Author response:

Reviewer #1 (Public review):

Summary:

The authors aim to investigate the relationship between low estrogen levels, postmenopausal hypertension, and the potential role of the molecule L-AABA as a biomarker for hypertension. By employing metabolomic analysis and various statistical methods, the study seeks to understand how estrogen deficiency affects blood pressure and identify key metabolites involved in this process, with a particular focus on L-AABA.

Strengths:

The study addresses a relevant and understudied area: the role of estrogen and metabolites in postmenopausal hypertension. It presents a novel hypothesis that L-AABA may serve as a protective factor against hypertension, which could have significant clinical implications if proven.

We appreciate the acknowledgment of our study’s focus on an important and understudied area. Our hypothesis regarding L-AABA’s role as a possible protective factor against hypertension indeed holds promise for advancing clinical implications.

Weaknesses:

The evidence linking L-AABA to hypertension is largely correlative, lacking experimental validation or mechanistic proof. Key limitations, such as the inadequacy of the ovariectomy model in replicating human menopause, are acknowledged but not addressed with alternative approaches. In summary, while the study offers an intriguing hypothesis, its conclusions are premature and require further experimental validation and human data to substantiate the claims.

We recognize the limitations regarding the correlative nature of our findings and the inadequacy of the OVX model in replicating human menopause. Future research will prioritize experimental validation and incorporate human studies to solidify our conclusions.

Reviewer #2 (Public review):

Summary:

In this manuscript, Dr. Yao Li et al. documented the metabolomic profile of the aorta from OVX rats and that from OVX plus E2. These conditions mimic post-menopause hypertension and hormonal replacement therapy.

Strengths:

The authors state that this is probably the first study to examine the metabolic changes in the aorta of post-menopause hypertension.

As pointed out by the reviewer, our study may be the first to investigate changes in aortic metabolism in postmenopausal hypertension. As an exploratory study, our goal is to depict the overall characteristics and explore possible research directions.

Weaknesses:

There are several weaknesses, and a few of them are quite serious.

(1) The aorta is not a resistant artery and has little to do with hypertension. The authors should have used resistant arteries for this study. The expression of several adrenergic receptors and cholinergic receptors in the aorta and resistant arteries are different. It is unknown whether the aorta metabolomic profile has any relevance to BP and whether they are similar to that of the resistant arteries. I understand the logistics issue of obtaining enough tissues from resistant arteries. At least, once some leads are discovered in the aorta, the authors should validate it in resistant arteries. This should be feasible.

We acknowledge the limitation of using the aorta and will aim to include studies on resistant arteries to validate our metabolomic findings.

(2) The aorta and all the arteries have three layers. It is critically important to know whether the metabolic changes occur in the intima or in the media, while the adventitia probably has little to do with vasoconstriction and hypertension. If the authors want to use the aorta to conduct the preliminary study, they should completely remove the adventitia and then use samples with and without their endothelium stripped and then assess their metabolomic profiles. After the leads are obtained from this preliminary profiling, they should be validated in endothelium and smooth muscles of the resistant artery. The current experiments are not appropriately designed.

Future studies will involve detailed profiling of specific arterial layers, focusing on the intima and media to enhance the relevance of our findings related to hypertension.

(3) The tail-cuff BP measurement is a technique of the last century. The current gold standard of BP measurement is by telemetry. The tail-cuff method is particularly problematic in this study because the 1-2 h restraining of the rats for more than 10 times BP measurement will cause significant stress in the animal, and their stress hormone secretion might cause biased metabolomic profiles in the OVX versus shames operated mice. The problem can be totally avoided by using telemetry.

We appreciate the suggestion and will consider telemetry for more accurate blood pressure measurements in future experiments to minimize stress-related bias.

(4) Although the L-AABA showed a high p-value (10^-4) of a decrease in the OVX rats, the fold change is small (2-3 folds). Such a small change should be validated using a different method to be convincing.

We plan to employ additional methods to validate the observed changes in L-AABA levels in the following research, ensuring robustness of our findings.

(5) The authors claim (or hypothesize) that the reduced AABA level in OVX can cause vascular remodeling. This can be easily validated by the histology of the OVX-resistant artery, and they should do that during the revision. The authors should also examine the M1 macrophage function from the OVX mice to validate their claimed link of AABA to M1.

We intend to conduct histological analyses and examine M1 macrophage function in OVX-resistant arteries to validate our hypothesis in the following research.

(6) As mentioned above, the authors need to pinpoint the changes of AABA to target cells, i.e., endothelial cells, SMC, or M1, and then use in vitro or in vivo cell biology approaches to assess whether these cells in the OVX rat indeed have an abnormality in function and, indeed, such functional changes are responsible for the BP phenotype.

Addressing these points, we aim to pinpoint specific cell types affected by AABA variations and conduct in vitro and in vivo studies to examine their physiological impacts in the following research.

(7) The results of the current study can be condensed into 1 or 2 figures that can serve as a base or a starting point for a deeper scientific study.

Thank you for your suggestion. As a omics research, our research approach may differ from traditional mechanism studies.

Summary

The experimental design of this manuscript is inappropriate, and the methods are not up to the current standards. The whole study is descriptive and rudimentary. It lacks validation and mechanism. The data from this manuscript might be of some value and can serve as the first step for more investigation of the mechanism of post-menopause hypertension.

Reviewer #3 (Public review):

Summary:

The decrease in estrogen levels is strongly associated with postmenopausal hypertension. Dr. Yao Li and colleagues aimed to investigate the metabolomic mechanisms of underlying postmenopausal hypertension using OVX and OVX+E2 rat models. They successfully established a correlation between reduced estrogen levels and the development of hypertension in rats. They identified L-alpha-aminobutyric acid (AABA) as a potential marker for postmenopausal hypertension. The research explored the metabolic alterations in aortic tissues and proposed several potential mechanisms contributing to postmenopausal hypertension.

Strengths:

The group performed a comprehensive enrichment analysis and various statistical analyses of the metabolomics data.

As summarized by the reviewer, our current study conducted a comprehensive analysis of metabolomics data. It is also a reliable foundation for further mechanism research.

Weaknesses:

(1) The manuscript is descriptive in nature, although they mentioned their primary objective is to explore the potential mechanisms linking low estrogen levels with postmenopausal hypertension. No mechanism insights have been interrogated in this study, which has been mentioned by the authors in the discussion. The connection between E2, AABA, and macrophage needs to be validated in endothelial cells, vascular smooth muscle cells, and other aortic tissue cells. Without such verification, the manuscript predominantly raises hypotheses only based on metabolomic data.

We have proposed research hypotheses based on detailed omics data. Further research on the mechanisms involving endothelial and vascular smooth muscle cells to validate the pathway connections between E2, AABA, and macrophages is undoubtedly the future direction of this study.

(2) The serum contains three forms of estrogen: Estradiol, Estrone, and Estriol. The authors used the Rat E2 ELISA kit. Ideally, all three forms of estrogen should be measured.

Future assays will aim to measure Estradiol, Estrone, and Estriol to capture a more comprehensive picture of estrogen’s role in postmenopausal hypertension.

Author response:

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

eLife Assessment

This useful study reports on the discovery of an antimicrobial agent that kills Neisseria gonorrhoeae. Sensitivity is attributed to a combination of DedA assisted uptake of oxydifficidin into the cytoplasm and the presence of a oxydifficidin-sensitive RplL ribosomal protein. Due to the narrow scope, the broader antibacterial spectrum remains unclear and therefore the evidence supporting the conclusions is incomplete with key methods and data lacking. This work will be of interest to microbiologists and synthetic biologists.

General comment about narrow scope: The broader antibacterial spectrum of oxydifficidin has been reported previously (S B Zimmerman et al., 1987). The main focus of this study is on its previously unreported potent anti-gonococcal activity and mode of action. While it is true that broad-spectrum antibiotics have historically played a role in effectively controlling a wide range of infections, we and others believe that narrow-spectrum antibiotics have an overlooked importance in addressing bacterial infections. Their advantage lies in their ability to target specific pathogens without markedly disrupting the human microbiota.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Kan et al. report the serendipitous discovery of a Bacillus amyloliquefaciens strain that kills N. gonorrhoeae. They use TnSeq to identify that the anti-gonococcal agent is oxydifficidin and show that it acts at the ribosome and that one of the dedA gene products in N. gonorrhoeae MS11 is important for moving the oxydifficidin across the membrane.

Strengths:

This is an impressive amount of work, moving from a serendipitous observation through TnSeq to characterize the mechanism by which Oxydifficidin works.

Weaknesses:

(1) There are important gaps in the manuscript's methods.

The requested additions to the method describing bacterial sequencing and anti-gonococcal activity screening will be made. However, we do not think the absence of these generic methods reduces the significance of our findings.

(2) The work should evaluate antibiotics relevant to N. gonorrhoeae.

(1) It is not clear to us why reevaluating the activity of well characterized antibiotics against known gonorrhoeae clinical strains would add value to this manuscript. The activity of clinically relevant antibiotics against antibiotic-resistant N. gonorrhoeae clinical isolates is well described in the literature. Our use of antibiotics in this study was intended to aid in the identification of oxydifficidin’s mode of action. This is true for both Tables 1 and 2.

(2) If the reviewer insists, we would be happy to include MIC data for the following clinically relevant antibiotics: ceftriaxone (cephalosporin/beta-lactam), gentamicin (aminoglycoside), azithromycin (macrolide), and ciprofloxacin (fluoroquinolone).

(3) The genetic diversity of dedA and rplL in N. gonorrhoeae is not clear, neither is it clear whether oxydifficidin is active against more relevant strains and species than tested so far.

(1) We thank the reviewer for this suggestion. We aligned the DedA sequence from strain MS11 with DedA proteins from 220 N. gonorrhoeae strains that have high-quality assemblies in NCBI. The result showed that there are no amino acid changes in this protein. Using the same method, we observed several single amino acid changes in RplL. This included changes at A64, G25 and S82 in 4 strains with one change per strain. These sites differ from R76 and K84, where we identified changes that provide resistance to oxydifficidin. Notably, in a similar search of representative Escherichia, Chlamydia, Vibrio, and Pseudomonas NCBI deposited genomes, we did not identify changes in RplL at position R76 or K84.

(2) While the usefulness of screening more clinically relevant antibiotics against clinical isolates as suggested in comment 2 was not clear to us, we agree that screening these strains for oxydifficidin activity would be beneficial. We have ordered Neisseria gonorrhoeae strain AR1280, AR1281 (CDC), and Neisseria meningitidis ATCC 13090. They will be tested when they arrive.

Reviewer #2 (Public Review):

Summary:

Kan et al. present the discovery of oxydifficidin as a potential antimicrobial against N. gonorrhoeae, including multi-drug resistant strains. The authors show the role of DedA flippase-assisted uptake and the specificity of RplL in the mechanism of action for oxydifficidin. This novel mode of action could potentially offer a new therapeutic avenue, providing a critical addition to the limited arsenal of antibiotics effective against gonorrhea.

Strengths:

This study underscores the potential of revisiting natural products for antibiotic discovery of modern-day-concerning pathogens and highlights a new target mechanism that could inform future drug development. Indeed there is a recent growing body of research utilizing AI and predictive computational informatics to revisit potential antimicrobial agents and metabolites from cultured bacterial species. The discovery of oxydifficidin interaction with RplL and its DedA-assisted uptake mechanism opens new research directions in understanding and combating antibiotic-resistant N. gonorrhoeae. Methodologically, the study is rigorous employing various experimental techniques such as genome sequencing, bioassay-guided fractionation, LCMS, NMR, and Tn-mutagenesis.

Weaknesses:

The scope is somewhat narrow, focusing primarily on N. gonorrhoeae. This limits the generalizability of the findings and leaves questions about its broader antibacterial spectrum. Moreover, while the study demonstrates the in vitro effectiveness of oxydifficidin, there is a lack of in vivo validation (i.e., animal models) for assessing pre-clinical potential of oxydifficidin. Potential SNPs within dedA or RplL raise concerns about how quickly resistance could emerge in clinical settings.

(1) Spectrum/narrow scope: The broader antibacterial spectrum of oxydifficidin has been reported previously (S B Zimmerman et al., 1987). The focus of this study is on its previously unreported potent anti-gonococcal activity and its mode of action. While it is true that broad-spectrum antibiotics have historically played a role in effectively controlling a wide range of infections, we and others believe that narrow-spectrum antibiotics have an overlooked importance in addressing bacterial infections. Their advantage lies in their ability to target specific pathogens without markedly disrupting the human microbiota.

(2) Animal models: We acknowledge the reviewer’s insight regarding the importance of in vivo validation to enhance oxydifficidin’s pre-clinical potential. However, due to the labor-intensive process needed to isolate oxydifficidin, obtaining a sufficient quantity for animal studies is beyond the scope of this study. Our future work will focus on optimizing the yield of oxydifficidin and developing a topical mouse model for subsequent investigations.

(3) Potential SNPs: Please see our response to Reviewer #1’s comment 3. We acknowledge that potential SNPs within dedA and rplL raise concerns regarding clinical resistance, which is a common issue for protein-targeting antibiotics. Yet, as pointed out in the manuscript, obtaining mutants in the lab was a very low yield endeavor.

Reviewer #3 (Public Review):

Summary:

The authors have shown that oxydifficidin is a potent inhibitor of Neisseria gonorrhoeae. They were able to identify the target of action to rplL and showed that resistance could occur via mutation in the DedA flippase and RplL.

Strengths:

This was a very thorough and clearly argued set of experiments that supported their conclusions.

Weaknesses:

There was no obvious weakness in the experimental design. Although it is promising that the DedA mutations resulted in attenuation of fitness, it remains an open question whether secondary rounds of mutation could overcome this selective disadvantage which was untried in this study.

We thank the reviewer for the positive comment. We agree that investigating factors that could compensate for the fitness attenuation caused by DedA mutation would enhance our understanding of the role of DedA.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) The use of the term "N. gonorrhoeae wildtype" should not be used. It is uninformative, as the species contains a large amount of diversity. Instead, please name the strain. From Figure 1, it looks like the authors used MS11. Since MS11 is a longstanding lab strain and likely does not reflect circulating N. gonorrhoeae, and since H041 is no longer in circulation, the authors should ideally test the compound against more representative strains of N. gonorrhoeae. This includes panels of isolates available through the CDC, for example (https://www.cdc.gov/drugresistance/resistance-bank/index.html). I encourage the authors to include FC428 or another recently identified isolate with the penA 60 allele to demonstrate oxydifficidin's activity against contemporary concerning isolates/lineages.

(1) “N. gonorrhoeae MS11” is now used instead of “N. gonorrhoeae WT” in this manuscript.

(2) In our revised manuscript, we have added MIC data for recently identified Neisseria gonorrhoeae isolates AR#1280 and AR#1281 which contain the penA 60 allele (Table 1). The data shows oxydifficidin maintains its potent activity against these multidrug-resistant strains. We also added a description of this data to the results section as shown below.

Original text: “Oxydifficidin was more potent against N. gonorrhoeae MS11 than almost all other antibiotics we tested. In fact, it was only slightly less active than the highly optimized third-generation cephalosporin, ceftazidime.([18]) However, unlike third-generation cephalosporins, oxydifficidin retained activity against the multidrug resistant H041 clinical isolate (Table 1).([4]) H041 is resistant to the “standard of care” cephalosporin ceftriaxone (2 µg/mL) as well as a number of other antibiotics that are normally active against N. gonorrhoeae (penicillin G, 4 µg/mL; cefixime, 8 µg/mL; levofloxacin, 32 µg/mL).”

Changed to: “Oxydifficidin was more potent against N. gonorrhoeae MS11 than most other antibiotics we tested. Notably, unlike clinically used antibiotics such as ceftriaxone, azithromycin, and ciprofloxacin, oxydifficidin retained activity against all multidrug-resistant clinical isolates we examined (Table 1).” (Line 77-79)

(2) Does oxydifficidin have activity against N. meningitidis? It is the species most closely related to N. gonorrhoeae and the other pathogenic Neisseria.

Oxydifficidin has potent activity against N. meningitidis ATCC 13090. In our revised manuscript, we have included its MIC data in Figure 1c.

(3) Given claims that oxydifficidin activity in N. gonorrhoeae as compared to other Neisseria reflects N. gonorrhoeae's dedA and sensitive rplL, it would be good to assess the allelic diversity of these genes in N. gonorrhoeae. There are over 20,000 genomes from clinical isolates of N. gonorrhoeae in databases. It should be straightforward to check whether dedA and rplL allelic variants already exist in the population. Should variants be observed, oxydifficidin should be tested against the associated strains of N. gonorrhoeae.

Response: We thank the reviewer for this suggestion. We aligned the DedA sequence from strain MS11 with DedA proteins from 220 N. gonorrhoeae strains that have high-quality assemblies in NCBI. The result showed that there are no amino acid changes in this protein. Using the same method, we observed several single amino acid changes in RplL. This included changes at A64, G25 and S82 in 4 strains with one change per strain. These sites differ from R76 and K84, where we identified changes that provide resistance to oxydifficidin. Notably, in a similar search of representative Escherichia, Chlamydia, Vibrio, and Pseudomonas NCBI deposited genomes, we did not identify changes in RplL at position R76 or K84.

New text: “A survey of 220 N. gonorrhoeae strains with high-quality assemblies in NCBI found no mutations in the DedA protein.” (Line 104-105)

“These two mutations were not found in the survey of the same collection of N. gonorrhoeae strains used to look for DedA mutations.” (Line 143-144)

(4) Clinically relevant antibiotics for N. gonorrhoeae are penicillin, tetracycline, spectinomycin, gentamicin, ciprofloxacin, azithromycin, ceftriaxone; moreover, zoliflodacin and gepotidacin have reportedly successfully completed phase 3 trials. The authors should redo their MIC testing with these antibiotics (e.g., for Figures 1 and 2 and Tables 1 and 2), both because this will enable direct comparison with the many clinical isolates that have undergone testing and because these are the drugs most pertinent to clinical practice. Ampicillin, ceftazidime, chloramphenicol, bacitracin, and daptomycin are not relevant. Could the authors explain why they tested vancomycin, polymyxin B, irgasan, melittin, avilamycin, and thiostrepton?

Our use of antibiotics with diverse modes of action (e.g. vancomycin, polymyxin B, irgasan, melittin, avilamycin, and thiostrepton) in this study was intended to aid in the identification of oxydifficidin’s mode of action. This is true for both Tables 1 and 2.

To address the reviewer’s concern, in our revised manuscript, we have added MIC data for the following clinically relevant antibiotics: ceftriaxone (cephalosporin/beta-lactam), gentamicin (aminoglycoside), azithromycin (macrolide), and ciprofloxacin (fluoroquinolone) to Table 1.

(5) Please describe the characteristics of the transposon library (finding four transposons in a single strain does seem unexpected, given how most transposon libraries aim for one transposon insertion per strain).

We understand that one transposon insertion per strain is ideal for transposon libraries. This Bacillus strain proved to be recalcitrant to genetic manipulation. In the rare cases where we obtained resistance colonies upon electroporation with the transposon, all colonies contained multiple (≥ 4) transposon insertions. This made it impractical to build a library with one transposon insertion per library member.

We assumed that the anti-N. gonorrhoeae activity most likely originated from a natural product BGC, which typically range from 10-100 kb in size.

Based on the average of 50 kb per BGC, ~80 transposon insertions would be required to fully search the 4.2 Mb genome of Bacillus amyloliquefaciens BK for a BGC. At 4 mutations per transformant, 1x coverage of the genome would require only 20 library members.

After extensive electroporation of transposon into Bacillus amyloliquefaciens BK, we were able to obtain a library of 50 members, including one mutant (Tn5-3) that lacked anti-N. gonorrhoeae activity.

New text added to the methods section:

“A library containing 50 transposon mutants was obtained. In the mutants examined, each strain contained ≥4 transposon insertions” (Line 337-339)

(6) Please describe in the methods how you sequenced and annotated the genome of Bacillus amyloliquefaciens BK.

The sequencing method is now described in “Genomic Sequencing and annotation of Bacillus amyloliquefaciens” section. The genome of Bacillus amyloliquefaciens BK was not fully annotated. Mutations were identified as described in the updated methods section below.

New text:

“Genomic Sequencing and annotation of Bacillus amyloliquefaciens

Genomic DNA from Bacillus amyloliquefaciens BK WT and transposon mutant Tn5-3 was isolated using PureLink Microbiome DNA purification kit (Invitrogen) according to the manufacturer’s instructions.

The Bacillus amyloliquefaciens BK WT genome was assembled by mapping its sequencing data onto the annotated genome of Bacillus amyloliquefaciens FZB42 using Geneious Prime. Differences in the mutant strain Tn5-3 were identified by mapping its sequencing data onto the assembled Bacillus amyloliquefaciens BK WT genome. The mutated genes were then annotated using NCBI BLAST. The oxydifficidin BGC was annotated using the antiSMASH online server.” (Line 253-260)

(7) Please describe in the methods how you screened the library for strains that lacked anti-gonococcal activity.

The method is added to our revised manuscript as section “Screening of Bacillus Strains Lacking Anti-N. gonorrhoeae Activity”.

New text:

“Screening of Bacillus Strains Lacking Anti-N. gonorrhoeae Activity

The transposon mutants of Bacillus amyloliquefaciens BK were grown overnight in LB medium at 30 °C. Each overnight culture was then diluted 1:5000, and 1 μl of the diluted culture was spotted onto a GCB agar plate swabbed with N. gonorrhoeae cells. The plate was then incubated overnight at 37 °C with 5% CO2. The mutant strain (Tn5-3) lacking anti-N. gonorrhoeae activity was identified due to its failure to produce a zone of growth inhibition in the resulting N. gonorrhoeae lawn.” (Line 341-346)

(8) Was only one strain found that was a 'non-producer' of anti-N. gonorrhoeae activity? Line 68 suggests that this was only one of multiple non-producers. Is that correct? If so, did you work up the others, and did they also have disruptions in the same biosynthetic gene cluster?

Only one strain was identified as a “non-producer” of anti-N. gonorrhoeae activity. We have modified the text to clarify this point.

Original text: “The sequencing of one non-producer strain revealed that it surprisingly contained four transposon insertions and one frame shift mutation.”

Changed to: “The sequencing of the non-producer strain revealed that it surprisingly contained four transposon insertions and one frame shift mutation.” (Line 53-54 )

(9) All sequences (including Bacillus amyloliquefaciens BK) must be deposited in a public database (e.g., NCBI) and the accession numbers reported in the manuscript.

Genomic sequence data of Bacillus amyloliquefaciens BK has been deposited in GenBank, and its accession number (GCA_019093835.1) now appears in figure legend of Figure S1a.

Figure S1a legend:

“Genome-based phylogenetic tree containing Bacillus amyloliquefaciens BK and closely related Bacillus spp. The tree was built by Genome Clustering of MicroScope using neighbor-joining method. The NCBI accession numbers of Bacillus strains used in the tree are GCA_000196735.1, GCA_000204275.1, GCA_000015785.2, GCA_019093835.1, GCA_000009045.1, GCA_000011645.1, GCA_000172815.1, GCA_000008005.1, and GCA_000007845.1 (from top to bottom).”

Minor

(10) Statements in the article would benefit from fact-checking. For example:

- gonorrhea is not the second most prevalent sexually transmitted infection worldwide; it is the second most reported bacterial sexually transmitted infection.

- Treatment is ceftriaxone 500mg IM x1 in the US, but 1g IM x1 in the UK and Europe. The UK guidelines also permit ciprofloxacin, should sequencing indicate gyrA 91S. I suggest reviewing / specifying which treatment guidelines you're referring to.

We appreciate the reviewer’s corrections. The word “prevalent” is now changed to “reported”.

Original text: “Gonorrhea, which is caused by Neisseria gonorrhoeae, is the second most prevalent sexually transmitted infection worldwide.”

Changed to: “Gonorrhea, which is caused by Neisseria gonorrhoeae, is the second most reported sexually transmitted infection worldwide.” (Line 2-3)

Original text: “Gonorrhea is the second most prevalent sexually transmitted infection worldwide, its causative agent is the bacterium Neisseria gonorrhoeae.”

Changed to: “Gonorrhea is the second most reported sexually transmitted infection worldwide, its causative agent is the bacterium Neisseria gonorrhoeae.” (Line 18-19)

“In the USA” is now added to the sentence stating gonorrhea treatment.

Original text: “The high dose (500 mg) of the cephalosporin ceftriaxone is currently the only recommended therapy for treating gonorrhea infections.”

Changed to: “The high dose (500 mg) of the cephalosporin ceftriaxone is currently the only recommended therapy for treating gonorrhea infections in the USA.” (Line 20-22)

(11) Please make sure all results are in the results section. The report of cell morphology, for example, should be in the results, not the discussion.

In our revised manuscript, we have included the cell morphology data in the results section with the text changes below.

Original text: “Interestingly, not only was dedA deficient N. gonorrhoeae less susceptible to oxydifficidin, oxydifficidin also kills this mutant more slowly (Figure 2b) than WT N. gonorrhoeae MS11.”

Changed to: “Interestingly, not only was dedA deficient N. gonorrhoeae less susceptible to oxydifficidin, oxydifficidin also kills this mutant more slowly (Figure 2b) than WT N. gonorrhoeae MS11. The dedA deletion mutant also showed an altered cell morphology with reduced membrane integrity and lower formation of micro-colonies (Figure S4). (Line 100-104)

Original text: “The dedA deletion mutant also showed an altered cell morphology with reduced membrane integrity and lower formation of micro-colonies (Figure S4), indicating that it should show reduced pathogenesis and fitness, and, as a result, not accumulate in a clinical setting, which adds to the therapeutic appeal of oxydifficidin.”

Changed to: “The dedA deletion mutant exhibited altered cell morphology, characterized by diminished membrane integrity and reduced micro-colony formation, indicating that it should show reduced pathogenesis and fitness, and, as a result, not accumulate in a clinical setting, which adds to the therapeutic appeal of oxydifficidin” (Line 206-210)

(12) Tables 1 and 2 should be combined and should address the most relevant antibiotics

The MIC data of additional relevant antibiotics are now included in Table 1. However, we still believe that keeping Tables 1 and 2 separate enhances the clarity of the manuscript. Table 2 specifically focuses on diverse ribosomal targeting antibiotics, which highlights the unique binding site of oxydifficidin.

(13) Supplemental Figure 1a. The tree could be better resolved, and there are four entries with the identical listing of "Bacillus amyloliquefaciens subsp. plantarum" on different branches. In the methods or the legend, please indicate the accession numbers for these genomes. Also please specify how this tree was made-is it a maximum likelihood tree? Something else?

The tree is now better resolved and includes new entries. The requested information regarding accession numbers and tree construction method has been included in the figure legend.

New supplemental Figure 1a legend:

“a. Genome-based phylogenetic tree containing Bacillus amyloliquefaciens BK and closely related Bacillus spp. The tree was built by Genome Clustering of MicroScope using neighbor-joining method. The NCBI accession numbers of Bacillus strains used in the tree are GCA_000196735.1, GCA_000204275.1, GCA_000015785.2, GCA_019093835.1, GCA_000009045.1, GCA_000011645.1, GCA_000172815.1, GCA_000008005.1, and GCA_000007845.1 (from top to bottom).”

Reviewer #2 (Recommendations For The Authors):

The conclusions drawn in the manuscript are well-supported by the experimental data presented.

I have the below minor comments:

(1) "serendipitously identified" - I feel this wording should be avoided throughout the manuscript. The point of a research paper is to communicate methodology and experimental detail, and this language portrays the opposite.

While we agree that methodology and experimental procedures are paramount in scientific reporting, we believe it is equally important to convey, particularly to younger generations, that a part of the scientific process is often unplanned and can benefit from chance observations. Therefore, we would like to keep this wording.

(2) The introduction should include the biological roles/function of DedA proteins in bacteria.

DedA proteins perform a wide array of biological roles and functions in bacteria. In the results section (Line 107-116), we have described the most well-established of these functions, particularly the flippase activity, which appears to be directly related to oxydifficidin sensitivity. We believe that introducing this information in the results section enhances the manuscript’s clarity and flow.

(3) "When we screened this contaminant for antibacterial activity against lawns of other Gram-negative bacteria it did not produce a zone of growth of inhibition against any of the bacteria we tested (e.g., Escherichia coli, Vibrio cholerae, Caulobacter crescentus)." Can these data Figures be included in the Supplements?

This result was recorded in the lead author’s notebook, but no image was saved.

(4) Line 52: Was any base analyses performed on the Tn-mutants i.e., how many insertion-sites? Depth of mutants? Was a library constructed in this study or previously? Why were only BGC assessed?

Please see our response to Reviewer #1’s comment (5). We focused on BGCs because we believed the anti-N. gonorrhoeae activity most likely resulted from a molecule encoded by a natural product BGC.

(5) Line 98: Do the other 2 predicted DedA-like proteins also have a role in uptake of oxydifficidin? Is there some redundancy in uptake?

We generated knockout mutants for two other predicted DedA-like proteins in N. gonorrhoeae MS11, and the MIC of oxydifficidin for these mutants remained the same as for the N. gonorrhoeae MS11 wild type strain. Therefore, we believe that the DedA protein discussed in this manuscript is the primary transporter of oxydifficidin. However, we cannot completely rule out the possibility of redundancy in oxydifficidin uptake by other DedA-like proteins.

New text: “We also generated deletion mutants for two other predicted dedA-like genes, and the MIC of oxydifficidin for these mutants remained the same as for the N. gonorrhoeae MS11 wild type strain.” (Line 98-100)

Reviewer #3 (Recommendations For The Authors):

This is a well presented manuscript and I could not immediately see any issues with it.

We appreciate the reviewer’s positive feedback.

eLife Assessment

Kan et al. report the discovery of a Bacillus amyloliquifaciens strain that kills Nerisseria gonorrhoeae via oxydifficidin which targets ribosomal proteins. Resistance occurred via mutation in the DedA flippase to influence oxydifficidin uptake. The overall mechanism of action is well described making this an important study with implications for combating clinical antibiotic resistance. The evidence presented is convincing due to rigour employed in the methodological approach. The authors should consider performing a more comprehensive genetic analyses of DedA and RpIL in this clinically relevant strain. This work will be of broad interest to microbiologists and synthetic biologists.

Reviewer #1 (Public review):

Summary:

Kan et al. report the serendipitous discovery of a Bacillus amyloliquefaciens strain that kills N. gonorrhoeae. They use TnSeq to identify that the anti-gonococcal agent is oxydifficidin and show that it acts at the ribosome and that one of the dedA gene products in N. gonorrhoeae MS11 is important for moving the oxydifficidin across the membrane.

Strengths:

- This is an impressive amount of work, moving from a serendipitous observation through TnSeq to characterize the mechanism by which Oxydifficidin works.

Weaknesses:

- The genetic diversity of dedA and rplL in N. gonorrhoeae is still not clear, as the authors looked at diversity of these genes in only 220 isolates (of unclear relationship to each other).

It's not so much a weakness as a source of confusion: how did the authors choose to screen a tiny transposon library of 50 mutants? Since they were surprised to find 4 transposon insertions (if I'm reading it correctly), what was the motivation for even looking at this small library? And since the mutation that led them to the biosynthetic gene cluster wasn't even a transposon insertion but a frameshift, it seems they had another huge episode of serendipity.

Reviewer #2 (Public review):

Summary:

Kan et al. presents the discovery of oxydifficidin as a potential antimicrobial against N. gonorrhoeae, including multi-drug resistant strains. The authors show the role of DedA flippase assisted uptake and the specificity of RplL in the mechanism of action for oxydifficidin. This mode of action could potentially offer a new therapeutic avenue, providing a critical addition to the limited arsenal of antibiotics effective against gonorrhea.

Strengths:

This study shows the potential of revisiting anti-bacterial agents/products for antibacterial activity against modern-day-concerning pathogens and highlights a new anti-gonoccoal mechanism of action. Indeed there is a recent growing body of research to revisit potential antimicrobial agents and metabolites from cultured bacterial species. The discovery of oxydifficidin interaction with RplL and its DedA-assisted uptake mechanism opens new research directions in understanding and combating antibiotic resistant N. gonorrhoeae. The antimicrobial activity of oxydifficidin is also active against N. meningitidis, a closely related species. Methodologically, the study is rigorous employing various experimental techniques including Tn-mutagenesis (TraDIS, Tn-Seq).

Weaknesses:

While the study demonstrates the in vitro effectiveness of oxydifficidin, there is a lack of in vivo validation (i.e., animal models) for assessing pre-clinical potential of oxydifficidin. However, I acknowledge that this would be a tremendous amount of work and likely outside the scope of this study. Potential SNPs within dedA or RplL raises concerns about how quickly resistance could emerge in clinical settings.

Reviewer #3 (Public review):

Summary:

The authors have shown that oxydifficidin is a potent inhibitor of Neisseria gonorrhoeae. They were able to identify the target of action to rpsL and showed that resistance could occur via mutation in the DedA flippase and RpsL.

Strengths:

This was a very thorough and clearly argued set of experiments that supported their conclusions.

Weaknesses:

There was no obvious weakness in the experimental design. Although it is promising that the DedA mutations resulted in attenuation of fitness, it remains an open question whether secondary rounds of mutation could overcome this selective disadvantage which was untried in this study.

Comments on revisions:

All of my suggestions were considered and the responses to the other reviewer's appears sound and has improved the manuscript.

eLife Assessment

Protein and lipid homeostasis is essential for maintaining cellular functions but their crosstalk remains largely unknown. This important manuscript deals with this interesting topic and applies the powerful unbiased tools of somatic cell genetics to discover evidence suggesting a link between sphingolipids/cholesterol ester metabolism and lysosomal protein aggregation. The authors provide compelling orthogonal evidence to support their conclusions.

Reviewer #1 (Public Review):

In this manuscript, Yong and colleagues link perturbations in lysosomal lipid metabolism with the generation of protein aggregates resulting from proteosome inhibition. The main tool used is the ProteoStat stain to assess protein aggregate burden in native cells (i.e. cells under no exogenous or endogenous stress). They initially use CRISPR-based genome-wide screens to identify several genes that affect this aggregate burden. Interestingly, knockdown of genes involved in lysosomal acidification was a major signature which led to identification of other culprit lysosome-associated genes that included ones involved in lipid metabolism. Subsequent CRISPR screen focused on lipidomic analysis led to identification of sphingolipid and cholesterol esters as lipid classes with effects on proteostasis.

Comments on revised version:

They did a decent job addressing most of my comments and the new data (including LysoIP) makes for much more plausible conclusions.

They propose the idea that microautophagy is mediating the delivery of these aggregates to lysosomes.

It appears there are enough experiments and support now for their premise.

The lysosomal lipid metabolism link to proteostasis is still a lingering question in this work but they addressed each of the points I raised regarding it and revised the manuscript accordingly with pertinent discussion.

It is difficult to truly address the lipid link and I think we have to acknowledge that. But overall, looking at the effort and conclusions, this has been improved enough to be a valuable contribution to the field.

Reviewer #2 (Public Review):

In this paper, starting with unbiased CRISPRi screening, the authors found that perturbations in lipid homeostasis lead to proteostasis impairment. The screen and most follow-up experiments used the dye ProteoStat, which detects protein aggregates and the aggresome. Based upon their screen hits and subsequent analyses, the authors determined that increased levels of sphingolipids and cholesterol esters induce proteostasis defects, along with formation of protein aggregates that appear to be localized in the lysosome. The lysosome increases in content, but its function is not detectably perturbed.

Comments on revised version:

I am satisfied with the authors' actions in response to my public and specific suggestions, but not yet with the manuscript itself. I think that the paper would be improved if they showed the evidence arguing against an effect on proteasome activity but I can live with this omission. I think that the readability and ease of grasping the main points are improved by Figure 7. Inclusion of these simple but informative conceptual summaries is a must.

Author response:

We are submitting a revised manuscript with major additions that address the main concerns in the initial reviews. At the highest level, this revision provides i) orthogonal biochemical measurements that yield concrete evidence of lysosomal protein aggregates, and ii) a plausible mechanism linking lysosomal lipid handling and protein aggregation through disruption of ESCRT function. We believe these additions significantly improve the completeness of this study and the conclusions that can be drawn from the data.

Below are more specific highlights on the addition in this revision:

-       We included orthogonal techniques (thioflavin-T staining and Lyso-IP followed by differential extraction) and confirmed the accumulation of RIPA-insoluble protein aggregates at the lysosomes in cells under lipid perturbation (Figure 3).

-       We performed TMT-Proteomics and identified accumulation of insoluble ESCRT components at the lysosomes under lipid perturbation (Figure 4). Two new authors involved in this effort are added onto the manuscript.

-       The ESCRT result prompted us to revisit lysosomal membrane integrity. With improved imaging conditions and analysis we were able to see increased membrane permeabilization under lipid perturbation. VPS4A overexpression partially rescued this phenotype, suggesting that lipid accumulation impairs ESCRT disassembly (Figure 5).

-       Together, the results suggest that lipid perturbation impairs ESCRT function, compromising both lysosomal membrane repair and microautophagy, resulting in the accumulation of endogenous protein aggregates at the lysosomes (Graphical Abstract).

Reviewer #1 (Recommendations For The Authors):

(1) Perhaps the most prominent limitation of this work is the unilateral focus on native cells (i.e. cells under no endogenous or exogenous stress) as the model for protein aggregate formation. Furthermore, although the ProteoStat stain has been utilized by many investigators before, the sole reliance on this stain as the read-out for their assays is concerning. To compound the concern, the ProteoStat-positive puncta co-localize with lysosmal markers which was surprising even to the authors. All in all, it behooves the authors to test proteostasis in multiple parallel ways to actually define what they are studying. How is it possible that protein aggregates under native conditions are only co-localized with lysosomes? Are we really studying protein aggregates which should predominantly be cytoplasmic insoluble aggregates?

(a) They need to get away from a simple stain like ProteoStat and conduct co-stainings with other markers such as poly-ubiquitin antibodies and other chaperones to define what and where else exactly are these aggregates.

Co-staining with poly-ubiquitin was included in the original manuscript. We added orthogonal staining with another widely used amyloid dye, Thioflavin-T, and provided fine-grained quantification of lysosomal vs cytosolic localization of various signals (Figures S4A-C & 3A-B).

(b) They need to do Immunoblots with and without triton insolubility to see if these aggregates are insoluble as most would predict. They can do lysosomal isolation vs cytoplasmic to see if the insoluble aggregates are really lysosomal.

We performed Lyso-IP followed by differential detergent extraction to confirm the accumulation of insoluble proteins at the lysosomes (Figure 3C). Proteomic analysis identified some of these insoluble proteins as ESCRT subunits (Figure 4).

(c) They should compare aggregate formation in the native state versus cells with lysosomal inhibition via Bafilomycin or chloroquine versus cells with proteosomal inhibition. The lysosomal inhibition experiments are particularly informative given the lysosomal relevance they have uncovered.

We included other small molecule inhibitors and at different time points to compare the effect of different modes of proteostasis challenge (Figure S4A-D). Together with the ESCRT finding, our results suggest the role of microautophagy in our system, and provide a model of how ProteoStat- and/or ubiquitin- positive substrates become partitioned between the cytoplasm and lysosomes under different perturbations.

(d) Many protein aggregates which are too bulky for proteosome degradation will traditionally be dealt with by aggrephagy. Why is this not observed?

Knockdown of core macroautophagy components did not impact Proteostat intensity in our CRISPRi screen, suggesting that basal macroautophagy plays a negligible role in clearing endogenous amyloid-like structures in our experimental system. We provide an alternative model that these aggregates instead arrive at the lysosomes via microautophagy.

(2) After addressing #1, they can validate if the genes they identified by CRISPR screens are also important in modulation of protein aggregate burden in other systems. For example, if they inhibit lysosomes by Bafilo or Chloroquine to obtain protein aggregates and then Knockdown the identified genes in the CRISPR screens, will they get the same results?

We addressed the effect of different modes of proteostasis challenge as recommended above. Deacidifying the lysosomes alone causes intense protein aggregation (Figure S4A-D) and eventually cell death, and was thus not combined with other perturbations.

(3) They identify lysosomal lipid metabolism genes/pathways as the culprit for inducing proteostasis. In particular sphingolipid and cholesteryl ester species appear to be operational here. However, there are no specific lipids species or specific lipid metabolism gene that is causative. Rather, you have to knockdown entire processes to have an effect. This suggests that the focus on lysosome health (i.e. permeability, proteolysis, etc) is rudimentary. When you have to knockdown entire classes of lipids, this would indicate more broad effects on cellular lipids (including membrane lipids beyond the lysosome) and related cellular health?

We included data on the effect of knocking down MYLIP, PSAP, and as a comparison PSMD2 on the growth rate of K562 cells (Figure S5A). MYLIP and PSAP KDs, which cause predominantly an accumulation of lipids, do not impede cell growth. Increasing lipid uptake by MYLIP KD increases cell proliferation under our culture conditions, suggesting a general negative impact on cell health was not required for the association between lipid levels and protein aggregates.

(a) They conduct a superficial methyl-beta-cyclodextrin experiment with equivocal results. The use of MBCD for different time-courses to deplete various membrane cholesterol pools including the plasma membrane pool is important to ascertain what aspect of the cellular cholesterol is affecting proteostasis. MBCD +/- cholesterol reintroduction time-courses for rescue will also be key to determine the culprit cellular cholesterol pool.

The MBCD / Filipin experiment helped us determine that ProteoStat doesn’t directly stain cholesterol, nor any major plasma membrane components. Free cholesterol was implicated in neither the screen nor the lipidomics and was not the subject of targeted experiments.

(b) The same concept can be applied to sphingolipids. There are sphingolipids in abundance in multiple membrane compartments. Which ones are causal here? More nuanced evaluation of this with sphingolipid staining/tracking can be conducted.

We attempted experiments where sphingolipids were added back to cells grown in FBS-depleted media. Nevertheless, we were not able to consistently deliver these lipid species and doing so while ensuring the correct subcellular localization at physiologically relevant level would require substantial methods development.

(c) As part of this, are lipid rafts and/or caveolae being affected by the perturbations in cholesterol and sphingolipids? Lipid rafts are highly enriched in these 2 lipids which could link to their preteostasis observation.

Indeed, ceramides released from SM hydrolysis are proposed to self-assembled into microdomains with negative curvature that can promote the formation of intralumenal vesicles (Alonso and Goni, 2018; Niekamp et al 2022). We propose that SM accumulation may hinder this process by counteracting the negative membrane curvature and impede microautophagy.

(d) How about ER membrane lipids? The UPR and subsequent effects on proteostasis are intricately involved with ER lipid bilayer composition.

We did not perform lipidomics on ER membranes in this study, though we note that at steady state, sphingolipids and cholesterol esters are not expected to be enriched at the ER (Ikonen and Zhou, 2021). We checked whether lipid-related genetic perturbations induced the UPR in published perturb-seq data in K562 cells. Neither MYLIP nor PSAP knockdown induced a UPR.

In conclusion, the manuscript is interesting but the excitement over a link between lysosome-related lipid metabolism and proteostasis needs to be tamped until a more robust experimental approach is employed to generate supportive and corroborating results.

Reviewer #2 (Recommendations For The Authors):

- The paper has a number of grammatically awkward sentences. Editing these would enhance clarity.

- It is important to show the co-localization of aggregates with the lysosome. This is shown in supplements but should be in a main figure. Here the authors cite previous work indicating that ProteoStat puncta co-localize with ubiquitinated proteins and state that they do not see this, then essentially just move on. Is there an explanation for this discrepancy and can it be resolved? What do they think is really going on? What happens to levels of ubiquitinated proteins when lipid metabolism is perturbed as in these experiments?

We have included the lipid-induced lysosomal protein aggregation data in the main text (Figure 3A-B), and provided fine-grained quantification of the cytosolic-vs-lysosomal ProteoStat / Ub / ThT signals under different aggregate-inducing conditions (Figure S4A-D). We discuss these results in the main text and propose a model involving ESCRT-mediated microautophagy in the main text. This is supported further by the LysoIP-proteomics and LMP analysis.

- Please add an indicator of amino acid numbers to Fig. 3C.

These annotations are now included (now Figure S3C).

- The legend for 3D is mislabelled.

We have corrected the legend (now Figure S3D).

Reviewer #3 (Recommendations For The Authors):

Protein homeostasis and lipid homeostasis are both are important for maintaining cellular functions. However, the crosstalk remains largely unknown. The manuscript entitled as "Impairment of lipid homoeostasis causes accumulation of protein aggregates in the lysosome" deals with this interesting topic. An important link between lysosomal protein aggregation and sphingolipids/cholesterol esters metabolism were discovered. The topic belonging to the Cell Biology domain also falls into the aims and scope of eLife. Here are the revisions I recommend:

(1) From lipidomics analysis, a remarkable correlation between levels of sphingomyelin and cholesterol ester and ProteoStat staining was found. Could the authors explain how sphingomyelin and cholesterol ester are quantified? The two lipids are not included as internal standards from the lipidomics experiment.

Sphingomyelin and cholesterol ester internal standards are included in the Avanti 330707 SPLASH® LIPIDOMIX® Mass Spec Standard, which was supplied at 3% v/v to the MeOH/H2O cell lysis buffer. We have amended the Methods section to clarify this.

(2) Could the authors perhaps delete Figure 1B and show it on Figure 2A only? There is no need to show the same figure two times. The threshold of both False Discovery Rate and Median Enrichment needs to be added. From Figure 2A, the Lysosomal hydrolases (GBA, LIPA, GALC) seems located in statistically insignificant region. Based on previous studies, the GBA could have an effect on sphingolipid levels, then how to explain that sphingomyelin was highly correlated with ProteoSate staining?

We have combined the two volcano plots into a single figure (now Figure 1D), and added a line to help visualize the gene effects while considering the combined contribution of FDR and enrichment. Individual lysosomal hydrolases indeed have insignificant effects on ProteoStat and this is discussed in the main text as having relatively constrained impacts on the general lipidome. For example, while GBA and GALC KDs can lead to accumulation of their immediate substrates (glucosylceramide and galactosylceramide, respectively), they do not directly impinge on sphingomyelin.

(3) The authors show the corelation between ProteoState staining and different lipids/lipid classes in Figure 3B and Figure S3A. It is not necessary to show the corelation with individual lipids (such as sphingomyelin(d18:1/24:0) and cholesterol ester(18:2). The corelation with full collection of lipid classes would be more representative, which is only list in Figure 3B and Figure S3A. It is suggested to add the information of how many individual lipids in each chass are used for the correlation analysis. Replace Figure 3A to Figure S3A, and put Figure 3A as supplementary figure are suggested.

We decided to retain the correlation of two individual lipids (a sphingomyelin and a cholesterol ester species) with ProteoStat as examples to illustrate with clarity how we obtained the class-wide comparison. The number of individual lipids included in each class for correlation analysis is now included in Figures 2F and S3A.

(4) The authors state that lipid uptake and metabolism modulate proteostasis. However, only cholesterol and LDL were tested. It would be more precise to state as cholesterol uptake and metabolism modulate proteostasis. In addition, sphingolipids and cholesterol esters accumulate with increased lysosomal protein aggregation. It would be interesting to see the effects of sphingolipids uptake, since sphingolipids are correlated with proteostasis better than cholesterol.

We attempted to add back specific sphingolipids to assess sufficiency. However, we found it challenging to ensure that these lipids were distributed to the correct subcellular locations at physiologically relevant levels. Without this crucial information, it was difficult to draw any conclusions about the sufficiency of the sphingolipids we tested to impair proteostasis.

Alonso A, Goñi FM. 2018. The Physical Properties of Ceramides in Membranes. Annu Rev Biophys 47:633–654. doi:10.1146/annurev-biophys-070317-033309

Ikonen E, Zhou X. 2021. Cholesterol transport between cellular membranes: A balancing act between interconnected lipid fluxes. Dev Cell 56:1430–1436. doi:10.1016/j.devcel.2021.04.025

Niekamp P, Scharte F, Sokoya T, Vittadello L, Kim Y, Deng Y, Südhoff E, Hilderink A, Imlau M, Clarke CJ, Hensel M, Burd CG, Holthuis JCM. 2022. Ca2+-activated sphingomyelin scrambling and turnover mediate ESCRT-independent lysosomal repair. Nat Commun 13:1875. doi:10.1038/s41467-022-29481-4

eLife Assessment

This study provides a valuable look at genome-wide RNaseA-resistant RNA-DNA interactions in human embryonic stem cells. The research indicated that RNase treatment maintained long-range RNA-chromatin connections characterized by significant sequence conservation while abolishing permissive interactions. Interestingly, coding and non-coding RNA transcripts exhibited differing sensitivity to RNase treatment. Although the study findings reveal an intriguing RNase-inaccessible regulatory RNA-chromatin interactome, conclusions about the identity and regulatory significance of RNase-resistant RNA-chromatin interactions are incomplete and would benefit from more rigorous approaches that include additional computational and experimental controls.

Reviewer #1 (Public review):

Summary:

This manuscript constitutes further analysis of a dataset generated for a previously-published study from the same group. The experiments in the previous work use an RNA-DNA proximity assay to capture RNAs that interact with chromatin, especially beyond their site of transcription, by crosslinking-and proximity ligation. The previous work added one novel feature to this treatment, compared to other studies by the same group, where they treated the nuclei with RNase A prior to crosslinking. The initial study concluded that long-range chromatin interaction via chromatin looping is affected by RNase treatment. In the current manuscript, the group analyze the data from this experiment in more detail. They describe some notable features of RNAs that remain after RNase treatment and where they are associated within the genome. Overall, the further analyses are somewhat useful, with some exceptions for specific analyses that are not clear in the current manuscript. The work is very complementary to the previously published original study, to the point that it is surprising it was not included in that study.

Strengths:

(1) The analyses are a useful complement that fill in gaps from the Calandrelli et al paper. Some of the findings are suggestive of RNA-protein networks that operate at long distances to regulate promoters.

Weaknesses:

(1) The beginning of the Results section, and elsewhere, describes steps that likely were performed in the previous publication from which the data are being further analyzed and possibly partially reanalyzed. The current manuscript should more clearly describe if there are any aspects of the pipeline that have been modified from the Calandrelli study (which does not have much detail regarding iMARGI parameters in the published paper) for the further analysis in this manuscript.

(2) The RNase treatment approach is similar to that addressed in recent papers from the Jenner and Davidovich groups (https://doi.org/10.1016/j.celrep.2024.113856; https://doi.org/10.1016/j.celrep.2024.113858) where these groups found RNase treatment significantly affected solubility of chromatin, causing aggregation. The authors should address this work and place it in light of their current study.

(3) Figure 1f: it is not clear what it means for genes to be "non-differentially expressed" in this context. Isn't this also RNase-insensitive? And how is the "Ctrl specific" RNA set determined? This is confusing, since RNase is assumed to degrade most of the RNA in these samples.

(4) Figure 2a: The results are somewhat surprising, given that protein-coding genes are depleted more in the RNase treatment. Is the Ctrl set the same as in 1f? This emphasizes the importance of defining that population better.

(5) Figure 3a: The text references this figure in ways that do not match the figure, referencing at least nine column clusters when there are only six. Heatmaps of certain TFs and "RAH explained" percentages don't seem to match the Results section description, either. The authors claim EZH2 binding sites are the top TF overlap with RAHs and yet do not include EZH2 in Figure 3a. Suz12 (EZH2 binding partner) and H3K27me3 (EZH2 product) are also referenced in the text for this figure, but not included in the figure itself.

(6) The manuscript uses the term "non-diffusive RNA-chromatin interactome" which is not directly supported by data. The authors use the term initially to describe the RNase-resistant species in their previous work, but through the current study, they support a model where the RNase resistance is simply due to protection by protein binding, not by any constraints on diffusion in particular chromatin environments.

Reviewer #2 (Public review):

Summary:

In this manuscript, the authors re-analyze RNase-treated iMARGI data to systematically identify and analyze RNase-resistant RNA-chromatin interactions. In general

Strengths:

Analyses are well-thought-out and generally solid.

Weaknesses:

Conclusions are massively overstated, and though the analytical pipelines used are solid, the conclusions deriving from them lack the backing of solid computational and molecular controls.

Reviewer #3 (Public review):

Summary:

The study investigated stable RNA-chromatin interactions by applying RNase treatment before the iMARGI (in situ mapping of RNA-genome interactome) procedure to remove promiscuous, unprotected RNA transcripts and selectively enrich for RNA-inaccessible, potentially functional RNA-chromatin interactions (RNA-Transcription factor and RNA-histone). The researchers found that short-range interactions (<1kb) are RNase resistant, possibly due to the protection from RNA polymerases. They noticed that long-range RNA-chromatin interactions (>2Mbp or interchromosomal) were also enriched after RNase treatment, hypothesizing that these interactions are stabilized by chromatin-binding proteins. They found that genic caRNAs were sensitive, while repeat-derived caRNAs, such as rRNA and satellite repeats, were resistant to RNase. Long non-coding RNAs (lncRNAs), particularly those associated with diseases, were over-represented among RNase-insensitive transcripts, indicating their potential regulatory significance. Additionally, RNase-insensitive caRNAs exhibited higher evolutionary conservation, implying that they are protected by protein complexes, especially in long-range interactions. RNA Attachment Hot Zones (RAHs) enriched post-RNase treatment were found to localize in functional genomic regions such as promoters, transcription factor binding sites (TFBS), and histone modification sites. Importantly, RNase treatment amplified specific RNA-transcription factor interactions, with caRNA signals being preserved at TFBS for factors with RNA-binding capabilities, suggesting that direct RNA-protein binding helps protect caRNAs from degradation. They also found that different TFs are enriched with specific caRNA species, distinguishing them from their genomic footprints. In addition, transcripts with higher abundance tend to enrich at more TFBS. Overall, the study highlights the role of RNase-inaccessible caRNAs in chromatin regulation and provides insight into their functional significance in genome organization.

Strengths:

This study involves rigorous and comprehensive data analysis involving datasets with very high sequencing depth and appropriate statistical tests (e.g., chi-square tests to validate the association between caRNAs and TFBS statistically). This analysis was further strengthened by comparing their results with orthogonal datasets, such as RedChIP and fRIP-seq, providing robust, cross-validated evidence for the caRNA-TFBS associations. In addition to examining broad interactions, the authors identified specific long-range RNA-chromatin interactions and pinpointed specific transcription factors and histone modification markers that are associated with these interactions. The authors explored the evolutionary implications of RNase-insensitive caRNAs and their potential medical relevance, particularly by identifying caRNAs linked to disease-associated genes and long non-coding RNAs (lncRNAs). This combination of detailed analysis, along with functional relevance, broadens the scope of the research, making it a significant contribution to chromatin biology.

Weaknesses:

However, I have the following concerns regarding the studies:<br /> (1) I don't understand the logic behind calling promoters, enhancers, and similar regions "functionally important regions" when describing the enrichment of RNase-insensitive interactions. Genic regions that are RNase-sensitive are also functionally relevant. So, what makes promoters, enhancers, etc, unique in terms of functionality?<br /> (2) First, while the study offers strong evidence for associations between caRNAs, transcription factors, and chromatin markers, it lacks direct functional validation experiments such as RNA knockdown or CRISPR interference, to confirm the specific roles of these RNAs in gene regulation or chromatin structure modifications.<br /> (3) Another limitation is the incomplete investigation of caRNAs with short-range interactions (<1kb). The authors hypothesized that these are protected by RNA polymerases but did not provide supporting experimental evidence or references to previous studies. Offering either experimental validation or a rationale for excluding these short-range interactions would strengthen this hypothesis. The conclusion that authors drew on that "chromatin-associated RNAs (caRNAs) involved in short- to middle-range interactions are more susceptible to RNase treatment" was unclear for the specific "short-range" distance. The data shown in Supplementary Figure 2a contradicted the conclusion in the discussion that "long-distance RNA-chromatin interactions are preferentially preserved after RNase treatment, while short-range interactions are depleted." as well as the suggestion made linking RNase inaccessibility to evolutionarily conserved in the paper.<br /> (4) The study heavily relies on RNase treatment to isolate stable RNA-chromatin interactions, which might neglect important transient or weak interactions and overlook the functional relevance of RNase-sensitive interactions, hence missing the dynamic nature of RNA-chromatin interactions.<br /> (5) Tthe analysis is limited to human embryonic stem cells (H1 cells), which might restrict the generalizability of the findings. Expanding the study to include a cell type that represents a broader range of cell types or tissues will strengthen the conclusions.<br /> (6) The term "RNase A treatment" in the methods section could be clearer if specified as "RNase-treated iMARGI," which encompasses the standard iMARGI protocol.<br /> (7) There is some ambiguity regarding whether the researchers generated new data or reanalyzed existing datasets. While it is mentioned early on that previously published RNase-treated iMARGI datasets were reanalyzed, the text later states that "three biological replicates were generated for the RNase-treated samples." Clarifying whether the data were newly generated in this study or obtained from public datasets would improve the clarity.<br /> (8) The color scheme should be the same for heatmaps for control, and RNase-treated samples in Figure 4.

Author response:

We thank the editors and reviewers for their thorough evaluation of our manuscript. We appreciate the constructive feedback and insights provided. 

We acknowledge that some of our conclusions would benefit from more measured statements and additional computational controls. We will revise the manuscript to better reflect the scope and limitations of our analytical approach. While we cannot add new experimental validations at this stage, we will strengthen our computational analyses and clarify our methodology.

Below, we outline our planned revisions to address the major points raised in the public reviews:

Clarification of Terms and Definitions:

(1) We will make it clearer in our manuscript to emphasize that we reuse the same raw datasets from our previous study as described in Calendrilli et al, 2023, and there is no modification to the experimental methods or data. 

(2) We will provide clear definitions for:

- "Non-differentially expressed" genes

- "Ctrl specific" RNA sets

- The composition of control populations in different analyses

(3) We will revise the use of "non-diffusive RNA-chromatin interactome" and “RNase-resistant” terminology to better reflect our actual findings.

(4) We will also improve clarity regarding:

- The rationale for focusing on specific genomic regions

- The interpretation of evolutionary conservation data

(5) We will provide additional rationale on the exclusion of short-range interactions.

Figure Revisions:

(1) Figure 3a: We will correct any discrepancy between text references and figure content.

(2) Figure 4: We will standardize the color scheme between control and RNase-treated samples.

(3) We will follow the reviewer's suggestion to move figure 1g to the supplementary file. 

Additional Computational Analyses:

(1) We will consider adding controls for RNA length effects and integrate any existing knowledge on the protection extent variation across different RBP.

Discussions:

(1) We will carefully rephrase our conclusions to more accurately reflect the scope and limitations of our computational findings, ensuring we do not overstate the implications.

(2) We will expand the discussion of limitations, including:

- The focus on RNase-resistant interactions only

- The cell-type specificity of our findings

- The lack of functional validation

- The limited ability to discern and study the transient or weak RNA-chromatin interactions using the current dataset

(3) Regarding the recent papers from Jenner and Davidovich groups about RNase treatment effects on chromatin solubility:

- We will discuss these findings in our revised manuscript

- We will address potential limitations this may impose on our interpretations

eLife Assessment

The current study presents useful findings about the inhibition of a membrane pyrophosphatase by non-hydrolyzable phosphonate substrate analogs. The study proposes a model in which the two monomers in a functional dimer interact with the phosphonate molecules in an asymmetric fashion. While asymmetry has been previously demonstrated through other studies, the DEER spectroscopy data presented in the current study provide incomplete evidence of the proposed asymmetry near the binding site.

Reviewer #1 (Public review):

Summary:

This work examines the binding of several phosphonate compounds to a membrane-bound pyrophosphatase using several different approaches, including crystallography, electron paramagnetic resonance spectroscopy, and functional measurements of ion pumping and pyrophosphatase activity. The work attempts to synthesize these different approaches into a model of inhibition by phosphonates in which the two subunits of the functional dimer interact differently with the phosphonate.

Strengths:

This study integrates a variety of approaches, including structural biology, spectroscopic measurements of protein dynamics, and functional measurements. Overall, data analysis was thoughtful, with careful analysis of the substrate binding sites (for example calculation of POLDOR omit maps).

Weaknesses:

Unfortunately, the protein did not crystallize with the more potent phosphonate inhibitors. Instead, structures were solved with two compounds with weak inhibitory constants >200 micromolar, which limits the molecular insight into compounds that could possibly be developed into small molecule inhibitors. Likewise, the authors choose to focus the spectroscopy experiments on these weaker binders, missing an opportunity to provide insight into the interaction between more potent binders and the protein.

In general, the manuscript falls short of providing any major new insight into membrane-bound pyrophosphatases, which are a very well-studied system. Subtle changes in the structures and ensemble distance distributions suggest that the molecular conformations might change a little bit under different conditions, but this isn't a very surprising outcome. It's not clear whether these changes are functionally important, or just part of the normal experimental/protein ensemble variation.

The ZLD-bound crystal structure doesn't predict the DEER distances, and the conformation of Na+ binding site sidechains in the ZLD structure doesn't predict whether sodium currents occur. This might suggest that the ZLD structure captures a conformation that does not recapitulate what is happening in solution/ a membrane.

Reviewer #2 (Public review):

Summary:

Crystallographic analysis revealed the asymmetric conformation of the dimer in the inhibitor-bound state. Based on this result, which is consistent with previous time-resolved analysis, authors verified the dynamics and distance between spin introduced label by DEER spectroscopy in solution and predicted possible patterns of asymmetric dimer.

Strengths:

Crystal structures with inhibitor bound provide detailed coordination in the binding pocket thus useful information for the PPase field and maybe for drug development.

Weaknesses:

The distance information measured by DEER is advantageous for verifying the dynamics and structure of membrane protein in solution. However, regarding T211 data, which, as the authors themselves stated, lacks measurement precision, it is unclear for readers how confident one can judge the conclusion leading from these data for the cytoplasmic side.

The distance information for the luminal site, which the authors claim is more accurate, does not indicate either the possibility or the basis for why it is the ensemble of two components and not simply a structure with a shorter distance than the crystal structure.

Reviewer #3 (Public review):

Summary:

Membrane-bound pyrophosphatases (mPPases) are homodimeric proteins that hydrolyze pyrophosphate and pump H+/Na+ across membranes. They are attractive drug targets against protist pathogens. Non-hydrolysable PPi analogue bisphosphonates such as risedronate (RSD) and pamidronate (PMD) serve as primary drugs currently used. Bisphosphonates have a P-C-P bond, with its central carbon can accommodate up to two substituents, allowing a large compound variability. Here the authors solved two TmPPase structures in complex with the bisphosphonates etidronate (ETD) and zoledronate (ZLD) and monitored their conformational ensemble using DEER spectroscopy in solution. These results reveal the inhibition mechanism of these compounds, which is crucial for developing future small molecule inhibitors.

Strengths:

The authors show that seven different bisphosphonates can inhibit TmPPase with IC50 values in the micromolar range. Branched aliphatic and aromatic modifications showed weaker inhibition.

High-resolution structures for TmPPase with ETD (3.2 Å) and ZLD (3.3 Å) are determined. These structures reveal the binding mode and shed light on the inhibition mechanism. The nature of modification on the bisphosphonate alters the conformation of the binding pocket.

The conformational heterogeneity is further investigated using DEER spectroscopy under several conditions.

Weaknesses:

The authors observed asymmetry in the TmPPase-ELD structure above the hydrolytic center. The structural asymmetry arises due to differences in the orientation of ETD within each monomer at the active site. As a result, loop5-6 of the two monomers is oriented differently, resulting in the observed asymmetry. The authors attempt to further establish this asymmetry using DEER spectroscopy experiments. However, the (over)interpretation of these data leads to more confusion than any further understanding. DEER data suggest that the asymmetry observed in the TmPPase-ELD structure in this region might be funneled from the broad conformational space under the crystallization conditions.

DEER data for position T211R1 at the enzyme entrance reveal a highly flexible conformation of loop5-6 (and do not provide any direct evidence for asymmetry, Figure EV8). Similarly, data for position S521R1 near the exit channel do not directly support the proposed asymmetry for ETD. Despite the high quality of the data, they reveal a very similar distance distribution. The reported changes in distances are very small (+/- 0.3 nm), which can be accommodated by a change of spin label rotamer distribution alone. Further, these spin labels are located on a flexible loop, thereby making it difficult to directly relate any distance changes to the global conformation.

The interpretations listed below are not supported by the data presented:

(1) 'In the presence of Ca2+, the distance distribution shifts towards shorter distances, suggesting that the two monomers come closer at the periplasmic side, and consistent with the predicted distances derived from the TmPPase:Ca structure.'

Problem: This is a far-stretched interpretation of a tiny change, which is not reliable for the reasons described in the paragraph above.

(2) 'Based on the DEER data on the IDP-bound TmPPase, we observed significant deviations between the experimental and the in silico distances derived from the TmPPase:IDP X-ray structure for both cytoplasmic- (T211R1) and periplasmic-end (S525R1) sites (Figure 4D and Figure EV8D). This deviation could be explained by the dimer adopting an asymmetric conformation under the physiological conditions used for DEER, with one monomer in a closed state and the other in an open state.'

Problem: The authors are trying to establish asymmetry using the DEER data. Unfortunately, no significant difference is observed (between simulation and experiment) for position 525 as the authors claim (Figure 4D bottom panel). The observed difference for position 112 must be accounted for by the flexibility and the data provide no direct evidence for any asymmetry.

(3) 'Our new structures, together with DEER distance measurements that monitor the conformational ensemble equilibrium of TmPPase in solution, provide further solid experimental evidence of asymmetry in gating and transitional changes upon substrate/inhibitor binding.'

Problem: See above. The DEER data do not support any asymmetry.

(4) Based on these observations, and the DEER data for +IDP, which is consistent with an asymmetric conformation of TmPPase being present in solution, we propose five distinct models of TmPPase (Figure 7).

Problem: Again, the DEER data do not support any asymmetry and the authors may revisit the proposed models.

(5) 'In model 2 (Figure 7), one active site is semi-closed, while the other remains open. This is supported by the distance distributions for S525R1 and T211R1 for +Ca/ETD informed by DEER, which agrees with the in silico distance predictions generated by the asymmetric TmPPase:ETD X-ray structure'

Problem: Neither convincing nor supported by the data

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This work examines the binding of several phosphonate compounds to a membrane-bound pyrophosphatase using several different approaches, including crystallography, electron paramagnetic resonance spectroscopy, and functional measurements of ion pumping and pyrophosphatase activity. The work attempts to synthesize these different approaches into a model of inhibition by phosphonates in which the two subunits of the functional dimer interact differently with the phosphonate.

Strengths:

This study integrates a variety of approaches, including structural biology, spectroscopic measurements of protein dynamics, and functional measurements. Overall, data analysis was thoughtful, with careful analysis of the substrate binding sites (for example calculation of POLDOR omit maps).

Weaknesses:

Unfortunately, the protein did not crystallize with the more potent phosphonate inhibitors. Instead, structures were solved with two compounds with weak inhibitory constants >200 micromolar, which limits the molecular insight into compounds that could possibly be developed into small molecule inhibitors. Likewise, the authors choose to focus the spectroscopy experiments on these weaker binders, missing an opportunity to provide insight into the interaction between more potent binders and the protein.

We acknowledge the reviewer concern regarding the choice of weaker inhibitors. We attempted co-crystallization with all available inhibitors, including those with higher potency. However, despite numerous efforts, these potent inhibitors yielded low-resolution crystals, making them unsuitable for detailed structural analysis. Therefore, we chose to focus on the weaker binders, as we were able to obtain high-quality crystal structures for these compounds. This allowed us to perform DEER spectroscopy with the added advantage of accurately analyzing the data against structural models derived from X-ray crystallography. Using these weaker inhibitors enabled a more precise interpretation of the DEER data, thus providing reliable insights into the conformational dynamics and inhibition mechanism. However, as suggested by the reviewer, in the revised version, we will perform DEER analysis on the more potent inhibitors to provide additional insight into their interactions.

In general, the manuscript falls short of providing any major new insight into membrane-bound pyrophosphatases, which are a very well-studied system. Subtle changes in the structures and ensemble distance distributions suggest that the molecular conformations might change a little bit under different conditions, but this isn't a very surprising outcome. It's not clear whether these changes are functionally important, or just part of the normal experimental/protein ensemble variation.

We respectfully disagree with the reviewer. The scale of motions seen in this study correspond to those seen in the full panoply of crystal structures of mPPases. Some proteins undergo very large conformational changes during catalysis – such as the rotary ATPase. This one doesn’t, meaning that the precise motions we describe are likely to be relevant. Conformational changes in the ensemble, whether large or small, represent essential protein motions which underlie key mPPase catalytic function. Our DEER spectroscopy data demonstrate the sensitivity and resolution necessary to monitor these subtle changes in equilibria, even if these are only a few Angstroms. For several of the conditions we investigated by DEER in solution, corresponding x-ray structures have been solved, with the derived distances agreeing well with the DEER distributions. This further validates the biological relevance of the structures, including serial time-resolved ones that indicate asymmetry.

The ZLD-bound crystal structure doesn't predict the DEER distances, and the conformation of Na+ binding site sidechains in the ZLD structure doesn't predict whether sodium currents occur. This might suggest that the ZLD structure captures a conformation that does not recapitulate what is happening in solution/ a membrane.

We agree with the reviewer that the ZLD-bound crystal structure does not predict the DEER distances. However, we believe this discrepancy arises from the effect of the bulkiness of ZLD inhibitor, which prevents the closure of the hydrolytic centre. Additionally, the absence of Na+ at the ion gate in the ZLD-bound structure suggests that Na+ transport does not occur, a conclusion further supported by our electrometric measurements. We agree with the reviewer, that the distances observed in the DEER experiments might represent a potential new conformation in solution, which may not be captured by the static X-ray structure, thereby offering insights into the dynamic nature of the protein under physiological conditions. Finally, the static x-ray structures have not captured the asymmetric conformations that must exist to explain half-of-the-sites reactivity.

Reviewer #2 (Public review):

Summary:

Crystallographic analysis revealed the asymmetric conformation of the dimer in the inhibitor-bound state. Based on this result, which is consistent with previous time-resolved analysis, authors verified the dynamics and distance between spin introduced label by DEER spectroscopy in solution and predicted possible patterns of asymmetric dimer.

Strengths:

Crystal structures with inhibitor bound provide detailed coordination in the binding pocket thus useful information for the PPase field and maybe for drug development.

Weaknesses:

The distance information measured by DEER is advantageous for verifying the dynamics and structure of membrane protein in solution. However, regarding T211 data, which, as the authors themselves stated, lacks measurement precision, it is unclear for readers how confident one can judge the conclusion leading from these data for the cytoplasmic side.

We thank the reviewer for acknowledging the advantageous use of the DEER methodology for identifying dynamic states of membrane proteins in solution. We used two sites in our analysis: S525 (periplasm) and T211 (cytoplasm). As we clearly stated in the original manuscript, S525R1 yielded high-quality DEER data, while T211R1 yielded weak (or no) visual oscillations, leading to broad, though different distributions for the several conditions tested. Our main conclusions are based on the S525R1 data. We included the T211R1 data because, although it does not provide definitive evidence, it is consistent with our proposed model and offers additional insights into biologically relevant conditions. Furthermore, the shifts in the centre of mass (Fig EV8D) of the broad T211R1 distributions show a trend that is consistent with our model; although not proving it, it does not exclude it either. Lastly, these data do indeed confirm an important structural feature of mPPase in solution conditions which is the intrinsically high dynamic state of the loop5-6 where T211 is located, and consistent with our previous (Kellosalo et al., Science,  2012; Li et al., Nat. Commun, 2016; Vidilaseris et al., Sci. Adv., 2019; Strauss et al., EMBO Rep., 2024) and current x-ray crystallography data.

The distance information for the luminal site, which the authors claim is more accurate, does not indicate either the possibility or the basis for why it is the ensemble of two components and not simply a structure with a shorter distance than the crystal structure.

We thank the reviewer for pointing out this possibility and alternative interpretation of our DEER data. In the revised version, we will show that our DEER data are consistent with (and do not exclude) asymmetry and rephrase to be inclusive of other possibilities. Importantly, this additional possibility does not affect the current interpretation of the data in our manuscript.

Reviewer #3 (Public review):

Summary:

Membrane-bound pyrophosphatases (mPPases) are homodimeric proteins that hydrolyze pyrophosphate and pump H+/Na+ across membranes. They are attractive drug targets against protist pathogens. Non-hydrolysable PPi analogue bisphosphonates such as risedronate (RSD) and pamidronate (PMD) serve as primary drugs currently used. Bisphosphonates have a P-C-P bond, with its central carbon can accommodate up to two substituents, allowing a large compound variability. Here the authors solved two TmPPase structures in complex with the bisphosphonates etidronate (ETD) and zoledronate (ZLD) and monitored their conformational ensemble using DEER spectroscopy in solution. These results reveal the inhibition mechanism of these compounds, which is crucial for developing future small molecule inhibitors.

Strengths:

The authors show that seven different bisphosphonates can inhibit TmPPase with IC50 values in the micromolar range. Branched aliphatic and aromatic modifications showed weaker inhibition.

High-resolution structures for TmPPase with ETD (3.2 Å) and ZLD (3.3 Å) are determined. These structures reveal the binding mode and shed light on the inhibition mechanism. The nature of modification on the bisphosphonate alters the conformation of the binding pocket.

The conformational heterogeneity is further investigated using DEER spectroscopy under several conditions.

Weaknesses:

The authors observed asymmetry in the TmPPase-ELD structure above the hydrolytic center. The structural asymmetry arises due to differences in the orientation of ETD within each monomer at the active site. As a result, loop5-6 of the two monomers is oriented differently, resulting in the observed asymmetry. The authors attempt to further establish this asymmetry using DEER spectroscopy experiments. However, the (over)interpretation of these data leads to more confusion than any further understanding. DEER data suggest that the asymmetry observed in the TmPPase-ELD structure in this region might be funneled from the broad conformational space under the crystallization conditions.

See also the response below - We respectfully disagree with the reviewer. The asymmetry was previously established using serial time crystallography (Strauss et al., EMBO Rep, 2024) and biochemical assays (e.g. Malinen et al., Prot. Sci., 2022; Artukka et al., Biochem J, 2018; Luoto et al., PNAS, 2013) and also partially seen in one static structure (Vidilaseris et al., Sci Adv 2019). DEER data only show that the previously proposed asymmetry could also be present within the conformational ensemble in solution conditions. Indeed, our data do not (and cannot) exclude this possibility.

DEER data for position T211R1 at the enzyme entrance reveal a highly flexible conformation of loop5-6 (and do not provide any direct evidence for asymmetry, Figure EV8).

Please see relevant response above. We acknowledge that T211 is indeed situated on a highly dynamic loop, which is important for gating and our DEER data confirm its high flexibility. Given we have not observed oscillations of this site, leading to broad distributions, we have stated in the original manuscript that we will not establish the presence of any asymmetry in solution on the basis of T211, rather relying on the S525 site, for which we have acquired high-quality DEER data, as was also pointed out and have been commented on by all reviewers.

Similarly, data for position S521R1 near the exit channel do not directly support the proposed asymmetry for ETD.

The reviewer appears to suggest that we hold the S525R1 DEER data as direct proof of asymmetry; this is combative on the grounds that to directly prove asymmetry would require time-resolved DEER measurements, far beyond the scope of this work. Rather, we have applied DEER measurements to explore whether asymmetry (observed previously via time-resolved X-ray crystallography) is also present (or indeed a possibility) in solution. We simply state that the DEER data are consistent with asymmetry (i.e., that the mean distance increases in the presence of ETD compared to the apo-state). This is a restrained interpretation of the data.

Despite the high quality of the data, they reveal a very similar distance distribution. The reported changes in distances are very small (+/- 0.3 nm), which can be accommodated by a change of spin label rotamer distribution alone. Further, these spin labels are located on a flexible loop, thereby making it difficult to directly relate any distance changes to the global conformation

We thank the reviewer for recognising the high quality of our DEER data for the S525R1, where visual oscillations in the raw traces, as in our case, reportedly lead to highly accurate and reliable distributions, able to separate (in fortuitous cases) helical movements of only a few Angstroms. The ability of DEER/PELDOR offering near Angstrom resolution was previously demonstrated by the acquisition and solution of high resolution multi-subunit spin-labelled membrane protein structures (Pliotas at al., PNAS, 2012; Pliotas et al., Nat Struct Mol Biol, 2015; Pliotas, Methods Enzymol, 2017) as well as it ability in detecting small (and of similar to mPPase magnitude) conformational changes in different integral membrane proteins systems (Kapsalis et al., Nature Comms, 2019; Kubatova et al., PNAS, 2023; Schmidt et al., JACS, 2024; Lane et al., Structure, 2024; Hett et al., JACS, 2021; Zhao et al., Nature, 2024), occurring under different conditions and/or stimuli in solution and/or lipid environment. The changes here are not very small (e.g. ~ 7 Angstroms between the two mean distance extremes (Ca vs IDP)) for DEER’s proven detection sensitivity, and with all other conditions showing changes between those extremes.

These changes are relatively small, but they are expected for membrane ion pumps. Indeed, none of the mPPase structures show helical movements of greater than a half a turn, and that only in helices 6 and 12. There appear to be larger-scale loop closing motions of the 5-6 loop that includes T211, due to the presence of E217 which binds to one of the Mg2+ ions that coordinate the leaving group phosphate. (This is, inter alia, the reason that this loop is so flexible: it can not order before substrate is bound.) Here we have the resolution to detect such subtle differences by DEER, given there are clear shifts in our time domain data and these are reflected in the mean distances in the distributions. Therefore, our study demonstrates the sensitivity and resolution DEER offers in detecting subtle conformational transitions, key in membrane proteins pathways. To further belabour this point, we do not quantify the DEER data (for instance through parametric fitting) to extract populations of different conformational states and we appreciate that to do so would be highly prone to error; however we do (and can, we feel without overinterpretation) assert that the mean distances shift.

The interpretations listed below are not supported by the data presented:

(1) 'In the presence of Ca2+, the distance distribution shifts towards shorter distances, suggesting that the two monomers come closer at the periplasmic side, and consistent with the predicted distances derived from the TmPPase:Ca structure.' Problem: This is a far-stretched interpretation of a tiny change, which is not reliable for the reasons described in the paragraph above.

While the authors overall agree with the reviewer assessment that ±0.3 nm is a small (not a minor) change, there are literature examples quantifying (or using for quantification) distribution peaks separated by similar Δr. (Kubatova et al., PNAS, 2023; Schmidt et al., JACS, 2024; Hett et al., JACS, 2021; Zhao et al., Nature, 2024). In particular, none of the mPPase structures show helical movements of greater than a half a turn (in helices 6 and 12 in particular). There appear to be larger-scale loop closing motions of the 5-6 loop that includes T211, due to the presence of E217 which binds to one of the Mg2+ ions that coordinate the leaving group phosphate. (This is, inter alia, the reason that this loop is so flexible: it can not order before substrate is bound.)

Importantly, we have fitted Gaussians to the experimental distance distributions of 525R1 output by the Comparative Deer Analyzer 2.0 and observed a change in the distribution width in presence of Ca2+, implying the rotameric freedom of the spin label is restricted. However, the CW-EPR for 525R1 indicate that the rotational correlation time of the spin label is highly consistent between conditions (the spectra are almost identical); this cannot be explained simply by rotameric preference of the spin label (as asserted by the reviewer 3), as there is no (further) immobilisation observed from the CW-EPR of apo-state (Figure EV9) to that in presence of Ca2+. Furthermore, in the absence of conformational changes, it is reasonable to assume (and demonstrable from the CW-EPR data) that the rotamer cloud should not significantly change between conditions. However, Gaussian fits of the two extreme cases yielding the longest (i.e., in presence of IDP) and shortest (in presence of ZTD) mean distances for the 525R1 DEER data indicated significant (i.e., above the noise floor after Tikhonov validation) probability density for the IDP condition at 50 Å (P(r) = 0.18). This occurs at four standard deviations above the mean of the ZTD condition, which by random chance should occur with <0.007% probability. Indeed, one can say that to observe 18% probability density at four standard deviations above the mean by random chance would occur on the order of one in 4 x 10^6.

As in previous response the method can detect changes of such magnitude which are not small, but physiologically relevant and expected for integral membrane proteins, such as mPPases. Indeed, even in equal (or more) complex systems such as heptameric mechanosensitive channel proteins DEER provided sub-Angstrom accuracy, when a spin labelled high resolution XRC structure was solved (Pliotas et al., PNAS, 2012; Pliotas et al., Nat Struct Mol Biol, 2015). Despite this is ideal case where DEER accuracy was experimentally validated another high resolution structural method on modified membrane protein and is not very common it demonstrates the power of the method , especially when strong oscillations are present in the raw DEER data (as here for mPPase 525R1), even when multiple distances are present, Angstrom resolution is achievable in such challenging protein classes.

(2) 'Based on the DEER data on the IDP-bound TmPPase, we observed significant deviations between the experimental and the in silico distances derived from the TmPPase:IDP X-ray structure for both cytoplasmic- (T211R1) and periplasmic-end (S525R1) sites (Figure 4D and Figure EV8D). This deviation could be explained by the dimer adopting an asymmetric conformation under the physiological conditions used for DEER, with one monomer in a closed state and the other in an open state.'

Problem: The authors are trying to establish asymmetry using the DEER data. Unfortunately, no significant difference is observed (between simulation and experiment) for position 525 as the authors claim (Figure 4D bottom panel). The observed difference for position 112 must be accounted for by the flexibility and the data provide no direct evidence for any asymmetry.

Reviewer 3 is wrong in suggesting that we are trying to prove asymmetry through the DEER data. That is a well-known fact in the literature (eg Vidilaseris et al, Sci Adv 2019 where we show (1) that the exit channel inhibitor ATC (i.e., close to 525) binds better in solution to the TmPPase:PPi complex than the TmPPase:PPi2 complex, and (2) that ATC binds in an asymmetric fashion to the TmPPase:IDP2 complex with just one ATC dimer on one of the exit channels. We merely use the DEER data to support this well-established fact.

However, we agree that the DEER data in presence of IDP does not provide direct proof for asymmetry; particularly mutant T211R1 yields in silico distributions too short for measurement by DEER. It is possible that the deviations observed (and particularly likely for T211R1) arise from conformational heterogeneity in solution. We will rephrase this paragraph accordingly: “Owing to the broad nature of the T211R1 (cytoplasmic site) distance distributions, we refrain from interpreting shifts in this data. For the 525R1 (periplasmic site) for which we obtained data of high quality (as also pointed out by both reviewers 2 and 3) we observed deviations between the experimental and the in-silico distances derived from the TmPPase:IDP X-ray structure. While this deviation is less pronounced than for the +ZTD condition, the deviation is consistent with an asymmetric conformation in solution.”

(3) 'Our new structures, together with DEER distance measurements that monitor the conformational ensemble equilibrium of TmPPase in solution, provide further solid experimental evidence of asymmetry in gating and transitional changes upon substrate/inhibitor binding.'

Problem: See above. The DEER data do not support any asymmetry.

We feel that the reviewer comments here are somewhat unfounded. The DEER data (and we will limit discussion only to the 525R1 mutant in this regard) satisfy relevant criteria of the white paper (Schiemann et al., 2021, JACS) from the EPR community (signal-to-noise ratio w.r.t modulation depth of > 20 in all cases; replicates have been performed and will be added into the main-text or supplementary; near quantitative labelling efficiency (evidenced by lack of free spin label signal in the CW-EPR spectra); analysed using the CDA (now Figure EV10, this data we will promote to the main-text) to avoid confirmation bias).

While the DEER data do not prove asymmetry, we do not claim proof of asymmetry in the above sentence. We concede to rephrase the offending sentence above as: “Our new structures, together with DEER distance measurements that monitor the conformational ensemble of TmPPase in solution, do not exclude asymmetry in gating and transitional changes upon substrate/inhibitor binding and are consistent with our proposed model.” We feel that this reframed conjecture of asymmetry is well founded; indeed, comparing the experimental apo-state 525R1 distance distribution with in-silico modelling performed on the hybridised asymmetric structure (i.e., comprised of one monomer bound to Ca2+ and another bound to IDP) yields an overlap coefficient (Islam and Roux, JPC B, 2015) of >0.97. This implies the envelope of the modelled distance distribution is quantitatively inside the envelope of the experimental distance distribution. Thus, the DEER data do not exclude asymmetry (previously observed by time-resolved XRC) in solution. While we appreciate that ideally one would measure time-resolved DEER to directly correlate kinetics of conformational changes within the ensemble to the catalytic cycle of mPPase,(and this is something we aim to do in the future), it is beyond the the scope of this study.

Indeed, half-of-the-sites reactivity has been demonstrated in at least the following papers (Vidilaseris et al, Sci Acv. ,2019, Strauss et al, EMBO Rep. 2024, Malinen et al Prot Sci, 2022, Artukka et al Biochem J, 2018; Luoto et al, PNAS, 2013). Half-of-the sites activity requires asymmetry in the mechanism, and therefore asymmetric motions in the active site (viz 211) and exit channel (viz 525). As mentioned above, we have demonstrated this for other inhibitors (Vidilaseris et al 2019) and as part of a time-resolved experiment (Strauss et al 2024). In fact, given the wealth of evidence showing that the symmetrical crystal structures sample a non- or less-productive conformation of the protein, it would be quixotic to propose the DEER experiments - in solution - do not generate asymmetric conformations. It certainly doesn’t obey Occam’s razor of choosing the simplest possible explanation that covers the data.

(4) Based on these observations, and the DEER data for +IDP, which is consistent with an asymmetric conformation of TmPPase being present in solution, we propose five distinct models of TmPPase (Figure 7).

Problem: Again, the DEER data do not support any asymmetry and the authors may revisit the proposed models.

We respectfully disagree with the reviewer. Please see our detailed response above. However, in the revised version, we will clarify that the proposed models are not solely based on the DEER data but are grounded in both current and previously solved structures, with the DEER data providing additional consistency with these models.

(5) 'In model 2 (Figure 7), one active site is semi-closed, while the other remains open. This is supported by the distance distributions for S525R1 and T211R1 for +Ca/ETD informed by DEER, which agrees with the in silico distance predictions generated by the asymmetric TmPPase:ETD X-ray structure'

Problem: Neither convincing nor supported by the data

We respectfully disagree with the reviewer. However, owing to the conformational heterogeneity of T211R1, in the revised version, we will exclude it in the above sentence, to the effect: Please see our detailed response above.

eLife Assessment

This important study used whole genome data to investigate Beefalo ancestry for the first time. It provides insight into the genetics of Beefalo cattle, definitively challenging the long-held claim of 37.5% buffalo ancestry reported by the American Beefalo Association. This results are convincing, with a comprehensive range of well-established population genomics methods being used to estimate ancestry in these animals. This work will be of significant interest to evolutionary biologists, population geneticists, animal breeders, and those involved in the conservation genetics of bovine species.

Reviewer #1 (Public review):

Summary:

This study used whole genome data to investigate Beefalo ancestry for the first time, filling the gap in the field of Beefalo ancestry. The authors used preserved semen samples to generate genomic data on 47 registered Beefalo and 3 bison hybrids, further questioning the ABA's stated goal of ⅜ bison ancestry. In addition, the authors also show that ancestry profiles of Beefalo and bison hybrid genomes are consistent with repeated backcrossing to either parental species, demonstrating the value of genomic information in examining gene flow between species in the genus Bison. This is an interesting study that still has some major weaknesses that exist, but overall, the work demonstrates the utility of genomic information in validating specific breeding claims for a more complete understanding of gene flow and genetic variation among bovine species.

Strengths:

Numerous genetic analysis methods such as PCA, ADMIXTURE, F4 ratios, and local ancestry inference techniques revealed that no single Beefalo set meets the ancestry requirements set by the American Beefalo Association (ABA) and some beefalo had detectable indicine cattle ancestry.

Weaknesses:

While this study contributes to our knowledge of Beefalo ancestry, there are some key issues that need to be addressed in terms of analysing the specific results as well as writing the article.

Reviewer #2 (Public review):

Summary:

Shapiro et al. set out to verify the American Beefalo Association's claim that Beefalo cattle possess 37.5% bison ancestry. They employ a comprehensive range of well-established population genomics methods to estimate ancestry in these hybrid populations, including PCA, ADMIXTURE, D and F statistics, and local ancestry inference. Their findings conclusively demonstrate that most Beefalo lack the claimed bison ancestry, with only 8 out of 47 samples showing any detectable bison ancestry, ranging from 2 - 18%.

Strengths:

The primary strength of this analysis lies in the comprehensive dataset available to the authors, which includes important foundational Beefalo individuals and various reference populations. The rigorous and multi-faceted methodological approach employs several well-established techniques in population genomics for detecting and measuring admixture. Each method used has a firm basis in the field, providing consistent and robust results. The authors' approach of using PCA to initially assess the data within a global context, followed by more specific analyses using ADMIXTURE and D-statistics, provides a clear and logical progression of evidence. The presentation of these results in figures is particularly effective, clearly illustrating the key findings of the study. Additionally, the examination of both autosomal and sex chromosome ancestry offers a more complete understanding of Beefalo genetic composition and the mechanics of bison-cattle hybridisation.

Weaknesses:

One limitation of this analysis is the relatively low coverage (~2x) of many Beefalo samples. However, the authors have taken steps to mitigate biases that may arise from this. Another weakness is the limited sampling of contemporary Beefalo populations, as the study focuses primarily on historical samples. This may limit our understanding of how Beefalo genetics may have changed over time.

Appraisal:

The authors have clearly achieved their primary aim using a rigorous and comprehensive methodology. Their extensive dataset and multi-faceted analytical approach provide strong support for their conclusions. The study not only addresses its main research question but also reveals unexpected insights into Beefalo genetics, particularly the presence of zebu ancestry.

Discussion:

This study is valuable for several reasons beyond its primary findings. First, it definitively addresses and refutes the claim of 37.5% bison ancestry in Beefalo, providing crucial information for those studying these interspecies hybrids and the viability of their offspring. Second, it reveals the unexpected presence of zebu ancestry in many Beefalo, raising intriguing questions about the breed's development and the potential role of zebu cattle in achieving desired traits. This finding suggests that the distinctive appearance of Beefalo may be due in part to zebu admixture rather than bison ancestry. Third, the study highlights the significant barriers to admixture between bison and cattle, both in controlled breeding programs and potentially in wild populations. This has important implications for conservation genetics and our understanding of gene flow between these species. Lastly, the study demonstrates the power of genomic analysis in verifying breed claims and understanding the complex history of domestic animal breeds. These findings open new avenues for research in bovine genomics, breed development, and the dynamics of interspecies hybridisation.

Reviewer #3 (Public review):

Summary:

I really like this topic and study. But I think much can be more focused and tightened up. All the components are here - just some more refining to really make the storyline clear, the journey of discovery, and the impact of such knowledge.

Strengths:

The authors dive directly into the question of genomic ancestry as compared to the breed club's reported ancestry with heavy, quantitative data and critical analytical methods. The questioning line is direct and does not meander. The reader learns about the challenges of breeding associations, and values of understood ancestry, and presents a clear need of re-evaluating the breed standards and expectations of beefalo (if ancestry is indeed the primary goal instead of a phenotype-driven breed mission).

Weaknesses:

Much of the quantitative results are only referred to in the main text with qualitative language. Please incorporate more written quantitative results to highlight evidence that underlines the study narrative because it is quite an interesting study!

eLife Assessment

This study presents a valuable finding that MK2 inhibitor CMPD1 can inhibit the growth, migration, and invasion of breast cancer cells both in vitro and in vivo by inducing microtubule depolymerization, preferentially at the microtubule plus-end, leading to cell division arrest. The evidence supporting the conclusion of this paper is solid, although additional experiments and controls are needed to further strengthen the claim. This work will be of interest to breast cancer researchers.

Reviewer #1 (Public review):

In this paper, the authors reveal that the MK2 inhibitor CMPD1 can inhibit the growth, migration, and invasion of breast cancer cells both in vitro and in vivo by inducing microtubule depolymerization, preferentially at the microtubule plus-end, leading to cell division arrest, mitotic defects, and apoptotic cell death. They also showed that CMPD1 treatment upregulates genes associated with cell migration and cell death, and downregulates genes related to mitosis and chromosome segregation in breast cancer cells, suggesting a potential mechanism of CMPD1 inhibition in breast cancer. Besides, they used the combination of an MK2-specific inhibitor, MK2-IN-3, with the microtubule depolymerizer vinblastine to simultaneously disrupt both the MK2 signaling pathway and microtubule dynamics, and they claim that inhibiting the p38-MK2 pathway may help to enhance the efficacy of MTAs in the treatment of breast cancer. However, there are a few concerns, including:

(1) What is the effect of CMPD1 on breast cancer metastasis?

(2) The mechanism is lacking as to how MK2 inhibitors enhance the efficacy of MTAs.

Reviewer #2 (Public review):

Summary:

This study explores the potential of inhibiting the p38-MK2 signaling pathway to enhance the efficacy of microtubule-targeting agents (MTAs) in breast cancer treatment using a dual-target inhibitor.

Strengths:

The study identifies the p38-MK2 pathway as a promising target to enhance the efficacy of microtubule-targeting agents (MTAs), offering a novel therapeutic strategy for breast cancer treatment. In addition, the study employs a wide range of techniques, especially live-cell imaging, to assess the microtubule dynamics in TNBC cells.

Weaknesses:

The study primarily uses RPE1 cells as the control for normal cells, which may not fully capture the response of normal mammary epithelial cells. While CMPD1 is shown to be effective in suppressing tumor growth in MDA-MB-231 xenograft, the study lacks detailed toxicity data to confirm its safety profile in vivo.

Reviewer #3 (Public review):

Summary:

The authors demonstrated MK2i could enhance the therapeutic efficacy of MTAs. With Tumor xenograft and migration assay, the author suggested that the p38-MK2 pathway may serve as a promising therapeutic target in combination with MTAs in cancer treatment.

Strengths:<br /> The authors provided a potential treatment for breast cancer.

Weaknesses:

(1) In Figure 2, the authors used a human retinal pigment epithelial-1 (RPE1) cell line to show that breast cancer cells are more sensitive to CMPD1 treatment. MCF10A cells would be suggested here as a suitable control. Besides, to compare the sensitivity, IC50 indifferent cell lines should be measured.

(2) The data of MDA-MB-231 in Figure 1D is not consistent with CAL-51 and T47D, also not consistent with the data in Figures 2B-C.

(3) To support the authors' conclusion in Figure 5, an additional animal experiment performed by tail vein injection would be helpful.

(4) Page 14, to evaluate the combination result of MK2i and vinblastine, an in vivo animal assay must be performed.

(5) The authors used RNA-seq to show some pathways affected by CMPD1. What are the key/top genes that were affected? How about the mechanism?

(6) Line 127, more experiments should be involved to support the conclusion.

eLife Assessment

This important study highlights the key role of NK cells and PD-L1+ neutrophils in worsening sepsis responses in the context of of MASH (metabolic dysfunction-associated steatohepatitis). While the data are solid, the overall evidence for the role of neutrophils in mediating this effect, which is based on a choline-deficient high-fat diet model of various knockouts or selective ablation of immune cell types, remains incomplete. The study will be of interest to researchers in immunopathological disease mechanisms.

Reviewer #1 (Public review):

Summary:

By using an established NAFLD model, choline-deficient high-fat diet, Barros et al show that LPS challenge causes excessive IFN-γ production by hepatic NK cells which further induces recruitment and polarization of a PD-L1 positive neutrophil subset leading to massive TNFα production and increased host mortality. Genetic inhibition of IFN-γ or pharmacological blockade of PD-L1 decreases recruitment of these neutrophils and TNFα release, consequently preventing liver damage and decreasing host death.

Since NAFLD is often accompanied by chronic, low-grade inflammation, it can lead to an overactive but dysfunctional immune response and increase the body's overall susceptibility to infections, therefore this is a very important research question.

Weaknesses:

I have quite a lot of concerns with this manuscript. One of those is that the authors did not indeed show that the seen effect is really due to NAFLD itself. The role of choline is already known in the context of sepsis since its deficiency (which can be observed in about two weeks through deterioration of liver structure and function) leads to body organ dysfunction both in humans and animals. Nolan and Vilayat in 1968 showed that the hepatic injury and mortality due to endotoxinaemic shock induced by intraperitoneal injection of LPS was significantly increased in adult female Holtzman rats fed on a choline-deficient diet. Therefore, in order to really show that the effect is mediated due to NAFLD some other diet model must be used (e.g. high-fat, high-fructose, and high-cholesterol diet).

Reviewer #2 (Public review):

Summary:

This is an extremely interesting mouse study, trying to understand how sepsis is tolerated during obesity/NAFLD. The researchers combine a well-established model of NASH (Choline-deficiency with High Fat Diet) with a sepsis model (IP injection of 10mg/kg LPS), leading to dramatic mortality in mice. Using this model, they characterize the complex contributions of immune cells. Specifically, they find that NK-cells and Neutrophils contribute the most to mortality in this model due to IFNG and PD-L1+ Neutrophils.

Strengths:

The biggest strength of the manuscript is how clear the primary phenotypes/endpoints of their model are. Within 6 hours of LPS injection, there is a stark elevation of liver inflammation and damage, which is exacerbated by a High Fat/CholineDeficient diet (HFCD). And after 1 day, almost all of the mice die. Using these endpoints, the authors were able to identify which cells were critical for mortality in the model and the specific mediators involved.

Weaknesses:

A few key details regarding the experimental design and interpretation are missing.

Most important is the choice of a high-fat diet with choline deficiency. I believe this model was chosen because the experiments are shorter and typically result in a liver inflammatory phenotype with not as clear of an adipose/obesity phenotype. I actually think it is typically considered a NASH (Non-alcoholic Steatohepatitis) model. I don't think the manuscript includes any data regarding the physiology of these mice that you would expect in an obesity model: body weight, liver weight, blood glucose, etc.

You should include a description in the methods for how the survival studies were conducted. Were the mice just checked on once a day for death, or were there other endpoints for euthanasia, like severe weight loss?

The measurement of IFNG and TNF in tissue throughout the manuscript seemed inconsistent. For example, IFNG in Figure 3A is 0.05pg/g for Chow+LPS, and 0.15pg.g for HFCD+LPS. But in Figure 4H, Chow+LPS is 0.18pg/g and HFCD+LPS is 0.18pg/g, so there is no effect of HFCD in the IgG controls. Also, in Figure 4I and 4J, the TNF values are dramatically different for the controls (0.1 vs 1pg/g).

You can't conclude that CD4+ and CD8+ T cells or monocytes don't play a role in liver damage from your data, because you did not measure liver damage, only mortality. I understand using mortality as an endpoint, but without ALT/AST measurements or histology, it's hard to say what exactly happened in the livers.

I'm not sure the authors can conclude that neutrophils expressing PD-L1 live longer in the hepatic environment from an in vitro experiment. I think this is an interesting result in terms of crosstalk between these two cell types, but I'm not sure that in vivo the neutrophils would live longer.

Reviewer #3 (Public review):

Summary:

The authors investigated how non-alcoholic fatty liver disease (NAFLD) influences liver damage during endotoxemia (a condition characterized by elevated endotoxins, like lipopolysaccharide or LPS, in the bloodstream) using a mouse model. Mice with NAFLD were given a moderate dose of LPS, which intensified liver inflammation and mortality compared to controls. The study concludes that targeting neutrophil activity and TNF-α signaling could be a promising approach to reducing excessive inflammation and liver injury in NAFLD patients experiencing endotoxemia. This can have important implications for the treatment but I think the manuscript requires revisions.

Strengths:

(1) The study presents both in vivo and ex vivo assay and results to support their hypothesis.

(2) Several cell types and their interaction with each other have been analyzed.

(3) The authors made use of the publicly available databases.

Weaknesses:

(1) Some figures contradict each other.

(2) Some of the cause-and-effect presentations need additional experiments and different approaches to be proven correct.

(3) Candidate/mechanism selection strategies are not very clear.

eLife Assessment

This study reports that activation of TFEB promotes lysosomal exocytosis and clearance of cholesterol from lysosomes, the strength of evidence for which is solid and considered valuable in the context of Niemann-Pick Disease Type C. However, beyond this aspect of the study, the reviewers found the strength of the evidence to be incomplete. The manuscript also needs careful editing to improve readability.

Reviewer #1 (Public review):

Summary:

The authors are trying to determine if SFN treatment results in dephosphorylation of TFEB, subsequent activation of autophagy-related genes, exocytosis of lysosomes, and reduction in lysosomal cholesterol levels in models of NPC disease.

Strengths:

(1) Clear evidence that SFN results in translocation of TFEB to the nucleus.

(2) In vivo data demonstrating that SFN can rescue Purkinje neuron number and weight in NPC1-/- animals.

Weaknesses:

(1) Lack of molecular details regarding how SFN results in dephosphorylation of TFEB leading to activation of the aforementioned pathways. Currently, datasets represent correlations.

(2) Based on the manuscript narrative, discussion, and data it is unclear exactly how steady-state cholesterol would change in models of NPC disease following SFN treatment. Yes, there is good evidence that lysosomal flux to (and presumably across) the plasma membrane increases with SFN. However, lysosomal biogenesis genes also seem to be increasing. Given that NPC inhibition, NPC1 knockout, or NPC1 disease mutations are constitutively present and the cell models of NPC disease contain lysosomes (even with SFN) how could a simple increase in lysosomal flux decrease cholesterol levels? It would seem important to quantify the number of lysosomes per cell in each condition to begin to disentangle differences in steady state number of lysosomes, number of new lysosomes, and number of lysosomes being exocytosed.

(3) Lack of evidence supporting the authors' premise that "SFN could be a good therapeutic candidate for neuropathology in NPC disease".

Reviewer #2 (Public review):

Summary:

This study presents a valuable finding that the activation of TFEB by sulforaphane (SFN) could promote lysosomal exocytosis and biogenesis in NPC, suggesting a potential mechanism by SFN for the removal of cholesterol accumulation, which may contribute to the development of new therapeutic approaches for NPC treatment.

Strengths:

The cell-based assays are convincing, utilizing appropriate and validated methodologies to support the conclusion that SFN facilitates the removal of lysosomal cholesterol via TFEB activation.

Weaknesses:

(1) The in vivo experiments demonstrate the therapeutic potential of SFN for NPC. A clear dose-response analysis would further strengthen the proposed therapeutic mechanism of SFN. Additional data supporting the activation of TFEB by SFN for cholesterol clearance in vivo would strengthen the overall impact of the study

(2) In Figure 4, the authors demonstrate increased lysosomal exocytosis and biogenesis by SFN in NPC cells. Including a TFEB-KO/KD in this assay would provide additional validation of whether these effects are TFEB-dependent.

(3) For lysosomal pH measurement, the combination of pHrodo-dex and CF-dex enables ratiometric pH measurement. However, the pKa of pHrodo red-dex (according to Invitrogen) is ~6.8, while lysosomal pH is typically around 4.7. This discrepancy may account for the lack of observed lysosomal pH changes between WT and U18666A-treated cells. Notably, previous studies (PMID: 28742019) have reported an increase in lysosomal pH in U18666A-treated cells.

(4) The authors are also encouraged to perform colocalization studies between CF-dex and a lysosomal marker, as some researchers may be concerned that NPC1 deficiency could reduce or block the trafficking of dextran along endocytosis.

(5) In vivo data supporting the activation of TFEB by SFN for cholesterol clearance would significantly enhance the impact of the study. For example, measuring whole-animal or brain cholesterol levels would provide stronger evidence of SFN's therapeutic potential.

Reviewer #3 (Public review):

Summary:

The authors demonstrate that activation of TFEB facilitates cholesterol clearance in cell models of Niemann-Pick type C (NPC). This is done through a variety of approaches including activation of TFEB by sulforaphane (SFN), a naturally occurring small-molecule TFEB agonist. SFN induces TFEB nuclear translocation and promotes lysosomal exocytosis. In an NPC mouse model, SFN dephosphorylates/activates TFEB in the brain and rescues the loss of Purkinje cells.

Strengths:

NPC is a severe disease and there is little in the way of treatment. The manuscript points towards some treatment options. However, the title, the title "Small-molecule activation of TFEB Alleviates Niemann-Pick Disease..." is far too strong and should be changed.

Weaknesses:

(1) The manuscript is extremely hard to read due to the writing; it needs careful editing for grammar and English.

(2) There are a number of important technical issues that need to be addressed.

(3) The TFEB influence on filipin staining in Figure 1A is somewhat subtle. In the mCherry alone panels there is a transfected cell with no filipin staining and the mCherry-TFEBS211A cells still show some filipin staining.

(4) Figure 1C is impressive for the upregulation of filipin with U18666A treatment. However, SFN is used at 15 microM. This must be hitting multiple pathways. Vauzour et al (PMID: 20166144) use SFN at 10 nM to 1microM. Other manuscripts use it in the low microM range. The authors should repeat at least some key experiments using SFN at a range of concentrations from perhaps 100 nM to 5 microM. The use of 15 microM throughout is an overall concern.

Author Response:

Thank you for your interest in our paper. We would also like to thank the anonymous reviewers for their critical and constructive comments. Although the reviewers found our work interesting, they raised several important concerns about our study. To address these concerns, mostly we will perform new experiments as following.

1. Examine whether antioxidant-NAC can block SFN-induced TFEB-nuclear translocation in NPC cells;

2. Examine whether calcineurin inhibitor (FK506+CsA) or Ca 2+ inhibitor (Bapta-AM) can block SFN-induced TFEB-nuclear translocation in NPC cells.

3. Investigate whether cholesterol was cleared by activation of TFEB by SFN in vivo tissues.

4. Investigate whether SFN-evoked the lysosomal exocytosis is TFEB-dependent by using TFEB-KO cells.

5. Examine the effect of NPC1 deficiency on dextran trafficking by studying the localization of CF- dex and Lamp1.

6. Perform cytotoxicity experiments to examine whether SFN used in this study is cytotoxic in various cell lines

In addition, according to the reviewers’ suggestions, we will make clarifications and corrections wherever appropriate in the manuscript. Below please find our point-by-point responses and plans to the reviewers’ comments.

Reviewer #1 (Public review):

Summary:

The authors are trying to determine if SFN treatment results in dephosphorylation of TFEB, subsequent activation of autophagy-related genes, exocytosis of lysosomes, and reduction in lysosomal cholesterol levels in models of NPC disease.

Strengths:

(1) Clear evidence that SFN results in translocation of TFEB to the nucleus.

(2) In vivo data demonstrating that SFN can rescue Purkinje neuron number and weight in NPC1-/- animals.

Thank you for the support!

Weaknesses:

(1) Lack of molecular details regarding how SFN results in dephosphorylation of TFEB leading to activation of the aforementioned pathways. Currently, datasets represent correlations.

Thank you for this constructive comment. The reviewer is right that in this manuscript the molecular mechanism of SFN-activated TFEB has not been discussed in details. Because previously we have shown that SFN induces TFEB nuclear translocation via a Ca 2+ - dependent but MTOR (mechanistic target of rapamycin kinase)-independent mechanism through a moderate increase in reactive oxygen species (ROS). And calcineurin-mediated TFEB dephosphorylation underlies SFN-induced TFEB activation. These data have been published in 2021 autophagy (Li, Shao et al. 2021) . Therefore, in this study we did not mention this part. We will add the molecular mechanism of TFEB activation by SFN in the discussion part. And to further confirm this mechanism in NPC cells, we will also perform experiments including: 1) examine whether antioxidant-NAC can block SFN-induced TFEB-nuclear translocation in NPC cells; 2) examine whether calcineurin inhibitor (FK506+CsA) can block SFN-induced TFEB-nuclear translocation in NPC cells.

(2) Based on the manuscript narrative, discussion, and data it is unclear exactly how steady-state cholesterol would change in models of NPC disease following SFN treatment. Yes, there is good evidence that lysosomal flux to (and presumably across) the plasma membrane increases with SFN. However, lysosomal biogenesis genes also seem to be increasing. Given that NPC inhibition, NPC1 knockout, or NPC1 disease mutations are constitutively present and the cell models of NPC disease contain lysosomes (even with SFN) how could a simple increase in lysosomal flux decrease cholesterol levels? It would seem important to quantify the number of lysosomes per cell in each condition to begin to disentangle differences in steady state number of lysosomes, number of new lysosomes, and number of lysosomes being exocytosed.

Thank you for the suggestion. It is important to define the three states 1) original number of lysosomes, 2) number of new lysosomes, and 3) number of lysosomes being exocytosis. However, we have checked literature, so far it seems that there is no good method that could clearly differentiate the three states of lysosomes.

(3) Lack of evidence supporting the authors' premise that "SFN could be a good therapeutic candidate for neuropathology in NPC disease".

Suggestion was taken! We will investigate whether cholesterol was reduced by activation of TFEB by SFN in vivo to strength the point that SFN could be a potential therapeutic compound for NPC treatment. And to avoid confusion, we have removed this sentence.

Reviewer #2 (Public review):

Summary:

This study presents a valuable finding that the activation of TFEB by sulforaphane (SFN) could promote lysosomal exocytosis and biogenesis in NPC, suggesting a potential mechanism by SFN for the removal of cholesterol accumulation, which may contribute to the development of new therapeutic approaches for NPC treatment.

Strengths:

The cell-based assays are convincing, utilizing appropriate and validated methodologies to support the conclusion that SFN facilitates the removal of lysosomal cholesterol via TFEB activation.

Weaknesses:

(1) The in vivo experiments demonstrate the therapeutic potential of SFN for NPC. A clear dose-response analysis would further strengthen the proposed therapeutic mechanism of SFN. Additional data supporting the activation of TFEB by SFN for cholesterol clearance in vivo would strengthen the overall impact of the study

We understand the reviewer’s point. We examined two doses of SFN-30 and 50mg/kg. As shown in Fig.6, SFN (50mg/kg), but not 30mg/kg prevents a degree of Purkinje cell loss in the lobule IV/V of cerebellum, suggesting a dose-correlated preventive effect of SFN. In vivo experiments with higher concentrations of SFN and optimized dosage form of SFN were planned in the future study, but will not be included in this study.

We will investigate whether cholesterol was cleared by activation of TFEB by SFN in vivo.

(2) In Figure 4, the authors demonstrate increased lysosomal exocytosis and biogenesis by SFN in NPC cells. Including a TFEB-KO/KD in this assay would provide additional validation of whether these effects are TFEB-dependent.

Thank you for this valuable suggestion. We will investigate whether SFN-evoked the lysosomal exocytosis is TFEB-dependent by using TFEB-KO cells.

(3) For lysosomal pH measurement, the combination of pHrodo-dex and CF-dex enables ratiometric pH measurement. However, the pKa of pHrodo red-dex (according to Invitrogen) is ~6.8, while lysosomal pH is typically around 4.7. This discrepancy may account for the lack of observed lysosomal pH changes between WT and U18666A-treated cells. Notably, previous studies (PMID: 28742019) have reported an increase in lysosomal pH in U18666A-treated cells.

We understand the reviewer’s point. But we used pHrodo™ Green-Dextran (P35368, Invitrogen), but not pHrodo red-dex to measure the lysosomal luminal acidity. According to the product information from Invitrogen, pHrodo Green-dex conjugates are non-fluorescent at neural pH, but fluorescence bright green at acidic pH ranges 4-9, such as those in endosomes and lysosomes. Therefore, pHrodo Green-dex can be used to monitor the acidity of lysosome (Hu, Li et al. 2022) . We also used LysoTracker Red DND-99 (Thermo Scientific, L7528) to measure lysosomal pH (Fig. 4G, H), which is consistent with results of pHrodo Green/CF measurement. Overall, in our hands, we have not detected pH change of lysosomes in U18666A-treated NPC1 cell models.

(4) The authors are also encouraged to perform colocalization studies between CF-dex and a lysosomal marker, as some researchers may be concerned that NPC1 deficiency could reduce or block the trafficking of dextran along endocytosis.

Suggestion was taken! We will examine the effect of NPC1 deficiency on dextran trafficking by studying the localization of CF-dex and Lamp1.

(5) In vivo data supporting the activation of TFEB by SFN for cholesterol clearance would significantly enhance the impact of the study. For example, measuring whole-animal or brain cholesterol levels would provide stronger evidence of SFN's therapeutic potential.

We really appreciate the reviewer’s suggestions. We will investigate whether cholesterol was cleared by activation of TFEB by SFN in vivo.

Reviewer #3 (Public review):

Summary:

The authors demonstrate that activation of TFEB facilitates cholesterol clearance in cell models of Niemann-Pick type C (NPC). This is done through a variety of approaches including activation of TFEB by sulforaphane (SFN), a naturally occurring small-molecule TFEB agonist. SFN induces TFEB nuclear translocation and promotes lysosomal exocytosis. In an NPC mouse model, SFN dephosphorylates/activates TFEB in the brain and rescues the loss of Purkinje cells.

Strengths:

NPC is a severe disease and there is little in the way of treatment. The manuscript points towards some treatment options. However, the title, the title "Small-molecule activation of TFEB Alleviates Niemann-Pick Disease..." is far too strong and should be changed.

Weaknesses:

(1) The manuscript is extremely hard to read due to the writing; it needs careful editing for grammar and English.

We will thoroughly check grammar to improve the manuscript.

(2) There are a number of important technical issues that need to be addressed.

We will address the technical issues mentioned in the following.

(3) The TFEB influence on filipin staining in Figure 1A is somewhat subtle. In the mCherry alone panels there is a transfected cell with no filipin staining and the mCherry-TFEBS211A cells still show some filipin staining.

We understand the reviewer’s point. We will investigate whether cholesterol is cleared by activation of TFEB by SFN in vivo.

(4) Figure 1C is impressive for the upregulation of filipin with U18666A treatment. However, SFN is used at 15 microM. This must be hitting multiple pathways. Vauzour et al (PMID: 20166144) use SFN at 10 nM to 1microM. Other manuscripts use it in the low microM range. The authors should repeat at least some key experiments using SFN at a range of concentrations from perhaps 100 nM to 5 microM. The use of 15 microM throughout is an overall concern.

We understand the reviewer’s point. See RESPONSE #1, previously we have shown that SFN (10–15 μM, 2–9 h) induces robust TFEB nuclear translocation in a dose- and time-dependent manner in HeLa GFP-TFEB stable cells as well as in other human cell lines without cytotoxicity (Li, Shao et al. 2021) . According to previous results, in this study, we chose SFN (15 μM) to examine its effect on cholesterol clearance. We will add the information in the discussion part. In this study, we will perform dose-response TFEB nuclear translocation in NPC model cells as well as cytotoxicity experiments to examine whether the concentrations of SFN used in various cell lines are toxic.

References:

Hu, M. Q., P. Li, C. Wang, X. H. Feng, Q. Geng, W. Chen, M. Marthi, W. L. Zhang, C. L. Gao, W. Reid, J. Swanson, W. L. Du, R. Hume and H. X. Xu (2022). "Parkinson's disease-risk protein TMEM175 is a proton-activated proton channel in lysosomes.” Cell 185(13): 2292-+.

Li, D., R. Shao, N. Wang, N. Zhou, K. Du, J. Shi, Y. Wang, Z. Zhao, X. Ye, X. Zhang and H. Xu (2021). “Sulforaphane Activates a lysosome-dependent transcriptional program to mitigate oxidative stress.” Autophagy 17(4): 872-887.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The work from Petazzi et al. aimed at identifying novel factors supporting the differentiation of human hematopoietic progenitors from induced pluripotent stem cells (iPSCs). The authors developed an inducible CRISPR-mediated activation strategy (iCRISPRa) to test the impact of newly identified candidate factors on the generation of hematopoietic progenitors in vitro. They first compared previously published transcriptomic data of iPSCderived hemato-endothelial populations with cells isolated ex vivo from the aorta-gonadmesonephros (AGM) region of the human embryo and they identified 9 transcription factors expressed in the aortic hemogenic endothelium that were poorly expressed in the in vitro differentiated cells. They then tested the activation of these candidate factors in an iPSCbased culture system supporting the differentiation of hematopoietic progenitors in vitro. They found that the IGF binding protein 2 (IGFBP2) was the most upregulated gene in arterial endothelium after activation and they demonstrated that IGFBP2 promotes the generation of functional hematopoietic progenitors in vitro.

Strengths:

The authors developed an extremely useful doxycycline-inducible system to activate the expression of specific candidate genes in human iPSC. This approach allows us to simultaneously test the impact of 9 different transcription factors on in vitro differentiation of hematopoietic cells, and the system appears to be very versatile and applicable to a broad variety of studies.

The system was extensively validated for the expression of 1 transcription factor (RUNX1) in both HeLa cells and human iPSC, and a detailed characterization of this test experiment was provided.

The authors exhaustively demonstrated the role of IGFBP2 in promoting the generation of functional hematopoietic progenitors in vitro from iPSCs. Even though the use of IGFBP2interacting proteins IGF1 and IGF2 have been previously reported in human iPSC-derived hematopoietic differentiation in vitro (Ditadi and Sturgeon, Methods 2016; Ng et al., Nature Biotechnology 2016), and IGFBP-2 itself has been shown to promote adult HSC expansion ex vivo (Zhang et al., Blood 2008), its role on supporting in vitro hematopoiesis was demonstrated here for the first time.

Weaknesses:

Although the authors performed a very thorough characterization of the system in proof-ofprinciple experiments activating a single transcription factor, the data provided when 9 independent factors were used is not sufficient to fully validate the experimental strategy. Indeed, in the current version of the manuscript, it is not clear whether the results presented in both the scRNAseq analysis and the functional assays are the consequence of the simultaneous activation of all 9 TF or just a subset of them. This is essential to establish whether all the proposed factors play a role during embryonic hematopoiesis, and a more complete analysis of the scRNAseq dataset could help clarify this aspect.

Similarly, the data presented in the manuscript are not sufficient to clarify at what stage of the endothelial-to-hematopoietic transition (EHT) the TF activation has an impact. Indeed, even though the overall increase of functional hematopoietic progenitors is fully demonstrated, the assays proposed in the manuscript do not clarify whether this is due to a specific effect at the endothelial level or to an increased proliferation rate of the generated hematopoietic progenitors. Similar conclusions can be applied to the functional validation of IGFBP2 in vitro.

The overall conclusions are sometimes vague and not always supported by the data. For instance, the authors state that the CRISPR activation strategy resulted in transcriptional remodeling and a steer in cell identity, but they do not specify which cell types are involved and at what level of the EHT process this is happening. In the discussion, the authors also claim that they provided evidence to support that RUNX1T1 could regulate IGFBP2 expression. However, this is exclusively based on the enrichment of RUNX1T1 gRNA in cells expressing higher levels of IGFBP2 and it does not demonstrate any direct or indirect association of the two factors.

We thank the reviewer for the positive comments about the importance of our work and have now addressed the points raised as weaknesses by performing additional analysis and experiments, adding a new schematic of the mechanism, and rewording our claims.

We have clarified the different effects mediated by the activation and the IGFBP2 addition in a summary section at the end of the results and added Figure 6, showing this in visual form. We have also clearly stated the limitations related to the correlation between RUNX1T1 and IGFBP2 in the discussion and toned down our claims regarding this throughout the entire paper. We have also reworded the text to clarify the specific cell types identified in the sequencing data that we refer to.

Reviewer #2 (Public Review):

To enable robust production of hematopoietic progenitors in-vitro, Petazzi et al examined the role of transcription factors in the arterial hemogenic endothelium. They use IGFBP2 as a candidate gene to increase the directed differentiation of iPSCs into hematopoietic progenitors. They have established a novel induced-CRISPR mediated activation strategy to drive the expression of multiple endogenous transcription factors and show enhanced production of hematopoietic progenitors through expansion of the arterial endothelial cells. Further, upregulation of IGFBP2 in the arterial cells facilitates the metabolic switch from glycolysis to oxidative phosphorylation, inducing hematopoietic differentiation. While the overall study and resources generated are good, assertions in the manuscript are not entirely supported by the experimental data and some claims need further experimental validation.

We thank the reviewer for the positive comments, and we have provided new data and analysis to make sure that all our assertations are clearly supported and also reworded those where limitations were identified by the reviewers.

Recommendations for the authors:

Reviewing Editor (Recommendations For The Authors):

The assessment could change from "incomplete" to "solid" if the authors: i) improve data analysis (for both scRNAseq and functional assays) by providing additional information that could strengthen their conclusions, as suggested in the specific comments by both reviewers; ii) either provide new functional evidence supporting their mechanistic conclusion or alternatively tone down the claims that are not fully supported by data and acknowledge the limitations raised by reviewers in the discussion; (iii) the issue of paracrine signaling to expand only hematopoietic progenitors needs to be addressed.

We have now improved the data analysis and provided additional functional tests to strengthen our conclusions and toned down those that were identified by the reviewers as not supported enough and included a discussion on these limitations. We have also reworded the section about the paracrine signaling throughout the paper.

Reviewer #1 (Recommendations For The Authors):

Figure 1 contains exclusively published data. It might be more appropriate to use it as a supplementary figure or as part of a more exhaustive figure (maybe combining Figures 1 and 2 together?).

Figure 1 contained novel bioinformatic analyses that represent the base of our research and it has a different content and focus to figure 2, which is already a large figure. We therefore believe it is better to keep it as a separate figure, containing a new panel now too. 

It seems there is an issue with Figure S3 labelling:

• In line 112, Figure S2A-B does not display genomic PCR and sequencing results;

• In line 123, Figure S3D-E does not show viability and proliferation data;

• In line 127, Figure S3G does not show mCherry expression in response to DOX;

We apologies for the confusion with the numbers, we have now correctly labelled the figures.

It would be more informative to include gates and frequency on flow cytometry plots in Figure S3, to be able to evaluate the extent of the reduction in mCherry expression.

We have now included the gating and frequency of mCherry-expressing cells in Supplementary Figure 3D.

It is not clear from the text and figures whether the SB treatment was maintained throughout the hematopoietic differentiation protocol (line 122):

• If so, it would be important to confirm that HDAC treatment does not affect EHT cultures

• If not, can the authors provide some evidence that transgene silencing is not occurring during hematopoietic differentiation?

We have clarified that we decided to treat the cells with SB exclusively in maintenance condihons because HDACs have been shown to be essenhal for the EHT (lines 138-142). We have now also included addihonal data showing the high expression of the mCherry tag reporhng the iSAM expression on day 8 (Supplementary Figure 4F).

Can the authors provide a simple diagram summarizing the experimental strategy for each differentiation experiment in the respective supplementary figure? For instance, at what stage of the protocol was DOX added in Figure 3? Or at what stage IGFBP2 was added in Figure 5? It would be a very useful addition to the interpretation of the results.

We have now included three schemahcs for all the experiments in the manuscript in supplementary figure 4 A-C.

In Figure 3, the authors should provide more detailed information about the data filtering of the scRNAseq experiment, and more specifically:

• How many cells were included in the analysis for each library after QC and filtering?

• How "cells in which the gRNAs expression was detected" were selected? Do they include only cells showing expression of gRNAs for all 9 TF?

This informahon is now included in the method sechon lines 773-781; the detailed code is available on the GitHub link provided in the same sechon. We have filtered the cells expressing one gRNA for the non-targehng gRNA (iSAM_NT) control and more than one for the iSAM_AGM sample. 

In Figure 3A, it is not clear whether the expression of the 9 factors is consistently detected in all cells or just a subset of them, and the heatmap in Figure 3A does not provide this information. It would be more accurate to provide expression on a per-cell basis, for instance, as a violin plot displaying single dots representing each cell. 

We have now included this violin plot in Supplementary Figure 4G as requested. However, this visualisation is difficult to interpret because some of the target genes’ expression seems variable in both experimental and control conditions. We had envisaged that this could have been the case and so this is why we had included the three different controls.  For this reason we chose to show the normalised expression which takes all the different variables into account (Figure 3A). 

In Figure 3B-C, it seems that clusters EHT1 and EHT2 do not express endothelial markers anymore. Are these fully differentiated hematopoietic cells rather than cells undergoing EHT? In general, it would be quite important to provide evidence of expressed marker genes characterizing each cluster (eg. heatmap summarizing top DEG in the supplementary figure?). 

We have now provided a spreadsheet containing the clusters’ markers that we used in

Supplementary Table 1) a heatmap in Figure 3E. Furthermor,e we have now edited Figure 3C to include Pan Endothelial markers (PECAM1 and CDH5). These data show that the EHT1 and EHT2 cluster both express endothelial markers but are progressively downregulated as expected during endothelial to hematopoietic transition. We have also included and discussed this in the manuscript lines 192-195 and a schematic for the mechanism in Figure 6.

In Figure 3E, displaying the proportion of clusters within each sample/library would be a more accurate way of comparing the cell types present in each library (removing potential bias introduced by loading different numbers of cells in each sample).

We have now included the requested data in Supplementary Figure 4I and it confirms again the expansion of arterial cells in the activated cells.    

In Figure 3G, by plating 20,000 total CD34+, the assay does not account for potential differences in sample composition. It is then hard to discriminate between the increased number of progenitors in the input or an enhanced ability of HE to undergo EHT. This is an important aspect to consider to precisely identify at what level the activation of the 9 factors is acting. A proper quantification of flow cytometry data summarizing the % of progenitors, arterial cells, etc. would be useful to interpret these results.

Lines 204-205 reworded. We are very much aware of the fact that the CD34+ cell population consists of a range of cells across the EHT process and this is precisely why we carried out this single cell sequencing analyses.  We purposely tested the effect of the observed changes in composition by colony assays

In Figure 3G, it seems that NT cells w/o DOX have very little CFU potential (if any). Can the authors provide an explanation for this?

We think that the limited CFU potential is due to the extensive genetic manipulation and selection that the cells underwent for the derivation of all the iSAM lines but this did not impede us from observing an effect of gene activation on CFU numbers. This is one of the primary reasons that we then validated our overall findings using the parental iPSC line in control condition and with the addition of IGFBP2. We show that the parental iPSC line gives rise to hematopoietic progenitor, both immunophenotypically (Figure 4D) and functionally, at expected levels (Figure 4B left column).

Figure 4A shows an upregulation of IGFBP2 in arterial cells as a result of TF activation. However, from the data presented here, it is not possible to evaluate whether this is specific to the arterial cluster, or it is a common effect shared by all cell types regardless of their identity. 

Data has now been included in Supplementary Figure 4H, which shows that all the cells show an increase in IGFBP2, but arterial cells show the highest increase. We have now edited the text to reflect this, in lines 228-230.

In Figure 5A-B only a minority of arterial cells express RUNX1 in response to IGFBP2 treatment. Is this sufficient to explain the very significant increase in the generation of functional hematopoietic progenitors described in Figure 4? Quantification and statistical analysis of RUNX1 upregulation would strengthen this conclusion.

We have now provided the statistical analysis showing significant upregulation of RUNX1 upon IGFBP2 addition. The p values are now provided in the figure 5 legend.

In Figure 5 the authors conclude that IGFBP2 remodels the metabolic profile of endothelial cells. However, it is not clear which cell types and clusters were included in the analysis of Figure 5C-G. Is the switch from Glycolysis to Oxidative Phosphorylation specific to endothelial cells? Or it is a more general effect on the entire culture, including hematopoietic cells? 

We based this conclusion on the fact that the single-cell RNAseq allows to verify that the metabolic differences are obtained in the endothelial cells. Given that we sorted the adherent cells, the majority of these are endothelial cells as shown in Figure 5A. The Seahorse pipeline includes a number of washing steps resulting in the analyses being performed on the adherent compartment which we know consists primarily of endothelial cells. We cannot exclude some contamination from non-endothelial cells but we highlight to this reviewer that the initial observation of the metabolic changes was identified in endothelial cells in the single cell sequencing data. Taken together, we believe that this implies that metabolic changes are specific to this population. We have clarified this in the line 317.

In the discussion, the authors conclude that they "provide evidence to support the hypothesis that RUNX1T1 could regulate IGFBP2 expression". To further support this conclusion, the authors could provide a correlation analysis of the expression of the two genes in the cell type of interest. 

Following the observation of the IGFBP2 high expression across clusters, we have now reworded this sentence in lines 382-385  We have tried to perform the correlation analysis but we believe this not to be appropriate due to the detection level of the gRNA, we have now included this as a limitation point in the discussion lines 416-427, and also toned down the conclusion we did draw about RUNX1T1 throughout the whole manuscript.

As mentioned by the authors, IGFBP2 binds IGF1 and IGF2 modulating their function. Both IGF1 (http://dx.doi.org/10.1016/j.ymeth.2015.10.001) and IGF2 (doi:10.1038/nbt.3702) have been used in iPSC differentiation into definitive hematopoietic cells. It would be relevant to discuss/reference this in the discussion.

We have now included the suggested reference in the section where we discuss the role of IGFBP2 in binding IGF1 and IGF2.

Reviewer #2 (Recommendations For The Authors):

(1) Figure 1 compares the transcriptome of human AGM and in-vitro derived hemogenic endothelial cells (HECs). It is not clear why only the genes downregulated in the latter were chosen. Are there any significantly upregulated genes, knockdown/knockout which could also serve a similar purpose? Single-cell transcriptome database analysis is very preliminary. A detailed panel with differences in cluster properties of HECs between the two systems should be provided. A heatmap of all differentially expressed genes between the two samples must be generated, along with a logical explanation for choosing the given set of genes. 

We have now included another panel in figure 1 to better clarify the logic behind the strategy used to identify our target genes (Figure 1A).

(2) Figure 2 - a panel describing the workflow of gRNA design and targeting for the 9 candidate genes, along with lentiviral packaging and transduction would make it easier to follow. 

We have now included three schematics for all the experiments in the manuscript in supplementary figure 4 A-C. 

(3) Figure 3- to assess the effect of arterial cell expansion on the emergence of hematopoietic progenitors, CD34+ Dll4+ cells should be sorted for OP9 co-culture assay.

Using only CD34+ cells does not answer the question raised. Also, the CFU assay performed does not fully support the claim of enhanced hematopoietic differentiation since only CFU-E and CFU-GM colonies are increased in Dox-treated samples, with no effect on other colony types. OP9 co-culture assay with these cells would be required to strengthen this claim. 

We wanted to clarify that the effect on the methylcellulose coming from the activated cells was not limited to CFU-E, as the reviewer reported; instead, it also affected CFU-GM and CFU-M. 

We have now performed additional experiments where we sorted the CD34+ compartment into DLL4- and DLL4+ in Supplementary Figure 5D-E, which we discussed in lines 250-258. 

(4) In Figure 3F, there appears to be a lot of variation in the DLL4% fold change values for

DOX treated iSAM_AGM sample, which weakens the claim of increased arterial expansion.

Can the authors explain the probable reason? It is suggested that the two other controls (iSAM_+DOX and iSAM_-DOX) should be included in this analysis. It is imperative to also show % populations rather than just fold change to gain confidence.

We agree that there is a lot of variability. That is because differentiation happens in 3D in embryoid bodies, which contain many different cell types that differentiate in different proportions across independent experiments. We have now included the raw data in Supplementary Figure 4 D, with additional statistical analysis to show the expansion of arterial cells including also the suggested additional controls.

(5) How does activation of these target genes cause increased arterialization? Is the emergence of non-HE populations suppressed? Or is it specific to the HE? The data on this should be clarified and also discussed. ANTO/Lesley text

We have provided additional data clarifying the connection between increased arterialisation and hemogenic potential. We showed that the activation induces increased arterialisation and that IGFBP2 acts by supporting the acquisition of hemogenic potential. We have discussed this in lines 326-348 and provided a new figure to explain this in detail (figure 6)

(6) Considering that IGFBP2 was chosen from the activated target gene(s) cluster, can the authors explain why the reduced CFU-M phenomenon observed in Figure 3G does not appear in the MethoCult assay for IGFBP2 treated cells (Figure 4B)?

The difference could be explained by the fact that in Figure 3G, the cells underwent activation of multiple genes, while in Figure 4B, they were only exposed to IGFBP2. Our results show that IGFBP2 could at least partially explain the phenotype that we see with the activation, but we believe that during the activation experiments, there might be other signals available that might not be induced by IGFBP2 alone. We have also added a summary section and a figure to clarify the different mechanisms of action of the gene activation and IGFBP2.

(7) Figure 4- while the experiments conducted support the role of IGFBP2 in increasing hematopoietic output, there is no experimental evidence to prove its function through paracrine signalling in HECs. The authors need to provide some evidence of how IGFBP2 supplementation specifically expands only the hematopoietic progenitors. Experimental strategies involving specifically targeting IGFBP2 in hemogenic/arterial endothelial cells are required to prove its cell type specific function. Additionally, assessing the in vivo functional potential of the hematopoietic cells generated in the presence of IGFBP2, by bone-marrow transplantation of CD34+ CD43+ cells, is essential. 

The role of IGFBP2 in the context of HSC production and expansion was not the topic of our research, and we have not claimed that IGFBP2  affects the long-term repopulating capacity of HSPCs. Therefore, we believe that the requested experiments are not required to support the specific claims that we do make. We have now provided more experiments and bioinformatic analysis that support the role of IGFBP2 in inducing the progression of EHT from arterial cells to hemogenic endothelium, and to avoid misunderstandings, we have toned down our claims by editing the text regarding its paracrine effect s. 

(8) Figure 4C-D -It is recommended to plot % populations along with fold change value. As this is a key finding, it is important to perform flow cytometry for additional hematopoietic markers- CD144, CD235a and CD41a to demonstrate whether this strategy can also expand erythroid-megakaryocyte progenitors. Telma

Figure 4C already shows the percentage values; we have now added the percentage for Figure 4D in SF5C. We have also performed additional analysis as requested and added the data obtained to Supplementary Figure 5D.

(9) In Figure 5, analysis showing the frequency of cells constituting different clusters, between untreated and IGFBP2-treated samples in the single-cell transcriptome analysis is essential. Additional experiments are required to validate the function of IGFBP2 through modulation of metabolic activity. Inhibition of oxidative phosphorylation in the IGFBP2treated cells should reduce the hematopoietic output. Authors should consider doing these experiments to provide a stronger mechanistic insight into IGFBP2-mediated regulation of hematopoietic emergence.

We have now included the requested cluster composition in Supplementary Figure 5F. We decided not to include further tests on the metabolic profile of IGFBP2 as we already discussed in other papers that showed, using selective inhibitors, that the EHT coincides with a glycol to OxPhos switch. 

(10) It is very striking to see that IGFBP2 supplementation changes the transcriptional profile of developing hematopoietic cells by increasing transcription of OXPHOS-related genes with concomitant reduction of glycolytic signatures, particularly at Day 13. However, the mitochondrial ATP rate measurements do not seem convincing. The bioenergetic profiles show that when mitochondrial inhibitors are added, both groups exhibit decreased OCR values and, on the other hand, higher ECAR. This indicates that both groups have the capability to utilize OXPHOS or glycolysis and may only differ in their basal respiration rates.

Differences in proliferation rate can cause basal respiration to change. There is no information on how the bioenergetic profile was normalized (cell no./protein amount). Given that IGFBP2 has been shown to increase proliferation, it is very likely that the cells treated with IGFBP2 proliferated faster and therefore have higher OCR. The data needs to be normalized appropriately to negate this possibility.

We have previously tested whether IGFBP2 causes an increase in proliferation by analysing the cell cycle of cells treated with it, as we initially thought this could be a mechanism of action. We have now provided the quantification of the cell cycle in the cells treated with IGFBP2, showing no effect was observed in cell cycle Supplementary Figure 4E. Following this analysis, we decided to plate the same number of cells and test their density under the microscope before running the experiment; each experiment was done in triplicate for each condition. We have now added this info to the method sections lines 806-813.  We did not comment on the basal difference, which we agree might be due to several factors, but we only compared the difference in response to the inhibitors, which isn’t affected by the basal level but exclusively by their D values. We have also included the formulas used to calculate the ATP production rate.

Overall, it appears that IGFBP2 does not seem to primarily cause metabolic changes, but simply accelerates the metabolic dependency on OXPHOS. Hence, the term 'metabolic remodelling' must be avoided unless IGFBP2 depletion/loss of function analysis is shown.

We thank the reviewer for suggesting how to interpret the data about the dependency on OXPHOS. We have now changed the conclusions and claims about the effect of IGFBP2. We have also included a cell cycle analysis of the hematopoietic cells derived upon IGFBP2 addition to show that they don’t show differences in proliferation that could cause the increase in colony formation we observed. Regarding the assay, we have plated the same number of cells for each group to make sure we were comparing the same number of cells, which we also assessed in the microscope before the test, and we eliminated the suspension cells during the washes that preceded the measurement. The review is correct in indicating that there is a basal difference in the value of OCR and ECAR where the IGFBP2 is lower at the start and not higher, which would not conceal higher proliferation. Finally, the ATP production rate is calculated on the variation of OCR and ECAR upon the addition of inhibitors, which normalizes for the basal differences.

eLife Assessment

This study presents useful findings to inform and improve the in vitro differentiation of hematopoietic progenitor cells from human induced pluripotent stem cells. Relying on a well-characterised technical approach, the data analysis is overall solid and reasonably supports the main conclusions.

Reviewer #1 (Public Review):

Summary:

The work from Petazzi et al. aimed at identifying novel factors supporting the differentiation of human hematopoietic progenitors from induced-pluripotent stem cells (iPSCs). The authors developed an inducible CRISPR-mediated activation strategy (iCRISPRa) to test the impact of newly identified candidate factors on the generation of hematopoietic progenitors in vitro. They first compared previously published transcriptomic data of iPSC-derived hemato-endothelial populations with cells isolated ex vivo from the aorta-gonad-mesonephros (AGM) region of the human embryo and they identified 9 transcription factors expressed in the aortic hemogenic endothelium that were poorly expressed in the in vitro differentiated cells. They then tested the activation of these candidate factors in an iPSC-based culture system supporting the differentiation of hematopoietic progenitors in vitro. They found that the IGF binding protein 2 (IGFBP2) was the most upregulated gene in arterial endothelium after activation and they demonstrated that IGFBP2 promotes the generation of functional hematopoietic progenitors in vitro.

Strengths:

The authors developed a very useful doxycycline-inducible system to activate the expression of specific candidate genes in human iPSC. This approach allows us to simultaneously test the impact of 9 different transcription factors on in vitro differentiation of hematopoietic cells, and the system appears to be very versatile and applicable to a broad variety of studies. Using this approach, the authors exhaustively demonstrated the role of IGFBP2 in promoting the generation of functional hematopoietic progenitors in vitro from iPSCs.

Weaknesses:

The authors performed a very thorough characterization of the system in proof-of-principle experiments activating a single transcription factor. However, when 9 independent factors were used, it is not always clear whether the observed results were the consequence of the simultaneous activation of all 9 TF or just a subset of them.

Author response:

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

Reviewer #1 (Recommendations For The Authors):

Summary:

In this manuscript, the molecular mechanism of interaction of daptomycin (DAP) with bacterial membrane phospholipids has been explored by fluorescence and CD spectroscopy, mass spectrometry, and RP-HPLC. The mechanism of binding was found to be a two-step process. A fast reversible step of binding to the surface and a slow irreversible step of membrane insertion. Fluorescence-based titrations were performed and analysed to infer that daptomycin bound simultaneously two molecules of PG with nanomolar affinity in the presence of calcium. Conformational change but not membrane insertion was observed for DAP in the presence of cardiolipin and calcium.

Strengths:

The strength of the study is skillful execution of biophysical experiments, especially stoppedflow kinetics that capture the first surface binding event, and careful delineation of the stoichiometry.

Weaknesses:

The weakness of the study is that it does not add substantially to the previously known information and fails to provide additional molecular details. The current study provides incremental information on DAP-PG-calcium association but fails to capture the complex in mass spectrometry. The ITC and NMR studies with G3P are inconclusive. There are no structural models presented. Another aspect missing from the study is the reconciliation between PG in the monomer, micellar, and membrane forms.

Besides the two-stage process, another important finding in the current work is the stable complex that plays a critical role in the drug uptake both in vitro and in B. subtilis. This complex has been shown to be a stable species in HPLC and its binding stoichiometry and affinity have been quantitatively characterized. The complex may not be stable enough in gas phase to be detected in the MS analysis, which was designed to detect the phospholipid and Dap components, not the complex itself. The structural model of this complex is clearly proposed and presented in Figure 6. 

The NMR and ITC studies have a very clear conclusion that Dap has a weak interaction with the PG headgroup alone, which is unable to account for the Dap-PG interaction observed in the fluorescence studies. Thus, the whole PG molecule has to be involved in the interaction, leading to the discovery of the stable complex.  

Reviewer #2 (Recommendations For The Authors):

(1) I appreciate and agree with the comment that there are stages of daptomycin insertion, and these might involve the formation of different complexes with different binding partners (e.g. pre-insertion vs quaternary vs bactericidal). However, it seems like lipid II is an apparent participant in daptomycin membrane dynamics (Grein et al. Nature Communications 2020). It's not clear why this was excluded from analysis by the authors, or what basis there is for the discussion statement that the quaternary complex can shift into the bactericidal complex by exchanging 1 PG for lipid II. 

We agree that lipid II and other isoprenyl lipids may be involved in the uptake and insertion of daptomycin into membrane according to the results of the Nat. Comm. paper. However, these isoprenyl lipids are very small components of the membrane in comparison to PG and their contribution to the drug uptake is thus expected to be much less significant. Nonetheless, we included farnesyl pyrophosphate (FPP) as an analog of bactoprenol pyrophosphate (C55PP), which was reported to have the same promoting effect as lipid II in the previous study, in our study but found no promoting effect in the fluorescence assay (Fig. 2B). In addition, no complex was formed when FPP replaced PG in our preparation and analysis of the drug-lipid complex. In consideration of these negative results and the expected small contribution, other isoprenyl lipids or their analogs were not included in the study.

The statement of forming the proposed bactericidal complex from the identified complex is a speculation that is possible only when lipid II has a higher affinity for Dap than a PG ligand. To avoid confusion, we deleted the sentence’ in the revision. 

(2) The detailed examination of daptomycin dynamics, particularly on the millisecond scale, in this paper is ideal for characterizing the effect of lipid II on daptomycin insertion. It would be helpful to either include lipid II in some analyses (micelle binding, fluorescence shifts, CD) or at least address why it was excluded from the scope of this work.

As mentioned in the response to the first comment, we did not exclude isoprenyl lipids in our study but used some of their analogs in the fluorescence assay. Besides FPP mentioned above, we also tested geranyl pyrophosphate and geranyl monophosphate but obtained the same negative results. Lipid II was not directly used because it is one of the three isoprenyl lipids reported to have the same promoting effects in the Nat. Comm. paper and also because its preparation is not easy. Even if lipid II were different from other isoprenyl lipids in promoting membrane binding, its contribution is likely negligible at the reversible stage compared to the phospholipids because of its minuscule content in bacterial membrane. This is the main reason we did not use the isoprenyl lipids in the fast kinetic study (this stage only involves reversible binding, not insertion). 

(3) Grein et al. 2020 saw that PG did not have a strong effect on daptomycin interaction with membranes. I believe this discrepancy is more likely due to the complex physical parameters of supported bilayers versus micelles/vesicles or some other methodological variable, but if the authors have more insight on this, it would be valuable commentary in the discussion.

We totally agree that the discrepancy is likely due to the different conditions in the assays. It is hard to tell exactly what causes the difference. Thus, we did not attempt to comment on the cause of this difference in the discussion.

(4) Isolation of the daptomycin complex from B. subtilis cells clearly had different traces from the in vitro complex; is it possible that lipid II is present in the B. subtilis complex? If not, a time-course extraction could be useful to support the model that different complexes have different activities. Isolates from early-stage incubation with daptomycin may lack lipid II but isolates from longer incubations may have lipid II present as the complex shifts from insertion to bactericidal.

From the day we isolated the complex from B. subtilis, we have been looking for evidence for the previously proposed lipid complexes containing lipid II or other isoprenyl lipids but have not been successful. We did not see any sign of lipid II or other isoprenyl lipids in the MALDI or ESI mass spectroscopic data. The minute peaks in the HPLC traces are not the expected complexes in separate LC-MS analysis. However, this does not mean that such complexes are not present in the isolated PG-containing complex because: (1) the amount of such complexes may be too small to be detected due to the low content of the isoprenyl lipids; (2) the isoprenyl lipids, particularly lipid II, are not easily ionizable due to their size and unique structure for detection in mass spectrometry. 

We don’t think the drug treatment time is the reason for the failure in detecting lipid II or other isoprenyl lipids. In our reported experiment, the cells were treated with a very high dose of Dap for 2 hours before extraction. In a separate experiment done recently, we treated B. subtilis at 1/3 of the used dose under the same condition and found all treated cells were dead after 1 hour in a titration assay, consistent with the results from reported time-killing assays in the literature. From this result, the proposed bactericidal lipid-containing complex should have been formed in the treated cells used in our extraction and isolated along with the PG-containing complex. It was not detected likely due to the reasons discussed above. To avoid the interference of the PG-containing complex, a large amount of bacterial cells might have to be treated at a low dose to isolate enough amount of the lipid II-containing complex for identification. However, isolation or identification of the lipid II-containing complex is outside the scope of the current investigation and is therefore not pursued. 

(5) Part of the daptomycin mechanism of interacting with bacterial membranes involves the flipping of daptomycin from one leaflet to another. There was some mentioned work on the consistency of results between micelles and vesicles, but the dynamics or existence of a flipping complex in the bilayer system wasn't addressed at all in this paper.

The current investigation makes no attempt to solve all problems in the daptomycin mode of action and is limited to the uptake of the drug, up to the point when Dap is inserted into the membrane. Within this scope, flipping of the complex is not yet involved and is thus irrelevant to the study. How the complex is flipped and used to kill the bacteria is what should be investigated next.  

(6) The authors mention data with phosphatidylethanolamine in the text, but I could not find the data in the main or supplemental figures. I recommend including it in at least one of the figures.

It is much appreciated that this error is identified. The POPE data was lost when the graphic (Fig. 2B) was assembled in Adobe to create Figure 2. We re-draw the graphic and reassemble the figure to solve this problem. Fig. 2B has also been modified to use micromolar for the concentration of the lipids.

(7) Readability point: I'd suggest some consistency in the concentrations mentioned. Making the concentrations either all molar-based or all percentage-based would make comparison across figures easier.

As suggested, we have changed the % into micromolar concentrations in Fig. 2B and also in Fig. 3A. 

(8) The model figure is quite difficult to interpret, particularly the final stage of the tail unfolding. I recommend the authors use a zoomed-in inset for this stage, or at least simplify the diagram by removing the non-participating lipid structures. The figure legend for the model figure should also have a brief description of the events and what the arrows mean, particularly the POPS PG arrow in the final panel of the figure. I am assuming here the authors are implying that daptomycin can transiently interact with one lipid species and move to another, but the arrow here suggests that daptomycin is moving through the lipid headgroup space.

We really appreciate the suggestions. As suggested, we put an inset to show the preinsertion complex more clearly. In addition, we have removed the green arrows originally intended to show the re-organization/movement of the phospholipids. Moreover, the legend is changed to ‘Proposed mechanism for the two-phased uptake of Dap into bacterial membrane. In the first phase, Dap reversibly binds to negative phospholipids with a hidden tail in the headgroup region, where it combines with two PG molecules to form a pre-insertion complex. In the second phase, the hidden tail unfolds and irreversibly inserts into the membrane. The inset shows the headgroup of the pre-insertion complex with the broad arrow showing the direction for the unfolding of the hidden tail. The red dots denote Ca2+.’  

(9) The authors listed the Kd for daptomycin and 2 PG as 7.2 x 10-15 M2. Is this correct? This is an affinity in the femtomolar range.

Please note that this Kd is for the simultaneous binding of two PG molecules, not for the binding of a single ligand that we usually refer to. Assuming that each PG contributes equally to this interaction, the binding affinity for each ligand is then the squared root of 7.2 x 10-15 M2, which equals to 8.5 x 10-8 M. This is equivalent to a nanomolar affinity for PG and is a reasonably high affinity.

Reviewer #3 (Recommendations For The Authors):

(1) The authors reported an increase in daptomycin intensity with the increasing amount of negatively charged DMPG. A similar observation has been reported for GUVs, however, the authors did not refer to this paper in their manuscript: E. Krok, M. Stephan, R. Dimova, L. Piatkowski, Tunable biomimetic bacterial membranes from binary and ternary lipid mixtures and their application in antimicrobial testing, Biochim. Biophys. Acta - Biomembr. 1865 (2023) [1]. This paper is also consistent with the authors' observation that there is negligible fluorescence detected for the membranes composed of PC lipids upon exposure to the Dap treatment.

As suggested, this paper is cited as ref. 29 in the revision by adding the following sentence at the end of the section ‘Dependence of Dap uptake on phosphatidylglycerol.’: ‘PG-dependent increase of the steady-state fluorescence was also observed in giant unilamellar vesicles (GUVs).29’. The numbering is changed accordingly for the remaining references.  

(2) Please include the plot of the steady-state Kyn fluorescence vs the content of POPA (Figure 2C shows traces for DMPG, CL, and POPS). Both POPA and POPS lipids are negatively charged, however, POPS seems to interact with Dap, while POPA does not. In my opinion, this observation is really interesting and might deserve a more thorough discussion. The authors might want to describe what could be the mechanism behind this lipid-specific mode of binding.

As suggested, a plot is now added for POPA in Fig. 2C, which is basically a flat line without significant increase for the Kyn fluorescence. Indeed, the different effect of the negative phospholipids is very interesting, indicating that the reversible binding of Dap to the lipid surface is dependent not only on the Ca2+-mediated ionic interaction but also the structure of the headgroup. In other words, Dap recognizes the phospholipids at the surface binding stage. Considering this headgroup specificity, the last sentence in the second paragraph in “Discussion’ is changed from ‘In addition, due to the low lipid specificity, this reversible binding likely involves Ca2+-mediated ionic interaction between Dap and the phosphoryl moiety of the headgroups.’ to ‘In addition, due to the specificity for negative phospholipids (Fig. 2B and 2C), this reversible binding of Dap likely involves both a nonspecific Ca2+-mediated ionic interaction and a specific interaction with the remaining part of the headgroups.’

(3) The authors write that they propose a novel mechanism for the Ca2+-dependent insertion of Dap to the bacterial membrane, however, they rather ignored the already published findings and hypotheses regarding this process. In fact the role of Ca2+, as well as the proposed conformational changes of Dap, which allow its deeper insertion into the membrane are well known:

The role of Ca2+ ions in the mechanism of binding is actually three-fold: (i) neutralization of daptomycin charge [2], (iii) creating the connection between lipids and daptomycin and (iii) inducing two daptomycin conformational changes. It should be noted that the interactions between calcium ions and daptomycin are 2-3 orders of magnitude stronger than between daptomycin and PG lipids [3,4]. Thus, upon the addition of CaCl2 to the solution, the divalent cations of calcium bind preferentially to the daptomycin, rather than to the negatively charged PG lipids, which results in the decrease of daptomycin net negative charge but also leads to its first conformational change [4]. Upon binding between calcium ions and two aspartate residues, the area of the hydrophobic surface increases, which allows the daptomycin to interact with the negatively charged membrane. In the next step, Ca2+ acts as a bridge connecting daptomycin with the anionic lipids. This event leads to the second conformational change, which enables deeper insertion of daptomycin into the lipid membrane and enables its fluorescence [4]. The overall mechanism has a sequential character, where the binding of daptomycin-Ca2+ complex to the negatively charged PG (or CA) occurs at the end.

The authors should focus on emphasizing the novelty of their manuscript, keeping in mind the already published paper.

We agree with the comments on the three general roles of calcium ion in the Dap interaction with membrane. The current investigation does not ignore the previous findings, which involve many more works than mentioned above, but takes these findings as common knowledge. Actually, the role of calcium ion is not the focus of current work. Instead, the current work focuses on how the drug is taken up and inserted into the membrane in the presence of the ion and how its structure changes in this process. With the known roles of calcium ion in mind, we propose an uptake mechanism (Fig. 6) that shows no conflict with the common knowledge.

We would like to point out that the ‘deeper insertion into the membrane’ in the comment is different from the membrane insertion referred to in our manuscript. This ‘deeper insertion’ still remains in the reversible stage of binding to the membrane surface because all negative phospholipids can do this (causing a conformational change and fluorescence increase, as quantified in Fig.2C) but now we know that only PG can enable irreversible membrane insertion because of our work. In addition, the comment that calcium binding to daptomycin causes first conformational change is not supported by our finding that no conformational change is found for Dap in the presence of calcium in a lipid-free environment (Fig. 3B). One important aspect of novelty and contribution of our work is to clear up some of these ambiguities in the literature. Another contribution of our work is to demonstrate the formation of a stable complex between Dap and PG with a defined stoichiometry and its crucial role in the drug uptake. 

(4) One paragraph in the section "Ca2+- dependent interaction between Dap and DMPG" is devoted to a discussion of the formation of precipitate upon extraction of DMPG-containing micelles, exposed to Dap in the calcium-rich environment. Contrary, in the absence of Dap, no precipitate was detected. The authors did not provide any visual proof for their statement. Please include proper photographs in the supplementary information.

The precipitate formed upon extraction of the DMPG-containing micelles was too little to be visually identifiable but could be collected by centrifugation and detected by fluorescence or HPLC after dissolving in DMSO. For visualization, we show below the precipitate formed using higher amount of Dap and DMPG. The Dap-DMPG-Ca2+ complex (left tube) was formed by mixing 1 mM Dap, 2 mM DMPG and 1 mM Ca2+ and the control (right tube) was a mixture of 2 mM DMPG and 1 mM Ca2+. This is now added as Fig. S7 in the supplementary information (the index is modified accordingly) and cited in the main text.

(5) The authors wrote that it is not clear how many calcium ions are bound to Dap-2PG complex (page 11, Discussion section). There are already reports discussing this issue. I recommend citing the paper discussing that exactly two Ca2+ ions bind to a single Dap molecule: R. Taylor, K. Butt, B. Scott, T. Zhang, J.K. Muraih, E. Mintzer, S. Taylor, M. Palmer, Two successive calcium-dependent transitions mediate membrane binding and oligomerization of daptomycin and the related antibiotic A54145, Biochim. Biophys. Acta - Biomembr. 1858, (2016) 1999-2005 [5]

We were aware of the cited work that shows binding of two Ca2+ but also noted that there are more works showing one Ca2+ in the binding, such as the paper in [Ho, S. W., Jung, D., Calhoun, J. R., Lear, J. D., Okon, M., Scott, W. R. P., Hancock, R. E. W., & Straus, S. K. (2008), Effect of divalent cations on the structure of the antibiotic daptomycin. European Biophysics Journal, 37(4), 421–433.]. That was the reason we said ‘it is not clear how many calcium ions are bound to Dap-2PG complex’. Now, both papers are cited (as Ref. #33, 34) to support this statement.

(6) The authors wrote two contradictory statements:

-  PG cannot be found in mammalian cell membranes:

"Moreover, the complete dependence of the membrane insertion on PG also explains why Dap selectively attacks Gram-positive bacteria without affecting mammalian cells, because PG is present only in bacterial membrane but not in mammalian membrane. " (Page 10, Discussion section, last sentence of the first paragraph)

"However, Dap absorbed on bacterial surface is continuously inserted into the acyl layer via formation of complex with PG in a time scale of minutes, whereas no irreversible insertion of Dap occurs on mammalian membrane due to the absence of PG while the bound Dap is continuously released to the circulation as the drug is depleted by the bacteria." (Page 13, Discussion section)

-  PG in trace amounts is present in mammalian membranes:

"The proposed requirement of the pre-insertion quaternary complex increases the threshold of PG content for the membrane insertion to happen and thus makes it impossible on the surface of mammalian cells even if their plasma membrane contains a trace amount of PG." (Page 13, Discussion section).

In fact, phosphatidylglycerol comprises 1-2 mol% of the mammalian cell membranes. Please, correct this information, which in this form is misleading to the readers.

We appreciate the comments about the PG content in mammalian cells. Changes are made as listed below:

(1) p10, the sentence is changed to ‘Moreover, the complete dependence of the membrane insertion on PG also explains why Dap selectively attacks Gram-positive bacteria without affecting mammalian cells, because PG is a major phospholipid in bacterial membrane but is a minor component in mammalian membrane.’ 

(2) p13, the sentence is changed to ‘However, Dap absorbed on bacterial surface is continuously inserted into the acyl layer via formation of complex with PG in a time scale of minutes, whereas little irreversible insertion of Dap occurs on mammalian membrane due to the low content of PG while the bound Dap is continuously released to the circulation as the drug is depleted by the bacteria.’

(3) p13, another sentence is modified to ‘The proposed requirement of the pre-insertion quaternary complex increases the threshold of PG content for the membrane insertion to happen and thus makes it less likely on the surface of mammalian cells that contain PG at a low level in the membrane.’ 

(7) Please include information that Dap is effective only against Gram-positive bacteria and does not show antimicrobial properties against Gram-negative strains. The authors focused on emphasizing that Dap does not affect mammalian membranes, most likely due to the low PG content, however even membranes of Gram-negative bacteria are not susceptible to the Dap, despite the relatively high content of negatively charged PG in the inner membrane (e.g. inner cell membrane of E. coli has ~20% PG).

The requested information is already included in ‘Introduction’. In this part, Dap is introduced to be only active against Gram-positive bacteria, implicating that it is not active against Gram-negative bacteria. The reason Dap is inactive against E. coli or other Gramnegative bacteria is because the outer membrane prevents the antibiotic from accessing the PG in the inner membrane to cause any harm. When the outer membrane is removed, Dap will also attack the plasma membrane of Gram-negative bacteria. 

Literature cited in the comments:

(1) E. Krok, M. Stephan, R. Dimova, L. Piatkowski, Tunable biomimetic bacterial membranes from binary and ternary lipid mixtures and their application in antimicrobial testing, Biochim. Biophys. Acta - Biomembr. 1865 (2023). https://doi.org/10.1101/2023.02.12.528174.

(2) S.W. Ho, D. Jung, J.R. Calhoun, J.D. Lear, M. Okon, W.R.P. Scott, R.E.W. Hancock, S.K. Straus, Effect of divalent cations on the structure of the antibiotic daptomycin, Eur. Biophys. J. 37 (2008) 421-433. https://doi.org/10.1007/S00249-007-0227-2/METRICS.

(3) A. Pokorny, P.F. Almeida, The Antibiotic Peptide Daptomycin Functions by Reorganizing the Membrane, J. Membr. Biol. 254 (2021) 97-108. https://doi.org/10.1007/s00232-02100175-0.

(4) L. Robbel, M.A. Marahiel, Daptomycin, a bacterial lipopeptide synthesized by a nonribosomal machinery, J. Biol. Chem. 285 (2010) 2750127508. https://doi.org/10.1074/JBC.R110.128181.

(5) R. Taylor, K. Butt, B. Scott, T. Zhang, J.K. Muraih, E. Mintzer, S. Taylor, M. Palmer, Two successive calcium-dependent transitions mediate membrane binding and oligomerization of daptomycin and the related antibiotic A54145, Biochim. Biophys. Acta - Biomembr. 1858 (2016) 1999-2005. https://doi.org/10.1016/J.BBAMEM.2016.05.020.

eLife Assessment

This valuable study describes the molecular mechanism of daptomycin insertion into bacterial membranes. The authors provide solid in vitro evidence for the early events of daptomycin interaction with phospholipid headgroups and stronger, specific interaction with phosphatidylglycerol. This work will be of interest to bacterial membrane biologists and biochemists working in the antimicrobial resistance field.

Reviewer #3 (Public review):

Summary:

Machhua et al. in their work focused on unravelling the molecular mechanism of daptomycin binding and interaction with bacterial cell membranes. Daptomycin (Dap) is an acidic, cyclic lipopeptide composed of 13 amino acids, known for preferential binding to anionic lipids, particularly phosphatidylglycerol (PG), which are prevalent components in the membranes of Gram-positive bacteria. The process of binding and antimicrobial efficacy of Dap are significantly influenced by the ionic composition of the surrounding environment, especially the presence of Ca2+ ions. The authors underscore the presence of significant knowledge gaps in our understanding of daptomycin's mode of action. Several critical questions remain unanswered, including the basis for selective recognition and accumulation in membranes of Gram-positive strains, the specific role of Ca2+ ions in this process, and the mechanisms by which daptomycin binds to and inserts into the cell membrane.

Dap is intrinsically fluorescent due to its kynurenine residue (Kyn-13) and this property allows direct imaging of Dap binding to model cell membranes without the need of additional labeling. Taking advantage of this Dap autofluorescence, authors monitored the emission intensity of micelles, composed of varying DMPG content upon their exposure to Dap and compared it with the kinetics of fluorescence observed for zwitterionic DMPC and other negatively charged lipids such as cardiolipin (CA), POPA and POPS. The authors noted that the linear relationship between DMPG content and Dap fluorescence is strongly lipid-specific, as it was not observed for other anionic lipids. The manuscript sheds light on the specificity of Dap's interaction with CA and DMPG lipids. Through Ca2+ sequestration with EGTA, the authors demonstrated that the binding of Dap with CA is reversible, while its interaction with DMPG results in the irreversible insertion of Dap into the lipid membrane structure, caused by the significant conformational change of this lipopeptide. The formation of a stable DMPG-Dap complex was also verified in bacterial cells isolated from Gram-positive bacteria B. subtilis, where Dap exhibited a permanent binding to PG lipids.

Altogether, the authors endeavored to illuminate novel insights into the molecular basis of Dap binding, interaction, and the mechanism of insertion into bacterial cell membranes. Such understanding holds promise for the development of innovative strategies in combating drug resistance and the emerging of the so-called superbugs.

Strengths:

- The manuscript by Machhua et al. provides a comprehensive analysis of the Dap mechanism of binding and interaction with the membrane. It discusses various aspects of this, only apparently trivial interaction such as the importance of PG presence in the membrane, the impact of Ca2+ ions, and different mechanisms of Dap binding with other negatively charged lipids.<br /> - The authors focused not only on model membranes (micelles) but also extended their research to bacterial cell membranes obtained from B. subtilis<br /> - The research is not only a report of the experimental findings but tries to give potential hypotheses explaining the molecular mechanisms behind the observed results

Weaknesses:

- The authors overestimate their findings, stating that they propose a novel mechanism of Dap interaction with bacterial cell membranes. This research is the extension of the hypotheses that have already been reported.<br /> - The literature study and overall discussion about the mechanism of action of Ca2+ ions or conformational changes of daptomycin could be improved.

eLife Assessment

This important study provides empirical evidence of the effects of genetic diversity and species diversity on ecosystem functions across multi-trophic levels in an aquatic ecosystem. The support for these findings is solid, but a more nuanced interpretation of the results could strengthen the conclusions. The work will be of interest to ecologists working on multi-trophic relationships and biodiversity.

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 stated that species diversity had negative effects on ecosystem functions, while genetic diversity had positive effects. Additionally, 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 genetic diversity of the dominant species in each trophic level, biomass production, decomposition rates, and environmental data. The writing is logical and easy to follow.

Weaknesses:

The two main conclusions-(1) species diversity had negative effects on ecosystem functions, while genetic diversity had positive effects, and (2) these effects were observed only within each trophic level, not across the three levels-are overly generalized. Analysis of the raw data shows that species and genetic diversity have different effects depending on the ecosystem function. For example, neither affected invertebrate biomass, but species diversity positively influenced fish biomass, while genetic diversity had no effect. Furthermore, Table S2 reveals that only four effect sizes were significant (P < 0.05): one positive genetic effect, one negative genetic effect, and two negative species effects, with two effects within a trophic level and two across trophic levels. Additionally, using a P < 0.2 threshold to omit lines in the SEMs is uncommon and was not adequately justified. A more cautious interpretation of the results, with acknowledgment of the variability observed in the raw data, would strengthen the manuscript.

Author response:

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

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:

(1) 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. We now provide individual SEMs (Figure 3), although we limit causal relationships to those for which the p-value was below 0.2 for the sake of graphical clarity. We also provide the percentage of explained variance for each ecosystem function. We detail the graph in the Results section (see l. 317-328) and discuss them (see l. 387-398). Note that we do not detail each function separately as this would (in our opinion) result in a long descriptive paragraph from which it might be difficult to get some key information. Rather, we summarize the percentage of explained variance for each function and discuss the strength of environmental vs biodiversity effects for some examples. In the Discussion, we explain why environmental effects (on functions and biodiversity) are relatively weak. We mainly attribute this to the sampling scheme that follows an East-West gradient (weak altitudinal range) rather than an upstream-downstream gradient as it is traditionally done in rivers. The reasoning behind this sampling scheme is explained in our companion paper (Fargeot et al. Oikos 2023) to which we now refer more explicitly in the MS. Briefly, using an upstream-downstream gradient would have certainly push up the effects of the environment, but this would have made extremely complex the inference of biodiversity effects due to strong collinearity among environmental and biodiversity parameters.

(2) 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 now provide more details. However, we are unfortunately not sure that this helped reaching some stronger explanation regarding underlying mechanisms. To be frank, we did not succeed in improving mechanistic inferences based on the outputs of the SEM models. We explored visually some additional relationships (e.g. relationships between the biomass of the focal species and that of other species in the assemblage) that we now discuss a bit more, but again, this did not really help in better understanding processes. We realize this is a limitation of our study and that this can be frustrating for readers. Nonetheless, as said in the Discussion, field-based study must be taken for what they are; observational studies forming the basis for future mechanistic studies. Although we failed to explain mechanisms, we still think that we provide important field-base evidence for the importance of biodiversity (as a whole) for ecosystem functions.

3) 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 is now addressed, see our response to your first comment. We now explain (result section) and discuss environmental effects. As explained in the MS, environmental effects are similar in strength to those of biodiversity and are not that high, which is partly explained by the sampling scheme (see Fargeot et al. 2023). This is a choice we’ve made at the onset of the experiment, as we wanted to focus on biodiversity effects and avoid strong collinearity as it is generally the case in rivers (which impedes any proper and strong statistical inferences).

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:

(4) 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 have addressed this remark and we ask the reviewers and the readers to refer to our response to comment 1 from reviewer 1. Regarding co-variation among biodiversity estimates (SGDCs according to Vellend’s framework), we have addressed these issues in a companion paper that we now cite and expand further in the MS (Fargeot et al. Oikos, 2023). Given the size of the dataset and its complexity (and associated analyses), we have decided to focus on patterns of species and genetic biodiversity in a first paper (Oikos paper) and then on the link between biodiversity and functions (this paper). As it can be read in the Oikos’s paper, there are no co-variation in term of biodiversity estimates; species diversity is not correlated to genetic diversity, and within facet, there are not co-variation among species. In addition, environmental predictors are highly estimate-specific (i.e. environmental predictors sustaining species and genetic estimates are idiosyncratic). As a result (see the new Figure 3), environmental effects are relatively weak (the same intensity that those of biodiversity) and collinearity among parameters is relatively weak. The second point is important, as this permit to better infer parameters from models, and this allows to discuss direct relationships (as observed in Figure 3, indirect environmental effects are relatively rare). We provide in the Discussion a bit more explanation about the absence of co-variation among biodiversity estimates (see l. 433-440).

(5) 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.

We now provide p-values (Table S2) and the seven models (Figure 3).

(6) 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”. We changed the formulation.

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.

(7) 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 have modified the MS to make this clear.

(8) 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.

Fish species number varies from 1 to 11, invertebrate family number varies from 15 to 42 and the tree species number varies from 7 to 20 (see Fargeot et al. 2023 for details). We have added this information in the M&M. The gradients are hence relatively large and do not cover a restricted set of values. There is a variance in species number among sites, even if sites are collected along a relatively weak altitudinal gradient. This is obviously complex to compare to SNP (genomic) diversity. Genetic and species effects are similar in effect sizes (percentage of explained variance), so it does not seem we have biased one of the two gradients of biodiversity.

(9) 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 represents 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 if 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 provide an explanation (l. 217-219).

(10) 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 now cite some of these studies (Li et al 2020, Moi et al. 2021, Seibold et al. 2018).

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Introduction:

The introduction of the manuscript is generally well-structured, and the scientific questions are clearly presented. However, in each paragraph where specific aspects are introduced, the authors do not focus sufficiently on the given points. The current introduction discusses the weaknesses of previous studies extensively but lacks detailed explanations of mechanisms and a clear anticipation of this study's contributions.

For example:

L72-77: The authors mention that "genetic diversity may functionally compensate for a species loss," but this point is not highly relevant to the main analyses of this study, which focus on comparing the relative effects of species diversity and genetic diversity.

Yes true, we understand the point made by the reviewers. We deleted this part of the sentence.

L87-95: As previously noted, "whether environmental variation decreases or enhances the relative influence of genetic and species diversity on ecosystem functions" was not addressed in this study. Additionally, the last sentence seems unnecessary here, as it does not relate to "environmental variation." The phrase "generate insightful knowledge for future mechanistic models" is vague. It would be helpful to specify what kind of knowledge and what types of future mechanistic models are being referred to.

We modified these two sentences. We now posit the prediction that what has been observed under controlled conditions (that genetic and species have effects of similar magnitude) might not be the norm under fluctuating environments (because it has been shown that environmental variation modulates the strength of interspecific BEFS and create huge variance).

L96-116: The use of "for instance" three times in this paragraph makes the structure seem scattered, as only examples are provided. Improving the transition words can help the text focus better on the main point.

We have modified some parts of this section to better reflect predictions

L115-116: Again, it would be beneficial to specify what kind of insightful information can be provided.

We have modified this sentence by making more explicit some of the information that may be gained.

L117-134: Stating clear expectations can help the introduction focus on the mechanisms and assist readers in following the results.

We now provide some predictions. We were reluctant to make predictions in the first version of the MS as we have the feeling that predictions can go on very different direction depending on how we set the scene. We therefore stick to predictions that we think are the most logical (the simplest ones). This illustrates the lack of theoretical papers on these issues.

Methods:

L287-293: The method for estimating the standard effect size is unclear. I assume it was derived from the SEM models? This needs further clarification.

Yes, it is derived from the standardized estimate from each pSEM. This is now explained in the MS.

Results:

As mentioned in the public review, it is very important to show the results of analyzing raw data.

Done, see Figure 3 and Results section.

Table 1: The font and format of the PCA table are different from other tables and appear vague, resembling a picture rather than a table.

Changed.

Table 2 (and supplementary table): "D.f." is not explained in the table legend. Is 1 the numerator df and 30 the denominator df? Is the denominator the residual? Additionally, the table legend mentions "magnitude and direction." ANOVA only tests if the biodiversity effects are significantly different between species or genetic diversity, but not the magnitude. For example, -0.5 and 0.5 are very different, but their effect magnitudes are the same.

This is a mistake; sorry the format of the Table was from a previous version of the MS in which we used linear models rather that linear mixed models (both lead to the same results). The ANOVA used to test the significance of fixed terms in linear mixed model are based on Wald chi-sqare tests, and it should have been read “Chi-value” rather than “F-value” in both tables and the only degree of freedom in this test is the one at the numerator. This has been changed. We have changed the caption of the Table (“ANOVA table for the linear mixed model testing whether the relationships between biodiversity and ecosystem functions measured in a riverine trophic chain differ between the biodiversity facets (species or genetic diversity) and the types of BEF (within- or between-trophic levels)”)

Minor:

There should always be a space between a number and a unit. In the manuscript, spaces are inconsistently used between numbers and units.

Corrected

Reviewer #2 (Recommendations for the authors):

(1) In the introduction, the authors could focus more and build out what they predicted/hypothesized as well as what has been found in the manipulated experiments that examined the role of species and genetic diversity. That would enhance the background information for a more general audience, and highlight expected results and why.

We modified the Introduction according to comments made by reviewer 1 and clarified the predictions as best as we can.

(2) Similarly, the discussion is fairly big picture, but this dataset focused exclusively on this 3-trophic interaction in a riverine system. It could be beneficial to dig into the ecology to find out why the opposite effects of species and genetic diversity are seen within trophic levels in this system.

We have added some explanations based on the specific pSEM (see our responses to the public reviews for details). But as said in the responses to the public reviews, even with mode detailed models, it is hard to tease apart mechanisms. One important point is that genetic and species diversity do not correlate one to each other (they do not co-vary over space), which means the effect of one facet is independent from the other. However, apart from that, we can’t really tell more without more mechanistic approaches. We understand this is frustrating, but this is the nature of field-based data. This does not mean they are useless. On the contrary, they confirm and expand patterns found under controlled conditions (which for ecologists is quite important as nature is our playground), but they are limited in inferences of mechanisms.

(3) It would also be informative if the authors specified what positive and negative Z scores mean. It seems counterintuitive in Figure 3. For example, in the upper left, it's denoted as a larger intraspecific effect - which I'd assume is higher genetic (within species) diversity - but is this not where species diversity effects are higher? In theory this figure could be similar to Figure 1 from Des Roches et al. 2018 - where showing the 1:1 line of where species and genetic diversity effects are similar and then how some are more impacted by SD or GD as that links to the overall question, right?

For example: Figure 3 makes it seem that GD effects are stronger (more positive) for within trophic responses (which is reflected in the text), but in that quadrant, it states that the interspecific effect is larger?

yes, you’re true Figure 3 (now Figure 4) is not ideal. We added an explicit explanation for interpreting Zr in the main text. In addition, we modified the text in the quadrat as this was not correct. Note that it cannot be directly be compared to that of DesRoches et al. In DesRoches et al., there is a single effect size (ES) per situation (which is roughly expressed as “ES = effect of species - effect of genotypes”). Here, there are two ES per situation, one for the species effect, the other for the genetic effect, which makes the biplot more complex (as species and genetic can be similar in magnitude, but opposite in direction, e.g., 0.5 and -0.5). We may have done as DesRoches et al. (“ES = effect of species - effect of genotypes”), but as we don’t have absolute ES (as in DesRoches) the resulting signs of the ES are non sensical…Not easy for us to find a clever solution (or said differently, we were not clever enough to find an easy solution).  Nonetheless, we tried another visualization by including “sub-quadrats” into the four main quadrats. We hope this will be clearer

(4) It's unclear why authors included both a simplified linear mixed model with diversity type and biodiversity facet as fixed factors, and then a second linear model that included trophic level (with those other 2 factors and interactions), but only showed results of trophic level from that more complex model. It is unclear why they include two models when the more complex one would have evaluated all aspects of their research question and shown the same patterns.

You’re true, the more complex model evaluates both aspects. Nonetheless, as the hypotheses were strictly separated, we thought it is simpler to associate one model to one hypothesis. We agree that this duplicates information, but we would like to keep the two models to make the text more gradual.

eLife Assessment

This valuable work provides novel insights into the substrate binding mechanism of a tripartite ATP-independent periplasmic (TRAP) transporter, which may be helpful for the development of specific inhibitors. The structural analysis is convincing, but additional work will be required to establish the transport mechanism as well as well as binding sites for all ligands. This study will be of interest to the membrane transport and bacterial biochemistry communities.

Reviewer #1 (Public review):

Summary:

This manuscript reports the substrate-bound structure of SiaQM from F. nucleatum, which is the membrane component of a Neu5Ac-specific Tripartite ATP-dependent Periplasmic (TRAP) transporter. Until recently, there was no experimentally derived structural information regarding the membrane components of TRAP transporter, limiting our understanding of the transport mechanism. Since 2022, there have been 3 different studies reporting the structures of the membrane components of Neu5Ac-specific TRAP transporters. While it was possible to narrow down the binding site location by comparing the structures to proteins of the same fold, a structure with substrate bound has been missing. In this work, the authors report the Na+-bound state and the Na+ plus Neu5Ac state of FnSiaQM, revealing information regarding substrate coordination. In previous studies, 2 Na+ ion sites were identified. Here, the authors also tentatively assign a 3rd Na+ site. The authors reconstitute the transporter to assess the effects of mutating the binding site residues they identified in their structures. Of the 2 positions tested, only one of them appears to be critical to substrate binding.

Strengths:

The main strength of this work is the capture of the substrate bound state of SiaQM, which provides insight into an important part of the transport cycle.

Weaknesses:

The main weakness is the lack of experimental validation of the structural findings. The authors identified the Neu5Ac binding site, but only test 2 residues for their involvement in substrate interactions, which is quite limited. However, comparison with previous mutagenesis studies on homologues supports the location of the Neu5Ac binding site. The authors tentatively identified a 3rd Na+ binding site, which if true would be an impactful finding, but this site was not sufficiently experimentally tested for its contribution to Na+ dependent transport. This lack of experimental validation prevents the authors from unequivocally assigning this site as a Na+ binding site. However, the reporting of these new data is important as it will facilitate follow up studies by the authors or other researchers.

Comments on revisions:

Overall, the authors have done a good job of addressing the reviewers' comments. It's good to know that the authors are working on the characterisation of the potential metal binding site mutants - characterising just a few of these will provide much needed experimental support for this potential Na+ site.<br /> The new MD simulations provide some additional support for the new Na+ site and could be included. However, as the authors know, direct experimental characterisation of mutants is the ideal evidence of the Na+ site.

Aside from the characterisation of mutants, which seems to be held up by technical issues, the only remaining issue is the comparison of the Na+- and Na+/Neu5Ac-bound states with ASCT2.<br /> It still does not make sense to me why the authors are not directly comparing their Na+ only and Na+/Neu5Ac states with the structures of VcINDY in the Na+-only and Na+/succinate bound states. These VcINDY structures also revealed no conformational changes in the HP loops upon binding succinate, as the authors see for SiaQM. Therefore, this comparison is very supportive. It is understood that the similarity to the DASS structure is mentioned on p.17, but it is also interesting and useful to note that TRAP and DASS transporters also share a lack of substrate-induced local conformational changes, to the extent these things have been measured.

Reviewer #3 (Public review):

The manuscript by Goyal et al report substrate-bound and substrate-free structures of a tripartite ATP independent periplasmic (TRAP) transporter from a previously uncharacterized homolog, F. nucleatum. This is one of most mechanistically fascinating transporter families, by means of its QM domain (the domain reported in his manuscript) operating as a monomeric 'elevator', and its P domain functioning as a substrate-binding 'operator' that is required to deliver the substrate to the QM domain; together, this is termed an 'elevator with an operator' mechanism. Remarkably, previous structures had not demonstrated the substrate Neu5Ac bound. In addition, they confirm the previously reported Na+ binding sites, and report a new metal binding site in the transporter, which seems to be mechanistically relevant. Finally, they mutate the substrate binding site and use proteoliposomal uptake assays to show the mechanistic relevance of the proposed substrate binding residues.

Strengths:

The structures are of good quality, the presentation of the structural data has improved, the functional data is robust, the text is well-written, and the authors are appropriately careful with their interpretations. Determination of a substrate bound structure is an important achievement and fills an important gap in the 'elevator with an operator' mechanism.

Weaknesses:

Although the possibility of the third metal site is compelling, I do not feel it is appropriate to model in a publicly deposited PDB structure without directly confirming experimentally. The authors do not extensively test the binding sites due to technical limitations of producing relevant mutants; however, their model is consistent with genetic assays of previously characterized orthologs, which will be of benefit to the field. Finally, some clarifications of EM processing would be useful to readers, and it would be nice to have a figure visualizing the unmodeled lipid densities - this would be important to contextualize to their proposed mechanism.

Author response:

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

Reviewer #1 (Public Review):

This manuscript reports the substrate-bound structure of SiaQM from F. nucleatum, which is the membrane component of a Neu5Ac-specific Tripartite ATP-dependent Periplasmic (TRAP) transporter. Until recently, there was no experimentally derived structural information regarding the membrane components of the TRAP transporter, limiting our understanding of the transport mechanism. Since 2022, there have been 3 different studies reporting the structures of the membrane components of Neu5Ac-specific TRAP transporters. While it was possible to narrow down the binding site location by comparing the structures to proteins of the same fold, a structure with substrate bound has been missing. In this work, the authors report the Na+-bound state and the Na+ plus Neu5Ac state of FnSiaQM, revealing information regarding substrate coordination. In previous studies, 2 Na+ ion sites were identified. Here, the authors also tentatively assign a 3rd Na+ site. The authors reconstitute the transporter to assess the effects of mutating the binding site residues they identified in their structures. Of the 2 positions tested, only one of them appears to be critical to substrate binding.

Strengths:

The main strength of this work is the capture of the substrate-bound state of SiaQM, which provides insight into an important part of the transport cycle.

Weaknesses:

The main weakness is the lack of experimental validation of the structural findings. The authors identified the Neu5Ac binding site, but only tested 2 residues for their involvement in substrate interactions, which was very limited. The authors tentatively identified a 3rd Na+ binding site, which if true would be an impactful finding, but this site was not tested for its contribution to Na+ dependent transport, and the authors themselves report that the structural evidence is not wholly convincing. This lack of experimental validation undermines the confidence of the findings. However, the reporting of these new data is important as it will facilitate follow-up studies by the authors or other researchers.

The main concern, also mentioned by other reviewers, is the lack of mutational data and functional studies on the identified binding sites. Two other structures of TRAP transporters have been determined, one from Haemophilus influenzae (Hi) and the other from Photobacterium profundum (Pp). We will refer to the references in this paper as [1], Peter et al. as [2], and Davies et al. as [3]. The table below lists all the mutations made in the Neu5Ac binding site, including direct polar interactions between Neu5Ac and the side chains, as well as the newly identified metal sites.

The structure of Fusobacterium nucleatum (Fn) that we have reported shows a significant sequence identity with the previously reported Hi structure. When we superimpose the Pp and Fn structures, we observe that nearly all the residues that bind to the Neu5Ac and the third metal site are conserved. This suggests that mutagenesis and functional studies from other research can be related to the structure presented in our work.

The table below shows that all three residues that directly interact with Neu5Ac have been tested by site-directed mutagenesis for their role in Neu5Ac transport. Both D521 and S300 are critical for transport, while S345 is not. We do not believe that a mutation of D521A in Fn, followed by transport studies, will provide any new information.

However, Peter et al. have mutated only one of the 5 residues near the newly identified metal binding site, which resulted in no transport. The rest of the residues have not been functionally tested. We propose to mutate these residues into Ala, express and purify the proteins, and then carry out transport assays on those that show expression. We will include this information in the revised manuscript.

Author response table 1.

Reviewer #2 (Public Review):

In this exciting new paper from the Ramaswamy group at Purdue, the authors provide a new structure of the membrane domains of a tripartite ATP-independent periplasmic (TRAP) transporter for the important sugar acid, N-acetylneuraminic acid or sialic acid (Neu5Ac). While there have been a number of other structures in the last couple of years (the first for any TRAP-T) this is the first to trap the structure with Neu5Ac bound to the membrane domains. This is an important breakthrough as in this system the ligand is delivered by a substrate-binding protein (SBP), in this case, called SiaP, where Neu5Ac binding is well studied but the 'hand over' to the membrane component is not clear. The structure of the membrane domains, SiaQM, revealed strong similarities to other SBP-independent Na+-dependent carriers that use an elevator mechanism and have defined Na+ and ligand binding sites. Here they solve the cryo-EM structure of the protein from the bacterial oral pathogen Fusobacterium nucleatum and identify a potential third (and theoretically predicted) Na+ binding site but also locate for the first time the Neu5Ac binding site. While this sits in a region of the protein that one might expect it to sit, based on comparison to other transporters like VcINDY, it provides the first molecular details of the binding site architecture and identifies a key role for Ser300 in the transport process, which their structure suggests coordinates the carboxylate group of Neu5Ac. The work also uses biochemical methods to confirm the transporter from F. nucleatum is active and similar to those used by selected other human and animal pathogens and now provides a framework for the design of inhibitors of these systems.

The strengths of the paper lie in the locating of Neu5Ac bound to SiaQM, providing important new information on how TRAP transporters function. The complementary biochemical analysis also confirms that this is not an atypical system and that the results are likely true for all sialic acid-specific TRAP systems.

The main weakness is the lack of follow-up on the identified binding site in terms of structure-function analysis. While Ser300 is shown to be important, only one other residue is mutated and a much more extensive analysis of the newly identified binding site would have been useful.

Please see the comments above.

Reviewer #3 (Public Review):

The manuscript by Goyal et al reports substrate-bound and substrate-free structures of a tripartite ATP-independent periplasmic (TRAP) transporter from a previously uncharacterized homolog, F. nucleatum. This is one of the most mechanistically fascinating transporter families, by means of its QM domain (the domain reported in his manuscript) operating as a monomeric 'elevator', and its P domain functioning as a substrate-binding 'operator' that is required to deliver the substrate to the QM domain; together, this is termed an 'elevator with an operator' mechanism. Remarkably, previous structures had not demonstrated the substrate Neu5Ac bound. In addition, they confirm the previously reported Na+ binding sites and report a new metal binding site in the transporter, which seems to be mechanistically relevant. Finally, they mutate the substrate binding site and use proteoliposomal uptake assays to show the mechanistic relevance of the proposed substrate binding residues.

The structures are of good quality, the functional data is robust, the text is well-written, and the authors are appropriately careful with their interpretations. Determination of a substrate-bound structure is an important achievement and fills an important gap in the 'elevator with an operator' mechanism. Nevertheless, I have concerns with the data presentation, which in its current state does not intuitively demonstrate the discussed findings. Furthermore, the structural analysis appears limited, and even slight improvements in data processing and resulting resolution would greatly improve the authors' claims. I have several suggestions to hopefully improve the clarity and quality of the manuscript.

We appreciate your feedback and will make the necessary modifications to the manuscript incorporating most of the suggestions. We will submit the revised version once the experiments are completed. We are also working on improving the quality of the figures and have made several attempts to enhance the resolution using CryoSPARC or RELION, but without success. We will continue to explore newer methods in an effort to achieve higher resolution and to model more lipids, particularly in the binding pocket.

Reviewing Editor (Recommendations for the Authors):

After discussing the reviews, the reviewers and reviewing editor have agreed on a list of the most important suggested revisions for the authors, which, if satisfactorily addressed, would improve the assessment of the work. These suggested revisions are listed below. We also include the full Recommendations For The Authors from each of the individual reviewers.

(1) The authors tentatively identified a 3rd Na+ binding site, which if true would be an impactful finding, but this site was not tested for its contribution to Na+ dependent transport, and the authors themselves report that the structural evidence is not wholly convincing. Additional mutagenesis and activity experiments to test the contribution of this site to transport would strengthen the manuscript. Measuring Na+ concentration-response relations and calculating Hill slopes in WT vs. an M site mutant would be a good experiment. Given the lack of functional data and poor density, it does not seem appropriate to build the M site sodium in the PDB model.

The density is well defined to suggest a metal bound (waters would not be clearly defined at this resolution).  While our modeling of the site as a Na+ is arbitrary, this was done to satisfy the refinement programs where we have a known scatterer modeled.  We could model this density with other metals, but unlike crystallographic refinement, real-space refinement of cryoEM maps does not produce a difference map that might allow us to identify the metal but not conclusively.   The density of the maps is good (we have added better figures to demonstrate this).  We tried making multiple mutations to test for activity – unfortunately, we are still struggling to express proteins with mutations in this site in sufficient quantities to carry out transport assays.

In the absence of being able to do the experiments, we did MD simulations (carried out by Senwei Quan and Jane Allison at University of Auckland).  Our results are shown below – we are not certain without further studies that these should be included in the current paper (we will add them as authors if the editor feels that this evidence is critical).

Author response table 2.

We are showing this for review to suggest that K+, Ca2+, and Na+ were tried, and only Na+ stays stably in the binding pocket. The rest of the results will also have to be explained, which would change the focus of the paper.

We also provided the sequence to Alphafold3 and asked it to identify the possible metal binding sites—when the input was Na+, it found all three binding sites. 

Summary:  Both our experimental data and computational studies suggest the observed metal binding site is real but at the moment, it is not possible to refine the structure and put an unidentified metal.  Computational studies suggest that this is a high-probability Na+ site. 

Demonstration of cooperativity between the Na+ site and transport require carrying out these experiments with mutations in these sites in a concentration-dependent manner. Unfortunately, our inability to produce well-expressed and purified proteins with mutations in a short time frame failed. 

(2) The authors identified the Neu5Ac binding site but only tested 2 residues for their involvement in substrate interactions, which was very limited. Given that the major highlight of this paper is the identification of the Neu5Ac binding site, it would strengthen the manuscript if the authors provided a more extensive series of mutagenesis experiments - testing at least the effect of D521A would be important. One inconsistency is Ser345 mutagenesis not affecting transport, and the authors should further discuss in the text why they think that is.

D521A has been tested in H. influenzae, and this mutation results in loss of transport.  This residue is highly conserved and occupies the same position. We expect the result to remain the same. 

We have added a few extra lines to discuss Serine 345: “Ser 345 OG is 3.5Å away from the C1-carboxylate oxygen – a distance that would result in a weak interaction between the two groups. It is, therefore, not surprising that the mutation into Ala did not affect transport. The space created by the mutation can be occupied by a water molecule.”

(3) The purification and assessment of the stability of the protein are described in text alone with no accompanying data. It would be beneficial to include these data (e.g. in the Supplementary info) as it allows the reader to evaluate the protein quality.

This is now added as Supplementary Figure 2.

(4) The structural figures throughout the paper could benefit from more clarity to better support the conclusions. Specific critiques are listed below:

- Figure 1: since the unbound map has a similar reported resolution, displaying the unbound structure's substrate binding site with the same contour would clearly demonstrate that the appearance of this density is substrate-dependent.

- Figure 1: the atomic fit of the ligand to the density, and the suggested coordination by side chain and backbone residues, would be useful in this figure.

- Figure 1: I think it would be more intuitive to compare apo and bound structures with the same local resolution scale.

We have remade Figure 1 “Architecture of FnSiaQM with nanobody. (A and B) Cryo-EM maps of FnSiaQM unliganded and sialic acid bound at 3.2 and 3.17 Å, respectively. The TM domain of FnSiaQM is colored using the rainbow model (N-terminus in blue and C-terminus in red). The nanobody density is colored in purely in red. The density for modeled lipids is colored in tan and the unmodelled density in gray. The figures were made with Chimera at thresholds of 1.2 and 1.3 for the unliganded and sialic acid-bound maps. (C and D) The cytoplasmic view of apo and sialic acid bound FnSiaQM, respectively. Color coding is the same as in panels A and B. The density corresponding to sialic acid and sodium ions are in purple. The substrate binding sites of apo and sialic acid bound FnSiaQM are shown with key residues labeled. The density (blue mesh) around these atoms was made in Pymol with 2 and 1.5 s for the apo and the sialic acid, respectively, with a carve radius of 2 Å.”

The local resolution maps have been moved to Supplementary Figure 3.

- Figure 3, Figure 5a: The mesh structures throughout the manuscript are blocky and very difficult to look at and interpret, especially for the ion binding sites, which are currently suggestive of but not definitively ion densities. Either using transparent surfaces, higher triangle counts, or smoothing the surface might help this.

We have made Figure 3 again with higher triangle counts.  We tried all three suggestions and this provided the best figure. We have replaced Figure 5A with density for Neu5Ac and residues around it.

- Figure 5A: It would be important to show the densities of the entire binding pocket, especially coordinating side chains, to show the reader what is and isn't demonstrated by this structure.

- It's not clear how Figure 5D is supposed to show that the cavity can accommodate Neu5Gc, as suggested by the text - please make the discussed cavity clearer in the Figure.

We have now marked with an arrow the Methyl Carbon where the hydroxyl group is added.  We have mentioned that in the legend.  It is open to the periplasmic side of the cavity.

- Supplementary Figure 4: Please label coordinating residue sites.

Labels have been added to Supplementary Figure 6 which was earlier Supplementary Figure 4.

(5) Intro section: the authors should introduce the work on HiSiaP around the role of the R147 residue in high-affinity Neu5Ac binding, which coordinates the carboxylate of Neu5Ac, and which is a generally conserved mechanism for organic acid binding in other TRAP transporters. This context will help magnify their discovery later that in the membrane domains, it is a key serine and not an arginine that coordinates the carboxylate group (probably as the local concentration of Neu5Ac is high and tight binding site is not desirable for rapid transport, which is mentioned in the discussion).

Thank you for pointing this out. We have added a new sentence to the introduction.

“All the SiaP structures show the presence of a conserved Arginine that binds to the C1-carboxylate of Neu5Ac, and this Arg residue is critical as the high electrostatic affinity may be important to have a strong binding affinity that sequesters the small amounts that reach the bacterial periplasmic space  (Glaenzer et al., 2017).”

(6) TRAP transporters exist for many organic compounds and not just sialic acid, which might be nice to make the reader aware of.

We initially did not do this as this is an advance paper and this was discussed in the earlier paper (Currie et. al., 2024). However, we have now added a sentence to the introduction. “Additionally, amino acids, C4-dicarboxylates, aromatic substrates and alpha-keto acids are also transported by TRAP transporters (Vetting et al., 2015). “

(7) On p. 12, the authors describe the Neu5Ac binding site as a large solvent-exposed vestibule, having previously described the substrate-bound state as occluded. These descriptions should be adjusted to make clear which structure is being referenced. The clarity of this would be substantially improved if the authors included a figure that showed this occlusion - currently none of the structure figures clearly demonstrate what the authors are referring to. There are several conspicuous unmodeled densities proximal to the substrate, reminiscent of lipids (in between transport and scaffold domain) and possibly waters/ions. Given this, it is really surprising that the substrate binding site is described as "solvent-exposed" since the larger molecules seem to occlude the pocket. The authors should further process their dataset and discuss the implications of these surrounding densities.

We have processed the data sets carefully both with cryosparc and relion and the resolution described here is same with both software with the cryosparc maps slightly better in terms of interpretability of peripheral helices and described in the manuscript. The current sample (FnTRAP) with the nanobody is a relatively stable sample (in our experience with other similar proteins) as evident from the number of images and particles to achieve a decent resolution and thus the workflow is straightforward and simple.  There are number of non-protein densities, which in principle can be modelled but we have chosen a conservative approach not to model these extra densities (except for the two lipids, few ions) due to limit of the resolution. It is possible that increasing the number of particles will result in an increase in resolution but from the estimated B-factor (125 or 135 Å2 for unliganded and liganded), this will certainly require lot of more images with no guarantee of increased resolution.

The question of outward open Vs outward occluded is a valid point. We have now modified this in the manuscript. “The Neu5Ac binding site has a large solvent-exposed vestibule towards the cytoplasmic side, while its periplasmic side is sealed off. Cryo-EM map shows the presence of multiple densities that could be modeled as lipids, possibly preventing the substrate from leaving the transporter. However, the densities are not well defined to model them as specific lipids, hence they have not been modeled.  We describe this as the “inward-facing open state” with the substrate-bound.”

(8) On p.15, the activity of FnSiaPQM in liposomes is reported, although the impetus for this study is not clear. Presumably, the reason for its inclusion is to ensure that the structurally characterized protein is active. It would be useful to say this at the start of the section if this is the case. This study nicely shows that the energetics and requirements of transport are identical to all the previous studies on Neu5Ac TRAP transporters - it would be good to acknowledge this somewhere in this section as well.

These changes have been incorporated.  We have added a line to say why we did this and added as the last line that this is similar to other SiaPQM’s characterized.

(9) Figure 5C. The authors show the transport activity with and without valinomycin. The authors do not explain the rationale for testing and reporting both conditions for these mutants; an explanation is required, or the data should be simplified. The expected membrane potential induced by valinomycin should be mentioned in the legend.

We have simplified Figure 5C and added the expected membrane potential value.

(10) The authors state that the S300A mutant is inactive. However, unless the authors also measured the background binding/transport of radiolabelled substrate in the absence of protein, then the accuracy of this statement is not clear because Figure 5C does indicate some activity for S300A, albeit much lower than WT. This is an important point in light of the authors' suggestion that the membrane protein does not need a binding site of high affinity or stringent selectivity.

We thank the reviewer for pointing this out we have now added a line in the experimental protocols “The experimental values were corrected by subtracting the control, i.e. the radioactivity taken up in liposomes reconstituted in the absence of protein. The radioactivity associated with the control samples, i.e. empty liposomes was less than 10% with respect to proteoliposomes.”.

(11) There are several issues and important omissions in the work cited:

- It is not normal practice to cite a reference in the abstract and the citation is only to the second structure of HiSiaQM, which does not fairly reflect previous work in the field by only referring to their own work. Also throughout the article, it is normal practice with in-text citations to order them chronologically, i.e. earliest first. Please update this.

This article was submitted as an “Research advance article”.  The instructions specifically say that “Research advance article should cite the article in eLife this paper advances.  Hence the citation of the “second structure of HiSiaQM”.  In fact, in the manuscript we explicitly say “The first structure of _Hi_SiaQM (4.7 Å resolution) demonstrated that it is composed of 15 transmembrane helices and two helical hairpins.”   We are following the policy laid out.  

Zotero organizes multiple references in alphabetical order, we did not choose to do it that way – the suggestion of bias is not true. The final version of the accepted paper will have numbers, and this argument will automatically be corrected.

- Intro: please cite the primary papers discovering other families of sialic acid transporters.

- Intro: When introducing information on the binding site, dissociation constant of Neu5Ac, and thermodynamics of ligand binding to SiaP, the authors should also include references to the work done by others in addition to their own work.

The Setty et al. paper was the first to demonstrate that the two-component systems are distinct, and that the binding protein of the TRAP system binds enthalpically while the binding protein of the ABC system binds entropically (SiaP vs SatA). As the reviewer points out, this is significant because it highlights how the Arg binding to the carboxylate, which is the enthalpic driver in this case and contributes to the difference between sugar binding to SiaP and SatA. Many studies have published binding affinities of molecules to SiaP, but this paper offers valuable insight into the differences between these systems. We have cited a number of the SiaP papers from other groups, including acknowledging the first structure of SiaP from H. influenzae by Muller et al., in 2006.

- p.5 "TRAP transporters are postulated to employ an elevator-type mechanism...". This postulation has been experimentally tested and published, so should be discussed and referenced (Peter et al. 2024. https://doi.org/10.1038/s41467-023-44327-3).

We have now corrected this error. We removed “are postulated to” and added the reference.

- p.5 "Notably, the transport of Neu5Ac by TRAP transporters requires at least two sodium ions (Davies et al., 2023)." The requirement for at least 2 Na+ ions for Neu5Ac transport was first demonstrated in Mulligan et al. PNAS 2009, so should also be cited (for completion, so should Mulligan et al. JBC. 2012 and Currie et al. elife 2023, which have also shown this requirement is a commonality amongst all Neu5Ac TRAP transporters).

Added.

- P.12, Mulligan et al, JBC, 2012 should be added to the citations in the first sentence.

Added.

- p.19 "Interestingly, even the dicarboxylate transporter from V. cholerae (VcINDY) binds to its ligand via electrostatic interactions with both carboxylate groups". Other references are more appropriate than the one used to support this statement.

Also added references for Mancusso et. al, 2012, Nie et.al, 2017 and Sauer et.al., 2022 here.

- p.19. "The structure of the protein in the outward-facing conformation is unknown". The authors do not discuss the mechanistic findings from Peter et al 2024 Nat Comm here. The work described in that paper revealed an experimentally verified model of the OFS of HiSiaQM, so really needs to be included.

This is not an experimentally determined 3D structure. They have shown the possible existence of this by microscopy, but the structure is not determined. The work mentioned is a wonderful piece of work, but it does not report the three-dimensional structure of the protein in the outward-facing conformation to allow us to understand the nature of the molecular interactions. 

- The reference to Kinz-Thompson et al 2022 on p. 6 is not appropriate - neither the HiSiaQM papers nor the PpSiaQM paper makes reference to this work when identifying the binding site. More suitable references are used, for example, Mancusso et al 2012, Nie et al 2017 and Sauer et al 2022; this should be reported accurately.

Added the suggested references.  We think the paper (Kinz-Thomposin et al 2022) is relevant and have also kept that reference.

- Garaeva et al report the opposite of what the authors mention - "In the human neutral amino acid transporter (ASCT2), which also uses the elevator mechanism, the HP1 and HP2 loops have been proposed to undergo conformational changes to enable substrate binding and release (Garaeva et al., 2019)." In fact, this paper suggested a one-gate model of transport (HP2), where HP1 seems uninvolved in gating.

The Reviewer is correct.  We were wrong and not clear.  The entire paragraph has been rewritten.

“While, both the HP1 and HP2 loops have been hypothesized to be involved in gating, in the human neutral amino acid transporter (ASCT2), (which also uses the elevator mechanism), only the HP2 loops have been shown to undergo conformational changes to enable substrate binding and release (Garaeva et al., 2019). Hence, it is suggested that there is a single gate that controls substrate binding. Superposition of the _Pp_SiaQM and _Hi_SiaQM structures do not reveal any change in these loop structures upon substrate binding. For TRAP transporters, the substrate is delivered to the QM protein by the P protein; hence, these loop changes may not play a role in ligand binding or release. This may support the idea that there is minimal substrate specificity within SiaQM and that it will transport the cargo delivered by SiaP, which is more selective.”

- p.19 "suggesting that SSS transporters have probably evolved to transport nine-carbon sugars such as Neu5Ac (Wahlgren et al, 2018)." Surely this goes without saying since Wahlgren et al 2018 demonstrated that SiaT, an SSS, could transport sialic acid? It's unclear why this was included here - perhaps it needs to be rewritten to make the point more clearly, but as it stands, this statement appears self-evident. Furthermore, these proteins can transport all kinds of molecules (see TCDB 2.A.21). This statement needs to be clarified. 

This was a comparison to other Neu5Ac binding sites in other Neu5Ac transporters. We have modified the sentence. “The polar groups bind to both the C1-caboxylate side of the molecule and the C8-C9 carbonyls, suggesting that Proteus mirabilis Neu5Ac transporter (SSS type) evolved specifically to transport nine-carbon sugars such as Neu5Ac (Wahlgren et al., 2018)”.  These were arguments we were making to suggest that the lack of tight binding could also mean reduced specificity.

- The authors reconstitute the FnSiaQM and measure transport with SiaP, which resembles closely what is known for both HiSiaPQM, VcSiaPQM, which is not cited (https://doi.org/10.1074/jbc.M111.281030).

- Regarding lipids between transport and scaffold domains: there is precedent for such lipids in the elevator transporter GltPh, Wang, and Boudker (eLife 2020) proposed similar displacements during transport and would be appropriate to cite here.

We have now cited the reference to the Mulligan et al., 2012 paper.  We also added a sentence on the findings of GltPh paper by Wang and Boudker.  Thank you for pointing this out.

(12) p.9 "TRAP transporters, as their name suggests, comprise three units: a substrate-binding protein (SiaP) and two membrane-embedded transporter units (SiaQ and SiaM) (Severi et al., 2007)." This is somewhat odd phrasing because the existence of fused membrane components has been well-documented for a long time. The addition of "Many" at the start of the sentence fixes this.

Added Many.

(13) On p.12 the authors compare the ligand-induced conformational changes of FnSiaQM with ASCT2, citing Garaeva et al, 2019. This comparison does not make sense considering TRAP transporters and ASCT2 do not share a common fold. A far superior comparison is with DASS transporters, which actually do have the same fold as TRAP transporters. And, importantly, the Na+ and substrate-induced conformational changes have been investigated for DASS transporters revealing a unique mechanism likely shared by TRAP transporters (Sauer et al, Nat Comm, 2022). The text on p.12 should be adjusted to replace the ASCT comparison with a VcINDY comparison.

The purpose of citing the ASCT2 paper was only concerning the HP1 and HP2 gates.  The authors show that HP2 changes conformation only.  Comparing the two FnSiaQM structures – with and without ligand, we see no change in either the HP1 or the HP2 loops.  On Page 17, when we describe the structure, we do specifically mention that the overall architecture is similar to VcINDY and the DASS transporters.

(14) p.12 "For TRAP transporters, the substrate is delivered to the QM protein by the SiaP" protein;" "SiaP protein" should be "P protein"

Corrected.

(15) p.18. "periplasmic membrane" should be "cytoplasmic membrane".

Corrected.

(16) p.19. "This prevents Neu5Ac from binding..." There is no evidence for this so this needs to be softened, e.g. "This likely prevents Neu5Ac from...".

Agree – Modified.

(17) Figure 2B is rather small, cramped, and difficult to see. We suggest that the authors make that panel larger, or include it as a stand-alone supplementary figure.

We have moved this figure into a supplementary figure as suggested by the reviewer.

(18) The authors describe the Neu5Ac binding site in SiaQM. It would be helpful if the authors provided a figure in support of the statement that the Neu5Ac binding site architecture is similar to dicarboxylate in VcINDY (especially as Neu5Ac is a monocarboxylate).

The Neu5Ac binding site is NOT similar to the VcINDY binding site. But, we understand the origin of the comment. We have now changed the sentence: “The overall architecture of the Neu5Ac binding site is similar to that of citrate/malate/fumarate in the di/tricarboxylate transporter of V. cholerae (Vc_INDY), but the residues involved in providing specificity are different (Kinz-Thompson _et al., 2022; Mancusso et al., 2012; Nie et al., 2017; Sauer et al., 2022). Neu5Ac binds to the transport domain without direct interactions with the residues in the scaffold domain. The majority of the interactions are with residues in the HP1 and HP2 loops of the transport domain (Figure 5B). Asp521 (HP2), Ser300 (HP1), and Ser345 (helix 5) interact with the substrate through their side chains, except for one interaction between the main chain amino group of residue 301 and the C1-carboxylate oxygen of Neu5Ac. Mutation of the residue equivalent to Asp521 has been shown to result in loss of transport (Peter et al., 2022). To evaluate the role of residues Ser-300 and Ser-345, we mutated them to alanine and performed the transport assays.”  

(19) When comparing the binding modes of Neu5Ac to different proteins in Figure 6, it would be helpful to include the structure in this paper as well.

The Neu5Ac binding site is present in figure 5. We would prefer not to show it again in Figure 6.

Additionally, there is a clear binding mode of Neu5Ac in Figure 1 as well.

(20) The manuscript would benefit from a more detailed comparison between Na+-bound (described as apo) and Na+/Neu5Ac structures, especially the prospective gates. If this transporter behaves anything like the archetypical ion-coupled glutamate transporters, some structural changes in the gates might be expected to facilitate transport domain movement when the substrate is loaded, but not when only Na+ is bound. It would be important to discuss and visualize these changes.

We have described in the manuscript that there is NO change in the HP1 and HP2 gates between the unliganded structure and the Neu5Ac bound structure. The major difference we observe is the ordering of the third metal binding site.

A figure comparing the substrate binding pockets between the different high-resolution structures would also be informative. Do the bonding distances between ligands and side chains significantly change between homologs?

This is the only Neu5Ac bound structure.  Since the specificity to the substrate comes from the variability of the residues that interact it, we do not believe that this figure would not add much value.  

(21) A supplementary figure (or an inset to Figure 2) showing pairwise percent identity between different characterized QM transporters would be useful.

We have now added a Supplementary Figure 4 showing the comparison of the three QM sequences whose structures have been determined.

(22) There is relatively minimal EM processing. More rigorous processing would require relatively little effort and could boost resolution, making this a vastly improved manuscript with a much more confident interpretation of structures.

We described the overall workflow. The processing was rigorous. After obtaining the first maps, we created templates with the structure and did template-based picking.  We then did several rounds of 2D classification followed by homogenous refinement, Non-Uniform Refinement.  We then made masks and carried out local refinement.  We then got the best maps and did a 3D classification. Refined the 3D classes independently.  Then, we regrouped them based on how similar they were. We then went back and picked particles again (we used different methods of particle picking, but template-based picking resulted in the final set of particles used) and went through the whole process again.  At the end of the refinement, we carried out global and local CTF refinement followed by reference-based motion correction. The final refinement was then done with the Bayesian polished particles.  The final refinement was local refinement with a mask over only the transporter and the nano-body. After the reviews came, we tried multi-body refinement in Relion5.  It did not improve resolution. We have expanded the legend to supplementary Figure 2 (without listing all the different things we tried). The best resolution we obtained for the structure was 3.1 Å. However, it is important to note that the local resolution of the map around the ligand is good. 

We realized this is not easy to depict in a local resolution map.  So, we wrote a script to take every atom, then take a radius of 5 Å (again we tried different radii and used the optimal one; we are preparing a manuscript to describe this), take all the local resolution values within the 5 Å spere and average it and add it as B-factor that atom. We have moved the local resolution map figure to the supplement and replaced Figure 1 with a Cartoon, where the color represents the local resolution in which the atom is. 

(23) Calling the structure without Neu5Ac bound an "apo" structure is confusing since it indeed has the ligand Na+ present and bound. "Na+" and "Na+/Neu5Ac" structures would be more appropriate.

Changed all “apo” to “unliganded”.

eLife Assessment

This valuable study reports a potential connection between the seminal microbiome and sperm quality/male fertility. The data are generally convincing. This study will be of interest to clinicians and biomedical researchers who work on microbiome and male fertility.

Reviewer #1 (Public review):

Summary:

The authors analyzed the bacterial colonization of human sperm using 16S rRNA profiling. Patterns of microbiota colonization were subsequently correlated with clinical data, such as spermiogram analysis, presence of reactive oxygen species (ROS), and DNA fragmentation. The authors identified three main clusters dominated by Streptococcus, Prevotella, and Lactobacillus & Gardnerella, respectively, which aligns with previous observations. Specific associations were observed for certain bacterial genera, such as Flavobacterium and semen quality. Overall, it is a well-conducted study that further supports the importance of the seminal microbiota.

Strengths:

- The authors performed the analysis on 223 samples, which is the largest dataset in semen microbiota analysis so far<br /> - Inclusion of negative controls to control contaminations.<br /> - Inclusion of a positive control group consisting of men with proven fertility.

Weaknesses:

- The manuscript needs comprehensive proofreading for language and formatting. In many instances spaces are missing or not required.<br /> - Could the authors explore correlation network analyses to get additional insights in the structure of different clusters?<br /> - The github link is not correct.<br /> - It is not possible to access the dataset on ENA.<br /> - Add the graphs obtained with decontam analysis as a supplementary figure.<br /> - There is nothing about the RPL group in the results section, while the authors discuss this issue in the introduction. What about the controls with proven fertility?<br /> - While correctly stated in the title, the term microbiota should be used throughout the manuscript instead of "microbiome"

Comments on revised version:

Discussion: Could the authors discuss more the findings about Flavobacterium? Has it ever been associated with the urogenital tract? What is the relative abundance in the present study: this type of bacterium has been previously associated with contaminations (PMID: 25387460, 30497919).

Figure 1: Increase the size of panel A.

Figure 3: Can the authors indicate the relative abundance of each genus/species by the size of the node?

Supplementary data: I don't see anywhere the decontam plots.

Author response:

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

Reviewer 1:

- The manuscript needs comprehensive proofreading for language and formatting. In many instances, spaces are missing or not required.

Thank you for your comments. The manuscript has been thoroughly proofread for errors in language and formatting.

- Could the authors explore correlation network analyses to get additional insights into the structure of different clusters? 

We have added a co-occurrence analysis (at species taxonomic level) based on SparCC to the manuscript (Figure 2).

This is described on Page 9 line 141-148

- The GitHub link is not correct. 

The github repository has now been made public.

- It is not possible to access the dataset on ENA. 

We have changed the ENA study PRJEB57401 status to open.

- Add the graphs obtained with decontam analysis as a supplementary figure. 

We have added the outputs of decontam (.csv files with feature lists of ASVs that were filtered based on the prevalence and frequency tests) to the github repository.

- There is nothing about the RPL group in the results section, while the authors discuss this issue in the introduction. What about the controls with proven fertility? 

Thank you. We have amended the manuscript to compare characteristics between the RPL, unexplained subfertility and controls groups.

Line 1279-130 page 8:  

“The study group represented 85% of samples with high sperm DNA fragmentation, 85% of samples with elevated ROS and 79% of samples with oligospermia. Rates of abnormal seminal parameters including low sperm concentration, reduced progressive motility and ROS concentrations were found to be highest in the MFI group (Supplementary Figure 1). Baseline characteristics between the RPL, unexplained subfertility and controls groups were similar.

Line 150-154 Page 9: 

“Bacterial richness, diversity and load were similar between all patient groups examined in the study (Supplementary Figure 4).

- While correctly stated in the title, the term microbiota should be used throughout the manuscript instead of "microbiome" 

Thank you. This misnomer has been amended throughout the manuscript.

Minor corrections:

Line 25: provoke is not a good term here. 

Thank you. The term ‘provoke’ has been removed

Line 26: why does semen culture have a limited scope? 

Thank you. Line 40-41 Page 3 has been amended:

“It is therefore plausible that asymptomatic seminal infections may be associated with impaired reproductive function in some men. Since semen culture has a limited scope for studying the seminal microbiota due to its inability to identify all present microbiota next generation sequencing (NGS) approaches have been reported recently by a growing number of investigators (13, 14, 15, 16, 17, 18, 19)”.

Line 68: write μl correctly

Thank you. This has been corrected

Line 131: several organisms at the genus level. 

Thank you. This has been corrected

Line 136: what are the relative abundances of these genera? Is this relevant? 

The mean relative abundances for the key taxa mention in each cluster are all above 20%. This information has been added to the manuscript text on page 9, line 153.

Line 173: Molina et al. 

Thank you. This has been corrected

Line 173: the contaminations are referred to the low biomass nature of testicular samples. If present, bacteria of accessory gland secretions are an integral part of the seminal microbiota itself. Please review these sentences. 

Thank you. This had been reworked to highlight the important of urethral contamination, which you later allude to as a limitation of our study is the failure to provide paired urine and semen samples.

Page 11 line 194-196

“Molina et al report that 50%-70% of detected bacterial reads may be environmental contaminants in a sample from extracted testicular spermatozoa (35); with the addition of passage along the urethra it is likely that contamination of ejaculated semen would be much higher.”

Table 1: remove results interpretation from table caption. 

Thank you this has been acted upon.

Table 1: why in some cases, like in DNA fragmentation index, the total is not equal to n=223? 

This is due to missing data/ analysis not possible for some men due to the requirement of a minimum number of sperm in the ejaculate to perform DNA fragmentation testing.

Table 1: "frag" is not defined. 

Thank you, this has been amended

Tables 2, 3 & 4: bacterial genera in italics. 

Thank you, this has been amended

Figure 1A: add the fertility status information above the cluster colors. 

Thank you, this has been amended in Figure 1.

Figure 1C: the color code is confusing. Use different colors for each cluster. 

Figure 1 legend: bacterial genera in italics. 

Figures 1 & 2: the authors should use similar chart formatting in the two tables. 

Thank you, this has been amended

Reviewer 2:

(1) The patient groups have different diagnoses and should be handled as different groups, and not fused into one 'patient' group in analyses. <br /> Why are the data in tables presented as controls and cases? I would consider men from couples with recurrent pregnancy loss, unexplained infertility, and male factor infertility to have different seminal parameters (not to fuse them into one group). This means, that the statistical analyses should be performed considering each group separately, and not to fuse 3 different infertility diagnoses into one patient group. 

We have conducted detailed analyses, requested by the reviewer, comparing seminal DNA, ROS and microbiota characteristics between each individual patient groups (Supplimental figures 1 and 4). No specific taxa (at either genera or species-level) were found to differ in relative abundance between the diagnostic groups. However, we expect associations between parameters such as reactive oxygen species, or DNA fragmentation, and relative abundance of bacterial species, to be general and not restricted to or specific to each diagnostic group. Therefore, we also conducted further analyses aggregating data from all patient groups to investigate relationships common to these different forms of male reproductive dysfunction.

(2) Were any covariables included in the statistical analyses, e.g. age, BMI, smoking, time of sexual abstinence, etc? 

Covariates were not included in the statistical analyses. This has been added in the manuscript to the limitations.

Page 14 line 267-268

“Additionally, we did not have other covariables such as smoking status with which to include in further analyses”.

(3) Furthermore, it is known that 16S rRNA gene analysis does not provide sensitive enough detection of bacteria on the species level. How much do the authors trust their results on the species level? 

The limitations of taxonomic assignment using 16S rRNA gene metataxonomics are well documented. However, the capacity to assign sequence amplicons at species level depends on the sequence variability of the 16S rRNA gene for each of the taxa reported and the specific gene region chosen. In this study, amplification of the V1-V2 region was performed using a mixed 28f primer set (see methods for details) that enables resolution and assignment of several bacterial species highly relevant to the reproductive tract including Lactobacillus spp., such as L. crispatus and L. iners, (e.g. https://doi.org/10.3389/fcell.2021.641921, https://doi.org/10.1128/msystems.01039-23, https://doi.org/10.1186/s12915-023-01702-2). In this study, we report the presence of L. iners, but not L. crispatus in semen samples, and we have also identified a specific association/co-occurrence between Gardnerella vaginalis and Lactobacillus iners, similar to that observed in vaginal bacterial communities.

(4) Were the analyses of bacterial genera and species abundances with seminal quality parameters controlled for diagnosis and other confounders? 

As stated in point 2, no adjustment was made for co-variates. No differences in microbiome composition were observed among the three diagnostic groups, so no adjustments were made to our analysis.

(5) The authors stress that their study is the biggest on the microbiome in semen. However, when considering that the study consists of 4 groups (with n=46-63), it does not stand out from previous studies. 

Our study is overall the largest investigating interactions between the seminal microbiome and male reproductive dysfunction. Other studies have included greater numbers of men with infertility.

(6) Weaknesses: There is a lack of paired seminal/urinal samples. 

Thank you. This limitation has been added.

Page 14 line 266-267

“A further limitation of this study, and others, is the lack of reciprocal genital tract microbiota testing of the female partners, or paired seminal and urinary samples from male participants”.

Recommendation for authors to consider:

Including previous classical reviews in the introduction: DOI:10.1097/MOU.0000000000000742 <br /> DOI: 10.1038/s41585-019-0250-y 

Thank you. This has been added.

Mentioning in the M&M section that there is a supplementary text with a more detailed M&M part. 

Thank you. This has been added. Further methodological detail can be found in supplementary text.

Revising the use of 'microbiota' and 'microbiome', they are not synonyms. When talking of 16S rRNA gene analysis, we consider 'microbiome' analysis. 

Thank you. This misnomer has been amended throughout the manuscript.

Revising the text, there are several erratas (e.g. verb missing, etc). 

Thank you for your comments. The manuscript has been thoroughly proofread for errors in language and formatting.

Author response:

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

Public Reviews: 

Reviewer #1 (Public Review):

Summary: 

In the manuscript entitled "Magnesium modulates phospholipid metabolism to promote bacterial phenotypic resistance to antibiotics", Li et al demonstrated the role of magnesium in promoting phenotypic resistance in V. alginolyticus. Using standard microbiological and metabolomic techniques, the authors have shown the significance of fatty acid biosynthesis pathway behind the resistance mechanism. This study is significant as it sheds light on the role of an exogenous factor in altering membrane composition, polarization, and fluidity which ultimately leads to antimicrobial resistance. 

Strengths: 

(1) The experiments were carried out methodically and logically. 

(2) An adequate number of replicates were used for the experiments. 

Weaknesses: 

(1) The introduction section needs to be more informative and to the point.  

Thank you so much for your suggestion. We have revised the introduction to make it more informative and to the point as following:

“Non-inheritable antibiotic or phenotypic resistance represents a serious challenge for treating bacterial infections. Phenotypic resistance does not involve genetic mutations Phenotypic resistance does not involve genetic mutations and is transient, allowing bacteria to resume normal growth. Biofilm and bacterial persisters are two phenotypic resistance types that have been extensively studied (Brandis et al., 2023; Corona & Martinez, 2013). Biofilms have complex structures, containing elements that impede antibiotic diffusion, sequestering and inhibiting their activity (Ciofu et al., 2022). Biofilm-forming bacteria and persisters also have distinct metabolic states that significantly reduce their antibiotic susceptibility (Yan & Bassler, 2019). These two types of phenotypic resistance share the common feature in their retarded or even cease of growth in the presence of antibiotics (Corona & Martinez, 2013). However, specific factors that promote phenotypic resistance and allow bacteria to proliferate in the presence of antibiotics remain poorly defined.

Metal ions have a diverse impact on the chemical, physical, and physiological processes of antibiotic resistance  (Booth et al, 2011; Lu et al, 2020; Poole, 2017). This includes genetic elements that confer resistance to metals and antibiotics (Poole, 2017) and metal cations that directly hinder (or enhance) the activity of specific antibiotic drugs (Zhang et al., 2014). The metabolic environment can also impact the sensitivity of bacteria to antibiotics (Jiang et al., 2023; Lee & Collins, 2012; Peng et al., 2015; Zhang et al., 2020; Zhao et al., 2021). Light metal ions, such as magnesium, sodium, and potassium, can behave as cofactors for different enzymes (Du et al., 2016) and influence drug efficacy. Heavy metal ions, including Cu2+ and Zn2+, confer resistance to antibiotics (Yazdankhah et al., 2014; Zhang et al., 2018). Recent reports suggest that sodium negatively regulates redox states to promote the antibiotic resistance of Vibrio alginolyticus (Yang et al., 2018), while actively growing Bacillus subtilis cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al, 2019). In Gram-negative bacteria, by contrast, zinc enhances antibiotic efficacy by potentiating carbapenem, fluoroquinolone, and β-lactam-mediated killing (Isaei et al., 2016; Zhang et al., 2014). Magnesium influences bacterial structure, cell motility, enzyme function, cell signaling, and pathogenesis (Wang et al., 2019). This mineral also modulates microbiota to harvest energy from the diet (Garcia-Legorreta et al., 2020), allowing Bacillus subtilis to cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al., 2019). However, the role of magnesium in promoting phenotypic resistance is less well understood.

Vibrios inhabit seawater, estuaries, bays, and coastal waters, regions full of metal ions such as magnesium (Kumarage et al., 2022). Magnesium is the second most dissolved element in seawater after sodium. At a salinity of 3.5% seawater, the magnesium concentration is about 54 mM (Potis, 1968), and in deep seawater, can be as high as 2,500 mM (Wang et al., 2024). Vibrio parahaemolyticus and V. alginilyticus are two representative Vibrio pathogens that infect humans and aquatic animals, resulting in illness and economic loss, respectively (Grimes, 2020). (Fluoro)quinolones such as balofloxacin are used to treat Vibrio infection, however, resistance has emerged due to overuse (Suyamud et al., 2024). Indeed, (fluoro)quinolones are one of China's two primary residual chemicals associated with aquaculture (Liu et al., 2017). Vibrio can develop quinolone resistance through mutations in the DNA gyrase gene or through plasmid-mediated mechanisms (Dutta et al., 2021). Thus, the use of V. parahaemolyticus and V. alginilyticus as bacterial representatives, and balofloxacin as a quinolone-based antibacterial representative, can help to define novel magnesiumdependent phenotypic resistance mechanisms of pathogenic Vibrio species. 

The current study evaluated whether magnesium induces phenotypic resistance in Vibrio species and defined the molecular/genetic basis for this resistance. Genetic approaches, GC-MS analysis of metabolite and membrane remodeling upon antibiotic exposure, membrane physiology, and extensive antimicrobial susceptibility testing were used for the evaluations.”

(2) The weakest point of this paper is in the logistics through the results section. The way authors represented the figures and interpreted them in the results section (or the figure legends) does not match. The figures are difficult to interpret and are not at all self-explanatory. 

Thank you so much for your suggestion. We have followed your suggestion to check the match between result and figures. They are now revised. 

(3) There are too many mislabeling of the figure panels in the main text which makes it difficult to find out which figures the authors are explaining. There should be more explanation on why and how they did the experiments and how the results were interpreted. 

Thank you so much for your suggestion. We have checked the figures and main text to ensure that we make every figure clearly stated.  

Reviewer #2 (Public Review): 

Summary: 

In this study, the authors aimed to identify if and how magnesium affects the ability of two particular bacteria species to resist the action of antibiotics. In my view, the authors succeeded in their goals and presented a compelling study that will have important implications for the antibiotic resistance research community. Since metals like magnesium are present in all lab media compositions and are present in the host, the data presented in this study certainly will inspire additional research by the community. These could include research into whether other types of metals also induce multi-drug resistance, whether this phenomenon can be observed in other bacterial species, especially pathogenic species that cause clinical disease, and whether the underlying molecular determinants (i.e. enzymes) of metal-induced phenotypic resistance could be new antimicrobial drug targets themselves. 

Strengths: 

This study's strengths include that the authors used a variety of methodologies, all of which point to a clear effect of exogenous Mg2+ on drug resistance in the targeted species. I also commend the authors for carrying out a comprehensive study, spanning evaluation of whole cell phenotypes, metabolic pathways, genetic manipulation, to enzyme activity level evaluation. The fact that the authors uncovered a molecular mechanism underlying Mg2+-induced phenotypic resistance is particularly important as the key proteins should be studied further.

Weaknesses: 

I believe there are weaknesses in the manuscript, however. The authors take for granted that the reader is familiar with all the assays utilized, and do not properly explain some experiments, and thus I highly suggest that the authors add a brief statement in each situation describing the rationale for each selected methodology (more details are in the private review to the authors). The Results section is also quite long and bogs down at times, and I suggest that the authors reduce its length by 10 to 20%. In contrast, the Introduction is sparse and lacks key aspects, for example, there should be mention of the study's main purpose and approaches, plus an introduction to the authors' choice of species and their known drug resistance properties, as well as the drug of choice (balofloxacin). Another notable weakness is that the authors evaluated Mg2+-induced phenotypic resistance only against two closely related species, and thus the generalizability of this mechanism of drug resistance is not known. The paper would be strengthened if the authors could demonstrate this type of phenotypic resistance in at least one more Gram-negative species and at least one Gram-positive species (antimicrobial susceptibility evaluations would suffice), each of which should be pathogenic to humans. Demonstrating magnesium-induced phenotypic drug resistance in the WHO Priority Bacterial Pathogens would be particularly important. 

In general, the conclusions drawn by the authors are justified by the data, except for the interpretation of some experiments. Importantly, this paper has discovered new antimicrobial resistance mechanisms and has also pointed to potential new targets for antimicrobials. 

Thank you so much for your suggestion! We followed your idea the revise the manuscript as following:

(1) We added a brief statement in the situation to explain the result and methodology according to your suggestion in the private review.

(2) To make the streamline of the story more logic, we moved the whole second result to supplementary text and supplementary figure. 

(3) We revised the introduction part by adding additional information to make it informative and to the point as following:

“Non-inheritable antibiotic or phenotypic resistance represents a serious challenge for treating bacterial infections. Phenotypic resistance does not involve genetic mutations Phenotypic resistance does not involve genetic mutations and is transient, allowing bacteria to resume normal growth. Biofilm and bacterial persisters are two phenotypic resistance types that have been extensively studied (Brandis et al., 2023; Corona & Martinez, 2013). Biofilms have complex structures, containing elements that impede antibiotic diffusion, sequestering and inhibiting their activity (Ciofu et al., 2022). Biofilm-forming bacteria and persisters also have distinct metabolic states that significantly reduce their antibiotic susceptibility (Yan & Bassler, 2019). These two types of phenotypic resistance share the common feature in their retarded or even cease of growth in the presence of antibiotics (Corona & Martinez, 2013). However, specific factors that promote phenotypic resistance and allow bacteria to proliferate in the presence of antibiotics remain poorly defined.

Metal ions have a diverse impact on the chemical, physical, and physiological processes of antibiotic resistance  (Booth et al, 2011; Lu et al, 2020; Poole, 2017). This includes genetic elements that confer resistance to metals and antibiotics (Poole, 2017) and metal cations that directly hinder (or enhance) the activity of specific antibiotic drugs (Zhang et al., 2014). The metabolic environment can also impact the sensitivity of bacteria to antibiotics (Jiang et al., 2023; Lee & Collins, 2012; Peng et al., 2015; Zhang et al., 2020; Zhao et al., 2021). Light metal ions, such as magnesium, sodium, and potassium, can behave as cofactors for different enzymes (Du et al., 2016) and influence drug efficacy. Heavy metal ions, including Cu2+ and Zn2+, confer resistance to antibiotics (Yazdankhah et al., 2014; Zhang et al., 2018). Recent reports suggest that sodium negatively regulates redox states to promote the antibiotic resistance of Vibrio alginolyticus (Yang et al., 2018), while actively growing Bacillus subtilis cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al, 2019). In Gram-negative bacteria, by contrast, zinc enhances antibiotic efficacy by potentiating carbapenem, fluoroquinolone, and β-lactam-mediated killing (Isaei et al., 2016; Zhang et al., 2014). Magnesium influences bacterial structure, cell motility, enzyme function, cell signaling, and pathogenesis (Wang et al., 2019). This mineral also modulates microbiota to harvest energy from the diet (Garcia-Legorreta et al., 2020), allowing Bacillus subtilis to cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al., 2019). However, the role of magnesium in promoting phenotypic resistance is less well understood.

Vibrios inhabit seawater, estuaries, bays, and coastal waters, regions full of metal ions such as magnesium (Kumarage et al., 2022). Magnesium is the second most dissolved element in seawater after sodium. At a salinity of 3.5% seawater, the magnesium concentration is about 54 mM (Potis, 1968), and in deep seawater, can be as high as 2,500 mM (Wang et al., 2024). Vibrio parahaemolyticus and V. alginilyticus are two representative Vibrio pathogens that infect humans and aquatic animals, resulting in illness and economic loss, respectively (Grimes, 2020). (Fluoro)quinolones such as balofloxacin are used to treat Vibrio infection, however, resistance has emerged due to overuse (Suyamud et al., 2024). Indeed, (fluoro)quinolones are one of China's two primary residual chemicals associated with aquaculture (Liu et al., 2017). Vibrio can develop quinolone resistance through mutations in the DNA gyrase gene or through plasmid-mediated mechanisms (Dutta et al., 2021). Thus, the use of V. parahaemolyticus and V. alginilyticus as bacterial representatives, and balofloxacin as a quinolone-based antibacterial representative, can help to define novel magnesiumdependent phenotypic resistance mechanisms of pathogenic Vibrio species. 

The current study evaluated whether magnesium induces phenotypic resistance in Vibrio species and defined the molecular/genetic basis for this resistance. Genetic approaches, GC-MS analysis of metabolite and membrane remodeling upon antibiotic exposure, membrane physiology, and extensive antimicrobial susceptibility testing were used for the evaluations.”

(4) We examined the effect of magnesium in WHO listed priority strains, which confirmed the results as following:

“Importantly, exogenous MgCl2 also increased MICs of clinic isolates, carbapenemresistant Escherichia coli, carbapenem-resistant Klebsiella pneumoniae, carbapenemresistant Pseudomonas aeruginosa and carbapenem-resistant Acinetobacter baumannii to balofloxacin (Fig 1G).”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

(1) There are many grammatical mistakes to point out. The manuscript needs proofreading and editing.

We appreciate this comment! The manuscript has been revised by a native speaker.

(2) The introduction could be more informative. A little more description of magnesium - such as what it does to antibiotics and how it's known to affect the microbiome - might be helpful for the general readers. The question remains why out of all the metal ions that might affect antibiotic resistance (many of them are less explored), authors particularly decided to work on the effect of magnesium. The introduction should cover the rationale of their hypothesis. Also, the authors might want to briefly talk about the model organisms (V. algonolyticus and V. parahemolyticus) describing how threatening they are and how they are becoming resistant to antibiotics. 

We appreciate this comment! We revise the introduction by providing additional information as following:

“In Gram-negative bacteria, by contrast, zinc enhances antibiotic efficacy by potentiating carbapenem, fluoroquinolone, and β-lactam-mediated killing (Isaei et al., 2016; Zhang et al., 2014). Magnesium influences bacterial structure, cell motility, enzyme function, cell signaling, and pathogenesis (Wang et al., 2019). This mineral also modulates microbiota to harvest energy from the diet (Garcia-Legorreta et al., 2020), allowing Bacillus subtilis to cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al., 2019). However, the role of magnesium in promoting phenotypic resistance is less well understood.

Vibrios inhabit seawater, estuaries, bays, and coastal waters, regions full of metal ions such as magnesium (Kumarage et al., 2022). Magnesium is the second most dissolved element in seawater after sodium. At a salinity of 3.5% seawater, the magnesium concentration is about 54 mM (Potis, 1968), and in deep seawater, can be as high as 2,500 mM (Wang et al., 2024). Vibrio parahaemolyticus and V. alginilyticus are two representative Vibrio pathogens that infect humans and aquatic animals, resulting in illness and economic loss, respectively (Grimes, 2020). (Fluoro)quinolones such as balofloxacin are used to treat Vibrio infection, however, resistance has emerged due to overuse (Suyamud et al., 2024). Indeed, (fluoro)quinolones are one of China's two primary residual chemicals associated with aquaculture (Liu et al., 2017). Vibrio can develop quinolone resistance through mutations in the DNA gyrase gene or through plasmid-mediated mechanisms (Dutta et al., 2021). Thus, the use of V. parahaemolyticus and V. alginilyticus as bacterial representatives, and balofloxacin as a quinolone-based antibacterial representative, can help to define novel magnesiumdependent phenotypic resistance mechanisms of pathogenic Vibrio species. 

The current study evaluated whether magnesium induces phenotypic resistance in Vibrio species and defined the molecular/genetic basis for this resistance. Genetic approaches, GC-MS analysis of metabolite and membrane remodeling upon antibiotic exposure, membrane physiology, and extensive antimicrobial susceptibility testing were used for the evaluations. ”

(3) Figure 1C is mislabeled as 1B (line 100). Line 101: The sentence is not clear and very confusing. What is meant by 15.6mM - 62.4 mM? Are they talking about the concentration of BLFX (though in the figure the concentration was shown in µg)? Please rewrite the sentence in a simplified way. Also, the zone of inhibition was decreased with increasing MgCl2, not increased. 

We appreciate this comment! These have been revised, including that Fig 1B is now corrected as Fig. 1C. Line 101, which is now Line 122. The sentence was revised as following:

“At balofloxacin doses of 1.56, 3.125, 6.25, and 12.5 µg, the zone of inhibition decreased with increasing MgCl2 (Fig 1D)”

(4) In the western blot images, it would be nice to indicate the MW of the protein bands shown. The loading control used for the experiments should be clearly mentioned in the figure legends. 

We appreciate this comment! The MWs are indicated in the western-blot image throughout the manuscript. 

The loading control is clearly stated in the figure legend as following:

“Whole cell lysates resolved by SDS-PAGE gel was stained with Coomassie brilliant blue as loading control.”. 

(5) Figures 2 B and C: the figure legend does not explain what the authors wanted to show. It's not clear how they plotted the inhibitory curve, or the binding efficacy. These panels need an explanation of how the analysis was done.

We appreciate this comment! The figure 2 is now removed to Suppl. Fig 2, and the description of figure 2 is moved to Suppl. Text. We revise the description of the result as following, which is in Suppl. Text:

“Prior studies suggest that the chelation of antibiotics by magnesium ions inhibits antibiotic uptake (Deitchman et al., 2018; Lunestad and Goksøyr, 1990). To investigate whether magnesium binds to balofloxacin, balofloxacin was pre-incubated with magnesium, and zone of inhibition (ZOI) analysis was conducted. Six different concentrations of balofloxacin were separately incubated with six different concentrations of MgCl2, and then spotted on filter paper so that a defined amount of balofloxacin could be used for ZOI. While lower concentrations of MgCl2, (0.78, 3.125, or 12.5 mM) did not alter the ZOI, higher concentrations, including 50 and 200 mM MgCl2, decreased the ZOI (Suppl. Fig 2A), suggesting that even high doses of magnesium had only a partial effect on balofloxacin through direct binding. For example, at 200 mM MgCl2 and 5 or 10 μg/mL balofloxacin, the balofloxacin ZOI was 53.2 and 70.3% of the ZOI at 0 mM MgCl2, suggesting that  50% of the antibiotics were still functional. Intracellular BLFX also decreased with increasing MgCl2 (Suppl. Fig 2B), while exogenous Mg2+ increased intracellular Mg2+ levels in a dose-dependent manner. For example, exogenous 50 and 200 mM MgCl2 increased intracellular Mg2+ levels to 1.21 and 1.31 mM, respectively (Suppl. Fig 2C). The relationship between TolC, an efflux pump that transports quinolones from bacterial cells, and Mg2+ was also assessed (Kobylka et al., 2020; Song et al., 2020). The expression of TolC/tolC was unaffected by Mg2+ (Suppl. Fig 2D). Magnesium is critical for LPS stability. LPS levels increased at 200 mM Mg2+ (Suppl. Fig 2E), however, the loss of waaF, lpxA, and lpxC, three key genes involved in LPS biosynthesis, did not influence balofloxacin sensitivity/resistance in the presence of Mg2+ (Suppl. Fig 2F). These findings suggest that magnesium-induced LPS biosynthesis does not contribute directly to BLFX resistance and demonstrate that Mg2+ influx is involved in balofloxacin resistance.”

(6) For the metabolomics results, it will help immensely if the authors provide a volcano plot of the identified metabolites and plot the heat map according to the -log2 metabolite intensities. In Figure 3A, it's not clear what information is conveyed through Euclidean distance calculations of the heat map. In Figure 3 B, the authors mentioned that the OPLS-DA test was conducted, although the figure shows a PCA plot, so it's not clear how these two are connected. Figure 3 E: the figure legend says scattered plot, but the panel represents color-coded numerical values, not a scattered plot. Also, it's not clear how they got those values. 

We appreciate this comment! We quite agree with you that if the differential metabolites could be shown as volcano plot. However, we didn’t adopt volcano plot in this study because this is a magnesium concentration-dependent metabolomes that includes 6 groups in parallel. Volcano plots may give a complex view of the comparison among different groups. We also tried to plot the heat map according to the -log2 metabolite intensities. Although this analysis cluster 200 mM and 50 mM groups better, the data of low magnesium concentrations was not consistent, which may be due to the minor metabolic change of low concentrations magnesium. Thank you for your understanding. 

For Euclidean distance calculations, we explain in the figure legend as following:

“Euclidean distance calculations were used to generate a heatmap that shows clustering of the biological and technical replicates of each treatment.” 

In Figure 2B, which was Figure 3B in previous version, it has been replaced with OPLS-DA analysis in the revised version. 

In Figure 2E, which was Figure 3E in previous version, it is revised as following:

“E. Areas of the peaks of palmitic acid and stearic acid generated by GC-MS analysis.” 

(7) In Figure 4, the figure legends (as well as the in the text) are not properly referred to. Please make sure to refer to the correct panel. 

We appreciate this comment! The figure legends have been corrected to match the panel and text. 

Figure 4F: how was the synergy analysis done? In the methods section, the authors described the antibiotic bactericidal assay protocol, but there was no clear indication of how they generated the isobologram. 

We appreciate this comment! We provide additional information in the Figure 3F legend, which was Figure 4F in previous version,  as following: 

“Synergy analysis for BFLX with palmitic acid for V. alginolyticus. Synergy was performed by comparing the dose needed for 50% inhibition of the synergistic agents (white) and non-synergistic (i.e., additive) agents (purple).”

(8) Figure 5 A: the scatter plot is plotted according to the area along the Y axis: which "area" is represented here? There is absolutely no explanation, neither in the results nor in the figure legends. Using box plots might be a better option than using a scattered plot.

We appreciate this comment! “Area” has been noted in the revised manuscript as following:

“The area indicates the area of the peak of the metabolite in total ion chromatography of GC-MS.” 

(9) In Figure 6 A, the heat map is plotted according to the column Z scores. What is meant by "column Z score"? The corresponding figure legend says, "heat map showing differential abundance of lipid". Z scores do not represent an abundance of a variable, so the conclusion might not be appropriate here. 

We appreciate this comment! In Figure 5A, which was Figure 6A in previous version, column Z score shows the abundance of metabolites analyzed, which is automatically generated in the heat map analysis to give a sign of these metabolites tested. The legend has been revised as following: 

“Heatmap showing changes in differential lipid levels at the indicated concentration of MgCl2.”  

(10) Line 313-314: it should be Figure EV6C.  

We appreciate this comment! The citation has been corrected.

(11) The authors have shown that Mg+2 does not alter the LPS transport system, however, there was some significant increase in LPS expression at 200mM MgCl2. It would be interesting if the authors could also check if Mg+2 has any effect on the outer membrane protein (OMP) integrity (by checking OMP components BamA and LptD).  

We appreciate this comment!  We have carefully examined the membrane permeability in Figure 7. We thus didn’t perform additional experiment here to see the change of BamA and LptD. Thank you very much for your understanding.

(12) I wonder if the authors could check the effect of extracellular Mg+2 during the co-treatment of palmitic acid, linoleic acid, and balofloxacin. Will there still be the antagonistic effect or the presence of Mg+2 could change the phenotype? 

We appreciate this comment! Additional experiments is performed as following:

“Furthermore, magnesium had a minimal effect on the antagonistic effect of palmitic acid, linolenic acid, and balofloxacin (Fig 4G), suggesting that this mineral functions through lipid metabolism.” 

Reviewer #2 (Recommendations For The Authors): 

(1) As mentioned in the Public Review, I strongly believe that the impact of this study will be more significant if magnesium-induced phenotypic drug resistance could be demonstrated in at least one other Gram-negative and one other Grampositive species, both of which should be human pathogens. The full suite of experiments would not be necessary for this suggestion; evaluation of the effect of Mg concentration in growth media on the drug resistance of other species, testing the different antibiotic types used in this study, would be sufficient. 

We appreciate this comment! Additional experiments have performed to test this idea. Mg2+ has the similar effect on carbapenem-resistant Escherichia coli, carbapenem-resistant Klebsiella pneumoniae, carbapenem-resistant Pseudomonas aeruginosa and carbapenem-resistant Acinetobacter baumannii as the similar as on the Vibrio species in shown in Figure 1G. These have been described following as

“Importantly, exogenous MgCl2 also increased MICs of clinic isolates, carbapenemresistant Escherichia coli, carbapenem-resistant Klebsiella pneumoniae, carbapenemresistant Pseudomonas aeruginosa and carbapenem-resistant Acinetobacter baumannii to balofloxacin (Fig 1G).”

(2) I recommend that the Introduction section be expanded. I recommend one or two sentences introducing the two Vibrio species selected for study. I.e. why did the authors choose these two species? What is known about their phenotypic drug resistance in the literature? Why did the authors select balofloxacin for their studies, is it a common antimicrobial used vs Vibrios? As well, the end of the Introduction section ends abruptly with no transition to the present study itself. The end of the introduction should include one or two sentences introducing the main purpose of the study, its approach, and the techniques undertaken. For example, "In this study, we evaluated whether magnesium induces phenotypic resistance in Vibrio species and the molecular/genetic basis for such resistance. We used genetic approaches, GC-MS analysis of metabolite and membrane remodeling upon antibiotic exposure, membrane physiology, and extensive antimicrobial susceptibility evaluations." 

We appreciate this comment! We revise the introduction by providing additional information as following:

“In Gram-negative bacteria, by contrast, zinc enhances antibiotic efficacy by potentiating carbapenem, fluoroquinolone, and β-lactam-mediated killing (Isaei et al., 2016; Zhang et al., 2014). Magnesium influences bacterial structure, cell motility, enzyme function, cell signaling, and pathogenesis (Wang et al., 2019). This mineral also modulates microbiota to harvest energy from the diet (Garcia-Legorreta et al., 2020), allowing Bacillus subtilis to cope with ribosome-targeting antibiotics by modulating ion flux (Lee et al., 2019). However, the role of magnesium in promoting phenotypic resistance is less well understood.

Vibrios inhabit seawater, estuaries, bays, and coastal waters, regions full of metal ions such as magnesium (Kumarage et al., 2022). Magnesium is the second most dissolved element in seawater after sodium. At a salinity of 3.5% seawater, the magnesium concentration is about 54 mM (Potis, 1968), and in deep seawater, can be as high as 2,500 mM (Wang et al., 2024). Vibrio parahaemolyticus and V. alginilyticus are two representative Vibrio pathogens that infect humans and aquatic animals, resulting in illness and economic loss, respectively (Grimes, 2020). (Fluoro)quinolones such as balofloxacin are used to treat Vibrio infection, however, resistance has emerged due to overuse (Suyamud et al., 2024). Indeed, (fluoro)quinolones are one of China's two primary residual chemicals associated with aquaculture (Liu et al., 2017). Vibrio can develop quinolone resistance through mutations in the DNA gyrase gene or through plasmid-mediated mechanisms (Dutta et al., 2021). Thus, the use of V. parahaemolyticus and V. alginilyticus as bacterial representatives, and balofloxacin as a quinolone-based antibacterial representative, can help to define novel magnesiumdependent phenotypic resistance mechanisms of pathogenic Vibrio species. 

The current study evaluated whether magnesium induces phenotypic resistance in Vibrio species and defined the molecular/genetic basis for this resistance. Genetic approaches, GC-MS analysis of metabolite and membrane remodeling upon antibiotic exposure, membrane physiology, and extensive antimicrobial susceptibility testing were used for the evaluations. ”

(3) The authors introduce the acronym AWST but never use it again in the paper, instead they use SWT. The authors should introduce SWT only for consistency. 

We appreciate this comment! We have corrected all the “SWT” to “ASWT”

(4) Line 76 is not clear: what is meant by "some of which could influence drug efficacy" - the enzymes that utilize light metal ions are co-factors? Or the metals directly?  

We appreciate this comment! The information we wanted to deliver is that light metal ions can serve as cofactors to catalyze biochemical reaction. Such chemical reaction would alter the drug efficacy, e.g. the Fe-S cluster are metallocofactor for proteins which regulates redox chemistry including antibioticinduced redox change. However, this information is not appropriate for this manuscript, so we delete this sentence. 

(5) Line 90: add a reference corroborating that this chemical composition is a mimic of marine water. The NaCl concentration used in particular looks quite low. 

We appreciate this comment! It was a typo error. The NaCl concentration was 210 mM as shown in Suppl. Table 1. We also provide details of the chemical composition of the marine water as following:

“Marine environments and agriculture, where antibiotics are commonly used, are rich in magnesium. To investigate whether this mineral impacts antibiotic activity, the minimal inhibitory concentration (MIC) of V. alginolyticus ATCC33787 and V. parahaemolyticus VP01, which we referred as ATCC33787 and VP01 afterwards, isolated from marine aquaculture, to balofloxacin (BLFX) in Luria-Bertani medium

(LB medium) plus 3% NaCl as LBS medium and “artificial seawater” (ASWT) medium that included the major ion species in marine water (Wilson, 1975) (LB medium plus 210 mM NaCl, 35 mM Mg2SO4, 7 mM KCl, and 7 mM CaCl2) were assessed”

(6) Line 98 and Figure 1B. M9 is indicated in the text but does not appear in the figure, the figure only shows SWT. This should be checked. Line 99: based on Figure 1C, the authors are adding MgCl2 to SWT, SWT should be mentioned in this line. Line 100: I believe this is referring to Figure 1C, which should be checked. 

We appreciate this comment! 

Line 98, which is now Line 118: We have corrected M9 to ASWT as following:

“However, the MIC for BLFX was higher in ASWT medium supplemented with Mg2SO4 or MgCl2 than in LB medium (Fig 1B).”

Line 99, which is now Line 133: the sentence is corrected as following:

“The MIC for BLFX increased at higher concentrations of MgCl2 in ASWT”

Line 100, which is now Line 135: we have corrected Fig 1B to Fig. 1C.

(7) Line 101: text and Figure 1D are not consistent, as Figure 1D does not show this level of precision in added MgCl2 as indicated in the text (15.6 - 62.4 mM).  

We appreciate this comment! The sentence has been corrected as following: “At balofloxacin doses of 1.56, 3.125, 6.25, and 12.5 µg, the zone of inhibition decreased with increasing MgCl2 (Fig 1D)””.  

(8) MgCl2 clearly induces increasing levels of BLFX resistance, and to high levels, but not for every antibiotic. For example, the level of increased resistance to blactams is low (ceftriaxone) and plateaus (ceftazidime). As well, resistance to gentamicin plateaus at a lower level than the other aminoglycosides. These observations do not take away from the conclusion that Mg induces multi-drug resistance, but since the behaviour of the MICs for these drugs is different than the other drugs, they should be mentioned. Also, Figure 1F - tetracyclines (plural) is used for vertical axis label - does this refer to the tetracycline itself or the class itself, and if the class, which one was tested? 

We appreciate this comment! We revise the description as following: “Notably, magnesium had a reduced effect on ceftriaxone and gentamicin than other antibiotics.”

The tetracyclines is labeled as “Oxytetracycline” in the revised manuscript. 

- The magnesium chelation experiments presented in Figure 2 are not clear. The authors should briefly mention how this was done around line 128, and what data underlies the values in Figure 2C. Figure 2B is also not clear to me at all. Similarly, how the authors measured intracellular balofloxacin and Mg2+ is not clear and should be mentioned briefly around lines 130-132. 

We appreciate this comment! These have been rewritten following as  “To investigate whether magnesium binds to balofloxacin, balofloxacin was preincubated with magnesium, and zone of inhibition (ZOI) analysis was conducted. Six different concentrations of balofloxacin were separately incubated with six different concentrations of MgCl2, and then spotted on filter paper so that a defined amount of balofloxacin could be used for ZOI. While lower concentrations of MgCl2, (0.78, 3.125, or 12.5 mM) did not alter the ZOI, higher concentrations, including 50 and 200 mM MgCl2, decreased the ZOI (Suppl. Fig 2A), suggesting that even high doses of magnesium had only a partial effect on balofloxacin through direct binding. For example, at 200 mM MgCl2 and 5 or 10 μg/mL balofloxacin, the balofloxacin ZOI was 53.2 and 70.3% of the ZOI at 0 mM MgCl2, suggesting that  50% of the antibiotics were still functional. Intracellular BLFX also decreased with increasing MgCl2 (Suppl. Fig 2B), while exogenous Mg2+ increased intracellular Mg2+ levels in a dose-dependent manner. For example, exogenous 50 and 200 mM MgCl2 increased intracellular Mg2+ levels to 1.21 and 1.31 mM, respectively (Suppl. Fig 2C). The relationship between TolC, an efflux pump that transports quinolones from bacterial cells, and Mg2+ was also assessed (Kobylka et al., 2020; Song et al., 2020). The expression of TolC/tolC was unaffected by Mg2+ (Suppl. Fig 2D). Magnesium is critical for LPS stability. LPS levels increased at 200 mM Mg2+ (Suppl. Fig 2E), however, the loss of waaF, lpxA, and lpxC, three key genes involved in LPS biosynthesis, did not influence balofloxacin sensitivity/resistance in the presence of Mg2+ (Suppl. Fig 2F). These findings suggest that magnesium-induced LPS biosynthesis does not contribute directly to BLFX resistance and demonstrate that Mg2+ influx is involved in balofloxacin resistance.”

- Line 135: LPS cannot be "expressed", as the authors word it here. This should be corrected. Also, the inspection of Figure 2G actually shows the levels of LPS increase with increased Mg2+. The authors should re-evaluate these results and change their description around this area of the Results. 

We appreciate this comment! We have removed the whole Figure 2 to Supplementary Text and Supplementary Figure 2. We rewrite this part as following: “The relationship between TolC, an efflux pump that transports quinolones from bacterial cells, and Mg2+ was also assessed (Kobylka et al., 2020; Song et al., 2020). The expression of TolC/tolC was unaffected by Mg2+ (Suppl. Fig 2D). Magnesium is critical for LPS stability. LPS levels increased at 200 mM Mg2+ (Suppl. Fig 2E), however, the loss of waaF, lpxA, and lpxC, three key genes involved in LPS biosynthesis, did not influence balofloxacin sensitivity/resistance in the presence of Mg2+ (Suppl. Fig 2F). These findings suggest that magnesium-induced LPS biosynthesis does not contribute directly to BLFX resistance and demonstrate that Mg2+ influx is involved in balofloxacin resistance.”

- Section: MgCl2 affects bacterial metabolism. Authors switched to M9 medium - why? This contrasts with other sections using SWT and should be explained. Also, I cannot evaluate whether the statistical analysis of the data here was performed correctly and was appropriate for this type of experiment. I advise the authors to move the details in lines 166-169 to the Materials and Methods and replace this section instead with a more accessible description of the statistical analysis that a non-expert would be able to appreciate. Furthermore, analysis of Figure 3A indicates that the levels of asparagine, 4-hydroxybutyric acid, uracil, cystathionine, fumaric acid, and aminoethanol have significantly changed at high MgCl2, but these are not mentioned in the text. I suggest the authors mention these if they are relevant to the 12 enriched pathways, especially the biosynthesis of fatty acids. 

We appreciate this comment! 

We indicate the reason we use M9 medium as following:

“To better understand how magnesium affects bacterial metabolism” for explaining why the M9 medium was used.”

The information lines 166-169 indicated has been removed to M &M. 

We have carefully examined the abundance of the metabolites and the enriched pathway. Among the listed metabolites, only fumarate is within the enriched pathways. We mention this point in our revised manuscript as following:

“The increase in fatty acid biosynthesis could be partially explained by an imbalanced pyruvate cycle/TCA cycle, in which fumarate levels increased at higher Mg2+ while succinate levels increased at lower Mg2+ (Suppl. Fig 5B). These findings indicated that glycolysis fluxes into fatty acid biosynthesis rather than the pyruvate cycle/TCA cycle. The relevance of fatty acids and BLFX was demonstrated by the observation that exogenous palmitic acid increased bacterial resistance to balofloxacin (Fig 2F). These results suggest that fatty acid metabolism may be critical to magnesium-based phenotypic resistance.”

- Line 211 appears to refer to Figure 4F and should be checked. Similarly in line 216 - appears this should be Figure 4H, and line 218 should be Figure 4H. Line 226: add a reference to Fig 4I (after arcA was decreased). Line 227: what are genes N646_1004 and N646_1885? Based on Fig 4J these are crp - authors should add to line 227. Line 228 appears to refer to Figure 4J, not Figure 4I. Line 229 - should be Figure 4K, not Figure 4I. Line 231 - should be 4L, not 4K. Line 239 - should be 4M.

We appreciate this comment! The text and figure is now matched. 

- Line 312: the descriptions of "11 lipids, 32 lipids, and 53", and then "26 lipids, 52 lipids, and 107 lipids" are not clear at all and should be corrected. 

We appreciate this comment! The sentence is revised as following:

“The abundance of 11, 32, and 53 lipids was increased in 3.125, 50, and 200 mM MgCl2-treated bacteria, respectively, while the abundance of 26, 52, and 107 lipids was decreased in 3.125, 50, and 200 mM MgCl2-treated bacteria, respectively (Suppl. Fig 7C)”

- Line 340. What is the assay the authors are using to measure the levels of the PGS and PSS enzymes? This is not mentioned or clear in this part of the Results.  

We appreciate this comment!  We provide the information in the manuscript as following:

“Levels of PGS and PSS were quantified by ELISA kits according to manufacture’s instruction (Shanghai Fusheng Industrial Co., Ltd., China)”

- Line 372: What is the assay for measuring membrane depolarization? This is not mentioned and I suggest it should be. Line 374: Figure 7B does not show time dependence, only dose dependence, this should be corrected, it is assumed the authors are referring to Fig 7C for the time dependence data. 

We appreciate this comment! We provide the information in the result as following:  

“The voltage-sensitive dye, DiBAC4(3) showed that 12.5–200 mM MgCl2 promoted membrane depolarization in a dose-dependent manner (Fig 6A)”

We also explain how DiBAC4(3) can be used to measure membrane depolarization in the Materials and Methods section as following:

“DiBAC4(3) is a s voltage-sensitive probe that penetrates depolarized cells, binding intracellular proteins or membranes exhibiting enhanced fluorescence and red spectral shift.”

To make it clear the specific figure, we revise the sentence as following:

“Meanwhile, MgCl2 had a dose-dependent (Fig 6B) and time-dependent (Fig 6C) effect on proton motive force (PMF).”

- Line 384: mention how FM5-95 measures membrane permeability. The authors should also clarify how this reagent is used to measure membrane fluidity, and it is not clear if the data for this is presented in Figure 7 - please clarify. Regarding SYTO9 dye experiment: the authors should briefly explain the experimental design - how SYTO9 dye operates and why FACS was chosen. What is labeled with FITC?  

We appreciate this comment! We clarify the reason we use FM5-95 in the Methods and Materials section as following:

“Measurement of fluidity by fluorescence microscopy

Measurement of membrane fluidity is performed as previously described (Wen et al., 2022). Briefly, ATCC33787 were cultured in medium with indicated concentrations of MgCl2, collected and then adjusted to OD 0.6. Aliquot of 100 μL bacteria cells of each sample were diluted to 1 mL and 10 μL (10 mg/mL) FM5-95 (Thermo Fisher

Scientific, USA) was added. FM5-95 is a lipophilic styryl dye that insert into the outer leaflet of bacterial membrane and become fluorescence. This dye preferentially bind to the microdomains with high membrane fluidity(Wen et al., 2022). After incubated for 20 min at 30 ℃ at vibration without light, the sample was centrifuged for 10 min at 12,000 rpm. The pellets were resuspended with 20 μL of 3% NaCI. Aliquot of 2 μL sample was dropped on the agarose slide, and take photos under the inverted fluorescence microscope.”

This data is presented as micrographs in Fig. 6D, which shows the decreased FM5-95 staining with increasing concentrations of MgCl2. We make this description clear with the following revision:

“FM5-95 staining decreased with increasing concentrations of Mg2+, and no staining was observed in the presence of 200 mM Mg2+ (Fig 6D).”

We explain the reason why we use SYTO9 as following:

“SYTO9, a green fluorescent dye that binds to nucleic acid, enters and stains bacteria cells when there is an increase in membrane permeability (Lehtinen et al., 2004; McGoverin et al., 2020). Staining decreased with increasing MgCl2, indicating that bacterial membrane permeability declined in an Mg2+ dose-dependent manner (Fig 6E).”

We didn’t use FACS in this study, while we only analyze the fluorescence distribution with the equipment. To make it clear, we revise the sentence as following:

“After incubated for 15 min at 30 ℃ at vibration without light, the mixtures were filtered and measured by flow cytometry (BD FACSCalibur, USA).”

- Lines 391-397. The statement that palmitic acid shifts the peaks in Figure 7F is not supported by the data. There is essential no change in the major peak position within each MgCl2 concentration set with increasing palmitic acid. For the linolenic acid data, it is clear that linolenic acid increases permeability only at 50 mM MgCl2-this should be mentioned in the text. 

We appreciate this comment! We revise the sentence as following:

“Exogenous palmitic acid also shifted the fluorescence signal peaks to the left in an MgCl2-dependent manner while palmitic acid only slightly shifted the peaks (Fig 6F). In contrast, exogenous linolenic acid shifted the peak to the right in a dose-dependent manner at 50 mM MgCl2 (Fig 6G).” 

- Line 404-405 - as mentioned earlier, the assay for the update of BLFX should be mentioned (if it is done so earlier in the text, then it does not need to be here).  

We appreciate this comment! It has been mentioned in the introduction.  

- Discussion: CpxA/R-OmprF pathway is mentioned here for the first time. Is this one of the pathways modified by MgCl2 as determined during the course of the study? If so, this should be reworded to mention that. If not, the relevance of this particular pathway as it relates to light metals and phenotypic resistance should be discussed.

We appreciate this comment! Since it is not relevant to the discussion of Mg2+ and fatty acid biosynthesis, we delete this sentence in the revised manuscript.  

-The following grammatical errors should be corrected:

-line 55 change to: "genetic mutations; instead, this type of resistance is transient, and bacteria resume normal growth"

-line 57: change to "resistance types are biofilm" 

-line 61: change to "states that significantly" 

-line 63: change to "resistance share the common feature in they retard or even cease in the presence" 

-line 65: change to "resistance that allow bacteria to proliferate" 

-line 81: change "But whether" to "Whether" 

-line 178: change to "may be critical to the Mg-based phenotypic resistance"

-line 86: change to "Marine environments and agriculture are rich in magnesium, where..." 

-line 93: change in to vs

-line 154: insert space after metabolism 

-line 158: change 'identified" to "focused on the levels of" 

-line 160: change "The levels of forty-one metabolites" 

-line 198: change shared to share 

-line 310: increased is duplicated, delete one 

-line 451: add "the" before ratio 

-line 453: gram should be capitalized 

-line 462: "the regulation" should be reworded to "More importantly, the effect of exogenous MgCl targets the..." 

-line 469: add dash between Mg2+ and limited

-line 478: change "the crucial" to "a crucial" 

-there are numerous locations in the manuscript where the word "magnetism" is used when clearly the word is supposed to be magnesium - this should be corrected

We appreciate this comment! These have been corrected or revised. 

Editors comments:

Page 2 line 27; Page 25 line number 426; page 27 line number 481: In the abstract and discussion, only Vibrio alginolyticus was mentioned, even though two Vibrio species were used in the study. It would be helpful to understand the rationale behind the focus on this particular species.

We appreciate this comment! We have revised the introduction to provide additional information as following:

“Vibrios inhabit seawater, estuaries, bays, and coastal waters, regions full of metal ions such as magnesium (Kumarage et al., 2022). Magnesium is the second most dissolved element in seawater after sodium. At a salinity of 3.5% seawater, the magnesium concentration is about 54 mM (Potis, 1968), and in deep seawater, can be as high as 2,500 mM (Wang et al., 2024). Vibrio parahaemolyticus and V. alginilyticus are two representative Vibrio pathogens that infect humans and aquatic animals, resulting in illness and economic loss, respectively (Grimes, 2020). (Fluoro)quinolones such as balofloxacin are used to treat Vibrio infection, however, resistance has emerged due to overuse (Suyamud et al., 2024). Indeed, (fluoro)quinolones are one of China's two primary residual chemicals associated with aquaculture (Liu et al., 2017). Vibrio can develop quinolone resistance through mutations in the DNA gyrase gene or through plasmid-mediated mechanisms (Dutta et al., 2021). Thus, the use of V. parahaemolyticus and V. alginilyticus as bacterial representatives, and balofloxacin as a quinolone-based antibacterial representative, can help to define novel magnesium-dependent phenotypic resistance mechanisms of pathogenic Vibrio species.”

On Page 2, line 34: The abstract contains some undefined abbreviations, such as 'PE' and 'PG', which should be explained. 

We appreciate this comment! We explain the PE and PG in the revised abstract as following:

“phosphatidylethanolamine (PE) biosynthesis is reduced and phosphatidylglycerol (PG)”

On Page 2, line 31-32: For the statement "Exogenous supplementation of fatty acids confirm the role of fatty acids in antibiotic resistance…" it would be beneficial to specify whether the fatty acids were saturated or unsaturated. 

Response, We appreciate this comment! We revise the sentence as following:

“Exogenous supplementation of unsaturated and saturated fatty acids increased and decreased bacterial susceptibility to antibiotics, respectively, confirming the role of fatty acids in antibiotic resistance.”

The potential effects of the specific ions (SO4 and Cl2) present in the Mg2SO4 and MgCl2 compounds used in the study were not discussed. It would be useful to understand if these ions had any influence on the observed outcomes.

We appreciate this comment! We revise the sentence as following:

“However, the MIC for BLFX was higher in ASWT medium supplemented with Mg2SO4 or MgCl2 than in LB medium (Fig 1B). And Mg2SO4 or MgCl2 had no

difference on MIC, suggesting it is Mg2+ not other ions contribute to the MIC change.”

On Page 8, line 141: The heading of Figure 2, "Mg2+ elevates intracellular Mg2+," seems redundant and could be revised for clarity or modified. 

We appreciate this comment! Figure 2 is now moved to supplementary figure as Suppl. Fig 2. The title is revised as following:

“Figure 2. Mg2+ decreases balofloxacin uptake.”

On Page 4, line 91: some terms/abbreviations, such as 'LB' and 'M9,' require expansion or definition to ensure the reader's understanding.

We appreciate this comment! We include the expansion for LB and M9 in the  revised manuscript as following:

“Luria-Bertani medium (LB medium)” and “M9 minimal medium (M9 medium)”

Page 4, line 92: The real seawater composition used in the experiments should be supported by a reference.

We appreciate this comment! We provide the reference in the revised manuscript as following:

““artificial seawater” (ASWT) medium that included the major ion species in marine water (Wilson, 1975) (LB medium plus 210 mM NaCl, 35 mM Mg2SO4, 7 mM KCl, and 7 mM CaCl2)”

Page 4 line, number 93: the he full names of the bacterial strains (e.g., ATCC33787 and VP01) should be provided instead of just the strain numbers.

We appreciate this comment! We revised the sentence as following:

“To investigate whether this mineral impacts antibiotic activity, the minimal inhibitory concentration (MIC) of V. alginolyticus ATCC33787 and V. parahaemolyticus VP01, which we referred as ATCC33787 and VP01 afterwards,”

Finally, there appears to be a potential contradiction between the statements on page 12, lines 211-212 and 214-216, regarding the effects of Mg2+ on the synthesis of unsaturated fatty acids. Further explanation may be needed to reconcile these seemingly contradictory points.

We appreciate this comment! For line 221-226, which was previously line 211-212, is about the gene expression for fatty acid biosynthesis. While, Line 228 and 233, which was previously line 214-216 is about the gene expression for fatty acid degradation. We agree that the previous description is a little bit confuse. We revise the sentence to emphasize that we focus on fatty acid degradation so that the readers can tell them apart. 

In the text, we revised it as following:

“In addition, we also quantified gene expression during fatty acid degradation to determine whether Mg2+ affects this process”  In the figure legend, we also indicate that 

“H. qRT-PCR for the expression of genes encoding fatty acid degradation in the absence or presence of the indicated concentrations of MgCl2”

eLife Assessment

The study explored the influence of magnesium on phenotypic antibiotic resistance in two strains of Vibrios: V. alginolyticus ATCC33787 and V. parahaemolyticus VP01. This research is fundamental for revealing the phenotypic antibiotic resistance mechanism utilized by the specified model bacteria in elevated levels of magnesium. The study produced convincing evidence indicating that in high concentrations of magnesium, the efficacy of selected antibiotics was diminished due to decreased biosynthesis of unsaturated fatty acids and PE, along with an increase in the biosynthesis of PG.

Reviewer #1 (Public review):

Summary:

In the manuscript entitled "Magnesium modulates phospholipid metabolism to promote bacterial phenotypic resistance to antibiotics", Li et al demonstrated the role of magnesium in promoting phenotypic resistance in V. alginolyticus. Using standard microbiological and metabolomic techniques, the authors have shown the significance of fatty acid biosynthesis pathway behind the resistance mechanism. This study is significant as it sheds light on the role of an exogenous factor in altering membrane composition, polarization and fluidity which ultimately leads to antimicrobial resistance.

Strengths:

Authors have used different approaches to demonstrate the effect of Mg+2 on drug resistance in Vibrio alginolyticus. The revised version of the manuscript is much improved, with a very informative introduction and a variety of methodologies with clear explanation of the experiments performed. Also, additional experiments were performed as suggested by the reviewers which certainly enhanced the quality of the paper. I believe the findings of this study will be of high impact in the bacterial community.

Weaknesses:

There are a few grammatical mistakes.

Comments on revisions:

The authors have done a comprehensive job of addressing all my concerns in their revised version.

Reviewer #2 (Public review):

Summary:

In this study, the authors aimed to identify if and how magnesium affects the ability of two particular bacteria species to resist the action of antibiotics. In my view, the authors succeeded in their goals and present a compelling study that will have important implications for the antibiotic resistance research community. Since metals like magnesium are present in all lab media compositions and are present in the host, the data presented in this study certainly will inspire additional research by the community. These could include research into whether other types of metals also induce multi-drug resistance, whether this phenomenon can be observed in other bacterial species, especially pathogenic species that cause clinical disease, and whether the underlying molecular determinants (i.e. enzymes) of metal-induced phenotypic resistance could be new antimicrobial drug targets themselves.

Strengths:

This study's strengths include that the authors used a variety of methodologies, all of which point to a clear effect of exogenous Mg2+ on drug resistance in the targeted species. I also comment the authors for carrying out a comprehensive study, spanning evaluation of whole cell phenotypes, metabolic pathways, genetic manipulation, to enzyme activity level evaluation. The fact that the authors uncovered a molecular mechanism underlying Mg2+-induced phenotypic resistance is particularly important as the key proteins should be studied further.

Weaknesses:

I thank the authors for improving their manuscript based on my previous suggestions. I still believe the Results section is long and bogs down at times.

In general, the conclusions drawn by the authors are justified by the data, except for the interpretation of some experiments. Importantly, this paper has discovered new antimicrobial resistance mechanisms and has also pointed to potential new targets for antimicrobials.

Comments on revisions:

I just wanted to thank the authors for addressing most of my previous comments.

eLife Assessment

This valuable study investigates the role of Complement 3a Receptor 1 (C3aR) in the pathogenesis of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) using mouse models with specific target deletions in various cell types. While the general relevance of C3aR in inflammatory contexts has been established before, the authors provide solid evidence here that C3aR does not contribute significantly to MASLD pathogenesis in their models. The work will be of interest to colleagues studying diseases of the liver and the intersection with inflammation.

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.

Comments on revisions:

The revised version fulfilled my queries.

Reviewer #2 (Public review):

Summary:

Homan et al. examined the effect of macrophage- or Kupffer cell-specific C3aR1 KO on MASLD/MASH-related metabolic or liver phenotypes.

Strengths:

Established macrophage- or Kupffer cell-specific C3aR1 KO mice, and showing comparable liver metabolic phenotypes between WT and macrophage-specific C3aR1KO mice in response to normal chow diet or MASH diet feeding.

Weaknesses:

Insufficient data showing the effects of C3aR1KO on liver macrophage phenotypes, such as hepatic macrophage profiles, macrophage activation status, etc, which are important for the development of liver steatosis and fibrosis.

Author response:

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

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. These experimental models have been clearly defined in the revised manuscript on pages 5 and 7 and in the methods section (page 10). The reviewer’s point is well taken that the 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. This has been added to the discussion (page 9).

Reviewer #2 (Public review):

Summary:

Homan et al. examined the effect of macrophage- or Kupffer cell-specific C3aR1 KO on MASLD/MASH-related 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). This point has been added to the text on page 5.

(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. This has been added to the discussion (page 8).

(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 have included additional data in the revised manuscript in Figure S4. These results show no to minimal development of a MASLD/MASH gene signature.

(5) Would C3aR1KO result in differences in liver phenotypes, including macrophage population/activation, liver inflammation, lipogenesis, in lean mice?

We have provided additional data further characterizing liver inflammation, lipogenesis and macrophages in macrophage C3ar1 KO mice under lean/regular diet conditions in Figure 2K. These results show a potential trend but no substantial development of a MASLD/MASH gene signature.

(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 has been clarified and explicitly stated in the revised manuscript on page 7 and in the methods section.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

These data should be repeated using a more established model of Kupffer cell target deletion via Clec4-F mice.

Our data with Kupffer cell C3ar1 deletion is indeed done with Clec4f-Cre transgenic mice. This has been clarified in the revised manuscript on page 7 and in the methods section.

Reviewer #2 (Recommendations for the authors):

(1) Typo: "iver" in the abstract

(2) Line 97, "GAN diet I" should be "GAN diet"?

These points have been corrected in the revised manuscript.

eLife Assessment

This is an important study suggesting that neuron-specific loss of function of the RNA splicing factor Ptbp1 in striatal neurons induces dopaminergic markers and alleviates motor defects in a 6-hydroxydopamine (6-OHDA) mouse model of Parkinson's Disease. The evidence supporting the rescue of motor deficits following Ptbp1 manipulation is solid, and, while additional characterization of dopaminergic neuronal identity may be required in future studies, these results have clear implications for Parkinson's disease therapeutics. The study also addresses recent controversial literature on cell reprogramming in Parkinson's Disease and will be of interest to researchers with a focus on the application of gene therapy to rescue neurodegeneration.

Reviewer #2 (Public review):

Summary:

The manuscript by Bock and colleagues describes generation of an AAV-delivered adenine base editing strategy to knockdown PTBP1 and the behavioral and neurorestorative effects of specifically knocking down striatal or nigral PTBP1 in astrocytes or neurons in a mouse model of Parkinson disease. The authors found that knocking down PTBP1 in neurons, but not astrocytes, and in striatum, but not nigra, results in the phenotypic reorganization of neurons to TH+ cells sufficient to rescue motor phenotypes, though insufficient to normalize responses to dopaminomimetic drugs.

Strengths:

The manuscript is well-written and adds to the growing literature challenging previous findings by Qian et al., 2020 and Zhou et al., 2020 indicating that astrocytic downregulation of PTBP1 can induce conversion to dopaminergic neurons in the midbrain and improve parkinsonian symptoms. The base editing approach is interesting and potentially more therapeutically relevant than previous approaches.

Weaknesses:

The animal model utilized, the 6-OHDA model, though useful to examine dopaminergic cell loss, exhibits accelerated neurodegeneration and none of the typical pathological hallmarks (synucleinopathy, Lewy bodies, etc.) compared to the typical etiology of Parkinson disease, limiting its translational interpretation. The identity of the converted neurons is unclear. Though the immunohistochemical methodology indicates they may be MSNs and/or interneurons, a more comprehensive identity is still lacking. There remains no real evidence that these cells actually release dopamine. Since striatal dopamine was assessed by whole-tissue analysis, which is not necessarily reflective of synaptic dopamine availability, it is difficult to assess whether the ~10% increase in TH+ cells in the striatum was sufficient to improve dopamine function. However, the improvement in motor activity suggests that it was.

Reviewer #3 (Public review):

This study explores the use of an adenine base editing strategy to knock down PTBP1 in astrocytes and neurons of a Parkinson's disease mouse model, as a potential AAV-BE therapy. The results indicate that editing Ptbp1 in neurons, but not astrocytes, leads to the formation of tyrosine hydroxylase (TH)+ cells, rescuing some motor symptoms.

Several aspects of the manuscript stand out positively. Firstly, the clarity of the presentation. The authors communicate their ideas and findings in a clear and understandable manner, making it easier for readers to follow.

The Materials and Methods section is well-elaborated, providing sufficient detail for reproducibility.

The logical flow of the manuscript makes sense, with each section building upon the previous one coherently.

The ABE strategy employed by the authors appears sound, and the manuscript presents a coherent and well-supported argument.

Positively, some of the data in this study effectively counteracts previous work in line with more recent publications, demonstrating the authors' ability to contribute to the ongoing conversation in the field.

Comments on revisions:

The authors have adequately addressed all the previous questions and suggestions by providing new data and/or adding necessary clarifications and deeper discussions. The newly presented data convincingly fills the gaps identified during the initial review process. The additional discussions and clarifications enhance both the clarity and transparency of the manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review): 

Summary: 

Recent years have seen spectacular and controversial claims that loss of function of the RNA splicing factor Ptbp1 can efficiently reprogram astrocytes into functional neurons that can rescue motor defects seen in 6-hydroxydopamine (6-OHDA)-induced mouse models of Parkinson's disease (PD). This latest study is one of a series that fails to reproduce these observations, but remarkably also reports that neuronal-specific loss of function of Ptbp1 both induces expression of dopaminergic neuronal markers in striatal neurons and rescues motor defects seen in 6-OHDA-treated mice. The claims, if replicated, are remarkable and identify a straightforward and potentially translationally relevant mechanism for treating motor defects seen in PD models. However, while the reported behavioral effects are strong and were collected without sample exclusion, other claims made here are less convincing. In particular, no evidence that Ptbp1 loss of function actually occurs in striatal neurons is provided, and the immunostaining data used to claim that dopaminergic markers are induced in striatal neurons is not convincing. Furthermore, no characterization of the molecular identity of Ptbp1-deficient striatal neurons is provided using single-cell RNA-Seq or spatial transcriptomics, making it difficult to conclude that these cells are indeed adopting a dopaminergic phenotype. 

Overall, while the claims of behavioral rescue of 6-OHDA-treated mice appear compelling, it is essential that these be independently replicated as soon as possible before further studies on this topic are carried out. Insights into the molecular mechanisms by which neuronalspecific loss of function of Ptbp1 induces behavioral rescue are lacking, however. Moreover, the claims of induction of neuronal identity in striatal neurons by Ptbp1 require considerable additional work to be convincing.

We thank the reviewer for the detailed analysis of our study. Please find our answers to the points raised by the reviewer below in blue.

Strengths of the study: 

(1) The effect size of the behavioral rescue in the stepping and cylinder tests is strong and significant, essentially restoring 6-OHDA-lesioned mice to control levels.

(2) Since the neurotoxic effects of 6-OHDA treatment are highly variable, the fact that all behavioral data was collected blinded and that no samples were excluded from analysis increases confidence in the accuracy of the results reported here. 

We appreciate the reviewer’s feedback and acknowledgement of the strengths of our study. We undertook several optimization steps in the surgery, post-operative care, and handling of the animals for behavior experiments to ensure high reproducibility of our experiments.

Weaknesses of the study:  

(1) Neurons express relatively little Ptbp1. Indeed, cellular expression levels as measured by scRNA-Seq are substantially below those of astrocytes and other non-neuronal cell types, and Ptbp1 immunoreactivity has not been observed in either striatal or midbrain neurons (e.g. Hoang, et al. Nature 2023). This raises the question of whether any recovery of Th expression is indeed mediated by the loss of function of Ptbp1 rather than by off-target effects. AAVmediated rescue of Ptbp1 expression could help clarify this.

In the original manuscript, we delivered control vectors that only express the ABE to 6-OHDAlesioned mice (labeled as AAV-ctrl) and did not detect TH positive cells in the midbrain or striatum of control mice or rescue of spontaneous motor skills. We can therefore exclude that the delivery procedure, AAV-PHP.eB capsid, or ABE expression caused adverse effects leading to induction of TH expression and functional rescue of spontaneous motor behaviors in PD mice. To further exclude that these effects were caused by off-target editing, we experimentally determined off-target binding sites of our sgRNA (sgRNA-ex3) using GUIDEseq and subsequently analyzed these sites in treated animals by NGS (Figure 3 – supplement 3). While two off-target sites were identified, it is unlikely that base editing at these sites caused the observed phenotypes. One off-target site was identified in the myopalladin (Mypn) gene, which encodes for a muscle-specific protein that plays a role in regulating the structure and growth of skeletal and cardiac muscle (Filomena et al., 2021, 2020).  The other site is not located in a coding region, but in an intron of the ankyrin-1 (Ank1) gene, encoding for an adaptor protein linking membrane proteins to the underlying cytoskeleton (Cunha and Mohler, 2009). Even though this gene is also expressed in neurons, base editing within this intronic region did not lead to changes in transcript levels (Figure 3 – supplement 3). Thus, the induction of TH expression upon adenine base editing with sgRNA-ex3 is likely a direct consequence of PTBP1 downregulation.

Further supporting this conclusion, in the revised manuscript we additionally show PTBP1 downregulation at the RNA and protein level in the SNc and striatum after base editor treatment (Figure 2 – figure supplement 5; figure 3 – supplement 2).

(2) It is not clear why dopaminergic neurons, which are not normally found in the striatum, are observed following Ptbp1 knockout. This is very similar to the now-debunked claims made in Zhou, et al. Cell 2020, but here performed using the hSyn rather than GFAP mini promoter to control AAV expression. While this is the most dramatic and potentially translationally relevant claim of the study, this claim is extremely surprising and lacks any clear mechanistic explanation for why it might happen in the first place.  

We agree with the reviewer that our study does not provide mechanistic insights into how Ptbp1 downregulation in neurons leads to the induction of dopaminergic markers in the striatum. As we believe that this is not within the scope of a revision, we discuss potential follow-up experiments in the discussion section of the revised manuscript.

This observation is even more surprising in light of reports that antisense oligonucleotidemediated knockdown of Ptbp1, which should have affected both neuronal and glial Ptbp1 expression, failed to induce expression of dopaminergic neuronal markers in the striatum (Chen, et al. eLife 2022). Selective loss of function of Ptbp1 in striatal and midbrain astrocytes likewise results in only modest changes in gene expression. 

Using 6-OHDA lesioned Aldh1l1-CreERT2;Rpl22lsl-HA mice, the Chen et al. study (eLife 2022) assessed potential astrocyte to neuron conversion by quantifying the presence of HA-labeled neurons after ASO-mediated knockdown of Ptbp1. Even though they did not detect HApositive neurons in the SNc, suggesting absence of astrocyte to neuron conversion, the images in Figure 4D reveal TH positive cells in the lesioned hemisphere, similar to our observations in Figure 2B-D. While it cannot be excluded that these TH positive cells are remnants from an incomplete 6-OHDA lesion, they could also be endogenous neurons with induced expression of dopaminergic markers after ASO-mediated knockdown of Ptbp1. Furthermore, Chen et al. performed the apomorphine test to assess changes in motor skills, which did not reveal an improvement in our study either.

It is critically important that this claim be independently replicated, and that additional data be provided to conclusively show that striatal neurons are indeed expressing dopaminergic markers.

Our behavior and immunofluorescence experiments involving mice injected into the striatum were performed with two independently generated cohorts of 6-OHDA mice. In detail, the 6OHDA mice were generated by two independent surgeons from different labs (>6 months between experiments of these cohorts), leading to comparable behavioral outcomes before and after treatment. Subsequent behavior and immunofluorescence experiments with each cohort were performed and analyzed by two independent and blinded researchers, showing comparable results.

(3) More generally, since multiple spectacular and irreproducible claims of single-step glial-toneuron reprogramming have appeared in high-profile journals in recent years, a consensus has emerged that it is essential to comprehensively characterize the identity of "transformed" cells using either single-cell RNA-Seq or spatial transcriptomics (e.g. Qian, et al. FEBS J 2021; Wang and Zhang, Dev Neurobiol 2022). These concerns apply equally to claims of neuronal subtype conversion such as those advanced here, and it is essential to provide these same datasets. 

In the revised version, we have analyzed the expression of additional neuronal markers in TH positive cells of the striatum using 4i imaging. Briefly, our results showed that the vast majority of TH-expressing cells also expressed the markers DAT and NEUN, further corroborating the neuronal and dopaminergic identity of these cells. Additional analysis revealed that this TH/DAT/NEUN expressing cell population expressed markers of GABAergic neurons, either of medium spiny neurons (~50%) and various types of interneurons (~50%). While our 4i analysis has allowed us to broadly classify these TH-expressing populations, we agree that detailed transcriptional analysis at the single cell level is required to understand the molecular mechanisms underlying the generation of TH positive cells. These analyses are, however, not within the scope of a revision and would require a thorough dedicated study. We have added these results and discussion points to the revised manuscript.

(4) Low-power images are generally lacking for immunohistochemical data shown in Figures 3 and 4, which makes interpretation difficult. DAPI images in Figure 3C do not appear nuclear. Immunostaining for Th, DAT, and Dcx in Figure 4 shows a high background and is difficult to interpret. 

We thank the reviewer for closely evaluating these images and suggestions for improvement. In the revised manuscript, we provide low power images and higher magnification insets as requested to allow for easier interpretation.

(5) Insights into the mechanism by which neuronal-specific loss of Ptbp1 function induces either functional recovery, or dopaminergic markers in striatal neurons, is lacking.

In the revised manuscript, we provide a more detailed discussion of mechanisms that could potentially be involved in the functional recovery or expression of dopaminergic markers. However, deciphering the exact molecular mechanisms underlying these observations requires thorough transcriptional analysis at the single cell level, which is out of scope of this revision.

Reviewer #2 (Public Review):

Summary: 

The manuscript by Bock and colleagues describes the generation of an AAV-delivered adenine base editing strategy to knockdown PTBP1 and the behavioral and neurorestorative effects of specifically knocking down striatal or nigral PTBP1 in astrocytes or neurons in a mouse model of Parkinson's disease. The authors found that knocking down PTBP1 in neurons, but not astrocytes, and in striatum, but not nigra, results in the phenotypic reorganization of neurons to TH+ cells sufficient to rescue motor phenotypes, though insufficient to normalize responses to dopaminomimetic drugs.

Strengths: 

The manuscript is generally well-written and adds to the growing literature challenging previous findings by Qian et al., 2020 and Zhou et al., 2020 indicating that astrocytic downregulation of PTBP1 can induce conversion to dopaminergic neurons in the midbrain and improve parkinsonian symptoms. The base editing approach is interesting and potentially more therapeutically relevant than previous approaches.

Weaknesses: 

The manuscript has several weaknesses in approach and interpretation. In terms of approach, the animal model utilized, the 6-OHDA model, though useful to examine dopaminergic cell loss, exhibits accelerated neurodegeneration and none of the typical pathological hallmarks (synucleinopathy, Lewy bodies, etc.) compared to the typical etiology of Parkinson's disease, limiting its translational interpretation. 

We thank the reviewer for the detailed assessment of our study and pinpointing its current weaknesses. Please find our answers to all comments below in blue.

We agree with the reviewer that the 6-OHDA model lacks the typical pathological hallmarks of PD. Nevertheless, we chose this model for two reasons:

i) The 6-OHDA model was used by both Qian et al. (2020) and Zhou et al. (2020). To allow comparison of our results to these studies, it was crucial to use the same model. Notably, the 6-OHDA model was also used by Chen et al. (2022) and Hoang et al. (2023) for comparison to the two studies from 2020.

ii) The 6-OHDA model is straightforward to generate and displays robust motor impairments for evaluation of potential therapeutic effects of neuroregeneration treatment approaches. We therefore believe that the model is well-suited to analyze the cellular and behavioral effects (specifically motor skills) of PTBP1 downregulation. 

In future studies, it would be critical to include models that also display typical pathological hallmarks of the disease to further evaluate the therapeutic effect of this base editing approach. These experiments are, however, not within the scope of this study, which was aimed to focus on the cellular and behavioral effects of PTBP1 downregulation. 

In addition, there is no confirmation of a neuronal or astrocytic knockdown of PTBP1 in vivo; all base editing validation experiments were completed in cell lines. 

In the revised manuscript, we assess in vivo base editing efficiencies at the Ptbp1 target site in the SNc (AAV-hsyn, 15.6%) and striatum (AAV-hysn, 21.1%). Furthermore, we assessed in vivo Ptbp1 downregulation at the RNA and protein level to complement our in vitro data (Figure 2 – figure supplement 5; figure 3 – supplement 2).

Finally, it is unclear why the base editing approach was used to induce loss-of-function rather than a cell-type specific knockout, if the goal is to assess the effects of PTBP1 loss in specific neurons. 

We expressed base editors under cell-type specific promoter to induce a reliable loss-offunction mutation at the Ptbp1 exon-intron junction in neurons or astrocytes. Performing these mutations with Cas9 nucleases instead would have had potential limitations and risks, including i) indel mutations do not always lead to a frameshift and loss-of-function despite high indel formation at the targeted site, ii) nucleases induce DNA double strand breaks, which can have serious side effects (e.g. chromosomal rearrangements or translocations), and iii) ‘mosaicisms’ as edited cells contain different indel mutations, which may result in different effects and thus complicate analysis of the downstream effects. We discuss these points in the revised manuscript.  

In terms of interpretation, the conclusion by the authors that PTBP1 knockdown has little likelihood to be therapeutically relevant seems overstated, particularly since they did observe a beneficial effect on motor behavior. We know that in PD, patients often display negligible symptoms until 50-70% of dopaminergic input to the striatum is lost, due to compensatory activity of remaining dopaminergic cells. Presumably, a small recovery of dopaminergic neurons would have an outsized effect on motor ability and may improve the efficacy of dopaminergic drugs, particularly levodopa, at lower doses, averting many problematic side effects. Since striatal dopamine was assessed by whole-tissue analysis, which is not necessarily reflective of synaptic dopamine availability, it is difficult to assess whether the ~10% increase in TH+ cells in the striatum was sufficient to improve dopamine function. However, the improvement in motor activity suggests that it was.

As pointed out by the reviewer, it is difficult to estimate the therapeutic effect and importance of a ~10% increase in TH+ cells for PD patient. Guided by the reviewer’s suggestion, we have included a more in-depth discussion of our results and its potential therapeutic value as well as outstanding questions for future studies in the revised manuscript.

Reviewer #3 (Public Review):

This study explores the use of an adenine base editing strategy to knock down PTBP1 in astrocytes and neurons of a Parkinson's disease mouse model, as a potential AAV-BE therapy. The results indicate that editing Ptbp1 in neurons, but not astrocytes, leads to the formation of tyrosine hydroxylase (TH)+ cells, rescuing some motor symptoms.

Several aspects of the manuscript stand out positively. Firstly, the clarity of the presentation. The authors communicate their ideas and findings in a clear and understandable manner, making it easier for readers to follow. 

The Materials and methods section is well-elaborated, providing sufficient detail for reproducibility. 

The logical flow of the manuscript makes sense, with each section building upon the previous one coherently.

The ABE strategy employed by the authors appears sound, and the manuscript presents a coherent and well-supported argument.

Positively, some of the data in this study effectively counteracts previous work in line with more recent publications, demonstrating the authors' ability to contribute to the ongoing conversation in the field.

We thank the reviewer for appreciating the effort we have put into this study. Please find below a point-by-point reply to the weaknesses raised by the reviewer. 

However, while the in vitro data yields promising results, it may have been overly optimistic to assume that the efficiencies observed in dividing cells will directly translate to in vivo conditions. This consideration is important given the added complexities of vector optimization, different cell types targeted in vitro versus in vivo, as well as unknown intrinsic limitations of the base editing technology. 

We agree with the reviewer that in vitro base editing efficiencies might not directly translate to in vivo editing outcomes. We therefore assessed in vivo base editing efficiencies at the Ptbp1 locus and PTBP1 downregulation in the striatum and midbrain. Our data revealed that in vivo base editing activity was lower than in our in vitro setting (in vitro: Figure 1; figure 1 – figure supplement 2; in vivo: figure 2 – figure supplement 5; figure 3 – supplement 2). However, we believe that these rates are slightly underestimated since we sequenced DNA isolated from the whole tissue (striatum or SNc) and not from purified astrocytes or neurons. Moreover, we could demonstrate that editing led to a reduction of Ptbp1 transcript and PTBP1 protein level (Figure 2 – figure supplement 5; figure 3 – supplement 2).

In addition, certain aspects of the manuscript would benefit from a more in-depth and comprehensive discussion rather than being only briefly touched upon. Such a discussion would enhance the relevance of the obtained results and provide the foundation for improvement when using similar approaches.

Following the reviewer’s suggestion, we included a more in-depth discussion of our results in the revised manuscript.

Recommendations for the authors:

Reviewing Editor (Recommendations for the Authors):

A summary of key recommendations that might improve the eLife assessment in a subsequent submission are provided below, as a guide to help the authors focus on changes that might enhance the strength of evidence (e.g., from "incomplete" to "solid").

(1) Provide further explanation of the mechanistic relationship between the downregulation of Ptbp1 and TH+ dopaminergic neuron reprogramming. Additional discussion of this topic should also be included.

(2) Demonstrate proof of editing in the intended targeted cells in vitro and/or in vivo.

(3) Show evidence of successful Base Editor delivery in vivo.

(4) Perform a deeper characterization of TH+ cells in vivo and provide a more thorough discussion of the identity of the targeted cells. This may include an exploration of whether TH+ cells detected are TH+ interneurons and/or establish their identity based on transcriptomics or a similar approach.

(5) Provide better-quality representative images supporting the quantitative data.

(6) Please include full statistical reporting including exact p-values wherever possible alongside the summary statistics (test statistic and df) and 95% confidence intervals. These should be reported for all key questions and not only when the p-value is less than 0.05 in the main manuscript.

In the revised manuscript, we provided 1) suggestions of the mechanistic relationship between Ptbp1 knockdown, dopamine synthesis, and the functional rescue of spontaneous behaviors, 2) proof of in vivo base editing and successful base editor delivery, 3) deeper characterization of TH-expressing cells in vivo using 4i imaging, 4) better quality images, and 5) full statistical reporting.  

Individual Reviewer recommendations for the authors are included below.

Reviewer #1 (Recommendations For The Authors):

Confirm loss of Ptbp1 function in infected striatal neurons. Single-cell RNA-Seq or spatial transcriptomic analysis must be performed to characterize the identity of the edited striatal neurons. The quality of the immunostaining in Figures 3 and 4 needs to be improved, and lowpower images provided. Were eLife a conventional journal, I would have insisted on all these being included prior to publication. Please also arrange for independent replication of the behavioral rescue and induction of dopaminergic marker gene expression in the striatum. 

In the revised manuscript, we confirmed Ptbp1 downregulation at the tissue level in the SNc and striatum by RT-qPCR and western blot and included low-power images for easier interpretation. Additionally, we assessed expression of additional neuronal markers on striatal sections using 4i imaging and found that TH/DAT/NEUN positive populations either expressed markers of medium spiny neurons or interneurons. We have included these results in the revised manuscript.

Our behavioral and imaging experiments involving mice injected into the striatum were in fact performed with two independently generated cohorts of 6-OHDA mice. In detail, the 6OHDA mice were generated by two independent surgeons from different labs (>6 months between experiments of these two cohorts), leading to comparable behavioral outcomes before and after treatment. The experiments with each cohort were performed and analyzed by two independent and blinded researchers, yielding comparable results. 

Reviewer #2 (Recommendations For The Authors):

(1) In the introduction, lines 43-45: This statement is inaccurate. Current treatment strategies do not focus on slowing or halting disease progression. There is currently no accepted therapy that does this. Dopaminergic therapies and deep brain stimulation can compensate for circuitry dysfunction as a result of dopamine cell loss but do not slow the disease. The referenced paper used is older and does not refer to new treatments for PD and is a summary article for a special issue of the Disease Models and Mechanisms journal. Please ensure that all references used are appropriate for the statement they are attached to.

We thank the reviewer for pointing this out. We have rephrased this statement accordingly and provided an appropriate reference describing current treatment strategies.

(2) The number of TH+ cells in the intact nigra seems low compared to published data. Suggest a stereological approach may be better than the Abercrombie method.

Following the reviewer’s suggestion, we re-quantified the number of TH positive cells using a stereological approach (Nv:Vref method). We have included these results in the revised manuscript. 

(3) Have the authors considered that the striatal TH+ cells could be TH+ striatal interneurons? 

In the revised manuscript, we performed additional 4i imaging experiments to further analyze the identity of the TH positive cells in the striatum. Briefly, we found that TH/DAT/NEUN positive populations either expressed markers of GABAergic medium spiny neurons or interneurons. We have added these results to the revised manuscript (Figure 4). 

(4) The Western blot shown in Figure 1 C for C8-D1A has some abnormalities and makes it difficult to judge the bands. Also, for 1B, the legends are difficult to see.

In the revised manuscript, we have repeated the respective western blot to make interpretation of the bands easier, and adapted the legends in Figure 1B for better visibility.

(5) Figure 2: Please show representative images for the GFAP-targeted editing.

Representative images of the GFAP-targeted groups can be found in Figure 2 – figure supplement 3.

(6) Figure 2, Supplement 3: Please include quantification.

The quantifications for these images can be found in Figure 2D and 2F. 

(7) Figure 1, Supplement 2: The gene name in A is misspelled.

Thank you for point this out. In the revised manuscript, we added the correct gene name.

(8) Line 267-276: As previously indicated, the statement here is overstated based on the data provided. In addition, the citation provided to justify this claim (Kannari et al., 2000) is an odd choice as the dosage of L-DOPA utilized was not therapeutically relevant (50 mg/kg). A better indication of efficacy would be the return to basal, unaffected levels rather than the fold increase in dopamine levels. A better comparison would be Lindgren et al., 2010 who showed that L-DOPA-treated animals with a physiologically relevant dose (6 mg/kg) that did not induce dyskinesia, showed a return to basal, non-lesioned dopamine levels in the striatum after LDOPA by microdialysis. To really support this claim, the authors would need to use an approach that could measure synaptic dopamine availability, rather than whole-tissue dopamine levels, such as microdialysis, fiber photometry, or an equivalent.

Following the reviewer’s suggestions, we replaced this reference with Lindgren et al. (2010) and provide a more detailed interpretation of our results and remaining questions for future studies.  

Reviewer #3 (Recommendations For The Authors):

Major and minor issues are discussed below by section.

INTRODUCTION and AIM - Lines 36-73

- The authors effectively contextualize the aim of their study by providing comprehensive background information on previous research regarding cell 'reprogramming' into dopaminergic neurons in the SNc. However, the introduction lacks contextualization of TH+ cells and PD. For readers who may not be well-versed in the Parkinson's field, understanding the importance of TH (Tyrosine Hydroxylase) may be challenging, since the term "TH+ cells" is mentioned only once by the end of the introduction (line 71), to then become a key element in the entire study.

- Providing a brief explanation of the role of Tyrosine Hydroxylase in the synthesis of L-DOPA would facilitate the reader's comprehension of why the presence of TH+ cells following Base Editing treatment is relevant.

- Further elaboration on the relationship between the downregulation of the general RNA binding protein, PTBP1, and the specific dopaminergic-related readout, TH, would improve coherence and strengthen the linkage between the introductory section and the results.

We thank the reviewer for the constructive suggestions. In the introduction of the revised manuscript, we describe the meaning and importance of TH in the context of dopamine synthesis and PD. Likewise, we briefly outlined the importance of the PTBP1/nPTBP regulatory loops during neuronal differentiation and maturation. 

RESULTS 

Result Section 1 - Line 75-109

- Thorough screening of sgRNAs targeting splice junctions across the Ptbp1 gene in HEPA cells, shows the achievement of high levels of editing (80-90%) with sgRNA-ex3 and sgRNAex7. 

- The data also indicates that editing translates into significant reductions in ptbp1 expression, along with an increase in the expression of genes repressed by PTBP1.

- Despite obtaining lower percentages of editing events in N2a neuroblastoma cells and the C8-D1A astroglial cell line, the differential expression levels of ptbp1 and the readout genes remain significant. However, the gRNA screening assay is performed in immortalized, dividing cells. 

- Providing proof that Adenosine Base Editing of Ptbp1 is successful in non-dividing cells (such as SNc and/or striatal primary neurons) would strengthen the case for the potential therapy in the intended cell type.

Following the reviewer’s comment, we show in vivo base editing rates in the SNc and striatum of treated PD mice in the revised manuscript (Figure 2 – figure supplement 5; figure 3 – supplement 2).

- Moreover, assessing the expression levels of tyrosine hydroxylase by qPCR after Ptbp1 base editing in vitro could help contextualize the use of TH+ detection as an in vivo readout and may help explain why the total number of TH+ cells is low after ABE treatment in vivo - as shown in following sections.

In the revised manuscript, we now provide quantifications of in vivo base editing efficiencies in the SNc (~15%) and striatum (~20%). As expected from these lower in vivo base editing rates, downregulation of Ptbp1 at the transcript and protein level was less pronounced compared to our in vitro experiments. It seems likely that higher base editing efficiency and more pronounced downregulation of Ptbp1 could lead to a larger population of TH expressing cells. We have added these results and interpretations to the revised manuscript.

- Furthermore, although ABEs are less prone to generating bystander and other nucleotide changes compared to CBEs, it is still possible. Figures 1 (line 811) and 1-supplement 2 (line 842) only show a brief window of the Sanger sequencing trace. Updating these figures to display a wider view of the sequencing trace would enhance transparency. If unwanted edits are detected, while they may not significantly alter the relevance, impact, or structure of the paper, they may become an important aspect of the discussion. 

Indeed, ABEs can induce bystander edits and we also detected such edits at the Ptbp1 target site. However, since our base editing strategy was designed to yield a loss of Ptbp1 function, bystander editing at the splice site was not a primary focus in our analysis. Nevertheless, we included CRISPResso output images showing the specific editing outcomes in a wider analysis window in the revised manuscript (Figure 3 – figure supplement 2). 

Result Section 2 - Lines 110-159

A split intein system is used in vivo with sgRNA-ex3, after updating the promoter to make it cell-specific: hSyn to restrict expression to neurons and GFAP to restrict expression to astrocytes. 

However, no other assay is performed to assess whether a) the promoter change and/or b) splitting Cas9 may affect the editing efficiency compared to their initial in vitro approach.

In the revised manuscript, we assessed the performance of the in vivo AAV vectors encoding the split intein ABE with sgRNA-ex3 in vitro in N2a and C8-D1A cells. Our results show that all vectors are functional and result in base editing at the target locus.

-  Addressing whether this is the case may explain the low number of TH+ cells observed in vivo. 

- The authors could also consider staining for Cas9 to address whether the low number of TH+cells could be attributed to a poor Cas9 delivery.

To confirm successful in vivo base editor delivery, we quantified in vivo base editing efficiencies in the SNc and striatum of PD mice. Our analysis revealed in vivo base editing efficiencies at both tissue sites, confirming that base editors were successfully delivered. Editing efficiencies were, however, substantially lower (Figure 2 – figure supplement 5; figure 3 – supplement 2).  than in our in vitro cell line setting (Figure 1; figure 1 – figure supplement 2). Even though tissue editing rates likely underestimate the cell type-specific editing rates in astrocytes or neurons, higher base editing rates would have likely resulted in a higher number of TH positive cells. We have added these results and their implications to the revised manuscript. 

-  Moreover, despite the presence of TH, in Figure 2 E,F authors examine the striatal innervation from newly generated TH+ cells in the SNc by Fluorescence Intensity (FI) to conclude that the edited cells do not form projections towards the striatum. Considering the low levels of TH+ positive cells obtained, the accumulation of gross FI might not be the most accurate way to assess the presence or absence of cell projections.

- Using another marker that stains the projections rather than the cell soma, and that is a marker of dopaminergic neurons, might be a better way to address this.

To address the reviewer’s comment, we analyzed the presence of potential dopaminergic fibers in the mfb, where projections are more concentrated (around the injection coordinates of 6-OHDA), using the dopaminergic marker DAT. In line with our previous observations in the striatum, we did not detect an increase in DAT fluorescence intensity upon treatment on the lesioned hemisphere (Figure 2 – figure supplement 4).  

Result Section 3 - Line 160-182

Minor issue

- The same dual split intein system is used in the striatum. However, in Figure 3 - Figure Supplement 1 - line 958 and in Figure 3 - Figure Supplement 4 - line 1000authors show the injection of 2x the viral genomes indicated along the manuscript. In previous experiments the SNc 2x108vg/animal was used whereas this figure shows 4x108vg/animal injected in the striatum. 

- The authors should clarify if the vg injected in the striatum was different from what they previously indicated.

Compared to injection in the SNc, the volume of vector injected in the striatum was doubled since the region is significantly larger. We clarified that the injected vector genomes were different between striatum and SNc in the revised manuscript.

Result Section 4- Line 183-220

In this section, the authors thoroughly examine the neuronal nature of TH+ cells through NeuN co-staining and iterative immunofluorescence imaging (4i). BrdU experiments are conducted to determine the origin of these cells, leading to the conclusion that TH+ cells derive from nondividing cells and express the neuronal marker DAT, characteristic of dopamine-producing neurons (DANs). Cell shape of the TH+ cells in the striatum and SNc is also evaluated measuring their Feret's diameter and their cell surface. Authors conclude there's heterogeneity in the TH+ cell population due to the presence of TH+/Neun- as well as differences in cell shape. 

However, their explanation of this heterogeneity is solely attributed to differences in the microenvironment and lacks further elaboration. Similarly, their observation that almost half the number of TH+ striatal cells after treatment express CTIP2 (Line 213 and Figure 4B), a marker for GABAergic medium spiny neurons, which they state as "interesting" (line 213) is not developed further. Delving deeper into these topics could strengthen the discussion.

In the revised manuscript, we provided a more in-depth discussion of the 4i imaging results and potential therapeutic implications. Additionally, we suggest follow-up experiments to analyze the identity, function, and molecular mechanisms underlying the expression of TH upon PTBP1 downregulation in future studies. 

Result Section 5- Line 221-243

Two drug-free and two drug-induced behavioral tests are conducted in control and treated animals to evaluate the restoration of motor functions following treatment. Consistent with their previous findings, only the treatment targeted to neurons resulted in the restoration of motor functions in drug-free behavioral tests. The rationale behind each test and its evaluation is clearly explained.

DISCUSSION 

- In the discussion section, the authors effectively re-examine their results contextualizing their data with previous studies in the field. However, it would be helpful at this point in the manuscript to reconsider the use of the term 'cell reprogramming,' as this study does not involve actual cell reprogramming. The concept "reprograming" entails the process of transforming adult cells into a stem cell-like state, to then differentiate them into a different cell type. As proven in section 4 by a BrdU proliferation assay, the targeted cells are differentiated neurons. Considering BrdU is administered 5 days after ABE treatment, if true cell reprogramming was taking place, there should be evidence of BrdU incorporation. Cell reprogramming or reprograming is mentioned 4 times in the manuscript (line 34, line 54, line 265, line 277). Therefore, using another terminology would be more accurate.

Following the reviewer’s suggestion, we removed the term “cell reprograming” from the manuscript and rather describe it as induction of TH expression in endogenous neurons.

- As noted in the comments of section 4, a more thorough discussion about the various possibilities for heterogeneity would enhance the manuscript's contribution to the PD field.

In the revised manuscript, we provided a more in-depth discussion of the 4i imaging results and potential therapeutic implications. 

- Despite observing low numbers of TH+ cells, no significant rescue of drug-induced behaviors, and low levels of released dopamine, the authors merely state that these results make the therapy non-viable, but there is no further exploration or discussion. Whether the limitations lie in the ABE strategy itself, such as its efficiency in targeting and editing of differentiated neurons; or if the issues lie on the injection and delivery, is never discussed. A deeper argumentation on the possible underlying reasons for these challenges would greatly enhance the manuscript and contribute to the advancement of ABE therapies in the brain.

We believe that the efficacy of our base editing approach could be significantly enhanced by optimizing the delivery. Currently, we are using a dual AAV approach to deliver intein-split ABEs. Since this approach relies on the delivery of higher AAV doses to achieve cotransduction of a cell by two different AAVs, the efficiency could be significantly enhanced by using smaller Cas9 orthologues that can be delivered as a single AAV. Furthermore, in this study we performed a single injection into the dorsal striatum to deliver ABE-expressing AAVs. Performing multiple injections into the rostral, medial, and caudal regions of the striatum might allow us to transduce more cells and induce TH expression in a larger population of striatal neurons. We have included these points in the revised manuscript.

- While drug-induced behaviors are not recovered, the data demonstrates a rescue of spontaneous behaviors. Further discussion on the potential differences in circuitry underlying these variations in behavioral rescue would also enrich the manuscript's discussion.

In the revised manuscript, we provide suggestions for potential mechanisms involved in the rescue of spontaneous behavior vs. absence of rescue of drug-induced behaviors. 

FIGURES AND FIGURE SUPPLEMENTS

General minor issue - low magnification images in the following figures, make it difficult to visualize positive cells in tissue sections: Figure 2; Figure 2- supplement 1; Figure 2 - supplement 3, Figure 3- supplement 1. Adding a higher magnification imaging of positive cells in tissue sections of SNc and striatum might help with the visualization. 

As suggested by the reviewer, we included higher magnification images in the corresponding figures to improve interpretation of our results.

eLife Assessment

This important study reports findings on the GnRH pulse generator's role in androgen-exposed mouse models, providing further insights into PCOS pathophysiology and advancing the field of reproductive endocrinology. The experimental data were collected using cutting-edge methodologies and are solid. The findings, while interesting, are primarily applicable to mouse models, and their translation to human physiology requires cautious interpretation and further validation. This work will be of interest to endocrinologists and reproductive biologists.

Reviewer #2 (Public review):

Summary:

The authors aimed to investigate the functionality of the GnRH (gonadotropin-releasing hormone) pulse generator in different mouse models to understand its role in reproductive physiology and its implications for conditions like polycystic ovary syndrome (PCOS). They compared the GnRH pulse generator activity in control mice, peripubertal androgen (PPA) treated mice, and prenatal androgen (PNA) exposed mice. The study sought to elucidate how androgen exposure affects the GnRH pulse generator and subsequent LH (luteinizing hormone) secretion, contributing to the pathophysiology of PCOS.

Strengths:

(1) Comprehensive Model Selection: The use of both PPA and PNA mouse models allows for a comparative analysis that can distinguish the effects of different timings of androgen exposure.

(2) Detailed Methodology: The methods employed, such as photometry recordings and serial blood sampling, are robust and allow for precise measurement of GnRH pulse generator activity and LH secretion.

(3) Clear Results Presentation: The experimental results are well-documented with appropriate statistical analyses, ensuring the findings are reliable and reproducible.

(4) Relevance to PCOS: The study addresses a significant gap in understanding the neuroendocrine mechanisms underlying PCOS, making the findings relevant to both basic science and potentially clinical research.

Weaknesses

(1) Model Limitations: While the PNA mouse model is suggested as the most appropriate for studying PCOS, the authors acknowledge that it does not completely replicate the human condition, particularly the elevated LH response seen in women with PCOS.

(2) Complex Data Interpretation: The reduced progesterone feedback and its effects on the GnRH pulse generator in PNA mice add complexity to data interpretation, making it challenging to draw straightforward conclusions.

(3) Machine Learning (ML) Selection and Validation: While k-means clustering is a useful tool for pattern recognition, the manuscript lacks detailed justification for choosing this specific algorithm over other potential methods. The robustness of clustering results has not been validated.

(4) Biological Interpretability: Although the machine learning approach identified cyclical patterns, the biological interpretation of these clusters in the context of PCOS is not thoroughly discussed. A deeper exploration of how these clusters correlate with physiological and pathological states could enhance the study's impact.

(5) Sample Size: The study uses a relatively small number of animals (n=4-7 per group), which may limit the generalisability of the findings. Larger sample sizes could provide more robust and statistically significant results.

(6) Scope of Application: The findings, while interesting, are primarily applicable to mouse models. The translation to human physiology requires cautious interpretation and further validation.

Comments on revised version:

I did not find the response to my main concerns regarding justification for the choice of the number of clusters (k) and providing evidence of cluster robustness satisfactory at all. It sounds contradictory to me to state that the authors have used unsupervised ML approach when at the same time had clear understanding of the data and the features they wanted to capture. Unsupervised approaches are meant to reveal features that are not apparent by eye... however in their response the authors state, "...our aim was to develop an unsupervised approach that would automatically detect the onset and existence of the key features of pulse generator cyclicity that were apparent by eye...". This sounds like a rather supervised ML approach to me.<br /> Furthermore, I am still unsure why did the authors choose k=5, i.e. assumed there are 5 clusters in the data, and did they explore other possible values for k?<br /> - If not why not? How does this fit with the claims that their ML approach is unsupervised, in other words purely data-driven without making any assumptions?<br /> - If yes did they compare the robustness of their clustering results obtained for different values of k?

Reviewer #3 (Public review):

Summary:

Zhou and colleagues elegantly used pre-clinical mouse models to understand the nature of abnormally high GnRH/LH pulse secretion in polycystic ovary syndrome (PCOS), a major endocrine disorder affecting female fertility worldwide. This work brings a fundamental question of how altered gonadotropin secretion takes place upstream within the GnRH pulse generator core, which is defined by arcuate nucleus kisspeptin neurons.

Strengths:

Authors use state-of-the-art in vivo calcium imaging with fiber photometry and important physiological manipulations and measurements to dissect the possible neuronal mechanisms underlying such neuroendocrine derangements in PCOS. The additional use of unsupervised k-means clustering analysis for the evaluation of calcium synchronous events greatly enhances the quality of their evidence. The authors nicely propose that neuroendocrine dysfunction in PCOS might involve different setpoints through the hypothalamic-pituitary-gonadal (HPG) axis, and beyond kisspeptin neurons, which importantly pushes our field forward toward future investigations.

Weaknesses:

The reviewer agrees that the authors provide important evidence and have improved the quality of the manuscript following first-round revisions. However, they seem resistant to show frequency and amplitude averages in Figure 1 or as supplemental data. Whether the amplitude is dependent on fiber position and its influences on the analysis should be a point of discussion and not data omission. A more detailed analysis of frequency data would enhance the quality of their manuscript.

Comments on revised version:

This comment is related to Reviewer 3's comment # 2 (major) response:

The response does not justify why authors could simply show frequency and amplitude averages in Figure 1 or as supplemental data. Whether the amplitude is dependent on fiber position and its influences on the analysis should be a point of discussion and not data omission.

Author response:

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

Reviewer #1 (Public Review):

The manuscript involves 11 research vignettes that interrogate key aspects of GnRH pulse generator in two established mouse models of PCOS (peripubertal and prenatal androgenisation; PPA and PNA) (9 of the vignettes focus on the latter model).

A key message of this paper is that the oft-quoted idea of rapid GnRH/LH pulses associated with PCOS is in fact not readily demonstrable in PNA and PPA mice. This is an important message to make known, but when established dogmas are being challenged, the experiments behind them need to be robust. In this case, underpowered experiments and one or two other issues greatly limit the overall robustness of the study.

General critiques

(1) My main concern is that many/most of the experiments were limited to 4-5 mice per group (PPA experiments 1 and 2, PNA experiments 3, 5, 6, 8, and 9). This seems very underpowered for trying to disprove established dogmas (sometimes falling back on "non-significant trends" - lines 105 and 239).

For the key characterization of GnRH pulse generator activity and LH pulsatility in intact PNA mice (Fig.3, 4, 6), we used 6-8 animals in each experiment which we believe to be sufficient. 

It is pertinent to explore the “established dogma”. While there is every expectation that the PNA model should have increased LH pulsatility, in fact there is only a single study (Moore, Prescott et al. 2015) that has shown this. The two other reports that have examined this issue find no change in LH pulse frequency (McCarthy, Dischino et al. 2021 and ours). Hence, we would suggest that expectations rather than evidence presently maintains the PNA “dogma”. For the PPA model, there is in fact not a single paper reporting increased LH pulse frequency.

(2) Page 133-142: it is concerning that the PNA mice didn't have elevated testosterone levels, and this clearly isn't the fault of the assay as this was re-tested in the laboratory of Prof Handelsman, an expert in the field, using LCMS. The point (clearly made in lines 315-336 of the Discussion) that elevated testosterone in PNA mice has been shown in some but not other publications is an important concern to describe for the field. However, the fact remains that it IS elevated in numerous studies, and in the current study it is not so, yet the authors go on to present GnRH pulse generator data as characteristic of the PNA model. Perhaps a demonstration of elevated testosterone levels (by LCMS?) should become a standard model validation prerequisite for publishing any PNA model data.

We provide a Table below showing the huge inconsistencies in testosterone levels reported in the PNA mouse model. If anything, these inconsistencies might be explained by age, although again this is very variable between studies. Much the same as the “dogma” related to LH pulsatility in the PNA model, we would question whether there is any robust increase in testosterone levels in this model. There is no question that women with PCOS have elevated testosterone but whether the PNA mouse is a good model for this is debatable. We have noted this caution and the need for further LC-MS studies in the Discussion.

Author response table 1.

*Same ELISA used in the current study.

(3) Line 191-196: the lack of a significant increase in LH pulse frequency in PNA mice is based on measurements using reasonable group sizes (7-8), although the sampling frequency is low for this type of analysis (10-minute intervals; 6-minute intervals would seem safer for not missing some pulses). The significance of the LH pulse frequency results is not stated (looks like about p=0.01). The authors note that LH concentration IS elevated (approximately doubled), and this clearly is not caused by an increase in amplitude (Figure 4 G, H, I). These things are worth commenting on in the discussion.

We have included the p-value of the LH pulse frequency results and included the relevant discussion.

(4) An interesting observation is that PNA mice appear to continue to have cyclical patterns of GnRH pulse generator activity despite reproductive acyclicity as determined by vaginal cytology (lines 209-241). This finding was used to analyse the frequency of GnRH pulse generator SEs in the machine-learning-identified diestrous-like stage of PNA mice and compare it to diestrous control mice (as identified by vaginal cytology?) (lines 245-254). The idea of a cycle stage-specific comparison is good, but surely the only valid comparison would be to use machine-learning to identify the diestrous-like stage in both groups of mice. Why use machine learning for one and vaginal cytology for the other?

As “machine learning-defined” diestrus is based on the control vaginal cytology information, the diestrous mice are in fact defined by the same machine learning parameters. We have now noted this.

Specific points

(5) With regard to point 2 above, it would be helpful to note the age at which the testosterone samples were taken.

We have included the age in the method.

(6) Lines 198-205 and 258-266: I think these are repeated measures of ANOVA data? If so, report the main relevant effect before the post hoc test result.

We have included the relevant main effect in the manuscript.

(7) Line 415: I don't think the word "although" works in this sentence.

We have changed the wording accordingly.

(8) Lines 514-518: what are the limits of hormone detection in the LCMS assay?

These were originally stated in the figure legend but have now been included in the Methods.

Reviewer #2 (Public Review):

Summary

The authors aimed to investigate the functionality of the GnRH (gonadotropin-releasing hormone) pulse generator in different mouse models to understand its role in reproductive physiology and its implications for conditions like polycystic ovary syndrome (PCOS). They compared the GnRH pulse generator activity in control mice, peripubertal androgen (PPA) treated mice, and prenatal androgen (PNA) exposed mice. The study sought to elucidate how androgen exposure affects the GnRH pulse generator and subsequent LH (luteinizing hormone) secretion, contributing to the pathophysiology of PCOS.

Strengths

(1) Comprehensive Model Selection: The use of both PPA and PNA mouse models allows for a comparative analysis that can distinguish the effects of different timings of androgen exposure.

(2) Detailed Methodology: The methods employed, such as photometry recordings and serial blood sampling, are robust and allow for precise measurement of GnRH pulse generator activity and LH secretion.

(3) Clear Results Presentation: The experimental results are well-documented with appropriate statistical analyses, ensuring the findings are reliable and reproducible.

(4) Relevance to PCOS: The study addresses a significant gap in understanding the neuroendocrine mechanisms underlying PCOS, making the findings relevant to both basic science and potentially clinical research.

Weaknesses

(1) Model Limitations: While the PNA mouse model is suggested as the most appropriate for studying PCOS, the authors acknowledge that it does not completely replicate the human condition, particularly the elevated LH response seen in women with PCOS.

We agree.

(2) Complex Data Interpretation: The reduced progesterone feedback and its effects on the GnRH pulse generator in PNA mice add complexity to data interpretation, making it challenging to draw straightforward conclusions.

We agree.

(3) Machine Learning (ML) Selection and Validation: While k-means clustering is a useful tool for pattern recognition, the manuscript lacks detailed justification for choosing this specific algorithm over other potential methods. The robustness of clustering results has not been validated.

Please see below.

(4) Biological Interpretability: Although the machine learning approach identified cyclical patterns, the biological interpretation of these clusters in the context of PCOS is not thoroughly discussed. A deeper exploration of how these clusters correlate with physiological and pathological states could enhance the study's impact.

It is presently difficult to ascribe specific functions of the various pulse generator states to physiological impact. While it is reasonable to suggest that Cluster_0 activity (representing very infrequent SEs) is responsible for the estrous/luteal-phase pause in pulsatility, we remain unclear on the physiological impact of multi-peak SEs on LH secretion, even in normal mice (see Vas et al., Endo 2024). Thus, for the moment, it is most appropriate to simply state that pulse generator activity remains cyclical in PNA mice without any unfounded speculation.

(5) Sample Size: The study uses a relatively small number of animals (n=4-7 per group), which may limit the generalisability of the findings. Larger sample sizes could provide more robust and statistically significant results.

For the key characterization of GnRH pulse generator activity and LH pulsatility in intact PNA mice (Fig.3, 4, 6), we used 6-8 animals in each experiment which we believe to be sufficient. Some of the subsequent experiments do have smaller N numbers and we are particularly aware of the progesterone treatment study that only has N=3 for the PNA group. However, as this was sufficient to show a statistical difference we did not generate more mice.

(6) Scope of Application: The findings, while interesting, are primarily applicable to mouse models. The translation to human physiology requires cautious interpretation and further validation.

We agree.

Reviewer #2 (Recommendations For The Authors):

(1) The validation of clustering results through additional metrics or comparison with other algorithms would strengthen the methodology. Specifically, the authors selected k=5 for k-means clustering without providing an explicit rationale or evidence of exploratory data analysis (EDA) to support this choice. They refer to their previous publication (Vas, Wall et al. 2024), which does not provide any EDA regarding the choice of a number of clusters nor their robustness. The arbitrary selection of "k" without justification can undermine confidence in the clustering results since clustering results heavily depend on "k". The authors also choose to use Euclidean distance as the "numerical measure" setting in the RapidMiner Studio's software without justification given the chosen features used for clustering and their properties. The lack of exploratory analysis to determine the optimal number of clusters, "k", to be considered means that the authors might have missed identifying the true structure of the data. Common cluster robustness methods, like the elbow method or silhouette analysis, are crucial for justifying the number of clusters. An inappropriate choice could lead to incorrect conclusions about the synchronisation patterns of ARN kisspeptin neurons and their implications for the study's hypotheses. Including EDA and other validation techniques (e.g., silhouette scores, elbow method) would have strengthened the manuscript by providing empirical support for the chosen algorithm and settings.

It is important to clarify that we did not start this exercise with an unknown or uncharacterised data set and that the objective of the clustering was not to provide any initial pattern to the data. Rather, our aim was to develop an unsupervised approach that would automatically detect the onset and existence of the key features of pulse generator cyclicity that were apparent by eye e.g. the estrous stage slowing and the presence of multi-peak SEs in metestrous. As such, our optimization was driven by the data as well as observation while retaining the unsupervised nature of k-means clustering. We started by assessed 10 variables describing all possible features of the recordings and through a process of elimination found that just 5 were sufficient to describe the key stages of the cycle. While we appreciate that the use of multiple different algorithms would progressively increase the robustness of the machine learning approach, it is evident that the current k-means approach with k=5 is already very effective at reporting the estrous cyclicity of the pulse generator in normal mice (Vas et al., Endo 2024). Having validated this approach, we have now used it here to compare the cyclical patterns of activity of PNA- and vehicle-treated mice.

(2) The data and methods presented in this study could be valuable for the research community studying reproductive endocrinology and neuroendocrine disorders provided the authors address my comments above regarding the application of ML methods. The insights gained from this work could potentially inform clinical research aiming to develop better diagnostic and therapeutic strategies for PCOS.

Reviewer #3 (Public Review):

Summary:

Zhou and colleagues elegantly used pre-clinical mouse models to understand the nature of abnormally high GnRH/LH pulse secretion in polycystic ovary syndrome (PCOS), a major endocrine disorder affecting female fertility worldwide. This work brings a fundamental question of how altered gonadotropin secretion takes place upstream within the GnRH pulse generator core, which is defined by arcuate nucleus kisspeptin neurons.

Strengths:

The authors use state-of-the-art in vivo calcium imaging with fiber photometry and important physiological manipulations and measurements to dissect the possible neuronal mechanisms underlying such neuroendocrine derangements in PCOS. The additional use of unsupervised k-means clustering analysis for the evaluation of calcium synchronous events greatly enhances the quality of their evidence. The authors nicely propose that neuroendocrine dysfunction in PCOS might involve different setpoints through the hypothalamic-pituitary-gonadal (HPG) axis, and beyond kisspeptin neurons, which importantly pushes our field forward toward future investigations.

Weaknesses:

Although the authors provide important evidence, additional efforts are required to improve the quality of the manuscript and back up their claims. For instance, animal experiments failed to detect high testosterone levels in PNA female mice, a well-established PCOS mouse model. Considering that androgen excess is a hallmark of PCOS, this highly influences the subsequent evaluation of calcium synchronous events in arcuate kisspeptin neurons and the implications for neuroendocrine derangements.

Please see our response to Reviewer 1. It will be important to establish a robust PCOS mouse model in the future that has elevated pulse generator activity in the presence of elevated testosterone concentrations.

Authors also may need to provide LH data from another mouse model used in their work, the peripubertal androgen (PPA) model. Their claims seem to fall short without the pairing evidence of calcium synchronous events in arcuate kisspeptin neurons and LH pulse secretion.

We have demonstrated that ARN-KISS neuron SEs are perfectly correlated with pulsatile LH secretion in intact and gonadectomized male and female mice on many occasions. Given that the pulse generator frequency slows by 50% in PPA mice, it is very hard to imagine how this could result in an elevated LH pulse frequency. While we were undertaking these studies the first paper (to our knowledge) looking at pulsatile LH secretion in the PPA model was published; no change was found.

Another aspect that requires reviewing, is further exploration of their calcium synchronous events data and the increase of animal numbers in some of their experiments.

Please see below.

Reviewer #3 (Recommendations For The Authors):

The reviewer believes that this work will greatly contribute to the field and, to provide better manuscript quality, there might be only a few minor and major revisions to be included in the future version.

Minor:

(1) Line 17: I would change the sentence to "One in ten women in their reproductive age suffer from PCOS" to adapt to more accurate prevalence studies.

We have revised the sentence as recommended.

(2) Line 18 and 19: Although the evidence indeed points to a high LH pulse secretion in PCOS, I would change it to "with increased LH secretion" as most studies show mean values and not LH pulse release data.

While we agree that most human studies show a mean increase in LH, when assessed with sufficient temporal resolution, this results from elevated LH pulse frequency. As such, and to keep the manuscript focussed on the pulse generator, we would like the retain the present wording.

(3) Line 47: Please correct "polycystic ovaries" to polycystic-like ovarian morphology to adapt to the current AEPCOS guidelines.

We have revised the sentence as recommended.

(4) Line 231: Authors stated that "These PNA mice exhibited a cyclical pattern of activity similar to that of control mice" (Figure 5C and D). Please, include the statistical tests here for this claim. Although they say there aren't differences, the colored fields do not reflect this and seem quite different. Could the authors re-evaluate these claims or provide better examples in the figure?

We used Sidak’s multiple comparisons tests for this analysis (as stated in Results). The key data for assessing overall cyclical activity in PNA and control mice is Fig 5B which suggest very little difference. We accept that the individual traces of activity (Fig.5D) do not look identical to controls and, indeed, they are representative of the data set. The key point is they remain cyclical in an acyclic mouse. We have made sure that this is clear in the text.

(5) Subheadings 6 and & of the result section: It sounds confusing to read the foremost claims of the absence of SE differences and next have a clear SE frequency difference in Figures 6 C and D. The reviewer suggests that authors could reorganize the text and figures to make their rationale flow better for future readers.

We have considered this point carefully but find that re-organization creates its own problems with having to use the machine learning algorithm before describing it. It will always be problematic to incorporate this type of data-reanalysis in an original paper but think this present sequence is the best that can be achieved.

(6) Discussion: If PNA female mice did not have elevated testosterone levels, how can the authors compare their results to the current literature? Could this be the case for lacking a more robust ARNKISS neuronal activity output in their experiments? The reviewer recommends a better discussion concerning these aspects.

Please refer to our response to Reviewer #1 comment (2).

(7) Discussion: the authors claim that diestrous PNA mice exhibited highly variable patterns of ARNKISS neuron activity. Would these differences be due to different circulating sex steroid levels or intrinsic properties? Would the inclusion of future in vitro calcium imaging (brain slices) studies contribute to their research question and conclusions? The reviewer recommends a better discussion concerning these aspects.

We have tried to clarify that the highly variable patterns of activity in “diestrous” PNA mice come from the fact that we are actually randomly recording from ARN-KISS neurons at metestrus, diestrus, proestrus and estrus.  The pulse generator is cycling but we only have the acyclic “diestrous” smear to go by. This also makes brain slice studies difficult as we would never know the actual cycle stage.

Major:

(1) Results section: The reviewer strongly recommends that the LH pulse secretion data for the PPA group be included in the manuscript. If the SEs represent the central mechanism of pulse generation, would the LH pulse frequency match those events? If not, could a mismatch be explained by androgen-mediated negative feedback at the pituitary level? What is the pituitary LH response to exogenous GnRH (i.p. injection) in the PPA group?

Our initial observation showed the frequency of ARNKISS neuron SEs was halved in PPA mice compared to controls. Additionally, one study reported pulsatile LH secretion to be unchanged in this animal model (Coyle, Prescott et al. 2022). Both pieces of evidence clearly indicate that the PPA mouse does not provide an appropriate PCOS model of elevated pulse generator activity. Therefore, we do not see the value of pursuing further experiments in this animal model.

(2) Although the evaluation of relative frequency and normalized amplitude indicate the dynamic over time, the authors should include the average amplitudes and frequencies of events within the recording session. For instance, looking at Figures 1 A and B and Figures 3 A and B, a reader can observe differences in the amplitude due to different scaling axes. Perhaps, using a Python toolbox such as GuPPy or any preferred analysis pipeline might help authors include these parameters.

The amplitude of recorded SEs for each mouse depends primarily on the fiber position. As such, it has only ever been possible to assess SE amplitude changes within the same mouse. It is not possible to assess differences in SE amplitude between mice.

(3) Line 144-156: (Immunoreactivity results): Authors should proceed with caution when describing these results and clearly state that results show a software-based measurement of immunoreactive signal intensity. In addition, the small sample size of the PNA group (N = 4) compared to controls (N = 6-7) seems to mask possible differences. Could the authors increase the N of the PNA group and re-evaluate these results?

We have clarified that the immunoreactive signal intensity is based on software-based measurement. The N number for PNA mice in these studies varies from 4 to 6 depending on brain section availability for the different immunohistochemistry runs. The scatter of data is such that any new data points would need to be at the extreme of the distributions to likely have any impact on statistical significance. As a minor part of the paper, we did not feel that the use of further mice was warranted.

(4) Considering the great variability of PNA's number of SE/hr, the review suggests increasing the N in this group, thus, authors can re-evaluate their findings and draw better analysis/ conclusion.

We have n=6 for the PNA group in the study. As noted above, the variability in SE/hr in Figure 3 comes from assessing the pulse generator at random times within the estrous cycle. Once we separate “diestrous-like” stage for the PNA animals, the variability is decreased as shown in Figure 6.

eLife Assessment

This paper presents a new method called MINT that is effective at BCI-style decoding tasks. The authors show convincing evidence to support their claims regarding how MINT is a new method that produces excellent decoding performance relative to the state-of-the-art. This work is important and will be of broad interest to neuroscientists and neuroengineers.

Reviewer #1 (Public Review):

Summary:

This paper presents an innovative decoding approach for brain-computer interfaces (BCIs), introducing a new method named MINT. The authors develop a trajectory-centric approach to decode behaviors across several different datasets, including eight empirical datasets from the Neural Latents Benchmark. Overall, the paper is well written and their method shows impressive performance compared to more traditional decoding approaches that use a simpler approach. While there are some concerns (see below), the paper's strengths, particularly its emphasis on a trajectory-centric approach and the simplicity of MINT, provide a compelling contribution to the field.

Strengths:

The adoption of a trajectory-centric approach that utilizes statistical constraints presents a substantial shift in methodology, potentially revolutionizing the way BCIs interpret and predict neural behaviour. This is one of the strongest aspects of the paper.

The thorough evaluation of the method across various datasets serves as an assurance that the superior performance of MINT is not a result of overfitting. The comparative simplicity of the method in contrast to many neural network approaches is refreshing and should facilitate broader applicability.

Weaknesses:

Scope: Despite the impressive performance of MINT across multiple datasets, it seems predominantly applicable to M1/S1 data. Only one of the eight empirical datasets comes from an area outside the motor/somatosensory cortex. It would be beneficial if the authors could expand further on how the method might perform with other brain regions that do not exhibit low tangling or do not have a clear trial structure (e.g. decoding of position or head direction from hippocampus)

When comparing methods, the neural trajectories of MINT are based on averaged trials, while the comparison methods are trained on single trials. An additional analysis might help in disentangling the effect of the trial averaging. For this, the authors could average the input across trials for all decoders, establishing a baseline for averaged trials. Note that inference should still be done on single trials. Performance can then be visualized across different values of N, which denotes the number of averaged trials used for training.

Comments on revisions:

I have looked at the responses and they are thorough and answer all of my questions.

Reviewer #2 (Public Review):

Summary:

The goal of this paper is to present a new method, termed MINT, for decoding behavioral states from neural spiking data. MINT is a statistical method which, in addition to outputting a decoded behavioral state, also provides soft information regarding the likelihood of that behavioral state based on the neural data. The innovation in this approach is neural states are assumed to come from sparsely distributed neural trajectories with low tangling, meaning that neural trajectories (time sequences of neural states) are sparse in the high-dimensional space of neural spiking activity and that two dissimilar neural trajectories tend to correspond to dissimilar behavioral trajectories. The authors support these assumptions through analysis of previously collected data, and then validate the performance of their method by comparing it to a suite of alternative approaches. The authors attribute the typically improved decoding performance by MINT to its assumptions being more faithfully aligned to the properties of neural spiking data relative to assumptions made by the alternatives.

Strengths:

The paper did an excellent job critically evaluating common assumptions made by neural analytical methods, such as neural state being low-dimensional relative to the number of recorded neurons. The authors made strong arguments, supported by evidence and literature, for potentially high-dimensional neural states and thus the need for approaches that do not rely on an assumption of low dimensionality.

The paper was thorough in considering multiple datasets across a variety of behaviors, as well as existing decoding methods, to benchmark the MINT approach. This provided a valuable comparison to validate the method. The authors also provided nice intuition regarding why MINT may offer performance improvement in some cases and in which instances MINT may not perform as well.

In addition to providing a philosophical discussion as to the advantages of MINT and benchmarking against alternatives, the authors also provided a detailed description of practical considerations. This included training time, amount of training data, robustness to data loss or changes in the data, and interpretability. These considerations not only provided objective evaluation of practical aspects but also provided insights to the flexibility and robustness of the method as they relate back to the underlying assumptions and construction of the approach.

Impact:

This work is motivated by brain-computer interfaces applications, which it will surely impact in terms of neural decoder design. However, this work is also broadly impactful for neuroscientific analysis to relate neural spiking activity to observable behavioral features. Thus, MINT will likely impact neuroscience research generally. The methods are made publicly available, and the datasets used are all in public repositories, which facilitates adoption and validation of this method within the greater scientific community.

Author response:

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

Summary of reviewers’ comments and our revisions: 

We thank the reviewers for their thoughtful feedback. This feedback has motivated multiple revisions and additions that, in our view, have greatly improved the manuscript. This is especially true with regard to a major goal of this study: clearly defining existing scientific perspectives and delineating their decoding implications. In addition to building on this conceptual goal, we have expanded existing analyses and have added a new analysis of generalization using a newly collected dataset. We expect the manuscript will be of very broad interest, both to those interested in BCI development and to those interested in fundamental properties of neural population activity and its relationship with behavior.

Importantly, all reviewers were convinced that MINT provided excellent performance, when benchmarked against existing methods, across a broad range of standard tasks:

“their method shows impressive performance compared to more traditional decoding approaches” (R1) 

“The paper was thorough in considering multiple datasets across a variety of behaviors, as well as existing decoding methods, to benchmark the MINT approach. This provided a valuable comparison to validate the method.” (R2) 

“The fact that performance on stereotyped tasks is high is interesting and informative…” (R3)

This is important. It is challenging to design a decoder that performs consistently across multiple domains and across multiple situations (including both decoding and neural state estimation). MINT does so. MINT consistently outperformed existing lightweight ‘interpretable’ decoders, despite being a lightweight interpretable decoder itself. MINT was very competitive with expressive machine-learning methods, yet has advantages in flexibility and simplicity that more ‘brute force’ methods do not. We made a great many comparisons, and MINT was consistently a strong performer. Of the many comparisons we made, there was only one where MINT was at a modest disadvantage, and it was for a dataset where all methods performed poorly. No other method we tested was as consistent. For example, although the GRU and the feedforward network were often competitive with MINT (and better than MINT in the one case mentioned above), there were multiple other situations where they performed less well and a few situations where they performed poorly. Moreover, no other existing decoder naturally estimates the neural state while also readily decoding, without retraining, a broad range of behavioral variables.

R1 and R2 were very positive about the broader impacts of the study. They stressed its impact both on decoder design, and on how our field thinks, scientifically, about the population response in motor areas: 

“This paper presents an innovative decoding approach for brain-computer interfaces” (R1)

“presents a substantial shift in methodology, potentially revolutionizing the way BCIs interpret and predict neural behaviour” (R1)

“the paper's strengths, particularly its emphasis on a trajectory-centric approach and the simplicity of MINT, provide a compelling contribution to the field” (R1)

“The authors made strong arguments, supported by evidence and literature, for potentially high-dimensional neural states and thus the need for approaches that do not rely on an assumption of low dimensionality” (R2)

“This work is motivated by brain-computer interfaces applications, which it will surely impact in terms of neural decoder design.” (R2)

“this work is also broadly impactful for neuroscientific analysis... Thus, MINT will likely impact neuroscience research generally.” (R2)

We agree with these assessments, and have made multiple revisions to further play into these strengths. As one example, the addition of Figure 1b (and 6b) makes this the first study, to our knowledge, to fully and concretely illustrate this emerging scientific perspective and its decoding implications. This is important, because multiple observations convince us that the field is likely to move away from the traditional perspective in Figure 1a, and towards that in Figure 1b. We also agree with the handful of weaknesses R1 and R2 noted. The manuscript has been revised accordingly. The major weakness noted by R1 was the need to be explicit regarding when we suspect MINT would (and wouldn’t) work well in other brain areas. In non-motor areas, the structure of the data may be poorly matched with MINT’s assumptions. We agree that this is likely to be true, and thus agree with the importance of clarifying this topic for the reader. The revision now does so. R1 also wished to know whether existing methods might benefit from including trial-averaged data during training, something we now explore and document (see detailed responses below). R2 noted two weaknesses: 1) The need to better support (with expanded analysis) the statement that neural and behavioral trajectories are non-isometric, and 2) The need to more rigorously define the ‘mesh’. We agree entirely with both suggestions, and the revision has been strengthened by following them (see detailed responses below).

R3 also saw strengths to the work, stating that:

“This paper is well-structured and its main idea is clear.” 

“The fact that performance on stereotyped tasks is high is interesting and informative, showing that these stereotyped tasks create stereotyped neural trajectories.” 

“The task-specific comparisons include various measures and a variety of common decoding approaches, which is a strength.”

However, R3 also expressed two sizable concerns. The first is that MINT might have onerous memory requirements. The manuscript now clarifies that MINT has modest memory requirements. These do not scale unfavorably as the reviewer was concerned they might. The second concern is that MINT is: 

“essentially a table-lookup rather than a model.”

Although we don’t agree, the concern makes sense and may be shared by many readers, especially those who take a particular scientific perspective. Pondering this concern thus gave us the opportunity to modify the manuscript in ways that support its broader impact. Our revisions had two goals: 1) clarify the ways in which MINT is far more flexible than a lookup-table, and 2) better describe the dominant scientific perspectives and their decoding implications.

The heart of R3’s concern is the opinion that MINT is an effective but unprincipled hack suitable for situations where movements are reasonably stereotyped. Of course, many tasks involve stereotyped movements (e.g. handwriting characters), so MINT would still be useful. Nevertheless, if MINT is not principled, other decode methods would often be preferable because they could (unlike MINT in R3’s opinion) gain flexibility by leveraging an accurate model. Most of R3’s comments flow from this fundamental concern: 

“This is again due to MINT being a lookup table with a library of stereotyped trajectories rather than a model.”

“MINT models task-dependent neural trajectories, so the trained decoder is very task-dependent and cannot generalize to other tasks.”

“Unlike MINT, these works can achieve generalization because they model the neural subspace and its association to movement.”

“given that MINT tabulates task-specific trajectories, it will not generalize to tasks that are not seen in the training data even when these tasks cover the exact same space (e.g., the same 2D computer screen and associated neural space).”

“For proper training, the training data should explore the whole movement space and the associated neural space, but this does not mean all kinds of tasks performed in that space must be included in the training set (something MINT likely needs while modeling-based approaches do not).”

The manuscript has been revised to clarify that MINT is considerably more flexible than a lookup table, even though a lookup table is used as a first step. Yet, on its own, this does not fully address R3’s concern. The quotes above highlight that R3 is making a standard assumption in our field: that there exists a “movement space and associated neural space”. Under this perspective, one should, as R3 argues fully explore the movement space. This would perforce fully explore the associated neural subspace. One can then “model the neural subspace and its association to movement”. MINT does not use a model of this type, and thus (from R3’s perspective) does not appear to use a model at all. A major goal of our study is to question this traditional perspective. We have thus added a new figure to highlight the contrast between the traditional (Figure 1a) and new (Figure 1b) scientific perspectives, and to clarify their decoding implications.

While we favor the new perspective (Figure 1b), we concede that R3 may not share our view. This is fine. Part of the reason we believe this study is timely, and will be broadly read, is that it raises a topic of emerging interest where there is definitely room for debate. If we are misguided – i.e. if Figure 1a is the correct perspective – then many of R3’s concerns would be on target: MINT could still be useful, but traditional methods that make the traditional assumptions in Figure 1a would often be preferable. However, if the emerging perspective in Figure 1b is more accurate, then MINT’s assumptions would be better aligned with the data than those of traditional methods, making it a more (not less) principled choice.

Our study provides new evidence in support of Figure 1b, while also synthesizing existing evidence from other recent studies. In addition to Figure 2, the new analysis of generalization further supports Figure 1b. Also supporting Figure 1b is the analysis in which MINT’s decoding advantage, over a traditional decoder, disappears when simulated data approximate the traditional perspective in Figure 1a.

That said, we agree that the present study cannot fully resolve whether Figure 1a or 1b is more accurate. Doing so will take multiple studies with different approaches (indeed we are currently preparing other manuscripts on this topic). Yet we still have an informed scientific opinion, derived from past, present and yet-to-be-published observations. Our opinion is that Figure 1b is the more accurate perspective. This possibility makes it reasonable to explore the potential virtues of a decoding method whose assumptions are well-aligned with that perspective. MINT is such a method. As expected under Figure 1b, MINT outperforms traditional interpretable decoders in every single case we studied. 

As noted above, we have added a new generalization-focused analysis (Figure 6) based on a newly collected dataset. We did so because R3’s comments highlight a deep point: which scientific perspective one takes has strong implications regarding decoder generalization. These implications are now illustrated in the new Figure 6a and 6b. Under Figure 6a, it is possible, as R3 suggests, to explore “the whole movement space and associated neural space” during training. However, under Figure 6b, expectations are very different. Generalization will be ‘easy’ when new trajectories are near the training-set trajectories. In this case, MINT should generalize well as should other methods. In contrast, generalization will be ‘hard’ when new neural trajectories have novel shapes and occupy previously unseen regions / dimensions. In this case, all current methods, including MINT, are likely to fail. R3 points out that traditional decoders have sometimes generalized well to new tasks (e.g. from center-out to ‘pinball’) when cursor movements occur in the same physical workspace. These findings could be taken to support Figure 6a, but are equally consistent with ‘easy’ generalization in Figure 6b. To explore this topic, the new analysis in Figure 6c-g considers conditions that are intended to span the range from easy to hard. Results are consistent with the predictions of Figure 6b. 

We believe the manuscript has been significantly improved by these additions. The revisions help the manuscript achieve its twin goals: 1) introduce a novel class of decoder that performs very well despite being very simple, and 2) describe properties of motor-cortex activity that will matter for decoders of all varieties.

Reviewer #1: 

Summary: 

This paper presents an innovative decoding approach for brain-computer interfaces (BCIs), introducing a new method named MINT. The authors develop a trajectory-centric approach to decode behaviors across several different datasets, including eight empirical datasets from the Neural Latents Benchmark. Overall, the paper is well written and their method shows impressive performance compared to more traditional decoding approaches that use a simpler approach. While there are some concerns (see below), the paper's strengths, particularly its emphasis on a trajectory-centric approach and the simplicity of MINT, provide a compelling contribution to the field. 

We thank the reviewer for these comments. We share their enthusiasm for the trajectory-centric approach, and we are in complete agreement that this perspective has both scientific and decoding implications. The revision expands upon these strengths.

Strengths: 

The adoption of a trajectory-centric approach that utilizes statistical constraints presents a substantial shift in methodology, potentially revolutionizing the way BCIs interpret and predict neural behaviour. This is one of the strongest aspects of the paper. 

Again, thank you. We also expect the trajectory-centric perspective to have a broad impact, given its relevance to both decoding and to thinking about manifolds.

The thorough evaluation of the method across various datasets serves as an assurance that the superior performance of MINT is not a result of overfitting. The comparative simplicity of the method in contrast to many neural network approaches is refreshing and should facilitate broader applicability. 

Thank you. We were similarly pleased to see such a simple method perform so well. We also agree that, while neural-network approaches will always be important, it is desirable to also possess simple ‘interpretable’ alternatives.

Weaknesses:  

Comment 1) Scope: Despite the impressive performance of MINT across multiple datasets, it seems predominantly applicable to M1/S1 data. Only one of the eight empirical datasets comes from an area outside the motor/somatosensory cortex. It would be beneficial if the authors could expand further on how the method might perform with other brain regions that do not exhibit low tangling or do not have a clear trial structure (e.g. decoding of position or head direction from hippocampus) 

We agree entirely. Population activity in many brain areas (especially outside the motor system) presumably will often not have the properties upon which MINT’s assumptions are built. This doesn’t necessarily mean that MINT would perform badly. Using simulated data, we have found that MINT can perform surprisingly well even when some of its assumptions are violated. Yet at the same time, when MINT’s assumptions don’t apply, one would likely prefer to use other methods. This is, after all, one of the broader themes of the present study: it is beneficial to match decoding assumptions to empirical properties. We have thus added a section on this topic early in the Discussion: 

“In contrast, MINT and the Kalman filter performed comparably on simulated data that better approximated the assumptions in Figure 1a. Thus, MINT is not a ‘better’ algorithm – simply better aligned with the empirical properties of motor cortex data. This highlights an important caveat. Although MINT performs well when decoding from motor areas, its assumptions may be a poor match in other areas (e.g. the hippocampus). MINT performed well on two non-motor-cortex datasets – Area2_Bump (S1) and DMFC_RSG (dorsomedial frontal cortex) – yet there will presumably be other brain areas and/or contexts where one would prefer a different method that makes assumptions appropriate for that area.”

Comment 2) When comparing methods, the neural trajectories of MINT are based on averaged trials, while the comparison methods are trained on single trials. An additional analysis might help in disentangling the effect of the trial averaging. For this, the authors could average the input across trials for all decoders, establishing a baseline for averaged trials. Note that inference should still be done on single trials. Performance can then be visualized across different values of N, which denotes the number of averaged trials used for training. 

We explored this question and found that the non-MINT decoders are harmed, not helped, by the inclusion of trial-averaged responses in the training set. This is presumably because the statistics of trialaveraged responses don’t resemble what will be observed during decoding. This statistical mismatch, between training and decoding, hurts most methods. It doesn’t hurt MINT, because MINT doesn’t ‘train’ in the normal way. It simply needs to know rates, and trial-averaging is a natural way to obtain them. To describe the new analysis, we have added the following to the text.

“We also investigated the possibility that MINT gained its performance advantage simply by having access to trial-averaged neural trajectories during training, while all other methods were trained on single-trial data. This difference arises from the fundamental requirements of the decoder architectures: MINT needs to estimate typical trajectories while other methods don’t. Yet it might still be the case that other methods would benefit from including trial-averaged data in the training set, in addition to single-trial data. Alternatively, this might harm performance by creating a mismatch, between training and decoding, in the statistics of decoder inputs. We found that the latter was indeed the case: all non-MINT methods performed better when trained purely on single-trial data.”

Reviewer #2:

Summary: 

The goal of this paper is to present a new method, termed MINT, for decoding behavioral states from neural spiking data. MINT is a statistical method which, in addition to outputting a decoded behavioral state, also provides soft information regarding the likelihood of that behavioral state based on the neural data. The innovation in this approach is neural states are assumed to come from sparsely distributed neural trajectories with low tangling, meaning that neural trajectories (time sequences of neural states) are sparse in the high-dimensional space of neural spiking activity and that two dissimilar neural trajectories tend to correspond to dissimilar behavioral trajectories. The authors support these assumptions through analysis of previously collected data, and then validate the performance of their method by comparing it to a suite of alternative approaches. The authors attribute the typically improved decoding performance by MINT to its assumptions being more faithfully aligned to the properties of neural spiking data relative to assumptions made by the alternatives. 

We thank the reviewer for this accurate summary, and for highlighting the subtle but important fact that MINT provides information regarding likelihoods. The revision includes a new analysis (Figure 6e) illustrating one potential way to leverage knowledge of likelihoods.

Strengths:  

The paper did an excellent job critically evaluating common assumptions made by neural analytical methods, such as neural state being low-dimensional relative to the number of recorded neurons. The authors made strong arguments, supported by evidence and literature, for potentially high-dimensional neural states and thus the need for approaches that do not rely on an assumption of low dimensionality. 

Thank you. We also hope that the shift in perspective is the most important contribution of the study. This shift matters both scientifically and for decoder design. The revision expands on this strength. The scientific alternatives are now more clearly and concretely illustrated (especially see Figure 1a,b and Figure 6a,b). We also further explore their decoding implications with new data (Figure 6c-g).

The paper was thorough in considering multiple datasets across a variety of behaviors, as well as existing decoding methods, to benchmark the MINT approach. This provided a valuable comparison to validate the method. The authors also provided nice intuition regarding why MINT may offer performance improvement in some cases and in which instances MINT may not perform as well. 

Thank you. We were pleased to be able to provide comparisons across so many datasets (we are grateful to the Neural Latents Benchmark for making this possible).

In addition to providing a philosophical discussion as to the advantages of MINT and benchmarking against alternatives, the authors also provided a detailed description of practical considerations. This included training time, amount of training data, robustness to data loss or changes in the data, and interpretability. These considerations not only provided objective evaluation of practical aspects but also provided insights to the flexibility and robustness of the method as they relate back to the underlying assumptions and construction of the approach. 

Thank you. We are glad that these sections were appreciated. MINT’s simplicity and interpretability are indeed helpful in multiple ways, and afford opportunities for interesting future extensions. One potential benefit of interpretability is now explored in the newly added Figure 6e. 

Impact: 

This work is motivated by brain-computer interfaces applications, which it will surely impact in terms of neural decoder design. However, this work is also broadly impactful for neuroscientific analysis to relate neural spiking activity to observable behavioral features. Thus, MINT will likely impact neuroscience research generally. The methods are made publicly available, and the datasets used are all in public repositories, which facilitates adoption and validation of this method within the greater scientific community. 

Again, thank you. We have similar hopes for this study.

Weaknesses (1 & 2 are related, and we have switched their order in addressing them): 

Comment 2) With regards to the idea of neural and behavioral trajectories having different geometries, this is dependent on what behavioral variables are selected. In the example for Fig 2a, the behavior is reach position. The geometry of the behavioral trajectory of interest would look different if instead the behavior of interest was reach velocity. The paper would be strengthened by acknowledgement that geometries of trajectories are shaped by extrinsic choices rather than (or as much as they are) intrinsic properties of the data. 

We agree. Indeed, we almost added a section to the original manuscript on this exact topic. We have now done so:

“A potential concern regarding the analyses in Figure 2c,d is that they require explicit choices of behavioral variables: muscle population activity in Figure 2c and angular phase and velocity in Figure 2d. Perhaps these choices were misguided. Might neural and behavioral geometries become similar if one chooses ‘the right’ set of behavioral variables? This concern relates to the venerable search for movement parameters that are reliably encoded by motor cortex activity [69, 92–95]. If one chooses the wrong set of parameters (e.g. chooses muscle activity when one should have chosen joint angles) then of course neural and behavioral geometries will appear non-isometric. There are two reasons why this ‘wrong parameter choice’ explanation is unlikely to account for the results in Figure 2c,d. First, consider the implications of the left-hand side of Figure 2d. A small kinematic distance implies that angular position and velocity are nearly identical for the two moments being compared. Yet the corresponding pair of neural states can be quite distant. Under the concern above, this distance would be due to other encoded behavioral variables – perhaps joint angle and joint velocity – differing between those two moments. However, there are not enough degrees of freedom in this task to make this plausible. The shoulder remains at a fixed position (because the head is fixed) and the wrist has limited mobility due to the pedal design [60]. Thus, shoulder and elbow angles are almost completely determined by cycle phase. More generally, ‘external variables’ (positions, angles, and their derivatives) are unlikely to differ more than slightly when phase and angular velocity are matched. Muscle activity could be different because many muscles act on each joint, creating redundancy. However, as illustrated in Figure 2c, the key effect is just as clear when analyzing muscle activity. Thus, the above concern seems unlikely even if it can’t be ruled out entirely. A broader reason to doubt the ‘wrong parameter choice’ proposition is that it provides a vague explanation for a phenomenon that already has a straightforward explanation. A lack of isometry between the neural population response and behavior is expected when neural-trajectory tangling is low and output-null factors are plentiful [55, 60]. For example, in networks that generate muscle activity, neural and muscle-activity trajectories are far from isometric [52, 58, 60]. Given this straightforward explanation, and given repeated failures over decades to find the ‘correct’ parameters (muscle activity, movement direction, etc.) that create neural-behavior isometry, it seems reasonable to conclude that no such isometry exists.”

Comment 1) The authors posit that neural and behavioral trajectories are non-isometric. To support this point, they look at distances between neural states and distances between the corresponding behavioral states, in order to demonstrate that there are differences in these distances in each respective space. This supports the idea that neural states and behavioral states are non-isometric but does not directly address their point. In order to say the trajectories are non-isometric, it would be better to look at pairs of distances between corresponding trajectories in each space. 

We like this idea and have added such an analysis. To be clear, we like the original analysis too: isometry predicts that neural and behavioral distances (for corresponding pairs of points) should be strongly correlated, and that small behavioral distances should not be associated with large neural distances. These predictions are not true, providing a strong argument against isometry. However, we also like the reviewer’s suggestion, and have added such an analysis. It makes the same larger point, and also reveals some additional facts (e.g. it reveals that muscle-geometry is more related to neural-geometry than is kinematic-geometry). The new analysis is described in the following section:

“We further explored the topic of isometry by considering pairs of distances. To do so, we chose two random neural states and computed their distance, yielding dneural1. We repeated this process, yielding dneural2. We then computed the corresponding pair of distances in muscle space (dmuscle1 and dmuscle2) and kinematic space (dkin1 and dkin2). We considered cases where dneural1 was meaningfully larger than (or smaller than) dneural2, and asked whether the behavioral variables had the same relationship; e.g. was dmuscle1 also larger than dmuscle2? For kinematics, this relationship was weak: across 100,000 comparisons, the sign of dkin1 − dkin2 agreed with dneural1 − dneural2 only 67.3% of the time (with 50% being chance). The relationship was much stronger for muscles: the sign of dmuscle1 − dmuscle2 agreed with dneural1 − dneural2 79.2% of the time, which is far more than expected by chance yet also far from what is expected given isometry (e.g. the sign agrees 99.7% of the time for the truly isometric control data in Figure 2e). Indeed there were multiple moments during this task when dneural1 was much larger than dneural2, yet dmuscle1 was smaller than dmuscle2. These observations are consistent with the proposal that neural trajectories resemble muscle trajectories in some dimensions, but with additional output-null dimensions that break the isometry [60].”

Comment 3) The approach is built up on the idea of creating a "mesh" structure of possible states. In the body of the paper the definition of the mesh was not entirely clear and I could not find in the methods a more rigorous explicit definition. Since the mesh is integral to the approach, the paper would be improved with more description of this component. 

This is a fair criticism. Although MINTs actual operations were well-documented, how those operations mapped onto the term ‘mesh’ was, we agree, a bit vague. The definition of the mesh is a bit subtle because it only emerges during decoding rather than being precomputed. This is part of what gives MINT much more flexibility than a lookup table. We have added the following to the manuscript.

“We use the term ‘mesh’ to describe the scaffolding created by the training-set trajectories and the interpolated states that arise at runtime. The term mesh is apt because, if MINT’s assumptions are correct, interpolation will almost always be local. If so, the set of decodable states will resemble a mesh, created by line segments connecting nearby training-set trajectories. However, this mesh-like structure is not enforced by MINT’s operations.

Interpolation could, in principle, create state-distributions that depart from the assumption of a sparse manifold. For example, interpolation could fill in the center of the green tube in Figure 1b, resulting in a solid manifold rather than a mesh around its outer surface. However, this would occur only if spiking observations argued for it. As will be documented below, we find that essentially all interpolation is local”

We have also added Figure 4d. This new analysis documents the fact that decoded states are near trainingset trajectories, which is why the term ‘mesh’ is appropriate.

Reviewer #3:

Summary:  

This manuscript develops a new method termed MINT for decoding of behavior. The method is essentially a table-lookup rather than a model. Within a given stereotyped task, MINT tabulates averaged firing rate trajectories of neurons (neural states) and corresponding averaged behavioral trajectories as stereotypes to construct a library. For a test trial with a realized neural trajectory, it then finds the closest neural trajectory to it in the table and declares the associated behavior trajectory in the table as the decoded behavior. The method can also interpolate between these tabulated trajectories. The authors mention that the method is based on three key assumptions: (1) Neural states may not be embedded in a lowdimensional subspace, but rather in a high-dimensional space. (2) Neural trajectories are sparsely distributed under different behavioral conditions. (3) These neural states traverse trajectories in a stereotyped order.  

The authors conducted multiple analyses to validate MINT, demonstrating its decoding of behavioral trajectories in simulations and datasets (Figures 3, 4). The main behavior decoding comparison is shown in Figure 4. In stereotyped tasks, decoding performance is comparable (M_Cycle, MC_Maze) or better (Area 2_Bump) than other linear/nonlinear algorithms

(Figure 4). However, MINT underperforms for the MC_RTT task, which is less stereotyped (Figure 4).  

This paper is well-structured and its main idea is clear. The fact that performance on stereotyped tasks is high is interesting and informative, showing that these stereotyped tasks create stereotyped neural trajectories. The task-specific comparisons include various measures and a variety of common decoding approaches, which is a strength. However, I have several major concerns. I believe several of the conclusions in the paper, which are also emphasized in the abstract, are not accurate or supported, especially about generalization, computational scalability, and utility for BCIs. MINT is essentially a table-lookup algorithm based on stereotyped task-dependent trajectories and involves the tabulation of extensive data to build a vast library without modeling. These aspects will limit MINT's utility for real-world BCIs and tasks. These properties will also limit MINT's generalizability from task to task, which is important for BCIs and thus is commonly demonstrated in BCI experiments with other decoders without any retraining. Furthermore, MINT's computational and memory requirements can be prohibitive it seems. Finally, as MINT is based on tabulating data without learning models of data, I am unclear how it will be useful in basic investigations of neural computations. I expand on these concerns below.  

We thank the reviewer for pointing out weaknesses in our framing and presentation. The comments above made us realize that we needed to 1) better document the ways in which MINT is far more flexible than a lookup-table, and 2) better explain the competing scientific perspectives at play. R3’s comments also motivated us to add an additional analysis of generalization. In our view the manuscript is greatly improved by these additions. Specifically, these additions directly support the broader impact that we hope the study will have.

For simplicity and readability, we first group and summarize R3’s main concerns in order to better address them. (These main concerns are all raised above, in addition to recurring in the specific comments below. Responses to each individual specific comment are provided after these summaries.)

(1) R3 raises concerns about ‘computational scalability.’ The concern is that “MINT's computational and memory requirements can be prohibitive.” This point was expanded upon in a specific comment, reproduced below:

I also find the statement in the abstract and paper that "computations are simple, scalable" to be inaccurate. The authors state that MINT's computational cost is O(NC) only, but it seems this is achieved at a high memory cost as well as computational cost in training. The process is described in section "Lookup table of log-likelihoods" on line [978-990]. The idea is to precompute the log-likelihoods for any combination of all neurons with discretization x all delay/history segments x all conditions and to build a large lookup table for decoding. Basically, the computational cost of precomputing this table is O(V^{Nτ} x TC) and the table requires a memory of O(V^{Nτ}), where V is the number of discretization points for the neural firing rates, N is the number of neurons, τ is the history length, T is the trial length, and C is the number of conditions. This is a very large burden, especially the V^{Nτ} term. This cost is currently not mentioned in the manuscript and should be clarified in the main text. Accordingly, computation claims should be modified including in the abstract.

The revised manuscript clarifies that our statement (that computations are simple and scalable) is absolutely accurate. There is no need to compute, or store, a massive lookup table. There are three tables: two of modest size and one that is tiny. This is now better explained:

“Thus, the log-likelihood of , for a particular current neural state, is simply the sum of many individual log-likelihoods (one per neuron and time-bin). Each individual log-likelihood depends on only two numbers: the firing rate at that moment and the spike count in that bin. To simplify online computation, one can precompute the log-likelihood, under a Poisson model, for every plausible combination of rate and spike-count. For example, a lookup table of size 2001 × 21 is sufficient when considering rates that span 0-200 spikes/s in increments of 0.1 spikes/s, and considering 20 ms bins that contain at most 20 spikes (only one lookup table is ever needed, so long as its firing-rate range exceeds that of the most-active neuron at the most active moment in Ω). Now suppose we are observing a population of 200 neurons, with a 200 ms history divided into ten 20 ms bins. For each library state, the log-likelihood of the observed spike-counts is simply the sum of 200 × 10 = 2000 individual loglikelihoods, each retrieved from the lookup table. In practice, computation is even simpler because many terms can be reused from the last time bin using a recursive solution (Methods). This procedure is lightweight and amenable to real-time applications.”

In summary, the first table simply needs to contain the firing rate of each neuron, for each condition, and each time in that condition. This table consumes relatively little memory. Assuming 100 one-second-long conditions (rates sampled every 20 ms) and 200 neurons, the table would contain 100 x 50 x 200 = 1,000,000 numbers. These numbers are typically stored as 16-bit integers (because rates are quantized), which amounts to about 2 MB. This is modest, given that most computers have (at least) tens of GB of RAM. A second table would contain the values for each behavioral variable, for each condition, and each time in that condition. This table might contain behavioral variables at a finer resolution (e.g. every millisecond) to enable decoding to update in between 20 ms bins (1 ms granularity is not needed for most BCI applications, but is the resolution used in this study). The number of behavioral variables of interest for a particular BCI application is likely to be small, often 1-2, but let’s assume for this example it is 10 (e.g. x-, y-, and z-position, velocity, and acceleration of a limb, plus one other variable). This table would thus contain 100 x 1000 x 10 = 1,000,000 floating point numbers, i.e. an 8 MB table. The third table is used to store the probability of s spikes being observed given a particular quantized firing rate (e.g. it may contain probabilities associated with firing rates ranging from 0 – 200 spikes/s in 0.1 spikes/s increments). This table is not necessary, but saves some computation time by precomputing numbers that will be used repeatedly. This is a very small table (typically ~2000 x 20, i.e. 320 KB). It does not need to be repeated for different neurons or conditions, because Poisson probabilities depend on only rate and count.

(2) R3 raises a concern that MINT “is essentially a table-lookup rather than a model.’ R3 states that MINT 

“is essentially a table-lookup algorithm based on stereotyped task-dependent trajectories and involves the tabulation of extensive data to build a vast library without modeling.”

and that,

“as MINT is based on tabulating data without learning models of data, I am unclear how it will be useful in basic investigations of neural computations.”

This concern is central to most subsequent concerns. The manuscript has been heavily revised to address it. The revisions clarify that MINT is much more flexible than a lookup table, even though MINT uses a lookup table as its first step. Because R3’s concern is intertwined with one’s scientific assumptions, we have also added the new Figure 1 to explicitly illustrate the two key scientific perspectives and their decoding implications. 

Under the perspective in Figure 1a, R3 would be correct in saying that there exist traditional interpretable decoders (e.g. a Kalman filter) whose assumptions better model the data. Under this perspective, MINT might still be an excellent choice in many cases, but other methods would be expected to gain the advantage when situations demand more flexibility. This is R3’s central concern, and essentially all other concerns flow from it. It makes sense that R3 has this concern, because their comments repeatedly stress a foundational assumption of the perspective in Figure 1a: the assumption of a fixed lowdimensional neural subspace where activity has a reliable relationship to behavior that can be modeled and leveraged during decoding. The phrases below accord with that view:

“Unlike MINT, these works can achieve generalization because they model the neural subspace and its association to movement.”

“it will not generalize… even when these tasks cover the exact same space (e.g., the same 2D computer screen and associated neural space).”

“For proper training, the training data should explore the whole movement space and the associated neural space”

“I also believe the authors should clarify the logic behind developing MINT better. From a scientific standpoint, we seek to gain insights into neural computations by making various assumptions and building models that parsimoniously describe the vast amount of neural data rather than simply tabulating the data. For instance, low-dimensional assumptions have led to the development of numerous dimensionality reduction algorithms and these models have led to important interpretations about the underlying dynamics”

Thus, R3 prefers a model that 1) assumes a low-dimensional subspace that is fixed across tasks and 2) assumes a consistent ‘association’ between neural activity and kinematics. Because R3 believes this is the correct model of the data, they believe that decoders should leverage it. Traditional interpretable method do, and MINT doesn’t, which is why they find MINT to be unprincipled. This is a reasonable view, but it is not our view. We have heavily revised the manuscript to clarify that a major goal of our study is to explore the implications of a different, less-traditional scientific perspective.

The new Figure 1a illustrates the traditional perspective. Under this perspective, one would agree with R3’s claim that other methods have the opportunity to model the data better. For example, suppose there exists a consistent neural subspace – conserved across tasks – where three neural dimensions encode 3D hand position and three additional neural dimensions encode 3D hand velocity. A traditional method such as a Kalman filter would be a very appropriate choice to model these aspects of the data.

Figure 1b illustrates the alternative scientific perspective. This perspective arises from recent, present, and to-be-published observations. MINT’s assumptions are well-aligned with this perspective. In contrast, the assumptions of traditional methods (e.g. the Kalman filter) are not well-aligned with the properties of the data under this perspective. This does not mean traditional methods are not useful. Yet under Figure 1b, it is traditional methods, such as the Kalman filter, that lack an accurate model of the data. Of course, the reviewer may disagree with our scientific perspective. We would certainly concede that there is room for debate. However, we find the evidence for Figure 1b to be sufficiently strong that it is worth exploring the utility of methods that align with this scientific perspective. MINT is such a method. As we document, it performs very well.

Thus, in our view, MINT is quite principled because its assumptions are well aligned with the data. It is true that the features of the data that MINT models are a bit different from those that are traditionally modeled. For example, R3 is quite correct that MINT does not attempt to use a biomimetic model of the true transformation from neural activity, to muscle activity, and thence to kinematics. We see this as a strength, and the manuscript has been revised accordingly (see paragraph beginning with “We leveraged this simulated data to compare MINT with a biomimetic decoder”).

(3) R3 raises concerns that MINT cannot generalize. This was a major concern of R3 and is intimately related to concern #2 above. The concern is that, if MINT is “essentially a lookup table” that simply selects pre-defined trajectories, then MINT will not be able to generalize. R3 is quite correct that MINT generalizes rather differently than existing methods. Whether this is good or bad depends on one’s scientific perspective. Under Figure 1a, MINT’s generalization would indeed be limiting because other methods could achieve greater flexibility. Under Figure 1b, all methods will have serious limits regarding generalization. Thus, MINT’s method for generalizing may approximate the best one can presently do. To address this concern, we have made three major changes, numbered i-iii below:

i) Large sections of the manuscript have been restructured to underscore the ways in which MINT can generalize. A major goal was to counter the impression, stated by R3 above, that: 

“for a test trial with a realized neural trajectory, [MINT] then finds the closest neural trajectory to it in the table and declares the associated behavior trajectory in the table as the decoded behavior”.

This description is a reasonable way to initially understand how MINT works, and we concede that we may have over-used this intuition. Unfortunately, it can leave the misimpression that MINT decodes by selecting whole trajectories, each corresponding to ‘a behavior’. This can happen, but it needn’t and typically doesn’t. As an example, consider the cycling task. Suppose that the library consists of stereotyped trajectories, each four cycles long, at five fixed speeds from 0.5-2.5 Hz. If the spiking observations argued for it, MINT could decode something close to one of these five stereotyped trajectories. Yet it needn’t. Decoded trajectories will typically resemble library trajectories locally, but may be very different globally. For example, a decoded trajectory could be thirty cycles long (or two, or five hundred) perhaps speeding up and slowing down multiple times across those cycles.

Thus, the library of trajectories shouldn’t be thought of as specifying a limited set of whole movements that can be ‘selected from’. Rather, trajectories define a scaffolding that outlines where the neural state is likely to live and how it is likely to be changing over time. When we introduce the idea of library trajectories, we are now careful to stress that they don’t function as a set from which one trajectory is ‘declared’ to be the right one:

“We thus designed MINT to approximate that manifold using the trajectories themselves, rather than their covariance matrix or corresponding subspace. Unlike a covariance matrix, neural trajectories indicate not only which states are likely, but also which state-derivatives are likely. If a neural state is near previously observed states, it should be moving in a similar direction. MINT leverages this directionality.

Training-set trajectories can take various forms, depending on what is convenient to collect. Most simply, training data might include one trajectory per condition, with each condition corresponding to a discrete movement. Alternatively, one might instead employ one long trajectory spanning many movements. Another option is to employ many sub-trajectories, each briefer than a whole movement. The goal is simply for training-set trajectories to act as a scaffolding, outlining the manifold that might be occupied during decoding and the directions in which decoded trajectories are likely to be traveling.”

Later in that same section we stress that decoded trajectories can move along the ‘mesh’ in nonstereotyped ways:

“Although the mesh is formed of stereotyped trajectories, decoded trajectories can move along the mesh in non-stereotyped ways as long as they generally obey the flow-field implied by the training data. This flexibility supports many types of generalization, including generalization that is compositional in nature. Other types of generalization – e.g. from the green trajectories to the orange trajectories in Figure 1b – are unavailable when using MINT and are expected to be challenging for any method (as will be documented in a later section).”

The section “Training and decoding using MINT” has been revised to clarify the ways in which interpolation is flexible, allowing decoded movements to be globally very different from any library trajectory.

“To decode stereotyped trajectories, one could simply obtain the maximum-likelihood neural state from the library, then render a behavioral decode based on the behavioral state with the same values of c and k. This would be appropriate for applications in which conditions are categorical, such as typing or handwriting. Yet in most cases we wish for the trajectory library to serve not as an exhaustive set of possible states, but as a scaffolding for the mesh of possible states. MINT’s operations are thus designed to estimate any neural trajectory – and any corresponding behavioral trajectory – that moves along the mesh in a manner generally consistent with the trajectories in Ω.”

“…interpolation allows considerable flexibility. Not only is one not ‘stuck’ on a trajectory from Φ, one is also not stuck on trajectories created by weighted averaging of trajectories in Φ. For example, if cycling speed increases, the decoded neural state could move steadily up a scaffolding like that illustrated in Figure 1b (green). In such cases, the decoded trajectory might be very different in duration from any of the library trajectories. Thus, one should not think of the library as a set of possible trajectories that are selected from, but rather as providing a mesh-like scaffolding that defines where future neural states are likely to live and the likely direction of their local motion. The decoded trajectory may differ considerably from any trajectory within Ω.”

This flexibility is indeed used during movement. One empirical example is described in detail:

“During movement… angular phase was decoded with effectively no net drift over time. This is noteworthy because angular velocity on test trials never perfectly matched any of the trajectories in Φ. Thus, if decoding were restricted to a library trajectory, one would expect growing phase discrepancies. Yet decoded trajectories only need to locally (and approximately) follow the flow-field defined by the library trajectories. Based on incoming spiking observations, decoded trajectories speed up or slow down (within limits).

This decoding flexibility presumably relates to the fact that the decoded neural state is allowed to differ from the nearest state in Ω. To explore… [the text goes on to describe the new analysis in Figure 4d, which shows that the decoded state is typically not on any trajectory, though it is typically close to a trajectory].”

Thus, MINT’s operations allow considerable flexibility, including generalization that is compositional in nature. Yet R3 is still correct that there are other forms of generalization that are unavailable to MINT. This is now stressed at multiple points in the revision. However, under the perspective in Figure 1b, these forms of generalization are unavailable to any current method. Hence we made a second major change in response to this concern…  ii) We explicitly illustrate how the structure of the data determines when generalization is or isn’t possible. The new Figure 1a,b introduces the two perspectives, and the new Figure 6a,b lays out their implications for generalization. Under the perspective in Figure 6a, the reviewer is quite right: other methods can generalize in ways that MINT cannot. Under the perspective in Figure 6b, expectations are very different. Those expectations make testable predictions. Hence the third major change… iii) We have added an analysis of generalization, using a newly collected dataset. This dataset was collected using Neuropixels Probes during our Pac-Man force-tracking task. This dataset was chosen because it is unusually well-suited to distinguishing the predictions in Figure 6a versus Figure 6b. Finding a dataset that can do so is not simple. Consider R3’s point that training data should “explore the whole movement space and the associated neural space”. The physical simplicity of the Pac-Man task makes it unusually easy to confirm that the behavioral workspace has been fully explored. Importantly, under Figure 6b, this does not mean that the neural workspace has been fully explored, which is exactly what we wish to test when testing generalization. We do so, and compare MINT with a Wiener filter. A Wiener filter is an ideal comparison because it is simple, performs very well on this task, and should be able to generalize well under Figure 1a. Additionally, the Wiener filter (unlike the Kalman Filter) doesn’t leverage the assumption that neural activity reflects the derivative of force. This matters because we find that neural activity does not reflect dforce/dt in this task. The Wiener filter is thus the most natural choice of the interpretable methods whose assumptions match Figure 1a.

The new analysis is described in Figure 6c-g and accompanying text. Results are consistent with the predictions of Figure 6b. We are pleased to have been motivated to add this analysis for two reasons. First, it provides an additional way of evaluating the predictions of the two competing scientific perspectives that are at the heart of our study. Second, this analysis illustrates an underappreciated way in which generalization is likely to be challenging for any decode method. It can be tempting to think that the main challenge regarding generalization is to fully explore the relevant behavioral space. This makes sense if a behavioral space has “an associated neural space”. However, we are increasingly of the opinion that it doesn’t. Different tasks often involve different neural subspaces, even when behavioral subspaces overlap. We have even seen situations where motor output is identical but neural subspaces are quite different. These facts are relevant to any decoder, something highlighted in the revised Introduction:

“MINT’s performance confirms that there are gains to be made by building decoders whose assumptions match a different, possibly more accurate view of population activity. At the same time, our results suggest fundamental limits on decoder generalization. Under the assumptions in Figure 1b, it will sometimes be difficult or impossible for decoders to generalize to not-yet-seen tasks. We found that this was true regardless of whether one uses MINT or a more traditional method. This finding has implications regarding when and how generalization should be attempted.”

We have also added an analysis (Figure 6e) illustrating how MINT’s ability to compute likelihoods can be useful in detecting situations that may strain generalization (for any method). MINT is unusual in being able to compute and use likelihoods in this way.

Detailed responses to R3: we reproduce each of R3’s specific concerns below, but concentrate our responses on issues not already covered above.

Main comments: 

Comment 1. MINT does not generalize to different tasks, which is a main limitation for BCI utility compared with prior BCI decoders that have shown this generalizability as I review below. Specifically, given that MINT tabulates task-specific trajectories, it will not generalize to tasks that are not seen in the training data even when these tasks cover the exact same space (e.g., the same 2D computer screen and associated neural space). 

First, the authors provide a section on generalization, which is inaccurate because it mixes up two fundamentally different concepts: 1) collecting informative training data and 2) generalizing from task to task. The former is critical for any algorithm, but it does not imply the latter. For example, removing one direction of cycling from the training set as the authors do here is an example of generating poor training data because the two behavioral (and neural) directions are non-overlapping and/or orthogonal while being in the same space. As such, it is fully expected that all methods will fail. For proper training, the training data should explore the whole movement space and the associated neural space, but this does not mean all kinds of tasks performed in that space must be included in the training set (something MINT likely needs while modeling-based approaches do not). Many BCI studies have indeed shown this generalization ability using a model. For example, in Weiss et al. 2019, center-out reaching tasks are used for training and then the same trained decoder is used for typing on a keyboard or drawing on the 2D screen. In Gilja et al. 2012, training is on a center-out task but the same trained decoder generalizes to a completely different pinball task (hit four consecutive targets) and tasks requiring the avoidance of obstacles and curved movements. There are many more BCI studies, such as Jarosiewicz et al. 2015 that also show generalization to complex realworld tasks not included in the training set. Unlike MINT, these works can achieve generalization because they model the neural subspace and its association to movement. On the contrary, MINT models task-dependent neural trajectories, so the trained decoder is very task-dependent and cannot generalize to other tasks. So, unlike these prior BCIs methods, MINT will likely actually need to include every task in its library, which is not practical. 

I suggest the authors remove claims of generalization and modify their arguments throughout the text and abstract. The generalization section needs to be substantially edited to clarify the above points. Please also provide the BCI citations and discuss the above limitation of MINT for BCIs. 

As discussed above, R3’s concerns are accurate under the view in Figure 1a (and the corresponding Figure 6a). Under this view, a method such as that in Gilja et al. or Jarosiewicz et al. can find the correct subspace, model the correct neuron-behavior correlations, and generalize to any task that uses “the same 2D computer screen and associated neural space”, just as the reviewer argues. Under Figure 1b things are quite different.

This topic – and the changes we have made to address it – is covered at length above. Here we simply want to highlight an empirical finding: sometimes two tasks use the same neural subspace and sometimes they don’t. We have seen both in recent data, and it is can be very non-obvious which will occur based just on behavior. It does not simply relate to whether one is using the same physical workspace. We have even seen situations where the patterns of muscle activity in two tasks are nearly identical, but the neural subspaces are fairly different. When a new task uses a new subspace, neither of the methods noted above (Gilja nor Jarosiewicz) will generalize (nor will MINT). Generalizing to a new subspace is basically impossible without some yet-to-be-invented approach. On the other hand, there are many other pairs of tasks (center-out-reaching versus some other 2D cursor control) where subspaces are likely to be similar, especially if the frequency content of the behavior is similar (in our recent experience this is often critical). When subspaces are shared, most methods will generalize, and that is presumably why generalization worked well in the studies noted above.

Although MINT can also generalize in such circumstances, R3 is correct that, under the perspective in Figure 1a, MINT will be more limited than other methods. This is now carefully illustrated in Figure 6a. In this traditional perspective, MINT will fail to generalize in cases where new trajectories are near previously observed states, yet move in very different ways from library trajectories. The reason we don’t view this is a shortcoming is that we expect it to occur rarely (else tangling would be high). We thus anticipate the scenario in Figure 6b.

This is worth stressing because R3 states that our discussion of generalization “is inaccurate because it mixes up two fundamentally different concepts: 1) collecting informative training data and 2) generalizing from task to task.” We have heavily revised this section and improved it. However, it was never inaccurate. Under Figure 6b, these two concepts absolutely are mixed up. If different tasks use different neural subspaces, then this requires collecting different “informative training data” for each. One cannot simply count on having explored the physical workspace.

Comment 2. MINT is shown to achieve competitive/high performance in highly stereotyped datasets with structured trials, but worse performance on MC_RTT, which is not based on repeated trials and is less stereotyped. This shows that MINT is valuable for decoding in repetitive stereotyped use-cases. However, it also highlights a limitation of MINT for BCIs, which is that MINT may not work well for real-world and/or less-constrained setups such as typing, moving a robotic arm in 3D space, etc. This is again due to MINT being a lookup table with a library of stereotyped trajectories rather than a model. Indeed, the authors acknowledge that the lower performance on MC_RTT (Figure 4) may be caused by the lack of repeated trials of the same type. However, real-world BCI decoding scenarios will also not have such stereotyped trial structure and will be less/un-constrained, in which MINT underperforms. Thus, the claim in the abstract or lines 480-481 that MINT is an "excellent" candidate for clinical BCI applications is not accurate and needs to be qualified. The authors should revise their statements according and discuss this issue. They should also make the use-case of MINT on BCI decoding clearer and more convincing. 

We discussed, above, multiple changes and additions to the revision that were made to address these concerns. Here we briefly expand on the comment that MINT achieves “worse performance on MC_RTT, which is not based on repeated trials and is less stereotyped”. All decoders performed poorly on this task. MINT still outperformed the two traditional methods, but this was the only dataset where MINT did not also perform better (overall) than the expressive GRU and feedforward network. There are probably multiple reasons why. We agree with R3 that one likely reason is that this dataset is straining generalization, and MINT may have felt this strain more than the two machine-learning-based methods. Another potential reason is the structure of the training data, which made it more challenging to obtain library trajectories in the first place. Importantly, these observations do not support the view in Figure 1a. MINT still outperformed the Kalman and Wiener filters (whose assumptions align with Fig. 1a). To make these points we have added the following:

“Decoding was acceptable, but noticeably worse, for the MC_RTT dataset… As will be discussed below, every decode method achieved its worst estimates of velocity for the MC_RTT dataset. In addition to the impact of slower reaches, MINT was likely impacted by training data that made it challenging to accurate estimate library trajectories. Due to the lack of repeated trials, MINT used AutoLFADS to estimate the neural state during training. In principle this should work well. In practice AutoLFADS may have been limited by having only 10 minutes of training data. Because the random-target task involved more variable reaches, it may also have stressed the ability of all methods to generalize, perhaps for the reasons illustrated in Figure 1b.

The only dataset where MINT did not perform the best overall was the MC_RTT dataset, where it was outperformed by the feedforward network and GRU. As noted above, this may relate to the need for MINT to learn neural trajectories from training data that lacked repeated trials of the same movement (a design choice one might wish to avoid). Alternatively, the less-structured MC_RTT dataset may strain the capacity to generalize; all methods experienced a drop in velocity-decoding R2 for this dataset compared to the others. MINT generalizes somewhat differently than other methods, and may have been at a modest disadvantage for this dataset. A strong version of this possibility is that perhaps the perspective in Figure 1a is correct, in which case MINT might struggle because it cannot use forms of generalization that are available to other methods (e.g. generalization based on neuron-velocity correlations). This strong version seems unlikely; MINT continued to significantly outperform the Wiener and Kalman filters, which make assumptions aligned with Figure 1a.”

Comment 3. Related to 2, it may also be that MINT achieves competitive performance in offline and trial-based stereotyped decoding by overfitting to the trial structure in a given task, and thus may not generalize well to online performance due to overfitting. For example, a recent work showed that offline decoding performance may be overfitted to the task structure and may not represent online performance (Deo et al. 2023). Please discuss. 

We agree that a limitation of our study is that we do not test online performance. There are sensible reasons for this decision:

“By necessity and desire, all comparisons were made offline, enabling benchmarked performance across a variety of tasks and decoded variables, where each decoder had access to the exact same data and recording conditions.”

We recently reported excellent online performance in the cycling task with a different algorithm

(Schroeder et al. 2022). In the course of that study, we consistently found that improvements in our offline decoding translated to improvements in our online decoding. We thus believe that MINT (which improves on the offline performance of our older algorithm) is a good candidate to work very well online. Yet we agree this still remains to be seen. We have added the following to the Discussion:

“With that goal in mind, there exist three important practical considerations. First, some decode algorithms experience a performance drop when used online. One presumed reason is that, when decoding is imperfect, the participant alters their strategy which in turn alters the neural responses upon which decoding is based. Because MINT produces particularly accurate decoding, this effect may be minimized, but this cannot be known in advance. If a performance drop does indeed occur, one could adapt the known solution of retraining using data collected during online decoding [13]. Another presumed reason (for a gap between offline and online decoding) is that offline decoders can overfit the temporal structure in training data [107]. This concern is somewhat mitigated by MINT’s use of a short spike-count history, but MINT may nevertheless benefit from data augmentation strategies such as including timedilated versions of learned trajectories in the libraries”

Comment 4. Related to 2, since MINT requires firing rates to generate the library and simple averaging does not work for this purpose in the MC_RTT dataset (that does not have repeated trials), the authors needed to use AutoLFADS to infer the underlying firing rates. The fact that MINT requires the usage of another model to be constructed first and that this model can be computationally complex, will also be a limiting factor and should be clarified. 

This concern relates to the computational complexity of computing firing-rate trajectories during training. Usually, rates are estimated via trial-averaging, which makes MINT very fast to train. This was quite noticeable during the Neural Latents Benchmark competition. As one example, for the “MC_Scaling 5 ms Phase”, MINT took 28 seconds to train while GPFA took 30 minutes, the transformer baseline (NDT) took 3.5 hours, and the switching nonlinear dynamical system took 4.5 hours.

However, the reviewer is quite correct that MINT’s efficiency depends on the method used to construct the library of trajectories. As we note, “MINT is a method for leveraging a trajectory library, not a method for constructing it”. One can use trial-averaging, which is very fast. One can also use fancier, slower methods to compute the trajectories. We don’t view this as a negative – it simply provides options. Usually one would choose trial-averaging, but one does not have to. In the case of MC_RTT, one has a choice between LFADS and grouping into pseudo-conditions and averaging (which is fast). LFADS produces higher performance at the cost of being slower. The operator can choose which they prefer. This is discussed in the following section:

“For MINT, ‘training’ simply means computation of standard quantities (e.g. firing rates) rather than parameter optimization. MINT is thus typically very fast to train (Table 1), on the order of seconds using generic hardware (no GPUs). This speed reflects the simple operations involved in constructing the library of neural-state trajectories: filtering of spikes and averaging across trials. At the same time we stress that MINT is a method for leveraging a trajectory library, not a method for constructing it. One may sometimes wish to use alternatives to trial-averaging, either of necessity or because they improve trajectory estimates. For example, for the MC_RTT task we used AutoLFADS to infer the library. Training was consequently much slower (hours rather than seconds) because of the time taken to estimate rates. Training time could be reduced back to seconds using a different approach – grouping into pseudo-conditions and averaging – but performance was reduced. Thus, training will typically be very fast, but one may choose time-consuming methods when appropriate.”

Comment 5. I also find the statement in the abstract and paper that "computations are simple, scalable" to be inaccurate. The authors state that MINT's computational cost is O(NC) only, but it seems this is achieved at a high memory cost as well as computational cost in training. The process is described in section "Lookup table of log-likelihoods" on line [978-990]. The idea is to precompute the log-likelihoods for any combination of all neurons with discretization x all delay/history segments x all conditions and to build a large lookup table for decoding. Basically, the computational cost of precomputing this table is O(V^{Nτ} x TC) and the table requires a memory of O(V^{Nτ}), where V is the number of discretization points for the neural firing rates, N is the number of neurons, τ is the history length, T is the trial length, and C is the number of conditions. This is a very large burden, especially the V^{Nτ} term. This cost is currently not mentioned in the manuscript and should be clarified in the main text. Accordingly, computation claims should be modified including in the abstract. 

As discussed above, the manuscript has been revised to clarify that our statement was accurate.

Comment 6. In addition to the above technical concerns, I also believe the authors should clarify the logic behind developing MINT better. From a scientific standpoint, we seek to gain insights into neural computations by making various assumptions and building models that parsimoniously describe the vast amount of neural data rather than simply tabulating the data. For instance, low-dimensional assumptions have led to the development of numerous dimensionality reduction algorithms and these models have led to important interpretations about the underlying dynamics (e.g., fixed points/limit cycles). While it is of course valid and even insightful to propose different assumptions from existing models as the authors do here, they do not actually translate these assumptions into a new model. Without a model and by just tabulating the data, I don't believe we can provide interpretation or advance the understanding of the fundamentals behind neural computations. As such, I am not clear as to how this library building approach can advance neuroscience or how these assumptions are useful. I think the authors should clarify and discuss this point. 

As requested, a major goal of the revision has been to clarify the scientific motivations underlying MINT’s design. In addition to many textual changes, we have added figures (Figures 1a,b and 6a,b) to outline the two competing scientific perspectives that presently exist. This topic is also addressed by extensions of existing analyses and by new analyses (e.g. Figure 6c-g). 

In our view these additions have dramatically improved the manuscript. This is especially true because we think R3’s concerns, expressed above, are reasonable. If the perspective in Figure 1a is correct, then R3 is right and MINT is essentially a hack that fails to model the data. MINT would still be effective in many circumstances (as we show), but it would be unprincipled. This would create limitations, just as the reviewer argues. On the other hand, if the perspective in Figure 1b is correct, then MINT is quite principled relative to traditional approaches. Traditional approaches make assumptions (a fixed subspace, consistent neuron-kinematic correlations) that are not correct under Figure 1b.

We don’t expect R3 to agree with our scientific perspective at this time (though we hope to eventually convince them). To us, the key is that we agree with R3 that the manuscript needs to lay out the different perspectives and their implications, so that readers have a good sense of the possibilities they should be considering. The revised manuscript is greatly improved in this regard.

Comment 7. Related to 6, there seems to be a logical inconsistency between the operations of MINT and one of its three assumptions, namely, sparsity. The authors state that neural states are sparsely distributed in some neural dimensions (Figure 1a, bottom). If this is the case, then why does MINT extend its decoding scope by interpolating known neural states (and behavior) in the training library? This interpolation suggests that the neural states are dense on the manifold rather than sparse, thus being contradictory to the assumption made. If interpolation-based dense meshes/manifolds underlie the data, then why not model the neural states through the subspace or manifold representations? I think the authors should address this logical inconsistency in MINT, especially since this sparsity assumption also questions the low-dimensional subspace/manifold assumption that is commonly made. 

We agree this is an important issue, and have added an analysis on this topic (Figure 4d). The key question is simple and empirical: during decoding, does interpolation cause MINT to violate the assumption of sparsity? R3 is quite right that in principle it could. If spiking observations argue for it, MINT’s interpolation could create a dense manifold during decoding rather than a sparse one. The short answer is that empirically this does not happen, in agreement with expectations under Figure 1b. Rather than interpolating between distant states and filling in large ‘voids’, interpolation is consistently local. This is a feature of the data, not of the decoder (MINT doesn’t insist upon sparsity, even though it is designed to work best in situations where the manifold is sparse).

In addition to adding Figure 4d, we added the following (in an earlier section):

“The term mesh is apt because, if MINT’s assumptions are correct, interpolation will almost always be local. If so, the set of decodable states will resemble a mesh, created by line segments connecting nearby training-set trajectories. However, this mesh-like structure is not enforced by MINT’s operations. Interpolation could, in principle, create state-distributions that depart from the assumption of a sparse manifold. For example, interpolation could fill in the center of the green tube in Figure 1b, resulting in a solid manifold rather than a mesh around its outer surface. However, this would occur only if spiking observations argued for it. As will be documented below, we find that essentially all interpolation is local.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

I appreciate the detailed methods section, however, more specifics should be integrated into the main text. For example on Line 238, it should additionally be stated how many minutes were used for training and metrics like the MAE which is used later should be reported here.

Thank you for this suggestion. We now report the duration of training data in the main text:

“Decoding R^2 was .968 over ~7.1 minutes of test trials based on ~4.4 minutes of training data.”

We have also added similar specifics throughout the manuscript, e.g. in the Fig. 5 legend:

“Results are based on the following numbers of training / test trials: MC\_Cycle (174 train, 99 test), MC\_Maze (1721 train, 574 test), Area2\_Bump (272 train, 92 test), MC\_RTT (810 train, 268 test).”

Similar additions were made to the legends for Fig. 6 and 8. Regarding the request to add MAE for the multitask network, we did not do so for the simple reason that the decoded variable (muscle activity) has arbitrary units. The raw MAE is thus not meaningful. We could of course have normalized, but at this point the MAE is largely redundant with the correlation. In contrast, the MAE is useful when comparing across the MC_Maze, Area2_Bump, and MC_RTT datasets, because they all involve the same scale (cm/s).

Regarding the MC_RTT task, AutoLFADS was used to obtain robust spike rates, as reported in the methods. However, the rationale for splitting the neural trajectories after AutoLFADS is unclear. If the trajectories were split based on random recording gaps, this might lead to suboptimal performance? It might be advantageous to split them based on a common behavioural state? 

When learning neural trajectories via AutoLFADS, spiking data is broken into short (but overlapping) segments, rates are estimated for each segment via AutoLFADs, and these rates are then stitched together across segments into long neural trajectories. If there had been no recording gaps, these rates could have been stitched into a single neural trajectory for this dataset. However, the presence of recording gaps left us no choice but to stitch together these rates into more than one trajectory. Fortunately, recording gaps were rare: for the decoding analysis of MC_RTT there were only two recording gaps and therefore three neural trajectories, each ~2.7 minutes in duration. 

We agree that in general it is desirable to learn neural trajectories that begin and end at behaviorallyrelevant moments (e.g. in between movements). However, having these trajectories potentially end midmovement is not an issue in and of itself. During decoding, MINT is never stuck on a trajectory. Thus, if MINT were decoding states near the end of a trajectory that was cut short due to a training gap, it would simply begin decoding states from other trajectories or elsewhere along the same trajectory in subsequent moments. We could have further trimmed the three neural trajectories to begin and end at behaviorallyrelevant moments, but chose not to as this would have only removed a handful of potentially useful states from the library.

We now describe this in the Methods:

“Although one might prefer trajectory boundaries to begin and end at behaviorally relevant moments (e.g. a stationary state), rather than at recording gaps, the exact boundary points are unlikely to be consequential for trajectories of this length that span multiple movements. If MINT estimates a state near the end of a long trajectory, its estimate will simply jump to another likely state on a different trajectory (or earlier along the same trajectory) in subsequent moments. Clipping the end of each trajectory to an earlier behaviorally-relevant moment would only remove potentially useful states from the libraries.”

Are the training and execution times in Table 1 based on pure Matlab functions or Mex files? If it's Mex files as suggested by the code, it would be good to mention this in the Table caption.

They are based on a combination of MATLAB and MEX files. This is now clarified in the table caption:

“Timing measurements taken on a Macbook Pro (on CPU) with 32GB RAM and a 2.3 GHz 8-Core Intel Core i9 processor. Training and execution code used for measurements was written in MATLAB (with the core recursion implemented as a MEX file).”

As the method most closely resembles a Bayesian decoder it would be good to compare performance against a Naive Bayes decoder. 

We agree and have now done so. The following has been added to the text:

“A natural question is thus whether a simpler Bayesian decoder would have yielded similar results. We explored this possibility by testing a Naïve Bayes regression decoder [85] using the MC_Maze dataset. This decoder performed poorly, especially when decoding velocity (R2 = .688 and .093 for hand position and velocity, respectively), indicating that the specific modeling assumptions that differentiate MINT from a naive Bayesian decoder are important drivers of MINT’s performance.”

Line 199 Typo: The assumption of stereotypy trajectory also enables neural states (and decoded behaviors) to be updated in between time bins. 

Fixed

Table 3: It's unclear why the Gaussian binning varies significantly across different datasets. Could the authors explain why this is the case and what its implications might be? 

We have added the following description in the “Filtering, extracting, and warping data on each trial” subsection of the Methods to discuss how 𝜎 may vary due to the number of trials available for training and how noisy the neural data for those trials is:

“First, spiking activity for each neuron on each trial was temporally filtered with a Gaussian to yield single-trial rates. Table 3 reports the Gaussian standard deviations σ (in milliseconds) used for each dataset. Larger values of σ utilize broader windows of spiking activity when estimating rates and therefore reduce variability in those rate estimates. However, large σ values also yield neural trajectories with less fine-grained temporal structure. Thus, the optimal σ for a dataset depends on how variable the rate estimates otherwise are.”

An implementation of the method in an open-source programming language could further enhance the widespread use of the tool. 

We agree this would be useful, but have yet not implemented the method in any other programming languages. Implementation in Python is still a future goal.

Reviewer #2 (Recommendations For The Authors): 

- Figures 4 and 5 should show the error bars on the horizontal axis rather than portraying them vertically. 

[Note that these are now Figures 5 and 6]

The figure legend of Figure 5 now clarifies that the vertical ticks are simply to aid visibility when symbols have very similar means and thus overlap visually. We don’t include error bars (for this analysis) because they are very small and would mostly be smaller than the symbol sizes. Instead, to indicate certainty regarding MINT’s performance measurements, the revised text now gives error ranges for the correlations and MAE values in the context of Figure 4c. These error ranges were computed as the standard deviation of the sampling distribution (computed via resampling of trials) and are thus equivalent to SEMs. The error ranges are all very small; e.g. for the MC_Maze dataset the MAE for x-velocity is 4.5 +/- 0.1 cm/s. (error bars on the correlations are smaller still).

Thus, for a given dataset, we can be quite certain of how well MINT performs (within ~2% in the above case). This is reassuring, but we also don’t want to overemphasize this accuracy. The main sources of variability one should be concerned about are: 1) different methods can perform differentially well for different brain areas and tasks, 2) methods can decode some behavioral variables better than others, and 3) performance depends on factors like neuron-count and the number of training trials, in ways that can differ across decode methods. For this reason, the study examines multiple datasets, across tasks and brain areas, and measures performance for a range of decoded variables. We also examine the impact of training-set-size (Figure 8a) and population size (solid traces in Fig. 8b, see R2’s next comment below). 

There is one other source of variance one might be concerned about, but it is specific to the neuralnetwork approaches: different weight initializations might result in different performance. For this reason, each neural-network approach was trained ten times, with the average performance computed. The variability around this average was very small, and this is now stated in the Methods.

“For the neural networks, the training/testing procedure was repeated 10 times with different random seeds. For most behavioral variables, there was very little variability in performance across repetitions. However, there were a few outliers for which variability was larger. Reported performance for each behavioral group is the average performance across the 10 repetitions to ensure results were not sensitive to any specific random initialization of each network.”

- For Figure 6, it is unclear whether the neuron-dropping process was repeated multiple times. If not, it should be since the results will be sensitive to which particular subsets of neurons were "dropped". In this case, the results presented in Figure 6 should include error bars to describe the variability in the model performance for each decoder considered. 

A good point. The results in Figure 8 (previously Figure 6) were computed by averaging over the removal of different random subsets of neurons (50 subsets per neuron count), just as the reviewer requests. The figure has been modified to include the standard deviation of performance across these 50 subsets. The legend clarifies how this was done.

Reviewer #3 (Recommendations For The Authors): 

Other comments: 

(1) [Line 185-188] The authors argue that in a 100-dimensional space with 10 possible discretized values, 10^100 potential neural states need to be computed. But I am not clear on this. This argument seems to hold only in the absence of a model (as in MINT). For a model, e.g., Kalman filter or AutoLFADS, information is encoded in the latent state. For example, a simple Kalman filter for a linear model can be used for efficient inference. This 10^100 computation isn't a general problem but seems MINT-specific, please clarify. 

We agree this section was potentially confusing. It has been rewritten. We were simply attempting to illustrate why maximum likelihood computations are challenging without constraints. MINT simplifies this problem by adding constraints, which is why it can readily provide data likelihoods (and can do so using a Poisson model). The rewritten section is below:

“Even with 1000 samples for each of the neural trajectories in Figure 3, there are only 4000 possible neural states for which log-likelihoods must be computed (in practice it is fewer still, see Methods). This is far fewer than if one were to naively consider all possible neural states in a typical rate- or factor-based subspace. It thus becomes tractable to compute log-likelihoods using a Poisson observation model. A Poisson observation model is usually considered desirable, yet can pose tractability challenges for methods that utilize a continuous model of neural states. For example, when using a Kalman filter, one is often restricted to assuming a Gaussian observation model to maintain computational tractability “

(2) [Figure 6b] Why do the authors set the dropped neurons to zero in the "zeroed" results of the robustness analysis? Why not disregard the dropped neurons during the decoding process? 

We agree the terminology we had used in this section was confusing. We have altered the figure and rewritten the text. The following, now at the beginning of that section, addresses the reviewer’s query: 

“It is desirable for a decoder to be robust to the unexpected loss of the ability to detect spikes from some neurons. Such loss might occur while decoding, without being immediately detected. Additionally, one desires robustness to a known loss of neurons / recording channels. For example, there may have been channels that were active one morning but are no longer active that afternoon. At least in principle, MINT makes it very easy to handle this second situation: there is no need to retrain the decoder, one simply ignores the lost neurons when computing likelihoods. This is in contrast to nearly all other methods, which require retraining because the loss of one neuron alters the optimal parameters associated with every other neuron.”

The figure has been relabeled accordingly; instead of the label ‘zeroed’, we use the label ‘undetected neuron loss’.

(3) Authors should provide statistical significance on their results, which they already did for Fig. S3a,b,c but missing on some other figures/places. 

We have added error bars in some key places, including in the text when quantifying MINT’s performance in the context of Figure 4. Importantly, error bars are only as meaningful as the source of error they assess, and there are reasons to be careful given this. The standard method for putting error bars on performance is to resample trials, which is indeed what we now report. These error bars are very small. For example, when decoding horizontal velocity for the MC_Maze dataset, the correlation between MINT’s decode and the true velocity had a mean and SD of the sampling distribution of 0.963 +/- 0.001. This means that, for a given dataset and target variable, we have enough trials/data that we can be quite certain of how well MINT performs. However, we want to be careful not to overstate this certainty. What one really wants to know is how well MINT performs across a variety of datasets, brain areas, target variables, neuron counts, etc. It is for this reason that we make multiple such comparisons, which provides a more valuable view of performance variability.

For Figure 7, error bars are unavailable. Because this was a benchmark, there was exactly one test-set that was never seen before. This is thus not something that could be resampled many times (that would have revealed the test data and thus invalidated the benchmark, not to mention that some of these methods take days to train). We could, in principle, have added resampling to Figure 5. In our view it would not be helpful and could be misleading for the reasons noted above. If we computed standard errors using different train/test partitions, they would be very tight (mostly smaller than the symbol sizes), which would give the impression that one can be quite certain of a given R^2 value. Yet variability in the train/test partition is not the variability one is concerned about in practice. In practice, one is concerned about whether one would get a similar R^2 for a different dataset, or brain area, or task, or choice of decoded variable. Our analysis thus concentrated on showing results across a broad range of situations. In our view this is a far more relevant way of illustrating the degree of meaningful variability (which is quite large) than resampling, which produces reassuringly small but (mostly) irrelevant standard errors.

Error bars are supplied in Figure 8b. These error bars give a sense of variability across re-samplings of the neural population. While this is not typically the source of variability one is most concerned about, for this analysis it becomes appropriate to show resampling-based standard errors because a natural concern is that results may depend on which neurons were dropped. So here it is both straightforward, and desirable, to compute standard errors. (The fact that MINT and the Wiener filter can be retrained many times swiftly was also key – this isn’t true of the more expressive methods). Figure S1 also uses resampling-based confidence intervals for similar reasons.

(4) [Line 431-437] Authors state that MINT outperforms other methods with the PSTH R^2 metric (trial-averaged smoothed spikes for each condition). However, I think this measure may not provide a fair comparison and is confounded because MINT's library is built using PSTH (i.e., averaged firing rate) but other methods do not use the PSTH. The author should clarify this. 

The PSTH R^2 metric was not created by us; it was part of the Neural Latents Benchmark. They chose it because it ensures that a method cannot ‘cheat’ (on the Bits/Spike measure) by reproducing fine features of spiking while estimating rates badly. We agree with the reviewer’s point: MINT’s design does give it a potential advantage in this particular performance metric. This isn’t a confound though, just a feature. Importantly, MINT will score well on this metric only if MINT’s neural state estimate is accurate (including accuracy in time). Without accurate estimation of the neural state at each time, it wouldn’t matter that the library trajectory is based on PSTHs. This is now explicitly stated:

“This is in some ways unsurprising: MINT estimates neural states that tend to resemble (at least locally) trajectories ‘built’ from training-set-derived rates, which presumably resemble test-set rates. Yet strong performance is not a trivial consequence of MINT’s design. MINT does not ‘select’ whole library trajectories; PSTH R2 will be high only if condition (c), index (k), and the interpolation parameter (α) are accurately estimated for most moments.”

Author response:

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

Public Reviews:

Reviewer #1 (Public Review): 

In the presented manuscript, the authors investigate how neural networks can learn to replay presented sequences of activity. Their focus lies on the stochastic replay according to learned transition probabilities. They show that based on error-based excitatory and balance-based inhibitory plasticity networks can selforganize towards this goal. Finally, they demonstrate that these learning rules can recover experimental observations from song-bird song learning experiments. 

Overall, the study appears well-executed and coherent, and the presentation is very clear and helpful. However, it remains somewhat vague regarding the novelty. The authors could elaborate on the experimental and theoretical impact of the study, and also discuss how their results relate to those of Kappel et al, and others (e.g., Kappel et al (doi.org/10.1371/journal.pcbi.1003511))). 

We agree with the reviewer that our previous manuscript lacked comparison with previously published similar works. While Kappel et al. demonstrated that STDP in winner-take-all circuits can approximate online learning of hidden Markov models (HMMs), a key distinction from our model is that their neural representations acquire deterministic sequential activations, rather than exhibiting stochastic transitions governing Markovian dynamics. Specifically, in their model, the neural representation of state B would be different in the sequences ABC and CBA, resulting in distinct deterministic representations like ABC and C'B'A', where ‘A’ and ‘A'’ are represented by different neural states (e.g., activations of different cell assemblies). In contrast, our network learns to generate stochastically transitioning cell assemblies which replay Markovian trajectories of spontaneous activity obeying the learned transition probabilities between neural representations of states. For example, starting from reactivation from assembly ‘A’, there may be an 80% probability to transition to assembly ‘B’ and 20% to ‘C’. Although Kappel et al.'s model successfully solves HMMs, their neural representations do not themselves stochastically transition between states according to the learned model. Similar to the Kappel et al.'s model, while the models proposed in Barber (2002) and Barber and Agakov (2002) learn the Markovian statistics, these models learned a static spatiotemporal input patterns only and how assemblies of neurons show stochastic transition in spontaneous activity has been still unclear. In contrast with these models, our model captures the probabilistic neural state trajectories, allowing spontaneous replay of experienced sequences with stochastic dynamics matching the learned environmental statistics.

We have included new sentences for explain these in ll. 509-533 in the revised manuscript.

Overall, the work could benefit if there was either (A) a formal analysis or derivation of the plasticity rules involved and a formal justification of the usefulness of the resulting (learned) neural dynamics; 

We have included a derivation of our plasticity rules in ll. 630-670 in the revised manuscript. Consistent with our claim that excitatory plasticity updates the excitatory synapse to predict output firing rates, we have shown that the corresponding cost function measures the discrepancy between the recurrent prediction and the output firing rate. Similarly, for inhibitory plasticity, we defined the cost function that evaluates the difference between the excitatory and inhibitory potential within each neuron. We showed that the resulting inhibitory plasticity rule updates the inhibitory synapses to maintain the excitation-inhibition balance.

and/or (B) a clear connection of the employed plasticity rules to biological plasticity and clear testable experimental predictions. Thus, overall, this is a good work with some room for improvement. 

Our proposed plasticity mechanism could be implemented through somatodendritic interactions. Analogous to previous computational works (Urbanczik and Senn., 2014; Asabuki and Fukai., 2020; Asabuki et al., 2022), our model suggests that somatic responses may encode the stimulus-evoked neural activity states, while dendrites encode predictions based on recurrent dynamics that aim to minimize the discrepancy between somatic and dendritic activity. To directly test this hypothesis, future experimental studies could simultaneously record from both somatic and dendritic compartments to investigate how they encode evoked responses and predictive signals during learning (Francioni et al., 2022).

We have included new sentences for explain these in ll. 476-484 in the revised manuscript.

Reviewer #2 (Public Review): 

Summary: 

This work proposes a synaptic plasticity rule that explains the generation of learned stochastic dynamics during spontaneous activity. The proposed plasticity rule assumes that excitatory synapses seek to minimize the difference between the internal predicted activity and stimulus-evoked activity, and inhibitory synapses try to maintain the E-I balance by matching the excitatory activity. By implementing this plasticity rule in a spiking recurrent neural network, the authors show that the state-transition statistics of spontaneous excitatory activity agree with that of the learned stimulus patterns, which are reflected in the learned excitatory synaptic weights. The authors further demonstrate that inhibitory connections contribute to well-defined state transitions matching the transition patterns evoked by the stimulus. Finally, they show that this mechanism can be expanded to more complex state-transition structures including songbird neural data. 

Strengths: 

This study makes an important contribution to computational neuroscience, by proposing a possible synaptic plasticity mechanism underlying spontaneous generations of learned stochastic state-switching dynamics that are experimentally observed in the visual cortex and hippocampus. This work is also very clearly presented and well-written, and the authors conducted comprehensive simulations testing multiple hypotheses. Overall, I believe this is a well-conducted study providing interesting and novel aspects of the capacity of recurrent spiking neural networks with local synaptic plasticity. 

Weaknesses: 

This study is very well-thought-out and theoretically valuable to the neuroscience community, and I think the main weaknesses are in regard to how much biological realism is taken into account. For example, the proposed model assumes that only synapses targeting excitatory neurons are plastic, and uses an equal number of excitatory and inhibitory neurons. 

We agree with the reviewer. The network shown in the previous manuscript consists of an equal number of excitatory and inhibitory neurons, which seems to lack biological plausibility. Therefore, we first tested whether a biologically plausible scenario would affect learning performance by setting the ratio of excitatory to inhibitory neurons to 80% and 20% (Supplementary Figure 7a; left). Even in such a scenario, the network still showed structured spontaneous activity (Supplementary Figure 7a; center), with transition statistics of replayed events matching the true transition probabilities (Supplementary Figure 7a; right). We then asked whether the model with our plasticity rule applied to all synapses would reproduce the corresponding stochastic transitions. We found that the network can learn transition statistics but only under certain conditions. The network showed only weak replay and failed to reproduce the appropriate transition (Supplementary Fig. 7b) if the inhibitory neurons were no longer driven by the synaptic currents reflecting the stimulus, due to a tight balance of excitatory and inhibitory currents on the inhibitory neurons. We then tested whether the network with all synapses plastic can learn transition statistics if the external inputs project to the inhibitory neurons as well. We found that, when each stimulus pattern activates a non-overlapping subset of neurons, the network does not exhibit the correct stochastic transition of assembly reactivation (Supplementary Fig. 7c). Interestingly, when each neuron's activity is triggered by multiple stimuli and has mixed selectivity, the reactivation reproduced the appropriate stochastic transitions (Supplementary Fig. 7d).

We have included these new results as new Supplementary Figure 7 and they are explained in ll.215-230 in the revised manuscript.

The model also assumes Markovian state dynamics while biological systems can depend more on history. This limitation, however, is acknowledged in the Discussion. 

We have included the following sentence to provide a possible solution to this limitation: “Therefore, to learn higher-order stochastic transitions, recurrent neural networks like ours may need to integrate higher-order inputs with longer time scales.” in ll.557-559 in the revised manuscript. 

Finally, to simulate spontaneous activity, the authors use a constant input of 0.3 throughout the study. Different amplitudes of constant input may correspond to different internal states, so it will be more convincing if the authors test the model with varying amplitudes of constant inputs. 

We thank the reviewer for pointing this out. In the revised manuscript, we have tested constant input with three different strengths. If the strength is moderate, the network showed accurate encoding of transition statistics in the spontaneous activity as we have seen in Fig.2. We have additionally shown that the weaker background input causes spontaneous activity with lower replay rate, which in turn leads to high variance of encoded transition, while stronger inputs make assembly replay transitions more uniform. We have included these new results as new Supplementary Figure 6 and they are explained in ll.211214 in the revised manuscript.

Reviewer #3 (Public Review): 

Summary: 

Asabuki and Clopath study stochastic sequence learning in recurrent networks of Poisson spiking neurons that obey Dale's law. Inspired by previous modeling studies, they introduce two distinct learning rules, to adapt excitatory-to-excitatory and inhibitory-to-excitatory synaptic connections. Through a series of computer experiments, the authors demonstrate that their networks can learn to generate stochastic sequential patterns, where states correspond to non-overlapping sets of neurons (cell assemblies) and the state-transition conditional probabilities are first-order Markov, i.e., the transition to a given next state only depends on the current state. Finally, the authors use their model to reproduce certain experimental songbird data involving highly-predictable and highly-uncertain transitions between song syllables. 

Strengths: 

This is an easy-to-follow, well-written paper, whose results are likely easy to reproduce. The experiments are clear and well-explained. The study of songbird experimental data is a good feature of this paper; finches are classical model animals for understanding sequence learning in the brain. I also liked the study of rapid task-switching, it's a good-to-know type of result that is not very common in sequence learning papers. 

Weaknesses: 

While the general subject of this paper is very interesting, I missed a clear main result. The paper focuses on a simple family of sequence learning problems that are well-understood, namely first-order Markov sequences and fully visible (nohidden-neuron) networks, studied extensively in prior work, including with spiking neurons. Thus, because the main results can be roughly summarized as examples of success, it is not entirely clear what the main point of the authors is. 

We apologize the reviewer that our main claim was not clear. While various computational studies have suggested possible plasticity mechanisms for embedding evoked activity patterns or their probability structures into spontaneous activity (Litwin-Kumar et al., Nat. Commun. 2014, Asabuki and Fukai., Biorxiv 2023), how transition statistics of the environment are learned in spontaneous activity is still elusive and poorly understood. Furthermore, while several network models have been proposed to learn Markovian dynamics via synaptic plasticity (Brea, et al. (2013); Pfister et al. (2004); Kappel et al. (2014)), they have been limited in a sense that the learned network does not show stochastic transition in a neural state space. For instance, while Kappel et al. demonstrated that STDP in winner-take-all circuits can approximate online learning of hidden Markov models (HMMs), a key distinction from our model is that their neural representations acquire deterministic sequential activations, rather than exhibiting stochastic transitions governing Markovian dynamics. Specifically, in their model, the neural representation of state B would be different in the sequences ABC and CBA, resulting in distinct deterministic representations like ABC and C'B'A', where ‘A’ and ‘A'’ are represented by different neural states (e.g., activations of different cell assemblies). In contrast, our network learns to generate stochastically transitioning cell assemblies that replay Markovian trajectories of spontaneous activity obeying the learned transition probabilities between neural representations of states. For example, starting from reactivation from assembly ‘A’, there may be an 80% probability to transition to assembly ‘B’ and 20% to ‘C’. Although Kappel et al.'s model successfully solves HMMs, their neural representations do not themselves stochastically transition between states according to the learned model. Similar to the Kappel et al.'s model, while the models proposed in Barber (2002) and Barber and Agakov (2002) learn the Markovian statistics, these models learned a static spatiotemporal input patterns only and how assemblies of neurons show stochastic transition in spontaneous activity has been still unclear. In contrast with these models, our model captures the probabilistic neural state trajectories, allowing spontaneous replay of experienced sequences with stochastic dynamics matching the learned environmental statistics.

We have explained this point in ll.509-533 in the revised manuscript.

Going into more detail, the first major weakness I see in this paper is the heuristic choice of learning rules. The paper studies Poisson spiking neurons (I return to this point below), for which learning rules can be derived from a statistical objective, typically maximum likelihood. For fully-visible networks, these rules take a simple form, similar in many ways to the E-to-E rule introduced by the authors. This more principled route provides quite a lot of additional understanding on what is to be expected from the learning process. 

We thank the reviewer for pointing this out. To better demonstrate the function of our plasticity rules, we have included the derivation of the rules of synaptic plasticity in ll. 630-670 in the revised manuscript. Consistent with our claim that excitatory plasticity updates the excitatory synapse to predict output firing rates, we have shown that the corresponding cost function measures the discrepancy between the recurrent prediction and the output firing rate. Similarly, for inhibitory plasticity, we defined the cost function that evaluates the difference between the excitatory and inhibitory potential within each neuron. We showed that the resulting inhibitory plasticity rule updates the inhibitory synapses to maintain the excitation-inhibition balance.

For instance, should maximum likelihood learning succeed, it is not surprising that the statistics of the training sequence distribution are reproduced. Moreover, given that the networks are fully visible, I think that the maximum likelihood objective is a convex function of the weights, which then gives hope that the learning rule does succeed. And so on. This sort of learning rule has been studied in a series of papers by David Barber and colleagues [refs. 1, 2 below], who applied them to essentially the same problem of reproducing sequence statistics in recurrent fully-visible nets. It seems to me that one key difference is that the authors consider separate E and I populations, and find the need to introduce a balancing I-to-E learning rule. 

The reviewer’s understanding that inhibitory plasticity to maintain EI balance is one of a critical difference from previous works is correct. However, we believe that the most striking point of our study is that we have shown numerically that predictive plasticity rules enable recurrent networks to learn and replay the assembly activations whose transition statistics match those of the evoked activity. Please see our reply above.

Because the rules here are heuristic, a number of questions come to mind. Why these rules and not others - especially, as the authors do not discuss in detail how they could be implemented through biophysical mechanisms? When does learning succeed or fail? What is the main point being conveyed, and what is the contribution on top of the work of e.g. Barber, Brea, et al. (2013), or Pfister et al. (2004)? 

Our proposed plasticity mechanism could be implemented through somatodendritic interactions. Analogous to previous computational works (Senn, Asabuki), our model suggests that somatic responses may encode the stimulusevoked neural activity states, while dendrites encode predictions based on recurrent dynamics that aim to minimize the discrepancy between somatic and dendritic activity. To directly test this hypothesis, future experimental studies could simultaneously record from both somatic and dendritic compartments to investigate how they encode evoked responses and predictive signals during learning.

To address the point of the reviewer, we conducted addionnal simulations to test where the model fails. We found that the model with our plasticity rule applied to all synapses only showed faint replays and failed to replay the appropriate transition (Supplementary Fig. 7b). This result is reasonable because the inhibitory neurons were no longer driven by the synaptic currents reflecting the stimulus, due to a tight balance of excitatory and inhibitory currents on the inhibitory neurons. Our model predicts that mixed selectivity in the inhibitory population is crucial to learn an appropriate transition statistics (Supplementary Fig. 7d). Future work should clarify the role of synaptic plasticity on inhibitory neurons, especially plasticity at I to I synapses. We have explained this result as new supplementary Figure7 in the revised manuscript.

The use of a Poisson spiking neuron model is the second major weakness of the study. A chief challenge in much of the cited work is to generate stochastic transitions from recurrent networks of deterministic neurons. The task the authors set out to do is much easier with stochastic neurons; it is reasonable that the network succeeds in reproducing Markovian sequences, given an appropriate learning rule. I believe that the main point comes from mapping abstract Markov states to assemblies of neurons. If I am right, I missed more analyses on this point, for instance on the impact that varying cell assembly size would have on the findings reported by the authors.

The reviewer’s understanding is correct. Our main point comes from mapping Markov statistics to replays of cell assemblies. In the revised manuscript, we performed additional simulations to ask whether varying the size of the cell assemblies would affect learning. We ran simulations with two different configurations in the task shown in Figure 2. The first configuration used three assemblies with a size ratio of 1:1.5:2. After training, these assemblies exhibited transition statistics that closely matched those of the evoked activity (Supplementary Fig.4a,b). In contrast, the second configuration, which used a size ratio of 1:2:3, showed worse performance compared to the 1:1.5:2 case (Supplementary Fig.4c,d). These results suggest that the model can learn appropriate transition statistics as long as the size ratio of the assemblies is not drastically varied.

Finally, it was not entirely clear to me what the main fundamental point in the HVC data section was. Can the findings be roughly explained as follows: if we map syllables to cell assemblies, for high-uncertainty syllable-to-syllable transitions, it becomes harder to predict future neural activity? In other words, is the main point that the HVC encodes syllables by cell assemblies? 

The reviewer's understanding is correct. We wanted to show that if the HVC learns transition statistics as a replay of cell assemblies, a high-uncertainty syllable-to-syllable transition would make predicting future reactivations more difficult, since trial-averaged activities (i.e., poststimulus activities; PSAs) marginalized all possible transitions in the transition diagram.

(1) Learning in Spiking Neural Assemblies, David Barber, 2002. URL: https://proceedings.neurips.cc/paper/2002/file/619205da514e83f869515c782a328d3c-Paper.pdf  

(2) Correlated sequence learning in a network of spiking neurons usingmaximum likelihood, David Barber, Felix Agakov, 2002. URL: http://web4.cs.ucl.ac.uk/staff/D.Barber/publications/barber-agakovTR0149.pdf  

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

In more detail: 

A) Theoretical analysis 

The plasticity rules in the study are introduced with a vague reference to previous theoretical studies of others. Doing this, one does not provide any formal insight as to why these plasticity rules should enable one to learn to solve the intended task, and whether they are optimal in some respect. This becomes noticeable, especially in the discussion of the importance of inhibitory balance, which does not go into any detail, but rather only states that its required, both in the results and discussion sections. Another unclarity appears when error-based learning is discussed and compared to Hebbian plasticity, which, as you state, "alone is insufficient to learn transition probabilities". It is not evident how this claim is warranted, nor why error-based plasticity in comparison should be able to perform this (other than referring to the simulation results). Please either clarify formally (or at least intuitively) how plasticity rules result in the mentioned behavior, or alternatively acknowledge explicitly the (current) lack of intuition. 

The lack of formal discussion is a relevant shortcoming compared to previous research that showed very similar results with formally more rigorous and principled approaches. In particular, Kappel et al derived explicitly how neural networks can learn to sample from HMMs using STDP and winner-take-all dynamics. Even though this study has limitations, the relation with respect to that work should be made very clear; potentially the claims of novelty of some results (sampling) should be adjusted accordingly. See also Yanping Huang, Rajesh PN Rao (NIPS 2014), and possibly other publications. While it might be difficult to formally justify the learning rules post-hoc, it would be very helpful to the field if you very clearly related your work to that of others, where learning rules have been formally justified, and elaborate on the intuition of how the employed rules operate and interact (especially for inhibition). 

Lastly, while the importance of sampling learned transition probabilities is discussed, the discussion again remains on a vague level, characterized by the lack of references in the relevant paragraphs. Ideally, there should be a proof of concept or a formal understanding of how the learned behaviour enables to solve a problem that is not solved by deterministic networks. Please incorporate also the relation to the literature on neural sampling/planning/RL etc. and substantiate the claims with citations. 

We have included sentences in ll. 691-696 in the revised manuscript to explain that for Poisson spiking neurons, the derived learning rule is equivalent to the one that minimizes the Kullback-Leibler divergence between the distributions of output firing and the dendritic prediction, in our case, the recurrent prediction (Asabuki and Fukai; 2020). Thus, the rule suggests that the recurrent prediction learns the statistical model of the evoked activity, which in turn allows the network to reproduce the learned transition statistics.

We have also added a paragraph to discuss the differences between previously published similar models (e.g., Kappel et al.). Please see our response above.

B) Connection to biology 

The plasticity rules in the study are introduced with a vague reference to previous theoretical studies of others. Please discuss in more detail if these rules (especially the error-based learning rule) could be implemented biologically and how this could be achieved. Are there connections to biologically observed plasticity? E.g. for error-based plasticity has been discussed in the original publication by Urbanzcik and Senn, or more recently by Mikulasch et al (TINS 2023). The biological plausibility of inhibitory balance has been discussed many times before, e.g. by Vogels and others, and a citation would acknowledge that earlier work. This also leaves the question of how neurons in the songbird experiment could adapt and if the model does capture this well (i.e., do they exhibit E-I balance? etc), which might be discussed as well. 

Last, please provide some testable experimental predictions. By proposing an interesting experimental prediction, the model could become considerably more relevant to experimentalists. Also, are there potentially alternative models of stochastic sequence learning (e.g., Kappel et al)? How could they be distinguished? (especially, again, why not Hebbian/STDP learning?) 

We have cited the Vogels paper to acknowledge the earlier work. We have also included additional paragraphs to discuss a possible biologically plausible implementation of our model and how our model differs from similar models proposed previously (e.g., Kappel et al.). Please see our response above.

Other comments 

As mentioned, a derivation of recurrent plasticity rules is missing, and parameters are chosen ad-hoc. This leaves the question of how much the results rely on the specific choice of parameters, and how robust they are to perturbations. As a robustness check, please clarify how the duration of the Markov states influences performance. It can be expected that this interacts with the timescale of recurrent connections, so having longer or shorter Markov states, as it would be in reality, should make a difference in learning that should be tested and discussed.

We thank the reviewer for pointing this out. To address this point, we performed new simulations and asked to what extent the duration of Markov states affect performance. Interestingly, even when the network was trained with input states of half the duration, the distributions of the durations of assembly reactivations remain almost identical to those in the original case (Supplementary Figure 3a). Furthermore, the transition probabilities in the replay were still consistent with the true transition probabilities (Supplementary Figure 3b). We have also included the derivation of our plasticity rule in ll. 630-670 in the revised manuscript. 

Similarly, inhibitory plasticity operates with the same plasticity timescale parameter as excitatory plasticity, but, as the authors discuss, lags behind excitatory plasticity in simulation as in experiment. Is this required or was the parameter chosen such that this behaviour emerges? Please clarify this in the methods section; moreover, it would be good to test if the same results appear with fast inhibitory plasticity. 

We have performed a new simulation and showed that even when the learning rate of inhibitory plasticity was larger than that of excitatory plasticity, inhibitory plasticity still occurred on a slower timescale than excitatory plasticity. We have included this result in a new Supplementary Figure 2 in the revised manuscript.

What is the justification (biologically and theoretically) for the memory trace h and its impact on neural spiking? Is it required for the results or can it be left away? Since this seems to be an important and unconventional component of the model, please discuss it in more detail. 

In the model, it is assumed that each stimulus presentation drives a specific subset of network neurons with a fixed input strength, which avoids convergence to trivial solutions. Nevertheless, we choose to add this dynamic sigmoid function to facilitate stable replay by regulating neuron activity to prevent saturation. We have explained this point in ll.605-611 in the revised manuscript.

Reviewer #2 (Recommendations For The Authors): 

I noticed a couple of minor typos: 

Page 3 "underly"->"underlie" 

Page 7 "assemblies decreased settled"->"assemblies decreased and settled"

We have modified the text. We thank the reviewer for their careful review.

I think Figure 1C is rather confusing and not intuitive. 

We apologize that the Figure 1C was confusing. In the revised figure, we have emphasized the flow of excitatory and inhibitory error for updating synapses.

Reviewer #3 (Recommendations For The Authors): 

One possible path to improve the paper would be to establish a relationship between the proposed learning rules and e.g. the ones derived by Barber. 

When reading the paper, I was left with a number of more detailed questions I omitted from the public review: 

(1) The authors introduce a dynamic sigmoidal function for excitatory neurons, Eq. 3. This point requires more discussion and analysis. How does this impact the results? 

In the model, it is assumed that each stimulus presentation drives a specific subset of network neurons with a fixed input strength, which avoids convergence to trivial solutions. Nevertheless, we choose to add this dynamic sigmoid function to facilitate stable replay by regulating neuron activity to prevent saturation. We have explained this point in ll.605-611 in the revised manuscript.

(2) For Poisson spiking neurons, it would be great to understand what cell assemblies bring (apart from biological realism, i.e., reproducing data where assemblies can be found), compared to self-connected single neurons. For example, how do the results shown in Figure 2 depend on assembly size? 

We have changed the cell assembly size ratio and how it affects learning performance in a new Supplementary Figure 4. Please see our reply above.

(3) The authors focus on modeling spontaneous transitions, corresponding to a highly stochastic generative model (with most transition probabilities far from 1). A complementary question is that of learning to produce a set of stereotypical sequences, with probabilities close to 1. I wondered whether the learning rules and architecture of the model (in particular under the I-to-E rule) would also work in such a scenario. 

We thank the reviewer for pointing this out. In fact, we had the same question, so we considered a situation in which the setting in Figure 2 includes both cases where the transition matrix is very stochastic (prob=0.5) and near deterministic (prob=0.9).

(4) An analysis of what controls the time so that the network stays in a certain state would be welcome. 

We trained the network model in two cases, one with a fast speed of plasticity and one with a slow speed of plasticity. As a result, we found that the duration of assembly becomes longer in the slow learning case than in the fast case. We have included these results as Supplementary Figure 5 in the revised manuscript.

Regarding the presentation, given that this is a computational modeling paper, I wonder whether *all* the formulas belong in the Methods section. I found myself skipping back and forth to understand what the main text meant, mainly because I missed a few key equations. I understand that this is a style issue that is very much community-dependent, but I think readability would improve drastically if the main model and learning rule equations could be introduced in the main text, as they start being discussed. 

We thank the reviewer for the suggestion. To cater to a wider audience, we try to explain the principle of the paper without using mathematical formulas as much as possible in the main text.

eLife Assessment

This is an important study that investigates how neural networks can learn to stochastically replay presented sequences of activity according to learned transition probabilities. The authors use error-based excitatory plasticity to minimize the difference between internally predicted activity and stimulus-driven activity, and inhibitory plasticity to maintain E-I balance. The approach is solid but the choice of learning rules and parameters is not always always justified, with some unclear aspects to the formal derivation.

Reviewer #2 (Public review):

Summary:

This work proposes a synaptic plasticity rule which explains the generation of learned stochastic dynamics during spontaneous activity. The proposed plasticity rule assumes that excitatory synapses seek to minimize the difference between the internal predicted activity and stimulus-evoked activity, and inhibitory synapses tries to maintain the E-I balance by matching the excitatory activity. By implementing this plasticity rule in a spiking recurrent neural network, the authors show that the state-transition statistics of spontaneous excitatory activity agrees with that of the learned stimulus patterns, which is reflected in the learned excitatory synaptic weights. The authors further demonstrate that inhibitory connections contribute to well-defined state-transitions matching the transition patterns evoked by the stimulus. Finally, they show that this mechanism can be expanded to more complex state-transition structures including songbird neural data.

Strengths:

This study makes an important contribution to computational neuroscience, by proposing a possible synaptic plasticity mechanism underlying spontaneous generations of learned stochastic state-switching dynamics that are experimentally observed in the visual cortex and hippocampus. This work is also very clearly presented and well-written, and the authors conducted comprehensive simulations testing multiple hypotheses. Overall, I believe this is a well-conducted study providing interesting and novel aspects on the capacity of recurrent spiking neural networks with local synaptic plasticity.

Weaknesses:

This study is very well-thought out and theoretically valuable to the neuroscience community, and I think the main weaknesses are in regard to how much biological realism is taken into account. For example, the proposed model assumes that only synapses targeting excitatory neurons are plastic, and uses an equal number of excitatory and inhibitory neurons.<br /> The model also assumes Markovian state dynamics while biological systems can depend more on history. This limitation, however, is acknowledged in the Discussion.<br /> Finally, to simulate spontaneous activity, the authors use a constant input of 0.3 throughout the study. Different amplitudes of constant input may correspond to different internal states, so it will be more convincing if the authors test the model with varying amplitudes of constant inputs.

Comments on revisions:

The authors have addressed all of the previously raised concerns satisfactorily, by running extra simulations with a biologically plausible composition of excitatory and inhibitory neurons, plasticity assumed for all synapses, and varied amounts of constant inputs representing internal states or background activities. While in some of these cases the stochastic dynamics during spontaneous activity change or do not replicate those of the learned stimulus patterns as well as before, these extended studies provide thorough evaluations of the strengths and limitations of the proposed plasticity rule as the underlying mechanism of stochastic dynamics during spontaneous activity. Overall, the revision has strengthened the paper significantly.

Reviewer #3 (Public review):

Summary:

Asabuki and Clopath study stochastic sequence learning in recurrent networks of Poisson spiking neurons that obey Dale's law. Inspired by previous modeling studies, they introduce two distinct learning rules, to adapt excitatory-to-excitatory and inhibitory-to-excitatory synaptic connections. Through a series of computer experiments, the authors demonstrate that their networks can learn to generate stochastic sequential patterns, where states correspond to non-overlapping sets of neurons (cell assemblies) and the state-transition conditional probabilities are first-order Markov, i.e., the transition to a given next state only depends on the current state. Finally, the authors use their model to reproduce certain experimental songbird data involving highly-predictable and highly-uncertain transitions between song syllables. While the findings are only moderately surprising, this is a well-written and welcome detailed study that may be of interest to experts of plasticity and learning in recurrent neural networks that respect Dale's law.

Strengths:

This is an easy-to-follow, well-written paper, whose results are likely easy to reproduce. The experiments are clear and well-explained. In particular, the study of the interplay between excitation and inhibition (and their different plasticity rules) is a highlight of the study. The study of songbird experimental data is another good feature of this paper; finches are classical model animals for understanding sequence learning in the brain. I also liked the study of rapid task-switching, it's a good-to-know type of result that is not very common in sequence learning papers.

Weaknesses:

One weakness I see in this paper is the derivation of the learning rules, which is semi-heuristic. The paper studies Poisson spiking neurons, for which learning rules can be derived from a statistical objective, typically maximum likelihood, as previously done in the cited literature. The authors provide a brief section connecting the learning rules to gradient descent on objective functions, but the link is only heuristic or at least not entirely presented. The reason is that the neural network state is not fully determined by (or "clamped to") the target during learning (for instance, inhibitory neurons do not even have a target assigned). So, the (total) gradient should take into account the recurrent contributions from other neurons, and equation 13 does not appear to be complete/correct to me. Moreover, the target firing rate is a mixture of external currents with currents arising from other neurons in the recurrent network. The authors ideally should start from an actual distribution matching objective (e.g., KL divergence, and not such a squared error), so that their main claims immediately follow from the mathematical derivations. Along the same line, it would be excellent to get some additional insights on the interaction of the two distinct plasticity rules, one of the highlights of the study. This could be naturally achieved by relating their distinct rules to a common principled objective.

The other major weakness (albeit one that is clearly discussed by the authors) is that the study assumes that every excitatory neuron is directly given its target state when learning. In machine learning language, there are no 'hidden' excitatory neurons. While this assumption greatly simplifies the derivation of efficient and biologically-plausible learning rules that can be mapped to synaptic plasticity, it also limits considerably the distributions that can be learned by the network, more precisely to those that satisfy the Markov property.

eLife Assessment

This fundamental study combines Global Positioning System tracking and the analysis of social interactions among feral pigs, to provide insights into the likelihood of disease transmission based on contact rates both within and between sounders. The method used for data collection is compelling, but the varying sample sizes across populations could be a potential source of bias. With the potential biases from varying sample sizes strengthened this paper would be of interest to the fields of Veterinary Medicine, Public Health, and Epidemiology.

Reviewer #1 (Public review):

Summary:

The authors aimed to quantify feral pig interactions in eastern Australia to inform disease transmission networks. They used GPS tracking data from 146 feral pigs across multiple locations to construct proximity-based social networks and analyze contact rates within and between pig social units.

Strengths:

(1) Addresses a critical knowledge gap in feral pig social dynamics in Australia.

(2) Uses robust methodology combining GPS tracking and network analysis.

(3) Provides valuable insights into sex-based and seasonal variations in contact rates.

(4) Effectively contextualizes findings for disease transmission modeling and management.

(5) Includes comprehensive ethical approval for animal research.

(6) Utilizes data from multiple locations across eastern Australia, enhancing generalizability.

Weaknesses:

(1) Limited discussion of potential biases from varying sample sizes across populations

(2) Some key figures are in supplementary materials rather than the main text.

(3) Economic impact figures are from the US rather than Australia-specific data.

(4) Rationale for spatial and temporal thresholds for defining contacts could be clearer.

(5) Limited discussion of ethical considerations beyond basic animal ethics approval.

The authors largely achieved their aims, with the results supporting their conclusions about the importance of sex and seasonality in feral pig contact networks. This work is likely to have a significant impact on feral pig management and disease control strategies in Australia, providing crucial data for refining disease transmission models.

Reviewer #2 (Public review):

Summary:

The paper attempts to elucidate how feral (wild) pigs cause distortion of the environment in over 54 countries of the world, particularly Australia.

The paper displays proof that over $120 billion worth of facilities were destroyed annually in the United States of America.

The authors have tried to infer that the findings of their work were important and possess a convincing strength of evidence.

Strengths:

(1) Clearly stating feral (wild) pigs as a problem in the environment.

(2) Stating how 54 countries were affected by the feral pigs.

(3) Mentioning how $120 billion was lost in the US, annually, as a result of the activities of the feral pigs.

(4) Amplifying the fact that 14 species of animals were being driven into extinction by the feral pigs.

(5) Feral pigs possessing zoonotic abilities.

(6) Feral pigs acting as reservoirs for endemic diseases like brucellosis and leptospirosis.

(7) Understanding disease patterns by the social dynamics of feral pig interactions.

(8) The use of 146 GPS-monitored feral pigs to establish their social interaction among themselves.

Weaknesses:

(1) Unclear explanation of the association of either the female or male feral pigs with each other, seasonally.

(2) The "abstract paragraph" was not justified.

(3) Typographical errors in the abstract.

Reviewer #3 (Public review):

Summary:

The authors sought to understand social interactions both within and between groups of feral pigs, with the intent of applying their findings to models of disease transmission. The authors analyzed GPS tracking data from across various populations to determine patterns of contact that could support the transmission of a range of zoonotic and livestock diseases. The analysis then focused on the effects of sex, group dynamics, and seasonal changes on contact rates that could be used to base targeted disease control strategies that would prioritize the removal of adult males for reducing intergroup disease transmission.

Strengths:

It utilized GPS tracking data from 146 feral pigs over several years, effectively capturing seasonal and spatial variation in the social behaviors of interest. Using proximity-based social network analysis, this work provides a highly resolved snapshot of contact rates and interactions both within and between groups, substantially improving research in wildlife disease transmission. Results were highly useful and provided practical guidance for disease management, showing that control targeted at adult males could reduce intergroup disease transmission, hence providing an approach for the control of zoonotic and livestock diseases.

Weaknesses:

Despite their reliability, populations can be skewed by small sample sizes and limited generalizability due to specific environmental and demographic characteristics. Further validation is needed to account for additional environmental factors influencing social dynamics and contact rates

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors aimed to quantify feral pig interactions in eastern Australia to inform disease transmission networks. They used GPS tracking data from 146 feral pigs across multiple locations to construct proximity-based social networks and analyze contact rates within and between pig social units.

Strengths:

(1) Addresses a critical knowledge gap in feral pig social dynamics in Australia.

(2) Uses robust methodology combining GPS tracking and network analysis.

(3) Provides valuable insights into sex-based and seasonal variations in contact rates.

(4) Effectively contextualizes findings for disease transmission modeling and management.

(5) Includes comprehensive ethical approval for animal research.

(6) Utilizes data from multiple locations across eastern Australia, enhancing generalizability.

Weaknesses:

(1) Limited discussion of potential biases from varying sample sizes across populations

This is a really good comment, and we will address this in the discussion as one of the limitations of the study.

(2) Some key figures are in supplementary materials rather than the main text.

We will move some of our supplementary material to the main text as suggested.

(3) Economic impact figures are from the US rather than Australia-specific data.

We included the impact figures that are available for Australia (for FDM), and we will include the estimated impact of ASF in Australia in the introduction.

(4) Rationale for spatial and temporal thresholds for defining contacts could be clearer.

We will improve the explanation of why we chose the spatial and temporal thresholds based on literature, the size of animals and GPS errors.

(5) Limited discussion of ethical considerations beyond basic animal ethics approval.

This research was conducted under an ethics committee's approval for collaring the feral pigs. This research is part of an ongoing pest management activity, and all the ethics approvals have been highlighted in the main manuscript.

The authors largely achieved their aims, with the results supporting their conclusions about the importance of sex and seasonality in feral pig contact networks. This work is likely to have a significant impact on feral pig management and disease control strategies in Australia, providing crucial data for refining disease transmission models.

Reviewer #2 (Public review):

Summary:

The paper attempts to elucidate how feral (wild) pigs cause distortion of the environment in over 54 countries of the world, particularly Australia.

The paper displays proof that over $120 billion worth of facilities were destroyed annually in the United States of America.

The authors have tried to infer that the findings of their work were important and possess a convincing strength of evidence.

Strengths:

(1) Clearly stating feral (wild) pigs as a problem in the environment.

(2) Stating how 54 countries were affected by the feral pigs.

(3) Mentioning how $120 billion was lost in the US, annually, as a result of the activities of the feral pigs.

(4) Amplifying the fact that 14 species of animals were being driven into extinction by the feral pigs.

(5) Feral pigs possessing zoonotic abilities.

(6) Feral pigs acting as reservoirs for endemic diseases like brucellosis and leptospirosis.

(7) Understanding disease patterns by the social dynamics of feral pig interactions.

(8) The use of 146 GPS-monitored feral pigs to establish their social interaction among themselves.

Weaknesses:

(1) Unclear explanation of the association of either the female or male feral pigs with each other, seasonally.

This will be better explain in the methods.

(2) The "abstract paragraph" was not justified.

We have justified the abstract paragraph as requested by the reviewer.

(3) Typographical errors in the abstract.

Typographical errors have been corrected in the Abstract.

Reviewer #3 (Public review):

Summary:

The authors sought to understand social interactions both within and between groups of feral pigs, with the intent of applying their findings to models of disease transmission. The authors analyzed GPS tracking data from across various populations to determine patterns of contact that could support the transmission of a range of zoonotic and livestock diseases. The analysis then focused on the effects of sex, group dynamics, and seasonal changes on contact rates that could be used to base targeted disease control strategies that would prioritize the removal of adult males for reducing intergroup disease transmission.

Strengths:

It utilized GPS tracking data from 146 feral pigs over several years, effectively capturing seasonal and spatial variation in the social behaviors of interest. Using proximity-based social network analysis, this work provides a highly resolved snapshot of contact rates and interactions both within and between groups, substantially improving research in wildlife disease transmission. Results were highly useful and provided practical guidance for disease management, showing that control targeted at adult males could reduce intergroup disease transmission, hence providing an approach for the control of zoonotic and livestock diseases.

Weaknesses:

Despite their reliability, populations can be skewed by small sample sizes and limited generalizability due to specific environmental and demographic characteristics. Further validation is needed to account for additional environmental factors influencing social dynamics and contact rates

This is a good point, and we thank the reviewer for pointing out this issue. We will discuss the potential biases due to sample size in our discussion. We agree that environmental factors need to be incorporated and tested for their influence on social dynamics, and this will be added to the discussion as we have plans to expand this research and conduct, the analysis to determine if environmental factors are influencing social dynamics.

eLife Assessment

This valuable study uses extensive comparative analysis to examine the relationship between plasma glucose levels, albumin glycation levels, and diet and life history, within the framework of the "pace of life syndrome" hypothesis. The evidence that glucose and glycation levels are broadly correlated is convincing. However, concerns about the consistency of the data quality across species and some aspects of data analysis make the key conclusion about higher glycation resistance in species with higher glucose levels currently incomplete. Still, as the first extensive comparative analysis of glycation rates, life history, and glucose levels in birds, the study has potential to be of interest to evolutionary ecologists and the aging research community more broadly.

Reviewer #1 (Public review):

The paper explored cross-species variance in albumin glycation and blood glucose levels in the function of various life-history traits. Their results show that<br /> (1) blood glucose levels predict albumin gylcation rates<br /> (2) larger species have lower blood glucose levels<br /> (3) lifespan positively correlates with blood glucose levels and<br /> (4) diet predicts albumin glycation rates.

The data presented is interesting, especially due to the relevance of glycation to the ageing process and the interesting life-history and physiological traits of birds. Most importantly, the results suggest that some mechanisms might exist that limit the level of glycation in species with the highest blood glucose levels.

While the questions raised are interesting and the amount of data the authors collected is impressive, I have some major concerns about this study:

(1) The authors combine many databases and samples of various sources. This is understandable when access to data is limited, but I expected more caution when combining these. E.g. glucose is measured in all samples without any description of how handling stress was controlled for. E.g glucose levels can easily double in a few minutes in birds, potentially introducing variation in the data generated. The authors report no caution of this effect, or any statistical approaches aiming to check whether handling stress had an effect here, either on glucose or on glycation levels.

(2) The database with the predictors is similarly problematic. There is information pulled from captivity and wild (e.g. on lifespan) without any confirmation that the different databases are comparable or not (and here I'm not just referring to the correlation between the databases, but also to a potential systematic bias (e.g. captivate-based sources likely consistently report longer lifespans). This is even more surprising, given that the authors raise the possibility of captivity effects in the discussion, and exploring this question would be extremely easy in their statistical models (a simple covariate in the MCMCglmms).

(3) The authors state that the measurement of one of the primary response variables (glycation) was measured without any replicability test or reference to the replicability of the measurement technique.

(4) The methods and results are very poorly presented. For instance, new model types and variables are popping up throughout the manuscript, already reporting results, before explaining what these are e.g. results are presented on "species average models" and "model with individuals", but it's not described what these are and why we need to see both. Variables, like "centered log body mass", or "mass-adjusted lifespan" are not explained. The results section is extremely long, describing general patterns that have little relevance to the questions raised in the introduction and would be much more efficiently communicated visually or in a table.

Reviewer #2 (Public review):

Summary

In this extensive comparative study, Moreno-Borrallo and colleagues examine the relationships between plasma glucose levels, albumin glycation levels, diet, and life-history traits across birds. Their results confirmed the expected positive relationship between plasma blood glucose level and albumin glycation rate but also provided findings that are somewhat surprising or contradicting findings of some previous studies (relationships with lifespan, clutch mass, or diet). This is the first extensive comparative analysis of glycation rates and their relationships to plasma glucose levels and life history traits in birds that are based on data collected in a single study and measured using unified analytical methods.

Strengths

This is an emerging topic gaining momentum in evolutionary physiology, which makes this study a timely, novel, and very important contribution. The study is based on a novel data set collected by the authors from 88 bird species (67 in captivity, 21 in the wild) of 22 orders, which itself greatly contributes to the pool of available data on avian glycemia, as previous comparative studies either extracted data from various studies or a database of veterinary records of zoo animals (therefore potentially containing much more noise due to different methodologies or other unstandardised factors), or only collected data from a single order, namely Passeriformes. The data further represents the first comparative avian data set on albumin glycation obtained using a unified methodology. The authors used LC-MS to determine glycation levels, which does not have problems with specificity and sensitivity that may occur with assays used in previous studies. The data analysis is thorough, and the conclusions are mostly well-supported (but see my comments below). Overall, this is a very important study representing a substantial contribution to the emerging field of evolutionary physiology focused on the ecology and evolution of blood/plasma glucose levels and resistance to glycation.

Weaknesses

My main concern is about the interpretation of the coefficient of the relationship between glycation rate and plasma glucose, which reads as follows: "Given that plasma glucose is logarithm transformed and the estimated slope of their relationship is lower than one, this implies that birds with higher glucose levels have relatively lower albumin glycation rates for their glucose, fact that we would be referring as higher glycation resistance" (lines 318-321) and "the logarithmic nature of the relationship, suggests that species with higher plasma glucose levels exhibit relatively greater resistance to glycation" (lines 386-388). First, only plasma glucose (predictor) but not glycation level (response) is logarithm transformed, and this semi-logarithmic relationship assumed by the model means that an increase in glycation always slows down when blood glucose goes up, irrespective of the coefficient. The coefficient thus does not carry information that could be interpreted as higher (when <1) or lower (when >1) resistance to glycation (this only can be done in a log-log model, see below) because the semi-log relationship means that glycation increases by a constant amount (expressed by the coefficient of plasma glucose) for every tenfold increase in plasma glucose (for example, with glucose values 10 and 100, the model would predict glycation values 2 and 4 if the coefficient is 2, or 0.5 and 1 if the coefficient is 0.5). Second, the semi-logarithmic relationship could indeed be interpreted such that glycation rates are relatively lower in species with high plasma glucose levels. However, the semi-log relationship is assumed here a priori and forced to the model by log-transforming only glucose level, while not being tested against alternative models, such as: (i) a model with a simple linear relationship (glycation ~ glucose); or (ii) a log-log model (log(glycation) ~ log(glucose)) assuming power function relationship (glycation = a * glucose^b). The latter model would allow for the interpretation of the coefficient (b) as higher (when <1) or lower (when >1) resistance in glycation in species with high glucose levels as suggested by the authors.

Besides, a clear explanation of why glucose is log-transformed when included as a predictor, but not when included as a response variable, is missing.

The models in the study do not control for the sampling time (i.e., time latency between capture and blood sampling), which may be an important source of noise because blood glucose increases because of stress following the capture. Although the authors claim that "this change in glucose levels with stress is mostly driven by an increase in variation instead of an increase in average values" (ESM6, line 46), their analysis of Tomasek et al.'s (2022) data set in ESM1 using Kruskal-Wallis rank sum test shows that, compared to baseline glucose levels, stress-induced glucose levels have higher median values, not only higher variation.

Although the authors calculated the variance inflation factor (VIF) for each model, it is not clear how these were interpreted and considered. In some models, GVIF^(1/(2*Df)) is higher than 1.6, which indicates potentially important collinearity; see for example https://www.bookdown.org/rwnahhas/RMPH/mlr-collinearity.html). This is often the case for body mass or clutch mass (e.g. models of glucose or glycation based on individual measurements).

It seems that the differences between diet groups other than omnivores (the reference category in the models) were not tested and only inferred using the credible intervals from the models. However, these credible intervals relate to the comparison of each group with the reference group (Omnivore) and cannot be used for pairwise comparisons between other groups. Statistics for these contrasts should be provided instead. Based on the plot in Figure 4B, it seems possible that terrestrial carnivores differed in glycation level not only from omnivores but also from herbivores and frugivores/nectarivores.

Given that blood glucose is related to maximum lifespan, it would be interesting to also see the results of the model from Table 2 while excluding blood glucose from the predictors. This would allow for assessing if the maximum lifespan is completely independent of glycation levels. Alternatively, there might be a positive correlation mediated by blood glucose levels (based on its positive correlations with both lifespan and glycation), which would be a very interesting finding suggesting that high glycation levels do not preclude the evolution of long lifespans.

Author response:

Reviewer #1:

(1) This concern is addressed in the ESM6, and partly in the ESM1. Indeed, many of the concerns raised by the reviewer later are already addressed on the multiple supplementary materials provided, so we kindly ask the reviewer to read them before moving forward into the discussion.

(2) This concern is reasonable, but its solution is not "extremely easy", as the reviewer states. The reviewer indicates the use of captive-based versus non-captive-based sources, remarking maximum lifespan, the main variable that is clearly expected to be systematically biased by the source of the data. Nevertheless, except for the ZIMS database, which includes only captive individuals, and some sources, as CNRS databases and EURING, which exclusively includes wild populations, the remaining databases, which are indeed where the vast majority of the data was collected from (i.e. Amniotes database, Birds of the World and AnAge) do not make any distinction. This means that they include just the maximum lifespan from the species as known by the authors of such databases' entries, regardless of provenance, which is also not usually made explicit by the database. Therefore, correcting for this would imply checking all the primary sources. Considering that these databases sometimes do not cite the primary source, but a secondary one, and that on several occasions such source is a specialized book that is not easily accessible, and still these referenced datasets may not indicate the source of the data, tracing all of this information becomes an arduous task, that would even render the usage of databases themselves useless. We will include some details about the concerns of database usage in the discussion to address this.

Furthermore, it remains relevant to indicate that what we discuss later about the possible effects of captivity is about our usage of animals that come from both sources, not about the provenance of the literature-extracted data used (i.e. captive or wild maximum lifespan, for example), which is an independent matter. We can test for the first for next submission, but very difficultly could we test for the second (as the reviewer seems to be pointing to). In any case, as we do not have in any case the same species from both a captive and a wild source, it would be difficult to determine if the effect tested comes from captivity or from species-specific differences.

(3) We will add data on the replicability of the glycation measurement in the next manuscript version. The CV for several individuals of different species measured repeated times is quite low (always below 2%).

(4) The reviewer remarks reported here are already addressed on the supplementary material (ESM6), given the lack of space in the main manuscript. We therefore kindly ask the reviewer to read the supplementary material added to the submission. If the editors agree, all or a considerable part of this could be transferred to the main text for clarity, but this would severely extend the length of a text that the reviewer already considered very long.

Reviewer #2:

Thanks for spotting this issue with the coefficient, as it is actually a redaction mistake. It is a remnant of a previous version of the manuscript in which a log-log relation was performed instead. Previous reviewers raised concerns about the usage of log transformation for glycation, this variable being (theoretically) a proportion variable (to which we argue that it does not behave as such), which they considered not to be transformed with a logarithm. After this, we still finally took the decision of not to transform this variable. In this line, the transformations of variables were decided generally by preliminary data exploration. In this particular case, both approaches lead to the same conclusion of higher glycation resistance in the species with higher glucose. Nevertheless, we will consider exploring the comparison of different versions for the resubmission.

About the issue related to handling time, this variable is not available, for the reasons already exposed in the answer to the other reviewer. Moreover, Kruskal-Wallis test, by its nature, does not determine differences in medians between groups per se, as the reviewer claims, but just differences in ranks-sums. It can be equivalently used for that purpose when the groups' distributions are similar, but not when they differ, as we see here with a difference in variance. What a significant outcome in a Kruskal-Wallis test tells us, thus, is just that the groups differ (in their ranks-sums), which here is plausibly caused by the higher variance in the stressed individuals. Even if we conclude that the average is higher in those groups, mere comparisons of averages for groups with very different variances render different interpretations than when homoscedasticity is met, particularly more so when the distribution of groups overlaps. For example, in a case like this, where the data is left censored (glucose levels cannot be lower than 0), most of this higher variance is related to many values in the stressed groups lying above all the baseline values. This, of course, would increase the average, but such a parameter would not mean the same as if the distributions did not overlap.

Regarding the GVIFs, why the values are above 1.6 is not well known, but we do not consider this a major concern, as the values are never above 2.2, level usually considered more worrying. We will include a brief explanation of this in the results section. Also, we explicitly calculated life history variables adjusted for body mass, which should eliminate their otherwise strong correlation. There exist other biological and interpretational reasons justified in the ESM6 for using the residuals on the models, instead of the raw values, despite previously raised concerns.

Given the asseveration by the reviewer that credible intervals are not to be used for the post hoc comparisons, as this is what the whiskers shown in Figure 4B represent, the affirmation of this graph suggesting any difference between groups remains doubtful. New comparisons have now been made with the function HPDinterval() applied to the differences between each diet category calculated from the posterior values of each group, confirming no significant differences exist.

We do not understand the suggestion made in relation to the model shown in Table 2. Removing glucose from the model could have two results, as the reviewer indicates: 1. Maximum lifespan (ML) relates with glycation, potentially spuriously through the effect of glucose (in this case not included) on both; 2. ML does not relate to glycation, and therefore "high glycation levels do not preclude the evolution of long lifespans", which is what we are already showing with the current model, which also controls for glucose, in an attempt to determine if not just raw glycation values, but glycation resistance, relates to longevity. This is intended to asses if long-lived species may show mechanisms that avoid glycation, by showing levels lower than expected for a non-enzymatic reaction.

eLife Assessment

This study aims to investigate the RNA binding activities of a conserved heterochromatin protein (Swi6) and proposes an entirely new model for how heterochromatin formation is initiated in fission yeast. While the concept is interesting, the data provided are inadequate, both for support of the claims regarding the new RNA binding activities and for support of the new model. The paper requires extensive editing as well as the inclusion of numerous experiments with appropriately controlled conditions.

Reviewer #1 (Public review):

Summary:

This manuscript explores the RNA binding activities of the fission yeast Swi6 (HP1) protein and proposes a new role for Swi6 in RNAi-mediated heterochromatin establishment. The authors claim that Swi6 has a specific and high affinity for short interfering RNAs (siRNAs) and recruits the Clr4 (Suv39h) H3K9 methyltransferases to siRNA-DNA hybrids to initiate heterochromatin formation. These claims are not in any way supported by the incomplete and preliminary RNA binding or the in vivo experiments that the authors present. The proposed model also lacks any mechanistic basis as it remains unclear (and unexplored) how Swi6 might bind to specific small RNA sequences or RNA-DNA hybrids. Work by several other groups in the field has led to a model in which siRNAs produced by the RNAi pathway load onto the Ago1-containing RITS complex, which then binds to nascent transcripts at pericentromeric DNA repeats and recruits Clr4 to initiate heterochromatin formation. Swi6 facilitates this process by promoting the recruitment of the RNA-dependent RNA polymerase leading to siRNA amplification.

Weaknesses:

(1) The claims that Swi6 binds to specific small RNAs or to RNA-DNA hybrids are not supported by the evidence that the authors present. Their experiments do not rule out non-specific charged-based interactions. Claims about different affinities of Swi6 for RNAs of different sizes are based on a comparison of KD values derived by the authors for a handful of S. pombe siRNAs with previous studies from the Buhler lab on Swi6 RNA binding. The authors need to compare binding affinities under identical conditions in their assays. The regions of Swi6 that bind to siRNAs need to be identified and evidence must be provided that Swi6 binds to RNAs of a specific length, 20-22 mers, to support the claim that Swi6 binds to siRNAs. This is critical for all the subsequent experiments and claims in the study.

(2) The in vivo results do not validate Swi6 binding to specific RNAs, as stated by the authors. Swi6 pulldowns have been shown to be enriched for all heterochromatic proteins including the RITS complex. The sRNA binding observed by the authors is therefore likely to be mediated by Ago1/RITS.

Most of the binding in Figure S8C seems to be non-specific.

In Figure S8D, the authors' data shows that Swi6 deletion does not derepress the rev dh transcript while dcr1 delete cells do, which is consistent with previous reports but does not relate to the authors' conclusions.

Previous results have shown that swi6 delete cells have 20-fold fewer dg and dh siRNAs than swi6+ cells due to decreased RNA-dependent RNA polymerase complex recruitment and reduced siRNA amplification.

(3) The RIP-seq data are difficult to interpret as presented. The size distribution of bound small RNAs, and where they map along the genome should be shown as for example presented in previous Ago1 sRNA-seq experiments.

It is also unclear whether the defects in sRNA binding observed by the authors represent direct sRNA binding to Swi6 or co-precipitation of Ago1-bound sRNAs.

The authors should also sequence total sRNAs to test whether Swi6-3A affects sRNA synthesis, as is the case in swi6 delete cells.

(4) The authors examine the effects of Swi6-3A mutant by overexpression from the strong nmt1 promoter. Heterochromatin formation is sensitive to the dosage of Swi6. These experiments should be performed by introducing the 3A mutations at the endogenous Swi6 locus and effects on Swi6 protein levels should be tested.

(5) The authors' data indicate an impairment of silencing in Swi6-3A mutant cells but whether this is due to a general lower affinity for nucleosomes, DNA, RNA, or as claimed by the authors, siRNAs is unclear. These experiments are consistent with previous findings suggesting an important role for basic residues in the HP1 hinge region in gene silencing but do not reveal how the hinge region enhances silencing.

(6) RNase H1 overexpression may affect Swi6 localization and silencing indirectly as it would lead to a general reduction in R loops and RNA-DNA hybrids across the genome. RNaseH1 OE may also release chromatin-bound RNAs that act as scaffolds for siRNA-Ag1/RITS complexes that recruit Clr4 and ultimately Swi6.

(7) Examples of inaccurate presentation of the literature.<br /> a. The authors state that "RNA binding by the murine HP1 through its hinge domains is required for heterochromatin assembly (Muchardt et al, 2002). The cited reference provides no evidence that HP1 RNA binding is required for heterochromatin assembly. Only the hinge region of bacterially produced HP1 contributes to its localization to DAPI-stained heterochromatic regions in fixed NIH 3T3 cells.<br /> b. "... This scenario is consistent with the loss of heterochromatin recruitment of Swi6 as well as siRNA generation in rnai mutants (Volpe et al, 2002)." Volpe et al. did not examine changes in siRNA levels in swi6 mutant cells. In fact, no siRNA analysis of any kind was reported in Volpe et al., 2002.

Reviewer #2 (Public review):

The aim of this study is to investigate the role of Swi6 binding to RNA in heterochromatin assembly in fission yeast. Using in vitro protein-RNA binding assays (EMSA) they showed that Swi6/HP1 binds centromere-derived siRNA (identified by Reinhardt and Bartel in 2002) via the chromodomain and hinge domains. They demonstrate that this binding is regulated by a lysine triplet in the conserved region of the Swi6 hinge domain and that wild-type Swi6 favours binding to DNA-RNA hybrids and siRNA, which then facilitates, rather than competes with, binding to H3K9me2 and to a lesser extent H3K9me3.

However, the majority of the experiments are carried out in swi6 null cells overexpressing wild-type Swi6 or Swi63K-3A mutant from a very strong promoter (nmt1). Both swi6 null cells and overexpression of Swi6 are well known to exhibit phenotypes, some of which interfere with heterochromatin assembly. This is not made clear in the text. Whilst the RNA binding experiments show that Swi6 can indeed bind RNA and that binding is decreased by Swi63K-3A mutation in vitro (confusingly, they only much later in the text explained that these 3 bands represent differential binding and that II is likely an isotherm). The gels showing these data are of poor quality and it is unclear which bands are used to calculate the Kd. RNA-seq data shows that overall fewer siRNAs are produced from regions of heterochromatin in the Swi63K-3A mutant so it is unsurprising that analysis of siRNA-associated motifs also shows lower enrichment (or indeed that they share some similarities, given that they originate from repeat regions).

The experiments are seemingly linked yet fail to substantiate their overall conclusions. For instance, the authors show that the Swi63K-3A mutant displays reduced siRNA binding in vitro (Figure 1D) and that H3K9me2 levels at heterochromatin loci are reduced in vivo (Figure 3C-D). They conclude that Swi6 siRNA binding is important for Swi6 heterochromatin localization, whilst it remains entirely possible that heterochromatin integrity is impaired by the Swi63K-3A mutation and hence fewer siRNAs are produced and available to bind. Their interpretation of the data is really confusing.

The authors go on to show that Swi63K-3A cells have impaired silencing at all regions tested and the mutant protein itself has less association with regions of heterochromatin. They perform DNA-RNA hybrid IPs and show that Swi63K-3A cells which also overexpress RNAseH/rnh1 have reduced levels of dh DNA-RNA hybrids than wild-type Swi6 cells. They interpret this to mean that Swi6 binds and protects DNA-RNA hybrids, presumably to facilitate binding to H3K9me2. The final piece of data is an EMSA assay showing that "high-affinity binding of Swi6 to a dg-dh specific RNA/DNA hybrid facilitates the binding to Me2-K9-H3 rather than competing against it." This EMSA gel shown is of very poor quality, and this casts doubt on their overall conclusion.

Unfortunately, the manuscript is generally poorly written and difficult to comprehend. The experimental setups and interpretations of the data are not fully explained, or, are explained in the wrong order leading to a lack of clarity. An example of this is the reasoning behind the use of the cid14 mutant which is not explained until the discussion of Figure 5C, but it is utilised at the outset in Figure 5A.

Another example of this lack of clarity/confusion is that the abstract states "Here we provide evidence in support of RNAi-independent recruitment of Swi6". Yet it then states "We show that...Swi6/HP1 displays a hierarchy of increasing binding affinity through its chromodomain to the siRNAs corresponding to specific dg-dh repeats, and even stronger binding to the cognate siRNA-DNA hybrids than to the siRNA precursors or general RNAs." RNAi is required to produce siRNAs, so their message is very unclear. Moreover, an entire section is titled "Heterochromatin recruitment of Swi6-HP1 depends on siRNA generation" so what is the author's message?

The data presented, whilst sound in some parts is generally overinterpreted and does not fully support the author's confusing conclusions. The authors essentially characterise an overexpressed Swi6 mutant protein with a few other experiments on the side, that do not entirely support their conclusions. They make the point several times that the KD for their binding experiments is far higher than that previously reported (Keller et al Mol Cell 2012) but unfortunately the data provided here are of an inferior quality and thus their conclusions are neither fully supported nor convincing.

Author response:

In this manuscript, we have addressed one of the possible modes of recruitment of Swi6 to the putative heterochromatin loci.

Our investigation was guided by earlier work showing ability of HP1 a to bind to a class of RNAs and the role of this binding in recruitment of HP1a to heterochromatin loci in mouse cells (Muchardt et al). While there has been no clarity about the mechanism of Swi6 recruitment given the multiple pathways being involved, the issue is compounded by the overall lack of understanding as to how Swi6 recruitment occurs only at the repeat regions. At the same time, various observations suggested a causal role of RNAi in Swi6 recruitment.

Thus, guided by the work of Muchardt et al we developed a heuristic approach to explore a possibly direct link between Swi6 and heterochromatin through RNAi pathway. Interestingly, we found that the lysine triplet found in the hinge domain in HP1, which influences its recruitment to heterochromatin in mouse cells, is also present in the hinge domain of Swi6, although we were cautious, keeping in mind the findings of Keller et al showing another role of Swi6 in binding to RNAs and channeling them to the exosome pathway. 

Accordingly, we envisaged that a mode of recruitment of Swi6 through binding to siRNAs to cognate sites in the dg-dh repeats shared among mating type, centromere and telomere loci could explain specific recruitment as well as inheritance following DNA replication. In accordance we framed the main questions as follows: i) Whether Swi6 binds specifically and with high affinity to the siRNAs and the cognate siRNA-DNA hybrids and whether the Swi63K-3A mutant is defective in this binding, ii) whether this lack of binding of Swi63K-3A affects its localization to heterochromatin, iii) whether the this specificity is validated by binding of Swi6 but not Swi63K-3A  to siRNAs and siRNA-DNA hybrids in vivo and iv) whether the binding mode was qualitatively and quantitatively different from that of Cen100 RNA or random RNAs, like GFP RNA.

We think that our data provides answers to these lines of inquiry to support a model wherein the Swi6-siRNA mediated recruitment can explain a cis-controlled nucleation of heterochromatin at the cognate sites in the genome. We have also partially addressed the points raised by the study by Keller et al by invoking a dynamic balance between different modes of binding of Swi6 to different classes of RNA to exercise heterochromatin formation by Swi6 under normal conditions and RNA degradation under other conditions.

While we aver about our hypothesis, we do acknowledge the need for more detailed investigation both to buttress our hypothesis and address the dynamics of siRNA binding and recruitment of Swi6  and how Swi6 functions fit in the context of other components of heterochromatin assembly, like the HDACs and Clr4 on one hand and exosome pathway on the other. Our future studies will attempt to address these issues.

Public Reviews:

Reviewer #1 (Public review):

Summary:

This manuscript explores the RNA binding activities of the fission yeast Swi6 (HP1) protein and proposes a new role for Swi6 in RNAi-mediated heterochromatin establishment. The authors claim that Swi6 has a specific and high affinity for short interfering RNAs (siRNAs) and recruits the Clr4 (Suv39h) H3K9 methyltransferases to siRNA-DNA hybrids to initiate heterochromatin formation. These claims are not in any way supported by the incomplete and preliminary RNA binding or the in vivo experiments that the authors present. The proposed model also lacks any mechanistic basis as it remains unclear (and unexplored) how Swi6 might bind to specific small RNA sequences or RNA-DNA hybrids. Work by several other groups in the field has led to a model in which siRNAs produced by the RNAi pathway load onto the Ago1-containing RITS complex, which then binds to nascent transcripts at pericentromeric DNA repeats and recruits Clr4 to initiate heterochromatin formation. Swi6 facilitates this process by promoting the recruitment of the RNA-dependent RNA polymerase leading to siRNA amplification.

Weaknesses:

(1) a) The claims that Swi6 binds to specific small RNAs or to RNA-DNA hybrids are not supported by the evidence that the authors present. Their experiments do not rule out non-specific charged-based interactions.

We disagree. We have used synthetic siRNAs of 20-22 nt length to do EMSA assay, as mentioned in the manuscript. Further, we have sequenced the small RNAs obtained after RIP experiments to validate the enrichment of siRNA in Swi6 bound fraction as compared to the mutant Swi6-bound fraction. These results are internally consistent regardless of the mode of binding. In any case the binding occurs primarily through the chromodomain although it is influenced by the hinge domain (see below).

Furthermore, we have carried out EMSA experiments using Swi6 mutants carrying all three possible double mutations of the K residues in the KKK triplet and found that there was no difference in the binding pattern as compared to the wt Swi6: only the triple mutant “3K-3A” showed the effect. These results suggest that that the bdining is not completely dependent on the basic residues. These results will be included in the revised version.

We also have some preliminary data from SAXS study showing that the CD of wt Swi6 shows a change in its structure upon binding to the siRNA, while the “3K-3A” mutant of Swi6 has a compact, folded structure that occludes the binding site of Swi6 in the chromodomain.” We propose to mention this preliminary finding in the revised version as unpublished data.

b) Claims about different affinities of Swi6 for RNAs of different sizes are based on a comparison of KD values derived by the authors for a handful of S. pombe siRNAs with previous studies from the Buhler lab on Swi6 RNA binding. The authors need to compare binding affinities under identical conditions in their assays.

Thus, the EMSA data do suggest sequence specificity in binding of Swi6 to specific siRNA sequences (Figure S5) and implies specific residues in Swi6 being responsible for that. Thus, Identification of the residues in Swi6 involved in siRNA binding in the CD would definitely be interesting, as also the experimental confirmation of the consensus siRNA sequence. It may however be noted that as against the binding of Swi6 to siRNAs occurs through CD, that of Cen100 or GFP RNA was shown be through the hinge domain by Keller et al.

The estimation of Kd by the Buhler group was based on NMR study, which we are not in a position to perform in the near future. Nonetheless, we did carry out EMSA study using the ‘Cen100’ RNA, same as the one used by the Keller et al study. Surprisingly, in contrast with the result of EMSA in agarose gel showing binding of Swi6 to “Cen100” RNA as reported by Keller et al, we fail to observe any binding in EMSA done in acrylamide gel. (The same is true of the RevCen 100). While this raises issues of why the Keller et al chose to do EMSA in agarose gel instead of the conventional approach of using acrylamide gel, it does lend support to our claim of stronger binding of Swi6 to siRNAs. Another relevant observation of binding of Swi6 to the “RevCen” RNA precursor RNAs but a detectable binding to siRNAs denoted as VI-IX (as measured by competition experiments, that are derived from RevCen RNA; Figure S4 and S7), which are derived by Dcr1 cleavage of the ‘’RevCen’’ RNA.

We also disagree that we carried out EMSA with a small bunch of siRNAs. As indicated in Figure 1 and S1, we synthesized nearly 12 siRNAs representing the dg-dh repeats at Cen, mat and tel loci and measured their specificity of binding to Swi6 using EMSA assay by labeling the ones labelled “D”, “E” and “V” directly and those of the remaining ones by the latter’s ability to compete against the binding (Figure 1, S4). These results point to presence of a consensus sequence in siRNAs that shows highly specific and strong binding to Swi6 in the low micromolar range.

Further, our claim of binding of Swi6 and not Swi63K>3A to siRNA in vivo is validated by RIP experiments, as shown in Fig 2 and S9.

c) The regions of Swi6 that bind to siRNAs need to be identified and evidence must be provided that Swi6 binds to RNAs of a specific length, 20-22 mers, to support the claim that Swi6 binds to siRNAs. This is critical for all the subsequent experiments and claims in the study.

We have provided both in vitro data, which is va;idiated in vivo by RIP experiments, as mentioned above. However, we agree that it wpuld be very interesting to identify the residues in Swi6 chromdomain responsible for binding to siRNA. However, such an investigation is beyond the scope of the present study.

(2) a) The in vivo results do not validate Swi6 binding to specific RNAs, as stated by the authors. Swi6 pulldowns have been shown to be enriched for all heterochromatic proteins including the RITS complex. The sRNA binding observed by the authors is therefore likely to be mediated by Ago1/RITS.

We disagree with the first comment. Our RIP experiments do validate the in vitro results (Fig 1, 2, S4 and S9), as argued above. The observation alluded to by the reviewer “Swi6 pulldowns have been shown to be enriched for all heterochromatic proteins including the RITS complex” is not inconsistent with our observation; it is possible that the siRNA may be released from the RITS complex and transferred to Swi6, possibly due to its higher affinity.

Thus, we would like to suggest that the role of Swi6 is likely to be coincidental or subsequent to that of Ago1/RITS (see below). We think that the binding by Swi6 to the siRNA and siRNA-DNA hybrid and could be also carried out in cis at the level of siRNA-DNA hybrids.

This point needs to be addressed in future studies.

b) Most of the binding in Figure S8C seems to be non-specific.

We would like to point out that the result in Figure S8C needs to be examined together with the Figure S8B, which shows RNA bound by Swi6 but not Swi63K-3A to hybridize with dg, dh and dh-k probes.

c) In Figure S8D, the authors' data shows that Swi6 deletion does not derepress the rev dh transcript while dcr1 delete cells do, which is consistent with previous reports but does not relate to the authors' conclusions.

The purpose of results shown in Figure S8D is just to compare the results of Swi6 with that of Swi63K-3A.

d) Previous results have shown that swi6 delete cells have 20-fold fewer dg and dh siRNAs than swi6+ cells due to decreased RNA-dependent RNA polymerase complex recruitment and reduced siRNA amplification.

This result is consistent with our results invoking a role of Swi6 in binding to, protecting and recruiting siRNAs to homologous sites.

To find if the overall production of siRNA is compromised in swi6 3K->3A mutant, we i) calculated the RIP-Seq read counts for swi6 3K->3A , swi6+ and vector control in 200 bp genomic bins , ii) divided the Swi6 3K->3A and swi6+ signals by that of control, iii) removed the background using the criteria of signal value < 25% of max signal, and iv) counted the total reads (in excess to control) in all peak regions in both samples.  This revealed a total count of 10878 and 8994 respectively for Swi6 3K->3A  and swi6+ samples, possibly implying that the overall siRNA production is not compromised in the Swi6 3K->3A mutant.

(3) a) The RIP-seq data are difficult to interpret as presented. The size distribution of bound small RNAs, and where they map along the genome should be shown as for example presented in previous Ago1 sRNA-seq experiments.

Please see the response to 2(d).

b) It is also unclear whether the defects in sRNA binding observed by the authors represent direct sRNA binding to Swi6 or co-precipitation of Ago1-bound sRNAs.

The correspondence between our in vivo and in vitro results suggests that the binding to Swi6 would be direct. We do not observe a complete correspondence between the Swi6- and Ago-bound siRNAs. We think Swi6 binding may be coincident with or following RITS complex formation.

This point will be discussed in the Revision.

The authors should also sequence total sRNAs to test whether Swi6-3A affects sRNA synthesis, as is the case in swi6 delete cells.

Please see response to 2(d) above.

(4) The authors examine the effects of Swi6-3A mutant by overexpression from the strong nmt1 promoter. Heterochromatin formation is sensitive to the dosage of Swi6. These experiments should be performed by introducing the 3A mutations at the endogenous Swi6 locus and effects on Swi6 protein levels should be tested.

Although we agree, we think that the heterochromatin formation is occurring in presence of nmt1-driven Swi6 but not Swi63K>3A, as indicated by the phenotype and Swi6 enrichment at otr1R::ade6, imr1::ura4 and his3-telo (Figure 3) and mating type (Fig. S10). Furthermore, the both GFP-Swi6 and GFPSwi63K>3A are expressed at similar level (Fig. S8A).

(5) The authors' data indicate an impairment of silencing in Swi6-3A mutant cells but whether this is due to a general lower affinity for nucleosomes, DNA, RNA, or as claimed by the authors, siRNAs is unclear. These experiments are consistent with previous findings suggesting an important role for basic residues in the HP1 hinge region in gene silencing but do not reveal how the hinge region enhances silencing.

Our study aims to correlate the binding of Swi6 but not Swi63K-3A to siRNA with its localization to heterochromatin. A similar difference in binding of Swi6 but not Swi63K-3A to siRNA-DNA hybrid, together with sensitivity of silencing and Swi6 localization to heterochromatin to RNaseH support the above correlations as being causally connected.

In terms of mechanism of binding, we need to clarify that the primary mode of binding is through the CD and not the hinge domain, although the hinge domain does influence this binding. This result is different from those of Keller et al.

We have some structural data based on preliminary SAXS experiment supporting binding of siRNA to the CD and influence of the hinge domain on this binding. However, this line of investigation need to be extended and will be subject of future investigations.

(6) RNase H1 overexpression may affect Swi6 localization and silencing indirectly as it would lead to a general reduction in R loops and RNA-DNA hybrids across the genome. RNaseH1 OE may also release chromatin-bound RNAs that act as scaffolds for siRNA-Ag1/RITS complexes that recruit Clr4 and ultimately Swi6.

These are formal possibilities. However, the correlation between swi6 binding to siRNA-DNA hybrid and delocalization upon RNase H1 treatment argues for a more direct link.

(7) Examples of inaccurate presentation of the literature.

a) The authors state that "RNA binding by the murine HP1 through its hinge domains is required for heterochromatin assembly (Muchardt et al, 2002). The cited reference provides no evidence that HP1 RNA binding is required for heterochromatin assembly. Only the hinge region of bacterially produced HP1 contributes to its localization to DAPI-stained heterochromatic regions in fixed NIH 3T3 cells.

Noted. Statement will be corrected.

b) "... This scenario is consistent with the loss of heterochromatin recruitment of Swi6 as well as siRNA generation in rnai mutants (Volpe et al, 2002)." Volpe et al. did not examine changes in siRNA levels in swi6 mutant cells. In fact, no siRNA analysis of any kind was reported in Volpe et al., 2002.

Correct.  We only say that Swi6 recruitment is reduced in rnai mutants and correlate it with ability of SWi6 to bind to siRNA generated by RNAi and subsequently to siRNA-DNA hybrid.

Reviewer #2 (Public review):

The aim of this study is to investigate the role of Swi6 binding to RNA in heterochromatin assembly in fission yeast. Using in vitro protein-RNA binding assays (EMSA) they showed that Swi6/HP1 binds centromere-derived siRNA (identified by Reinhardt and Bartel in 2002) via the chromodomain and hinge domains. They demonstrate that this binding is regulated by a lysine triplet in the conserved region of the Swi6 hinge domain and that wild-type Swi6 favours binding to DNA-RNA hybrids and siRNA, which then facilitates, rather than competes with, binding to H3K9me2 and to a lesser extent H3K9me3.

However, the majority of the experiments are carried out in swi6 null cells overexpressing wild-type Swi6 or Swi63K-3A mutant from a very strong promoter (nmt1). Both swi6 null cells and overexpression of Swi6 are well known to exhibit phenotypes, some of which interfere with heterochromatin assembly. This is not made clear in the text.

We think that the argument is not valid as we show that swi6 but not Swi63K-3A could restore silencing at imr1::ura4, otr1::ade6 and his3-telo (Fig 3) and mating type (Fig. S10), when transformed into a swi6D strain.

Whilst the RNA binding experiments show that Swi6 can indeed bind RNA and that binding is decreased by Swi63K-3A mutation in vitro (confusingly, they only much later in the text explained that these 3 bands represent differential binding and that II is likely an isotherm). The gels showing these data are of poor quality and it is unclear which bands are used to calculate the Kd.

We disagree with the comment about the quality of EMSA data. We think it is of similar quality or better than that of Keller et al, except in some cases, like Fig 1D, a shorter exposure shown to distinguish the slowest shifted band has caused the remaining bands to look fainter.

RNA-seq data shows that overall fewer siRNAs are produced from regions of heterochromatin in the Swi63K-3A mutant so it is unsurprising that analysis of siRNA-associated motifs also shows lower enrichment (or indeed that they share some similarities, given that they originate from repeat regions).

Please see response to comment 2(d) of the first reviewer above.

It is not clear which bands are being alluded to. However, we‘ll rectify any gaps in information in the revision.

The experiments are seemingly linked yet fail to substantiate their overall conclusions. For instance, the authors show that the Swi63K-3A mutant displays reduced siRNA binding in vitro (Figure 1D) and that H3K9me2 levels at heterochromatin loci are reduced in vivo (Figure 3C-D). They conclude that Swi6 siRNA binding is important for Swi6 heterochromatin localization, whilst it remains entirely possible that heterochromatin integrity is impaired by the Swi63K-3A mutation and hence fewer siRNAs are produced and available to bind. Their interpretation of the data is really confusing.

Our argument is that the lack of binding by Swi63K>3A to siRNA can explain the loss of recruitment to heterochromatin loci and thus affect the integrity of heterochroamtin; the recruitment of Swi6 can occur possibly by binding initially to siRNA and thereafter as siRNA-DNA hybrid. However, the overall level of siRNAs is not affected, as in 2(D) above. This interpretation is supported by results of ChIP assay and confocal experiments, as also by the effect of RNaseH1 in the recruitment of Swi6.

The authors go on to show that Swi63K-3A cells have impaired silencing at all regions tested and the mutant protein itself has less association with regions of heterochromatin. They perform DNA-RNA hybrid IPs and show that Swi63K-3A cells which also overexpress RNAseH/rnh1 have reduced levels of dh DNA-RNA hybrids than wild-type Swi6 cells. They interpret this to mean that Swi6 binds and protects DNA-RNA hybrids, presumably to facilitate binding to H3K9me2. The final piece of data is an EMSA assay showing that "high-affinity binding of Swi6 to a dg-dh specific RNA/DNA hybrid facilitates the binding to Me2-K9-H3 rather than competing against it." This EMSA gel shown is of very poor quality, and this casts doubt on their overall conclusion.

We do agree with the reviewer about the quality of EMSA (Fig. 5B). However, as may be noticed in the EMSA for siRNA-DNA hybrid binding  (Fig 4A), the bands of Swi6-bound siRNA-DNA hybrid are extremely retarded. Hence the EMSA for subsequent binding by H3-K9-Me peptides required a longer electrophoretic run, which led to reduction in the sharpness of the bands. Nevertheless, the data does indicate binding efficiency in the order H3K9-Me2> H3-K9-Me3 > H3-K9-Me0. Having said that, we plan to repeat the EMSA or address the question by other methods, like SPR.

Unfortunately, the manuscript is generally poorly written and difficult to comprehend. The experimental setups and interpretations of the data are not fully explained, or, are explained in the wrong order leading to a lack of clarity. An example of this is the reasoning behind the use of the cid14 mutant which is not explained until the discussion of Figure 5C, but it is utilised at the outset in Figure 5A.

We tend to agree somewhat and will attempt to submit a revised version with greater clarity, as also the explanation of experiment with cid14D strain.

Another example of this lack of clarity/confusion is that the abstract states "Here we provide evidence in support of RNAi-independent recruitment of Swi6". Yet it then states "We show that...Swi6/HP1 displays a hierarchy of increasing binding affinity through its chromodomain to the siRNAs corresponding to specific dg-dh repeats, and even stronger binding to the cognate siRNA-DNA hybrids than to the siRNA precursors or general RNAs." RNAi is required to produce siRNAs, so their message is very unclear. Moreover, an entire section is titled "Heterochromatin recruitment of Swi6-HP1 depends on siRNA generation" so what is the author's message?

The reviewer has correctly pointed out the error. Indeed, our results actually indicate an RNAi-dependent rather than independent mode of recruitment. Rather, we would like to suggest an H3-K9-Me2-indpendnet recruitment of Swi6. We will rectify this error in our revised manuscript.

The data presented, whilst sound in some parts is generally overinterpreted and does not fully support the author's confusing conclusions. The authors essentially characterise an overexpressed Swi6 mutant protein with a few other experiments on the side, that do not entirely support their conclusions. They make the point several times that the KD for their binding experiments is far higher than that previously reported (Keller et al Mol Cell 2012) but unfortunately the data provided here are of an inferior quality and thus their conclusions are neither fully supported nor convincing.

We have used the method of Heffler et al (2012) to compute the Kd from EMSA data.

eLife Assessment

This manuscript provides a valuable in-depth biochemical analysis of p53 isoforms, highlighting their aggregation propensity, interaction with chaperones, and potential dominant-negative effects on p53 family members. The study presents solid evidence of isoform-specific properties, which may contribute to protein misfolding and impaired cellular function in cancer. While highly informative, the findings would benefit from further discussion of physiological relevance, given the high isoform expression levels used, and addressing prior evidence of isoform-specific transcriptional activity. Overall, this work significantly advances our understanding of p53 isoform biochemistry and its implications for cancer research.

Reviewer #1 (Public review):

Summary:

Brdar, Osterburg, Munick, et al. present an interesting cellular and biochemical investigation of different p53 isoforms. The authors investigate the impact of different isoforms on the in-vivo transcriptional activity, protein stability, induction of the stress response, and hetero-oligomerization with WT p53. The results are logically presented and clearly explained. Indeed, the large volume of data on different p53 isoforms will provide a rich resource for researchers in the field to begin to understand the biochemical effects of different truncations or sequence alterations.

Strengths:

The authors achieved their aims to better understand the impact/activity of different p53 is-forms, and their data will support their statements. Indeed, the major strengths of the paper lie in its comprehensive characterization of different p53 isoforms and the different assays that are measured. Notably, this includes p53 transcriptional activity, protein degradation, induction of the chaperone machinery, and hetero-oligomerization with wtp53. This will provide a valuable dataset where p53 researchers can evaluate the biological impact of different isoforms in different cell lines. The authors went to great lengths to control and test for the effect of (1) p53 expression level, (2) promotor type, and (3) cell type. I applaud their careful experiments in this regard.

Weaknesses:

One thing that I would have liked to see more of is the quantification of the various pull-down/gel assays - to better quantify the effect of, e.g., hetero-oligomerization among the various isoforms. In addition, a discussion about the role of isoforms that contain truncations in the IDRs is not available. It is well known that these regions function in an auto-inhibitory manner (e.g. work by Wright/Dyson) and also mediate many PPIs, which likely have functional roles in vivo (e.g. recruiting p53 to various complexes). The discussion could be strengthened by focusing on some of these aspects of p53 as well.

Reviewer #2 (Public review):

Summary:

In this manuscript entitled "p53 isoforms have a high aggregation propensity, interact with chaperones and lack 1 binding to p53 interaction partners", the authors suggest that the p53 isoforms have high aggregation propensity and that they can co-aggregate with canonical p53 (FLp53), p63 and p73 thus exerting a dominant-negative effect.

Strengths:

Overall, the paper is interesting as it provides some characterization of most p53 isoforms DNA binding (when expressed alone), folding structure, and interaction with chaperones. The data presented support their conclusion and bring interesting mechanistic insight into how p53 isoforms may exert some of their activity or how they may be regulated when they are expressed in excess.

Weaknesses:

The main limitation of this manuscript is that the isoforms are highly over-expressed throughout the manuscript, although the authors acknowledge that the level of expression is a major factor in the aggregation phenomenon and "that aggregation will only become a problem if the expression level surpasses a certain threshold level" (lines 273-274 and results shown in Figures S3D, 6E). The p53 isoforms are physiologically expressed in most normal human cell types at relatively low levels which makes me wonder about the physiological relevance of this phenomenon.

Furthermore, it was previously reported that some isoforms clearly induce transcription of target genes which are not observed here. For example, p53β induces p21 expression (Fujita K. et al. p53 isoforms Delta133p53 and p53beta are endogenous regulators of replicative cellular senescence. Nat Cell Biol. 2009 Sep;11(9):1135-42), and Δ133p53α induces RAD51, RAD52, LIG4, SENS1 and SOD1 expression (Gong, L. et al. p53 isoform D113p53/D133p53 promotes DNA double-strand break repair to protect cell from death and senescence in response to DNA damage. Cell Res. 2015, 25, 351-369. / Gong, L. et al. p53 isoform D133p53 promotes the efficiency of induced pluripotent stem cells and ensures genomic integrity during reprogramming. Sci. Rep. 2016, 6, 37281. / Horikawa, I. et al. D133p53 represses p53-inducible senescence genes and enhances the generation of human induced pluripotent stem cells. Cell Death Differ. 2017, 24, 1017-1028. / Gong, L. p53 coordinates with D133p53 isoform to promote cell survival under low-level oxidative stress. J. Mol. Cell Biol. 2016, 8, 88-90. / Joruiz et al. Distinct functions of wild-type and R273H mutant Δ133p53α differentially regulate glioblastoma aggressiveness and therapy-induced senescence. Cell Death Dis. 2024 Jun 27;15(6):454.) which demonstrates that some isoforms can induce target genes transcription and have defined normal functions (e.g. Cellular senescence or DNA repair).

However, in this manuscript, the authors conclude that isoforms are "largely unfolded and not capable of fulfilling a normal cellular function" (line 438), that they do not have "well defined physiological roles" (line 456), and that they only "have the potential to inactivate members of the p53 protein family by forming inactive hetero complexes with wtp53" (line 457-458).

Therefore, I think it is essential that the authors better discuss this major discrepancy between their study and previously published research.

eLife Assessment

This study examines age-related, sex-specific gene expression and alternative splicing in humans using the GTEx dataset. Solid evidence is provided to demonstrate that alternative splicing was affected by both sex and age across many tissues in this dataset. Although the authors performed comprehensive computational analyses with useful 'transcriptomic' changes with sex/age, they did not validate their findings with independent longitudinal datasets. This limits the wide impact of this study but can be used as a starting point to examine sex- and age differences in the transcriptome due to alternative splicing.

Reviewer #1 (Public review):

Summary:

Wang et al. investigate sexual dimorphic changes in the transcriptome of aged humans. This study relies upon analysis of the Genotype-Tissue Expression dataset that includes 54 tissues from human donors. The authors investigate 17,000 transcriptomes from 35 tissues to investigate the effect of age and sex on transcriptomic variation, including the analysis of alternative splicing. Alternative splicing is becoming more appreciated as an influence in the aging process, but how it is affected by sexual dimorphism is still largely unclear. The authors investigated multiple tissues but ended up distilling brain tissue down to four separate regions: decision, hormone, memory, and movement. Building upon prior work, the authors used an analysis method called principal component-based signal-to-variation ratio (pcSVR) to quantify differences between sex or age by considering data dispersion. This method also considers differentially expressed genes and alternative splicing events.

Strengths:

(1) The authors investigate sexual dimorphism on gene expression and alternative splicing events with age in multiple tissues from a large publicly available data set that allows for reanalysis.

(2) Furthermore, the authors take into account the ethnic background of donors. Identification of aging-modulating genes could be useful for the reanalysis of prior data sets.

Weaknesses:

The models built off of the GTEx dataset should be tested in another data set (ex. Alzheimer's disease) where there are functional changes that can be correlated. Gene-length-dependent transcription decline, which occurs with age and disease, should also be investigated in this data set for potential sexual dimorphism.

Reviewer #2 (Public review):

Summary:

In this manuscript, Wang et al analyze ~17,000 transcriptomes from 35 human tissues from the GTEx database and address transcriptomic variations due to age and sex. They identified both gene expression changes as well as alternative splicing events that differ among sexes. Using breakpoint analysis, the authors find sex dimorphic shifts begin with declining sex hormone levels with males being affected more than females. This is an important pan-tissue transcriptomic study exploring age and sex-dependent changes although not the first one.

Strengths:

(1) The authors use sophisticated modeling and statistics for differential, correlational, and predictive analysis.

(2) The authors consider important variables such as genetic background, ethnicity, sampling bias, sample sizes, detected genes, etc.

(3) This is likely the first study to evaluate alternative splicing changes with age and sex at a pan-tissue scale.

(4) Sex dimorphism with age is an important topic and is thoroughly analyzed in this study.

Weaknesses:

(1) The findings have not been independently validated in a separate cohort or through experiments. Only selective splicing factor regulation has been verified in other studies.

(2) It seems the authors have not considered PMI or manner of death as a variable in their analysis.

(3) The manuscript is very dense and sometimes difficult to follow due to many different types of analyses and correlations.

(4) Short-read data can detect and quantify alternative splicing events with only moderate confidence and therefore the generalizability of these findings remains to be experimentally validated.

Reviewer #3 (Public review):

Summary:

In this study, Wang et al utilized the available GTEx data to compile a comprehensive analysis that attempt to reveal aging-related sex-dimorphic gene expression as well as alternative splicing changes in humans.

The key conclusions based on their analysis are that

(1) extensive sex-dimorphisms during aging with distinct patterns of change in gene expression and alternative splicing (AS), and

(2) the male-biased age-associated AS events have a stronger association with Alzheimer's disease, and

(3) the female-biased events are often regulated by several sex-biased splicing factors that may be controlled by estrogen receptors. They further performed break-point analysis and revealed that in males there are two main breakpoints around ages 35 and 50, while in females, there is only one breakpoint at 45.

Strengths:

This study sets an ambitious goal, leveraging the extensive GTEx dataset to investigate aging-related, sex-dimorphic gene expression and alternative splicing changes in humans. The research addresses a significant question, as our understanding of sex-dimorphic gene expression in the context of human aging is still in its early stages. Advancing our knowledge of these molecular changes is vital for identifying therapeutic targets for age-related diseases and extending the human health span. The study is highly comprehensive, and the authors are commendable for their attempted thorough analysis of both gene expression and alternative splicing - an area often overlooked in similar studies.

Weaknesses:

Due to the inherent noise within the GTEx dataset - which includes numerous variables beyond aging and sex - there are significant technical concerns surrounding this study. Additionally, the lack of cross-validation with independent, existing data raises questions about whether the observed gene expression changes genuinely reflect those associated with human aging. For instance, the break-point analysis in this study identifies two major breakpoints in males around ages 35 and 50, and one breakpoint in females at age 45; however, these findings contradict a recent multi-omics longitudinal study involving 108 participants aged 25 to 75 years, where breakpoint at 44 and 60 years was observed in both male and females (Shen et al, 2024). These issues cast doubt on the robustness of the study's conclusions. Specific concerns are outlined below:

(1) The primary method used in this study is linear regression, incorporating age, sex, and age-by-sex interactions as covariates, alongside other confounding factors (such as ethnicity) as unknown variables. However, the analysis overlooks two critical known variables in the GTEx dataset: time of death (TOD) and postmortem interval (PMI). Both TOD and PMI are recorded for each sample and account for substantial variance in gene expression profiles. A recent study by Wucher et al.(Wucher et al, 2023) demonstrated the powerful impact of TOD on gene expression by using it to reconstruct human circadian and even circannual datasets. Similarly, Ferreira et al. (Ferreira et al, 2018) highlighted PMI's influence on gene expression patterns. Without properly adjusting for these two variables, confidence in the study's conclusions remains limited at best.

(2) To demonstrate that their analysis is robust and that the covariates TOD and PMI are otherwise negligible - the authors should cross-validate their findings with independent datasets to confirm that the identified gene expression changes are reproducible for some tissues. For instance, the recent study by Shen et al. (Shen et al., 2024) in Nature Aging offers an excellent dataset for cross-validation, particularly for blood samples. Comparing the GTEx-derived results with this longitudinal transcriptome dataset would enable verification of gene expression changes at both the individual gene and pathway levels. Without such validation, confidence in the study's conclusions remains limited.

(3) As a demonstration of the lack of such validation, in the Shen et al. study (Shen et al., 2024), breakpoints at 44 and 60 years were observed in both males and females, while this study identifies two major breakpoints in males around ages 35 and 50, and one breakpoint in females at age 45. What caused this discrepancy?

(4) Although the alternative splicing analysis is intriguing, the authors did not differentiate between splicing events that alter the protein-coding sequence and those that do not. Many splicing changes occurring in the 5' UTR and 3' UTR regions do not impact protein coding, so it is essential to filter these out and focus specifically on alternative splicing events that can modify protein-coding sequences.

(5) One of the study's main conclusions - that "male-biased age-associated AS events have a stronger association with Alzheimer's disease" - is not supported by the data presented in Figure 4A, which shows an association with "regulation of amyloid precursor formation" only in female, not male, alternative splicing genes. Additionally, the gene ontology term "Alzheimer's disease" is absent from the unbiased GO analysis in Figure S6. These discrepancies suggest that the focus on Alzheimer's disease may reflect selective data interpretation rather than results driven by an unbiased analysis.

(6) The experimental data presented in Figures 5E - I merely demonstrate that estrogen receptor regulates the expression of two splicing factors, SRSF1 and SRSF7, in an estradiol-dependent manner. However, this finding does not support the notion that this regulation actually contributes to sex-dimorphic alternative splicing changes during human aging. Notably, the authors do not provide evidence that SRSF1 and SRSF7 expression changes actually occur in a sex-dependent manner with human aging (in a manner similar to TIA1). As such, this experimental dataset is disconnected from the main focus of the study and does not substantiate the conclusions on sex-dimorphic splicing during human aging. The authors performed RNA-seq in wild-type and ER mutant cells, and they should perform a comprehensive analysis of ER-dependent alternative splicing and compare the results with the GTEx data. It should be straightforward.

References:

Ferreira PG, Muñoz-Aguirre M, Reverter F, Sá Godinho CP, Sousa A, Amadoz A, Sodaei R, Hidalgo MR, Pervouchine D, Carbonell-Caballero J et al (2018) The effects of death and post-mortem cold ischemia on human tissue transcriptomes. Nature Communications 9: 490.

Shen X, Wang C, Zhou X, Zhou W, Hornburg D, Wu S, Snyder MP (2024) Nonlinear dynamics of multi-omics profiles during human aging. Nature Aging.

Wucher V, Sodaei R, Amador R, Irimia M, Guigó R (2023) Day-night and seasonal variation of human gene expression across tissues. PLOS Biology 21: e3001986.

eLife Assessment

This useful study informs the transcriptional mechanisms that promote stem cell differentiation and prevent degeneration in the adult eye. Through inducible mouse mutagenesis, the authors uncover a dual role for a transcription factor (Sox9) in stem cell differentiation and prevention of retinal degeneration. The data at hand provide solid support to the main conclusions with several minor weaknesses identified as well. The study will be of general interest to the fields of neuronal development and neurodegeneration.

Reviewer #1 (Public review):

Summary:

Hurtado et al. show that Sox9 is essential for retinal integrity, and its null mutation causes the loss of the outer nuclear layer (ONL). The authors then show that this absence of the ONL is due to apoptosis of photoreceptors and a reduction in the numbers of other retinal cell types such as ganglion cells, amacrine cells, and horizontal cells. They also describe that Müller Glia undergoes reactive gliosis by upregulating the Glial Fibrillary Acidic Protein. The authors then show that Sox9+ progenitors proliferate and differentiate to generate the corneal cells through Sox9 lineage-tracing experiments. They validate Sox9 expression and characterize its dynamics in limbal stem cells using an existing single-cell RNA sequencing dataset. Finally, the authors argue that Sox9 deletion causes progenitor cells to lose their clonogenic capacity by comparing the sizes of control and Sox9-null clones. Overall, Hurtado et al. underline the importance of Sox9 function in retinal and corneal cells.

Strengths:

The authors have characterized a myriad of striking phenotypes due to Sox9 deletion in the retina and limbal stem cells which will serve as a basis for future studies.

Weaknesses:

Hurtado et al. investigate the importance of Sox9 in the retina and limbal stem cells. However, the overall experimental narrative appears dispersed.

The authors begin by characterizing the phenotype of Sox9 deletion in the retina and show that the absence of the ON layer is due to photoreceptor apoptosis and a reduction in other retinal cell types. The authors also note that Müller glia undergoes gliosis in the Sox9 deletion condition. These striking observations are never investigated further, and instead, the authors switch to lineage-tracing experiments in the limbus that seem disconnected from the first three figures of the paper. Another example of this disconnect is the comparison of Sox9 high and Sox9 low populations using an existing scRNA-seq dataset and the subsequent GO term analysis, which does not directly tie in with the lineage-tracing data of the succeeding Sox9∆/∆ experiments.

A major concern is that a single Sox9∆/∆ limbal clone has a sufficiently large size, comparable to wild-type clones, as seen in Figure 6D. This singular result is contrary to their conclusion, which states that Sox9-deficient stem cells minimally contribute to the maintenance of the cornea.

Reviewer #2 (Public review):

Summary:

Sox9 is a transcription factor crucial for development and tissue homeostasis, and its expression continues in various adult eye cell types, including retinal pigmented epithelium cells, Müller glial cells, and limbal and corneal basal epithelia. To investigate its functional roles in the adult eye, this study employed inducible mouse mutagenesis. Adult-specific Sox9 depletion led to severe retinal degeneration, including the loss of Müller glial cells and photoreceptors. Further, lineage tracing revealed that Sox9 is expressed in a basal limbal stem cell population that supports stem cell maintenance and homeostasis. Mosaic analysis confirmed that Sox9 is essential for the differentiation of limbal stem cells. Overall, the study highlights that Sox9 is critical for both retinal integrity and the differentiation of limbal stem cells in the adult mouse eye.

Strengths:

In general, inducible genetic approaches in the adult mouse nervous system are rare and difficult to carry out. Here, the authors employ tamoxifen-inducible mouse mutagenesis to uncover the functional roles of Sox9 in the adult mouse eye.

Careful analysis suggests that two degeneration phenotypes (mild and severe) are detected in the adult mouse eye upon tamoxifen-dependent Sox9 depletion. Phenotype severity nicely correlates with the efficiency of Cre-mediated Sox9 depletion.

Molecular marker analysis provides strong evidence of Mueller cell loss and photoreceptor degeneration.

A clever genetic tracing strategy uncovers a critical role for Sox9 in limbal stem cell differentiation.

Weaknesses:

The Introduction can be improved by explaining clearly what was previously known about Sox9 in the eye. A lot of this info is mentioned in a single, 3-page long paragraph in the Discussion. However, the current study's significance and novelty would become clearer if the authors articulated in more detail in the Introduction what was already known about Sox9 in retina cell types (in vitro and in vivo).

Because a ubiquitous tamoxifen-inducible CreER line is employed, non-cell autonomous mechanisms possibly contribute to the observed retina degeneration. There is precedence for this in the literature. For example, RPE-specific ablation of Otx2 results in photoreceptor degeneration (PMID: 23761884). Have the authors considered the possibility of non-cell autonomous effects upon ubiquitous Sox9 deletion?

Given the similar phenotypes between animals lacking Otx2 and Sox9 in specific cell types of the eye, the authors are encouraged to evaluate Otx2 expression in the tamoxifen-induced Sox9 adult retina.

The most parsimonious explanation for the dual role of Sox9 in retinal cell types and limbal stem cells is that the cell context is different. For example, Sox9 may cooperate with TF1 in photoreceptors, TF2, in Mueller cells, and TF3 in limbal stem cells, and such cell type-specific cooperation may result in different outcomes (retinal integrity, stem cell differentiation). The authors are encouraged to add a paragraph to the discussion and share their thoughts on the dual role of Sox9.

One more molecular marker for Mueller glial cells would strengthen the conclusion that these cells are lost upon Sox9 deletion.

Using opsins as markers, the authors conclude that the photoreceptors are lost upon Sox9 deletion. However, an alternate possibility is that the photoreceptors are still present and that Sox9 is required for the transcription of opsin genes. In that case, Sox9 (like Otx2) may act as a terminal selector in photoreceptor cells. This point is particularly important because vertebrate terminal selectors (e.g., Nurr1, Otx2, Brn3a) initially affect neuron type identity and eventually lead to cell loss.

Quantification is needed for the TUNEL and GFAP analysis in Figure 3.

Line 269-320: The authors examined available scRNA-Seq data on adult retina. This data provides evidence for Sox9 expression in distinct cell types. However, the dataset does not inform about the functional role of Sox9 because Sox9 mutant cells were not analyzed with RNA-Seq. Hence, all the data that claim that this experiment provides insights into possible Sox9 functional roles must be removed. This includes panels F, G, and H in Figure 5. In general, this section of the paper (Lines 269-320) needs a major revision. Similarly, lines 442-446 in the Discussion should be removed.

eLife Assessment

This study presents numerical results on a framework for understanding the dynamics of subthreshold waves in a network of electrical synapses modeled on the connectome data of the C elegans nematode. The strength of the evidence presented in favor of interference effects being a major component in subthreshold wave dynamics is inadequate and the approach is flawed. Substantial methodological issues are present, including altering the original network structure of the connectome without a clear justification and providing little motivation for the choice of numerical parameters values that were used.

Reviewer #1 (Public review):

Summary:

This work investigates numerically the propagation of subthreshold waves in a model neural network that is derived from the C. elegans connectome. Using a scattering formalism and tight-binding description of the network -- approximations which are commonplace in condensed matter physics -- this work attempts at showing the relevance of interference phenomena, such as wavenumber-dependent propagation, for the dynamics of subthreshold waves propagating in a network of electrical synapses.

Strengths:

The primary strength of the work is in trying to use theoretical tools from a far-away corner of fundamental physics to shed light on the properties of a real neural system.

Weaknesses:

The authors provide a good introduction and motivation for studying the propagation of subthreshold oscillations in the inferior olive nuclei. However, they chose to use the C elegans connectome for their study, and the implications of this work for C elegans neuroscience remain unclear by the end of the preprint. The authors should also give more evidence for the claim that their study may give a mechanism for synchronized rhythmic activity in the mammalian inferior olive nucleus, or refrain from making this conclusion. In the same vein, since the work emphasizes the dependence on the wavenumber for the propagation of subthreshold oscillations, they should make an attempt at estimating the wavenumber of subthreshold oscillations in C elegans if they were to exist and be observed. Next, the presence of two "mobility edges" in the transmission coefficient calculated in this work is unmistakably due to the discrete nature of the system, coming from the tight-binding approximation, and it is unclear to me if this approximation is justified in the current system. Similarly, it is possible that the wavenumber-dependent transmission observed depends strongly on the addition of a large number of virtual nodes (VNs) in the network, which the authors give little to no motivation for. As these nodes are not present in the C elegans connectome, the authors should explain the motivation for their inclusion in the model and should discuss their consequences on the transmission properties of the network. As it stands, I think the work would only have a very limited impact on the understanding of subthreshold oscillations in the rat or in C elegans. Indeed, the preprint falls short of relating its numerical results to any phenomena which could be observed in the lab.

Reviewer #2 (Public review):

This manuscript addresses an interesting and important question: the basic mechanisms underlying subthreshold intrinsic oscillations in the inferior olive. Instead of a direct investigation of the questions, the authors decide to study subthreshold oscillations in the C-elegance, where the connectivity pattern is known but does not exhibit sub-threshold oscillations. Furthermore, instead of the common description of gap-junction coupling by resistors, the authors decide to represent the system as a tight-binding Anderson Hamiltonian.

Weaknesses:

The authors study an architecture of the C-elegance instead of that of the inferior olive of mammals because the architecture of C-elegance is known.

No subthreshold oscillations were identified in the C-elegance.<br /> Instead of representing electrical coupling via resistors that connect neurons, the authors use a quantum formalism and introduce the tight-binding Anderson Hamiltonian. Why?

Equally spaced two virtual nodes were added between cells connected by a gap junction. Why?

Comments on revised version:

Last time, I recommended that the authors should represent electrical coupling via resistors that connect neurons instead of via the quantum formalism. The authors have not tested this direction.

Author response:

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

Joint Public Review:

(1) This work investigates numerically the propagation of subthreshold waves in a model neural network that is derived from the C. elegans connectome. Using a scattering formalism and tight-binding description of the network -- approximations which are commonplace in condensed matter physics -- this work attempts to show the relevance of interference phenomena, such as wavenumber-dependent propagation, for the dynamics of subthreshold waves propagating in a network of electrical synapses.

(2) The primary strength of the work is in trying to use theoretical tools from a far-away corner of fundamental physics to shed light on the properties of a real neural system. While a system composed of neurons and synapses is classical in nature, there are occasions in which interference or localization effects are useful for understanding wave propagation in complex media [review, van Rossum & Nieuwenhuizen, 1999]. However, it is expected that localization effects only have an impact in some parameter regimes and with low phase dissipation. The authors should have addressed the existence of this validity regime in detail prior to assuming that interference effects are important.

The theoretical concept and tool used in this study are not situated in a far-away corner of fundamental physics but hold one of the central positions in condensed matter physics and statistical physics. In fact, the non-scientific statement about where the theoretical concept and tool employed by the researchers are positioned within the realm of fundamental physics is irrelevant. The fundamental physics governs the foundations of all natural phenomena, and thus it provides indispensable principles for interpreting not only neural systems but also all life phenomena. One such principle explored in our study is the interference and localization of waves.

Specifically, in the third paragraph of the Introduction, we introduced that the interference effect of subthreshold oscillating waves, beyond being a theoretical possibility, is a phenomenon actually observed in neural tissue (Chiang and Durand, 2023; Gupta et al., 2016). Moreover, according to Devor and Yarom (2002), the propagation of subthreshold oscillations observed in the inferior olivary nucleus extended beyond a distance of 0.2 mm. Therefore, considering the propagation of subthreshold waves and the resulting interference in the connectome of C. elegans, which has a total body length of less than 1 mm, a diameter of about 0.08 mm, and most neurons distributed in the ring structure near its neck, provides sufficient validity for the initiation of theoretical and computational studies.

The primary objective of our study is to investigate which regimes of signal transmission/localization and interference phenomena are valid within the network of electrical synapses in C. elegans, the only system for which the neural connectome structure is perfectly known. As the Reviewer rightly pointed out in the question, this is exactly the issue that the Reviewer is curious about. Therefore, the existence of this validity regime cannot be addressed prior to conducting the study but can only be identified as a result of performing the research. And we have conducted such a study.

(3) An additional approximation that was made without adequate justification is the use of a tight-binding Hamiltonian. This can be a reasonable approximation, even for classical waves, in particular in the presence of high-quality-factor resonators, where most of the wave amplitude is concentrated on the nodes of the network, and nodes are coupled evanescently with each other. Neither of these conditions were verified for this study.

The tight-binding Anderson Hamiltonian we used in this study originally consisted of the on-site energy at each node and the hopping matrix between nodes. When the on-site energy is relatively much more stable (i.e., has a large negative value) compared to the hopping matrix, most of the wave amplitude becomes concentrated on the nodes as the Reviewer mentioned. However, as is well-known from reference papers (Anderson, 1958; Chang et al., 1995; Meir et al., 1989; Shapir et al., 1982; Thomas and Nakanishi, 2016), in this study, we also removed the on-site energy to prevent the waves from being concentrated on the nodes. Therefore, the tight-binding Hamiltonian we used in this study ensures that waves propagate through edges in the network where the values of the hopping matrix exist.

To assist the Reviewer in better understanding the model used in this study, we provide additional explanations as follows. In the manuscript, we have already provided detailed descriptions of the setup using the tight-binding Anderson Hamiltonian in the Method section under “Construction of our circuit model” and the explanation of Figure 1. In the model we used, the edges represented by solid lines are perfect conductors, while the dotted lines representing gap junctions act as potential barriers (Fig. 1B). Therefore, when electric signals propagate, we are dealing with the phenomenon where signals transmitted through the edges encounter potential barriers, causing scattering or attenuation. The model described by the Reviewer is indeed a commonly used model in condensed matter physics, but we did not use the exact model mentioned by the Reviewer. Instead, as is common in well-known reference papers, we modified it to suit our purposes. We hope this explanation helps the Reviewer gain a better understanding.

(4) The motivation for this work is to understand the basic mechanisms underlying subthreshold intrinsic oscillations in the inferior olive, but detailed connectivity patterns in this brain area are not available. The connectome is known for C elegans, but sub-threshold oscillations have not been observed there, and the implications of this work for C elegans neuroscience remain unclear. The authors should also give more evidence for the claim that their study may give a mechanism for synchronized rhythmic activity in the mammalian inferior olive nucleus, or refrain from making this conclusion.

We agree with the Reviewer's point. In this study, we do not provide additional analysis on the mammalian inferior olive nucleus beyond what is already known from previous research. What we intended to discuss in the Discussion section was to suggest that within our model, there is a “possibility” that a group of cells exchanging wave signals of a specific wavenumber with high transmittance may show synchronized rhythmic activity. Therefore, to avoid any misunderstanding for the reader, we have revised the corresponding sentence in the Discussion as follows.

In the Discussion, “The plausible possibility according to our model study is that the constructive interference of subthreshold membrane potential waves with a specific wavenumber may generate the synchronized rhythmic activation.

(5) In the same vein, since the work emphasizes the dependence on the wavenumber for the propagation of subthreshold oscillations, they should make an attempt at estimating the wavenumber of subthreshold oscillations in C elegans if they were to exist and be observed. Next, the presence of two "mobility edges" in the transmission coefficient calculated in this work is unmistakably due to the discrete nature of the system, coming from the tight-binding approximation, and it is unclear if this approximation is justified in the current system.

In this study, we modeled the propagation of subthreshold waves on the electrical synapse network of C. elegans, but we did not explain the generation of subthreshold oscillations themselves. Here, we simply injected wave signals with various wavenumber values into the network using a hypothetical device called an "Injector." As the Reviewer pointed out, estimating the wavenumbers of subthreshold oscillations that may exist or be observed in C. elegans would require a comprehensive investigation of the membrane potential dynamics occurring in the membranes of individual neurons. However, this is beyond the scope of this study and would require considerable effort to accomplish.

As for the use of the tight-binding Hamiltonian, we have addressed that in our response to the third paragraph in the Joint Public Review above.

(6) Similarly, it is possible that the wavenumber-dependent transmission observed depends strongly on the addition of a large number of virtual nodes (VNs) in the network, which the authors give little to no motivation for. As these nodes are not present in the C elegans connectome, the authors should explain the motivation for their inclusion in the model and should discuss their consequences on the transmission properties of the network.

As mentioned in our response to the third paragraph in the Joint Public Review above, in our model, a node is simply a pathway for waves to pass through. Therefore, inserting virtual nodes between two neurons that are connected in the C. elegans connectome does not alter the actual connection structure. In other words, virtual nodes do not create new connections between cells that didn’t exist in the connectome. The virtual nodes we introduced are merely a way to divide the sections—axon, gap junction, dendrite—through which the wave passes when it is transmitted between two neurons. As we have already explained in Fig. 1B, the edge connected by two virtual nodes, represented by a dotted line, is motivated to depict the gap junction acting as a potential barrier. We hope this explanation helps the Reviewer better understand the model used in this study.

(7) As it stands, the work would only have a very limited impact on the understanding of subthreshold oscillations in the rat or in C elegans. Indeed, the preprint falls short of relating its numerical results to any phenomena which could be observed in the lab.

In this study, we proposed a minimalistic model built using the currently available but limited C. elegans connectome information. Specifically, our model is not a phenomenological one that adjusts parameters to accurately predict experimental measurements, but rather an attempt at a novel conceptual approach to theoretically possible scenarios. While the model may not be satisfactory enough to explain experimental phenomena at present, it is a theoretical/computational study that someone needs to undertake. We believe this is the path of scientific progress. Therefore, as the Reviewer has expressed concern, it is entirely understandable that reproducing the numerical results measured in actual experiments is difficult in this study. Nevertheless, we believe that this study makes a basic contribution to the conceptual understanding of subthreshold signal propagation in C. elegans’ electric synapses.

Rather than offering a stretched opinion, we maintain a positive hope that future researchers in this field will improve the model by incorporating more detailed and extensive biological data through follow-up studies, allowing us to get closer to describing real phenomena.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The word "Sensory" was misspelled in Figures 2, 4 and 5.

We appreciate the feedback from Reviewer #1. We have corrected the mentioned typos in Figures 2, 4, and 5 of the revised manuscript.

eLife Assessment

This important study uses calcium imaging to show an increase in the selectivity of the sensory-evoked response in the apical dendritic tuft of layer 5 barrel cortex neurons as mice learn a whisker-dependent discrimination task. The evidence supporting the conclusions is compelling, and this work will be of great interest to neuroscientists working on reward-based learning and sensory processing.

Reviewer #1 (Public review):

What neurophysiological changes support the learning of new sensorimotor transformations is a key question in neuroscience. Many studies have attempted to answer this question at the neuronal population level - with varying degrees of success - but few, if any, have studied the change in activity of the apical dendrites of layer 5 cortical neurons. Neurons in the layer 5 of the sensory cortex appear to play a key role in sensorimotor transformations, showing important decision and reward-related signals, and being the main source of cortical and subcortical projections from the cortex. In particular, pyramidal track (PT) neurons project directly to subcortical regions related to motor activity, such as the striatum and brainstem, and could initiate rapid motor action in response to given sensory inputs. Additionally, layer 5 cortical neurons have large apical dendrites that extend to layer 1 where different neuromodulatory and long-range inputs converge, providing motor and contextual information that could be used to modulate layer 5 neurons output and/or to establish the synaptic plasticity required for learning a new association.

In this study, the authors aimed to test whether the learning of a new sensorimotor transformation could be supported by a change in the evoked response of the apical dendrites of layer 5 neurons in the mouse whisker primary somatosensory cortex. To do this, they performed longitudinal functional calcium imaging of the apical dendrites of layer 5 neurons while mice learned to discriminate between two multiwhiskers stimuli. The authors used a simple conditioning task in which one whisker stimulus (upward or backward air puff, CS+) is associated with reward after a short delay, while the other whisker stimulus (CS-) is not. They found that task learning (measured by the probability of anticipatory licking just after the CS+) was not associated with a significant change of the average population response evoked by the CS+ or the CS-, nor change in the average population selectivity. However, when considering individual dendritic tufts, they found interesting changes in selectivity, with approximately equal numbers of dendrites becoming more selective for CS+ and dendrites becoming more selective for CS-.

One of the major challenges when assessing changes in neural representation during the learning of such Go/NoGo tasks is that the movements and rewards themselves may elicit strong neural responses that may be a confounding factor, that is, inexperienced mice do not lick in response to the CS+, while trained mice do. In this study, the authors addressed this issue in three ways: first, they carefully monitor the orofacial movements of mice and show that task learning is not associated with changes in evoked whisker movements. Second, they show that whisking or licking evokes very little activity in the dendritic tufts compared to whisker stimuli (CS+ and CS-). Finally, the authors introduced into the design of their task a post-conditioning session after the last conditioning session during which the CS+ and the CS- are presented but no reward is delivered. During this post-session, the mice gradually stopped licking in response to the CS+. A better design might have been to perform the pre-conditioning and post-conditioning sessions in non-water-restricted, unmotivated mice to completely exclude any lick response, but the fact that the change in selectivity persists after the mice stopped licking in the last blocks of the post-conditioning session (in mice relying only on their whiskers to perform the task) is convincing.

The clever task design and careful data analysis provide compelling evidence that learning this whisker discrimination task does not result in a massive change in sensory representation in the apical dendritic tufts of layer 5 neurons in the primary somatosensory cortex on average. Nevertheless, individual dendritic tufts do increase their selectivity for one or the other sensory stimulus, likely enhancing the ability of S1 neurons to accurately discriminate the two stimuli and trigger the appropriate motor response (to lick or not to lick).

One limitation of the present study is the lack of evidence for the necessity of the primary somatosensory cortex in the learning and execution of the task. As the authors have strongly emphasized in their previous publications, the primary somatosensory cortex may not be necessary for the learning and execution of simple whisker detection tasks, especially when the stimulus is very salient. Although this new task requires the discrimination between two whisker stimuli, the simplicity and salience of the whisker stimuli used could make this task cortex independent. Especially when considering that some mice seem to not rely entirely on their whiskers to execute the task.

Nevertheless, this is an important result that shows for the first-time changes in the selectivity to sensory stimuli at the level of individual apical dendritic tufts in correlation with the learning of a discrimination task. This study sheds new light on the cortical cellular substrates of reward-based learning, and opens interesting perspectives for future research in this area. In future studies, it will be important to determine whether the change in selectivity of dendritic calcium spikes is causally involved in the learning the task or whether it simply correlates with learning, as a consequence of changes in synaptic inputs caused by reward. The dendritic calcium spikes may be involved in the establishment of synaptic plasticity required for learning and impact the output of layer 5 pyramidal neurons to trigger the appropriate motor response. It would be important also to study the changes in selectivity in the apical dendrite of the identified projection neurons.

Comments on revisions:

The authors have addressed all my questions. I have no further recommendations.

Reviewer #2 (Public review):

Summary:

The authors did not find an increased representation of CS+ throughout reinforcement learning in the tuft dendrites of Rbp4-positive neurons from layer 5B of the barrel cortex, as previously reported for soma from layer 2/3 of the visual cortex.<br /> Alternatively, the authors observed an increased selectivity to both stimuli (CS+ and CS-) during reinforcement learning. This feature 1) was not present in repeated exposures (without reinforcement), 2) was not explained by animal's behaviour (choice, licking and whisking) and 3) was long-lasting, being present even when the mice disengaged from the task.<br /> Importantly, increased selectivity was correlated with learning (% correct choices), and neural discriminability between stimuli increased with learning.

In conclusion, the authors show that tuft dendrites from layer 5B of the barrel cortex increase the representation of conditioned (CS+) and unconditioned stimuli (CS-) applied to the whiskers, during reinforcement learning.

Strengths:<br /> The results presented are very consistent throughout the entire study, and therefore very convincing:

(1) The results observed are very similar using two different imaging techniques (using 2-photon -planar imaging- and SCAPE - volumetric imaging). Fig. 3 and Fig.4 respectively.<br /> (2) The results are similar using "different groups" of tuft dendrites for the analysis (e.g. initially unresponsive and responsive pre- and post-learning). Fig. 5.<br /> (3) The results are similar from a specific set of trials (with the same sensory input, but different choices). Fig.7.<br /> (4) Additionally, the selectivity of tuft dendrites from layer 5B of the barrel cortex was higher in the mice that exclusively used the whisker to respond to the stimuli (CS+ and CS-).

The results presented are controlled against a group of mice that received the same stimuli presentation, except the reinforcement (reward).

Additionally, the behaviour outputs, such as choice, whisking and licking could not account for the results observed.

Although there are no causal experiments, the correlation between selectivity and learning (% of correct choices), as well as the increased neural discriminability with learning, but not in repeated exposure, are very convincing.

Weaknesses:

The biggest weakness is the absence of causality experiments. Although inhibiting specifically tuft dendritic activity in layer 1 from layer 5 pyramidal neurons is very challenging, tuft dendritic activity in layer 1 could be silenced through optogenetic experiments as in Abs et al. 2018. By manipulating NDNF-positive neurons the authors could specifically modify tuft dendritic activity in the barrel cortex during CS presentations, and test if silencing tuft dendritic activity in layer 1 would lead to the lack of selectivity and an impairment of reinforcement learning. Additionally, this experiment will test if the selectivity observed during reinforcement learning is due to changes in the local network, namely changes in local synaptic connectivity, or solely due to changes in the long-range inputs.

Author response:

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

Reviewer #1 (Public Review):

What neurophysiological changes support the learning of new sensorimotor transformations is a key question in neuroscience. Many studies have attempted to answer this question at the neuronal population level - with varying degrees of success - but few, if any, have studied the change in activity of the apical dendrites of layer 5 cortical neurons. Neurons in layer 5 of the sensory cortex appear to play a key role in sensorimotor transformations, showing important decision and reward-related signals, and being the main source of cortical and subcortical projections from the cortex. In particular, pyramidal track (PT) neurons project directly to subcortical regions related to motor activity, such as the striatum and brainstem, and could initiate rapid motor action in response to given sensory inputs. Additionally, layer 5 cortical neurons have large apical dendrites that extend to layer 1 where different neuromodulatory and long-range inputs converge, providing motor and contextual information that could be used to modulate layer 5 neurons output and/or to establish the synaptic plasticity required for learning a new association. 

In this study, the authors aimed to test whether the learning of a new sensorimotor transformation could be supported by a change in the evoked response of the apical dendrites of layer 5 neurons in the mouse whisker primary somatosensory cortex. To do this, they performed longitudinal functional calcium imaging of the apical dendrites of layer 5 neurons while mice learned to discriminate between two multi-whisker stimuli. The authors used a simple conditioning task in which one whisker stimulus (upward or backward air pu , CS+) is associated with a reward after a short delay, while the other whisker stimulus (CS-) is not. They found that task learning (measured by the probability of anticipatory licking just after the CS+) was not associated with a significant change in the average population response evoked by the CS+ or the CS-, nor a change in the average population selectivity. However, when considering individual dendritic tufts, they found interesting changes in selectivity, with approximately equal numbers of dendrites becoming more selective for CS+ and dendrites becoming more selective for CS-. 

One of the major challenges when assessing changes in neural representation during the learning of such Go/NoGo tasks is that the movements and rewards themselves may elicit strong neural responses that may be a confounding factor, that is, inexperienced mice do not lick in response to the CS+, while trained mice do. In this study, the authors addressed this issue in three ways: first, they carefully monitored the orofacial movements of mice and showed that task learning is not associated with changes in evoked whisker movements. Second, they show that whisking or licking evokes very little activity in the dendritic tufts compared to whisker stimuli (CS+ and CS-). Finally, the authors introduced into the design of their task a post-conditioning session after the last conditioning session during which the CS+ and the CS- are presented but no reward is delivered. During this post-session, the mice gradually stopped licking in response to the CS+. A better design might have been to perform the pre-conditioning and post-conditioning sessions in nonwater-restricted, unmotivated mice to completely exclude any lick response, but the fact that the change in selectivity persists after the mice stopped licking in the last blocks of the post-conditioning session (in mice relying only on their whiskers to perform the task) is convincing. 

The clever task design and careful data analysis provide compelling evidence that learning this whisker discrimination task does not result in a massive change in sensory representation in the apical dendritic tufts of layer 5 neurons in the primary somatosensory cortex on average. Nevertheless, individual dendritic tufts do increase their selectivity for one or the other sensory stimulus, likely enhancing the ability of S1 neurons to accurately discriminate the two stimuli and trigger the appropriate motor response (to lick or not to lick). 

One limitation of the present study is the lack of evidence for the necessity of the primary somatosensory cortex in the learning and execution of the task. As the authors have strongly emphasized in their previous publications, the primary somatosensory cortex may not be necessary for the learning and execution of simple whisker detection tasks, especially when the stimulus is very salient. Although this new task requires the discrimination between two whisker stimuli, the simplicity and salience of the whisker stimuli used could make this task cortex-independent. Especially when considering that some mice seem to not rely entirely on their whiskers to execute the task. 

Nevertheless, this is an important result that shows for the first time changes in the selectivity to sensory stimuli at the level of individual apical dendritic tufts in correlation with the learning of a discrimination task. This study sheds new light on the cortical cellular substrates of reward-based learning and opens interesting perspectives for future research in this area. In future studies, it will be important to determine whether the change in selectivity of dendritic calcium spikes is causally involved in the learning of the task or whether it simply correlates with learning, as a consequence of changes in synaptic inputs caused by reward. The dendritic calcium spikes may be involved in the establishment of synaptic plasticity required for learning and impact the output of layer 5 pyramidal neurons to trigger the appropriate motor response. It would be important also to study the changes in selectivity in the apical dendrite of the identified projection neurons.  

Reviewer #2 (Public Review):

Summary: 

The authors did not find an increased representation of CS+ throughout reinforcement learning in the tuft dendrites of Rbp4-positive neurons from layer 5B of the barrel cortex, as previously reported for soma from layer 2/3 of the visual cortex. 

Alternatively, the authors observed an increased selectivity to both stimuli (CS+ and CS-) during reinforcement learning. This feature: 

(1) was not present in repeated exposures (without reinforcement), 

(2) was not explained by the animal's behaviour (choice, licking, and whisking), and 

(3) was long-lasting, being present even when the mice disengaged from the task. 

Importantly, increased selectivity was correlated with learning (% correct choices), and neural discriminability between stimuli increased with learning. 

In conclusion, the authors show that tuft dendrites from layer 5B of the barrel cortex increase the representation of conditioned (CS+) and unconditioned stimuli (CS-) applied to the whiskers, during reinforcement learning. 

Strengths: 

The results presented are very consistent throughout the entire study, and therefore very convincing: 

(1) The results observed are very similar using two different imaging techniques (2-photon planar imaging- and SCAPE-volumetric imaging). Figure 3 and Figure 4 respectively. 

(2) The results are similar using "different groups" of tuft dendrites for the analysis (e.g.

initially unresponsive and responsive pre- and post-learning). Figure 5. 

(3) The results are similar from a specific set of trials (with the same sensory input, but di erent choices). Figure 7. 

(4) Additionally, the selectivity of tuft dendrites from layer 5B of the barrel cortex was higher in the mice that exclusively used the whisker to respond to the stimuli (CS+ and CS-).  The results presented are controlled against a group of mice that received the same stimuli presentation, except for the reinforcement (reward). 

Additionally, the behaviour outputs, such as choice, whisking, and licking could not account for the results observed. 

Although there are no causal experiments, the correlation between selectivity and learning (percentage of correct choices), as well as the increased neural discriminability with learning, but not in repeated exposure, are very convincing. 

Weaknesses: 

The biggest weakness is the absence of causality experiments. Although inhibiting specifically tuft dendritic activity in layer 1 from layer 5 pyramidal neurons is very challenging, tuft dendritic activity in layer 1 could be silenced through optogenetic experiments as in Abs et al. 2018. By manipulating NDNF-positive neurons the authors could specifically modify tuft dendritic activity in the barrel cortex during CS presentations, and test if silencing tuft dendritic activity in layer 1 would lead to the lack of selectivity and an impairment of reinforcement learning. Additionally, this experiment will test if the selectivity observed during reinforcement learning is due to changes in the local network, namely changes in local synaptic connectivity, or solely due to changes in the long-range inputs.    

We agree that such causal manipulations are a logical next step. Such manipulations are unfortunately not specific to layer 5 apicals, so the results would be difficult to interpret. We now discuss the challenge of such manipulations in the Discussion section.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors): 

Overall, the study is solid and the article is well and clearly written. I have no suggestion for other experiments that would fall within the scope of this article. I would like only to suggest some additional analyses and clarifications in the writing. 

Additional analyses: 

Obviously, the main confounding factor in this type of data comes from the acquired motor response which follows - with a short latency - the sensory stimulus. This is particularly problematic for functional calcium imaging which has very low temporal resolution. The authors have addressed this question to some extent by showing that motor-evoked activity does not account for the change in selectivity acquired with learning and through the use of a post-conditioning session during which no reward was delivered. Figures 8C-D show that mice gradually stop licking in response to CS+ in this session and that the distribution of the selectivity index remains similar in these last blocks. Perhaps a more convincing analysis would be to simply select Miss and Correct rejection trials in which mice did not lick in response to the CS+ and CS-, respectively. Ideally, if the number of trials is sufficient, one could even select trials devoid of any evoked movement (no licking and no whisking).  

We agree it would be interesting to compare Miss and Correct rejection trials to further rule out effects of a motor response, but there were never enough Miss trials to conduct such an analysis. Even in very early learning, there are few Miss trials (see Figure 1, session 2). We found that in early learning, animals would lick in most trials. Then, over the course of conditioning, they would learn to withhold licks during CS- presentation. Thus, we were able to examine Hits, Correct rejections, and False alarms (Figure 7), but not Miss trials. We have added text suggesting a future experiment in which the stimulus strengths are substantially reduced to drastically increase the error rates.

The fact that changes in selectivity occur in both directions overall is really interesting. However, in the way the data are presented currently, one may wonder about mice/field of view vs single cell effect. i.e., do di erent dendritic tufts in the same field of view show opposite changes in selectivity? If we were to replot Figure 3A for a single mouse, would we obtain the same picture?  

We appreciate this very good suggestion and have added scatter plots and selectivity index histograms for individual conditioned animals in Supplementary figure 2. These data demonstrate that different dendritic tufts in the same field of view exhibit opposite changes in selectivity.

The authors point out that they observed no change in the mean response or selectivity during learning, but did find changes in selectivity at the level of individual dendritic tufts. This suggests that, at the population level, the ability to discriminate between the two stimuli should improve. A possible complementary analysis would be to show that the ability to decode stimulus identity from dendritic tuft population activity increases with learning.  

Given the substantial change in individual tuft selectivity and that the tuft events occur are not rare, the population result is guaranteed. If individual tufts increase selectivity, the population will also increase its selectivity on a trial-by-trial basis. We have nevertheless included a new supplementary figure with a population analysis using SVMs to demonstrate this.

Clarification: 

The authors should make it clear from the beginning that mice are still water-restricted during the post-conditioning session and actually do keep licking for many CS+ trials. Therefore, this session is not devoid of motor response. 

We have clarified this in the text.

Did mice in the repeated exposure condition receive any reward during the recording sessions? If so when were rewards delivered? 

We previously described in the Methods that these mice received water in their home cage, but we now additionally clarify this in the Results section.

Minor: 

Figure 2Aii, the labels of the Alpha and Betta barrels should be swapped. 

Fixed

Line 218: I believe this sentence should read "Using SCAPE microscopy, ...". 

Corrected.

Line 665: 'Reconstruction from 50' does that refer to the single cell reconstruction on the left panel? 

Yes – Clarified in legend

Reviewer #2 (Recommendations For The Authors): 

Minor suggestions: 

The 'summary' should mention from which brain area the results were acquired. Otherwise, it is misleading, giving the idea that the results described a generic feature, which is still unknown.  

Added to the text.

Please correct sentence 219: "SCAPE microscopy, we image tuft activity of additional mice..." 

Added to the text.

In the same sentence (219) it would be good to provide the number of additional mice imaged (2). 

Added to the text.

Regarding Supplementary Figure 1, it would be interesting to correlate the second peak after reward and learning rate, to provide further support to the sentences 109 to 113. 

We agree this would be interesting to examine, but only four animals exhibited this second peak, which is too small of a sample to observe a meaningful correlation. We now clarify this in the text.

In Figure 3, why not present the correlation between 'neural discriminability' and % of correct choices? 

We appreciate the suggestion and have added this plot to Figure 3.

The 'results' section will benefit tremendously if the authors consistently indicate the figures to which the results are being described, or 'data not shown' if it is the case. To give a few examples: 

Sentence 108 - "averaged 28% ΔF/F" - From which figure is this result coming from?  Sentence 123 - "(p = 0.62, 0.64, respectively)" - comparison not shown, but see Figures 2E and D respectively? 

Sentence 125 - "(CS+ responsive (...) across all sessions)" - From which figure is this result coming from? 

Sentence 130 - "during pre-conditioning (p=0.66) or post-conditioning sessions (p=0.44) - From which figure? 

Sentence 154 - "(Pre: p=0.20; last rewarded: p=0.43; Post: p=0.64, sign-rank test)" - From which figure? 

Sentence 175 - "(-0.049, -0.001, and 0.003" - From which figure? Please show the graph that shows that the mean SI is not different. It can be supplementary. The distribution of SI will be strengthened by it.  

We added this plot to supplementary figure 2.

Sentence 244 - "(conditioned: 458/603; repeated exposure: 334/457) - From Figure 5E. 

Sentence 256 - "(p=0.04, 2-sample t-test comparison mice) - From Figure 5B.  Sentence 258 - "(p=0.03, paired t-test) - from Figure 5B  Sentences 370 to 378 - No reference to the figure. 

The 'discussion' section (sentences 459 to 494) refers to the differences between the current and previous studies (references 1,3,5), namely soma vs. dendrites and layer 2/3 vs. layer 5. However, it should also mention the difference between the nature of the stimuli and the brain area recorded (visual cortex vs. barrel cortex).

We have addressed these issues in the text.

Author response:

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

Reviewer 1:

Authors reject the substance of Reviewer 1’s feedback primarily due to clear lack of understanding of typical parameterization practices used to avoid overfitting. To ensure the Spearman-rank correlation accuracy, 70% of all data was withheld from the optimization process and used solely for testing to yield figure 6. Data was withheld prior to model parameterization and therefore avoids Reviewer 1’s charge of “artificially forcing the correlation”. Authors did appreciate the request for clarification of additional definitions and minor reorganization suggestions. Below we provide specific responses to each numbered point (note: multiple responses are provided for some of the reviewer points).

Point 1: Clarify Metrics Definition and Evaluation

Authors clarified the description of biodiversity metrics. The metrics associated with manual methods are detailed in the third paragraph of the Materials and Methods: Data Analysis section, while the sensor-based metric is described in the second paragraph, and summarized in its last sentence.

Text Additions:

Authors added clarification to the introduction’s first paragraph defining biodiversity metrics, including species richness.

Authors added detailed definitions of community metrics and their significance in community ecology in the Materials and Methods section (3rd paragraph of “Data Analysis” section). The discussion was updated to include a reference to community ecology and the benefits of big data, specifically highlighting the potential of autonomous optical sensors in entomology.

Methods Reorganization

We have reorganized the Methods section for clarity. Updated section clarifies metrics studied, location, dates, a description and methods around optical sensors, Malaise traps, and sweep netting.

Text Additions:

An overview paragraph was added to “Data analysis” (3rd paragraph) detailing key metrics used, specifying metrics such as abundance, richness, Shannon index, and Simpson index.

Visualization methods for sensor data to deliver analogous metrics of abundance, richness, and diversity indices was added to “Data analysis” section.

Supplementary Table 1 and the first paragraph of the Materials and Methods section cover location, dates, and other general information.

Detailed descriptions and methods for optical sensors, Malaise traps, and sweeping are provided.

Integration of Metrics

Authors integrated two paragraphs explaining the fundamental differences between conventional methods in the 3rd paragraph of the discussion and the presented method of biodiversity measurement.

Point 2: Body-to-Wing Ratio Calculation

The backscattered optical cross-section is now clearly defined as the value measured at the maximum point of the event. Specifically, we have added the word ‘maximum’ to our methods section for clarity.

Point 3: Ecosystem Services Paragraph

We have shortened and edited this paragraph for clarity. The revised text is now more straightforward and comprehensible.

Point 4: Results Section Structure

We believe restructuring the results section around each metric would result in redundancy. The value of our analysis is in the comparison of different methods; therefore, instead of talking about methods in isolation, we provide an integrated discussion and comparison of all three methods across all metrics. Instead, we have maintained our current structure but ensured that the metrics are consistently described and analyzed.

Point 5: Abundance Correlation

We agree that the lack of a correlation between methods for abundance remains an open question. However, we maintain that fitting a linear model would be inappropriate and potentially misleading in the absence of significant correlation. We have clarified this in our manuscript.

Point 6: Richness and Diversity Evaluations

The authors disagree with Reviewer 1's feedback, citing a clear misunderstanding of standard parameterization practices used to prevent overfitting. Specifically, authors implemented a 30/70 Training/Testing split. Therefore only 30% of the data was used to fit the model and 70% of the dataset was reserved for testing to ensure the validity and reliability of our clustering results. By validating with a 70% testing dataset, we ensure that the clustering model can accurately group new data points and is robust against overfitting. This process helps verify that the identified clusters are meaningful and consistent across different subsets of the data.  Spearman's rho converts the data values into ranks and does not assume a linear relationship between the variables or require the data to follow a normal distribution. Spearman's rank correlation offers robustness against non-linearity and outliers by focusing on ranks. This approach is explained in the 4th paragraph of the “Data Analysis” section.

Point 7: Clustering Method Credibility

Authors acknowledge the variability in optical sensor features. However, the Law of Large Numbers supports increased insect measurement accuracy and stability occurs from optical insect sensors due to the increased number of observations made by the optical sensors compared to conventional methods. The manuscript now includes a detailed discussion of these aspects in the 3rd paragraph of discussion, emphasizing the correlation observed despite variability.

Reviewer 2:

Authors appreciate Reviewer 2’s feedback especially regarding contextualization. While authors disagree with the need for more specific experimental questions in a methods paper and the suggested need for more complex analysis, we agree with the essence of the review and added additional text regarding potential questions, method applications, and ecosystem processes for contextualization.

Point 1: Larger Question Framing

We present this article as a methodological paper rather than asking a specific experimental question. This approach is justified by the generalizable nature of methods papers, akin to those describing ImageJ or mass spectrometers. The method is widely applicable to a range of scientific questions. 

We provided a discussion on how this technology could be applied in community ecology, conservation, and managed ecological systems like agriculture.

In the Conclusion section we provided elaboration on the potential research questions and applications.

Point 2: Complex Analyses

While complex analyses like NMDS are useful for specific questions, this paper aims to establish the method. Once established, this method can be applied to various research questions in future studies. Therefore, as we are not directly asking an experimental question, more complex analysis is unnecessary.

Point 3: Ecosystem Process (Granivory) Assay

We have improved the contextualization and explanation of the ecosystem process assay throughout the manuscript, ensuring it is well-integrated and clear to readers.

eLife Assessment

The authors propose a new methodology to survey insects, using new sensors and analytical capabilities that could be valuable for addressing urgent conservation challenges. While the results of the optical sensors appear to be comparable to those obtained with classical survey methodologies, current analyses are considered incomplete.

Reviewer #2 (Public review):

Summary:

The manuscript proposes a new technology to survey insects. They deployed optical sensors in agricultural landscapes and contrast their results to those in classical malaise and sweep nets survey methodologies. They found the results of optical sensors to be comparable with classical survey methodologies. The authors discuss pros and cons of their near-infrared sensor.

Strengths:

Contrasting the results with optical sensors with those in classical malaise and sweep nets was a clever idea.

Weaknesses:

The submitted materials on Revision 1 (in particular the response to reviewers) are difficult to follow. I encourage the authors to provide a point-by-point response to the first set of comments, as well as to this second review.

A new version of the manuscript needs to make sure that variability in the system (different crops) is taken into consideration. Also, stronger analysis including our current understanding of biodiversity metrics (including measures of sample coverage, sample completeness, Hill numbers, among others) will be important to make sure your new methodology is properly capable to be used as a new standard methodology.

While this new version is stronger and much clearer, I also agree with Reviewer 1 that the usage of terminology is weak. The paper and the new methodology is sound. It is is the application to real ecosystems/questions and datasets that is not properly addressed in the manuscript.

eLife Assessment

This paper explores the important question of how two major inhibitory interneuron classes in the neocortex differentially affect cortical dynamics. Using a linearized fixed point approach, they provide convincing evidence that the existence of multiple interneuron classes can explain the counterintuitive finding that inhibitory modulation can increase the gain of the excitatory cell population while also increasing the stability of the circuit's state to minor perturbations. Support for the main conclusions is solid, but could be strengthened by additional analyses.

Reviewer #1 (Public review):

Summary:

This paper explores how diverse forms of inhibition impact firing rates in models for cortical circuits. In particular, the paper studies how the network operating point affects the balance of direct inhibition from SOM inhibitory neurons to pyramidal cells, and disinhibition from SOM inhibitory input to PV inhibitory neurons. This is an important issue as these two inhibitory pathways have largely been studies in isolation. Support for the main conclusions is generally solid, but could be strengthened by additional analyses.

Strengths

The paper has improved in revision, and the new intuitive summary statements added to the end of each results section are quite helpful.

Weaknesses

The concern about whether the results hold outside of the range in which neural responses are linear remains. This is particularly true given the discontinuity observed in the stability measure. I appreciate the concern (provided in the response to the first round of reviews) that studying nonlinear networks requires a lot of work. A more limited undertaking would be to test the behavior of a spiking network at a few key points identified by your linearization approach. Such tests could use relatively simple (and perhaps imperfect) measures of gain and stability. This could substantially enhance the paper, regardless of the outcome.

Reviewer #2 (Public review):

Summary:

Bos and colleagues address the important question of how two major inhibitory interneuron classes in the neocortex differentially affect cortical dynamics. They address this question by studying Wilson-Cowan-type mathematical models. Using a linearized fixed point approach, they provide convincing evidence that the existence of multiple interneuron classes can explain the counterintuitive finding that inhibitory modulation can increase the gain of the excitatory cell population while also increasing the stability of the circuit's state to minor perturbations. This effect depends on the connection strengths within their circuit model, providing valuable guidance as to when and why it arises.

Overall, I find this study to have substantial merit. I have some suggestions on how to improve the clarity and completeness of the paper.

Strengths:

(1) The thorough investigation of how changes in the connectivity structure affect the gain-stability relationship is a major strength of this work. It provides an opportunity to understand when and why gain and stability will or will not both increase together. It also provides a nice bridge to the experimental literature, where different gain-stability relationships are reported from different studies.

(2) The simplified and abstracted mathematical model has the benefit of facilitating our understanding of this puzzling phenomenon. (I have some suggestions for how the authors could push this understanding further.) It is not easy to find the right balance between biologically-detailed models vs simple but mathematically tractable ones, and I think the authors struck an excellent balance in this study.

Weaknesses:

(1) The fixed-point analysis has potentially substantial limitations for understanding cortical computations away from the steady-state. I think the authors should have emphasized this limitation more strongly and possibly included some additional analyses to show that their conclusions extend to the chaotic dynamical regimes in which cortical circuits often live.

(2) The authors could have discussed -- even somewhat speculatively -- how VIP interneurons fit into this picture. Their absence from this modelling framework stands out as a missed opportunity.

(3) The analysis is limited to paths within this simple E, PV, SOM circuit. This misses more extended paths (like thalamocortical loops) that involve interactions between multiple brain areas. Including those paths in the expansion in Eqs. 11-14 (Fig. 1C) may be an important consideration.

Comments on revisions:

I think the authors have done a reasonable job of responding to my critiques, and the paper is in pretty good shape. (Also, thanks for correctly inferring that I meant VIP interneurons when I had written SST in my review! I have updated the public review accordingly.)

I still think this line of research would benefit substantially from considering dynamic regimes including chaotic ones. I strongly encourage the authors to consider such an extension in future work.

Reviewer #3 (Public review):

Summary:

Bos et al study a computational model of cortical circuits with excitatory (E) and two subtypes of inhibition - parvalbumin (PV) and somatostatin (SOM) expressing interneurons. They perform stability and gain analysis of simplified models with nonlinear transfer functions when SOM neurons are perturbed. Their analysis suggests that in a specific setup of connectivity, instability and gain can be untangled, such that SOM modulation leads to both increases in stability and gain, in contrast to the typical direction in neuronal networks where increased gain results in decreased stability.

Strengths:

- Analysis of the canonical circuit in response to SOM perturbations. Through numerical simulations and mathematical analysis, the authors have provided a rather comprehensive picture of how SOM modulation may affect response changes.<br /> - Shedding light on two opposing circuit motifs involved in the canonical E-PV-SOM circuitry - namely, direct inhibition (SOM -> E) vs disinhibition (SOM -> PV -> E). These two pathways can lead to opposing effects, and it is often difficult to predict which one results from modulating SOM neurons. In simplified circuits, the authors show how these two motifs can emerge and depend on parameters like connection weights.<br /> - Suggesting potentially interesting consequences for cortical computation. The authors suggest that certain regimes of connectivity may lead to untangling of stability and gain, such that increases in network gain are not compromised by decreasing stability. They also link SOM modulation in different connectivity regimes to versatile computations in visual processing in simple models.

Weaknesses

Computationally, the analysis is solid, but it's very similar to previous studies (del Molino et al, 2017). Many studies in the past few years have done the perturbation analysis of a similar circuitry with or without nonlinear transfer functions (some of them listed in the references). This study applies the same framework to SOM perturbations, which is a useful computational analysis, in view of the complexity of the high-dimensional parameter space.

Link to biology: the most interesting result of the paper with regard to biology is the suggestion of a regime in which gain and stability can be modulated in an unconventional way - however, it is difficult to link the results to biological networks:<br /> - A general weakness of the paper is a lack of direct comparison to biological parameters or experiments. How different experiments can be reconciled by the results obtained here, and what new circuit mechanisms can be revealed? In its current form, the paper reads as a general suggestion that different combinations of gain modulation and stability can be achieved in a circuit model equipped with many parameters (12 parameters). This is potentially interesting but not surprising, given the high dimensional space of possible dynamical properties. A more interesting result would have been to relate this to biology, by providing reasoning why it might be relevant to certain circuits (and not others), or to provide some predictions or postdictions, which are currently missing in the manuscript.<br /> - For instance, a nice motivation for the paper at the beginning of the Results section is the different results of SOM modulation in different experiments - especially between L23 (inhibition) and L4 (disinhibition). But no further explanation is provided for why such a difference should exist, in view of their results and the insights obtained from their suggested circuit mechanisms. How the parameters identified for the two regimes correspond to different properties of different layers?<br /> - One of the key assumptions of the model is nonlinear transfer functions for all neuron types. In terms of modelling and computational analysis, a thorough analysis of how and when this is necessary is missing (an analysis similar to what has been attempted in Figure 6 for synaptic weights, but for cellular gains). A discussion of this, along with the former analysis to know which nonlinearities would be necessary for the results, is needed, but currently missing from the study. The nonlinearity is assumed for all subtypes because it seems to be needed to obtain the results, but it's not clear how the model would behave in the presence or absence of them, and whether they are relevant to biological networks with inhibitory transfer functions.<br /> - Tuning curves are simulated for an individual orientation (same for all), not considering the heterogeneity of neuronal networks with multiple orientation selectivity (and other visual features) - making the model too simplistic.

Author response:

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

Reviewer #1 (Public Review):

Summary:

This paper explores how diverse forms of inhibition impact firing rates in models for cortical circuits. In particular, the paper studies how the network operating point affects the balance of direct inhibition from SOM inhibitory neurons to pyramidal cells, and disinhibition from SOM inhibitory input to PV inhibitory neurons. This is an important issue as these two inhibitory pathways have largely been studies in isolation. Support for the main conclusions is generally solid, but could be strengthened by additional analyses.

Strengths:

A major strength of the paper is the systematic exploration of how circuit architecture effects the impact of inhibition. This includes scans across parameter space to determine how firing rates and stability depend on effective connectivity. This is done through linearization of the circuit about an effective operating point, and then the study of how perturbations in input effect this linear approximation.

Weaknesses:

The linearization approach means that the conclusions of the paper are valid only on the linear regime of network behavior. The paper would be substantially strengthened with a test of whether the conclusions from the linearized circuit hold over a large range of network activity. Is it possible to simulate the full network and do some targeted tests of the conclusions from linearization? Those tests could be guided by the linearization to focus on specific parameter ranges of interest.

We agree with the reviewer that it would be interesting to test if our results hold in a nonlinear regime of network behaviour (i.e. the chaotic regime, see also comment 1 by reviewer 2). As mentioned above, this requires a different type of model (either rate-based or spiking model with multiple neurons instead of modelling the mean population rate dynamics) which, in our opinion, exceeds the scope of this manuscript. Furthermore, the core measures of our study, network gain, and stability require linearization. In a chaotic regime where the linearization approach is impossible, we would need to consider/define new measures to characterize network response/activity. Therefore, while certainly being an interesting question to study, the broad scope of the studying networks in a nonlinear regime is better tackled in a separate study. We now acknowledge in the discussion of our manuscript that the linearization approach is a limitation in our study and that it would be an interesting future direction to investigate chaotic dynamics.

The results illustrated in the figures are generally well described but there is very little intuition provided for them. Are there simplified examples or explanations that could be given to help the results make sense? Here are some places such intuition would be particularly helpful:

page 6, paragraph starting ”In sum ...”

Page 8, last paragraph

Page 10, paragraph starting ”In summary ...”

Page 11, sentence starting ”In sum ...”

We agree with the reviewer that we didn’t provide enough intuition to our results. We now extended the paragraphs listed by the reviewer with additional information, providing a more intuitive understanding of the results presented in the respective chapter.

Reviewer #2 (Public Review):

Summary:

Bos and colleagues address the important question of how two major inhibitory interneuron classes in the neocortex differentially affect cortical dynamics. They address this question by studying Wilson-Cowan-type mathematical models. Using a linearized fixed point approach, they provide convincing evidence that the existence of multiple interneuron classes can explain the counterintuitive finding that inhibitory modulation can increase the gain of the excitatory cell population while also increasing the stability of the circuit’s state to minor perturbations. This effect depends on the connection strengths within their circuit model, providing valuable guidance as to when and why it arises.

Overall, I find this study to have substantial merit. I have some suggestions on how to improve the clarity and completeness of the paper.

Strengths:

(1) The thorough investigation of how changes in the connectivity structure affect the gain-stability relationship is a major strength of this work. It provides an opportunity to understand when and why gain and stability will or will not both increase together. It also provides a nice bridge to the experimental literature, where different gain-stability relationships are reported from different studies.

(2) The simplified and abstracted mathematical model has the benefit of facilitating our understanding of this puzzling phenomenon. (I have some suggestions for how the authors could push this understanding further.) It is not easy to find the right balance between biologically detailed models vs simple but mathematically tractable ones, and I think the authors struck an excellent balance in this study.

Weaknesses:

(1) The fixed-point analysis has potentially substantial limitations for understanding cortical computations away from the steady-state. I think the authors should have emphasized this limitation more strongly and possibly included some additional analyses to show that their conclusions extend to the chaotic dynamical regimes in which cortical circuits often live.

We agree with the reviewer that it would be interesting to test if our results hold in a chaotic regime of network behaviour (see also comment by reviewer 1). As mentioned above, this requires a different type of model (either rate-based or spiking model with multiple neurons instead of modelling the mean population rate dynamics) which, in our opinion, exceeds the scope of this manuscript. Furthermore, the core measures of our study, network gain, and stability require linearization. In a chaotic regime where the linearization approach is impossible, we would need to consider/define new measures to characterize network response/activity. Therefore, while certainly being an interesting question to study, the broad scope of the studying networks in a nonlinear regime is better tackled in a separate study. We now acknowledge in the discussion of our manuscript that the linearization approach is a limitation in our study and that it would be an interesting future direction to investigate chaotic dynamics.

(2) The authors could have discussed – even somewhat speculatively – how SST interneurons fit into this picture. Their absence from this modelling framework stands out as a missed opportunity.

We believe that the reviewer wanted us to speculate about VIP interneurons (and not SST interneurons, which we already do extensively in the manuscript). Previous models have included VIP neurons in the circuit (e.g. del Molino et al., 2017; Palmigiano et al., 2023; Waitzmann et al., 2024). While we do not model VIP cells explicitly, we implicitly assume that a possible source of modulation of SOM neurons comes from VIP cells. We have now added a short discussion on VIP cells in the last paragraph in our discussion section.

(3) The analysis is limited to paths within this simple E,PV,SOM circuit. This misses more extended paths (like thalamocortical loops) that involve interactions between multiple brain areas. Including those paths in the expansion in Eqs. 11-14 (Fig. 1C) may be an important consideration.

We agree with the reviewer that our framework can be extended to study many other different paths, like thalamocortical loops, cortical layer-specific connectivity motifs, or circuits with VIP or L1 inhibitory neurons. Studying these questions, however, are beyond the scope of our work. In our discussion, we now mention the possibility of using our framework to study those questions.

Reviewer #3 (Public Review):

Summary:

Bos et al study a computational model of cortical circuits with excitatory (E) and two subtypes of inhibition parvalbumin (PV) and somatostatin (SOM) expressing interneurons. They perform stability and gain analysis of simplified models with nonlinear transfer functions when SOM neurons are perturbed. Their analysis suggests that in a specific setup of connectivity, instability and gain can be untangled, such that SOM modulation leads to both increases in stability and gain. This is in contrast with the typical direction in neuronal networks where increased gain results in decreased stability.

Strengths:

- Analysis of the canonical circuit in response to SOM perturbations. Through numerical simulations and mathematical analysis, the authors have provided a rather comprehensive picture of how SOM modulation may affect response changes.

- Shedding light on two opposing circuit motifs involved in the canonical E-PV-SOM circuitry - namely, direct inhibition (SOM → E) vs disinhibition (SOM → PV → E). These two pathways can lead to opposing effects, and it is often difficult to predict which one results from modulating SOM neurons. In simplified circuits, the authors show how these two motifs can emerge and depend on parameters like connection weights.

- Suggesting potentially interesting consequences for cortical computation. The authors suggest that certain regimes of connectivity may lead to untangling of stability and gain, such that increases in network gain are not compromised by decreasing stability. They also link SOM modulation in different connectivity regimes to versatile computations in visual processing in simple models.

Weaknesses:

The computational analysis is not novel per se, and the link to biology is not direct/clear.

Computationally, the analysis is solid, but it’s very similar to previous studies (del Molino et al, 2017). Many studies in the past few years have done the perturbation analysis of a similar circuitry with or without nonlinear transfer functions (some of them listed in the references). This study applies the same framework to SOM perturbations, which is a useful and interesting computational exercise, in view of the complexity of the high-dimensional parameter space. But the mathematical framework is not novel per se, undermining the claim of providing a new framework (or ”circuit theory”).

In the introduction we acknowledge that our analysis method is not novel but is rather based on previous studies (del Molino et al., 2017; Kuchibhotla et al., 2017; Kumar et al., 2023, Litwin-Kumar et al., 2016; Mahrach et al., 2020; Palmigiano et al., 2023; Veit et al., 2023; Waitzmann et al., 2024). We now rewrote parts of the introduction to make sure that it does not sound like the computational analysis has been developed by us, but that we rather use those previously developed frameworks to dissect stability and gain via SOM modulation.

Link to biology: the most interesting result of the paper with regard to biology is the suggestion of a regime in which gain and stability can be modulated in an unconventional way - however, it is difficult to link the results to biological networks: - A general weakness of the paper is a lack of direct comparison to biological parameters or experiments. How different experiments can be reconciled by the results obtained here, and what new circuit mechanisms can be revealed? In its current form, the paper reads as a general suggestion that different combinations of gain modulation and stability can be achieved in a circuit model equipped with many parameters (12 parameters). This is potentially interesting but not surprising, given the high dimensional space of possible dynamical properties. A more interesting result would have been to relate this to biology, by providing reasoning why it might be relevant to certain circuits (and not others), or to provide some predictions or postdictions, which are currently missing in the manuscript.

- For instance, a nice motivation for the paper at the beginning of the Results section is the different results of SOM modulation in different experiments - especially between L23 (inhibition) and L4 (disinhibition). But no further explanation is provided for why such a difference should exist, in view of their results and the insights obtained from their suggested circuit mechanisms. How the parameters identified for the two regimes correspond to different properties of different layers?

As pointed out by the reviewer, the main goal of our manuscript is to provide a general understanding of how gain and stability depend on different circuit motifs (ie different connectivity parameters), and how circuit modulations via SOM neurons affect those measures. However, we agree with the reviewer that it would be useful to provide some concrete predictions or postdictions following from our study.

An interesting example of a postdiction of our model is that the firing rate change of excitatory neurons in response to a change in the stimulus (which we define as network gain, Eq. 2) depends on firing rates of the excitatory, PV, and SOM neurons at the moment of stimulus presentation (Fig. 3ii; Fig. 4Aii,Bii,Cii; Fig. 5Aii, Bii, Cii). Hence any change in input to the circuit can affect the response gain to a stimulus presentation, in line with experimental evidence which suggests that changes in inhibitory firing rates and changes in the behavioral state of the animal lead to gain modifications (Ferguson and Cardin 2020).

Another recent concrete example is the study of Tobin et al., 2023, in which the authors show that optogenetically activating SOM cells in the mouse primary auditory cortex (A1) decreases the excitatory responses to auditory stimuli. In our framework, this corresponds to the case of decreases in network gain (gE) for positive SOM modulation, as seen in the circuit with PV to SOM feedback connectivity (Suppl. Fig. S1).

Another example is the study by Phillips and Hasenstaub 2016, in which the authors study the effect of optogenetic perturbations of SOM (and PV) cells on tuning curves of pyramidal cells in mouse A1. While they find large heterogeneity in additive/subtractive or multiplicative/divisive tuning curve changes following SOM inactivation, most cells have a purely multiplicative or purely additive component (and none of the cells have a divisive component). In our study, we see that large multiplicative responses of the excitatory population follow from circuits with strong E to SOM feedback connectivity.

We note that in future computational studies, it would be useful to apply our framework with a focus on a specific brain region and add all relevant cell types (at a minimum E, PV, SOM, and VIP) plus a dendritic compartment, in order to formulate much more precise experimental predictions.

We have now added additional information to the discussion section.

- Another caveat is the range of parameters needed to obtain the unintuitive untangling as a result of SOM modulation. From Figure 4, it appears that the ”interesting” regime (with increases in both gain and stability) is only feasible for a very narrow range of SOM firing rates (before 3 Hz). This can be a problem for the computational models if the sweet spot is a very narrow region (this analysis is by the way missing, so making it difficult to know how robust the result is in terms of parameter regions). In terms of biology, it is difficult to reconcile this with the realistic firing rates in the cortex: in the mouse cortex, for instance, we know that SOM neurons can be quite active (comparable to E neurons), especially in response to stimuli. It is therefore not clear if we should expect this mechanism to be a relevant one for cortical activity regimes.

We agree with the reviewer that it’s important to test the robustness of our results. As suggested by the reviewer, we now include a new supplementary figure (Suppl. Fig. S2) which measures the percentage of data points in the respective quadrant Q1-Q4 when changing the SOM firing rates (as done in Fig. 5). We see that the quadrants in which the network gain and stability change in the same direction (Q2 and Q3) remain high in the case for E to SOM feedback (Suppl. Fig. S2A) over SOM rates ranging over 0-10 Hz (and likely beyond).

- One of the key assumptions of the model is nonlinear transfer functions for all neuron types. In terms of modelling and computational analysis, a thorough analysis of how and when this is necessary is missing (an analysis similar to what has been attempted at in Figure 6 for synaptic weights, but for cellular gains). In terms of biology, the nonlinear transfer function has experimentally been reported for excitatory neurons, so it’s not clear to what extent this may hold for different inhibitory subtypes. A discussion of this, along with the former analysis to know which nonlinearities would be necessary for the results, is needed, but currently missing from the study. The nonlinearity is assumed for all subtypes because it seems to be needed to obtain the results, but it’s not clear how the model would behave in the presence or absence of them, and whether they are relevant to biological networks with inhibitory transfer functions.

It is true that the nonlinear transfer function is a key component in our model. We chose identical transfer functions for E, PV, and SOM (; Eq. 4) to simplify our analysis. If the transfer function of one of the neuron types would be linear (β \= 1), then the corresponding b terms (the slope of the nonlinearity at the steady state; b \= dfX/dqX; Fig. 1B; Eq. 4) would be equal to α. Therefore, if neurons had a linear transfer function in our model, there would not be a dependence of network gain on E and PV firing rate as studied in Fig. 3-5. This is because the relationship between PV rates and their gain would be constant (bP \= α) in Fig. 1B (bottom).

If all the transfer functions were linear, changes in firing rates would not have an impact on network gain or stability. Changing the nonlinear transfer function by changing the α or β terms in Eq. 4 would only scale the way a change in the rates affects the b terms and hence the results presented in Fig. 3-5. More interesting would be to study how different types of nonlinearities, like sigmoidal functions or sublinear nonlinearities (i.e. saturating nonlinearities), would change our results. However, we think that such an investigation is out of scope for this study. We now added a comment to the Methods section.

Experimentally, F-I curves have been measured also for PV and SOM neurons. For example, Romero-Sosa et al., 2021 measure the F-I curve of pyramidal, PV and SOM neurons in mouse cortical slices. They find that similar to pyramidal neurons, PV and SOM neurons show a nonlinear F-I curve. We now added the citation of Romero-Sosa et al., 2021 to our manuscript.

- Tuning curves are simulated for an individual orientation (same for all), not considering the heterogeneity of neuronal networks with multiple orientation selectivity (and other visual features) - making the model too simplistic.

The reviewer is correct that we only study changes in tuning curves in a simplistic model. In our model, the excitatory and PV populations are tuned to a single orientation (in the case of Fig. 7 to θ \= 90). While this is certainly an oversimplification, it allows us to understand how additive/subtractive and multiplicative/divisive changes in the tuning curves come about in networks with different connectivity motifs. To model heterogeneity of tuning responses within a network, it requires more complex models. A natural choice would be to extend a classical ring attractor model (Rubin et al., 2015) by splitting the inhibitory population into PV and SOM neurons, or study the tuning curve heterogeneity that occurs in balanced networks (Hansel and van Vreeswijk 2012). However, this model has many more parameters, like the spatial connectivity profiles from and onto PV and SOM neurons. While highly valuable, we believe that studying such models exceeds the scope of our current manuscript. We now added a paragraph in the discussion section, mentioning this as an interesting future direction.

Reviewer #1 (Recommendations For The Authors):

The last sentence of the abstract is hard to interpret before reading the rest of the paper - suggest replacing or rephrasing.

We rephrased the sentence to make more clear what we mean.

Page 3, last full paragraph: I think this assumes that phi is positive. What is the justification for that assumption? More generally, I think you could say a bit more about phi in the main text since it is a fairly complicated term.

The reviewer is correct, for a stable system phi is always positive. We now clarify this and explain phi in more detail in the main text.

Fig 1D: It would be helpful to identify when the stimulus comes on and be clearer about what the stimulus is. I assume it’s a step increase in S input at 0.05 s or so - but that should be immediately apparent looking at the figure.

We agree with the reviewer and we added a dashed line at the time of stimulus onset in Fig. 1D.

Page 5: ”To motivate our analysis we compare ... (Fig. 2A)” - Figure 2A does not show responses without modulation, so this sentence is confusing.

The dashed lines in Fig. 2A (and Fig. 2C) actually represents the rate change without modulation.

Page 6: sentence “The central goal of our study ...” seems out of place since this is pretty far into the results, and that goal should already be clear.

We agree with the reviewer, hence we updated the sentence.

Page 10, top: the green curve in panel Aii always has a negative slope - so I am confused by the statement that increasing wSE decreases both gain and stability.

We thank the reviewer for pointing out this mistake. We now fixed it in the text.

Figure 6: in general it is hard to see what is going on in this figure (the green and blue in particular are hard to distinguish). Some additional labels would be helpful, but I would also see if the color scheme can be improved.

We added a zoom-in to the panels which were hard to distinguish.

Reviewer #2 (Recommendations For The Authors):

Major recommendations:

(1) The authors should explain early on in the results section what the key factor(s) is that differentiates SOM from PV cells in their model. E.g., in Fig. 1A, the only obvious difference is that SOM cells don’t inhibit themselves. However, later on in the paper, the difference in external stimulus drive to these interneuron classes is more heavily emphasized. Given the importance of that difference (in external stim drive), I think this should be highlighted early on.

We now mention the key factors that differentiate PV and SOM neurons already when describing Fig. 1A.

(2) The result in Figs. 5,6 demonstrate that recurrent SOM connectivity is important for achieving increases in both gain and stability. This observation could benefit from some intuitive explanation. Perhaps the authors could find this explanation by looking at their series expansion (Eqs. 11-14, Fig. 1C) and determining which term(s) are most important for this effect. The corresponding paths through the circuit – the most important ones – could then be highlighted for the reader.

We agree with the reviewer that our results benefit from more intuitive explanations. This has also been pointed out by reviewer 1 in their public review. We now extended the concluding paragraphs in the context of Fig. 4-6 with additional information, providing a more intuitive understanding of the results presented in the respective chapter. While it is possible to gain an intuitive understanding of how the network gain depends on rate and weight parameters (Eq. 2), this understanding is unfortunately missing in the case of stability. The maximum eigenvalue of the system have a complex relationship with all the parameters, and often have nonlinear dependencies on changes of a parameter (e.g. as we show in Fig. 3iv or one can see in Fig. 6). We now discuss this difficulty at the end of the section “Influence of weight strength on network gain vs stability”.

(3) I think the authors should consider including some analyses that do not rely on the system being at or near a fixed point. I admit that such analysis could be difficult, and this could of course be done in a future study. Nevertheless, I want to reiterate that this addition could add a lot of value to this body of work.

As outlined above, we decided to not include additional analysis on network behaviour in nonlinear regimes but we now acknowledge in the discussion of our manuscript that the linearization approach is a limitation in our study and that it would be an interesting future direction to investigate chaotic dynamics.

Minor recommendations:

(1) At the top of P. 6, when the authors first discuss the stability criterion involving eigenvalues, they should address the question ”eigenvalues of what?”. I suggest introducing the idea of the Jacobian matrix, and explaining that the largest eigenvalue of that matrix determines how rapidly the system will return to the fixed point after a small perturbation.

We included an additional sentence in the respective paragraph explaining the link between stability and negative eigenvalues, and we also added a sentence in the Methods section stating the the largest real eigenvalue dominates the behavior of the dynamical system.

(2) The panel labelling in Fig. 3 is unnecessarily confusing. It would be simpler (and thus better) to simply label the panels A,B,C,D, or i,ii,iii,iv, instead of the current labelling: Ai, Aii, Aiii, Aiv. (There are currently no panels ”B” in Fig. 3).

We updated the figure accordingly.

Reviewer #3 (Recommendations For The Authors):

• Suggestions for improved or additional experiments, data or analyses.

Analysis of the effect of different nonlinear transfer functions is necessary.

Please see our detailed answer to the reviewer’s comment in the public review above.

Analysis of gain modulation in models with more realistic tuning properties.

Please see our detailed answer to the reviewer’s comment in the public review above.

Mathematical analysis of the conditions to obtain ”untangled” gain and stability:

One of the promises of the paper is that it is offering a computational framework or circuit theory for understanding the effect of SOM perturbation. However, the main result, namely the untangling of gain and stability, has only been reported in numerical simulations (e.g. Fig. 6). Different parameters have been changed and the results of simulations have been reported for different conditions. Given the simplified model, which allows for rigorous mathematical analysis, isn’t it possible to treat this phenomenon more analytically? What would be the conditions for the emergence of the untangled regime? This is currently missing from the analyses and results.

We agree with the reviewer that our results benefit from more intuitive explanations. This has also been pointed out by reviewer 1 in their public review. We now extended the concluding paragraphs in the context of Fig. 4-6 with additional information, providing a more intuitive understanding of the results presented in the respective chapter. While it is possible understand analytically of how the network gain depends on rate and weight parameters (Eq. 2), this understanding is unfortunately missing in the case of stability. The maximum eigenvalue of the system have a complex relationship with all the parameters, and often have nonlinear dependencies on changes of a parameter (e.g. as we show in Fig. 3iv or one can see in Fig. 6). This doesn’t allow for a a deep analytical understanding of the entangling of gain and stability. We now discuss this difficulty at the end of the section “Influence of weight strength on network gain vs stability”.

• Recommendations for improving the writing and presentation. The Results section is well written overall, but other parts, especially the Introduction and Discussion, would benefit from proof reading - there are many typos and problems with sentence structures and wording (some mentioned below).

We have gone through the manuscript again and improved the writing.

The presentation of the dependence on weight in Figure 6 can be improved. For instance, the authors talk about the optimal range of PV connectivity, but this is difficult to appreciate in the current illustration and with the current colour scheme.

We added a zoom in to the panels which were hard to distinguish.

• Minor corrections to the text and figures. Text:

We thank the reviewer for their thorough reading of our manuscript. We fixed all the issues from below in the manuscript.

Some examples of bad structure or wording:

From the Abstract:

”We show when E - PV networks recurrently connect with SOM neurons then an SOM mediated modulation that leads to increased neuronal gain can also yield increased network stability.” From Introduction:

Sentence starting with ”This new circuit reality ...”

”Inhibition is been long identified as a physiological or circuit basis for how cortical activity changes depending upon processing or cognitive needs ...”

Sentence starting with ”Cortical models with both ...”

”... allowing SOM neurons the freedom to ..”

From Results:

”... affects of SOM neurons on E ..”

”seem in opposition to one another, with SOM neuron activity providing either a source or a relief of E neuron suppression”. The sentence after is also difficult to read and needs to be simplified.

P. 7: ”We first remark that ...”

Difficult to read/understand - long and badly structured sentence.

P. 8: ”adding a recurrent connection onto SOM neurons from the E-PV subcircuit” It’s from E (and not PV) to be more precise (Fig. 5).

Discussion:

”Firstly, E neurons and PV neurons experience very similar synaptic environments.” What does it mean?

”Fortunately, PV neurons target both the cell bodies and proximal dendrites” Fortunately for whom or what? ”in line with arge heterogeneity”

Methods:

Matrix B is never defined - the diagonal matrix of b (power law exponents) I assume.

Some of the other notations too, e.g. bs, etc (it’s implicit, but should be explained).

Structure of sentence:

”Network gain is defined as ...” (p. 17)

Figure:

The schematics in Figure 4 can be tweaked to highlight the effect of input (rather than other components of the network, which are the same and repetitive), to highlight the main difference for the reader.

eLife Assessment

This useful manuscript presents findings on Tom1p's interaction with Spt6p and its role in chromatin dynamics, supported by structural analysis through CryoEM. The evidence for the conclusions is currently incomplete, lacking key experiments including continuation in vivo interaction and orthogonal binding assays (e.g., SPR, MST, ITC) to fully support the proposed mechanism. While the results are promising, further validation is needed to strengthen the evidence and improve the manuscript's overall cohesion.

Reviewer #1 (Public review):

Summary:

In this preprint, Madrigal et al present "Tom1p ubiquitin ligase structure, interaction with Spt6p, and function in maintaining normal transcript levels and the stability of chromatin in promoters" which describes the identification of Tom1p, a conserved ubiquitin ligase, as a potential binding partner for the transcription elongation/histone chaperone Spt6p, and reveal the Tom1p structure as determined by CryoEM. Tom1p is a homolog of human HUWE1, which has been implicated in decay for a variety of basic protein substrates such as ribosomal proteins and histones. Structure-function analyses identify regions required for Spt6p interaction, suggesting that the interaction with Spt6p is phosphorylation dependent, and for interactions with histones, the latter of which confers phenotypes in vivo when mutated, suggesting that the Tom1p acidic region is important for its function. What is less clear is the function or interaction with Spt6p. The manuscript speculates that Spt6p-Tom1p interactions may tune Tom1p localization, and it is shown that Tom1p is recruited to transcribed genes by chromatin IP. In addition, the Tom1p structure will be valuable to those trying to understand the mechanisms of this very large ubiquitin ligase. Here, structures of homologs from other organisms have already been described elsewhere, however, the authors here indicate some details potentially not previously visualized in other structures.

Strengths:

It has not previously been known that the Spt6p tSH2 had any additional targets. Interaction with a ubiquitin ligase already implicated in histone turnover given Spt6p's role as histone chaperone is interesting. A structure of Tom1p also provides insight into this very large, conserved protein and structure-function analysis in a model system is a good start towards mechanistic dissection.

Weaknesses:

Some aspects of the manuscript seem less cohesive in that there are two halves of the manuscript and both don't quite solidify insights into the Spt6p relationship to Tom1p or deepen our understanding of Tom1p mechanism extensively, though results are a great start on both sides of the paper. There are several points that are less clear in that it is not known if Spt6p interacts with Tom1p and in what context. The interaction surface of Spt6p able to interact with Tom1p is the identical tSH2 that would be predicted to be occupied by phosphorylated RNAPII when Spt6p is incorporated into the RNAPII elongation complex. This means how and when Spt6p might be available to interact with Tom1p is not clear. Previous work from the Hill and Formosa groups on the tSH2 domain and its RNAPII linker target have suggested that phenotypes of mutants in the two are similar, suggesting that their main function is to interact with each other. A simple test of examining Tom1p interaction with genes in the tSH2 mutant was not done. Additionally, the Spt6p interacting surface on Tom1p is not narrowed to a specific putatively phosphorylated residue that it might target. It remains possible that mutations in other regions of Tom1p affect potential phosphorylation of this target, and therefore it is possible that some mutations that alter Spt6p interaction could do so indirectly. Finally, the authors might consider additional models for their discussion where Spt6p potentially could function to deliver histones to Tom1p.

Reviewer #2 (Public review):

Summary:

Madrigal et al identified Tom1, a E3 ubiquitin ligase previously known to be involved in ribosome biogenesis, as a protein that binds to terminal tandem Src-homology 2 (tSH2) domain of Spt6. They mapped this interaction to the acid region of Tom1, which is also known to interact with histones. Cells with tom1 mutants that cannot bind Spt6 did not show temperature sensitive phenotypes displaced for tom1 null mutant. Using ChIP assays, they showed that Tom1 is enriched at gene bodies of highly transcribed genes, and a loss of tom1 leads to reduced nucleosomal changes at gene promoters. Finally, they also solved the structures of Tom1 lacking the acidic region and found that Tom1p can adopt a compact a-solenoidal "basket" similar to the previously described structure of HUWE1. Overall, this is an interesting study and I have the following suggestions to improve the manuscript.

Major concerns:

(1) Promoter regions are in general nucleosome free. How does Tom1 mutant affect nucleosome-sized fragments at the promoter regions?

(2) While Tom1 antibodies may not specific, could the author perform Tom1 ChIP-seq in wild type and tom 1 null cells? This dataset may be more informative than tagged Tom1 that may not be functional.

Reviewer #3 (Public review):

Summary:

The authors report a novel, direct interaction of Spt6p tSH2 domain to Tom1p. This extends the function of Spt6p from communication with factors associated with RNAPII transcription to processes of ubiquitination. Tom1p is known to ubiquitinate a large variety of substrates, but it is unknown how substrate recognition is done in a specific manner. The team identified a conserved central acidic region of Tom1p which is essential for in vivo functions and binds to histones and nucleosomes, as well as Spt6p. They further describe the Tom1p occupancy pattern on chromatin, assigning it a stabilizing effect on nucleosomes near promotors and a destabilizing effect on nucleosomes within the gene bodies. The authors were able to resolve two different conformational states of Tom1p which are likely connected to its activity, and possibly substrate selectivity.

Overall, the authors show that an intrinsically disordered region in Tom1p is important for substrate interaction and function of Tom1p. The protein is further involved in chromatin architecture and structural transitions control its activity.

Strengths:

By revealing the interaction of Spt6p and Tom1p, the authors discover a novel connection between transcriptional elongation and processes of ubiquitination.<br /> In recent years, disordered regions of MDa protein complexes have become a focus of research projects. The effects of disordered regions on protein localization and specificity of binding interactions have been discussed in great extent, including proteins that are involved in chromatin remodeling and transcription. Adding to these current efforts, the authors assign a function to a highly conserved disordered region of Tom1p in technically clean experiments. Furthermore, with their data, they pin down a specific functional region in Tom1p which is relevant for the previously observed temperature sensitivity caused by Tom1p deletion in yeast.<br /> The team performs a thorough and complete analysis of the cryo-EM structure and they nicely model the hinge motion and details of an open and closed conformation.

Weaknesses:

Despite the high number of interesting findings, there is little connection between the individual sections of the manuscript. For example, many experiments are not related to Spt6p binding although this protein is presented as a major actor in this manuscript during the introduction. Furthermore, the structural analysis is well done, but it is also not quite clear how structural rearrangements are connected to Spt6 binding or chromatin remodeling. Some experimental results lack novelty, as similar data has previously been presented for the human homolog.

To confirm the novel, direct binding interaction of Spt6p and Tom1p, no orthogonal binding assays (SPR, MST, ITC) have been performed to confirm the interaction. To me, this is insufficient, especially since the team has purified both proteins to high quality levels, or could use peptides to test the function of the relevant regions.<br /> Additionally, interaction of Tom1p with Spt6p in the context of transcription elongation is proposed. Yet it is not clear on the mechanistic level how this is regulated if Tom1p and Rpb1p bind in a competitive manner. How is Tom1p tethered to the elongation complex if not through Spt6p? In addition to WT vs. knockout, the authors should further perform the genetic analyses with the intΔ11 mutant. This way they might be able pin down which interactions on chromatin are mediated by Spt6 vs. by other factors and could strengthen the overall model involving Spt6P.

Although the authors try to describe a final model in the discussion, this section is not easy to follow and needs more explanation, ideally drawn as a Figure of the proposed mechanism.

Author response:

We thank the reviewers’ for their helpful comments.  We will make several minor edits to the text to improve clarity. Further experiments are beyond the scope of the current study.

eLife Assessment

By combining the 'pinging' technique with fMRI-based multivariate decoding, this important study examined the nature of the representation of the attentional template during preparation. While the findings are very interesting and the experimental evidence is solid, the methodological (e.g., the manipulation of attention, the potential cross-contamination between attention and working memory, and the representational distance analysis) and interpretation confounds (e.g., more thorough clarification of "pinging" and dual-format attentional templates) need to be addressed. The work will be of interest to researchers in psychology, cognitive science, and neuroscience.

Reviewer #1 (Public review):

Summary:

The aim of the experiment reported in this paper is to examine the nature of the representation of a template of an upcoming target. To this end, participants were presented with compound gratings (consisting of tilted to the right and tilted to the left lines) and were cued to a particular orientation - red left tilt or blue right tilt (counterbalanced across participants). There are two directly compared conditions: (i) no ping: where there was a cue, that was followed by a 5.5-7.5s delay, then followed by a target grating in which the cued orientation deviated from the standard 45 degrees; and (ii) ping condition in which all aspects were the same with the only difference that a ping (visual impulse presented for 100ms) was presented after the 2.5 seconds following the cue. There was also a perception task in which only the 45 degrees to the right or to the left lines were presented. It was observed that during the delay, only in the ping condition, were the authors able to decode the orientation of the to-be-reported target using the cross-task generalization. Attention decoding, on the other hand, was decoded in both ping and non-ping conditions. It is concluded that the visual system has two different functional states associated with a template during preparation: a predominantly non-sensory representation for guidance and a latent sensory-like for prospective stimulus processing.

Strengths:

There is so much to be impressed with in this report. The writing of the manuscript is incredibly clear. The experimental design is clever and innovative. The analysis is sophisticated and also innovative - the cross-task decoding, the use of Mahalanobis distance as a function of representational similarity, the fact that the question is theoretically interesting, and the excellent figures.

Weaknesses:

While I think that this is an interesting study that addresses an important theoretical question, I have several concerns about the experimental paradigm and the subsequent conclusions that can be drawn.

(1) Why was V1 separated from the rest of the visual cortex, and why the rest of the areas were simply lumped into an EVC ROI? It would be helpful to understand the separation into ROIs.

(2) It would have been helpful to have a behavioral measure of the "attended" orientation to show that participants in fact attended to a particular orientation and were faster in the cued condition. The cue here was 100% valid, so no such behavioral measure of attention is available here.

(3) As I was reading the manuscript I kept thinking that the word attention in this manuscript can be easily replaced with visual working memory. Have the authors considered what it is about their task or cognitive demand that makes this investigation about attention or working memory?

(4) If I understand correctly, the only ROI that showed a significant difference for the cross-task generalization is V1. Was it predicted that only V1 would have two functional states? It should also be made clear that the only difference where the two states differ is V1.

(5) My primary concern about the interpretation of the finding is that the result, differences in cross-task decoding within V1 between the ping and no-ping condition might simply be explained by the fact that the ping condition refocuses attention during the long delay thus "resharpening" the template. In the no-ping condition during the 5.5 to 7.5 seconds long delay, attention for orientation might start getting less "crisp." In the ping condition, however, the ping itself might simply serve to refocus attention. So, the result is not showing the difference between the latent and non-latent stages, rather it is the difference between a decaying template representation and a representation during the refocused attentional state. It is important to address this point. Would a simple tone during the delay do the same? If so, the interpretation of the results will be different.

(6) The neural pattern distances measured using Mahalanobis values are really great! Have the authors tried to use all of the data, rather than the high AMI and low AMI to possibly show a linear relationship between response times and AMI?

(7) After reading the whole manuscript I still don't understand what the authors think the ping is actually doing, mechanistically. I would have liked a more thorough discussion, rather than referencing previous papers (all by the co-author).

Reviewer #2 (Public review):

Summary:

In the present study, the authors investigated the nature of attentional templates during the preparatory period of goal-directed attention. By combining the use of 'pinging' the neural activity with a visual impulse and fMRI-based multivariate decoding, the authors found that the nature of the neural representations of the prospective feature target during the preparatory period was contingent on the presence of the 'pinging' impulse. While the preparatory representations contained highly similar information content as the perceptual representations when the pinging impulse was introduced, they fundamentally differed from perceptual representations in the absence of the pinging impulse. Based on these findings, the authors proposed a dual-format mechanism in which both a "non-sensory" template and a latent "sensory" template coexisted during attentional preparation. The former actively guides activity in the preparatory state, and the latter is utilized for future stimulus processing.

Strengths:

Overall, I think this is an interesting study that introduced a novel perspective concerning the nature of neural representations during attentional processing. Methodologically, the present study combines an innovative utilization of the pinging technique in working memory studies and fMRI-based multivariate pattern analysis. The method is sound and the results are convincing. While I appreciate the conceptual elegance of the dual-format idea proposed by the authors, there are several questions that need to be addressed more thoroughly to clarify some of the potential ambiguities of the results and to increase the plausibility of the author's theory.

Weaknesses:

(1) The origin of the latent sensory-like representation. By 'pinging' the neural activity with a high-contrast, task-irrelevant visual stimulus during the preparation period, the authors identified the representation of the attentional feature target that contains the same information as perceptual representations. The authors interpreted this finding as a 'sensory-like' template is inherently hosted in a latent form in the visual system, which is revealed by the pinging impulse. However, I am not sure whether such a sensory-like template is essentially created, rather than revealed, by the pinging impulses. First, unlike the classical employment of the pinging technique in working memory studies, the (latent) representation of the memoranda during the maintenance period is undisputed because participants could not have performed well in the subsequent memory test otherwise. However, this appears not to be the case in the present study. As shown in Figure 1C, there was no significant difference in behavioral performance between the ping and the no-ping sessions (see also lines 110-125, pg. 5-6). In other words, it seems to me that the subsequent attentional task performance does not necessarily rely on the generation of such sensory-like representations in the preparatory period and that the emergence of such sensory-like representations does not facilitate subsequent attentional performance either. In such a case, one might wonder whether such sensory-like templates are really created, hosted, and eventually utilized during the attentional process. Second, because the reference orientations (i.e. 45 degrees and 135 degrees) have remained unchanged throughout the experiment, it is highly possible that participants implicitly memorized these two orientations as they completed more and more trials. In such a case, one might wonder whether the 'sensory-like' templates are essentially latent working memory representations activated by the pinging as was reported in Wolff et al. (2017), rather than a functional signature of the attentional process.

(2) The coexistence of the two types of attentional templates. The authors interpreted their findings as the outcome of a dual-format mechanism in which 'a non-sensory template' and a latent 'sensory-like' template coexist (e.g. lines 103-106, pg. 5). While I find this interpretation interesting and conceptually elegant, I am not sure whether it is appropriate to term it 'coexistence'. First, it is theoretically possible that there is only one representation in either session (i.e. a non-sensory template in the no-ping session and a sensory-like template in the ping session) in any of the brain regions considered. Second, it seems that there is no direct evidence concerning the temporal relationship between these two types of templates, provided that they commonly emerge in both sessions. Besides, due to the sluggish nature of fMRI data, it is difficult to tell whether the two types of templates temporally overlap.

(3) The representational distance. The authors used Mahalanobis distance to quantify the similarity of neural representation between different conditions. According to the authors' hypothesis, one would expect greater pattern similarity between 'attend leftward' and 'perceived leftward' in the ping session in comparison to the no-ping session. However, this appears not to be the case. As shown in Figures 3B and C, there was no major difference in Mahalanobis distance between the two sessions in either ROI and the authors did not report a significant main effect of the session in any of the ANOVAs. Besides, in all the ANOVAs, the authors reported only the statistic term corresponding to the interaction effect without showing the descriptive statistics related to the interaction effect. It is strongly advised that these descriptive statistics related to the interaction effect should be included to facilitate a more effective and intuitive understanding of their data.

Reviewer #3 (Public review):

This paper discusses how non-sensory and latent, sensory-like attentional templates are represented during attentional preparation. Using multivariate pattern analysis, they found that visual impulses can enhance the decoding generalization from perception to attention tasks in the preparatory stage in the visual cortex. Furthermore, the emergence of the sensory-like template coincided with enhanced information connectivity between V1 and frontoparietal areas and was associated with improved behavioral performance. It is an interesting paper with supporting evidence for the latent, sensory-like attentional template, but several problems still need to be solved.

(1) The title is "Dual-format Attentional Template," yet the supporting evidence for the non-sensory format and its guiding function is quite weak. The author could consider conducting further generalization analysis from stimulus selection to preparation stages to explore whether additional information emerges.

(2) In Figure 2, the author did not find any decodable sensory-like coding in IPS and PFC, even during the impulse-driven session, indicating that these regions do not represent sensory-like information. However, in the final section, the author claimed that the impulse-driven sensory-like template strengthens informational connectivity between sensory and frontoparietal areas. This raises a question: how can we reconcile the lack of decodable coding in these frontoparietal regions with the reported enhancement in network communication? It would be helpful if the author provided a clearer explanation or additional evidence to bridge this gap.

(3) Given that the impulse-driven sensory-like template facilitated behavior, the author proposed that it might also enhance network communication. Indeed, they observed changes in informational connectivity. However, it remains unclear whether these changes in network communication have a direct and robust relationship with behavioral improvements.

(4) I'm uncertain about the definition of the sensory-like template in this paper. Is it referring to the Ping impulse-driven condition or the decodable performance in the early visual cortex? If it is the former, even in working memory, whether pinging identifies an activity-silent mechanism is currently debated. If it's the latter, the authors should consider whether a causal relationship - such as "activating the sensory-like template strengthens the informational connectivity between sensory and frontoparietal areas" - is reasonable.

eLife Assessment

What makes one member of the species behave differently from another? This is a core problem in behavioral neuroscience. This valuable study seeks an answer for the specific case of the fruit fly expressing preferences for one odor over another. By a combination of behavioral measurements, neurophysiology, and network modeling, the authors find solid evidence for at least one locus of individuality in the peripheral olfactory system.

Joint Public Review:

Summary:

The authors aimed to identify the neural sources of behavioral variation in fruit flies deciding between odor and air, or between two odors.

Strengths:

- The question is of fundamental importance.<br /> - The behavioral studies are automated, and high-throughput.<br /> - The data analyses are sophisticated and appropriate.<br /> - The paper is clear and well-written aside from some initially strong wording.<br /> - The figures beautifully illustrate their results.<br /> - The modeling efforts mechanistically ground observed data correlations.

Weaknesses:

-The correlations between behavioral variations and neural activity/synapse morphology are relatively weak, and sometimes overstated in the wording that describes them.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors seek to establish what aspects of nervous system structure and function may explain behavioral differences across individual fruit flies. The behavior in question is a preference for one odor or another in a choice assay. The variables related to neural function are odor responses in olfactory receptor neurons or in the second-order projection neurons, measured via calcium imaging. A different variable related to neural structure is the density of a presynaptic protein BRP. The authors measure these variables in the same fly along with the behavioral bias in the odor assays. Then they look for correlations across flies between the structure-function data and the behavior.

Strengths:

Where behavioral biases originate is a question of fundamental interest in the field. In an earlier paper (Honegger 2019) this group showed that flies do vary with regard to odor preference, and that there exists neural variation in olfactory circuits, but did not connect the two in the same animal. Here they do, which is a categorical advance, and opens the door to establishing a correlation. The authors inspect many such possible correlations. The underlying experiments reflect a great deal of work, and appear to be done carefully. The reporting is clear and transparent: All the data underlying the conclusions are shown, and associated code is available online.

We are glad to hear the reviewer is supportive of the general question and approach.

Weaknesses:

The results are overstated. The correlations reported here are uniformly small, and don't inspire confidence that there is any causal connection. The main problems are

Our revision overhauls the interpretation of the results to prioritize the results we have high confidence in (specifically, PC 2 of our Ca++ data as a predictor of OCT-MCH preference) versus results that are suggestive but not definitive (such as PC 1 of Ca++ data as a predictor of Air-OCT preference).

It’s true that the correlations are small, with R2 values typically in the 0.1-0.2 range. That said, we would call it a victory if we could explain 10 to 20% of the variance of a behavior measure, captured in a 3 minute experiment, with a circuit correlate. This is particularly true because, as the reviewer notes, the behavioral measurement is noisy.

(1) The target effect to be explained is itself very weak. Odor preference of a given fly varies considerably across time. The systematic bias distinguishing one fly from another is small compared to the variability. Because the neural measurements are by necessity separated in time from the behavior, this noise places serious limits on any correlation between the two.

This is broadly correct, though to quibble, it’s our measurement of odor preference which varies considerably over time. We are reasonably confident that more variance in our measurements can be attributed to sampling error than changes to true preference over time. As evidence, the correlation in sequential measures of individual odor preference, with delays of 3 hours or 24 hours, are not obviously different. We are separately working on methodological improvements to get more precise estimates of persistent individual odor preference, using averages of multiple, spaced measurements. This is promising, but beyond the scope of this study.

(2) The correlations reported here are uniformly weak and not robust. In several of the key figures, the elimination of one or two outlier flies completely abolishes the relationship. The confidence bounds on the claimed correlations are very broad. These uncertainties propagate to undermine the eventual claims for a correspondence between neural and behavioral measures.

We are broadly receptive to this criticism. The lack of robustness of some results comes from the fundamental challenge of this work: measuring behavior is noisy at the individual level. Measuring Ca++ is also somewhat noisy. Correlating the two will be underpowered unless the sample size is huge (which is impractical, as each data point requires a dissection and live imaging session) or the effect size is large (which is generally not the case in biology). In the current version we tried in some sense to avoid discussing these challenges head-on, instead trying to focus on what we thought were the conclusions justified by our experiments with sample sizes ranging from 20 to 60. Our revision is more candid about these challenges.

That said, we believe the result we view as the most exciting — that PC2 of Ca++ responses predicts OCT-MCH preference — is robust. 1) It is based on a training set with 47 individuals and a test set composed of 22 individuals. The p-value is sufficiently low in each of these sets (0.0063 and 0.0069, respectively) to pass an overly stringent Bonferroni correction for the 5 tests (each PC) in this analysis. 2) The BRP immunohistochemistry provides independent evidence that is consistent with this result — PC2 that predicts behavior (p = 0.03 from only one test) and has loadings that contrast DC2 and DM2. Taken together, these results are well above the field-standard bar of statistical robustness.

In our revision, we are explicit that this is the (one) result we have high confidence in. We believe this result convincingly links Ca++ and behavior, and warrants spotlighting. We have less confidence in other results, and say so, and we hope this addresses concerns about overstating our results.

(3) Some aspects of the statistical treatment are unusual. Typically a model is proposed for the relationship between neuronal signals and behavior, and the model predictions are correlated with the actual behavioral data. The normal practice is to train the model on part of the data and test it on another part. But here the training set at times includes the testing set, which tends to give high correlations from overfitting. Other times the testing set gives much higher correlations than the training set, and then the results from the testing set are reported. Where the authors explored many possible relationships, it is unclear whether the significance tests account for the many tested hypotheses. The main text quotes the key results without confidence limits.

Our primary analyses are exactly what the reviewer describes, scatter plots and correlations of actual behavioral measures against predicted measures. We produced test data in separate experiments, conducted weeks to months after models were fit on training data. This is more rigorous than splitting into training and test sets data collected in a single session, as batch/environmental effects reduce the independence of data collected within a single session.

We only collected a test set when our training set produced a promising correlation between predicted and actual behavioral measures. We never used data from test sets to train models. In our main figures, we showed scatter plots that combined test and training data, as the training and test partitions had similar correlations.

We are unsure what the reviewer means by instances where we explored many possible relationships. The greatest number of comparisons that could lead to the rejection of a null hypothesis was 5 (corresponding to the top 5 PCs of Ca++ response variation or Brp signal). We were explicit that the p-values reported were nominal. As mentioned above, applying a Bonferroni correction for n=5 comparisons to either the training or test correlations from the Ca++ to OCT-MCH preference model remains significant at alpha=0.05.

Our revision includes confidence intervals around ⍴signal for the PN PC2 OCT-MCH model, and for the ORN Brp-Short PC2 OCT-MCH model (lines 170-172, 238)

Reviewer #2 (Public Review):

Summary:

The authors aimed to identify the neural sources of behavioral variation in a decision between odor and air, or between two odors.

Strengths:

-The question is of fundamental importance.

-The behavioral studies are automated, and high-throughput.

-The data analyses are sophisticated and appropriate.

-The paper is clear and well-written aside from some strong wording.

-The figures beautifully illustrate their results.

-The modeling efforts mechanistically ground observed data correlations.

We are glad to read that the reviewer sees these strengths in the study. We hope the current revision addresses the strong wording.

Weaknesses:

-The correlations between behavioral variations and neural activity/synapse morphology are (i) relatively weak, (ii) framed using the inappropriate words "predict", "link", and "explain", and (iii) sometimes non-intuitive (e.g., PC 1 of neural activity).

Taking each of these points in turn:

i) It would indeed be nicer if our empirical correlations are higher. One quibble: we primarily report relatively weak correlations between measurements of behavior and Ca++/Brp. This could be the case even when the correlation between true behavior and Ca++/Brp is higher. Our analysis of the potential correlation between latent behavioral and Ca++ signals was an attempt to tease these relationships apart. The analysis suggests that there could, in fact, be a high underlying correlation between behavior and these circuit features (though the error bars on these inferences are wide).

ii) We worked to ensure such words are used appropriately. “Predict” can often be appropriate in this context, as a model predicts true data values. Explain can also be appropriate, as X “explaining” a portion of the variance of Y is synonymous with X and Y being correlated. We cannot think of formal uses of “link,” and have revised the manuscript to resolve any inappropriate word choice.

iii) If the underlying biology is rooted in non-intuitive relationships, there’s unfortunately not much we can do about it. We chose to use PCs of our Ca++/Brp data as predictors to deal with the challenge of having many potential predictors (odor-glomerular responses) and relatively few output variables (behavioral bias). Thus, using PCs is a conservative approach to deal with multiple comparisons. Because PCs are just linear transformations of the original data, interpreting them is relatively easy, and in interpreting PC1 and PC2, we were able to identify simple interpretations (total activity and the difference between DC2 and DM2 activation, respectively). All in all, we remain satisfied with this approach as a means to both 1) limit multiple comparisons and 2) interpret simple meanings from predictive PCs.

No attempts were made to perturb the relevant circuits to establish a causal relationship between behavioral variations and functional/morphological variations.

We did conduct such experiments, but we did not report them because they had negative results that we could not definitively interpret. We used constitutive and inducible effectors to alter the physiology of ORNs projecting to DC2 and DM2. We also used UAS-LRP4 and UAS-LRP4-RNAi to attempt to increase and decrease the extent of Brp puncta in ORNs projecting to DC2 and DM2. None of these manipulations had a significant effect on mean odor preference in the OCT-MCH choice, which was the behavioral focus of these experiments. We were unable to determine if the effectors had the intended effects in the targeted Gal4 lines, particularly in the LRP experiments, so we could not rule out that our negative finding reflected a technical failure.

Author response image 1.

We believe that even if these negative results are not technical failures, they are not necessarily inconsistent with the analyses correlating features of DC2 and DM2 to behavior. Specifically, we suspect that there are correlated fluctuations in glomerular Ca++ responses and Brp across individuals, due to fluctuations in the developmental spatial patterning of the antennal lobe. Thus, the DC2-DM2 predictor may represent a slice/subset of predictors distributed across the antennal lobe. This would also explain how we “got lucky” to find two glomeruli as predictors of behavior, when we were only able to image a small portion of the glomeruli.

Reviewer #3 (Public Review):

Churgin et. al. seeks to understand the neural substrates of individual odor preference in the Drosophila antennal lobe, using paired behavioral testing and calcium imaging from ORNs and PNs in the same flies, and testing whether ORN and PN odor responses can predict behavioral preference. The manuscript's main claims are that ORN activity in response to a panel of odors is predictive of the individual's preference for 3-octanol (3-OCT) relative to clean air, and that activity in the projection neurons is predictive of both 3-OCT vs. air preference and 3-OCT vs. 4-methylcyclohexanol (MCH). They find that the difference in density of fluorescently-tagged brp (a presynaptic marker) in two glomeruli (DC2 and DM2) trends towards predicting behavioral preference between 3-oct vs. MCH. Implementing a model of the antennal lobe based on the available connectome data, they find that glomerulus-level variation in response reminiscent of the variation that they observe can be generated by resampling variables associated with the glomeruli, such as ORN identity and glomerular synapse density.

Strengths:

The authors investigate a highly significant and impactful problem of interest to all experimental biologists, nearly all of whom must often conduct their measurements in many different individuals and so have a vested interest in understanding this problem. The manuscript represents a lot of work, with challenging paired behavioral and neural measurements.

Weaknesses:

The overall impression is that the authors are attempting to explain complex, highly variable behavioral output with a comparatively limited set of neural measurements.

We would say that we are attempting to explain a simple, highly variable behavioral measure with a comparatively limited set of neural measurements, i.e. we make no claims to explain the complex behavioral components of odor choice, like locomotion, reversals at the odor boundary, etc.

Given the degree of behavioral variability they observe within an individual (Figure 1- supp 1) which implies temporal/state/measurement variation in behavior, it's unclear that their degree of sampling can resolve true individual variability (what they call "idiosyncrasy") in neural responses, given the additional temporal/state/measurement variation in neural responses.

We are confident that different Ca++ recordings are statistically different. This is borne out in the analysis of repeated Ca++ recordings in this study, which finds that the significant PCs of Ca++ variation contain 77% of the variation in that data. That this variation is persistent over time and across hemispheres was assessed in Honegger & Smith, et al., 2019. We are thus confident that there is true individuality in neural responses (Note, we prefer not to call it “individual variability” as this could refer to variability within individuals, not variability across individuals.) It is a separate question of whether individual differences in neural responses bear some relation to individual differences in behavioral biases. That was the focus of this study, and our finding of a robust correlation between PC 2 of Ca++ responses and OCT-MCH preference indicates a relation. Because behavior and Ca++ were collected with an hours-to-day long gap, this implies that there are latent versions of both behavioral bias and Ca++ response that are stable on timescales at least that long.

The statistical analyses in the manuscript are underdeveloped, and it's unclear the degree to which the correlations reported have explanatory (causative) power in accounting for organismal behavior.

With respect, we do not think our statistical analyses are underdeveloped, though we acknowledge that the detailed reviewer suggestions included the helpful suggestion to include uncertainty in the estimation of confidence intervals around the point estimate of the strength of correlation between latent behavioral and Ca++ response states – we have added these for the PN PC2 linear model (lines 170-172).

It is indeed a separate question whether the correlations we observed represent causal links from Ca++ to behavior (though our yoked experiment suggests there is not a behavior-to-Ca++ causal relationship — at least one where odor experience through behavior is an upstream cause). We attempted to be precise in indicating that our observations are correlations. That is why we used that word in the title, as an example. In the revision, we worked to ensure this is appropriately reflected in all word choice across the paper.

Recommendations for the Authors:

Reviewer #1 (Recommendations for the Authors):

Detailed comments: Many of the problems can be identified starting from Figure 4, which summarizes the main claims. I will focus on that figure and its tributaries.

Acknowledging that the strength of several of our inferences are weak compared to what we consider the main result (the relationship between PC2 of Ca++ and OCT-MCH preference),we have removed Figure 4. This makes the focus of the paper much clearer and appropriately puts focus on the results that have strong statistical support.

(1) The process of "inferring" correlation among the unobserved latent states for neural sensitivity and behavioral bias is unconventional and risky. The larger the assumed noise linking the latent to the observed variables (i.e. the smaller r_b and r_c) the bigger the inferred correlation rho from a given observed correlation R^2_cb. In this situation, the value of the inferred rho becomes highly dependent on what model one assumes that links latent to observed states. But the specific model drawn in Fig 4 suppl 1 is just one of many possible guesses. For example, models with nonlinear interactions could produce different inference.

We agree with the reviewer’s notes of caution. To be clear, we do not intend for this analysis to be the main takeaway of the paper and have revised it to make this clear. The signal we are most confident in is the simple correlation between measured Ca++ PC2 and measured behavior. We have added more careful language saying that the attempt to infer the correlation between latent signals is one attempt at describing the data generation process (lines 166-172), and one possible estimate of an “underlying” correlation.

(2) If one still wanted to go through with this inference process and set confidence bounds on rho, one needs to include all the uncertainties. Here the authors only include uncertainty in the value of R^2_c,b and they peg that at +/-20% (Line 1367). In addition there is plenty of uncertainty associated also with R^2_c,c and R^2_b,b. This will propagate into a wider confidence interval on rho.

We have replaced the arbitrary +/- 20% window with bootstrapping the pairs of (predicted preference by PN PC2, measured preference) points and getting a bootstrap distribution of R2c,b, which is, not surprisingly, considerably wider. Still, we think there is some value in this analysis as the 90% CI of 𝜌signal under this model is 0.24-0.95. That is, including uncertainty about the R2b,b and R2c,c in the model still implies a significant relationship between latent calcium and behavior signals.

(2.1) The uncertainty in R^2_cb is much greater than +/-20%. Take for example the highest correlation quoted in Fig 4: R^2=0.23 in the top row of panel A. This relationship refers to Fig 1L. Based on bootstrapping from this data set, I find a 90% confidence interval of CI=[0.002, 0.527]. That's an uncertainty of -100/+140%, not +/-20%. Moreover, this correlation is due entirely to the lone outlier on the bottom left. Removing that single fly abolishes any correlation in the data (R^2=0.04, p>0.3). With that the correlation of rho=0.64, the second-largest effect in Fig 4, disappears.

We acknowledge that removal of the outlier in Fig 1L abolishes the correlation between predicted and measured OCT-AIR preference. We have thus moved that subfigure to the supplement (now Figure 1 – figure supplement 10B), note that we do not have robust statistical support of ORN PC1 predicting OCT-AIR preference in the results (lines 177-178), and place our emphasis on PN PC2’s capacity to predict OCT-MCH preference throughout the text.

(2.2) Similarly with the bottom line of Fig 4A, which relies on Fig 1M. With the data as plotted, the confidence interval on R^2 is CI=[0.007, 0.201], again an uncertainty of -100/+140%. There are two clear outlier points, and if one removes those, the correlation disappears entirely (R^2=0.06, p=0.09).

We acknowledge that removal of the two outliers in Fig 1M between predicted and measured OCT-AIR preference abolishes the correlation. We have also moved that subfigure to the supplement (now Figure 1 – figure supplement 10F) and do not claim to have robust statistical support of PN PC1 predicting OCT-AIR preference.

(2.3) Similarly, the correlation R^2_bb of behavior with itself is weak and comes with great uncertainty (Fig 1 Suppl 1, panels B-E). For example, panel D figures prominently in computing the large inferred correlation of 0.75 between PN responses and OCT-MCH choice (Line 171ff). That correlation is weak and has a very wide confidence interval CI=[0.018, 0.329]. This uncertainty about R^2_bb should be taken into account when computing the likelihood of rho.

We now include bootstrapping of the 3 hour OCT-MCH persistence data in our inference of 𝜌signal.

(2.4) The correlation R^2_cc for the empirical repeatability of Ca signals seems to be obtained by a different method. Fig 4 suppl 1 focuses on the repeatability of calcium recording at two different time points. But Line 625ff suggests the correlation R^2_cc=0.77 all derives from one time point. It is unclear how these are related.

Because our calcium model predictors utilize principal components of the glomerulus-odor responses (the mean Δf/f in the odor presentation window), we compute R2c,c through adding variance explained along the PCs, up to the point in which the component-wise variance explained does not exceed that of shuffled data (lines 609-620 in Materials and Methods). In this revision we now bootstrap the calcium data on the level of individual flies to get a bootstrap distribution of R2c,c, and propagate the uncertainty forward in the inference of 𝜌signal.

(2.5) To summarize, two of the key relationships in Fig 1 are due entirely to one or two outlier points. These should not even be used for further analysis, yet they underlie two of the claims in Fig 4. The other correlations are weak, and come with great uncertainty, as confirmed by resampling. Those uncertainties should be propagated through the inference procedure described in Fig 4. It seems possible that the result will be entirely uninformative, leaving rho with a confidence interval that spans the entire available range [0,1]. Until that analysis is done, the claims of neuron-to-behavior correlation in this manuscript are not convincing.

It is important to note that we never thought our analysis of the relationship between latent behavior and calcium signals should be interpreted as the main finding. Instead, the observed correlation between measured behavior and calcium is the take-away result. Importantly, it is also conservative compared to the inferred latent relationship, which in our minds was always a “bonus” analysis. Our revisions are now focused on highlighting the correlations between measured signals that have strong statistical support.

As a response to these specific concerns, we have propagated uncertainty in all R2’s (calcium-calcium, behavior-behavior, calcium-behavior) in our new inference for 𝜌signal, yielding a new median estimate for PN PC 2 underlying OCT-MCH preference of 0.68, with a 90% CI of 0.24-0.95. (Lines 171-172 in results, Inference of correlation between latent calcium and behavior states section in Materials and Methods).

(3) Other statistical methods:

(3.1) The caption of Fig 4 refers to "model applied to train+test data". Does that mean the training data were included in the correlation measurement? Depending on the number of degrees of freedom in the model, this could have led to overfitting.

We have removed Figure 4 and emphasize the key results in Figure 1 and 2 that we see statistically robust signal of PN PC 2 explaining OCT-MCH preference variation in both a training set and a testing set of flies (Fig 2 – figure supplement 1C-D).

(3.2) Line 180 describes a model that performed twice as well on test data (31% EV) as it did on training data (15%). What would explain such an outcome? And how does that affect one's confidence in the 31% number?

The test set recordings were conducted several weeks after the training set recordings, which were used to establish PN PC 2 as a correlate of OCT-MCH preference. The fact that the test data had a higher R2 likely reflects sampling error (these two correlation coefficients are not significantly different). Ultimately this gives us more confidence in our model, as the predictive capacity is maintained in a totally separate set of flies.

(3.340 Multiple models get compared in performance before settling on one. For example, sometimes the first PC is used, sometimes the second. Different weighting schemes appear in Fig 2. Do the quoted p-values for the correlation plots reflect a correction for multiple hypothesis testing?

For all calcium-behavior models, we restricted our analysis to 5 PCs, as the proportion of calcium variance explained by each of these PCs was higher than that explained by the respective PC of shuffled data — i.e., there were at most five significant PCs in that data. We thus performed at most 5 hypothesis tests for a given model. PN PC 2 explained 15% of OCT-MCH preference variation, with a p-value of 0.0063 – this p-value is robust to a conservative Bonferroni correction to the 5 hypotheses considered at alpha=0.05.

The weight schemes in Figure 2 and Figure 1 – figure supplement 10 reflect our interpretations of the salient features of the PCs and are follow-up analysis of the single principal component hypothesis tests. Thus they do not constitute additional tests that should be corrected. We now state in the methods explicitly that all reported p-values are nominal (line 563).

(3.4) Line 165 ff: Quoting rho without giving the confidence interval is misleading. For example, the rho for the presynaptic density model is quoted as 0.51, which would be a sizeable correlation. But in fact, the posterior on rho is almost flat, see caption of Fig 4 suppl 1, which lists the CI as [0.11, 0.85]. That means the experiments place virtually no constraint on rho. If the authors had taken no data at all, the posterior on rho would be uniform, and give a median of 0.5.

We now provide a confidence interval around 𝜌signal for the PN PC 2 model (lines 170-172). But per above, and consistent with the new focus of this revision, we view the 𝜌signal inference as secondary to the simple, significant correlation between PN PC 2 and OCT-MCH preference.

(4) As it stands now, this paper illustrates how difficult it is to come to a strong conclusion in this domain. This may be worth some discussion. This group is probably in a better position than any to identify what are the limiting factors for this kind of research.

We thank the reviewer for this suggestion and have added discussion of the difficulties in detecting signals for this kind of problem. That said, we are confident in stating that there is a meaningful correlation between PC 2 of PN Ca++ responses and OCT-MCH behavior given our model’s performance in predicting preference in a test set of flies, and in the consistent signal in ORN Bruchpilot.

Reviewer #3 (Recommendations for the Authors):

Two major concerns, one experimental/technical and one conceptual:

(1) I appreciate the difficulty of the experimental design and problem. However, the correlations reported throughout are based on neural measurements in only 5 glomeruli (~10% of the olfactory system) at early stages of olfactory processing.

We acknowledge that only imaging 5 glomeruli is regrettable. We worked hard to develop image analysis pipelines that could reliably segment as many glomeruli as possible from almost all individual flies. In the end, we concluded that it was better to focus our analysis on a (small) core set of glomeruli for which we had high confidence in the segmentation. Increasing the number of analyzed glomeruli is high on the list of improvements for subsequent studies. Happily, we are confident that we are capturing a significant, biologically meaningful correlation between PC 2 of PN calcium (dominated by the responses in DC2 and DM2) and OCT-MCH preference.

3-OCT and MCH activate many glomeruli in addition to the five studied, especially at the concentrations used. There is also limited odor-specificity in their response matrix: notably responses are more correlated in all glomeruli within an individual, compared to responses across individuals (they note this in lines 194-198, though I don't quite understand the specific point they make here). This is a sign of high experimental variability (typically the dynamic range of odor response within an individual is similar to the range across individuals) and makes it even more difficult to resolve underlying individual variation.

We respectfully disagree with the reviewer’s interpretation here. There is substantial odor-specificity in our response matrix. This is evident in both the ORN and PN response matrices (and especially the PN matrix) as variation in the brightness across rows. Columns, which correspond to individuals, are more similar than rows, which correspond to odor-glomerulus pairs. The dynamic range within an individual (within a column, across rows) is indeed greater than the variation among individuals (within a row, across columns).

As an (important) aside, the odor stimuli are very unusual in this study. Odors are delivered at extremely high concentrations (variably 10-25% sv, line 464, not exactly sure what "variably' means- is the stimulus intensity not constant?) as compared to even the highest concentrations used in >95% of other studies (usually <~0.1% sv delivered).

We used these concentrations for a variety of reasons. First, following the protocol of Honegger and Smith (2020), we found that dilutions in this range produce a linear input-output relationship, i.e. doubling or halving one odorant yields proportionate changes in odor-choice behavior metrics. Second, such fold dilutions are standard for tunnel assays of the kind we used. Claridge-Chang et al. (2009) used 14% and 11% for MCH and OCT respectively, for instance. Finally, the specific dilution factor (i.e., within the range of 10-25%) was adjusted on a week-by-week basis to ensure that in an OCT-MCH choice, the mean preference was approximately 50%. This yields the greatest signal of individual odor preference. We have added this last point to the methods section where the range of dilutions is described (lines 442-445).

A parsimonious interpretation of their results is that the strongest correlation they see (ORN PC1 predicts OCT v. air preference) arises because intensity/strength of ORN responses across all odors (e.g. overall excitability of ORNs) partially predicts behavioral avoidance of 3-OCT. However, the degree to which variation in odor-specific glomerular activation patterns can explain behavioral preference (3-OCT v. MCH) seems much less clear, and correspondingly the correlations are weaker and p-values larger for the 3-OCT v. MCH result.

With respect, we disagree with this analysis. The correlation between ORN PC 1 and OCT v. air preference (R2 \= 0.23) is quite similar to that of PN PC 2 and OCT vs MCH preference (R2 \= 0.20). However, the former is dependent on a single outlying point, whereas the latter is not. The latter relationship is also backed up by the BRP imaging and modeling. Therefore in the revision we have de-emphasized the OCT v. air preference model and emphasized the OCT v. MCH preference models.

(2) There is a broader conceptual concern about the degree of logical consistency in the authors' interpretation of how neural variability maps to behavioral variability. For instance, the two odors they focus on, 3-OCT and MCH, barely activate ORNs in 4 of the 5 glomeruli they study. Most of the correlation of ORN PC1 vs. behavioral choice for 3-OCT vs. air, then, must be driven by overall glomerular activation by other odors (but remains predictive since responses across odors appear correlated within an individual). This gives pause to the interpretation that 3-OCT-evoked ORN activity in these five glomeruli is the neural substrate for variability in the behavioral response to 3-OCT.

Our interpretation of the ORN PC1 linear model is not that 3-OCT-evoked ORN activity is the neural substrate for variability – instead, it is the general responsiveness of an individual’s AL across multiple odors (this is our interpretation of the the uniformly positive loadings in ORN PC1). It is true that OCT and MCH do not activate ORNs as strongly as other odorants – our analysis rests on the loadings of the PCs that capture all odor/glomerulus combinations available in our data. All that said, since a single outlier in Figure 1L dominates the relationship, therefore we have de-emphasized these particular results in our revision.

This leads to the most significant concern, which is that the paper does not provide strong evidence that odor-specific patterns of glomerular activation in ORNs and PNs underlie individual behavioral preference between different odors (that each drive significant levels of activity, e.g. 3-OCT v. MCH), or that the ORN-PN synapse is a major driver of individual behavioral variability. Lines 26-31 of the abstract are not well supported, and the language should be softened.

We have modified the abstract to emphasize our confidence in PN calcium correlating with odor-vs-odor preference (removing the ORN & odor-vs-air language).

Their conclusions come primarily from having correlated many parameters reduced from the ORN and PN response matrices against the behavioral data. Several claims are made that a given PC is predictive of an odor preference while others are not, however it does not appear that the statistical tests to support this are shown in the figures or text.

For each linear model of calcium dynamics predicting preference, we restricted our analysis to the first 5 principal components. Thus, we do not feel that we correlated many parameters against the behavioral data. As mentioned below, the correlations identified by this approach comfortably survive a conservative Bonferroni correction. In this revision, a linear model with a single predictor – the projection onto PC 2 of PN calcium – is the result we emphasize in the text, and we report R2 between measured and predicted preference for both a training set of flies and for a test set of flies (Figure 1M and Figure 2 – figure supplement 1).

That is, it appears that the correlation of models based on each component is calculated, then the component with the highest correlation is selected, and a correlation and p-value computed based on that component alone, without a statistical comparison between the predictive values of each component, or to account for effectively performing multiple comparisons. (Figure 1, k l m n o p, Figure 3, d f, and associated analyses).

To reiterate, this was our process: 1) Collect a training data set of paired Ca++ recordings and behavioral preference scores. 2) Compute the first five PCs of the Ca++ data, and measure the correlation of each to behavior. 3) Identify the PC with the best correlation. 4) Collect a test data set with new experimental recordings. 5) Apply the model identified in step 3. For some downstream analyses, we combined test and training data, but only after confirming the separate significance of the training and test correlations.

The p-values associated with the PN PC 2 model predicting OCT-MCH preference are sufficiently low in each of the training and testing sets (0.0063 and 0.0069, respectively) to pass a conservative Bonferroni multiple hypothesis correction (one hypothesis for each of the 5 PCs) at an alpha of 0.05.

Additionally, the statistical model presented in Figure 4 needs significantly more explanation or should be removed- it's unclear how they "infer" the correlation, and the conclusions appears inconsistent with Figure 3 - Figure Supplement 2.

We have removed Figure 4 and have improved upon our approach of inferring the strength of the correlation between latent calcium and behavior in the Methods, incorporating bootstrapping of all sources of data used for the inference (lines 622-628). At the same time, we now emphasize that this analysis is a bonus of sorts, and that the simple correlation between Ca++ and behavior is the main result.

Suggestions:

(1) If the authors want to make the claim that individual variation in ORN or PN odor representations (e.g. glomerular activation patterns) underlie differences in odor preference (MCH v. OCT), they should generalize the weak correlation between ORN/PN activity and behavior to additional glomeruli and pair of odors, where both odors drive significant activity. Otherwise, the claims in the abstract should be tempered.

We have modified the abstract to focus on the effect we have the highest confidence in: contrasting PN calcium activation of DM2 and DC2 predicting OCT-MCH preference.

(2) One of the most valuable contributions a study like this could provide is to carefully quantify the amount of measurement variation (across trials, across hemispheres) in neural responses relative to the amount of individual variation (across individuals). Beyond the degree of variation in the amplitude of odor responses, the rank ordering of odor response strength between repeated measurements (to try to establish conditions that account for adaptation, etc.), between hemispheres, and between individuals is important. Establishing this information is foundational to this entire field of study. The authors take a good first step towards this in Figure 1J and Figure 1, supplement 5C, but the plots do not directly show variance, and the comparison is flawed because more comparisons go into the individual-individual crunch (as evidenced by the consistently smaller range of quartiles). The proper way to do this is by resampling.

We do not know what the reviewer means by “individual-individual crunch,” unfortunately. Thus, it is difficult to determine why they think the analysis is flawed. We are also uncertain about the role of resampling in this analysis. The medians, interquartile ranges and whiskers in the panels referenced by the reviewer are not confidence intervals as might be determined by bootstrap resampling. Rather, these are direct statistics on the coding distances as measured – the raw values associated with these plots are visualized in Figure 1H.

In our revision we updated the heatmaps in Figure 1 – figure supplement 3 to include recordings across the lobes and trials of each individual fly, and we have added a new supplementary figure, Figure 1 – figure supplement 4, to show the correspondence between recordings across lobes or trials, with associated rank-order correlation coefficients. Since the focus of this study was whether measured individual differences predict individual behavioral preference, a full characterization of the statistics of variation in calcium responses was not the focus, though it was the focus of a previous study (Honegger & Smith et al., 2019).

To help the reader understand the data, we would encourage displaying data prior to dimensionality reduction - why not show direct plots of the mean and variance of the neural responses in each glomerulus across repeats, hemispheres, individuals?

We added a new supplementary figure, Figure 1 – figure supplement 4, to show the correspondence between recordings across lobes or trials.

A careful analysis of this point would allow the authors to support their currently unfounded assertion that odor responses become more "idiosyncratic" farther from the periphery (line 135-36); presumably they mean beyond just noise introduced by synaptic transmission, e.g. "idiosyncrasy" is reproducible within an individual. This is a strong statement that is not well-supported at present - it requires showing the degree of similarity in the representation between hemispheres is more similar within a fly than between flies in PNs compared to ORNs (see Hige... Turner, 2015).

Here are the lines in question: “PN responses were more variable within flies, as measured across the left and right hemisphere ALs, compared to ORN responses (Figure 1 – figure supplement 5C), consistent with the hypothesis that odor representations become more idiosyncratic farther from the sensory periphery.”

That responses are more idiosyncratic farther from the periphery is therefore not an “unfounded assertion.” It is clearly laid out as a hypothesis for which we can assess consistency in the data. We stand by our original interpretation: that several observations are consistent with this finding, including greater distance in coding space in PNs compared to ORNs, particularly across lobes and across flies. In addition, higher accuracy in decoding individual identity from PN responses compared to ORN responses (now appearing as Figure 1 – figure supplement 6A) is also consistent with this hypothesis.

Still, to make confusion at this sentence less likely, we have reworded it as “suggesting that odor representations become more divergent farther from the sensory periphery.” (lines 139-140)

(3) Figure 3 is difficult to interpret. Again, the variability of the measurement itself within and across individuals is not established up front. Expression of exogenous tagged brp in ORNs is also not guaranteed to reflect endogenous brp levels, so there is an additional assumption at that level.

Figure 3 – figure supplement 1 Panels A-C display the variability of measurements (Brp volume, total fluorescence and fluorescence density) both within (left/right lobes) and across individuals (the different data points). We agree that exogenous tagged Brp levels will not be identical to endogenous levels. The relationship appears significant despite this caveat.

Again there are statistical concerns with the correlations. For instance, the claim that "Higher Brp in DM2 predicted stronger MCH preference... " on line 389 is not statistically supported with p<0.05 in the ms (see Figure 3 G as the closest test, but even that is a test of the difference of DM2 and DC2, not DM2 alone).

We have changed the language to focus on the pattern of the loadings in PC 2 of Brp-Short density and replaced “predict.” (lines 366-369).

Can the authors also discuss what additional information is gained from the expansion microscopy in the figure supplement, and how it compares to brp density in DC2 using conventional methods?

The expansion microscopy analysis was an attempt to determine what specific aspect of Brp expression was predictive of behavior, on the level of individual Brp puncta, as a finer look compared to the glomerulus-wide fluorescence signal in the conventional microscopy approach. Since this method did not yield a large sample size, at best we can say it provided evidence consistent with the observation from confocal imaging that Brp fluorescent density was the best measure in terms of predicting behavior.

I would prefer to see the calcium and behavioral datasets strengthened to better establish the relationship between ORN/PN responses and behavior, and to set aside the anatomical dataset for a future work that investigates mechanisms.

We are satisfied that our revisions put appropriate emphasis on a robust result relating calcium and behavior measurements: the relationship between OCT-MCH preference and idiosyncratic PN calcium responses. Finding that idiosyncratic Brp density has similar PC 2 loadings that also significantly predict behavior is an important finding that increases confidence in the calcium-behavior finding. We agree with the reviewer that these anatomical findings are secondary to the calcium-behavior analyses, but think they warrant a place in the main findings of the study. As the reviewer suggests, we are conducting follow-on studies that focus on the relationship between neuroanatomical measures and odor preference.

(4) The mean imputation of missing data may have an effect on the conclusions that it is possible to draw from this dataset. In particular, as shown in Figure 1, supplemental figure 3, there is a relatively large amount of missing data, which is unevenly distributed across glomeruli and between the cell types recorded from. Strikingly, DC2 is missing in a large fraction of ORN recordings, while it is present in nearly all the PN recordings. Because DC2 is one of the glomeruli implicated in predicting MCH-OCT preference, this lack of data may be particularly likely to effect the evaluation of whether this preference can be predicted from the ORN data. Overall, mean imputation of glomerulus activity prior to PCA will artificially reduce the amount of variance contributed by the glomerulus. It would be useful to see an evaluation of which results of this paper are robust to different treatments of this missing data.

We confirmed that the linear model of predicted OCT-MCH using PN PC2 calcium was minimally altered when we performed imputation via alternating least squares using the pca function with option ‘als’ to infill missing values on the calcium matrix 1000 times and taking the mean infilled matrix (see MATLAB documentation and Figure 1 – figure supplement 5 of Werkhoven et al., 2021). Fitted slope value for model using mean-infilled data presented in article: -0.0806 (SE = 0.028, model R2 \= 0.15), fitted slope value using ALS-imputed model: -0.0806 (SE 0.026, model R2 \= 0.17).

Additional comments:

(1) On line 255 there is an unnecessary condition: "non-negative positive".

Thank you – non-negative has been removed.

(2) In Figure 4 and the associated analysis, selection of +/- 20% interval around the observed $R^2$ appears arbitrary. This could be based on the actual confidence interval, or established by bootstrapping.

We have replaced the +/- 20% rule by bootstrapping the calculation of behavior-behavior R2, calcium-calcium R2, and calcium-behavior R2 and propagating the uncertainties forward (Inference of correlation between latent calcium and behavior states section in Materials and Methods).

(3) On line 409 the claim is made "These sources of variation specifically implicate the ORN-PN synapse..." While the model recapitulates the glomerulus specific variation of activity under PN synapse density variation, it also occurs under ORN identity variation, which calls into question whether the synapse distribution itself is specifically implicated, or if any variation that is expected to be glomerulus specific would be equally implicated.

We agree with this observation. We found that varying either the ORNs or the PNs that project to each glomeruli can produce patterns of PN response variation similar to what is measured experimentally. This is consistent with the idea that the ORN-PN synapse is a key site of behaviorally-relevant variation.

(4) Line 214 "... we conclude that the relative responses of DM2 vs DC2 in PNs largely explains an individual's preference." is too strong of a claim, based on the fact that using the PC2 explains much more of the variance, while using the stated hypothesis noticeable decreases the predictive power ($R^2$ = 0.2 vs $R^2$ = 0.12 )

We have changed the wording here to “we conclude that the relative responses of DM2 vs DC2 in PNs compactly predict an individual’s preference.” (lines 192-193)

Author response:

Reviewer #1:

We thank the reviewer for recognizing the impact of our work on the pivotal roles of N-glycan-dependent ERQC in cellular fitness and pathogenicity and providing valuable comments to be considered to improve the manuscript. As suggested, we will rearrange data, reduce text volume, and discuss the possibility of how ERQC mutation decreases EV secretion without significant defect in conventional secretion. Regarding the proteomics data, we have already initiated a comparative analysis of total intracellular and EV-associated proteins to determine whether the reduced cargo loading in the Ugg1 mutant is specific to EV-associated proteins. Additionally, we may extend the analysis to include total secretion, enabling a clearer comparison between classical secretion and EV-mediated secretion to better evaluate the extent of classical secretion defects in the Ugg1 mutant.

Reviewer #2:

We sincerely thank the reviewer for the positive evaluation of our work. As recommended, we will reduce the text and reorganize the data to enhance the manuscript's readability.

Reviewer #3:

We sincerely thank the reviewer for the high appreciation of our work. As recommended, we will provide a more detailed explanation of the results with improved interpretation, strongly grounded on the obtained data.

eLife Assessment

This important study confirms the molecular function of putative components of the N-glycan-dependent endoplasmic reticulum protein quality control (ERQC) system in the pathogen Cryptococcus neoformans. The study demonstrates an involvement in fitness, virulence, and the secretion and composition of extracellular vesicles, albeit in ways that are not yet fully understood. The evidence provided is largely convincing, with rigorous well-controlled assays and the use of complemented strains.

Reviewer #1 (Public review):

Summary:

Using gene deletion analysis, the authors confirm the molecular function of putative components of an N-glycan-dependent endoplasmic reticulum protein quality control (ERQC) system (UGG1, MNS1, MNS101, MNL1, and MNL2), in the basidiomycetous fungal pathogen Cryptococcus neoformans. Specifically, they confirm the essential role of these components in the ERQC system and their role in ER stress which contributes to cellular fitness and pathogenicity.

The second part of the study links the components to secretion, mainly EV biogenesis and composition. However, this part of the study is less convincing.

Strengths:

Although it is unclear why ER stress in the mutants would not manifest into a classical secretion defect, this is a rigorous, well-controlled study, with the use of complemented strains that demonstrate phenotypic restoration. The diagram in Figure 1 is very useful in orientating the reader to a complex subject matter, although the legend could be more descriptive.

Weaknesses:

A major weakness is the sheer volume of data presented (in the main text and supplement), which makes the results difficult to follow and retain: the work could essentially be two separate studies.<br /> Another major weakness is the lack of mechanistic insight into the role of the ERQC system in EV secretion and its disconnection to "classical" secretion, which is difficult to reconcile. Some insight into why EV secretion is decreased, and classical secretion is unaffected, would strengthen the significance of the findings. No mechanism is provided to explain why the ERQC mutants (Ugg1 mutant in particular) would have reduced and heterogeneously sized EVs. Furthermore, it is not convincing that the EV content changes would greatly impact fitness and virulence. The proteomics data showing reduced cargo in the Ugg1 mutant is not convincing and difficult to follow.

Reviewer #2 (Public review):

Summary:

This study investigates the molecular function of the N-glycan-dependent endoplasmic reticulum protein quality control system (ERQC) in Cryptococcus neoformans and correlates this pathway with key features of C. neoformans virulence, especially those mediated by extracellular vesicle transport. The findings provide valuable insights into the connection between this pathway and the biogenesis of C. neoformans extracellular vesicles.

Strengths:

The strength of this study lies primarily in the careful selection of appropriate and current methodologies, which provide a solid foundation for the authors' results and conclusions across all presented data. All experiments are supported by well-designed and established controls in the study of C. neoformans, further strengthening the validity of the results and conclusions drawn from them. The study presents novel data on this important pathway in C. neoformans, establishing its connection with C. neoformans virulence. Interestingly, the findings led the authors to understand the relationship between this pathway and the transport of key fungal virulence factors via extracellular vesicles. This was demonstrated in the study, paving the way for a deeper understanding of extracellular vesicle biogenesis-a field still filled with gaps but one to which this study contributes solid data, helping to clarify aspects of this process.

Weaknesses:

I do not see significant weaknesses in this study. The experiments are well-grounded, and the results are clearly presented. I believe the only weakness is that the paper could be condensed. Sections like the discussion, for instance, are extremely lengthy, which may make reading and, consequently, understanding more challenging for many readers. Regarding the presentation of the results, while clear, the figures contain a lot of information, and I believe that some of this content could be moved to supplementary figures.