individuals, which we use in our extension to the 3-sibling scenario, and cite Shai Carmi’s work where this is used. The joint distribution is a more straightforward way to derive the conditional sibling distribution, but our derivation based on considering mid-parents is generalisable to cases where assumptions of random mating, Gaussian population trait distribution and no selection do not hold. We also think that our mid-parent based derivation will be more intuitive to many readers, leading to greater understanding and potential for extension. Therefore, overall we believe that its presentation is worthwhile and we have now elaborated on this in the Methods.

R.2.3: Equation 8: this probability should be conditional on s1 

A.19: Corrected – we thank the reviewer for spotting this.

R.2.4: The empirical application to UKB data is lacking methodological details. Also, the number of siblings used is low compared to the number of available sibling pairs. Around 19k sibling pairs are available in the UKB white British subsample, but only 10k were used for height. Why? Also, why are extreme values excluded? Isn't this removing the signal the authors are looking to explain?

A.20: We have now provided more methodological details throughout the Methods section, in particular in relation to the samples used and quality control performed. The removal of individuals with extreme values, in particular, is because unusually low/high trait values are more likely to be due to measurement error (e.g. due to imperfect measuring device, or storage/assaying) than for typical values, and so while this may also result in some loss in power (albeit small due to few individuals having values +/- 8 s.d. trait means) we consider it worth it for the potential reduction in type I error. In performing our newly expanded analysis (described above), and accounting for the reviewer’s point here about sample size, we did find a bug in our pipeline that meant that we did not include as many sibling pairs as available. We thank the reviewer for spotting this, since this contributed to our new analysis being substantially more powerful than the original (including up to ~17k sibling pairs depending on completeness of trait data).

Benjamin C Haller, Phillip W Messer. SLiM 3: Forward Genetic Simulations Beyond the Wright–Fisher Model. Molecular Biology and Evolution. 2019. 36(3): 632-637.

SD Whiteman, SM McHale, A Soli. Theoretical Perspectives on Sibling Relationships. J Fam Theory Rev. 2011 Jun 1;3(2):124-139.

Nicholas H Barton, Alison M Etheridge, and Amandine Véber. The infinitesimal model: Definition, derivation, and implications. Theoretical population biology, 118:50–73, 2017.

Valentin Hivert et al. “Estimation of non-additive genetic variance in human complex traits from a large sample of unrelated individuals.” American journal of human genetics vol. 108,5 (2021)

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors demonstrate that it is possible to carry out eQTL experiments for the model eukaryote S. cerevisiae, in "one pot" preparations, by using single-cell sequencing technologies to simultaneously genotype and measure expression. This is a very appealing approach for investigators studying genetic variation in single-celled and other microbial systems, and will likely inspire similar approaches in non-microbial systems where comparable cell mixtures of genetically heterogeneous individuals could be achieved.

Strengths:

While eQTL experiments have been done for nearly two decades (the corresponding author's lab are pioneers in this field), this single-cell approach creates the possibility for new insights about cell biology that would be extremely challenging to infer using bulk sequencing approaches. The major motivating application shown here is to discover cell occupancy QTL, i.e. loci where genetic variation contributes to differences in the relative occupancy of different cell cycle stages. The authors dissect and validate one such cell cycle occupancy QTL, involving the gene GPA1, a G-protein subunit that plays a role in regulating the mating response MAPK pathway. They show that variation at GPA1 is associated with proportional differences in the fraction of cells in the G1 stage of the cell cycle. Furthermore, they show that this bias is associated with differences in mating efficiency.

Weaknesses:

While the experimental validation of the role of GPA1 variation is well done, the novel cell cycle occupancy QTL aspect of the study is somewhat underexploited. The cell occupancy QTLs that are mentioned all involve loci that the authors have identified in prior studies that involved the same yeast crosses used here. It would be interesting to know what new insights, besides the "usual suspects", the analysis reveals. For example, in Cross B there is another large effect cell occupancy QTL on Chr XI that affects the G1/S stage. What candidate genes and alleles are at this locus? And since cell cycle stages are not biologically independent (a delay in G1, could have a knock-on effect on the frequency of cells with that genotype in G1/S), it would seem important to consider the set of QTLs in concert.

We thank the reviewer for this suggested clarification. We have modified the text to make it clear that cell cycle occupancy is a compositional phenotype. Like the reviewer, we also noticed the distal trans eQTL hotspot on Chr XI in Cross B, but we were not able to identify compelling candidate gene(s) or variant(s) despite extensive effort.

Reviewer #2 (Public Review):

Boocock and colleagues present an approach whereby eQTL analysis can be carried out by scRNA-Seq alone, in a one-pot-shot experiment, due to genotypes being able to be inferred from SNPs identified in RNA-Seq reads. This approach obviates the need to isolate individual spores, genotype them separately by low-coverage sequencing, and then perform RNA-Seq on each spore separately. This is a substantial advance and opens up the possibility to straightforwardly identify eQTLs over many conditions in a cost-efficient manner. Overall, I found the paper to be well-written and well-motivated, and have no issues with either the methodological/analytical approach (though eQTL analysis is not my expertise), or with the manuscript's conclusions.

I do have several questions/comments.

393 segregant experiment:

For the experiment with the 393 previously genotyped segregants, did the authors examine whether averaging the expression by genotype for single cells gave expression profiles similar to the bulk RNA-Seq data generated from those genotypes? Also, is it possible (and maybe not, due to the asynchronous nature of the cell culture) to use the expression data to aid in genotyping for those cells whose genotypes are ambiguous? I presume it might be if one has a sufficient number of cells for each genotype, though, for the subsequent one-pot experiments, this is a moot point.

As mentioned in our preliminary response, while it is possible to expand the analysis along these lines, this is not relevant for the subsequent one-pot experiments. We have made all the data available so that anyone interested can try these analyses.

Figure 1B:

Is UMAP necessary to observe an ellipse/circle - I wouldn't be surprised if a simple PCA would have sufficed, and given the current discussion about whether UMAP is ever appropriate for interpreting scRNA-Seq (or ancestry) data, it seems the PCA would be a preferable approach. I would expect that the periodic elements are contained in 2 of the first 3 principal components. Also, it would be nice if there were a supplementary figure similar to Figure 4 of Macosko et al (PMID 26000488) to indeed show the cell cycle dependent expression.

We have added two new figures (S2 and S3) that represent alternative visualizations of the cell-cycle that are not dependent on UMAP. Figure S2 shows plots of different pairs of principal components, with each cell colored by its assigned cell-cycle stage. We do not observe a periodic pattern in the first 3 principal components as the reviewer expected, but when we explore the first 6 principal components, we see combinations of components that clearly separate the cell cycle clusters. We emphasize that the clusters were generated using the Louvain algorithm and assigned to cell-cycle stages using marker genes, and that UMAP was used only for visualization.

We could not create a figure similar to Macosko et al. because of differences between the cell cycle categories we used and those of Spellman et al (PMID 9843569). We instead created Figure S3 to address the reviewer's comment. This figure uses a heatmap in a style similar to that of Macosko et al. to display cell-cycle-dependent expression of the 22 genes we used as cell cycle markers across each of the five cell cycle stages (M/G1, G1, G1/S, S, G2/M).

We have renumbered the supplementary figures after incorporating these two additional supplementary figures into the manuscript.

Aging, growth rate, and bet-hedging:

The mention of bet-hedging reminded me of Levy et al (PMID 22589700), where they saw that Tsl1 expression changed as cells aged and that this impacted a cell's ability to survive heat stress. This bet-hedging strategy meant that the older, slower-growing cells were more likely to survive, so I wondered a couple of things. It is possible from single-cell data to identify either an aging, or a growth rate signature? A number of papers from David Botstein's group culminated in a paper that showed that they could use a gene expression signature to predict instantaneous growth rate (PMID 19119411) and I wondered if a) this is possible from single-cell data, and b) whether in the slower growing cells, they see markers of aging, whether these two signatures might impact the ability to detect eQTLs, and if they are detected, whether they could in some way be accounted for to improve detection.

As mentioned in our preliminary response, we are not sure how to look for gene expression signatures of aging in yeast scRNA-seq data. We believe that the proposed analyses are beyond the scope of the current paper. As noted above, we have made all the data available so that anyone interested can explore these hypotheses.

AIL vs. F2 segregants:

I'm curious if the authors have given thought to the trade-offs of developing advanced intercross lines for scRNA-Seq eQTL analysis. My impression is that AIL provides better mapping resolution, but at the expense of having to generate the lines. It might be useful to see some discussion on that.

We thank the reviewer for the comments. We believe that a discussion of trade-offs between different approaches for constructing mapping populations, such as AIL and F2 segregants, is beyond the scope of this paper.

10x vs SPLit-Seq

10x is a well established, but fairly expensive approach for scRNA-Seq - I wondered how the cost of the 10x approach compares to the previously used approach of genotyping segregants and performing bulk RNA-Seq, and how those costs would change if one used SPLiT-Seq (see PMID 38282330).

We thank the reviewer for the comments. We believe that a discussion of cost trade-offs between 10x and other approaches is beyond the scope of this paper, especially given the rapidly evolving costs of different technologies.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Throughout the results section the authors point to File S1 for additional information. This file is a tarball with about 20 Excel documents in it, each with several sheets embedded. The authors should provide a detailed README describing how to understand the organizations of the files in File S1 and the many embedded sheets in each file. Statements made in the manuscript about File S1 should explicitly direct the reader to a specific spreadsheet and table to refer to.

We have added an additional README file to the tarball that explains the organization of File S1 and describes the data contained in each sheet. Throughout the text, we now reference specific spreadsheets to assist the reader. In addition, these spreadsheets have been added to a github repository https://github.com/theboocock/finemapping_spreadsheets_single_cell

Neither of the two GitHub repositories referenced under "Code availability" has adequate documentation that would allow a reader to try and reproduce the analyses presented here. The one entitled https://github.com/joshsbloom/single_cell_eQTL has no functional README, while https://github.com/theboocock/yeast_single_cell_post_analysis is somewhat better but still hard to navigate. Basic information on expected inputs, file formats, file organization, output types, and formats, etc. is required to get any of these pipelines to run and should be provided at a minimum.

We thank the reviewer for the comment. In response, we have refactored both GitHub repositories and added extensive documentation to improve usability. We updated the versions of software and packages, this has been reflected in the methods section.

S. cerevisiae strains are preferentially diploid in nature and many genes involved in the mating pathway are differentially regulated in diploids vs haploids. Have the authors explored the fitness effects of the GPA1 82R allele in diploids? What is the dominance relationship between 82W and 82R?

We thank the reviewer for the comment. In diploid yeast, the mating pathway is repressed, and thus we would not expect there to be any fitness consequences due to the presence of different alleles of GPA1.

The diploid expression profiling (page 5 and Table S9) doesn't implicate GPA1; can you the authors comment on this in light of their finding in haploids?

The mating pathway, including GPA1, is repressed in diploids, and hence the expression of GPA1 cannot be studied in these strains (PMID: 3113739). In addition, allele-specific expression differences only identify cis-regulatory effects. We know that the GPA1 variant results in a protein-coding change, which may or may not influence the levels of mRNA in cis, so that even if GPA1 were expressed in diploids, there would be no expectation of an allele-specific difference in expression.

With respect to the candidate CYR1 QTL -- note that strains with compromised Cyr1 function also generally show increased sporulation rates and/or sporulation in rich media conditions (cAMP-PKA signaling represses sporulation). Is this the case in diploids with the CBS2888 allele at CYR1? If the CBS2888 allele is a CYR1 defect one might expect reduced cAMP levels. It is possible to estimate adenylate cyclase levels using a fairly straightforward ELISA assay. This would provide more convincing evidence of the causal mechanism of the alleles identified.

We thank the reviewer for the comment, and we agree that a functional study of the CYR1 alleles would provide more convincing evidence for the causal mechanism of the connection between cell cycle occupancy, cAMP levels, and growth. However, we believe that the proposed experiments are beyond the scope of our current study. The evidence we provide is sufficient to establish that CYR1 is a strong candidate gene for the eQTL hotspot.

Re: CYR1 candidate QTL -- The authors should reference the work of [Patrick Van Dijck] (https://pubmed.ncbi.nlm.nih.gov/?sort=date&term=Van+Dijck+P&cauthor_id= 20924200) and [Johan M Thevelein] (https://pubmed.ncbi.nlm.nih.gov/?sort=date&term=Thevelein+JM&cauth or_id=20924200) on CYR1 allelic variation, and other papers besides the Matsumoto/ Ishikawa papers, as the effects of cAMP-PKA signaling on stress can be quite variable. cAMP pathway variants, including in CYR1, have popped up in quite a few other yeast QTL mapping and experimental evolution papers. These should be referenced as well.

We thank the reviewer for these references; we have added a comment about the relationship between stress tolerance and CYR1 variation, and cited the relevant references accordingly.

Figure S10 - the subfigure showing the frequency of the GPA 82R compared to 82W suggests a fairly large and deleterious fitness effect of this allele; on the order of 7-8% fewer cells per cell cycle stage than the 82W allele. Can the authors reconcile this with the more modest growth rate effect they report on page 8?

Figure S12C displays the allele frequency of the 82R allele across the cell cycle in the single-cell data from allele-replacement strains. These strains were grown separately and processed using two individual 10x chromium runs. The resulting sequenced library had 11,695 cells with the 82R allele and 14,894 cells with the 82W allele. The 7-8% difference in the number of cells is due to slight differences in the number of captured cells per run, not due to growth differences, because we attempted to pool cells in equal numbers from separate mid-log cultures.

The proportion of cells in G1 increases by ~3% in strains with the 82R allele relative to the baseline proportion of cells in the experiment, which, to the reviewers point, is still larger than the ~1% growth difference we observed. Cell cycle occupancy is a compositional phenotype. As shown in figure S12C, the 82R variant increases the fraction of cells in G1 and slightly decreases the fraction of cells in M/G1. There is no obvious expectation for quantitatively translating a change in cell cycle occupancy to a change in growth rate.

The authors refer to the Lang et al. 2009 paper w/respect to GPA1 variant S469I but that paper seems to have explored a different GPA1 allele, GPA1-G1406T, with respect to growth rates.

We thank the reviewer for their comment. The S469I variant is the same as the G1406T variant, one denoting the amino acid change at position 469 in the protein and the other denoting the corresponding nucleotide change at position 1406 in the DNA coding sequence. We have altered the text to make this clear to the reader.

Reviewer #2 (Recommendations For The Authors):

I make no recommendations as to additional work for the authors. The manuscript is complete. I suggested some things I would like to see in my review, but it's up to them to decide whether they think any of those would further enhance the manuscript.

However, I do have I have some pedantic formatting notes:

- Microliters are variously presented as uL, ul, and µl - it should be µL

- Similarly, milliliters are presented as ml and ML - it should be mL

- Also, there should be a space between the number and the unit, e.g. 10 µL

- Some gene names in the manuscript are not italicized in all instances, e.g., GPA1

We thank the reviewer for these formatting suggestions, we have made these changes throughout the text.

Author response:

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

Response to Public Reviews:

We thank the reviewers for their kind comments have implemented many of the suggestion their suggestions. Our paper has greatly benefited from their advice.  Like Reviewer 1, we acknowledge that while the exact involvement of Ih in allowing smooth transitions is likely not universal across all systems, our demonstration of the ways in which such currents can affect the dynamics of the response of complex rhythmic motor networks provides valuable insight. To address the concerns of Reviewer 2, we included a sentence in the discussion to highlight the fact that cesium neither increased the pyloric frequency nor caused consistent depolarization in intracellular recordings. We also highlighted that these observations suggest both that cesium is not indirectly raising [K+]outside and support the conclusion that the effects of cesium are primarily through blockade of Ih rather than other potassium channels.

Reviewer 3 raised some important points about modeling. While the lab has models that explore the effects of temperature on artificial triphasic rhythms, these models do not account for all the biophysical nuances of the full biological system. We have limited data about the exact nature of temperature-induced parameter changes and the extent to which these changes are mediated by intrinsic effects of temperature on protein structure versus protein interactions/modification by processes such as phosphorylation. With respects to the A current, Tang et al., 2010 reported that the activation and inactivation rates are differentially temperature sensitive but we do not have the data to suggest whether or not the time courses of such sensitivities are different. As such, we focus our discussion on the properties we know are modulated by temperature, i.e. activation rates. Within the discussion we now include the suggestion that future, more comprehensive modeling may be appropriate to further elucidate the ways in which reducing Ih may produce the here reported experimentally observed effects.

Reviewer #1 (Recommendations For The Authors):

Suggested revisions:

A figure showing examples of the voltage-clamp traces for the critical measurements of the extent of Ih block by 5 mM CsCl in PD and LP neurons at the temperature extremes in these preparations is not shown, and the authors should consider including such a figure, perhaps as a supplemental figure.

We have added Supplemental Figure 1 containing voltage-clamp traces demonstrating the extent of Ih block by 5mM CsCl in PD and LP neurons at 11 and 21°C.  Due to technical concerns, different preparations were used in the measurements at 11°C and 21°C, but the point that the H-current is reduced is demonstrated in all cases.

Reviewer #2 (Recommendations for The Authors):

Specific (Minor) Comments:

(1) Line 83: In Cs+ "at 11°C, the pyloric frequency was significantly decreased compared to control conditions (Saline: 1.2± 0.2 Hz; Cs+ 0.9± 0.2 Hz)".

As above, the authors often report that cesium generally reduces pyloric frequency. Figure 5A demonstrates this action quite nicely. However, cesium's effect on pyloric frequency at 11°C seems less robust in Figure 1C. Why the discrepancy?

There is variability in the effects of Cs+ on the pyloric frequency.  As noted, the standard deviation in frequency in both conditions is 0.2Hz.  As such, there are some cases in which the initial frequency drop in Cs+ compared to control was relatively small.  1C is one such case, but was selected as an example because of its clear reduction in temperature sensitivity. 

(2) I don't understand what the arrows/dashed lines are trying to convey in Figure 3C.

The arrows/dashed lines represent the criteria used to define a cycle as “decreasing in frequency” (Temperature Increasing) or “increasing in frequency” (Temperature Stable).  We have amended lines 130 and 137 in the text to hopefully clarify this point, as well as the figure legend.

(3) Lines 118/168. The description of cesium's specific action on the depolarizing portion of PD activity is a bit confusing. In my mind, "depolarization phase" refers to the point at which PD is most depolarized. Perhaps restating the phrase to "elongation of the depolarizing trajectory" is less confusing. The authors may also want to consider labeling this trajectory in Figure 2C.

We have changed “depolarization phase” to “depolarizing phase” to highlight that this is the period during which the cell is depolarizing, rather than at its most depolarized.  We consider the plateau of the slow wave and spiking (the point at which PD is most depolarized) to be the “bursting phase”.  We have labeled these phases in Figure 2C as suggested.

(4) Figure 3C legend: a few words seem to be missing. I suggest "the change in mean frequency was more likely TO decrease IN Cs+ than in saline".

Thank you for catching this typo, it has been corrected.

(5) Line 165: Awkward phrasing. “In one experiment, the decrease in frequency while temperature increased and subsequent increase in frequency after temperature stabilized was particularly apparent in Cs+ PTX”.

How about: “One Cs+ PTX experiment wherein elevating the temperature transiently decreased pyloric frequency is shown in Figure 4F.”

We have amended this sentence to read, “One Cs++PTX experiment in which elevating the temperature produced a particularly pronounced transient decrease in frequency is shown in Figure 4F.”

(6) Line 186: Awkward phrasing. "LP OFF was also significantly advanced in Cs+, although duty cycle (percent of the period a neuron is firing) was preserved".

The use of the word "although" seems a bit strange. If both LP onset and LP offset phase advance by the same amount, then isn't an unchanged duty cycle expected?

“Although” has been changed to “and subsequently”.

Reviewer #3 (Recommendations For The Authors):

Major comments:

(1) I know the Marder lab has detailed models of the pyloric rhythm. I am not saying they have to add modeling to this already extensive and detailed paper, but it would be useful to know how much of these temperature effects have been modeled, for example in the following locations.

(2) Line 259 - "Mathematically..." - Is there a computational model of H current that has shown this decrease in frequency in pyloric neurons? If you are working on one for the future, you could mention this.

There is not currently a model in which the reduction of the H-current results in the non-minimum phase dynamics in the frequency response to temperature seen experimentally. It should be noted that our existing models of pyloric activity responses to temperature are not well suited to investigate such dynamics in their current iterations.  Further work is necessary to demonstrate the principles observed experimentally in computational modeling, and we have added a sentence to the paper to reflect this point (Line 268).

(3) Line 318 - "therefore it remains unclear" - I thought they had models of the circuit rhythmicity. Do these models include temperature effects? Can they comment on whether their models of the circuit show an opposite effect to what they see in the experiment? I'm not saying they have to model these new effects as that is probably an entirely different paper, but it would be interesting to know whether current models show a different effect.

We have some models of the pyloric response to temperature, but these models were specifically selected to maintain phase across the range of temperature.  When Ih was reduced in these models, a variety of effects on phase and duty cycle were seen.  These models were selected to have the same key features of behavior as the pyloric rhythm, but do not capture all the biophysical nuances of the complete system, and therefore should not necessarily be expected to reflect the experimental findings in their current iterations.  Furthermore, these models are meant to have temperature as a static, rather than dynamic input, and thus are ill-suited to examine the conditions of our experiments.  The models in their current state are not sufficiently relevant to these experimental findings that we they can illuminate the present paper `2.

(4) "If deinactivation is more accelerated or altered by temperature than inactivation...While temperature continued to change, the difference in parameters would continue to grow" - This is described as a difference in temperature sensitivity, but it seems like it is also a function of the time course of the response to change in temperature (i.e. the different components could have the same final effect of temperature but show a different time course of the change).

We know from Tang et al, 2010, that activation and inactivation rates of the A current are differentially temperature sensitive. We have no evidence to suggest that the time course of the response to temperature of various parameters differ.  The physical actions of temperature on proteins are likely to be extremely rapid, making a time course difference on the order of tens of seconds less unlikely, though not impossible. Modeling of the biophysics might illuminate the relative plausibility of these different mechanisms of action, but we feel that our current suggested explanation is reasonable based on existing information.

(5) Is it known how temperature is altering these channel kinetics? Is it via an intrinsic rearrangement of the protein structure, or is it a process that involves phosphorylation (that could explain differences in time course?). Some mention of the mechanism of temperature changes would be useful to readers outside this field.

It is not known exactly how temperature alters channel parameters.  Invariably some, if not all, of it is due to an intrinsic rearrangement of protein structure, and our current models treat all parameter changes as an instantaneous consequence.  However, it is possible that some effects of temperature are due to longer timescale processes such as phosphorylation or cAMP interactions.  Current work in the lab is actively exploring these questions, but there is no definitive answer. Given that this paper focuses on the phenomenon and plausible biomolecular explanations based on existing data, we have not altered the paper to include more exhaustive  coverage of all the possible avenues by which temperature may alter channel properties.

Specific comments:

Title: misspelling of "Cancer" ?

We are unsure how that extra “w” got into the earliest version of the manuscript and have removed it.

Line 66 "We used 5mM CsCl" - might mention right up front that this was a bath application of the substance.

We have altered this line to read “used bath application of 5mM CsCl”.  

Figure 4 - "The only feedback synapse to the pacemaker kernel neurons, LP to PD, and is blocked by picrotoxin" - I think the word "and" should be removed from this phrase in the figure legend.

Fixed

Figure 4 legend - "Reds denote temperature...yellows denote..." - I think it should be "Red dots denote temperature...yellow dots denote...".

Done

Figure 4B - Why does the change in frequency in cesium look so different in Figure 4B compared to Figure 1C or Figure 3B? In the earlier figures, the increase of frequency is smaller but still present in cesium, whereas, in Figure 4B, cesium seems to completely block the increase in frequency. I'm not sure why this is different, but I guess it's because 3B and 4B are just mean traces from single experiments. Presumably, 4B is showing an experiment in which the cesium was subsequently combined with picrotoxin?

Figures 1C, 3B, and 4B are indeed all from different single experiments. As acknowledged in our concluding paragraph, there was substantial variability in the exact response of the pyloric rhythm to temperature while in cesium.  The most consistent effect was that the difference in frequency between cesium and saline at a particular temperature increased, as demonstrated across 21 preparations in Figure 1D. It may be noted in Figure 1E that the Q10 was not infrequently <1, meaning that there was a net decrease in frequency as temperature increased in some experiments such as seen in the example of Figure 4B.  The “fold over” (initial increase in steady-state frequency with temperature, then decrease at higher temperatures) has been observed at higher temperatures (typically around 23-30 degrees C) even under control conditions but has not been highlighted in previous publications.  The example in 4B was chosen because it demonstrated both the similarity in jags between Cs+ and Cs++PTX and an overall decrease in temperature sensitivity, even though in this instance the steady-state change in frequency with temperature was not monotonic. 

Figure 6A - "Phase 0 to 1.0" - The y-axis should provide units of phase. Presumably, these are units of radians so 1.0=2*pi radians (or 360 degrees, but probably best to avoid using degrees of phase due to confusion with degrees of temperature).

Phase, with respect to pyloric rhythm cycles, does not traditionally have units as it is a proportion rather than an angle. As such, we have not changed the figure.

Line 275 - "the pacemaker neuron can increase" - Does this indicate that the main effects of H current are in the follower neurons (i.e. LP and PY versus the driver neuron PD)?

Not necessarily.  We posit in the next paragraph that the effect of the H current on the temperature sensitivity could be due to its phase advance of LP, but that phase advance of LP is not particularly expected to increase frequency.  We favor the possibility that temperature increases Ih in the pacemaker, which in turn advances the PRC of the rhythm, allowing the frequency increase seen under normal conditions.  In Cs+, this advance does not occur, resulting in the lower temperature sensitivity.  In Cs++PTX, the lack of inhibition from LP means compensatory advance of the pacemaker PRC by Ih is unnecessary to allow increased frequency.

Line 285 - "either increase frequency have no effect" - Is there a missing "or" in this phrase?

Thank you, we have added the “or”.

Author response:

eLife assessment

This potentially valuable study examines the role of IL17-producing Ly6G PMNs as a reservoir for Mycobacterium tuberculosis to evade host killing activated by BCG immunisation. The authors report that IL17-producing polymorphonuclear neutrophils harbour a significant bacterial load in both wild-type and IFNg-/- mice and that targeting IL17 and Cox2 improved disease outcomes whilst enhancing BCG efficacy. Although the authors suggest that targeting these pathways may improve disease outcomes in humans, the evidence as it stands is incomplete and requires additional experimentation for the study to realise its full impact.

Thank you for evaluating our manuscript. We understand the concern related to the direct role of Ly6G+Gra-derived IL17 in TB pathogenesis. For the revised manuscript, we will provide additional experimental evidence through direct regulation of IL-17 production in Mtb-infected mice and its impact on improving BCG efficacy.

Public Reviews:

Reviewer #1 (Public review):

Summary:

Recruitment of neutrophils to the lungs is known to drive susceptibility to infection with M. tuberculosis. In this study, the authors present data in support of the hypothesis that neutrophil production of the cytokine IL-17 underlies the detrimental effect of neutrophils on disease. They claim that neutrophils harbor a large fraction of Mtb during infection, and are a major source of IL-17. To explore the effects of blocking IL-17 signaling during primary infection, they use IL-17 blocking antibodies, SR221 (an inverse agonist of TH17 differentiation), and celecoxib, which they claim blocks Th17 differentiation, and observe modest improvements in bacterial burdens in both WT and IFN-γ deficient mice using the combination of IL-17 blockade with celecoxib during primary infection. Celecoxib enhances control of infection after BCG vaccination. 

Thank you for the summary.

Strengths:

The most novel finding in the paper is that treatment with celecoxib significantly enhances control of infection in BCG-vaccinated mice that have been challenged with Mtb. It was already known that NSAID treatments can improve primary infection with Mtb.

Thank you.

Weaknesses:

The major claim of the manuscript - that neutrophils produce IL-17 that is detrimental to the host - is not strongly supported by the data. Data demonstrating neutrophil production of IL17 lacks rigor. 

Our response: Neutrophil production of IL-17 is supported by two independent methods/ techniques in the current version: 

(1) Through Flow cytometry- a large fraction of Ly6G+CD11b+ cells from the lungs of Mtb-infected mice were also positive for IL-17 (Fig. 3C).

(2) IFA co-staining of Ly6G + cells with IL-17 in the lung sections from Mtb-infected mice (Fig. 3 E_G and Fig. 4H, Fig. 5I).

However, to further strengthen this observation, we plan to analyse sorted Ly6G+Gra from the lungs of infected mice using IL-17 ELISPOT assay. This will unequivocally prove the Ly6+Gra production of IL-17. Several publications support the production of IL-17 by neutrophils (Li et al. 2010; Katayama et al. 2013; Lin et al. 2011). For example, neutrophils have been identified as a source of IL-17 in human psoriatic lesions (Lin et al. 2011), in neuroinflammation induced by traumatic brain injury (Xu et al. 2023) and in several mouse models of infectious and autoimmune inflammation (Ferretti et al. 2003; Hoshino et al. 2008) (Li et al. 2010). However, ours is the first study reporting neutrophil IL-17 production during Mtb pathology.

The experiments examining the effects of inhibitors of IL-17 on the outcome of infection are very difficult to interpret. First, treatment with IL-17 inhibitors alone has no impact on bacterial burdens in the lung, either in WT or IFN-γ KO mice. This suggests that IL-17 does not play a detrimental role during infection. Modest effects are observed using the combination of IL-17 blocking drugs and celecoxib, however, the interpretation of these results mechanistically is complicated. Celecoxib is not a specific inhibitor of Th17. Indeed, it affects levels of PGE2, which is known to have numerous impacts on Mtb infection separate from any effect on IL-17 production, as well as other eicosanoids. 

The reviewer correctly says that Celecoxib is not a specific inhibitor of Th17. However, COX-2 inhibition does have an effect on IL-17 levels, and numerous reports support this observation (Paulissen et al. 2013; Napolitani et al. 2009; Lemos et al. 2009). We elaborate on the results below for better clarity.

Firstly, in the WT mice, Celecoxib treatment led to a complete loss of IL-17 production in the lungs of Mtb-infected mice (Fig. 5D). Interestingly, IL-17 production independent of IL-23 is known to require PGE2 (Paulissen et al. 2013; Polese et al. 2021). In the WT or IFNγ KO mice, we rather noted a decline in IL-23 levels post-infection, suggesting a possible role of PGE2 in IL-17 production. However, in the lung homogenates of Mtb-infected IFNγ KO mice, Celecoxib had no effect on IL-17 levels in the lung homogenates. Thus, celecoxib controls IL-17 levels only in the Mtb-infected WT mice. Including celecoxib with anti-IL17 in the IFNγ KO mice controls pathology and extends its survival.

Second, the reviewer’s observation is only partially correct that IL-17 inhibition has a modest effect on the outcome of infection. While IL-17 neutralization and inhibition alone in the IFNγ KO mice and WT mice, respectively, did not bring down the lung CFU burden significantly, in both these cases, there was an improvement in the lung pathology. The reduced pathology coincided with reduced neutrophil recruitment and a reduced Ly6G+Graresident Mtb population in the WT mice. IL-17 neutralization alone improved IFNγ KO mice survival by ~10 days (Fig. 4F-G). 

Third, regarding the SR2211 and Celecoxib combination study, we agree with the reviewer that Celecoxib has roles independent of IL-17 regulation. However, in the results presented in this study, there are three key aspects- 1) neutrophil-derived IL-17-dependent neutrophil recruitment, 2) the presence of a large proportion of intracellular Mtb in the neutrophils and 3) dissemination of Mtb to the spleen. Celecoxib treatment alone helps reduce lung Mtb burden in the WT mice. However, SR2211 fails to do so. It is evident that celecoxib is doing more than just inhibiting IL-17 production. The result shows that celecoxib blocks neutrophil recruitment (which could be an IL-17-dependent mechanism) and also controls the intraneutrophil bacterial population. Finally, either SR2211 or celecoxib could block dissemination to the spleen. The role of neutrophils in TB dissemination is only beginning to emerge (Hult et al. 2021). We will revise the description in the results and discussion section for this data to make it easier to understand.

Finally, we have also done experiments with SR2211 in BCG-vaccinated animals, which shows the direct impact of IL-17 inhibition on the BCG vaccine efficacy. We will add this result in the revised version.

Finally, the human data simply demonstrates that neutrophils and IL-17 both are higher in patients who experience relapse after treatment for TB, which is expected and does not support their specific hypothesis. 

We disagree with the above statement. Why a higher IL-17 is expected in patients who show relapse, death or failed treatment outcomes? Classically, IL-17 is believed to be protective against TB, and the reviewer also points to that in the comments below. A very limited set of studies support the non-protective/pathological role of IL-17 in tuberculosis (Cruz et al. 2010). High IL-17 and neutrophilia at the baseline in the human subjects (i.e. at the time of recruitment in the study) highlight severe pathology in those subjects, which could have contributed to the failed treatment outcome. This observation in the human cohort strongly supports the overall theme and central observation in this study.

The use of genetic ablation of IL-17 production specifically in neutrophils and/or IL-17R in mice would greatly enhance the rigor of this study. 

The reviewer’s point is well-taken. Having a genetic ablation of IL-17 production, specifically in the neutrophils, would be excellent. At present, however, we lack this resource, and therefore, it is not feasible to do this experiment within a defined timeline. Instead, for the revised manuscript, we will present the data with SR2211, a direct inhibitor of RORgt and, therefore, IL-17, in BCG-vaccinated mice.

The authors do not address the fact that numerous studies have shown that IL-17 has a protective effect in the mouse model of TB in the context of vaccination.

Yes, there are a few articles that talk about the protective effect of IL-17 in the mouse model of TB in the context of vaccination (Khader et al. 2007; Desel et al. 2011; Choi et al. 2020). This part was discussed in the original manuscript (in the Introduction section). For the revised manuscript, we will also provide results from the experiment where we blocked IL-17 production by inhibiting RORgt using SR2211 in BCG-vaccinated mice. The results clearly show IL-17 as a negative regulator of BCG-mediated protective immunity. We believe some of the reasons for the observed differences could be 1) in our study, we analysed IL-17 levels in the lung homogenates at late phases of infection, and 2) most published studies rely on ex vivo stimulation of immune cells to measure cytokine production, whereas we actually measured the cytokine levels in the lung homogenates. We will elaborate on these points in the revised version.

Finally, whether and how many times each animal experiment was repeated is unclear.

We will provide the details of the number of experiments in the revised version. Briefly, the BCG vaccination experiment (Figure 1) and BCG vaccination with Celecoxib treatment experiment (Figure 6) were performed twice and thrice, respectively. The IL-17 neutralization experiment (Figure 4) and the SR2211 treatment experiment (Figure 5) were done once. We will add another SR2211 experiment data in the revised version. 

Reviewer #2 (Public review):

Summary:

In this study, Sharma et al. demonstrated that Ly6G+ granulocytes (Gra cells) serve as the primary reservoirs for intracellular Mtb in infected wild-type mice and that excessive infiltration of these cells is associated with severe bacteremia in genetically susceptible IFNγ/- mice. Notably, neutralizing IL-17 or inhibiting COX2 reversed the excessive infiltration of Ly6G+Gra cells, mitigated the associated pathology, and improved survival in these susceptible mice. Additionally, Ly6G+Gra cells were identified as a major source of IL-17 in both wild-type and IFNγ-/- mice. Inhibition of RORγt or COX2 further reduced the intracellular bacterial burden in Ly6G+Gra cells and improved lung pathology.

Of particular interest, COX2 inhibition in wild-type mice also enhanced the efficacy of the BCG vaccine by targeting the Ly6G+Gra-resident Mtb population.

Thank you for the summary.

Strengths:

The experimental results showing improved BCG-mediated protective immunity through targeting IL-17-producing Ly6G+ cells and COX2 are compelling and will likely generate significant interest in the field. Overall, this study presents important findings, suggesting that the IL-17-COX2 axis could be a critical target for designing innovative vaccination strategies for TB.

Thank you for highlighting the overall strengths of the study.  Weaknesses:

However, I have the following concerns regarding some of the conclusions drawn from the experiments, which require additional experimental evidence to support and strengthen the overall study.

Major Concerns:

(1) Ly6G+ Granulocytes as a Source of IL-17: The authors assert that Ly6G+ granulocytes are the major source of IL-17 in wild-type and IFN-γ KO mice based on colocalization studies of Ly6G and IL-17. In Figure 3D, they report approximately 500 Ly6G+ cells expressing IL-17 in the Mtb-infected WT lung. Are these low numbers sufficient to drive inflammatory pathology? Additionally, have the authors evaluated these numbers in IFN-γ KO mice? 

Thank you for pointing out about the numbers in Fig. 3D. It was our oversight to label the axis as No. of IL17+Ly6G+Gra/lung. For this data, only a part of the lung was used. For the revised manuscript, we will provide the number of these cells at the whole lung level from Mtb-infected WT mice. Unfortunately, we did not evaluate these numbers in IFN-γ KO mice through FACS. 

For the assertion that Ly6G+Gra are the major source of IL-17 in TB, we have used two separate strategies- a) IFA and b) FACS. 

However, as described above in response to the first reviewer, for the revision, we propose to perform an IL-17 ELISpot assay on the sorted Ly6G+Gra from the lungs of Mtb-infected WT mice.

(2) Role of IL-17-Producing Ly6G Granulocytes in Pathology: The authors suggest that IL17-producing Ly6G granulocytes drive pathology in WT and IFN-γ KO mice. However, the data presented only demonstrate an association between IL-17+ Ly6G cells and disease pathology. To strengthen their conclusion, the authors should deplete neutrophils in these mice to show that IL-17 expression, and consequently the pathology, is reduced.

Thank you for this suggestion. Others have done neutrophil depletion studies in TB, and so far, the outcomes remain inconclusive. In some studies, neutrophil depletion helps the pathogen (Rankin et al. 2022; Pedrosa et al. 2000; Appelberg et al. 1995), and in others, it helps the host (Lovewell et al. 2021; Mishra et al. 2017) ). One reason for this variability is the stage of infection when neutrophil depletion was done. However, another crucial factor is the heterogeneity in the neutrophil population. There are reports that suggest neutrophil subtypes with protective versus pathological trajectories (Nwongbouwoh Muefong et al. 2022; Lyadova 2017; Hellebrekers, Vrisekoop, and Koenderman 2018; Leliefeld et al. 2018). Depleting the entire population using anti-Ly6G could impact this heterogeneity and may impact the inferences drawn. A better approach would be to characterise this heterogeneous population, efforts towards which could be part of a separate study.

For the revised manuscript, we will provide results from the SR2211 experiment in BCG-vaccinated mice and other results to show the role of IL-17-producing Ly6G+Gra in TB pathology.   

(3) IL-17 Secretion by Mtb-Infected Neutrophils: Do Mtb-infected neutrophils secrete IL-17 into the supernatants? This would serve as confirmation of neutrophil-derived IL-17. Additionally, are Ly6G+ cells producing IL-17 and serving as pathogenic agents exclusively in vivo? The authors should provide comments on this.

We have not directly measured IL-17 secretion by neutrophils in our experiments. However, Hu et al have reported IL-17 secretion by Mtb-infected neutrophils in vitro (Hu et al. 2017). Whether there are a few neutrophil roles exclusively seen under in vivo condition is an interesting proposition. We do have some observations that suggest in vitro phenotype of Mtb-infected neutrophils is different from in vivo.

(4) Characterization of IL-17-Producing Ly6G+ Granulocytes: Are the IL-17-producing Ly6G+ granulocytes a mixed population of neutrophils and eosinophils, or are they exclusively neutrophils? Sorting these cells followed by Giemsa or eosin staining could clarify this.

This is a very important point. While usually eosinophils do not express Ly6G markers in laboratory mice, under specific contexts, including infections, eosinophils can express Ly6G. Since we have not characterized these potential Ly6G+ sub-populations, that is one of the reasons we refer to the cell types as Ly6G+ granulocytes, which do not exclude Ly6G+ eosinophils. A detailed characterization of these subsets could be taken up as a separate study.

Reviewer #3 (Public review):

Summary:

The authors examine how distinct cellular environments differentially control Mtb following BCG vaccination. The key findings are that IL17-producing PMNs harbor a significant Mtb load in both wild-type and IFNg-/- mice. Targeting IL17 and Cox2 improved disease and enhanced BCG efficacy over 12 weeks and neutrophils/IL17 are associated with treatment failure in humans. The authors suggest that targeting these pathways, especially in MSMD patients may improve disease outcomes.

Thank you.

Strengths:

The experimental approach is generally sound and consists of low-dose aerosol infections with distinct readouts including cell sorting followed by CFU, histopathology, and RNA sequencing analysis. By combining genetic approaches and chemical/antibody treatments, the authors can probe these pathways effectively.

Understanding how distinct inflammatory pathways contribute to control or worsen Mtb disease is important and thus, the results will be of great interest to the Mtb field.

Thank you.

Weaknesses:

A major limitation of the current study is overlooking the role of non-hematopoietic cells in the IFNg/IL17/neutrophil response. Chimera studies from Ernst and colleagues (PMCID: PMC2807991) previously described this IDO-dependent pathway following the loss of IFNg through an increased IL17 response. This study is not cited nor discussed even though it may alter the interpretation of several experiments.

Thank you for pointing out this earlier study, which we concede we missed discussing. We disagree on the point that results from that study may alter the interpretation of several experiments in our study. On the contrary, the main observation that loss of IFNγ causes severe IL-17 levels is aligned in both studies.

IDO1 is known to alter Th cell differentiation towards Tregs and away from Th17 (Baban et al. 2009). It is absolutely feasible for the non-hematopoietic cells to regulate these events. However, that does not rule out the neutrophil production of IL-17 and the downstream pathological effect shown in this study. We will discuss and cite this study in the revised manuscript.

Several of the key findings in mice have previously been shown (albeit with less sophisticated experimentation) and human disease and neutrophils are well described - thus the real new finding is how intracellular Mtb in neutrophils are more refractory to BCGmediated control. However, given there are already high levels of Mtb in PMNs compared to other cell types, and there is a decrease in intracellular Mtb in PMNs following BCG immunization the strength of this finding is a bit limited.

The reviewer’s interpretation of the BCG-refractory Mtb population in the neutrophil is interesting. The reviewer is right that neutrophils had a higher intracellular Mtb burden, which decreased in the BCG-vaccinated animals. Thus, on that account, the reviewer rightly mentions that BCG is able to control Mtb even in neutrophils. However, BCG almost clears intracellular burden from other cell types analysed, and therefore, the remnant pool of intracellular Mtb in the lungs of BCG-vaccinated animals could be mostly those present in the neutrophils. This is a substantial novel development in the field and attracts focus towards innate immune cells for vaccine efficacy. 

References:

Appelberg, R., A. G. Castro, S. Gomes, J. Pedrosa, and M. T. Silva. 1995. 'SuscepBbility of beige mice to Mycobacterium avium: role of neutrophils', Infect Immun, 63: 3381-7.

Baban, B., P. R. Chandler, M. D. Sharma, J. Pihkala, P. A. Koni, D. H. Munn, and A. L. Mellor. 2009. 'IDO activates regulatory T cells and blocks their conversion into Th17-like T cells', J Immunol, 183: 2475-83.

Choi, H. G., K. W. Kwon, S. Choi, Y. W. Back, H. S. Park, S. M. Kang, E. Choi, S. J. Shin, and H. J. Kim. 2020. 'AnBgen-Specific IFN-gamma/IL-17-Co-Producing CD4(+) T-Cells Are the Determinants for ProtecBve Efficacy of Tuberculosis Subunit Vaccine', Vaccines (Basel), 8.

Cruz, A., A. G. Fraga, J. J. Fountain, J. Rangel-Moreno, E. Torrado, M. Saraiva, D. R. Pereira, T. D. Randall, J. Pedrosa, A. M. Cooper, and A. G. Castro. 2010. 'Pathological role of interleukin 17 in mice subjected to repeated BCG vaccination after infection with Mycobacterium tuberculosis', J Exp Med, 207: 1609-16.

Desel, C., A. Dorhoi, S. Bandermann, L. Grode, B. Eisele, and S. H. Kaufmann. 2011. 'Recombinant BCG DeltaureC hly+ induces superior protection over parental BCG by simulating a balanced combination of type 1 and type 17 cytokine responses', J Infect Dis, 204: 1573-84.

Ferreg, S., O. Bonneau, G. R. Dubois, C. E. Jones, and A. Trifilieff. 2003. 'IL-17, produced by lymphocytes and neutrophils, is necessary for lipopolysaccharide-induced airway neutrophilia: IL-15 as a possible trigger', J Immunol, 170: 2106-12.

Hellebrekers, P., N. Vrisekoop, and L. Koenderman. 2018. 'Neutrophil phenotypes in health and disease', Eur J Clin Invest, 48 Suppl 2: e12943.

Hoshino, A., T. Nagao, N. Nagi-Miura, N. Ohno, M. Yasuhara, K. Yamamoto, T. Nakayama, and K. Suzuki. 2008. 'MPO-ANCA induces IL-17 production by activated neutrophils in vitro via classical complement pathway-dependent manner', J Autoimmun, 31: 79-89.

Hu, S., W. He, X. Du, J. Yang, Q. Wen, X. P. Zhong, and L. Ma. 2017. 'IL-17 ProducBon of Neutrophils Enhances AnBbacteria Ability but Promotes ArthriBs Development During Mycobacterium tuberculosis InfecBon', EBioMedicine, 23: 88-99.

Hult, C., J. T. Magla, H. P. Gideon, J. J. Linderman, and D. E. Kirschner. 2021. 'Neutrophil Dynamics Affect Mycobacterium tuberculosis Granuloma Outcomes and DisseminaBon', Front Immunol, 12: 712457.

Katayama, M., K. Ohmura, N. Yukawa, C. Terao, M. Hashimoto, H. Yoshifuji, D. Kawabata, T. Fujii, Y. Iwakura, and T. Mimori. 2013. 'Neutrophils are essential as a source of IL-17 in the effector phase of arthritis', PLoS One, 8: e62231.

Khader, S. A., G. K. Bell, J. E. Pearl, J. J. Fountain, J. Rangel-Moreno, G. E. Cilley, F. Shen, S. M. Eaton, S. L. Gaffen, S. L. Swain, R. M. Locksley, L. Haynes, T. D. Randall, and A. M. Cooper. 2007. 'IL-23 and IL-17 in the establishment of protective pulmonary CD4+ T cell responses after vaccination and during Mycobacterium tuberculosis challenge', Nat Immunol, 8: 369-77.

Leliefeld, P. H. C., J. Pillay, N. Vrisekoop, M. Heeres, T. Tak, M. Kox, S. H. M. Rooijakkers, T. W. Kuijpers, P. Pickkers, L. P. H. Leenen, and L. Koenderman. 2018. 'DifferenBal antibacterial control by neutrophil subsets', Blood Adv, 2: 1344-55.

Lemos, H. P., R. Grespan, S. M. Vieira, T. M. Cunha, W. A. Verri, Jr., K. S. Fernandes, F. O. Souto, I. B. McInnes, S. H. Ferreira, F. Y. Liew, and F. Q. Cunha. 2009. 'Prostaglandin mediates IL-23/IL-17induced neutrophil migraBon in inflammation by inhibiting IL-12 and IFNgamma production', Proc Natl Acad Sci U S A, 106: 5954-9.

Li, L., L. Huang, A. L. Vergis, H. Ye, A. Bajwa, V. Narayan, R. M. Strieter, D. L. Rosin, and M. D. Okusa. 2010. 'IL-17 produced by neutrophils regulates IFN-gamma-mediated neutrophil migration in mouse kidney ischemia-reperfusion injury', J Clin Invest, 120: 331-42.

Lin, A. M., C. J. Rubin, R. Khandpur, J. Y. Wang, M. Riblen, S. Yalavarthi, E. C. Villanueva, P. Shah, M. J. Kaplan, and A. T. Bruce. 2011. 'Mast cells and neutrophils release IL-17 through extracellular trap formation in psoriasis', J Immunol, 187: 490-500.

Lovewell, R. R., C. E. Baer, B. B. Mishra, C. M. Smith, and C. M. Sasseg. 2021. 'Granulocytes act as a niche for Mycobacterium tuberculosis growth', Mucosal Immunol, 14: 229-41.

Lyadova, I. V. 2017. 'Neutrophils in Tuberculosis: Heterogeneity Shapes the Way?', Mediators Inflamm, 2017: 8619307.

Mishra, B. B., R. R. Lovewell, A. J. Olive, G. Zhang, W. Wang, E. Eugenin, C. M. Smith, J. Y. Phuah, J. E. Long, M. L. Dubuke, S. G. Palace, J. D. Goguen, R. E. Baker, S. Nambi, R. Mishra, M. G. Booty, C. E. Baer, S. A. Shaffer, V. Dartois, B. A. McCormick, X. Chen, and C. M. Sasseg. 2017. 'Nitric oxide prevents a pathogen-permissive granulocytic inflammation during tuberculosis', Nat Microbiol, 2: 17072.

Napolitani, G., E. V. Acosta-Rodriguez, A. Lanzavecchia, and F. Sallusto. 2009. 'Prostaglandin E2 enhances Th17 responses via modulation of IL-17 and IFN-gamma production by memory CD4+ T cells', Eur J Immunol, 39: 1301-12.

Nwongbouwoh Muefong, C., O. Owolabi, S. Donkor, S. Charalambous, A. Bakuli, A. Rachow, C. Geldmacher, and J. S. Sutherland. 2022. 'Neutrophils Contribute to Severity of Tuberculosis Pathology and Recovery From Lung Damage Pre- and Posnreatment', Clin Infect Dis, 74: 1757-66.

Paulissen, S. M., J. P. van Hamburg, N. Davelaar, P. S. Asmawidjaja, J. M. Hazes, and E. Lubberts. 2013. 'Synovial fibroblasts directly induce Th17 pathogenicity via the cyclooxygenase/prostaglandin E2 pathway, independent of IL-23', J Immunol, 191: 1364-72.

Pedrosa, J., B. M. Saunders, R. Appelberg, I. M. Orme, M. T. Silva, and A. M. Cooper. 2000. 'Neutrophils play a protective nonphagocytic role in systemic Mycobacterium tuberculosis infection of mice', Infect Immun, 68: 577-83.

Polese, B., B. Thurairajah, H. Zhang, C. L. Soo, C. A. McMahon, G. Fontes, S. N. A. Hussain, V. Abadie, and I. L. King. 2021. 'Prostaglandin E(2) amplifies IL-17 production by gamma-delta T cells during barrier inflammation', Cell Rep, 36: 109456.

Rankin, A. N., S. V. Hendrix, S. K. Naik, and C. L. Stallings. 2022. 'Exploring the Role of Low-Density Neutrophils During Mycobacterium tuberculosis InfecBon', Front Cell Infect Microbiol, 12: 901590.

Xu, X. J., Q. Q. Ge, M. S. Yang, Y. Zhuang, B. Zhang, J. Q. Dong, F. Niu, H. Li, and B. Y. Liu. 2023. 'Neutrophil-derived interleukin-17A participates in neuroinflammation induced by traumatic brain injury', Neural Regen Res, 18: 1046-51.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary: 

In this paper Homan et al used mouse models of Metabolic Dysfunction-Associated Steatotic Liver Disease and different specific target deletions in cells to rule out the role of Complement 3a Receptor 1 in the pathogenesis of disease. They provided limited evidence and only descriptive results that despite C3aR being relevant in different contexts of inflammation, however, these tenets did not hold true. 

Weaknesses: 

(1) The results are based on readouts showing that C3aR is not involved in the pathogenesis of liver metabolic disease. 

(2) The description of the mouse models they used to validate their findings is not clear. Lysm-cre mice - which are claimed to delete C3aR in (?) macrophages are not specific for these cells, and the genetic strategy to delete C3aR in Kupffer cells is not clear. 

(3) Taking this into account, it is very challenging to determine the validity of these data, also considering that they are merely descriptive and correlative. 

We generated 2 different cohorts of mice using LysM-Cre (Jackson Strain #004781) to drive deletion in all macrophages and Clec4f-Cre (Jackson Strain #033296) to specifically ablate C3ar1 in Kupffer cells. We will ensure that experimental models will be clearly defined in the revised manuscript. The reviewer’s point is well taken that LysM-Cre transgene can also be active in granulocytes and some dendritic cells. Even so, despite deletion of C3ar1 in macrophages and other granulocytes, we do not see a major effect on hepatic steatosis and fibrosis in this GAN diet induced model of MASLD/MASH. This was a somewhat surprising finding. We do not agree that our findings are correlative. We specifically ablated C3aR1 in macrophages or Kupffer cells and found no significant differences in the major readouts of steatosis and fibrosis for MASLD/MASH between control and knockout mice. It is possible that in other models of liver injury that we did not test (e.g., short-term treatment with a hepatotoxin such as carbon tetrachloride), there may be differences in liver injury in mice lacking C3ar1 in macrophages, but the GAN diet model has been shown to better parallel the gene expression changes in human MAFLD/MASH.

Reviewer #2 (Public review):

Summary:

Homan et al. examined the effect of macrophage- or Kupffer cell-specific C3aR1 KO on MASLD/MASHrelated metabolic or liver phenotypes. 

Strengths:

Established macrophage- or Kupffer cell-specific C3aR1 KO mice. 

Weaknesses:

Lack of in-depth study; flaws in comparisons between KC-specific C3aR1KO and WT in the context of MASLD/MASH, because MASLD/MASH WT mice likely have a low abundance of C3aR1 on KCs. 

Homan et al. reported a set of observation data from macrophage or Kupffer cell-specific C3aR1KO mice. Several questions and concerns as follows could challenge the conclusions of this study: 

(1) As C3aR1 is robustly repressed in MASLD or MASH liver, GAN feeding likely reduced C3aR1 abundance in the liver of WT mice. Thus, it is not surprising that there were no significant differences in liver phenotypes between WT vs. C3aR1KO mice after prolonged GAN diet feeding. It would give more significance to the study if restoring C3aR1 abundance in KCs in the context of MASLD/MASH. 

GAN diet feeding resulted in higher liver C3ar1 compared to regular diet (Figure 1H). This thus became an impetus for studying the effects of C3ar1 deletion in macrophages or Kupffer cells, which are responsible for the majority of liver C3ar1 expression, in MASLD/MASH (Figures 2B and 3H).  

(2) Would C3aR1KO mice develop liver abnormalities after a short period of GAN diet feeding?  

We did not assess if short term GAN diet feeding resulted in significant differences in liver abnormalities in the C3ar1 macrophage or Kupffer cell knockout mice. Perhaps the reviewer’s point is that perhaps with shorter periods of GAN diet feeding there may be a phenotype in the KO mice. We agree that this is entirely possible, though with shorter feeding timeframes what is typically seen is hepatic steatosis without fibrosis. Nevertheless, the most important element in our opinion for a disease preventing or modifying model lies with the longer-term GAN diet feeding. With long term GAN diet feeding that has been previously shown to model human MASLD/MASH, we did not observe significant differences in liver abnormalities with the KO mice.

(3) What would be the liver macrophage phenotypes in WT vs C3aR1KO mice after GAN feeding? 

Similar to the above point, given the lack of a major MASLD/MASH phenotype in hepatic steatosis and fibrosis, we did not further profile the liver macrophage profiles of the macrophage or Kupffer cell C3ar1 KO mice with GAN feeding.  

(4) In Fig 1D, >25wks GAN feeding had minimal effects on female body weight gain. These GAN-fed female mice also develop NASLD/MASH liver abnormalities? 

We thank the reviewer for this question. In general, female GAN-fed mice develop milder MASLD/MASH abnormalities. We will include additional data in the revised manuscript.

(5) Would C3aR1KO result in differences in liver phenotypes, including macrophage population/activation, liver inflammation, lipogenesis, in lean mice? 

Likewise, we will include data further characterizing liver inflammation, lipogenesis and macrophages in macrophage C3ar1 KO mice under lean/regular diet conditions.

(6) The authors should provide more information regarding the generation of KC-specific C3aR1KO. Which Cre mice were used to breed with C3aR1 flox mice? 

Clec4f-Cre transgenic mice were used to generate Kupffer cell specific KO of C3ar1. This will be clarified and explicitly stated in the revised manuscript.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study puts forth the model that under IFN-B stimulation, liquid-phase WTAP coordinates with the transcription factor STAT1 to recruit MTC to the promoter region of interferon-stimulated genes (ISGs), mediating the installation of m6A on newly synthesized ISG mRNAs. This model is supported by strong evidence that the phosphorylation state of WTAP, regulated by PPP4, is regulated by IFN-B stimulation, and that this results in interactions between WTAP, the m6A methyltransferase complex, and STAT1, a transcription factor that mediates activation of ISGs. This was demonstrated via a combination of microscopy, immunoprecipitations, m6A sequencing, and ChIP. These experiments converge on a set of experiments that nicely demonstrate that IFN-B stimulation increases the interaction between WTAP, METTL3, and STAT1, that this interaction is lost with the knockdown of WTAP (even in the presence of IFN-B), and that this IFN-B stimulation also induces METTL3-ISG interactions.

Strengths:

The evidence for the IFN-B stimulated interaction between METTL3 and STAT1, mediated by WTAP, is quite strong. Removal of WTAP in this system seems to be sufficient to reduce these interactions and the concomitant m6A methylation of ISGs. The conclusion that the phosphorylation state of WTAP is important in this process is also quite well supported.

Weaknesses:

The evidence that the above mechanism is fundamentally driven by different phase-separated pools of WTAP (regulated by its phosphorylation state) is weaker. These experiments rely relatively heavily on the treatment of cells with 1,6-hexanediol, which has been shown to have some off-target effects on phosphatases and kinases (PMID 33814344).

Given that the model invoked in this study depends on the phosphorylation (or lack thereof) of WTAP, this is a particularly relevant concern.

Related to this point, it is also interesting (and potentially concerning for the proposed model) that the initial region of WTAP that was predicted to be disordered is in fact not the region that the authors demonstrate is important for the different phase-separated states. Taking all the data together, it is also not clear to me that one has to invoke phase separation in the proposed mechanism.

We are grateful for the Reviewer’s positive comment and constructive feedback. In this article, we claim a novel and important mechanism that de-phosphorylation-driven solid to liquid phase transition of WTAP mediates its co-transcriptional m6A modification. We first observed that WTAP underwent phase transition during virus infection and IFN-β stimulation, and confirmed the phase transition driven force of WTAP through multiple experiments. Besides 1,6‐hexanediol (1,6-hex) treatment, we also introduced S/T to D/A mutations to mimic the phosphorylation and de-phosphorylation WTAP in vitro and in cells, identified 5ST-D mutant as SLPS mutant, and 5ST-A mutant as LLPS mutant. We then performed 1,6-hex experiment to confirm the importance of phase separation for WTAP function, and revealed that 5ST-D SLPS mutant and 5ST-A LLPS mutant had different influence on WTAP-promoter region interaction and co-transcriptional m6A modification. Following the reviewer’s suggestion, we need to further clarify the phosphorylation of WTAP phase separation. We plan to repeat the experiments by introducing potent PP4 inhibitor, fostriecin, and performed further experiments to explore the effect of WTAP IDR domain, which is reported to play a critical role for its phase separation.

1,6-hex was initially considered as the inhibitor of hydrophobic interaction which involved in various kinds of protein-protein interaction, indicating that off-target effects of 1,6-hex was inevitable. It is reported that 1,6-hex impaired RNA pol II CTD specific phosphatase and kinase activity at 5% concentration3. However, 1,6-hex is still widely used in the LLPS-associated functional studies despite its off-target effect. Related to this article, 10% 1,6-hex was reported to dissolve WTAP phase separation droplets2. Beside WTAP, 1,6-hex (5%-10% w/v) was also used to explore the phase separation characteristic and function on phosphorylated protein or even kinase, including p‐tau441, TAZ, HSF1 and so on4-6. 10% 1,6-hex inhibited the crucial role of phosphorylation-driven HSF1 LLPS in chromatin binding and transcriptional process presented by RNA-seq dataset6, indicating the function on kinase or phosphatase of 1,6-hex might not a global effect. To avoid the 1,6-hex-mediated kinase/phosphatase impairment in this project, we introduced the WTAP SLPS mutation and LLPS mutation besides 1,6-hex treatment to explore the m6A modification function of WTAP phase transition. We plan to repeat the experiments by lower the 1,6-hex concentration, check the WTAP phosphorylation status after 1,6-hex treatment, and discuss them in the discussion part.

A considerable number of proteins undergo phase separation via interactions between intrinsically disordered regions (IDRs). IDR contains more charged and polar amino acids to present multiple weakly interacting elements, while lacking hydrophobic amino acids to show flexible conformations7. In our article, we used PLAAC websites (http://plaac.wi.mit.edu/) to predict IDR domain of WTAP, and a fragment (234-249 amino acids) was predicted as prion-like domain. However, deletion of this fragment failed to abolish the phase separation properties of WTAP, which might be the main confusion to reviewers. To explain this issue, we checked the WTAP structure (within part of MTC complex) from protein data bank (https://www.rcsb.org/structure/7VF2) and found that prediction of IDR has been renewed due to the update of different algorithm. IDR of WTAP has expanded to 245-396 amino acids, containing the whole CTD region. According to our results, lack of CTD inhibited WTAP liquid-liquid phase separation both in vitro and in cells, while the phosphorylation status on CTD had dramatic impact on WTAP phase transition, which was consistent with the LLPS-regulating function of IDR. Therefore, we will revise our description on WTAP IDR, and performed further experiment to test its function.

Taken together, given the highly association between WTAP phosphorylation with phase separation status and its function during IFN-β stimulation, it is necessary to involve WTAP phase separation in our mechanism. We will perform further experiments to propose more convincing evidence and perfect our project.

Reviewer #2 (Public review):

In this study, Cai and colleagues investigate how one component of the m6A methyltransferase complex, the WTAP protein, responds to IFNb stimulation. They find that viral infection or IFNb stimulation induces the transition of WTAP from aggregates to liquid droplets through dephosphorylation by PPP4. This process affects the m6A modification levels of ISG mRNAs and modulates their stability. In addition, the WTAP droplets interact with the transcription factor STAT1 to recruit the methyltransferase complex to ISG promoters and enhance m6A modification during transcription. The investigation dives into a previously unexplored area of how viral infection or IFNb stimulation affects m6A modification on ISGs. The observation that WTAP undergoes a phase transition is significant in our understanding of the mechanisms underlying m6A's function in immunity. However, there are still key gaps that should be addressed to fully accept the model presented.

Major points:

(1) More detailed analyses on the effects of WTAP sgRNA on the m6A modification of ISGs:

a. A comprehensive summary of the ISGs, including the percentage of ISGs that are m6A-modified. merip-isg percentage

b. The distribution of m6A modification across the ISGs. topology

c. A comparison of the m6A modification distribution in ISGs with non-ISGs. topology

In addition, since the authors propose a novel mechanism where the interaction between phosphorylated STAT1 and WTAP directs the MTC to the promoter regions of ISGs to facilitate co-transcriptional m6A modification, it is critical to analyze whether the m6A modification distribution holds true in the data.

We appreciate the reviewer‘s summary of our manuscript and the constructive assessment. We plan to perform the related analysis accordingly to present the m6A modification in ISGs in our model. 

(2) Since a key part of the model includes the cytosol-localized STAT1 protein undergoing phosphorylation to translocate to the nucleus to mediate gene expression, the authors should focus on the interaction between phosphorylated STAT1 and WTAP in Figure 4, rather than the unphosphorylated STAT1. Only phosphorylated STAT1 localizes to the nucleus, so the presence of pSTAT1 in the immunoprecipitate is critical for establishing a functional link between STAT1 activation and its interaction with WTAP.

We plan to repeat the immunoprecipitation experiments to clarify the function of pSTAT1 in WTAP interaction and m6A modification as the reviewer suggested.

(3) The authors should include pSTAT1 ChIP-seq and WTAP ChIP-seq on IFNb-treated samples in Figure 5 to allow for a comprehensive and unbiased genomic analysis for comparing the overlaps of peaks from both ChIP-seq datasets. These results should further support their hypothesis that WTAP interacts with pSTAT1 to enhance m6A modifications on ISGs.

We first performed the MeRIP-seq and RNA-seq and explored the critical role of WTAP in ISGs m6A modification and expression. By immunoprecipitation and immunofluorescence experiments, we found phase transition of WTAP enhanced its interaction to pSTAT1. These results indicate that WTAP mediated ISGs m6A modification and expression by enhanced its interaction with pSTAT1 during virus infection and IFN-β stimulation. However, we were still not sure how WTAP-mediated m6A modification related to pSTAT1-mediated transcription. By analyzing METTL3 ChIP-seq data or caPAR-CLIP-seq data, several researches have revealed the recruitment of m6A methylation complex (MTC) to transcription start sites (TSS) of coding genes and R-loop structure by interacting with transcriptional factors STAT5B or DNA helicase DDX21, indicating the engagement of MTC mediated m6A modification on nascent transcripts at the very beginning of transcription 8-10. Thus, we proposed that phase transition of WTAP could be recruited to the ISGs promoter region by pSTAT1, and verified this hypothesis by pSTAT1/WTAP-ChIP-qPCR. We believe ChIP-seq experiment is a good idea to explore the mechanism in depth, but the results in this article for now are enough to explain our mechanism. We will continuously focus on the whole genome chromatin distribution of WTAP and explore more functional effect of transcriptional factor-dependent WTAP-promoter region interaction in t.

Minor points:

(1) Since IFNb is primarily known for modulating biological processes through gene transcription, it would be informative if the authors discussed the mechanism of how IFNb would induce the interaction between WTAP and PPP4.

(2) The authors should include mCherry alone controls in Figure 1D to demonstrate that mCherry does not contribute to the phase separation of WTAP. Does mCherry have or lack a PLD?

(3) The authors should clarify the immunoprecipitation assays in the methods. For example, the labeling in Figure 2A suggests that antibodies against WTAP and pan-p were used for two immunoprecipitations. Is that accurate?

(4) The authors should include overall m6A modification levels quantified of GFPsgRNA and WTAPsgRNA cells, either by mass spectrometry (preferably) or dot blot.

We thank reviewer for raising these useful suggestions. We will perform related experiments and revised the manuscript carefully the as reviewer suggested.

Reviewer #3 (Public review):

Summary:

This study presents a valuable finding on the mechanism used by WTAP to modulate the IFN-β stimulation. It describes the phase transition of WTAP driven by IFN-β-induced dephosphorylation. The evidence supporting the claims of the authors is solid, although major analysis and controls would strengthen the impact of the findings. Additionally, more attention to the figure design and to the text would help the reader to understand the major findings.

Strength:

The key finding is the revelation that WTAP undergoes phase separation during virus infection or IFN-β treatment. The authors conducted a series of precise experiments to uncover the mechanism behind WTAP phase separation and identified the regulatory role of 5 phosphorylation sites. They also succeeded in pinpointing the phosphatase involved.

Weaknesses:

However, as the authors acknowledge, it is already widely known in the field that IFN and viral infection regulate m6A mRNAs and ISGs. Therefore, a more detailed discussion could help the reader interpret the obtained findings in light of previous research.

It is well-known that protein concentration drives phase separation events. Similarly, previous studies and some of the figures presented by the authors show an increase in WTAP expression upon IFN treatment. The authors do not discuss the contribution of WTAP expression levels to the phase separation event observed upon IFN treatment. Similarly, METTL3 and METTL14, as well as other proteins of the MTC are upregulated upon IFN treatment. How does the MTC protein concentration contribute to the observed phase separation event?

How is PP4 related to the IFN signaling cascade?

In general, it is very confusing to talk about WTAP KO as WTAPgRNA.

We are grateful for the positive comments and the unbiased advice by reviewer. To interpret the findings in previous research, we will revise the manuscript carefully and preform more detailed discussion on ISGs m6A modification during virus infection or IFN stimulation. As previous reported, WTAP protein level will be induced by long time IFN-β stimulation or LPS stimulation, while LPS-induced WTAP expression promoted its phase separation ability2,11. Although there was no significant upregulation of WTAP expression level in our short time treatment, we hypothesized that WTAP phase separation will be promoted due to higher protein concentration after long time IFN stimulation, enhancing m6A modification deposition on ISGs mRNA, revealing a feedback loop between WTAP phase separation and m6A modification during specific stimulation. To discuss the effect of MTC protein concentration in our proposed event, we will perform immunoblotting experiments of MTC proteins and check the phase separation effect in different WTAP concentration.

Protein phosphatase 4 (PP4) is a multi-subunit Ser/Thr phosphatase complex that participate in diverse cellular pathways including DDR, cell cycle progression, and apoptosis12. Protein phosphatase 4 catalytic subunit 4C (PPP4C) is one of the components of PP4 complex. Previous research showed that knockout of PPP4C enhanced IFN-β downstream signaling and gene expression, which was consistent with our findings that knockdown of PPP4C impaired WTAP-mediated m6A modification, enhanced the ISGs expression. Since there was no significant enhancement in PPP4C expression level during IFN-β stimulation in our results, we will consider to explore the post-translation modification that may influence the protein-protein interaction, such as ubiquitination.

In this project, all the WTAP-deficient THP-1 cells were bulk cells treated with WTAPsgRNA, but not monoclonal knockout cells. We confirmed that WTAP expression was efficiently knockdown in WTAPsgRNA THP-1 cells, and the m6A modification level has been impaired, avoiding the compensatory effect on m6A modification by other possible proteins. Thus, we prefer to call it WTAPsgRNA THP-1 cells rather than WTAP KO THP-1 cells.  

References

(1) Raja, R., Wu, C., Bassoy, E.Y., Rubino, T.E., Jr., Utagawa, E.C., Magtibay, P.M., Butler, K.A., and Curtis, M. (2022). PP4 inhibition sensitizes ovarian cancer to NK cell-mediated cytotoxicity via STAT1 activation and inflammatory signaling. J Immunother Cancer 10. 10.1136/jitc-2022-005026.

(2) Ge, Y., Chen, R., Ling, T., Liu, B., Huang, J., Cheng, Y., Lin, Y., Chen, H., Xie, X., Xia, G., et al. (2024). Elevated WTAP promotes hyperinflammation by increasing m6A modification in inflammatory disease models. J Clin Invest 134. 10.1172/JCI177932.

(3) Duster, R., Kaltheuner, I.H., Schmitz, M., and Geyer, M. (2021). 1,6-Hexanediol, commonly used to dissolve liquid-liquid phase separated condensates, directly impairs kinase and phosphatase activities. J Biol Chem 296, 100260. 10.1016/j.jbc.2021.100260.

(4) Wegmann, S., Eftekharzadeh, B., Tepper, K., Zoltowska, K.M., Bennett, R.E., Dujardin, S., Laskowski, P.R., MacKenzie, D., Kamath, T., Commins, C., et al. (2018). Tau protein liquid-liquid phase separation can initiate tau aggregation. The EMBO journal 37. 10.15252/embj.201798049.

(5) Lu, Y., Wu, T., Gutman, O., Lu, H., Zhou, Q., Henis, Y.I., and Luo, K. (2020). Phase separation of TAZ compartmentalizes the transcription machinery to promote gene expression. Nat Cell Biol 22, 453-464. 10.1038/s41556-020-0485-0.

(6) Zhang, H., Shao, S., Zeng, Y., Wang, X., Qin, Y., Ren, Q., Xiang, S., Wang, Y., Xiao, J., and Sun, Y. (2022). Reversible phase separation of HSF1 is required for an acute transcriptional response during heat shock. Nat Cell Biol 24, 340-352. 10.1038/s41556-022-00846-7.

(7) Hou, S., Hu, J., Yu, Z., Li, D., Liu, C., and Zhang, Y. (2024). Machine learning predictor PSPire screens for phase-separating proteins lacking intrinsically disordered regions. Nat Commun 15, 2147. 10.1038/s41467-024-46445-y.

(8) Hao, J.D., Liu, Q.L., Liu, M.X., Yang, X., Wang, L.M., Su, S.Y., Xiao, W., Zhang, M.Q., Zhang, Y.C., Zhang, L., et al. (2024). DDX21 mediates co-transcriptional RNA m(6)A modification to promote transcription termination and genome stability. Mol Cell 84, 1711-1726 e1711. 10.1016/j.molcel.2024.03.006.

(9) Barbieri, I., Tzelepis, K., Pandolfini, L., Shi, J., Millan-Zambrano, G., Robson, S.C., Aspris, D., Migliori, V., Bannister, A.J., Han, N., et al. (2017). Promoter-bound METTL3 maintains myeloid leukaemia by m(6)A-dependent translation control. Nature 552, 126-131. 10.1038/nature24678.

(10) Bhattarai, P.Y., Kim, G., Lim, S.C., and Choi, H.S. (2024). METTL3-STAT5B interaction facilitates the co-transcriptional m(6)A modification of mRNA to promote breast tumorigenesis. Cancer Lett 603, 217215. 10.1016/j.canlet.2024.217215.

(11) Ge, Y., Ling, T., Wang, Y., Jia, X., Xie, X., Chen, R., Chen, S., Yuan, S., and Xu, A. (2021). Degradation of WTAP blocks antiviral responses by reducing the m(6) A levels of IRF3 and IFNAR1 mRNA. EMBO Rep 22, e52101. 10.15252/embr.202052101.

(12) Dong, M.Z., Ouyang, Y.C., Gao, S.C., Ma, X.S., Hou, Y., Schatten, H., Wang, Z.B., and Sun, Q.Y. (2022). PPP4C facilitates homologous recombination DNA repair by dephosphorylating PLK1 during early embryo development. Development 149. 10.1242/dev.200351.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Although this study provides a comprehensive outlook on the ETC function in various tissues, the main caveat is that it's too technical and descriptive. The authors didn't invest much effort in putting their findings in the context of the biological function of the tissue analyzed, i.e., some tissues might be more glycolytic than others and have low ETC activity.

To better contextualize our results, we have added substantial amount of new information to the Discussion Section.

Also, it is unclear what slight changes in the activity of one or the other ETC complex mean in terms of mitochondrial ATP production.

Unfortunately, the method we used can only determine oxygen consumption rate through complex I (CI), CII, or CIV. It cannot tell us about ATP production. This method only measures maximal uncoupled respiration.

Likely, these small changes reported do not affect the mitochondrial respiration.

We are indeed looking at mitochondrial respiration. Some changes are more dramatic while others are much more modest. We are looking at the normal aging process across tissues (focusing on mitochondrial respiration) and not pathological states. As such, we expect many of the changes in mitochondrial respiration across tissues to be mild or relatively modest. After all, aging is slow and progressive. In fact, the variations we observed in mitochondrial respiration across tissues are consistent with the known heterogenous rate of aging across tissues.

With such a detailed dataset, the study falls short of deriving more functionally relevant conclusions about the heterogeneity of mitochondrial function in various tissues. In the current format, the readers get lost in the large amount of data presented in a technical manner.

We agree that the paper contains a large amount of information. In the revised manuscript, we did our best to contextualize our results by substantially expanding the Discussion Section.

Also, it is highly recommended that all the raw data and the values be made available as an Excel sheet (or other user-friendly formats) as a resource to the community.

We included all the data in two excel sheets (Figure 1 – data source 1; Figure 1 – data source 2). We presented them in such as way that it will be easy for other investigators to follow and re-use our dataset in their own studies for comparison.

Major concerns

(1) In this study, the authors used the method developed by Acin-Perez and colleagues (EMBO J, 2020) to analyze ETC complex activities in mitochondria derived from the snap-frozen tissue samples. However, the preservation of cellular/mitochondrial integrity in different types of tissues after being snap-frozen was not validated.

All the samples are actually maximally preserved due to being snap frozen. Freezing the samples disrupts the mitochondria to produce membrane fragments. Subsequent thawing, mincing, and homogenization in a non-detergent based buffer (mannose-sucrose) ensures that all tissue samples are maximally disrupted into fragments which contain ETC units in various combinations. This allows the assay to give an accurate representation of maximal respiratory capacity given the ETC units present in a tissue sample.

Since aging has been identified as the most important effector in this study, it is essential to validate how aging affects respiration in various fresh frozen tissues. Such analysis will ensure that the results presented are not due to the differential preservation of the mitochondrial respiration in the frozen tissue. In addition, such validations will further strengthen the conclusions and promote the broad usability of this "new" method.

The reason we adopted this method is because it has been rigorously validated in the original publication (PMID: 32432379) and a subsequent methods paper (PMID: 33320426). The authors in the original paper benchmarked their frozen tissue method with freshly isolated mitochondria from the same set of tissues. Their work showed highly comparable mitochondrial respiration from frozen tissues and isolated mitochondria. For this reason, we did not repeat those validation studies.

(2) In this study, the authors sampled the maximal activity of ETC complex I, II, and IV, but throughout the manuscript, they discussed the data in the context of mitochondrial function.

We apologize that we did not make it clearer in our manuscript. We corrected this in our revised manuscript (the Discussion Section). Our method we measure respiration starting at Complex I (CI; via NADH), starting at CII (via succinate), or starting at CIV (using TMPD and ascorbate). Regardless of whether electrons (donated by the substrate) enter the respiratory chain through CI, CII or CIV, oxygen (as the final electron acceptor) is only consumed at CIV. Therefor, the method measures mitochondrial respiration and function through CI, CII, or CIV. This high-resolution respirometry analysis method is different from the classic enzymatic method of assessing CI, CII, or CIV activity individually; the enzymatic method does not actually measure oxygen consumption due to electrons flowing through the respiratory complexes.

However, it is unclear how the changes in CI, CII, and CIV activity affect overall mitochondrial function (if at all) and how small changes seen in the maximal activity of one or more complexes affect the efficiency and efficacy of ATP production (OxPhos).

Please see the preceding response to the previous question. The method is measuring mitochondrial respiration through CI, CII or CIV. The limitation of this method is that it is maximal uncoupled respiration; namely, mitochondrial respiration is not coupled to ATP synthesis since the measurements are not performed on intact mitochondria. As such, we cannot say anything about the efficiency and efficacy of ATP production. This will be an interesting future studies to further investigating tissue level variations of mitochondrial OXPHOS.

The authors report huge variability between the activity of different complexes - in some tissues all three complexes (CI, CII, and CIV) and often in others, just one complex was affected. For example, as presented in Figure 4, there is no difference in CI activity in the hippocampus and cerebellum, but there is a slight change in CII and CIV activity. In contrast, in heart atria, there is a change in the activity of CI but not in CII and CIV. However, the authors still suggest that there is a significant difference in mitochondrial activity (e.g., "Old males showed a striking increase in mitochondrial activity via CI in the heart atria....reduced mitochondrial respiration in the brain cortex..." - Lines 5-7, Page 9). Until and unless a clear justification is provided, the authors should not make these broad claims on mitochondrial respiration based on small changes in the activity of one or more complexes (CI/CII/CIV). With such a data-heavy and descriptive study, it is confusing to track what is relevant and what is not for the functioning of mitochondria.

We have attempted to address these issues in the revised Discussion section.

(3) What do differences in the ETC complex CI, CII, and CIV activity in the same tissue mean? What role does the differential activity of these complexes (CI, CII, and CIV) play in mitochondrial function? What do changes in Oxphos mean for different tissues? Does that mean the tissue (cells involved) shift more towards glycolysis to derive their energy? In the best world, a few experiments related to the glycolytic state of the cells would have been ideal to solidify their finding further. The authors could have easily used ECAR measurements for some tissues to support their key conclusions.

We have attempted to address these issues in the revised Discussion section. The frozen tissue method does not involve intact mitochondria. As such, the method cannot measure ECAR, which requires the presence of intact mitochondria.

(4) The authors further analyzed parameters that significantly changed across their study (Figure 7, 98 data points analyzed). The main caveat of such analysis is that some tissue types would be represented three or even more times (due to changes in the activity of all three complexes - CI, CII, and CIV, and across different ages and sexes), and some just once. Such a method of analysis will skew the interpretation towards a few over-represented organ/tissue systems. Perhaps the authors should separately analyze tissue where all three complexes are affected from those with just one affected complex.

Figure 7 summarizes the differences between male vs female, and between young vs old. All the tissue-by-tissue comparisons (data separated by CI-linked respiration, CII-linked respiration, and CIV-linked respiration) can be found in earlier figures (Figure 1-6).

The focus of Figure 7 is to helps us better appreciate all the changes seen in the preceding Figure 1-6:

Panel A and B indicate all changes that are considered significant

Panel C indicates total tissues with at least one significantly affected respiration

Panel D indicates total magnitude of change (i.e., which tissue has the highest OCR) offering a non-relative view

Panel E indicates whole body separations

Panel F indicates whole body separations and age vs sex clustering

(5) The current protocol does not provide cell-type-specific resolution and will be unable to identify the cellular source of mitochondrial respiration. This becomes important, especially for those organ systems with tremendous cellular heterogeneity, such as the brain. The authors should discuss whether the observed changes result from an altered mitochondria respiratory capacity or if changes in proportions of cell types in the different conditions studied (young vs. aged) might also contribute to differential mitochondrial respiration.

We agree with the reviewer that this is a limitation of the method. We have addressed this issue in the revised Discussion section.

(6) Another critical concern of this study is that the same datasets were repeatedly analyzed and reanalyzed throughout the study with almost the same conclusion - namely, aging affects mitochondrial function, and sex-specific differences are limited to very few organs. Although this study has considerable potential, the authors missed the chance to add new insights into the distinct characteristics of mitochondrial activity in various tissue and organ systems. The author should invest significant efforts in putting their data in the context of mitochondrial function.

We have attempted to address these issues in the revised Discussion section.

Reviewer #2 (Public Review):

Summary:

The authors utilize a new technique to measure mitochondrial respiration from frozen tissue extracts, which goes around the historical problem of purifying mitochondria prior to analysis, a process that requires a fair amount of time and cannot be easily scaled up.

Strengths:

A comprehensive analysis of mitochondrial respiration across tissues, sexes, and two different ages provides foundational knowledge needed in the field.

Weaknesses:

While many of the findings are mostly descriptive, this paper provides a large amount of data for the community and can be used as a reference for further studies. As the authors suggest, this is a new atlas of mitochondrial function in mouse. The inclusion of a middle aged time point and a slightly older young point (3-6 months) would be beneficial to the study.

We agreed with the reviewer that inclusion of additional time points (e.g., 3-6 months) would further strengthen the study. However, the cost, labor, and time associated with another set of samples (660 tissue samples from male and female mice and 1980 respirometry assays) are too high for our lab with limited budget and manpower. Regrettably, we will not be able to carry out the extra work as requested by the reviewer.  

Reviewer #3 (Public Review):

The aim of the study was to map, a) whether different tissues exhibit different metabolic profiles (this is known already), what differences are found between female and male mice and how the profiles changes with age. In particular, the study recorded the activity of respirasomes, i.e. the concerted activity of mitochondrial respiratory complex chains consisting of CI+CIII2+CIV, CII+CIII2+CIV or CIV alone.

The strength is certainly the atlas of oxidative metabolism in the whole mouse body, the inclusion of the two different sexes and the comparison between young and old mice. The measurement was performed on frozen tissue, which is possible as already shown (Acin-Perez et al, EMBO J, 2020).

Weakness:

The assay reveals the maximum capacity of enzyme activity, which is an artificial situation and may differ from in vivo respiration, as the authors themselves discuss. The material used was a very crude preparation of cells containing mitochondria and other cytosolic compounds and organelles. Thus, the conditions are not well defined and the respiratory chain activity was certainly uncoupled from ATP synthesis. Preparation of more pure mitochondria and testing for coupling would allow evaluation of additional parameters: P/O ratios, feedback mechanism, basal respiration, and ATP-coupled respiration, which reflect in vivo conditions much better. The discussion is rather descriptive and cautious and could lead to some speculations about what could cause the differences in respiration and also what consequences these could have, or what certain changes imply.

Nevertheless, this study is an important step towards this kind of analysis.

We have attempted to address some of these issues in the revised Discussion Section. The frozen tissue method can only measure maximal uncoupled respiration. Because we are not measuring mitochondrial respiration using intact mitochondria, several of the functional parameters the reviewer alluded to (e.g., P/O ratios, feedback mechanism, basal respiration, and ATP-coupled respiration) simply cannot be obtained with the current set of samples. Nevertheless, we agree that all the additional data (if obtained) would be very informative.

Reviewer #1 (Recommendations For The Authors):

(1) For most of the comparative analysis, the authors normalized OCR/min to MitoTracker Deep RedFM (MTDR) fluorescence intensity. Why was the data normalized to the total protein content not used for comparative analysis? Is there a correlation between MTDR fluorescence and the protein content across different tissues?

Given that we used the crude extract method, total protein content does not equal total mitochondrial protein content. This is why the MTDR method was used, as this represents a high throughput method of assessing mitochondrial mass in this volume of samples. In general, the total protein concentration is used to ensure the respiration intensity was approximately the same across all samples loaded into the Seahorse machine.

(2) To test the mitochondrial isolation yield, the authors should run immunoblot against canonical mitochondrial proteins in both homogenates and mitochondrial-containing supernatants and show that the protocol followed effectively enriched mitochondria in the supernatant fraction. This would also strengthen the notion that the "µg protein" value used to normalize the total mitochondrial content comes from isolated mitochondria and not other extra-mitochondrial proteins.

Because we are using crude tissue lysate (from frozen tissue), the total ug protein content does not come from isolated mitochondria; for this reason, it was not used and this is why MTDR was. Total mitochondrial protein content is subject to change depending on tissue for non-mitochondrial reasons. This method does not use isolated mitochondria; we only use tissue lysates enriched for mitochondrial proteins. This method has been rigorously validated in the original study (PMID: 32432379) and a subsequent methods paper (PMID: 33320426). In those studies, the authors had performed requisite quality checks the reviewer has asked for (e.g., immunoblot against canonical mitochondrial proteins in both homogenates and mitochondrial-containing supernatants to show effective enrichment of mitochondrial proteins). For this reason, we did not repeat this.

(3) MitoTracker loads into mitochondria in a membrane potential-dependent manner. The authors should rule out the possibility that samples from different ages and sexes might have different mitochondrial membrane potentials and exhibit a differential MitoTracker loading capacity. This becomes relevant for data normalization based on MTDR (MTDR/µg protein) since it was assumed that loading capacity is the same for mitochondria across different tissue and age groups.

MitoTracker Deep Red is not membrane potential dependent and can be effectively used to quantify mitochondrial mass even when mitochondrial membrane potential is lost. This is highlighted in the original study (PMID: 32432379).

(4) Page 11, line 3 typo - across, not cross.

Response: We have fixed the typo.

Reviewer #2 (Recommendations For The Authors):

If possible, I would include a middle aged time point between 12 and 14 months of age.

We agreed with reviewer that inclusion of additional time points (e.g., 3-6 months) would further strengthen the study. However, the cost, labor, and time associated with another set of samples (660 tissue samples from male and female mice and 1980 respirometry assays) are too high for our lab with limited budget and manpower. Regrettably, we will not be able to carry out the extra work as requested by the reviewer. 

Reviewer #3 (Recommendations For The Authors):

Overall, the work is well done and the data are well processed making them easy to understand. Some minor adjustments would improve the manuscript further:

- Significance OCR in Figure 2, maybe add error bars?

We have added the error bars and statistical significance to revised Figure 2.

- Tissue comparison A-C, right panel: graphs are cropped

We are not sure what the reviewer meant here. We have double checked all our revised figures to make sure nothing is accidentally cropped.

- Heart ventricle: Old males and females have higher CI- and CII-dependent respiration than young males and females? Only CIV respiration is lower?

Comparing old to young male or female heart ventricle respiration via CI or CII shows an increase in maximal capacity with age. CIV-linked respiration is in the upward direction as well, although not significant, when comparing old to young. When comparing the respiration values among themselves within a mouse, i.e. old male CI- or CII-linked respiration compared to old male CIV- linked respiration, we can see that the old male CIV-linked respiration is very similar. When comparing the same in the old female mouse, there appears to be something special about electrons entering through CI as compared to CII or CIV, as CI-linked respiration appears to be elevated compared to both CII and CIV. Although we do not know if this is significantly different, the trend in the data is clear. We do not know the exact reason as to why this occurred in the heart ventricles. To differing degrees, the connected nature of CI-, CII-, and CIV-linked respirations seems to be in a generally similar style in most skeletal muscles as well, and the old male heart atria. Again, the root of this discrepancy is unknown and potentially indicates an interesting physiologic trait of certain types of muscle and merits further exploration.

- What is plotted in Fig.3: The mean of all OCR of all tissues? A,B,C: Plot with break in x-axis to expand the violin, add mean/median values as numbers to the graph (same for Fig4)

The left most side of Figure 3 A, B, and C shows the average OCR/MTDR value across all tissues in a group. Each tissue assayed is represented in the violin plot as an open circle.

- Fig. 3D: add YM/YF to graph for better understanding, same in following figures

This is in the scale bar next to all heat maps presented in the figures. We also added to the revised figure as well to improve clarity.

- Additional figures: x-axis title (time) is missing in OCR graphs

Time has been added to the x axis of all additional figures for clarity.

- Also a more general question is: where the concentrations of substrates and inhibitors optimized before starting the series of experiments?

All the details of assay optimization was carried out in the original study (PMID: 32432379) and the subsequent methods paper (PMID: 33320426). Because we had to survey 33 different tissues, we tested and optimized the “optimal” protein concentrations we need to use; the primary goal of this was to balance enough respiration signal without too much respiration signal across all tissue types as to keep all the diverse tissues analyzed under the Seahorse machine’s capabilities of detection. Through our optimization of mostly the very high respiring tissues like heart and kidney, we were also able to prove that all substrates and inhibitors were in saturating concentrations since we could get respiration to go higher if more sample was added and that all signal could be lost in these samples with the same amount of inhibitors.

Author response:

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

Reviewer 2:

In addition, it is still unacceptable for me that the number of ovulated oocytes in mice at 6 months of age is only one third of young mice (10 vs 30; Fig. S1E). The most of published literature show that mice at 12 months of age still have ~10 ovulated oocytes.

We disagree with the reviewer’s comment, and the concerns raised were not shared by the other reviewers.  We have reported our data with full transparency (each data point is plotted). In the current study, we observed an intermediate phenotype in gamete number (assessed by both ovarian follicle counts and ovulated eggs) when comparing 6 month old mice to 6 week or 10 month old mice; this is as expected. It is well accepted that follicle counts are highly mouse strain dependent.  Although the reviewer mentions that mice at 12 months have ~10 ovulated oocytes, no actual references are provided nor are the mouse strain or other relevant experimental details mentioned.  Therefore, we do not know how these quoted metrics relate to the female FVB mice used in our current study.   As clearly explained and justified in our manuscript, we used mice at 6 months and 10 months to represent a physiologic aging continuum. 

Moreover, based on the follicle counting method used in the present study (Fig. S1D), there are no antral follicles observed in mice at 6 months and 10 months of age, which is not reasonable.

This statement is incorrect. Antral follicles were present at 6 and 10 months of age, but due to the scale of the y-axis and the normalization of follicle number/area in Fig. S1D, the values are small.  The absolute number of antral follicles per ovary (counted in every 5th section) was 31.3 ± 3.8 follicles for 6-week old mice, 9.3 ± 2.3 follicles for 6-month old mice, and 5.3 ± 1.8 follicles for 10-month old mice.  Moreover, it is important to note that these ovaries were not collected in a specific stage of the estrous cycle, so the number of antral follicles may not be maximal.  In addition, as described in the Materials and Methods, antral follicles were only counted when the oocyte nucleus was present in a section to avoid double counting.  Therefore, this approach (which was applied consistently across samples) could potentially underestimate the total number.

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This manuscript by Bomba-Warczak describes a comprehensive evaluation of long-lived proteins in the ovary using transgenerational radioactive labelled 15N pulse-chase in mice. The transgenerational labeling of proteins (and nucleic acids) with 15N allowed the authors to identify regions enriched in long-lived macromolecules at the 6 and 10-month chase time points. The authors also identify the retained proteins in the ovary and oocyte using MS. Key findings include the relative enrichment in long-lived macromolecules in oocytes, pregranulosa cells, CL, stroma, and surprisingly OSE. Gene ontology analysis of these proteins revealed enrichment for nucleosome, myosin complex, mitochondria, and other matrix-type protein functions. Interestingly, compared to other post-mitotic tissues where such analyses have been previously performed such as the brain and heart, they find a higher fractional abundance of labeled proteins related to the mitochondria and myosin respectively.

Response: We thank the reviewer for this thoughtful summary of our work.  We want to clarify that our pulse-chase strategy relied on a two-generation stable isotope-based metabolic labelling of mice using 15N from spirulina algae (for reference, please see (Fornasiero & Savas, 2023; Hark & Savas, 2021; Savas et al., 2012; Toyama et al., 2013)).  We did not utilize any radioactive isotopes.

Strengths:

A major strength of the study is the combined spatial analyses of LLPs using histological sections with MS analysis to identify retained proteins.

Another major strength is the use of two chase time points allowing assessment of temporal changes in LLPs associated with aging.

The major claims such as an enrichment of LLPs in pregranulosa cells, GCs of primary follicles, CL, stroma, and OSE are soundly supported by the analyses, and the caveat that nucleic acids might differentially contribute to this signal is well presented.

The claims that nucleosomes, myosin complex, and mitochondrial proteins are enriched for LLPs are well supported by GO enrichment analysis and well described within the known body of evidence that these proteins are generally long-lived in other tissues.

Weaknesses:

Comment 1: One small potential weakness is the lack of a mechanistic explanation of if/why turnover may be accelerating at the 6-10 month interval compared to 1-6.

Response 1: At the 6-month time point, we detected more long lived proteins than the 10 month time point in both the ovary and the oocyte.  We anticipated this because proteins are degraded over time, and substantially more time has elapsed at the later time point.  Moreover, at the 6–10-month time point, age-related tissue dysfunction is already evident in the ovary.  For example, in 6-9 month old mice, there is already a deterioration of chromosome cohesion in the egg which results in increased interkinetochore distances (Chiang et al., 2010), and by 10 months, there are multinucleated giant cells present in the ovarian stroma which is consistent with chronic inflammation (Briley et al., 2016).  Thus, the observed changes in protein dynamics may be another early feature of aging progression in the ovary.  

Comment 2: A mild weakness is the open-ended explanation of OSE label retention. This is a very interesting finding, and the claims in the paper are nuanced and perfectly reflect the current understanding of OSE repair. However, if the sections are available and one could look at the spatial distribution of OSE signal across the ovarian surface it would interesting to note if label retention varied by regions such as the CLs or hilum where more/less OSE division may be expected. 

Response 2: We agree that the enrichment of long-lived molecules in the OSE is interesting. To make interpretable conclusions about the dynamics of long-lived molecules in the OSE, we would need to generate a series of samples at precise stages of the estrous cycle or ideally across a timecourse of ovulation to capture follicular rupture and repair.  These samples do not currently exist and are beyond the scope of this study. However, this idea is an important future direction and it has been added to the discussion (lines 221-223). Furthermore, from a practical standpoint, MIMS imaging is resource and time intensive. Thus, we are not able to readily image entire ovarian sections.  Instead, we focused on structures within the ovary and took select images of follicles, stroma, and OSE.  We, therefore, do not have a comprehensive series of images of the OSE from the entire ovarian section for each mouse analyzed.

Reviewer #2 (Public Review):

Summary:

The manuscript by Bomba-Warczak et al. applied multi-isotope imaging mass spectrometry (MIMS) analysis to identify the long-lived proteins in mouse ovaries during reproductive aging, and found some proteins related to cytoskeletal and mitochondrial dynamics persisting for 10 months.

Response: We thank the reviewer for their summary and feedback.

Strengths:

The manuscript provides a useful dataset about protein turnover during ovarian aging in mice.

Weaknesses:

Comment 1: The study is pretty descriptive and short of further new findings based on the dataset. In addition, some results such as the numbers of follicles and ovulated oocytes in aged mice are not consistent with the published literature, and the method for follicle counting is not accurate. The conclusions are not fully supported by the presented evidence.

Response 1: We agree with the reviewer that this study is descriptive. Our goal, as stated, was to use a discovery-based approach to define the long-lived proteome of the ovary and oocyte across a reproductive aging continuum.  As the prominent aging researcher, Dr. James Kirkland, stated: “although ‘descriptive’ is sometimes used as a pejorative term…descriptive or discovery research leading to hypothesis generation has become highly sophisticated and of great relevance to the aging field (Kirkland, 2013).”  We respectfully disagree with the reviewer that our study is short of new findings. In fact, this is the first time that a stable two-generation stable isotope-based metabolic labelling of mice in combination with two different state-of-the-art mass spectrometry methods has been used to identify and localize long lived molecules in the ovary and oocyte along this particular reproductive aging continuum in an unbiased manner.  We have identified proteins groups that were previously not known to be long lived in the ovary and oocyte.  Our hope is that this long-lived proteome will become an important hypothesis-generating resource for the field of reproductive aging.

The age-dependent decline in number of follicles and eggs ovulated in mice has been well established by our group as well as others (Duncan et al., 2017; Mara et al., 2020).  Thus, we are unclear about the reviewer’s comments that our results are not consistent with the published literature.  The absolute numbers of follicles and eggs ovulated as well as the rate of decline with age are highly strain dependent.  Moreover, mice can have a very small ovarian reserve and still maintain fertility (Kerr et al., 2012).  In our study, we saw a consistent age-dependent decrease in the ovarian reserve (Figure 1 – figure supplement 1 D), the number of oocytes collected from large antral follicles following hyperstimulation with PMSG (used for LC-MS/MS), and the number of eggs collected from the oviduct following hyperstimulation and superovulation with PMSG and hCG (Figure 1 – figure supplement 1 E and F).  In all cases, the decline was greater in 10 month old compared to 6 month old mice demonstrating a relative reproductive aging continuum even at these time points.

Our research team has significant expertise in follicle classification and counting as evidenced by our publication record (Duncan et al., 2017; Kimler et al., 2018; Perrone et al., 2023; Quan et al., 2020).  We used our established methods which we have further clarified in the manuscript text (lines 395-397).  Follicle counts were performed on every 5th tissue section of serial sectioned ovaries, and 1 ovary from 3 mice per timepoint were counted. Therefore, follicle counts were performed on an average of 48-62 total sections per ovary. The number of follicles was then normalized per total area (mm2) of the tissue section, and the counts were averaged. Figure 1 – figure supplement 1 C and D represents data averaged from all ovarian sections counted per mouse.   It is important to note that the same criteria were applied consistently to all ovaries across the study, and thus regardless of the technique used, the relative number of follicles or oocytes across ages can be compared.

Reviewer #3 (Public Review):

Summary:

In this study, Bomba-Warczak et al focused on reproductive aging, and they presented a map for long-lived proteins that were stable during reproductive lifespan. The authors used MIMS to examine and show distinct molecules in different cell types in the ovary and tissue regions in a 6 month mice group, and they also used proteomic analysis to present different LLPs in ovaries between these two timepoints in 6-month and 10-month mice. The authors also examined the LLPs in oocytes in the 6-months mice group and indicated that these were nuclear, cytoskeleton, and mitochondria proteins.

Response: We thank the reviewer for their summary and feedback.

Strengths:

Overall, this study provided basic information or a 'map' of the pattern of long-lived proteins during aging, which will contribute to the understanding of the defects caused by reproductive aging.

Weaknesses:

Comment 1: The 6-month mice were used as an aged model; no validation experiments were performed with proteomics analysis only.  

Response 1:  We did not select the 6-month time point to be representative of the “aged model” but rather one of two timepoints on the reproductive aging continuum – 6 and 10 months.  In the manuscript (Figure 1 – figure supplement 1) we have demonstrated the relevance of the two timepoints by illustrating a decrease in follicle counts, number of fully grown oocytes collected, and number of eggs ovulated as well as a tendency towards increased stromal fibrosis (highlighted in the main text lines 78-85).  Inclusion of the 6-month timepoint ultimately turned out to be informative and essential as many long-lived proteins were absent by the 10 month timepoint. These results suggest that important shifts in the proteome occur during mid to advanced reproductive age.  The relevance of these timepoints is mentioned in the discussion (lines 247-270).

Two independent mass spectrometry approaches (MIMS and LC-MS/MS) were used to validate the presence of long-lived macromolecules in the ovary and oocyte. Studies focused on the role of specific long-lived proteins in oocyte and ovarian biology as well as how they change with age in terms of function, turnover, and modification are beyond the scope of the current study but are ongoing.  We have acknowledged these important next steps in the manuscript text (lines 286-288, 311-312).

It is important to note, that oocytes are biomass limited cells, and their numbers decrease with age.  Thus, we had to select ages where we could still collect enough from the mice available to perform LC-MS/MS. 

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Comment 1: The writing and figures are beautiful - it would be hard to improve this manuscript.

Response 1: We greatly appreciate this enthusiastic evaluation of our work.

Comment 2: In Fig S1E/F it would help to list the N number here. Why are there 2 groups at 6-12 wk?

Response 2:  We did not have 6 month and 10-month-old mice available at the same time to be able to run the hyperstimulation and superovulation experiment in parallel.  Therefore, we performed independent experiments comparing the number of eggs collected from either 6-month-old or 10 month old mice relative to 6-12 week old controls.  In each trial, eggs were collected from pooled oviducts from between 3-4 mice per age group, and the average total number of eggs per mouse was reported.  Each point on the graph corresponds to the data from an individual trial, and two trials were performed.  This has been clarified in the figure legend (lines 395-397).  Of note, while addressing this reviewer’s comments, we noticed that we were missing Materials and Methods regarding the collection of eggs from the oviduct following hyperstimulation and superovulation with PMSG and hCG.  This information has now been added in Methods Section, lines 477-481.

Comment 3: The manuscript would benefit from an explanation of why the pups were kept on a 1-month N15 diet after birth, since the oocytes are already labeled before birth, and granulosa at most by day 3-4. Would ZP3 have not been identified otherwise?

Response 3:   The pups used in this study were obtained from fully labeled female dams that were maintained on an15N diet.  These pups had to be kept with their mothers through weaning.  To limit the pulse period only through birth, the pups would have had to be transferred to unlabeled foster mothers.  However, this would have risked pup loss which would have significantly impacted our ability to conduct the studies given that we only had 19 labeled female pups from three breeding pairs.  We have clarified this in the manuscript text in lines 78-80.  It is hard to know, without doing the experiment, whether we would have detected ZP3 if we only labeled through birth.  The expression of ZP3 in primordial follicles, albeit in human, would suggest that this protein is expressed quite early in development.

Comment 4: What is happening to the mitochondria at 6-10 months? Does their number change in the oocyte? Is there a change in the rate of fission? Any chance to take a stab at it with these or other age-matched slides?

Response 4:  The reviewer raises an excellent point.  As mentioned previously in the Discussion (lines 290-301), there are well documented changes in mitochondrial structure and function in the oocyte in mice of advanced reproductive age.  However, there is a paucity of data on the changes that may happen at earlier mid-reproductive age time points.  From the oocyte mitochondrial proteome perspective, our data demonstrate a prominent decline in the persistence of long-lived proteins between 6 and 10 months, and this occurs in the absence of a change in the total pool of mitochondrial proteins (both long and short lived populations) as assessed by spectral counts or protein IDs (figure below).  These data, which we have added into Figure 3 – figure supplement 1 and in the manuscript text (lines 164-170) are suggestive of similar numbers of mitochondria at these two timepoints. It would be informative to do a detailed characterization of oocyte mitochondrial structure and function within this window to see if there is a correlation with this shift in long lived mitochondrial proteins.  Although this analysis is beyond the scope of the current manuscript, it is an important next line of inquiry which we have highlighted in the manuscript text (lines 255-257 and 311-312).

Reviewer #2 (Recommendations For The Authors):

Several concerns are raised as shown below.

Comment 1: In Fig. 2F, it is surprising that ZP3 disappeared in the ovary from mice at the age of 10 months by MIMS analysis, because quite a few oocytes with intact zona pellucida can still be obtained from mice at this age. Notably, ZP would not be renewed once formed.

Response 1: To clarify, Figure 2F shows LC-MS/MS data and not MIMS data.  As mentioned in the Discussion, the detection of long-lived pools of ZP3 at 6 months cannot be derived from newly synthesized zona pellucidae in growing follicles because they would not have been present during the pulse period.  The only way we could detect ZP3 at 6 months is if it forms a primitive zona scaffold in the primordial follicle or if ZPs from atretic follicles of the first couple of waves of folliculogenesis incorporate into the extracellular matrix of the ovary.  The lack of persistence of ZP3 at 10 months could be due to protein degradation. Should ZP3 indeed form a primitive zona, its loss at 10 months would be predicted to result in poor formation of a bona fide zona pellucida upon follicle growth.  Interestingly, aging has been associated with alterations in zona pellucida structure and function.   These data open novel hypotheses regarding the zona pellucida (e.g. a primitive zona scaffold and part of the extracellular matrix) and will require significant further investigation to test. These points are highlighted in the Discussion lines 227-245.

Comment 2: To determine whether those proteins that can not be identified by MIMS at the time point of 10 months are degraded or renewed, the authors should randomly select some of them to examine their protein expression levels in the ovary by immunoblotting analysis.

Response 2: To clarify, proteins were identified by LC-MS/MS and not MIMS which was used to visualize long lived macromolecules.   Each protein will be comprised of old pools (15N containing) and newly synthesized pools (14N containing).  Degradation of the old pool of protein does not mean that there will be a loss of total protein.  Moreover, immunoblotting cannot distinguish old and newly synthesized pools of protein. Where overall peptide counts are listed for each protein identified at both time points.  As peptides derive from proteins, the table provided with the manuscript reflects what immunoblotting would, but on a larger and more precise scale.

Comment 3: I think those proteins that can be identified by MIMS at the time point of 6 months but not 10 months deserve more analyses as they might be the key molecules that drive ovarian aging.

Response 3:  This comment conflicts with comment 2 from Reviewer #3 (Recommendations For The Authors).  This underscores that different researchers will prioritize the value and follow up of such rich datasets differently.  We agree that the LLP identified at 6 months are of particular interest to reproductive aging, and we are planning to follow up on these in future studies.

Comment 4:  Figure 1 – figure supplement 1 C-F, compared with the published literature, the numbers of follicles at different developmental stages and ovulated oocytes at both ages of 6 months and 10 months were dramatically low in this study. For 6-month-old female mice, the reproductive aging just begins, thus these numbers should not be expected to decrease too much. In addition, follicle counting was carried out only in an area of a single section, which is an inaccurate way, because the numbers and types of follicles in various sections differ greatly. Also, the data from a single section could not represent the changes in total follicle counts.

Response 4: We have addressed these points in response to Comment 1 in the Reviewer #2 Public Review, and corresponding changes in the text have been noted.    

Comment 5:  The study lacks follow-up verification experiments to validate their MIMS data.

Response 5: Two independent mass spectrometry approaches (MIMS and LC-MS/MS) were used to validate the presence of long-lived macromolecules in the ovary and oocyte. Studies focused on the role of specific long-lived proteins in oocyte and ovarian biology as well as how they change with age in terms of function, turnover, and modification are beyond the scope of the current study but ongoing.  We have acknowledged these important next steps in the manuscript text (lines 286-288 and 311-312).

Reviewer #3 (Recommendations For The Authors):

Comment 1: The authors used the 6-month mice group to represent the aged model, and examined the LLPs from 1 month to 6 months. Indeed, 6-month-old mice start to show age-related changes; however, for the reproductive aging model, the most widely accepted model is that 10-month-old age mice start to show reproductive-related changes and 12-month-old mice (corresponding to 35-40 year-old women) exhibit the representative reproductive aging phenotypes. Therefore, the data may not present the typical situation of LLPs during reproductive aging.

Response 1: As described in the response to Comment 1 in the Reviewer #3 Public Review, there were clear logistical and technical feasibility reasons why the 6 month and 10-month timepoints were selected for this study.  Importantly, however, these timepoints do represent a reproductive aging continuum as evidenced by age-related changes in multiple parameters.  Furthermore, there were ultimately very few LLPs that remained at 10 months in both the oocyte and ovary, so inclusion of the 6-month time point was an important intermediate.  Whether the LLPs at the 6-month timepoint serve as a protective mechanism in maintaining gamete quality or whether they contribute to decreased quality associated with reproductive aging is an intriguing dichotomy which will require further investigation.  This has been added to the discussion (lines 247-257).

Comment 2:  Following the point above, the authors examined the ovaries in 6 months and 10 months mice by proteomics, and found that 6 months LLPs were not identical compared with 10 months, while there were Tubb5, Tubb4a/b, Tubb2a/b, Hist2h2 were both expressed at these two time points (Fig 2B), why the authors did not explore these proteins since they expressed from 1 month to 10 months, which are more interesting.

Response 2:  The objective of this study was to profile the long-lived proteome in the ovary and oocyte as a resource for the field rather than delving into specific LLPs at a mechanistic level.  That being said, we wholeheartedly agree with the reviewer that the proteins that were identified at both 6 month and 10 months are the most robust and long lived and worthy of prioritizing for further study.  Interestingly, Tubb5 and Tubb4a have high homology to primate-specific Tubb8, and Tubb8 mutations in women are associated with meiosis I arrest in oocytes and infertility (Dong et al., 2023; Feng et al., 2016).  Thus, perturbation of these specific proteins by virtue of their long-lived nature may be associated with impaired function and poor reproductive outcomes.  We have highlighted the importance of these LLPs which are present at both timepoints and persist to at least 10 months in the manuscript text (lines 259-270).

Comment 3:  The authors also need to provide a hypothesis or explanation as to why LLDs from 6 months LLPs were not identical compared with 10 months.

Response 3:  We agree that LLDs identified at 10 months should be also identified as long-lived at 6 months. This is a common limitation of mass spectrometry-based proteomics where each sample is prepared and run individually, which introduces variability between biological replicates, especially when it comes to low abundant proteins. It is key to note that just because we do not identify a protein, it does not mean the protein is not there – it merely means that we were not able to detect it in this particular experiment, but low levels of the protein may still be there. To compensate for this known and inherent variability, we have applied stringent filtering criteria where we required long-lived peptides to be identified in an independent MS scan (alternative is to identify peptide in either heavy or light scan and use modeling to infer FA value based on m/z shift), which gave us peptides of highest confidence. Ideally, these experiments would be done using TMT (tandem mass tag) approach. However, TMT-based experiments typically require substantial amount of input (80-100ug per sample) which unfortunately is not feasible with oocytes obtained from a limited number of pulse-chased animals.  We have added this explanation to the discussion (lines 265-270).

Comment 4:  The reviewer thinks that LLPs from 6 months to 10 months may more closely represent the long-lived proteins during reproductive aging.

Response 4:  We fully agree that understanding the identity of LLPs between the 6 month and 10 month period will be quite informative given that this is a dynamic period when many of LLPs get degraded and thus might be key to the observed decline in reproductive aging. This is a very important point that we hope to explore in future follow-up studies.

Comment 5: The authors used proteomics for the detection of ovaries and oocytes, however, there are no validation experiments at all. Since proteomics is mainly for screening and prediction, the authors should examine at least some typical proteins to confirm the validity of proteomics. For example, the authors specifically emphasized the finding of ZP3, a protein that is critical for fertilization.

Response 5:  Thank you, we agree that closer examination of proteins relevant and critical for fertilization is of importance.  However, a detailed analysis of specific proteins fell outside of the scope of this study which aimed at unbiased identification of long-lived macromolecules in ovaries and oocytes. We hope to continue this important work in near future.

Comment 6: For the oocytes, the authors indicated that cytoskeleton, mitochondria-related proteins were the main LLPs, however, previous studies reported the changes of the expression of many cytoskeleton and mitochondria-related proteins during oocyte aging. How do the authors explain this contrary finding?   

Response 6:  Our findings are not contrary to the studies reporting changes in protein expression levels during oocyte aging – the two concepts are not mutually exclusive. The average FA value at 6-month chase for oocyte proteins is 41.3 %, which means that while 41.3% of long-lived proteins pool persisted for 6 months, the other 58.7% has in fact been renewed. With the exception of few mitochondrial proteins (Cmkt2 and Apt5l), and myosins (Myl2 and Myh7), which had FA values close to 100% (no turnover), most of the LLPs had a portion of protein pools that were indeed turned over. Moreover, we included new data analysis illustrating that we identify comparable number of mitochondrial proteins between the two time points, indicating that while the long-lived pools are changing over time, the total content remains stable (Figure 3 – figure supplement 1E-G).

Comment 7:  The authors also should provide in-depth discussion about the findings of the current study for long-lived proteins. In this study, the authors reported the relationship between these "long-lived" proteins with aging, a process with multiple "changes". Do long-lived proteins (which are related to the cytoskeleton and mitochondria) contribute to the aging defects of reproduction? or protect against aging?

Response 7: This is a very important comment and one that needs further exploration. The fact is – we do not know at this moment whether these proteins are protective or deleterious, and such a statement would be speculative at this stage of research into LLPs in ovaries and oocytes. Future work is needed to address this question in detail.

Briley, S. M., Jasti, S., McCracken, J. M., Hornick, J. E., Fegley, B., Pritchard, M. T., & Duncan, F. E. (2016). Reproductive age-associated fibrosis in the stroma of the mammalian ovary. Reproduction, 152(3), 245-260. https://doi.org/10.1530/REP-16-0129

Chiang, T., Duncan, F. E., Schindler, K., Schultz, R. M., & Lampson, M. A. (2010). Evidence that Weakened Centromere Cohesion Is a Leading Cause of Age-Related Aneuploidy in Oocytes. Current Biology, 20(17), 1522-1528. https://doi.org/10.1016/j.cub.2010.06.069

Dong, J., Jin, L., Bao, S., Chen, B., Zeng, Y., Luo, Y., Du, X., Sang, Q., Wu, T., & Wang, L. (2023). Ectopic expression of human TUBB8 leads to increased aneuploidy in mouse oocytes. Cell Discov, 9(1), 105. https://doi.org/10.1038/s41421-023-00599-z

Duncan, F. E., Jasti, S., Paulson, A., Kelsh, J. M., Fegley, B., & Gerton, J. L. (2017). Age-associated dysregulation of protein metabolism in the mammalian oocyte. Aging Cell, 16(6), 1381-1393. https://doi.org/10.1111/acel.12676

Feng, R., Sang, Q., Kuang, Y., Sun, X., Yan, Z., Zhang, S., Shi, J., Tian, G., Luchniak, A., Fukuda, Y., Li, B., Yu, M., Chen, J., Xu, Y., Guo, L., Qu, R., Wang, X., Sun, Z., Liu, M., . . . Wang, L. (2016). Mutations in TUBB8 and Human Oocyte Meiotic Arrest. N Engl J Med, 374(3), 223-232. https://doi.org/10.1056/NEJMoa1510791

Fornasiero, E. F., & Savas, J. N. (2023). Determining and interpreting protein lifetimes in mammalian tissues. Trends Biochem Sci, 48(2), 106-118. https://doi.org/10.1016/j.tibs.2022.08.011

Hark, T. J., & Savas, J. N. (2021). Using stable isotope labeling to advance our understanding of Alzheimer's disease etiology and pathology. J Neurochem, 159(2), 318-329. https://doi.org/10.1111/jnc.15298

Kerr, J. B., Hutt, K. J., Michalak, E. M., Cook, M., Vandenberg, C. J., Liew, S. H., Bouillet, P., Mills, A., Scott, C. L., Findlay, J. K., & Strasser, A. (2012). DNA damage-induced primordial follicle oocyte apoptosis and loss of fertility require TAp63-mediated induction of Puma and Noxa. Mol Cell, 48(3), 343-352. https://doi.org/10.1016/j.molcel.2012.08.017

Kimler, B. F., Briley, S. M., Johnson, B. W., Armstrong, A. G., Jasti, S., & Duncan, F. E. (2018). Radiation-induced ovarian follicle loss occurs without overt stromal changes. Reproduction, 155(6), 553-562. https://doi.org/10.1530/REP-18-0089

Kirkland, J. L. (2013). Translating advances from the basic biology of aging into clinical application. Exp Gerontol, 48(1), 1-5. https://doi.org/10.1016/j.exger.2012.11.014

Mara, J. N., Zhou, L. T., Larmore, M., Johnson, B., Ayiku, R., Amargant, F., Pritchard, M. T., & Duncan, F. E. (2020). Ovulation and ovarian wound healing are impaired with advanced reproductive age. Aging (Albany NY), 12(10), 9686-9713. https://doi.org/10.18632/aging.103237

Perrone, R., Ashok Kumaar, P. V., Haky, L., Hahn, C., Riley, R., Balough, J., Zaza, G., Soygur, B., Hung, K., Prado, L., Kasler, H. G., Tiwari, R., Matsui, H., Hormazabal, G. V., Heckenbach, I., Scheibye-Knudsen, M., Duncan, F. E., & Verdin, E. (2023). CD38 regulates ovarian function and fecundity via NAD(+) metabolism. iScience, 26(10), 107949. https://doi.org/10.1016/j.isci.2023.107949

Quan, N., Harris, L. R., Halder, R., Trinidad, C. V., Johnson, B. W., Horton, S., Kimler, B. F., Pritchard, M. T., & Duncan, F. E. (2020). Differential sensitivity of inbred mouse strains to ovarian damage in response to low-dose total body irradiationdagger. Biol Reprod, 102(1), 133-144. https://doi.org/10.1093/biolre/ioz164

Savas, J. N., Toyama, B. H., Xu, T., Yates, J. R., 3rd, & Hetzer, M. W. (2012). Extremely long-lived nuclear pore proteins in the rat brain. Science, 335(6071), 942. https://doi.org/10.1126/science.1217421

Toyama, B. H., Savas, J. N., Park, S. K., Harris, M. S., Ingolia, N. T., Yates, J. R., 3rd, & Hetzer, M. W. (2013). Identification of long-lived proteins reveals exceptional stability of essential cellular structures. Cell, 154(5), 971-982. https://doi.org/10.1016/j.cell.2013.07.037

Author response:

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

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

This is a fine paper that serves the purpose to show that the use of light sheet imaging may be used to provide whole brain imaging of axonal projections. The data provided suggest that at this point the technique provides lower resolution than with other techniques. Nonetheless, the technique does provide useful, if not novel, information about particular brain systems. 

Strengths: 

The manuscript is well written. In the introduction a clear description of the functional organization of the barrel cortex is provided provides the context for applying the use of specific Cre-driver lines to map the projections of the main cortical projection types using whole brain neuroanatomical tracing techniques. The results provided are also well written, with sufficient detail describing the specifics of how techniques were used to obtain relevant data. Appropriate controls were done, including the identification of whisker fields for viral injections and determination of the laminar pattern of Cre expression. The mapping of the data provides a good way to visualize low resolution patterns of projections. 

Weaknesses: 

(1) The results provided are, as stated in the discussion, "largely in agreement with previously reported studies of the major projection targets". However it must be stated that the study does not "extend current knowledge through the high sensitivity for detecting sparse axons, the high specificity of labeling of genetically defined classes of neurons and the brain wide analysis for assigning axons to detailed brain regions" which have all been published in numerous other studies. ( the allen connectivity project and related papers, along with others). If anything the labeling of axons obtained with light sheet imaging in this study does not provide as detailed mapping obtained with other techniques. Some detail is provided of how the raw images are processed to resolve labeled axons, but the images shown in the figures do not demonstrate how well individual axons may be resolved, of particular interest would be to see labeling in terminal areas such as other cortical areas, striatum and thalamus. As presented the light sheet imaging appears to be rather low resolution compared to the many studies that have used viral tracing to look at cortical projections from genetically identified cortical neurons. 

We agree with the reviewer that the resolution of imaging should be further improved in future studies, as also mentioned in the original manuscript. On P. 17 of the revised manuscript we write “Probably most important for future studies is the need to increase the light-sheet imaging resolution perhaps combined with the use of expansion microscopy to provide brain-wide micron-resolution data (Glaser et al., 2023; Wassie et al., 2019).” However, even at somewhat lower resolution, through bright sparse labelling, individual axonal segments can nonetheless be traced through machine learning to define axonal skeletons, whose length can be quantified as we do in this study. This methodology highlights sparse wS1 and wS2 innervation of a large number of brain areas, some of which are not typically considered, and our anatomical results might therefore help the neuronal circuit analysis underlying various aspects of whisker sensorimotor processing. Despite impressive large-scale projection mapping projects such as the Allen connectivity atlas, there remains relatively sparse cell typespecific projection map data for the representations of the large posterior whiskers in wS1 and wS2, and our data in this study thus adds to a growing body of cell-type specific projection mapping with the specific focus on the output connectivity of these whisker-related neocortical regions of sensory cortex.

In the revised manuscript, we now provide an additional supplementary figure (Figure 1 – figure supplement 2) showing examples of the axonal segmentation from further additional image planes including branching axons in the key innervation regions mentioned by the reviewer, namely “other cortical areas, striatum and thalamus”.

(2) Amongst the limitations of this study is the inability to resolve axons of passage and terminal fields. This has been done in other studies with viral constructs labeling synaptophysin. This should be mentioned. 

The reviewer brings up another important point for future methodological improvements to enhance connectivity mapping. Indeed, we already mentioned this in our original submission near the end of the first paragraph under the Limitations and future perspectives section. In the revised manuscript on P. 17, we write “Future studies should also aim to identify neurotransmitter release sites along the axon, which could be achieved by fluorescent labeling of prominent synaptic components, such as synaptophysin-GFP (Li et al., 2010).”

(3) There is no quantitative analysis of differences between the genetically defined neurons projecting to the striatum, what is the relative area innervated by, density of terminals, other measures. 

The reviewer raises an interesting question, and in the revised manuscript, we now present a more detailed analysis of cell class-specific axonal projections focusing specifically on the striatum. Following the reviewer’s suggestion, in a new supplementary figure (Figure 7 – figure supplement 1), we now report spatial axonal density maps in the striatum from SSp-bfd and SSs, finding potentially interesting differences comparing the projections of Rasgrf2-L2/3, Scnn1a-L4 and Tlx3-L5IT neurons. On P. 12 of the revised manuscript, we now write “We also investigated the spatial innervation pattern of Rasgrf2-L2/3, Scnn1a-L4 and Tlx3-L5IT neurons in the striatum (Figure 7 – figure supplement 1), where we found that axonal density from Rasgrf2-L2/3 neurons in both SSp-bfd and SSs was concentrated in a posterior dorsolateral part of the ipsilateral striatum, whereas Tlx3-L5IT neurons had extensive axonal density across a much larger region of the striatum, including bilateral innervation by SSp-bfd neurons. Striatal innervation by Scnn1a-L4 neurons was intermediate between Rasgrf2-L2/3 and Tlx3-L5IT neurons.” We think the reviewer’s comment has helped reveal further interesting aspects of our data set, and we thank the reviewer.

(4) Figure 5 is an example of the type of large sets of data that can be generated with whole brain mapping and registration to the Allen CCF that provides information of questionable value. Ordering the 50 plus structures by the density of labeling does not provide much in terms of relative input to different types of areas. There are multiple subregions for different functional types ( ie, different visual areas and different motor subregions are scattered not grouped together. Makes it difficult to understand any organizing principles.

We agree with the reviewer, and fully support the importance of considering subregions within the relatively coarse compartmentalization of the current Allen CCF. In order to provide some further information about connectivity that may help give the reader further insights into the data, we have now added further quantification of cortex-specific axonal density ranked according to functional subregions in a new supplementary figure (Figure 5 – figure supplement 2). 

(5) The GENSAT Cre driver lines used must have the specific line name used, not just the gene name as the GENSAT BAC-Cre lines had multiple lines for each gene and often with very different expression patterns. Rbp4_KL100, Tlx3_PL56, Sim1_KJ18, Ntsr1_ GN220. 

In the revised manuscript, we now write out a fuller description of the mouse lines the first time they are mentioned in the Results section on P. 7. The full mouse line names, accession numbers and references were of course already described in the methods section, which remains the case in the revised manuscript.

Reviewer #2 (Public Review): 

Summary: 

This study takes advantage of multiple methodological advances to perform layer-specific staining of cortical neurons and tracking of their axons to identify the pattern of their projections. This publication offers a mesoscale view of the projection patterns of neurons in the whisker primary and secondary somatosensory cortex. The authors report that, consistent with the literature, the pattern of projection is highly different across cortical layers and subtype, with targets being located around the whole brain. This was tested across 6 different mouse types that expressed a marker in layer 2/3, layer 4, layer 5 (3 sub-types) and layer 6.  Looking more closely at the projections from primary somatosensory cortex into the primary motor cortex, they found that there was a significant spatial clustering of projections from topographically separated neurons across the primary somatosensory cortex. This was true for neurons with cell bodies located across all tested layers/types. 

Strengths: 

This study successfully looks at the relevant scale to study projection patterns, which is the whole brain. This is achieved thanks to an ambitious combination of mouse lines, immunohistochemistry, imaging and image processing, which results in a standardized histological pipeline that processes the whole-brain projection patterns of layer-selected neurons of the primary and secondary somatosensory cortex. 

This standardization means that comparisons between cell-types projection patterns are possible and that both the large-scale structure of the pattern and the minute details of the intra-areas pattern are available. 

This reference dataset and the corresponding analysis code are made available to the research community. 

Weaknesses: 

One major question raised by this dataset is the risk of missing axons during the postprocessing step. Indeed, it appears that the control and training efforts have focused on the risk of false positives (see Figure 1 supplementary panels). And indeed, the risk of overlooking existing axons in the raw fluorescence data id discussed in the article. 

Based on the data reported in the article, this is more than a risk. In particular, Figure 2 shows an example Rbp4-L5 mouse where axonal spread seems massive in Hippocampus, while there is no mention of this area in the processed projection data for this mouse line. 

In Figure 2, we show the expression of tdTomato in double-transgenic mice in which the Cre-driver lines were crossed with a Cre-dependent reporter mouse expressing cytosolic tdTomato. In addition to the specific labelling of L5PT neurons in the somatosensory cortex, Rbp4-Cre mice also express Cre-recombinase in other brain regions including the hippocampus. In the reporter mice crossed with Rbp4-Cre mice, tdTomato is expressed in neurons with cell bodies in the hippocampus which is clearly visualized in Figure 2. Because our axonal labelling is based on localized viral vector expression of tdTomato in SSp-bfd and SSs, the expression of Cre in hippocampus does not affect our analysis. In order to clarify to the reader, in the legend to Figure 2D, we now specifically write “As for panel A, but for Rbp4-L5 neurons. Note strong expression of Cre in neurons with cell bodies located in the hippocampus, which does not affect our analysis of axonal density based on virus injected locally into the neocortex.” Consistent with this observation, the Allen Institute’s ISH data support

expression of Rbp4 in neurons of the hippocampus e.g. https://mouse.brainmap.org/gene/show/19425 and https://mouse.brainmap.org/experiment/show/68632655.

Similarily, the Ntsr1-L6CT example shows a striking level of fluorescence in Striatum, that does not reflect in the amount of axons that are detected by the algorithms in the next figures.  These apparent discrepancies may be due to non axonal-specific fluorescence in the samples. In any case, further analysis of such anatomical areas would be useful to consolidate the valuable dataset provided by the article. 

As pointed out above, Figure 2 shows cytosolic tdTomato fluorescence in transgenic crosses of the Cre-driver mice with Cre-dependent tdTomato reporter mice. For the Ntsr1-Cre x LSL-tdTomato mice, all corticothalamic L6CT neurons from across the entire cortex drive tdTomato expression. The axon of each neuron must traverse the striatum giving rise to fluorescence in the striatum. As discussed above, labelling of synaptic specialisations will be important in future studies to separate travelling axon from innervating axon. However, the overall impact of the axons traversing the striatum is again mitigated in our study by considering the axonal projections from local sparse infections in SSp-bfd and SSs rather than from cortex-wide tdTomato expression.

Reviewer #3 (Public Review): 

Summary: 

The paper offers a systematic and rigorous description of the layer-and sublayer specific outputs of the somatosensory cortex using a modern toolbox for the analysis of brain connectivity which combines: 1) Layer-specific genetic drivers for conditional viral tracing; 2) whole brain analyses of axon tracts using tissue clearing and imaging; 3) Segmentation and quantification of axons with normalization to the number of transduced neurons; 4) registration of connectivity to a widely used anatomical reference atlas; 5) functional validation of the connectivity using optogenetic approaches in vivo. 

Strengths: 

Although the connectivity of the somatosensory cortex is already known, precise data are dispersed in different accounts (papers, online resources,) using different methods. So the present account has the merit of condensing this information in one very precisely documented report. It also brings new insights on the connectivity, such as the precise comparison of layer specific outputs, and of the primary and secondary somatosensory areas. It also shows a topographic organization of the circuits linking the somatosensory and motor cortices. The paper also offers a clear description of the methodology and of a rigorous approach to quantitative anatomy. 

Weaknesses: 

The weakness relates to the intrinsic limitations of the in toto approaches, that currently lack the precision and resolution allowing to identify single axons, axon branching or synaptic connectivity. These limitations are identified and discussed by the authors. 

We agree with the reviewer.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

No additional comment 

OK

Reviewer #2 (Recommendations For The Authors): 

In Figure 8, we don't get to see much raw data, while the diversity of functional responses pattern to the primary and supplementary S1 activations is highly intriguing (and this diversity exists as suggested by the results in Figure 8E, LRPT). 

Can Figure 8C be less blurred? Maybe give more space to individual examples, such as an overlay of the delineations of the activated area across the tested mice? 

Also, can we have a view on the time dynamics of the functional activation and integration window? 

Raw data - We have now added a new supplementary figure (Figure 8 – figure supplement 1) to show data from individual mice, as well as plotting the time-course of the evoked jRGECO fluorescence signals in the frontal cortex hotspot. 

Image blur - Each pixel represents 62.5 x 62.5 um on the cortical surface. The images in Figure 8B&C were averaged across mice, which causes some additional spatial blurring. However, the most likely explanation for the ‘blurred’ impression, is the overall large horizontal extent of the axonal innervation as well as likely rapid lateral spread of excitation both at the stimulation area and in the target region, as for example also indicated in rapid voltage-sensitive imaging experiments (Ferezou et al., 2007).  

Reviewer #3 (Recommendations For The Authors): 

At the time being, the abstract is really centred on the methodology which is no longer very novel as it has actually been already been described previously by other groups. In my view the paper would gain visibility, and be a useful tool for the community if amended to better point out the significant results of the study, for instance, i) the layer and sub-layer specificity of the outputs, using the listed genetic drivers; ii) the comparison of primary and secondary somatosensory areas, iii) the functional validation. The layer specificity of each cre- line should be indicated in the abstract. 

We have tried to improve the writing of the abstract along the lines suggested by the reviewer. Specifically, we have now added layer and projection class of the various Cre-lines, and we now also highlight the most obvious differences in the innervation patterns.

There is some degree of redundancy in the description in the result section. One suggestion, for an easier flow of reading, would be to join the paragraphs " Laminar characterization of the Cre-lines.." and: "Axonal projections...". Start for each Cre-line with a description of the laminar localisation of recombination in the somatosensory cortices, followed therefrom by the description of outputs from SSp-bfd and SSs; Then the general description/overview of the outputs can be summarized as a legend to Figure 5-supplementary 2, which could appear as a main figure. 

Although we agree with the reviewer that there is some level of redundancy in the text, the results of the characterization of the Cre-line (Figure 2) is quite a different experiment compared to the viral injections described in other figures, and we therefore prefer to keep these sections separate.

Other minor points: 

In the text; Indicate the genetic background of the transgenic mouse lines. 

On P. 18, we now indicate that all mice were “back-crossed with C57BL/6 mice”.

Keep consistency in the designation of the areas, S1 appears sometimes as SSp-bfd or as SSp 

We thank the reviewer for pointing out the inconsistent nomenclature, which we have now corrected in the revised manuscript. ‘SSp’ remains used on P. 9 and P. 16 of the revised manuscript to indicate a region including SSp-bfd but also extending beyond.

Figure 1 supplement 2 is not really necessary to show (as the viral tools have previously been validated) can just be stated in the text. Conversely one would like to see a higher resolution image of the injection sites that allowed to do the cell counts used for normalization, as this can be pretty tricky. 

In response to the reviewer’s suggestion, we have now added a new supplemental figure to show an example of how cells in the injection site were counted (Figure 1 – figure supplement 3).

Figure 2: the most important here is the higher magnification to show the precise laminar localisation of the recombination, rather than the atlas landmarks that is already shown in Figure 1. This would allow more space for clearer higher magnification panels comparing SSs and SSp. The present image hints to some real differences, but difficult to appreciate with the current resolution. The legend should also comment on the labelling seen in layer 1, in the Tlx2 and Rbp4 lines. Could be dendritic labelling, but this needs a word of clarification.

We think both the overview images as well as the high-resolution images are of value to the reader. Following the reviewer’s comment, in the legends to Figure 2C&D, we have now added text suggesting that the layer 1 fluorescence is likely axonal or dendritic in origin : “Labelling in layer 1 is likely of axonal or dendritic origin, and no cell bodies were labelled in this layer.” In addition, we have added a new supplemental figure which shows the cortical labelling in SSp and SSS in a more magnified view (Figure 2 – figure supplement 1).

Figure 3: the comparison of the 3 transgenic lines labelling layer 5 and showing sublaminar identities is really interesting in showing the heterogeneity of this layer and possible regional differences. However, the cases shown for illustration for Rbp4 and Tlx3 seem pretty massive in comparison with the other drivers. Maybe cases with smaller injections could be chosen for illustration. 

Figure 3 shows grand average axonal density maps across different mice normalized to the number of neurons in the injection site. The large amount of axon per neuron observed in Rbp4 and Tlx3 mice therefore shows their long, wide-ranging axons compared to other neuronal classes.

Figure 6A could be a supplementary figure in my view; 6B is clearer. 

We think both representations are useful, and we think different readers might better appreciate either of the two analyses.

Author response:

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

eLife assessment

This work is potentially useful because it has generated a mineable yield of new candidate immune inhibitory receptors, which can serve both as drug targets and as subjects for further biological investigation. It is noted however that the argument of the work is rather incomplete, in that it does very little to validate the putative new receptors, and merely makes a study of their putative distribution across cell types. Experimental follow-up to demonstrate the claimed properties for the proteins identified, or mining existing experimental data sources on gene expression across tissues to at least show that the pipeline correctly identified genes likely to be specific to immune cells (or something along these lines), would make this work more complete and compelling. 

We thank the editors for their critical reading and assessment of our manuscript. We acknowledge that the present study is limited by a lack of experimental follow-up. However, we purposely chose to make this pipeline of putative novel inhibitory receptors public at this early stage for our work to be a starting point for further functional investigation of these targets by the scientific community.   

Public Reviews:

Reviewer #1 (Public Review):

This manuscript proposes a new bioinformatics approach identifying several hundreds of previously unknown inhibitory immunoreceptors. When expressed in immune cells (such as neutrophils, monocytes, CD8+, CD4+, and T-cells), such receptors inhibit the functional activity of these cells. Blocking inhibitory receptors represents a promising therapeutic strategy for cancer treatment.

As such, this is a high-quality and important bioinformatics study. One general concern is the absence of direct experimental validation of the results. In addition to the fact that the authors bioinformatically identified 51 known receptors, providing such experimental evaluation (of at least one, or better few identified receptors) would, in my opinion, significantly strengthen the presented evidence.

I will now briefly summarize the results and give my comments.

First, using sequence comparison analysis, the authors identify a large set of putative receptors based on the presence of immunoreceptor tyrosine-based inhibitory motifs (ITIMs), or immunoreceptor tyrosinebased switch motifs (ITSMs). They further filter the identified set of receptors for the presence of the ITIMs or ITSMs in an intracellular domain of the protein. Second, using AlphaFold structure modeling, the authors select only receptors containing ITIMs/ITSMs in structurally disordered regions. Third, the evaluation of gene expression profiles of known and putative receptors in several immune cell types was performed. Fourth, the authors classified putative receptors into functional categories, such as negative feedback receptors, threshold receptors, threshold disinhibition, and threshold-negative feedback. The latter classification was based on the available data from Nat Rev Immunol 2020. Fifth, using publicly available single-cell RNA sequencing data of tumor-infiltrating CD4+ and CD8+ cells from nearly twenty types of cancer, the authors demonstrate that a significant fraction of putative receptors are indeed expressed in these datasets.

In summary, in my opinion, this is an interesting, important, high-quality bioinformatics work. The manuscript is clearly written and all technical details are carefully explained.

One comment/suggestion regarding the methodology of evaluating gene expression profiles of putative receptors: perhaps it might be important to look at clusters of genes that are co-expressed with putative inhibitory receptors. 

We thank the reviewer for their comments and suggestions.  We acknowledge that looking at co-expressed genes and subsequently at gene ontology enrichment could be an interesting approach to prioritize the inhibitory receptors. However, since there are many ways to approach the results of the gene coexpression networks, which also depend on the cell type and activation status of interest, we have chosen to discuss the implications of these networks in the discussion with the following paragraph, rather than reporting all these different approaches in the paper:

“To further prioritize inhibitory receptors in immune cell subsets or diseases of interest, gene coexpression networks of putative inhibitory receptors could be assessed. On the one hand, the cooccurrence of putative inhibitory receptors with known inhibitory receptors within a module could be one approach, while on the other hand the presence of putative inhibitory receptors in a different module could suggest novel regulation of different biological functions than the known receptors. The location of the putative inhibitory receptors in the network could also change depending on the cell type and the activation status of the cell. Additionally, one could look at the co-expression of candidates with other genes within a gene module to look at potential biological function, and at co-expression with signalling molecules known to interact with inhibitory receptors, such as Csk, SHP-1, SHP-2 and SHIP1, although their regulation might be more post-translationally regulated rather than at mRNA level.”

Reviewer #2 (Public Review):

Summary:

The authors developed a bioinformatic pipeline to aid the screening and identification of inhibitory receptors suitable as drug targets. The challenge lies in the large search space and lack of tools for assessing the likelihood of their inhibitory function. To make progress, the authors used a consensus protein membrane topology and sequence motif prediction tool (TOPCOS) combined with both a statistical measure assessing their likelihood function and a machine learning protein structural prediction model (AlphaFold) to greatly cut down the search space. After obtaining a manageable set of 398 high-confidence known and putative inhibitory receptors through this pipeline, the authors then mapped these receptors to different functional categories across different cell types based on their expression both in the resting and activated state. Additionally, by using publicly available pan-cancer scRNA-seq for tumor-infiltrating T-cell data, they showed that these receptors are expressed across various cellular subsets.

Strengths:

The authors presented sound arguments motivating the need to efficiently screen inhibitory receptors and to identify those that are functional. Key components of the algorithm were presented along with solid justification for why they addressed challenges faced by existing approaches. To name a few:

• TOPCON algorithm was elected to optimize the prediction of membrane topology.

• A statistical measure was used to remove potential false positives.

• AlphaFold is used to filter out putative receptors that are low confidence (and likely intrinsically disordered).

To examine receptors screened through this pipeline through a functional lens, the authors proposed to look at their expression of various immune cell subsets to assign functional categories. This is a reasonable and appropriate first step for interpreting and understanding how potential drug targets are differentially expressed in some disease contexts.

Weaknesses:

The paper has strength in the pipeline they presented, but the weakness, in my opinion, lies in the lack of concrete demonstration on how this pipeline can be used to at least "rediscover" known targets in a

disease-specific manner. For example, the result that both known and putative immune inhibitory receptors are expressed across a wide variety of tumor-infiltrating T-cell subsets is reassuring, but this would have been more informative and illustrative if the authors could demonstrate using a disease with known targets, as opposed to a pan-cancer context. Additionally, a discussion that contrasts the known and putative receptors in the context above would help readers better identify use cases suitable for their research using this pipeline. Particularly,

• For known receptors, does the pipeline and the expression analysis above rediscover the known target in the disease of interest?

• For putative receptors, what do the functional category mapping and the differential expression across various tumor-infiltrating T-cell subsets imply on a potential therapeutic target?

We thank the reviewer for their assessment and comments. The primary purpose of the bioinformatics pipeline was to identify putative inhibitory receptors in a disease-agnostic manner and allow the scientific community to further explore targets in their specific diseases of interest. We performed our pan-cancer expression analysis as a preliminary proof of concept and agree that exploring targets in specific diseases, cancer or otherwise, could be more informative. To validate that we rediscovered known immunotherapeutic targets, we analyzed the expression of known inhibitory receptors on tumorinfiltrating T cells of melanoma patients using the same dataset as figure 3. We find high expression of known therapeutic targets, such as PD-1, in addition to other known inhibitory receptors that are being targeted in clinical trials, one of which being TIGIT. We have added this information to the results section and added the corresponding graph as supplementary figure 5. 

For the putative inhibitory receptors, we believe the functional categorization can assist in selecting targets that are more likely to be successful in a therapeutic context. As we previously proposed in our perspective on functional categorization of inhibitory receptors (Rumpret et al., Nat Imm, 2020), it might be beneficial to target inhibitory receptors of different functional categories in cancer immunotherapy. Targeting a threshold receptor to lower the threshold for activation and a negative feedback receptor to lengthen and strengthen the cellular response might therefore be more effective than targeting two receptors of a single functional category. Even though we realize RNA sequencing data of in vitro stimulated immune cells is not identical to data from TILs, we have tried to characterize the functional categories expressed by TILs by extrapolating the defined functional categorization per gene from figure 2, and added the corresponding graphs as supplementary figure 4. This shows that mainly threshold receptors and some (threshold-)negative feedback receptors are expressed by the different T cell subsets, which would open the possibility of using the proposed therapeutic strategy of targeting different functional categories. However, we acknowledge that this will require further validation of expression patterns in vivo in different cancers and immune cell subsets. 

Reviewer #1 (Recommendations For The Authors):

One comment/suggestion regarding the methodology of evaluating gene expression profiles of putative receptors: perhaps it might be important to look at clusters of genes that are co-expressed with putative inhibitory receptors.

See our reply to the suggestion above.

Reviewer #2 (Recommendations For The Authors):

Results section

(a) "Putative ITIM/ITSM-bearing immune inhibitory receptors can be found in the human genome"

i. Figure 1 could benefit from additional labeling. For example, in B, the grey line indicates 5%, etc. Additionally, in panel B&C, I assume by "predicted" the author meant using TOPCONS?

ii. Figure 1B doesn't seem to be consistent with this sentence "However, for 10 out of 51, we observed ITIM/ITSM sequences in the permutated sequence up to ~25% of the time" [page 2, line 1-3], as all 51 data points in Figure 1B (under "Known" panel) are below the 0.25 horizontal line?

i. We have adjusted the figure legend to better indicate the information provided in the figures. The predicted genes are all unknown transmembrane candidates that contain an ITIM or ITSM in their intracellular domain, as determined using TOPCONS.

ii. Due to the nature of permutation testing, there is some variation in the individual likelihood values for each protein sequence. However, as they were generally below 0.25 in any given iteration, we decided to define this value as a threshold for inclusion. 

(b) "AlphaFold structure predictions can assist in identifying likely functional ITIM/ITSMs"

i. Readability would increase if the author indicate how pLDDT score is computed and in what range is it (between 0 and 100.)

ii. Third paragraph. Can the author comment on why 80 pLDDT is chosen as the cutoff? The first sentence of this paragraph states "We found that 99 out of 101 ITIM/ITSMs of the 51 known receptors had low confidence score, i.e., less than 80 pLDDT, with an average confidence score of 49.3 pLDDT..." However, it was later stated in the Discussion, page 10, starting Line 11 "We determined a threshold of 80 pLDDT based on the average prediction scores of the ITIM/ITSMs in known inhibitory receptors....". If 99 out of 101 ITIM/ITSMs had pLDDT<80, then it seems strange that the average of the 101 is at 80pLDDT, even in the extreme where the remaining 101-99=2 ITIM/ITSMs attain the maximum pLDDT score at 100, unless the distribution of those 99 is narrowly centered around 80? A distribution of the pLDDT would help clarify.

i. The pLDDT scores are computed by AlphaFold as a way to determine how well a specific residue and/or region is expected to be modelled in three-dimensional space. We now refer to the corresponding AlphaFold publications and references therein to clarify this (10.1093/nar/gkab1061, 10.1038/s41586021-03819-2, 10.1093/bioinformatics/btt473). We also have now included the range (i.e., 0-100) in the text.

ii. The threshold of 80 pLDDT was chosen as this still encompasses all known inhibitory receptors and was not calculated based on an average of the prediction scores. In this way, we still included ITIM/ITSMs with a relatively high pLDDT, such as those observed in PD-1 and LAIR-1. The previous text ‘average prediction scores of the ITIM/ITSMs in known inhibitory receptors’ referred to the averaging of the confidence score for each of the six amino acids encompassing the ITIM/ITSM into one overall score per ITIM/ITSM. We have adjusted the text to better reflect this.

(c) "Putative inhibitory receptors are expressed across immune cell subsets"

Figure S2, the last sentence in the caption (relevant for panel C) states "Cell subsets without uniquely expressed putative inhibitory receptors i.e., B cells and T cell, are excluded from the panel for clarity", but B cells and T cells are present in panel C?

Indeed, but they are only included for the cases where the cell subsets share receptor expression with other immune cell subsets. The B and T cells do not express any unique putative multi-spanning receptors, all receptors are shared with at least one other immune cell subset. 

(d) "Known and putative inhibitory receptors are expressed on tumour infiltrating T cells"

i. Missing panel C label in Figure 3 and S3.

ii. By comparing Figure 3 and S3, it looks to me that there's not a big difference between single-spanning and multi-spanning inhibitory receptors. I wonder if the authors can comment or speculate on this similarity in addition to differences of expression among T-cell subsets. Would the similarities and differences above be explained by cancer type?

i. Figure 3 and S3 do not contain a panel C, but panel B consists of a lower (CD8+) and an upper (CD4+) subpanel, we have more clearly indicated this in the figure legend in the revised manuscript. 

ii. While some T cell subsets, such as exhausted CD8+ T cells and CD4+ regulatory T cells, appear to not differ much in their expression of either single- or multi-spanning receptors, we do observe that, for example, effector memory CD4+ T cells or EMRA CD8+ T cells express single-spanning inhibitory receptors to a higher extent than multi-spanning inhibitory receptors. It is possible that these differences and similarities reflect some of the roles multi-spanning inhibitory receptors could play in regulating immune cells, for example in response to chemokines, as many chemokine receptors are multi-spanning proteins. 

Data and Code availability

Although the Methods section provides some context for the computational analysis and citations for relevant data, software availability and a data availability statement are lacking.

We have included a data availability statement to the data files and code in the revised manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript by Jang et al. describes the application of new methods to measure the localization of GTP-binding signaling proteins (G proteins) on different membrane structures in a model mammalian cell line (HEK293). G proteins mediate signaling by receptors found at the cell surface (GPCRs), with evidence from the last 15 years suggesting that GPCRs can induce G-protein mediated signaling from different membrane structures within the cell, with variation in signal localization leading to different cellular outcomes. While it has been clearly shown that different GPCRs efficiently traffic to various intracellular compartments, it is less clear whether G proteins traffic in the same manner, and whether GPCR trafficking facilitates "passenger" G protein trafficking. This question was a blind spot in the burgeoning field of GPCR localized signaling in need of careful study, and the results obtained will serve as an important guidepost for further work in this field. The extent to which G proteins localize to different membranes within the cell is the main experimental question tested in this manuscript. This question is pursued through two distinct methods, both relying on genetic modification of the G-beta subunit with a tag. In one method, G-beta is modified with a small fragment of the fluorescent protein mNG, which combines with the larger mNG fragment to form a fully functional fluorescent protein to facilitate protein trafficking by fluorescent microscopy. This approach was combined with the expression of fluorescent proteins directed to various intracellular compartments (different types of endosomes, lysosome, endoplasmic reticulum, Golgi, mitochondria) to look for colocalization of G-beta with these markers. These experiments showed compelling evidence that G-beta co-localizes with markers at the plasma membrane and the lysosome, with weak or absent co-localization for other markers. A second method for measuring localization relied on fusing G-beta with a small fragment from a miniature luciferase (HiBit) that combines with a larger luciferase fragment (LgBit) to form an active luciferase enzyme. Localization of Gbeta (and luciferase signal) was measured using a method known as bystander BRET, which relies on the expression of a fluorescent protein BRET acceptor in different cellular compartments. Results using bystander BRET supported findings from fluorescence microscopy experiments. These methods for tracking G protein localization were also used to probe other questions. The activation of GPCRs from different classes had virtually no impact on the localization of G-beta, suggesting that GPCR activation does not result in the shuttling of G proteins through the endosomal pathway with activated receptors.

Strengths:

The question probed in this study is quite important and, in my opinion, understudied by the pharmacology community. The results presented here are an important call to be cognizant of the localization of GPCR coupling partners in different cellular compartments. Abundant reports of endosomal GPCR signaling need to consider how the impact of lower G protein abundance on endosomal membranes will affect the signaling responses under study.

The work presented is carefully executed, with seemingly high levels of technical rigor. These studies benefit from probing the experimental questions at hand using two different methods of measurement (fluorescent microscopy and bystander BRET). The observation that both methods arrive at the same (or a very similar) answer inspires confidence about the validity of these findings.

Weaknesses:

The rationale for fusing G-beta with either mNG2(11) or SmBit could benefit from some expansion. I understand the speculation that using the smallest tag possible may have the smallest impact on protein performance and localization, but plenty of researchers have fused proteins with whole fluorescent proteins to provide conclusions that have been confirmed by other methods. Many studies even use G proteins fused with fluorescent proteins or luciferases. Is there an important advantage to tagging G-beta with small tags? Is there evidence that G proteins with full-size protein tags behave aberrantly? If the studies presented here would not have been possible without these CRISPR-based tagging approaches, it would be helpful to provide more context to make this clearer. Perhaps one factor would be interference from newly synthesized G proteins-fluorescent protein fusions en route to the plasma membrane (in the ER and Golgi).

There are several advantages to using small peptide tags that we did not fully explain. From a practical standpoint the most important advantage of using the HiBit tag instead of full-length Nanoluc is that it allows us to restrict luminescence output to cells transiently transfected with LgBit. In this way untransfected cells contribute no background signal. Although we did not take advantage of it here, this also applies to fluorescent protein complementation, and will be useful for visualizing proteins in individual cells within tissues. The HiBit tag also allows PAGE analysis by probing membranes with LgBit (as in Fig. 1). We are not aware of evidence that tagging Gb or Gg subunits on the N terminus results in aberrant behavior, while there is some evidence that Ga subunits tagged with full-size protein tags (in some positions) have altered functional properties (PMID: 16371464). We do think that editing endogenous genes is critical, as studies using transient overexpression (usually driven by strong promoters) have sometimes reported accumulation of tagged G proteins in the biosynthetic pathway (e.g., PMID: 17576765), as the reviewer suggests. Ga and Gbg appear to be mutually dependent on each other for appropriate trafficking to the plasma membrane (reviewed in PMID: 23161140), therefore the native (presumably matched) stoichiometry is likely to be critical.

To clarify this context the revised manuscript includes the following:

“For bioluminescence experiments we added the HiBit tag (Schwinn et al., 2018) and isolated clonal “HiBit-b1“ cell lines. An advantage of this approach over adding a full-length Nanoluc luciferase is that it requires coexpression of LgBit to produce a complemented luciferase. This limits luminescence to cotransfected cells and thus eliminates background from untransfected cells.”

“Some studies using overexpressed G protein subunits have suggested that a large pool of G proteins is located on intracellular membranes, including the Golgi apparatus (Chisari et al., 2007; Saini et al., 2007; Tsutsumi et al., 2009), whereas others have indicated a distribution that is dominated by the plasma membrane (Crouthamel et al., 2008; Evanko, Thiyagarajan, & Wedegaertner, 2000; Marrari et al., 2007; Takida & Wedegaertner, 2003). A likely factor contributing to these discrepant results is the stoichiometry of overexpressed subunits, as neither Ga nor Gbg traffic appropriately to the plasma membrane as free subunits (Wedegaertner, 2012). Our gene-editing approach presumably maintains the native subunit stoichiometry, providing a more accurate representation of native G protein distribution.”

As noted by the authors, they do not demonstrate that the tagged G-beta is predominantly found within heterotrimeric G protein complexes. If there is substantial free G-beta, then many of the conclusions need to be reconsidered. Perhaps a comparison of immunoprecipitated tagged G beta vs immunoprecipitated supernatant, with blotting for other G protein subunits would be informative.

We do think that HiBit-b1 exists predominantly within heterotrimeric complexes, for several reasons. First, overexpression studies have shown that Gbg requires association with Ga to traffic to the plasma membrane, and that by itself Gbg is retained on the endoplasmic reticulum

(PMID: 12609996; PMID: 12221133). We find almost no endogenous Gb1 on the endoplasmic reticulum, and a high density on the plasma membrane. Second, we are able to detect large increases in free HiBit-Gbg after G protein activation using free Gbg sensors (e.g. Fig. 1). Third, many proteins that bind to free Gbg are found entirely in the cytosol of HEK 293 cells (e.g. PMID: 10066824), suggesting there is not a large population of free Gbg. We have added discussion of these points to the revised manuscript as follows:

“Endogenous Ga and Gb subunits are expressed at approximately a 1:1 ratio, and Gb subunits are tightly associated with Gg and inactive Ga subunits (Cho et al., 2022; Gilman, 1987; Krumins & Gilman, 2006). Moreover, proteins that bind to free Gbg dimers are found in the cytosol of unstimulated HEK 293 cells, suggesting at most only a small population of free Gbg in these cells. Therefore, we assume that the large majority of mNG-b1 and HiBit-b1 subunits in unstimulated cells are part of heterotrimers.”

“Notably, when Gbg dimers are expressed alone they accumulate on the endoplasmic reticulum

(Michaelson et al., 2002; Takida & Wedegaertner, 2003). That we detect almost no endogenous Gbg on the endoplasmic reticulum supports our conclusion that the large majority of Gbg in unstimulated HEK 293 cells is associated with Ga, although we cannot rule out a small population of free Gbg.”

We do not entirely understand the suggested experiment, as free Gbg will still be largely associated with the membrane fraction. Notably, we find almost no HiBit-b1 in the supernatant after lysis in hypotonic buffer and preparation of membrane fractions, and the small amount that we do find does not change if Ga is overexpressed.

Additional context and questions:

(1) There exists some evidence that certain GPCRs can form enduring complexes with G-betagamma (PubMed: 23297229, 27499021). That would seem to offer a mechanism that would enable receptor-mediated transport of G protein subunits. It would be helpful for the authors to place the findings of this manuscript in the context of these previous findings since they seem somewhat contradictory.

We agree. In our original submission we noted “It is possible that other receptors will influence G protein distribution using mechanisms not shared by the receptors we studied.” In the revised manuscript we have added:

“For example, a few receptors are thought to form relatively stable complexes with Gbg, which could provide a mechanism of trafficking to endosomes (Thomsen et al., 2016; Wehbi et al., 2013).”

(2) There is some evidence that GaS undergoes measurable dissociation from the plasma membrane upon activation (see the mechanism of the assay in PubMed: 35302493). It seems possible that G-alpha (and in particular GaS) might behave differently than the G-beta subunit studied here. This is not entirely clear from the discussion as it now stands.

Indeed, there is abundant evidence that some Gas translocates away from the plasma membrane upon activation. We referred to translocation of “some Ga subunits” in the introduction, although we did not specify that Gas is by far the most studied example. In a previous study (PMID: 27528603) we found that overexpressed Gas samples many intracellular membranes upon activation and returns to the plasma membrane when activation ceases. This is similar to activation-dependent translocation of free Gbg dimers. Because these translocation mechanisms depend on activation and are reversible they are unlikely to be a major source of inactive heterotrimers for intracellular membranes.

We did a poor job of making it clear that we intentionally avoided translocation mechanisms that operate only during receptor and G protein stimulation. In the revised manuscript we have added new data showing reversible activation-dependent translocation of endogenous HiBitGb1.

(3) The authors say "The presence of mNG-b1 on late endosomes suggested that some G proteins may be degraded by lysosomes". The mechanism of lysosomal degradation by proteins on the outside of the lysosome is not clear. It would be helpful for the authors to clarify.

We agree we didn’t connect the dots here. Our initial idea was that G proteins on the surface of late endosomes might reach the interior of late endosomes and then lysosomes by involution into multivesicular bodies. However, the reviewer correctly points out that much of the G protein associated with lysosomes still appears to be on the cytosolic surface, where it would not be subject to degradation. In fact, since lysosomes can fuse with the plasma membrane under certain circumstances, this could even represent a pathway for recycling G proteins to the plasma membrane.

We have revised the text to avoid giving the impression that lysosomes degrade G proteins, since we have scant evidence that this occurs. In the revised discussion we point out that we do not know the fate of G proteins located on the surface of lysosomes and speculate that these could be returned to the plasma membrane:

“We do not know the fate of G proteins located on the surface of lysosomes. Since lysosomes may fuse with the plasma membrane under certain circumstances (Xu & Ren, 2015), it is possible that this represents a route of G protein recycling to the plasma membrane.”

(4) Although the authors do a good job of assessing G protein dilution in endosomal membranes, it is unclear how this behavior compares to the measurement of other lipidanchored proteins using the same approach. Is the dilution of G proteins what we would expect for any lipid-anchored protein at the inner leaflet of the plasma membrane?

This is a great question. To begin to address it we have studied a model lipid-anchored protein consisting of mNeongreen2 anchored to the plasma membrane by the C terminus of HRas, which is palmitoylated and prenylated. We find that this protein is also diluted on endocytic vesicles, although to a lesser degree than heterotrimeric G proteins. We have added a section to the results and a new figure supplement describing these results:

“To test if other peripheral membrane proteins are similarly depleted from endocytic vesicles, we performed analogous experiments by overexpressing mNG bearing the C-terminal membrane anchor of HRas (mNG-HRas ct). We found that mNG-HRas ct was also less abundant on FM464-positive endocytic vesicles than expected based on plasma membrane abundance, although not to the same extent as mNG-b1 (Figure 4 - figure supplement 2); mNG-HRas ct density on FM4-64-positive vesicles was 64 ± 17% (mean ± 95% CI; n=78) of the nearby plasma membrane.”

Reviewer #2 (Public Review):

This is an interesting method that addresses the important problem of assessing G protein localization at endogenous levels. The data are generally convincing.

Specific comments

Methods:

The description of the gene editing method is unclear. There are two different CRISPR cell lines made in two different cell backgrounds. The methods should clearly state which CRISPR guides were used on which cell line. It is also not clear why HiBit is included in the mNG-β1 construct. Presumably, this is not critical but it would be helpful to explicitly note. In general, the Methods could be more complete.

We have added the following to the methods to clarify that the same gRNA was used to produce both cell lines:

“The human GNB1 gene was targeted at a site corresponding to the N-terminus of the Gb1 protein; the sequence 5’-TGAGTGAGCTTGACCAGTTA-3’ was incorporated into the crRNA, and the same gRNA was used to produce both HiBit-b1 and mNG-b1 cell lines.”

We have added the following to the methods to clarify why HiBit is included in the mNG-b1 construct:

“HiBit was included in the repair template for producing mNG-b1 cells to enable screening for edited clones using luminescence.”

Results:

The explanation of validation experiments in Figures 1 C and D is incomplete and difficult to follow. The rationale and explanation of the experiments could be expanded. In addition, because this is an interesting method, it would be helpful to know if the endogenous editing affects normal GPCR signaling. For example, the authors could include data showing an Isoinduced cAMP response. This is not critical to the present interpretation but is relevant as a general point regarding the method. Also, it may be relevant to the interpretation of receptor effects on G protein localization.

We have expanded the rationale and explanation of experiments in Figures 1C and D by adding:

“For example, we observed agonist-induced BRET between the D2 dopamine receptor and mNG-b1, an interaction that requires association with endogenous Ga subunits (Figure 1C). Similarly, we observed BRET between HiBit-b1 and the free Gbg sensor memGRKct-Venus after activation of receptors that couple Gi/o, Gs, and Gq heterotrimers, indicating that HiBit-b1 associated with endogenous Ga subunits from these three families (Figure 1D).”

We have done the suggested cAMP experiment and provide the data in a new figure supplement:

“We also found that cyclic AMP accumulation in response to stimulation of endogenous b adrenergic receptors was similar in edited cell lines and their unedited parent lines (Figure 1 - figure supplement 1).”

Discussion:

The conclusion that beta-gamma subunits do not redistribute after GPCR activation seems new and different from previous reports. Is this correct? Can the authors elaborate on how the results compare to previous literature?

Many previous studies have indeed shown that free Gbg dimers can redistribute after GPCR activation and sample intracellular membranes. Our initial focus was on possible changes in heterotrimer distribution after GPCR activation, but in retrospect we should have directly addressed free Gbg translocation and made the distinction clear. 

In the revised manuscript we show that during stimulation we observe changes consistent with modest translocation of endogenous Gbg from the plasma membrane and sampling of intracellular compartments. To our knowledge this is the first demonstration of endogenous Gbg translocation.

We have added:

“With overexpressed G proteins free Gbg dimers translocate from the plasma membrane and sample intracellular membrane compartments after activation-induced dissociation from Ga subunits. Consistent with this, we observed small decreases in bystander BRET at the plasma membrane and small increases in bystander BRET at intracellular compartments during activation of GPCRs, suggesting that endogenous Gbg subunits undergo similar translocation (Figure 5- figure supplement 1). Notably, these changes occurred at room temperature, suggesting that endocytosis was not involved, and developed over the course of minutes. The latter observation and the small magnitude of agonist-induced changes are both consistent with expression of primarily slowly-translocating endogenous Gg subtypes in HEK 293 cells. Moreover, as shown previously for overexpressed Gbg, the changes we observed with endogenous Gbg were readily reversible (Figure 5- figure supplement 1), suggesting that most heterotrimers reassemble at the plasma membrane after activation ceases.”

Can the authors note that OpenCell has endogenously tagged Gβ1 and reports more obvious internal localization? Can the authors comment on this point?

OpenCell has tagged GNB1 and the Leonetti group kindly provided a parent cell line we used to add a slightly different tag. Although their study did not identify any specific intracellular compartments, our impression is that most of the internal structures visible in their images are likely to be lysosomes, as they are large, round and often have a clear lumen. Overall their images and ours are comfortingly similar. We have added:

“Unsurprisingly, our images are quite similar to those made as part of previous study that labeled Gb1 subunits with mNG2 (Cho et al., 2022).”

Notably, the Leonetti group has recently reported the subcellular distribution of many untagged proteins using a proteomic approach. They find that Gb1 is enriched on the plasma membrane and lysosomes but is not enriched on endosomes, the Golgi apparatus, endoplasmic reticulum or mitochondria (https://www.biorxiv.org/content/10.1101/2023.12.18.572249v1). We have cited this work in the revised manuscript.

Is this the first use of CRISPR / HiBit for BRET assay? It would be helpful to know this or cite previous work if not. Also, as this is submitted as a tools piece, the authors might say a little more about the potential application to other questions.

The only previous study we are aware of utilizing a similar combination of methods is a 2020 report from the group of Dr. Stephen Hill, in which the authors studied binding of fluorescent ligands to HiBit-tagged GPCRs. This work is now cited.

We have also added the following to our previous brief statement about potential applications:

“In addition, it may also be possible to use these cells in combination with targeted sensors to study endogenous G protein activation in different subcellular compartments. More broadly, our results show that subcellular localization of endogenous membrane proteins can be studied in living cells by adding a HiBit tag and performing bystander BRET mapping. Applied at large scale this approach would have some advantages over fluorescent protein complementation, most notably the ability to localize endogenous membrane proteins that are expressed at levels that are too low to permit fluorescence microscopy.”

Reviewer #3 (Public Review):

Summary:

This article addresses an important and interesting question concerning intracellular localization and dynamics of endogenous G proteins. The fate and trafficking of G protein-coupled receptors (GPCRs) have been extensively studied but so far little is known about the trafficking routes of their partner G proteins that are known to dissociate from their respective receptors upon activation of the signaling pathway. The authors utilize modern cell biology tools including genome editing and bystander bioluminescence resonance energy transfer (BRET) to probe intracellular localization of G proteins in various membrane compartments in steady state and also upon receptor activation. Data presented in this manuscript shows that while G proteins are mostly present on the plasma membrane, they can be also detected in endosomal compartments, especially in late endosomes and lysosomes. This distribution, according to data presented in this study, seems not to be affected by receptor activation. These findings will have implications in further studies addressing GPCR signaling mechanisms from intracellular compartments.

Strengths:

The methods used in this study are adequate for the question asked. Especially, the use of genome-edited cells (for the addition of the tag on one of the G proteins) is a great choice to prevent the effects of overexpression. Moreover, the use of bystander BRET allowed authors to probe the intracellular localization of G proteins in a very high-throughput fashion. By combining imaging and BRET authors convincingly show that G proteins are very low abundant on early endosomes (also ER, mitochondria, and medial Golgi), however seem to accumulate on membranes of late endosomal compartments.

Weaknesses:

While the authors provide a novel dataset, many questions regarding G protein trafficking remain open. For example, it is not entirely clear which pathway is utilized to traffic G proteins from the plasma membrane to intracellular compartments. Additionally, future studies should also address the dynamics of G protein trafficking, for example by tracking them over multiple time points.

We agree, there is much more to do.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

On page 7 the text says "the difference did reach significance (Figure 5D)". It looks like the difference did not reach significance. Please check on this.

Thank you, this was an unfortunately significant typo.

Reviewer #3 (Recommendations For The Authors):

This article addresses an important and interesting question concerning intracellular localization and dynamics of endogenous G proteins. While the posed question is indeed a grand one and the methods used by the authors are novel, I believe that the data presented in this manuscript are still insufficient to support all claims posed by the authors. Below I list my major concerns:

(1) The authors claim that they provide a "detailed subcellular map of endogenous G protein distribution", however, the map is in my opinion not sufficiently detailed (e.g. trans-Golgi network is not included) and not quantitative enough (e.g. % of proteins present on one compartment vs. the other as authors claim that BRET signals "cannot be directly compared between different compartments"). To strengthen this statement, except for providing more extensive and quantitative data, it would be beneficial to provide such a "map" as an illustration based on the findings presented in this article.

“Detailed” is certainly a subjective term. While we maintain that our description of endogenous G protein distribution is far more detailed than any previous study, we now simply claim to provide a “subcellular map”. We have added images of TGNP (TGN46; TGOLN2), showing that endogenous G proteins are readily detectable on the structures labeled by this marker. These data are now provided in Figure 3 – figure supplement 7.

We did not claim that our study was quantitative- we did not try to count G proteins. However, if we use published estimates of total G proteins and surface area for HEK 293 cells we estimate that there are roughly 2,500 G proteins µm-2 on the plasma membrane and 500 G proteins µm-2 on endocytic vesicles. For other intracellular compartments relative density can be approximated by inspecting images, but a truly quantitative estimate would require a surface area standard analogous to FM4-64 for each compartment. The percentage of the total G protein pool on a given compartment is, in our opinion, less important than the density of G proteins on that compartment, as the latter is more likely to affect the efficiency of local signal transduction. Since we do not claim to have accurate G protein density estimates for many intracellular compartments, we prefer to provide several raw images for each compartment rather than a schematized map.

Bystander BRET values cannot be compared directly across compartments due to differences in expression and energy transfer efficiency of different markers and compartment surface area. This method is well suited for following changes in distribution as a function of time or after perturbations and for sensitive detection of weak colocalization but can only provide approximate “maps” of absolute distribution.

(2) Probing of the intracellular distribution of these proteins, especially after GPCR activation, includes a single chosen timepoint. I believe that the manuscript would greatly benefit from including some dynamic data on internalization and intracellular trafficking kinetics. What is the turnover of tested G proteins? What is the fraction that is going to recycling compartments and/or lysosomes? Authors could perhaps turn to other methods to be able to dynamically track proteins over time e.g. via photoconversion techniques.

Because G protein trafficking appears to be largely constitutive there is no easy way for us to assess how long it takes G proteins to transit various intracellular compartments, although we agree this would be interesting. As the reviewer suggests, dynamic data on constitutive trafficking would require methods (such as photoconversion) not currently available to us for endogenous G proteins. Accordingly, we have made no claims regarding the kinetics of G protein trafficking. As for possible redistribution after GPCR activation, in the revised manuscript we have added 5- and 15-minute timepoints after agonist stimulation for our bystander BRET mapping (Figure 5- figure supplement 2). These timepoints were chosen to correspond to persistent signaling mediated by internalized receptors. 

(3) Exemplary images with cells showing significant colocalization with lysosomal compartments seem to contain more intracellular vesicles visible in the mNG channel than in the case of the other compartment. Is it an effect of the treatment to stain lysosomes? It would be beneficial to compare it with some endogenous marker e.g. LAMP1 without additional treatments.

The visibility of intracellular vesicles in our lysosome images likely reflects our selection of cells and regions with visible and abundant lysosomes, specifically peripheral regions directly adhered to the coverslip, rather than treatment with lysosomal stains (LV 633 and dextran). As suggested, we now include images of cells expressing LAMP1 as an alternative lysosome marker (Figure 3 - figure supplement 6).

(4) The authors probe an abundance of G proteins along the constitutive endocytic pathway. However, to prove that G proteins are not de-palmitoylated rather than endocytosed authors should perform control experiments where endocytosis is blocked e.g. pharmacologically or via a knockdown approach. Additionally, various endocytic pathways can be probed.

We did not claim that depalmitoylation plays no role in delivery of G proteins to internal compartments. In fact, we pointed out that we cannot at present rule out other pathways and delivery mechanisms. Importantly, if some of the G proteins that we detect along the endocytic pathway do arrive there by trafficking through the cytosol this would only strengthen our major conclusion that endocytosis is inefficient.

Having said this, we have now conducted extensive experiments investigating the role of palmitate cycling in the trafficking of heterotrimeric G proteins and the small G protein H-Ras. Our results suggest that a depalmitoylation-repalmitoylation cycle is not important for the distribution of heterotrimers, but these findings will be the subject of a separate publication focused on this specific question for both large and small G proteins.

We agree that it will be interesting to probe different endocytic pathways, as suggested using a genetic approach. Our main interest here was in endocytic membranes that were defined functionally (with FM4-64 or internalized receptors) rather than biochemically.

Minor comments:

(5) "Imaging" paragraph in the Methods section refers to a non-existent figure called "SI Appendix S9".

Thank you.

(6) It is not clear what was used as a "control" in Figure 5E.

“Control” refers to DPBS vehicle alone. This information is now added to the legend for Figure 5E.

Author response:

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

Reviewer 1:

Comment 1 and 2: “The pipeline relies on a large number of hard-coded conditions: size of Gaussian blur (Gaussian should be written in uppercase), values of contrast, size of filters, levels of intensity, etc. Presumably, the authors followed a heuristic approach and tried values of these and concluded that the ones proposed were optimal. A proper sensitivity analysis should be performed. That is, select a range of values of the variables and measure the effect on the output.”

“Linked to the previous comments. Other researchers that want to follow the pipeline would have either to have exactly the same acquisition conditions as the manuscript or start playing with values and try to compensate for any difference in their data (cell diameter, fluorescent intensity, etc.) to see if they can match the results of the manuscript.”

We thank the Reviewer for his insightful comments. We have modified the “Usage” section of the GitHub page (https://github.com/ieoresearch/cellcycle-image-analysis) to include, for each step of the image processing, a paragraph explaining the significance of the operation and a paragraph named “Suggested Values Range” where tips for optimal parameter settings are given and examples with different parameter settings are shown. We believe that these new paragraphs help researchers easily customize the pipeline to their own data.

Reviewer 2:

Comment 1: “It would be useful to include frames from the movie showing a G1/S cell in Figures 1 and S1 with some indication of how long that cell is present. From Figure S4 it looks like it is substantially less than an hour.

It would definitely be nice to validate this observation. A brief pulse of EdU together with the FUCCI colors could allow you to do that in a culture of cycling cells. It appears that the green color as cells enter S-phase develops slowly (and maybe gets brighter continuously) as does the red color as cells progress through G1. It would be nice to validate what the color the cells are when they actually initiate DNA replication.”  

We thank the Reviewer for the opportunity to further investigate our results and clarify points that were unclear in the first version of the manuscript. As suggested, we have included all acquired frames depicting the G1 to S transition/early S phase of three cells: the Kasumi-1 untreated cell and the PF-06873600 treated NB4 cell shown in Fig. 1A, and the MDA-MB-231 cell shown in Fig. S1; they are shown in panels D of Fig. 4 and S5, respectively.

For the Kasumi-1 and NB4 cells, the G1 to S transition/early S phase, defined in the pipeline refinement step as a yellow phase appearing before the S phase, is visible at the 12-hour frame. Conversely, the MDA-MB-231 cell shown in Fig. S5D does not exhibit the G1 to S or early S phase, yellow; it transitions abruptly from red to green within our acquisition timeframe (30 min in this case), producing a green early S phase. This observation supports the Reviewer's suggestion that the G1 to S yellow transition is often shorter than one hour and it is not identifiable in all cells.

To further investigate this point, we also conducted the EdU experiments kindly suggested by the Reviewer. Kasumi-1 and MDA-MB-231 cells expressing the FUCCI(CA)2 probes were exposed to a pulse of EdU, and subsequently analyzed using flow cytometry and confocal microscopy. A new paragraph titled “The workflow allows the identification of the G1 to S phase transition” has been added to the Results section, with the corresponding data presented in Fig. 4 and Fig. S5 for Kasumi-1 and MDA-MB-231 cells, respectively. The Methods section has also been updated describing the new experiments.

Additionally, in BOX1 under the 'Cell phase assignment' paragraph, point (III), we have removed point 'a. Re-assign the G2/M frames to G1'. Although theoretically possible according to the pipeline, this reassignment is incorrect in practice because mVenus fluorescence indicates that the cells are starting or have already initiated DNA replication.

All the modifications we made in the text and Figure captions are highlighted in red. We would be thankful if the co-first authorship of Kourosh Hayatigolkhatmi, Chiara Soriani and Emanuel Soda is acknowledged in the final published version of the article.

We believe that the revisions have strengthened our manuscript, and we hope that it now meets the reviewers' suggestions for greater clarity.

Author response:

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

Reviewer #1 (Public Review):

Summary:

In the present study, Rincon-Torroella et al. developed ME3BP-7, a microencapsulated formulation of 3BP, as an agent to target MCT1 overexpressing PDACs. They provided evidence showing the specific killing of PDAC cells with MCT1 overexpressing in vitro, along with demonstrating the safety and anti-tumor efficacy of ME3BP-7 in PDAC orthotopic mouse models.

Strengths:

* Developed a novel agent.

* Well-designed experiments and an organized presentation of data that support the conclusions drawn.

Weaknesses:

There are some minor issues that could enhance the clarity and completeness of the study:

(1) Statistical results should be visually presented in Figure 4 and Figure S1.

(2) Given the tumor heterogeneity and the identification of focal high expression of MCT1 in Figure 7 and Figure S5B, it is suggested that the authors include the results of immunohistochemical (IHC) analysis of MCT1 expression in both control and ME3BP-7 treated tumor tissues. This addition may offer insight into whether the remaining tumors are composed of PDAC cells with negative MCT1 expression, while the cells with relatively high levels of MCT1 expression were eliminated by ME3BP-7 treatment.

(3) The authors are encouraged to discuss the future directions for improving the efficacy of this study. For example, exploring the combination of ME3BP-7 with a glutaminase-1 inhibitor (PMID 37891897) could be a valuable avenue for further research.

We thank the reviewer for pointing these out. We have addressed these individually in detail in the next section

Reviewer #2 (Public Review):

Summary:

In the manuscript by Rincon-Torroella et al, the authors evaluated the therapeutic potential of ME3BP-7, a microencapsulated formulation of 3BP which specifically targets MCT-1 high tumor cells, in pancreatic cancer models. The authors showed that, compared to 3BP, ME3BP-7 exhibited much-enhanced stability in serum. In addition, the authors confirmed the specificity of ME3BP-7 toward MCT-1 high tumor cells and demonstrated the in vivo anti-tumor effect of ME3BP-7 in orthotopic xenograft of human PDAC cell line and PDAC PDX model.

Strengths:

(1) The study convincingly demonstrated the superior stability of ME3BP-7 in serum.

(2) The specificity of ME3BP-7 and 3BP toward MCT-1 high PDAC cells was clearly demonstrated with CRISPR-mediated knockout experiments.

Weaknesses:

The advantage of ME3BP-7 over 3BP under an in vivo situation was not fully established.

This is a helpful observation indeed and we have attempted to address this in the revised manuscript as well as clarified the details in the following section in detail.

Reviewer #1 (Recommendations For The Authors):

There are some minor issues that could enhance the clarity and completeness of the study:

We appreciate these comments and have addressed them to the best of our abilities in the revised manuscript.

(1) Statistical results should be visually presented in Figure 4 and Figure S1.

Figure 4 and S1 have been updated to include visual representation of statistical results.

(2) Given the tumor heterogeneity and the identification of focal high expression of MCT1 in Figure 7 and Figure S5B, it is suggested that the authors include the results of immunohistochemical (IHC) analysis of MCT1 expression in both control and ME3BP-7 treated tumor tissues. This addition may offer insight into whether the remaining tumors are composed of PDAC cells with negative MCT1 expression, while the cells with relatively high levels of MCT1 expression were eliminated by ME3BP-7 treatment.

This is an excellent suggestion, but unfortunately, we were unable to implement it.   We identified a single antibody that showed specificity in our MCT1 knockout isogenic panel after testing 6 different commercial anti-MCT1 antibodies. While the chosen antibody (sc-365501) worked well on fixed human pancreatic cancer samples, it exhibited significant cross-reactivity against background mouse tissue, rendering it difficult to effectively visualize the orthotopically implanted PDx samples.  

(3) The authors are encouraged to discuss the future directions for improving the efficacy of this study. For example, exploring the combination of ME3BP-7 with a glutaminase-1 inhibitor (PMID 37891897) could be a valuable avenue for further research.

We have included potentially useful combinations of ME3BP-7 in the discussion section.

Reviewer #2 (Recommendations For The Authors):

The overall study is straightforward with translational significance. However, additional clarification is needed to determine the novelty of the study. As cited by the authors, the same group previously published a paper in Clinical Cancer Research, demonstrating the anti-tumor effect of beta-CD-3BP which is also a microencapsulated form of 3BP prepared with succinyl-beta-cyclodextrin. Please clarify what is the major difference between the ME3BP-7 and beta-CD-3BP.

We designed the first generation of beta-CD-3BP and presented the preliminary results in the Clinical Cancer Research paper.  Over the last several years, we sought to optimize the formulation so that it would be a a robust clinical candidate. The current manuscript describes our in-depth exploration.

We used a combination of SEC HPLC analyses (representative chromatogram in Fig. 3A) along with a newly developed assay to assess serum stability (representative data in Fig 3B) of a panel of ME-3BP complexes. The panel was created by varying the molar ratios of three different beta-CDs (succinyl beta-CD, native beta-CD and hydroxypropyl beta CD) to 3BP.   We discovered that an excess of succinyl-beta-CD (1.2 :1) resulted in the most stable agent with no noticeable batch effects, and this formulation was dubbed ME3BP-7).

The study clearly demonstrated the superior stability of ME3BP-7 in serum compared to 3BP. To further support the advantage of ME3BP-7, it will be important to include the same dose of 3BP as a control in the in vivo treatment experiment to evaluate the difference in both toxicity and anti-tumor effect.

We wanted to include a control arm in our study wherein the same dose of 3BP was used. However, in toxicity studies on three different species of mice, we found that infusion of 3BP at the identical dose was highly toxic, killing the animals within a few days.  We have highlighted this toxicity of the non-microencapsulated 3BP in the revised manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript by Kim et al. describes a role for axonal transport of Wnd (a dual leucine zipper kinase) for its normal degradation by the Hiw ubiquitin ligase pathway. In Hiw mutants, the Wnd protein accumulates dramatically in nerve terminals compared to the cell body of neurons. In the absence of axonal transport, Wnd levels rise and lead to excessive JNK signaling that makes neurons unhappy.

Strengths:

Using GFP-tagged Wnd transgenes and structure-function approaches, the authors show that palmitoylation of the protein at C130 plays a role in this process by promoting golgi trafficking and axonal localization of the protein. In the absence of this transport, Wnd is not degraded by Hiw. The authors also identify a role for Rab11 in the transport of Wnd, and provide some evidence that Rab11 loss-of-function neuronal degenerative phenotypes are due to excessive Wnd signaling. Overall, the paper provides convincing evidence for a preferential site of action for Wnd degradation by the Hiw pathway within axonal and/or synaptic compartments of the neuron. In the absence of Wnd transport and degradation, the JNK pathway becomes hyperactivated. As such, the manuscript provides important new insights into compartmental roles for Hiw-mediated Wnd degradation and JNK signaling control.

Weaknesses:

It is unclear if the requirement for Wnd degradation at axonal terminals is due to restricted localization of HIW there, but it seems other data in the field argues against that model. The mechanistic link between Hiw degradation and compartmentalization is unknown. 

We thank the Reviewer for valuable comments. In our revised manuscript, we have addressed reviewer ‘s comments and clarified confusions. We did not intent to imply that Rab11 directly mediates anterograde Wnd protein transport towards axon terminals. We re-worded related text throughout our manuscript to avoid confusion. Additionally, to strengthen the link between Rab11 and Wnd, we have added additional data that heterozygous mutation of wnd could rescue the eye degeneration phenotypes caused by Rab11 loss-of-function (new Figure 7C).

It is unclear if the requirement for Wnd degradation at axonal terminals is due to restricted localization of HIW there, but it seems other data in the field argues against that model. The mechanistic link between Hiw degradation and compartmentalization is unknown.

We believe that the mechanistic understanding on how Wnd protein turnover is restricted to axon/axon terminals is beyond the scope of current manuscript. We are actively investigating this interesting research question – please see our point-by-point response for details.

Reviewer #2 (Public Review):

Summary:

Utilizing transgene expression of Wnd in sensory neurons in Drosophila, the authors found that Wnd is enriched in axonal terminals. This enrichment could be blocked by preventing palmitoylation or inhibiting Rab1 or Rab11 activity. Indeed, subsequent experiments showed that inhibiting Wnd can prevent toxicity by Rab11 loss of function.

Strengths:

This paper evaluates in detail Wnd location in sensory neurons, and identifies a novel genetic interaction between Rab11 and Wnd that affects Wnd cellular distribution.

Weaknesses:

The authors report low endogenous expression of wnd, and expressing mutant hiw or overexpressing wnd is necessary to see axonal terminal enrichment. It is unclear if this overexpression model (which is known to promote synaptic overgrowth) would be relevant to normal physiology.

We agree that most of our subcellular localization studies were conducted using transgenes, which may not accurately reflect endogenous protein localization. Albeit with this technical limitation, our work addresses an important mechanistic link between DLK’s axonal localization and protein turnover, in neuronal stress signaling and neurodegeneration. 

Additionally, most of our experiments were done using a kinase-dead form of Wnd or with DLKi treatment (DLK kinase inhibitor). Neurons do not display synaptic overgrowth phenotypes under these experimental conditions. Thus, the changes in Wnd axonal localization are likely independent of synaptic overgrowth phenotypes.

Palmitoylation of the Wnd orthologue DLK in sensory neurons has previously been identified as important for DLK trafficking in a cell culture model.

Palmitoylation of DLK has been studied in previous works including Holland et al. 2015. These are important works. However, there are significant differences from our findings. First, inhibiting DLK palmitoylation caused cytoplasmic localization of DLK. It has been reported that expression levels of wild-type and the palmitoylation-defective DLK (DLK-CS) in axons are not different in cultured sensory neurons (Holland 2015, Figure 2A and 2B). This could be simply because DLK-CS is entirely cytoplasmic and can readily diffuse into axons – which led to the conclusion that DLK palmitoylation is essential for DLK localization on motile axonal puncta. Second, because of this cytoplasmic localization, DLK-CS failed to induce downstream signaling (Holland 2015).

However, the behavior of Wnd-CS from our study is entirely different. Wnd-CS does not show diffuse cytoplasmic localization, rather shows discrete localizations in neuronal cell bodies (Figure 2E, Figure 2-supplement 1). Furthermore, Wnd-CS is able to induce downstream signaling (Figure 4 – supplement 1 and 2). Thus, our manuscript is not an extension of previously published work. Rather, our manuscript took advantage of this unique behavior of Wnd-CS and elucidated biological function of the axonal localization of Wnd.

The authors find genetic interaction between Wnd and Rab11, but these studies are incomplete and they do not support the authors' mechanistic interpretation.

Our model describes that Wnd is constantly transported to axon terminals for protein degradation (protein turnover), and that this process is essential to keep Wnd activity at low levels to prevent unwanted neuronal stress signal. Based on this model, a failure in Wnd transport to axon terminals – as seen in Wnd-C130S or by Rab11 loss-of-function – would compromises protein degradation of Wnd, hence, results in excessive abundance of Wnd proteins. This was clearly demonstrated for Wnd-C130S (Figure 3) and for Rab11 mutants (Figure 6E), which support our model.

To strengthen the link between Rab11 and Wnd, we have added additional data in our revised manuscript, which showed that heterozygous mutation of wnd significantly rescued the eye degeneration phenotypes caused by Rab11 loss-of-function (new Figure 7C).

We did not intent to imply that Rab11 directly mediates anterograde Wnd protein transport towards axon terminals. We re-worded related text throughout our manuscript to avoid confusion.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) It would be interesting to overexpress Hiw in C4da neurons to see if this can degrade the C130S Wnd protein and reduce ERK signaling, or overexpress Hiw in the Rab11 mutant background to see if this can reduce the accumulation of Wnd or total Wnd levels. This could address the question of whether the reduction in Wnd turnover is due to Hiw's inaccessibility to Wnd.

Thank you for your comment. We believe this question warrants an independent line of study. Although this is beyond the scope of current work, we would like to share our findings here. We have found that overexpressing Hiw did not suppress the transgenic expression of Wnd-KD in C4da neurons regardless of cellular locations. However interestingly, the same Hiw overexpression suppressed increased Wnd-KD expression by hiw mutations in C4da neuron axon terminals. Thus, it seems that endogenous levels of Hiw in wild-type was sufficient to suppress transgenic expression of Wnd-KD, and that excessive Hiw expression does not further enhance this effect. Currently, we do not know the mechanisms underlying these observations. One possibility is that Hiw functions exclusively in the context of E3 ubiquitin ligase complex. Wu et al. (2007) found that DFsn is synaptically enriched and acts as an F-box protein of Hiw E3 ligase complex. It is possible that DFsn or some other components of Hiw E3 ligase complex determine the subcellular specificity of Hiw function. We are actively pursuing this research question currently.

(2) The authors claim that Rab11 transports Wnd to the axon terminals. However, they do not see reliable colocalization of Rab11 and Wnd at axon terminals. Can the authors see Rab11-enriched vesicles with Wnd in nerve bundles, or is the role only to sort Wnd onto a post-recycling endosome compartment that moves to axonal terminals without Rab11?

We apologize for the confusion. We did not intend to claim that Rab11 directly transports Wnd along axons. We suggested that Rab11 is necessary for axonal localization of Wnd by acting at the somatic recycling endosomes since Rab11 and Wnd extensively colocalize in the cell body but not in the axon terminals (Figure 6 and Figure 6 supplement 1). In our new “Figure 6 supplement 1”, we have now added Rab11 and Wnd colocalization in axons (segmental nerves). We also revised the text (line 294-298) “On the other hand, we did not detect any meaningful colocalization between YFP::Rab11 and Wnd-KD::mRFP in C4da axon terminals or in axons (Manders’ coefficient 0.34 ± 0.14 and 0.41 ± 0.10 respectively) (Figure 6 – supplement 1). These suggest that Rab11 is involved in Wnd protein sorting at the somatic REs rather than transporting Wnd directly.” And in Discussion (line 396-398) “These further suggest that Rab11 is not directly involved in the anterograde long-distance transport of Wnd proteins, rather is responsible for sorting Wnd into the axonal anterograde transporting vesicles.”.

(3) The authors mis-cite the Tortosa et al 2022 study which shows the exact opposite of what the authors state. Tortosa et al show DLK recruitment to vesicles through phosphorylation and palmitoylation is essential for its signaling, not the opposite, so the authors should reword that or remove the citation.

We believe the citation is correct. Tortosa et al (2022) “Stress‐induced vesicular assemblies of dual leucine zipper kinase are signaling hubs involved in kinase activation and neurodegeneration” describes that membrane association of DLK rather than palmitoylation itself is sufficient for DLK signaling activation. This is achieved by DLK palmitoylation for mammalian DLK. However, when artificially targeted to cellular membranes, palmitoylation defective DLK (mammalian DLK-CS in their study) was able to induce DLK signaling. Specifically, in their Figure 2 (K-N), when targeted to the intracellular membranes of ER and mitochondria, DLK-CS (palmitoylation defective DLK) elicited DLK signaling as shown by c-Jun phosphorylation.

Reviewer #2 (Recommendations For The Authors):

Major Concerns:

(1) A concern is the overinterpretation of results. The authors find the accumulation of Wnd in axon terminals when they express hiw null or when they overexpress Wnd, but extrapolate that this occurs in "normal conditions" without evidence. Could the increase of Wnd in the axonal terminal be in the setting of known synaptic overgrowth associated with transgene expression?

Most of our work was conducted using a kinase-dead version of Wnd (Wnd-KD) in a wild-type background (Figure 1C and Figure 1 supplement 1). Moreover, Wnd kinase activity does not affect Wnd axonal localization in our experimental settings (Figure 1 supplement 1).

When using hiw mutant background, the larvae were treated with Wnd kinase inhibitor thus, prevented excessive axonal growth (Figure 1E, bottom right image – note that there is no axonal overgrowth in this condition). Additionally, Wnd-C130S is expressed lower levels in axon terminals than Wnd (Figure 3B) while exhibiting similar axon overgrowth (Figure 4 supplement 1B). Taken together, axonal overgrowth is unlikely affect axonal protein localization of Wnd.

(2) The interpretation of these results is based on a supposition that Rab11 anterogradely transports Wnd along axons without evidence for this. Indeed, it has been shown that Rab11 is excluded from axons in mature neurons, but can be mislocalized when overexpressed. This should be addressed in their discussion.

We apologize for the confusion. We did not intend to suggest that Rab11 directly transports Wnd along axons. We suggested that Rab11 is necessary for axonal localization of Wnd by acting at the somatic recycling endosomes since Rab11 and Wnd extensively colocalize in the cell body but not in the axon terminals (Figure 6 and Figure 6 supplement 1). In our new “Figure 6 supplement 1”, we have now added Rab11 and Wnd colocalization in axons (segmental nerves). We also revised the text (line 296-298) “On the other hand, we did not detect any meaningful colocalization between YFP::Rab11 and Wnd-KD::mRFP in C4da axon terminals or in axons (Manders’ coefficient 0.34 ± 0.14 and 0.41 ± 0.10 respectively) (Figure 6 – supplement 1). These suggest that Rab11 is involved in Wnd protein sorting at the somatic REs rather than transporting Wnd directly.” And in Discussion (line 396-398) “These further suggest that Rab11 is not directly involved in the anterograde long-distance transport of Wnd proteins, rather is responsible for sorting Wnd into the axonal anterograde transporting vesicles.”.

(3) In Figure 1, the authors should also show images of Wnd-GFSTF in wild-type (non-hiw mutations) to show endogenous Wnd levels in the axon terminal.

We have now added the figures of Wnd-GFSTF in wild-type (new Figure 1A). To show the comparable fluorescent intensities, we also re-performed hiw mutant experiment and replaced the old images.

(4) For Figure 1- Supplement, the authors state that the kinase-dead version of Wnd exhibited similar axonal enrichment in comparison to Wnd::GFP in the presence and absence of DLKi. This statement would be better supported with images specifically showing this (for example Wnd-KD::GFP compared to Wnd:GFP with DLKi and Wnd:GFP without DLKi).

We did not show the images from Wnd::GFP (DLKi) in this supplement figure because it would be redundant with Figure 1C. Rather, we presented the axonal enrichment index for Wnd::GFP (DLKi), Wnd-KD::GFP, Wnd-KD::GFP (DLKi), and Wnd-KD::GFP (DMSO) in Figure 1 supplement 1B.

Overexpressing catalytically active Wnd dramatically lowers ppk-GAL4 activity in C4da neurons thus prevents us from performing an experiment for Wnd::GFP without DLKi. In this condition, Wnd::GFP expression is barely detectable in C4da neurons.

(5) In Figure 2 - Supplement 3 the authors state that their data suggests that Wnd protein palmitoylation is catalyzed by HIP14 due to colocalization in the somatic Golgi and mutating HIP14 leads to less Wnd in the axon terminal. This statement would be better supported by evaluating Wnd's palmitoylation via immunoprecipitation in response to dHIP14 enzyme activity.

We appreciate reviewer’s comment. Although the exact identity of Wnd palmitoyltransferase might be of high interest, our study rather concerns about the biological role of Wnd axonal localization. Moreover, the identity of DLK palmitoyltransferase has been identified in mammalian cell culture and worm studies (Niu et al. 2020 “Coupled Control of Distal Axon Integrity and Somal Responses to Axonal Damage by the Palmitoyl Acyltransferase ZDHHC17”). ZDHHC17 is another name for HIP14. Our data together with these published works strongly suggest that Wnd, the Drosophila DLK might also be targeted by Drosophila HIP14 or dHIP14.

(6) The authors argue that palmitoylation of Wnd is essential for axonal localization of Wnd. If dHIP14 indeed palmitoylates Wnd as the authors claim, shouldn't there be a decrease in Wnd's palmitoylation within dHIP14 mutants, consequently resulting in its accumulation in the cell body rather than localization in the axonal terminal? However, Wnd is reduced at the axon terminal in dHip14 mutants, but it does not appear to increase in the cell body (Figure 2S3.C). This observation contradicts the results showing increased Wnd in the cell body presented in Figure 2. B and E. This discrepancy should be addressed.

Thank you for your comment. Our study concerns about the biological role of Wnd axonal localization. Although in an ideal model, dHIP14 mutations should prevent Wnd palmitoylation and causes subsequent cell body accumulation. However, it is highly likely that dHIP14 mutations affect a large number of protein palmitoylations – not just Wnd, which likely changes many aspect of cell functions. We envision that Wnd protein expression might be indirectly affected by these changes. In this context, mutating C130 in Wnd can be considered as more targeted approach – and our data clearly shows that such Wnd mutations render Wnd accumulation in cell bodies.

(7) Figure 3 - the authors show increased Wnd protein by Western blot in WndC130S:GFP compared to Wnd::GFP. qPCR experiments to show similar mRNA expression of these two transgenes would be an important control, if it's thought that the increase of protein is due to reduction of protein degradation.

Thank you for your comment. Expressing WndC130S::GFP vs Wnd::GFP was done by GAL4-UAS system – not through endogenous wnd promoter. Thus, we do not expect different mRNA abundance of WndC130S::GFP and Wnd::GFP. However, your concern is valid for Rab11 mutants. We measured wnd mRNA abundance by RT-qPCR and found that Rab11 mutations did not increase wnd mRNA levels (Figure 6 - Supplement 2). Rather, we observed consistent reduction in wnd mRNA levels by Rab11 mutant. Please note that total Wnd protein levels were significantly increased by Rab11 mutations. We currently do not have a clear explanation. We envision that the dramatic increase in Wnd signaling (ie, JNK signal, Figure 7A) induces a negative-feedback to reduce wnd mRNA levels (line 313-317).

(8) Figure 4 Supplement - the authors report that Wnd::GFP causes robust induction of Puc-LacZ. A control without Wnd::GFP expression would be necessary to support that there was an induction.

We have added control data of UAS-Wnd-KD::GFP (new Figure 4 supplement 1A). Since this required a new side-by-side comparison of fluorescent intensities, we re-performed the full set of experiments and replaced our old data sets.  The results confirmed that both Wnd::GFP and Wnd-C130S::GFP induces puc-lacZ expression. 

(9) Previously it was shown that inhibiting palmitoylation of DLK prevented activation of JNK signaling (Holland et al 2015), but the authors show in Figure 4A instead an increase of JNK signaling. This discrepancy should be addressed.

The use of Wnd palmitoylation-defective mutant in our study was only possible because of different behavior of Wnd-C130S from those of palmitoylation-defective DLK. Unlike diffuse cytoplasmic localization of the palmitoylation-defective DLK in mammalian cells or in C elegans neurons, Wnd-C130S exhibited clear puncta localization in neuronal cell bodies – which extensively co-localizes with somatic Golgi complex (Figure 2E and Figure 2 supplement 1). Tortosa et al (2022) showed that palmitoylation-defective DLK (DLK-CS) can trigger DLK signaling when artificially targeted to intracellular membranous organelles (Tortosa 2022, Figure 2 (K-N)). Thus, we reasoned that unlike the palmitoylation-defective DLK from mammalian and worms, Drosophila DLK, Wnd might be catalytically active when mutated on Cysteine 130 because of its puncta localization.

(10) Figure 6 Supplement - the Rab11 staining is not in a pattern that would be expected with endosomes. A control of just YFP would be useful to determine if this fluorescence signal is specific to Rab11. Can endogenous Rab11 be detected in axons or in the axonal terminal?

In our model system, endogenously tagged Rab11 (TI-Rab11) does not show clear puncta patterns in segmental nerves (axons) and neuropils (axon terminals), neither colocalize with Wnd-KD. This is indeed related to the reviewer’s comment #2, which suggests that Rab11 does not form endosomes in distal axons or axon terminals in mature neurons. Expressing Rab11 transgenes exhibited some puncta structures in axons (segmental nerves) (new Figure 6 supplement 1). However, they did not show meaningful colocalize with Wnd-KD. These are consistent with our model that Rab11 acts in neuronal cell bodies for Wnd axonal transport – likely via a sorting process.

(11) There is growing evidence that palmitoylation is important for cargo sorting in the Golgi, and Rab11 is also located at the Golgi and important for trafficking from the Golgi. A mechanism that could be considered from your data is that blocking palmitoylation impairs sorting at the Golgi and trafficking from the Golgi, as opposed to impairing fast axonal transport. Indeed, Rab11 has been shown to be blocked from axons in mature neurons, making Rab11 unlikely to be responsible for the fast axonal transport of Wnd. Direct evidence of Rab11 transporting Wnd in axons would be necessary for the claim that Rab11 constantly transports DLK to terminals.

We apologize for the confusion. We did not intend to suggest that Rab11 directly transports Wnd along the axons. We suggested that Rab11 is necessary for axonal localization of Wnd by acting at the somatic recycling endosomes since Rab11 and Wnd extensively colocalize in the cell body but not in the axon terminals (Figure 6 and Figure 6 supplement 1). In our new “Figure 6 supplement 1”, we have now added Rab11 and Wnd colocalization in axons (segmental nerves). We also revised the text (line 296-298) “On the other hand, we did not detect any meaningful colocalization between YFP::Rab11 and Wnd-KD::mRFP in C4da axon terminals or in axons (Manders’ coefficient 0.34 ± 0.14 and 0.41 ± 0.10 respectively) (Figure 6 – supplement 1). These suggest that Rab11 is involved in Wnd protein sorting at the somatic REs rather than transporting Wnd directly.” And in Discussion (line 394-398) “These further suggest that Rab11 is not directly involved in the anterograde long-distance transport of Wnd proteins, rather is responsible for sorting Wnd into the axonal anterograde transporting vesicles.”.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This work used a comprehensive dataset to compare the effects of species diversity and genetic diversity within each trophic level and across three trophic levels. The results showed that species diversity had negative effects on ecosystem functions, while genetic diversity had positive effects. These effects were observed only within each trophic level and not across the three trophic levels studied. Although the effects of biodiversity, especially genetic diversity across multi-trophic levels, have been shown to be important, there are still very few empirical studies on this topic due to the complex relationships and difficulty in obtaining data. This study collected an excellent dataset to address this question, enhancing our understanding of genetic diversity effects in aquatic ecosystems.

Strengths:

The study collected an extensive dataset that includes species diversity of primary producers (riparian trees), primary consumers (macroinvertebrate shredders), and secondary consumers (fish). It also includes the genetic diversity of the dominant species at each trophic level, biomass production, decomposition rates, and environmental data.

The conclusions of this paper are mostly well supported by the data and the writing is logical and easy to follow.

Weaknesses:

While the dataset is impressive, the authors conducted analyses more akin to a "meta-analysis," leaving out important basic information about the raw data in the manuscript. Given the complexity of the relationships between different trophic levels and ecosystem functions, it would be beneficial for the authors to show the results of each SEM (structural equation model).

We understand the point raised by the reviewer. Our objective was to focus the Results section on the main hypotheses, and for this we let away the raw statistics. We can definitively show the seven individual SEM, highlighting the major links, which may help understand some processes. This will be done in the next version of the manuscript.

The main results presented in the manuscript are derived from a "metadata" analysis of effect sizes. However, the methods used to obtain these effect sizes are not sufficiently clarified. By analyzing the effect sizes of species diversity and genetic diversity on these ecosystem functions, the results showed that species diversity had negative effects, while genetic diversity had positive effects on ecosystem functions. The negative effects of species diversity contradict many studies conducted in biodiversity experiments. The authors argue that their study is more relevant because it is based on a natural system, which is closer to reality, but they also acknowledge that natural systems make it harder to detect underlying mechanisms. Providing more results based on the raw data and offering more explanations of the possible mechanisms in the introduction and discussion might help readers understand why and in what context species diversity could have negative effects.

We hope you will be right. As said above, we will explore this possibility.

Environmental variation was included in the analyses to test if the environment would modulate the effects of biodiversity on ecosystem functions. However, the main results and conclusions did not sufficiently address this aspect.

This will be addressed by the more in-depth analysis of individual SEM, and we will discuss this further.

Reviewer #2 (Public review):

Summary:

Fargeot et al. investigated the relative importance of genetic and species diversity on ecosystem function and examined whether this relationship varies within or between trophic-level responses. To do so, they conducted a well-designed field survey measuring species diversity at 3 trophic levels (primary producers [trees], primary consumers [macroinvertebrate shredders], and secondary consumers [fishes]), genetic diversity in a dominant species within each of these 3 trophic levels and 7 ecosystem functions across 52 riverine sites in southern France. They show that the effect of genetic and species diversity on ecosystem functions are similar in magnitude, but when examining within-trophic level responses, operate in different directions: genetic diversity having a positive effect and species diversity a negative one. This data adds to growing evidence from manipulated experiments that both species and genetic diversity can impact ecosystem function and builds upon this by showing these effects can be observed in nature.

Strengths:

The study design has resulted in a robust dataset to ask questions about the relative importance of genetic and species diversity of ecosystem function across and within trophic levels.

Overall, their data supports their conclusions - at least within the system that they are studying - but as mentioned below, it is unclear from this study how general these conclusions would be.

Weaknesses:

(1) While a robust dataset, the authors only show the data output from the SEM (i.e., effect size for each individual diversity type per trophic level (6) on each ecosystem function (7)), instead of showing much of the individual data. Although the summary SEM results are interesting and informative, I find that a weakness of this approach is that it is unclear how environmental factors (which were included but not discussed in the results) nor levels of diversity were correlated across sites. As species and genetic diversity are often correlated but also can have reciprocal feedbacks on each other (e.g., Vellend 2005), there may be constraints that underpin why the authors observed positive effects of one type of diversity (genetic) when negative effects of the other (species). It may have also been informative to run SEM with links between levels of diversity. By focusing only on the summary of SEM data, the authors may be reducing the strength of their field dataset and ability to draw inferences from multiple questions and understand specific study-system responses.

We will address this issue by performing a more in-depth analysis of each individual SEMs, and provide directly these raw data. Regarding the comment on species-genomic diversity correlations (SGDCs), we would like to point out that this has already been addressed in a previous paper (Fargeot et al. Oikos, 2023). There is actually no correlations between genomic and species diversity in these dataset, which is merely explain by the selection of the sampling sites. The relationships between species diversity, genomic diversity and environmental factors are also detailed in Fargeot et al. (2023). We precisely published this paper first to focus here “only” on BEFs. But we realize we need to provide further information and discuss further these issues. This will be done in the next version of the manuscript.

(2) My understanding of SEM is it gives outputs of the strength/significance of each pathway/relationship and if so, it isn't clear why this wasn't used and instead, confidence intervals of Z scores to determine which individual BEFs were significant. In addition, an inclusion of the 7 SEM pathway outputs would have been useful to include in an appendix.

Yes, we can provide p-values. Results from p-values will provide the same information than 95%Cis, both yield very similar (if not exactly the same) results/conclusions. We wil provide the 7 SEMs in Appendices.

(3) I don't fully agree with the authors calling this a meta-analysis as it is this a single study of multiple sites within a single region and a specific time point, and not a collection of multiple studies or ecosystems conducted by multiple authors. Moreso, the authors are using meta-analysis summary metrics to evaluate their data. The authors tend to focus on these patterns as general trends, but as the data is all from this riverine system this study could have benefited from focusing on what was going on in this system to underpin these patterns. I'd argue more data is needed to know whether across sites and ecosystems, species diversity and genetic diversity have opposite effects on ecosystem function within trophic levels.

We agree. “Meta-regression” would perhaps be more adequate than “meta-analyses”. As said above, more details will be provided on the next version of the manuscript.

Reviewer #3 (Public review):

The manuscript by Fargeot and colleagues assesses the relative effects of species and genetic diversity on ecosystem functioning. This study is very well written and examines the interesting question of whether within-species or among-species diversity correlates with ecosystem functioning, and whether these effects are consistent across trophic levels. The main findings are that genetic diversity appears to have a stronger positive effect on function than species diversity (which appears negative). These results are interesting and have value.

However, I do have some concerns that could influence the interpretation.

(1) Scale: the different measures of diversity and function for the different trophic levels are measured over very different spatial scales, for example, trees along 200 m transects and 15 cm traps. It is not clear whether trees 200 m away are having an effect on small-scale function.

Trees identification and invertebrate (and fish) sampling are done on the same scale. Trees are spread along the river so that their leaves fall directly in the river. Traps have been installed all along the same transect in various micro-habitats. Diversity have been measured at the exact same scale for all organisms. We will try to be more precise.

(2) Size of diversity gradients: More information is needed on the actual diversity gradients. One of the issues with surveys of natural systems is that they are of species that have already gone through selection filters from a regional pool, and theoretically, if the environments are similar, you should get similar sets of species, without monocultures. So, if the species diversity gradients range from say, 6 to 8 species, but genetic diversity gradients span an order of magnitude more, you can explain much more variance with genetic diversity. Related to this, species diversity effects on function are often asymptotic at high diversity and so if you are only sampling at the high diversity range, we should expect a strong effect.

We will provide more information. The range of diversity also vary according to the trophic level; there are more invertebrate species than fish species. But overall the rage of species number is large.

(3) Ecosystem functions: The functions are largely biomass estimates (expect decomposition), and I fail to see how the biomass of a single species can be construed as an ecosystem function. Aren't you just estimating a selection effect in this case?

The biomass estimated for a certain area represent an estimate of productivity, whatever the number of species being considered. Obviously, productivity of a species can be due to environmental constraints; the biomass is expected to be lower at the niche margin (selection effect). But is these environmental effects are taken into account (which is the case in the SEMs), then the residual variation can be explained by biodiversity effects. We will try to make it more clear.

Note that the article claims to be one of the only studies to look at function across trophic levels, but there are several others out there, for example:

Thanks, we will cite some of these studies (and make our claim less strong)

Li, F., Altermatt, F., Yang, J., An, S., Li, A., & Zhang, X. (2020). Human activities' fingerprint on multitrophic biodiversity and ecosystem functions across a major river catchment in China. Global change biology, 26(12), 6867-6879.

Luo, Y. H., Cadotte, M. W., Liu, J., Burgess, K. S., Tan, S. L., Ye, L. J., ... & Gao, L. M. (2022). Multitrophic diversity and biotic associations influence subalpine forest ecosystem multifunctionality. Ecology, 103(9), e3745.

Moi, D. A., Romero, G. Q., Antiqueira, P. A., Mormul, R. P., Teixeira de Mello, F., & Bonecker, C. C. (2021). Multitrophic richness enhances ecosystem multifunctionality of tropical shallow lakes. Functional Ecology, 35(4), 942-954.

Wan, B., Liu, T., Gong, X., Zhang, Y., Li, C., Chen, X., ... & Liu, M. (2022). Energy flux across multitrophic levels drives ecosystem multifunctionality: Evidence from nematode food webs. Soil Biology and Biochemistry, 169, 108656.

And the case was made strongly by:

Seibold, S., Cadotte, M. W., MacIvor, J. S., Thorn, S., & Müller, J. (2018). The necessity of multitrophic approaches in community ecology. Trends in ecology & evolution, 33(10), 754-764.

Author response:

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

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

Line 127. Provide a few more words describing the voltage protocol. To the uninitiated, panels A and B will be difficult to understand. "The large negative step is used to first close all channels, then probe the activation function with a series of depolarizing steps to re-open them and obtain the max conductance from the peak tail current at -36 mV. "

We have revised the text as suggested (revision lines 127 to Line 131): “From a holding potential within the gK,L activation range (here –74 mV), the cell is hyperpolarized to –124 mV, negative to EK and the activation range, producing a large inward current through open gK,L channels that rapidly decays as the channels deactivate. We use the large transient inward current as a hallmark of gK,L. The hyperpolarization closes all channels, and then the activation function is probed with a series of depolarizing steps, obtaining the max conductance from the peak tail current at –44 mV (Fig. 1A).”

Incidentally, why does the peak tail current decay? 

We added this text to the figure legend to explain this: “For steps positive to the midpoint voltage, tail currents are very large. As a result, K+ accumulation in the calyceal cleft reduces driving force on K+, causing currents to decay rapidly, as seen in A (Lim et al., 2011).”

The decay of the peak tail current is a feature of gK,L (large K+ conductance) and the large enclosed synaptic cleft (which concentrates K+ that effluxes from the HC). See Govindaraju et al. (2023) and Lim et al. (2011) for modeling and experiments around this phenomenon.

Line 217-218. For some reason, I stumbled over this wording. Perhaps rearrange as "In type II HCs absence of Kv1.8 significantly increased Rin and tauRC. There was no effect on Vrest because the conductances to which Kv1.8 contributes, gA and gDR activate positive to the resting potential. (so which K conductances establish Vrest???). 

We kept our original wording because we wanted to discuss the baseline (Vrest) before describing responses to current injection.

->Vrest is presumably maintained by ATP-dependent Na/K exchangers (ATP1a1), HCN, Kir, and mechanotransduction currents. Repolarization is achieved by delayed rectifier and A-type K+ conductances in type II HCs.

Figure 4, panel C - provides absolute membrane potential for voltage responses. Presumably, these were the most 'ringy' responses. Were they obtained at similar Vm in all cells (i.e., comparisons of Q values in lines 229-230). 

We added the absolute membrane potential scale. Type II HC protocols all started with 0 pA current injection at baseline, so they were at their natural Vrest, which did not differ by genotype or zone. Consistent with Q depending on expression of conductances that activate positive to Vrest, Q did not co-vary with Vrest (Pearson’s correlation coefficient = 0.08, p = 0.47, n= 85).

Lines 254. Staining is non-specific? Rather than non-selective? 

Yes, thanks - Corrected (Line 264).

Figure 6. Do you have a negative control image for Kv1.4 immuno? Is it surprising that this label is all over the cell, but Kv1.8 is restricted to the synaptic pole? 

We don’t have a null-animal control because this immunoreactivity was done in rat. While the cuticular plate staining was most likely nonspecific because we see that with many different antibodies, it’s harder to judge the background staining in the hair cell body layer. After feedback from the reviewers, we decided to pull the KV1.4 immunostaining from the paper because of the lack of null control, high background, and inability to reproduce these results in mouse tissue. In our hands, in mouse tissue, both mouse and rabbit anti-KV1.4 antibodies failed to localize to the hair cell membrane. Further optimization or another method could improve that, but for now the single-cell expression data (McInturff et al., 2018) remain the strongest evidence for KV1.4 expression in murine type II hair cells.

Lines 400-404. Whew, this is pretty cryptic. Expand a bit? 

We simplified this paragraph (revision lines 411-413): “We speculate that gA and gDR(KV1.8) have different subunit composition: gA may include heteromers of KV1.8 with other subunits that confer rapid inactivation, while gDR(KV1.8) may comprise homomeric KV1.8 channels, given that they do not have N-type inactivation .”

Line 428. 'importantly different ion channels'. I think I understand what is meant but perhaps say a bit more. 

Revised (Line 438): “biophysically distinct and functionally different ion channels”.

Random thought. In addition to impacting Rin and TauRC, do you think the more negative Vrest might also provide a selective advantage by increasing the driving force on K entry from endolymph? 

When the calyx is perfectly intact, gK,L is predicted to make Vrest less negative than the values we report in our paper, where we have disturbed the calyx to access the hair cell (–80, Govindaraju et al., 2023, vs. –87 mV, here). By enhancing K+ accumulation in the calyceal cleft, the intact calyx shifts EK—and Vrest—positively (Lim et al., 2011), so the effect on driving force may not be as drastic as what you are thinking.

Reviewer #2 (Recommendations For The Authors): 

(1) Introduction: wouldn't the small initial paragraph stating the main conclusion of the study fit better at the end of the background section, instead of at the beginning? 

Thank you for this idea, we have tried that and settled on this direct approach to let people know in advance what the goals of the paper are.

(2) Pg.4: The following sentence is rather confusing "Between P5 and P10, we detected no evidence of a non-gK,L KV1.8-dependent.....". Also, Suppl. Fig 1A seems to show that between P5 and P10 hair cells can display a potassium current having either a hyperpolarised or depolarised Vhalf. Thus, I am not sure I understand the above statement. 

Thank you for pointing out unclear wording. We used the more common “delayed rectifier” term in our revision (Lines 144-147): “Between P5 and P10, some type I HCs have not yet acquired the physiologically defined conductance, gK,L.. N effects of KV1.8 deletion were seen in the delayed rectifier currents of immature type I HCs (Suppl. Fig. 1B), showing that they are not immature forms of the Kv1.8-dependent gK,L channels. ”

(3) For the reduced Cm of hair cells from Kv1.8 knockout mice, could another reason be simply the immature state of the hair cells (i.e. lack of normal growth), rather than less channels in the membrane? 

There were no other signs to suggest immaturity or abnormal growth in KV1.8–/– hair cells or mice. Importantly, type II HCs did not show the same Cm effect.

We further discussed the capacitance effect in lines 160-167: “Cm scales with surface area, but soma sizes were unchanged by deletion of KV1.8 (Suppl. Table 2). Instead, Cm may be higher in KV1.8+/+ cells because of gK,L for two reasons. First, highly expressed trans-membrane proteins (see discussion of gK,L channel density in Chen and Eatock, 2000) can affect membrane thickness (Mitra et al., 2004), which is inversely proportional to specific Cm. Second, gK,L could contaminate estimations of capacitive current, which is calculated from the decay time constant of transient current evoked by small voltage steps outside the operating range of any ion channels. gK,L has such a negative operating range that, even for Vm negative to –90 mV, some gK,L channels are voltage-sensitive and could add to capacitive current.”

(4) Methods: The electrophysiological part states that "For most recordings, we used .....". However, it is not clear what has been used for the other recordings.

Thanks for catching this error, a holdover from an earlier ms. version.  We have deleted “For most recordings” (revision line 466).

Also, please provide the sign for the calculated 4 mV liquid junction potential. 

Done (revision line 476).

Reviewer #3 (Recommendations For The Authors): 

(1) Some of the data in panels in Fig. 1 are hard to match up. The voltage protocols shown in A and B show steps from hyperpolarized values to -71mV (A) and -32 mV (B). However, the value from A doesn't seem to correspond with the activation curve in C.

Thank you for catching this.  We accidentally showed the control I-X curve from a different cell than that in A. We now show the G-V relation for the cell in A.

Also the Vhalf in D for -/- animals is ~-38 mV, which is similar to the most positive step shown in the protocol.

The most positive step in Figure 1B is actually –25 mV. The uneven tick labels might have been confusing, so we re-labeled them to be more conventional.

Were type I cells stepped to more positive potentials to test for the presence of voltage-activated currents at greater depolarizations? This is needed to support the statement on lines 147-148. 

We added “no additional K+ conductance activated up to +40 mV” (revision line 149-150).  Our standard voltage-clamp protocol iterates up to ~+40 mV in KV1.8–/– hair cells, but in Figure 1 we only showed steps up to –25 mV because K+ accumulation in the synaptic cleft with the calyx distorts the current waveform even for the small residual conductances of the knockouts. KV1.8–/– hair cells have a main KV conductance with a Vhalf of ~–38 mV, as shown in Figure 1, and we did not see an additional KV conductance that activated with a more positive Vhalf up to +40 mV.

(2) Line 151 states "While the cells of Kv1.8-/- appeared healthy..." how were epithelia assessed for health? Hair cells arise from support cells and it would be interesting to know if Kv1.8 absence influences supporting cells or neurons. 

We added our criteria for cell health to lines 477-479: “KV1.8–/– hair cells appeared healthy in that cells had resting potentials negative to –50 mV, cells lasted a long time (20-30 minutes) in ruptured patch recordings, membranes were not fragile, and extensive blebbing was not seen.”

Supporting cells were not routinely investigated. We characterized calyx electrical activity (passive membrane properties, voltage-gated currents, firing pattern) and didn’t detect differences between +/+, +/–, and –/– recordings (data not shown). KV1.8 was not detected in neural tissue (Lee et al., 2013). 

(3) Several different K+ channel subtypes were found to contribute to inner hair cell K+ conductances (Dierich et al. 2020) but few additional K+ channel subtypes are considered here in vestibular hair cells. Further comments on calcium-activated conductances (lines 310-317) would be helpful since apamin-sensitive SK conductances are reported in type II hair cells (Poppi et al. 2018) and large iberiotoxin-sensitive BK conductances in type I hair cells (Contini et al. 2020). Were iberiotoxin effects studied at a range of voltages and might calcium-dependent conductances contribute to the enhanced resonance responses shown in Fig. 4? 

We refer you to lines 310-317 in the original ms (lines 322-329 in the revised ms), where we explain possible reasons for not observing IK(Ca) in this study.

(4) Similar to GK,L erg (Kv11) channels show significant Cs+-permeability. Were experiments using Cs+ and/or Kv11 antagonists performed to test for Kv11? 

No. Hurley et al. (2006) used Kv11 antagonists to reveal Kv11 currents in rat utricular type I hair cells with perforated patch, which were also detected in rats with single-cell RT-PCR (Hurley et al. 2006) and in mice with single-cell RNAseq (McInturff et al., 2018).  They likely contribute to hair cell currents, alongside Kv7, Kv1.8, HCN1, and Kir. 

(5) Mechanosensitive ("MET") channels in hair cells are mentioned on lines 234 and 472 (towards the end of the Discussion), but a sentence or two describing the sensory function of hair cells in terms of MET channels and K+ fluxes would help in the Introduction too. 

Following this suggestion we have expanded the introduction with the following lines  (78-87): “Hair cells are known for their large outwardly rectifying K+ conductances, which repolarize membrane voltage following a mechanically evoked perturbation and in some cases contribute to sharp electrical tuning of the hair cell membrane.  Because gK,L is unusually large and unusually negatively activated, it strongly attenuates and speeds up the receptor potentials of type I HCs (Correia et al., 1996; Rüsch and Eatock, 1996b). In addition, gK,L augments a novel non-quantal transmission from type I hair cell to afferent calyx by providing open channels for K+ flow into the synaptic cleft (Contini et al., 2012, 2017, 2020; Govindaraju et al., 2023), increasing the speed and linearity of the transmitted signal (Songer and Eatock, 2013).”

(6) Lines 258-260 state that GKL does not inactivate, but previous literature has documented a slow type of inactivation in mouse crista and utricle type I hair cells (Lim et al. 2011, Rusch and Eatock 1996) which should be considered. 

Lim et al. (2011) concluded that K+ accumulation in the synaptic cleft can explain much of the apparent inactivation of gK,L. In our paper, we were referring to fast, N-type inactivation. We changed that line to be more specific; new revision lines 269-271: “KV1.8, like most KV1 subunits, does not show fast inactivation as a heterologously expressed homomer (Lang et al., 2000; Ranjan et al., 2019; Dierich et al., 2020), nor do the KV1.8-dependent channels in type I HCs, as we show, and in cochlear inner hair cells (Dierich et al., 2020).”

(7) Lines 320-321 Zonal differences in inward rectifier conductances were reported previously in bird hair cells (Masetto and Correia 1997) and should be referenced here.

Zonal differences were reported by Masetto and Correia for type II but not type I avian hair cells, which is why we emphasize that we found a zonal difference in I-H in type I hair cells. We added two citations to direct readers to type II hair cell results (lines 333-334): “The gK,L knockout allowed identification of zonal differences in IH and IKir in type I HCs, previously examined in type II HCs (Masetto and Correia, 1997; Levin and Holt, 2012).”

Also, Horwitz et al. (2011) showed HCN channels in utricles are needed for normal balance function, so please include this reference (see line 171). 

Done (line 184).

(8) Fig 6A. Shows Kv1.4 staining in rat utricle but procedures for rat experiments are not described. These should be added. Also, indicate striola or extrastriola regions (if known). 

We removed KV1.4 immunostaining from the paper, see above.

(9) Table 6, ZD7288 is listed -was this reagent used in experiments to block Gh? If not please omit. 

ZD7288 was used to block gH to produce a clean h-infinity curve in Figure 6, which is described in the legend.

(10) In supplementary Fig. 5A make clear if the currents are from XE991 subtraction. Also, is the G-V data for single cell or multiple cells in B? It appears to be from 1 cell but ages P11-505 are given in legend. 

The G-V curve in B is from XE991 subtraction, and average parameters in the figure caption are for all the KV1.8–/–  striolar type I hair cells where we observed this double Boltzmann tail G-V curve. I added detail to the figure caption to explain this better.

(11) Supplementary Fig. 6A claims a fast activation of inward rectifier K+ channels in type II but not type I cells-not clear what exactly is measured here.

We use “fast inward rectifier” to indicate the inward current that increases within the first 20 ms after hyperpolarization from rest (IKir, characterized in Levin & Holt, 2012) in contrast to HCN channels, which open over ~100 ms. We added panel C to show that the activation of IKir is visible in type II hair cells but not in the knockout type I hair cells that lack gK,L. IKir was a reliable cue to distinguish type I and type II hair cells in the knockout.

For our actual measurements in Fig 6B, we quantified the current flowing after 250 ms at –124 mV because we did not pharmacologically separate IKir and IH.

Could the XE991-sensitive current be activated and contributing?

The XE991-sensitive current could decay (rapidly) at the onset of the hyperpolarizing step, but was not contributing to our measurement of IKir­ and IH, made after 250 ms at –124 mV, at which point any low-voltage-activated (LVA) outward rectifiers have deactivated. Additionally, the LVA XE991-sensitive currents were rare (only detected in some striolar type I hair cells) and when present did not compete with fast IKir, which is only found in type II hair cells.

Also, did the inward rectifier conductances sustain any outward conductance at more depolarized voltage steps? 

For the KV1.8-null mice specifically, we cannot answer the question because we did not use specific blocking agents for inward rectifiers.  However, we expect that there would only be sustained outward IR currents at voltages between EK and ~-60 mV: the foot of IKir’s I-V relation according to published data from mouse utricular hair cells – e.g., Holt and Eatock 1995, Rusch and Eatock 1996, Rusch et al. 1998, Horwitz et al., 2011, etc.  Thus, any such current would be unlikely to contaminate the residual outward rectifiers in Kv1.8-null animals, which activate positive to ~-60 mV. 

(I-HCN is also not a problem, because it could only be outward positive to its reversal potential at ~-40 mV, which is significantly positive to its voltage activation range.)

Author response:

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

We edited the manuscript for clarity, added information described in new figure panels (below) and corrected typos.

In figure 1 we corrected a typo.

In figure 2, panel 2H, and Figure S2E, we included a new statistical analysis (mixed effect linear regression) to compare mutational burden in controls and AD patients.

In figure 3, and Figure S4B, we revised the western blots panels in Panel 3E,F, to improve presentation of controls and quantification.

we corrected typos.

In figure 5 we removed a panel (former 5D) which did not add useful information.

In Figure S1A we included information about sex and age from the control and patients analyzed. In Figure S2B, we added an analysis of the mutational burden in controls, distinguishing controls with and without cancer.

We modified Table S1 for completeness of information for all samples analyzed.

Reviewer #1:

Weaknesses: 

Even though the study is overall very convincing, several points could help to connect the seen somatic variants in microglia more with a potential role in disease progression. The connection of P-SNVs in the genes chosen from neurological disorders was not further highlighted by the authors. 

All P-SNVs are reported in Table S3.

We observed only two P-SNVs within genes associated to neurological disorders (brain panel in Table S2). - SQSTM1 (p.P392L) was identified in blood but not in brain from the patient AD48A.

- OPTN was identified (p.Q467P) in PU.1 from control 25.   

To highlight this point, we modified the first paragraph of the discussion as follow:

“We report here that microglia from a cohort of 45 AD patients with intermediate-onset sporadic AD (mean age 65 y.o) is enriched for clones carrying pathogenic/oncogenic variants in genes associated with clonal proliferative disorders (Supplementary Table 2) in comparison to 44 controls. Of note we did not observe microglia P-SNVs within genes reported to be associated with neurological disorders (Supplementary Table 2) in patients, and one such variant was identified in a control (Supplementary Table 3) “.

The authors show in snRNA-seq data that a disease-associated microglia state seems to be enriched in patients with somatic variants in the CBL ring domain, however, this analysis could be deepened. For example, how this knowledge may translate to patient benefits when the relevant cell populations appear concentrated in a single patient sample (Figure 5; AD52) is unclear; increasing the analyzed patient pool for Figure 5 and showcasing the presence of this microglia state of interest in a few more patients with driving mutations for CBL or other MAPK pathway associated mutations would lend their hypotheses further credibility. 

We acknowledge this limitation, but we respectfully submit that the analysis was performed in 2 patients. AD 53 also show a MAPK-associated inflammatory signature in the microglia clusters associated with mutations.

We performed the analysis on all FACS-purified PU.1+ nuclei samples that passed QC for single nuclei RNAseq. It should be noted that this analysis is extremely difficult with current technologies because microglia nuclei need to be fixed for PU.1 staining and FACS purification and the clones are small (~1% of microglia).

A potential connection between P-SNVs in microglia and disease pathology and symptoms was not further explored by the authors. 

At the population level, Braak/CERAD scores, the presence of Lewy bodies, amyloid angiopathy, tauopathy, or alpha synucleinopathy were not different between AD patients with or without pathogenic microglial clones (Figure S3 and Table S1). Of note, we studied here a homogenous population of AD patients.

At the tissue level, the roles of mutant microglia in plaques for example is being investigated, but we do not have results to present at this time.

A recent preprint (Huang et al., 2024) connected the occurrence of somatic variants in genes associated with clonal hematopoiesis in microglia in a large cohort of AD patients, this study is not further discussed or compared to the data in this manuscript. 

This pre-print supports the high frequency of detection of oncogenic variants associated with clonal proliferative disorders, they hypothesize that the mutations may be associated with microglia, but they only check a few mutations in purified microglia. Most of the study is performed in whole brain tissue. It does not really bring new information as compared to other study we cite in the introduction (and to our manuscript).

Reviewer #2 (Recommendations For The Authors): 

Suggestions for improved or additional experiments, data, or analyses: 

The authors can demonstrate that identified pathological SNVs from their AD cohort also lead to the activation of human microglia-like cells in vitro, but do not provide any data from histological examination of the patient cohort (e.g. accumulation at the plaque site, microglia distribution, and cell number). The study could be further supported by providing a histological examination of patients with and without P-SNVs to identify if microglia response to pathology, microglia accumulation, or phagocytic capacity are altered in these patients. 

We performed IBA1 staining in brain samples from control and from AD patients, with or without microglial clones and microglia density was not different between patient with and without mutations. In addition, histological reports from the brain bank (Braak/CERAD scores, Lewis bodies, amyloid angiopathy, tauopathy, or alpha synucleinopathy did not suggest differences between patient with and without mutations (Figure S3). These results are preliminary and further investigations are ongoing.

It would have been interesting to see if for example, transgenic AD mice with an introduced somatic mutation in microglia show an altered disease progression with alterations in amyloid pathology or cognition. 

We agree with the reviewer. We performed an in vivo study with mice expressing a  5xFAD transgene, an inducible microglia Cx3cr1CreERt2 BrafLSL-V600E transgene, or both, and performed survival, behavioral (Y-Maze and Novel Object Recognition), and histological analyses for β-Amyloid, p-Tau and Iba1 staining.

Microgliosis was increased in the group with the 2 transgenes, however the phenotype associated with the expression of a BrafV600E allele in microglia (Mass et al Nature 2017) was strongly dominant over the phenotype of 5xFAD mice, which did not allow us to conclude on survival and behavioral analyses.

Other studies with different transgenes are in progress but we have no results yet to include in this revised manuscript.

To connect the somatic mutations in microglia better to a potential contribution in neurodegeneration or neurotoxicity, the authors could provide further details on how to demonstrate if human microglia-like cells respond differentially to amyloid or induce neurotoxicity in a co-culture or slice culture model. 

These studies are undertaken in the laboratory, but unfortunately, we have no results as yet to include in this revised manuscript.

The number of samples analyzed for hippocampi, especially in the age-matched controls might be underpowered. 

Unfortunately, despite our best efforts, we were not able to analyze more hippocampus from control individuals. To control for bias in sampling as well as to other potential bias in our analysis, we investigated the statistical analysis of the cohorts for inclusion of age as a criterion (age matched controls), inclusion of a random effect structure, and possible confounding factor such as sex, brain bank site, and samples’ anatomical location (see revised Methods and revised Fig. 2C, F, and H, and S2B).

We first tested whether the inclusion of age is appropriate in a fixed-effects linear regression using a generalized linear model (GLM) with gaussian distribution. Compared to the baseline model, the model with age had significantly low AIC (from -66.6 to -71.9, P = 0.0067 by chi-square test). Therefore, the inclusion of age as a fixed effect is appropriate. We next tested multiple structures of mixed-effects linear modeling. We used donors as random effects, while utilizing age, disease status (neurotypical control vs. AD), or both as fixed effects. Fitting was performed using the lme function implemented in the nlme package with the maximum likelihood (ML) method. The incorporation of age and disease status significantly improved overall model fitting. Both age and AD are associated with a significant increase in SNV burden in this model (P<1x10^-4 and P=1x10^-4, respectively, by likelihood ratio test). The model's total explanatory power is substantial (conditional R^2=0.48). We also asked if the addition of potential confounding factors to the model is justified. Three factors were tested via the two above-mentioned methods: sex, brain bank site, and the anatomical location of the samples. In all cases, the AIC increased, and the P values by likelihood ratio tests were higher than 0.99. Therefore, from a statistical standpoint, the inclusion of these potential confounding factors does not seem to improve overall model fitting.

Minor corrections to the text and figures: 

The authors made a great effort to analyze various samples from one individual donor. One can get a bit confused by the sentence that "an average of 2.5 brains samples were analyzed for each donor". Maybe the authors could highlight more in the first paragraph of the results section and in Figure 1A, that there are multiple samples ("technical replicates") from one individual patient across different brain regions used. 

We removed the ‘2.5’ sentence and rewrote the paragraph for clarity. Samples information’s are now displayed in Table S1.

In the method section is a part included "Expression of target genes in microglia", it was very hard to allocate where these data from public data sets were actually used and for which analysis. Maybe the authors could clarify this again. 

AU response: we apologize and corrected the paragraph in the methods (page 6) as follow: “ Expression of target genes in microglia. To evaluate the expression levels of the genes identified in this study as target of somatic variants, we consulted a publicly available database (https://www.proteinatlas.org/), and also plotted their expression as determined by RNAseq in 2 studies (Galatro et al. GSE99074 33, and Gosselin et al. 34) (Table S3 and Figure S2). For data from Galatro et al. (GSE99074) 33, normalized gene expression data and associated clinical information of isolated human microglia (N = 39) and whole brain (N = 16) from healthy controls were downloaded from GEO. For data from Gosselin et al. 34, raw gene expression ­data and associated clinical information of isolated microglia (N = 3) and whole brain (N = 1) from healthy controls were extracted from the original dataset. Raw counts were normalized using the DESeq2 package in R 35.”

Table S3 is very informative, but also very complex. The reader could maybe benefit a lot from this table if it can be structured a bit easier especially when it comes to identifying P-SNVs and in which tissue sample they were found and if this was the same patient. The sorting function on top of the columns helps, but the color coding is a bit unclear. 

Despite our best efforts we agree that the table, which contain all sequencing data for all samples, is complex. The color coding (red) only highlights the presence of pathogenic mutation.

Reviewer #3 (Recommendations For The Authors): 

This is a well-done study of an important problem. I present the following minor critiques: 

At the bottom of Page 4 and into the top of Page 5, the authors state that 66 of the 826 variants identified in their panel sequencing experiment were found in multiple donors. Then the authors proceed to analyze the remaining 760 variants. It seems that the authors concluded that these multi-donor mosaics were artifacts, which is why they were excluded from further analysis. I think this is a reasonable assumption, but it should be stated explicitly so it is clear to the reader. Complicating this assumption, however, the authors later state that one of their CBL variants was found in two donors, and it is treated as a true mosaic. The authors should make it clear whether recurrent variants were filtered out of any given analysis. It remains possible that all recurrent variants are true mosaics that occurred in multiple donors. The authors should do a bit more to characterize these recurrent variants. Are they observed in the human population using a database like gnomAD, which, together with their recurrence, would strongly suggest they are germline variants? Are they in MAPK genes, or otherwise relevant to the study?

We apologize for the confusion. Our original intent for the ddPCR validation of variants (Figure 1E) was to count only 1 ‘unique’ variant for variants found for example in 1 brain sample and in the blood from the same patient, or in 2 brain regions from one patient, in order to avoid the criticism of overinflating our validation rate. This was notably the case for TET2 and DNMT3 variants. For example, validation of a TET2 variant found in 2 different brain areas and blood of the same donor is counted as 1 and not 3. We did not eliminate these variants from the analysis as they passed the criteria for somatic variants as presented in Methods.

In contrast, when a specific variant was found and validated in two different donors, we counted it as 2.

The characterization of variants included multiple parameters and databases, including for example AF and gnomAD, as indicated in Methods and reported in Table S3.

All ddPCR results can be found at the end of Table S3.

Figure 2B labels age-matched controls as "C", but Figure 2C labels age-matched controls as AM-C. Labels should be consistent throughout the manuscript. 

We corrected this in the revised version.

It is not clear if the "p:0.02" label in Figure 2F is referring to AM-C Cx vs. AD-Cx or AM-C vs. AD. Please clarify. 

We apologize for the confusion, and we corrected the legend. The calculated p value is for the comparison between Cortex from Controls (age-matched) and the Cortex from AD.

On Page 7, the authors state, "The allelic frequencies at which MAPK activating variants are detected in brain samples from AD patients range from ~1-6% of microglia (Fig. 3G), which correspond to clones representing 2 to 12% of mutant microglia in these samples, assuming heterozygosity." I understand what the authors mean here but I think it's a bit confusingly stated. I suggest something like "The allelic frequencies at which MAPK activating variants are detected in brain samples from AD patients range from ~1-6% in microglia (Figure 3G), which correspond to mutant clones representing 2 to 12% of all microglia in these samples, assuming heterozygosity." 

We thank the reviewer for this suggestion and re-wrote that sentence.

Is there any evidence that the transcriptional regulators mutated in AD microglia (MED12, SETD2, MLL3, DNMT3A, ASXL1, etc.) are involved in regulating MAPK genes? This would tie these mutations into the broader conclusions of the paper. 

This is a very interesting question, and indeed published studies indicate that some of the transcriptional /epigenetic regulators regulate expression of MAPK genes. However, in the absence of experimental evidence in microglia and patients, the argument may be too speculative to be included.

Do the authors have any thoughts as to whether germline variants in CBL are linked to AD? If not, why do they think germline mutations in CBL are not relevant to AD? 

This is also a very interesting question. As indicated in our manuscript, germline mutations in CBL (and other member of the classical MAPK genes, see Figure 3C) cause early onset (pediatric) and severe developmental diseases known as RASopathies, characterized by multiple developmental defects, and associated with frequent neurological and cognitive deficits.

It is possible that some other (and more frequent?) germline variants may be associated with a late-onset brain restricted phenotype, but we did not find germline pSNV in our patients. GWAS studies may be more appropriate to test this hypothesis.

Do any donors show multiple variants? I don't think this is addressed in the text. 

We do find donors with multiple variants (see Figure 3D and Figure S3), however at this stage, we did not perform single nuclei genotyping to investigate whether they are part of the same clone.

Figure S3 appears to be upside down. 

This was corrected

Figure 5C should have some kind of label telling the reader what gene set is being depicted. 

We added this information above the panel (it was in the corresponding legend).

At the top of Page 12, Lewy bodies are written as Lewis bodies. 

This was corrected

Many control donors died of cancer (Table S1). Is there any information on which, if any, chemotherapeutics or radiation these patients received? Might this impact the somatic mutation burden? The authors should compare controls with and without cancer or with and without cancer treatments to rule this out. 

As suggested by the reviewer, we analyzed the mutational load of age-matched controls with and without cancer (revised Figure S2B). As expected, we saw an increase in the mutational load in controls with cancer, particularly in their blood. This information was added in the result section.

This is most likely associated with the treatments received as well as possible cancer clones.

The formatting for Table S3 is odd. Multiple different fonts are used (this is also seen in Table S5). Column Q has no column ID. The word "panel" is spelled "pannel." The word "expressed" is spelled "expressd" in one of the worksheet labels. Columns BG-BN in the ALL-SNV worksheet are blank but seemingly part of the table. 

We fixed this error in Table S3.

Author response:

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

We thank the reviewers for their constructive reviews.  Taken together, the comments and suggestions from reviewers made it clear that we needed to focus on improving the clarity of the methods and results.  We have revised the manuscript with that in mind.  In particular, we have restructured the results to make the logic of the manuscript clearer and we have added details to the methods section.

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

The work of Muller and colleagues concerns the question of where we place our feet when passing uneven terrain, in particular how we trade-off path length against the steepness of each single step. The authors find that paths are chosen that are consistently less steep and deviate from the straight line more than an average random path, suggesting that participants indeed trade-off steepness for path length. They show that this might be related to biomechanical properties, specifically the leg length of the walkers. In addition, they show using a neural network model that participants could choose the footholds based on their sensory (visual) information about depth. 

Strengths: 

The work is a natural continuation of some of the researchers' earlier work that related the immediately following steps to gaze [17]. Methodologically, the work is very impressive and presents a further step forward towards understanding real-world locomotion and its interaction with sampling visual information. While some of the results may seem somewhat trivial in hindsight (as always in this kind of study), I still think this is a very important approach to understanding locomotion in the wild better. 

Weaknesses: 

The manuscript as it stands has several issues with the reporting of the results and the statistics. In particular, it is hard to assess the inter-individual variability, as some of the data are aggregated across individuals, while in other cases only central tendencies (means or medians) are reported without providing measures of variability; this is critical, in particular as N=9 is a rather small sample size. It would also be helpful to see the actual data for some of the information merely described in the text (e.g., the dependence of \Delta H on path length). When reporting statistical analyses, test statistics and degrees of freedom should be given (or other variants that unambiguously describe the analysis).

There is only one figure (Figure 6) that shows data pooled over subjects and this is simply to illustrate how the random paths were calculated. The actual paths generated used individual subject data. We don’t draw our conclusions from these histograms – they are instead used to generate bounds for the simulated paths.  We have made clear both in the text and in the figure legends when we have plotted an example subject. Other plots show the individual subject data. We have given the range of subject medians as well as the standard deviation for data illustrated in Figure (random vs chosen), we have also given the details of the statistical test comparing the flatness of the chosen paths versus the randomly generated paths.  We have added two supplemental figures to show individual walker data more directly: (Fig. 14) the per subject histograms of step parameters, (Fig. 18) the individual subject distributions for straight path slopes and tortuosity.

The CNN analysis chosen to link the step data to visual sampling (gaze and depth features) should be motivated more clearly, and it should describe how training and test sets were generated and separated for this analysis.

We have motivated the CNN analysis and moved it earlier in the manuscript to help clarify the logic the manuscript. Details of the training and test are now provided, and the data have been replotted. The values are a little different from the original plot after making a correction in the code, but the conclusions drawn from this analysis are unchanged. This analysis simply shows that there is information in the depth images from the subject’s perspective that a network can use to learn likely footholds. This motivates the subsequent analysis of path flatness.

There are also some parts of figures, where it is unclear what is shown or where units are missing. The details are listed in the private review section, as I believe that all of these issues can be fixed in principle without additional experiments. 

Several of the Figures have been replotted to fix these issues.

Reviewer #2 (Public Review): 

Summary: 

This manuscript examines how humans walk over uneven terrain using vision to decide where to step. There is a huge lack of evidence about this because the vast majority of locomotion studies have focused on steady, well-controlled conditions, and not on decisions made in the real world. The author team has already made great advances in this topic, but there has been no practical way to map 3D terrain features in naturalistic environments. They have now developed a way to integrate such measurements along with gaze and step tracking, which allows quantitative evaluation of the proposed trade-offs between stepping vertically onto vs. stepping around obstacles, along with how far people look to decide where to step. 

Strengths: 

(1) I am impressed by the overarching outlook of the researchers. They seek to understand human decision-making in real-world locomotion tasks, a topic of obvious relevance to the human condition but not often examined in research. The field has been biased toward well-controlled studies, which have scientific advantages but also serious limitations. A well-controlled study may eliminate human decisions and favor steady or periodic motions in laboratory conditions that facilitate reliable and repeatable data collection. The present study discards all of these usually-favorable factors for rather uncontrolled conditions, yet still finds a way to explore real-world behaviors in a quantitative manner. It is an ambitious and forward-thinking approach, used to tackle an ecologically relevant question. 

(2) There are serious technical challenges to a study of this kind. It is true that there are existing solutions for motion tracking, eye tracking, and most recently, 3D terrain mapping. However most of the solutions do not have turn-key simplicity and require significant technical expertise. To integrate multiple such solutions together is even more challenging. The authors are to be commended on the technical integration here.

(3) In the absence of prior studies on this issue, it was necessary to invent new analysis methods to go with the new experimental measures. This is non-trivial and places an added burden on the authors to communicate the new methods. It's harder to be at the forefront in the choice of topic, technical experimental techniques, and analysis methods all at once. 

Weaknesses: 

(1) I am predisposed to agree with all of the major conclusions, which seem reasonable and likely to be correct. Ignoring that bias, I was confused by much of the analysis. There is an argument that the chosen paths were not random, based on a comparison of probability distributions that I could not understand. There are plots described as "turn probability vs. X" where the axes are unlabeled and the data range above 1. I hope the authors can provide a clearer description to support the findings. This manuscript stands to be cited well as THE evidence for looking ahead to plan steps, but that is only meaningful if others can understand (and ultimately replicate) the evidence. 

We have rewritten the manuscript with the goal of clarifying the analyses, and we have re-labelled the offending figure.

(2) I wish a bit more and simpler data could be provided. It is great that step parameter distributions are shown, but I am left wondering how this compares to level walking.  The distributions also seem to use absolute values for slope and direction, for understandable reasons, but that also probably skews the actual distribution. Presumably, there should be (and is) a peak at zero slope and zero direction, but absolute values mean that non-zero steps may appear approximately doubled in frequency, compared to separate positive and negative. I would hope to see actual distributions, which moreover are likely not independent and probably have a covariance structure. The covariance might help with the argument that steps are not random, and might even be an easy way to suggest the trade-off between turning and stepping vertically. This is not to disregard the present use of absolute values but to suggest some basic summary of the data before taking that step. 

We have replotted the step parameter distributions without absolute values. Unfortunately, the covariation of step parameters (step direction and step slope) is unlikely to help establish this tradeoff.  Note that the primary conclusion of the manuscript is that works make turns to keep step slope low (when possible). Thus, any correlation that might exist between goal direction and step slope would be difficult to interpret without a direct comparison to possible alternative paths (as we have done in this paper). As such we do not draw our conclusions from them.  We use them primarily to generate plausible random paths for comparison with the chosen paths.  We have added two supplementary figures including distributions (Fig 15) and covariation of all the step parameters discussed in the methods (Fig 16).

(3) Along these same lines, the manuscript could do more to enable others to digest and go further with the approach, and to facilitate interpretability of results. I like the use of a neural network to demonstrate the predictiveness of stepping, but aside from above-chance probability, what else can inform us about what visual data drives that?

The CNN analysis simply shows that the information is there in the image from the subject’s viewpoint and is used to motivate the subsequent analysis.  As noted above, we have generally tried to improve the clarity of the methods.

Similarly, the step distributions and height-turn trade-off curves are somewhat opaque and do not make it easy to envision further efforts by others, for example, people who want to model locomotion. For that, clearer (and perhaps) simpler measures would be helpful. 

We have clarified the description of these plots in the main text and in the methods.  We have also tried to clarify why we made the choices that we did in measuring the height-turn trade-off and why it is necessary in order to make a fair comparison.

I am absolutely in support of this manuscript and expect it to have a high impact. I do feel that it could benefit from clarification of the analysis and how it supports the conclusions. 

Reviewer #3 (Public Review): 

Summary: 

The systematic way in which path selection is parametrically investigated is the main contribution. 

Strengths: 

The authors have developed an impressive workflow to study gait and gaze in natural terrain. 

Weaknesses: 

(1) The training and validation data of the CNN are not explained fully making it unclear if the data tells us anything about the visual features used to guide steering. It is not clear how or on what data the network was trained (training vs. validation vs. un-peeked test data), and justification of the choices made. There is no discussion of possible overfitting. The network could be learning just e.g. specific rock arrangements. If the network is overfitting the "features" it uses could be very artefactual, pixel-level patterns and not the kinds of "features" the human reader immediately has in mind. 

The CNN analysis has now been moved earlier in the manuscript to help clarify its significance and we have expanded the description of the methods. Briefly, it simply indicates that there is information in the depth structure of the terrain that can be learned by a network. This helps justify the subsequent analyses.  Importantly, the network training and testing sets were separated by terrain to ensure that the model was being tested on “unseen” terrain and avoid the model learning specific arrangements.  This is now clarified in the text.

(2) The use of descriptive terminology should be made systematic. 

Specifically, the following terms are used without giving a single, clear definition for them: path, step, step location, foot plant, foothold, future foothold, foot location, future foot location, foot position. I think some terms are being used interchangeably. I would really highly recommend a diagrammatic cartoon sketch, showing the definitions of all these terms in a single figure, and then sticking to them in the main text. 

We have made the language more systematic and clarified the definition of each term (see Methods). Path refers to the sequence of 5 steps. Foothold is where the foot was placed in the environment. A step is the transition from one foothold to the next.

(3) More coverage of different interpretations / less interpretation in the abstract/introduction would be prudent.  The authors discuss the path selection very much on the basis of energetic costs and gait stability. At least mention should be given to other plausible parameters the participants might be optimizing (or that indeed they may be just satisficing). That is, it is taken as "given" that energetic cost is the major driver of path selection in your task, and that the relevant perception relies on internal models. Neither of these is a priori obvious nor is it as far as I can tell shown by the data (optimizing other variables, satisficing behavior, or online "direct perception" cannot be ruled out). 

The abstract has been substantially rewritten.  We have adjusted our language in the introduction/discussion to try to address this concern.

Recommendations for the authors:

Reviewing Editor comments 

You will find a full summary of all 3 reviews below. In addition to these reviews, I'd like to highlight a few points from the discussion among reviewers. 

All reviewers are in agreement that this study has the potential to be a fundamental study with far-reaching empirical and practical implications. The reviewers also appreciate the technical achievements of this study. 

At the same time, all reviewers are concerned with the overall lack of clarity in how the results are presented. There are a considerable number of figures that need better labeling, text parts that require clearer definitions, and the description of data collection and analysis (esp. with regard to the CNN) requires more care. Please pay close attention to all comments related to this, as this was the main concern that all reviewers shared. 

At a more specific level, the reviewers discussed the finding around leg length, and admittedly, found it hard to believe, in short: "extraordinary claims need strong evidence". It would be important to strengthen this analysis by considering possible confounds, and by including a discussion of the degree of conviction. 

We have weakened the discussion of this finding and provided some an additional analyses in a supplemental figure (Figure 17) to help clarify the finding.

Reviewer #1 (Recommendations For The Authors): 

First, let me apologize for the long delay with this review. Despite my generally positive evaluation (see public review), I have some concerns about the way the data are presented and questions about methodological details. 

(1) Representation of results: I find it hard to decipher how much variability arises within an individual and how much across individuals. For example, Figure 7b seems to aggregate across all individuals, while the analysis is (correctly) based on the subject medians.

Figure 7b That figure was just one subject. This is now clarified.

It would be good to see the distribution of all individuals (maybe use violin plots for each observer with the true data on one side and the baseline data on the other, or simple histograms for each). To get a feeling for inter-individual and intra-individual variability is crucial, as obviously (see the leg-length analysis) there are larger inter-individual differences and representations like these would be important to appreciate whether there is just a scaling of more or less the same effect or whether there are qualitative differences (especially in the light of N=9 being not a terribly huge sample size). 

The medians for the individual subjects are now provided with the standard deviations between subjects to indicate the extent of individual differences. Note that the random paths were chosen from the distribution of actual step slopes for that subject as one of the constraints. This makes the random paths statistically similar to the chosen paths with the differences only being generated by the particular visual context. Thus the test for a difference between chosen and random is quite conservative

Similarly, seeing \DeltaH plotted as a function of steps in the path as a figure rather than just having the verbal description would also help. 

To simplify the discussion of our methods/results we have removed the analyses that examine mean slope as a function of steps.  Because of the central limit theorem the slopes of the chosen paths remain largely unchanged regardless of the choice path length.  The slopes of the simulated paths are always larger irrespective of the choice of path length.

(2) Reporting the statistical analyses: This is related to my previous issue: I would appreciate it if the test statistics and degrees-of-freedom of the statistical tests were given along with the p-values, instead of only the p-values. This at some points would also clarify how the statistics were computed exactly (e.g., "All subjects showed comparable difference and the difference in medians evaluated across subjects was highly significant (p<<0.0001).", p.10, is ambiguous to me). 

Details have been added as requested.

(3) Why is the lower half ("tortuosity less than the median tortuosity") of paths used as "straight" rather than simply the minimum of all viable paths)?

The benchmark for a straight path is somewhat arbitrary. Using the lower half rather than the minimum length path is more conservative.

(4) For the CNN analysis, I failed to understand what was training and what was test set. I understand that the goal is to predict for all pixels whether they are a potential foothold or not, and the AUC is a measure of how well they can be discriminated based on depth information and then this is done for each image and the median over all images taken. But on which data is the CNN trained, and on which is it tested? Is this leave-n-out within the same participant? If so, how do you deal with dependencies between subsequent images? Or is it leave-1-out across participants? If so, this would be more convincing, but again, the same image might appear in training and test. If the authors just want to ask how well depth features can discriminate footholds from non-footholds, I do not see the benefit of a supervised method, which leaves the details of the feature combinations inside a black box. Rather than defining the "negative set" (i.e., the non-foothold pixels) randomly, the simulated paths could also be used, instead. If performance (AUC) gets lower than for random pixels, this would confirm that the choice of parameters to define a "viable path" is well-chosen. 

This has been clarified as described above.

Minor issues: 

(5) A higher tortuosity would also lead a participant to require more steps in total than a lower tortuosity. Could this partly explain the correlation between the leg length and the slope/tortuosity correlation? (Longer legs need fewer steps in total, thus there might be less tradeoff between \Delta H and keeping the path straight (i.e., saving steps)). To assess this, you could give the total number of steps per (straight) distance covered for leg length and compare this to a flat surface.

The calculations are done on an individual subject basis and the first and last step locations are chosen from the actual foot placements, then the random paths are generated between those endpoints. The consequence of this is that the number of steps is held constant for the analysis.  We have clarified the methods for this analysis to try to make this more clear.

(6) As far as I understand, steps happen alternatingly with the two feet. That is, even on a flat surface, one would not reach 0 tortuosity. In other words, does the lateral displacement of the feet play a role (in particular, if paths with even and paths with odd number of steps were to be compared), and if so, is it negligible for the leg-length correlation? 

All the comparisons here are done for 5 step sequences so this potential issue should not affect the slope of the regression lines or the leg length correlation.

(7) Is there any way to quantify the quality of the depth estimates? Maybe by taking an actual depth image (e.g., by LIDAR or similar) for a small portion of the terrain and comparing the results to the estimate? If this has been done for similar terrain, can a quantification be given? If errors would be similar to human errors, this would also be interesting for the interpretation of the visual sampling data.

Unfortunately, we do not have the ground truth depth image from LIDAR.  When these data were originally collected, we had not imagined being able to reconstruct the terrain.  However, we agree with the reviewers that this would be a good analysis to do. We plan to collect LIDAR in future experiments. 

To provide an assessment of quality for these data in the absence of a ground truth depth image, we have performed an evaluation of the reliability of the terrain reconstruction across repeats of the same terrain both between and within participants.  We have expanded the discussion of these reliability analyses in the results section entitled “Evaluating Terrain Reconstruction”, as well as in the corresponding methods section (see Figure 10).

(8) The figures are sometimes confusing and a bit sloppy. For example, in Figure 7a, the red, cyan, and green paths are not mentioned in the caption, in Figure 8 units on the axes would be helpful, in Figure 9 it should probably be "tortuosity" where it now states "curviness". 

These details have been fixed.

(9) I think the statement "The maximum median AUC of 0.79 indicates that the 0.79 is the median proportion of pixels in the circular..." is not an appropriate characterization of the AUC, as the number of correctly classified pixels will not only depend on the ROC (and thus the AUC), but also on the operating point chosen on the ROC (which is not specified by the AUC alone). I would avoid any complications at this point and just characterize the AUC as a measure of discriminability between footholds and non-footholds based on depth features. 

This has been fixed.

(10) Ref. [16]is probably the wrong Hart paper (I assume their 2012 Exp. Brain Res. [https://doi.org/10.1007/s00221-012-3254-x] paper is meant at this point) 

Fixed

Typos (not checked systematically, just incidental discoveries): 

(11) "While there substantial overlap" (p.10) 

(12) "field.." (p.25) 

(13) "Introduction", "General Discussion" and "Methods" as well as some subheadings are numbered, while the other headings (e.g., Results) are not. 

Fixed

Reviewer #2 (Recommendations For The Authors): 

The major suggestions have been made in the Public Review. The following are either minor comments or go into more detail about the major suggestions. All of these comments are meant to be constructive, not obstructive. 

Abstract. This is well written, but the main conclusions "Walkers avoid...This trade off is related...5 steps ahead" sound quite qualitative. They could be strengthened by more specificity (NOT p-values), e.g. "positive correlation between the unevenness of the path straight ahead and the probability that people turned off that path." 

The abstract has been substantially rewritten.

P. 5 "pinning the head position estimated from the IMU to the Meshroom estimates" sounds like there are two estimates. But it does not sound like both were used. Clarify, e.g. the Meshroom estimate of head position was used in place of IMU? 

Yes that’s correct.  We have clarified this in the text.

Figure 5. I was confused by this. First, is a person walking left to right? When the gaze position is shown, where was the eye at the time of that gaze? There are straight lines attached to the blue dots, what do they represent? The caption says gaze is directed further along the path, which made me guess the person is walking right to left, and the line originates at the eye. Except the origins do not lie on or close to the head locations. There's also no scale shown, so maybe I am completely misinterpreting. If the eye locations were connected to gaze locations, it would help to support the finding that people look five steps ahead of where they step. 

We have updated the figure and clarified the caption to remove these confusions.  There was a mistake in the original figure (where the yellow indicated head locations, we had plotted the center of mass and the choice of projection gave the incorrect impression that the fixations off the path, in blue, were separated from the head).

The view of the data is now presented so the person is walking left to right and with a projection of the head location (orange), gaze locations (blue or green) and feet (pink).

Figure 6. As stated in the major comments, the step distributions would be expected to have a covariance structure (in terms of raw data before taking absolute values). It would be helpful to report the covariances (6 numbers). As an example of a simple statistical analysis, a PCA (also based on a data covariance) would show how certain combinations of slope/distance/direction are favored over others. Such information would be a simple way to argue that the data are not completely random, and may even show a height-turn trade-off immediately. (By the way, I am assuming absolute values are used because the slopes and directions are only positive, but it wasn't clear if this was the definition.) A reason why covariances and PCA are helpful is that such data would be helpful to compute a better random walk, generated from dynamics. I believe the argument that steps are not random is not served by showing the different histograms in Figure 7, because I feel the random paths are not fairly produced. A better argument might draw randomly from the same distribution as the data (or drive a dynamical random walk), and compare with actual data. There may be correlations present in the actual data that differ from random. I could be mistaken, because it is difficult or impossible to draw conclusions from distributions of absolute values, or maybe I am only confused. In any case, I suspect other readers will also have difficulty with this section. 

This has been addressed above in the major comments.

p. 9, "average step slope" I think I understand the definition, but I suggest a diagram might be helpful to illustrate this.

There is a diagram of a single step slope in Figure 6 and a diagram of the average step slope for a path segment in Figure 12.

Incidentally, the "straight path slope" is not clearly defined. I suspect "straight" is the view from above, i.e. ignoring height changes. 

Clarified

p. 11 The tortuosity metric could use a clearer definition. Should I interpret "length of the chosen path relative to a straight path" as the numerator and denominator? Here does "length" also refer to the view from above? Why is tortuosity defined differently from step slope? Couldn't there be an analogue to step slope, except summing absolute values of direction changes? Or an analogue to tortuosity, meaning the length as viewed from the side, divided by the length of the straight path? 

We followed the literature in the definition of tortuosity.  We have clarified the definition of tortuosity in the methods, but yes, you can interpret the length of the chosen path relative to a straight path, as the numerator and denominator, and length refers to 3D length.  We agree that there are many interesting ways to look at the data but for clarity we have limited the discussion to a single definition of tortuosity in this paper.

Figure 8 could use better labeling. On the left, there is a straight path and a more tortuous path, why not report the metrics for these? On the right, there are nine unlabeled plots. The caption says "turn probability vs. straight path slope" but the vertical axis is clearly not a probability. Perhaps the axis is tortuosity? I presume the horizontal axis is a straight path slope in degrees, but this is not explained. Why are there nine plots, is each one a subject? I would prefer to be informed directly instead of guessing. (As a side note, I like the correlations as a function of leg length, it is interesting, even if slightly unbelievable. I go hiking with people quite a bit shorter and quite a lot taller than me, and anecdotally I don't think they differ so much from each other.) 

We have fixed Figure 8 which shows the average “mean slope” as a function of tortuosity.  We have added a supplemental figure which shows a scatter plot of the raw data (mean slope vs. tortuosity for each path segment).  

Note that when walking with friends other factors (e.g. social) will contribute to the cost function. As a very short person my experience is that it is a problem. In any case, the data are the data, whatever the underlying reasons. It does not seem so surprising that people of different heights make different tradeoffs. We know that the preferred gait depends on individual’s passive dynamics as described in the paper, and the terrain will change what is energetically optimal as described in the Darici and Kuo paper.

Figure 9 presumably shows one data point per subject, but this isn't clear. 

The correlations are reported per subject, and this has been clarified. 

p. 13 CNN. I like this analysis, but only sort of. It is convincing that there is SOME sort of systematic decision-making about footholds, better than chance. What it lacks is insight. I wonder what drives peoples' decisions. As an idle suggestion, the AlexNet (arXiv: Krizhevsky et al.; see also A. Karpathy's ConvNETJS demo with CIFAR-10) showed some convolutional kernels to give an idea of what the layers learned. 

Further exploration of CNN’s would definitely be interesting, but it is outside the scope of the paper. We use it simply to make a modest point, as described above.

p. 15 What is the definition of stability cost? I understand energy cost, but it is unclear how circuitous paths have a higher stability cost. One possible definition is an energetic cost having to do with going around and turning. But if not an energy cost, what is it? 

We meant to say that the longer and flatter paths are presumably more stable because of the smaller height changes. You are correct that we can’t say what the stability cost is and we have clarified this in the discussion.

p. 16 "in other data" is not explained or referenced.

Deleted 

p. 10 5 step paths and p. 17 "over the next 5 steps". I feel there is very little information to really support the 5 steps. A p-value only states the significance, not the amount of difference. This could be strengthened by plotting some measures vs. the number of steps ahead. For example, does a CNN looking 1-5 steps ahead predict better than one looking N<5 steps ahead? I am of course inclined to believe the 5 steps, but I do not see/understand strong quantitative evidence here. 

We have weakened the statements about evidence for planning 5 steps ahead.

p. 25 CNN. I did not understand the CNN. The list of layers seems incomplete, it only shows four layers. The convolutional-deconvolutional architecture is mentioned as if that is a common term, which I am unfamiliar with but choose to interpret as akin to encoder-decoder. However, the architecture does not seem to have much of a bottleneck (25x25x8 is not greatly smaller than 100x100x4), so what is the driving principle? It's also unclear how the decoder culminates, does it produce some m x m array of probabilities of stepping, where m is some lower dimension than the images? It might be helpful also to illustrate the predictions, for example, show a photo of the terrain view, along with a probability map for that view. I would expect that the reader can immediately say yes, I would likely step THERE but not there. 

We have clarified the description of the CNN. An illustration is shown in Figure 11.

Reviewer #3 (Recommendations For The Authors): 

(This section expands on the points already contained in the Public Review). 

Major issues 

(1) The training and validation data of the CNN are not explained fully making it unclear if the data tells us anything about the visual features used to guide steering. A CNN was used on the depth scenes to identify foothold locations in the images. This is the bit of the methods and the results that remains ambiguous, and the authors may need to revisit the methods/results. It is not clear how or on what data the network was trained (training vs. validation vs. un-peeked test data), and justification of the choices made. There is no discussion of possible overfitting. The network could be learning just for example specific rock arrangements in the particular place you experimented. Training the network on data from one location and then making it generalize to another location would of course be ideal. Your network probably cannot do this (as far as I can tell this was not tried), and so the meaning of the CNN results cannot really be interpreted. 

I really like the idea, of getting actual retinotopic depth field approximations. But then the question would be: what features in this information are relevant and useful for visual guidance (of foot placement)? But this question is not answered by your method. 

"If a CNN can predict these locations above chance using depth information, this would indicate that depth features can be used to explain some variation in foothold selection." But there is no analysis of what features they are. If the network is overfitting they could be very artefactual, pixel-level patterns and not the kinds of "features" the human reader immediately has in mind. As you say "CNN analysis shows that subject perspective depth features are predictive of foothold locations", well, yes, with 50,000 odd parameters the foothold coordinates can be associated with the 3D pixel maps, but what does this tell us? 

See previous discussion of these issues.

It is true that we do not know the precise depth features used. We established that information about height changes was being used, but further work is needed to specify how the visual system does this. This is mentioned in the Discussion.

You open the introduction with a motivation to understand the visual features guiding path selection, but what features the CNN finds/uses or indeed what features are there is not much discussed. You would need to bolster this, or down-emphasize this aspect in the Introduction if you cannot address it. 

"These depth image features may or may not overlap with the step slope features shown to be predictive in the previous analysis, although this analysis better approximates how subjects might use such information." I do not think you can say this. It may be better to approximate the kind of (egocentric) environment the subjects have available, but as it is I do not see how you can say anything about how the subject uses it. (The results on the path selection with respect to the terrain features, viewpoint viewpoint-independent allocentric properties of the previous analyses, are enough in themselves!) 

We have rewritten the section on the CNN to make clearer what it can and cannot do and its role in the manuscript. See previous discussion.

(2) The use of descriptive terminology should be made systematic. Overall the rest of the methodology is well explained, and the workflow is impressive. However, to interpret the results the introduction and discussion seem to use terminology somewhat inconsistently. You need to dig into the methods to figure out the exact operationalizations, and even then you cannot be quite sure what a particular term refers to. Specifically, you use the following terms without giving a single, clear definition for them (my interpretation in parentheses): 

foothold (a possible foot plant location where there is an "affordance"? or a foot plant location you actually observe for this individual? or in the sample?) 

step (foot trajectory between successive step locations) 

step location (the location where the feet are placed) 

path (are they lines projected on the ground, or are they sequences of foot plants? The figure suggests lines but you define a path in terms of five steps. 

foot plant (occurs when the foot comes in contact with step location?) 

future foothold (?) 

foot location (?) 

future foot location (?) 

foot position (?) 

I think some terms are being used interchangeably here? I would really highly recommend a diagrammatic cartoon sketch, showing the definitions of all these terms in a single figure, and then sticking to them in the main text. Also, are "gaze location" and "fixation" the same? I.e. is every gaze-ground intersection a "gaze location" (I take it it is not a "fixation", which you define by event identification by speed and acceleration thresholds in the methods)? 

We have cleaned up the language. A foothold is the location in the terrain representation (mesh) where the foot was placed. A step is the transition from one foothold to the next. A path is the sequences of 5 steps. The lines simply illustrate the path in the Figures. A gaze location is the location in the terrain representation where the walker is holding gaze still (the act of fixating). See Muller et al (2023) for further explanation.

(3) More coverage of different interpretations / less interpretation in the abstract/introduction would be prudent. You discuss the path selection very much on the basis of energetic costs and gait stability. At least mention should be given to other plausible parameters the participants might be optimizing (or that indeed they may be just satisficing). Temporal cost (more circuitous route takes longer) and uncertainty (the more step locations you sample the more chance that some of them will not be stable) seem equally reasonable, given the task ecology / the type of environment you are considering. I do not know if there is literature on these in the gait-scene, but even if not then saying you are focusing on just one explanation because that's where there is literature to fall back on would be the thing to do. 

Also in the abstract and introduction you seem to take some of this "for granted". E.g. you end the abstract saying "are planning routes as well as particular footplants. Such planning ahead allows the minimization of energetic costs. Thus locomotor behavior in natural environments is controlled by decision mechanisms that optimize for multiple factors in the context of well-calibrated sensory and motor internal models". This is too speculative to be in the abstract, in my opinion. That is, you take as "given" that energetic cost is the major driver of path selection in your task, and that the relevant perception relies on internal models. Neither of these is a priori obvious nor is it as far as I can tell shown by your data (optimizing other variables, satisficing behavior, or online "direct perception" cannot be ruled out). 

We have rewritten the abstract and Discussion with these concerns in mind.

You should probably also reference: 

Warren, W. H. (1984). Perceiving affordances: Visual guidance of stair climbing. Journal of Experimental Psychology: Human Perception and Performance, 10(5), 683-703. https://doi.org/10.1037/0096-1523.10.5.683 

Warren WH Jr, Young DS, Lee DN. Visual control of step length during running over irregular terrain. J Exp Psychol Hum Percept Perform. 1986 Aug;12(3):259-66. doi: 10.1037//0096-1523.12.3.259. PMID: 2943854. 

We have added these references to the introduction.

Minor point 

Related to (2) above, the path selection results are sometimes expressed a bit convolutedly, and the gist can get lost in the technical vocabulary. The generation of alternative "paths" and comparison of their slope and tortuousness parameters show that the participants preferred smaller slope/shorter paths. So, as far as I can tell, what this says is that in rugged terrain people like paths that are as "flat" as possible. This is common sense so hardly surprising. Do not be afraid to say so, and to express the result in plain non-technical terms. That an apple falls from a tree is common sense and hardly surprising. Yet quantifying the phenomenon, and carefully assessing the parameters of the path that the apple takes, turned out to be scientifically valuable - even if the observation itself lacked "novelty". 

Thanks.  We have tried to clarify the methods/results with this in mind.

Author response:

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

We are grateful for the many positive comments. Moreover, we appreciate the recommendations to improve the manuscript; particularly, the important discussion points raised by reviewer 1 and the comments made by reviewer 2 concerning an extended quantification of how near-spike input conductances vary across individual spikes. We have performed several new detailed analyses to address reviewer 2’s comments. In particular, we now provide for all relevant postsynaptic cells the complete distributions of the excitatory and inhibitory input conductance changes that occur right before and after postsynaptic spiking, and we provide corresponding distributions of non-spiking regions as a reference. We performed these analyses separately for different baseline activity levels. Our new results largely support our previous conclusions but provide a much more nuanced picture of the synaptic basis of spiking. To the best of our knowledge, this is the first time that parallel information on input excitation, inhibition and postsynaptic spiking is provided for individual neurons in a biological network. We would argue that our new results further support the fundamental notion that even a reductionist neuronal culture model can give rise to sophisticated network dynamics with spiking – at least partially – triggered by rapid input fluctuations, as predicted by theory. Moreover, it appears that changes in input inhibition are a key mechanism to regulate spiking during spontaneous recurrent network activity. It will be exciting to test whether this holds true for neural circuits in vivo.

In the following section, we address the reviewers’ comments individually.

Reviewer 1:

In this study the authors develop methods to interrogate cultured neuronal networks to learn about the contributions of multiple simultaneously active input neurons to postsynaptic activity. They then use these methods to ask how excitatory and inhibitory inputs combine to result in postsynaptic neuronal firing in a network context.

The study uses a compelling combination of high-density multi-electrode array recordings with patch recordings. They make ingenious use of physiology tricks such as shifting the reversal potential of inhibitory inputs, and identifying inhibitory vs. excitatory neurons through their influence on other neurons, to tease apart the key parameters of synaptic connections.

We thank the reviewer for acknowledging our efforts to develop an approach to investigate the synaptic basis of spiking in biological neurons and for appreciating the technical challenges that needed to be overcome.

The method doesn't have complete coverage of all neurons in the culture, and it appears to work on rather low-density cultures so the size of the networks in the current study is in the low tens.

(1) It would be valuable to see the caveats associated with the small size of the networks examined here.

(2) It would be also helpful if there were a section to discuss how this approach might scale up, and how better network coverage might be achieved.

These are indeed very important points that we should have discussed in more detail. Maximizing the coverage of neurons is critical to our approach, as it determines the number of potential synaptic connections that can be tested. The number of cells that we seeded onto our HD-MEA chip was chosen to achieve monolayer neuronal cultures. As detailed in ‘Materials and Methods -> Electrode selection and long-term extracellular recording of network spiking’, the entire HD-MEA chip (all 26'400 electrodes) was scanned for activity at the beginning of each experiment, and electrodes that recorded spiking activity were subsequently selected. While it is possible that some individual neurons escape detection, since they were not directly adjacent to an electrode, we estimate that a large majority of the active neurons in the culture was covered by our electrode selection method. New generations of CMOS HD-MEAs developed in our laboratory and other groups feature higher electrode densities, larger recording areas, and larger sets of electrodes that can be simultaneously recorded from (e.g., DOI:

10.1109/JSSC.2017.2686580 & 10.1038/s41467-020-18620-4). These features will substantially improve the coverage of the network and also allow for using larger neuronal networks. As suggested by reviewer 1, we added these points to the Discussion section of the revised manuscript.

The authors obtain a number of findings on the conditions in which the dynamics of excitatory and inhibitory inputs permit spiking, and the statistics of connectivity that result in this. This is of considerable interest, and clearly one would like to see how these findings map to larger networks, to non-cortical networks, and ideally to networks in-vivo. The suite of approaches discussed here could potentially serve as a basis for such further development.

(3) It would be useful for the authors to suggest such approaches.

We are confident that our suite of approaches will open important avenues to study the E & I input basis of postsynaptic spiking in other circuits beyond the in vitro cortical networks studied here. In fact, CMOS HD-MEA probes have been successfully combined with patch clamping in vivo (DIO: 10.1101/370080) and, in principle, the strategies and software tools introduced in our study would be equally applicable in an in vivo context. However, currently available in vitro CMOS HD-MEAs still surpass their in vivo counterparts (e.g., Neuropixels probes) in terms of electrode count. Moreover, using in vitro neural networks enables easy access and better network coverage compared to in vivo conditions. These are the main reasons why we chose an in vitro network for our investigation. We added these points to the Discussion section of the revised manuscript.

(4) The authors report a range of synaptic conductance waveforms in time. Not surprisingly, E and I look broadly different. Could the authors comment on the implications of differences in time-course of conductance profiles even within E (or I) synapses? Is this functional or is it an outcome of analysis uncertainty?

We are grateful to the reviewer for raising this interesting point. On the one hand, the onsets of the synaptic conductance waveform estimates were strikingly different between E and I synapses (see Fig. 8D). Furthermore, the rise and decay phases of synaptic currents were distinct for E vs. I inputs (Fig. 4C). We think that these differences are not just due to analysis uncertainty because both these observations are consistent with previously described properties of E and I inputs: Synaptic GABAergic I currents are typically slower compared to Glutamatergic E currents with respect to both rising and decay phase (DOI: 10.1126/science.abj586). Moreover, the relatively small onset latencies for I inputs that we observed are consistent with the well-known local action of inhibition. This finding was also consistent with smaller PRE-POST distances and general differences in neurite characteristics of E compared to I cells (Fig. S2).

One of the challenges in doing such studies in a dish is that the network is simply ticking away without any neural or sensory context to work on, nor any clear idea of what its outputs might mean. Nevertheless, at a single-neuron level one expects that this system might provide a reasonable subset of the kinds of activity an individual cell might have to work on.

(5) Could the authors comment on what subsets of network activity is, and is not, likely to be seen in the culture?

(6) Could they indicate what this would mean for the conclusions about E-I summation, if the in-vivo activity follows different dynamics?

We agree that there are natural limitations to a reductionist model, such as a dissociated cell culture. One may argue that neuronal cultures bear some similarities with neural networks formed during early brain development, where network formation is primarily driven by intrinsic, self-organizational capabilities. While such a self-organization is likely constrained in a 2D culture, it has been shown that several important circuit mechanisms that are observed in vivo are preserved in 2D dissociated cultures. For example, dissociated neuronal cultures can maintain E-I balance and achieve active decorrelation (DOI: 10.1038/nn.4415). In addition, in terms of activity levels, the sequences of heightened and more quiescent network spiking bear similarities with cortical Up-Down state oscillations observed during slow-wave sleep. To what extent individual circuit connectivity motifs and more nuanced network dynamics, found in vivo, can be recapitulated in vitro, is still not clear. However, combining our and previous work (especially DOI: 10.1038/nn.4415), we believe that there is sufficient evidence to justify work such as ours. On the one hand, identifying in simple cell culture models features of network dynamics and microcircuits known (or predicted) to exist in vivo is a testimony of neuronal self-organizing capabilities. On the other hand, our in vitro results will allow for more directed testing of equivalent mechanisms in vivo.

Reviewer 2:

The authors had two aims in this study. First, to develop a tool that lets them quantify the synaptic strength and sign of upstream neurons in a large network of cultured neurons. Second, they aimed at disentangling the contributions of excitatory and inhibitory inputs to spike generation.

For the quantification of synaptic currents, their methods allows them to quantify excitatory and inhibitory currents simultaneously, as the sign of the current is determined by the neuron identity in the high-density extracellular recording. They further made sure that their method works for nonstationary firing rates, and they did a simulation to characterize what kind of connections their analysis does not capture. They did not include the possibility of (dendritic) nonlinearities or gap junctions or any kind of homeostatic processes.

Thank you for the concise summary of our aims and of the features of our method. Indeed, we did not model nonlinear synaptic interactions, short-term plasticity etc., as there is likely a spectrum of possible interaction rules. Importantly, non-linear synaptic interactions were reduced by performing synaptic measurements in voltage-clamp mode.

We do not anticipate that this would impact our connectivity inference per se. However, the presence of a significant number of nonlinear events would imply that some deviations between reconstructed and measured patch current traces were to be expected even if all incoming monosynaptic connections were identified. In the future, it will be exciting to add to our current experimental protocol a simultaneous HD-MEA & patch-clamp recording, in which the membrane potential is measured in current-clamp mode. Following application of our synaptic input-mapping procedure, one could, in this way, directly assess input-sequence dependent non-linear synaptic integration during spontaneous network activity.

I see a clear weakness in the way that they quantify their goodness of fit, as they only report the explained variance, while their data are quite nonstationary. It could help to partition the explained variance into frequency bands, to at least separate the effects of a bias in baseline, the (around 100 Hz) band of synaptic frequencies and whatever high-frequency observation noise there may be. Another weak point is their explanation of unexplained variance by potential activation of extrasynaptic receptors without providing evidence. Given that these cultures are not a tissue and diffusion should be really high, this idea could easily be tested by adding a tiny amount of glutamate to the culture media.

As suggested by the reviewer, we have now partitioned the current traces into frequency bands and separately assessed the goodness-of-fit. We have updated Fig. 3C accordingly:

The following sentence was added to the main text:

“We separately compared slow baseline changes (< 3 Hz), fast synaptic activity (3 - 200 Hz) and putative high-frequency noise (> 200 Hz), yielding a median variance explained of approximately 60% in the 3 - 200 Hz range (Fig. 3C).”

Importantly, the variance explained in the frequency range of synaptic activity remains high. We would also like to point out that, even if all synaptic input connections were identified, one would expect some deviations between measured and reconstructed current trace. This is because the reconstructed trace is based on average input current waveforms and in the measured trace there may be synaptic transmission failures.

We agree that the offered explanation for unexplained variance by activation of extrasynaptic receptors is fairly speculative. As it was not a crucial discussion point, we have therefore removed the statement.

For the contributions of excitation and inhibition to neuronal spiking, the authors found a clear reduction of inhibitory inputs and increase of excitation associated with spiking when averaging across many spikes. And interestingly, the inhibition shows a reversal right after a spike and the timescale is faster during higher network activity. While these findings are great and provide further support that their method is working, they stop at this exciting point where I would really have liked to see more detail.

Thank you for acknowledging our main results concerning the synaptic basis of spiking. We attempted to integrate in one manuscript a suite of new approaches, in addition to the respective applications. We, therefore, tried to strike the appropriate level of detail in presenting our findings. With regard to our analyses of which synaptic input events regulate postsynaptic spiking, we agree with reviewer 2’s assessment that more detail concerning the variability across individual spikes would be helpful. In the following parts, we detail multiple new analyses that we have included in the revised manuscript to address reviewer 2’s comments.

A concern, of course, is that the network bursts in cultures are quite stereotypical, and that might cause averages across many bursts to show strange behaviour. So what I am missing here is a reference or baseline or null hypothesis. How does it look when using inputs from neurons that are not connected? And then, it looks like the E/(E+I) curve has lots of peaks of similar amplitude (that could be quantified...), so why does the neuron spike where it does? If I would compare to the peak (of similar amplitude) right before or right after (as a reference) are there some systematic changes? Is maybe the inhibition merely defining some general scaffold where spikes can happen and the excitation causes the spike as spiking is more irregular?

The averaged trace reveals a different timescale for high and low activity states. But does that reflect a superposition of EPSCs in a single trial or rather a different jittering of a single EPSC across trials? For answering this question, it would be good to know the variance (and whether/ how much it changes over time). Maybe not all spikes are preceded by a decrease in inhibition. Could you quantitify (correlate, scatterplot?) how exactly excitation and inhibition contributions relate for single postsynaptic spikes (or single postsynaptic non-spikes)? After all, this would be the kind of detail that requires the large amount of data that this study provides.

First of all, we are very grateful for the reviewer’s thorough assessment of our work and for the many valuable suggestions to improve it. We are convinced that we have addressed with our new analyses and the updated manuscript all issues raised here. One of the main findings from our original manuscript was that a rapid and brief change in input conductance (and particularly a reduction in inhibition) is an important spike trigger/regulator. We followed the reviewer’s suggestion and now provide scatter plots and distributions of the pre- (and post-spike) changes in input excitation and inhibition for individual postsynaptic spikes. A quantification of the peaks in the noisy E/(E+I) traces was not always trivial, which is why we reasoned that an assessment of the respective E and I changes is better suited. Moreover, as an unbiased reference, we generated separately for each postsynaptic cell a corresponding distribution of changes in input conductance in non-spiking periods (using random time points). We included our new results and updated figures in our responses to the specific reviewer comments below.

For the first part, the authors achieved their goal in developing a tool to study synaptic inputs driving subthreshold activity at the soma, and characterizing such connections. For the second part, they found an effect of EPSCs on firing, but they barely did any quantification of its relevance due to the lack of a reference.

With the availability of Neuropixels probes, there is certainly use for their tool in in vivo applications, and their statistical analysis provides a reference for future studies.

The relevance of excitatory and inhibitory currents on spiking remains to be seen in an updated version of the manuscript.

Thank you. Please see our new analyses below. Our new findings are in agreement with the main conclusions of the original manuscript. We provide evidence that rapid pre-spike changes in input conductance are observed across most individual spikes and that these rapid changes occur significantly more often before measured spikes than in non-spiking periods.

I feel that specifically Figures 6 and 7 lack relevant detail and a consistent representation that would allow the reader to establish links between the different panels. The analysis shows very detailed examples, but then jumps into analyses that show population averages over averaged responses, losing or ignoring the variability across trials. In addition, while their results themselves pass a statistical test, it is crucial to establish some measure of how relevant these results are. For that, I would really want to know how much spiking would actually be restricted by the constraints that would be posed by these results, i.e. would this be reflected in tiny changes in spiking probabilities, or are there times when spiking probabilities are necessarily high, or do we see times when we would almost certainly get a spike, but neurons can fire during other times as well.

I would agree that a detailed, quantitative analysis of this question is beyond the scope of this paper, but a qualitative analysis is feasible and should be done.

Please see our revised Figure 6. We have rearranged some of the original panels and removed one example of mean conductance profiles. Moreover, we removed a panel with analysis results based on mean conductances that is now obsolete, as more detailed analyses are provided (which are in agreement with the original findings). Analyses from panels (A-F) are mostly unchanged. Panels (G-J) show the new results.

The following paragraphs, which were added to the main text of the revised manuscript, describe our new findings:

“For a more nuanced picture of which synaptic events are associated with postsynaptic spiking, we next quantified the changes in input excitation and inhibition that preceded individual postsynaptic spikes. In our analysis, we first focused on periods with high synaptic input activity. As previously discussed, cortical neurons in vivo typically receive and integrate barrages of input activation, similar to the high-activity events that we observed here (e.g., the event depicted in Fig. 6A, right). In Fig. 6G/H, individual pre-spike changes in input conductance are shown for two example postsynaptic neurons (plots labeled ‘spiking’, right). To assess how specific these conductance changes were to spiking periods, we also quantified the changes in input conductance that occurred during non-spiking periods as a reference (we used random time points from high-activity events excluding time points adjacent to measured spike times; we upscaled the number of measured spikes by 10x; the respective plots were labeled ‘non-spiking’). Spikes of both example neurons exhibited – compared to non-spiking regions – significantly more often a pre-spike decrease in inhibition, consistent with the mean conductance profiles. Precisely how an increase (top-right quadrants in Fig. 6G/H) or decrease (bottom-left quadrants) in both I and E conductance influenced the neuronal membrane potential is difficult to predict. However, if rapid changes in input conductance had a significant role in triggering spikes, one would expect that fewer spikes would exhibit a hyperpolarizing pre-spike increase in I and decrease in E (top-left quadrant) compared to the non-spiking period. Conversely, a decrease in I and an increase E (bottom-right quadrants) would likely result in a membrane potential depolarization so that more spikes should feature the corresponding pre-spike conductance changes compared to non-spiking periods. These relative shifts are precisely what can be observed in the plots of the two example neurons (Fig. 6G/H) and, in fact, across recordings (Fig. 6I). Finally, we compared the distributions of pre-spike changes in input inhibition and excitation of each postsynaptic neuron (Fig. 6J). Further indicating a pivotal role of inhibition in triggering spikes, 6 out of 7 neurons exhibited a clear decrease in the mean values (and medians) of pre-spike changes in inhibition compared to non-spiking periods. Interestingly, the 3 out of 7 neurons with an increase in excitation showed the smallest decrease in inhibition (or even an increase in inhibition in case of neuron #7). This latter observation suggests a matching of E and I inputs and cell-specific relative contributions of E and I conductance changes in triggering spikes.

Theoretically, neuronal spiking could be driven by a prolonged suprathreshold depolarization (Petersen and Berg 2016; Renart et al. 2007) or, in more favorable subthreshold regimes, by fast synaptic input fluctuations (Ahmadian and Miller 2021; Amit and Brunel 1997; Brunel 2000; Van Vreeswijk and Sompolinsky 1996). In this section, we demonstrated that the majority of investigated neurons featured – during high-activity periods – a significant number of spikes that were associated with rapid pre-spike changes in input conductances. These findings suggest that even simple neuronal cultures can self-organize to form circuits exhibiting sophisticated spiking dynamics.”

Our new analyses detailed in Fig. 6 show that there are also presumably depolarizing events (e.g., decrease in I and increase in E) in non-spiking regions. In future studies, it will be interesting to examine what distinguishes these events from spike-inducing events of similar magnitude – one possibility is a dependency on specific input-activation sequences.

During the first days and weeks of developing neuronal cultures, spiking activity rapidly shifts from synapse-independent activity patterns to spiking dynamics that do depend on synaptic inputs and are progressively organized in network-wide high-activity events (DOI: 10.1016/j.brainres.2008.06.022). In our study, cultures at days-in-vitro 15-18 were used, and approximately 15% of the spikes occurred during high-activity events with relatively strong E and I input activity. In addition, spikes that occurred during low-activity events were at least partially regulated by synaptic input (see answers below related to Fig. 7).

In the following, I am detailing what I would consider necessary to be done about these two Figures:

Figure 6C is indeed great, though I don't see why the authors would characterize synchrony as low. When comparing with Figure 4B, I'd think that some of these values are quite high. And it wouldn't help me to imagine error bars in panel 6D.

We have removed our characterization as ‘low’ from the text. One important difference between our synchrony measure (STTC) and the quantification of spike-transmission probability (STP) is the ‘lag’ of a few milliseconds for the STP quantification window to account for synaptic delay.

Figure 6B is useful, but could be done better: The autocovariance of a shotnoise process is a convolution of the autocovariance of underlying point process and the autocovariance of the EPSC kernel. So one would want to separate those to obtain a better temporal resolution. But a shotnoise process has well defined peaks, and the time of these local maxima can be estimated quite precisely. Now if I would do a peak triggered average instead of the full convolution, I would do half of the deconvolution and obtain a temporally asymmetric curve of what is expected to happen around an EPSC. Importantly, one could directly see expected excitation after inhibition or expected inhibition after excitation, and this visualization could be much better and more intuitively compared to panel 6E.

We appreciate the reviewer’s suggestion to present these results in a more sophisticated way. We would like to propose to stick with the original analysis to have it comparable with related analyses from the literature (e.g., DOI: 10.1038/nn.2105). Therefore, we hope the reviewer finds it acceptable that we leave the presentation of the data in its original form and potentially follow up in future work with the analysis strategy proposed by the reviewer.

Panel D needs some variability estimate (i.e. standard deviation or interquartile range or even a probability density) for those traces.

Figure 6E: Please use more visible colors. A sensitivity analysis to see traces for 2E/(2E+I) and E/(E+2I) would be great.

Figure 6F: with an updated panel B, we should be able to have a slope for average inhibition after excitation for each of these cells. A second panel / third column showing those slopes would be of interest. It would serve as a reference for what could be expected from E-I interactions alone.

With regard to the variability estimate in D, we now provide multiple panels characterizing the variability. For one, Fig. 6H contains a scatter plot of the pre-spike changes in input conductance across all individual postsynaptic spikes from the example cell shown in D. Moreover, in Fig. 7A, we show from the same example cell the standard deviations associated with the mean conductance traces separately for spikes that occurred during low- and high-activity states. For better visibility and because the separation according to activity states is more informative, we kept the original presentation of panel D (however, removing one example cell). In addition, we show the same mean traces from panel D with the respective standard deviations (across all spikes) in Supplementary Figure S3.

Colors in Fig. 6E are adjusted, as requested.

We have removed panel Fig. 6F as we now provide more detailed analyses at single-spike level (see Fig. 6G-J).

Figure 6G: Could the authors provide an interquartile range here?

With regard to the aligned input-output data from original panel Fig. 6G, now in panel Fig. 6F in the updated figure version, we show all individual traces that were averaged: the E/I traces from panel Fig. 6E and the three action potential waveforms from Supplementary Figure S5. Therefore, we chose to present the means only for better visibility.

Figure 7A: it may be hard to squeeze in variability estimates here, but the information on whether and how much variance might be explained is essential. Maybe add another panel to provide a variability estimate? The variability estimate in panel 7B and 7D only reflect variability across connections, and it would be useful to add panels for the time courses of the variability of g (or E/(E+I) respectively).

We now include the standard deviations across the input conductance traces in the updated Fig. 7A, as requested. We have also simplified Fig. 7 and performed the analysis using the 6 out of 7 neurons that, based on our new analysis (Fig. 6J) displayed a clear reduction in pre-spike inhibition, relative to the reference distribution. For a complete overview of the state-dependent changes in input conductance that are associated with individual postsynaptic spikes, we have included a new supplementary figure (Fig. S6). Fig. S6 also includes a characterization of the changes in input inhibition that occur right after postsynaptic spiking. In addition, Fig. S6D shows the standard deviations corresponding to the mean input conductance traces of all cells – separately for high- and low-activity periods.

We added the following paragraph to the main text of the revised manuscript:

“How can these deviations in the mean conductance profiles be explained? To answer this question, we further quantified – separately for low and high g states – the changes in input inhibition that occurred right before and after individual postsynaptic spikes (Fig. S6). This single-spike analysis suggested that, during high g states, most spikes experienced a post-spike increase and pre-spike decrease in inhibition (see also Fig. 6J). On the other hand, low g states were characterized by sparse synaptic input (e.g., see reconstruction in Fig. 6A). Therefore, many of the spikes that occurred during low g states were associated with little change in input conductance (note medians of approximately zero in Fig. S6A/C). Nevertheless, a considerable fraction of spikes (often > 25%) from low g states were also associated with a post-spike increase and pre-spike drop in inhibition. It, therefore, appears that even the sparse inhibitory inputs of low g states could influence spike timing. Moreover, the post-spike increases in input inhibition during low g states suggest that there were strong regulatory inhibitory circuits in place. However, limited activity levels during low g states presumably introduced an increased jitter of these spike-associated changes in input inhibition.

In summary, the input inhibition of high-conductance states provides reliable and narrow windows-of-spiking opportunity. In addition, even during periods of sparse activity, there are rudimentary synaptic mechanisms in place to regulate spike timing.”

As a suggestion for further analysis, though I am well aware that this is likely beyond the scope of this manuscript, I'd suggest the following analysis:

I would split the data into the high and low activity states. Then I would compute the average of E/(E+I) values for spikes. Assuming that spikes tend to happen for local maxima of E/(E+I) I would find local maxima for periods without spike such that their average is equal to the value for actual spikes. Finally, I would test for a systematic difference in either excitation or inhibition.

If there is no difference, you can make the claim that synaptic input does not guarantee a spike, and compare to a global average of E/(E+I).

We are grateful for the fantastic suggestions for future analysis. We look forward to conducting these analyses in a more detailed follow-up characterization.

In addition to the major alterations detailed above, we performed smaller corrections (e.g., spelling mistakes, inaccuracies) in some parts of the manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript, the authors use a large dataset of neuroscience publications to elucidate the nature of self-citation within the neuroscience literature. The authors initially present descriptive measures of self-citation across time and author characteristics; they then produce an inclusive model to tease apart the potential role of various article and author features in shaping self-citation behavior. This is a valuable area of study, and the authors approach it with an appropriate and well-structured dataset.

The study's descriptive analyses and figures are useful and will be of interest to the neuroscience community. However, with regard to the statistical comparisons and regression models, I believe that there are methodological flaws that may limit the validity of the presented results. These issues primarily affect the uncertainty of estimates and the statistical inference made on comparisons and model estimates - the fundamental direction and magnitude of the results are unlikely to change in most cases. I have included detailed statistical comments below for reference.

Conceptually, I think this study will be very effective at providing context and empirical evidence for a broader conversation around self-citation. And while I believe that there is room for a deeper quantitative dive into some finer-grained questions, this paper will be a valuable catalyst for new areas of inquiry around citation behavior - e.g., do authors change self-citation behavior when they move to more or less prestigious institutions? do self-citations in neuroscience benefit downstream citation accumulation? do journals' reference list policies increase or decrease self-citation? - that I hope that the authors (or others) consider exploring in future work.

Thank you for your suggestions and your generally positive view of our work. As described below, we have made the statistical improvements that you suggested.

Statistical comments:

(1) Throughout the paper, the nested nature of the data does not seem to be appropriately handled in the bootstrapping, permutation inference, and regression models. This is likely to lead to inappropriately narrow confidence bands and overly generous statistical inference.

We apologize for this error. We have now included nested bootstrapping and permutation tests. We defined an “exchangeability block” as a co-authorship group of authors. In this dataset, that meant any authors who published together (among the articles in this dataset) as a First Author / Last Author pairing were assigned to the same exchangeability block. It is not realistic to check for overlapping middle authors in all papers because of the collaborative nature of the field. In addition, we believe that self-citations are primarily controlled by first and last authors, so we can assume that middle authors do not control self-citation habits. We then performed bootstrapping and permutation tests in the constraints of the exchangeability blocks.

We first describe this in the results (page 3, line 110):

“Importantly, we accounted for the nested structure of the data in bootstrapping and permutation tests by forming co-authorship exchangeability blocks.”

We also describe this in 4.8 Confidence Intervals (page 21, line 725):

“Confidence intervals were computed with 1000 iterations of bootstrap resampling at the article level. For example, of the 100,347 articles in the dataset, we resampled articles with replacement and recomputed all results. The 95% confidence interval was reported as the 2.5 and 97.5 percentiles of the bootstrapped values.

We grouped data into exchangeability blocks to avoid overly narrow confidence intervals or overly optimistic statistical inference. Each exchangeability block comprised any authors who published together as a First Author / Last Author pairing in our dataset. We only considered shared First/Last Author publications because we believe that these authors primarily control self-citations, and otherwise exchangeability blocks would grow too large due to the highly collaborative nature of the field. Furthermore, the exchangeability blocks do not account for co-authorship in other journals or prior to 2000. A distribution of the sizes of exchangeability blocks is presented in Figure S15.”

In describing permutation tests, we also write (page 21, line 739):

“4.9 P values

P values were computed with permutation testing using 10,000 permutations, with the exception of regression P values and P values from model coefficients. For comparing different fields (e.g., Neuroscience and Psychiatry) and comparing self-citation rates of men and women, the labels were randomly permuted by exchangeability block to obtain null distributions. For comparing self-citation rates between First and Last Authors, the first and last authorship was swapped in 50% of exchangeability blocks.”

For modeling, we considered doing a mixed effects model but found difficulties due to computational power. For example, with our previous model, there were hundreds of thousands of levels for the paper random effect, and tens of thousands of levels for the author random effect. Even when subsampling or using packages designed for large datasets (e.g., mgcv’s bam function: https://www.rdocumentation.org/packages/mgcv/versions/1.9-1/topics/bam), we found computational difficulties.

As a result, we switched to modeling results at the paper level (e.g., self-citation count or rate). We found that results could be unstable when including author-level random effects because in many cases there was only one author per group. Instead, to avoid inappropriately narrow confidence bands, we resampled the dataset such that each author was only represented once. For example, if Author A had five papers in this dataset, then one of their five papers was randomly selected. We updated our description of our models in the Methods section (page 21, line 754):

“4.10 Exploring effects of covariates with generalized additive models

For these analyses, we used the full dataset size separately for First and Last Authors (Table S2). This included 115,205 articles and 5,794,926 citations for First Authors, and 114,622 articles and 5,801,367 citations for Last Authors. We modeled self-citation counts, self-citation rates, and number of previous papers for First Authors and Last Authors separately, resulting in six total models.

We found that models could be computationally intensive and unstable when including author-level random effects because in many cases there was only one author per group. Instead, to avoid inappropriately narrow confidence bands, we resampled the dataset such that each author was only represented once. For example, if Author A had five papers in this dataset, then one of their five papers was randomly selected. The random resampling was repeated 100 times as a sensitivity analysis (Figure S12).

For our models, we used generalized additive models from mgcv’s “gam” function in R 49. The smooth terms included all the continuous variables: number of previous papers, academic age, year, time lag, number of authors, number of references, and journal impact factor. The linear terms included all the categorical variables: field, gender affiliation country LMIC status, and document type. We empirically selected a Tweedie distribution 50 with a log link function and p=1.2. The p parameter indicates that the variance is proportional to the mean to the p power 49. The p parameter ranges from 1-2, with p=1 equivalent to the Poisson distribution and p=2 equivalent to the gamma distribution. For all fitted models, we simulated the residuals with the DHARMa package, as standard residual plots may not be appropriate for GAMs 51. DHARMa scales the residuals between 0 and 1 with a simulation-based approach 51. We also tested for deviation from uniformity, dispersion, outliers, and zero inflation with DHARMa. Non-uniformity, dispersion, outliers, and zero inflation were significant due to the large sample size, but small in effect size in most cases. The simulated quantile-quantile plots from DHARMa suggested that the observed and simulated distributions were generally aligned, with the exception of slight misalignment in the models for the number of previous papers. These analyses are presented in Figure S11 and Table S7.

In addition, we tested for inadequate basis functions using mgcv’s “gam.check()” function 49. Across all smooth predictors and models, we ultimately selected between 10-20 basis functions depending on the variable and outcome measure (counts, rates, papers). We further checked the concurvity of the models and ensured that the worst-case concurvity for all smooth predictors was about 0.8 or less.”

The direction of our results primarily stayed the same, with the exception of gender results. Men tended to self-cite slightly less (or equal self-citation rates) after accounting for numerous covariates. As such, we also modeled the number of previous papers to explain the discrepancy between our raw data and the modeled gender results. Please find the updated results text below (page 11, line 316):

“2.9 Exploring effects of covariates with generalized additive models

Investigating the raw trends and group differences in self-citation rates is important, but several confounding factors may explain some of the differences reported in previous sections. For instance, gender differences in self-citation were previously attributed to men having a greater number of prior papers available to self-cite 7,20,21. As such, covarying for various author- and article-level characteristics can improve the interpretability of self-citation rate trends. To allow for inclusion of author-level characteristics, we only consider First Author and Last Author self-citation in these models.

We used generalized additive models (GAMs) to model the number and rate of self-citations for First Authors and Last Authors separately. The data were randomly subsampled so that each author only appeared in one paper. The terms of the model included several article characteristics (article year, average time lag between article and all cited articles, document type, number of references, field, journal impact factor, and number of authors), as well as author characteristics (academic age, number of previous papers, gender, and whether their affiliated institution is in a low- and middle-income country). Model performance (adjusted R2) and coefficients for parametric predictors are shown in Table 2. Plots of smooth predictors are presented in Figure 6.

First, we considered several career and temporal variables. Consistent with prior works 20,21, self-citation rates and counts were higher for authors with a greater number of previous papers. Self-citation counts and rates increased rapidly among the first 25 published papers but then more gradually increased. Early in the career, increasing academic age was related to greater self-citation. There was a small peak at about five years, followed by a small decrease and a plateau. We found an inverted U-shaped trend for average time lag and self-citations, with self-citations peaking approximately three years after initial publication. In addition, self-citations have generally been decreasing since 2000. The smooth predictors showed larger decreases in the First Author model relative to the Last Author model (Figure 6).

Then, we considered whether authors were affiliated with an institution in a low- and middle-income country (LMIC). LMIC status was determined by the Organisation for Economic Co-operation and Development. We opted to use LMIC instead of affiliation country or continent to reduce the number of model terms. We found that papers from LMIC institutions had significantly lower self-citation counts (-0.138 for First Authors, -0.184 for Last Authors) and rates (-12.7% for First Authors, -23.7% for Last Authors) compared to non-LMIC institutions. Additional results with affiliation continent are presented in Table S5. Relative to the reference level of Asia, higher self-citations were associated with Africa (only three of four models), the Americas, Europe, and Oceania.

Among paper characteristics, a greater number of references was associated with higher self-citation counts and lower self-citation rates (Figure 6). Interestingly, self-citations were greater for a small number of authors, though the effect diminished after about five authors. Review articles were associated with lower self-citation counts and rates. No clear trend emerged between self-citations and journal impact factor. In an analysis by field, despite the raw results suggesting that self-citation rates were lower in Neuroscience, GAM-derived self-citations were greater in Neuroscience than in Psychiatry or Neurology.

Finally, our results aligned with previous findings of nearly equivalent self-citation rates for men and women after including covariates, even showing slightly higher self-citation rates in women. Since raw data showed evidence of a gender difference in self-citation that emerges early in the career but dissipates with seniority, we incorporated two interaction terms: one between gender and academic age and a second between gender and the number of previous papers. Results remained largely unchanged with the interaction terms (Table S6).

2.10 Reconciling differences between raw data and models

The raw and GAM-derived data exhibited some conflicting results, such as for gender and field of research. To further study covariates associated with this discrepancy, we modeled the publication history for each author (at the time of publication) in our dataset (Table 2). The model terms included academic age, article year, journal impact factor, field, LMIC status, gender, and document type. Notably, Neuroscience was associated with the fewest number of papers per author. This explains how authors in Neuroscience could have the lowest raw self-citation rates but the highest self-citation rates after including covariates in a model. In addition, being a man was associated with about 0.25 more papers. Thus, gender differences in self-citation likely emerged from differences in the number of papers, not in any self-citation practices.”

(2) The discussion of the data structure used in the regression models is somewhat opaque, both in the main text and the supplement. From what I gather, these models likely have each citation included in the model at least once (perhaps twice, once for first-author status and one for last-author status), with citations nested within citing papers, cited papers, and authors. Without inclusion of random effects, the interpretation and inference of the estimates may be misleading.

Please see our response to point (1) to address random effects. We have also switched to GAMs (see point #3 below) and provided more detail in the methods. Notably, we decided against using author-level effects due to poor model stability, as there can be as few as one author per group. Instead, we subsampled the dataset such that only one paper appeared from each author.

(3) I am concerned that the use of the inverse hyperbolic sine transform is a bit too prescriptive, and may be producing poor fits to the true predictor-outcome relationships. For example, in a figure like Fig S8, it is hard to know to what extent the sharp drop and sign reversal are true reflections of the data, and to what extent they are artifacts of the transformed fit.

Thank you for raising this point. We have now switched to using generalized additive models (GAMs). GAMs provide a flexible approach to modeling that does not require transformations. We described this in detail in point (1) above and in Methods 4.10 Exploring effects of covariates with generalized additive models (page 21, line 754).

“4.10 Exploring effects of covariates with generalized additive models

For these analyses, we used the full dataset size separately for First and Last Authors (Table S2). This included 115,205 articles and 5,794,926 citations for First Authors, and 114,622 articles and 5,801,367 citations for Last Authors. We modeled self-citation counts, self-citation rates, and number of previous papers for First Authors and Last Authors separately, resulting in six total models.

We found that models could be computationally intensive and unstable when including author-level random effects because in many cases there was only one author per group. Instead, to avoid inappropriately narrow confidence bands, we resampled the dataset such that each author was only represented once. For example, if Author A had five papers in this dataset, then one of their five papers was randomly selected. The random resampling was repeated 100 times as a sensitivity analysis (Figure S12).

For our models, we used generalized additive models from mgcv’s “gam” function in R 48. The smooth terms included all the continuous variables: number of previous papers, academic age, year, time lag, number of authors, number of references, and journal impact factor. The linear terms included all the categorical variables: field, gender affiliation country LMIC status, and document type. We empirically selected a Tweedie distribution 49 with a log link function and p=1.2. The p parameter indicates that the variance is proportional to the mean to the p power 48. The p parameter ranges from 1-2, with p=1 equivalent to the Poisson distribution and p=2 equivalent to the gamma distribution. For all fitted models, we simulated the residuals with the DHARMa package, as standard residual plots may not be appropriate for GAMs 50. DHARMa scales the residuals between 0 and 1 with a simulation-based approach 50. We also tested for deviation from uniformity, dispersion, outliers, and zero inflation with DHARMa. Non-uniformity, dispersion, outliers, and zero inflation were significant due to the large sample size, but small in effect size in most cases. The simulated quantile-quantile plots from DHARMa suggested that the observed and simulated distributions were generally aligned, with the exception of slight misalignment in the models for the number of previous papers. These analyses are presented in Figure S11 and Table S7.

In addition, we tested for inadequate basis functions using mgcv’s “gam.check()” function 48. Across all smooth predictors and models, we ultimately selected between 10-20 basis functions depending on the variable and outcome measure (counts, rates, papers). We further checked the concurvity of the models and ensured that the worst-case concurvity for all smooth predictors was about 0.8 or less.”

(4) It seems there are several points in the analysis where papers may have been dropped for missing data (e.g., missing author IDs and/or initials, missing affiliations, low-confidence gender assessment). It would be beneficial for the reader to know what % of the data was dropped for each analysis, and for comparisons across countries it would be important for the authors to make sure that there is not differential missing data that could affect the interpretation of the results (e.g., differences in self-citation being due to differences in Scopus ID coverage).

Thank you for raising this important point. In the methods section, we describe how the data are missing (page 18, line 623):

“4.3 Data exclusions and missingness

Data were excluded across several criteria: missing covariates, missing citation data, out-of-range values at the citation pair level, and out-of-range values at the article level (Table 3). After downloading the data, our dataset included 157,287 articles and 8,438,733 citations. We excluded any articles with missing covariates (document type, field, year, number of authors, number of references, academic age, number of previous papers, affiliation country, gender, and journal). Of the remaining articles, we dropped any for missing citation data (e.g., cannot identify whether a self-citation is present due to lack of data). Then, we removed citations with unrealistic or extreme values. These included an academic age of less than zero or above 38/44 for First/Last Authors (99th percentile); greater than 266/522 papers for First/Last Authors (99th percentile); and a cited year before 1500 or after 2023. Subsequently, we dropped articles with extreme values that could contribute to poor model stability. These included greater than 30 authors; fewer than 10 references or greater than 250 references; and a time lag of greater than 17 years. These values were selected to ensure that GAMs were stable and not influenced by a small number of extreme values.

In addition, we evaluated whether the data were not missing at random (Table S8). Data were more likely to be missing for reviews relative to articles, for Neurology relative to Neuroscience or Psychiatry, in works from Africa relative to the other continents, and for men relative to women. Scopus ID coverage contributed in part to differential missingness. However, our exclusion criteria also contribute. For example, Last Authors with more than 522 papers were excluded to help stabilize our GAMs. More men fit this exclusion criteria than women.”

Due to differential missingness, we wrote in the limitations (page 16, line 529):

“Ninth, data were differentially missing (Table S8) due to Scopus coverage and gender estimation. Differential missingness could bias certain results in the paper, but we hope that the dataset is large enough to reduce any potential biases.”

Reviewer #2 (Public Review):

The authors provide a comprehensive investigation of self-citation rates in the field of Neuroscience, filling a significant gap in existing research. They analyze a large dataset of over 150,000 articles and eight million citations from 63 journals published between 2000 and 2020. The study reveals several findings. First, they state that there is an increasing trend of self-citation rates among first authors compared to last authors, indicating potential strategic manipulation of citation metrics. Second, they find that the Americas show higher odds of self-citation rates compared to other continents, suggesting regional variations in citation practices. Third, they show that there are gender differences in early-career self-citation rates, with men exhibiting higher rates than women. Lastly, they find that self-citation rates vary across different subfields of Neuroscience, highlighting the influence of research specialization. They believe that these findings have implications for the perception of author influence, research focus, and career trajectories in Neuroscience.

Overall, this paper is well written, and the breadth of analysis conducted by authors, with various interactions between variables (eg. gender vs. seniority), shows that the authors have spent a lot of time thinking about different angles. The discussion section is also quite thorough. The authors should also be commended for their efforts in the provision of code for the public to evaluate their own self-citations. That said, here are some concerns and comments that, if addressed, could potentially enhance the paper:

Thank you for your review and your generally positive view of our work.

(1) There are concerns regarding the data used in this study, specifically its bias towards top journals in Neuroscience, which limits the generalizability of the findings to the broader field. More specifically, the top 63 journals in neuroscience are based on impact factor (IF), which raises a potential issue of selection bias. While the paper acknowledges this as a limitation, it lacks a clear justification for why authors made this choice. It is also unclear how the "top" journals were identified as whether it was based on the top 5% in terms of impact factor? Or 10%? Or some other metric? The authors also do not provide the (computed) impact factors of the journals in the supplementary.

We apologize for the lack of clarity about our selection of journals. We agree that there are limitations to selecting higher impact journals. However, we needed to apply some form of selection in order to make the analysis manageable. For instance, even these 63 journals include over five million citations. We better describe our rationale behind the approach as follows (page 17, line 578):

“We collected data from the 25 journals with the highest impact factors, based on Web of Science impact factors, in each of Neurology, Neuroscience, and Psychiatry. Some journals appeared in the top 25 list of multiple fields (e.g., both Neurology and Neuroscience), so 63 journals were ultimately included in our analysis. We recognize that limiting the journals to the top 25 in each field also limits the generalizability of the results. However, there are tradeoffs between breadth of journals and depth of information. For example, by limiting the journals to these 63, we were able to look at 21 years of data (2000-2020). In addition, the definition of fields is somewhat arbitrary. By restricting the journals to a set of 63 well-known journals, we ensured that the journals belonged to Neurology, Neuroscience, or Psychiatry research. It is also important to note that the impact factor of these journals has not necessarily always been high. For example, Acta Neuropathologica had an impact factor of 17.09 in 2020 but 2.45 in 2000. To further recognize the effects of impact factor, we decided to include an impact factor term in our models.”

In addition, we have now provided the 2020 impact factors in Table S1.

By exclusively focusing on high impact journals, your analysis may not be representative of the broader landscape of self-citation patterns across the neuroscience literature, which is what the title of the article claims to do.

We agree that this article is not indicative of all neuroscience literature, but rather the top journals. Thus, we have changed the title to: “Trends in Self-citation Rates in High-impact Neurology, Neuroscience, and Psychiatry Journals”. We would also like to note that compared to previous bibliometrics works in neuroscience (Bertolero et al. 2020; Dworkin et al. 2020; Fulvio et al. 2021), this article includes a wider range of data.

(2) One other concern pertains to the possibility that a significant number of authors involved in the paper may not be neuroscientists. It is plausible that the paper is a product of interdisciplinary collaboration involving scientists from diverse disciplines. Neuroscientists amongst the authors should be identified.

In our opinion, neuroscience is a broad, interdisciplinary field. Individuals performing neuroscience research may have a neuroscience background. Yet, they may come from many backgrounds, such as physics, mathematics, biology, chemistry, or engineering. As such, we do not believe that it is feasible to characterize whether each author considers themselves a neuroscientist or not. We have added the following to the limitations section (page 16, line 528):

“Eighth, authors included in this work may not be neurologists, neuroscientists, or psychiatrists. However, they still publish in journals from these fields.”

(3) When calculating self-citation rate, it is important to consider the number of papers the authors have published to date. One plausible explanation for the lower self-citation rates among first authors could be attributed to their relatively junior status and short publication record. As such, it would also be beneficial to assess self-citation rate as a percentage relative to the author's publication history. This number would be more accurate if we look at it as a percentage of their publication history. My suspicion is that first authors (who are more junior) might be more likely to self-cite than their senior counterparts. My suspicion was further raised by looking at Figures 2a and 3. Considering the nature of the self-citation metric employed in the study, it is expected that authors with a higher level of seniority would have a greater number of publications. Consequently, these senior authors' papers are more likely to be included in the pool of references cited within the paper, hence the higher rate.

While the authors acknowledge the importance of the number of past publications in their gender analysis, it is just as important to include the interplay of seniority in (1) their first and last author self-citation rates and (2) their geographic analysis.

Thank you for this thoughtful comment. We agree that seniority and prior publication history play an important role in self-citation rates.

For comparing First/Last Author self-citation rates, we have now included a plot similar to Figure 2a, where self-citation as a percentage of prior publication history is plotted.

(page 4, line 161): “Analyzing self-citations as a fraction of publication history exhibited a similar trend (Figure S3). Notably, First Authors were more likely than Last Authors to self-cite when normalized by prior publication history.

For the geographic analysis, we made two new maps: 1) that of the number of previous papers, and 2) that of the journal impact factor (see response to point #4 below).

(page 5, line 185): “We also investigated the distribution of the number of previous papers and journal impact factor across countries (Figure S4). Self-citation maps by country were highly correlated with maps of the number of previous papers (Spearman’s r\=0.576, P=4.1e-4; 0.654, P=1.8e-5 for First and Last Authors). They were significantly correlated with maps of average impact factor for Last Authors (0.428, P=0.014) but not Last Authors (Spearman’s r\=0.157, P=0.424). Thus, further investigation is necessary with these covariates in a comprehensive model.”

Finally, we included a model term for the number of previous papers (Table 2). We analyzed this both for self-citation counts and self-citation rates and found a strong relationship between publication history and self-citations. We also included the following section where we modeled the number of previous papers for each author (page 13, line 384):

“2.10 Reconciling differences between raw data and models

The raw and GAM-derived data exhibited some conflicting results, such as for gender and field of research. To further study covariates associated with this discrepancy, we modeled the publication history for each author (at the time of publication) in our dataset (Table 2). The model terms included academic age, article year, journal impact factor, field, LMIC status, gender, and document type. Notably, Neuroscience was associated with the fewest number of papers per author. This explains how authors in Neuroscience could have the lowest raw self-citation rates but the highest self-citation rates after including covariates in a model. In addition, being a man was associated with about 0.25 more papers. Thus, gender differences in self-citation likely emerged from differences in the number of papers, not in any self-citation practices.”

(4) Because your analysis is limited to high impact journals, it would be beneficial to see the distribution of the impact factors across the different countries. Otherwise, your analysis on geographic differences in self-citation rates is hard to interpret. Are these differences really differences in self-citation rates, or differences in journal impact factor? It would be useful to look at the representation of authors from different countries for different impact factors.

We made a map of this in Figure S4 (see our response to point #3 above).

(page 5, line 185): “We also investigated the distribution of the number of previous papers and journal impact factor across countries (Figure S4). Self-citation maps by country were highly correlated with maps of the number of previous papers (Spearman’s r=0.576, P=4.1e-4; 0.654, P=1.8e-5 for First and Last Authors). They were significantly correlated with maps of average impact factor for Last Authors (0.428, P=0.014) but not Last Authors (Spearman’s r=0.157, P=0.424). Thus, further investigation is necessary with these covariates in a comprehensive model.”

We also included impact factor as a term in our model. The results suggest that there are still geographic differences (Table 2, Table S5).

(5) The presence of self-citations is not inherently problematic, and I appreciate the fact that authors omit any explicit judgment on this matter. That said, without appropriate context, self-citations are also not the best scholarly practice. In the analysis on gender differences in self-citations, it appears that authors imply an expectation of women's self-citation rates to align with those of men. While this is not explicitly stated, use of the word "disparity", and also presentation of self-citation as an example of self-promotion in discussion suggest such a perspective. Without knowing the context in which the self-citation was made, it is hard to ascertain whether women are less inclined to self-promote or that men are more inclined to engage in strategic self-citation practices.

We agree that on the level of an individual self-citation, our study is not useful for determining how related the papers are. Yet, understanding overall trends in self-citation may help to identify differences. Context is important, but large datasets allow us to investigate broad trends. We added the following text to the limitations section (page 16, line 524):

“In addition, these models do not account for whether a specific citation is appropriate, as some situations may necessitate higher self-citation rates.”

Reviewer #3 (Public Review):

This paper analyses self-citation rates in the field of Neuroscience, comprising in this case, Neurology, Neuroscience and Psychiatry. Based on data from Scopus, the authors identify self-citations, that is, whether references from a paper by some authors cite work that is written by one of the same authors. They separately analyse this in terms of first-author self-citations and last-author self-citations. The analysis is well-executed and the analysis and results are written down clearly. There are some minor methodological clarifications needed, but more importantly, the interpretation of some of the results might prove more challenging. That is, it is not always clear what is being estimated, and more importantly, the extent to which self-citations are "problematic" remains unclear.

Thank you for your review. We attempted to improve the interpretation of results, as described in the following responses.

When are self-citations problematic? As the authors themselves also clarify, "self-citations may often be appropriate". Researchers cite their own previous work for perfectly good reasons, similar to reasons of why they would cite work by others. The "problem", in a sense, is that researchers cite their own work, just to increase the citation count, or to promote their own work and make it more visible. This self-promotional behaviour might be incentivised by certain research evaluation procedures (e.g. hiring, promoting) that overly emphasise citation performance. However, the true problem then might not be (self-)citation practices, but instead, the flawed research evaluation procedures that emphasis citation performance too much. So instead of problematising self-citation behaviour, and trying to address it, we might do better to address flawed research evaluation procedures. Of course, we should expect references to be relevant, and we should avoid self-promotional references, but addressing self-citations may just have minimal effects, and would not solve the more fundamental issue.

We agree that this dataset is not designed to investigate the downstream effects of self-citations. However, self-citation practices are more likely to be problematic when they differ across specific groups. This work can potentially spark more interest in future longitudinal designs to investigate whether differences in self-citation practices leads to differences in career outcomes, for example. We added the following text to clarify (page 17, line 565):

“Yet, self-citation practices become problematic when they are different across groups or are used to “game the system.” Future work should investigate the downstream effects of self-citation differences to see whether they impact the career trajectories of certain groups. We hope that this work will help to raise awareness about factors influencing self-citation practices to better inform authors, editors, funding agencies, and institutions in Neurology, Neuroscience, and Psychiatry.”

Some other challenges arise when taking a statistical perspective. For any given paper, we could browse through the references, and determine whether a particular reference would be warranted or not. For instance, we could note that there might be a reference included that is not at all relevant to the paper. Taking a broader perspective, the irrelevant reference might point to work by others, included just for reasons of prestige, so-called perfunctory citations. But it could of course also include self-citations. When we simply start counting all self-citations, we do not see what fraction of those self-citations would be warranted as references. The question then emerges, what level of self-citations should be counted as "high"? How should we determine that? If we observe differences in self-citation rates, what does it tell us?

Our focus is when the self-citation practices differ across groups. We agree that, on a case-by-case basis, there is no exact number for a self-citation rate that is “high.” With a dataset of the current size, evaluating whether each individual self-citation is appropriate is not feasible. If we observe differences in self-citation rate, this may tell us about broad (not individual-level) trends and differences in self-citing practice. If one group is self-citing much more highly compared to another group–even after covarying relevant variables such as prior publication history–then the self-citation differences can likely be attributed to differences in self-citation practices/behaviors.

For example, the authors find that the (any author) self-citation rate in Neuroscience is 10.7% versus 15.9% in Psychiatry. What does this difference mean? Are psychiatrists citing themselves more often than neuroscientists? First author men showed a self-citation rate of 5.12% versus a self-citation rate of 3.34% of women first authors. Do men engage in more problematic citation behaviour? Junior researchers (10-year career) show a self-citation rate of about 5% compared to a self-citation rate of about 10% for senior researchers (30-year career). Are senior researchers therefore engaging in more problematic citation behaviour? The answer is (most likely) "no", because senior authors have simply published more, and will therefore have more opportunities to refer to their own work. To be clear: the authors are aware of this, and also take this into account. In fact, these "raw" various self-citation rates may, as the authors themselves say, "give the illusion" of self-citation rates, but these are somehow "hidden" by, for instance, career seniority.

We included numerous covariates in our model. In addition, to address the difference between “raw” and “modeled” self-citation rates, we added the following section (page 13, line 384):

“2.10 Reconciling differences between raw data and models

The raw and GAM-derived data exhibited some conflicting results, such as for gender and field of research. To further study covariates associated with this discrepancy, we modeled the publication history for each author (at the time of publication) in our dataset (Table 2). The model terms included academic age, article year, journal impact factor, field, LMIC status, gender, and document type. Notably, Neuroscience was associated with the fewest number of papers per author. This explains how authors in Neuroscience could have the lowest raw self-citation rates but the highest self-citation rates after including covariates in a model. In addition, being a man was associated with about 0.25 more papers. Thus, gender differences in self-citation likely emerged from differences in the number of papers, not in any self-citation practices.”

Again, the authors do consider this, and "control" for career length and number of publications, et cetera, in their regression model. Some of the previous observations then change in the regression model. Neuroscience doesn't seem to be self-citing more, there just seem to be junior researchers in that field compared to Psychiatry. Similarly, men and women don't seem to show an overall different self-citation behaviour (although the authors find an early-career difference), the men included in the study simply have longer careers and more publications.

But here's the key issue: what does it then mean to "control" for some variables? This doesn't make any sense, except in the light of causality. That is, we should control for some variable, such as seniority, because we are interested in some causal effect. The field may not "cause" the observed differences in self-citation behaviour, this is mediated by seniority. Or is it confounded by seniority? Are the overall gender differences also mediated by seniority? How would the selection of high-impact journals "bias" estimates of causal effects on self-citation? Can we interpret the coefficients as causal effects of that variable on self-citations? If so, would we try to interpret this as total causal effects, or direct causal effects? If they do not represent causal effects, how should they be interpreted then? In particular, how should it "inform author, editors, funding agencies and institutions", as the authors say? What should they be informed about?

We apologize for our misuse of language. We will be more clear, as in most previous self-citation papers, that our analysis is NOT causal. Causal datasets do have some benefits in citation research, but a limitation is that they may not cover as wide of a range of authors. Furthermore, non-causal correlational studies can still be useful in informing authors, editors, funding agencies, and institutions. Association studies are widely used across various fields to draw non-causal conclusions. We made numerous changes to reduce our causal language.

Before: “We then developed a probability model of self-citation that controls for numerous covariates, which allowed us to obtain significance estimates for each variable of interest.”

After (page 3, line 113): “We then developed a probability model of self-citation that includes numerous covariates, which allowed us to obtain significance estimates for each variable of interest.”

Before: “As such, controlling for various author- and article-level characteristics can improve the interpretability of self-citation rate trends.”

After (page 11, line 321): “As such, covarying various author- and article-level characteristics can improve the interpretability of self-citation rate trends.”

Before: “Initially, it appeared that self-citation rates in Neuroscience are lower than Neurology and Psychiatry, but after controlling for various confounds, the self-citation rates are higher in Neuroscience.”

After (page 15, line 468): “Initially, it appeared that self-citation rates in Neuroscience are lower than Neurology and Psychiatry, but after considering several covariates, the self-citation rates are higher in Neuroscience.”

We also added the following text to the limitations section (page 16, line 526):

“Seventh, the analysis presented in this work is not causal. Association studies are advantageous for increasing sample size, but future work could investigate causality in curated datasets.”

The authors also "encourage authors to explore their trends in self-citation rates". It is laudable to be self-critical and review ones own practices. But how should authors interpret their self-citation rate? How useful is it to know whether it is 5%, 10% or 15%? What would be the "reasonable" self-citation rate? How should we go about constructing such a benchmark rate? Again, this would necessitate some causal answer. Instead of looking at the self-citation rate, it would presumably be much more informative to simply ask authors to check whether references are appropriate and relevant to the topic at hand.

We believe that our tool is valuable for authors to contextualize their own self-citation rates. For instance, if an author has published hundreds of articles, it is not practical to count the number of self-citations in each. We have added two portions of text to the limitations section:

(page 16, line 524): “In addition, these models do not account for whether a specific citation is appropriate, though some situations may necessitate higher self-citation rates.”

(page 16, line 535): “Despite these limitations, we found significant differences in self-citation rates for various groups, and thus we encourage authors to explore their trends in self-citation rates. Self-citation rates that are higher than average are not necessarily wrong, but suggest that authors should further reflect on their current self-citation practices.”

In conclusion, the study shows some interesting and relevant differences in self-citation rates. As such, it is a welcome contribution to ongoing discussions of (self) citations. However, without a clear causal framework, it is challenging to interpret the observed differences.

We agree that causal studies provide many benefits. Yet, association studies also provide many benefits. For example, an association study allowed us to analyze a wider range of articles than a causal study would have.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Statistical suggestions:

(1) To improve statistical inference, nesting should be accounted for in all of the analyses. For example, the logistic regression model using citing/cited pairs should include random effects for article, author, and perhaps subfield, in order for independence of observations to be plausible. Similarly, bootstrapping and permutation would ideally occur at the author level rather than (or in addition to) the paper level.

Detailed updates addressing these points are in the public review. In short, we found computational challenges with many levels of the random effects (>100,000) and millions of observations at the citation pairs level. As such, we decided to model citations rates and counts by paper. In this case, we found that results could be unstable when including author-level random effects because in many cases there was only one author per group. Instead, to avoid inappropriately narrow confidence bands, we resampled the dataset such that each author was only represented once. For example, if Author A had five papers in this dataset, then one of their five papers was randomly selected. We repeated the random resampling 100 times (Figure S12). We updated our description of our models in the Methods section (page 21, line 754).

For permutation tests and bootstrapping, we now define an “exchangeability block” as a co-authorship group of authors. In this dataset, that meant any authors who published together (among the articles in this dataset) as a First Author / Last Author pairing were assigned to the same exchangeability block. It is not realistic to check for overlapping middle authors in all papers because of the collaborative nature of the field. In addition, we believe that self-citations are primarily controlled by first and last authors, so we can assume that middle authors do not control self-citation habits. We then performed bootstrapping and permutation tests in the constraints of the exchangeability blocks.

(2) In general, I am having trouble understanding the structure of the regression models. My current belief is that rows are composed of individual citations from papers' reference lists, with the outcome representing their status as a self-citation or not, and with various citing article and citing author characteristics as predictors. However, the fact that author type is included in the model as a predictor (rather than having a model for FA self-citations and another for LA self-citations) suggests to me that each citation is entered as two separate rows - once noting whether it was a FA self-citation and once noting whether it was an LA self-citation - and then it is run as a single model.

(2a) If I am correct, the model is unlikely to be producing valid inference. I would recommend breaking this analysis up into two separate models, and including article-, author-, and subfield-level random effects. You could theoretically include a citation-level random effect and keep it as one model, but each 'group' would only have two observations and the model would be fairly unstable as a result.

(2b) If I am misunderstanding (and even if not), I would encourage you to provide a more detailed description of the dataset structure and the model - perhaps with a table or diagram

We split the data into two models and decided to model on the level of a paper (self-citation rate and self-citation count). In addition, we subsampled the dataset such that each author only appears once to avoid misestimation of confidence intervals (see point (1) above). As described in the public review, we included much more detail in our methods section now to improve the clarity of our models.

(3) I would suggest removing the inverse hyperbolic sine transform and replacing it with a more flexible approach to estimating the relationships' shape, like generalized additive models or other spline-based methods to ensure that the chosen method is appropriate - or at the very least checking that it is producing a realistic fit that reflects the underlying shape of the relationships.

More details are available in the public review, but we now use GAMs throughout the manuscript.

(4) For the "highly self-citing" analysis, it is unclear why papers in the 15-25% range were dropped rather than including them as their own category in an ordinal model. I might suggest doing the latter, or explaining the decision more fully

We previously included this analysis as a paper-level model because our main model was at the level of citation pairs. Now, we removed this analysis because we model self-citation rates and counts by paper.

(5) It would be beneficial for the reader to know what % of the data was dropped for each analysis, and for your team to make sure that there is not differential missing data that could affect the interpretation of the results (e.g., differences in self-citation being due to differences in Scopus ID coverage).

Thank you for this suggestion. We added more detailed missingness data to 4.3 Data exclusions and missingness. We did find differential missingness and added it to the limitations section. However, certain aspects of this cannot be corrected because the data are just not available (e.g., Scopus coverage issues). Further details are available in the public review.

Conceptual thoughts:

(1) I agree with your decision to focus on the second definition of self-citation (self-cites relative to my citations to others' work) rather than the first (self-cites relative to others' citations to my work). But it does seem that the first definition is relevant in the context of gaming citation metrics. For example, someone who writes one paper per year with a reference list of 30% self-citations will have much less of an impact on their H-index than someone who writes 10 papers per year with 10% self-citations. It could be interesting to see how these definitions interact, and whether people who are high on one measure tend to be high on the other.

We agree this would be interesting to investigate in the future. Unfortunately, our dataset is organized at the level of the paper and thus does not contain information regarding how many times the authors cite a particular work. We hope that we can explore this interaction in the future.

(2) This is entirely speculative, but I wonder whether the increasing rate of LA self-citation relative to FA self-citation is partly due to PIs over-citing their own lab to build up their trainees' citation records and help them succeed in an increasingly competitive job market. This sounds more innocuous than doing it to benefit their own reputation, but it would provide another mechanism through which students from large and well-funded labs get a leg-up in the job market. Might be interesting to explore, though I'm not exactly sure how :)

This is a very interesting point. We do not have any means to investigate this with the current dataset, but we added it to the discussion (page 14, line 421):

“A third, more optimistic explanation is that principal investigators (typically Last Authors) are increasingly self-citing their lab’s papers to build up their trainee’s citation records for an increasingly competitive job market.”

Reviewer #2 (Recommendations For The Authors):

(1) In regards to point 1 in the public review: In the spirit of transparency, the authors would benefit from providing a rationale for their choice of top journals, and the methodology used to identify them. It would also be valuable to include the impact factor of each journal in the S1 table alongside their names.

Given the availability and executability of code, it would be useful to see how and if the self-citation trends vary amongst the "low impact" journals (as measured by the IF). This could go in any of the three directions:

a. If it is found that self-citations are not as prevalent in low impact journals, this could be a great starting point for a conversation around the evaluation of journals based on impact factor, and the role of self-citations in it.

b. If it is found that self-citations are as prevalent in low impact journals as high impact journals, that just strengthens your results further.

c. If it is found that self-citations are more prevalent in low impact journals, this would mean your current statistics are a lower bound to the actual problem. This is also intuitive in the sense that high impact journals get more external citations (and more exposure) than low impact journals, as such authors (and journals) may be less likely to self-cite.

Expanding the dataset to include many more journals was not feasible. Instead, we included an impact factor term in our models, as detailed in the public review. We found no strong trends in the association between impact factor and self-citation rate/count. Another important note is that these journals were considered “high impact” in 2020, but many had lower impact factors in earlier years. Thus, our modeling allows us to estimate how impact factor is related to self-citations across a wide range of impact factors.

It is crucial to consider utilizing such a comprehensive database as Scopus, which provides a more thorough list of all journals in Neuroscience, to obtain a more representative sample. Alternatively, other datasets like Microsoft Academic Graph, and OpenAlex offer information on the field of science associated with each paper, enabling a more comprehensive analysis.

We agree that certain datasets may offer a wider view of the entire field. However, we included a large number of papers and journals relative to previous studies. In addition, Scopus provides a lot of detailed and valuable author-level information. We had to limit our calls to the Scopus API so restricted journals by 2020 impact factor.

(2) In regards to point 2 in the public review: To enhance the accuracy and specificity of the analysis, it would be beneficial to distinguish neuroscientists among the co-authors. This could be accomplished by examining their publication history leading up to the time of publication of the paper, and identify each author's level of engagement and specialization within the field of neuroscience.

Since the field of neuroscience is largely based on collaborations, we find that it might be impossible to determine who is a neuroscientist. For example, a researcher with a publication history in physics may now be focusing on computational neuroscience research. As such, we feel that our current work, which ensures that the papers belong to neuroscience, is representative of what one may expect in terms of neuroscience research and collaboration.

(3) In regards to point 3 in the public review: I highly recommend plotting self-citation rate as the number of papers in the reference list over the number of total publications to date of paper publication.

As described in the public review, we have now done this (Figure S3).

(4) In regards to point 5 in the public review: It would be useful to consider the "quality" of citations to further the discussion on self-citations. For instance, differentiating between self-citations that are perfunctory and superficial from those that are essential for showing developmental work, would be a valuable contribution.

Other databases may have access to this information, but ours unfortunately does not. We agree that this is an interesting area of work.

(5) The authors are to be commended for their logistic regression models, as they control for many confounders that were lacking in their earlier descriptive statistics. However, it would be beneficial to rerun the same analysis but on a linear model whereby the outcome variable would be the number of self-citations per author. This would possibly resolve many of the comments mentioned above.

Thank you for your suggestion. As detailed in the public review, we now model the number of self-citations. This is modeled on the paper level, not the author level, because our dataset was downloaded by paper, not by author.

Minor suggestions:

(1) Abstract says one of your findings is: "increasing self-citation rates of First Authors relative to Last Authors". Your results actually show the opposite (see Figure 1b).

Thank you for catching this error. We corrected it to match the results and discussion in the paper:

“…increasing self-citation rates of Last Authors relative to First Authors.”

(2) It might be interesting to compute an average academic age for each paper, and look at self-citation vs average academic age plot.

We agree that this would be an interesting analysis. However, to limit calls to the API, we collected academic age data only on First and Last Authors.

(3) It may be interesting to look at the distribution of women in different subfields within neuroscience, and the interaction of those in the context of self-citations.

Thank you for this interesting suggestion. We added the following analysis (page 9, line 305):

“Furthermore, we explored topic-by-gender interactions (Figure S10). In short, men and women were relatively equally represented as First Authors, but more men were Last Authors across all topics. Self-citation rates were higher for men across all topics.”

Reviewer #3 (Recommendations For The Authors):

- In the abstract, "flaws in citation practices" seems worded rather strongly.

We respectfully disagree, as previous works have shown significant bias in citation practices. For example, Dworkin et al. (Dworkin et al. 2020) found that neuroscience reference lists tended to under-cite women, even after including various covariates.

- Links of the references to point to (non-accessible) paperpile references, you would probably want to update this.

We apologize for the inconvenience and have now removed these links.

- p 2, l 24: The explanation of ref. (5) seems to be a bit strangely formulated. The point of that article is that citations to work that reinforce a particular belief are more likely to be cited, which *creates* unfounded authority. The unfounded authority itself is hence no part of the citation practices

Thank you for catching our misinterpretation. We have now removed this part of the sentence.

- p 3, l 16: "h indices" or "citations" instead of "h-index".

We now say “h-indices”.

- p 5, l 5: how was the manual scoring done?

We added the following to the caption of Figure S1.

“Figure S1. Comparison between manual scoring of self-citation rates and self-citation rates estimated from Python scripts in 5 Psychiatry journals: American Journal of Psychiatry, Biological Psychiatry, JAMA Psychiatry, Lancet Psychiatry, and Molecular Psychiatry. 906 articles in total were manually evaluated (10 articles per journal per year from 2000-2020, four articles excluded for very large author list lengths and thus high difficulty of manual scoring). For manual scoring, we downloaded information about all references for a given article and searched for matching author names.”

- p 5, l 23: Why this specific p-value upper bound of 4e-3? From later in the article, I understand that this stems from the 10000 bootstrap sample, with then taking a Bonferroni correction? Perhaps good to clarify this briefly somewhere.

Thank you for this suggestion. We now perform Benjamini/Hochberg false discovery rate (FDR) correction, but we added a description of the minimum P value from permutations (page 21, line 748):

“All P values described in the main text were corrected with the Benjamini/Hochberg 16 false discovery rate (FDR) correction. With 10,000 permutations, the lowest P value after applying FDR correction is P=2.9e-4, which indicates that the true point would be the most extreme in the simulated null distribution.”

- Fig. 1, caption: The (a) and (b) labelling here is a bit confusing, because the first sentence suggests both figures portray the same, but do so for different time periods. Perhaps rewrite, so that (a) and (b) are both described in a single sentence, instead of having two different references to (a) and (b).

Thank you for pointing this out. We fixed the labeling of this caption:

“Figure 1. Visualizing recent self-citation rates and temporal trends. a) Kernel density estimate of the distribution of First Author, Last Author, and Any Author self-citation rates in the last five years. b) Average self-citation rates over every year since 2000, with 95% confidence intervals calculated by bootstrap resampling.”

- p7, l 9: Regarding "academic age", note that there might be a difference between "age" effects and "cohort" effects. That is, there might be difference between people with a certain career age who started in 1990 and people with the same career age, but who started in 2000, which would be a "cohort" effect.

We agree that this is a possible effect and have added it to the limitations (page 16, line 532):

“Tenth, while we considered academic age, we did not consider cohort effects. Cohort effects would depend on the year in which the individual started their career.”

- p 7, l 15: "jumps" suggests some sort of sudden or discontinuous transition, I would just say "increases".

We now say “increases.”

- Fig. 2: Perhaps it should be made more explicit that this includes only academics with at least 50 papers. Could the authors please clarify whether the same limitation of at least 50 papers also features in other parts of the analysis where academic age is used? This selection could affect the outcomes of the analysis, so its consequences should be carefully considered. One possibility for instance is that it selects people with a short career length who have been exceptionally productive, namely those that have had 50 papers, but only started publishing in 2015 or so. Such exceptionally productive people will feature more highly in the early career part, because they need to be so productive in order to make the cut. For people with a longer career, the 50 papers would be less of a hurdle, and so would select more and less productive people more equally.

We apologize for the lack of clarity. We did not use this requirement where academic age was used. We mainly applied this requirement when aggregating by country, as we did not want to calculate self-citation rate in a country based on only several papers. We have clarified various data exclusions in our new section 4.3 Data exclusions and missingness.

- p 8, l 11: The affiliated institution of an author is not static, but rather changes throughout time. Did the authors consider this? If not, please clarify that this refers to only the most recent affiliation (presumably). Authors also often have multiple affiliations. How did the authors deal with this?

The institution information is at the time of publication for each paper. We added more detail to our description of this on page 19, line 656:

“For both First and Last Authors, we found the country of their institutional affiliation listed on the publication. In the case of multiple affiliations, the first one listed in Scopus was used.”

- p 10, l 6: How were these self-citation rates calculated? This is averaged per author (i.e. only considering papers assigned to a particular topic) and then averaged across authors? (Note that in this way, the average of an author with many papers will weigh equally with the average of an author with few papers, which might skew some of the results).

We calculate it across the entire topic (i.e., do NOT calculate by author first). We updated the description as follows (page 7, line 211):

“We then computed self-citation rates for each of these topics (Figure 4) as the total number of self-citations in each topic divided by the total number of references in each topic…”

- p 13, l 18: Is the academic age analysis here again limited to authors having at least 50 papers?

This is not limited to at least 50 papers. To clarify, the previous analysis was not limited to authors with 50 papers. It was instead limited to ages in our dataset that had at least 50 data points. e.g., If an academic age of 70 only had 20 data points in our dataset, it would have been excluded.

- Fig. 5: Here, comparing Fig. 5(d) and 5(f) suggests that partly, the self-citation rate differences between men and women, might be the result of the differences in number of papers. That is, the somewhat higher self-citation rate at a given academic age, might be the result of the higher number of papers at that academic age. It seems that this is not directly described in this part of the analysis (although this seems to be the case from the later regression analysis).

We agree with this idea and have added a new section as follows (page 13, line 384):

“2.10 Reconciling differences between raw data and models

The raw and GAM-derived data exhibited some conflicting results, such as for gender and field of research. To further study covariates associated with this discrepancy, we modeled the publication history for each author (at the time of publication) in our dataset (Table 2). The model terms included academic age, article year, journal impact factor, field, LMIC status, gender, and document type. Notably, Neuroscience was associated with the fewest number of papers per author. This explains how authors in Neuroscience could have the lowest raw self-citation rates by highest self-citation rates after including covariates in a model. In addition, being a man was associated with about 0.25 more papers. Thus, gender differences in self-citation likely emerged from differences in the number of papers, not in any self-citation practices.”

- Section 2.10. Perhaps the authors could clarify that this analysis takes individual articles as the unit of analysis, not citations.

We updated all our models to take individual articles and have clarified this with more detailed tables.

- p 18, l 10: "Articles with between 15-25% self-citation rates were 10 discarded" Why?

We agree that these should not be discarded. However, we previously included this analysis as a paper-level model because our main model was at the level of citation pairs. Now, we removed this analysis because we model self-citation rates and counts by paper.

- p 20, l 5: "Thus, early-career researchers may be less incentivized to 5 self-promote (e.g., self-cite) for academic gains compared to 20 years ago." How about the possibility that there was less collaboration, so that first authors would be more likely to cite their own paper, whereas with more collaboration, they will more often not feature as first author?

This is an interesting point. We feel that more collaboration would generally lead to even more self-citations, if anything. If an author collaborates more, they are more likely to be on some of the references as a middle author (which by our definition counts toward self-citation rates).

- p 20, l 15: Here the authors call authors to avoid excessive self-citations. Of course, there's nothing wrong with calling for that, but earlier the authors were more careful to not label something directly as excessive self-citations. Here, by stating it like this, the authors suggest that they have looked at excessive self-citations.

We rephrased this as follows:

Before: “For example, an author with 30 years of experience cites themselves approximately twice as much as one with 10 years of experience on average. Both authors have plenty of works that they can cite, and likely only a few are necessary. As such, we encourage authors to be cognizant of their citations and to avoid excessive self-citations.”

After: “For example, an author with 30 years of experience cites themselves approximately twice as much as one with 10 years of experience on average. Both authors have plenty of works that they can cite, and likely only a few are necessary. As such, we encourage authors to be cognizant of their citations and to avoid unnecessary self-citations.”

- p 22, l 11: Here again, the same critique as p 20, l15 applies.

We switched “excessively” to “unnecessarily.”

- p 23, l 12: The authors here critique ref. (21) of ascertainment bias, namely that they are "including only highly-achieving researchers in the life 12 sciences". But do the authors not do exactly the same thing? That is, they also only focus on the top high-impact journals.

We included 63 high-impact journals with tens of thousands of authors. In addition, some of these journals were not high-impact at the time of publication. For example, Acta Neuropathologica had an impact factor of 17.09 in 2020 but 2.45 in 2000. This still is a limitation of our work, but we do cover a much broader range of works than the listed reference (though their analysis also has many benefits since it included more detailed information).

- p 26, l 22-26: It seems that the matching is done quite broadly (matching last names + initials at worst) for self-citations, while later (in section 4.9, p 31, l 9), the authors switch to only matching exact Scopus Author IDs. Why not use the same approach throughout? Or compare the two definitions (narrow / broad).

Thank you for catching this mistake. We now use the approach of matching Scopus Author IDs throughout.

- S8: it might be nice to explore open alternatives, such as OpenAlex or OpenAIRE, instead of the closed Scopus database, which requires paid access (which not all institutions have, perhaps that could also be corrected in the description in GitHub).

Thank you for this suggestion. Unfortunately, switching databases would require starting our analysis from the beginning. On our GitHub page, we state: “Please email matthew.rosenblatt@yale.edu if you have trouble running this or do not have institutional access. We can help you run the code and/or run it for you and share your self-citation trends.” We feel that this will allow us to help researchers who may not have institutional access. In addition, we released our aggregated, de-identified (title and paper information removed) data on GitHub for other researchers to use.

Author response:

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

Reviewer #1 (Public Review):

[...] Based on these results, the authors support a model whereby kinetic regimes are encoded in the cis-regulatory sequences of a gene instead of imposed by an evolving trans-regulatory environment.

The question asked in this manuscript is important and the eve locus represents an ideal paradigm to address it in a quantitative manner. Most of the results are correctly interpreted and well-presented. However, the main conclusion pointing towards a potential "unified theory" of burst regulation during Drosophila embryogenesis should be nuanced or cross-validated.

Our results and those of others suggest that different developmental genes follow unified—yet different—transcriptional control strategies whereby different combinations of bursting parameters are regulated to modulate gene expression: burst frequency and amplitude for eve (Berrocal et al., 2020), and burst frequency and duration for gap genes (Zoller et al., 2018). In light of the aforementioned works, we can only claim that our results suggest a unified strategy for eve, our case of study, as we observe that eve regulatory strategies are robust to disruption of enhancers and binding sites. In the Discussion section of our revised manuscript, we will emphasize that the bursting control strategy we uncovered for eve does not necessarily apply to other genes, and speculate in more detail that genes that employ the same strategy of transcriptional bursting may be grouped in families that share a common molecular mechanism of transcription.

Manuscript updates:

We have emphasized in the Discussion section that our claim of unified strategies pertains exclusively to the bursting behavior of the gene even-skipped, and do not necessarily extend to other genes. To clarify this point, we referenced the findings of (Zoller, Little, and Gregor 2018) and (Chen et al. 2023), who observed that the bursting control strategy of Drosophila gap genes relies on the modulation of burst frequency and duration. Additionally, we cited the findings of (Syed, Duan, and Lim 2023), who reported a decrease in bursting amplitude and duration upon disruption of Dorsal binding sites on the snail minimal distal enhancer. Both examples describe bursting control strategies that differ from the modulation of burst frequency and amplitude observed for even-skipped.

In addition to the lack of novelty (some results concerning the fact that koff does not change along the A/P axis/the idea of a 'unified regime' were already obtained in Berrocal et al 2020),...

Unfortunately, we believe there is a misunderstanding in terms of what we construe as novelty in our work. In our previous work (Berrocal et al., 2020), we observed that the seven stripes of even-skipped (eve) expression modulate transcriptional bursting through the same strategy—bursting frequency and amplitude are controlled to yield various levels of mRNA synthesis, while burst duration remains constant. We reproduce that result in our paper, and do not claim any novelty. However, what was unclear is whether the observed eve bursting control strategy would only exist in the wild-type stripes, whose expression—we reasoned—is under strong selection due to the dramatic phenotypic consequences of eve transcription, or if eve transcriptional bursting would follow the same strategy under trans-regulatory environments that are not under selection to deliver specific spatiotemporal dynamics of eve expression. Our results—and here lies the novelty of our work—support the second scenario, and point to a model where eve bursting strategies do not result from adaptation of eve activity to specific trans-regulatory environments. Instead, we speculate that a molecular mechanism constrains eve bursting strategy whenever and wherever the gene is active. This is something that we could not have known from our first study in (Berrocal et al., 2020) and constitutes the main novelty of our paper. To put this in other words, the novelty of our work does not rest on the fact that both burst frequency and amplitude are modulated in the endogenous eve pattern, but that this modulation remains quantitatively indistinguishable when we focus on ectopic areas of expression. We will make this point clearer in the Introduction and Discussion section of our revised manuscript.

Manuscript updates:

We have clarified this point in both the Introduction and Discussion sections. In the updated Introduction, we state that while our previous work (Berrocal et al. 2020) examined bursting strategies in endogenous expression regions that are, in principle, subject to selection, the present study induced the formation of ectopic expression patterns to probe bursting strategies in regions presumably devoid of evolutionary pressures. In the Discussion section, we highlight that the novelty of our work lies in the insights derived from the comparative analysis between ectopic and endogenous regions of even-skipped expression, an aspect not addressed in our previous work.

… note i) the limited manipulation of TF environment;...

We acknowledge that additional genetic manipulations would make it possible to further test the model. However, we hope that the reviewer will agree with us that the manipulations that we did perform are sufficient to provide evidence for common bursting strategies under the diverse trans-regulatory environments present in wild-type and ectopic regions of gene expression. In the Discussion section of our revised manuscript, we will elaborate further on the kind of genetic manipulations (e.g., probing transcriptional strategies that result from swapping promoters in the context of eve-MS2 BAC; or quantifying the impact on eve transcriptional control after performing optogenetic perturbations of transcription factors and/or chromatin remodelers) that could shed further light on the currently undefined molecular mechanism that constrains eve bursting strategies, as a mean to motivate future work.

Manuscript updates:

In our Discussion section, we elaborated on proposed manipulations of the transcription factor environment to elucidate the molecular mechanisms behind even-skipped bursting control strategies. We began by listing studies linking transcription factor concentration to bursting control strategies, such as (Hoppe et al. 2020), who observed that the natural BMP (Bone Morphogenetic Protein) gradient shapes bursting frequency of target genes in Drosophila embryos. And (Zhao et al. 2023), who used the LEXY optogenetic system to modulate Knirps nuclear concentration and observed that this repressor acts on eve stripe 4+6 enhancer by gradually decreasing bursting frequency until the locus adopts a reversible quiescent state. Then, we proposed performing systematic LEXY-mediated modulation of critical transcription factors (Bicoid, Hunchback, Giant, Kruppel, Zelda) to understand the extent of their contribution to the unified even-skipped bursting strategies.

To better frame the hypothesis that the even-skipped promoter defines strategies of bursting control, we added a reference to the work of (Tunnacliffe, Corrigan, and Chubb 2018). This study surveyed 17 actin genes with identical sequences but distinct promoters in the amoeba Dictyostelium discoideum, and found that all genes display different bursting strategies. Their findings, together with the previously cited work by (Pimmett et al. 2021) and (Yokoshi et al. 2022), suggest a critical role of gene promoters in constraining the bursting strategies of eukaryotic genes.

… ii) the simplicity with which bursting is analyzed (only a two-state model is considered, and not cross-validated with an alternative approach than cpHMM) and…

Based on our previous work (Lammers et al., 2020), and as described in the SI Section of the current manuscript: Inference of Bursting Parameters, we selected a three-state model (OFF, ON1, ON2) under the following rationale: transcription of even-skipped in pre-gastrulating embryos occurs after DNA replication, and promoters on both sister chromatids remain paired. Most of the time these paired loci cannot be resolved independently using conventional microscopy. As a result, when we image an MS2 spot, we are actually measuring the transcriptional dynamics of two promoters. Thus, each MS2-fluorescent spot may result from none (OFF), one (ON1) or two (ON2) sister promoters being in the active state. Following our previous work, we analyzed our data assuming the three-state model (OFF, ON1, ON2), and then, for ease of presentation, aggregated ON1 and ON2 into an effective single ON state. As for the lack of an alternative model, we chose the simplest model compatible with our data and our current understanding of transcription at the eve locus. With this in mind, we do not rule out the possibility that more complex processes—that are not captured by our model—shape MS2 fluorescence signals. For example, promoters may display more than two states of activity. However, as shown in (Lammers et al., 2020 - SI Section: G. cpHMM inference sensitivities), model selection schemes and cross-validation do not give consistent results on which model is more favorable; and for the time being, there is not a readily available alternative to HMM for inference of promoter states from MS2 signal. For example, orthogonal approaches to quantify transcriptional bursting, such as smFISH, are largely blind to temporal dynamics. As a result, we choose to entertain the simplest two-state model for each sister promoter. We appreciate these observations, as they point out the need of devoting a section in the supplemental material of our revised manuscript to clarify the motivations behind model selection.

Manuscript updates:

We have devoted the new Supplemental Material section “Selection of a three-state model of promoter activity and a compound Hidden Markov Model for inference of promoter states from MS2 fluorescent signal” to clarify the rationale behind our selection of a three-state promoter activity model. Since transcription in pre-gastrulating Drosophila embryos occurs after DNA replication, each MS2-active locus contains two unresolvable sister promoters that can either be inactive (OFF), one active (ON1), or both active (ON2).

Next, we elaborated on the conversion of a three-state model into an effective two-state model for ease of presentation and described how the effective two-state model parameters—kon (burst frequency), koff-1 (burst duration), and r (burst amplitude)—were calculated.

Additionally, we acknowledged that while the three-state model of promoter activity is the simplest model compatible with our current understanding of transcription in the even-skipped locus, we do not rule out the possibility that even-skipped transcription may be described by more complex models that include multiple states beyond ON and OFF. Finally, we referenced (Lammers et al. 2020) who asserted that while all inferences of promoter states computed from confocal microscopy of MS2/PP7 fluorescence data rely on Hidden Markov models, cross-comparisons between one, two, or multiple-state Hidden Markov models do not yield consistent results regarding which is more accurate. We close the new section by proposing that state-of-the-art microscopy and deconvolution algorithms to improve signal-to-noise-ratio may offer alternatives to the inference of promoter states.

… iii) the lack of comparisons with published work.

We thank the reviewer for pointing this out. In the current discussion of our manuscript, we compare our findings to recent articles that have addressed the question of the origin of bursting control strategies in Drosophila embryos (Pimmett et al., 2021; Yokoshi et al., 2022; Zoller et al., 2018). Nevertheless, we acknowledge that we failed to include references that are relevant to our study. Thus, our revised Discussion section must include recent results by (Syed et al., 2023), which showed that the disruption of Dorsal binding sites on the snail minimal distal enhancer results in decreased amplitude and duration of transcription bursts in fruit fly embryos. Additionally, we have to incorporate the study by (Hoppe et al., 2020), which reported that the Drosophila bone morphogenetic protein (BMP) gradient modulates the bursting frequency of BMP target genes. References to thorough studies of bursting control in other organisms, like Dictyostelium discoideum (Tunnacliffe et al., 2018), are due as well.

Manuscript updates:

As mentioned in the updates above, our revised manuscript now includes long due references to studies by (Syed, Duan, and Lim 2023), (Hoppe et al. 2020), (Tunnacliffe, Corrigan, and Chubb 2018), and (Chen et al. 2023). All of which are relevant for our current workk.

Reviewer #2 (Public Review):

The manuscript by Berrocal et al. asks if shared bursting kinetics, as observed for various developmental genes in animals, hint towards a shared molecular mechanism or result from natural selection favoring such a strategy. Transcription happens in bursts. While transcriptional output can be modulated by altering various properties of bursting, certain strategies are observed more widely. As the authors noted, recent experimental studies have found that even-skipped enhancers control transcriptional output by changing burst frequency and amplitude while burst duration remains largely constant. The authors compared the kinetics of transcriptional bursting between endogenous and ectopic gene expression patterns. It is argued that since enhancers act under different regulatory inputs in ectopically expressed genes, adaptation would lead to diverse bursting strategies as compared to endogenous gene expression patterns. To achieve this goal, the authors generated ectopic even-skipped transcription patterns in fruit fly embryos. The key finding is that bursting strategies are similar in endogenous and ectopic even-skipped expression. According to the authors, the findings favor the presence of a unified molecular mechanism shaping even-skipped bursting strategies. This is an important piece of work. Everything has been carried out in a systematic fashion. However, the key argument of the paper is not entirely convincing.

We thank the reviewer, as these comments will enable us to improve the Discussion section and overall logic of our revised manuscript. We agree that the evidence provided in this work, while systematic and carefully analyzed, cannot conclusively rule out either of the two proposed models, but just provide evidence supporting the hypothesis for a specific molecular mechanism constraining eve bursting strategies. Our experimental evidence points to valuable insights about the mechanism of eve bursting control. For instance, had we observed quantitative differences in bursting strategies between ectopic and endogenous eve domains, we would have rejected the hypothesis that a common molecular mechanism constrains eve transcriptional bursting to the observed bursting control strategy of frequency and amplitude modulation. Thus, we consider that our proposition of a common molecular mechanism underlying unified eve bursting strategies despite changing trans-regulatory environments is more solid. On the other hand, while our model suggests that this undefined bursting control strategy is not subject to selection acting on specific trans-regulatory environments, it is not trivial to completely discard selection for specific bursting control strategies given our current lack of understanding of the molecular mechanisms that shape the aforesaid strategies. Indeed, we cannot rule out the hypothesis that the observed strategies are most optimal for the expression of eve endogenous stripes according to natural selection, and that these control strategies persist in ectopic regions as an evolutionary neutral “passenger phenotype” that does not impact fitness. We recognize the need to acknowledge this last hypothesis in the updated Introduction and Discussion sections of our manuscript. Further studies will be needed to determine the mechanistic and molecular basis of eve bursting strategies.

Manuscript updates:

In this work, we compared strategies of bursting control between endogenous and ectopic regions of even-skipped expression. Different strategies between both regions would suggest that selective pressure maintains defined bursting strategies in endogenous regions. Conversely, similar strategies in both ectopic and endogenous regions would imply that a shared molecular mechanism constrains bursting parameters despite changing trans-regulatory environments.

In our updated Discussion section, we acknowledge that while our work provides evidence supporting the second hypothesis, we cannot conclusively rule out the possibility that the observed strategies were selected as the most optimal for endogenous even-skipped expression regions and that ectopic regions retain such optimal bursting strategies as an evolutionary neutral “passenger phenotype”.

Reviewer #3 (Public Review):

In this manuscript by Berrocal and coworkers, the authors do a deep dive into the transcriptional regulation of the eve gene in both an endogenous and ectopic background. The idea is that by looking at eve expression under non-native conditions, one might infer how enhancers control transcriptional bursting. The main conclusion is that eve enhancers have not evolved to have specific behaviors in the eve stripes, but rather the same rates in the telegraph model are utilized as control rates even under ectopic or 'de novo' conditions. For example, they achieve ectopic expression (outside of the canonical eve stripes) through a BAC construct where the binding sites for the TF Giant are disrupted along with one of the eve enhancers. Perhaps the most general conclusion is that burst duration is largely constant throughout at ~ 1 - 2 min. This conclusion is consistent with work in human cell lines that enhancers mostly control frequency and that burst duration is largely conserved across genes, pointing to an underlying mechanistic basis that has yet to be determined.

We thank the reviewer for the assessment of our work. Indeed, evidence from different groups (Berrocal et al., 2020; Fukaya et al., 2016; Hoppe et al., 2020; Pimmett et al., 2021; Senecal et al., 2014; Syed et al., 2023; Tunnacliffe et al., 2018; Yokoshi et al., 2022; Zoller et al., 2018) is coming together to uncover commonalities, discrepancies, and rules that constrain transcriptional bursting in Drosophila and other organisms.

Additional updates to the manuscript

(1) In our current study, we observed the appearance of a mutant stripe of even-skipped expression beyond the anterior edge of eve stripe 1, which we refer to as eve stripe 0. This stripe appeared in embryos with a disrupted eve stripe 1 enhancer. In a previous study, (Small, Blair, and Levine 1992) reported a “head patch” of even-skipped expression while assaying the regulation of reporter constructs carrying the minimal regulatory element of eve stripe 2 enhancer alone. In our updated manuscript, we state that it is tempting to identify our eve stripe 0 with the previously reported head patch. (Small, Blair, and Levine 1992) speculated that this head patch of even-skipped expression appeared as a result of regulatory sequences present in the P-transposon system they used for genomic insertions. However, P-transposon sequences are not present in our experimental design. Thus, the appearance of eve stripe 0 indicates a repressive role of the eve stripe 1 enhancer at the anterior end of the embryo and may imply that the minimal regulatory element of the eve stripe 2 enhancer, as probed by (Small, Blair, and Levine 1992), can drive the expression of the head patch/eve stripe 0 when the eve stripe 1 enhancer is not present.

(2)  In our current analysis, we observed that the disruption of Gt-binding sites on the eve stripe 2 enhancer synergizes with the deletion of the eve stripe 1 enhancer, as double mutant embryos display more ectopic expression in their anterior regions than embryos with only disrupted Gt-binding sites. While this may indicate that the repressive activity of eve stripe 1 enhancer synergizes with the repression exerted by Giant, other unidentified transcription factors may be involved in this repressive synergy. In the updated manuscript we clarified that unidentified transcription factors may bind in the vicinity of Gt-binding sites. The hypothesis that Gt-binding sites recognize other transcription factors was proposed by (Small, Blair, and Levine 1992), as they observed that the anterior expansion of eve stripe 2 resulting from Gt-binding site deletions was “somewhat more severe” than expansion observed in embryos carrying null-Giant alleles.

Author response:

We will address all the textual suggestions, including rectifying any typos and incorporating the most recent literature.

We will conduct longitudinal studies to determine whether the phenotype worsens or improves over time in liver-specific SMN-depleted mice. In this regard, we will present data from P60 animals, such as histological analyses for the characterization of the liver and pancreas.